DRA VIDIAN COMPARATIVE GRAMMAR A COMPARATIVE GRAMMAR OF THE DRA VIDIAN OR SOUTH-INDIAN FAMILY OF LANGUAGES RT. REV. ROBERT CALDWELL, D.D., LL.D. HONOKARY MEMBER OF THE ROYAL ASIATIC SOCIETY, FELLOW OF THE UNIVERSITY OF MADRAS, BISHOP IN TINNEVELLY, SOUTHERN INDIA. ZTbirD BMtion, 1Revf6e& anD JEDfteD b^ tbe REV. J. L. WYATT, M.A. LATE MISSIONARY S.P.G., EDEYANGOODY, TINNEVELLY, AND TRICHINOPOLY, READER IN TAMIL, UNIVERSITY OF CAMBRIjXJE, RECTOR OF BRANDON, SUFFOLK, AND T. RAMAKRISHNA PILLAL B.A. FELLOW OF THE ROYAL HISTORICAL SOCIETY, MEMBER AMERICAN ORIENTAL SOCIETY, FELLOW OK THE UNIVERSITY OF MADRAS, CHAIRMAN, DRAVIDIAN BOARD OF STUDIES AND TAMIL BOARD OF EXAMINERS, MADRAS UNIVERSITY. J.ONDON REGAN PAUL, TRENCH, TRUBNER & CO., LTD BROADWAY HOUSE, 68-74 CARTER LANE, E.G. 1913 (V^l^^jtuioeA EDITORS' PREFACE TO THE THIRD EDITION Bishop Caldwell's Comparative Grammar of the Dravidian or South-Indian Family of Languages is so well known a classic of Indian philology as to need no introduction to readers who are interested in the ethnology or linguistics of India. But a few words of introduction seem necessary to explain the origin and scope of the present edition. As a justification for attempting a new edition, it is only necessary to say that the book was originally published so long ago as 185G, and that the only other edition produced by the author dates from 1875, and has long been out of print. Con- sequently it has become rare and expensive, a matter of importance to Indian scholars and students, few of whom are in a position to pay the prices now demanded for a work indispensable to their studies. No further explanation seems called for to justify the issue of a third edition of a work which has no rivals. We may indeed hope that students of the Dravidian tongues will not only welcome this reissue of Dr. Caldwell's famous work, but will be surprised that they have had to wait nearly forty years to greet its publication. It only remains, then, to explain briefly in what respects this edition differs from its predecessors. We have added some statistical and other notes. For those embodying the latest census figures we have to express our gratitude to Mr. E. A. Gait, c.i.e.. Census Com- missioner of India. We have also transliterated all Greek words for the benefit of those who may be unacquainted with that language. We have decided to omit so much of Bishop Caldwell's Introduction as relates to the History of Dravidian Literature. In doing so we have been guided by two considerations. An account of Dravidian Literature is not strictly germane to the main purpose of the book, as Bishop Caldwell himself candidly admitted. In the second place, some of the authot's conclusions as to the dates of the older books have been rendered obsolete by the researches of Indian scholars and by the investigations of the Government Archaeo- logical Departments. ^ J ) '"^i ' i vi editors' preface We have also omitted many pages of purely controversial matter, in which Bishop Caldwell was at pains to controvert the views of writers now forgotten or negligible. In other respects the present edition is a reprint, revised and brought up to date, of a work which has so far found no successor, and will in no case be rendered wholly superfluous by the labours of other scholars. As the pioneer effort of Dravidian scholarship on European lines it will always have its own interest and importance, even if others should hereafter build on the foundations so solidly laid by the most distinguished investigator of Dravidian philology. Our task has been, in the main, one of pious conservation, and our sole object has been to enable students to obtain access to so much of the author's work as retains a permanent value, in view of the increased attention paid to the study of Oriental languages in general, and the solicitude shown by the Madras Government for the study of the languages of South India in particular. AUTHOR'S PREFACE TO THE SECOND EDITION It is now nearly nineteen years since the first edition of this book was published, and a second edition ought to have appeared long ere this. The first edition was soon exhausted, and the desirableness of bringing out a second edition was often suggested to me. But as the book was a first attempt in a new field of research and necessarily very imperfect, I could not bring myself to allow a second edition to appear without a thorough revision. It was evident, however, that the preparation of a thoroughly revised edition, with the addition of new matter wherever it seemed to be necessary, would entail upon me more labour than I was likely for a long time to be able to undertake. The duties devolving upon me in India left me very little leisure for extraneous work, and the exhaustion arising from long residence in a tropical climate left me very little surplus strength. For eleven years, in addition to my other duties, I took part in the Revision of the Tamil Bible, and after that great work had come to an end, it fell to my lot to take part for one year more in the Revision of the Tamil Book of Common Prayer. I suffered also for some time from a serious illness of such a nature that it seemed to render it improbable that I should ever be able to do any literary work again. Thus year after year elapsed, and year after year the idea of setting myself to so laborious a task as that of preparing a second edition of a book of this kind grew more and more distasteful to me. I began to hope that it had become no longer necessary to endeavour to rescue a half-forgotten book from oblivion. At this juncture it was considered desirable that I should return for a time to my native land for the benefit of my health ; and at the same time I was surprised to receive a new and more urgent request that I should bring out a second edition of this book — for which I was informed that a demand still existed. Ac- cordingly I felt that I had now no opjtion left, and arrived reluctantly at the conclusion that as^ihe first edition was brought out during the period of my first return to this country on furlough, so it had become necessary that the period of my second furlough should be devoted to the preparation and publication of a second edition. Vlll AUTHOR S PREFACE The first edition — chiefly on account of the novelty of the under- taking — was received with a larger amount of favour than it ap- peared to me to deserve. I trust that this second edition, revised and enlarged, will be found more really deserving of favour. Though reluctant to commence the work, no sooner had I entered upon it than my old interest in it revived, and I laboured at it con amore. I have endeavoured to be accurate and thorough throughout, and to leave no difficulty unsolved, or at least uninvestigated ; and yet, notwithstanding all my endeavours, I am conscious of many deficiencies, and feel sure that I must have fallen into many errors. Of the various expressions of approval the first edition received, the one which gratified me most, because I felt it to be best deserved, was that it was evident I had treated the Dra vidian languages " lovingly." I trust it will be apparent that I have given no smaller amount of loving care and labour to the preparation of this second edition. The reader must be prepared, however, to find that many of the particulars on which I have laboured most " lovingly," though exceedingly interesting to persons who have made the Dravidian languages their special study, possess but little interest for persons whose special studies lie in the direction of some other family of languages, or who are interested, not in the study of any one language or family of languages in particular, but only in philological studies in general, or in discussions respecting the origin of language in general. It is now more than thirty-seven years since I commenced the study of Tamil, and I had not proceeded far in the study before I came to the conclusion that much light might be thrown on Tamil by comparing it with Telugu, Canarese, and the other sister idioms. On proceeding to make the comparison I found that my supposition was verified by the result, and also, as it appeared to me, that Tamil imparted still more light than it received. I have become more and more firmly persuaded, as time has gone on, that it is not a theory, but a fact, that none of these languages can be thoroughly understood and appreciated without some study of the others, and hence that a Comparative Grammar of the Dravidian Languages may claim to be regarded not merely as something that is useful in its way, but as a necessity. I trust it will be found that I have not left much undone that seemed to be necessary for the elucidation of Tamil ; but I hope this branch of work will now be taken up by persons who have made Telugu, Canarese, Malayalam, or Tulu their special study, so that the whole range of the Dravidian languages and dialects may be fully elucidated. One desideratum at present seems to be author's , PREFACE ix a Comparative Vocabulary of the Dravidian Languages, distin- guishing the roots found, say, in the four most distinctive languages — Tamil, Telugu, Canarese, and Malayalam — from those found only in three, only in two, or only in one. An excellent illustiation of what may be done in this direction has been furnished by Dr. Gundert, whose truly scientific " Dictionary of Malayalam " has given a fresh stimulus to Dravidian philology. Another thing which has long appeared to me to be a desideratum is a more thorough examination of all the South Indian alphabets, ancient and modern, with a careful comparison of them, letter by letter, not only with the alphabets of Northern India, ancient and modern, but also, and especially, with the characters found in ancient in- scriptions in Ceylon, Java, and other places in the further East. It has been announced that a work on this subject, by Dr. Burnell, M.C.S., entitled " South-Indian Palaeography," is about to be pub- lished in Madras, but I regret that a copy of it has not yet arrived. It has been my chief object throughout this work to promote a more systematic and scientific study of the Dravidian languages themselves — for their own sake, irrespective of theories respecting their relationship to other languages — by means of a careful inter- comparison of their grammars. Whilst I have never ceased to regard this as my chief object, I have at the same time considered it desirable to notice, as opportunity occurred, such principles, forms, and roots as appeared to bear any afiinity to those of any other language or family of languages, in the hope of contributing thereby to the solution of the question of their ultimate relationship. That question has never yet been scientifically solved, though one must hope that it will be solved some day. It has not yet got beyond the region of theories, more or less plausible. My own theory is that the Dravidian languages occupy a position of their own between the languages of the Indo-European family and those of the Turanian or Scythian group — not quite a midway position, but one consider- ably nearer the latter than the former. The particulars in which they seem to me to accord with the Indo-European languages are numerous and remarkable, and some of them, it will be seen, are of such a nature that it is impossible, I think, to suppose that they have been accidental ; but the relationship to which they testify — in so far as they do testify to any real relationship — appears to me to be very indefinite, as well as very remote. On the other hand the particulars in w^ich they seem to me to accord with most of the so-called Scythian languages are not only so numerous, but are so distinctive and of so essential a nature, that they appear to me to amount to what is called a family likeness, and therefore X author's preface naturally to suggest the idea of a common descent. The evidence is cumulative. It seems impossible to suppose that all the various remarkable resemblances that will be pointed out, section after section, in this work can have arisen merely from similarity in mental development — of which there is no proof — or similarity in external circumstances and history — of which also there is no proof — much less without any common cause whatever, but merely from the chapter of accidents. The relationship seems to me to be not merely morphological, but — in some shape or another, and however it may be accounted for — genealogical. The genealogical method of investigation has produced remarkable results in the case of the Indo-European family of languages, and there seems no reason why it should be discarded in relation to any other family or group ; but this method is applicable, as it appears to me, not merely to roots and forms, but also to principles, contrivances, and adapta- tions. I have called attention to the various resemblances I have noticed, whether apparently important or apparently insignificant — not under the supposition that any one of them, or all together, will suffice to settle the difficult question at issue, but as an aid to inquiry, for the purpose of helping to point out the line in which further research seems likely — or not likely — to be rewarded with success. An ulterior and still more difficult question will be found to be occasionally discussed. It is this : Does there not seem to be reason for regarding the Dra vidian family of languages, not only as a link of connection between the Indo-European and Scythian groups, but — in some particulars, especially in relation to the pronouns — as the best surviving representative of a period in the history of human speech older than the Indo-European stage, older than the Scythian, and older than the separation of the one from the other ? Whilst pointing out extra-Dravidian affinities wherever they appeared to exist, it has always been my endeavour, as far as possible, to explain Dravidian forms by means of the Dravidian languages themselves. In this particular I think it will be found that a fair amount of progress has been made in this edition in comparison with the first — for which I am largely indebted to the help of Dr. Gundert's suggestions. A considerable number of forms which w^ere left unexplained in the first edition have now, more or less conclusively, been shown to have had a Dravidian origin, and possibly this process will be found to be capable of being carried further still. The Dravidian languages having been culti- vated from so early a period, and carried by successive stages of progress to so high a point of refinement, we should be prepared to author's preface xi expect that in supplying themselves from time to time with in- flexional forms they had availed themselves of auxiliary words already in use, with only such modifications in sound or meaning as were necessary to adapt them to the new purposes to which they were applied. Accordingly it does not seem necessary or desirable to seek for the origin of Dravidian forms out of the range of the Dravidian languages themselves, except in the event of those languages failing to afford us a tolerably satisfactory explanation. Even in that event, it must be considered more probable that the evidence of a native Dravidian origin has been obliterated by lapse of time than that the Dra vidians, when learning to inflect their words, borrowed for this purpose the inflexional forms of their neighbours. It is a different question whether some of the Dravidian forms and roots may not have formed a portion of the linguistic inheritance which appears to have descended to the earliest Dra- vidians from the fathers of the human race. I should be inclined, however, to seek for traces of that inheritance only in the narrow area of the simplest and most necessary, and therefore probably the most primitive, elements of speech. In preparing the second edition of this book, as in preparing the first, I have endeavoured to give European scholars, whether resident in Europe or in India, such information respecting the Dravidian languages as might be likely to be interesting to them. I have thought more, however, of the requirements of the natives of the country, than of those of foreigners. It has been my earnest and constant desire to stimulate the natives of the districts in which the Dra- vidian languages are spoken to take an intelligent interest in the comparative study of their own languages ; and I trust it will be found that this object has in some measure been helped forward. Educated Tamilians have studied Tamil — educated Telugus have studied Telugu — the educated classes in each language-district have studied the language and literature of that district — with an earnestness and assiduity which are highly creditable to them, and which have never been exceeded in the history of any of the languages of the world — except, perhaps, by the earnestness and assiduity with which Sanskrit has been studied by the Brahmans. One result of this long-continued devotion to grammatical studies has been the development of much intellectual acuteness ; another result has been the progressive refinement of the languages themselves ; and these results have acf^d and reacted one upon another. Hence, it is impossible for any European who has acquired a competent knowledge of any of the Dravidian languages — say Tamil — to regard otherwise than wdth respect the intellectual capacity of a xii author's peeface people amongst whom so wonderful an organ of thought has been developed. On the other hand, in consequence of the almost ex- clusive devotion of the native literati to grammatical studies they have fallen considerably behind the educated classes in Europe in grasp and comprehensiveness. What they have gained in acute- ness, they have lost in breadth. They have never attempted to compare their own languages with others — not even with other languages of the same family. They have never grasped the idea that such a thing as a family of languages existed. Consequently the interest they took in the study of their languages was not an intelligent, discriminating interest, and proved much less fruitful in results than might fairly have been expected. Their philology, if it can be called by that name, has remained up to our own time as rudimentary and fragmentary as it was ages ago. Not having become comparative, it has not become scientific and progressive. The comparative method of study has done much, in every depart- ment of science, for Europe ; might it not be expected to do much for India also ? If the natives of Southern India began to take an interest in the comparative study of their own languages and in comparative philology in general, they would find it in a variety of ways much more useful to them than the study of the grammar of their own language alone ever has been. They would cease to content themselves with learning by rote versified enigmas and harmonious platitudes. They would begin to discern the real aims and objects of language, and realise the fact that language has a history of its own, throwing light upon all other history, and ren- dering ethnology and archaeology possible. They would find that philology studied in this manner enlarged the mind instead of cramp- ing it, extended its horizon, and provided it with a plentiful store of matters of wide human interest. And the consequence probably would be that a more critical, scholarly habit of mind, showing itself in a warmer desire for the discovery of truth, would begin to prevail. Another result — not perhaps so immediate, but probably in the end as certain — a result of priceless value — would be the development of a good, readable, respectable, useful, Dravidian literature — a literature written in a style free at once from pedantry and from vulgarisms, and in matter, tone, and tendency, as well as in style, worthy of so intelligent a people as the natives of Southern India undoubtedly are. I trust the interest taken in their language, literature, and antiqui- ties by foreigners will not be without its effect in kindling amongst the natives of Southern India a little wholesome, friendly rivalry. If a fair proportion of the educated native inhabitants of each district AUTHORS PREFACE Xlll were ooly to apply themselves to the study of the philology and archaeology of their district with anything like the same amount of zeal with which the philology and archaeology of Europe are studied by educated Europeans, the result would probably be that many questions which are now regarded as insoluble would speedily be solved, and that pursuits now generally regarded as barren would be found full of fruit. Native pandits have never been surpassed in patient labour or in an accurate knowledge of details. They require in addition that zeal for historic truth and that power of discrimination, as well as of generalisation, which have hitherto been supposed to be special characteristics of the European mind. Both these classes of qualities seem to me to be combined in a remarkable degree in the articles recently contributed by learned natives to the Bombay Indian Antiquary on subjects connected with the languages and literature of Northern India ; and those articles appear to me to be valuable not only in themselves, but also as giving the world a specimen of the kind of results that might be expected if learned natives of Southern India entered, in the same critical, careful spirit, on the cultivation of the similar, though hitherto much-neglected, field of literary labour, which may be regarded as specially their owm I was much gratified last year on finding that this Comparative Grammar of the Dravidian Languages had ceased to be the only Indian Comparative Grammar that had appeared. Mr. Beames has followed up this line of philological research by the publication of the first volume of a Comparative Grammar of the Modern Aryan Languages of India — that is, the North-Indian Vernaculars. I regret that the second volume of that valuable work has not yet been published. A Comparative Grammar of the Kolarian tongues, the third great Indian family, has probably not yet been contem- plated ; but I am inclined to think that it would be found to be productive of important and interesting results. I have endeavoured to make the second edition of this work more easily available for reference, as well as more complete, than the former one, by providing the reader with a full table of contents and an index of proper names, together with paradigms of nouns, numerals, pronouns, verbs, &c. I have also given a list of the books and papers bearing, directly or indirectly, on Dravidian philology which have appeared since the first edition of this work, and which have been referred to or made use of in this edition. I have much pleasure in acknowledging the valuable help I hav« received from many friends, ximongst them are the following : — xiv author's preface Rev. J. Brigel ; C. P. Brown, Esq. ; A. C. Burnell, Esq., Ph.D. ; Rev. J. Clay ; T. W. Rhys Davids, Esq. ; Rev. E. Diez ; Prof. Eggeling ; Sir Walter Elliot, K.c.S.i. ; the late C. Gover, Esq.; Rev. F. Kittel ; Rev. F. Metz ; Prof. Max MuUer ; N. P. Narasim- miengar, Esq. ; Rev. Dr. Pope ; P. Le Page Renouf, Esq. ; Dr. Rost ; Prof. Teza ; Dr. Ernest Trumpp. I have especially to thank Colonel Yule, c.B., for much interesting and valuable information on points connected with topography and history ; and the Rev. Dr. Gundert for the invaluable help he was so kind as to render me in connection with every department of this work. I beg to thank the Indian and Colonial Governments and the various officers entrusted with the management of the late Indian census for the information with which I have been favoured respecting the numbers of the people speaking the various Dravidian languages. R. CALDWELL. Office of the Society for the Propagation OF THE Gospel, 19 Del AH AY Street, Westminster, London, 1875. CONTENTS PAGE INTRODUCTION ..... .1 Object in view, investigation and illustration of grammatical struc- ture of Dravidian languages. Those languages the vernaculars of Southern India, 1. Position of Sanskrit and Hindustani, 2. Position of English. Use of the Common Term ' Dravidian ' . . . , 3-6 Dra vidian Languages at one time styled ' Tamulian.' Kumarila- bhatta's term, Andhra-Drdvida bhdshd. Reasons for choosing the term Dravidian : Manu's use of Dravida.' Use of ' Dravidi ' by philological writers. Enumeration of Dravidian Languages . . , . 6-40 Six Cultivated Dialects. Six Uncultivated Dialects. I. Tamil ........ 6-16 Where spoken. Name of Madras ; spelling of ' Tamil.' Tamil erroneously called ' Malabar ' ; origin of the error ; Professor Max Miiller ; Dr. Hunter. Colebrooke ; first book printed in Tamil. ' Dravida ' corresponds to ' Tamil ' in Sanskrit ; proof of this ; Varaha- mihira, Taranatha, Mahawanso. Asoka's inscription ; Peutinger Tables ; Ravenna geographer. Derivation of native pandits ; names of three subdivisions of Tamil people ; Pandya ; Singhalese traditions, Mahabharata ; Pandyas on Malabar Coast, Pandyas as known to the Greeks. Pliny's references to the Pandyas ; Chola. — Asoka's in- scription, Ptolemy, Hwen Thsang ; capital of the Cholas, extent of their power. Ch^ra. — Various shapes of this name ; original identity of the three subdivisions of the Tamil people ; native tradition, representations in Sanskrit. Why is Tamil called ' Aravam ' ? Various theories. II. Malayalam ....... 16-20 Where spoken. Origin of the name ' Malayalam.' Different shapes of the name Kerala ; identity with * Chera ' ; meaning of ' Koiigu.' Cosmos * Indicopleustes' MaX^ ; period of separation of Malayajam from Tamil. Configuration of the country. Origin of the term ' Coromandel ' . . . . . 20-22 Era Paulino's supposition ; use of ' Choramandala ' by the first Por- tuguese ; equivalent of Ma' bar. Derivation from name of village of Coromandel inadmissible : Colonel Yule's communication. XVI CONTENTS Origin of the term ' Malabar '..... Use of first part of tho name amongst Greeks and Arabians ; use of the affix bar amongst Arabians and early Europeans ; origin of bar. Suggestion of Dr. Gundert ; Colonel Yule's communication. Maldives ; Persian bar ; origin of war of Kattywar, &c. ; Dr. Trumpp. III. Telugu ....... Where spoken. Eastern ' Klings ' ; Sanskrit Andhra ; Andhras in the Vedas and the Greek writers. Derivation of the name Telugu ; native derivation regarded by Mr. C. P. Brown as inaccurate. Traces of Trilingam : traces of Trikalinga ; meaning of Vadugu. IV. Canarese ....... Where spoken. Derivation of the name Karnataka ; different applications of the name. V. TuLU ........ Where spoken ; Tu|u a highly-developed language ; to which Dra vi- dian language most nearly alHed ? VI. KuDAGU or COOKG ...... Where spoken ; which Dra vidian language it resembles most ; doubtful whether it should be placed amongst the cultivated class. VII. TUDA Where spoken ; Tudas the smallest of Dravidian tribes ; books about the Tudas and their language. VIII. K6ta Where spoken ; characteristics of the language. IX. GOND ........ Gondwana ; numbers of the Gonds ; different tribes ; Koitors. X. Khond or Ku ....... Where spoken ; human sacrifices ; origin of name. XI. Maler or RIjmahal ...... Where spoken ; language different from that of the Santals. XII. Oraon ....... Relationships of this tribe and their language. Amount of the Dravidian element in the Maler and Oraon not clearly ascertained. Census of peoples and tribes speaking Dravidian languages. Tribes not enumerated; Kolarian tribes. Tribes of the North-Eastern frontier ; Brahui contains a Dravidian element : Dravidians seem to have entered India from the North-West. The Dravidian idioms not merely provincial dialects of the same language People not mutually understood ; Tamil and Telugu furthest apart. The Dravidian Languages independent of Sanskrit Supposition of the northern pandits that the South-Indian vernacu- lars were derived from Sanskrit erroneous. List of sixty words in Sanskrit and Tamil. Ancient dialect of Tamil contains Httle San- skrit. Relation of EngUsh to Latin, and of Tamil to Sanskrit, illus- trated by a comparison of Ten Commandments in English and Tamil. CONTENTS XVU FAQB Archbishop Trench's expressions. Tamil less studied than other dialects by Brahmans. Thirteen particulars in which the Dravidian languages differ essentially from Sanskrit. Are there traces of Scythian influences in Sanskrit itself ? Mr. Edkins's " China's Place in Philology." Is there a Dravidian element in the Vernacular Languages of Northern India ? ....... 52-61 Hypothesis that the corruption of Sanskrit out of which the Northern vernaculars have arisen was due to the Dravidian languages con- sidered ; general conclusion that the modifying influences, though probably Scythian or non- Aryan, do not appear to have been dis- tinctively Dravidian. To what group of Languages are the Dravidian idioms to be affiliated ? . 61-77 Professor Rask's opinion. Meaning of the term ' Scythian ; ' Pro- fessor Max Miiller. Intercomparison of the Scythian languages themselves should be carried further. Some of the resemblances incapable of being accounted for by accident. The original unity of languages probable. Confirmation of the Scythian theory by the Behistun Tablets. Principal points of resemblance between the language of the Tablets and the Dravidian languages. The existence of any analogy between the Dravidian languages and the Finno- Ugrian tends to confirm the argument for the original oneness of the human race. Note. — Professor Hunfalvy. Indo-European languages not so prolific of differences as Scythian, Relationship of Dravidian languages to Scythian not universally admitted ; Dr. Pope's remarks. Dr. Bleek's remarks ; possibility of developments ah intra. List of primitive Indo-Europeanisms discoverable in the Dravidian languages. Position between Indo-European and Scythian languages occupied by Dravidian ; existence of a few Semitic analogies. Australian affinities. Resemblances discoverable in an African language. * Which language or dialect best represents the primitive condition of the Dravidian tongues ?...... ll-IS No one dialect implicitly to be followed ; a comparison of all existing dialects our safest guide. 1. Literary, classical dialects of the Dravidian languages : to what extent may they be regarded as representing the primitive condition of those languages ? . . . . . . 78-81 As soon as the Indian languages begin to be cultivated, the literary style has a tendency to become a literary language. Illustrations from Northern India ; the same tendency in the Dravidian languages. High Tamil. 2. High antiquity of the literary cultivation of Tamil. . . 81-88 Six reasons for inferring its relatively high antiquity. The San- skrit words contained in I'b.mil belong to three different periods. Remarkable corruptions of certain Sanskrit words. Tamil inscriptions. Characters in which those inscriptions are written ; character of Jewish and Christian tablets. XVlll CONTENTS PAGE Earliest extant Written Relics of the Dravidian Languages 88-105 Dravidian words in the Hebrew of the Old Testament. Earliest Dravidian word in Greek, Ctesias's name for cinnamon. Largest stock of Dravidian words found in names of places mentioned by- Ptolemy and the other Greek geographers. List of these words. Con- dition of the Dravidian languages scarcely at all changed since the time of the Greeks. Note. — Roman coins ; dates of Greek geographers. Words of the Turks of the Altai preserved by the Chinese ; period when the Dravidian speech divided into dialects. Political and Social Relation of the Primitive Dravidians to THE Aryan and Pr^-Aryan Inhabitants of Northern India 106-109 Were the Dravidians identical with the aborigines whom the Aryans found in India ? Relations of the Dravidians to the Aryans seem to have been always peaceable. Dravidians may have been pre- ceded by another Scythian race. Mr. Curzon's opinion ; immigrations from India to Ceylon and back again. Original Use and Progressive Extension of the term ' Sudra ' 109-113 Were the Sudras of the same race as the Aryans, or of a different race ? Lassen's supposition. Aryanisation of the Dravidians the result, not of conquest, but of colonisation. Connection of the Pandyas with the Pandavas. Note — Professor Max Miiller's remarks. Dravidians called Sudras by the Brahmans ; Sudra has a higher meaning in the South than in the North. Pr^-Aryan Civilisation of the Dravidians . . 11.3, 114 Testimony of the Dravidian vocabulary, when freed from its San- skrit. Probable date of Aryan civilisation of the Dravidians . . 114-118 First city and state of the Dravidians probably Kolkei on the Tam- raparni ; Agastya, the traditional leader of the first Brahman colony. Agastya's age ; references to Dravidas, &c., in Manu and the Maha- bharata. Note. — Name of Agastya's mountain. References to early Dravidians in Maha-wanso. Inference from Kumarila-bhatta's - reference to the Dravidians : names of places recorded by the Greeks Brahmanical ; suppositions respecting earliest Dravidian characters. COMPAEATIVE GEAMMAE Note on Transliteration ...... 121, 122 Reasons for using Roman characters. How vowels are to be repre- sented and pronounced. How cerebral consonants and nasals are to be represented. How some consonants are to be represented when single, and how when doubled. Tendency to pronounce e like ye, and o like wo. This usage not ancient, and not observed in this work. CONTENTS XIX PART I PAOB SOUNDS . . . ' . . . . 122-190 Dbavidian Alphabets . . . . • . 123-132 Three Alphabets in use. Their origin. Mr. Ellis's theory. Mr. E. Thomas's theory ; alphabet of the Malabar Inscriptions. Dr. Bumell's theory considered. Characters of Chera Inscriptions ; fur- ther research needed. Differences amongst the existing Alphabets. Peculiarities of Tamil Alphabet. Comparative View of Devanagari and Tamil Alphabets. Dbavidian System of Sounds — I. Voivels ....... 132-137 Weakening of a and d. Origin of the diphthong ei ; Kumarila- bhatta ; Mr. Beames. Enunciative u. Short e and o. Attraction of certain vowels. II. Consonants ....... 137-150 Convertibility of surds and sonants. Hebrew dagesh ; Finnish aw. Gutturals and palatals. Telugu pronunciation of palatals. Cerebrals and dentals ; conjunction of nasals and sonants. Labials and semi-vowels ; Tamil rule. Vocalic r. Cerebral I ; rough r. Pronunciation of Tamil nr ; which is radical ? Sibilants and aspirates. Origin of the Cebebbal Sounds — Excursus . . . 147 Reasons which lead me to suppose these sounds borrowed by San- skrit from the Dravidian languages. Mr. Norris's opinion ; Vedic Sanskrit I. Dialectic Interchange of Consonants .... 151-164 Interchange of Gutturals. Palatals. Linguals. Dentals. Labials. Semi-vowels. Sibilants. EupJionic Permutation of Consonants .... 165-167 Some permutations imitate Sanskrit. Others independent ; initial surds, when softened ? Assimilation of concurrent consonants. Euphonic Nunnation or Nasalisation .... 167-173 Insertion of a nasal before formative suffixes. This accounts for shape of certain Tamil adverbs of place. Suffixes with t and d nasal- ised. Origin of demonstrative adjectives ; Dr. Gundert's view. Insertion of a nasal before the d of the preterite in Tamil. Use of nunnation in other languages. Prevention of Hiatus ...... 173-180 Hiatus, how prevented in Indo-European Languages ? how in Dra- vidian ? Use of V, y, and n. Use of m. Use of ti in Tamil also. Origin of the n in certain numerals. Usage of Tulu. Euphonic insertion of r and d. • Harmonic Sequence of Vowels . . . . . 180-182 This law in the Turanian languages. Similar law of attraction in Telugu. In Canarese also. XX CONTENTS PAGE PrincifHes of Syllabation . . . . . . 182-186 Dra vidian dislike of compound consonants. Sanskrit double con- sonants, how dealt with in Tamil. The same peculiarity in Scythian languages ; similar instances in other languages ; Professor Max Miiller's illustrations. Kesemblance of Prakrit rules to Dravidian. Minor Dialectic Peculiarities ..... 186-190 1. Euphonic displacement of consonants. 2. Euphonic displace- ment of vowels. 3. Rejection of radical consonants. 4. Accent. Changes which Sanskrit words undergo when Dravidianised. PART II ROOTS . . . . ... . 191-217 Languages of Europe and Asia admit of being arranged into classes, in accordance with the changes effected in Roots by the addition of grammatical forms ; monosyllabic, intro- mutative, and agglutinative languages. Arrangement of Dravidian Roots into Classes — 1. Verbal Roots . . . . . .192 2. Nouns ....... 193-195 Illustrations of the formation of nouns from verbal roots. Some nouns remain which cannot be traced to any ulterior source. Dravidian Roots originalhj Monosyllabic . . . .196 Successive accretions ; illustration. Euphonic Lengthening of Roots ..... 197 Crude roots lengthened by the addition of enunciative vowels. Formative Additions to Roots ..... 198-203 Originally formatives of verbal nouns ; used now to distinguish intransitive verbs from transitives. Examples of use and force of formatives. l,ku; 2,su; 3,du. Origin of wmc^w, to swim ; transitive suffix preferred as a formative. 4, bu ; euphonisation of formatives ; quality possessed in common by adjectives and transitive verbs. First part of the word alone generally contains the root ; examples. Reduplication of Final Consonant of Root .... 204 Four purposes for which this is done in Tamil ; rationale. Particles of Specialisation ...... 205-209 Use of such particles in Semitic languages. Resemblance of Dra- vidian root-system to Semitic in this particular. Illustrations ; groups which radiate from the base syllables ad and aii. List of specialising particles ending in a consonant. Another set of groups of roots ; Max Miiller, Aryan instances. Changes in Root-Vowels . . . . . .211 Root-vowel generally unalterable. Exceptions — Internal Changes in Root^ . . . ^ 211-217 1. Euphonic changes. 2. Changes pertaining to grammatical expres- sion ; root-vowels of pronouns. 3. Strengthening of root-vowel CONTENTS XXI PAGE of verb to form verbal noun ; examples ; this usage not likely to have been derived from Sanskrit. Class of nouns so formed used adjec- tivally ; root-vowels of numerals ; shorter form older. Origin of pern, Tam. green. 4. Shortening of root-vowel in the preterite tense of certain verbs ; Tamil verbs vd and td ; Dr. Pope's opinion. Exceptions to the stability of root-vowels found also in the Scythian languages. PART III THE NOUN ....... 218-308 Section I. — Gender and Number . . . 218-252 1. Gender. Dravidian laws of gender accord more closely with Scythian than with Indo-European tongues. Indo-European laws of gender how dif- ferent from Scythian. Dravidian nouns divided into two classes, denoting rational bein and things without reason. Note. — Mind and body. Primitive laws of gender faithfully retained by Malayalam ; Telugu and Gond destitute of feminine singular. Canarese and Malayalam agree in this particular with Tamil. 2. Number ....... 222 Only two numbers, singular and plural. (1.) Masculine Singular ...... 222-227 Masculine singular suffixes in Tamil ; formation of appellatives. Subdivisions of appellatives. Canarese and Telugu suffixes. Ultimate identity of these with Tamil. (2.) Feminine Singular ...... 227,228 No suffix of the feminine singular in Telugu and Gond ; a formative sometimes used resembling the suffix of Tamil-Canarese, and probably of the same origin. Telugu mode of forming feminine singular appellatives ; a woman ; another feminine suffix possibly Sanskritic. (3.) Neuter Singular ...... 229-231 Dravidian nouns naturally neuter. Neuter suffixes rarely required ; suffix of neuter singular of demonstrative pronouns and appellative nouns. Affinities of neuter singular suffix in d possibly Indo-European. The Plural : Principles of Pluralisation .... 231-239 In Indo-European tongues number is denoted by the terminations ; in the Scythian number is generally left indefinite. Neuters plu- raHsed in Telugu, but rarely in Tamil. Progress of pluralisation. Sign of pluraUty distinct from case-sign ; added directly to the crude base. Paradigm of a noun in Hungarian and Tamil. Pluralisa- tion of masculine and feminine nouns ; no distinction of sex in plural ; analogies to other languages. Double plurals in Telugu. Double plurals in Tamil. (1.) Epicene Pluralising Particle ..... 239-244 Origin of epicene plural suffix ar, &c. Origin of mar in Tam.- Mal. ; fornlative in var. Dr. Gundert's explanation ; origin of verbal XXll CONTENTS PAOB terminations in mar, &c. Relationship to pluralising particles in other families of languages. Resemblance in use more important than resemblance in sound. (2.) Pluralising Particle of the Neuter . ' . . 244-252 1. The neuter plural suffix gal, with its varieties, gal appears as lu in Telugu. Gond particle ; particles used in High Asian lan- guages. Origin of gal. 2. Neuter plural suffix in a. Illustrations of use. Neuter plural of verb ; of possessive adjectives ; of Malay alam demonstratives. Lapse of a into ei. Telugu and Gond peculiarities. Relationship of neuter plural suffix a ; Indo-European affinities ; grammatical gender more fully developed in the Dra vidian than in any other family of languages. Section II. — Formation of Cases .... 252-308 Principles of Case-formation . . . . . 252-254 In this particular the Indo-European and Scythian families originally in agreement. Case-signs in both originally postpositional words ; case-terminations of the plural different from those of the singular in the Indo-European ; identical in the Scythian group. Dravidian languages follow the Scythian plan. Number of Declensions ...... 254 Only one declension, properly speaking, in Dravidian languages ; no difiference in signs of case. Number of Dravidian cases. The Nominative — Absence of '^Nominative Case-terminations . . 255-259 Dravidian nominative the noun itself. Apparent exceptions exist. (1.) Neuter termination am might be supposed to be a nominative case-sign, but is not ; origin of this am. Probably am was an ancient form of the demonstrative pronoun ; alternates with an. (2.) Final n of personal pronoun does not make it a nominative. (3.) Lengthening of vowel of personal pronoun in the nominative looks hke a case in point ; but probably vowel lengthened for sake of emphasis. Inflexion or Inflexional Base of the Oblique Cases . . . 259 In many instances the noun itseK used as the inflexional base. Gene- rally the base receives some augmentation. Signs of case added to this inflected form. (L) The inflexional increment ITS , with its dialectic varieties . . 260-262 Illustrations. In Telugu ni. in originally a locative. (2.) The inflexional increments ad and ar . . . . 262-264 These are most used in Canarese ; are they identical in origin ? Tamil sometimes uses neuter demonstrative adu in a similar manner. (3.) The inflexional increment Ti . . . . . 264 This the most common increment of neuter nouns in Telugu ; pro- bably ti, not ti ; connection of this with neuter demonstrative. (4.) The inflexional increment attu or attru . . 265-267 TamU nouns in am take this increment, attu used by the singular alone ; attru used instead of attu by a few neuter plural pronominals ; aitu and attru virtually identical. Origin of the r of attru. Dr. Gundert's views respecting its origin. CONTENTS XXlll PAGE (5.) The formation of the inflexion by means of doubling and hardening the final consonant ...... 267, 268 Explanation of this doubling ; Dr. Gundert's view. In Telugu, final consonant hardened, but not doubled. (6.) The inflexional increment i. Origin. Euphonic links of connection between the base and the inflexion .... 268-270 In Tamil, euphonic u. Use of v and y. The Accusative or Second Case ..... 270, 271 In Indo-European languages, accusative a sign of passivity ; in Dra- vidian, accusative case-sign originally a formative of neuter ab- stracts ; nominative much used instead. The same in Telugu as to things without life. (l.) Accusative Case-signs -EI, B, and A . . . . 271 In Tamil ei ; in Malayalam e or a. With what case-signs in other languages this may be compared. (2.) Accusative Case-signs am, annu, anna, nu, ci;c. . 272-274 am the Old Canarese sign, annu the modern ; change of m into n. Tulu case-sign nu or n; in Telugu nu or ni. Comparison with case- signs in other languages ; Indo-European m or n ; origin of Dravidian case-sign ; identical with am, the formative of neuter nouns. Change of am into a. The Instrumental or Third Case, properly so called . . 275-277 Different particles used in the different dialects. Telugu instru- mental ; Canarese, origin of this. Tamil and Malayalam dl, an ; origin of this ; Dr. Gundert's explanation. No affinities to other languages ; periphrastic modes of forming instrumental case. The Conjunctive or Social Case ..... 277 Ought to have a place in the list of Dravidian cases ; difference from instrumental Tam.-Mal. case-signs. Telugu case-sign ; is this allied to the Tam.-Mal. ? Tulu communicative case. The Dative or Fourth Case ...... 279-283 In North Indian languages dative postpositions substantially the same as accusative. Dravidian dative ; resemblance between Dravidian case-sign and North Indian, how to be accounted for ; Dr. Trumpp's explanation of Hindu led ; Mr. Beames's explanation. Antiquity of Dravidian ku ; Scythian dative case-signs bear some resemblance to Dravidian. Behistun-Scythian case-signs ; Malayalam seems to have two case-signs ; Dr. Gundert's view of origin of nnu. Can a Dravidian origin be discovered for ku ? The Ablative of Motion or Fifth Case .... 283-286 This case included in th%list out of deference to Sanskrit grammar- ians ; not really different from locative. Change of place expressed by addition of verb of motion ; Tamil suffixes il and in ; Old Canarese im. Were il and in originally identical ? Compound abla- tive suffixes in Canarese. Explanation of Telugu ablative ; Tulu. XXIV CONTENTS PAGE The Genitive or Sixth Case ..... 286-302 (1.) The abbreviated Pronominal Genitive ^ . . . 286 This may be explained as a pronominal adjective. Similar abbre- viation in the case of some of the numerals, (2.) The Neuter Inflexional Genitive .... 287-289 Neuter suffixes used for the genitive originally signs of the locative ; Dr. Trumpp's view ; adjectival use of these suffixes arose from their use as genitives. Connection between locative, genitive, and adjective ; Max Miiller's view. Inflexional suffixes used as signs of genitive in Telugu ; not so in Canarese. (3.) The Neuter Demonstrative Genitives .... 289 adu and ddu in Tamil, how used. Followed by the singular alone ; Telugu use. This suffix appended to the inflexion. (4.) The Possessive Suffix IN, and its varieties . . . 291-296 Tamil in and Telugu ni originally locative suffixes ; in the most com- mon of all possessive suffixes in Tamil ; Max Miilier. Other case- suffixes generally appended to this in. Adjectival force of in ; use of am resembles that of in. Indo-European analogies to the use of in as a genitive. Scythian analogies. (5.) The Genitival Suffix A . .... .296-300 Probably identical with the a which forms the relative participle ; a the only genitive case-sign in Canarese. So also in Telugu and Tulu ; adjectival a of some Telugu nouns identical with possessive a ; a little used now in Tamil, though first in the Ust. Its use generally confined to poetical plurals. Indo-European affinities of this a, especially in the later dialects. (6.) The 3Ialaydlam Genitive Suffix HIE or B-E . . . 300-302 This takes the shape of indre or inde. Some resemblances to this illusory ; Dr. Stevenson ; Hindustani and Persian rd, Sec. Identical with Tamil adu. (7.) Auxiliary Suffixes of the Genitive in Telugu and Tamil . . 302-303 (i.) Telugu yoha ; origin of this word, (ii.) Tamil udeiya means literally that which is the property of ; Mai. ude. Locative or Seventh Case ...... 303-306 il the most common sign of this case in Tamil ; any word signifying ' place ' may be used. Canarese suffixes 61 and alii ; Telugu andu and 16. Note. — Resemblances between Tamil il and Latin in. Telugu na ; use of the inflexion as a sign of the locative ; fusion of the meaning of genitives and locatives. Note. — Radical element in il is i; Max Miilier. The Vocative or Eighth Case ..... 306 No case-sign of the vocative in Dravidian languages ; modes in which the vocative is formed. Compound Case-signs ...... 307 Two or more case-signs occasionally compounded into one. CONTENTS XXV PAQB Possessive Compounds ..... 307 The absence of this class of compounds in the Dravidian languages remarkable. Section III. — Adjectives, or Nouns used Adjectivally . 308-318 Adjectives in grammatical agreement with substantives in the Indo- European tongues ; in the Scji^hian tongues independent nouns of qua- lity. 1. Dravidian adjectives also nouns of quality. 2. How- Sanskrit derivatives become Dravidian adjectives. 3. How nouns ending in hard consonants double those consonants when used as adjectives. Soft finals, how changed. 4. Each of the inflexional incre- ments used for converting substantives into adjectives. 5. Relative participles of verbs largely used as adjectives. 6. Past verbal participle used as an adjective in Telugu. 7. Many Dravidian adjectives formed by the addition to nouns of the suffixes by which relative participles are formed; (1.) Addition of the suffix iya / origin of this ; (2.) Addi- tion of the suffix a. Explanation of origin of certain adjectives ; (3.) Addition of the suffix of the future relative participle. 8. Nouns may become adjectives by the addition of the relative participle of the verb to become. Certain words erroneously styled adjectives. Comparison of Adjectives ..... 316-317 Mode of comparison different from that used in Indo-European lan- guages ; resembles Semitic and Scythian mode. Addition of con- junctive particle ^^m, &c., as an intransitive. Formation of superla- tive ; attempt of Robert de Nobilibus. Postpositions . . . . . . .318 AU postpositions nouns, in the locative case understood. Comparative paradigm of a Neuter Dravidian Noun^ sing, and plur. 318 PART IV THE NUMERALS 321-356 Each cardinal number has two shapes, that of a neuter noun of num- ber and that of a numeral adjective ; in the colloquial dialects the former sometimes used instead of the latter. Primitive form that of the numeral adjective . . . . . .321 One. — Two forms in existence, oJca in Telugu, oru in all other dialects. 1. Basis of oru is or. ondu or onn' at first sight resembles Indo-European ' one.' Origin of ondu from oru ; similar changes in other words. Dr. Gundert's opinion ; Mr. Kittel's. Origin of Telugu word for one, oka. Scythian analogies to oka ; are oka and or related ? Dravidian indefinite article. The numeral adjective for ' one ' used as a sort of indefinite article ...... 322-328 Two. — Neuter nouns differ slightly in the various dialects ; numeral adjective, ir ; the same in all.. Canarese form of neuter ; Tamil form nasaUsed. Radical form without a nasal ; origin of ir ; Dr. Gundert's opinion ; Mr. Kittel's. No analogies in any Indo-European language. Brahui word. No Scythian analogies . . . 329-332 XXVI CONTENTS PAQB Three. — Neuter noun ; numeral adjective, mu ? or mu ? Brahui word. Origin of word for three. Dr. Gundert ; Mr. Kittel . . 332-333 Four. — Neuter noun ; numeral adjective. Origin of nal. No Indo-European analogy ; Ugro-Finnish analogies remarkably close 334-335 Five. — Neuter noun ; numeral adjective, in all the dialects ei. Resemblance between Sans, panchau and Tam.-Mal. anju. How this resemblance has arisen. Radical meaning of ei ; Mr. Kittel's explanation ....... 335-340 Six. — Neuter noun and numeral adjective nearly alike ; root- meaning of aru. No analogy with other languages discoverable . 340-341 Seven. — Neuter noun and numeral adjective nearly alike. No resemblance to word for seven in other languages . . . 341-342 Eight. — Tamil neuter noun ettu resembles Indo-European octo, &c. ; this resemblance disappears on examination. Radical shape en ; explanation of Telugu word enimidi ; Telugu numeral adjective ewa, Origin of midi. Origin of en ; Max Miiller ; Mr. Clay ; origin of en ; similar derivation of a numeral in Lappish . . . 342-346 Nine. — In all Dravidian languages nine a compound number ; principal forms which nine assumes ; difference between meaning of word nine and Dravidian word ; second member of the word means ten. First member appears to mean ' one,' but probably means ' before.' Mode in which compounds into which nine enters are formed. No affinity between Tamil word and Greek . . 346-350 Ten. — The word for ten virtually the same in all Dravidian dialects, Changes which take place. Dr. Gundert's opinion ; comparison of Sanskrit pankti with Dravidian word. Malayalam word for twelve. Root of Dravidian word for ten ..... 350-354 A Hundred. — Sameness of word for a hundred in all Indo-European languages a proof of intellectual culture and unity ; one and the same word used by all Dravidian languages ; derivation . . 354 A Thousand. — Generally used Dravidian word a Sanskrit deriva- tive ; Telugu word ; derivation ..... 354 Ordinal Numbers ....... 354-356 Derivation of Dravidian ordinaL number first ; forms of ordinal suffixes of other numbers ; do. of adverbial numbers. Affiliation . . . . . ... 356-357 No evidence of Indo-European descent. Existence of Scythian analogies, especially as to the number four ; Professor Hunfalvy's opinion ; arithmetical faculty of Scythians not strongly developed. Dravidian Numerals in the Five Principal Dialects : Paradigm . 358 CONTENTS XXVll PART V PAGE THE PRONOUN ....... 359-444 Light thrown by pronouns on relationship of languages. Personal pronouns the most persistent of all words. Peculiarity of Japanese. Section I. — Personal Pronouns ..... 359-419 1. Pronoun of the First Person Singular . . . 359-383 Comparison of Dialects ...... 359-373 Primitive form. Classical and colloquial dialects to be com- pared ; inflexional forms and plurals to be compared, not nominative singular only. Written form of the word represents oldest pronuncia- tion ; forms of this pronoun in Tamil. Malayalam and Canarese forms. Telugu and Tulu ; minor dialects : which was the primitive form, nan or ydn ? Opinion expressed in former edition. Dr. Gundert's opinion ; Dr. Pope's " Outlines of Tuda "; the late Mr. Gover's paper. Relationship of ydn to nan ; changeableness of y. Malayalam middle point nan ; both initial and final n changeable. Both ydn and nan very ancient ; illustration from Sanskrit, asme, and vayam, yuslime and yuyam. Included vowel a ox e ? a weakened to e ; origin of final n ; a sign of number. Is n identical with m, the final of neuter singular nouns ? Only essential difference between pronouns of first and second persons consists in difference of included vowels a and i. ■ What is the explanation of this ? These cannot be the demonstrative vowels ; an explanation suggested. Chinese ; Mr. Edkins ; first three simple vowels utilised. Extra- Dravidian Relationship ..... 373 All pronouns of the first person traceable to one of two roots, ah and ma. 1. Semitic Analogies ...... 374 Sir H. Rawlinson. 2. Indo-European Analogies . . . . . 375-379 Comparison of pronouns and pronominal terminations of verb. Can any analogy to Dravidian pronoun be traced ? (1.) m of ma often changes to n. Instances of change of m into n. (2.) This m changes also into v. (3.) ma also changes into a ; were the Indo-European and the Dravidian words originally related ? Scythian Analogies ...... 379-383 Interesting analogies exist. (1.) Nominative, as well as base of obhque cases, derived from ma. Illustrations from various Scythian languages ; m the equivalent of ma. m occasionally changes into n ; instances. In some Scythian languages this pronoun almost identical with Dravidian. (2.) Some traces of the softening of na into a ; probability of a common origin of all these forms. Professor Hunfalvy's paper read at Laternational Congress of Orientalists. 2. Pronoun of the Second Person Singular . . 384-395 Comparison of Dialects ...... 384-388 Tamil forms of this pronoun. Second person of verb ; Beschi's error. Plurals ; Canarese and Telugu forms. Minor dialects. XXVlll CONTENTS PAGE Relative antiquity of existing forms ; ni very old, but * probably older. Oldest shape of the vowel, i or n ? probably i. Extra- Dravidian Relationship ..... 389-395 Dravidian pronoun of the second person singular more distinctively non-Aryan than the first : most prevalent form in both classes of lan- guages has t for its basis ; the other is founded on n. yu, base of the Aryan plural. Origin of yu from tu. Mr. Edkins' suggestion ; t gene- rally changed into s. s more prevalent in Scythian tongues than t. Euphonic final n ; instances. Another pronoun in n, not t, in some Scythian languages, apparently identical with the Dravidian ; Chinese. Behistun tablets, Brahui, Bornu ; allied forms in Ostiak, &c. Traces discoverable in Finnish, Turkish, &c. Himalayan dialects ; Australian. 3. The Reflexive Pronoun 'Self' .... 395-402 This pronoun, tan, more regular and persistent than any other of the Dravidian personal pronouns ; has a wider application than the corre- sponding Aryan reflexives. Used honorifically ; from which use a class of words has arisen. List of such words, with explanations : tamhirdn, tagappan, tandei, tdy, tammei, tannei, tameiyan, tamuk- kei, iambi, tangei, nambi ; Coorg instances ; use of tan as basis of abstract noun for quality, l^ote,. — Meaning of spinster and duhitri. Origin of ta, the base of this pronoun, from some demonstrative root; Sanskrit and Greek demonstratives in t. Use of tan in the word for quality, like Sans, tad, a confirmation. 4. Pluralisation of the Personal and Reflexive Pronotjns 402-415 Comparison of Dialects . . . . . . 402 Tamil plurals ; double plural in colloquial dialects. Telugu double plural ; similar usage in Gaurian languages ; Mr. Beames ; plurals of verbal inflexions. Canarese and Telugu plurals. Change of initial n in Telugu into w. Harmonic changes. Origin of Pluralising Particles ..... 406-408 (1.) Origin of R. ni-{y)-ir may mean thou + these people = you. Saius. yushme ; alternative explanation from »>, two. (2.) Origin of m; this m a relic of the copulative um ; used like Latin que ; nd-um, I + and = we. Verbs similarly pluralised. Eztra-Dravidian Relationship ..... 409-414 Finno-Ugrian analogies ; remarkable Aryan analogies ; n in the singular of pronouns and m in the plural in North Indian vernaculars ; Pah-Prakrit ; Mr. Beames in Indian Antiquary. Mr. Gover's opinion ; Dr. Pope's ; resemblance great, but only apparent. Oldest forms of Greek and Sanskrit plurals of personal pronouns. Explanation of sme ; sma found in singular. In third person also. Twofold Plural of the Dravidian Pronoun of the First Person . 414-415 Plural used as honorific singular ; two plurals, the plural inclusive and the plural exclusive ; similar distinction found in two North Indian languages ; not found in Indo-European family ; found every- where in Central Asia. Usage in different Dravidian dialects ; conclu- sion ; results exhibited in foflowing tables. CONTENTS XXIX PAGE Paradigms . , . . . . . 415-419 Dra vidian Pronoun of the First Person. „ „ „ Second Person. Pronoun of the First Person, in Seventeen Dialects of Central India ; Dr. Hunter's " Comparative Dictionary." Pronoun of the Second Person, in Seventeen Dialects of Central India ; Dr. Hunter's " Comparative Dictionary." Section II. — Demonstrative and Interrogative Pronouns . 420-444 Difficult to treat these two classes of pronouns separately. 1. Demonstrative and Interrogative Bases . . . 420-421 1. Demonstrative Bases . . . . ' . . 420 Dra vidian languages use for pronouns of the third person demonstra- tives signifying ' this ' and ' that,' man, &c. ; words which signify man, &c., have shrunk into terminations ; four demonstrative bases recog- nised — remote, proximate, intermediate, and emphatic. 2. Interrogative Bases . . . . . .421 Two classes of interrogatives — one an interrogative prefix, the other suffixed or added to the end of the sentence ; (a) e the most common interrogative prefix. 1. Paradigm of Demonstrative and Interrogative Prefixes . . 422 Beautiful regularity ; Dravidian demonstratives, not borrowed from Sanskrit, but much older ; Old Japhetic bases ; (b) ?/a, the other inter- rogative base ; e probably weakened from yd. Change of yd in Canarese into da ; uses of this interrogative. 2. Demonstrative and Interrogative Pronouns . . . 424-429 Bases best seen in neuter singular ; suffixes ; euphonic links of con- nection. In Tamil v and n ; Telugu usage ; Tulu. Tulu pecu- liarities ; Tamil abstract demonstrative and interrogative nouns. Neuter interrogative pronoun ; m or n use as a formative. Origin of the copulative conjunction U7n ; Dr. Gundert. 3. Demonstrative and Interrogative Adjectives . . . 429-431 Demonstrative and interrogative bases, when prefixed to substan- tives, acquire the meaning of adjectives ; initial consonant of sub- stantive doubled, or prefixed vowel lengthened. Tamil demonstrative adjectives anda, that, &c. Telugu triplet. 4. Demonstrative and Interrogative Adverbs . . . 431-436 These formed by annexing formative suffixes to vowel bases. Classes of adverbs arranged according to their formatives. List. (1.) Formative k, g, h ; (2.) Formative ch, j, n ; (3.) Formative t, d, n ; (4.) Formative t, d, n, also ndr ; (5.) Formative mh ; (6.) Forma- tive I, I Demonstratives and interrogatives formed from I found in Telugu and Canarese ; are they also found in Tamil ? Four meanings of eZ in Tamil. Traces of il and al used as demonstratives ; their use as negatives XXX CONTENTS PAGE Affiliation of Demonstrative Bases : Extra-Dravidian Affinities . 436-438 North Indian vernaculars ; Scythian languages ; closest analogies in Indo-European languages. New Persian. Affiliation of Interrogative Bases : Extra-Dravidian Relationship . 438 No relationship apparent. Emphatic . . . . . . . .439 Use of this particle. Tamil ; Tulu ; Hebrew ' he paragogic,' &c. Honorific Demonstrative Pronouns ..... 440-441 Canarese and Telugu ; suspicion of Aryan influences. Syntactic Interrogatives 1 and 6 . . . . . 442-443 Particles used for putting inquiries like ' Is there ? ' use of these particles ; 6 instead of a in Malayalam ; 6 generally an expression of doubt. 6 perhaps derived from a ; possible origin of the interroga- tive a from the demonstrative a ; difference in location. Distributive Pronouns ...... 443 How formed. III. Relative Pronouns ...... 444 Noticeable fact that this class of pronouns does not exist in the Dra- vidian languages ; relative participles used instead. PART VI THE VERB ....... 445-560 Remarks on structure of Dravidian verb ; 1. Many roots used either as verbs or nouns ; 2. Formative particles often added to roots ; 3. Structure of verb agglutinative ; 4. Second person singular of impera- tive the shortest form ; 5. But one conjugation and few irregularities ; moods and tenses few ; Tulu and Gond exceptional. Conjugation does not equal that of ancient Scythian verb in simplicity ; Remusat. Antiquity of Tamilian culture ; origin of conjugational forms ; 6. Compounds of verbs with prepositions unknown ; preposition-hke words really nouns. New shades of meaning imparted by gerunds. Section I. — Classification . . . . . 449-481 1. Teansitives and Intransitives . . . . 449-455 Two classes of Dravidian verbs ; Hungarian objective and subjec- tive verbs. Three modes in which i ntransitive verbs are converted into transitives ; 1. By hardening and doubling consonant of formative. Illustrations ; Telugu ; apparent resemblance to Sanskrit. Hebrew dagesh forte ; 2. By doubling and hardening initial consonant of signs of tense ; illustrations. Intransitives sometimes do the same, in Tamil only ; 3. By adding a particle of transition to root ; origin of this particle. 4. By doubling and hardening certain final con- sonants. 2. Causal Verbs ...... 455-462 Causals different from transitives. Indo-European languages here fall behind Dravidian ; double accusatives. Causals formed from transitives ; one and the same causal particle in all the dialects, except CONTENTS XXXI PAGE! Tulu and Gond ; this appears to be i. Explanation of chu in Telugu inchu ; explanation of p of pinchu. Canarese causal particle isu ; identity of Telugu and Canarese particles. Causal particle in Tamil i preceded by v, b, or pp ; origin of these preceding letters. Tamil future tense-signs throw light on those letters ; Tamil future originally an abstract verbal noun. Origin of Dravidian Causal Particle i . . 462 Probably from i, to give. 3. Frequentative Verbs . . . . 462 No peculiarity in their conjugation. 463 463 . 463-467 anguages ; in Dravidian languages 4. Intensive Verb 5. Inceptive Verb 6. The Passive Voice . Passive voice in Indo-European no passive voice, properly so called. How the meaning of the passive is expressed ; 1. It is expressed by the use of the intransitive verb ; 2. By appending auxiliary verbs meaning to become, to go, &c. ; verbal nouns much used in these passives ; third person neuter required ; similar mode in Bengali ; use of active verbs as passives ; relative participial noun. 3. Passive in Gond ; 4. Formed by using the verb ' to eat ' as an auxiliary ; this singular idiom in the Northern vernaculars also ; 5. Much use is made of the auxiliary verb ' to suffer.' This compound rather a phrase than a passive voice. 7. The Middle Voice ...... 467 Only a few traces of such a voice appear. 8. The Negative Voice ...... 468-477 Combination of negative particle with verbal themes a Scythian peculiarity ; forms like Sanskrit ndsti very rare in Indo-European languages ; Dravidian negative verb generally destitute of tenses ; Tulu and Gond exceptions. Rationale of absence of signs of tense ; Tamil pecuUarity. Telugu shows that the negative particle is a ; apparent exceptions. Other dialects ; participial and imperative formatives. Mr. A. D. Campbell, Dr. Stevenson; explanation of Telugu ku and ka. Prohibitive particle in classical Tamil. Gond manni ; resemblance to Tamil min ; explanation of this. Origin of a, the Dravidian Negative Particle . . . 474-476 Not related to alpha privative ; equivalent to al, the particle of negation ; illustrations ; Dr. Gundert. a probably the primitive shape, al the secondary ; Dr. Gundert, al a negative in itself, not merely when followed by a vowel ; illustrations of force of ol and U in Tamil ; prohibitive particles in other languages. 9. Appellative Verbs, or Conjugated Nouns . . 477-480 Appellative compounds in Ugrian languages ; Mordvin. Agree- ment with Dravidian appellative verbs remarkable ; Professor Hun- falvy, illustrations. Telugu appellative verb ; Tamil more highly developed. Adjectives as well as nouns formed into appellatives. XXXll CONTENTS PAGE Section II. — Conjugation al System .... 481-553 Mode of annexing 'pronominal Signs .... 481-486 Pronominal terminations suffixed, not directly to root, but to signs of tense. 1. Personal signs suffixed, not prefixed ; position of pro- noun in old Turanian dialects ; position in Buriat, in Semitic, in modem Indo-European dialects. Position in Malayalam. 2. Dravidian personal signs suffixed, not to root, but to temporal particles ; three elements in every Tamil verb. In Indo-European languages pro- nominal signs not appended to participles : Turkish, BengaU. 3. In Telugu third person sometimes left destitute of conjugational signs ; similar usage in several other languages. 4. Traces in Tamil and Can- arese of very primitive system of conjugation. Dravidian verb^ appears to have been originally uninflected. 5. Distinctions of gender in Dravidian verb peculiarly minute. Formation of the Tenses ..... 486-490 Participles must first be investigated. Verbal Participles, their Signification and Force . . . 487 Verbal participles explained ; name not quite appropriate. 1. Present Verbal Participle ; illustration. 2. Preterite do. do. do. . . . . 487 Sanskrit participle in tvd ; Dravidian participles continuative ; native definition ; Turanian participles ; Mr. Edkins. 1. The Present Terise . . . . ... 488 (1.) How formed in poetical Tamil ; (2.) Tamil and Malayalam seem formerly to have had a present participle ; (3.) Canarese usage ; (4.) Telugu usage. Formation of the Present ..... 490-495 Canarese participle in ut ; Mr. Kittel's explanation ; Old Canarese participle in dap. Mr. Kittel ; Telugu present participle ; Tulu. Sign of present tense in Tamil and Malayalam ; Old Tamil inscrip- tion ; Malayalam form the same, somewhat modified. Which is the more ancient Tamil form, gixn ? or gindru ? Explanation of gindru ; Dr. Graul's " Outlines of Tamil Grammar "; present tense seldom used in Tamil poetry ; Tuda. The Preterite Tense ...... 495 Semitic and Indo-European modes of forming preterite ; Dravidian mode. Use of participles. 1. Formation of Preterite by Reduplication of Final Consonant 496 This mode confined to a small number of verbs ; how it differs from Indo-European reduplication. 2. Formation of Preterite by suffixing Particle or Sign of Past Time ....... 497-507 Each dialect to be examined seriatim, (l.) The Canarese Preterite . . . . .497-500 Signs of past time i ov d ; d the more characteristic. CONTENTS XXXUl PAGE (2.) The Tamil Preterite ...... 500-604 The same signs of time as in Canarese, , , (3.) The Malay dlam Preterite ..... 504 Substantially as in Tamil ; misleading spelling ; in Dr. Gundert's Grammar and Dictionary, and Brigel's " Grammar of Tulu," Lepsius's method adopted. (4.) The Telugu Preterite . ... . . . 505-606 Originally resembled Tamil. (5.) The Tulu Preterite ...... 607 Difference between imperfect and perfect. (6.) Preterites of Minor Dialects . ' . . . . 607 Tuda ; Kota ; Mr. Metz ; Dr. Pope ; Gond. Conclusion ; d, or some modification of it, the most characteristic sign of Dra vidian pre- terite. Origin of the Dra vidian Sign of Past Time . ' . . 607-620 1. Origin of I ....... 507 Originally a vowel of conjunction ; compare Sanskrit and Latin. 2. Origin of B ....... 608-512 Is it remotely connected with Indo-European suffix of passive parti- ciple ? certainly not borrowed from it ; Bengali preterite I ; Max Miiller ; Bopp. New Persian ; modern Teutonic preterite d ; Turkish preterite di or d ; Hungarian d ; Finnish t. May not this sign of the preterite have had its origin in the Dravidian languages them- selves ? Dr. Graul's " Outlines of Tamil Grammar " ; the d of adu, the demonstrative. Explanation of Turkish preterite di ; Max Miiller ; Mongolian gerund in d; Mr. Edkins. 3. The Future Tense ...... 613-517 Difference between formation of preterite and that of future ; two futures ; future the least distinctive tense ; form of the Tamil future surviving in the poets. Ordinary mode of forming the future. Aoristic future in um. Future formed on the basis of the formed verbal theme ; altogether impersonal, u instead of um ; probably the basis of the conjunctive. Future Verbal Participle . . . . . . 617-618 Use of the participle in classical Tamil and Malayalam. Changes in its initial consonant ; Canarese and Telugu aoristic futures. 2. The more Distinctive Future . . . . 519 Telugu and Canarese forms. Affinities of the Sign of the Future . . . . . 619 Bengali future ; Latin fut^jre ; Max Miiller j Ugrian afi&nities ; no affinities reliable. 4. Compound Tenses ...... 520 Mode of formation. XX XIV CONTENTS PAGE The Relative Participle ...... 520-525 Dravidian languages have no relative pronoun ; use a participle in- stead ; how North Indian vernaculars express meaning of relative. Explanation ; suffix of relative participle ; a most largely used ; Can- arese use. Adjectives formed by means of the same suffix. Origin of the Relative Suffixes ..... 623 A possessive case-sign originally ; Manchu illustrates this ; Chinese ; Mr. Edkins ; light thrown on this part of speech by non-Aryan languages of Asia. Use of relative pronoun, in Turkish and Finnish. Formation of Moods ...... 525-539 Properly speaking, only one mood. 1. The Conditional or Subjunctive . . . . . 525-530 Dravidian subjunctive formed by postfixing a particle expressing condition ; two forms in Tulu. Telugu conditionals eni and ^. Ancient Tamil conditional in il or in ; use of dgil. Third form postfixes kdl ; meaning of kdl ; fourth form in dl. al sign of instru- mental case ; origin bi al. 2. The Imperative ...... 530-533 Second person singular imperative identical with root. Impera- tive of transitives differing from that of intransitives ; particles added to imj)erative in Telugu and Tamil ; Canarese imperative. Tamil imperative second person plural ; um, used as a conjunctive and as a continuative ; plural imperative in classical Tamil. Tam. and Mai. in and Old Canarese im identical ; Dr. Gundert ; Gesenius ; Hebrew imperative. 3. The Infinitive . . . . . . . 533-537 The true Dravidian infinitive a verbal noun incapable of being de- clined. Various forms of the infinitive ; Max Miiller's supposition. Formation of infinitive, a alone the normal formative of Dravidian infinitive ; origin of infinitive in ga in classical Tamil. Telugu and Canarese infuiitives. Origin of the Infinitive Suffix a . . . . . 637 Probably identical with a, the demonstrative base ; connection be- tween a and al. Use of the Infinitive . , . . . . . 533 Used in five ways ; illustrations of each. Connection between infinitive and verbal noun in al ; Gond infinitive ; Armenian affinity. Formation op Verbal Nouns ..... 640-663 Two classes of Dravidian verbal nouns — participial and verbal nouns, properly so called. 1. Participial Nouns . . . . . . 540-642 Formation of participial nouns ; neuter singular used in three different significations. Analogy between these nouns and infinitives ; abstract participial nouns in Tamil and Malayalam ; abstract appella- tive nouns. CONTENTS XXXV PAGE 2. Verbal Nouns ...... 542-543 /^ Such nouns express the act, not the abstract ; derivative nouns different from verbal nouns ; illustrations. 3. Derivative Nouns . . . . . 544-562 Various classes ; mode of formation of each class. Four purposes served by the doubling of final consonants ; mode of formation of derivative nouns, continued. Alphabetical list of formatives used in the formation of derivatives, with illustrations. 4. Nouns of Agency ...... 552-553 i, the suffix of Dra vidian nouns of agency, resembles Sanskrit, but not borrowed from it. Adverbs ....... 553 Every Dra vidian adverb either a noun or a verb. Comparative Paradigm of a Dravidian Verb . . . 554r-564 PART VII GLOSSARIAL AFFINITIES 565-624 Comparison of vocables of less importance than comparison of gram- matical forms and structure, but useful when carefully conducted ; testimony of comparative vocabulary as to position occupied by Dravidian tongues. Section I. — Indo-European Affinities . . . . 565-605 1. Indebtedness of Sanskrit to the Dravidian Languages . . 565-579 Extraneous questions to be set aside. Statement of the question at issue ; British words in English ; Greek and Latin in Sanskrit. Six rules for detecting Dravidian words in Sanskrit lexicons. Words 'probably borrowed by Sanskrit from the Dravidian tongues, alphabetically arranged. Names of places not included ; origin of name Malaya. Dr. Gundert's views in Jo^irnal of German Oriental Society. Selections from Dr. Gunderfs list of words, alphabetically arranged. Selections from a list of similar words by Mr. Kittel in the Indian Anti- quary. 2. Sanskrit Affinities ...... 579-587 Words which appear to be the common property of Sanskrit and the Dravidian languages. List of such words alphabetically arranged. 3. Extra-Sanskritic or West Indo-European Affinities . . 587-605 List of words, alphabetically arranged, which appear to bear a closer resemblance to the non-Sanskritic members of the Indo-European family than to Sanskrit. Section II. — Semitic AFFiNmES ..... 605-609 Resemblances between Dravidian and Hebrew words interesting, but scarcely such as to establish relationship. Alphabetical list of such words. XXXVl CONTENTS ' PAGE Section III. — Scythiai^ Affinities . : . . 610-622 These affinities clearly and more direct than Indo-European or Semitic affinities ; vocabularies of the Scythian languages present extraordinary divergences. Alphabetical list of words. Hungarian affinities ; Dr. Gundert's. Chinese, Japanese, and Mongolian affi- nities ; Mr. Edkins's book ; Max Miiller's remarks ; these affinities adduced as aids to inquiry. APPENDIX ....... 625-635 1. Tuda. Information derived from Dr. Pope, Mr. Metz, and Colonel Marshall. Dr. Pope's conclusions respecting Tuda. 2. Kota ; who are the Kotas ? paradigm of pronoun and verb ; resemblance to Ancient Canarese. 3. Gond ; publications by Mr. Driberg and Mr. Dawson ; particulars in which Gond agrees with Telugu and Canarese ; more numerous particulars in which it agrees with Tamil. Particulars in which it takes a course of its own. 4. Ku ; Mr. Latchmaji's Grammar. Note. — Agreements and disagreements with other idioms. 5. RdjmaMl ; list of words defective ; contains Dra vidian element. 6. Ordon ; Mr. Batsch's " Grammar and Vocabulary " ; Oraon more distinctively Dra vidian. Dra vidian words in Oraon. 7. Dravidian element in Brahui. Dr. Bellew's book. Brahui contains many Scythian elements, some distinctively Dravidian. Illustrations of the Dravidian element. Difference between Brahui and languages of the North-Eastern frontier. Index ........ 636-640 J \y BOOKS AND PAPEES bearing on Dravidian Comparative Philology, published subsequently to the first edition OF this work, and quoted or referred to in this edition. Arden. — Progressive Grammar of the Telugu Language. By the Rev. A. H. Arden, M.A. MasuHpatam, and Triibners, London, 1872. Baer. — Historische Fragen mit Hiilfe der Naturwissenschaften Beantwortet Von Dr. Carl Ernst v. Baer. St. Petersburg, 1873. Batsch. — Brief Grammar and Vocabulary of the Oraon Language. By Rev. F. Batsch. Journal of the Bengal Asiatic Society, vol. xxxv. Calcutta. Beames. — Comparative Grammar of the Modern Aryan Languages of India. By J. Beames, Esq., B.C.S. Triibners, London, 1872. Beames. — On the Present Position of Old Hindi in Oriental Philology. By J. Beames, Esq., B.C.S. Indian Antiquary for January 1872. Bombay, and Triibners, London. Beames. — Kirtans ; or Hymns from the Earliest Bengali Poets. By J. Beames, Esq., B.C.S. Indian Antiquary for November 1872. Bombay, and Triibners, London. Bellew. — From the Indus to the Tigris (including a Grammar and Vocabulary of the Brahui Language). By Dr. Bellew. Triibners, London, 1873. Bleek. — Comparative Grammar of the South African Languages. By W. H. J. Bleek, Esq., Ph.D. Triibners, London, 1862. Bleek. — On the Position of the Australian Languages. By W. H. J. Bleek, Esq., Ph.D. Journal of the Anthropological Society. London, 1871. Bower. — On the Tamil Language and Literature. By the Rev. H. Bower, D.D. Calcutta Review, vol. xxv. Bower. — Lecture on Auveyar, a Tamil female poet. By the Rev. H. Bower, D.D. Madras. Brigel. — Tulu Grammar. By the Rev. J. Brigel. Mangalore, 1872. Buehler. — On the Origin of Sanskrit Linguals. By Dr. George Buehler. Madras Journal of Literature, July 1864. Burnell. — An interesting passage in Kamarila-Bhatta's Tantravarttika. By A. C. Burnell, Esq., Ph.D., M.C.S. Indian Antiquary for October 1872. Burnell. — The Oldest-known South Indian Alphabet. By A. C. Burnell, Esq., Ph.D., M.C.S. Indian Antiquary for November 1872. Burnell. — Specimen of South Indian Dialects : 1. Konkani ; 2. Coorg {Kodagu) ; Mappila Malayalam. Mangalore, 1872. (In progress.) Caldwell. — On the Substitution of the Roman for the Indian Characters. By the Rev. Dr. Caldw|;ll. Madras Journal of Lit.ratur,i for 1858-9. Campbell. — Ethnology of India. By Sir George Campbell, K.C.S.I. Journal of the Bengal Asiatic Society, vol. xxxv. Campbell. — Specimens of Languages of India. By Sir George Campbell, K.C.S.I. Calcutta, 1874. XXXVlll BOOKS AND PAPERS Chitty. — The Tamil Plutarch. By Simon Casie Chitty, Esq., Jaffna. Ceylon, 1859. - Cole. — Coorg Grammar. By Major R. A. Cole. Bangalore, 1867. . Cunningham. — The Ancient Geography of India. By General Alexander Cunningham. London, 1871. ■Dalton. — Descriptive Ethnology of Bengal. By Colonel E. T. Dalton, C.S.I. Calcutta, 1872. Davids. — Conquest of South India in the Twelfth Century, by Parakrama Baliu the Great, king of Ceylon. By T. W. Rhys Davids, Esq. Journal of the Bengal Asiatic Society, 1872. N. .. -■ Dawson. — Brief Grammar and Vocabulary of the Gond Language. By the Rev. J. Dawson. Journal of the Bengal Asiatic Society, 1871. Duff, Sir M. E. Grant. — Cojitemporary Review, September 1891. Edkins. — China's Place in Philology. By the Rev. Joseph Edkins, D.D. Peking, 1870. Eggeling. — On the Chera and Chalukya Dynasties, A paper read at the Inter- national Congress of OrientaHsts by Dr. Eggeling, Secretary of the Royal Asiatic Society, London, 1874. 'Frye. — On the Uriya and Khond Population of Orissa. By Lieut. J. P. Feye. Journal of the Royal Asiatic Society. London, 1872. Gait, E. A., CLE., Census Commisioner of India. Gazetteer. — The Central Provinces Gazetteer. Nagpur, second edition, 1870. J Cover. — The Folk-Songs of Southern India. By C. E. Gover, Esq. Madras, 1871. Graeter. — Coorg Songs, with Outlines of Coorg Grammar. By Rev. A. Geaeter. Mangalore, 1870. Graul. — Outlines of Tamil Grammar. By the Rev. C. Geaul. D.D. Leipzig, 1856. Graul. — Der Kural des Tiruvalluvar. By the Rev. C. Geaul, D.D. Leipzig, 1856. Graul. — Reise nach Ostindien. By the Rev. C. Geaul, D.D, Three vols. Leipzig, 1856. 'Growse. — On the Non-Aryap Element in Hindi Speech. By F. S. Geowse, Esq., M.A., B.C.S. Indian Antiquary for April 1872. Gundert. — Malayalam Grammar. By the Rev. H. Gundeet, Ph.D. Second edition. Mangalore, 1868. ^Gundert. — Malayalam Dictionary. By the Rev. H. Gundeet, Ph.D. Man- galore, 1872. « Gundert. — On the Dravidian Elements in Sanskrit. By the Rev. H. Gundeet, Ph.D. Journal of the German Oriental Society for 1869. Hislop. — Papers relating to the Aboriginal Tribes of the Central Provinces, left in MS. by the late Rev. S. Hislop. Edited, with Notes, by Sir RiCHAED Temple. Nagpur, 1866. Hodgson. — Essays on the Languages, Literature, and Religion of Nepal, Tibet, and adjacent Countries. By Beian Hodgson, Esq., late British Resident, Nepal. Hodson. — Canarese Grammar. By the Rev. T. Hodson. Second edition. Banga- lore, 1864. Munfalvy. — On the Study of the Turanian Languages. A paper read at the International Congress of Orientalists, London, 1874, by Professor Hunfalvy. W BOOKS AND PAPERS XXXIX ^Hunter. — Comparative Dictionary of the Non-Aryan Languages of India and High Asia. By W. W. Hunteb, LL.D., B.C.S. Triibners, London, 1868. Kennet.—^oteB, on Early-printed Tamil Books. By the Rev. C. E. Kennet. Indian Antiquary for June 1873. Kittel. — On the Dravidian Element in Sanskrit Dictionaries. By the Rev. F. KiTTEL. Indian Antiquary for August 1872. i^< Kittel. — Notes concerning the Numerals of the Ancient Dra vidians. By the Rev. F. Kittel. Indian Antiquary for January 1873. -^Kittel. — Kesiraja's Jewel Mirror of Grammar (Sabda mani darpana), a Grammar of Ancient Canarese. By the Rev. F. Kittel. Mangalore, 1872. Kittel. — Article on Old Canarese Literature, by the Rev. F. Kittel, in Indian Antiquary for January 1875. ' Koelle. — Grammar of the Bomu or Kanuri Language. By the Rev. S. W. Koelle, London. , Marshall. — A Phrenologist among the Tudas. By Lieut. -Col. Maeshall. Triibners, London, 1873. ' Metz. — The tribes inhabiting the Neilgherry Hills. By the Rev. F. Metz. Second edition. Mangalore, 1864. Jf%ir.— Original Sanskrit Texts. By J. Mum, Esq., D.C.L., LL.D., late B.C.S. Second edition, five vols. Triibners, London, 1868. Mueller. — Reise der Fregatte Novara, Linguistischer Theil. By Professor Friedeich Muellee. Vienna, 1868. Mueller. — Lectures on the Science of Language. By Professor Max Muellee. Two vols. London, 1864. Murdoch. — Classified Catalogue of Tamil-printed Books, with introductory notices. By J. Muedoch, Esq., LL.D. Madras, 1865. Nelson. — The Madura Country. A Manual compiled by order of the Madras Government. By J. H. Nelson, Esq., M.A., M.C.S. Madras, 1868. —Phillips. — TumuH in the Salem District. By the Rev. Maueice Phillips. Indian Antiquary, 1873. Pope. — A Larger Grammar of the Tamil Language in both dialects, with the V^ Nanniil and other native authorities. By the Rev. G. U. Pope, D.D. Second edition. Madras, 1859. Pope. — Tamil Handbook. By the Rev. G. U. Pope, D.D. Second edition. Madras, 1859. Pope. — One Alphabet for all India. By the Rev. G. U. Pope, D.D. Madras, 1859. Pope. — The Sermon on the Mount, in English, Tamil, Malayalam, Canarese, and Telugu, in the Roman character. By the Rev. G. U. Pope, D.D. Madras, 1860. Pope. — Outlines of the Grammar of the Tuda Language. By the Rev. G. U. Pope, D.D., included in Colonel Marshall's " Phrenologist among the Tudas." Trubners, London, 1873. 'Priaulx. — India and Rome. Indian Travels of Apollonius of Tyana, and the Indian Embassi^ to Rome. By P. de B. Peiaulx, Esq. London, 1873. Prinsep. — Essays on Indian Antiquities. By the late J. T. Peinsep, Esq., B.C.S. Edited with Notes by Edwaed Thomas, Esq., F.R.S., late B.C.S. Two vols. London, 1858. xl BOOKS AND PAPERS ' Sanderson. — Canarese Dictiopary, by the Rev. W. Reeves. Revised and en- larged by the Rev. D. S. Sanderson. Bangalore, 1858. Sayce. — Principles of Comparative Philology. By A. H. Sayce. London, ^ Trubner & Co., 1874. . Quatrefages. — Etude sur les Todas. Par M. de Quatbefages de Br^au. Journal des Savants, December 1873-January 1874. Paris. Tdrandtha's History of the Propagation of Buddhism in India ; Tibetan and German. St. Petersburg, 1870. ^Tennent. — Ceylon. By Sir Emerson Tennent. Two vols. London, 1860. 'Tickell. — Brief Grammar and Vocabulary of the Ho, a Kolarian Language. ^ By Colonel Tickell. Journal of the Bengal Asiatic Society, vol. XXXV. Travancore. — Inscriptions in Tinnevelly and South Travancore. By His Highness Rama Varma, First Prince of Travancore. Indian Anti- quary for December 1873. 'Trumpp. — Grammar of the Sindhi Language. By Dr. Ernest Trumpp. Triib- ners, London, 1872. Yule. — Marco Polo, newly translated and edited with Notes, by Lieut.-Col. H. Yule, C.B. Two vols. London, 1871. Tw^e.— Cathay and the Way Thither. By Lieut.-Col. H. Yule, C.B. Hakluyt Society, London, 1866. Yule. — Map of Ancient India, with accompanying Memoir, in Dr. Wm. Smith's Atlas of Ancient Classical Geography. London, 1875. Weber. — Indian Pronunciation of Greek, and Greek Pronunciation of Hindu Words. By Dr. A. Weber. Translated by E. Rehatsek, Esq. Indian Antiquary for May 1873. Williams. — Application of the Roman Alphabet to the Languages of India. By Professor Monieb Williams. London, 1859. y^ i Winslow. — Tamil Dictionary. Completed and edited by the Rev. M. Winslow, D.D. Madras, 1862. DRAVIDIAN COMPARATIVE GRAMMAR INTEODUCTION It is the object of the following work to examine and compare the grammatical principles and forms of the various Dra vidian languages, in the hope of contributing to a more thorough knowledge of their primitive structure and distinctive character. In pursuing this object, it will be the writer's endeavour to point out everything which appears likely to throw any light on the question of the relation which this family of languages bears to the principal families or groups into which the languages of Europe and Asia have been divided. Whilst the grammatical structure of each Dra vidian language and dialect will be investigated and illustrated in a greater or less degree, in proportion to its importance and to the writer's acquaintance with it, it will be his special and constant aim to throw light upon the structure of Tamil — a language which he has for more than thirty- seven years studied and used in the prosecution of his missionary labours, and which is probably the earliest cultivated, and most highly developed, of the Dra vidian languages — in many respects the representative language of the family. The idioms which are included in this word under the general term ' Dra vidian,' constitute the vernacular speech of the great majority of the inhabitants of Southern India. With the exception of Orissa, and those districts of Western India and the Dekhan in which Gujarat! and Marathi are spoken, the whole of the peninsular portion of Indii. from the Vindhya mountains and the river Nerbudda (Narmada) to Cape Comorin (Kumari)^is peopled, and from the earliest period appears to have been peopled, by different branches of one and the same race, speaking different dialects of one and the same language — the language to which the term ' Dravidian ' is here applied ; and 2 INTRODUCTION scattered offshoots from the same stem may be traced still farther north, as far as the Eajmahal hills in Bengal, and even as far as the mountain fastnesses of Beluchistan. / Gujarati, Marathi (with its offshoot, Konkani), and Oriya, the Panguage of Odra-desa, or Orissa, idioms which are derived from 'Hhe decomposition of Sanskrit, form the vernacular speech of the Hindu population in the peninsular portion of India within their respective limits : besides which, and besides the Dravidian lan- guages, various idioms which cannot be termed indigenous or verna- cular are spoken or occasionally used by particular classes resident in Peninsular India. \ Sanskrit, though it is improbable that it ever was the vernacular language of any district or country, whether in the north or in the south, is in every southern district read, and to some extent under- stood, by the Brahmans — the descendants of those Brahmanical I colonists of early times to whom the Dra vidians appear to have been i indebted to some extent for the higher arts of life and a considerable I portion of their literary culture. Such of the Brahmans as not only ) retain the name, but also discharge the functions of the priesthood, >and devote themselves to professional studies, are generally able to understand and interpret Sanskrit writings, though the vernacular language of the district in which they reside is that which they use in their families, and with which they are most familiar. They are styled, with reference to the language of their adopted district; Dravida Brahmans, Andhra Brahmans, Karnataka Brahmans, &c. ; and the Brahmans of the several language-districts have virtually become distinct castes ; but they are all undoubtedly descended from one and the same stock, and Sanskrit, though now regarded only as an accomplishment or as a professional acquirement, is properly the literary dialect of their ancestral tongue. Hindustani is the distinctive language of the Muhammedan portion of the population in the Dekhan — most of which consists, of the descendants of those warlike Patans, or Afghans, and other Mu- hammedans from Northern India by whom most of the peninsula was overrun some centuries ago. It may almost be regarded as the vernacular language in some parts of the Hyderabad country ; but generally throughout Southern India the middle and lower classes of the Muhammedans make as much use of the language of the dis- trict in which they reside as of their ancestral tongue, if not more. Hindustani was never the ancestral language of the class of southern Muhammedans, generally called by the English ' Lubbais,' but by SOUTH INDIAN VERNACULARS 3 Indians on the eastern coast S'onagas (Yavanas), and by those on the western coast Mappillas. These are descendants of Arab merchants and their native converts, and speak Tamil or Malayalam. Hebrew is used by the small colony of Jews resident in Cochin and the neighbourhood, in the same manner and for the same pur- poses as Sanskrit is used by the Brahmans. Gujarati and Marathi are spoken by the Gujarati bankers and the Parsi shopkeepers who reside in the principal towns in the peninsula. The mixed race of ' country-born ' Portuguese are rapidly forgetting (except in the territory of Goa itself) the corrupt Portuguese which their fathers and mothers were accustomed to speak, and learning EngHsh instead ; whilst French still retains its place as the language of the French employes and their descendants in the settlements of Pondicherry (Puduchcheri), Carrical (Kareikkal), and Mahe (Mayjruri), which still belong to France. Throughout the British territories in India, English is not only the language of the governing race, and of its ' East-Indian,' Eurasian, or ' Indo-British ' offshoot, but is also used to a considerable and rapidly increasing extent by the natives of the country in the administration of justice and in commerce ; and in the Presidency of Madras and the principal towns it has already won its way to the position which was formerly occupied by Sanskrit as the vehicle of all higher learning. Neither English, however, nor any other foreign tongue appears to have the slightest chance of becoming the ver- nacular speech of any portion of the inhabitants of Southern India. The indigenous Dravidian languages, which have maintained their ground for more than two thousand years against Sanskrit, the language of a numerous, powerful, and venerated sacerdotal race, may be expected successfully to resist the encroachments of every other tongue. Use of the Common Term ' Dravidian.' I have designated the languages now to be subjected to comparison by a common term, because of the essential and distinctive gram- matical characteristics which they all possess in common, and in virtue of which, joined to the possession in common of a large number of roots of primary importance, they justly claim to be considered as springing from a common origin, and as forming a distinct family of tongues. This family was at one time styled by European writers ' Tamulian * 4 INTRODUCTION or * Talmulic ' ; but as Tamil is the oldest and most highly cultivated member of the family, and that which contains the largest proportion of the family inheritance of forms and roots, and as it is desirable to reserve the terms ' Tamil ' and ' Tamilian ' (or as they used sometimes to be erroneously written ' Tamul ' and ' Tamulian ') to denote the Tamil language itself and the people by whom it is spoken, I have preferred to designate this entire family by a term which is capable of a wider application. One of the earliest terms used in Sanskrit to designate the family seems to have been that of Andhra-Drdvida-bhdshd, ' the Telugu- Tamil language,' or rather, perhaps, ' the language of the Telugu and Tamil countries.' This term is used by Kumarila-bhatta, a con- troversial Brahman writer of eminence, who is supposed to have lived at the end of the seventh century a.d. ; and, though vague, it is not badly chosen, Telugu and Tamil being the dialects spoken by the largest number of people in Southern India. Canarese was probably supposed to be included in Telugu and Malayalam in Tamil ; and yet both dialects, together with any sub-dialects that might be included in them, were evidently regarded as forming but one hhdshd (language). The word I have chosen is ' Dra vidian,' from Dravida, the adjecti- val form of Dravida. This term, it is true, has sometimes been used, and is still sometimes used, in almost as restricted a sense as that of Tamil itself, so that though on the whole it is the best term I can find, I admit that it is not perfectly free from ambiguity. It is a term, however, which has already been used more or less distinctively by Sanskrit philologists, as a generic appellation for the South Indian peoples and their languages, and it is the only single term they seem ever to have used in this manner. I have, therefore, no doubt of the propriety of adopting it. Manu says (x. 43, 44) : " The following tribes of Kshatriyas have gradually sunk into the state of Vrishalas (outcasts), from the extinction of sacred rites and from having no communication with Brahmans, viz. — Paundrakas, Odras, Dravidas, Kambojas, Yavanas, S'akas, Paradas, Pahlavas, Chinas, Kiratas, Daradas, and Kha^as." Of the tribes here mentioned the only tribe belonging to Southern India is that of the Dravidas. This name, therefore, appears to have been supposed to denote the whole of the South Indian tribes. If any of those tribes were not intended to be included, it would pro- bably be the Andhras, the Telugus of the interior, who had already been mentioned byname in the Aitareya Brahmana, and classed with USE OF THE COMMON TERM DRAVIDIAN 5 Pundras, Sabaras, and Pulindas, as degraded descendants of Visva- mitra. The same statement is made in the Maha-bharata ; and in the two lists of degraded Kshatriyas therein given, the Dravidas are the only South Indian tribe mentioned. It must be concluded, therefore, that the term is generically used, seeing that the more specific names of Pandyas, Cholas, &c., had become well known in Northern India by that time. Doubtless it is in the same sense that Satyavrata, the Indian Noah, is called in the Bhagavata Purana ' the lord of Dravida ' (Muir's " Sanskrit Texts," vol. i.).* The more distinctively philological writers of a later period used the term Dravida in what, appears to be substantially the same sense as that in which I propose that it should be used. The principal Prakrits — that is, colloquial dialects — of ancient India were the Maharashtri, the Sauraseni, and the Magadhi. Amongst minor or less-known Prakrit dialects the Dravidi, or language of the Dravidas, was included. A Sanskrit philologist quoted by Muir (vol. ii. 46) speaks of the language of Dravida as a vibhdshd, or minor Prakrit ; and another (p. 50) speaks of ' the language proper to l!)ravidas ' (in which persons of that race should be represented as speaking in dramas) as the Dravidi. It is evident that we have here to under- stand not the Tamil alone, or any other South Indian language alone, but the Dra vidian languages generally, supposed in a vague manner by North Indian writers to constitute only one tongue. This lan- guage of the Dravidas was evidently included in what was called the Paisachi Prakrit, a name which appears to have been applied promiscuously to a great number of provincial dialects, including dialects so widely differing from one another as ' the language of the Pandyas ' (Tamil), and ' that of the Bhotas ' (Tibetan). The only property these languages can have possessed in common must have been the contempt in which they were held by Brahman philologists, in virtue of which it must have been that they were styled also Paisachi, the language of fisdchas, or demons. The more accurate term Dravidi has continued to be used occasionally by northern scholars up to our own time. As late as 1854, the learned Hindu philologist Babu Rajendra Lai Mitra (quoted by Muir, vol. ii. 127), speaks of the ' Dravidi ' as one of the recognised Prakrits, equally with the Sauraseni, and as being, like it, the parent of some of the * The tradition is recorded in the ancient Tamil classics which speaks of a large continent which once existed contiguous to Southern India, and which was submerged hy the ocean during a certain inundation not far removed from human recollection. — The Tamilian Anti'juarij. 6 INTRODUCTION present vernaculars of India. It thus appears that the word ' Dravida,' from which the term ' Dra vidian ' has been formed, though sometimes used in a restricted sense, as equivalent to Tamil, is better fitted, notwithstanding, for use as a generic term ; inasmuch as it not only has the advantage of being more remote from ordinary usage, and somewhat more vague, but has also the further and special advantage of having already been occasionally used by Indian philologists in a generic sense. By the adoption of this term ' Dra vidian,' the word ' Tamilian ' has been left free to signify that which is distinctively Tamil. Enumeration of Dravidian Languages. The idioms which I designate as ' Dravidian ' are twelve in number, exclusive of the Brahui. They are as follows : — 1. Cultivated Dialects. 1. Tamil .^ 2. Malayalani. 3. Tehicru." 4. Canarcse. 6. Tulu. 6. Kudai'u or Cooi". 2. Uncultivated Dialects. 1. Tada. 2. Kota. 3. Gond. 4. Kliond or Ku. 5. Oraon. 6. Rajmahal. I. Tamil. — This language being probably the earhest cultivated of all the Dravidian idioms, the most copious, and that which contains the largest portion and the richest variety of indubitably ancient forms, it is deservedly placed at the head of the list. It includes two dialects, the classical and the colloquial, or the ancient and the modern, called respectively the ' S'en-Damir ' and the ' Kodun- Damir,' which differ one from the other so widely that they might almost be regarded as different languages. The Tamil language is spoken throughout the vast plain of the Carnatic, or country below the Ghauts, from Pulicat to Cape Comorin, and from the Ghauts, or central mountain range of Southern India, to the Bay of Bengal. It is also spoken in the southern part of the Travancore country on the western side of the Ghauts, from Cape Comorin to the neighbourhood of Trivandrum ; and in the northern and north-western parts of Ceylon, where Tamilians commenced to form settlements prior even to the Christian era, and from whence they have gradually thrust out the Singhalese. All throughout Ceylon the coolies in the coffee ENUMERATION OF DRA VIDIAN LANGUAGES 7 plantations are Tamilians ; the majority of the money-making classes even in Colombo are Tamilians ; and it seems not unlikely that ere long the Tamilians will have excluded the Singhalese from almost every profitable employment in their own island. The majority of the domestic servants of Europeans and of the camp- followers in every part of the Presidency of Madras being Tamil people, Tamil is the prevailing language in all the military canton- ments in Southern India, whatever be the vernacular language of the district. Hence, at Cannanore in the Malayalam country, at Bangalore in the Canarese country, at Bellary in the Telugu country, and at Secunderabad, where Hindustani may be considered as the vernacular, the language which most frequently meets the ear in the bazaars is Tamil. The majority of the Klings (Kalingas), or Hindus, who are found in Pegu, Penang, Singapore, and other places in the further east, are Tamilians : a large proportion of the coolies who have emigrated in such numbers to the Mauritius and to the West Indian colonies are Tamilians ; in short, wherever money is to be made, wherever a more apathetic or a more aristocratic people is waiting to be pushed aside, thither swarm the Tamilians, the Greeks or Scotch of the east, the least superstitious and the most enterprising and persevering race of Hindiis. Including Tamilians resident in military stations and distant colonies, and the Tamilian inhabitants of South Travan- core, and Northern Ceylon, and excluding not only Muhammedans, &c., but also people of Telugu origin who are resident in the Tamil country, and who form probably ten per cent of the whole popula tion, the people who speak the Tamil language may be estimated at about twenty-one millions. Madras, the chief city in the Tamil country, is also the chief city in the South Indian Presidency. The name by which it is known amongst natives everywhere is, not Madras, but Chennappa- pattanam, abbreviated into Chenna-pattanam, a name which it derived from Chennappa Nayakkar, father-in-law of the Nayakkar of Chinglepat, a petty local chieftain, a feudatory of the Chandragiri Kaja, from whom the English obtained possession of a little fort on the coast which they converted into a fortified factory. The origin of the name by which it appears always to have been called by Europeans — Madras (ofhcklly Madraspatam) — has never been made out with certainty. Perhaps the most probable derivation is from the Telugu niaduru (Tamil madil), the surrounding wall of a fort, a rampart. There is a neighbouring town, Sadras, originally a Dutch 8 INTRODUCTION settlement, tlie name of which closely resembles Madras. Sadras is an European corruption from Sadurei, which is an abbreviation of Sadurangam (=Sans. Chaturanga), the four constituent arms of an army. I have not been able to discover any authority for the statement sometimes made that Madras is derived from Mandraj- pattanam. The proper spelling of the name Tamil is Tamir, but through the change of r into I it is often pronounced Tamil ; and is often (though erroneously) written Tamul by Europeans. Tamul is the mode of writing the name which appears to have been introduced by the French ; but the name given to the language by the Portuguese, and by which it was generally known amongst the earlier Europeans, was neither Tamul nor Tamil, but ' the Malabar ' — a name founded on a misapprehension.* Colebrooke, though writing in Northern India, was aware of the identity of Malabar with Tamil. He says (" Essay on the Sanskrit and Prakrit Languages "), " The language of the province is the Tamel, to which Europeans have given the name of Malabar." The identity, however, of the two languages was known at a much earlier date to persons who had the opportunity of acquiring local know- ledge. In the very first book ever printed in Tamil characters — at Ambalakkadu, on the Malabar coast, in 1577 or 9 — the language of the book is styled ' Malavar or Tamul.' The writer apparently regarded Tamil as the more correct word. See " Sounds : Alphabet." The Sanskrit name corresponding to Tamil is Dravida, a word which denotes both the country inhabited by the people called Dravidas and the language spoken by them ; and I have come to the conclusion that the words Tamir and Dravida, though they seem to differ a good deal, are identical in origin. Supposing them to be one and the same word, it will be found much easier to derive Tamir from Dravida than Dravida from Tamir. It might naturally seem improbable at the outset that a Dravidian people residing in the extreme south should call themselves and be called by their neigh- bours, not by a Dravidian, but by a Sanskrit name ; but Pandya, the name of the southernmost portion of the Dra vidians, is Sanskrit, and a similar peculiarity meets us with regard to almost all the names of the South Indian peoples — Cholas, Keralas, Andhras, Kalingas, &c. — which, so far as is known at present, are Sanskrit, not Dra- * But compare the title of 'The Chief of the Malabars (Tamils),' conferred on the nephew and successor of Ananda Runga Pillai, Dupleix's Dewana at Pondi- cherry, by King Louis XVI of France in 1766. — Editors. ENUMERATION OF DRAVIDIAN LANGUAGES 9 vidian. The name Karnataka alone appears to have a Dra vidian origin. If the other names were originally Dravidian, as this seems to have been, and as it might naturally be supposed they all must have been, their original shape and root-meaning have disappeared. What adds to the difficulty is, that though these words have a place in Sanskrit dictionaries and are accepted as Sanskrit by the Dra vidian people, Sanskrit fails as completely as the Dravidian languages to furnish us with a clue to their original meaning. When we have traced them back to Sanskrit we are obliged to leave them there. The name Andhra appears, as has already been mentioned, in one of the Brahmanas, but, like most of the Vedic proper names, it is incapable of explanation. May it not be, indeed, that those proper names belonged originally to some old North Indian vernacular — some prse- Aryan, though not necessarily non- Aryan — speech, which had disappeared before the literary history of Sanskrit commenced. If this were the case, it would be in vain to expect the derivation of such words as Dravida to be cleared up now. The compound dr is quite un-Dravidian. It would be tira in Tamil ; but even if we suppose some such word as Tiravida or Tiramida to have been con- verted into Dravida by the Sanskrit-speaking people, we get no nearer to an explanation of the original meaning of the word'. The oldest form of Dravida — or, at least, the form which appears to have been most widely in use — appears to have been Dramida ; and this is the first step towards identifying the two words, Dravida and Tamir. Both forms of the word are known in Tamil, but Dramida (written Tiramida) is preferred by the classics, and is placed first in ancient Tamil vocabularies. In Varaha-mihira's Brihat- samhita, according to Dr Kern, some manuscripts give Dramida, instead of Dravida. Through the change of d into I, the Dravidas are called Dramilas in Taranatha's Tibetan " History of the Propagation of Buddhism in India " (a.d. 1573), and Dr Gundert informs me that this is the form in which the word occurs again and again in the old Malayalam versions of the Puranas. In the Pali of the Mahawanso the form used is Damilo, the derivative of which is Damilo ; and as initial d becomes t by rule in Tamil, we now reach the ordinary Tamil mode of writing the word, Tamir or Tamil. Each of the changes that have taken place is in accordance with a -recognised Dravidian law of sound. Initial dr is alw%ys softened in the Prakrits into d — e.g., droha becomes doho. In the same manner sr becomes s, an example of which we have in the word S'ramana, a Buddhist or Jaina ascetic, which in Tamil has become Samana (in Pali, Sammana ; in the lO INTRODUCTION Greek of Clemens Alexandrinus the plural is 2tt/xavaiot and ^efivoi (Samanaiyoi or Semnoi) ). The change of v into m or of m into v, even in Sanskrit itself, is seen in such words as dhmdnksha, Sans, a crow, instead of dhvdnksha, and especially in the affixes mat and vat, nian and van, min and vin. Perhaps the most considerable change is from d in Dravida to r in Tamir ; but this also is quite in accord- ance with usage, as will appear in the chapter on " Sounds." Com- pare nddi. Sans, a measure, with the Tam.-Mal. yidri or ndli. A good illustration of this change is furnished by the name of one of the nations included under the general name of Tamil — viz., that of the Cholas. This name in the Sanskrit of one of Anoka's inscriptions is Choda, in ordinary Sanskrit Chola, in Tamil S'ora, in Telugu Chola. In Telugu inscriptions it is often Choda as in Asoka's. The change of d to tin the beginning of a word is unavoidable in Tamil, but we have a reminiscence perhaps of the original sound in the name given to the language by the first Danish missionaries — viz.. Lingua Damulica. In the Indian segment of the very interesting set of Roman maps, called, from the name of the discoverer, the Peutinger Tables — (this segment at least seems to me anterior to Ptolemy's Geography) — we find a considerable portion of the country covered by two names — Andre Indi and Damirice. We can scarcely err in identifying these names with the Telugu and Tamil countries — the languages of which were called, as we have seen, by Kumarila-bhatta, some centuries later, the languages of the Andhras and Dravidas. If so, the earliest appearance of the name Tamil in any foreign document will be found also to be most perfectly in accordance with the native Tamil mode of spelling the name. Damirice evidently means Damir-ice. Com- pare the ApLOLKi] {Aridhi) of Ptolemy and the Aryaka of Varaha- mihira. In another place in the same map a district is called Scytia Dymirice ; and it appears to have been this word which, by a mistake of A for A, Ptolemy wrote Avjupiia] {Lumirihi). The D retains its place, however, in the Cosmography of the anonymous geographer of Ravenna, who repeatedly mentions Dimirica as one of the three divisions of India, and the one farthest to the east. He shows also that the Tamil country must have been meant by the name, by mentioning Modura as one of the cities it contained. There can be little doubt that the name Tamil may also be identified with the Tchi-mo-lo of Hwen Thsang, a word which may also be read Dimala or even Dimara. It is remarkable that native Tamil scholars, though generally willing enough to trace every word to a Sanskrit origin, have failed ENUMERATION OF DRAVIDIAN LANGUAGES II to see in Tamir — or Tamira, as it is also sometimes written — a tadbhava of Dravida or Dramida, and have invented for the name of their language (like their neighbours the Telugu people — though perhaps with less reason), the meaning of ' sweetness or fragrance ' — a meaning of the word Tamir which has nothing to support or com- mend it, but its agreement with the estimate formed by the Tamilians of the euphoniousness of their native tongue. I accept their estimate rof their language as in the main correct, but cannot accept their derivation of the word. A discussion respecting the origin of the word Tamil would not be complete without some reference to the names of the three great subdivisions into which the Tamil people were divided in ancient times — Cheras, Cholas, and Pandyas. The arrangement of the names is climatic, and denotes that the Pandyas were supposed in those times to have the pre-eminence — a supposition which appears to be in accordance with the facts of the case. Pandya. — The Singhalese traditions preserved in the Mahawanso represent Vijaya, the first sovereign of Ceylon, as marrying a daughter of the Pandya king, in consequence of which his son was called Panduvamsadeva. Arjuna also, one of the five Panda va brothers, is related in the Maha-bharata to have married a daughter of the king of the Pandyas in the course of his many wanderings. There is no certainty in these traditions ; but it is certain that about the time of Pliny and the Periplus a portion of the Malabar coast was ruled over by the Pandyas, a proof that their power had considerably extended itself from its original seats ; and I regard it as nearly certain that the Indian king who sent an embassy to Augustus* was not Porus, but Pandion — i.e., the king of the Pandyas, called in Tamil Pandiyan. If this be admitted, it is an interesting proof of the advanced social position occupied by the Pandyas — (probably in consequence of the foreign trade they carried on in connection with their settlements on the Malabar coast) — that after * Compare the following by Sir M. E. Grant Duff in the "Contemporary Review" of September, 1891: "There is a coin in the Madras Museum, the finding of which was I tliink one of the most interesting epigrams of events with which I am acquainted. It is none other than the Aureus of Claudius, which was struck to commemorate the conquest of Britain, and it was found in the Madura District. ... I myself possess a coin of the Arian Emperor Valens, which was found in the Vaigai river in the same district. Such unlooked for links between ideas and associations separated by half the world are very curious. " — Editors. 1 2 INTRODUCTION the termination of the political relations that subsisted between the successors of Alexander and the princes of Northern India, the Pandyas were the only Indian princes who perceived the advantages of an European alliance. The Sanskrit Pandya is written in Tamil Pandiya, but the more completely Tamilised form Pandi is still more commonly used all over Southern India. I derive Pdndi, not from the Tamil and Malayajam pandu, ancient, though that is a very tempting deriva- tion, but — as native scholars always derive the word — from the Sanskrit Pandu, the name of the father of the Pandava brothers. This very form Pandya, in the sense of a descendant of Pandu, is mentioned, as I am informed by Professor Max Miiller, by Katya- yana, the immediate successor of Panini. The second and most celebrated capital of the Pandyas-— (the first was Kolkei on the Tamraparni) — was Madurei, in English Madura, which is the Tamil mode of writing Mathura [the Muttra of our maps, and the MeOopa {Methora) of the Greeks] the name of the city which remained in the possession of the Panda vas at the conclusion of the great war. The Madura of the Pandyas is appropriately called in the Hariva?>isa, ' the Southern Mathura.' There is another (Matura) in Ceylon, and a fourth (Madura) in the Eastern Archipelago. The Singhalese annalists in the Mahawanso call the king of the Pandyas sometimes Pandyava, sometimes Pandu ; and this shows that there cannot be any doubt of the connection of the name of the Pandyas with that of the heroes of the great war, though the origin and nature of that connection cannot now be ascertained. Pandya must at first have been the name of the ruHng family only. Its extension to the people followed the course which dynastic names have often taken in other parts of the world. Megasthenes speaks of a country in India w^hich was called IlavSatr; {Pandau), after the name of the only daughter of the Indian Hercules — that is, of Krishna. I have no doubt that the country referred to was that of the Pandyas. A writer who had heard of the Andarse and Calinga3 could not but have heard of the Pandyas also. He partly, it is true, misapprehended the legends related to him ; but he was right in deriving the name of the Pandya country from the name of its rulers, and in connecting their name — in some fashion, however erroneously — with mythological heroes and heroines. The myth really current at that time — if we may suppose the substance of the Mahabharata in its present shape then in existence — was that Arjuna, one of the Pandava brothers and Krishna's chief friend, had in the course of his wanderings in the ENUMERATION OF DRAVIDIAN LANGUAGES . 1 3 south married a daughter of the king of the Pandyas. Everything related by Megasthenes respecting this country, especially the state- ment that it was there that pearls were procured, serves to identify it with the Pandya country. Pliny, apparently following another passage of Megasthenes, enumerates amongst the Indian nations a nation called Pandae. It is not clear where he supposed their country was situated, but we cannot doubt that the Pandyas of Madura, wherever he thought they were located, were the people referred to. His statement that the Pandse alone amongst Indian nations were ruled by women, though not correct (so far as is now known), if supposed to relate to the Pandyas of Madura, may be regarded as sufficiently applicable to the peculiar social usages of the Malabar coast, where almost every inheritance still runs in the female line, and where, in Pliny's own times at least, if not also in those of Megasthenes, the Pandyas of Madura had colonies. Pliny expressly mentions that a portion of the western coast was then under the rule of king Pandion, " far away from his mediterranean em- porium of Modura " ; yet he remarks also that this name, with others in the same neighbourhood, was new to him. He evidently had no idea that the subjects of king Pandion were identical with the Pandae he himself had already referred to. Chola, the name of the Tamil people placed second in the list, is a word of unknown origin. It appears as Choda in one of Asoka's inscriptions, and also in the Telugu inscriptions of the Chalukya dynasty. In modern Telugu this word appears as Chola, in Tamil as Chora or Sora. We have here doubtless the Swpat (Sorai), &c., of Ptolemy. It is difficult to identify the country called Choliya by Hwen Thsang with the country inhabited by the Cholas, but it seems probable that the names are identical ; and we know that the Northern Circars were ruled by an offshoot of the Cholas in the eleventh century. The original seat of the Cholas seems to have been the extensive, fertile valley of the Kaveri, including the Tan j ore and Trichinopoly districts ; but subsequently they ruled over the whole of the Tamil country north of the Kaveri. Their capital city in the earliest period was JJtieyur (literally the ' city of habitation '), called also Kori, which appears to have been nearly identical with the modern Trichinopoly {J'irisirdppalli). In the eleventh century the Cholas reached the zenith of their power, and ruled — as is ascertained by inscriptions — over the whole Tamil country, including not only the country north of the Kaveri, but also the country of the 14 INTRODUCTION Pandyas, South Travancore, the northern districts in Ceylon, and a portion of the Telugu country. Chera, the name of the third Tamilian people, is a word which presents itself to us in many shapes, as will be seen when we proceed to consider the Malayalam language. The language of the Cholas never differed from that of the Pandyas ; and originally the language of the Cheras also differed but little from that of the other two portions of the Tamil people, as appears from the Syrian and Jewish inscriptions of the eighth century. By whatever local or dynastic names they may have called themselves, they all — whether Cheras, Cholas, or Pandyas — continued to be called Dravidas, and the language they spoke in common was everywhere called by the one name of Dravida or Tamil. This idea of the original identity of the Cheras, or people of Kerala, with the Cholas and Pandyas, is quite in accordance with native traditions. According to Tamil tradition, Cheran, Choran, and Pdndiyan were three royal brothers, who at first lived and ruled in common at Kolkei, on the Tamraparni, a river in Tinnevelly re- nowned in ancient song, on the banks of which the earliest civilisation in Southern India appears to have been built up. Eventually a separation took place : Pandiyan remained at home ; Cheran and Choran went forth to seek their fortunes, and founded kingdoms of their own to the north and west. We have a similar representation, perhaps merely an echo of the Tamil tradition, in the Hari-vamsa and several Puranas (see Muir's " Sanskrit Texts," vols. i. and ii.), in which Pandya, Kerala, Kola, and Cliola are represented as the four sons of Akrida, or of Dushyanta, the adopted son of Turvasu, a prince of the lunar line of the Kshatriyas. Who the Kolas of this list were is not clear. The term is supposed by some to have been intended to denote the Canarese people, Karnata being given in this connection instead of Kola by several Puranas. The Canarese people, however, are never called Kolas either by themselves or by their Dra vidian neighbours ; and it seems most probable that the Kols or Kolarians were referred to, perhaps under the impression (if so, an erroneous one, except in so far as the Oraons and Malers are con- cerned) that they also were Dra vidians. The Tamil language is called Aravam by the Mussulmans of the Dekhan, the Telugus, and the Canarese. What is the derivation of this term Aravam ? Its origin appears to me very uncertain. Dr ENUMERATION OF DRA VIDIAN LANGUAGES I 5 Gundert suggested that as Tamil literature excelled other literatures in ethics, it might have been perhaps from this circumstance that Tamilians were called Aravas. Aravas on this supposition would signify moralists, for aiam in Tamil means virtue ; it might mean perhaps even Buddhists, for Avavan, Tam. ' the virtuous one,' is a name of Buddha. It would not be a valid objection to this deriva- tion that the r of the Telugu and Canarese word Aravam is the ordinary liquid or semi-vowel, whilst the r of the Tamil aram is the hard, rough r, for the hard r of Tamil generally changes into r in Telugu and Canarese ; and this very word aram, Tam. virtue, is aravu in Canarese. Another theory derives the term from arivu, the Tamil word for knowledge, the Tamil people being supposed to be distinguished amongst the people of the south for their intelligence. Another derivation is from Aruvd, the name of an unknown district somewhere in the Tamil country, which was reckoned one of the twelve districts in which, according to the Tamil grammarians, bad Tamil was spoken. A formidable, if not a fatal, objection to these derivations is, that they have all a Tamil origin, whereas Aravam is absolutely unknown in Tamil itself as a name either of the people or of their language. It is by the Telugus, Canarese, and Dekhanis that the name is used, and its derivation must, therefore, be sought out of the Tamil country. The opinion of the best Telugu pandits I have consulted is that Arava is a Sanskrit, not a Dra vidian, word. It is to be divided as a-rava, destitute of sound ; and this name has been given, they suppose, to Tamil by the northern neighbours of the Tamilians on account of its being destitute of aspirates. Being the only language in India totally without aspirates, it was despised by outsiders for what was regarded as a defect, and was called in consequence Arava, which may be rendered ' unsonorous.' It was not likely, if this were the origin of the word, that the Tamil people would apply it to their own tongue. Aravam-u having come to be used in Telugu as the name of the language, the Telugu people went in time a step further, and called the people who spoke the language Arava-lu, Aravas. The Telugu word Aravam-u, ' the Tamil language,' is not to be confounded with the Tamil word arava7n, sound. It is a curious circumstance that the latter word means sound, whilst the former means being without sound. The initial a of the Tamil word is not, as it might r^dily be supposed to be, the Sanskrit a privative, but is one of the devices employed in Tamil to render it possible for Tamil organs to pronounce an initial r. (Comp. arasan, king, from Sanskrit raja.) It may also be noticed that whilst the 1 6 INTRODUCTION Sanskrit word rava means a loud sound, a noise, the Tamil form of the same word, aravam, means a very slight noise.* II. Malayalam. — This language claims to be placed next to Tamil in the list of Dravidian tongues, on account of the peculiarly close relationship to Tamil in which it stands. Malayalam is spoken along the Malabar coast, on the western side of the Ghauts, or Malaya range of mountains, from the vicinity of Chandragiri, near Mangalore, where it supersedes Canarese and Tulu, to Trivandrum (Tiru- vanantapuram), where it begins to be superseded by Tamil. The people by whom this language is spoken in the native states of Travancore (Tiruvidankodu or Tiravankodu) and Cochin (Kochchi), and in the British Indian districts of Malabar and Canara, may be estimated at 6,750,000. All along the Malabar coast Tamil inter- twines itself with Malayalam. Though that coast was for many ages more frequented by foreigners than any other part of India, though Phoenicians, Greeks, Jews, Syrian or Persian Christians, and Arabs, traded in succession to the various ports along the coast, and though permanent settlements were formed by the last three classes, yet the Malayalam people continue to be of all Dravidians the most exclusive and superstitious, and shrink most sensitively from contact with foreigners. Hence the lines and centres of communication have been occupied, and a considerable portion of the commerce and public business of the Malabar States has been monopolised, especially in Travancore, by the less scrupulous and more adroit Tamilians. Malayalam is also called Malaydrma, another form of which is Malay dyma ; but both words are substantially the same. The first part of each word is not the Sanskrit Malaya, ' a range of mountains ' (probably identical with the Western and Southern Ghauts), but the Dravidian mala, ' a mountain,' from which doubtless the Sanskrit malaya itself was derived. The second part of the word, dlam or drmxi, is an abstract neuter noun, between wxila and which y is inserted by rule to prevent hiatus, dlam is plainly a verbal deriva- tive from the root dl, ' to possess,' ' to use,' ' to rule ' (not to be confounded with dram, ' depth,' from the root dr, ' to be deep '). It bears the same relation to drma, originally dlma (Tam. dnmei, euphonised from dlmei), that tanam (Mai. -Tam. ' quality ') does to tanma, Mai. (Tam. tanmei) ; that is, it is more commonly used, but * It is noteworthy, perhaps, that the people who are represented by Ptolemy as oi^cnj.yiug, according to Colonel Yule, the portion of the Coromandel coast j.ear Ts. .lore, are called by him the Arvarni. ENUMERATION OF DRAVIDIAN LANGUAGES 1/ is reckoned less elegant. dr7}m is softened from alma, as in Tarn. velldlan, a cultivator, is sometimes softened into velldran. More frequently r changes to I, but the change of I to r is also known. This r is further softened in Malayalam to y, in consequence of which Malay dr ma becomes Malay dy ma. In colloquial Tamil this softening process is sometimes carried so far that the I disappears altogether and leaves no trace behind. Thus, velldnmei, Tam. cultivation, becomes in Malayalam velldyma, but in colloquial Tamil velldmei; nattdmnei, Tam. the headship of a village, from nddu and dnmei, becomes in Malayalam ndttdyma, but in colloquial Tamil ndttdmei ndttdnma is also found in Malayalam ; and this supplies us with a clear proof of the descent of dyma, through dnma, from dlma. Per- haps the best rendering of the term Malayalam or Malayarma is the ' mountain region.' If we had a word in English for a mountain district ending in ' ship ' like ' township,' it would come still nearer. When used as an abstract term in compounds anmei means use or possession — e.g., villdnmei, the use of the bow, from vil, bow. The appellative noun connected with this word dnmei is dlan or dli, each of which forms is in ordinary use both in Tamil and Malayajam — e.g., villdlan =villdli, Tam. -Mai. a bowman. The appellative noun corresponding to Malaydlam or Malaydrma is Malaydli, a man of Malayalam, a mountaineer. The Malayalam language is not distinguished from Tamil by San- skrit writers, the term Dravida, as used by them, including both tongues ; but the Malayalam country has a name of its own in Sanskrit, with special names for the various districts included in it, from Gokarnam to Cape Comorin. The general name of this entire region in Sanskrit is Kerala, a term which appears in the Kapur Di Giri version of Asoka's edict, in the third century B.C., in which the king of this country is called Keralamputra. Keralam is found in all the Dravidian dialects in one shape or another. In Tamil, through the softening of k into s, c, or ch, this word sometimes becomes Seralam, more commonly still Seram. Where the initial k is retained unchanged, it is followed by the Dravidian I — e.g., Keralam — and this is the case also in Telugu and Canarese. In Malayalam we find Keralam, Cheralam, and Cheram, as in Tamil, and also Keram, A man of Keralam is called sometimes Kelan or Kelu, and though this is evidently a contracl^n of Keralan, it must be one of great antiquity, for we find it in Pliny's name of the king of the country Celobotras, a form of the w^ord which is thus seen to be as accurate as Ptolemy's K^jpo^oBpos (Kerohothros). 1 8 INTIIODUCTION The Kerala of the ancients seems to have divided itself into two portions, one of which, the district lying along the sea coast, has always retained the Sanskritic name of Kerala, whilst it also called itself by the Tamil name of Chera ; the other, an inland district, including Coimbatore, Salem, and a portion of Mysore, seems to have dropped the name of Kerala altogether, and called itself exclusively either Chera or Kongu. It is to the latter district that the papers of Professor Dowson and Dr Eggeling on the Chera dynasty refer. Though, however, the districts and dynasties differed, I have no doubt that the names Kerala and Chera were originally one and the same, and it is certain that they are always regarded as synony- mous in native Tamil and Malayalam lists of synonyms. In the various lists of the boundaries of Chera given by Tamil writers, the Malabar coast from Calicut southward — that is, the whole of southern Kerala — is invariably included. Probably Kera was the earliest form of the word, Kerala a Sanskritic derivative. The word Kongu, one of the names of the Chera country, means, like Kudagu (Coorg), crooked, curved, and is evidently a name derived from the con- figuration of the country. The meaning of Keram is not so certain. One meaning of this word in Malayalam is ' a cocoa-nut palm.' This would furnish us with a very natural origin for the name of the country ; but, unfortunately, it seems to be only a secondary meaning, the name of the country itself being probably the origin of this name of its most characteristic tree. No word allied to Malay- a|am, the native name of the language, and the name most commonly used now for the country, seems to have been known to the earlier Greeks. A portion of the name appears for the first time in the " Christian Topography " of Cosmas Indicopleustes, about 545 a.d., who, "ivriting especially about Ceylon, mentions amongst the adjacent countries, " MaAe, {Male) whence the pepper comes." This form of the word is evidently identical with the Tamil malei, a hill, the hill country, a word which would be in common use then, as now, amongst the Tamil settlers in Ceylon. The distinctively Malayajam form of the same word is mala. Malayalam being, as I conceive, a very ancient offshoot of Tamil, differing from it chiefly at present by its disuse of the personal termi- nations of the verbs and the larger amount of Sanskrit derivatives it has availed itself of, it might perhaps be regarded rather as a dialect of Tamil, than as a distinct member of the Dra vidian family. Though its separation from Tamil must have taken place at a very early period, yet it seems to have participated, as time went on, in the ENUMERATION OF DRAVIDIAN LANGUAGES 1 9 progressive cultivation and refinement of Tamil, — possibly through the political influence the Tamilians acquired on the western coast in early times, an illustration of which we have seen in the fact that the author of the " Periplus " represents Nelkynda, one of the most important emporia on the western coast, as belonging to the Pandya king of Madura, the principal potentate in the Tamil country. The oldest Malayalam poetry, as I learn from Dr Gundert, imitated Tamil rather than Sanskrit. It eschewed all letters not included in the thirty-two adopted by Tamil, and the character employed was a character often used in inscriptions in the Tamil country, particularly in the south, and differing very widely from the Malayalam character now in use. The " Rama Charita," probably the oldest poem in the language, though not, after all, of any very great antiquity, was composed before the introduction of the Sanskrit alphabet, and exhibits substantially the same phase of the language as the Jewish and Syrian S'asanas (inscriptions). Bearing this in mind, it is remarkable that the Brahmanisation of the language and literature should now have become so complete. This process appears to have been carried on systematically only during the last two or three centuries, yet one of the most marked characteristics of the Malay- alam language, as we now find it, is the quantity of Sanskrit it contains. The proportion of Sanskrit words adopted by the Dravidian languages is least in Tamil, greatest in Malayalam ; and the modern Malayalam character seems to have been derived in the main from the Grantha, the character in which Sanskrit is written in the Tamil country. In consequence of these things, the difference between Malayalam and Tamil, though originally slight, has progressively increased, so that the claim of Malayalam, as it now stands, to be considered, not as a mere dialect of Tamil, but as a sister language, cannot be called in question. Originally, it is true, I consider it to have been not a sister of Tamil, but a daughter. It may best be described as a much-altered offshoot. The descent of Malayalam from Tamil may be illustrated by the word it uses to denote east. This is Jcirakku, meaning beneath, downwards, a word which corresponds to that which is used to denote west, viz., melku, above, upwards ; both of which words necessarily originated, not in the western coast, but in the Tamil country, or the country on the eastern side of the Ghauts, where a lofty range of mountains rises everywhere to the westward, and where, conse- quently, to go westward is to go upwards, whilst to the eastward 20 INTRODUCTION the country slopes downwards to the sea. The configuration of the Malayalam country, as of tlie whole of the western coast, is directly the reverse of this, the mountain range being to the eastward, and the sea to the westward. Notwithstanding this, the Malayalam word for east is identical with the Tamil word ! To what can this coin- cidence point but the original identity of Malayalam with Tamil ? The people by whom Malayajam is spoken must originally have been a colony of Tamilians. They must have entered the Malayalam country by the Paulghaut or Coimbatore gap, and from thence spread themselves along the coast, northward to the Chandragiri river, southward to the Neyyaru river near Trivandrum, at each of which points their further progress seems to have been stopped by settlements of colonists of a kindred race, who had already reached the western coast by different routes. Dr Gundert (Introduction to "Malayalam Dictionary"), whilst admitting Tamil and Malayalam to be very nearly related, appears to be unwilling to consider Malayalam as an offshoot of Tamil. He argues (in a private communication) that the words used in Malayalam for east and west cannot safely be regarded as proving the immigration of the Malayalam people from the east, and that if the analogous progress of the Aryans to the south be considered, it will appear probable that the Dravidians, like the Aryans, formed settlements on the western coast first, and after- wards made their acquaintance with the eastern. It is true, as he observes, that padinndru, properly fadinndyivu, meaning the setting sun, is more commonly used in Malayalam for west than melku, but fadunndyiru is also a Tamil word, and Dr Gundert admits that both melku and hirakku must have originated in the Tamil country. The argument from the analogy of the Aryan immigration appears to prove too much. It would require us to regard the whole Tamil people as immigrants from the western coast, and the Tamil language as an offshoot from Malayalam, the geographical and philological difii- culties in the way of both which suppositions appear to me to be insuperable. Origin of the terms ' Coroviandel ' and ' Malabar.' — Before passing on to the rest of the Dravidian languages, it may be desirable to inquire into the origin of the names ' Coromandel ' coast and ' Mala- bar ' coast, by which the eastern and western coasts of the southern portion of the Indian peninsula, in which the Tamil and Malayalam languages are spoken, are usually designated. 1. Coromandel. — The best derivation of Coromandel is from the Tamil Choramandalam,, the Chola country, from Chora, the Tamil ENUMERATION OF DRA VIDIAN LANGUAGES 21 form of the name which is best known in its Sanskrit form of Chola, and mandalam (a Sanskrit tadbhava), ' a district of country.' Un- doubtedly Fra Paulino a St Bartolomaso was wrong in supposing Cholamandalam to have meant ' the millet country.' The first word, Choram, though often pronounced like Ch6}am (' maize,' not ' millet '), is always written in Tamil Choram, and the compound Chora-mandalam, 'the country of the Choras,'like Pandya-mandalam, ' the country of the Pandyas,' has been in common use for ages. The first Portuguese, as I learn from Dr Gundert, always called by the name of Choramandala the fifth province of the Rayar's empire (the empire of the so-called Rayulu or Telugu kings of Vijayanagara), which they represented as extending from the frontiers of Quilon (that is, from near Cape Comorin) to Orissa. The Portuguese evi- dently adopted this name as the equivalent of Ma' bar, the name by which the greater part of the Coromandel coast had up to that time been generally called by the Muhammedans and those Europeans who derived their information from them. (See Ibn Batuta and Marco Polo.) This name Ma' bar, literally a ford or passage, was used originally to denote the coast of Madura, from which there was an easy passage by Rama's bridge to Ceylon. The application of the name was then indefinitely extended northwards. The change from Choramandala to Coromandel is one which would easily be made. The middle point appears to be Choromandel, the mode in which the name was written by the early Dutch. In the first edition of this work, whilst assigning this origin to the term Coromandel coast, I suggested also that it was difficult to see how the first mariners could have become acquainted with this somewhat high-flown classical word. It seemed to me desirable, therefore, to seek for some more trite and easy derivation of the word Coromandel — some derivative that would suit the circum- stances of mariners and factors ; and this, I said, I think we find in Karu-mayial (literally, black sand), the name of a small village on the eastern coast, near Pulicat (the first settlement of the Dutch), which is invariably pronounced and written Coromandel by the Europeans who are resident in Madras, some of whom annually take refuge in Karumanal or Coromandel during the hot land winds. Coromandel is often the first point which is sighted by ships from Europe bound to Madras ; and the objects ^n which my own eyes first rested on approaching the coast, in January 1838, were the cocoanut trees of Coromandel and the distant Nagari hills. I fear, however, this easy derivation must be given up, and the more ancient one, which carries 22 INTRODUCTION US back to the first arrival of the Portuguese in India, retained. I learn also from Mr C. P. Brown, that in a map of the Jaghire of Madras in " Kitchin's Atlas " (about 1790), the name of the village in question is written, not Coromandel, but Karri ynannel, so that the application of the name Coromandel to this village by the English must be of recent date. One of the names given to the eastern coast in Telugu is Kharamandalam, from khara, Sans, hot ; but this name has never been used so widely along the coast as to render it likely that it was the origin of the name Coromandel. Besides, this name w^as never used, as Choramandalam was, as a political designation. I am indebted to Colonel Yule, the learned editor of Marco Polo, for additional information regarding the use of the term Coromandel by the early Portuguese. He says — " It certainly w^as a name in use when the Portuguese arrived in India. This appears from its use in the short narrative of Hieronimo de Sto Stefano, dated in 1499, which is published at the end of Major's ' India ' in the fifteenth century. After mentioning Ceylon he says, ' departing thence after twelve days we reached another place called Coromandel.' The city of Choromandel appears in ' Vaithema's Travels' (published in 1510); and in Barbosa, the most complete of the early Portuguese accounts, we have the country of Charamandel (in the Portuguese edition), Coromandel (in Ramusio's Italian), Cholmendel and Cholmender in a Spanish MS. translated by Lord Stanley of Alderley in the Hakluyt series. I believe both Spanish and Portuguese pronounce the ch as we do, so I should think it probable that the Italian Co was written ^o. This Cholmendel is remarkable, as the MS. is supposed to date about 1510, too early for theories about Chola-mandala. I had given up the hope of finding proof of the use of this name by the Muhammedans, but on turning to Rowlandson's translation of the ' Tohfat al Majahidin, or History of the Muhammedans in Malabar,' I have found (p. 153) that the Franks had built fortresses ' at Mielapoor and Nagapatam, and other seaports of Solmondul,' and the name occurs again in the next page." Colonel Yule, in mentioning this in the Bombay Antiquary for August, 1874, adds — " The occurrence of this name in this form and in a Muhammedan writer upsets a variety of theories as to the origin of the name." The Coromandel coast is evidently the HapaAta l\o/)ojrwi/ {Paralia Soroton) (or 2w/Diywv) {Sorigon) of Ptolemy, and also the district ttJs tStws Xeyovjicvr]^ UapakLus Tojptyyojv (Tls idios legoumefies ENUMERATION OF DKAVIDIAN LANGUAGES 2 3 paralias Toringon), (or Swptyycov) (Soringon), in which the mouth of the Xa/Srjpos (KaMros), the Kaveri, was situated. These seem remarkable anticipations of the name by which the coast was known in later times. 2. Malabar. — The origin of the name Malabar has hitherto been enveloped in greater obscurity than that of the corresponding name Coromandel. The first part of the name (Mala) is evidently the Malayalam word for mountain, as in the name Malayalam itself, and we can scarcely err in concluding it to have been a perpetuation of the Male of the later Greeks. I learn from Colonel Yule that in the relations of the Arabian navigators the name Male held its place, nearly as Cosmos has it, without any such suffix as bar, down to the eleventh or twelfth century. In 851 A. d. it occurs, he says, as Malai or Kulam-Malai, in 1150 as Mali and also Maliah. It is interesting to find the name of Quilon (Kulam, properly Kollam) as early as 851 associated with the name of the coast, in the compound term Kulam- Malai ; but Colonel Yule has found Quilon mentioned by name prior even to 660,* which tends to show, as he observes, that the Quilon era (the first year of which corresponds to a.d. 824-5) did not in reality take its origin, as has been supposed, from the foundation of the city. The first appearance of the affix bar is in 1150, and from the time of its appearance, the word to which it is affixed — the first part of the compound — is frequently found to change. Colonel Yule gives the following Arabian forms — Malibar, Manibar, Mulibar, Miinibar, Malibar ; and the following as the forms used by early European travellers, &c. — Minibar, Milibar, Melibar (Marco Polo), Minubar, Melibaria. From the time of the arrival of the Portuguese in India it seems always to have been called Malabar, as by ourselves, and in this form of the word Mala, mountain, is coi*rectly given. It has been more difficult to ascertain the origin and meaning of the affix bar, Lassen explained it as identical with the Sanskrit vara, in the sense of 'a region' ; Malaya-vara= Malabar = the region of Malaya, the Western Ghauts. The difficulty in the way of accepting this is that Malaya-vara is a factitious word, not really found in Sanskrit, and never actually. used by the people of the Malabar coast. The same * A letter in Assemaiii's Bii^liotheca, from the Patriarch Jesajabus (died a.d. 660) to Simon, Metropolitan of Persia, blames his neglect of duty, saying that in consequence, not only is India, *' which extends from the coast of the kingdom of Persia to Colon, a distance of 1200 parasangs, deprived of a regular ministry, but Persia itself is lying in darkness." — Colonel Yule. 24 INTRODUCTION difficulty stands in the way of Mala-varam, Tarn. -Mai. the foot of the mountains, and Malappadu, the mountain district. These derivations might be regarded at first sight as admissible ; but they are Indian vernacular words, and if the name Malabar had been derived from them, we should expect to find them in use in India itself, whereas there is no trace of either of them having ever actually been used by any Indian people. Dr Gundert suggested to me the possibility of the derivation of bar from the Arabic barr, continent, as he considered it probable that the name of Malabar had first been brought into use by the Arabian navigators. Colonel Yule arrived independently at a similar conclusion. He preferred, however, the Persian bar to the Arabic barr, and has given illustrations of the use of this Persian affix by the Arabs, which appear to me to carry conviction. He says (in one of the private communications with which he has favoured me), " This affix bar seems to have been much used by navigators. We have ZsLUzi-bdr (the country of the blacks), Kala-bar (see the ' Arabic Eelations,' by Eeinaud, I. 17, where it is explained that ' the word bar signifies either a coast or a kingdom ') ; and even according to Johnson's ' Persian Arabic Dictionary,' Hindu-bar. Burton says {Journal of the Royal Geographical Society, vol. xxix. p. 30) that at Zanzibar, in distinguishing the mainland from the island, they call the former Barr-el-Moli, or ' continent.' And in a note he adds, ' The word MoH, commonly used in the corrupt Arabic of Zanzibar, will vainly be sought in the dictionaries.' Query, if this word Moli for continent may not have shaped some of the forms of the name of Malabar that we have above. I suppose bar itself is rather Persian than Arabic, and may be radically the same affix that we have in so many Indian names of countries, Marwar, Rajwar, &c." This Persian derivation seems to me so satisfactory that it may safely be accepted, bar, country, may have been added to Male to dis- tinguish the mainland from the adjacent islands, the Maldives and the Laccadives. The Maldives may have been the dives or islands of Male, whilst Malaga/ was the continent or mainland of Male. Colonel Yule informs me that Pyrard de la Val and Moresby agree in calling the principal island Male ; the first vowel of this name may be either lon^ or short. In Singhalese the islands are called the Maldives, but in Tamil they are called If aZdives ; and this Tamil mdl differs con- siderably from Mala, the name of the Malabar coast, whilst it agrees perfectly with the name given to the islands by Ibn Batuta, who calls them Dhibat-al-mahal, from the name of the ' atoll ' where the ENUMERATION OF DRAVIDIAN LANGUAGES 2$ sultan of the islands lived — viz., Al-mahal. Mahal is always cor- rupted into mdl in Tamil. The Persian bar, one of the meanings of which is ' a country,' is regarded by Vuller (" Lexicon Pers.-Lat.") as identical in origin with the Sanskrit vara, a noun of multitude. It does not follow, however, that it is identical with the affix vdr which we find in so many Indian names of countries, as Marwar, Dharwar, Kattywar, &c. The apparent resemblance between this war and the Persian bar and especially the Sanskrit vara disappears on investigation. This war is written vdd ; and Dr Trumpp assures me that its lineal descent from the Sanskrit vdta (vdta, vdd, vdr) is capable of proof. vdta, Sans., means not only ' an enclosure,' but also ' a district ' — e.g., Prdchyavdta, the eastern district. Dr Eggeling informs me that he has found Dharwar written Dhara-varsha in an inscription of the seventh century. According to Dr Trumpp, however, the war of the modern Dharwar must have had a different origin, as varsJia becomes in the Prakrit, not vdr, but variso or varakhi. III. Telugu. — In respect of antiquity of culture and glossarial copiousness, Telugu is generally considered as ranking next to Tamil in the list of Dravidian idioms, whilst in point of euphonic sweetness it justly claims to occupy the first place. This language was some- times called by the Europeans of the last generation the ' Gentoo,' from the Portuguese word for heathens or ' gentiles,' a term which was used at first to denote all Hindus or ' natives,' but which came in time to mean the Telugus alone. The use of the term Gentoo for Telugu, like that of Malabar for Tamil, has now nearly disappeared. Telugu is spoken all along the eastern coast of the Peninsula, from the neighbourhood of Pulicat, where it supersedes Tamil, to Chicacole, where it begins to yield to the Oriya, and inland it prevails as far as the eastern boundary of the Maratha country and Mysore, including within its range the ' Ceded districts ' and Karnul, a con- siderable part of the territories of the Nizam, or the Hyderabad country, and a portion of the Nagpur country and Gondvana. The district thus described was called Telingana by the Muhammedans. The Telugu people, though not at present the most enterprising or migratory, are undoubtedly the most numerous branch of the Dravidian race. Including the Nayudus (Tam. Nayakkas = Sans. Nayakas), Eeddis, and other Telugu tribes settled in the Tamil country, who are chiefly the descendants of those soldiers of fortune by whom the Pandya and Chola kingdoms were subverted, and who 26 INTKODUCTION number not much less than a million of souls ; and including also the Telugu settlers in Mysore, and the indigenous Telugu inhabitants of the native states, the people who speak the Telugu language may- be estimated as amounting to at least twenty-three millions. The Telugu-speaking people in the Nizam's territory number, according to the latest census, more than six millions. Though the Telugu people cannot at present be described as the most migratory portion of the Dravidians, there was a time when they appear to have exhibited this quality more conspicuously than any other branch of the race. Most of the Klings, or Hindus, found in the eastern archipelago in our times, are, it is true, Tamilians ; but the Tamilians, in trading and forming settlements in the East, have entered on a field formerly occupied by the Telugus, and not only so, but have actually inherited the name by which their Telugu predecessors were known. ' Kling ' stood for ' Kalinga,' and Kalinga meant the seaboard of the Telugu country. The Hindus, who in the early centuries of the Christian era formed settlements, built temples, and exercised dominion in Sumatra and Java, appear to have been Telugus, not Tamilians ; and whilst the Tamil country was overrun by the Telugus in the sixteenth and seventeenth centuries, no corresponding settlement of Tamilians in the Telugu country to any considerable extent seems to have followed the establishment in that country (or at least in the portion of it specially called Kalinga) of a dynasty of Chola kings in the eleventh or twelfth centuries. Telugu is called Andhra by Sanskrit writers — that is, the language of the Andhras, one of the two nations into which the Telugu people seems from the earliest times to have been divided. The other nation was the Kalingas. The Andhras seem to have been better known than the Kalingas to the early Aryans. They are mentioned as early as in the " Aitareya Brahmana of the Rig-veda," though represented therein as an uncivilised race ; and in Puranic times a dynasty of Andhra kings is represented to have reigned in Northern India. The Andarse are represented by Pliny (after Megasthenes) as a powerful people, and the Andre Indi have a place in the " Peutinger Tables " (north of the Ganges !) amongst the few Indian nations of which the author of those tables had heard. The first reference to their lan- guage I find made by any foreigner is in the memoirs of Hwen Thsang, the Chinese pilgrim, about the middle of the seventh century A.D., who states that the language of the Andhras differed from that of Central India, whilst the forms of the written characters were ENUMERATION OF DRA VIDIAN LANGUAGES 2/ for the most part the same. It is clear from this that Telugu culture had already made considerable progress, especially amongst the Andhra branch of the nation. Hence it naturally happened that the name of the Andhras, instead of that of the KaHngas, who in- habited the more remote seaboard, and were perhaps less cultured, was given by Sanskrit writers to the language which both branches of the nation spoke in common. It occupies the first place — not Kalinga or Trilinga — in the compound term, Andhra-Dravida- bhasha, by which Kumarila-bhatta, shortly after Hwen Thsang's date, designated what he appears to have supposed to be the one language spoken by the Dra vidians. Telugu is the name by which the language is called by the Telugu people themselves ; other forms of the name are Telungu, Telinga, Taihnga, Tenugu, and Tenungu. The name has been corrupted still further in various directions by Muhammedans and other foreigners. One of the above-mentioned forms, Tenugu or Tenungu, is sometimes represented by Telugu pandits as the original form of the word, and the meaning they attribute to it is sweetness. This derivation seems to have been an afterthought, suggested by the resemblance of the word to tene, honey ; but there is more reason for it — both on account of the resemblance between the two words, and also on account of the exceedingly me/Zvfluous character of the Telugu language, than for the corresponding afterthought of the Tamil pandits, respecting the meaning of the word Tamir. The favourite derivation of Telugu pandits for Telugu or Telungu, the ordinary name of their language, is from TriHnga, ' the language of the three lingas ' ; that is, as they represent, of the country of which three celebrated linga temples constituted the boundaries. This derivation was accepted by Mr A. D. Campbell, but is rejected by Mr C. P. Brown, who affirms it to be an invention of modern poets, and regards the name Telugu as devoid of any known root. Probably so much of the theory as is built on the connection of the name with certain temples may be unceremoniously discarded ; but the derivation of the name itself from trilinga (without committing ourselves to the determination of the sense in which the word linga is used) may perhaps be found to be deserving of a better fate. If the derivation of Telugu from TriHnga be an invention, it must be admitted to have at least tSe merit of being an ingenious invention ; for though it is quite true, as Mr Brown observes, that Trilinga, as a name of a country, is not found in any of the Usts of Indian countries contained in the Puranas, yet the existence of such a name seems 28 INTRODUCTION capable of being established by reliable evidence derived from other sources. Taranatha, the Tibetan author already referred to, who derived his information, not from modern Telugu poets or pandits, but from Indian Buddhistical narratives (which, having been written before Buddhism disappeared from India, must have been of con- siderable antiquity), repeatedly designates the Telugu country Trilinga, and describes Kalinga as a portion of Trilinga, and Kalinga- pura as its capital. The name of Trilinga had reached Ptolemy himself at a time anterior probably to the date of the Puranas. It is true his TptyXvirTOv (TpiyXvcfiOV ?) to Kal TpiXtyyov /?a(rtAetov (Triglypton [Triglyphon] to hai Trilmgon Basileion) is placed by him to the east of the Ganges ; but the names of places mentioned by Ptolemy seem generally much more reliable than the positions he assigns to them ; and it is conceivable that the mariners or merchants from whom he derived his information spoke of the place in question merely as beyond the Ganges, without being certain whether it was east or south. We have seen that in like manner the '' Peutinger Tables " place the Andre Indi — about whose identity with the Telugu people there can be no doubt — beyond the Ganges. The foreign name Trilingam must have been the name by which the place was called by the natives of the place, whilst Triglypton or Tri- glyphon must have been a translation of the name which had come into use amongst the Greeks. Hence the antiquity of Trilinga, as the name of a state, or of the capital city of a state, situated some- where in India in Ptolemy's time, must be admitted to be established. The word linga forms the second portion of the name of several Indian nations mentioned by Pliny (after Megasthenes), as the BoUngae, and the Maccocalingse, a various reading of which is Maccolinga). Another name mentioned by Pliny, Modogalingam, involves some difficulty. He says — " Insula in Gauge est magnse magnitudinis gentem continens unam, Modogalingam nomine." Mr A. D. Campbell, in the Introduction to his " Telugu Grammar," repre- sented the modoga of this name as the ancient Telugu word for three, and hence argued that Modogalingam was identical with Trilingam. If this identification were admitted, not only would the antiquity of Trilingam be firmly established, but also the opinion of the pandits that the original name of their language was Trilinga, and that this Trilinga became gradually Telinga, Telungu, Telugu, and Tenugu, would be confirmed. The Telugu word for ' three,' however, is not modoga, but mudu ; ^nudugu might be used ; but it is a poetical ENUMERATION OF DRAVIDIAN LANGUAGES 29 form, the use of which would be pedantic. Mr C. P. Brown prefers to write the name of the nation referred to by Pliny (after a MS. in Sillig's edition) " modo Galingam," and considers this Galingam equivalent to Galingam. The change of c (k) into g in such a con- nection would be quite in accordance with Telugu laws of sound, provided modo, as well as Galingam, were a Telugu word ; and if it were Telugu it would more naturally represent mudu, three, than anything else. On this supposition, modo-Galingam would mean, not indeed ' the three lingas,' but ' the three Kalingas ' ; and it is remarkable that the corresponding expression Tri-kalinga has been found in actual use in India. General Cunningham, in his " Ancient Geography of India," mentions an inscription in which a line of kings assumed the title of ' lords of Tri-kalinga.' Dr Kern also, in his translation of Varaha-mihira's " Brihat-samhita," men- tions that the name Tri-kalinga is found in one of the Puranas ; and the same name has recently been found in an inscription on a copper plate referred to in the proceedings of the Bengal Asiatic Society for 1872, p. 171. General Cunningham thinks it probable that there is a reference to these three Kalingas in the circumstance that Pliny mentions the Macco-Calingae and the Gangarides-Calingse as separate nations from the Calingse ; and that the Maha-bharata mentions the Kalingas three times, and each time in connection with different neighbours. The circumstance that Modogalingam is represented as an ' island in the Ganges ' presents no insuperable obstacle to its identification with Tri-kalinga or Telingana. The term island has often been used very vaguely. Taranatha calls the Tamil country an island ; and Kalinga was supposed to be a Gangetic country by Sanskrit writers themselves, who generally agreed in representing it as the last of the districts visited by the Ganges. It is also to be remembered that the Godavari is often supposed by natives to be somehow identical with the Ganges. General Cunningham thinks Telinga derived, not from Trilinga, but from Tri-kalinga, but this derivation of the word needs to be historically confirmed. Kalinga and linga may probably in some way be connected, but the nature and history of the connection have not as yet been made out. One of the names by which the Telugu language is known in the Tamil country is Vadugu, and a Telugu man, especially if a member of the Nayakka caste, is caMed a Vadugan. The root of this is vada, north, the Telugu country lying to the north of the Tamil. This word explains the name ' Badages,' by which certain marauding hordes were designated by the early Portuguese, and in the letters of 30 INTRODUCTION St Francis Xavier. Mr C. P. Brown informs me that the early French missionaries in the Guntur country wrote a vocabulary " de la langue Talenga, dite vulgairement le Badega." IV. — Canarese. — The next place is occupied by Canarese, properly the Kannada, or Karnataka, which is spoken throughout the plateau of Mysore, in the southern Mahratta country, and in some of the western districts of the Nizam's territory, as far north as Beder. It is spoken also (together with Malayalam, Tulu, and Konkani, but more extensively than any of them) in the district of Canara, properly Kannadiyam, on the Malabar coast, a district which was subjected for centuries to the rule of Canarese princes, and hence acquired the name by which it is at present known. The speech of the Badagas (' people from the north '), commonly called by the English Burghers, the most numerous class of people in- habiting the Neilgherry hills, is undoubtedly an ancient Canarese dialect. The Canarese, properly so called, includes, like the Tamil, two dialects — classical, commonly called Ancient Canarese, and the colloquial or modern ; of which the former differs from the latter, not — as classical Telugu and Malayalam differ from the colloquial dialects of those languages — by containing a larger infusion of Sanskrit derivatives, but by the use of different inflexional termina- tions. The dialect called Ancient Canarese is not to be confounded with the character denoted by that name, which is found in many ancient inscriptions in the Maratha country, as well as in Mysore. The language of all really ancient inscriptions in the Hala Kannada, or Ancient Canarese character, is Sanskrit, not Canarese. The people that speak the Canarese .language may be estimated at ten millions and a half. The term Karnata or Karnataka is said to have been a generic term, including both the Telugu and Canarese peoples and their languages, though it is admitted that it usually denoted the latter alone, and though it is to the latter that the abbreviated form Kanna- dam has been appropriated. Karnataka (that which belongs to Karnata) is regarded as a Sanskrit word by native pandits, but I agree with Dr Gundert in preferring to derive it from the Dra vidian words kar, black, ndd-u (the adjectival form of which in Telugu is ndt-i), country — that is, the black country — a term very suitable to designate the " black, cotton soil," as it is called, of the plateau of the Southern Dekhan. The use of the term is of considerable antiquity, as we find it in Varahamihira at the beginning of ENUMERATION OF DRAVIDIAN LANGUAGES 3 1 the fifth century a.d. Taranatha also mentions Karnata. The word Karnata or Karnataka, though at first a generic term, became in process of time the appellation of the Canarese people and of their language alone, to the entire exclusion of the Telugu. Karnataka has now got into the hands of foreigners, who have given it a new and entirely erroneous application. When the Muham- medans arrived in Southern India, they found that part of it with which they first became acquainted — the country above the Ghauts, including Mysore and part of Telingana — called the Karnataka country. In course of time, by a misapplication of terms, they applied the same name, the Karnatak, or Carnatic, to designate the country below the Ghauts, as well as that which was above. The English have carried the misapplication a step further, and restricted the name to the country below the Ghauts, which never had any right to it whatever. Hence the Mysore country, which is properly the Carnatic, is no longer called by that name by the English ; and what is now geographically termed ' the Carnatic ' is exclusively the country below the Ghauts, on the Coromandel coast, including the whole of the Tamil country, and the district of Nellore only in the Telugu country. The word Karnataka was further corrupted by the Canarese people themselves into Kannada or Kannara, from which the language is styled by the English ' Canarese.' V. TuLU. — Next in the list of cultivated Dravidian languages stands Tulu or Tul.uva. The claim of this pecuHar and very in- teresting language to be ranked amongst the cultivated members of the family may perhaps be regarded as open to question, seeing that it is destitute of a literature in the proper sense of the term, and never had a character of its own. The Canarese character having been used by the Basle missionaries in the Tulu books printed by them at Mangalore — the only books ever printed in Tu]u — that character has now become inseparably associated with the language. Not- withstanding its want of a literature, Tulu is one of the most highly developed languages of the Dravidian family. It looks as if it had been cultivated for its own sake, and it is well worthy of careful study. This language is spoken in a very limited district and by a very small number of people. The Chandragiri and Kalyanapuri rivers, in the district of Cai^ra, are regarded as its ancient boundaries and it does not appear ever to have extended much beyond them. The number of the Tulu-speaking people has been found to exceed half a million, and their country is broken in upon to such a degree by 32 INTRODUCTION other languages that Tulu might be expected soon to disappear. All Tulii Christians are taught Canarese as well as Tulu. Tulu, however, shows, it is said, no signs of disappearing, and the people have the reputation of being the most conservative portion of the Dravidian race. The name Tulu means, according to Mr Brigel, mild, meek, humble, and is to be regarded therefore as properly denoting the people, not their language. Tulu was supposed by Mr Ellis to be merely a dialect of Malayajam; but although Malayajam characters were, and still are, ordinarily employed by Tulu Brahmans in writing Sanskrit, in consequence of the prevalence of Malayalam in the vicinity, the supposition that Tulu was a dialect of Malayalam can no longer be entertained. The publication of Mr Brigel's " Tulu Grammar " has thrown much new light on this peculiarly interesting language. It differs far more widely from Malayalam than Malayalam does from Tamil. It differs widely, but not so widely, from Canarese ; still less so from Coorg. The dialect from which it differs most widely is Tamil. There is a tradition mentioned by Mr EUis, in his treatise on Mirasi right, to the effect that the ancient Kurumbars or nomadic shepherds, in the neighbourhood of Madras, were expelled and their lands given to Vellalas from Tuluva ; and this tradition is confirmed by the fact that certain Vellala families in that neighbourhood call themselves, and are called by others, Tuluva Vellalas. Probably, however, the number of Tuluva immigrants was not very considerable, for there is no trace of any infusion of the peculiarities of Tulu into the colloquial Tamil of Madras, which, if it differs in any degree from the Tamil spoken in the rest of the Tamil country, differs, not in a Tulu, but in a Telugu direction. VI. KuDAGU or Coorg. — Last in the list of cultivated Dravidian languages is the language of Coorg ; but though I have thought it best to give this language a place amongst the cultivated members of the family, the propriety of doing so seems to me still more doubtful than that of placing Tu]u in this list. Coorg is a small but interesting district, formerly an independent principality, beauti- fully situated amongst the ridges of the Western Ghauts, between Mysore on the east and North Malabar and South Canara on the west. The native spelling of Coorg is usually Kodagu, properly Kudagu, from kuda, west, a meaning of the word which is usual in Ancient Tamil. In the first edition of this work this language had not assigned to it a place of its own, but w^as included under the head of ENUMERATION OF DR A VIDIAN LANGUAGES 33 Canarese. It had been generally considered rather as an uncultivated dialect of Canarese, modified by Tulu, than as a distinct language. I mentioned then, however, that Dr Mogling, a German missionary, who had resided for some time amongst the Coorgs, was of opinion that their language was more closely allied to Tamil and Malayalam than to Canarese. ' It is not quite clear to me yet to which of the Dra vidian dialects it is most closely allied. On the whole, however, it seems safest to regard it as standing about midway between Old Canarese and Tulu. Like Tulu it has the reputation of puzzling strangers by the peculiarities of its pronunciation. A grammar of the Coorg language has been published by Major Cole, Superintendent of Coorg, and some specimens of Coorg songs, with an epitome of the grammar by the Rev. B. Grater of Mangalore. " Like the similar dialects spoken by the tribes of the Nilagiris, there can be no doubt that this language has preserved its form comparatively free from change ow4ng to the retired position of the people who speak it. That the inhabitants of Coorg early settled on the Western Ghauts is shown by the primitive Dravidian custom of polyandria which they still follow. They are as yet far from being Brahmanised, and they have no literature in the proper sense of the word." — Burnell's " Specimens of South Indian Dialects," No. 3. The six languages which follow differ from those that have been mentioned in that they are entirely uncultivated, destitute of written characters, and comparatively little known. VII. TuDA. — Toda, properly Tuda, is the language of the Tudas or Tudavars, a primitive and peculiarly interesting tribe inhabiting the Neilgherry (Nilagiri) hills. It is now regarded as certain that the Tudas were not the original inhabitants of those hills, though it is still far from certain who the original inhabitants were. Their numbers could not at any time have exceeded a few thousands, and at present, probably through opium-eating and polyandria, and through the prevalence amongst them at a former period of female infanticide, they do not, it has been ascertained, number more than about 700 souls. I have to thank, the Rev. F. Metz, the veteran missionary among the Neilgherry tribes, for much information respecting the Tudas and their language ; and an interesting book has lately been written by Colonel Marshall, entitled " A Phreno- logist among the Todas," if! which everything that is known of this people is fully described. The same book contains a valuable epitome of the grammar of their language by the Rev. Dr Pope. 34 INTRODUCTION Dr Pope connects the name of the Todas with the Tamil word tora, a herd ; but the d of Tuda is not the lingual d, but the dental, which has no relationship to r or I. The derivation of the name may be regarded as at present unknown. VIII. KoTA. — The language of the Kotas, a small tribe of helot craftsmen inhabiting the Neilgherry hills, and numbering about twelve hundred souls.* This language may be considered as a very old and very rude dialect of the Canarese, which was carried thither by a persecuted low-caste tribe at some very remote period. Besides the languages of the Todas and Kotas, two other languages are vernacular on the Neilgherry hills — viz., the dialect spoken by the Burghers or Badagars (the northern people), an ancient but organised dialect of the Canarese ; and the rude Tamil spoken by the Irulars (* people of the darkness ') and Kuruburs (Can. Kuruharu, Tam. Kurumbar, shepherds), who are occasionally stumbled upon by adventurous sportsmen in the denser, deeper jungles, and the smoke of whose fires may occasionally be seen rising from the lower gorges of the hills. IX. GoND. — The language of the indigenous inhabitants of the extensive hilly and jungly tracts in Central India, formerly called Gondwana. According to the recent census the various tribes included under the general name of Gonds number one million and a half. The Marias are regarded as the purest, and are certainly the wildest, tribe of Gonds. They sometimes call themselves Kohitur, a name which is evidently identical with Koitor, the name by which four out of the twelve tribes of Gonds call themselves. It has been asserted, indeed, that all the Gonds, when speaking of themselves in their own language, prefer to call themselves Koitors. This word is a plural appellative regularly formed from Koi. Much valuable information concerning the Gonds is contained in Colonel Dalton's " Ethnology of Bengal " ; in the papers left in MS. by the late Eev. S. Hislop, edited by Sir R. Temple ; and in the Gazetteer of the Central Provinces. A grammar and vocabulary of the Gond language were published by the Rev. J. G. Driberg, at Bishop's College, Calcutta, in 1849. A translation of the Gospels of St Matthew and St Mark into Gond by the Rev. J. Dawson, published at Allahabad in 1872-73, furnishes us with a still more valuable contribution to the knowledge of the language. Mr Dawson has also recently * The latest census figures show that there are nine of these in Assam. — Editors. ENUMEIJATION OF DRA VIDIAN LANGUAGES 35 published a brief grammar and vocabulary of the language in the Journal of the Bengal Asiatic Society. X. Khond ; more properly Ku. This is the language of the people who have hitherto been commonly called Khonds. By their neigh- bours in Orissa their name is said to be pronounced Kandhs ; but by themselves they are called, it is said, Kus. They are a primitive race supposed to be allied to the Gonds, and inhabit the eastern parts of Gondwana, Gumsur, and the hilly ranges of Orissa, which con- stitute the Tributary Mahals. Colonel Dalton says they are not found further north than the 22nd degree of latitude, and that they extend south as far as Bastar, whence their position as the aboriginal people is taken up by the Savaras or Sauras. They acquired a bad notoriety for a long time, through their horrid prac- tice of stealing the children of their neighbours of the plains, and offer- ing them up in sacrifice — a practice now entirely suppressed. The meaning of the name of this people is involved in obscurity. Some consider Khond a kindred word with Gond, and derive both names from the Tamil word hundru, a hill, literally a small hill, the Telugu form of which is Jconda. This would be a very natural derivation for the name of a hill people ; but, unfortunately, their nearest neighbours, the Telugus, call them, not Konds or Gonds, but Gonds, also Kods ; and as they call themselves Kus, according to Mr Latchmaji, the author of the grammar of their language, the existence of any connection between their name and hundru or konda, a hill, seems very doubtful. The term Ku is evidently allied to Koi, the name by which the Gonds call themselves, and which they are fond of lengthening into Koitor. The Khonds, according to the late census, number nearly 530,000 souls. XI. The Maler, commonly called the Rajmahal, the language of the Paharias, or hill people, wdio seem to have been the original inhabitants of the Rajmahal hills in Bengal. The brief vocabulary of the language of this tribe contained in the " Asiatic Researches," vol. v., and the somewhat fuller lists of words belonging to the same language contained in Mr Hodgson's and Sir George Campbell's collections and in Colonel Dalton's " Ethnology of Bengal," lead to the supposition that the R^mahal idiom is in its basis Dravidian. This language is not to be confounded with the speech of the Santals, a branch of the extensive Kol family inhabiting at present the skirts of the Rajmahal hills (but said to be mostly emigrants from the ^6 INTRODUCTION Hazarabagh district), who belong to a stock totally different from that of the Malers. Unfortunately very little is known of the grammatical structure of this language. The numbers of the people by whom it is spoken have been ascertained to amount to 64,000. XII. Oraon. — The Oraons of Chutia Nagpur and the neighbouring districts are estimated to amount to 800,000. Colonel Dalton has given a very full and interesting account of this tribe in his " Ethno- logy of Bengal." They have preserved, like the Malers, the rudi- ments of a language substantially Dravidian, as appears from the lists of words collected by Mr Hodgson and Colonel Dalton, and especially from an epitome of the grammar of their language pre- pared by the Rev. F. Batsch. Their traditions are said to connect them with the Konlcan, from which it is supposed they derive the name Khurnk, by which they invariably call themselves. They assert that for many generations they were settled on the Rohtas and adjoining hills in the Patna district, and that when driven out from thence, one party emigrated to the Rajmahal hills, the other went south-eastward till they arrived in the highlands of Chutia Nagpur. This tradition of the original identity of the Malers and the Oraons is borne out by the evident affinity of their languages, and, as Colonel Dalton mentions, by the similarity of their customs. According to their traditions, the Oraons arrived in Chutia Nagpur later than the Mundas and other Kolarians. Tuda, Kota, Gond, and Ku, though rude and uncultivated, are undoubtedly to be regarded as essentirJly Dravidian dialects, equally with the Tamil, the Canarese, and the Telugu. I feel some hesitation in placing in the same category the Rajmahal and the Oraon, seeing that they appear to contain so large an admixture of roots and forms belonging to some other family of tongues, probably the Kolarian. I venture, however, to classify them as in the main Dravidian, be- cause the Dravidian roots they contain are roots of primary import- ance, including the pronouns and the first four numerals, from which it may fairly be inferred that these dialects belonged originally to the Dravidian family. The Oraon was considered by Mr Hodgson as a connecting link between the K61 dialects and the Maler ; the Maler as a connecting link between the K61 and the distinctively Tamilian families. The Maler seems to me, on the whole, less distinctively Dravidian than the Oraon, perhaps because the Malers, or hill men of Rajmahal, are locally more remote than the Oraons ENUMERATION OF DRA VIDIAN LANGUAGES 3/ from the present seats of the Dra vidian race. Sir George Campbell's lists of words belonging to the Maler and Or^on dialects appear to contain a larger proportion of words that can be recognised as distinctively Dra vidian than any previous lists. The existence of a distinctively Dra vidian element in two at least of these aboriginal dialects of the Central Provinces and Bengal being established, the Dravidian race can now be traced as far north as the confines of Bengal, if not also to the banks of the Ganges ; and the supposition that this race was diffused at an early period through the greater part of India is thereby confirmed. Colonel Dalton carries the Dravidian element still further than I have ventured to do. He says (" Ethnology of Bengal," p. 243), *' The Dravidian element enters more largely into the composition of the population of Bengal than is generally supposed. I believe that a large majority of the tribes described as Hinduised aborigines might with propriety have been included in this group. The people called Bhuiyas, diffused through most of the Bengal districts, and massed in the jungle and tributary estates of Chutia Nagpiir and Orissa, certainly belong to it ; and if I am right in my conjecture regarding the Kocch nation, they are of the same stock. I roughly estimate the Bhuiyas at two and a half millions, and the Kocch at a million and a half, so that we have in these two peoples about one-tenth of the Bengal population, who in all probability should be classed as Dravidian." I hesitate for the present to endorse this supposition, in the absence of lingual affinities of any kind and of physical characteristics — if there are any such even amongst the Dra vidians themselves — that can be regarded as distinctly Dravidian. Leaving these doubtful races out of account, the numbers, as far as can be ascertained by the census of 1911, of the various peoples and tribes by whom distinctively Dravidian languages are spoken are here exhibited. 1. Tamil .... 19,189,740* 2. Telugu .... 23,542,859 3. Canarese . . . . 10,525,739 4. Malavalam . . . 6,792,277 Carryforward* . . 60,050,615 * This does not include the Eastern Archipelago, Mauritius, and other places. The peoples speaking the Tamil language may thus be roughly esti- mated at twenty- one millions. 38 INTRODUCTION Brought forward 60,050,615 5. Tulu 531,498 6. Kudagii or Coorg 42,881 7. Tuda . ... 730 8. Kota 1,280 9. Gond 1,527,157 10. Khond or Ku 530,476 11. Rajmahal 64,875 12. Oraon ... 800,328 63,549,840 According to this estimate the Dravidian-speaking peoples amount to nearly sixty-four millions of souls. In this enumeration of the Dra vidian languages the idioms of the Ramusis, the Lambadis, and various other wandering, predatory, or forest tribes have not been included. The Lambadis, the gipsies of the peninsula, speak a dialect of Hindustani ; the Ramusis a fatois of Telugu ; the tribes inhabiting the hills and forests, cor- rupted dialects of the languages of the contiguous plains. None of these dialects is found to differ essentially from the speech of the more cultivated classes residing in the same neighbourhood. The Male- arasas, ' hill-kings ' (in Malayajam, Mala-arayas), the hill tribe inhabiting the Southern Ghauts, speak corrupt Malayalam in the northern part of the range, where Malayalam is the prevailing lan- guage, and corrupt Tamil, with a tinge of Malayalam, in the southern, in the vicinity of Tamil-speaking districts. In the above list of the Dravidian languages I have not included the Ho, the Munda, or any of the rest of the languages of the Kols, the Savaras, and other rude tribes of Central India and of Bengal, called ' Kolarian ' by Sir George Campbell, and included by Mr Hodgson under the general term Tamulian. These languages might naturally be supposed to be allied to Gond or Ku, to Oraon or Rajmahal, and consequently to be of Dravidian origin ; but though a few Dra- vidian words may perhaps be detected in some of them, their grammatical structure shows that they belong to a totally different family of languages. Without the evidence of similarity in gram- matical structure, the discovery of a small number of similar words seems to prove only local proximity, or the existence of mutual intercourse at an earlier or later period, not the original relationship either of races or of languages. ENUMERATION OF DRAVIDIAN LANGUAGES 39 I leave also out of acconnt the languages of the north-eastern frontier of India, which are spoken by the Bodos, Dhimals, and other tribes inhabiting the mountains and forests between Kumaon and Assam. These were styled Tamulian by Mr Hodgson, on the supposition that all the aborigines of India, as distinguished from the Aryans, or Sanskrit-speaking race and its offshoots, belonged to one and the same stock ; and that of this aboriginal race, the Tamilians of Southern India were to be considered the best representatives. But as the relationship of those north-eastern idioms to the languages of the Dravidian family, is unsupported by the evidence either of similarity in grammatical structure or of a similar vocabulary, and is founded only on such general grammatical analogies as are common to the whole range of the Scythian group of languages, it seems to me almost as improper to designate those dialects Tamilian or Dravidian, as it would be to designate them Turkish or Tungusian. Possibly they may form a link of connection between the Indo-Chinese or Tibetan family of tongues, and the Kolarian ; but even this is at present little better than an assumption. Professor Max Miiller proposed to call all the non- Aryan languages of India, including the Sub-Himalayan, the K61, and the Tamilian families, Nishada languages, the ancient aborigines being often termed Nishadas in the Puranas. Philologically, I think, the use of this common term is to be deprecated, inasmuch as the Dravidian languages differ so widely from the others, that they possess very few features in com- mon. For the present, I have no doubt that the safest common appellation is the negative one, non- Aryan, or non-Sanskritic. Brahui, the language of the mountaineers in the khanship of Kelat in Beluchistan, contains not only some Dravidian words, but a considerable infusion of distinctively Dravidian forms and idioms ; in consequence of which this language has a better claim to be regarded as Dravidian or TamiHan than any of the languages of the Nepal and Bhutan frontier, which had been styled ' Tamulian ' by Mr Hodgson. I have not included, however, the Brahui in the list of Dravidian languages which are to be subjected to systematic comparison (though I shall refer to it occasionally for illustration), because the Dravidian element contained in it bears but a small proportion to the rest of its component elements.* • * The census of 1911 includes the Brahui in the Dravidian family, and there are 170,998 persons speaking that language. There are also two other languages included in the group, namely, the Malhar spoken by 236 persons and the Kolami spoken by 24,074 persons, — Editors. 40 INTKODUCTION It is true that the great majority of the words in the Brahui language seem altogether unconnected with Dravidian roots ; but it will be evident from the analogies in structure, as well as in the vocabulary, that this language contains many grammatical forms essentially and distinctly Dravidian, together with a small proportion of important Dravidian words. The Brahuis state that their fore- fathers came from Haleb (Aleppo) ; but even if this tradition could be regarded as a credible one, it would apply to the secondary or conquering race, apparently of Indo-European origin, not to their Dravidian predecessors. The previous existence of the latter race seems to have been forgotten, and the only evidence that they ever existed is that which is furnished by the Dravidian element which has been discovered in the language of their conquerors. The Brahui enables us to trace the Dravidian race beyond the Indus to the southern confines of Central Asia. The Brahui language, considered as a whole, seems to be derived from the same source as the Panjabi and Sindhi, but it evidently contains a Dravidian element ; and the discovery of this Dravidian element in a language spoken beyond the Indus tends to show that the Dravidians, like the Aryans, the Grseco-Scythians, and the Turco-Mongolians, must have entered India by the north-western route. . The Dravidian Idioms not merely Provincial Dialects of the SAME Language. Though I have described the twelve vernacular idioms mentioned in the foregoing list as dialects or varieties of one and the same original Dravidian language, it would be erroneous to consider them as dialects in the popular sense of the term — viz., as provincial peculiarities or varieties of speech. Of all those idioms no two are so nearly related to each other that persons who speak them can be mutually understood. The most nearly related are Tamil and Malayalam ; and yet it is only the simplest and most direct sen- tences in the one language that are intelligible to those who speak only the other. Involved sentences in either language, abounding in verbal and nominal inflexions, or containing conditions and reasons, will be found, by those who speak only the other language, to be unintelligible. Tamil, Malayalam, Telugu, and Canarese, have each a distinct and independent literary culture ; and each of the three former — Tamil, Malayalam, and Telugu — has a system of written characters peculiar to itself. The modern Canarese character has been DRA VIDIAN IDIOMS NOT MERELY PROVINCIAL DIALECTS 4 1 borrowed from that of the Telugu, and differs but slightly from it ; but the Canarese language differs even more widely from Telugu than it does from Tamil ; and the Ancient Canarese character is exceed- ingly unlike the character of the Telugu. Of the six cultivated Dra vidian dialects mentioned above — Tamil, Telugu, Canarese, Malayalam, Tulu, Kuduga — the farthest removed from each other are Tamil and Telugu. The great majority of the roots in both languages are, it is true, identical ; but they are often so disguised in composition by peculiarities of inflexion and dialectic changes, that not one entire sentence in the one language is in- telligible to those who are acquainted only with the other. The various Dravidian idioms, though sprung from a common origin, are therefore to be considered, not as mere provincial dialects of the same speech, but as distinct though affiliated languages. They are as distinct one from the other as Spanish from Italian, Hebrew from Aramaic, Sindhi from Bengali. If the cultivated Dravidian idioms differ so materially from each other, it will naturally be supposed that the uncultivated idioms — Tuda, Kota, Gond, Khond, and the Oraon — must differ still more widely both from one another and from the cultivated languages. This supposition is in accordance with facts. So many and great are the differences and peculiarities observable amongst these rude dialects, that it has seemed to me to be necessary to prove, not that they differ, but that they belong, notwithstanding their differences, to the same stock as the more cultivated tongues, and that they have an equal right to be termed Dravidian. The Dravidian Languages independent of Sanskrit. It was supposed by the Sanskrit Pandits (by whom everything with which they were acquainted was referred to a Brahmanical origin), and too hastily taken for granted by the earlier European scholars, that the Dravidian languages, though differing in many particulars from the North Indian idioms, were. equally with them derived from the Sanskrit. They could not but see that each of the Dravidian languages to which their attention had been drawn con- tained a certain proportion of Sanskrit words, some of which were quite unchanged, though s<5hie were so much altered as to be recog- nised with difficulty ; and though they observed clearly enough that each language contained also many non-Sanskrit words and forms, they did not observe that those words and forms constituted the 42 INTRODUCTION bulk of the language, or that it was in them that the living spirit of the language resided. Consequently they contented themselves with ascribing the non-Sanskrit portion of these languages to an ad- mixture of a foreign element of unknown origin. According to this view there was no essential difference between the ' Draviras ' and the ' Gauras '; for the Bengali and other languages of the Gaurian group appear to contain also a small proportion of non-Sanskritic words and forms, whilst in the main they are corruptions of Sanskrit. This representation fell far short of the real state of the case, and the supposition of the derivation of the Dravidian languages from San- skrit, though entertained in the past generation by a Colebrooke, a Carey, and a Wilkins, is now known to be entirely destitute of founda- tion. The orientalists referred to, though deeply learned in Sanskrit, and well acquainted with the idioms of Northern India, were un- acquainted, or but very slightly acquainted, with the Dravidian languages. No person who has any acquaintance with the principles of comparative philology, and who has carefully studied the gram- mars and vocabularies of the Dravidian languages, and compared them with those»of Sanskrit, can suppose the grammatical structure and inflexional forms of those languages and the greater number of their more important roots capable of being derived from Sanskrit by any process of development or corruption whatsoever. . The hypothesis of the existence of a remote original affinity be- tween the Dravidian languages and Sanskrit, or rather between those languages and the Indo-European family of tongues, inclusive of Sanskrit, of such a nature as to allow us to give the Dravidian lan- guages a place in the Indo-European group, is altogether different from the notion of the direct derivation of those languages from Sanskrit. The hypothesis of a remote original affinity is favoured by some interesting analogies both in the grammar and in the vocabu- lary, which will be noticed in their place. Some of those analogies are best accounted for by the supposition of the retention by the Dravidian family, as by Finnish and Turkish, of a certain number of roots and forms belonging to the prse-Aryan period, the period which preceded the final separation of the Indo-European group of tongues from the Scythian. I think I shall also be able to prove, with respect to one portion at least of the analogies referred to, that instead of the Dravidian languages having borrowed them from Sanskrit, or both having derived them from a common source, Sanskrit has not disdained to borrow them from its Dravidian neigh- bours. Whatever probabiHties may be in favour of the hypothesis DRAVIDIAN LANGUAGES INDEPENDENT OF SANSKRIT 43 now mentioned, the older supposition of the direct derivation of the Dravidian languages from Sanskrit, in the same manner as Hindi, Bengali, and the other Gaurian dialects are directly derived from it, was certainly erroneous. (1.) It overlooked the circumstance that the non-Sanskritic portion of the Dravidian languages was very greatly in excess of the Sanskrit. (2.) It overlooked the still more material circumstance that the pronouns and numerals of the Dra- vidian languages, their verbal and nominal inflexions, and the syntactic arrangement of their words — everything, in short, which constitutes the living spirit of a language — were originally and radically different from Sanskrit. (3.) The orientalists who held the opinion of the derivation of the Dravidian languages from Sanskrit, relied mainly on the circumstance that all dictionaries of Dravidian languages contained a large number of Sanskrit words scarcely at all altered, and a still larger number which, though much altered, were evidently Sanskrit derivatives. They were not, however, aware that such words are never regarded by native scholars as of Dravidian origin, but are known and acknowledged to be derived from Sanskrit, and that they are arranged in classes, according to the degree in which they have been corrupted, or with reference to the medium through which they have been derived. They were also unaware that true Dravidian words, which form the great majority of the words in the southern vocabularies, are placed by native grammarians in a different class from the above-mentioned derivatives from Sanskrit, and honoured with the epithets ' national words ' and ' pure words.' In general no difficulty is felt in distinguishing Sanskrit derivatives from the ancient Dravidian roots. There are a few cases only in which it may be doubtful whether particular words are Sanskrit or Dravidian — e.g., nir, water, and min, fish, are claimed as component parts of both languages, though I believe that both are of Dravidian origin. 44 INTRODUCTION COMPARATIVE LIST of Sixty Words of Primary Importance (not iNCLUDiNa Pronouns and Numerals) in Sanskrit AND Tamil. Sanskrit. Tamil. Sanskrit. Tamil. father, fitri, aj)pa{n). dog, svan. ndy. mother. mdtri, dyi cat, viddla, funei. son, sunu, maga{n). tiger. vydghra. kadu-vdy. daughter, duhitri, maga{l). deer. mriga,, mdn. head. siras, talei. monkey, kapi. kurang-u. eye, aksJii, kan. bear. sriksha. karadi. ear. karna, sevi. hog. sukara, panti. mouth, mukha, vdy. snake,' sarpa, pdmhu. tooth, danta, pal. bird. vayas. favavei. hair, kesa, mayir. black, kdla, kar-u. hand, l hasta, kara,^ kci. white, sukla, vet. red, rakta. he. foot. fad, kdl. great, TYiahat, fer-u. sun. surya. ndyiv-u. small, alpa. sit-u. moon. chandra, tingal. sweet. madhura, in. sky, div. van. sour, . amla. puli. day. divasa. ndl. salt, lavana, uppu. night, nak, iravu. cat. bhaksh. tin. fire, agni. tt. drink, fd, kudi. water. ap, nira,^ nir. come, e. vd. fish, { matsya, ] 7nina,* J mm. go. stand. gam,, sthd, p6. nil. hill. farvata, malei. sit, ds. ir-u. tree. drunia, maram. walk. char, eg-u. stone. asman, kal. run, dru. 6d-u. house. vesman. it. sleep. svap, uvang-u. \ village. grama, ur. hear, sru, kel. elephant, hastin, dnei. tell, vad, sol. horse. asva. kudirei. laugh, has, nagei. j cow. go, d. weep, rud, ar-u. buffalo, maliisha, erumei. kill. han, kol. See "Glossarial Affinities," I. t See "Glossarial Affinities," II. DRAVIDIAN LANGUAGES INDEPENDENT OF SANSKRIT 45 (4.) The orientalists who supposed the Dra vidian languages to be derived from Sanskrit were not aware of the existence of uncultivated languages of the Dra vidian family, in which Sanskrit words are not at all, or but very rarely, employed ; and they were also not aware that some of the Dravidian languages which make use of Sanskrit derivatives, are able to dispense with those derivatives altogether, such derivatives being considered rather as luxuries or articles of finery than as necessaries. It is true it would now be difficult for Telugu to dispense with its Sanskrit : more so for Canarese ; and most of all for Malayalam : — those languages having borrowed from Sanskrit so largely, and being so habituated to look up to it for help, that it would be scarcely possible for them now to assert their inde- pendence. Tamil, however, the most highly cultivated ah intra of all Dravidian idioms, can dispense with its Sanskrit altogether, if need be, and not only stand alone, but flourish without its aid. The ancient or classical dialect of the Tamil languages, called Shen- Tamil (S'en-Damir) or correct Tamil, in which nearly all the literature has been written, contains exceedingly little Sanskrit ; and differs from the colloquial dialect, or the language of prose, chiefly in the sedulous and jealous care with which it has rejected the use of San- skrit derivatives and characters, and restricted itself to pure Ancient Dravidian sounds, forms, and roots. So completely has this jealousy of Sanskrit pervaded the minds of the educated classes amongst the Tamilians, that a Tamil poetical composition is regarded as in accordance with good taste and worthy of being called classical, not in proportion to the amount of Sanskrit it contains, as would be the case in some other dialects, but in proportion to its freedom from Sanskrit ! The speech of the very lowest classes of the people in the retired country districts accords to a considerable extent with the classical dialect in dispensing with Sanskrit derivatives. In every country it is in the poetry and in the speech of the peasantry that the ancient condition of the language is best studied. It is in studied Tamil prose compositions, and in the ordinary speech of the Brahmans and the more learned Tamilians, that the largest infusion of Sanskrit is contained ; and the words that have been borrowed from Sanskrit are chiefly those which express abstract ideas of philo- sophy, science, and religion, together with the technical terms of the more elegant arts. Even in prose compositions on religious sub- jects, in which a larger amount of Sanskrit is employed than in any other department of literature, the proportion of Sanskrit which has found its way into Tamil is not greater than the amount of Latin 46 INTRODUCTION contained in corresponding compositions in English. Let us, for example, compare the amount of Sanskrit contained in the Tamil translation of tlie Ten Commandments with the amount of Latin which is contained in the English version of the same formula, and which has found its way into it, either directly from ecclesiastical Latin, or indirectly, through the medium of Norman-French. Of forty- three nouns and adjectives in the English version twenty-nine are Anglo-Saxon, fourteen Latin : of fifty- three nouns and adjectives in Tamil (the difference in idiom causes this difference in the number) thirty-two are Dra vidian, twenty-one Sanskrit. Of twenty verbs in English, thirteen are Anglo-Saxon, seven Latin : of thirty-four verbs in Tamil, twenty-seven are Dravidian, and only seven Sanskrit. Of the five numerals which are found in English, either in their cardinal or their ordinal shape, all are Anglo-Saxon : of the six numerals found in Tamil, five are Dravidian, one (' thousand ') is Sanskrit. Putting all these numbers together for the purpose of ascertaining the percentage, I find that in the department of nouns, numerals, and verbs, the amount of the foreign element is in both instances the same — viz., as nearly as possible forty-five per cent. In both instances, also, all the pronouns, prepositions, adverbs, and con- junctions, and all the inflexional forms and connecting particles, are the property of the native tongue. Archbishop Trench's expressions respecting the character of the contributions which our mother-English has received from Anglo* Saxon and from Latin respectively, are exactly applicable to the relation and proportion which the native Dravidian element bears to the Sanskrit contained in Tamil. " All its joints, its whole arti- culation, its sinews, and its ligaments, the great body of articles, pronouns, conjunctions, prepositions, numerals, auxiliary verbs, all smaller words which serve to knit together, and bind the larger into sentences, these, not to speak of the grammatical structure of the language, are exclusively Anglo-Saxon (Dravidian). The Latin (Sanskrit) may contribute its tale of bricks, yea, of goodly and polished hewn stones, to the spiritual building, but the mortar, with all that holds and binds these together, and constitutes them into a house, is Anglo-Saxon (Dravidian) throughout." Though the proportion of Sanskrit which we find to be contained in the Tamil version of the Ten Commandments happens to corre- spond so exactly to the proportion of Latin contained in the English version, it would be an error to conclude that the Tamil language is as deeply indebted to Sanskrit as English is to Latin. Tamil can DRAVIDIAN LANGUAGES INDEPENDENT OF SANSKRIT 47 readily dispense with the greater part or the whole of its Sanskrit, and by dispensing with it rises to a purer and more refined style ; whereas English cannot abandon its Latin without abandoning perspicuity. Anglo-Saxon has no synonyms of its own for many of the words it has borrowed from Latin ; so that if it were obliged to dispense with them, it would, in most cases, be under the necessity of using a very awkward periphrasis instead of a single word. Tamil, on the other hand, is peculiarly rich in synonyms ; and generally it is not through any real necessity, but from choice and the fashion of the age, that it makes use of Sanskrit. If the Ten Commandments were expressed in the speech of the lower classes of the Tamil people, the proportion of Sanskrit would be very greatly diminished ; and if we wished to raise the style of the translation' to a refined and classical pitch, Sanskrit would almost entirely disappear. Of the entire number of words contained in this formula there is only one which could not be expressed with faultless propriety and poetic ele- gance in equivalents of pure Dra vidian origin. That word is ' image ' ! Both word and thing are foreign to primitive Tamil usages and habits of thought, and were introduced into the Tamil country by the Brahmans, with the Puranic system of religion and the worship of idols. Through the predominant influence of the religion of the Brahmans, the majority of the words expressive of religious ideas in actual use in modern Tamil are of Sanskrit origin, and though there are equivalent Dravidian words which are equally appropriate, and in some instances more so, such words have gradually become obsolete, and are now confined to the poetical dialect ; so that the use of them in prose compositions would sound affected and pedantic. This is the real and only reason why Sanskrit derivatives are so generally used in Tamil religious compositions. In the other Dravidian languages, whatever be the nature of the composition or subject-matter treated of, the amount of Sanskrit employed is considerably larger than in Tamil ; and the use of it has acquired more of the character of a necessity. This is in con- sequence of the literature of those languages having chiefly been cultivated by Brahmans. Even in Telugu the principal grammatical writers and the most celebrated poets have been Brahmans. There is only one work of note in that language which was not composed by a member of the sacred caste ; and indeed the Telugu S'udras, who constitute far excellence th^ Telugu people, seem almost entirely to have abandoned to the Brahmans the culture of their own lan- guage, with every other branch of literature and science. In Tamil, 48 INTRODUCTION on the contrary, few Brahmans have written anything worthy of preservation. The language has been cultivated and developed with immense zeal and success by native Tamilians ; and the highest rank in Tamil literature which has been reached by a Brahman is that of a commentator.* The commentary of Parimelaragar on the Kural of Tiruvaljuvar (supposed to have been a Pariar, yet the acknow- ledged and deified prince of Tamil authors) is the most classical production written in Tamil by a Brahman. Professor Wilson observes that the spoken languages of the South were cultivated in imitation of Sanskrit, and but partially aspired to an independent literature ; that the principal compositions in Tamil, Telugu, Canarese, and Malayalam, are translations or para- phrases from Sanskrit works, and that they largely borrow the phraseology of their originals. This representation is not perfectly correct, in so far as Tamil is concerned ; for the compositions that are universally admitted to be the finest in the language, viz., the Kural and the Chintamani, are perfectly independent of Sanskrit, and original in design as well as in execution ; and though it is true that Tamil writers have imitated — I cannot say translated — the Rama- yana, the Maha-bharata, and similar works, they boast that the Tamil Ramayana of their own Kambar is greatly superior to the Sanskrit original of Valmiki. (5.) Of all evidences of identity or diversity of languages the most conclusive are those which are furnished by a comparison of their grammatical structure ; and by such a comparison the inde- pendence of the Dravidian languages of Sanskrit will satisfactorily and conclusively be established. By the same comparison (at the risk of anticipating a question which will be discussed more fully in the body of the work), the propriety of placing these languages, if not in the Scythian group, yet in a position nearer that group than the Indo-European, will be indicated. The most prominent and essential differences in point of grammati- cal structure between the Dravidian languages and Sanskrit, are as follows : — (i.) In the Dravidian languages all nouns denoting inanimate sub- stances and irrational beings are of the neuter gender. The dis- tinction of male and female appears only in the pronouns of the third person ; in the adjectives (properly appellative nouns) which denote rational beings, and are formed by suffixing the pronominal termina- * This is not strictly accurate. Brahmins have contributed also to Tamil literature, devotional as well as philosophical. — Editors. DRAVIDIAN LANGUAGES INDEPENDENT OP SANSKRIT 49 tions ; and in the third person of the verb, which, being formed by suffixing the same pronominal terminations, has three forms in the singular and two in the plural, to distinguish the several genders, in accordance with the pronouns of the third person. In all other cases where it is required to mark the distinction of gender, separate words signifying ' male ' and ' female ' are prefixed ; but, even in such cases, though the object denoted be the male or female of an animal, the noun which denotes it does not cease to be considered neuter, and neuter forms of .the- pronoun and verb are required to be con- joined with it. This rule presents a marked contrast to the rules respecting gender which we find in the vivid and highly imaginative Sanskrit, and in the other Indo-European languages, but it accords with the usage of the languages of the Scythian group. (ii.) Dra vidian nouns are inflected, not by means of case-termina- tions, but by means of suffixed post-positions and separable particles. The only difference between the declension of the plural and that of the singular, is that the inflexional signs are annexed in the singular to the base, in the plural to the sign of plurality, exactly as in the Scythian languages. After the pluralising particle has been added to the base, all nouns, irrespective of number and gender, are declined in the same manner as in the singular. (iii.) Dra vidian neuter nouns are rarely pluralised ; neuter plurals eve still more rare in the inflexions of the verb. (iv.) The Dravidian dative ku, hi, or ge, bears no analogy to any dative case-termination which is found in Sanskrit or in any of the Indo-European languages ; but it corresponds to the dative of the Oriental Turkish, to that of the language of the Scythian tablets of Behistun, and to that of several of the languages of the Finnish family. (v.) In those connections in which prepositions are used in the Indo-European languages, the Dravidian languages, with those of the Scythian group, use post-positions instead — which post-positions do not constitute a separate part of speech, but are simply nouns of relation or quality, adopted as auxiliaries. All adverbs are either nouns or the gerunds or infinitives of verbs, and invariably precede the verbs they qualify. (vi.) In Sanskrit and the Indo-European tongues, adjectives are declined like substantives, and agree with the substantives to which they are conjoined in gender, number, and case. In the Dravidian languages, as in the Scythian, adjectives are incapable of declension. When used separately as abstract nouns of quality, which is the 50 INTKODUCTION _ ^ original and natural character of Dra vidian adjectives, they are sub- ject to all the inflexions of substantives ; but when they are used adjectivally — i.e., to qualify other substantives — they do not admit any inflexional change, but are simply prefixed to the nouns which they qualify. (vii.) It is also a characteristic of these languages, as of the Mon- golian, the Manchu, and several other Scythian languages, in contra- distinction to the languages of the Indo-European family, that, wherever it is practicable, they use as adjectives the relative parti- ciples of verbs, in preference to nouns of quality, or adjectives properly so called ; and that in consequence of this tendency, when nouns of quality are used, the formative termination of the relative participle is generally suffixed to them, through which suffix they partake of the character both of nouns and of verbs. (viii.) The existence of two pronouns of the first person plural, one of which includes, the other excludes, the party addressed, is a pecu- liarity of the Dravidian dialects, as of many of the Scythian lan- guages ; but is unknown to Sanskrit and the languages of the Indo-European family. The only thing at all resembling it in these languages is their use of the dual. (ix.) The Dravidian languages have no passive voice. The passive is expressed by auxiliary verbs signifying ' to suffer,' &c. (x.) The Dravidian languages like the Scythian, but unlike the Indo-European, prefer the use of continuative participles to con- junctions. (xi.) The existence of a negative as well as an affirmative voice in the verbal system of these languages, constitutes another essential point of difference between them and Sanskrit : it equally con- stitutes a point of agreement between them and the Scythian tongues. (xii.) It is a marked peculiarity of these languages, as of the Mon- golian and the Manchu, and in a modified degree of many other Scythian languages, that they make use of relative participles in- stead of relative pronouns. There is no trace of the existence of a relative pronoun in any Dravidian language except the Gond alone, which seems to have lost its relative participle, and uses instead the relative pronoun of the Hindi. The place of such pronouns is supplied in the Dravidian languages, as in the Scythian tongues mentioned above, by relative participles, which are formed from the present, preterite, and future participles of the verb by the addition of a formative suffix ; which suffix is in general identical with the sign of the possessive case. Thus, * the person who came ' is in Tamil DRAVIDIAN LANGUAGES INDEPENDENT OF SANSKRIT . 5 I vand-a dl, literally 'the who-came person'; vand-u, .Wi^ preterite verbal participle signifying ' having come,' being converted into a relative participle, equivalent to ' the-who-came,' by the addition of the old possessive and adjectival suffix a. (xiii.) The situation of the governing word is characteristic of each of these families of languages. In the Indo-European family it usually precedes the word governed : in the Dra vidian and in all the Scythian languages, it is invariably placed after it ; in consequence of which the nominative always occupies the first place in the sen- tence, and the one finite verb the last. The adjective precedes the substantive : the adverb precedes the verb : the substantive which is governed by a verb, together with every word that depends upon it or qualifies it, precedes the verb by which it is governed : the relative participle precedes the noun on which it depends : the negative branch of a sentence precedes the affirmative : the noun in the genitive case precedes that which governs it : the 2?^e-position changes places with the noun and becomes a ^^os^position in virtue of its governing a case : and finally the sentence is concluded by the one, all-governing, finite verb. In each of these important and highly characteristic peculiarities of syntax, the Dravidian languages and the Scythian are thoroughly agreed.* Many other differences in grammatical structure, and many differ- ences also in regard to the system of sounds, will be pointed out here- after, in the course of the analysis ; but in the important particulars which are mentioned above, the Dravidian languages evidently differ so considerably from the languages of the Indo-European family, and in particular from Sanskrit (notwithstanding the predominance for so many ages of the social and religious influence of the Sanskrit- speaking race), that it can scarcely be doubted that they belong to a totally different family of tongues. They are neither derived from Sanskrit, nor are capable of being affiliated to it : and it cannot have escaped the notice of the student that in every one of those parti- culars in which the grammatical structure of the Dravidian languages differs from Sanskrit, it agrees with the structure of the Scythian languages, or the languages of Central and Northern Asia. In some particulars — as might be expected from the contact into which the Sanskrit-speaking race was brought with the aboriginal races of India— Sanskrit anpears to differ less widely than the other * The only exceptions to the rule respecting the position of the governing word in the Dravidian languages are found in poetical compositions, in which, occasionally, for the sake of effect, the order of words required by rule is transposed. 52 . INTRODUCTION Indo-European tongues from the languages of the Scythian group. One of these particulars — the appearance in Sanskrit of consonants of the cerebral series — will he discussed further on in connection with the Dra vidian system of sounds. Mr Edkins, in his " China's Place in Philology," has opened up a new line of inquiry in regard to the existence of Turanian influences in the grammatical structure of Sanskrit. He regards the inflexion of nouns by means of case- endings alone, without prepositions in addition, as the adoption by Sanskrit of a Turanian rule. He thinks also the position of the words in a Sanskrit prose sentence is Turanian rather than Aryan. It is an invariable law of the distinctively Turanian tongues that related sentences precede those to which they are related. It is another invariable law that the finite verb is placed at the end of the sentence. In both these particulars Mr Edkins thinks that Sanskrit has yielded to Turanian influences. This certainly seems to be the case with regard to the vernaculars which have been developed out of the old colloquial Sanskrit ; but in so far as the Sanskrit of literature is concerned, the Turanian rule is far from being universally followed. Mr Edkins himself gives an illustration from a Sanskrit prose story (p. 315), which shows that a relative clause sometimes succeeds, instead of preceding, the indicative clause, and that the position of the finite verb is not always at the end of the sentence. Perhaps all that can be said with certainty is that in Sanskrit prose and in prosaic verse related sentences generally precede, and the finite verb generally comes last. Up to this point, therefore, it may perhaps fairly be held that Turanian influences have made themselves felt even in Sanskrit. We are safer, however, in dealing with facts than with causes ; for on this theory it might be necessary to hold that Latin syntax is more ' Turanian ' than Greek, and German more * Turanian ' than English. Is THERE A Dra VIDIAN ELEMENT IN THE VERNACULAR LANGUAGES OF Northern India ? The hypothesis of the direct derivation of the Dravidian tongues from Sanskrit, with the admixture of a proportion of words and forms from an unknown source, having been found untenable, some oriental scholars adopted an opposite hypothesis, and attributed to the influence of the Dravidian languages that corruption of Sanskrit out of which the vernaculars of Northern India have arisen. It was supposed by the Rev. Dr Stevenson, of Bombay,* Mr Hodgson, of * Journal of the Asiatic Society of Bombay. NON-SANSKRIT ELEMENT IN NORTHERN VERNACULARS 53 Nepal,* and some other orientalists, (1) that the North-Indian vernaculars had been derived from Sanskrit, not so much by the natural process of corruption and disintegration, as through the overmastering, remoulding power of the non-Sanskritic element contained in them ; and (2) that this non-Sanskritic element was identical with the Dravidian speech, which they supposed to have been the speech of the ancient Nishadas, and other aborigines of India. The first part of this hypothesis appears to rest upon a better foundation than the second ; but even the first part appears to me to be too strongly expressed, and to require considerable modification ; for in some important particulars the corruption of Sanskrit into Hindi, Bengali, &c., has been shown to have arisen from that natural process of change which we see exemplified in Europe, in the corruption of Latin into Italian and Spanish. Nevertheless, on comparing the grammatical structure and essential character of. Sanskrit with those of the vernaculars of Northern India, I feel persuaded — though here I am off my own ground, and must express myself with difiidence — that the direction in which those vernaculars have been differentiated from Sanskrit has to a considerable extent been non- Aryan, and that this must have been owing, in what way soever it may have been brought about, to the operation of non- Aryan influences. The modifications which the grammar of the North Indian lan- guages has received, being generally of one and the same character, and in one and the same direction, it may be concluded that there must have been a common modifying cause ; and as the non- Sanskritic portion of those languages, which Professor Wilson styles '' a portion of a primitive, unpolished, and scanty speech, the relics of a period prior to civilisation," has been calculated to amount to one-tenth of the whole, and in Marathi to a fifth, it seems reasonable to infer that it was, in part at least, from that extraneous element that the modifying influences proceeded. It is admitted that before the arrival of the Aryans, or Sanskrit- speaking colony of Brahmans, Kshatriyas, and Vaisyas, the greater part of Northern India was peopled by rude aboriginal tribes, called by Sanskrit writers Dasyus, Nishadas, Mlechchas, &c. ; and it is the received opinion that those aboriginal tribes were of Scythian, or at least of non- Aryan, origin. On the irruption of the Aryans, it would naturally happen that the copious and expressive Sanskrit of the * Journal of the Asiatic Society of Bengal; also "Aborigines of India," Cal- cutta, 1849. 54 INTRODUCTION conquering race would alniost overwhelm the vocabulary of the rude Scythian tongues spoken by the aboriginal tribes. Nevertheless, as the grammatical structure of the Scythian tongues possesses peculiar stability and persistency, and as the prse-Aryan tribes, who were probably more numerous than the Aryans, were not annihilated, but only reduced to a dependent position, and eventually, in most instances, incorporated in the Aryan community, it would seem almost necessarily to follow that they would modify, whilst they adopted, the language of their conquerors, and that this modification would consist, partly in the addition of new words, and partly also in the introduction of a new spirit and tendency. This hypothesis seems to have the merit of according better than any other with existing phenomena. Seeing that the northern verna- culars possess, with the words of the Sanskrit, a grammatical structure which in the main appears to be Scythian, it seems more correct to represent those languages as having a Scythian basis, with a large and almost overwhelming Sanskrit addition, than as having a Sanskrit basis, with a small admixture of a Scythian element. The existence of a ' Tartarean or Chaldee,' that is, of a Scythian, element in the colloquial dialects of Northern India was first asserted by Sir W. Jones C Asiatic Eesearches," vol. i.), and till of late has been generally admitted. It has recently been called in question in the Indian Antiquary (April 1872), in a paper by Mr Growse, B.C.S. His observations are confined to Hindi, and deal, not with its gram- matical principles, but with the vocabulary only ; but they prove the necessity of more extended research before the existence of any considerable amount of non-Sanskritic elements in that dialect can be regarded as certain. The second part of the hypothesis of Dr Stevenson, viz., the identity of the non-Sanskritic element contained in those languages — supposing the existence of such an element established — with the languages of the Dravidian family, rests on a different foundation, and appears to me to be less defensible. According to the supposition in question, the Scythian or Dravidian element is substantially one and the same in all the vernacular languages of India, whether northern or southern, but is smallest in amount in those districts of Northern India which were first conquered by the Aryans ; greater in the remoter districts of the Dekhan, Telingana, and Mysore ; and greatest of all in the Tamil country, at the southern extremity of the peninsula, to which the aggressions of the Brahmanical race had scarcely extended in the age of Manu and the Eamayana. NON-SANSKRIT ELEMENT IN NORTHERN VERNACULARS 55 This hypothesis certainly appears at first sight to accord with the current of events in the ancient history of India ; but whatever relationship, in point of blood and race, may originally have subsisted between the northern aborigines and the southern — whatever ethnological evidences of their identity may be supposed to exist — when we view the question philologically, and with reference to the evidence furnished by their languages alone, the hypothesis of their identity does not appear to me to have been established. It may be true that various analogies in point of grammatical structure appear to connect the non-Sanskritic element contained in the North- Indian idioms with the Scythian tongues. This connection, however (if it really exists), amounts only to a general relationship to the entire group of Scythian languages ; and scarcely any special re- lationship to the Dra vidian languages, in cohtra-distinction to those of the Turkish, the Finnish, or any other Scythian family, has yet been shown to exist. Indeed I conceive that the non-Aryan sub- stratum of the North-Indian idioms presents as large a number of points of agreement with the oriental Turkish, or with that Scythian tongue or family of tongues by which the New Persian has been modified, as with any of the Dra vidian languages. The principal particulars in which the grammar of the North- Indian idioms accords with that of the Dravidian languages are as follows : — (1), the inflexion of nouns by means of separate post-fixed particles added to the oblique form of the noun ; (2), the inflexion of the plural by annexing to the unvarying sign of plurality the same suffixes of case as those by which the singular is inflected ; (3), the use in several of the northern idioms of two pronouns of the first person plural, the one including, the other excluding, the party addressed ; (4), the use of post-positions, instead of prepositions ; (5), the formation of verbal tenses by means of participles ; (6), the situation of the relative sentence before the indicative ; (7), the situation of the governing word after the word governed. In the particulars above-mentioned, the grammar of the North-Indian idioms undoubtedly resembles that of the Dravidian family : but the argument founded upon this general agreement is to a con- siderable extent neutralised by the circumstance that those idioms accord in the same particulars, and to the same extent, with several other families of the Scytkian group. None of those particulars in which the Dravidian languages differ from the Turkish or the Mon- golian (and there are many such points of difference) has as yet been discovered, so far as I am aware, in the North-Indian idioms. For 56 INTRODUCTION instance, those idioms contain no trace of the relative participle which is used in all the Dra vidian tongues, except the Gond, instead of a relative pronoun ; they are destitute of the regularly inflected negative verb of the Dravidian languages ; and they contain not one of the Dravidian pronouns or numerals — not even those which we find in the Medo-Scythic tablets of Behistun, and which still survive even in the languages of the Ostiaks, the Chinese, and the Lapps. If the non-Sanskritic element contained in the northern vernaculars had been Dravidian, we might also expect to find in their vocabularies a few primary Dravidian roots — such as the words for ' head,' ' foot,' ' eye,' ' ear,' &c. ; but I have not been able to discover any reliable analogy in words belonging to this class. The only resemblances which have been pointed out are those which Dr Stevenson traced in a few words remote from ordinary use, and on which, in the absence of analogy in primary roots, and especially in grammatical structure, it is impossible to place any dependence. The wideness of the difference between the Dravidian vocabulary and that of the languages of Northern India with respect to primary roots, together with the essential agreement of all the Dravidian vocabularies one with another, will appear from the following com- parative view of the pronouns of the first and second persons singular. It sometimes happens that where one form of the pronoun is used in the nominative, another survives in the oblique cases, and a third in the verbal inflexions : it also sometimes happens that the ancient form of the pronoun differs from the modern. Where such is the case I have given all extant forms a place in the list, for the purpose of facilitating comparison. Pronoun or the Gaurian Idioms. (Sanskrit primary form, aham ; secondary forms, ma, mi, m ; Turkish primary form, man.) First Person Singular. Dravidian Idioms. Tamil, nan, ydn, en, en. Canarese, an, ydn, nd, ndnu, en, ene. Tulu, ydn, yen, e. Malayalam, ndn, en, en, ena, eni, ini. - Telugu, nenu, ne, enu, e, nd, nu, ni. Hindi, main. Tuda, an, en, eni, ini. Bengali, mui. Kota, dne, en, e. Marathi^-' mi. Gond, annd, nd, dn, na. Gujarati, hun. Ku, dnu, nd, in, e. Sindhi, man. Rajmahal, en. Oraon, enan. NON-SANSKRIT ELEMENT IN NORTHERN VERNACULARS 57 Pronoun of the Second Person Singular. Gaurian Idioms. Dra vidian Idioms. (Sanskrit primary forms, Tamil, nt, nin, nun, ei, i, ay, 6y. tvam, tav, te : secondary Canarese form, si, s; Turkish pri- mary form, sen.) nin, mnu, m, nm, ay e, lye, I, i. Tulu, I, nin, ni. Malayalam, ni, nin. Hindi, tu, tun. te. Telugu, nivu, ivu, ni, nin, vu, vi. Bengali, tui, to. Tuda, ni, nin, i. Marathi, tun, tu, to.. Kota, ni, nin, i. Gujarati, tun, ta. Gond, imma, ni, i. Sindhi, tun, to. Ku, Oraon, Eajmahal, Brahui, inu, ni, i. nien. nin. ni, nd. Scythic of the Behistun tablets, ni. From the striking dissimilarity existing between the Gaurian pro- nouns and the Dra vidian, it is obvious that, whatever may have been the nature and origin of the influences by which the Gaurian languages were modified, those influences do not appear to have been dis- tinctively Dravidian. In the pronouns of almost all the North- Indian languages we may notice the Scythic termination — the obscure n, which forms the final of most of the pronouns. We cannot fail also to notice the entire disappearance of the nominative of the Sanskrit pronoun of the first person singular, and the substitution for it of the Turkish-like main or man ; but in no connection, in no number or case, in no compound or verbal inflexion, do we see any trace of the peculiar personal pronouns of the Dravidian family. Possibly further research may disclose the existence in the northern vernaculars of distinctively Dravidian forms and roots ; but their existence does not appear to me as yet to be proved ; for most of Dr Stevenson's analogies take too wide a range, and where they are supposed to be distinctly Dravidian they disappear on examina- tion. I conclude, therefore, that the non-Sanskritic portion of the northern languages cannot safely be placed in the same category with the southern, except perhaps in the sense of both being Scythian rather than Aryan. Thus far I had written in the first edition of this work. Since then the subject has been much discussed, especially in Muir's " Sanskrit 58 INTRODUCTION Texts," vol. ii., and in Beames's " Comparative Grammar of the Modern Aryan Languages of India." The general result appears to be that it remains as certain as ever — it could scarcely become more certain — that few, if any, traces of distinctively Dravidian elements are discernible in the North-Indian vernaculars. On the one hand, Dr Gundert argues strongly — not indeed for the existence of Dra- vidian elements in those vernaculars, as distinguished from their existence in Sanskrit — but for the existence of such elements in Sanskrit itself. See his remarks on this subject (from the Journal of the German Oriental Society for 1869), in the section on Glossarial Affinities. On the other hand, Mr Growse thus concludes a dis- cussion on the question of the existence of traces of a non-Aryan element in the northern vernaculars — " The foregoing considerations demonstrate the soundness of the proposition laid down in the outset, viz., that the proportion of words in the Hindi vocabulary not con- nected with Sanskrit forms is exceedingly inconsiderable ; such fact appearing — first, from the silence of the early grammarians as to the existence of any such non-Sanskritic element ; secondly, from the discovery that many of the words hastily set down as barbarous are in reality traceable to a classic source ; and, thirdly, from the unconscious adherence of the modern vernacular to the same laws of formation as influenced it in an admittedly Sanskritic stage of development." The following more extended remarks in confirmation of the same view of the subject are from Mr Beames's " Comparative Grammar " (Introduction, pp. 9-10,* § 3) : — " Next comes the class of words described as neither Sanskritic nor Aryan, but x. It is known that on entering India the Aryans found that country occupied by races of a different family from their own. With these races they waged a long and chequered warfare, gradually pushing on after each fresh victory, till at the end of many centuries they obtained possession of the greater part of the territories they now enjoy. Through these long ages, periods of peace alternated with those of war, and the contest between the two races may have been as often friendly as hostile. The Aryans exercised a powerful influence upon their opponents, and we cannot doubt but that they themselves were also, but in a less degree, subject to some influence from them. There are conse- quently to be found even in Sanskrit some words which have a very non- Aryan look, and the number of such words is much greater still * "A Comparative Grammar of the Modern Non- Aryan Languages of India," &c., by John Beames, Esq., B.C.S. London, 1872. NON-SANSKRIT ELEMENT IN NORTHERN VERNACULARS 59 in the modern languages, and there exists, therefore, a temptation to attribute to non- Aryan sources any words whose origin it is diffi- cult to trace from Aryan beginnings, " It may be as well here to point'out certain][simple and almost- obvious limitations to the application of the theory that the Aryans borrowed from their alien predecessors. Verbal resemblance is, unless supported by other arguments, the most unsafe of all grounds on which to base an induction in philology. Too many writers, in other respects meritorious, seem to proceed on Fluellen's process, ' There is a river in Macedon, and there is also moreover a river in Monmouth, and there is salmon in both.' A certain Tamil word contains a P, so does a certain Sanskrit word, and ergo, the latter is derived from the former ! Now, I would urge, that, in the first place, the Aryans were superior morally as well as physically to the aborigines, and probably therefore imparted to them more than they received from them. Moreover, the Aryans were in possession of a copious language before they came into India ; they would there- fore not be likely to borrow words of an ordinary, usual description, such as names for their clothing, weapons, and utensils, or for their cattle and tools, or for the parts of their bodies, or for the various relations in which they stood to each other. The words they would be likely to borrow would be names for the new plants, animals, and natural objects which they had not seen in their former abodes, and even this necessity would be reduced by the tendency inherent in all races to invent descriptive names for new objects. A third limitation is afforded by geographical considerations. Which were the tribes that the Aryans mixed with, either as friends or foes ? Could the bulk of them have come into frequent and close contact with the Dra vidians ; and if so, when and how ? These are questions which it is almost impossible to answer in the present state of our knowledge, but they are too important to be altogether set aside ; and it may be therefore pointed out, merely as a contribution to the subject, that the tribes driven out of the valley of the Ganges by the Aryans were almost certainly Kols to the south, and semi-Tibetans to the north. It is fair to look with suspicion on an etymology which takes us from Sanskrit to Tamil, without exhibiting a con- necting series of links through the intervening Kol tribes. If the above limitations are rigidlf applied, they will narrow very much the area within which non- Aryan forms are possible in Sanskrit and its descendants, and will force us to have recourse to a far more ex- tensive and careful research within the domain of Sanskrit itself 6o INTRODUCTION than has hitherto been made, with a view to finding in that language the origin of modern words." I coincide generally in the above remarks, especially in so far as they bear on the question of the influence of the Dra vidian languages, properly so called, on the North-Indian or Aryan vernaculars. That influence, as I have always held, must have been but slight. It is a different question whether the influences by which the Aryan vernaculars have been moulded into their present shape may not have been in some degree Scythian or at least non-Aryan. Dravidian, Scythian, and non-Aryan are not convertible terms. Mr Beames himself says, in his chapter on " Vowel Changes," p. 128, " I am not in a position to point out how far, or in what direction, Aryan vocalism has been influenced by these alien races (on the northern and eastern frontier, in Central India, and on the south) ; but that some sort of influence has been at work is almost beyond a doubt." In treating of ' the breaking down of a and a into e ' in the northern vernaculars, he says, " this seems to be one of those points where non- Aryan influences have been at work." — P. 140. In treating also of the cerebral I, he says, " This curious heavy I is very widely employed in the Dravidian group of languages, where it interchanges freely with r and d, and it is also found in the Kole family in Central India. The Marathas and Oriyas are perhaps of all the Aryan tribes those which have been for the longest time in contact with Koles and Dra vidians, and it is not surprising, therefore, to find the cerebral I more freely used by them than by others." — ^P. 245. Dr Ernest Trumpp, in his " Grammar of the Sindhi Language," maintains that the northern vernaculars exhibit decided traces of non- Aryan influences. He thinks we shall be able " to trace out a certain residuum of vocables, which we must allot to an old aboriginal language, of which neither name nor extent is now known to us, but which in all probability was of the Tatar stock of languages, and spread throughout the length and breadth of India before the irruption of the Aryan race." In confirmation of this view he ad- duces the preference of cerebral consonants to dentals. " Nearly three-fourths," he thinks, "of the Sindhi words which commence with a cerebral are taken from some aboriginal non-Aryan idiom which in recent times has been termed Scythian, but which he would prefer to call Tatar." " And this," he proceeds to say, " seems to be very strong proof that the cerebrals have been borrowed from some idiom anterior to the introduction of the Aryan languages." In noticing the aversion of the Prakrit to aspirates, he remarks that " this AFFILIATION OF DRAVIDIAN LANGUAGES 6 1 aversion seems to point to a Tatar underground current in the mouth of the common people, the Dravidian languages of the south being destitute of aspirates." He attributes also to Dravidian influences the pronunciation of cli and / in certain connections as ts and dz, by Marathi as by Telugu. To WHAT Group of Languages are the Dravidian Idioms to be AFFILIATED ? From the commencement of my Tamil studies I felt much inter- ested in the problem of the ulterior relationship of the Dravidian family of languages ; and before I was aware of the opinion which Professor Rask of Copenhagen was the first to express, I arrived by a somewhat similar process at a similar conclusion — viz., that the Dravidian languages are to be affiliated not so much to the Indo- European as to the Scythian group of tongues. I described the conclusion I arrived at as similar to Rask's, not the same, because I did not think it safe to place the Dravidian idioms unconditionally in the Scythian group, but preferred considering them more closely allied to the Scythian than to the Indo-European. In using the word ' Scythian,' I use it in the wide, general sense in which it was used by Rask, who first employed it to designate that group of tongues which comprises the Finnish, the Turkish, the Mongolian, and the Tun- gusian famiHes. All these languages are formed on one and the same grammatical system, and in accordance with the same general laws. They all express grammatical relation by the simple agglutination of auxiliary words or particles ; whilst in the Semitic languages grammatical relation is expressed by variations in the internal vowels of the roots, and in the Chinese and other isolative, monosyllabic languages, by the position of words in the sentence alone. The Indo- European languages appear to have been equally with the Scythian agglutinative in origin ; but they have come to require to be formed into a class by themselves, through their allowing their agglutinated auxiliary words to sink into the position of mere signs of inflexion. The Scythian languages have been termed by some the Tatar family of tongues, by others the Finnish, the Altaic, the MongoHan, or the Turanian ; but as these terms have often been appropriated to designate one or two families, to the exclusion of the rest, they seem too narrow to be safely employed as common designations of the entire group. The term ' Scythian ' having already been used by the classical writers in a vague, undefined sense, to denote generally the barbarous tribes of unknown origin that inhabited the northern 62 INTRODUCTION parts of Asia and Europe, it seemed to me to be the most appropriate and convenient word which was available. Professor Eask, who was the first to suggest that the Dravidian languages were probably Scythian, did little more than suggest this relationship. The evidence of it was left both by him and by the majority of succeeding writers in a very defective state. General statements of the Scythian relationship of the Dravidian languages, with a few grammatical illustrations, occupy a place in Prichard's " Researches," and have been repeated in several more recent works. Prichard himself wished to see the problem, not merely stated, but solved ; but I believe it can never be definitely solved without pre- viously ascertaining, by a careful intercomparison of dialects, what were the most ancient grammatical forms and the most essential characteristics of the Dravidian languages and of the various families of languages included in the Scythian group respectively. It was not till after I had commenced to carry the first edition of this work through the press that I became acquainted with Professor Max Miiller's treatise " On the Present State of our Knowledge of the Turanian Languages," included in Bunsen's " Outlines of the Philosophy of Universal History." Notwithstanding the great excellence of that treatise, I did not find my own work forestalled by the Professor's. His was a general survey of the whole field. It was my object to endeavour to cultivate more thoroughly one por- tion of the field, or at least to prepare it for thorough cultivation. Whilst the principal features of the Dravidian tongues are strongly marked, and whilst their grammatical principles and syntactic arrangement are of too peculiar a nature to be easily mistaken, there is much in the phonic system of these languages, in their dialectic interchanges and displacements, and in their declensional and conjugational forms, which cannot be understood without special study. In the course of the grammatical analysis and comparison of the Dravidian languages on which we are about to enter, I hope to help forward the solution of the problem of their ulterior relationship. It is a problem which has often, up to a certain point, been ingeniously elucidated, but which has never yet been thoroughly investigated. I am very far from regarding anything contained in the following work as a thorough investigation of this problem. The chief object I have in view is to contribute to a better knowledge of the Dravidian languages themselves. However interesting the question of affilia- tion may be, I regard that question as quite subsidiary to the object AFFILIATION OF DRAVIDIAN LANGUAGES 63 of the work in hand. Besides, I believe it will be found necessary for the satisfactory solution of the question, that the intercomparison of the various languages and families of languages of which the Scythian group is composed, should be carried much further than it has been carried as yet. An excellent beginning has been made in Boiler's treatises : " Die Finnischen Sprachen " and " Die Conjuga- tion in den Finnischen Sprachen," Schott's treatise " t)ber das Finnish-Tatarische Sprachengeschlecht," and Castren's " De Affixis Personalibus Linguarum Altaicarum " ; in addition to which we have now Professor Hunfalvy's paper " On the Study of the Turanian Languages," in which he carefully compares the Hungarian, Vogul, Ostiak, and Finnish, and proves that the vocabularies of those four languages are of a common origin, and that their grammars are closely related. Till, however, the comparative study of the whole of these languages has been carried still further, one term of the comparison will always be liable to be misapprehended. My know- ledge of the Scythian languages is only at second hand, and I am fully conscious of the truth of Bohtlingk's dictum, that "it is dangerous to write on languages of which we do not possess the most accurate knowledge." I trust, therefore, it will be remembered that if I advocate any particular theory on this question of affiliation, I do so with considerable diffidence. Professors Pott and Friedrich Miiller, followed by an increasing number of philologists, are unwilling to admit that the various lan- guages of the so-called Scythian or Turanian class or group have had a common origin. They admit them to be morphologically or physio- logically related, but do not concede to them any genealogical relationship. Dr Black also (Journal of the Anthropological Society, 1871) thinks it " not impossible that some or all of the Turanian languages exhibit only certain stages of development in one particular direction, taken either by members of different families, or by differ- ent branches of the same family." On the whole, however, the resemblances apparent amongst these languages, both in structure and vocabulary, as pointed out by Castren and the other writers referred to, seem to me too numerous and essential to admit of any other conclusion than that of their original oneness. " These lan- guages " appear to me, to use Professor Max Miiller's words, to " share elements in common ^hich they must have borrowed from the same source, and their formal coincidences, though of a different character from those of the Aryan and Semitic families, are such that it would be impossible to ascribe them to mere accident " {" Lecture 64 INTRODUCTION I.," 301). " The only coincidences we are likely to find," lie says, " in agglutinative languages long separated, are such as refer to ' the radical materials of language, or to those parts of speech which it is most difficult to reproduce — pronouns, numerals, and prepositions. It is astonishing rather that any words of a conventional meaning should have been discovered as the common property of the Turanian languages than that most of their words and forms should be peculiar to each.' " The various particulars which I adduced in the preceding section to prove that the Dra vidian languages are essentially different from, and independent of, Sanskrit (each of which will be considered more fully under its own appropriate head) may also be regarded as con- tributing to show, both that the various languages of the Scythian group have sprung from a common origin, and also that the Dravi- dian languages — if not actually to be included in the Scythian group — stand to that group in some sort of relationship. In some import- ant particulars the Dra vidian languages have undoubtedly approxi- mated to the Indo-European, "especially in this, that instead of continuing to be purely agglutinative they have become partly inflexional. Seyeral of the words of relation used as auxili- aries in declension and conjugation have ceased to be capable of being used as independent words. Still, it would be unnecessary on this account alone to disconnect these languages wholly from the Scythian group, for those auxili«,ry words, though they have now in some instances shrunk into the condition of fossihsed relics, are always separable from the roots to which they are appended. They have never so far coalesced with the roots — as such words have generally done in the Indo-European languages — as to form with the roots only one integral word, in which it is almost impossible to determine which is the root and which is the modificatory element. It is also to be remembered that the Turkish, Finnish, Hungarian, and Japanese languages, though in many particulars distinctively Turanian, have become still more inflexional than the Dravidian. Mr Edkins, in his " China's Place in Philology," has warmly sup- ported both the positions I have advocated — viz., the original unity of all the Scythian languages and the affiliation of the Dravidian languages on the whole to the Scythian group. A considerable number of the minute coincidences on which he relies will probably disappear on further investigation ; but the more this branch of philology is studied the more I think it will be evident that the main lines of his argument — especially with regard to the resem- AFFILIATION OF DRAVIDIAN LANGUAGES 65 blances between the Dra vidian languages and the Mongolian — are correct. I cannot say that I think the resemblances of the Dra vidian languages to the Chinese very numerous. Mr Edkins holds the original unity, not only of the Scythian languages, but of all the languages of Europe and Asia, and argues that " what arc called families of languages are only dialects of an earlier speech." This general principle seems to me to be in accordance, on the whole, with such facts as are known to us respecting the history of human speech, but it will probably be a considerable time before it is scientifically established. I may add that, to my own mind, the light which is thrown on the structure of the Dravidian languages by the study of the languages of the Scythian group has always seemed a strong confirmation of the theory of the existence in them of a Scjrthian element. The relative participle is one of the most dis- tinguishing features of the Dravidian verb ; but I never clearly understood the principle of the formation of that participle till I saw how it was formed in the Mongolian and Manchu ; and no per- son, however reluctant to see a Scythian element in the Dravidian languages, has ever, so far as I am aware, objected to the explanation of the origin of the relative participle given in the first edition of this work, or suggested another. (See " The Relative Participle," in Part v., on " The Verb.") A remarkable confirmation, on the whole, of the Scythian theory has been furnished by the translation of the Behistun tablets. The inscriptions discovered at Behistun or Baghistan, in western Media, record the poHtical autobiography of Darius Hystaspes in the Old Persian, in the Babylonian, and also in the language of the Scythians of the Medo-Persian empire ; and the translation of the Scythian portion of those inscriptions has thrown a new light on the connection of the Dravidian languages with the Scythian group. The language of the second series of tablets was shown in Mr Norris's paper (in the J ournalofthelioyal Asiatic Society, Yol. xv.) to be distinctively Scythian. Professor Oppert holds that the people by whom this language was spoken were Medians, but agrees with Mr Norris in considering the language Scythian — that is, Turanian. We are now enabled, there- fore, to compare the Dravidian idioms with a fully developed language of the Scythian family, as spoken in the fifth century B.C. : and whilst the language oithe tablets has been shown to belong generally to the Scythian group, it has been found to bear a special relationship to a particular family included in that group — the Ugro- Finnish — a family which the Dravidian dialects have long appeared 66 INTRODUCTION to me to resemble. The principal points of resemblance between the Dravidian dialects and the language of the tablets are as follows : — (1.) The language of the tablets appears to accord with the Dravi- ,dian tongues in the use of consonants of the cerebral class, t, d, and n. These sounds exist also in Sanskrit, but I have long suspected that Sanskrit borrowed them from the indigenous Dravidian languages (vide the section on '' Sounds ") ; and I find that Mr Norris has expressed the same opinion. (2.) The language of the tablets agrees with Tamil in regarding the same consonant as a surd in the beginning of a word, and as a sonant in the middle, and in pronouncing the same consonant as a sonant when single, and as a surd when doubled. (See in the section on " Sounds " illustrations of the Tamil rule.) (3.) The genitive case of the language of the tablets is formed by suffixing the syllables na, nina, or inna. The analogous forms of the Dravidian languages are ni in the Telugu, na or a in Gond or Brahui, and in in Tamil. (4.) The dative of the tablets is ikki or ikha. There are analogies to this both in the Tatar-Turkish and in the Ugrian families ; but the form which is most perfectly in accordance with it is that of the Dra- vidian dative suffix hu, hi, ha, Sec, preceded as the suffix generally is in Tamil and Malayalam, by an euphonic u or i, and a consequent doubling of the h. Compare nin-ihha, to thee, in the language of the tablets, with the corresponding nin-a-ge, in Canarese, and especially the Malay a]am nin-a-hhu. (5.) The pronouns of the language of the tablets form their accusa- tive by suffixing un, in, or n. Compare the Telugu accusative in- flexion nu or ni, and the Canarese mn, ann-u, &c. (6.) The only numeral written in letters in the Scythian tablets is hir, one, with which appears to be connected the numeral adjective, or indefinite article, ra, or irra. In Telugu, ' one ' is oka, and in Tamil or. The Ku numeral adjective ' one ' is ra, corresponding to the Tamil oru, but more closely to the ra or irra of the tablets. In the language of the tablets all ordinal numbers end in im, in Tamil in dm, in Samoiede in im. (7.) The pronoun of the second person is exactly the same in the language of the inscriptions as in the Dravidian languages. In all it is ni ; the oblique form, which is also the accusative, is nin. Unfortu- nately the plural of this pronoun is not contained in the tablets — the singular having been used instead of the plural in addressing inferiors. AFFILIATION OF DRA VIDIAN LANGUAGES 6^ (8.) The language of the tablets, like the Dravidian languages, makes use of a relative participle. A relative pronoun is used in addition to the relative participle ; but Mr Norris supposes the \lse of this pronoun to be owing to the imitation of the Persian original. The particular particle which is used in the tablets in forming the relative participle differs from that which is generally used in the Dra- vidian languages ; but the position and force of this particle, and the manner in which the participle formed by it is employed, are in perfect harmony with Dravidian usage. Perhaps the use of this relative participle is the most remarkable and distinctive charac- teristic of the grammar of every unaltered dialect of the Scythian family. (9.) The negative imperative, or prohibitive, particle of the tablets is inni, in Gond minni. The conjugational system of the language of the tablets accords with that of the Hungarian, the Mordvin, and other languages of the Ugrian family, but differs considerably from the Dravidian lan- guages, which form their tenses in a simpler manner, by the addition of particles of time to the root, and which form the persons of their verbs by the addition of the ordinary pronominal terminations to the particles of time. Notwithstanding this discrepancy in the inflexions of the verbs, the resemblances shown to subsist between the language of the tablets and the Dravidian idioms, most of which are in particulars of primary importance, seem to establish the existence of a radical, though very remote, connection. From the discovery of these analogies, we are led to conclude that the Dravi- dian race, though resident in India from a period long prior to the commencement of history, originated in the central tracts of Asia — the seed-plot of nations ; and that from thence, after parting company with the Aryans and the Ugro-Turanians, and leaving a colony in Beluchistan, they entered India by way of the Indus. Whilst I regard the grammatical structure and prevailing charac- teristics of the Dravidian idioms as in the main Scythian, I claim for them also, and have always claimed, as will be seen further on, the possession of certain remarkable affinities to the Indo-European family. In so far as they may be regarded as Scythian, they are allied not to the Turkish family, or to the Ugrian, or to the Mongolian or to the Tungusian (each of which families differs materially from the others, notwithstanding generic points of resemblance), but to the group or class in which all these families are comprised. The 6S INTRODUCTION Scythian family to which, on the whole, the Dravidian languages may be regarded as most nearly allied, is the Finnish or Ugrian, with some special affinities, as it appears, to the Ostiak branch of that family ; and this supposition, which I had been led to entertain from the comparison of grammars and vocabularies alone, derives some confirmation from the fact brought to light by the Behistun tablets, that the ancient Scythic race, by which the greater part of Central Asia was peopled prior to the irruption of the Medo-Persians, belonged not to the Turkish, or to the Mongolian, but to the Ugrian stock. If we can venture to take for granted, at present, the con- clusiveness of the evidence on which this hypothesis rests, the result at which we arrive is one of the most remarkable that the study of comparative philology has yet realised. How remarkable that distinct affinities to the speech of the Dravidians of inter-tropical India should be discoverable in the language of the Finns of Northern Europe, and of the Ostiaks and other Ugrians of Siberia ; and, conse- quently, that the prae-Aryan inhabitants of the Dekhan should appear, from the evidence furnished by their language alone, in the silence of history, in the absence of all ordinary probabilities, to be alUed to the tribes that appear to have overspread Europe before the arrival of the Teutons and the Hellenes, and even before the arrival of the Celts !* What a confirmation of the statement that " God hath made of one blood all nations of men, to dwell upon the face of the whole earth " ! In weighing the reasons which may be adduced for affiliating the Dravidian languages in the main to the Scythian group, it should be borne in mind that whilst the generic characteristics of the Scythian languages are very strongly marked and incapable of being mistaken, in a vast variety of minor particulars, and especially in their vocabu- laries, the languages comprised in this family differ from one another more widely than the various idioms of the Indo-European family mutually differ. Thus, whilst in nearly all the Indo-European lan- guages the numerals are not only similar, but the same — (the San- skrit word for one being the only real exception to the rule of general identity) — not only do the numerals of every Scythian family differ so widely from those of every other as to present few or no points of resemblance, but even the numerals of any two languages of the same * Professor Himfalvy does not admit that the Finno-Ugrian race arrived in Europe before the Celts, Teutons, and Slavonians. I adhere, however, to the ordinary belief prevailing amongst ethnologists, which appears to me in the main well-grounded. The late arrival of the Magyars in Hungary is of course admitted. AFFILIATION OF DRAVIDIAN LANGUAGES 69 family are found to differ very widely. So great, indeed, is the diver- sity existing amongst the Scythian tongues, that, whilst the Indo- European idioms form but one family, the Scythian tongues form not so much a family as a group of families — a group held together not by the bond of identity in details, but only by the bond of certain general characteristics which they all possess in common. The Indo-European languages may be regarded as forming but a single genus, of which each language — (Sanskrit, Zend, Old Persian, Greek, Latin, Gothic, Lithuanian, Slavonic, Celtic) — forms a species ; whilst the languages of the Scythian group, more prolific in differences, comprise at least five or six authenticated genera, each of which includes as many species as are contained in the solitary Indo-European genus, besides twenty or thirty isolated languages, which have up to this time resisted every effort to classify them. This remarkable dift'erence between the Indo-European languages and those of the Scythian stock seems to have arisen partly from the higher mental gifts and higher capacity for civilisation, with which the Indo-European tribes appear to have been endowed from the beginning, and still more from the earlier literary culture of their languages, and the better preservation, in consequence, of their forms and roots. It seems also to have arisen in part from their settled habits, in comparison with the wandering, nomadic life led by most of the Scythian tribes. But, from whatever cause this difference may have arisen, it is obvious that in weighing evidences of relationship this circumstance must be taken into account ; and that so minute an agreement of long-separated sister dialects of the Scythian stock is not to be expected as in parallel cases amongst the Indo-European dialects. Professor Max Miiller, in his " Lectures on the Science of Language," adduces many instances of the rapidity and extent of the divergence which takes place between un- cultivated dialects of the same language. Bishop Patteson also says, " In most cases the languages of two neighbouring islands may show their common derivation in their structure (the safest proof of all, I imagine), but nearly all the words will be different." — " Letter from Bishop Patteson to Professor Max Miiller." Appendix to Life. The relationship of the Dtavidian languages to the languages of the Scythian group, — whether the relation of lineal descent, or the relation of sisterhood, or the wider relationship for which I plead, — has not been universally admitted by students of Dravidian philo- 70 INTRODUCTION logy. From the brief remarks bearing on this question contained in Dr Pope's various publications, it is evident that that eminent Dra- vidian scholar considers the Dra vidian languages in the main Indo- European. In the introduction to his " Tamil Hand-Book " (Madras, 1859), he says : " The more deeply they (the South Indian languages) are studied, the more close will their affinity to Sanskrit be seen to be, and the more evident it will appear that they possess a primitive and very near relationship to the languages of the Indo- European group. Yet they are certainly not mere Prakrits, or corruptions of Sanskrit. I have always supposed that their place was among the members of the last mentioned family, and that they were probably disjecta membra of a language coeval with Sanskrit, and having the same origin with it. They certainly contain many traces of a close connection with the Greek, the Gothic, the Persian, and the other languages of the same family, in points even where Sanskrit presents no parallel." In the introduction to his " Sermon on the Mount," in four Dra vidian languages, with comparative vocabulary and inflexional tables (Madras, 1860), he says : " The writer would direct the attention of philologists to the deep-seated, radical affinities between these languages and the Celtic and Teutonic languages. Had leisure and space permitted, he was prepared to have exhibited in detail these analogies. In a next edition, or in some future work, he yet cherishes the hope of doing so. The subject of the affiliation of these languages is one which requires that further elucidation which nothing but a complete comparative lexicon could afford." The last reference he makes to the subject is in a prefatory notice to his " Outlines of the Grammar of the Tuda Language " (Bangalore, 1872), in which he says : " While agreeing in the main with Dr Caldwell, I yet think that the remarkable analogies between the Celtic and the Dravidian languages merit a more thorough investigation." I trust Dr Pope will ere long have time to favour philologers with the thorough investigation which this question undoubtedly merits. I may remark here, how- ever, that in everything he says respecting the existence of ' analo- gies,' and ' affinities,' and ' traces of a close connection ' between the Dravidian languages and various members of the Indo-European family, I not only perfectly coincide with him, but pointed out many of those particulars of agreement or resemblance myself (yet without deducing from them precisely the same conclusion) in every section of the first edition of this work. The theory I advocate, indeed, takes account of both sets of relationships — the Scythian and AFFILIATION OF DRAVIDIAN LANGUAGES /I the Indo-European — though it regards the former as, on the whole, closer and more essential. With regard to Celtic affinities in par- ticular, it is to be remembered that of all the members of the Indo- European family the Celtic is that which appears to have most in common with the Scythian group, and especially with the languages of the Finnish family — languages which may possibly have been widely spoken in Europe previously to the arrival of the Celts. It will be necessary, therefore, in each case to inquire whether the Celtic affinity may not also be a Scythian affinity. At the very outset of my own inquiries, I thought I observed in the Dravidian languages the Indo-European analogies to which I have referred ; and, rejecting affinities which are unreal and which dis- appear on investigation (such as the connection of the Tamil numerals ondru or onnu, one ; anju, five ; ettu, eight ; with un-us, panch-an, and asht-an, — a connection which looks very plausible, but appears to me to be illusory (see section on " Numerals ") — I think it highly probable that a small number of the grammatical forms of the Dravidian languages and a more considerable number of their roots, are to be regarded as of cognate origin with corre- sponding forms and roots in the Indo-European languages. Not- withstanding the existence of a few analogies of this character, the most essential features of the grammar of the Dravidian idioms seem to me to be undoubtedly Scythian, and therefore I think the propriety of placing those idioms in the Scythian group is indicated. Though many Hebrew roots have been shown to be allied to Sanskrit, yet the Hebrew language does not cease to be regarded as Semitic rather than Indo-European ; so, notwithstanding many interesting analogies with Sanskrit, Greek, Gothic, Celtic, and Persian, which may be discovered on a careful examination of the Dravidian tongues and which will be pointed out in their order in each of the succeeding sections, the essential characteristics of those tongues are such as seem to me to require us to regard them as in the main Scythian. Dr Gustave Schlegel, in his *' Sinico-Aryaca " (Batavia, 1872), a treatise on Chinese and Aryan affinities, endeavours to establish the existence of an ultimate relationship between the Chinese roots and those of the Aryan languages. Supposing this point established, it would not follow that Chinese is an Aryan tongue . It would only follow that it had succeeded in preserving certain exceedingly primitive forms of speech, which had also been preserved in the languages of the Aryan family. Not Chinese only, but Sanskrit and Hebrew, are now known to have been originally monosyllabic ; and the mono- ^2 INTRODUCTION syllabic character of most Dravidian roots, if not of all, will appear in every section of this work. Dr Bleek (in a paper in the Journal oj the Anthropological Society for 1871) has thrown out the idea that the Aryan family of languages may possibly have been exposed at an early period to Dra vidian influences. He says : " The Aryan are distinguished from the other sex-denoting languages by the pos- session of a neuter gender. The Dra vidian languages possess a neuter gender, which has as wide a range as in English, the most logically arranged of the Aryan languages. The distinctive marks of the neuter gender, in the Dravidian languages, even agree with those of our own languages to so great an extent that it does not appear probable that these two circles of languages (which are the only ones known to possess this threefold gender — i.e., mas- culine, feminine, and neuter) should have developed the neuter gender quite independently of each other. The Dravidian languages have not as yet been proved to belong to our own sex-denoting family of languages ; and although it is not impossible that they may be shown ultimately to be a member of this family, yet it may also be that at the time of the formation of the Aryan languages a Dravidian influence was exerted upon them, to which this, among other similarities, is due." The Dravidian languages had a neuter pronoun of the third person at the earliest period to which their forms can be traced ; but I suspect it was at a later period of their history that gender made its appearance in the verb. When the Dra vidians entered India their verb must, I think, have been without personal terminations, and therefore without gender. It will be seen hereafter that gender is more fully and systematically de- veloped in the verb of the Dravidian literary dialects than in any other language in the world. This could not have been owing to the influence of Sanskrit, but must have been ah intra. In stating that the Dravidian languages contain certain roots and forms allied to Sanskrit, and to the Indo-European languages gene- rally, it is necessary to preclude misapprehension. During the long period of the residence of the Dravidian and Aryan races in the same country, the Dravidian vocabularies have borrowed largely from Sanskrit. It is necessary therefore to remind the reader that the analogies to which I refer are not founded on the existence in the Dravidian tongues of Sanskrit derivatives, but are such as are discoverable in the original structure and primitive vocabulary of those languages. Whilst the Dravidian languages have confessedly borrowed much from their more wealthy neighbours, Sanskrit, in AFFILIATION OF DRAVIDIAN LANGUAGES 73 some instances, has not disdained to borrow from the Dra vidian : but in general there is no difficulty in distinguishing what the one language has borrowed from the other ; and the statement I have now made relates not to derivatives, or words which may be supposed to be derivatives, but to radical, deep-seated analogies which it is difficult to explain on any supposition but that of a partial or distant relationship. In most instances the words and forms in which analogies are discoverable are allied not to Sanskrit alone, but to the entire Indo-European family : in not a few instances analogies are discoverable in Greek and Latin, which are not found in Sanskrit ; and in many of those instances in which Sanskrit appears to exhibit the closest analogy, it is not the euphonised, systematised Sanskrit (Samskrita) of written -compositions, but the crude, original Sanskrit, which is discoverable by analysis and comparison — the Vor-Sanskrit of W. von Humboldt. I subjoin here a few illustrations of what I mean by primitive, un- derived Indo-Europeanisms discoverable in the Dravidian languages. (1.) The use of n, as in Greek, to prevent hiatus. (2.) The existence of gender in the pronouns of the third person and in verbs, and in particular the existence of a neuter gender. (3.) The use of d or t as the sign of the neuter singular of demon- strative pronouns or pronouns of the third person. (4.) The existence of a neuter plural, as in Latin, in short a. (5.) The formation of the remote demonstrative from a base in a, the proximate from the base in i. (6.) The formation of most preterites, as in Persian, by the addition of d. (7.) The formation of some preterites by the reduplication of a por- tion of the root. (8.) The formation of a considerable number of verbal nouns by lengthening the vowel of the verbal root. See also " Glossarial Affinities." The illustrations given above form only a small portion of the analogous forms which will be adduced in the grammatical analysis and in the glossarial affinities : they will, however, suffice to render it probable that Indo-European analogies are really discoverable in the Dravidian languages. They also serve to illustrate the statement that, though Sanskrit has^iong been the nearest neighbour of the Dravidian tongues, there are not a few Dravidian roots which seem more nearly allied to the w^estern Indo-European idioms than to the Sanskritic or eastern. If therefore the Dravidian languages may be 74 INTKODUCTION classified, as I am still inclined to classify them, as essentially and in the main Scythian, I must add that I consider them as of all Scythian tongues those which present the most numerous, ancient, and inter- esting analogies to the Indo-European languages. The position which this family occupies, if not midway between the two groups, seems to me to lie on that side of the Scythian group on which the Indo-European appears to have been severed from it, and on which the most distinct traces of the original identity of the families still remain. If this view be correct (as I think it will be shown to be), the Indo-Europeanisms discoverable in the Dravidian languages carry us back to a period beyond all history, beyond all mythology, not only prior to the separation of the western branches of the Indo- European race from the eastern, but prior also to the separation of the yet undivided Indo-European race from that portion of the common stock which was afterwards styled Scythian. It is a curious circumstance that in the vocabulary of the Dra- vidian languages, especially in that of Tamil, a few Semitic analogies may also be discovered. In some instances the analogous roots are found in the Indo-European family, as well as in Hebrew, though the Hebrew form of the root is more closely analogous. For example, though we find in Latin ave-o, to desire, and in Sanskrit av, of which ' to desire ' is a subordinate meaning ; yet the corresponding Tamil words avd, desire, and aval (signifying also desire, a verbal noun from a lost verb dv-u, to desire) seem still more directly allied to the Hebrew dvah, to desire, and the verbal noun avvdh, desire. In addition, however, to such general analogies as pervade several families of tongues, including the Dravidian, there are a few roots discoverable, I think, both in the Dravidian languages and in Hebrew to which I am not aware of the existence of any resemblance in any language of the Indo-European family. Illustrations of these special analogies will be found under the head of '' Glossarial Affinities : Semitic." The Semitic analogies observable in Tamil are neither so numerous nor so important as the Indo-European, nor do they carry with them such convincing evidence ; but taking them in connection with that more numerous and important class of analogous roots which are found in the Indo-European languages, as well as in Hebrew, but of which the Hebrew form is more closely allied to the Dravidian (see the " Glossarial Affinities "), these analogies, such as they are, constitute an additional element of interest in the problem of the origin and prse-historic connections of the Dravidian race. I do not AFFILIATION OF DR A VIDIAN LANGUAGES 75 adduce these analogies for the purpose of endeavouring to prove the existence of any relationship between the Dravidian language and Hebrew. Aware of the danger of proving nothing by attempting to prove too much, I content myself with merely stating those analogies, without attempting to deduce any inference from them. The Indo-European analogies are so intimately connected with the individuality and vital essence of the Dravidian languages, that it seems difficult to suppose them to be merely the result of early association, however intimate. It is only on the supposition of the existence of a remote or partial relationship that they appear to be capable of being fully explained. In the case of the Semitic analogies, however, the supposition of a relationship between the two families of tongues does not appear to be necessary. The analogies that appear to exist may be only accidental, or they can be accounted for on the hypothesis — a very easy and natural one — that the primitive Dravidians were at some early period before their arrival in India associated with a people speaking a Semitic language. It seems proper here to notice the remarkable general resemblance which exists between the Dravidian pronouns and those of the aboriginal tribes of southern and western Australia. In whatever way it may be explained, the existence of a general resemblance seems to be unquestionable ; but it has not hitherto been observed that the Australian pronouns of the first person are more nearly allied to the Tibetan than to the Dravidian. This will appear from the following comparative view of the pronoun of the first person singular. Dravidian. Australian. Tibetan. Chinese. I, nan, ydn, nd, nga, ngaii, ngatsa, nga, nge, nged. ngo. en. nganya. Whilst the base of this pronoun seems to be closely allied to the corresponding pronoun in Tibetan, and in the Indo-Chinese family generally, the manner in which it is pluralised in the Australian dialects bears a marked resemblance to the Dravidian, and especially to Telugu. Telugu forms its plurals by suffixing lu to the singular ; the Australian dialects by a similar addition of lu, li, dlu, dli, &c. In this particular some of the dialects of the north-eastern frontier of India exhibit also an agrefment with Telugu — e.g., compare Dhimal nd, thou, with nyel, you. In the Australian dialects I find the follow- ing plurals and duals of the pronoun of the first person — we, or we two, ngalu, ngadlu, ngadli, ngalata, &c. Compare this with the ^6 INTRODUCTION manner in which the Telugii forms its plural — e.g., vdnd'u, he, vdfidlu, they ; and even with the Tamil ' plural exclusive ' of the pronoun of the first person — e.g., nan, I, ndngal, we. The resemblance between the Australian pronouns of the second person, both singular and plural, and those of the Dra vidian languages is more distinct and special, and is apparent, not only in the suffixes, but in the pronominal base itself. The normal forms of these pronouns in the Dra vidian languages are — singular, nin, plural, nim. The per- sonality resides in the crude root n%, thou, which is the same in both numbers, with the addition of a singular formative n [nin, thou), and a pluralising formative m {m-m, thous, or you). In some cases the pluralising particle m has been displaced, and r, which I regard as properly the sign of the epicene plural of the third person, has been substituted for it — e.g., mr, you (in Telugu mir-u). This abnormal form nvr is most used as a nominative, the older and more regular nim retains its place in the compounds. Whilst i is the vowel which is almost invariably found in the singular of the pronoun of the second person, it is found that in the plural i often gives place to u, as in the classical Tamil nmna, your, and the Brahui nwn, you. It is to be noticed also that the modern Canarese has softened nim into nivu or nuvu, in the nominative. It is singular, in whatever way it may be accounted for, that in each of the particulars now mentioned the Australian dialects resemble the Dra vidian. See the following comparative view. Under the Australian head I class the dual together with the plural, as being substantially the same. Dravidian. Australian. thou, nm, nin. ninna, nginne, ngintoa, ningte. you, nim, nim, nir, num, ntvu. nimedoo, nura, niwa, ngurle. Compare also the accusative of the first person singular in Tamil, ennei, me, with the Australian accusative emmo. The grammatical structure of the Australian dialects exhibits a general agreement with the languages of the Scythian group. In the use of postpositions instead of prepositions ; in the use of two forms of the first person plural, one inclusive of the party addressed, the other exclusive ; in the formation of inceptive, causative, and reflective verbs by the addition of certain particles to the root; and, generally, in the agglutinative structure of words and in the position of words in a sentence, the dialects of Australia resemble the Dravidian — as also the Turkish, the Mongolian, and other Scythian THE REPRESENTATIVE DIALECT 77 languages ; and in the same particulars, with one or two exceptions, they differ essentially from the dialects which are called Polynesian. The vocabularies of the Australian dialects which have been compiled do not appear to furnish additional confirmation to the resemblances pointed out above ; but it is difficult to suppose these resemblances to be unreal or merely accidental, and it is obvious that the Aus- tralian dialects demand (and probably will reward) further examination.* It is singular also, and still more difficult to be accounted for, that some resemblances may be traced between the Dravidian languages and the Bornu, or rather the Kanuri, one of the languages spoken in the Bornu country, in Central Africa. Most of the resemblances are, it is true, of a general nature — e.g., the Kanuri is agglutinative in structure, it uses postpositions instead of prepositions, it adds to nouns and sentences syllables expressive of doubt, interrogation, and emphasis, in a peculiarly Dravidian manner, and its verb has a negative voice. It has an objective verb, as well as a subjective, like the Hungarian. The most distinctive resemblance to the Dravidian languages I notice is in the pronoun of the second person, which is ni, as in each of the Dravidian dialects. Even this, however, as has been shown, is common to the Dravidian with Brahui, Chinese, the language of the second Behistun tablets, and the Australian dialects. The Kanuri language differs so remarkably from the rest of the African tongues, that it is very desirable that its relationship should be fully investigated. See Koelle's " Grammar of Bornu." Which Language or Dialect best represents the Primitive Condition of the Dravidian Tongues ? Before entering upon the grammatical comparison of the Dra- vidian dialects, it seems desirable to ascertain where we should look for their earliest characteristics. Some persons have been of opinion that what is called Shen-Tamil {^en-Damir), or the classical dialect of the Tamil language, is to be regarded as the best representative of the primitive Dravidian speech. Without underestimating the great value of the Shen-Tamil, I am convinced that no one dialect can be implicitly accepted as a mirror of Dravidian antiquity. A comparison of all the dialects that exist will be found our best and safest guide to a knowlec^e of the primitive speech from which the * See a paper " On the position of the Australian languages," by W. H. J. Bleek, Esq., Ph.D., read at a meeting of the Anthropological Society. London, 1871. 78 INTRODUCTION various existing dialects have diverged; and not only the Shen-Tamil, but every existing dialect, even the rudest, will be found to con- tribute its quota of help towards this end. The Tamil pronouns of the first and second persons cannot be understood without a know- ledge of Ancient or Classical Canarese ; and the Khond or Ku, one of the rudest dialects, the grammar of which was reduced to writing only a few years ago, is the only dialect which throws light on the masculine and feminine terminations of the Dra vidian pronouns of the third person. Still it is unquestionable that the largest amount of assistance towards ascertaining the primitive condition of the Dra vidian languages will be afforded by Tamil, and in particular by Shen-Tamil ; and this naturally follows from the circumstance that of all the Dravidian idioms Tamil appears to have been the earliest cultivated. (1.) Literary, classical dialects of the Dravidian Languages : To tvhat extent 7nay they he regarded as representing the primitive condition of those Languages ? It is a remarkable peculiarity of the Indian languages that, as soon as they begin to be cultivated, the literary style evinces a tendency to become a literary dialect distinct from the dialect of common life, with a grammar and vocabulary of its own. This is equally char- acteristic of the speech of the Aryans of the north and of that of the Dra vidians of the south. The relation in which Sanskrit stands to the Prakrits and the modern vernaculars is not identical with the relation in which the dead languages of Europe stand to the living languages descended from them. The so-called dead languages of Europe were at one time living tongues, spoken nearly as they were written, as, e.g., the speeches of Demosthenes and Cicero testify. When we call those languages dead, we merely mean to describe them as the speech of the dead past, not that of the living present. Sanskrit cannot properly be called a dead language in this sense. Probably it was never the actual, everyday speech of any portion of the Aryans of India at any period of their history, however remote. Its name SamsJcrita, the elaborated or developed speech, illustrates its origin. It was the language not of any race or district, but of a class — the class of bards and priests, the literary men of the first ages ; or rather it was the language of literature ; and as literary culture made progress, the language of literature became ever more THE REPRESENTATIVE DIALECT 79 copious, euphonious, and refined. If life means growth, and if growth means change, Sanskrit must be regarded as having for a long period been, not a dead, but a living tongue ; though it must be admitted that it changed slowly, like everything else in India — more slowly, doubtless, than the colloquial dialects. The Sanskrit of the Puranas differed from the Sanskrit of the Vedas ; and in the Vedas themselves the style of the later hymns differed from that of the earlier. The earliest Sanskrit extant is evidently the result of a process of refinement, originating in the literary activity of a still earlier period, of which no records survive. A composition is not necessarily ancient because written in Sanskrit ; for all through the ages, down to very recent times, all the literati of Northern and Western India, with the exception of the Buddhists, together with a considerable proportion of the literati of the South, have been accustomed to regard Sanskrit as the most orthodox vehicle for the expression of every variety of orthodox thought. " The great reformer Buddha, in the sixth century before Christ, adopted the popular speech as the vehicle of his teachings ; his successors were infected with an unbounded cacoethes scribendi (evil habit of writing), and have left behind a literature of enormous extent. Here again, however, the fatal mistake common to all Indian writers was committed. No sooner had Prakrit become the language of the Buddhists' scriptures, than it was at once regarded as sacred, and carefully preserved from change or development. It took with regard to the popular speech the same position that Sanskrit had taken in the earlier centuries. This seems to be the fate of all Indian languages : when once committed to writing they assume a literary type, and have a tendency to draw away from the vulgar living tongue of the people. In the present day we see the same process going on in Bengal. Few Bengali writers, save those whose minds have been to some extent moulded on English models of thought and feeling, are content to write as they speak. They must have something more elaborate and refined when they take pen in hand and fill their pages with pompous and artificial Sanskrit words, which they readily admit are not ' understanded of the people.' " This state of things is not peculiar to Northern India. We find precisely the same tendencies, with the same results, in the South. Each of the four cultivated Dravidian languages has split up into two dialects more or less distinct — a literary, classical dialect ; and a popular, colloquial dialect. Classical Canarese is usually 8o INTRODUCTION called ' Old Canarese ' ; but it may more properly be regarded neither as new nor as old, but simply as the language of Canarese literature, seeing that it is the language in which literary compositions seem always to have been written, at least from the twelfth century, when Ke^ava's grammar was composed, down to the present day. ' Old Malayajam ' seems to have a better title than Old Canarese to be called ' old,' inasmuch as it contains a considerable number of obsolete forms. Moreover, whilst modern Malayajam literature is intensely Sanskritic, the older literature was pervaded with the characteristics of the older or classical Tamil. The language of Telugu poetry differs considerably from that of everyday life, but it is not regarded as a different dialect, or designated by any special name. It is regarded by native Telugu scholars as differing from ordinary Telugu only in being purer and more elevated. The most appropriate name for any of the literary dialects, as it appears to me, is that by which the higher dialect of Tamil is designated. It is called Shen-Tamil (^en-Damir) — that is, classical or correct Tamil, literally ' straight Tamil,' by which nafne it is meant to be distinguished not merely from the colloquial Tamil of the masses, but still more from certain rude local dialects, said to be twelve in number, mentioned by grammarians by name, and included under the generic designation of Kodun-Damir — that is, literally, ' crooked Tamil.' The name ordinarily given by Europeans to the literary dialect of Tamil is ' High Tamil ' ; and this appears to me to be a more accurate term, on the whole, than that ordinarily given to the literary dialect of the Canarese ; for though there is a sense in which each of these literary dialects may be described as ' old,' their most essential characteristic is the extraordinary amount of polish and refinement they have received. Classical Tamil bears nearly the same relation to the actual speech of the people that Sanskrit (that is, classical Indo-Aryan) did to the ancient Prakrits, and now does to the modern Gaurian vernaculars. Even at the time the oldest extant High Tamil compositions were written, there was probably almost as wide a difference between the language of the vulgar and that affected by the literati as there is at present. It is in- conceivable that so elaborately refined and euphonised a style of language as that of the classical poems and grammars, can ever have been the actual everyday speech of any class of the people. It contains, it is true, many ancient forms ; but forms that had come to be regarded as vulgar by the time that literary culture had com- menced (no matter how great their antiquity), seem to have been ANTIQUITY OF THE TAMIL 8 1 systematically rejected. The speech of the masses may therefore contain forms and words as old as, or even older than, the corre- sponding forms and words of the literature ; and yet there is an important difference between the two to be borne in mind. No argument in favour of the antiquity of a word or form can be founded merely on the fact of its existence in the colloquial dialect ; whereas the existence of a word or form in the classical dialect, especially in the grammars and vocabularies of that dialect, proves at least that it was in existence when that dialect was fixed, which certainly cannot have been less than a thousand years ago. There is an additional presumption in favour of its antiquity in the circumstance that all poets, even the earliest, have been accustomed to regard expressions that were considered more or less archaic in their own time, as peculiarly suitable to poetical compositions. (2.) High antiquity of the literary cultivation of Tamil. The relatively high antiquity of the literary cultivation of Tamil being a matter of interest considered in itself, irrespective of its bearings on the question of Dravidian comparative grammar, I shall here adduce a few of the evidences on which this conclusion rests. 1. Classical Tamil, which not only contains all the refinements which the Tamil has received, but also exhibits to some extent the primitive condition of the language, differs more from the colloquial Tamil than the classical dialect of any other Dravidian idiom differs from its ordinary dialect. It differs from colloquial Tamil so con- siderably that it might almost be considered as a distinct language : for not only is classical Tamil poetry as unintelligible to the unlearned Tamilian as the iEneid of Virgil to a modern Italian peasant, but even prose compositions written in the classical dialect might be read for hours in the hearing of a person acquainted only with the colloquial idiom, without his understanding a single sentence. Notwithstanding this, classical Tamil contains less Sanskrit, not more, than the colloquial dialect. It affects purism and national independence ; and its refinements are all ah intra. As the words and forms of classical Tamil cannot have been invented all at once by the poets, but must have come into use slowly and gradually, the degree in which colloquial Tamil has diverged from the poetical dialect, notwithstanding the slowness with which language, like everything else, changes in the East, seems to me a proof of the high antiquity of the literary cultivation of Tamil. 82 INTRODUCTION 2. Another evidence consists in the extraordinary copiousness of the Tamil vocabulary, and the number and variety of the gramma- tical forms of Shen-Tamil. The Shen-Tamil grammar is a crowded museum of obsolete forms, cast-off inflexions, and curious anomalies. Many of these will be pointed out from time to time in the body of this work. I may here refer especially to the extreme and almost naked simplicity of some of the conjugational forms of the oldest Tamil, particularly to the existence of an uninflected form of the verb, and of another form in which only the first rudimentary traces of inflexion are seen. These particulars, as will be shown in the Part " on the Verb," seem to me to point to the arrest of the develop- ment of the Tamil verb at a very early period by the invention of writing, as in the still more remarkable instance of Chinese. The extraordinary copiousness of the Tamil vocabulary is shown by the fact that a school lexicon of the Tamil language, published by the American missionaries at Jaffna, contains no less than 58,500 words ; notwithstanding which, it would be necessary to add several thou- sands of technical terms, besides provincialisms, and thousands upon thousands of authorised compounds, in order to render the list complete. Nothing strikes a Tamil scholar more, on examining the dictionaries of the other Dravidian dialects, than the paucity of their lists of synonyms in comparison with those of Tamil. The Tamil vocabulary contains not only those words which may be regarded as appropriate to the language, inasmuch as they are used by Tamil alone, but also those which may be considered as the property of Telugu, Canarese, &c. Thus, the word used for ' house ' in ordinary Tamil is vidu ; but the vocabulary contains also, and occasionally uses, the word appropriate to Telugu, il (Tel. illu), and the dis- tinctive Canarese word, manei (Can. mana) ; besides another synonym, Jcudi, which it has in common with Sanskrit and the whole of the Finnish languages. The grammar and vocabulary of Tamil are thus to a considerable extent the common repository of Dra- vidian forms and roots. We may conclude, therefore, that the literary cultivation of Tamil dates from a period prior to that of the other idioms, and not long subsequent to the final breaking up of the language of the ancient Dra vidians into dialects. 3. Another evidence of the antiquity and purity of Tamil consists in the agreement of the ancient Canarese, the ancient Malayalam, the Tulu, and also the Tuda, Gond, and Ku, with Tamil, in many of the particulars in which modern Canarese and modern Telugu differ from it. ANTIQUITY OF THE TAMIL 83 4. The fact that in many instances the forms of Telugu roots and inflexions have evidently been softened down from the forms of Tamil, is a strong confirmation of the higher antiquity of the Tamilian forms. Instances of this will be given in the section on the phonetic system of these languages. It will suffice now to adduce, as an illustration of what is meant, the transposition of vowels in the Telugu demonstrative pronouns. The true Dra vidian demonstrative bases are a, remote, and i, proximate ; to which are suffixed the formatives of the genders, with v euphonic, to prevent hiatus. The Tamil demonstratives are avan, ille, and ivan, hie. The Telugu masculine formative answering to the Tamil an, is du, udu, or adu ; and hence the demonstratives in Telugu, answering to the Tamil avan, ivan, might be expected to be avadu, and ivadu, instead of which we find vddu, ille, and vidu, hie. Here the demonstrative bases a and i have shifted from their natural position at the beginning of the word to the middle, whilst by coalescing with the vowel of the formative, or as a compensation for its loss, their quantity has been increased. The altered, abnormal form of the Telugu is evidently the later one ; but as even the high dialect of the Telugu contains no other form, the period when the Telugu grammar was rendered permanent by written rules and the aid of written compositions, must have been subsequent to the origin of the corruption in ques- tion, and therefore subsequent to the literary cultivation of Tamil. 5. Another evidence of antiquity consists in the great corruption of many of the Sanskrit tadhhavas or derivatives found in Tamil. The Sanskrit contained in Tamil may be divided into three portions of different dates. (1.) The most recent portion was introduced by the three religious schools which divide amongst them the allegiance of the mass of the Tamil people. These are the school of the S'aiva-Siddhanta, or that of the philosophy of the Agamas, the most popular system amongst the Tamil S'udras, the school of S'ankara Acharya, the apostle of Advaita, and the chief rival of both, the school of S'ri Vaishnava, founded by Ramanuja Acharya. The period of the greatest activity and influence of those sects seems to have extended from about the eleventh century a.d. to the sixteenth ; and the Sanskrit derivatives introduced by the adherents of these systems (with the exception of a |ew points wherein change was unavoidable) are pure, unchanged Sanskrit. (2.) The school of writers, partly preceding the above and partly contemporaneous with them, by which the largest portion of the 84 INTRODUCTION Sanskrit derivatives found in Tamil were introduced, was that of the Jainas, which flourished from about the ninth or tenth century a.d. to the thirteenth.* The period of the predominance of the Jainas (a predominance in intellect and learning — rarely a predominance in political power) was the Augustan age of Tamil literature, the period when the Madura College, a celebrated literary association, appears to have flourished, and when the Kura], the Chintamani, and the classical vocabularies and grammars were written. The Sanskrit derivatives found in the writings of this period are very considerably altered, so as to accord with Tamil euphonic rules. Thus loha, Sans. the world, is changed into ulagu ; raj a, a king, into arasu. Nearly the whole of the Sanskrit derivatives found in Telugu, Canarese and Malayalam belong to the periods now mentioned, or at least they accord on the whole with the derivatives found in the Tamil of those two periods, especially the former or more recent. They are divided, according to the degree of permutation or cor- ruption to which they have been subjected, into the two classes of tat-sama, the same with it — i.e., words which are identical with Sanskrit — and tad-hhava, of the same nature with it = derived from it — i.e., words which are derived from a Sanskrit origin, but have been more or less corrupted or changed by local influences. The former class, or tatsama words, are scarcely at all altered, and generally look like words which have been used only by Brahmans, or which had been introduced into the vernaculars at a period when the Sanskrit alphabetical and phonetic systems had become naturalised, through the predominance of the later forms of Hin- duism. Sanskrit derivatives of the second class which have been altered more considerably, or tadhhava words, do not appear to have been borrowed direct from Sanskrit, but are represented by Telugu and Canarese grammarians themselves as words that have been borrowed from the Prakrits, or colloquial dialects of the Sanskrit, spoken in ancient times in the contiguous Gaura provinces. (3.) In addition to the Sanskrit tatsama and tadhhava derivatives of the two periods now mentioned — the modern Vedantic, Saiva, and Vaishnava periods, and the Jaina period — Tamil contains many derivatives belonging to the very earliest period of the literary cul- ture of the language — derivatives which are probably of an earlier date than the introduction of Sanskrit into the other dialects. The derivatives of this class were not borrowed from the northern Prakrits (though much more corrupted than even the derivatives * Modern researches point to a much earlier date than that given here. — Editors, ANTIQUITY OF THE TAMIL . 85 borrowed from those Prakrits by Canarese and Telugu), but appear to have been derived from oral intercourse with the first Brah- manical priests, scholars, and astrologers, and probably remained unwritten for a considerable time. The Sanskrit of this period is not only greatly more corrupted than that of the period of the Jainas, but its corruptions are of a different character. The Jainas altered the Sanskrit which they borrowed in order to bring it into accord- ance with Tamil euphonic rules ; whereas in the Sanskrit of the period now under consideration — the earliest period — the changes that have been introduced seem to be in utter defiance of rule. The following are instances of derivatives of this class : (a.) The Sans, srt, sacred, was altered into tiru, whilst a more recent alteration of the Sanskrit word is into sirt, siti, and si. (b.) The Sans, karman, a work, is in the Tamil of the more modern periods altered into karumam and kanmani ; but in the older Tamil it was corrupted into kam. (c.) Several of the names of the Tamil months supply us with illustrations of early corruptions of Sanskrit. The Tamil months, though now solar-siderial, are named from the old lunar asterisms, the names of which asterisms, and still more the names of the months borrowed from them, are greatly corrupted. E.g., the asterism furva-dsliddam is changed into purddam : ashddmn, also, is changed into ddam, from which is formed ddi, the Tamil name of the month July — August. The name of the asterism asvini has been corrupted into eifpasi, which is the Tamil name of the month October — November. The change of furva hhadra-pada, the Sanskrit name of one of the asterisms, into furattdsi is still more extraordinary. Purva-hhadra-pada was first changed into purattddi, the name of the corresponding asterism in Tamil ; and this, again, by the shortening of the first syllable and the change of di into si, became purattdsi, the Tamil month September — October. The corresponding names of the asterisms and months in Telugu, Canarese, &c., are pure, unchanged Sanskrit ; and hence the greater antiquity of the introduction of those words into Tamil, or at least the greater antiquity of their use in Tamil written compositions, may safely be concluded. 6. The higher antiquity of the literary cultivation of Tamil may also be inferred from Tamil inscriptions. In Karnataka and Telin- gana, every inscription of an early date and the majority even of modern inscriptions are written in Sanskrit. Even when the characters employed are those of the ancient Canarese or Telugu 86 INTRODUCTION ^characters wliicli have been arranged to express the peculiar sounds of Sanskrit), Sanskrit is the language in which the inscription is found to be written, if it is one of any antiquity. In the Tamil country, on the contrary, all inscriptions belonging to an early period are written in Tamil ; and I have not met with, or heard of, a single Sanskrit inscription in the Tamil country which appears to be older than the fourteenth century a.d., though I have obtained fac-similes of all the inscriptions I could hear of in South Tinnevelly and South Travancore — integral portions of the ancient Pandyan kingdom. The number of inscriptions I have obtained is about a hundred and fifty. They were found on the walls and floors of tem- ples, and on rocks and pillars. The latest are written in Grantha, or the character in which Sanskrit is written by the Dravida Brah- mans ; those of an earlier age either in an old form of the existing Tamil character, or in a still older character, which appears to have been common to the Tamil and the ancient Malayalam coun- tries, and is the character in which the ancient sdsanas or docu- mentary tablets in the possession of the Jews at Cochin and of the Syrian Christians in Travancore are written. This character is still used with some variations by the Muhammedan colonists in North Malayalam. It presents some points of resemblance to the modern Telugu-Canarese character, and also to the character in which some undeciphered inscriptions in Ceylon and the Eastern Islands are written.* The language of all the more ancient of these inscriptions is Tamil, and the style in which they are written is that of the classical dialect, without any of those double plurals {e.g., ningal, yous, instead of nir, you), and other unauthorised novelties by which modern Tamil is disfigured ; but it is free also from the affected brevity and involutions of the poetical style. As no inscription of any antiquity in Telingana or Karnataka is found to be written in the Canarese or the Telugu language, whatever be the character employed, the priority of Tamil literary culture, as well as its national independence to a considerable extent, may fairly be concluded. I may here remark that the Cochin and Travancore sdsanas or tablets which are referred to above, and which have been translated by Dr Gundert, prove amongst other things the substantial identity of ancient Malayalam with ancient Tamil. The date of these docu- ments is probably not later than the ninth century a.d., nor earlier than the seventh ; for the technical terms of solar-siderial chron- * Journal of the Madras Literary Society, vol. xiii. ANTIQUITY OF THE TAMIL Sy ology (derived from the Surya-Siddhanta of Arya-bhatta) which are employed in these inscriptions were not introduced till the seventh century. The sdsanas were written at a time when the Kerala dynasty was still predominant on the Malabar coast ; but though words and forms which are peculiar to Malayalam may be detected in them, the general style of the language in which they are written is Tamil, the inflexions of the nouns and verbs are Tamil, and the idiom is mostly Tamil ; and we are therefore led to infer that at that period Tamil was the language at least of the court and of the edu- cated classes in the Malayalam country, and that what is now called Malayalam, if it then existed at all, was probably nothing more than a patois current among the inhabitants of the hills and jungles in the interior. The fact that the sdsanas which were given by the ancient Malayalam kings to the Jews and Syrian Christians are in the Tamil language, instead of what is now called Malayalam, cannot be accounted for by the supposition of the subjection at that time of any part of the Malayalam country to the ancient kings of Madura ; for the kings in question were Kerala, not Pandya kings, with Kerala names, titles, and insignia ; and it is evident from the Greek geographers themselves, from whom alone we know anything of an ancient Pandya conquest, that it was only a few isolated places, on or near the Malabar coast, that w^ere really under the rule of the Pandyas. The only part of the Malayalam country which at that period could have belonged bond fide to the Pandyas, was the southern part of the country of the Aii or Paralia, i.e., South Travancore, a district which has always been inhabited chiefly by Pandis, and where to the present day the language of the entire people is Tamil, not Malayalam. From the various particulars mentioned above, it appears clear that the Tamil language was of all the Dra vidian idioms the earliest cultivated ; it also appears highly probable that in the endeavour to ascertain the characteristics of the primitive Dra vidian speech, from which the various existing dialects have divaricated, most assistance will be furnished by Tamil. The amount and value of this assistance will appear in almost every portion of the grammatical comparison on which we are about to enter. It must, however, be borne in mind, as has already been intimated, that neither Tamil nor any other single diale(?^, ancient or modern, can be implicitly adopted as a faithful representative of the primitive Dravidian tongue. A careful comparison of the peculiarities of all the dialects will carry us up still further, probably up to the period of their 88 INTRODUCTION mutual divergence, a period long anterior to that of grammars and vocabularies ; and it is upon the result of such a comparison that most dependence is to be placed. Earliest extant Written Relics of the Dra vidian Languages. The Dravidian words which are contained in the Ramayana, the Maha-bharata, and other Sanskrit poems of undoubted antiquity, are so few that they throw no light whatever upon the ancient condition of the Dravidian languages prior to the ninth or tenth centuries a.d., the earliest date to which any extant Tamil compositions can safely be attributed.* The oldest Dravidian word found in any written record in the world appears to be the word for ' peacock ' in the Hebrew text of the Books of Kings and Chronicles, in the list of the articles of merchandise brought from Tarshish or Ophir in Solomon's ships, about 1000 B.C. This word is tuki in Kings, tuki in Chronicles. The ordinary name at present for the peacock on the Malabar coast and in Tamil is mayil (Sans, mayura) ; it is also sometimes called sihi (Sans. HhJii), a name given to it on account of its crest ; but the ancient, poetical, purely Tamil-Malayalam name of the peacock is tokei, the bird with the (splendid) tail. §ikhi=avis cristata ; tokei = avis caudata. The verbal root of the word tokei is to¥ or t6k\ tuk or tuk\ to hang ; hence ' a scarf,' ' a skirt border,' is called tohkei. The vowel of the root librates between u and o : half the derivatives have the one vowel, half the other. Hence there is no reason to suppose the Phoenicians in error when they represented tuk as the radical part of the word. That the vowel is short in Kings and long in Chronicles is also quite in accordance with the fact that in Tamil- Malayalam the vowel is sometimes short, sometimes long. Though tokei, as a verbal noun, is a derivative from toM or tuk\ yet the ultimate root appears to have been to or tu. Judging from analogy the final k^ or ku must have been a formative. A primary root with this addition becomes a verbal noun, and in the next stage of the language this verbal noun becomes in its turn a new, secondary verbal root. It is interesting to be able to trace the use of this Tamil-Malayalam formative ¥ or ku so early as the time of the Phoenician trade with India. Max Mliller, speaking of this etymol- ogy (Lect. p. 209), remarks : " If this etymology be right, it would * This statement must be modified. Later researches point to their having been in existence at a much earlier period. — Editors. EARLIEST TRACES OF THE DRAVIDIAN LANGUAGES 89 be an important confirmation of the antiquity of the Tamulic languages spoken in India before the advent of the Aryan tribes." I have no doubt that this etymology is right, and that the inference deduced from it is well founded. It may here be added that from the Dravidian tokei, pronounced togei, would naturally be derived the Arabic tawas, the Greek raws {taos), and ultimately the Latin pavo and our own pea-iowl. Minayeff has discovered in the Buddhistical writings a reference to voyages made by ancient Indian merchants to Babylon (called 'Baveru' =01d Cuneiform Persian ' Babiru '), in the second of which voyages they took thither the first peacock for sale. (See paper by Professor Weber in the Indian Antiquary for May 1873.) Of the names of the other articles of merchandise mentioned in Kings and Chronicles, kof, an ape, has generally been identified with the Sanskrit kapi ; and the Greek ktjttos (kepos), and even the English ape, have been supposed to have the same origin. It seems more probable, however, that the word has been derived from the old Egyptian kdf, an ape, a word which Mr Le Page Kenouf informs me is in very common use in Egyptian inscriptions, and which he says is to all appearance as ancient as the language itself. The origin of the word used for ' ivory ' {shen lidbhim, the tooth of the hdbh) still seems to me somewhat doubtful. On the whole, the most probable derivation seems to be from the old Egyptian ah, ivory. Algum may perhaps be the Sanskrit valguka, sandal wood, another meaning of which is ' beautiful,' a word which seems to be identical with, or derived from, the Tamil-Malayalam aragu or alagu, beauty. If so, algum will be more correct than almug. The fragrant wood called ' aloes ' in Proverbs vii. 17, &c., was the Aquilaria Agallocha, the Hebrew word for which, alialiyn or ahaloth, is evidently derived rather from the Tamil-Malayalam form of the word, aghil, than from the Sanskrit agaru, though both are ultimately identical. The Greek word opv^a {oruza), rice, must be one of considerable antiquity. It dates from the period, whenever that was, when rice was first introduced from India into Europe ; and it cannot be doubted that we have here the Tamil word arisi, rice deprived of the husk, this being the condition in which rice was then, as now, brought up in India for exportation to Europe. The distinctly Malayalam form of the wofd, ari, seems a corruption. The earliest Dravidian word in Greek of which we know the date is KapTTLov (karpion), Ctesias's name for cinnamon. Herodotus describes cinnamon " as the Kdp(f)€a [karphea), (dry sticks), which we, 90 INTRODUCTION after the Phoenicians, call Kti/va/xw/xov {kinnamomon)." Liddell and Scott say, in loc. Kapt^o^ [karpJios), plural Kapcji^a {karphea), " this word bears a curious resemblance to the Arabic words kerf at, kirfahy This resemblance, however, must, I think, be accidental, seeing that Herodotus considered ' cinnamon ' alone as a foreign word, and that Kapcfios is naturally derived from KapcfiU) (karphS), to wither. The word mentioned by Ctesias seems, however, to have a real resemblance to the Arabic word, and also to a Dra vidian one. Ctesias, the author of the earliest Greek treatise on India, describes an odorous oil produced from an Indian tree having flowers like the laurel, which the Greeks called pivpopoSa (muroroda), but which in India was called Kap-n-Lov (karpion). From Ctesias's description (making allowance for its exaggerations) it is evident that cinnamon oil was meant, and in this opinion Wahl agrees. Uranius, a writer quoted by Stephen of Byzantium, mentions Kkpiradov (kerpatlion) as one of the productions of the Abaseni, the Arabian Abyssinians, by which we are doubtless to understand not so much the products of their country as the articles in which they traded. From the connection in which it is found, KcpTraOov (kerpathon) would appear to be cinnamon, and we can scarcely err in identifying with it kerfat, or, more properly, kirfah, one of the names which cinnamon has received in Arabic. Some Arabic scholars derive kirfah from karafa, ' decortavit ' ; but Mr Hassoun does not admit this derivation, and considers kirfah a foreign word. We are thus brought back to Ctesias's Kapiriov {karpion), or the Indian word which Kapiriov (karpidn) represented. As this is a word of which we know the antiquity, the supposition that the Greeks or Indians borrowed it from the Arabs is quite inadmissible. What then is the Indian word Ctesias referred to ? Not, as has been supposed, kurundhu, the Sin- ghalese name for cinnamon, derived from the Sanskrit karunta ; but the Tamil-Malayalam word karuppu or kdrppu — e.g., karappa-{t)tai' lam, Mai. oil of cinnamon. Other forms of this word are karappu, karuva, and karuvd, the last of which is the most common form in modern Tamil. Rheede refers to this form of the word when he says that " in his time in Malabar oils in high medical estimation were made from both the root and the leaves of the karua or wild cinnamon of that coast." There arc two meanings of karu in Tamil-Malayalam, ' black ' and ' pungent,' and the latter doubtless supplies us with the ex- planation of karuppu, ' cinnamon.' A word with a related meaning to this is karukku, ' a medicinal preparation.' This name may have EARLIEST TRACES OF THE DRAVIDIAN LANGUAGES 9 1 been given to cinnamon from what has been described as ' the sweet burning taste ' of the bark, and especially of the oil. Wild cinnamon grows freely in Malabar, in the very region in which Ctesias's name for it, and the name adopted by the Arabians, is still in use. The cinnamon now grown in Ceylon is, it is true, of a much finer quality, but it is doubtful whether the cultivation of it had been introduced into Ceylon at that early period, and even if it had, it should be remembered that Ctesias, who derived all his information about India from Persian and Babylonian merchants, seems to have known nothing of Ceylon. I have little doubt that the Sanskrit karpura, ' camphor,' is substantially the same as the Tamil-Malayalam karuppu and Ctesias's KapTnov (karpion), seeing that it does not seem to have any root in Sanskrit, and that camphor and cinnamon are nearly related. The camphor of commerce is from a cinnamon tree, the camphora officinarum. If the identity of Ctesias's word with the Tamil-Malayalam karwppa be admitted, it follows that we have here the earliest Dravidian word quoted by the Greeks, and that at that early time Tamil roots were sometimes converted into verbal nouns by the addition of the formative pu, as they are at present, just as we have seen in the Hebrew tuki, the alternative formative ku or kei, used, as at present, for the same purpose. It is a remarkable circumstance that the largest stock of primitive Dravidian words contained in any written documents of ancient time — the earliest authentic extant traces of the existence of the Dravidian languages, as distinguished from Sanskrit — are those which are found in the notices of the Greek geographers Ptolemy and the author of the " Periplus Maris Erythrsei " ; including also the " Natural History " of Pliny. Many of the names of places and tribes recorded by those geographers, not long after the commence- ment of the Christian era, are identical, letter for letter, with the names now in use. Several of those names have become obsolete, or cannot now be identified ; but the signification of the compound words of which they consist is generally apparent, and in several of them we can detect the operation of some interesting dialectic peculiarity or euphonic rule which is still characteristic of these languages. I subjoin a few examples of Dravidian words of this class recorded by the Greeks, beginning with the names of Dravidian peoples and princes. * (1.) 6 HavSiijov (ho Pandion) — ^ X^^P^ HavSiovajv (he chora Pan- dionon), KavSioviov (Kandionon) is evidently an error — the Pandya king and people. This name is, as we have seen, of Sanskrit origin, 92 INTKODUCTION and Pandae, the form which Pliny, after Megasthenes, gives in his list of Indian nations, comes very near the Sanskrit. The more recent local information of Pliny himself, as well as the notices of Ptolemy and the Periplus, supply us with the Dravidian form of the word. The Tamil sign of the masculine singular is an, and Tamil inserts * euphonically after nd, consequently IlavStwv {Pandion), and still better, the plural form of the word HavSioves (Pandiones) faith- fully represents the Tamil masculine singular Pandiyan. Ptolemy is quite correct in giving the same name to the people and their prince. The people were Pandyas, the prince the Pandya, or the Pandya-deva. The form of the masculine singular in ancient Canarese, corresponding to the Tamil an, is am ; in Telugu it is udu, so that Pandiyudu in Telugu answers to Pandiyan in Tamil. Consequently we learn, that as early as the Christian era, Tamil differed dialectically from the other Dravidian idioms, and in particular that its mode of forming the masculine singular was then the same as it is now. We also learn from the expression MoSov/oa /Saa-tXeiov IlavSiovis (Modoura Basileion Pandionis) that the Pandyas had transferred their capital from Kolkei on the Tamra- parni to Madura on the Veigei before the Christian era. Mo5ov/oa itself (in PHny Modura) is the Sanskrit Mathura, pronounced in the Tamil manner. The corresponding city in Northern India, Muttra, is written by the Greeks IslkSopa (Methora). (2.) 6 Kr]po/366pos{hoKerohothros). The prince called by this name by Ptolemy is called 6 Kv^Tryoo/^orpos (ho Keprohotros) by the author of the Periplus. The insertion of tt is clearly an error, but more likely to be an error of a copyist than that of the author, who himself had visited the territories of the prince in question. He is called Cselobothras in Pliny's text, but one of the MSS. gives it more cor- rectly as Celobotras. The name in Sanskrit, and in full, is Kerala- futra, but both Kera and Kela are Dravidian abbreviations of Kerala. They are Malayalam, however, not Tamil abbreviations ; and the district over which Keralaputra ruled is that in which the Malayalam language is now spoken. (3.) 2w/3at vo/xaSes (Sorai nomades) — 'ApKaroD /Saa-tXeLOV ^wpai {ArJcatou Basileion Sorai) — "OpOovpa jSaaiX^iov 2w/oi/ayo§ (Orthoura Basileion Sornagos) — Ila/jaAia 2w/)'>^twi/ {Paralia Soreton) [or 2w/)tywv (Sorigon)] ; also IlapaAia Twptyywv (Paralia Toringon) which should evidently have been Sw/aiyywt' (Soringon), seeing that it included the mouth of the river Xa/3r)pos(Chaheros) ]. Without enter- ing here on any minute topographical discussions with regard to details, EARLIEST TEACES OF THE DRA VIDIAN LANGUAGES 93 it seems evident to me that the word Swpa (Sora), which we meet alone and in various combinations in these notices, represents the name of the northern portion of the Tamilian nation. This name is Chola in Sanskrit, Chola in Telugu ; but in Tamil Sora or Chora. Ptolemy's accuracy, or rather perhaps that of his informants, with regard to the name of this people is remarkable ; for in Tamil they appear not only as Soras, but also as Soragas and Soriyas, and even as Soringas ; their country also is called Soragam. The r of the Tamil word Sora is a peculiar sound, not contained in Telugu, in which it is generally represented by d, nor in Sanskrit and Pali, in which it is represented by d or I. The transliteration of this letter by the Greeks as p seems to show that then, as now, the use of this peculiar r was a dialectic peculiarity of Tamil. The Indian equivalent of the name of the king Sornax has not survived — as those of 6 IlavStwv {ho Pandion) and 6 Yir]pofi69pos{hoKer6bothros) have — and it is fruitless to guess what it may have been ; but as we know from native poems that the name of the ancient capital of the Soras was Ureiyur (pro- nounced Oreiyur), we may safely identify this name with Ptolemy's "OpOovpa (Orthoura), the capital of the UapaXU Sw/o^^twv {Paralia Soreton). (4.) 'ApKarov /Sao-iXeiov 2w/)a (Arhatou BasiUion Sora). "ApKaro^ (Arkatos) is here represented, not as a country, people, or city, but as the name of a prince. As General Cunningham has pointed out, liopa (Sora) is represented as the name of a city, where a king called "ApKaro? {Arkatos) reigned. Though this was evidently Ptolemy's meaning, yet one is strongly tempted to suppose that here the names given by the natives of the country to his informants had got transposed. The name 2wpa {Sora) is identical with that of the people of the district, whom Ptolemy himself calls Sw/oat vo/xa8es {Sorai noniades), and "ApKaros {Arkatos) answers exceed- ingly well, in situation as well as in sound, to Arcot, the capital of the Carnatic in Muhammedan times. There is a distinct tradition that the inhabitants of that part of the Chola or Sora country which lies between Madras and the Ghauts, including Arcot as its centre, were Kurumbars or wandering shepherds — nomads — for several centuries after the Christian era. General Cunningham objects to this identi- fication that Arcot is quite a modern name ; but it must, as Colonel Yule has pointed out, be at l^ast as old as 1340 a.d., for it is men- tioned by Ibn Batuta. The name is properly dt'-kad, Tam. the six forests, and the Hindus of the place regard it as an ancient city, though not mentioned by name in the Puranas, and point out the 94 INTRODUCTION ' six forests ' in which six of the rishis of the ancient period had their hermitages. If this identification be admitted, we have here another instance of the antiquity of the dialectic peculiarities of Tamil, for the oblique form of the word McU is hdtt\ and the word ordinarily used in Telugu for forest is not Md\ but adavi or atavi. (5.) Kdpovpa j3aa-iA€Lov Ki-]pop6Bpov (Karoura Basileion Kero- hothrou). Karur is mentioned in Tamil traditions as the ancient capital of the Chera, Kera, or Kerala kings, and is generally identified with Karur, an important town in the Coimbatore district, originally included in the Chera kingdom. Karur means' the black town, and I consider it identical with Kdragam and Kaddram, names of places which I have frequently found in inscriptions in the Tamil country, and which are evidently the poetical equivalents of Karur. The meaning of each of the names is the same. Ptolemy's word Kdpovpa {Karoura) represents the Tamil name of the place with perfect accuracy ; har means black, and ur (sometimes pronounced wr-w), a town. Neither of these words seems to have altered in the least in sound or signification for 1800 years. (6.) Modogalingam nomine, Pliny. I have already, in p. 28, dis- cussed the meaning of this name. I add here that if modo be regarded as a Telugu word, meaning three, we have here an interesting illus- tration of the antiquity of Dravidian dialectic pecuHarities ; for three is in Telugu mudu, in Tamil mundru, in Canarese muru, in Tulu muji. (7.) Damirice, and also Scytia Dymirice, Peutinger Tables ; Dimi- rica, in the Ravenna Cosmography, see p. 10. The Dymir of Dymi- rice was supposed by Dr Burnell to represent the word Tamir, and if so, the Damir of Damirice will come still nearer thereto. The portion of the Malabar coast immediately to the north of Dymirice is called, by Ptolemy and the author of the " Periplus," 'kpiaKy] (AriaJce), and it seems probable that this was the district to which the name of Aryaka was given by Varaha-mihira several centuries afterwards {Journal of the Royal Asiatic Society, vol. v.). It appears probable, therefore, that the difference between the Aryans and the Dra vidians can be traced in the names given by the Greeks to those por- tions of the Malabar coast which we know from other sources of information have always been inhabited by Aryans and Dravidians respectively. (8.) I content myself with simply noting the following names of places on the Malabar coast. Mov^ipt'i {Mouziris) appears to be the Muyiri of Muyiri-cotta ; TiVSis {Tyndis) is Tundi ; and the EARLIEST TRACES OF THE DBA VIDIAN LANGUAGES 95 Kynda of Nelkynda [or as Ptolemy has it MeA-Kvi/Sa (Mel-Kynda), i.e., probably Western Kynda] seems to be Kannettri, the southern boundary of Kerala proper. One MS. of Pliny writes the second part of this word not cyndon, but canidon. The first of these places was identified by Dr Gundert ; for the remaining two we are in- debted to Dr Burnell. (9.) Cottonara, Pliny ; KoTTovapcKyj {Kottonarike), Perip. ; the district where the best pepper was produced. It is singular that this district was not mentioned by Ptolemy. Cottonara was evidently the name of the district ; KorrovapiKov (kottonarikon), the name of the pepper for which the district was famous. Dr Buchanan identifies Cottonara with Kadatta-nadu, the name of a district in the Calicut country celebrated for its pepper. Dr Burnell identifies it with Kolatta-nadu, the district about Tellicherry, which he says is the pepper district, kadatta, in Malayalam, means transport, conveyance ; nddu, Tam.-MaL, means a district. (10.) Idyyapa (Sangara). The author of the " Periplus " calls by this name the canoes formed out of single trees, in which pepper was brought from Cottonara to Barace. The Malayalam name of these boats is changddam, Tulu jangdla. Compare Sanskrit sam.- ghddam, a raft. I have never been able to explain KoXavhiof^iavra {kolandio'phonta), the name of the large vessels that sailed from the western coast to Ceylon and the Ganges. (11.) KoTTictpa {Kottiara). This is the name of a place in the country of the "A lot (Aioi) of Ptolemy, in the liapaXia (Par alia) of the author of the " Periplus," identical in part with South Travan- core. Apparently it is the Cottora of Pliny, and I have no doubt that it is the Cottara of the Peutinger Tables. It is not to be con- founded with Cottonara, the place mentioned above. It is called by Ptolemy Koma/oa Mr/r/aoTroAis (Kottiara Metrofolis), and must have been a place of considerable importance. The town referred to is probably KottaT-u, or as it is ordinarily written by Europeans, Kotaur, the principal town in South Travancore, and now, as in the time of the Greeks, distinguished for its commerce. The name of the place is derived from kod-u, Tam.-Mal. a fort, and dr-u, a river. It is a rule both in Tamil and in Malayalam that when a word like kod' is the first member of a compound, the final d must be doubled for the purpose of giving th^ word the force of an adjective : it is another rule that sonants when doubled become surds. Consequently the compound kod-u— dr-u becomes by rule Kott-dr-u. If the identi- fication of the place be correct, as it appears to me to be, we find here 96 INTRODUCTION an interesting proof that in the time of the Greeks the same phonetic rules were in operation as now. (12.) Ko/xa/)ia aKpov {Kotnaria akron), Ptol. ; Ko/xdp {Komar), Kojxapd {Komarei), Perip. Cape Comorin* has derived its name from the Sanskrit kumdn, a virgin, one of the names of the goddess Durga, the presiding divinity of the place ; but the shape this word has taken, especially in Ko[idp (Komar), is distinctively Tamilian. In ordinary Tamil kumdri becomes kumdri ; and in the vulgar dialect of the people residing in the neighbourhood of the Cape, a virgin is neither kumdri nor kumdri, but kmndr, pronounced komdr. It is remarkable that this vulgar corruption of the Sanskrit is identical with the name given to the place by the author of the " Periplus." He says, " After this there is another place called Kojidp {Komar), where there is a fipidpiov (briarion) — probably ^povpiov {Phrourion), a fort ; tepov {liieron) is less likely — and a harbour, where also people come to bathe and purify themselves, . . . for it is related that a goddess was once accustomed to bathe there monthly." This monthly bathing in honour of the goddess Durga is still continued at Cape Comorin, but is not practised to the same extent as in ancient times. Kumari formerly ranked as one of the five renowned sacred bathing places, a representation which accords with the statement of the author of the " Periplus." Through the continued encroachments of the sea, the harbour the Greek mariners found at Cape Comorin, and the fort (if that were meant), have com- pletely disappeared ; but a fresh-water well remains in the centre of a rock a little way out at sea. It is singular that Cape Comorin does not appear in any shape in the Peutinger Tables. (13.) liapaXia {Paralia). There are three Paralias mentioned by the Greeks, two by Ptolemy (the Paralia of the Soreti, and the Paralia properly so called, that of the Toringi), one by the author of the " Periplus." The Paralia mentioned by the latter corresponded to Ptolemy's country of the"A6oi {Aioi) and that of the Ka/O€ot (Kareoi), that is, to South Travancore and South Tinnevelly. It commenced at the Red Cliffs, south of Quilon, and included not only Cape Comorin, but also KoAxoi, (Kolchoi), where the pearl-fishing was carried on, and which belonged to King Pandion. Dr Burnell identifies TLapaXla (Paralia) with Purali, which he states is an old * Compare Cymri (Wales). "It is stated that the original home of the Cwmry, Cumri, or Cymry, was in Southern Hindustan, the southern ex- tremity of which, Cape Comorin, takes its name from the same root." — From a Historical Souvenir issued on the occasion of the meeting of the British Medical Association at Swansea, 1903. — Editors. EARLIEST TRACES OF THE DRAVIDIAN LANGUAGES 9/ name for Travancore, but I am not quite able to adopt this view. It is true that, if the Greeks found any part of the Travancore coast called Purali, they would naturally proceed to convert that name into a word of their own, bearing an intelligible and appropriate meaning ; but, on the other hand, it is not clear that any part of the coast was ever called by that name. Purali is stated by Dr Gundert (" Malayalam Dictionary " in loc.) to be the name of a fort belonging to the old kings of KoUayagam in tlie interior. Hence PuralUan, lord of Purali, was one of the titles of those kings. This title is now poetically applied to the kings of Travancore ; but it seems probable that it was adopted by them at a comparatively late period, on their gaining possession of the territory to which the title belonged, in the same manner as they adopted the title of Vanji- hhupati, lord of Vanji, a name of Karur, the ancient Chera or Kerala capital. It is also to be remembered that the Paralia of the " Peri- plus " included not only the coast of South Travancore, but also the coast of Tinnevelly as far as Kolkei. It appears to me, therefore, that Hapakta {Paralia) is to be taken as a Greek word, though possi- bly it may have corresponded in meaning, if not in sound, to some native word meaning coast. This will appear probable from the next item. (14.) ol KapeoL (hoi Kareoi). The Carei of Ptolemy inhabited the southern portion of Tinnevelly, between Cape Comorin and Kolkei ; consequently their country constituted the eastern portion of the Paralia of the " Periplus." Karei is the Tamil word for coast or shore from the verbal theme karei, to be melted down, to be washed away, and is obviously identical in meaning with the Greek UapaXla (Paralia). Up to the present time several portions of the Tinnevelly coast (including that part where I have myself lived and laboured for more than thirty years) are called Karei, the coast, or Karei- (ch)chuttru, the coast circuit, and a caste of fishermen further north are called Kareiydr, coast-people. There cannot be any doubt that the last portion of two names of places mentioned by Ptolemy represents the Tamil karei, coast, viz., KaXatKapias (Kalaikarias) and UepL-yKapei (Perinkarei) . If the latter word had been written HepvyKapet it would have been perfectly accurate Tamil, letter for letter. The meaning is great shore ; and 'perum, great, becomes perung before k by rule, ^erum itself, instead of peru, is a distinc- tively classical form. (15.) y] 1o)X.yv (he Solen). The Tamraparni, the chief river in Tinne- velly, must be the river intended to be denoted by Ptolemy by this 98 INTRODUCTION name, for it is the only river mentioned by him between Cape Comorin and the Kaveri, and it entered the sea south of KoAxot, (KolcJioi), the emporium of the pearl trade, which was certainly at the mouth of the Tamraparni. It is difficult, however, to explain how it came to be called SwA?^ (Solen). This word means in Greek a shell-fish, a mussel ; and it seems uncertain whether the Greeks called the river by this name, because the native name of it some- what resembled this, or because of the fishing for chanks, as well as pearls, then as now, carried on at its mouth. The name by which the river seems always to have been called in India is Tamraparni, a name which bears no resemblance whatever to Solen. In Tamil poetry it is often called the Porunei, which is merely a Tamilisation of the second portion of its Sanskrit name. Tdmrapamt, Sans., would naturally mean the tree with red or copper leaves ; applied to a river, it would seem to mean the river which resembles a red leaf. It is called by this name in the Maha-bharata, though whether the passage in which it is mentioned is older than Ptolemy may be regarded as uncertain. The name Tamra-parni being identical with the oldest name of Ceylon — Tambapanni in Pali, TaTrpofiavt] (Tapro- hane) in Greek — it might have been supposed, if the river had been called by this name in the time of the Greeks, that they would have called it the Taprobane, the name by which they called Ceylon. Solen cannot have any connection with Sylaur, erroneously represented in Lassen as the name of the principal tributary of the Tamraparni. This tributary is called the Chitra- nadi, commonly the Chittar, which means in Tamil the small river, and it is physically impossible that it ever can have been, as Lassen conjectured, the principal stream, the mountain district it drains being very much smaller than that which the Tamraparni drains. (16.) BryTTtyw {Bettigo). This, according to Ptolemy, was the name of the mountain range in which the SwAryi/ (Solen) — the Tamraparni — took its rise, in addition to two rivers on the western coast, the Bdpis {Baris) and ^€v^6crroiio (ne-6), the Latin no, nato, and also to nau. Sans, a boat, of which Sanskrit does not appear to contain the root. Derivative nouns formed from verbs which have formative suffixes always prefer as their formative the transitive suffix, or that which doubles and hardens the initial consonant. Thus from tiru- ndu, Tam. to become correct, is formed tiru-ttam, correction ; and from tu-ngu, to sleep, tu-kkwm, sleep (comp. tuyil, sleep). In some instances the crude root of a verb is used as the intransitive, whilst the transitive is formed by the addition of ttu to the root. E.g., padu, Tam. to lie down, padu-ttu, to lay ; tar, to be low, tdr-ttu, to lower ; nil (Tel. niki), to stand, niru-ttu (for nilu-ttu), to establish. In such cases Canarese uses du instead of the Tamil ttu — e.g., tdl-du, to lower, instead of tdr-ttu. This transitive formative is sometimes represented as a causal ; bu^ it will be shown in the section on " The Verb " that i is the only real causal in the Dra vidian languages. In all the cases now mentioned, where ttu is used as the formative of the transitive by Tamil, Telugu uses chu or pu. 202 ROOTS I class under the head of this formative all those nouns in which the cerebral consonants d, nd, and it are used in the same manner and for the same purpose as the dentals d, nd, and tt — e.g., kuru-du, blindness, adjectival form of the same, huru-ttu, blind ; ira-ndu, two, adjectival form, ira-Uu, double. Telugu hardens, but does not double, the final d of such nouns — e.g., 6d-u, a leak, oti, leaky. In some instances in Tamil the hard, rough r, when used as a final, seems to be equivalent to du, or du, and is doubled and pronounced with a t — e.g., kina-vu, a well, kifia-nu (pronounced kinattru), of a well. (4.) hu or mhu, with its transitive ffu. — In Canarese, hu, the original form of this intransitive suffix, has been softened into vu, and in Tamil, hu has universally been euphonised into mhu. This Tamilian formative mhu is in some instances softened in Telugu nouns into mu. The hu or mhu of Tamil verbs is superseded by vu or gu in Telugu ; and the forms answering to the Tamil transitive 'pfu are fu and mfu, rarely p'pu. Example of the use of this formative by a verb : niramhu, Tam. to be full, nira-ppu, to fill ; of which the crude base nir reappears in the related verbs nir-a, nir-avu, nir-ei, and nir-ei, to be full, to be level, &c. Telugu has nindu instead of niramhu ; but the transitive nimpu answers very nearly to the Tamil nirafpu. Example of a noun in 7nbu and 'ppu : iru-mhu, Tam. iron, adjectival form, iru-fpu, of iron — e.g., iruppu-{k)k6l, an iron rod. In Telugu irumhu is softened into inumu, adjectival form inwpa. Canarese still adheres to the original form of this suffix, generally softening h into v, but leaving it always unnasalised — e.g., Canarese hdvu, a snake, properly fdvu : Tamil pdmhu, nasalised from pdhu ; adjectival form fdppu — e.g., fdffu-{k)kodi, the serpent banner : Telugu, still further altered, fdynu. This example clearly illustrates the progressive euphonisation of the formative in question. It has been mentioned that Telugu uses fu or m,fu as a formative of transitive verbs where Tamil uses ffu. It should be added that even in those cases where Tamil uses the other formatives previously noticed, viz., kku and Itu, Telugu often prefers fu. Compare the following infinitives in Tamil and in Telugu — e.g., meykka, Tam. to feed cattle, mefa, Tel. ; nirutta, Tam. to establish, nilwpa, Tel. Where kku in Tamil, and fu in Telugu, are preceded by i, this formative becomes in Telugu either mfu or nchu — e.g., compare Offuvi'kka, Tamil, to deliver over, with the corresponding Telugu infinitive, opfagi-nifa, or oppagi-ncha. It appears from the various particulars now mentioned, that tran- FORMATIVE ADDITIONS TO ROOTS 2O3 sitive verbs and nouns used adjectivally must have been regarded by the primitive Tamilians as possessing some quality in common. The common feature possessed by each is doubtless the quality of transi- tion ; for it is evident that when nouns are used adjectivally there is a transition of the quality or act denoted by the adjectival noun to the noun substantive to which it is prefixed, which corresponds to the transition of the action denoted by the transitive verb to the accusative which it governs. It is manifest that the various particles which are used as forma- tives do not essentially differ from one another either in signification, in the purpose for which they are used, in the manner in which they are affixed, or in the manner in which they are doubled and hardened. It seems to have been euphony only that determined which of the sonants g, s, d, d, or b should be suffixed as a formative to any par- ticular verb or noun. The only particular in which a grammatical principle appears to exist, is the doubling of the initial consonant of the formative, to denote or correspond with the putting forth of energy, which is inherent in the idea of active or transitive verbs, as distinguished from intransitives. From the statements and examples given above, it may be con- cluded that wherever Dra vidian verbs or nouns are found to ter- minate in any of the syllables referred to, there is reason to suspect that the first part of the word alone constitutes or contains the root. The final syllables gu, ngu, kku ; su, chu ; du, ndu, ttu ; du, ndu, ttu ; hu, mhu, mfu, fu, fpu ; mu, vu, may as a general rule be rejected as formative additions. This rule will be found on examination to throw unexpected light on the derivation and relationship of many nouns which are commonly supposed to be primitive and independ- ent, but which, when the syllables referred to above are rejected, are found to be derived from or allied to verbal roots which are still in use. I adduce, as examples, the following Tamil words : — kombu, a branch, a twig ; vembu, the margosa-tree ; vambu, abuse ; pdmbu, a snake. As soon as the formative final, mhu, is rejected, the verbs from which these nouns are derived are brought to light. Thus, ko- mbu, a twig, is plainly derived from ko-y, to pluck off, to cut ; ve-mhu, the margosa-tree, is from ve-y, to screen or shade (the shade of this tree being peculiarly prized) ; va-mbu, abuse, is from vei, properly va-y (corresponding to the Canarese bayyu), to revile ; 'pd-mbu, a snake, is from pd-y, to spring. In these instances, the verbal base which is now in use ends in y, a merely euphonic addition, which does not belong to the root, and which disappears in the 204 ROOTS derivatives before the consonants which are added as formatives. The same principle applied to nouns ending in the other formative syllables will be found to yield similar results — e.g., marundu, a medical drug, from 7na,ru, to be fragrant ; and kirangu, a root, from kir, to be beneath, the ^ of which, though long in the Tamil Mr, is short in the Telugu kinda, below. . Reduplication of the Final Consonant of the Root. — The principle of employing reduplication as a means of producing gram- matical expression is recognised by the Dra vidian languages as well as by those of the Indo-European family, though the mode in which the reduplication is effected and the objects in view are different. It is in Tamil that this reduplication is most distinctly apparent, and it should here be borne in mind, that when a Tamil consonant is doubled it is changed from a sonant into a surd. The final consonant of a Tamil root is doubled — (1.) for the purpose of changing a noun into an adjective, showing that it qualifies another noun, or of putting it in the genitive case — e.g., from mddu, an ox, is formed mdtt-u{t)t6l, ox-hide ; (2.) for the purpose of converting an intransitive or neuter verb into a transitive — e.g., from 6d-u, to run, is formed ottu, to drive ; (3.) for the purpose of forming the preterite — e.g., tag-u, to be fit, takk-a, that was fit ; and (4.) for the purpose of forming derivative nouns from verbal themes — e.g., from erud-u, to write, is formed erutt-u, a letter. (See this subject further elucidated in the sections on " The Noun " and " The Verb.") It is remarkable that whilst the Indo-European tongues often mark the past tense by the reduplication of the first syllable, it is by the reduplication of the last letter that the Dravidian languages effect this purpose ; and also, that whilst the Tibetan converts a noun into a verb by doubling the last consonant, this should be a Dravidian method of converting a verb into a noun. The rationale of the Dravidian reduplication seems to be, that it was felt to be a natural way to express the idea of transition both in the act and in the result. In Hebrew also the doubling of a consonant is intensitive or causative. Up to this point it has been found that all Dravidian polysyllabic roots are traceable to a monosyllabic base, lengthened either by euphonic additions, or by the addition of formative particles. An important class of dissyllabic bases remains, of which the second syllable, whatever may have been its origin, is an inseparable particle of specialisation, into the nature and use of which we shall now inquire. PARTICLES OF SPECIALISATION 205 Particles of Specialisation. — The verbs and nouns belonging to the class of bases which are now under consideration, consist of a monosyllabic root or stem, containing the generic signification, and a second syllable, originally perhaps a formative addition, or perhaps the fragment of a lost root or lost postposition, by which the generic meaning of the stem is in some manner modified. The second syllable appears sometimes to expand and sometimes to restrict the signification, but in some instances, through the absence of syno- nyms, its force cannot now be ascertained. As this syllable is in- tended in some manner to specialise the meaning of the root, I call it ^' the particle of specialisation." It is certain in some cases, prob- able in many, that these particles of specialisation were originally formatives of verbal nouns. This will appear from a comparison of the verbs and nouns contained in the list of final particles which will be found near the end of this section. The principle involved in the use of these particles of specialisation, and the manner in which it is carried into effect, correspond in a certain degree to a characteristic feature of the Semitic languages, which it appears to be desirable to notice here. As far back as the separate existence of the Semitic family of languages can be traced, every root is found to consist of two syllables, comprising generally three consonants. When Semitic biliteral roots are compared with their synonyms, or corresponding roots, in the Indo-European languages, and especially with those which are found in Sanskrit, a simpler and more primitive root-system has been brought to light. It has been ascertained in a considerable number of instances that whilst the first syllable of the Hebrew root corresponds with Sanskrit, the second syllable does not in any manner correspond to any Indo- European synonym. It is found also that the second syllable has not any essential connection with the first, and that a considerable number of families of roots exist in which the first syllable is the same in each case, whilst the second continually varies. It is there- fore inferred that in such cases the first syllable alone (comprising two consonants, the initial and the final, together with the vowel used for enunciation) contains the radical base and generic significa- tion, and that the second syllable, perhaps the fragment of an obso- lete auxiliary verb, has been appended to the first and afterwards compounded with it, for tlie purpose of giving the generic significa- tion a specific and definite direction. According to this view, which appears to be in the main correct, Hebrew roots are to be regarded, not singly and separately, as independent monads, but as arranged 206 ROOTS generically in clusters or groups, exhibiting general resemblances and special differences. The family likeness resides in the first syllable, the radical base ; the individuality, or special peculiarity, in the second, the particle of specialisation. It is true that in some instances the second syllable of Semitic roots meets with its counterpart in the Indo-European languages, as well as the first, or even instead of the first ; but the peculiar rule or law now referred to is found to pervade so large a portion of the Hebrew roots, that it justly claims to be considered as a character- istic of the language. Thus, there is a family of Hebrew roots signifying generally to divide, to cleave, to separate, &c. The mem- bers of this family are fdldh, pdlah, fdlag, fold, pdlal ; and also (through the dialectic interchange of / with r) 'pdrash, pdras, Chaldee 'peras. It cannot be doubted that in all these instances the first syllable 'pdl or jictr, or rather p-r, f-l (for the vowel belongs not to the root, but to the grammatical relation), expresses merely the general idea of division ; whilst the second syllable (which is in some in- stances a reduplication of the final consonant of the biliteral) expresses, or is supposed to express, the particular mode in which the division or partition is effected. The first syllable, which is the same in all the members of this group of roots, is that which is to be compared with synonyms in other languages, whilst the second syllable is merely modal. In this instance we not only observe a distinct analogy between the Hebrew roots p-r, p-l, and the Greek TTop-o) {por-o), the Latin pars, par-tis, and the Sanskrit phal, to divide, but we also discover the existence of an analogy with the Dra vidian languages. Compare with the Hebrew p-r, p-l, the Tamil piri, to divide, and pdl, a part ; pila and por, to cleave ; as also pagir and pagu, to portion out, to divide. See also the " Glossarial Affinities." On turning our attention to the root-system of the Dra vidian lan- guages, we are struck with the resemblance which it bears to the Semitic root-system referred to above. We find in these languages groups of related roots, the first syllables of which are nearly or wholly identical, whilst their second syllables are different in each instance, and in consequence of this difference produce the required degree of diversity in the signification of each member of the group. We also find in these languages, as in Hebrew, that the generic par- ticle or common base, and the added particle of specialisation, are so conjoined as to become one indivisible etymon. The specialising particle, which was probably a separable suffix, formative, or post- position at first, has become by degrees a component part of the PARTICLES OF SPECIALISATION 207 word ; and this word, so compounded, constitutes the base to which all formatives, properly so called, and all inflexional particles are appended. This root-system exists in all the languages of the Dravidian family, but its nature and peculiarities are especially apparent in Tamil. Out of many such groups of related Tamil roots, I select as illustrations two groups which commence with the first letter of the alphabet. 1. Roots which radiate from the base syllable ad : — adu to come near ; also to cook, to kill, to unite, to belong to. 7 7 7 > to be contained, to enclose. adakku ) adi to drive in, commonly to beat, adi, as a noun, the basis of anything, a footstep, a sole. adei to attain, to get in, to roost ; transitive, to enclose. adeisu to stuff in. adar to be close together, to be crowded, to join battle. aduhhu to place one thing upon another, to pile up. This verb and adakku are properly aduk and adak, but final k in Tamil is always vocalised by the help of u, and often doubled, as in this instance, before receiving the u and a of the root. andu (Tel. antu), to approach. This verb seems to be identical with adu, the first in the list, and euphonised from it by the insertion of the nasal. Compare also the related verb an. It is obvious that all these roots are pervaded by a family resem- blance. All contain the generic notion of nearness, expressed by the first or base syllable ad ; whilst each, by means of the second syllable, or particle of specialisation, denotes some particular species of nearness. 2. Roots which radiate from the base syllable an : — anu, anugu to approach, to touch. ani to put on, to wear. anei to connect, to embrace ; as a noun, a weir, a dam. anavu to cleave to. annu to resort to, to lean upon. (From this verb is derived annal or ani^n, an elder brother, one to lean upon, a derivation which has at least the merit of being poetical.) The corresponding Telugu verb is dnuta. anmu to be near. 208 KOOTS The generic idea signified by the base syllable an is evidently that of contact ; and this group differs from the previous one as actual contact differs from contiguity or nearness. Probably am, a nail, a fastening, is derived from the same verb, and it appears probable also that this is the origin of the Sanskrit ani or dni, the pin of an axle. The illustrations given above prove that the second syllables of the various verbs now adduced have not been added merely for pur- poses of euphony, but have been appended in order to expand, to restrict, or in some manner to modify and specialise the signification. It was shown in a previous part of this section, that the vowels a, i, u, e, and ei are sometimes added euphonical ly to monosyllabic roots. It is obvious, however, that this is not the only purpose for which those vowel additions are used ; and it is of importance to know that when they are merely euphonic they are found to be interchangeable with other vowels, whereas when they are used as particles of specialisation they retain their individual character more firmly. Probably they had all a specialising signification at first, which they retain in some instances, but have lost in others. The examples already given may suffice to illustrate the use of appended vowels as specialising particles. Syllables ending in con- sonants, especially in I and r, are also used very frequently for this purpose ; and it seems desirable here to adduce examples of the use of particles of this class. As has already been observed in connection with " Formative Additions to Roots," all these syllables seem to have been originally formatives of verbal nouns, probably each of them with a specialising signification. Many of the verbal nouns so formed have then become secondary verbal themes. The following examples are mostly from Tamil, in which I and r may stand as finals. The other dialects add u to the final consonant of each of these particles. Tamil requires this euphonic addition of u only when a word ends in the hard, rough r, or in any consonant besides the nasals and semi- vowels. Each word being considered either as a verb or as a noun according to circumstances, I give examples of nouns as well as of verbs. Some of the following words, though used as verbs, are more commonly used as nouns, and some, though used as nouns, are more commonly used as verbs. Some of the examples, again, are used either as nouns only or as verbs only : — PARTICLES OF SPECIALISATION 209 Final Particles. Verbs. Nouns. ar valar, to grow. sudar, lustre. ir tulir, to sprout. ugir, a finger nail. ur nudur-u, Tel. the forehead. ar pugar, to praise. idar, a flower petal. ir magir, to rejoice. avir, a grain of rice. av-u idat-u, to trip. kinaT-u, a well. ii-u ndyir-u, the sun. al sural, to whirl. iral, the liver. il kuyil, to utter a sound. veyil, sunshine. ul fagul-u, Tel. to break. al tuval, to bend. tingal, the moon. il madil, a fort wall. ul urul, to roll. irul, darkness. Of all the thirteen specialising particles ending in consonants of which examples have now been adduced, only one appears occasion- ally to be used as an equivalent for a vowel addition : ar alternates with ei — e.g., amar, Tam. to rest, and amei, are apparently equiva- lent. The verb to grow, also, is in Tamil valar, and in Canarese bale, which in Tamil would be valei. The original meaning of most of the particles used as formative suffixes or particles of specialisation, is now unknown, but there are two of which the meaning appears nearly certain ; these are il, which survives as a substantive, meaning here or a house, the particle used as the most common case sign of the locative in Tamil-Malayalam, and ul, which is still used both as a noun and as a verb ; as a noun meaning within, and as a verb, to be. The force of these particles and their retention of the locative signification will appear in such instances as vdyil, a doorway, literally the mouth house (from vdy, mouth) ; veyil, the heat of the sun, literally, that in which heat resides (from vey, to be hot). Dr Gundert suggests also porul, wealth, which may come from poru, to unite ; arul, grace, from aru, to be scarce, precious ; and irul, darkness, from ir, to be dark, the root of ird, night. I here subjoin an example of another peculiar and interesting set of groups of roots found in the^Dravidian languages, which are formed upon a plan differing considerably from that which has now been explained. The roots referred to are dissyllabic, but they contain only one consonant, which is preceded and followed by a vowel. 2IO ROOTS This consonant appears to represent the ultimate or radical base, whilst the initial and final vowels alter in accordance with the particular shade of signification which it is desired to convey. When we compare idu, Tam. to press or crush, odu, to squeeze, to bring into a smaller compass, and idi, to bruise, to beat down, as also adi, to drive in, or odi, to break in two, and udei (pronounced odei), to break open, we cannot avoid the conclusion that the first four roots are closely related members of the same family or group ; that the last two are in like manner mutually related ; and that possibly the whole of them have an ulterior relationship, in virtue of their possessing in common the same nucleus or radical base, the central consonant d, and the same generic signification. The existence of clusters of roots, like these mentioned above, is not a peculiarity of the Dravidian languages alone. Max Miiller (Lectures, ii. 313) observes, " We find in Sanskrit and in all the Aryan languages clusters of roots, expressive of one common idea, and differing from each other merely by one or two additional letters, either at the end or at the beginning." In illustration of this he says, "To go would be expressed by sar, to creep by sarp ; to shout by nad, to rejoice by nand ; to join by yu or yuj, to glue together by yaut.'' In another place (i. 274) he says, " In the secondary roots we can generally observe that one of the consonants, in the Aryan languages generally the final, is liable to modification. The root retains its general meaning, which is slightly modified and deter- mined by the changes of the final consonants." " These secondary roots," he says, " stand to the primaries in about the same relation as the triliteral Semitic roots to the more primitive biliteral." In the Dravidian languages the change under consideration is as often in the vowel of the root as in the consonant, and it is hard to say whether the initial vowel is not even more subject to modification than the final vowel. Changes in Eoot Vowels. — As a general rule the vowels of Dra- vidian roots belong as essentially to the radical base as the con- sonants. They very rarely pertain, as in the Semitic languages, to the system of means by which grammatical relations are expressed, and they are still more rarely modified, as in the Indo-European languages, by the addition of inflexional forms, or in composition. In the Semitic languages the radical base is destitute of vowels, and by itself unpronounceable. The insertion of vowels not only vocalises the consonants of the root, but constitutes it a grammatically in- CHANGES IN ROOT- VOWELS 211 fleeted verb or noun, the signification of which varies with tiie varia- tion of the interior vowels. In the Indo-European languages grammatical modifications are generally produced by additions to the root ; and though in the earliest period of the history of those languages, the root, generally monosyllabic, is supposed to have remained unaltered by additions and combinations, yet the existence of that rigidity is scarcely capable of direct proof ; for on examining the Sanskrit, Greek, Latin, and German, the most faithful repre- sentatives of the early condition of those languages, we find that the root-vowels of a large proportion of the words have been modified by the addition of the suffixes of case and tense ; and in particular, that the reduplication of the root, by which the past tense appears usually to have been formed, is often found either to alter the quantity of the root-vowel, to change one vowel into another, or entirely to expunge it. In the Scythian family of tongues, not only does the vowel belong essentially to the roo', but in general it remains unalterable. It very rarely happens that the root-vowel sustains any change or modification on the addition to the root of the signs of gender, number, and case, or of person, tense, and mood ; which, as a rule, are successively agglutinated to the root,, not welded into combina tion with it. This rigidity or persistency is almost equally char- acteristic of the root- vowels of the Dra vidian languages. In general, whatever be the length or weight of the additions made to a Dra- vidian root, and whether it stands alone or is combined with other words in a construct state, it is represented as fully and faithfully in the oblique cases as in the nominative, in the preterite and future as in the present tense or in the imperative. I proceed to point out some noticeable exceptions to this rule. Exceptions. — Internal Changes in Roots. 1. One class of changes is purely euphonic. It has no relation to grammatical expression ; but it seems desirable to mention it here in order to give a complete view of the subject. It is connected with one of the minor dialectic peculiarities referred to in the chapter on sounds, and consists in the occasional softening or rejection of the medial consonant of a dissyllabic root or verbal noun, together with the coalescence of the vowels that preceded and followed it. It has been shown that g has a tendency to be softened into v and then to disappear, and that s sometimes changes in the same manner into y, when it sometimes becomes absorbed. When either of these con- sonants is a medial, it is apt to be thus softened down and rejected. 212 ROOTS Thus dogal-u, Can. skin, becomes in Tamil tol ; 'peJar, Can. a name, becomes in Tamil first 'peyar, and then per. So in Tamil, togup-pu, a collection, is softened into top-pu, which has the restricted meaning of a collection of trees, a tope. In like manner the medial v of the Tamil avan, he, disappears in the personal terminations of verbs, and the preceding and following vowels coalesce, when avan becomes an or on. So also the length of the demonstrative roots, a remote, and i proximate, varies in different dialects, and even in different connections in the same dialect, through considerations of euphony. 2. The exceptions that follow in this and the following paragraphs are not euphonic merely, but real. They pertain to grammatical expression. In most of the Dra vidian languages the quantity of the root-vowels of the pronouns of the first and second persons, both singular and plural, is short in the oblique cases. The nominatives of those pronouns are long — e.g., nan, Tamil, I, nam, we ; ni, thou, nir, you. But in Tamil, Canarese, Malayalam, and Tulu, in all the oblique cases the vowels are shortened before receiving the suffixed inflexional particles. Thus, in Canarese, to me is not ndn-a-ge, but ndn-a-ge ; to thee is not nin-a-ge, but nin-a-ge. Telugu, Gond, and Ku generally retain the quantity of the vowel of the nominative unaltered — e.g., in Telugu we find ni-ku, to thee, as well as m, thou ; but in the accusative, nin-u or ninn-u, thee, the quantity is altered. It is open to us to regard the shorter form of the pronouns as the original, and the longer as the form that has been altered ; and it will be seen, when the pronouns are under discussion, that this is the view I prefer. Singularly enough, this exception from the general rigidity of the root-vowels is a Scythian exception, as well as a Dra vidian one. In the Scythian version of the Behistun tablets, whilst the nominative of the pronoun of the second person is nt, thou, as in the Dravidian languages, the possessive case is ni, thy, and the accusative nin, thee, corresponding in quantity to the Dravidian oblique cases — e.g., Telugu nin-u, thee ; Tulu nin-a, thy, nin-an\ thee ; High Tamil nin, thy, and ninnei, thee. 3. Another class of exceptions consists of instances in which the quantity of a vowel is lengthened when a verbal root is formed, directly and without any extraneous addition, into a noun. The alteration which the root-vowel sustains is prior to any inflexional additions being made. If any formative particle is added to a verbal root to convert it into a noun, the quantity of the root- vowel remains unchanged. The lengthening of the root- vowel to which I refer takes place only in (some of) those cases in which the CHANGES IN ROOT- VOWELS 21^ verbal base itself is used as a noun. Thus, tlie verb ked-u, to destroy or to become destroyed, may become a verbal noun by the addition of the formative di — e.g., kedudi, destruction, in which event the root- vowel remains unaltered ; but the verbal base may also be used without addition as a verbal noun, in which case ked-u is lengthened into ked-u. The following Tamil examples of the lengthening of each of the five primary vowels will suffice to illustrate this usage : — From 'pad-u, to suffer, is formed fdd-u, a suffering ; from min, to shine, min, a star ; from sud-u, to burn, sud-u, heat ; from pev-u, to obtain, pei-u, a benefit obtained ; and from kol, to receive, kol, reception. I am not aware of the existence of a similar rule in any of the Scythian languages, but it is well known in Sanskrit {e.g., compare vach, to speak, with vdch, a word ; niar {mri), to die, with mdra, death). Nevertheless, I can scarcely think it likely that it is from Sanskrit that the Dra vidian languages have derived a usage which prevails among them to so great an extent, and which has every appearance of being an original feature of their own. If it is not to be regarded as an independently developed peculiarity, arising out of the same mental and lingual habitudes as those out of which the cor- responding Sanskrit usage was developed, it is probably to be re- garded as a relic of those pre-Sanskrit influences of which many traces seem to be discoverable in these languages. In one particular the Dra vidian rule differs from the Sanskrit. In Sanskrit the root- vowel is often not only lengthened, but changed, according to certain rules, into another — e.g., from vid, to know, comes veda, knowledge, the Veda ; whereas in the Dravidian languages the rule is that the root-vowel is simply lengthened — e.g., from vid-u, Tam. to set free, comes vH-u, emancipation, a house (meaning probably a tax-free tenement). Dr Gundert derives ver, Tam. a root, from vir, the radical part of viri, to expand (compare viral, a finger). If this derivation be accepted as correct, as I think it may, it will furnish an instance of the operation of the Sanskrit law in question. Another derivation which I regard as still more probable is that of ner, Tam. straight, from nira, to be level. These very rare exceptions, however, do not nulHfy the rule. • I must here notice a class of verbal nouns formed after this manner which are much used adjectivally. All Dravidian adjectives, gramma- tically considered, are nouns, but some of them are used indiscrimi- 214 ROOTS nately either as nouns or as adjectives ; some exclusively as ad- jectives, some exclusively as nouns. The three adjectives fer, large, kdr, black, and dr, precious, furnish good illustrations of the class of verbal nouns to which I refer, fer and dr are used exclusively as adjectives, hdr both as an adjective and as a noun. As an adjective it means black, as a noun, blackness, a cloud, the rainy season, &c. The radical forms of these words are also in use. These are fer-u, to be large, Jcar-u, to be black, and ar-u, to be precious. The final u is, as usual, merely enunciative ; the roots are per, kar, and ar. When we find a Dravidian root in two shapes, one with a longer, the other with a shorter vowel, it may generally be assumed, and can often be proved, that the shorter form is the radical one. Where both forms are in use, as in the case of these three words, the longer form is considered more elegant, and is much used in combinations, especially before words beginning with a vowel. It is to the shorter and probably more ancient form that mei, the formative of abstract nouns, like our English nouns ending in ness, is suffixed — e.g., aru- mei, preciousness. The same change in the internal vowel of the root is apparent in some of the numerals. The radical forms of the Tamil numerals one and two seem to be or and ir, and these are often lengthened, when the numeral is used not as a substantive but as an adjective, into or and ir. There are also two forms of the numerals three, six, and seven (mu and mu, avu and diu, eru and eru), but in these instances it is the shorter forms that are used adjectivally. These shorter forms cannot stand alone, they can be used only as adjectives, whereas the longer ones are used as numeral substantives. The formation of verbal nouns by means of the lengthening of the root-vowel throws as much light on the original meaning of some adjectives, or nouns of quality, as we have seen that it does (in the previous part of this section) in the case of certain nouns exclusively used as substantives. For instance, far (Tam.), desolate, is evidently a verbal noun from par-u, to grow old. To grow mature or ripe is a secondary meaning, from which we have param, a ripe fruit. Another form used adjectivally is para, old. A verb of the secondary forma- tion is paragu, to become used to anything. When the final consonant of the crude root belongs to this class of hard letters, it cannot be enunciated by Dravidian organs, whether the preceding vowel be long or short, without the aid of a final euphonic u. Thus paSu, Tam. to be green, when lengthened becomes, not pas (as per, Mr, &c.), but pdsu, green. A change sometimes takes place in the internal vowel of this word which has been supposed to CHANGES IN ROOT- VOWELS 21 5 accord with the Sanskrit change of a short vowel into a longer one of a different order, and of a naturally long vowel into a diphthong, on the change of a noun or verbal-root into an adjective, fasum, green (another form of pasu), is changed in certain conjunctions into peim — e.g., feim-fon (Tarn.), excellent, literally green, gold. This change, however, is merely euphonic. It has already been shown that s, when medial, has a tendency to soften into y, and then to disappear, and when this takes place the preceding and following vowels coalesce. In consequence of this tendency, fasum naturally becomes fayum, and this again, by a change which is almost im- perceptible in pronunciation, feim. We have a parallel instance of this in the noun kasuffu (Tam.), bitterness, which may optionally be written and pronounced keippu ; kasuppu changing first into kayuffu and then into keippu. It should also be observed that peim has not in the least superseded pasum. The one may be optionally used instead of the other, and this proves that both forms are grammatically equivalent. I should be prepared to admit that in these and similar instances y may possibly be older than s. The process, on this supposition, would have to be reversed ; fei, properly payu, would become pasu, but the result would be the same. The change in the internal vowel would still be owing merely to the euphonic substitution of one consonant for another. I may here remark that forms like pasum, green, do not appear to me to be derived, as Beschi, following native grammarians, supposed, from pasumei, greenness, by the omission of the final ei ; for mei, not ei, is the particle by which abstract nouns of quality are formed, and the initial m is the most essential portion of that particle. Pasum is evidently derived from pas, the crude verbal root, with the addition of um, the sign of the aoristic future, by means of which it becomes an aoristic relative participle, a class of participles which the Dra vidian tongues delight to use as adjectives. 4. Another class of internal changes appears in those instances in which Tamil shortens the quantity of the root-vowel in the pre- terite tense of verbs. This shortening is observed in Canarese also, but the following illustrations are furnished by Tamil — e.g., ve, to burn, has for its preterite participle, not vendu, but vendu ; no, to be in pain, has for its preterite, not nondu, but nondu ; kdn, to see, becomes, not kdndu, but Jmndu. Another instance is sd, to die, which takes not sdttu, but settu. The Malayalam and Canarese form of this participle, sattu or chattu, represents the root-vowel more accurately than the Tamil. In some instances Tamil retains in the 2l6 ROOTS preterite the long vowel of the root, whilst Canarese shortens it — e.g., i, to give, has for its preterite in Tamil indu, in Canarese ittu. There are two verbs in Tamil, vd, to come, and td, to give, which involve peculiarities of which it is difficult to give a satisfactory explanation. Each of them is regularly conjugated, except in the preterite and imperative, as if from roots in var and tar {e.g., varu- given, I come, tarugiven, I give) ; each takes the root with the long vowel without r for its imperative singular, and inserts r between this form of the root and the personal termination in the imperative plural {e.g., vd, come, td, give ; vdrum, come ye, tdrum, give ye) ; and each forms its preterite by shortening the vowel without inserting r, as if from roots in vd and td, after the manner described in the previous paragraph {e.g., vanden, I came, tanden, I gave, like nonden, I felt pain, from the root no). Dr Pope, in his " Tamil Handbook," p. 52, considers the r of these verbs euphonically inserted to prevent hiatus and the whole of the tenses built upon the roots in vd and td. I should have no objection to this view if the r made its appearance in the plural imperative only, as in kdrum, protect ye, from M, to protect, the only other instance I know of r being used for this purpose in Tamil, and one which I have already mentioned in the chapter on " Prevention of Hiatus." On the other hand, the appearance of the roots in var and tar, in every part of the verb, except the preterite and the singu- lar imperative alone, and in all the verbal nouns without exception {e.g., varal, varattu, varuttu, varudal, varavu, varugei, each of them meaning a coming), leads to the conclusion that var and tar (whatever be the origin of their difference from vd and td) are treated in Tamil as verbal themes. If r were not a portion of the root, we should expect to find the present, future, infinitive, negative voice, verbal nouns, &c., formed from vd and td, with the addition of ^ or i; as a formative suffix, as w^e find to be the case with the parallel verbs no, &c. Compare noga, infinitive ; novu, nodal, &c., verbal nouns ; nogd, negative. The Canarese roots are bar and tar. In Telugu the im- perative singular is vd, the plural mmmti, and this seems to me to confirm the supposition that r is an essential part of the root. If the Telugu r represented only the supposed euphonic r of the Tamil, the root-consonant would be left without any representative at all. It appears to me improbable, moreover, because unsupported by usage, that the Tamil v has been changed into r in Telugu. It seems more in accordance with usage to recognise here a change similar to that which has converted the Tamil iladu, there is not, into ledu in Telugu, and ird, night, into re. See the chapter on " Euphonic CHANGES IN ROOT- VOWELS 21/ Displacement of Vowels." Notwithstanding this, I am not disposed to regard the forms in vd and td as having found their way into the conjugation of the verbs by mistake. It is evident that vd and td, not var and tar, are the themes from which the preterites vanden and tanden have been formed, and which we find pure in the im- peratives. We seem therefore driven to adopt Dr Gundert's sug- gestion that vd and var, and td and tar, are alternative roots — perhaps it would be preferable to say, different forms of the same root. This supposition need not be reHnquished in consequence of its being regarded as probable that td is identical with the Indo-European root dd, to give. The Dravidian tar may have sprung from a related form of the same root, of which possibly a trace may survive in the Greek SQpov [doron) and the Hebrew tan. I may add that though the change in the length of the vowel in the preterite has a grammatical significance, its change of length in the imperative, from rd, Tel. singular, to rammu, honorific singular (plural), and from vd, Tam. singular, to High Tam. vammm, plural, appears to be purely euphonic. The changes in the internal vowels of Dravidian roots exhibited in the last three classes of instances mentioned in this section as exceptions to the ordinary stability of the Dravidian root-vowels, evidently accord, as far as they go, with usages prevalent in the Indo- European languages, inasmuch as one of the classes referred to furnishes us with instances of the lengthening of the root- vowel, when the verb is converted into a noun, whilst the other classes furnish us with examples of the shortening of the interior vowels of the root on receiving the addition of inflexional particles, to compensate for the additional weight thus imposed on the root-vowel, or for the purpose of distinguishing one tense from another. In regard, however, to changes in root-vowels, it would be erroneous to suppose the rule of the Scythian languages essentially and universally dis- similar to the Indo-European. In the Scythian languages, as in the Dravidian, stability in the root-vowels is the rule, change the exception. But exceptions exist {e.g., compare olen, Finnish, I am, from the root ol, to be, with lienen, if I be ; compare also Hungarian leven, from the same root, being, with volt, having been, and lenni, to be). In consequence of the existence of such exceptions as these, it is impossible to erect the difference between the two families of language, in this particular, into a hard and fast law of distinction. It would also be unsafe, on this ground alone^ to disconnect the Dravidian languages from the languages of the Scythian group and to connect them with the Indo-European. PART III, THE NOUN. In this section it will be my endeavour to investigate the nature and inflexions of the Dravidian noun, with the view of ascertaining its method of expressing the relations of gender and number, and the principles on which that method proceeds, together with the charac- teristics and origin of its case system, or system of means for ex- pressing the relationship of nouns with other parts of speech. It will be shown at the close of the section on " The Verb," how derivative nouns are formed from verbal roots ; and the various classes of participial nouns will then also be investigated. SECTION I.— GENDER AND NUMBER. 1. Gender. When the Indo-European laws of gender are compared with those of the Scythian group of tongues, it will appear that in this point, as in many others, the Dravidian languages accord more closely with the Scythian than with the Indo-European family. In all the more primitive Indo-European languages, not only are words that denote rational beings and living creatures regarded as masculine or feminine, according to the sex of the objects referred to, but also inanimate objects and even abstract ideas have similar sexual distinctions attributed to them ; so that many nouns which denote objects naturally destitute of gender, and which ought therefore to be regarded as neuters, are treated by the grammars of those languages as if the objects they denote were males and females, and are fitted not with neuter, but with masculine or feminine case terminations, and with pronouns of corresponding genders. This peculiar system is a proof of the highly imaginative and poetical character of the Indo-European mind, by which principles of resemblance were discerned in the midst of the greatest differences, and all things that exist were not only animated, but personified. lb 218 GENDER 219 is from this personification that most of the ancient mythologies are supposed to have arisen. A similar remark applies to the Semitic languages also, in which the same or a similar usage respecting gender prevailed. In the progress of the corruption of the primitive Indo- European languages, a less imaginative but more natural usage gained ground. Nevertheless, in a majority of the modern colloquial dialects of this family, both in Europe and in India, the gender of nouns is still an important and difficult section of the grammar, and a standing impediment in the way of the idiomatic use of those languages by foreigners. On the other hand, in the Manchu, Mongolian, Turkish, and Finnish families of tongues — the principal families of the Scythian group — a law or usage respecting the gender of nouns universally prevails, which is generically different from that of the Indo-Euro- pean and the Semitic idioms. In those families, not only are all things which are destitute of reason and life denoted by neuter nouns, but no nouns whatever — not even nouns which denote human beings — are regarded as in themselves masculine or feminine. All nouns, as such, are neuter, or rather are destitute of gender. In those languages there is no mark of gender inherent in, or in- separably annexed to, the nominative of any noun (the crude root being generally the nominative) ; and in none of the oblique cases, or postpositions used as case terminations, is the idea of gender at all involved. The unimaginative Scythians reduced all things, whether rational or irrational, animate or inanimate, to the same dead level, and regarded them all as impersonal. They prefixed to common nouns, wherever they found it necessary, some word denot- ing sex, equivalent to male or female, he or she ; but they invariably regarded such nouns as in themselves neuters, and generally they supplied them with neuter pronouns. The only exceptions to this rule in the Scythian languages consist in a few words, such as God, man, woman, husband, wife, which are so highly instinct with personality that of themselves, and without the addition of any word denoting sex, they necessarily convey the signification of masculine or feminine. When our attention is turned to the Dravidian languages we find that, whilst their rules respecting gender differ widely from those of the Indo-European group, ^hey are not quite identical with those of the Scythian. It seems probable, however, that the particulars in which the Dravidian rules respecting gender differ from those of the Scythian languages, and evince a tendency in the Indo-European 220 THE NOUN direction, are not the result of direct Sanskritic influences, of which no trace is perceptible in this department of Dravidian grammar, but have arisen either from the progressive mental cultivation of the Dravidians themselves, or from an inheritance of prse-Sanskritic elements. Dravidian nouns are divided into two classes, which Tamil gram- marians denote by the technical terms of high-caste and casteless nouns, but which are called by Telugu grammarians mahdt, majors, and a-mahdt, minors. High-caste nouns, or majors, are those which denote " the celestial and infernal deities and human beings," or, briefly, all things endowed with reason ; and in all the Dravidian dialects (with a peculiar exception which is found only in Telugu and Gond) nouns of this class are treated in the singular as masculines or feminines respectively, and in the plural as epicenes, that is, without distinguishing between masculines and feminines, but distinguishing both from the neuter. The other class of nouns, called casteless, or minors, includes everything which is destitute of reason, whether animate or inanimate. This classification of nouns, though not so imaginative as that of the Indo-European and Semitic tongues, is decidedly more philosophical ; for the difference between rational beings and beings or things which are destitute of reason is more momentous and essential than any difference that exists between the sexes. The new Persian, which uses one pluralising particle for nouns that denote animated beings, and another and different one for things that are destitute of life, is the only non-Dra vidian language in which nouns are classified in a manner which is in any degree similar to the Dravidian system. The peculiar Dravidian law of gender which has now been described would appear to be a result of progressive intellectual and grammatical cultivation ; for the masculine, feminine, and epicene suffixes which form the termina- tions of Dravidian high-caste nouns, are properly fragments of pro- nouns or demonstratives of the third person, as are also most of the neuter formatives. It may, indeed, be stated as a general rule that all primitive Dravidian nouns are destitute of gender, and that every noun or pronoun in which the idea of gender is formally expressed, being a compound word, is necessarily of later origin than the uncompounded primitives. The technical term by which such nouns are denoted by Tamil grammarians is pagu-padam, divisible words, i.e., compounds. Hence the poetical dialects, which retain many of the primitive landmarks, are fond of discarding the ordinary suffixes of gender or rationality, and treating all nouns as far as possible as GENDER 221 abstract neuters. Thus, in poetical Tamil Dev-u, God, a crude noun destitute of gender, is reckoned more classical than Dev-an, the corresponding masculine noun. This word is a Sanskrit derivative ; but the same tendency to fall back upon the old Scythian rule appears in the case of many other words which are primitive Dra vidian nouns — e.g., itei, a king, a word which is destitute of gender, is more classical than irei-{v)-an, the commoner form, which possesses the masculine singular termination. In the modern Tamil spoken by the educated classes, the words which denote sun and moon {suriy-an and sandir-an, derived from the Sanskrit surya and chandra) are of the masculine gender, in accordance with Sanskrit usage and with the principles of the Brah- manical mythology ; but in the old Tamil of the poets and the peasants, ndyitu, the sun, also forudu, and tingal, the moon, also nild, all pure Dra vidian words, are neuter. All true Dravidian names of towns, rivers, &;c., are in like manner destitute of every mark of personality or gender. In some few instances Malayalam and Canarese retain the primitive laws of gender more faithfully than Tamil. Thus, in the Tamil word peiyan, a boy, we find the masculine singular termination an ; whereas Malayalam (with which agrees Canarese) uses the older word peidal, a word (properly a verbal noun) which is destitute of gender, to which it prefixes in a thoroughly Scythian manner words that signify respectively male and female to form compounds signifying boy and girl — e.g., an peidal, a boy, fen feidal, a girl. The nature and origin of the terminations which are used to signify gender in the various Dravidian dialects will be inquired into under the head of " Number," with the consideration of which this subject is inseparably connected. Under this head I restrict myself to a statement of the general principles respecting gender which characterise the Dravidian languages. A peculiarity of Telugu, which appears also in Gond, should here be mentioned. Whilst those dialects agree with the other members of the Dravidian family in regarding masculines and feminines, and both combined, as constituting in the plural a common or epicene gender, they differ from the other dialects in this respect, that they are wholly or virtually destitute of a feminine singular, and instead of the feminine singular use the singular of the neuter. This rule includes in its operation psonouns and verbs as well as substantives, and applies to goddesses and queens, as well as to ordinary women. The Telugu possesses, it is true, a few forms which are appropriate to the feminine singular, but they are rarely used, and that only in 222 THE NOUN certain rare combinations and conjunctures. He and it are the only pronouns of the third person singular which are ordinarily made use of by more than twenty millions of the Telugu people ; and the colloquial dialect does not even possess any pronoun, equivalent to our pronoun she, which is capable of being applied to women of the lower as well as of the higher classes. Ordinarily every woman is spoken of in Telugu as a chattel or a thing, or as we are accustomed to speak of very young children {e.g., it did so and so), apparently on the supposition either that women are destitute of reason, or that their reason, like that of infants, lies dormant. Whilst each woman taken singly is treated by Telugu grammar as a chattel or as a child, women taken collectively are regarded with as much respect as by the other Dra vidian dialects. In the plural they are honoured with the same high-caste or rational suffixes and pronouns that are applied to men and gods. Canarese and Malayalam agree in this point with Tamil, and regard women, not in the plural only but also in the singular, as pertaining to the class of rationals : accordingly in those languages there is a feminine singular pronoun equivalent to she, which corre- sponds in the principle of its formation to the masculine he. With those languages agrees Ku, which, though the near neighbour of Telugu and Gond, pursues in this respect a politer course than either. In the idioms of the Tudas and Kotas, the rude aborigines of the Nilgherry hills, there is, properly speaking, only one pronoun of the third person, and that is without distinction of gender or number. atham,, remote, itham, proximate, mean indiscriminately he, she, it, they. The pronouns avan, aval, he, she, are also occasionally used, but Dr Pope thinks they have been recently introduced from the Tamil and Canarese. This usage reminds one of the employment in the old Hebrew of the same pronoun, hu, to signify both he and she, and still more of the use of the reflexive pronoun of the Latin se, for all genders and numbers. Compare ivuh, Hindustani, he, she. 2. Number. The Dra vidian languages recognise only two numbers, the singular and the plural. The dual, properly so called, is unknown, and there is no trace extant of its use at any previous period. Several of the languages of this family contain two plurals of the pronoun of the first person, one of which includes the party addressed as well as the party of the speaker, and which may therefore be considered as a NUMBER — THE MASCULINE SINGULAR 223 species of dual, whilst the other excludes the party addressed. As, however, this peculiarity is restricted to the personal pronouns, it will be examined in that connection. Under the head of " Number," we shall inquire into the Dravidian mode of forming the masculine, feminine, and neuter singular, and the epicene and neuter plural. (1.) Masculine Singular. — It has already been intimated that the formatives by which the gender of nouns is occasionally expressed are identical with the terminations of the demonstrative pronouns. From a very early period of the history of these languages, particles or formatives of gender were suffixed to the demonstrative bases, by the addition of which suffixes demonstrative pronouns were formed. Those formatives of gender were not originally appended to or combined with substantive nouns ; but their use was gradually extended as their utility was perceived, and nouns which included the idea of gender were made to express that idea by suffixing the gender terminations of the pronouns, whereby they became appella- tive nouns. The manner in which all these suffixes are added will be sufficiently illustrated by the instance of the masculine singular. The masculine singular suffix of the Tamil is an, an, or on. An, the shorter formative, is that which appears in the demonstrative pronoun avan {a-{v)-an), he ; and by suffixing any of these formatives to an abstract or neuter noun, the noun ceases to be abstract and becomes a concrete masculine-singular appellative. Thnsmupp-u, age, by the addition of an becomes muff -an, an elder, literally age-he, or age-man ; and from Tamir comes Tamir-an, a Tamilian, a Tamil-man. These and similar nouns are called generically " compound or divisible words " by Tamil grammarians. They are obviously com- pounded of a noun — generally a noun of quality or relation — and a suffix of gender, which appears also to have been a noun originally. In the instances which have been adduced, the suffix of gender is annexed to the nominative or casus rectus ; but in many cases it is annexed to the oblique case or inflexional base, viz., to that form of the noun to which the case signs are suffixed, and which, when used by itself, has the meaning of the genitive or locative. When the inflexion, or oblique case, is employed instead of the nominative in compounds of this nature, it generally conveys a possessive or locative signification — e.g., maleiyinan {malei-{y)-m-an), a moun- taineer, literally a man of oi^on the mountain ; pattinsittan {fattin'- aWdn), a citizen, literally a man of or in the city. Sometimes, how- ever, the inflexional " in " is merely added euphonically — e.g., there is no difference in meaning between villan (vill-an), a bowman, and 224 THE NOUN villinan {viW-m-am), which is considered a more elegant form. Words of this description are in some grammars called adjectives ; but they are never regarded as such by any native grammarians : they cannot be simply prefixed for the purpose of qualifying other words, and it is evident from their construction that they are merely appellative nouns. A subdivision of appellatives consists of words in which the suffixes of gender are annexed to adjectival forms — e.g., hodiya-n, a cruel man. I regard words of this class as participial nouns, and they will be investigated in the part on " The Verb," under the head of " Appellative Verbs " ; but whatever be the nature of kodiya (the first part of the compound), hodiya-n is certainly not an adjective, for before it can be used adjectivally we must append to it the relative participle ana, that is — e.g., hodiyan-dna, that is a cruel man ; and as the compound, cruel man, cannot be called an adjective in English, neither is hodiyan an adjective in Tamil : it is properly an appellative noun. It may be said that the neuter plural of this word, viz., kodiya, may be prefixed adjectivally to any substantive ; but kodiya, cruel things, the neuter plural of kodiyan, is not really identical with the adjective kodiya, cruel. It is totally distinct from it, though identical in appearance. The a of the former word is the neuter suffix of plurality ; whereas the a of the latter is that of the possessive case and of the relative participle, as will be shown at the close of this part (see " Adjectival Formatives ") and in the part on " Verbs." Another species of Tamil appellative nouns is said by Beschi to be formed by annexing suffixes of gender to verbal roots — e.g., oduvdn, a reader, from odu, to read ; but this, I believe, is an error. Those words are to be regarded as participial nouns, and oduvdn is literally he who will read, i.e., he who is accustomed to read. In the same manner, odinan is the participial noun of the preterite tense, and means he who read or is accustomed to read : odugindravan, the corresponding present participial noun, he who reads, belongs to the same class ; and these forms are not to be confounded with appella- tive nouns properly so called. On the other hand, such words as kdffan, a protector, are true appellatives ; but kdffan is not formed from the future tense of the verb (though kdppdn means he will protect), but from kdppu, protection, a derivative noun, of which the final and formative ppu is from the same origin as the corre- sponding final of muppu, old age. See the concluding section of the part on " The Verb." NUMBER — THE MASCULINE SINGULAR 2 25 The suffixes of gender which form the terminal portion of appella- tive nouns vary somewhat in form, but they are one and the same in origin, and their variations are merely euphonic. It is the vowel only that varies, never the consonant. When a neuter noun ends with a vowel which is essential to it, and is incapable of elision, and also when a noun happens to be a long monosyllable, an, or in poetry on, is more commonly suffixed than an. In some cases avan, he, the full demonstrative pronoun, is suffixed instead of its termination only, and this mode is thought peculiarly elegant. Thus, from vil or vill-u, a bow, we may form vill-an, vill-dn, and vill-6n, an archer, a bowman, and also vill-avan. Indeed, an and on have evidently been formed, not from an, but from a-\-v-^n, by the softening of the euphonic v, and the coalescence of the vowels. This corruption of avan into an appears systematically in the third person masculine singular of the colloquial Tamil verb — e.g., f6-(n)-dn (not 'p6-{ny avan), he went. The Canarese masculine singular suffix anu is identical with the Tamil an, the addition of u being merely a phonetic necessity of the modern dialect. In the older Canarese, the termination which was used was am, a particle which is to be regarded as the equivalent of an, n and m being interchangeable nasals. Malayalam is in this par- ticular perfectly identical with Tamil. The corresponding Telugu masculine singular formative is d-u, ud-u, or ad-u ; or rather nd-u, und-u, or andu, the obscure n being always pronounced, and being probably an essential part of the original form of the particle, and by suffixing the same formative to any substantive noun, it becomes a masculine singular — e.g., mag-andu, a husband, a word which seems to be identical in origin with the Tamil mag-an, a son (the primitive and proper meaning of each word being a male). The mascuHne singular suffix of Telugu often takes the shape of und-u, and in like manner the epicene plural suffix, which is in Tamil ar-u, is often ur-u in Telugu ; but in these instances a changes into u through attrac- tion. As Tamil forms masculine appellatives by suffixing the demon- strative pronoun avan, so does Telugu sometimes suffix its full demonstrative pronoun vdndu — e.g., chinna-vdndu, a boy (Tamil, Hnna-{v)-an), literally he who is little. It is probable that the Telugu masculine singular suffix was originally an or an-u, as in Tamil- Canarese. andu, und-u, or ndu, is found only in the nominative in correct Telugu, and it is replaced in all the oblique cases by ani or ni ; and that this ni is not merely an inflexional increment, but the repre- 226 THE NOUN sentative of an old masculine singular suffix, appears on comparing it with n, the corresponding oblique case suffix of the masculine-femi- nine plural, which is certainly formed from ar-u. When vdniki, to him, is compared with its plural vdriki, to them, it is evident that the former corresponds as closely to the Tamil avanukku as the latter to avarukku ; and consequently that the ni of vdniki must be significant of the masculine singular. Probably the same termination survives in the demonstrative, dyana, he, a form which is more rarely used than vdndu. The Telugu nd being thus found to be identical with the Tam., Can., Mai. n, and the old Can. m, the masculine suffixes an, am, and andu are also found to be identical. It is more difficult to determine the origin of this suffix an. an is sometimes used in Tam. and Mai. instead of am as a formative of neuter nouns, as will be shown here- after in the section on the Nominative — e.g., folan (Sans, phala), fruit, instead of palam ; but I cannot see how this can be identical in origin with the suffix an which denotes the masculine, the Dra vidian masculine being a distinctive one — that is, not merely a grammatical term, but a sign of sex. On looking around for an explanation of the origin of the masculine suffix, it appears to me that the Ku, though one of the most barbarous of the Dravidian dialects, throws more light than any other upon this point. It forms its demonstrative pronouns in a simple and truly primitive manner by prefixing d, the demonstrative base, to common nouns which signify man and woman. These nouns are dn-u, a man, and dl-u, a woman ; and ddn-u (compare Tam. a{v)an), literally that man, is used to signify he, and ddlu (compare Tam. a{v)al), that woman, to signify she. The Ku dn-u, a man, seems certainly identical with the Tam. noun dn, a male, and probably also with dl, a man, a person. In the use to which this primitive root is put in the Ku word d-dn-u, we may see, I think, the origin of an, the suffix of the masculine singular in most of the Dravidian dialects. The final u, of the Ku word dn-u, being merely euphonic, the root appears to be dn or dn ; and as n and n have been shown to be interchangeable, dn must be regarded as only another form of dn. n, again, is not only often euphonised by suffix- ing du [e.g., fen, Tam. a female, colloquially and poetically pend-u), but it is also sometimes directly changed into d, of which we have an instance in the classical Tamil ped-ei, a hen, a word which is derived by this process from, and is identical with pen, a female. Hence, the Telugu suffix and-u might naturally be derived from an older form in an, if it should appear that that form existed ; and that it did NUMBER — THE FEMININE SINGULAR 227 exist, appears from the vulgar use of the present day. of n instead of n in some of the oblique cases {e.q., vdnni, him, instead of vdni), and from the half anusvdra, or obscure nasal, which precedes du itself — e.g., vdndu, for vddu, he. A close connection appears thus to be established between the Tamil-Canarese an and the Telugu ad-u, through the middle point an. The only difficulty in the way of the perfect identification of the formative an with the Ku anu, a man, and with the Tamil an, a male, lies in the length of the vowel of the latter words. Here again Ku comes to our assistance ; for we find that the vowel was euphoni- cally shortened in some instances in the very dialect in which the origin of the word itself was discovered. In Ku the d of dn-u is long, both when it is used as an isolated word and in the demonstratives ddn-u, he, and ddl-u, she ; but when the demonstrative pronoun is appended to, and combined with, the relative participle of the verb, so as to form with it a participial noun, the d of dn-u is shortened into a, and in this shortened form the connection of the Ku formative with the Tamil-Canarese is seen to be complete. Compare the Ku participial noun gitdn-u, he who did, with the corresponding Canarese geyiddn-u ; gitdr-u, Ku, they who did, with geyidar-u, Can., and also gital-u, Ku, she who did, with geyidal-u, Can. (2.) Feminine Singular. — Though Telugu and Gond generally use the neuter singular to supply the place of the feminine singular, the other Dravidian dialects possess and constantly use a feminine singular formative which is quite distinct from that of the neuter. This formative is al in Tamil, Malayalam, and old Canarese, and by suffixing the sign of gender to the demonstrative base, the feminine singular demonstrative pronoun aval {a{v)al), she, is formed — a word which perfectly corresponds to avan (a{v)an), he. A numerous class of feminine singular appellative nouns is formed by suffixing the same particle to abstract or neuter nouns in their crude state — e.g., compare mag-al, Tam. a daughter, with mag-an, a son ; ill-dl, house- wife, a wife, and ill-dn, a husband, are formed from the addition of the pronouns aval and avan (euphonised into dl and dn) to il, a home. Telugu, in some connections, uses a feminine singular formative which appears to be identical with that of Tamil-Canarese. That formative is dl-u, which is used by Ku more largely than by Telugu ; and its identity with TamilCanarese al will be found to furnish us with a clue to the origin and literal meaning of the latter. As dn-u, in Ku, means a man, so dl-u means a woman : ddl-u, she, is literally that woman. The same word dl-u means a woman, a wife, in 228 THE NOUN poetical and vulgar Telugu also ; and in Gond there is a word which is apparently allied to it, dr, a woman. Even in Sanskrit we meet with dli, a woman's female friend. It is evident that dl-u would be shortened into al as easily as dn-u into an, and the constant occur- rence of a cerebral I in Tamil and Canarese, where Telugu has the medial /, fully accounts for the change of the one semi-vowel into the other. The unchanged form of this suffix appears in Telugu in such words as manama-{r)-dlu, a granddaughter, compared with manama- Tjdu, a grandson. The abbreviation of the vowel of the feminine suffix, which is characteristic of Tamil and Canarese, is exemplified in Telugu also, in the words maradal-u, a niece, and kodal-u, a daughter-in-law ; in which words the feminine suffix al-u is evi- dently identical both with Tamil-Canarese al or al-u, and also with dl-u, the older and more regular form of this suffix, which is capable of being used by itself as a noun. Probably the Telugu dd-u, adj. female, though now treated as a different word, is identical in origin with dl-u, through the very common interchange of d and I ; an illustration of which we have in kei-{y)-dlu, Tam. to use, which is converted in the colloquial dialect to kei-{y)-ddu. The feminine singular suffix al appears in Tamil and Canarese in the terminations of verbs as well as in those of pronouns. Telugu, on the other hand, which uses the neuter demonstrative instead of the feminine singular, uses the final fragment of the same demonstrative as the termination of the feminine singular of its verb. It may be remarked that in some of the Caucasian dialects, n and I are used as masculine and feminine terminals, exactly as in Tamil — e.g., in Avar, eynen is father, evel is mother. There is another mode of forming the feminine singular of appella- tive nouns, which is much used in all the Dravidian dialects, and which may be regarded as especially characteristic of Telugu. It consists in suffixing the Telugu neuter singular demonstrative, its termination, or a modification of it, to any abstract or neuter noun. The neuter singular demonstrative being used by Telugu instead of the feminine singular (it for she), this neuter suffix has naturally in Telugu supplied the place of a feminine suffix ; and though in the other dialects the feminine pronouns are formed by means of feminine suffixes, not by those of the neuter, yet the less respectful Telugu usage has crept into the department of their appellative nouns. In Tamil, this neuter-feminine suffix is atti or tti. This will appear on comparing velldl-atti, a woman of the cultivator caste, with velldl-an, a man of the same caste ; oru-tti, one woman, una, with oru-{v)-an. NUMBER — THE NEUTER SINGULAR 229 one man, unus ; and vannd-tti, a washerwoman, with vannd-n, a washerman, tt, a portion of this suffix, is sometimes erroneously used in vulgar Tamil as a component element in the masculine appellative noun oruttan, one man, instead of the classical and correct oruvan. With this exception its use is exclusively feminine. The same suffix is iti or ti in Canarese — e.g., arasiti, a queen (corre- sponding to the Tamil rdsdtti), okkalati, a farmer's wife. The Telugu uses adi or di — e.g., k67nati-{y)-adi or komati-di, a woman of the Komti caste ; mdla-di, a Paria woman ; chinna-di, a girl. It seems to me evident, not only that all these suffixes are identical, but that the Telugu form of the demonstrative neuter singular, viz., adi, it, which is used systematically by Telugu to signify she, is the root from whence they have all proceeded. Another feminine singular suffix of appellatives occasionally used in the Dra vidian languages may possibly have been derived from the imitation of Sanskrit. It consists in the addition of i to the crude or neuter noun ; and it is only in quantity that this i differs from the long I, which is so much used by Sanskrit as a feminine suffix. In the majority of cases it is only in connection with Sanskrit deriva- tives that this suffix is used ; but it has also come to be appended to some pure Dravidian nouns — e.g., talei-{v)-i, Tam. a lady (compare talei-{v)-an, a lord), from talei, a head ; compare also the Gond ferd- gal, a boy, with ferdgi, a girl. This feminine suffix is not to be con- founded with i, a suffix of agency, which is much used in the forma- tion of nouns of agency and operation, and which is used by all genders indiscriminately. See " Verbal Derivatives," at the close of the part on " The Verb." (3). Neuter Singular. — There is but little which is worthy of remark in the singular forms of neuter Dravidian nouns. Every Dravidian noun is naturally neuter, or destitute of gender, and it becomes mas- culine or feminine solely in virtue of the addition of a masculine or feminine suffix. When abstract Sanskrit nouns are adopted by the Dravidians, the neuter nominative form of those nouns (generally ending in am) is preferred. Sanskrit masculines, with the exception of those which denote rational beings, are made to terminate in am, being treated as neuters ; and there are also some neuter nouns of pure Dravidian origin which end in am, or take a7n as their formative. The Dravidian termination^m is not to be regarded, however, as a sign of the neuter, or a neuter suffix, though such is often its character in Sanskrit. It is merely one of a numerous class of formatives, of which much use is made by the Dravidian dialect, and by the 230 THE NOUN addition of which verbal roots are transformed into derivative nouns. Such formatives are to be regarded as forming a part of the noun itself, not of the inflexional additions. See "Verbal Derivatives," at the close of the section on " The Verb." All animated beings destitute of reason are placed by Dravidian grammarians in the caste-less, or neuter class, and the nouns that denote such animals, both in the singular and in the plural, are uni- formly regarded as neuter or destitute of gender, irrespective of the animal's sex. If it happen to be necessary to distinguish the sex of any animal that is included in this class, a separate word signifying male or female, he or she, is prefixed. Even in such cases, however, the pronoun with which the noun stands in agreement is neuter, and notwithstanding the specification of the animal's sex, the noun itself remains in the caste-less or neuter class. For this reason, suffixes expressive of the neuter gender, whether singular or plural, were not much required by Dravidian nouns. The only neuter singular sufiix of the Dravidian languages which is used in the same manner as the masculine an or adu, and the feminine al, is that which constitutes the termination of the neuter singular of demonstrative pronouns and appellative nouns. This pronoun is in Tamil, Canarese, and Malay- alam, adu, that, idu, this ; in Telugu adi, idi ; in Gond ad, id. In the Tulu pronoun the d has dropped out. The pronoun ' that ' is avu. Dr Gundert considers this simply a corruption, and he shows that the language had its neuter singular in d originally, like its sister languages, by adducing such words as att\ it is not, which was evi- dently aldu, originally, like the Tamil allaud (old Tam. andru = aldu), in which the suffix du or d is the formative of the neuter singular. The same neuter demonstrative, or in some instances its termina- tion only, is used in the conjugation of Dravidian verbs as the sign of the neuter singular of each tense, and in Telugu as the sign of the feminine singular also. The bases of the Dravidian demonstratives being a and i {a remote, i proximate), that part of each pronoun which is found to be annexed to those demonstrative vowels is evidently a suffix of number and gender ; and as the final vowels of ad-u, ad~i, id-u, id-i, are merely euphonic, and have been added only for the purpose of helping the enunciation, it is evident that d alone constitutes the sign of the neuter singular. This view is confirmed by the circumstance that d never appears in the neuter plural of this demonstrative, but is replaced by ei, u, i, or short a, with a preceding euphonic v or n — e.g.^ compare adu (a-d-u), Tam. that, with ava NUMBER — PLURALISATION 23 I (a-{v)-a), Malayalam, those. It will be shown afterwards that this final a is a sign of the neuter plural. Appellative nouns which form their mascuHne singular in Tamil in an, and their feminine singular in al, form their neuter singular by annexing du, with such euphonic changes as the previous consonant happens to require — e.g., nalla-du, a good thing ; al-du, euphonically andru, a thing that is not ; periya-du or peri-du, great, a great thing. This neuter singular suffix d is largely used in all the dialects in the formation of verbal nouns — e.g., fSgiia-du, Tam. the act of going, pona-du, the having gone, fova-du, the being about to go. This form has been represented by some, but erroneously, as an infinitive : it is a concrete verbal or participial noun of the neuter gender, which has gradually come to be used as an abstract. The affinities of the neuter singular suffix in d appear to be ex- clusively Indo-European, and they are found especially in the Indo- European pronouns and pronominals. We may observe this suffix in the Sanskrit tat, that ; in tyat, that ; in adas, a weakened form of adat, that ; in etat, this ; and in the relative pronoun yat, who, which, what. We find it also in the Latin illud, id, &c. (compare the Latin id with the Tamil id-u, this) ; and in our English demonstrative neuter it (properly hit), the neuter of he, as also in what, the neuter of who. Compare also the Vedic it, an indeclinable pronoun, described as " a petrified neuter," which combines with the negative particle na to form net, if not, apparently in the same manner as in Telugu the aoristic neuter ledu, there is not, is compounded of the negative la for ila, and the suffix du. Though the Dravidian languages appear in this point to be allied to the Sanskrit family, it would be unsafe to suppose that they borrowed this neuter singular suffix from Sanskrit. The analogy of the Dravidian neuter plural in a, which, though Indo-European, is foreign to Sanskrit, and that of the remote and proximate demonstrative vowels a and i, which, though known to the Indo-European family, are used more system- atically and distinctively by the Dravidian languages than by any other class of tongues, would lead to the supposition that these particles were inherited by the Dravidian family, in common with Sanskrit, from a primitive prse-Sanskrit source. The Plural : Principles of Pluralisation. — In the primitive Indo-European tongues, the plural is carefully distinguished from the singular ; and with the exception of a few nouns of quantity which have the form of the singular, but a plural signification, the 232 THE NOUN number of nouns is always denoted by their inflexional terminations. Nouns whose number is indefinite, like our modern English sheep, are unknown to the older dialects of this family. In the languages of the Scythian group a looser principle prevails, and number is gener- ally left indefinite, so that it is the connection alone which determines whether a noun is singular or plural. Manchu restricts the use of its pluralising particle to words which denote animated beings : all other words are left destitute of signs of number. Even the Tartar, or Oriental Turkish, ordinarily pluralises the pronouns alone, and leaves the number of other nouns indeterminate. In Brahui also, the number of nouns is generally left undefined ; and when it is desired to attach to any noun the idea of plurality, a word signifying many or several is prefixed to it. Notwithstanding this rule, Brahui verbs are regularly pluralised ; and the number of an inde- terminate noun may often be ascertained from the number of the verb with which it agrees. With respect to principles of pluralisation, most of the Dravidian tongues differ considerably from the Indo-European family, and accord on the whole with the languages of the Scythian stock. The number of Tamil nouns, especially of neuter nouns, is ordinarily in- definite ; and it depends upon the connection whether any noun is to be regarded as singular or as plural. It is true that when more persons than one are referred to, the high-caste or rational pronouns that are used are almost invariably plural, and that even neuter nouns themselves are sometimes pluralised, especially in polished prose compositions ; but the poets and the peasants, the most faithful guardians of antique forms of speech, rarely pluralise the neuter, and are fond of using the singular noun in an indefinite singular-plural sense, without specification of number, except in so far as it is expressed by the context. This rule is adhered to with especial strictness by Tamil, which in this, as in many other par- ticulars, seems to exhibit most faithfully the primitive condition of the Dravidian languages. Thus in Tamil, mddu, ox, means either an ox or oxen, according to the connection ; and even when a numeral which necessarily conveys the idea of plurality is prefixed, idiomatic speakers prefer to retain the singular or indefinite form of the noun. Hence they will rather say, ndlu mddu meygivadu, literally four ox is feeding, than ndlu ynddugal ^neygindrana, four oxen are feeding, which would sound stiff and pedantic. Telugu is an exception to this rule. In it neuter nouns are as regularly pluralised as masculines or feminines, and the verbs with which they agree are pluralised to NUMBER— PLURALISATION 233 correspond. In Tuda, on the other hand, the only words that appear to be ever pluralised are the pronouns and the verbs which have pronouns for their nominatives. In Coorg neuter nouns have no plural. We find a similar usage occasionally even in English, as Mr C. P. Brown points out, in the military phrases, a hundied foot, three hundred horse. In Tamil, even when a neuter noun is pluralised by the addition of a pluralising particle, the verb is rarely pluralised to correspond ; but the singular form of verb is still used for the plural — the number of the neuter singular being naturally indeterminate. This is almost invariably the practice in the speech of the lower classes ; and the colloquial style of even the best-educated classes exhibits a similar characteristic. Tamil contains, it is true, a plural form of the third person neuter of the verb ; but the use of this neuter plural verb is ordinarily restricted to poetry, and ^ven in poetry the singular number both of neuter nouns and of the verbs that correspond is much more commonly used than the plural. It should be remarked also, that the third person neuter of the Tamil future, or aorist, is altogether destitute of a plural. In this particular, therefore, the Tamil verb is more decidedly Scythian in character than the noun itself. Max Miiller supposes that a Dra vidian neuter plural noun, with its suffix of plurality, is felt to be a compound (like animal-mass for animals, or stone-heap for stones), and that it is on this account that it is followed by a verb in the singular. The explanation I have given seems to me preferable. The number of all Dravidian nouns, whether high-caste or caste-less, was originally indefinite : the singular, the primitive condition of every noun, was then the only number which was or could be recognised by verbal or nominal inflexions, and plurality was left to be inferred from the context. As civilisation made progress, the plural made its appearance, and effected a permanent settlement in the department of high-caste or masculine-feminine nouns and verbs ; whilst the number of caste-less or neuter nouns, whether suffixes of plurality were used or not, still remained generally unrecognised by the verb in the Dra- vidian languages. Even where the form exists it is little used. It is curious that in this point the Greek verb exhibits signs of Scythian influences, or of the influences of a culture lower than its own, viz., in the use of the singular v#rb for the neuter plural. The Dravidian languages ordinarily express the idea of singu- larity or oneness, not by the addition of a singular sufiix to nouns and pronouns, or by the absence of the pluralising particle (by which 234 THE NOUN number is still left indeterminate), but by prefixing the numeral adjective one. Thus niddu, Tam. ox, does not mean exclusively either an ox or oxen, but admits of either meaning according to circumstances ; and if we wish distinctly to specify singularity, we must say oru mddu, one or a certain ox. Europeans in speaking the Dravidian dialects use this prefix of singularity too frequently, misled by their habitual use of an indefinite article in their own tongues. They also make too free a use, in Tamil, of the distinctively plural form of neuter nouns, when the objects to which they wish to refer are plural. Occasionally, when euphony or usage recom- mends it, this is done by Tamilians themselves, but as a general rule the neuter singular is used instead of the neuter plural, and that not in Tamil only, but also in almost all the languages of the Scythian group. Another important particular in which the Indo-European lan- guages differ from the Scythian is, that in the former the plural has a different set of case-terminations from the singular, by the use of which the idea of plurality is not separately expressed, but is com- pounded with that of case-relation ; whilst in the latter family the plural uses the same set of case-terminations as the singular, and plurality is expressed by a sign of plurality common to all the cases, which is inserted between the singular, or crude form of the noun, and the case-terminations. I call it a sign of plurality, not a noun denoting plurality, for in many instances only a fraction of a word, perhaps only a single letter, remains. In the Indo- European languages, each inflexion includes the twofold idea of number and of case. Thus there are a genitive singular and a genitive plural, each of which is a complex idea ; but there is no inflexion which can be called genitive, irrespective of number ; and in many instances (this of the genitive being one) there is no apparent con- nection between the case-termination of the singular and that which is used in, and which constitutes, the plural. In those few cases in which the sign of number and the sign of case seem to have been originally distinct, and to have coalesced into one, the sign of case seems to have preceded that of number — e.g., the Gothic plural accusative ns is derived from n or m, the sign of the accusative singular, and s, the sign of plurality. When the Scythian family of languages is examined, it is found that each of their case- signs is fixed and unalterable. It expresses the idea of case and nothing more, and is the same in the plural as in the singular, with the exception of those few trivial changes which are required by NUMBER — PLURALISATION 235 euphony. The sign of plurality also is not only distinct from the case-sign, but is one and the same in all the cases. It is an unalter- able postposition — a fixed quantity ; and it is not post-fixed to the case-sign, much less compounded with it, as in the Indo-European languages, but is prefixed to it. It is attached directly to the root itself, and followed by the signs of the different cases. In the Dra vidian languages a similar simplicity -and rigidity of structure characterises the use of the particles of plurality. They are added directly to the crude base of the noun (which is equivalent to the nominative singular) and are the same in each of the oblique cases as in the nominative. The signs of case are the same in the plural as in the singular, the only real difference being that in the singular they are suffixed to the crude noun itself, in the plural to the pluralising particle, after the addition of that particle to the crude noun. The only exception to this rule is in Tulu, in which a, the sign of the genitive, keeps its place in the singular, as in the other dialects, but is weakened to e in the plural. In Hungarian, hdz, a house, is declined as follows : — Singular. Nom. hdz. Gen. hdz-nak. Dat. hdz-nak. Ace. hdz-at. Plural. Nom. hdzaJc. Gen. hdz-ak-nak. Dat. hdz-ak-nak. Ace. hdz-ak-at. In Tamil, manei, a house, is declined as follows :- Singular. Plural. Nom. manei. Nom. manei-gal. Ace. manei-{y)-ei. Ace. manei-gal-ei. Instr. manei-{y)-dl. Instr. manei-gal-dl. Conj. manei- {y)-6du. Conj. manei-gal- odu. Dat. 7)ianei-kku. Dat. 7nanei-gal-{u)-kku. Ablat. manei-{y)-il-irundu. Ablat. manei-gal-il-irundu Gen. manei-{y)-inadu. Gen. manei-gal-inadu . Locat. manei-{y)-idatt-il. Locat. manei-gal-idatt-il. Voc. manei-{y)-e. Voc. manei-gal-e. (See f aradigm of Nouns.) We here see that the particular signs which are used to express plurality and as exponents of case, in Tamil and Hungarian respec- 236 THE NOUN tively, are taken from the resources of each language ; whilst the manner in which they are used in both languages is precisely the same. The neuter of Dra vidian nouns being identical with the crude base, when the pluralising particle is attached to a neuter noun, it is at- tached to it not as a substitute for any suffix of the singular, but directly and without any change : it is attached to it pure and simple. In the case of masculine and feminine nouns, including pronouns, a somewhat different method of pluralisation is necessary. The singular of the masculine and feminine is formed, as has already been pointed out, by the addition to the root of particles denoting a male or a female. Hence, to pluralise those nouns, it is necessary either to add a pluralising particle to the masculine and feminine suffixes, or to substitute for those suffixes an epicene pluralising particle. In all the Dra vidian languages the primitive plan of pluralising these two classes of nouns seems to have been that of substituting for the masculine and feminine singular suffixes a suffix of plurality which applied in common to men and women, without distinction of sex. This is the mode which is still used in most of the dialects ; but in Telugu it retains its place only in connection with pronouns and verbs, and has disappeared from substantives, which form their plural by means of a neuter suffix. The classification of Dra vidian nouns into rationals and irrationals has already been explained ; it has also been shown that in the singular, the masculine of rational nouns is distinguished from the feminine. In the plural both those genders are combined ; the high-caste particle of plurality, or plural of rational beings, is the same for both genders, and includes men and women, gods and god- desses, without distinction of sex. Irrational or neuter nouns have a particle of plurality different from this, and in general peculiar to themselves. Hence the Dra vidian languages have one form of the plural which may be called epicene or masculine-feminine, and another which is ordinarily restricted to the neuter ; and by means of these pluralising particles, gender and number are conjointly expressed in the plural by one and the same termination. The masculine-feminine plural expresses the idea of plurality conjointly with that of rationality ; the neuter plural, the idea of plurality conjointly with that of irrationality. Arrangements of this kind for giving combined expression to gender and number are very commonly observed in the Indo- European family ; and even the plan of classing masculines and NUMBER — PLURALISATION 237 feminines together in the plural, without distinction of sex, is also very common. Thus, the Sanskrit plural in as is masculine-feminine ; so is the Latin plural in es, and the Greek in es. The chief difference with respect to this point between the Dravidian system and the Indo-European one lies in this, that in the Dravidian languages the masculine-feminine particle of plurality is carefully restricted to rational beings ; whereas in the Indo-European languages irrational and even inanimate objects are often complimented with inflexional forms and pluralising particles which imply the existence, not only of vitality, but even of personality — that is, of self-conscious intelligence. A still closer analogy to the Dravidian system is that which is exhibited by the New Persian. That dialect possesses two pluralising particles, of which one, an, is suffixed to nouns denoting living beings, the other, hd, to nouns denoting inanimate objects. The particles employed in Persian are different from those which are used in the Dravidian languages, but the principle is evidently analogous. The Persians specialise life, the Dravidians reason ; and both of them class the sexes together indiscriminately in the plural. In Telugu some confusion has been introduced between the epicene sign of plurality ar-u, and the neuter lu. The pronouns pluralise their masculines and feminines regularly by substituting ar-u for their mas- culine and feminine singular suffixes, whilst the substantives and some of the appellative nouns append lu, which is properly the neuter sign of plurality, instead of the more correct ar-u. Thus the Telugu demonstrative pronoun vdr-u, they (the plural of vdndu, he), corre- sponding to the Canarese avar-u, exhibits the regular epicene plural ; ■whilst tnagandu,3i'husb8ind (in Tamil magan,Si son), takes for its plural not magaru, but magalu ; and some nouns of this class add lu to the masculine or feminine singular suffix — e.g., alludu, a son-in-law, makes in the plural not alluru, nor even allulu, but allundlu, nasalised from alludlu ; and instead of vdru, they, vdndlu is colloquially used, a word which is formed on the same plan as the Low Tamil avangal, they, instead of avargal, or the higher and purer avar. One of the few cases in which the irrational pluralising particle is used in the higher dialect of the Tamil instead of the rational epicene, is that of makkal {maggal), mankind, people. This is not really, however, an exception to the rule, for makJcal is regarded by Tamil grammarians as the plural*of maga (from mag-u), and the primary meaning of this seems to be child, a naturally neuter noun. Another instance of this anomaly both in Tamil and Canarese, and one to which no exception can be taken, is that of the masculine noun guru 238 THE NOUN (Sans.), a teacher. The phiral of this word is in Tam. gurukkal, in Can. guru-galu. Tulu also has gurukulu. Tulu agrees with the other dialects in using er as its sign of plu- rality in personal nouns, but differs from most of them in using this form occasionally only, and using gal, or the shape which gal assumes in Tulu, as its ordinary plural of personal nouns, as well as of neuters. Thus, the plural pronouns of the third person in ordinary use in Tulu are dkulu, they {rem.), ynokulu, they {prox.). It uses also dr' (Tam. avar) for the former, and 7ner (Tam. ivar) for the latter, but rather as honorific singulars than as plurals. It also uses ntkulu for you, instead of ir\ the latter having come to be used as an honorific singular. The Ku rational plural is ngd, which is properly an irrational one. The pronouns and participial nouns form their rational plural by the addition of dru, which is identical with the ar of the other dialects. Modern colloquial Tamil seems to have been influenced in some de- gree by the usage of Telugu, and has adopted the practice of adding the irrational plural to the rational one, thereby systematically forming a double plural ar-gal, instead of the old rational plural ar — e.g., avan, he, and aval, she, properly take avar, they, as their plural ; but the plural preferred by modern Tamil is the double one avargal. So also the plural of the second person is properly ntr ; but the plural which is most commonly used is nin-gal (from nirn, an older form of nir, and gal), which is a double plural like avar-gal. Two forms of the epicene plural being thus placed at the disposal of the Tamil people (the classical nir and avar, and the colloquial nin-gal and avar-gal), they have converted the former, in colloquial usage and in prose compositions, into an honorific singular, and the same practice is not unknown in Canarese. This usage, though universally prevalent now, was almost unknown to the poets. I have not ob- served in the poets, or in any of the old inscriptions in my possession, any instance of the use of the epicene plural as an honorific singular, ■except in connection with the names and titles of the divinities, whether those names and titles are applied to the gods themselves, or are conferred honorifically upon kings. Even in those cases, however, the corresponding pronoun follows the ordinary rule, and is very rarely honorific. In modern Telugu a double plural, similar to that of the Tamil, has gained a footing — e.g., vdra-lu (for vdr-u), they, and mira-lu (for mir-u), you. In Malay alam, avar is still constantly used for the ordinary epicene plural, and avargal is used more commonly as an honorific singular. This use of avargal is also NUMBER — THE EPICENE PLURAL 239 common in Tamil, and the corresponding gdru equally so in Telugu. (Tam. durei-avargal = Te\. dora-gdru, the gentleman, literally the gentlemen, his honour.) In Canarese, avaru is commonly used simply as a plural ; dtanu is regarded as the honorific singular, though avaru also is sometimes used in this sense, nmgal in Tamil and Malayalam is both plural and honorific singular, like Can. nivu and Tel. 7niTu. Telugu, as has been observed, pluralises masculine and feminine substantive nouns by the addition, not of the rational, but of the neuter or irrational sign of plurality. By a similar inversion of idiom, Gond sometimes uses the rational plural to pluralise neuter nouns — e.g., kdwdlor, crows. Such usages, however, are evidently exceptions to the general and more distinctively Dravidian rule, according to which the neuter pluralising particle is restricted to neuter nouns, and the epicene particle to rational or personal nouns, i.e., masculines and feminines. We shall now consider in detail the pluralising particles them- selves. 1. Epicene Pluralising Particle. — This particle is virtually one and the same in all the dialects, and the different forms it has taken are owing merely to euphonic peculiarities. In Tamil nouns, pronouns, and verbs, it assumes the forms of ar, dr, or ; ir, ir : in Canarese and Telugu, aru, aru ; dre, eru ; ri, ru : in Tulu, er : in Ku, dru : in Gond, or. The lengthened forms include the assimilated demonstra- tive vowel of the pronoun. The Brahui also forms the second person plural of its verb in ere, ure. Sec, the third person in ur or ar. I regard ar (not simply r) as probably the primitive shape of this pluralising particle, from which the other forms have been derived by euphonic mutation. It is true that m, thou, forms its plural in modern Tamil by simply adding r ; but this does not prove that r alone was the primitive form of the epicene plural, for an older form of nir, you, is ni-{v)-ir or ni-{y)-ir, from which nir has evidently been derived. It might naturally be supposed that in this case ir is used instead of ar, through the attraction of the preceding long vowel i ; but we also find ir used as a pluralising particle in 7nagalir, High Tam. women, and also a longer form, ir, in magalir ; consequently ir has acquired a position of its own in the language, as well as ar. All that we can certainly conclude respecting the original shape of this particle is that the final r, which is plainly essential, was preceded by a vowel, and that that vowel was probably a. May we regard this a as identical with the demonstrative a ? On this supposition, ar would be simply an older form of a(v)ar, and would mean those persons ; 240 ' THE NOUN ir would mean these persons. On the other hand, may we venture to identify ir and ir with the second numeral ir and ir, two ? nir would on this supposition have been originally a dual, meaning ye two. It is not impossible, indeed, that the plural may in all languages have been developed out of the dual. In Bornu, we, ye, they, mean literally we two, ye two, they two. The chief difficulty in the way of accepting this as the origin of the Tamil ir or nir, you, is that the ar of avar, they, which is the form of the epicene plural most com- monly used, would have to be regarded as a corruption and a mistake, which it does not appear to be. The Canarese rational plural suffix andar — e.g., avandar-u (for avar-u), illi, and ivandar-u (for ivar-u), hi seems to be identical with the Tel. indefinite plural andar-u, indar-u, so many, the final ar of which is the ordinary suffix of epicene plural. In old Canarese, ir is a plural vocative of epicenes. Tamil and Malayalam have another particle of plurality applicable to rational beings, viz., mar, or in High Tamil mar, which has a considerable resemblance to ar, and is evidently allied to it. It is suffixed to the noun which it qualifies in a different manner from ar ; for whilst ar is substituted for the masculine and feminine suffixes of the singular, not added to them, mar is generally added to the singu- lar suffix by idiomatic writers and speakers. Thus in Tamil, furushan (Sans.), a man, a husband, when pluralised by suffixing ar becomes furushar ; but if mar is used instead of ar, it is not sub- stituted for an, the masculine singular suffix, but appended to it — e.g., furuslian-mdr, not furusha-mar. mdr, it is true, is sometimes added to ar — e.g., furusliar-mdr ; but this is considered unidio- matical. mar is also sometimes used as an isolated particle of plu- rality in a pecuharly Scythian manner — e.g., tdy-tagaf fan-mar, Tam. mothers and fathers, parents ; in which both mother and father are in the singular, and mdr is separately appended to pluralise both. Probably there was originally no difference in signification between ar and mar or mdr. In modern Tamil, mar is suffixed to nouns signifying parents, priests, kings, &c., as a plural of honour, but it may be suffixed, if necessary, to any class of nouns denoting rational beings. In Malayalam it is used with a wider range of application than in Tamil, and in cases in which an honorific meaning cannot be intended — e.g., kallan-mdr, thieves. The antiquity of many of the forms of the Malayalam grammar favours the supposition that in ancient Tamil, which was apparently identical with ancient Malay- alam, mar or mdr may generally have been used instead of ar, as the ordinary pluralising particle of high-caste nouns. NUMBER— THE EPICENE PLURAL 24 1 A few traces of the use of the particle mar, as the ordinary sign of epicene plurality, survive in classical Tamil, mar, which is evidently equivalent to mar, forms the epicene plural of a few nouns — e.g., enmar, eight persons. As ar is older than dr (the latter being euphon- ised from avar by the coalescence of the vowels), so in like manner it may be concluded that mar is older than mar. This mar again seems to have been derived from var, or to be an older form of it, m and v being sometimes found to change places. When the Tam. ndlvar, four persons, eivar, five persons, are compared with enmar, eight persons, it is evident that mar is equivalent to var, and probable that the use of m for v is an euphonic change, ndlmar would be im- possible in classical Tamil ; enmar is not only possible, but euphonic. var is a very common formative of epicene appellative nouns in Tamil and Malayalam, and often appears as avar, in which case we cannot but regard it as the pronominal avar, they, used as a plural formative — e.g., vinnavar, Tam. the heavenly ones, from vin, heaven, with avar affixed. Compare this form with participial nouns like seydavar, Tam. they who did, from seyd-(u), having done, and avar, they, and the identity in origin of the avar of vinnavar and that of seydavar will be evident. This avar, again, seems to have been abbreviated into var, like the Telugu avaru, they, into vdru. The v of eivar, five persons, might be regarded as simply euphonic, as a soft consonant inserted to prevent hiatus, but this explanation is inadmissible in the case of ndlvar, four persons, there being no hiatus here to be provided against. This var being identical in use with avar, it may safely be concluded to be identical with it in origin ; and if var is a pronominal form, an abbreviation of avar, may not mar be the same ? The example of the lengthening of ar into dr (i.e., the substitution of the plural pronoun itself in an euphonised form for the bare particle of plurality) would naturally lead to the lengthening of var into var (the origin of the v being by this time forgotten) ; and when once mar had established itself instead of var, this also would naturally be lengthened into mar. Thus tagappan-mdr would come to be used instead of tagap'pan-vdr. This suffixing of the plural formative to the singular noun, which seems so irregular, may be compared with the mode in which the singular is still honorifically pluralised by the addition of the plural pronoun — e.g., tagappan- avargal, father, and especially with the still more common tagap'pan- dr, forms which, though used as singular, are grammatically plurals. tagapfan-mdr is invariably used as a plural, but it seems not im- probable that it is identical in origin with tagappan-dr. 242 THE NOUN In this explanation of mar I have followed a suggestion of Dr Gun- dert ; but I find myself unable to follow him also in supposing the Tamil verbal terminations mar, mar, mundr, to be identical in origin with the pluralising particles mar, mar, though I admit that at first sight it seems impossible to suppose them to be otherwise. These are poetical forms of the future tense only, which do not make their appearance in any other part of the verb, and the m they contain will be found, I think, on examination, to have a futuric, not a pro- nominal signification. It appears to be identical with h or v, the sign of the future, and there appears no reason why m should not be used instead of v or 6 in this instance, as well as in others that have already been pointed out. The impersonal future of en, to say, in classical Tamil is enba. When the personal terminations of the third person plural are suffixed to the root, we find ' they will say ' repre- sented indifferently by enbar, or enmar, enbdr enmdr, or enmandr. The force of the future, according to Tamil grammarians, being con- veyed by each of these forms in m, precisely as by each of the forms in h, 1 conclude that this future m must be regarded as independent of the m of the pluralising particle, and the resemblance between the two, however complete, to be after all accidental. Dr Gundert suggests that the final dr of enmandr, preceded by an, may be ex- plained by a comparison of it with tagappan-dr, a form already referred to, and here I am disposed to coincide with him. We have now to inquire whether ar, dr, mar, and 7ndr, the Dra- vidian plurals of rationality, appear to sustain any relation to the plural terminations, or pluralidng suffixes, of other languages. It might at first sight be supposed that the formation of the plural by the addition of r to the singular which characterises some of the Teutonic tongues, is analogous to the use of / or ar in the Dravidian languages. In the Icelandic the most common plural is that which terminates in r — sometimes the consonant r alone, sometimes the syllables ar, ir, ur — e.g., konungur, kings. A relic of this plural may be traced in the vulgar English childer, for children. The same plural appears in the old Latin termination of the masculine plural in or which is found in the Eugubian tables — e.g., subator for subacti, and screhitor for scrifti. Compare also mas, the termination of the first per- son plural of verbs in Sanskrit, with mar, the corresponding termina- tion in Irish, answering to the Doric /xes {mes) and the ordinary Greek fxev {men). In these cases, however, the resemblance to the Dravidian plural ar is perhaps rather apparent than real ; for the final r of these forms has been hardened from an older s, and the s of the Sanskrit NUMBER — THE EPICENE PLURAL 243 nominative singular is hardened in some of the Teutonic tongues into r, equally with the as or s of the plural ; whilst there is no evidence, on which we can rely, of the existence of a tendency in the Dravidian languages to harden s into r, and therefore no evidence for the supposition that the Dravidian epicene ar has been derived from, or is connected with, the Sanskrit masculine-feminine as. It should also be noted that the Irish mar is a compound of two forms, ma, the representative of the singular of the personal pronoun I, and r, the hardened equivalent of the plural suffix s ; and that, therefore, it has no real resemblance to the Dravidian mar, which is entirely and exclusively a plural suffix of the third person. There is more probability perhaps of the Dravidian plural suffixes being related to the pluralising particles of some of the Scythian languages. The Turkish plural suffix, which is inserted, as in the Dravidian languages, between the crude noun and each of the case- terminations, is lar or ler — e.g., dn-lar, they. Dr Logan says, but on what authority does not appear, that nar is a plural suffix in K61. Mongolian nouns which end with a vowel are pluralised by the addition of nar or ner, a particle which is evidently related to, or identical with, the Turkish lar or ler : and the resemblance of this Mongol suffix nar to the Dravidian mar, both in the final ar and in the nasal prefix, is remarkable. It is well known that m evinces a ten- dency to be softened into n (witness the change of the Sanskrit mama, my, into mana in Zend) ; and in this manner it may perhaps be supposed that the Dravidian mar may be allied to the High Asian nar. The Tamil ileinar {ilei-nar), young people, a plural appellative noun, formed from ilei, youth, exhibits a form of pluralisation which at first sight seems very closely to resemble the Mongolian nar. Nay, nar is actually used in this very instance instead of nar by some of the poets, and it is certain that n and n often change places. Un- fortunately we find this n or n in the singular, as well as the plural ; which proves it to be inserted merely for euphony in order to prevent hiatus, and therefore ileinar must be re-divided, and represented not as ilei nar, but as ilei-{n)-ar or ^'fe^-(n)-ar, equivalent to ilei-{y)-ar. The resemblance of the final syllable nar, of the Tamil verb enmandr, already commented on, to the Mongolian plural suffix nar, seems more reliable, and yet that also seems to disappear on further examination. « Turkish, besides its ordinary plural lar or ler, uses 2; as a plural suffix of the personal pronouns, as may be observed in hiz, we, and siz, you ; and the Turkish terminal z corresponds to the r of some 244 THE NOUN other Scythian languages. Thus ydz, Turkish, summer, is in Magyar ydr or ndr (compare the Tamil ndyit-u, the sun). It would almost appear, therefore, that the Turkish suffix of plurality has undergone a process of change and comminution similar to that of the Tamil, and that the Turkish z and the Tamil r are remotely connected, as the last remaining representatives or relics of mar, nar, and far. Though I call attention to these and similar Scythian correspond- ences, I wish it to be understood that I do so only in the hope that they will be inquired into more thoroughly, and the existence or otherwise of a real relationship between them and the Dravidian forms with which they correspond ascertained. I attribute much more weight to the resemblance between the Dravidian languages and those of the Scythian group in the use they make of these particles of plurality, and the manner in which they connect them with the case-sign than to any resemblance, however close, that can be traced between the particles themselves. We should look, I think, not so much at the linguistic materials used by the Scythian languages and the Dravidian respectively, as at the use they severally make of those materials. 2, Pluralising Particles of the Neuter. — There are two neuter pluralising particles used by the Dravidian languages : — (1.) The Neuter Plural Suffix gal, ivith its Varieties. — It has already been noticed that gal is occasionally used in Tamil and Canarese as the plural suffix of rational nouns and pronouns ; and that the corresponding Telugu lu is still more systematically used in this manner. Nevertheless, I have no doubt that it was originally and is essentially a suffix of the neuter plural. This suffix is in both dialects of the Tamil gal — e.g., kei-gal, hands, with only such changes as are required by Tamilian rules of euphony. In accordance with one of those rules, when g, the initial consonant of gal, is doubled, or pre- ceded without an intermediate vowel by another consonant, gal is regularly hardened into kal or khal. Thus kal-gal, stones, is changed by rule into kax-kal. gal is occasionally lengthiened in Tamil poetry into gal. In Malayalam this particle is generally gal, kal, or kkal, but sometimes the initial k coalesces with a preceding nasal and becomes n — e.g., nin^ial, you, instead of nim-kal, in Tamil nm-gal. In modern Canarese we have gal-u, in ancient gal, as in Tamil. The three southern idioms are in perfect agreement with respect to this particle, but when we advance further north we shall find its shape considerably modified. In Telugu the corresponding neuter plural suffix is lu, of which the NUMBER — THE NEUTER PLURAL 245 I answers, as is usual in Telugu, to the lingual I of the other dialects : l-u, therefore, accords with the final syllable of the Canarese gal-u. The only real difference between the Telugu and the Tamil-Canarese consists in the omission by the former of the initial consonant k or g. Traces, however, exist, in Telugu, of the use of a vowel before lu. Thus, in gundlu, horses, the long d is derived from the combination of the short final a of the inflexional base gutta and a vowel, evidently a, which must have preceded lu. We thus arrive at al-u as the primitive form of the Telugu plural ; and it is obvious that al-u could easily have been softened from gal-u. Conjecture, however, is scarcely needed, for in some nouns ending in n-u, of which the Tamil equivalents end in m, the old Dravidian pluralising particle in gal is exhibited in Telugu almost as distinctly as in Tamil. Thus, holan-u a tank (Tamil kulam), takes as its plural kolan-kul-u, a word cited in this form by Nannaya Bhatta (Tamil kulan-gal), and gon-u, the name of a species of tree, forms its plural in gon-gul-u. When kul-u and gul-u are compared with the Tamil-Canarese forms kal, gal, and gal-u, it is obvious that they are not only equivalent but identical. An illustration of the manner in which the Telugu lu has been softened from gal-u, may be taken also from colloquial Tamil, in which avar- gal, they, is commonly pronounced aval ; Pirdynatmrgal, Brahmans, Pirdmandl. k or g is dropped or elided in a similar manner in many languages of the Scythian family. Tulu, though locally remote from Telugu, follows its example in many points, and amongst others in this. It often rejects the k or g of the plural, and uses merely lu, like Telugu. It uses the full form kulu more rarely. The same form of the pluralising particle appears in the languages of some of the tribes of the north-eastern frontier — languages which possibly form a link of comiection between the Dravidian and the Tibetan families. In the Miri or Abor-Miri dialect, no, thou, forms its plural in nolu, you ; and in the Dhimal, ne, thou, is pluralised into nyel, you. The pronoun of the Mikir is pluralised by adding li — e.g., na-li, you, whilst substantives have no plural form. In the Dhimal, substantive nouns are pluralised by the addition of galai, which is possibly the origin of the pronominal plural /, though this particle or word, galai, is not compounded with, or agglutinated to, the noun, but placed after it separately. Though it is used as a separate word, it does not seem to retain a^j signification of its own independent of its use as a postposition. The resemblance of galai to the Tamil- Canarese gal or galM, is distinct and remarkable. The pluralising particle of the Naga also is khala. 246 THE NOUN It is not an uncommon occurrence to find one portion of a much- used prefix or suffix in one language or dialect of a family, and another portion of it in another member of the same family. Seeing, there- fore, that the Telugu has adopted the latter portion of the particle kal, gal, or galu, and omitted the initial ka, ga, or k, we may expect to find this k used as a pluralising particle in some other Dravidian dialect, and the final lu or I omitted. Accordingly, in Gond we find that the plural neuter is commonly formed by the addition of k alone — e.g., nai, a dog, naik, dogs (compare Tamil ndykal, pronounced nay gal). The Seoni-Gond forms its plural by adding nk — e.g., neli, a field, nelnk, fields. The Ku dialect uses ngd, and also ska, of all which forms ^ or ^ constitutes the basis. k is sometimes found to interchange with t, especially in the lan- guages of High Asia. This interchange appears also in the Gond pluralising particle ; for whilst k is the particle in general use, the pronouns of the first and second persons form their plurals, or double plurals, by the addition of t to the nominative — e.g., amat, we, imat, you. The same interchange between k and t appears in Brahui. Though a separate word is usually employed by Brahui to denote plurality, a suffix in k is also sometimes used ; but this k is found only in the nominative plural, and is replaced by t in the oblique cases. When we turn to the grammatical forms of the Finnish family of languages, we find some tolerably distinct analogies to this Dra- vidian plural suffix. Compare with the Dravidian forms noticed above the Magyar plural in k or ak ; the Lappish in k, cJi, or li : also the t by which k is replaced in almost all the other dialects of the Finnish family ; and observe the reappearance of the sound of I in the Ostiak plural suffix tl. In Ostiak, the dual suffix is kan, or gan ; in Samoied-Ostiak, ga or ka ; in Kamass, gai. Castren supposes these suffixes to be derived from the conjunctive particle ka or ki, also ; but their resemblance to the Dravidian signs of plurality is worth noticing. Even Armenian forms its plural in k — e.g., tu, thou, tuk, you ; sir em, I love ; siremk, we love. In Turkish also, k is the sign of plurality in some forms of the first person plural of the verb — e.g.. Mum, I was, iduk^,vfQ were, t, on the other hand, is the sign of the plural in Mongolian, and in Calmuck is softened into d. Even in Zend, though a language of a different family, there is a neuter plural in t. Thus, for imdni (Sans.), these things, Zend has imat. In those instances of the interchange of t and k, in which it can be NUMBER — THE NEUTER PLURAL 247 ascertained with tolerable clearness which consonant was the one originally used and which was the corruption, t sometimes appears to be older than h. Thus, the Doric Trjvo<^ [tenos) is in better accordance with related words, and therefore probably older, than the iEolian ktJvos (kenos), the origin of e-zcetvos (e-keinos). The Semitic pronoun or pronominal fragment ta, thou (preserved in attd and antd), is also, I doubt not, a more accurate and older form than the equivalent or auxiliary suffix kd. In several of the Polynesian dialects, k is found instead of an apparently earlier Sanskrit or prse-Sanskrit t. On the other hand, as Dr Gundert points out, k sometimes appears to be older than t, particularly in Greek — e.g., compare Gr. ns [tis) with Sans. kas. If, in accordance with a portion of these precedents, where k and t are found to be interchanged, t is to be regarded as older than k, it would follow that kal, the Dravidian plural suffix now under con- sideration, may originally have been tal. I cannot think that the Dravidian gal has been derived, as Dr Stevenson supposed, from the Sanskrit sakala (in Tamil sagala), all. kal, the base of sa-kala, has been connected with oA-09 {hol-os) ; but el, the root signifying ' all,' which is found in all the Dravidian languages — Tel. ella ; Tam.-Mal. elld, elldm, elldvum (the conjunction um intensifies the meaning) — if it were related to any Indo-European word at all, which is doubtful, would be connected, not with the Gr. 6A {hoi), Heb. kol, Sans, sar-va, &c., but with the Germanic alia, Eng. all. The Dravidian tala, one of the meanings of which is a heap, a quantity, would suit very well ; but even this derivation of kal is destitute of evidence. The sup- posititious Dravidian tal may be compared with the Ostiak plural suffix tl ; but in the absence of evidence it is useless to proceed with conjectural analogies. The New Persian neuter plural, or plural of inanimate objects, which corresponds generally to the Dravidian neuter plural, is hd, a form which Bopp derives with much probabiHty from the Zend. It may here be mentioned, though I do not attach any importance to a resemblance which is certainly accidental, that the Tamil plural gal sometimes resembles ha in the pronunciation of the peasantry — e.g., irukkiidrgal, they are, is vulgarly pronounced irukkitdha. (2.) Neuter Plural Suffix in a. — In addition to the neuter plural in gal, with its varieties, we find in nearly all the Dravidian languages a neuter plural in short a, 0$ traces of the use of it at some former period, gal, though a neuter plural suffix, is occasionally used, especially in the modern dialects, as the plural suffix of rationals ; but in those dialects in which a is used, its use is invariably restricted 248 THE NOUN to neuters, and it seems therefore to be a more essentially neuter form than gal itself. We shall first examine the traces of the existence and use of this suffix which are contained in Tamil, gal is invariably used in Tamil as the plural suffix of uncompounded neuter nouns ; but a is pre- ferred in the classical dialect for pluralising neuter compounds, that is, appellative nouns, or those which are compounded of a base and a suffix of gender, together with demonstrative pronouns, pronominal adjectives, and participial nouns. Even -in the ordinary dialect, a is generally used as the suffix of the neuter plural in the conjugation of verbs. The second line in one of the distichs of Tiruvalluvar's " Kural " contains two instances of the use of a as a neuter plural of appellative nouns — e.g., dgula mra pira, vain shows (are all) other (things). The first of these three words is used adjectivally ; and in that case the final a is merely that which remains of the neuter termination am, after the regular rejection of m ; but the next two words, nira and pira, are undoubted instances of the use of a as a suffix of the neuter plural of appellatives. The much-used Tamil words pala, several, or many (things), and sila, some, or some (things), (from pal and HI), though commonly considered as adjectives, are in reality neuter plurals — e.g., fini pala, diseases (are) many ; pala-(v)-in-pdl, the neuter plural gender, literally the gender of the many (things). This is the case also in poetry in Malayalam. The use of these words adjectivally, and with the signification, not of the collective, but of the distributive plural, has led some persons to overlook their origin and real meaning, but I have no doubt that they are plurals. So also alia, not, is properly a plural appellative. It is formed from the root al, not, by the addition of a, the plural suffix, and literally means things that are not, and the singular that corresponds to alia is al-du, not, euphonically andru, literally a thing that is not. In the higher dialect of Tamil, all nouns of quality and relation may be, and very frequently are, converted into appellatives and pluralised by the addition of a — e.g., ariya (Kural), things that are difficult, difficilia. We have some instances in High Tamil of the use of a as the plural suffix even of substantive nouns — e.g., porula, substances, things that are real, realities (from the singular pontl, a thing, a substance) ; also porulana and forulavei — with the addition of ana and avei (for ava), the plural neuters of the demonstrative pronouns. The neuter plural of the third person of the Tamil verb, a form wliich i>; u:cd occayioncilly in ordinary prose as well as in the classical NUMBER — THE NEUTER PLURAL 249 dialect, ends in ana — e.g., irukkindrana, they (neut.) are. ana is un- doubtedly identical with ava (now avei), the neuter plural of the demonstrative pronoun, and is possibly an older form than ava. It is derived from the demonstrative base a, with the addition of a, the neuter plural suffix, and an euphonic consonant {n or v) to prevent hiatus — e.g., a-{n)-a or a-(v)-a. Sometimes in classical Tamil this a, the sign of the neuter plural, is added directly to the temporal suffix of the verb, without the addition of the demonstrative base of the pronoun — e.g., mmda, they (neut.) returned, instead of mmdana. This final a is evidently a sign of the neuter plural, and of that alone. Possibly we should also regard as a sign of the neuter plural the final a of the High Tamil possessive adjectives ena, my (things), mea ; nama, our (things), nostra. The final a of ena would, on this supposi- tion, be not only equivalent to the final a of the Latin mea, but really identical with it. These possessive adjectives are regarded by Tamil grammarians as genitives ; and it will be shown hereafter that a is undoubtedly the most essential sign of the genitive in the Dra vidian languages. The real nature of ena and nama will be discussed when the genitive case-terminations are inquired into. It should be stated, however, under this head, that Tamil grammarians admit that ena and nama, though, as they say, genitives, must be followed by nouns in the neuter plural — e.g., ena keigal, my hands ; and this, so far as it goes, constitutes the principal argument in favour of regarding the final a of these words, not as a genitive, but as the ordinary neuter plural suffix of the high dialect. In Malayalam, the oldest daughter of Tamil, and a faithful pre- server of many old forms, the neuter plurals of the demonstrative pronouns are ava, those (things), and iva, these (things). The exist- ence, therefore, in Tamil and Malayalam of a neuter plural in short a, answering to a neuter singular in d, is clearly established. In addition to ava and iva, avattrugal and ivattrugal are regularly used in Malay- alam, like the double plural aveigal, iveigal, in Tamil. Canarese appears to have originally agreed with Tamil in all the particulars and instances mentioned above ; but the neuter plural in a is now generally hidden in that dialect by the addition of euphonic u, or the addition of avu, they, neuter (corresponding to the Tamil avei), to the base. Thus fita, Tam. other (things), is in Canar- ese heravu. The neuter ^ural of the demonstrative pronoun is not ava, as it is in Malayalam, and as it must have been in primitive Tamil, but avu. Though, however, the nominative is avu, all the oblique cases in the ancient Canarese reject the final u before 250 THE NOUN receiving the case-suffixes, and must have been formed from the base of an older ava — e.g., avara (ava-ra), of those things. The Telugu plural neuters of the demonstratives are avi, those, ivi, these, answering to the singular neuters adi and idi. The oblique forms of the same demonstratives (or rather the bases of those oblique forms), to which the case-terminations are suffixed, are vd remote, and v% proximate (vdti, vUe), which are evidently formed (by that process of displacement peculiar to Telugu) from the primitive bases ava and iva, like vdru, from avaru, and vtru, from ivaru. The neuter plural of the Telugu verb is formed by suffixing avi or vi. Dr Gundert calls my attention here to the natural and easy transi- tion from one vowel to another apparent on comparing the Malay- alam and old Tamil ava with the modern Tamil avei, and finally with the Telugu avi. So also Malayalam and old Tamil ilia, none, is illei in modern Tamil. Final a constantly lapses in the Dravidian languages into a weaker sound. In Gond the singular demonstratives are ad and id ; the corre- sponding plurals av and iv. If Telugu and Gond were the only extant dialects of the Dravidian family, we should naturally conclude that as d is the sign of the neuter singular, so v is the sign of the neuter plural. When the other extant dialects, however (Tamil, Malayalam, and Canarese), are examined, we perceive that this v is not a sign of plurality, nor a sign of anything but of abhorrence of hiatus ; and that it is merely an euphonic link between the preceding and suc- ceeding vowels. Telugu and Gond must therefore yield to the over- powering weight of evidence which is adducible in proof of this point from their sister dialects. Nor is there anything opposed to analogy in the supposition that Telugu has changed the a, which was the sign of the neuter plural of its pronouns and verbs, into i, and then, to represent the idea of plurality, adopted a consonant which was used originally merely to prevent hiatus. In the case of avaru, they, illi, converted into vdru, and ivaru, they, Id, converted into wrw, t), though only euphonic in its origin, has become an initial and apparently a radical ; and the old initial and essentially demonstrative vowels a and i have been thrust into a secondary place. The conversion, therefore, of ava into vd, and of iva into vi (vdti, vUi), the oblique forms of the Telugu plural demonstratives, is directly in accordance with this analogy ; and thus Telugu cannot be considered as opposed to the concurrent testimony of the other dialects, which is to the effect that v is merely euphonic, and that a is the sign of the neuter plural of the demonstrative pronouns. NUMBER — THE NEUTER PLURAL 25 I I remarked it as a curious irregularity, that in Tulu v had become the sign of the neuter singular instead of d — e.g., avu, it. Dr Gundert says that the v is not written. The word is written au-u, and he considers it merely a softened pronunciation of adu, so that there is no irregularity here after all. It is written avu, however, in Brigel's Grammar. If short a be, as it has been shown to be, a sign of the neuter plural inherent in the Dravidian languages, and most used by the oldest dialects, we have now to inquire into the relationship which it apparently sustains to the neuter plural suffix of some of the Indo- European languages. I know of no plural in any of the Scythian tongues with which it can be compared ; and we appear to be obliged to attribute to it, as well as to d, the suffix of the neuter singular, an origin which is allied to that of the corresponding Indo-European forms. In the use of a as a neuter plural suffix, it is evident that the Dravidian family has not imitated, or been influ- enced by, the Sanskrit, and that it was not through the medium of Sanskrit that Indo-European influences made their way into this department of the Dravidian languages ; for the Dravidian neuter plural a differs widely from the Sanskrit neuter plural dni, and it is as certainly unconnected with the masculine-feminine plural as (softened in modern Sanskrit into ah). It is with the short a, which constitutes the neuter plural of Zend, Latin, and Gothic, that the Dravidian neuter plural a appears to be allied. Compare also the Old Persian neuter plural a. It will be evident on recapitulating the various particulars that have been mentioned in this section, that grammatical gender has been more fully and systematically developed in the Dravidian languages than in perhaps any other language, or family of languages, in the world. Properly speaking, there is no such thing as gender in the Scythian languages. Gender appears in the Indo-European languages in the pronouns and pronominals, but not in the verb. In the Semitic languages the verb distinguishes between the mascu- line and feminine in the singular ; but in the plural, as in the verb of the Indo-European languages, gender is ignored. In the Dravidian languages, on the other hand, not only is there a full equipment of sex-denoting pronouns, but there is the same develop- ment of gender in the v^rb also. We have verbal forms — without the necessity of using the separate pronouns as nominatives — for expressing he is, she is, it is, they {persons) are, they {things) are. This is a refinement of expressiveness in which the Dravidian Ian- 252 THE NOUN guages appear to stand alone. Sanskrit is far less highly developed in this particular, so that if there were any borrowing, the Dra vidian family must have been the lender, not the borrower. Probably, however, neither borrowed from the other, but both inherited elements of greater antiquity than either, which the Dravidian family has best preserved, and turned to best account. See Introduction. SECTION II.— FORMATION OF CASES. Princifles of Case-Formation. — The Indo-European and the Scythian families of tongues originally agreed in the principle of expressing the reciprocal relations of nouns by means of postpositions or auxiliary words. The difference between those families with respect to this point consists chiefly in the degree of faithfulness with which they have retained this principle. In the Scythian tongues, postpositions, that is, appended auxiliary words, have generally held fast their individuality and separate existence. In the Indo-European tongues, on the contrary, the old postpositions or suffixes have been welded into combination with the roots to which they were appended, and converted into mere technical case-signs or inflexional terminations ; whilst in the later corruptions to which those languages have been subjected, most of the case-terminations have been abandoned altogether, and pre- positions, as in the Semitic tongues, have generally come to be em- ployed instead of the older case-signs. It cannot reasonably be doubted that the case-terminations of the primitive dialects of the Indo-European family were originally postpositional words, which were added to the root to express relation, and at length blended into an inseparable union with it, through that love of composition by which every member of the family was characterised. In most instances the root and the original signification of those postpositions are now unknown, or they are ascertained with difficulty by means of analogy and comparison. Both in Greek and in Latin we find some postpositions still used in a manner which illustrates the conversion of a portion of this class of words into case-endings — e.g., in Latin nobiscum, and in Greek such words as dypoOi (agrothi), in the country ; aXaBe [halade), to the sea ; and ovpavoO^v (ouranothen), from heaven. The postpositional auxiliary words used in these instances are appended to their bases in a truly primitive manner. If there is any difference between them and the usage of the Scythian postpositions, it consists in CASE-FORMATION 253 this — that in most of the Scythian tongues di {thi), ^e {de), 9cv {then), would be written as separate words. One of the Greek postpositions quoted above, 8e (de), signifying direction to a place, has been supposed to be allied to de, the dative of the Manchu ; and the Greek Oev (then) has been conjectured to be allied to the Tartar ablative din or den. One may well be doubtful whether any such connection can be established ; but in the manner in which the particles are appended to their bases a distinct analogy may be observed. On turning our attention to the Dravidian languages, we find that the principle on which they have proceeded in the formation of cases is distinctively Scythian. All case-relations are expressed by means of postpositions, or postpositional suffixes. Most of the postpositions are, in reality, separate words ; and in all the Dravidian dialects the postpositions retain traces of their original character as auxiliary nouns. Several case-signs, especially in the more cultivated dialects, have lost the faculty of separate existence, and can only be treated now as case-terminations ; but there is no reason to doubt that they were all postpositional nouns originally. The dialect of the Tudas shows its want of literary cultivation in the paucity of its case-signs. There is no difference in it between the nominative, genitive, and accusative. There is another point in which the Scythian principles of case- formation differ materially from the Indo-European. In the Indo- European family the case-endings of the plural differ from those of the singular. It is true, that on comparing the case-terminations of all the members of the family, some traces have been discovered of the existence of an original connection between the singular and the plural terminations of some of the cases ; but in several instances — e.g., in the instrumental case — no such connection between the singular and the plural has been brought to light by any amount of investigation ; and it may be stated as a general rule that the languages of this family appear to have acted from the beginning upon the principle of expressing the case-relations of the singular by one set of forms, and the case-relations of the plural by another set. On the other hand, in all the languages of the Scythian group, the same case-signs are employed both in the singular and in the plural, without alteration, or with only such alterations as euphony is supposed to require. In the singular, the case-postpositions are appended directly to the nominative, which is identical with the base ; in the plural they are appended, not to the nominative or 254 THE NOUN base, but to the particle of pluralisation which has been suffixed to the base. In general, this is the only difference between the singular case-signs and those of the plural. The only exception of importance is, that in some of the Scythian tongues, especially in the languages of the Finnish family, the included vowel of the case-sign differs in the two numbers : it is generally a in the singular and e in the plural — a change which arises from the " law of harmonic sequences," by which those tongues are characterised, and which reappears, but little modified, in Telugu and Tulu. It has already been remarked that in Tulu the a of the singular becomes e in the plural. When the Dra vidian languages are examined, it is found that they differ from those of the Indo-European family, and are, in general, in perfect accordance with the Scythian group, in their use of the same signs of case in the plural as in the singular. The only excep- tions are the truly Scythian one apparent in Tulu, in the change in the case-sign vowel, mentioned above, from a in the singular to e in the plural, and the equally Scythian exception apparent in Telugu, in which the dative case-sign is either ki or ku, according to the nature of the vowel by which it is preceded or influenced ; in consequence of which it is generally ki in the singular and ku in the plural. This identity of the singular and plural case-endings in the languages of the Scythian group, as well as in those of the Dra vidian family, will be found greatly to facilitate the comparison of the case-signs of one language of either of those families with those of the other. Number of Declensions. — There is only one declension, I conceive, properly so-called, in the Dravidian languages, as in the Scythian family generally. Those varieties of inflexional increments which have been called declensions by some scholars, both native and European, especially with reference to Canarese, Tulu, and Telugu, appear to me to con- stitute but one declension ; for there is no difference between one so-called declension and another with respect to the signs of case. Those signs are precisely the same in all : the difference which exists relates solely to suffixes of gender, or to the euphonic and inflexional increments which are added to the bases before the addition of the case-signs. On proceeding to analyse the case-formation of the Dravidian languages, we shall follow the order in which they have been arranged by Dravidian grammarians, which is the same as that of the San- skrit. The imitation of Sanskrit in this particular was certainly an error ; for whilst in Sanskrit there are eight cases only, the number THE NOMINATIVE 255 of cases in Tamil, Teliigu, &c., is almost indefinite. Every post- position annexed to a noun constitutes, properly speaking, a new case ; and therefore the number of such cases depends upon the requirements of the speaker and the different shades of meaning he wishes to express. In particular, the " inflexion " or inflected form of the base, or oblique case, as it is sometimes called, which has some- times a possessive, sometimes a locative, and sometimes an adjectival signification, ought to have had a place of its own. So also the social and conjunctive case. (See the Inflexion and the Instrumental Case.) Notwithstanding this, the usage of Dravidian gramm.arians has restricted the number of cases to eight ; and though there are not a few disadvantages in this arrangement, it will conduce to perspi- cuity to adhere to the ordinary usage in the analysis on which we are about to enter. Tamil grammarians, in following the order of the Sanskrit cases, have also adopted or imitated the Sanskrit mode of denominating them — not by descriptive appellations, as dative or ablative, but by numbers. They have affixed a number to each case in the same order as in Sanskrit — e.g., first case, second case, &c., to eighth case. Though a nominative, or first case, stands at the head of the Dravidian list of cases, the only cases, properly so called, which are used by these languages, are the oblique cases. The Nominative — Absence of Nominative Case-terminations. — In the Scythian languages in which nouns are inflected, as in the Dra- vidian, the nominative is not provided with a case-termination. With regard to Japanese, this is expressed by saying that the noun has no nominative. The Dravidian nominative singular is simply feyar-e, the noun itself — the inflexional base of the noun — without addition or alteration ; but it necessarily includes the formative, if there be one. The nominative plural differs from the nominative singular only by the addition to it of the pluralising particle. There are three apparent exceptions to this rule, or instances in which the nominative might appear to have terminations peculiar to itself, which it is desirable here to inquire into. (1.) The neuter termination am might at first sight be supposed to be a nominative case-sign. In Sanskrit, am is the most common sign of the nominative neuter ; and in Tamil also, all nouns ending in am (in Telugu am-u), whether Sanskrit derivatives or pure Dra- vidian words, are neuter ab|tracts. In Sanskrit the accusative of the neuter is identical with the nominative, but in the other cases am disappears. In Tamil, am is discarded by all the oblique cases of the singular without exception : every case retains it in the plural, 256 THE NOUN but in the singular it is used by the nominative alone. This com- prises the sum total of the reasons for regarding am as a termination of the nominative. On the other hand, though am disappears in Tamil from the oblique cases in the singular, it retains its place in every one of the cases in the plural. The particle of plurality is regularly suffixed to am, and the signs of case are then suffixed to the particle of plurality ; which is a clear proof that, whatever am may be, it is not a mere termination or case-sign of the nominative. The Telugu regards am or am-u as part of the inflexional base, retains it in each case of both numbers alike, and suffixes to it in the singular the case-signs, in the plural the particle of plurality. Ancient Canarese uses am in the nominative and accusative singular of nouns ending in a, and discards it in the plural. In that dialect a tree is 7naram, as in Tamil ; but the plural nominative, trees, is not marangal (maram-gal), but maragal. Modern Canarese appears to make no use of aw whatever, either in the singular or the plural, but it is evident that the final vu of many Canarese nouns is a softened form of m. Compare Tam. mar am, a tree ; Can. maravu. Neuter nouns borrowed from Sanskrit by Tamil ordinarily retain (in the nominative alone, in the singular) the am of the Sanskrit nominative singular : this am is used in every one of the cases in the plural ; so that even in Sanskrit derivatives am is regarded in Tamil, not as a case-sign, but as a portion of the inflexional base. Whatever be the origin of the Tamil am, considered (as I think we must consider it) as a formative, not as a nominative case-sign, it does not appear to have been borrowed from Sanskrit, in which it is used for so different a purpose ; and I believe it springs from a source altogether independent of Sanskrit. We find it added to many of the purest Dravidian roots, and by the addition of it many verbs of that class are converted into nouns. Thus nil-am,, Tam. the ground, is from nil, to stand, dr-am, Tam. depth, is from dr, to be deep. See " Derivative Nouns," in the section on " The Verb." The best explanation of the origin of this am is probably that suggested by Dr Gundert, viz., that it is an obsolete demon- strative pronoun meaning 'it.' I am doubtful whether the Tamil demonstrative adjectives anda, that, inda, this, &c., and the demonstrative adverbs angu, there, &c., have originated in this supposed demonstrative pronoun am, because of the existence of equivalent forms (dndu, indu, &c.), in which the nasal m or n is evidently an euphonic insertion ; and also because the Tuju proximate demonstrative pronoun indu or undu, it, can THE NOMINATIVE 25/ clearly be identified with the unnasalised idu proximate, and udu intermediate, of Tamil and Canarese. (See section on " Euphonic Nomination.") In the case, however, of am, the suffix of so many Dra vidian neuter nouns, the supposition that this was an ancient form of the demonstrative pronoun, regularly formed from the demonstrative root a, that, appears best to suit the use to which it is applied. It cannot indeed be regarded as a perfectly satisfactory explanation of the particle ; for, given a supposed demonstrative am, formed from the demonstrative base a, it cannot fail to be asked, What, then, is the origin of the m of this supposititious am ? Still, without being able to answer this question, we may readily suppose that a demonstrative am, it, was at one time current as an equivalent to ad-u. A parallel instance will then enable us to see how it came to be used as a suffix to nouns. In Tamil poetry adu, it, is frequently appended to neuter nouns as a sort of suffix of emphasis — e.g., we may either say fon, gold, or fonnadu {fon, gold, adu, it). The only difference is that adu is separable from the word to which it is affixed, whereas wherever am was affixed, it seems to have adhered. The oblique cases of the Tamil reflexive pronouns, tan, tdm, are also suffixed to nouns in Tamil poetry instead of the oblique cases of those nouns themselves — e.g., marandanei{k) (instead of marattei) kanden, I saw the tree (accus.). The reflexive seems here to be used in a demonstrative sense. Though we do not now find a neuter de- monstrative pronoun in am or an holding an independent postion of its own in any of the Dravidian languages (as is the case with the neuter demonstrative ad-u), yet we may pretty safely conclude that such a form once existed. An evident trace of this ancient demon- strative am (or an, which would be quite equivalent to it) is found in the existence of the interrogative particles, or rather nouns, Tam. en, en, Tel. emi, what, why. If the interrogative edu, what, leads us necessarily to adu, that, may it not be regarded as almost equally certain that the interrogative em or en, what, points to a demon- strative am or an, that ? Whatever be the origin of the neuter formative am, we must assign the same origin to the an which is sometimes substituted for it. Thus we may say in Tamil either hadam or hadan, debt ; uram or uran, strength. When adu is appended to neuter nouns in Tamil as a separable formative, it can keep its place, if euphony «s supposed to require it, in the oblique cases as well as in the nominative, and to it the case-signs may be affixed. This is also the case with the formative an, and herein it differs in use, if not in origin, from am. Thus kadam in Tamil loses 258 , THE NOUN am in the accusative, takes attu instead, and thus forms its accusative kadattei ; whereas Jcadan retains an, and has kadanei for its accusa- tive. In Malayalam an sometimes alternates with ar as a formative of nouns — e.g., ulan or ular, being, equivalent to the more common ulava ; ulan-dgu, to be born. I find a corroboration of this sup- position of the original identity of am and adu in the use of attu, Tam., ad. Can., and ti, Tel., as inflexional increments or signs of the oblique cases of nouns, all these increments being, as it appears to me, only the different shapes which adu or adi takes in construction. In the inflexion of singular nouns in Tamil, attu, as in the example given above, is regularly used instead of the am of the nominative, from which we may conclude the identity of both am and attu {adu) in signi£cation, and probably in origin, as different forms of the same demonstrative. (2.) In Canarese the crude form of the personal pronouns is occasionally used instead of the nominative — e.g., nd, instead of ndnu, I, and td, instead of tdnu, self ; and hence it might be supposed that the final n or nu of those pronouns constitutes a nominative ter- mination. This supposition, however, is inadmissible ; for in all the oblique cases, without exception, the final n or nu retains its place, and it is to it that the signs of case are added. Consequently it is evident that n is not a sign of the nominative, but a formative, which has been compounded with the inflexional base, or annexed to it, though it is capable of occasional separation from it. (3.) In all the Dravidian languages, the quantity of the included vowels of the personal pronouns in some of the oblique cases (and in Tamil, Malayalam, and Canarese in all the oblique cases) differs from the quantity of the same vowels in the nominative. In the nominative the vowel is invariably long, in the oblique cases generally short — e.g., in Canarese we find 7idnu, I, nanna, my; ninu, thou, ninna, thy ; tdnu, self, tanna, of one's self. This is the only instance in these languages in which there is a difference between the nomina- tive and the oblique cases of such a nature as almost to constitute the nominative a case by itself. In this instance, however, it is uncertain whether the nominative has been lengthened for the sake of emphasis, and we are to seek the true form of the root in the oblique cases, or whether the nominative is the true base, and the shortening of the quantity of the vowel in the obHque cases, prior to the addition of postpositions, has arisen from the euphonic tendencies of the language. Telugu shortens the root-vowel in the accusative only. In Tamil the shortened form, without any inflex- THE INFLEXIONAL BASE 259 ional addition, is often used as a possessive — e.g., nin, thy, from the obsolete ntn, thou — a usage which is in accordance with the ordinary Dravidian rule that the inflected form of every noun, or the basis of the obHque cases, is to be regarded as of itself a possessive or adjective. See " Roots : Internal Changes." Before proceeding to consider the oblique case-signs seriatim, it is necessary to inquire into the changes which the base sustains prior to receiving the suffixes. Inflexion or Inflexional Base of the Oblique Cases. — In a very large number of instances that form of the Dravidian noun which con- stitutes the crude base, and which is used as the nominative, constitutes also the inflexional base. The nominative of this class of nouns and the base of the oblique cases are identical ; and the case-signs are added to the base or nominative without any link of connection, whether inflexional or euphonic, beyond the ordinary v or y, which is inserted to prevent hiatus between concurrent vowels. In a smaller number of instances (a number which constitutes, how- ever, a very large minority), the base or nominative undergoes some alteration before receiving the addition of the terminations, or case- signs, of the oblique cases. In the solitary instance of the personal pronouns, as pointed out under the preceding head, the nominative sustains a curtailment (viz., by the shortening of the quantity of the included vowel) on becoming the inflexional base, or base of the oblique cases : but in all other instances the alteration which the base sustains consists in an augmentation, which is sometimes optional and sometimes neces- sary ; and it is to this augmented form (augmented by the addition of some inflexional increment) that the case-signs are attached. This Dravidian rule may be illustrated by Hebrew. In Hebrew the personal and other suffixes of substantives and verbal nouns are attached, not to the base or nominative, but to the construct state — i.e., the state in which a noun stands when it is qualified by a subse- quent noun. Just so in the Dravidian languages, in that large class of nouns in which the inflexional base of the noun, or its adjectival form, differs from the crude form or nominative, the signs of case are attached, not to the crude, natural form of the noun, but to the altered, inflected form — viz., to that form which a Dravidian noun assumes when it qualifies or*is qualified by a subsequent noun, or when it stands to such noun in the relation of an adjective. This inflected form of the noun is frequently used by itself, without the addition of any case-termination, and when so used it has sometimes 26o THE NOUN a locative, sometimes a possessive or adjectival force. Tamil grammarians hold that the inflexion is not a case-sign, though they cannot but admit that for almost every purpose for which the possessive or locative case-signs are used, the oblique case, or inflected form of the noun, may be used instead. They admit that it is u^ed adjectivally : but it appears to me that its use as an adjectival formative is a secondary one, and that it was originally, like many other adjectival formatives in various languages, a sign of the possessive or locative. Its use eventually as the inflexional basis of all the cases is in perfect harmony with this view of its origin, and testifies to the existence of a period in the history of the language when each of the postpositions of case was known and felt to be a substantive, which required to be united to its base by a sign of localisation or relationship. At present, however, it is our object to seek out and arrange the various increments which are used for forming the inflexional base of the oblique cases, without reference to the other uses to which those increments are put. (1.) The Inflexional increment ' in ' ivith its dialectic varieties. — The particle in constitutes the inflexion of certain classes of nouns in Tamil-Canarese ; and the corresponding Telugu particles are ni and na. All these particles are, I believe, virtually one and the same. Tamil uses in in the singular and in the plural alike ; and its' original signification has been forgotten to such a degree that it is now often used merely as an euphonic link of connection between the base and its case-signs. For this reason its use both in Tamil and in Canarese is optional. In Telugu the corresponding particles are used only in the singular ; and where they are used, their use is not euphonic merely, but is intended to constitute the inflexion. Ku, which in this respect is more nearly allied to Tamil than Telugu is, and more regular, uses ni as the inflexion of the plural as well as of the singular of all classes of nouns. When in is used in Tamil as the inflexion of the neut. sing, demon- stratives aduj that, idu, this, it is apt to be confounded with an, a termination which those pronouns often take, especially in the oblique cases, instead of u. Instead of adu and idu, we may say in Tamil adan and idan. In the nominative these forms are very rarely used ; but the accusative, adan-ei, is more common, and the dative, adarJcu (adan-ku), still more so. id-in-dl, through this, ad-in-dl, through that, and cases similarly formed, must therefore be care- fully distinguished from idan-dl and adan-dl. The an of the latter is a formative, which is probably of the same origin as the am of INFLEXIONAL INCREMENTS 26 1 many neuter nouns (that am being often convertible into an) ; whereas in is an inflexional increment, and was probably a case- sign of the locative originally. The use of in as an inflexional increment effects no alteration in the meaning of the case-sign which is suffixed to it. Where it is not followed by a case-sign, it is generally found to be used as a mode of expressing the genitive ; but where a case-sign follows, it is merely euphonic, and its use is optional. Thus, we may say either Jceiydl {kei-{y)-dl), with the hand, or keiyin-dl {kei-(y)-in-dl) ; either Midi, with the foot, or kdlindl (kdl-in-dl). In the first of these instances {kei-(y)-in-dl), y is used to keep the initial vowel of in pure, in accordance with the ordinary rule of the language ; from the use of which, in this instance, it is evident that in, though merely euphonic in its present application, was in its origin something more than a mere euphonic expletive. in is not only attached as an inflexional increment to the crude base of Tamil nouns, but it is appended also to other inflexional increments, viz., to attu, and to the doubled final d and r of certain classes of nouns. Thus, by the addition of attu to mara-m, a tree, we form niarattu, the inflexional base of the oblique cases, by suffixing to which dl, the sign of the instrumental case, we form marattdl, by a tree ; but we may also attach in to attu, forming attin (att-in), a doubled and euphonised increment — e.g., marattindl {mara-attu-in-dl). As in when standing alone, without the suffix of any case-sign, has acquired the force of the genitive, so also has the double increment, attin — e.g., marattin signifies of a tree. In Tamil, in is the inflexion of all nouns except those which end in am, or in d-u or t-u : in Canarese in is much more rarely used than in Tamil ; but where it is used, its use is rather euphonic and optional than inflexional, and it cannot be used by itself to express the force of the genitive. As in Tamil guruvil, in a priest, and guruvinil are identical, so we may say in Canarese either guruvalli or guruvinalli. In Malayalam the use of in before il, as in the last instance now adduced, is found, Dr Gundert says, only in pedantic poetry. Before the other inflexional increments it is common enough. In Telugu the corresponding particles ni and na constitute the inflexion, or natural genitive of certain classes of nouns, and are also attached as inflexional increments to the base before suffixing the case-signs — e.g., diniki (di-ni-ki), to it, tammuniki {tammu-ni-ki), to a younger brother, guruva-na-ku, to a spiritual teacher. These increments are attached only to the singular in Telugu. They constitute the 262 THB NOUN singular inflexion — i.e., the genitival or adjectival base of the noun ; and though their use is now in many connections optional and merely euphonic, they doubtless contributed at the outset to grammatical expression ; nor are they to be regarded as the inflexion of masculine nouns and pronouns alone, though they are chiefly used by them, for ddniki, to that, dtniki, to this, are neuters. The Telugu ni, and the Tamil-Canarese in, are doubtless identical in origin. The change in the position of the vowel is in accordance with the change of il, Tam. the negative particle, into le in Telugu, and of ul, Tam. within, into 16 in Telugu. It also corresponds to the change of the position of the vowel which is apparent when in, the Latin preposition, is compared with the corresponding Sanskrit preposition ni. It will be seen that in is used not only as an inflexional increment, but as a genitive, an ablative, and a locative. We cannot be in error, therefore, I think, in regarding in in all these instances as one and the same particle, though in different connections it is used for different purposes, nor in concluding that originally it had only one meaning, and was used for only one purpose. A comparison of the various case- signs or increments appears to show that in was originally an equiva- lent form for il, and as il means ' here,' or a house {e.g., ko-v-il, Tam. God's house, a temple), it seems evident that the first use of il in the inflexion of nouns must have been as a sign of the locative. It appears probable therefore that its equivalent in must also have had at first a locative signification. Dr Gundert says, " The oblique cases would all seem to be modified forms of the locative, as expressing something happening in or about the noun, whilst the nominative pronounces its totality." in being used in so many connections and in so general a way, in course of time it came to be regarded in some connections as merely an inflexional increment, that is as an optional suffix to the base, and lastly, as little better than an euphonic expletive, which might be prefixed (its original meaning now having become obscured) to any case-sign, and even to il, its own earliest shape. (2.) The inflexional increments ' ad ' and ' ar.' — The particles ad and ar are extensively used by Canarese as inflexional increments. Their use exactly resembles that of in in the same language, though each is restricted to a particular class of words, in is used as an increment of the base in connection with nouns which end in u — e.g., guru, a priest ; and ad and ar are used in connection with neuter nouns and demonstratives, and with those alone. In the Canarese genitive case-endings, ara, ada, ina, and a, it will be seen that the real and only sign of the genitive is a, the final vowel of each ; and INFLEXIONAL INCREMENTS 263 therefore Dr Stevenson erred in comparing ara or ra (properly ar-a or ad-a) with the New Persian rd, ad and ar are prefixed to the signs of case, not by the genitive only, but by three cases besides — viz., by the accusative, the instrumental, and the locative. Thus we may say not only idara {id-ar-a), of this, and marada {mar-a-da), of a tree, but also idaralli (id-ar-alli), in this, and maradinda [inar-ad- inda), by a tree. Consequently ad and ar, whatever be their origin, do not appear to be signs of case, in so far as their use is concerned, but are used merely as increments of the base, or inflexional bonds of conjunction between the base and the case-signs, like in, ni, &c. Moreover, Canarese differs in its use of these increments from Telugu and Tamil in this, that it never suffixes them alone without the addition of the case-signs, and never gives them the signification of genitives or adjectival formatives. ad and ar are evidently related. Are they also identical ? Both are increments of the neuter alone ; and where Canarese uses ar, Tulu uses t. d and r are known to change places dialectically, as in the southern provinces of the Tamil country, in which adu, it, is pronounced axu ; and the Canarese increment ad is certainly, and ar probably, identical with that very word — viz., with the Tamil- Canarese demonstrative adu or ad, it. Dr Gundert thinks ar derived, not from adu, but from an, the equivalent of aru, I do not feel sure of this ; but it is certain that n changes into r before k — e.g., adatku, Tam. to that — and that n and r are sometimes found to change places — e.g., comp. pir-agu, afterwards, with pin, afterwards. Though Tamil has not regularly adopted the unchanged form of this demonstrative, adu, as an inflexional increment of the base in the declension of nouns, it makes use of it occasionally in a manner which perfectly illustrates the origin of the Canarese use of it. In classical Tamil, as I have already mentioned in discussing the origin of the increment am, the neuter demonstrative may optionally be added to any neuter noun in the singular, not for the purpose of altering the signification, but merely for the improvement of the euphony, and for the purpose of meeting the requirements of prosody, adu may thus be added even to the nominative — e.g., we may not only write pon, gold, but also poetically ponnadu, gold, etymologically gold-that — i.e., that (which is) gold. It is much more common, however, and more in accordance also \j^th the Canarese usage, to use ad-u in the oblique cases ; in which event it is inserted between the base and the case-sign, so as to become virtually (yet without losing its proper character) an inflexional increment — e.g., instead of 264 THE NOUN fonnei, the accusative of fon, gold, we may write ponnadei {ponn-ad-ei) . We may possibly connect with the Canarese ar, and therefore with ad, and ultimately with the neuter demonstrative itself, the euphonic consonant r, which is used in Telugu in certain instances to separate between a noun of quality used as an adjective and the feminine suffix dlu — e.g., sogasu-r-dlu, a handsome woman. This would be quite in accordance with the peculiar Telugu usage of employing the neuter demonstrative singular in place of the feminine singular. I should prefer, however, to regard this r as used simply to prevent hiatus. (3.) The inflexional increment ' ti.' — In Telugu ti or ti is the most common and characteristic inflexional increment of neuter singular nouns, and it is used in Telugu, like the corresponding attu in Tamil, not merely as an increment of the base, but as the inflexion, with the signification of the possessive case or of that of an adjective, as the context may require. Two instances of the use of this increment will suffice out of the very numerous class of neuter nouns which form their singular inflexion by the addition of ti or ti (or rather by the combination of that particle with their last syllable) — e.g., vdkili, a doorway, inflexion vdkiti ; nudum, the forehead, inflexion nuduti. In these instances of the use of ti or ti, the inflexional increment appears to be substituted for the last syllable ; but it is certainly to be considered as an addition to the word — as a particle appended to it ; and the blending of the increment with the base, instead of merely suffixing it, has arisen from the euphonic tendencies of the language. I have no doubt that the suffixed particle which constitutes the Telugu inflexional increment was originally ti, not ti — the dental, not the lingual. This would account for the circumstance that t alone follows words of which the final consonant is r or ^ ; for on the addi- tion of the dental t to r or I both consonants dialectically coalesce and become t ; the hard cerebral being regarded as euphonically equiva- lent to the two soft letters. In no case in Telugu is there a double t in the inflexional increment, tolli, antiquity, forms its inflexion not in tolti or tollinti, as might have been expected, but in tonti. Here, however, it is not the increment that is euphonised, but the final I of the base. Compare the Tamil tondru, antiquity, from the root tol. ti is evidently the equivalent of the Tamil dru, an euphonised form of du. The dental ti is used instead of the cerebral ti, as the inflexion of nouns ending in a pure vowel or in yu after a pure vowel — e.g., vdyu, INFLEXIONAL INCREMENTS 265 the mouth, inflexion vdti ; cM, the hand, inflexion cheti. This cir- cumstance proves that it was the dental ti which was originally used in all cases. The dental t, on being appended to consonants, changes naturally into the lingual ; whereas the lingual rarely, if ever, changes into the dental. If we now conclude, as I think we undoubt- edly may, that the Telugu inflexion was originally ti, not ti, this inflexional increment may at once be connected with the Telugu neuter demonstrative, adi, in the same manner as the Canarese ad and the Tamil attu are connected with the Tamil-Canarese neuter demonstrative adu. Though the identification of the inflexion and the neuter singular demonstrative could not easily be established from Telugu alone, or from any one dialect alone, yet the cumulative argument derived from a comparison of all the dialects has great force. An important link of evidence is furnished by the inflexion which follows. (4.) The inflexional increment ' attu ' or ' attru ' {atvu). — All Tamil nouns which end in am, whether Sanskrit derivatives or pure Tamil roots, reject am in the oblique cases in the singular, and take att-u instead ; and it is to this increment that the various case-signs are suffixed — e.g., the locative case-sign il is not added to dram, depth, but to the inflexional base dr-attu, so that in the depth is not drafn-il, but dr-att-il. This rule admits of no exception in the ordinary dialect of the Tamil ; but in the poetical dialect, which represents more or less distinctly an older condition of the language, attu is sometimes left unused, and the case-sign is added directly to the crude base — e.g., instead of kay-attu-kku, to the depth (from kayam, depth), kaya-kku is used in the Chintamani. When the increment attu is not followed by any sign of case, but by another noun, like the other inflexion in, and like the corresponding Telugu inflexion ti, it has ordinarily the force either of the genitive or of an adjective, sometimes that of a locative, which is perhaps the first use to which it was put — e.g., kul-attu nun may mean as a genitive, the fish of the tank as a locative, the fish in the tank, or as an adjective, tank fish. This inflexion, like ad and ar in Canarese, and ti or ti in Telugu, is used in connection with the singular alone, am, the formative of the base, which is used only by the nominative in the singular, is retained in the plural, not in the nominative only, but in all the oblique cases, ifo it the sign of plurality is appended, and the case-sign follows the sign of plurality — e.g., marangalil [maramgal-il), in trees. There are in Tamil a few naturally plural (neuter) pronominals and 266 THE NOUN nouns of relation [e.g., avei, those (things) ; sila, few ; pala, many ; elld, all ; compare Mai. silava, folava, elldva] which receive in their oblique cases the inflexional increment atru, pronounced attru. Thus, from elldm, all, which is properly elld-v-um or elld-um {um being the conjunctive and intensitive particle ' even,' and elld-um or elldm, signifying even all, all together), the locative which is formed by the Tamil is elldvattrilum {elld-(v)-aUr'-il-um), in all, literally, even in all. So also avei, they (neuter), forms its accusative, not by adding ei, the accusative case-sign, to avei, but by inserting attru, and adding ei thereto — e.g., avattrei (av-attr-ei), them ; in which instance ei (for a), the sign of the plural, is rejected, and its place is supplied by attru, the inflexional increment of this class of plurals. It is evident that the Tamil increments, attu and attru, are virtually identical. The difference in use is slight, and in pronunciation still slighter ; and in general attru is pronounced exactly like attu by the vulgar. We may therefore conclude that they are one and the same, and on examining Telugu we find additional confirmation of their identity. In Telugu, avi, they (neuter), answering to the Tamil avei, forms its inflexion in vdti (for avati). This Telugu (supposititious) avati is evidently identical with the Tamil avattru. The ti of this inflexion is certainly the same as the ti of Telugu nouns substantive : and if there is no difference in Telugu between the ti which forms the inflexional increment of neuter singular nouns and demonstratives and the plural inflexion ti of such words as vdti, we may also conclude that there is no real difference between the singular attu and the plural attru of the Tamil. Whence did the r which is included in anu or attr-u take its rise ? We see its origin, I think, in Canarese ; for in the ancient dialect ar or r forms the inflexional increment of every one of the plural pro- nominals which take avvu in Tamil — e.g., avara (corresponding Tam. avaiTu), of those things ; ellavara (Tam. elldvaitu), of all things ; kelavara (Tam. silavarru), of some (things). The Canarese r is prob- ably, as we have seen, derived from, and originally identical with, d, or t ; and hence Tamil in doubling r gives it the sound ttr. Thus, not only the Tamil increment att-u, but also avt-u, seems to be de- rived from the same origin as the Canarese ad or ar, and the Telugu ti — viz., from the neuter singular demonstrative. Both these inflexions have been formed also by the same process ; for ar, when doubled, becomes atv-u {attr-u), as naturally as ad, when doubled, becomes att-u ; and in each case the doubling arises from the ad- jectival use to which the suffixed pronoun is put. It is a recognised INFLEXIONAL INCKEMENTS 267 rule of Tamil that when a noun ending in d-u is used adjectivally, the d-u may either become d-in or tt-u — e.g., from erud-u, an ox, is formed either erud-in or erutt-u, of an ox. So also ad-u, it, which is now generally inflected by the addition of in, seems to have been inflected formerly as att-u. adu is vulgarly pronounced in the oblique cases as attu by the bulk of the northern Tamilians. The majority of the natives of Madras, for instance, use attei {attu-ei) as the accusative of adu, that, instead of adei ; and in the neuter singular pronominal sufiixes to the verb the same pronunciation is not only commonly heard, but is often written — e.g., instead of iruhhiiadukhu, to its being (the dative of iru-khiv-adu, it is, the being, or that which is), Madras Tamilians write irukkiiattukku ; in which compound attu is evidently used as the neuter demonstrative singular instead of adu. It is also deserving of notice, that the feminine singular sufiix of a large class of appellative nouns, which is di or adi in Telugu, and which has been shown to be identical with the neuter demonstrative, is in Tamil tti or atti. I explain in this way the Tamil neuter singular preterites in ttru, like dyittru {dyivtu), it becomes. This was dyidadu, which was abbreviated into dyiddu^dyittu (compare the correspond- ing change in Canarese), and this was euphonised into dyittru. Two instances will suffice to illustrate the identity of the Tamil attu and the Canarese ad, and thus supply the only link that is wanting to the perfect identification of attu with the Telugu ti, and of both with adu. The Tamil purv-att-il, in ancient times, is compounded of furv-am (Sans, deriv.), antiquity, att-u, the inflexional increment, and il, the sign of the locative. Compare this with the corresponding Canarese purv-ad-alli, in which it is evident that ad is used in the same manner as att-u, and perfectly agrees with it in signification. Again, the Tamil dyirattondru, a thousand and one, is formed from dyiram, a thousand (the inflexion of which is ay ir -attu), and ondru, one. When this is compared with the corresponding Canarese word sdvirad-ondu, from sdvira, a thousand (equivalent to the Tamil dyira) — inflexional form sdvir-ad — to which ondu, identical with ondru, is appended, it is evident that the Canarese increment ad' and the Tamil att' are one and the same ; and also that in this instance the Canarese ad' is used for precisely the same purpose as the Tamil att\ viz., as an inflexional increment with an adjectival signification. (5.) The formation ofthg inflexion by means of doubling and hard- ening the final consonant. — Tamil nouns ending in d-u and r-u form the basis of their oblique cases by doubling the final d and r : and the doubled d becomes by rule tt, and the doubled r, ttr (though spelled 268 THE NOUN rr) — e.g., from kdd-u, a jungle, is formed Mtt-{u)-kku, to a jungle ; from dt-u, a river, dttr-il (dTT-il), in a river. This inflexion, like all others, is supposed by Dr Gundert to have been originally a locative. I am doubtful of the propriety of this theory in this instance, and prefer the following explanation. This doubling of the final consonants of such nouns is to be re- garded, I think, as a sign of the transition of the meaning of the first noun to the succeeding one, just as when intransitive or neuter verbs ending in d-u or t-u acquire by doubling their vowels a transitive signification — e.g., from 6d-u, to run, is formed 6tt-u, to drive ; from ter-u, to become clear, comes tettr-u (tert-u), to clarify, to com- fort. Properly speaking, therefore, this doubling of the final is an adjectival formative, rather than an inflexional or case-sign basis ; but in this, as in many other cases, the same form appears to be used in two different connections, in consequence of the case-sign which is appended to the doubled final having originally been a noun, and still retaining in compounds the force of a noun. In Telugu the final consonant of nouns of this class is hardened, but not doubled, to form the inflexion or basis of the oblique cases — e.g., the inflexion of ev-u, a river, is not etti (em), but eti, of a river; and that of nddu, a country, is ndti, of a country. In some instances Telugu corresponds more closely to Tamil in forming the inflexion of nouns in rr by changing that into r^ — e.g., attu, the neck ; in- flexion of the same atti. If we regarded Telugu alone, we should consider these examples, not as instances of the doubling of a final d or r, but rather as instances of the incorporation of ti, the usual inflexional sufiix, with those finals ; and we should suppose this view to be confirmed by the circumstance that Telugu does not, like Tamil, double the final d-u or v-u of intransitive verbs on converting them into transitives, but adds a formative chu. Nevertheless, the Tamil rule is so clear and express, and so evidently founded upon grammatical reasons, and the Telugu words in question, nddi, &c., so exactly agree with the Tamil, that we cannot but recognise in them the operation of the same principle, though somewhat disguised. In other and parallel instances, though the Telugu hardens, it does not double — e.g., from pdd-u, Tam. and Tel. to sing, Tamil forms pdtt-u, a song, Telugu pdt-a. The final i of such Telugu inflexions as ndti, of a country (from ndd-u), instead of ndt-u, which Tamil would lead us to expect, is owing, I have no doubt, to the influence of ti, which is the ordinary suffix of the inflexion of neuter nouns. (6.) The inflexional increment ' i.' — The inflexion of the plural of INFLEXIONAL INCREMENTS 269 the Telugu epicene demonstrative pronoun consists in i — e.g., vdru (from avaru), those persons ; inflexion vdri, of them, their. The final u of vdr-u is merely euphonic, but the i of vdri is certainly an in- flexional increment ; and possibly the final i of the singular masculine demonstrative inflexional vdni is not to be regarded as a portion of ni, the ordinary inflexional increment of Telugu masculine nouns, but is identical with the final i of vdri. A small class of Telugu nouns form their singular inflexion also in i — e.g., Ml-i, of a foot, ter-i, of a car. What is the origin of this i? 1 think we are guided to a true idea of its origin by comparing it with the possessive pronoun vdridi, Tel. that which is theirs, which in Ku also is evdridi. When vdridi is com pared with the Tamil possessive avaradu, the meaning of which is exactly the same, we see that in each language the termination is that of the neuter demonstrative pronoun, which is adu in Tamil, adi in Telugu ; and we also see that the penultimate i of vdridi is derived by attraction, according to Telugu usage, from the succeeding i, which is that of the neuter demonstrative singular adi. The final i of vdri may therefore be regarded as an abbreviation of adi, or at least as derived from it. (7.) Telugu plural inflexional increment in ' a.' — In Telugu a con- stitutes the plural inflexion of most colloquial pronominals, and of all substantive nouns without exception, l-u, properly I, is the pluralising particle of all neuter nouns in Telugu, and of the majority of rational ones. The inflexion is effected by changing this l-u into la, or to speak more correctly, by suffixing a to I — the final vowel of lu being merely euphonic ; and it is to this incremental a, as to ni and ti, the singular infiexions, that all the case-signs are appended — e.g., kattulu, knives ; inflexion kattula ; instrumental hattula-cheta, by knives. I have no doubt that this inflexional increment a is identical with a, one of the Tamil-Canarese signs of the genitive, of the use of which as a genitive in the singular as well as in the plural, we have an illustration even in Telugu, in the reflexive pronouns tan-a, of self, tam-q, of selves. This increment also, therefore, is to be regarded as a genitive in origin, though in actual use merely an in- flexion ; and I have no doubt that each of the Dra vidian inflexions proceeds from some case-sign. Before leaving this subject, I should briefly refer to one which bears some relation to it, ^iz. : — EupJwnic links of connection between the base and the inflexion, the base and the case-signs, or the inflexion and the case-signs. In Tamil the dative case-sign ku is generally preceded by an 2/0 THE NOUN euphonic w, and through the influence of this u the k is doubled. Thus, from avan, he, is formed not avanku, to him, but avanukku (avan-u-kku). The personal pronouns, both in the singular and in the plural, make use of an euphonic a in this connection, instead of u — e.g., from nan (or rather from a weakened form, en), I, is formed the inflexion en ; and this takes as its dative not enku or enukku, to me, but enakku (en-a-kku). In the higher dialect of Tamil the dative case-sign ku is often directly attached to the noun, especially in those instances in which the noun terminates in a liquid or semi-vowel — e.g., we find in that dialect not avarukku (avar-u-kku), to them, but avarku. In ancient Canarese also, the dative case-sign was invariably attached in this manner. In Malayalam the personal pronouns require the insertion of an euphonic vowel, as in Tamil, between the inflected base and the case-sign. Thus, to thee, is enikk\ inikk\ enakk\ ninakF, or ninakk\ To us, namukku, namakku. Some of these forms are rarely used^ Whenever concurrent vowels meet in Tamil v and y are used, as has already been shown, to prevent hiatus ; and accordingly they are used between the final vowel of nouns and those inflexions or case- signs which begin with vowels — e.g., naduvil {nadu-(v)-il), in the middle ; variyil {vari-(y)-il), in the way. Compare this with the use of V for a similar purpose in Magyar — e.g., from 16, a horse, and at, the sign of the objective case, is formed not loat, but lovat, precisely as would be done in Tamil, v and y are used by Canarese in the same manner as by Tamil ; but in Telugu, as has already been shown, n is used as a preventive of hiatus instead of v. The way has now been prepared for the investigation of the Dra- vidian oblique cases, and of the signs of case properly so called. The Accusative or ' Second ' Case. — In the Indo-European lan- guages the case-sign of the accusative of neuter nouns is identical with that of the nominative case. This identity has arisen, I con- ceive, not from the nominative being used as an accusative, but vice versa from the accusative being used as a nominative. The accusative case-suffix is a sign of passivity, or of being acted upon ; and it appears to have been suffixed to masculine and feminine nouns to denote that in that instance they were to be regarded not as agents, but as objects. Subsequently, I conceive, it was adopted, because of this signification, as a general characteristic of the neuter, ob- jective, or dead class of nouns, and so came to be used as the nominative, or normal case-ending of nouns of that class. In the Dravidian languages, on the other hand, that which was in THE ACCUSATIVE 2/1 its origin a formative termination of abstract neuter nouns, seems to have been adopted as an accusative case-sign. The old Canarese accusative case-sign am seems to be identical with, and is probably derived from, the am which is so largely used as a formative by Dra- vidian neuters. Notwithstanding this, the use of the nominative, or rather of the simple, unformed base, as the accusative of neuter nouns, is the ordinary and almost universal colloquial usage of Tamil- Malayalam, and is often found even in classical compositions. The accusative case-termination may be suffixed whenever it appears to be desirable to do so, either for the sake of euphony or to prevent ambiguity ; but it is rarely employed except when it is required for those purposes. When this case-termination is used without necessity, it sounds stiiT and unidiomatic ; and this is one of the peculiarities by which the Tamil of foreigners is marked. Tamil- Malayalam masculine and feminine nouns and their corresponding pronouns invariably take the accusative case-suffix when they are governed by active verbs. This probably proceeds from the prin- ciple that it is more natural for rational beings to act than to be acted upon ; and hence when they do happen to be acted upon — when the nouns by which they are denoted are to be taken objec- tively — it becomes necessary, in order to avoid misapprehension, to suffix to them the objective case-sign. On the other hand, the difference between the nominative and the accusative of neuter nouns is often allowed to pass unnoticed, because such nouns, whether they act or are acted upon, are alike destitute of personality and inert. Whether the accusative is used as the nominative, as in the Indo-European languages, or whether, as is often the case in the Scythian tongues, the nominative is used for the accusative, the principle involved appears to be one and the same. In Telugu the use of the nominative for the accusative is confined to things without life. In the case of irrational animals, as in that of rational beings, the accusative must be expressed. As far as things without life are concerned, Telugu adheres to the ordinary Dra vidian rules. The dialect of the Tudas uses the nominative for the accusative and genitive in the case of all nouns, except the personal pronouns. The use of the nominative of neuter nouns for the accusative is not un- known to the North Indian vernaculars, and is one of those parti- culars in which those vernaculars appear to have participated in Dravidian or non- Aryan influences. (1.) Accusative case-signs ei, e, and a. — The only sign of the accusative which Tamil recognises is ei, which is suffixed to both 2/2 THE NOUN numbers and to all genders ; though, as has been mentioned, the accusative of neuter nouns is often identical with the nominative or base. Examples, avan-ei, him, aval-ei, her, ad-ei, it. The accusative case-sign of Malayalam is e, which evidently represents the Tamil ei. In ancient Malayalam, Dr Gundert says, a is often used instead. Canarese ordinarily uses either a or annu as its accusative case-sign ; but in some instances {e.g., nanna, me, ninna, thee), a seems to have been converted into na. This a seems to be equivalent to the Malay- alam e and the Tamil ei, into which the Canarese short a is often found to change by rule. The Tamil-Malayalam accusative case-sign e or a may be com- pared with he or e, the dative-accusative of Hindi pronouns ; with the Gujarathi dative-accusative singular e ; and with the pre- ponderance of the vowel e which is observed in the dative-accusatives of the Bengali and Sindhi. Compare also the Brahui dative- accusative ne or e, and the Malay e. On pushing the comparison amongst the fScythian tongues, not a few of their accusative case- signs are found to resemble the Tamil accusative. Thus the Wotiak accusative is formed by adding a to the root — e.g., ton, thou, ton-d, thee. The Turkish accusative is % or yi ; the Mongolian ^ after a consonant : dji, instead of the Turkish yi, after a vowel. The Turkish i is doubtless a softened form of the Oriental Turkish accusative case-sign ni, from which it has been derived by the same process by which the Turkish dative case-signed or yeh is undoubtedly derived from the old Oriental Turkish gd or ghdh. It would therefore appear that the Scythian accusative originally contained a nasal ; and in accordance with this supposition we find in the Calmuck pronouns an accusative case-sign corresponding to the Oriental Turkish n% — e.g., hida-ni, us, from hida, we, and also na-mai, me, and dzi-mdi, thee, from the bases na and dzi. With this we may again compare the Brahui dative-accusative ne or e. ni being evidently the basis of the Turkish and Mongolian sign of the accusative, if the Dra vidian ei or e be allied to it (though this can hardly be regarded as probable), this ei or e must originally have been preceded or fol- lowed by a nasal ; and in investigating the other Dravidian accusa- tive case-signs we shall discover some reasons for surmising this to have been actually the case. (2.) Accusative case-signs am, annu, anna, nu, &c. — am is the char- acteristic sign of the ancient Canarese accusative, and is used in connection with nouns and pronouns alike — e.g., aval-am, her. The more modern form of the Canarese accusative is annu — e.g., aval- THE ACCUSATIVE 2/3 annu, her ; and this annu is evidently identical with the older am. am has in other instances besides this evinced a tendency to change into an ; for ' he ' is avam in ancient Canarese, though avan in Tamil. The change of the old Indo-European m, the sign of the accusative in Latin and Sanskrit, into the Greek v {n) is also a parallel case. The ancient Canarese case-sign am no sooner changed into an, than it would irresistibly be impelled to euphonise an by the addition of nu Even in Tamil, ma%i, earth, is commonly pronounced mannu, and the corresponding Telugu word is mannu by rule. Hence we seem to be quite safe in deriving annu directly from an, and an from am. Another form of the Canarese accusative case-sign is anna, instead of annu, or simply nna or na — e.g., na-nna, me. The final u has in this instance been changed into a, through the attractive force of the primitive an ; or perhaps the entire euphonic appendage nu has been rejected, and the original case-sign an been softened to a, whilst the final n of the base has been doubled to augment or express the objectivity of the signification. The Tulu accusative case-sign is nu or n\ which is evidently iden- tical with the case-signs of the Telugu and Canarese. Compare the various accusatives of ' this ' — old Can. idam, modern Can. idannu ; Tulu unden ; Tel. dini. Probably the whole of these case-signs are altered forms of the old Can. am ; and this particle, as has already been suggested, under the head of the nominative, appears to have been originally a singular neuter demonstrative pronoun. When the Gond accusative differs from the dative it is denoted by un. In Telugu the neuter accusative is often the same as the nomina- tive, as in the other Dravidian dialects ; but when the noun denotes animals, or things possessed of life, whether rational or irrational, the accusative must be expressed by the addition of a sign of case. The accusative case-sign may optionally be suffixed, as in Tamil, to nouns denoting things without life ; but whether the noun denote a thing without life, or a being possessed of life, whether it be singular or plural, the sign of case must be suffixed to the inflexion, genitive, or oblique case basis, not to the nominative. When the inflexion is the same as the nominative, the noun to which the case- sign is attached is still regarded as the inflexion, so that in theory the rule admits of no exceptions. The sign of the accusative in Telugu is nu or ni. When precede by i it is ni — e.g., mti-ni, dom-um ; where it is preceded by any other vowel it is nu — e.g., hidda-nu, fuer-um. A similar ni or na is used in Telugu (but not so systemati- cally as the corresponding in in Tamil) as an euphonic inflexional 274 THE NOUN increment ; and na or ni is also a sign of the locative in Telugu. Probably those locative and genitive suffixes were originally, and are still to be regarded as one and the same ; but the sign of the accusa- tive, though nearly identical in sound, proceeds apparently from a different source. Comparing it with the Canarese, and especially with the Tulu, accusative nu or n\ we can scarcely avoid the con- clusion that, though in sound it is identical with the ordinary inflexional augment, it is to be regarded as a relic of the Canarese accusative case-sign annu or mn. The suffixes of the accusative of the Telugu personal pronouns can be explained on this supposition alone. The inflexions of those pronouns are essentially different from their accusatives, and incapable of being confounded with them ; and the accusatives of those pronouns take of necessity, and not merely for euphony, the nasal suffixes nu or nnu in the singular, and mu or mmu in the plural. Thus, whilst nd, of me, is the inflexion of nenu, I, its accusative is nanu or nannu, me ; the accusative of the second person is ninu or ninnu, thee, and their plurals are mamu or mammu, us, mimu or mimmu, you, whilst the inflexions of those plurals are md and mt. When these accusatives are compared with the Canarese and Tulu, especially with yanan' me, and ninan\ thee, in the latter, their virtual identity, and therefore the origin of them all from the ancient Canarese am, can scarcely be doubted. We may now proceed to compare this accusative case-sign am, an, annu, nu, or na, with the Gujarathi dative-accusative ne, with the Panjabi nu or num, and also with the Brahui ne or e, and the Turkish and Mongolian ni or i. In the Finnish tongues the greater number of singular accusatives are formed by suffixing en, an, &c., which are also used as signs of the genitive : in the plural there is rarely any difference between the nominative and the accusative. Ascending further towards the source of the Scythian tongues, we find in the language of the Scythian tablets at Behistun an unquestionable link of connection with the Dra vidian. The pronoun of the second person singular in that language is nt, thou, of which nin is the accusative ; and when this is compared with the Tulu nin-an\ thee, we cannot fail to be struck with the closeness of the resemblance. We should also notice the extensive use of m or n as an accusative case-sign in the languages of the Indo-European family. In Sanskrit, Latin, and Gothic, m predominates, in Greek n ; but these consonants are virtually identical, like the m of the ancient Canarese and the n of the modern. A similar form of the accusative being extensively prevalent, as we have seen, in the Scythian tongues, it would be THE INSTRUMENTAL 275 unreasonable to derive tlie Dravidian case-sign from the Indo- European. In this instance it would be safer to conclude that both families have retained a relic of their original oneness. If, as appears highly probable, the old Dravidian accusative in am is identical in origin with the am which is used as a sort of nominative neuter, or rather neuter formative, and if this aw was originally a demonstrative pronoun, formed from the demonstrative base a, we seem to find in the Dravidian languages, not only a relic of their original relationship with other families of tongues now widely divergent, but an index to the original meaning of the neuter accusative case-sign m or n, wherever found, and an explanation of the identity of the singular neuter accusative case-sign in so many Indo-European languages with the singular nominative case-sign am. Being a formative of neuter nouns, a class of nouns which more commonly denote things that are acted upon than things that act, it would naturally come to be used as an accusative case-sign — that is, as a sign of objectivity. It only remains to inquire whether the Tamil-Malayalam accusa- tive case-sign ei, e, or a, cannot be connected with the Canarese am, annu, and na. On comparing the ancient Canarese accusative ninnam, thee, with the more modern ninna^ it can scarcely be doubted that the latter is derived from the former by the ordinary process of the softening away of the final nasal. Through this very process the final am of many substantive nouns has been softened to a — e.g., maram, ancient Can. a tree, Tnara or mara-vu, modern Can. If, then, the sign of the accusative in ninna, thee, is not na, but a (instead of am), as is probably the case, there cannot be any difficulty in deriving from it the Tamil accusative case-sign ei, for the change of a into ei takes place so frequently that it may almost be considered as a dialectic one — e.g., compare old Tamil ila, not, with the modern Tamil illei. (1.) The Instrumental or ^ Third' Case, properly so called. — Different particles are used by different Dravidian dialects as suffixes of the instrumental case. In Telugu the most classical instrumental is identical with the inflexional locative, and consists in changing ti or ti, the inflexion, into ta or ta — e.g., rd-ta, with a stone, from rd-yi, a stone, the inflexion of which is rd-ti. This form of the instrumental was probably a locative in its original signification, and at all events it is identical with an old form of the locative — e.g., inta, in a house, from illu, a house, of which the inflexion is inti. The more commonly used instrumental of Telugu is formed by the addition to the inflexion 2/6 THE NOUN of any noun of che or chela, which is itself the instrumental form of ehe-yi, the hand, signifying by the hand (of) — e.g., nifpU'cMta, by fire, literally by the hand of fire. The inflexion, or genitive, without the addition of any special suffix, is also occasionally used in Telugu, as in High Tamil, to denote the instrumental case, as well as the ablative of motion, and the locative. The particle na is also sometimes suffixed to neuter nouns to denote all three ablatives. The old Canarese instrumental suffix hn is evidently identical in origin with in, the suffix of the Tamil ablative of motion, originally a locative. It has already been seen how easily m changes into n : and both in Canarese and in Tamil there is so close a connection between the ablative of motion and the instrumental, that the case-sign of the one is very often used for the other, especially by the poets — e.g., vdl-in di/a vadu, Tam. a wound inflicted by a sword, not from a sword. In Canarese also the ablative of motion is denoted more frequently by the suffix of the instrumental than by its own suffix. Through a similar tendency to confound these cases, the case-sign of the instru- mental has disappeared from Latin, Greek, &c., and the sign of the ablative has come to be used instead. Even in English, by, originally a locative (e.g., close by), is used at present to form the ablative, or more properly the instrumental. The instrumental case-sign in modern Canarese is inda, evidently an euphonised form of an, as are also the old Canarese suffixes indam. and inde. The instrumental suffix of the Tuda is edd. Dr Pope connects this with erd, past tense of er, to be ; but as he states that end is sometimes used instead of edd, I should prefer to consider edd derived from end by the same process by which ondu, one, in the other dialects, has become odd in Tuda, and end, identical with the Canarese inda, used by the Tudas' Badaga neighbours. The instru- mental case-sign of the Tulu is (Tdu, which Dr Gundert derives from a locative noun ede = idei, Tam. a place, to which the oblique case- sign or inflexion du, answering to the Canarese da, is added. I sus- pect the Tulu d'du has the same connection with the Canarese inda as the Tuda edd appears to have. In Tamil and Malayalam the suffix of t"he instrumental is dl ; in High Tamil an also, dl is the case-sign of the ablative or instrumental in Gond, though in Telugu, which is spoken between the Tamil country and the country of the Gonds, a different case-sign is used. This suffix dl may possibly be derived from, or allied to, Ml, Tam. a channel. In some dialects channel is a compound word (Tam. kdl- vdy ; Tel. kdlava ; Can. hdlive), and the only meaning of Ml is a foot. THE INSTKUMENTAL 277 This meaning is contained in Tamil, but that of a channel, which Tamil contains also, suits better the supposed use which is made of Ml, as a sign of the instrumental case. Ml may have lost its initial h in the same manner as Ml or gal, the neuter sign of plurality, is known to have done in Telugu and Tulu, in which it has become l-u, by corruption from Ml-u or gal-u. Compare also the corruption of avargal to dl in the colloquial Tamil aval, they. Here both g and r have disappeared. Compare also the disappearance of k from the Canarese Mmmdranu instead of MrmaMranu. In the Indo-European family of languages there are no signs of the instrumental case which at all resemble those that we have noticed in the Dravidian family. The only analogies which I have noticed (and probably they are illusory) are those which exist between the case-sign of the Tamil-Malayalam and the corresponding case-signs of the Finnish tongues. Compare dl with the instrumental suffix of the Magyar, which is al in the singular, el in the plural ; and with alia, ella, &c., the instrumental suffixes of the Finnish proper, and which are euphonically augmented forms of al and el. A secondary or periphrastic mode of forming the instrumental case, which obtains in the Dravidian languages, as also in the northern vernaculars, is by means of the preterite verbal participle of the verb to take, and the accusative or abstract nominative of any noun — e.g., Mttiyei (h)Mndu, Tam., with a knife, literally having taken a knife : compare the corresponding Bengali cliuri diyd, with (i.e., having taken) a knife. Various participles besides Mndu are used instead of the instrumental in Tamil and Malayalam, as knowing, doing, seeing, considering, putting, saying, &c. ; but Mndu, taking, is the one most commonly used. This has arisen from the repugnance of the Dravidian (as of the Scythian) languages to continue to make use of any inflexional form after it has ceased to express its original meaning, and has become a mere technical sign. When that has taken place, as in the instance of the Tamil dl, those languages are often found to abandon the old form, or let it fall gradually into disuse, and to adopt some word or phrase instead which has a distinct meaning of its own, and the use of which recommends itself at once to the intelligence of the speaker. (2.) The Conjunctive or Social Case. — Dravidian grammarians have arranged the case system of their nouns in the Sanskrit order, and in doing so have done violence to the genius of their own grammar. The Dravidian ablative of motion and the locative are evidently one and the same case, though represented as different by grammarians, in 2/8 THE NOUN deference to Sanskrit precedents ; and the Dra vidian social ablative, as some have called it, or rather, as it should be termed, the con- junctive case, though it takes an important position in the Dra vidian languages, has been omitted in each dialect from the list of cases, or added on to the instrumental case, simply because Sanskrit knows nothing of it as separate from the instrumental. The conjunctive, or social, stands in greater need of a place of its own in the list of cases in these languages than in Sanskrit, seeing that in these it has several case-signs of its own, Avhilst in Sanskrit it has none. The instrumental is best rendered in English by the preposition by, by means of ; the force of the conjunctive is that of the preposition ' with,' in the sense of the Latin cum, or together with. Sometimes the English preposition ' with ' is used in either sense — e.g., I cut it ivith a knife, I went ivith him ; but in the Dravidian languages the former ' with ' would be represented by the sign of the instrumental case, the latter by that of the conjunctive — e.g., katti-{y)-dl, Tam. by a knife, avan-odu, with him. Though Sanskrit and the Indo- European languages generally are destitute of this case, Latin evinces a tendency towards it in such forms as nohiscum. Whilst most of the Scythian tongues have a regularly formed conjunctive case equally with the Dravidian ; and den, the conjunctive case-sign of Calmuck, may even be compared (though doubtless the resemblance is accidental) with the Tamil conjunctive case-sign udan. The Tamil and Malay alam conjunctive case-signs are odu and odu (when emphasised, ode) ; also udan. odu is evidently a lengthened form, probably a verbal noun, from odu ; and the root meaning of odu, as is apparent from its derivative o^tu, adhesion, is to touch, or rather to touch so as to adhere. The particle odu, or odu, thus de- notes the closest kind of junction, and is appropriately used as the sign of the conjunctive case. Udan or udane, the other sign of the case in Tamil, is pronounced odan ; and in the Canarese odane, the initial o is written as well as heard. The final an being one of the ordinary formative particles of Tamil nouns, it appears probable that the root is od\' and if so, udan and odane are identical in origin, as in use, with odu and odu. Udan is still used poetically as a noun signifying conjunction, and commonly as an adjective with the meaning of joint — e.g., udan-pangdli, Tam. a joint sharer ; as an adverb, udane means immediately. The Tamil verb todu, to touch, with its derivative todar, to follow, seems to me to be closely allied to odu, to adhere to. The Telugu conjunctive case-sign is todu, of which to is an abbre- THE DATIVE 2/9 viated form. This todu appears to resemble the Tamil odu, and the Tel. adverb todanu, todene, at once ; it still more closely resembles the Tam.-Can odane. The resemblance, however, does not amount to identity ; for if the Telugu words into which todu enters in various shapes are compared, it will be found that the Tel. todu is identical, not with the Tamil odu, but with tora (as in toramei, companionship), the radical form of which is doubtless tor-u, a verb, of which the original meaning, probably ' to be together with,' survives in Tamil only in the verbal nouns torudi, a collection, and toru, a cow-stall. I quite agree with Dr Gundert in thinking that odu and toru cannot be identified ; but I still think them allied, through their common point todu. The Tamil odu and the Tel. todu (the lengthened forms of odu and todu = toru) are certainly not identical, and yet it is difficult to suppose the resemblance between them altogether accidental. I admit, however, that different postpositions for the different signs of case may be freely selected for use in the various dialects, just as Tamil and Malayalam use il, here, house, as the sign of the locative, whilst Tel. prefers 16 ^ul, within. Tulu has a case, which Brigel, in his '' Tulu Grammar," calls the communicative, which is used with some of the meanings of a dative, but which on the whole seems to have more of the force of a conjunctive. The case-sign is da or ta, and this particle seems naturally to connect itself, both in sound and signification, with odu, the Tam.-Mal. sign of the conjunctive. The Dative or ' Fourth ' Case. — In the North Indian dialects one and the same postposition or suffix is used more or less regularly as a sign of case both by the dative and by the accusative. In the Dra- vidian languages, with the exception of the Gond, not only is the difference between the dative and the accusative essential and strongly marked, but there is less discrepancy amongst the various Dra vidian dialects with respect to the particular suffix used to denote the dative, than with respect to any other case-sign. The accusa- tives, instrumentals, ablatives, and genitives, of the various dialects, exhibit material differences ; but in all the dialects of this family — in the rudest as well as in the most polished — there is but one suffix of the dative. The dative is formed in Tamil by suffixing ku (in construction Mu) ; in Malayalam kku ; in Telugu hu or hi, according to the nature of the preceding vowel — i.e., ki after a word ending in i, ku in all other con- nections ; in old Canarese ge or ke ; in the modern dialect ge or kke, and in construction ige ; Tulu, ku, gu, F, g' ; Tuda, k or g, generally 28o THE NOUN the latter. From a comparison of these forms it is obvious that the guttural k or g (generally followed by a vowel) constitutes the most essential part of this suffix ; and that, as the vowel seems to have been added chiefly for the purpose of helping the enunciation, it is of little moment what vowel in particular appears to be used for this purpose. In the primitive Indo-European tongues we discover no trace of any such dative suffix or case-sign as the Dravidian hu ; but ho, the dative-accusative of the Hindi (in Bengali ke, in Sindhi khe), resembles the Dravidian ku so much that it seemed to me highly probable that some relationship existed between them. Two recent writers, however, seem to have proved that the Gaurian ko has been derived from Sanskrit ; and if this be the case, its relationship to the Dravidian ku cannot be maintained. Dr Trumpp, in his " Sindhi Grammar,'' derives the Sindhi khe and the Bengali ke from the Sanskrit locative kr^te, for the sake of, in regard to. This form be- came in Prakrit first kite, then kie. It was then contracted into ke, which in Sindhi, by reason of the elided r, became khe. He derives the Hindi and Hindustani form of this postposition ko by a similar process from the Sanskrit kr'tam, which is used adverbially with the same signification as the locative krHe. In Prakrit, and still more in the modern dialects, the neuter is changed into the masculine. In accordance with this rule, we have first kito, then kio, and then the more modern contracted form ko. He thinks kom and kaum formed from ko by the addition of an euphonic anusvdra, to which the modern tongues have taken a great fancy. Dr Trumpp argues also that the fact that the Arian vernaculars, which border immediately on the Dravidian idioms, have not adopted the use of ko as a sign of the dative, shows that it is improbable that the dialects more to the north have been indebted for this form to the Dravidian idioms. Mr Beames, in his " Comparative Grammar of the Modern Aryan Languages of India " (Introduction, p. 48), attributes to the ko of the Hindi, &c., a different origin ; but the origin he assigns to it is as distinctively Sanskritic, and equally far removed from relationship to the Dravidian case-sign. He says, " It is demonstrable from actual written documents that the modern Hindu ko is a pure accusative or objective, and w^as in old Hindu kaun, which is the usual and regular form of the Sanskrit kcwi, the accusative of nouns in kah ; so that there does not appear to be the slightest reason for connecting it with anything but the cognate forms in its own group of languages." Though the derivation of the Hindi ko from the Dravidian ku cannot THE DATIVE 28 1 now be maintained, it does not follow that the Dra vidian form must be supposed to be derived from the Hindi one. The Dravidian hu, being found in every dialect of the family, however cultivated or however rude, has an antiquity of its own, greatly surpassing that of the change of kaun into ko in Hindi. Probably none of the written documents referred to by Mr Beames can pretend to an antiquity equalling that of the Syrian Christian inscription, in the Tamil of the period, on the Malabar coast, which has been ascertained to have been written in 774 a.d., and in which we find ku used as a dative (e.g., nagarattuKKV , to the city) precisely as it would be at the present day. All that can be said is that this resemblance of ko to ku is one of those cases of remarkably close resemblance which do not amount to, but which might readily be mistaken for relationship. The Singhalese dative is ghai ; in the Oraon, a Dravidian dialect strongly tinctured with Kolarian elements, it is gai ; in Tibetan gya ; in the language of the Bodos, a Bhutan hill tribe, it is kho, nearly identical with the Hindi. The suffix of the dative in the various languages of the Turkish family seems closely to correspond to the Dravidian dative. The forms of this suffix found in the Oriental Turkish are ke, ka, ge, ga, ghah, and also a. The Osmanli Turkish dative is yeh or eh, the initial k or g of the older dialect having been softened into y, and then discarded. The Manchu de and the Mongolian dou are possibly allied to the Tartar ke ; for it has already been remarked that the change of k into t or d, or vice versa, is not an uncommon one in this group of tongues, and that even amongst sister dialects belonging to the same family or sub -genus, the pluralising particle in one dialect is ek, and in another et. In the Finnish family of languages the Turko-Dra vidian dative reappears ; though the Finnish proper has le, not ke. In the Irtish and Surgutish dialects of the Ostiak the suffix of the dative is ga, corresponding to the Oriental Turkish ga or ge. The ordinary Ostiak has also a, softened, as in the Oriental Turkish itself, from ga. Compare also the Mordvin adessive suffix va or ga. The Cheremiss illative, which denotes motion into a place, is ska, ske, &c., but in adverbs and certain postpositions this is replaced by ke, ka, &c., signifying direction. The origin of this particle is considered identical with that of the particle ke or ge, which is used to form a social ablative. The syllable ka or ki is also a pai* of the case-sign of the ordinary ablative and the superlative. The Japanese sign of direction is ve, he, ye, e — e.g., Yedo-ve, Yedo- wards. Interesting and remarkable analogies have been brought to light 282 THE NOUN by the Scythian tablets of Behistun. We learn from those tablets that a dative suffix which is almost identical with the Dravidian, and also with the Turkish and Ostiak, was used by the oldest Scythian dialect of Central Asia of which any remains are extant. The dative case-sign or suffix which is most largely used in the Scythic tablets is ikki or ikka. Mr Norris noticed the resemblance of this suffix to the Magyar genitive-dative nek and the Telugu genitive postposition yokka ; but its resemblance to the dative suffix of the Telugu and of the other Dravidian dialects is still closer. The Tamil A^m becomes, as we have seen, akku or ukku in construction ; the Canarese ge becomes ige ; and the Malayalam ukku or kku becomes ikku ; which last form of the suffix is identical with the Scythian of Behistun. Compare, e.g., the cuneiform Scythian ni-ikka or ni-ikki, to thee, with the corresponding Malayalam nina-kku, the Telugu ni-ku, and the Tulu ni-kk\ It has thus been shown that the principal languages of the Scythian family accord very exactly with the Dravidian languages in the use of ka, ki, ku, or some related particle, as the suffix of the dative. It may be noticed also, that in the language of the Malays there is a prefix, ^a, which signifies 'towards.' 'To a place,' however, in Malay, is datan. It is difficult to determine whether the Finnish dative suffix le has any connection with ke. It certainly seems to resemble much more closely the Tibetan, Pushtu, and Marathi dative suffix Id — which Id is evidently equivalent to the New Persian ra. Compare, e.g., the Marathi tu-la, to thee, thee, with the corre- sponding Persian to-ra. Malayalam alone of all the Dravidian dialects appears to possess two suffixes of the dative, viz., kku, which is the suffix most largely used, and innu, nu, or u, which is occasionally used in the dative singular only. This innu is evidently a compound form, and seems to be euphonised and softened from in-ku. Tamil is fond of adding to the base of nouns which are to be declined the euphonic increment in (originally a locative), before suffixing the signs of case. The same practice prevails in Malayalam also. Consequently, this exceptional Malayalam dative is not inna, but nu, or simply u ; and the doubled n which sometimes precedes it (e.g., avannu, to him) may only be an euphonic compensation for the loss of the k. The k or g of ka or ga has been softened away in some dialects of the Turkish and Ostiak, precisely as I suppose it has been in Malayalam, Dr Gundert prefers to derive this peculiar dative case-sign innu from the possessive case-sign inadu. The Malayalam endre, my, is, I THE ABLATIVE OF MOTION 283 doubt not, to be resolved into enadu, and therefore trmrattindre, of a tree, into rnarattinadu. This inarattindre again may have been soft- ened into rnarattinu, just as the Mai. ninnu, standing, innu, to-day, are softened from the Tam. nindru and indru. Dr Gundert supposes, therefore, that in this form of the dative we have a relic of the possessive. He is doubtful, however, himself of the validity of this explanation, as nu is as common in old Malayalam as nnu — e.g., avanu, to him, as well as avannu. Here he thinks it most probable that the ku has been simply dropped. If the expression " softened away " were used instead of dropped, this explanation would be equivalent to mine, which is that innu, being a dative, is more likely to be a softened form of inku, which is in itself a true dative, than that it should be a softened form of inadu, which is in itself a possessive. Can a purely Dravidian origin be discovered for the Dravidian dative case-suffix ku? The locative suffixes il and in can be explained ab intra ; but I doubt whether ku is capable of an ah intra explana- tion. The only suggestion I can offer is as follows : — Looking at such nouns of direction as vadakku, north, and kirakku, east, we find the final ku, though a dative or directive in signification, indistinguishable in form from the ku which is one of the commonest formatives of verbal nouns, and from the ku, possibly the same ku, which is a sign of futurity in the oldest form of the Tamil verb. Can it be that in all three connections the ku is the same, and that the root idea in each case was transition ? This does not explain how ku came to mean transition ; but it may indicate the direction in which inquiry may be made. The Ablative of Motion or 'Fifth ' Case. — This case appears to have been included in the list of cases by Dravidian grammarians out of deference to the grammatical principles of the Sanskrit. It is true that if we look at the construction and meaning of a Dravidian sen- tence, the signification of an ablative of motion will be found to exist, and it will be found to be expressed much more clearly even than in Sanskrit ; but a distinction is to be drawn between the existence . of a case and the existence of a case-sign, or regular technical suftix of case. The Dravidian languages have undoubtedly an ablative of motion, and a great many other ablatives besides ; but I doubt whether they have any case-suffix which belongs exclusively to the ablative of motion. On comparing the sujfixes of the ablatives of motion (which are also used sometimes in an instrumental sense) with those of the locatives 284 THE NOUN in the various dialects of this family, no real difference is apparent between the one class and the other, or at least no adequate reason appears for regarding them as distinct and independent suffixes ; for whatever difference does exist is to be attributed, not to the signs of case, but to the verbs or verbal participles which are annexed to them. The object of the ablative of motion is to furnish an answer to the question whence ? and this answer is obtained in the Dra vidian tongues by suffixing to a noun of place the sign of the locative, and annexing to that sign a verb of motion. By this means the locative is converted into what is called the ablative, without changing its case-suffixes, and the idea of a change of place is thus naturally and necessarily educed. Native Tamil grammarians appear to hold that il, the ordinary suffix of the ablative, and il, the most largely used sign of the locative in the colloquial dialect, though written and pronounced alike, are different particles with different signifi- cations. I am persuaded, however, that this view is erroneous ; and that a natural system of case classification would determine that the Dravidian languages have no ablative, properly so called, but only a variety of locative and instrumental suffixes, which are capable of becoming ablatives by the addition of appropriate verbs. In Tamil, the suffixes which are used in forming the ' fifth ' case, or ablative of motion, are il and in. il (Tel. illu) signifies by itself a house, a place — e.g., k6-(v)-il, a temple, God's house ; its primitive meaning, however, appears to have been 'here,' 'in this place'; and it is therefore well suited for becoming a sign of the locative. Accordingly it has a place in the list of locative suffixes, as well as in those of the ablative ; and in the colloquial dialect it is used as a sign of the locative far more frequently than any other particle. The other suffix, in, is identical, I conceive, with i^n, the old Canarese sign of the instrumental : it is used as an instrumental in Tamil also ; but probably both im and in were previously locative suffixes. In old Canarese the proper suffix of the ablative is attanim (other forms of which are attanindam and attaninde), which is itself formed from the demonstrative adverb attana (identical with attal-u or atta, there, or attal, that side), by the addition of im, the old instrumental suffix, meaning originally ' here,' from which inda, the more modern suffix, is derived ; and this inda, though the ordinary sign of the instrumental, is also ordinarily used, with the addition of a verb of motion, as the sign of the ablative. Whilst I think that not only il, but also in and ifii were originally locative suffixes, it is more difficult to determine whether il and in THE ABLATIVE OF MOTION 285 were originally identical in sound and signification, a^^»vell as in application. In every instance in which il is used in Tamil, in may be substituted for it poetically ; and it is almost exclusively by the poets that in is used. Moreover, in Telugu, illu, a house, identical with il, is euphonised into in, in the inflexion inti, of a house. On the other hand, if we regard in as originally a locative, it will be found to have a far wider range of analogies than il, and may therefore be surmised to have sprung from a different root. In Finnish and Magyar we find an, en, and still more frequently in, used as signs of the locative. Even in Sanskrit we find in used as a locative case-sign of pronouns of the third person — e.g., tasmin, in him ; and though this in is supposed to have been euphonised from i, yet in the Latin locative preposition in and the Greek iv (en), corre- sponding to the Sanskrit ni, we find the existence of a remarkable analogy, il, on the other hand, has no apparent affinities out of the pale of the Dravidian family. It seems probable that in, one of the signs of the locative in Tamil, is identical with in, a sign of the genitive, or inflexional increment, in Tamil-Canarese ; and if so, a new and very wide range of affinities is disclosed, as will be seen when the case-signs of the genitive are inquired into. The Tamil il and in agree in this, that when they are used as sufiixes of the ablative, they both require to be followed by verbs of motion. In the spoken dialect of the Tamil, the verb of motion is preceded by the verbal participles nindru, standing, or irundu, being. The use of these participles strengthens the supposition that il and in are properly to be regarded as locatives. In the higher dialect, how- ever, they are ordinarily dispensed with, and il or in is followed by a verb of motion alone — e.g., malei-{y) in virum aruvi, the cataract which falls from the mountain. In this expression the idea of " motion from a place " is plainly implied in the aoristic relative participle virum, which falls ; and hence in, whatever it may have been in origin, acquires the force of a sign of the ablative of motion. In Canarese the compound ablative suffixes attanim and deseyinda are not so commonly used as inda, the terminal member of the second compound suffix ; and though inda is described to be the sign of the instrumental, I have no doubt that it is identical with *m and in, and a locative in origin. «(rhe first member of the Can. compound dese means a point of the compass (Sans, dis, Tam. tisei). inda is not only used by itself to form the ablative, but is also allied to alii or illi, the sign of the locative, for the purpose of denoting the 286 THE NOUN ablative. Compare the Canarese allinda or illinda, from, with the corresponding Tamil compound il-irundu or il-nindru. In Telugu the particle na, which corresponds to the Tamil in and the old Canarese im, is more distinctively a locative than an ablative of motion. This particle is ni after i ; and if this is its normal form it may at once be identified with the Tamil in. The Telugu ablative of motion is ordinarily formed by means of the verbal participle nundi or nunchi alone, without the aid of any such suffix as na, or ni, il or in ; consequently this ablative seems to have still less of the character of an independent case than in Tamil. On further exam- ination, however, it comes into accordance with the Tamil ablative. nundi or nunchi is regarded by Mr Clay, and I think correctly, as formed from undi, having been, the past participle of undu, to be, to which is prefixed the n of the locative case-sign, the full form of which is na or ni. Thus faralokamu-nundi vachchenu, he came from heaven, should be divided paral6kamu-n' -undi vachchenu ; literally, " he, having been in heaven, came." unchi is not found in the classics in this connection, and being the past participle of a transitive verb derived from the same root (meaning to place), its use as the suffix of the ablative of motion would be somewhat inappropriate. On the other hand, the use of undi in this connection is perfectly in accordance with the use in the Tamil ablative of motion of the corresponding form irundu, having been, or nindru, having stood, to which also in, the true case-sign, originally a case-sign of the locative, must be prefixed. The Tulu ablative of motion, which is also used as an instrumental, is d'du or d'd\ The corresponding form of the Tuda is edd, which is also pronounced end ; and as this is probably identical with the Canarese inda, it seems possible that the Tulu cVd' may have had the same origin. The Genitive or ' Sixth ' Case. — The genitive or possessive case is formed in the Dra vidian languages in various ways, and by means of various suffixes, each of which requires to be examined separately. The Tuda dialect uses the nominative for the genitive, as for the accusative. (1.) The abbreviated pronominal genitive. — The personal pronouns of the Tamil form their inflexion, or ordinary genitive, by shortening the included vowel of the root — e.g., ni (properly nin), thou, nin, thy ; nam, we, nam, our. This shortened form has the force of a genitive in Tamil without any suffix or addition whatever, though it is often strengthened by the addition of a sufiix in the other dialects — THE GENITIVE 287 e.g., in Canarese it requires to have a genitive suffix appended to it, and of itself it is merely an inflexional basis. In the Scythian of the Behistun tablets the nominative of the pronoun of the second person is long — viz., ni, whilst the inflexional form and enclitic possessive m is short, precisely as in Tamil-Canarese. We shall best, I think, understand the origin and force of this peculiar form of the genitive of personal pronouns, by considering it as a pronominal adjective. Every Dra vidian noun of quality or relation becomes an adjective on being prefixed to a noun-substantive for the purpose of quahfying it ; and ordinarily the only changes which it undergoes on becoming an adjective are such petty euphonic changes as are intended to facilitate the combined enunciation of the two words. The change in the quantity of the personal pronoun to which I have referred, appears to have this origin. I regard it as simply euphonic, and euphony is certainly promoted by this con- version of a long vowel into a short one prior to the addition of the case-suffixes, or of the governing substantive. We find apparently a similar euphonic shortening of the quantity of the vowel of the root, on the conversion of the abstract noun into an adjective. See the section on " Numerals " — e.g., dtu, Tam. six, diubadu, sixty ; eru, seven, eruhadu, seventy. There is room, however, as we shall see, for supposing that the process which has actually taken place may have been the reverse of this — viz., that the shorter form of these numerals is the radical one, and that the longer has been euphoni- cally lengthened. (2.) The neuter inflexional genitive. — The neuter inflexions attu, attru, ti, ti, &c., are largely used in forming the genitive in Tamil and Telugu. The various suffixes which are used to form the inflexion were originally, I conceive, signs of the locative case ; but in process of time they have come to convey more commonly either a possessive or an adjectival signification, according to the connection ; and in many cases, as has been shown, they have shrunk into inflexional increments of the base, or have become mere euphonic links of connection between the base and the case-suffix. Dr Trumpp con- siders the inflexion or formative of the North Indian vernaculars originally a genitive. The inflexion which is now under consideration is in Tamil attu, and is used by the singular of neuter nouns alone, attu, pronounced attru, is occasionally used by neuter pronominal plurals. The same inflexion — for I believe I have shown it to be the same — is in Telugu ti or ti. 288 THE NOUN The inflexional suffixes being, as I conceive, first locative then possessive suffixes in their origin, their adjectival use naturally flowed from their use in forming possessives. There is sometimes little difference in signification between the locative, the genitive, and the adjective ; and in several languages besides the Dravidian the adjec- tival formative either appears to have been derived from the possessive suffix, or to be identical with it. Thus, as we have already shown, in Tamil, it matters little whether hulattu mm (from kulam, a tank, and 7nin, fish) be translated adjectivally tank fish, or genitivally the fish of the tank, or locatively the fish in the tank. The adjectival rendering is ordinarily the more natural one, but if a few words be added to the compound expression, so as to bring out the full force of the inflexional suffixes, it will be evident that those suffixes must have been signs of case originally, and that their adjectival use is secondary to their use as signs of the possessive or locative. Thus, when we say in Tamil, i-(h)-hulattu mm perugittru, to render the sentence, this tank fish has increased, would not only be barbarous, but would partly fail to express the meaning, which is, the fish of this tank have increased. In this instance it is evident that the suffix attu is used as a sign of the genitive, though capable of acquir- ing in certain connections the force of an adjectival formative. This same suffix attu has sometimes in Tamil and Malayalam the force of a sign of the locative, properly so called, like the corresponding inflexional suffixes in Telugu ; and when used as a suffix of the locative, it is governed by a verb, not by a noun ; from which it is certain that it must be regarded as a case-suffix in origin. It is here to be noted that though attu may have had at first a locative signifi- cation, yet in such phrases as those given above, it is clear that it is not used as a locative. It has a locative signification only when the governing word is a verb. In these instances the governing word is a noun ; attu is therefore used as a possessive. Max Miiller appears to derive the genitive from the adjective, not the adjective from the genitive. He says (" Lectures," p. 110), "It can be proved etymologically that the termination of the genitive is, in most cases, identical with those derivative suffixes by which substantives are changed into adjectives." I have already mentioned the connection which subsists between the inflexional suffix attu and adu, it, the neuter singular demon- strative pronoun. It is deserving of notice in this place that adu (the very same demonstrative, I doubt not) is one of the recognised suffixes of the possessive case in Tamil, and is occasionally used as a THE GENITIVE 289 possessive in tlie other dialects also. Thus we may say in Tamil either marattu {k)-koppu (from maram, a tree, and koppu, a branch), the branch of a tree, or marattinadu koffu (mar' -attin-adu) . mara- madu may also be used, though not in ordinary use, because in- euphonic ; but the possessive case-sign adu is quite as frequently suffixed to the crude form of the noun, or the nominative, as to the oblique form — e.g., vdrei-(y)-adu param, the fruit of the plantain, is as common as vdrei-(y)-in-adu param, and is even more elegant. I have no doubt of the identity of the adu of vdrei-(y)-adu and the attu of marattu in origin. The old crude base of maram, a tree, is mara, as found in Canarese, the final am or m being a formative ; and on adu, the sign of the possessive (originally a demonstrative), being added to mara, we shall have mxiradu, of a tree (in Canarese marada) ; of which the d has only to be doubled (as it is colloquially by the Tamil people, many of whom say attu for adu), when the word becomes marattu, the very form in which we now find it. In old Canarese we find this form attu alternating with adu and atu in the possessives of the personal pronouns — e.g., instead of ninnadu, thine, we sometimes find ninatu or ninattu. In Telugu, the inflexional suffixes ti and ti are used without any additional particle as signs of the possessive or genitive even more frequently than in Tamil. The postposition yokka is but seldom added to it, and needs not ever be added. In Telugu also the connection subsisting between this suffix and the neuter demonstrative pronoun is still more obvious than in Tamil, adi, it, is systematically suffixed in Telugu to nouns and pronouns to convert them into possessives {e.g., vdridi, their or theirs), and the relation subsisting between adi (or di, as it is in some instances) and ti or ti is very close. In Canarese the corresponding particles ad and ar, though used as inflexional increments of the base, prior to the addition of several of the signs of case to certain classes of nouns, have not now of themselves a possessive signification. Their present use is purely euphonic, and does not contribute to gramma- tical expression. Nouns in which ad and ar are introduced form their possessives in ada and ara ; and in these forms the final a is that which contains and conveys the possessive signification, ad and ar have only the same incremental or euphonic force in ad-a and ar-a, that in has in in-a, which is a corresponding Canarese possessive. (3.) The neuter demonstmtive genitives. — adu, it, and its euphoni- cally leng-thened equivalent adu, are often used, especially in classical Tamil, as signs of the possessive, and they are ranked by native grammarians amongst genitive case-signs, adu is the neuter singular 290 THE NOUN demonstrative (derived from a, the remote demonstrative base, and d, the sign of the neuter singular). Its meaning when standing alone. is invariably that of a demonstrative pronoun, but by usage it has acquired the signification of a genitive or possessive, when annexed to any noun as a suffix, avan-adu is literally ' he+that,' that is, ' he-f- that which belongs to him,' but by usage it means ' his property,' his. This use of adu, as a possessive suffix, is derived from its use as the formative of nouns of possession. By the addition of this demonstrative to any noun or pronoun (generally it is added to the inflexion — in the case of pronouns it is always to the inflexion that it is added) a compound noun of posses- sion or relation is formed, which, like all Dra vidian nouns of relation, is capable of being used as an adjective ; and it seems to have been the use of nouns with this termination as possessive adjectives which has led to adu and its equivalents being regarded as signs of the possessive case. The noun to which adu is appended may be used, and often is used, without any addition or modification, as the nominative of a verb or of a sentence. Thus, enadu, Tam. (from en, my, and adu, that), signifies properly that (which is) mine ; and this compound possessive may either be used adjectivally — e.g., enadu Jcei, my hand, literally the hand that is mine (in which instance adu is called by grammarians a genitive case-sign) ; or it may be used as a possessive noun, and as such it becomes the nominative of a verb — e.g., enadu foyittru, mine (or my property) is gone. Thus adu which at first meant ' that,' became secondly the formative of a possessive noun (avan-adu, that which is his, literally he-j-that), thirdly the formative of a possessive adjective (avan-adu, his), and lastly a sign of the possessive case generally, signifying ' of ' or ' belonging to.' Another reason for regarding the genitive case-sign adu as originally and properly the formative of a noun or adjective of possession, is that it cannot be followed indiscriminately by any kind of noun, but by neuter nouns alone, and properly by the neuter singular alone. Thus we may say enadu kei, my hand, but not enadu keigal, my hands ; except indeed in the colloquial dialect, in which the singular is used for the plural more frequently than in the higher dialect or by the poets. The higher dialect would prefer in this instance ena keigal — ena instead of enadu — i.e., mea, instead of meum. adu is not only a formative, therefore, but is distinctively a neuter singular formative, employed to give a possessive signification to the noun to which it is suffixed. Like all other nouns, these possessive nouns in adu are THE GENITIVE , 29 1 capable of being used as adjectives, by being prefixed without alteration to other nouns ; and when so prefixed, adu came to be used and regarded as a possessive case-sign. This explanation seems to account for all the phenomena, and therefore is probably the true explanation. In Malayalam, this use of adu as a possessive case- sign, though common in the ancient poetry, has nearly disappeared from the popular dialect. It is scarcely discernible except in tanadu, enadu (from which come tandre and endre, its, my). The old Canarese possessive pronouns, answering to the Tamil enadu, &c., are ennadu, ninnadu, tannadu, mine, thine, its. These take also the shape of ninatu, &c., and also ninattu, &c. A similar use of the neuter singular of the demonstrative as a possessive suffix obtains in Telugu also — e.g., nddi, mine, literally that (which is) mine, from nd, my, and adi, that, a form which is exactly equivalent to the Tamil enadu. Telugu uses a similar suffix to form a plural possessive to correspond with enadu or nddi, viz., vi, which bears the same relation to avi, those (things), which di does to adi, that (thing) — e.g., vdrivi, theirs or the (things which are) theirs. In this respect Telugu acts more systematically than spoken Tamil. It is not so fond, however, of using these possessive nouns adjec- tivally as the Tamil, and therefore di and vi have not in Telugu come to be regarded as case-signs of the genitive. The Canarese and the Tamil not only form neuter possessive nouns and adjectives by adding to them the neuter demonstrative, but they form also masculine and feminine possessives, or possessive appellatives, of both numbers, by adding the masculine and feminine formatives to the genitive case or inflexion of nouns and pronouns. In the Tuda dialect, ad, the demonstrative base, appears sometimes to be added to the first of two nouns, when it is used adjectivally. All the Dra vidian dialects agree in appending the demonstrative possessive suffixes to the in- flexion, not to the nominative, as a general rule, wherever the nomina- tive differs considerably from the inflexion. When nouns receive in Tamil a double inflexional increment — e.g., attu and in (in combina- tion attin), the possessive suffix is added to this double increment — e.g., mar' -attin-adu Icoppu, the branch of a tree. (4.) The possessive suffix ' in,' and its varieties. — in in Tamil and ni in Telugu, and corresponding particles in the other dialects, are not only used as inflexional augments of the base and euphonic bonds of connection between the base and the case-signs, but also as suffixes of the possessive and as adjectival formatives. I have no doubt that in and ni, of themselves and originally, were locative 292 , THE NOUN suffixes, and that every other use to which they have been applied grew out of their use as signs of the locative. As Max Miiller says (p. 229), " A special case, such as the locative, may be generalised into the more general genitive, but not vice versa.'' Native Tamil grammarians do not include in amongst their case-signs, but describe it as a formative augment or adjectival increment alone : but on comparing its use in Tamil with its use in the other dialects, I am convinced that it was originally a sign of the locative, then adopted as a sign of the genitive, and that it is still to be regarded, notwith- standing its other uses, and its probable origin, as one of the most characteristic of the genitive suffixes. In Tamil, of all genitive suffixes, in is that which is most frequently used, attu is used in the neuter singular alone, and avvu (attru) in the neuter plural alone ; but in is used in connection with both numbers and with all genders. A similar use of in appears in the Malayalam. In Canarese, on the other hand, in is used only as an inflexional augment, not as a sign of case. One of the so-called de- clensions of the Canarese is said by grammarians to take ina as its genitive case-sign ; but in this instance the final a is the real sign of the genitive, as it invariably is in Canarese ; and this genitive a is found to be preceded by various euphonic increments — in, ad, ar, or V, according to circumstances. Doubtless the in of in-a, like the Tamil in, was a sign of the locative originally, then of the possessive ; but it has long ceased to contribute to grammatical expression, and therefore cannot now be regarded as a sign of case. In Telugu, na or ni, the dialectic equivalent of in, is used as a possessive suffix, as in Tamil, though not so frequently. The only difference in principle is that ni is used in Telugu in connection with the singular alone, and might be called a genitive singular case-sign, if the Telugu stood in an isolated position ; whereas in Tamil it is used in connection with plural nouns as frequently as with the singular. In Ku, which has special resemblances to the Telugu, ni constitutes the inflexion (in reality the genitive) of all classes of nouns, whether singular or plural, precisely like the Tamil in. The Gond uses as genitive case- signs na and nd, da and a — forms which are probably allied one to another, as well as to the Brahui nd, and to the Telugu and Gond ni and the Tamil in. Though in is not regarded by Tamil grammarians as a sign of the genitive, yet when those particles which are regarded as genitive case-signs are suffixed to any noun, in is ordinarily inserted between the noun and those case-signs ; so that all auxiliary or additional THE GENITIVE 293 particles are appended to this incremental in, not to the noun itself — e.g., from adu, it, is formed not ad'-udeiya, but ad'-in-udeiya, of it ; from tambi, a younger brother, is formed not tamhi-(y)-adu, but more commonly ta})ibi-{y)-in-adu, of a younger brother : and this rule seems to indicate that in, whatever its origin, has acquired more of the force of a genitive case-sign than the genitive particles which have subsequently been suffixed to it. The same inference is still more clearly deducible from the circumstance that in a large number of instances, both in the singular and in the plural, each of the case suffixes in succession is appended, not to the crude form of the noun, but to the increment in. These case-suffixes are not mere postpositional fragments, but were, or are still, nouns of relation ; and in, the particle by which they are united to the base, serves as a bond of connection, in virtue, as I conceive, of its signification as a suffix of the genitive. Thus, in the colloquial Tamil kallinidattil {kal{l)-in-idattil), in a stone, idattil, the local ablative or locative suffix, literally means ' in the place ' ; and this suffix evidently requires, or at least desires, the possessive in (with the signification ' of ') to connect it with the base. Hence kal(l)-in-idattil literally signifies ' in the place of (or occupied by) a stone.' The adjectival meaning of in, though not its only or original mean- ing, is one which is recognised by native grammarians, and which they prove by examples — e.g., fonnin {'pon{n)-in) kudam, a golden vessel. This adjectival use of in is not only allied to, but is derived from, its use as a suffix of the genitive, and in the illustration which has now been adduced it is evident that fonnin kudam might be rendered with equal propriety, a vessel of gold. It will be found also in the Indo-European analogies which will presently be adduced, that the similarity or identity of the adjectival formative and the genitive case-sign Vhich is apparent in this instance, has a wider range than that of the Dra vidian languages. There is another particle resembling in — viz., am, with its equivalent an, which is occasionally used in Tamil for both those purposes, and, like in, it is sometimes appended to the noun itself, and sometimes to the neuter inflexion. We see this fusion of the adjectival and the geni- tive signification of am in such forms as dlam (dV-am) pu, the banyan flower, or the flower of the banyan, and dttran karei {dttru, the inflexion of dru, a rivei^, the river-bank, or the bank of the river. The same adjectival formative is much used in Malayalam also — e.g., maV-am puli {mala-am puli), a mountain tiger, or a tiger of the mountain, a royal tiger. The final m of am changes by rule into 294 THE NOUN the nasal which corresponds to the first consonant of the word which follows it and with which it is compounded. Hence it changes into n when followed by a dental — e.g., panan-doppu panei-am-toppu), a palmyra tope. It must not be supposed, how- ever, that we have here to deal with an, the formative suffix of many Tamil nouns. In such words as adavhu, Tam. to it, for adan-ku, am is not considered a sign of case or even as an inflexional increment, but (as we have already seen in the section on " The Inflexional Increment ") as a formative sufiix, found in the nominative (though rarely), as well as in the oblique cases, am and an agree in this, that both are used as formative particles of nouns, am, however, is also used as a genitival or adjectival suffix in Tamil, whereas an is not. am and an are, I believe, identical in origin ; so also another pair of particles in and im (the latter the Canarese form), am and an I regard as demonstrative pronouns ; in and im as related to or derived from il, here, a house, the locative case-sign. We have now to inquire whether any trace of the genitive case-sign or adjectival formative in in, ni, am, or any related form, can be found beyond the circle of the Dravidian dialects. Of all the North Indian vernaculars the Gujarathi is the only one which contains a form of the genitive resembling that which we have been examining. That language has a genitive sufiix in n (no, ni, nun), which somewhat resembles the Telugu ni, nu, &c. In the language of the Bodos, a Himalayan tribe, the pronominal genitive is regularly formed by sufiixing ni — eg., an-ni, of me, nan-ni, of thee, hi-ni, of him. In Sanskrit the n which precedes the ah or as, of certain genitives, is undoubtedly euphonic ; but both in Sanskrit and in other members of the Indo-European family, we may observe distinct traces of the adjectival or genitival use of a particle of which the consonant n is the most essential element. With the Dravidian particle compare an-a, the Sanskrit adjectival formative, and an, the suffix of appellatives ; the Greek possessive suffix wv (on) ; the adjectival use of iv (in) in Greek words like \l6lv oi, in the Germanic woode/i; and also in, the Sanskrit sufiix of agency, which is preserved in the adjectives of the New Persian. These forms look as if they were reciprocally related ; and possibly also there may be some ulterior relationship between them and the Tamilian in. There are traces in the Indo-European family of languages themselves of the use of in as a distinctively genitival suffix. The Celtic forms its genitive systematically by means of n, an, en, &c. : nor is it the genitive plural only of the Celtic dialects which uses this case-sign (as in the THE GENITIVE 295 Sanskrit family), but it is employed to form the genitive singular also. It should be noticed too that in the ancient Egyptian n (alter- nating with m) was used to express all case relations, but particularly that of the genitive. Compare also the Sanskrit genitive or possessive mama (ma-mxi), of me, my, with the Zend mana, the Old Persian mand, and the Gothic meina, mine, theina, thine, seina, his ; in each of which examples the final na, or its Sanskrit equivalent ma, resembles the Dravidian in or ni, not only in sound, but also in the union of an adjectival signification with that of the possessive or genitive case. The Lithuanian goes further than any other Indo- European tongue in resemblance to the Tamil in this point, for it not only uses w as a sign of the pronominal possessive (of the first person), but it adopts this genitival man as the inflexional base of all the rest of the oblique cases of the same pronoun. In the languages of the Scythian stock we find a large number of still more essential analogies with the Dravidian genitival sufiix in or ni. Compare both with the Dravidian and with the Indo-European possessives the Mongolian and Manchu mini (mi-ni), of me, my ; and the Mongolian tchini and the Manchu sini (si-ni), of thee, thy. In the languages of tte Finnish family, the prevailing form of the genitive is that which corresponds to the Dravidian : it is n, an, en, un, &c., not only in pronominal inflexions, but universally. Thus in Mordvin and Cheremiss, the genitive is formed by suffixing n or en — e.g., kudo, a house, hudo-n, of a house. The genitive plural of the Mordvin is nen, possibly a reduplication of n, intended to symbolise the plural — e.g., kudot-nen, of houses. The Lappish genitive takes n or en in the singular, and i in the plural, e forms the ordinary possessive suffix of the Magyar. The Finnish proper forms the genitive by suffixing n, un, in, an, kc.—e.g., mind (min-d), I, min-un, of me, my. The prevailing form of the genitive in the Tratar or High Asian families corresponds to nen, the reduplicated suffix of the Mordvin plural, and to its equivalent reduplication in the old Scythian of the Behistun tablets ; but whilst the reduplicated suffix is very fre- quently used, it systematically alternates with the simpler suffix un or in. The Oriental Turkish forms its genitive by suffixing ning or nin, or nmg or nm. In the Ottoman Turkish the initial nasal is only occasionally used : the genitive plural is uniformly un ; the singular takes un or nun, according as the noun to which it is suffixed ends in a consonant or in a vowel. In the Mongolian, the sign of the genitive is u after the consonant n ; after every other consonant, un ; 296 THE NOUN and after a vowel in or yin. The personal pronouns, as has already been observed, form their possessive by suffixing nu or ni — e.g., mi-mu, or mi-ni, my. Compare the Mongolian kol-un, of the foot, with the ordinary Tamil genitive of the corresponding noun Ml-in, of the foot. The Calmuck dialect of the Mongolian forms its genitive by suffixing u 01 i to nouns ending in n, and in or yin to all other nouns. The Tibetan postfixes in like manner i or yin. The Manchu makes much use of a possessive relative suffix ngge or ningge, signifying ' which has' ; but it also forms genitives, properly so called, by suffixing ni or i. In Japanese ni is used generally as a sign of relation, with a still wider variety of meanings than the Tam. in. no, however, is the ordinary sign of the possessive, and is also used in the formation of adjectives. In the language of the Scythian tablets of Behistun, the genitive was ordinarily formed by suffixing na : the first personal pronoun formed its genitive by suffixing a reduplicated form of this particle, ni-na — e.g., hu-ni-na, of me ; whilst the genitive plural was generally formed by means of the addition of inna, probably softened from ni-na. The nearest direct resemblance to the Behistun-Scythian genitival na, is the Brahui na, and the Gond ndovd. This interesting record of the speech of the ancient Scythians furnishes us, I think, with a clue to the origin of nun or nin, the Tratar genitive suffix. In the Tratar tongues nun is interchangeable with and equivalent to un ; and un or in is also interchangeable with ni or nu ; in Mongolian yin and un are suffixed to substantives, ni to the personal pronouns. It appears from the Behistun tablets that na, the ordinary genitive suffix, was sometimes euphonically changed into ni-na, and that this again was softened into inna. I conceive that the Tratar un was in this same manner, by the reduplication of the nasal, converted into nun ; which in Manchu became ngge or ningge. Possibly also ni or nu was nasalised by the addition of a final n or ng, of the use of which we have an instance in point in the final euphonic n of the first and second personal pronouns in most of the Scythian languages. A parallel instance of the reduplication of a nasal is apparent in Telugu itself, in the conjunctive or copulative particle. This particle is um in Tamil, u in Canarese, and u in Telugu ; but this Telugu u becomes euphonically nu, and by reduplication nunnu in particular instances. (5.) The genitival suffix ' a.' — This sign of the genitive or possessive claims to be regarded not only as the most distinctively Dravidian suffix, but as the sole original one. It is little used in modern Tamil, though placed first in the list of genitive case-signs by Tamil gram- THE GENITIVE 297 marians ; but if we take all the Dra vidian idioms into consideration, in several of which it is the only sign in use, we shall find it more largely used than any other suffix of the genitive — a proof of the accuracy of the Tamil classification. I conceive this suffix to be identical with a, the formative of the most frequently used Dra vidian relative participle (see " The Verb "), but totally distinct in origin from a, the neuter particle of pluralisation which has already been investigated. In Canarese a is the only sign of the genitive which is ever used. It is sometimes euphonically lengthened to a, as the Tamil adu, of which the same a forms the most essential part, is sometimes length- ened to ddu. a is sometimes preceded by an euphonic consonant, which is inserted between it and the base, to form a link of con nection between them, viz., by v or y, the use of which is purely of an euphonic nature, and by m, ad, or ar, which are inflexional incre- ments of the base, and old petrified locatives or genitives — e.g., guru-{v)-a, of a priest ; kuri-(y)-a, of a sheep ; kus-in-a, of a child ; mar-ad-a, of a tree ; ad-ar-a, of that (thing), or of it. When this genitive a is added to the abbreviated inflexional form of the Canarese personal pronouns, the final nasal of those pronouns is doubled — e.g., nanna (from nan, I), of me ; namma (from nam, we), of us. A comparison of these forms with the Tamil and Tulu nama, of us, our, proves that the doubling of the final nasal arises from an euphonic source, a forms the genitive suffix not only of the singular of Canarese nouns and pronouns, but also of the plural, whether the noun belongs to the rational or to the irrational class — e.g., avar-a, of them (epicene), avugal-a, of them (neuter). These examples prove that a is the true Canarese genitive case-sign : and it is also to be noted that this case-sign is never used, like in in Tamil, as the common fulcrum of the suffixes of all the oblique cases, but is used solely as a case-sign of the genitive. In Tulu a is the only sign of the genitive, as in Canarese. The only difference is that in the plural a is weakened to e. In many instances in singular nouns a is preceded by d or t ; but this con- sonant is merely the equivalent of the Canarese ad or d, which has already been referred to ; and in the genitive of the personal pro- nouns a is preserved purer in Tulu than in Canarese. Thus, instead of the Canarese nanna, of ifle, the Tulu has yan-a ( =nan-a), and in- stead of ninna, of thee, it has nin-a. The language of the Kotas of the Nilgherry Hills forms all its genitives by suffixing a. In Telugu a forms the plural inflexion or genitive of all substantive 298 THE NOUN nouns without exception, lu, the pluralising particle, is changed into la ; and as the u of lu is added merely to facilitate enunciation, and I alone constitutes the suffix of the plural, it is evident that the a of la is a suffix of case. As the plural inflexion, a constitutes the fulcrum to which the other case-signs, or suffixes of the oblique cases, are added ; and as the genitive plural, it expresses the signification of the genitive, without any auxiliary or additional particle. The Telugu personal pronouns use their crude bases adjectivally as their inflexion and genitive. The pronouns of the third person, or the demonstratives, generally form their genitives, both in the singular and in the plural, by adding i to the root : in the singular a few of them suffix ni, as is done by the greater number of nouns in the singular. One of the Telugu pronouns uses a, both in the singular and in the plural, as the sign of the genitive, in complete accordance with the Canarese and Tulu. The genitive of the reflexive pronouns tdn-u, self, tdm-u, selves, is formed in Telugu by shortening the quantity of the radical vowel and suffixing a, as in Canarese — e.g., tan-a, of self, tam-a, of selves. The adjectival a of some Telugu substantives is evidently identical with this genitival a — e.g., ur-a Jcavi, a village poet, or a poet of the village. In Tamil, though a is placed first in the list of genitive sufiixes, it is now less used than any other sign of the genitive, and indeed is used only as a classical genitive of the personal and reflexive pro- nouns — e.g., nam-a, our (from nam, we), like the Sanskrit 7nama, my, and tava, thy. It is difficult, indeed, to determine whether this sufiix has retained in Tamil any genitival signification whatever. Whether it be attached to a singular or to a plural pronoun, it must be followed by, and be in agreement with, a neuter plural noun ; and this circumstance would lead to the conclusion that in Tamil it is used as a suffix of plurality, not as a sign of the genitive. On this supposition, in the words ena heigal, my hands, ena would signify not mei, of me, but mea, (the things that are) mine. It would be a pro- nominal adjective or possessive plural, not a genitive ; and the fact that a is largely used in classical Tamil as a sign of the neuter plural (e.g., sila, few, literally a few things ; fala, many, literally many things), shows that this supposition would be a very natural one. On the other hand, a was classed with genitive suffixes by the most ancient Tamil grammarians, and those grammarians, who were remarkably well acquainted with the principles of their own lan- guage, were perfectly aware that a was also a sign of the plural of " irrationals." Moreover, though it is stated by Tamil grammarians^ THE GENITIVE 299 that the genitive in a must always be in agreement with a plural noun, yet they admit that the noun with which it agrees is some- times singular in form though plural in signification — e.g., the expression nun-a str'adi, thy small foot, occurs in the Chintamani. They say that foot is here used for feet, and this is certainly true ; but it does not follow that nun-a is determined thereby to be a plural, for the use of the singular with a plural signification, yet with the declensional and conjugational forms of the singular, is a fixed usage of these languages. I think, therefore, that we may confidently regard this nun-a as an illustration of the use of a, even in Tamil, in connection with the singular. In Tamil, it is true, a is ordinarily followed by the neuter plural alone ; but in Canarese and Telugu it may be followed by any gender or number ; and the a of the Tamil tan-a, of self, is evidently identical with that of the corresponding Telugu tan-a ; whilst the a of nam-a, of us, our, is evidently identical with the Canarese namm-a. Hence, as the one a is unquestionably a genitive, so must the other have been originally ; and thus we are led to the supposition that the Tamil rule which requires a to be followed by the neuter plural is merely a secondary, recent, dialectic peculiarity, which has arisen from the influence of its accidental resemblance to the sign of the plural of irrationals. This peculiarity of the genitival a in Tamil may be compared with the somewhat parallel case of the use in Hindustani of one possessive suffix rather than another, according to the gender of the noun which follows and governs that to which it is suffixed. Though in grammatical Tamil a is always followed by the plural, yet the vulgar in the rural districts commonly use it without discrimination of number, as in Canarese and Telugu. Thus, they will say na7)ia (or more commonly, as in Canarese, namma) ur, our village ; and this confirms the supposition that in Tamil, as in the other dialects, the original use of this a was simply that of a suffix of the genitive. In the Ho, a Kolarian dialect, a is a common possessive suffix ; and it is also, as in Tamil, an adjectival formative. We have now to inquire whether there is any other language or family of languages with which this genitive suffix appears capable of being affiliated. There is no direct Scythian analogy for it, and the only affinities which I have observed are Indo-European. The most direct and reliable Indo-Eurdpean analogy is that which is presented by the personal pronouns, which in some of the Indo-European dialects have a possessive in a strongly resembling this Dravidian possessive. If we looked only at the Gothic meina, my, theina, thy, 300 THE NOUN seina, his or its, we should naturally conclude the sign of the pos sessive in these words to be, not a, but na (answering to the old Scythian and Brahui na, and to the Telugu ni) ; but on comparing the forms which this sign of the possessive assumes in various languages, it appears probable that a alone conveys the signification of the possessive; and that the nasal which precedes it in the Sanskrit mama, the Zend mana, and the Gothic meina, may merely have been inserted euphonically for the purpose of keeping the contiguous vowels pure. Compare mama, Sans, my (from 7na, I), with tava, thy (from tva, thou) ; and especially compare the Gothic tlieina, seina, with the corresponding Lithuanian possessives tava-s, sava-s. In these instances v euphonic is used as the equivalent of n. The Indo- European pronominal possessive in a is exceptional ; for the primitive languages of that family evince an almost perfect agree- ment in the use of as, or some closely related form, as the sign of the genitive singular, and of sdm or dm as the sign of the genitive plural. In the later Teutonic dialects, however, a genitive case-sign in a becomes exceedingly common, and is found in the plural as well as in the singular. Thus in the Frisian all plural substantives and such singulars as end in a vowel form their possessive by suffixing a ; in the Icelandic all plurals and all masculine and neuter singulars use a as their case-sign ; and in the Anglo-Saxon all plurals. Though the oldest Gothic possessives accorded with ordinary Sanskrit forms as and dm, yet the resemblance between the possessives of some of the Teutonic vernaculars and the Dravidian possessive is deserving of notice. The use of a as a sign of the possessive by all plural sub- stantives in Telugu is especially remarkable. Has the Dravidian a under consideration been softened from as (of which, however, there is not the smallest trace or analogical probability), or has it been softened from na, the old Scythian suffix ? The latter sup- position, though unsupported by evidence, is not an improbable one in itself ; for we have seen that the Gond nd alternates with d, the Scythian ni-na w^ith inna, the Turkish nun with unu. (6.) The Malaydlam genitive singular suffix ' re' or ' de.' — In most cases this Malayalam genitive takes the shape of indre or inde, of which in is the genitival suffix and inflexional increment, which has already been described. In en-de, my, the inflexional base is of itself a genitive, and the addition of in is not required ; hence it appears that de or dre is an auxiliary genitive suffix, like the adu which is so often added to in in Tamil, and is probably from the same origin. This suffix is written re ; but it is always added to n, and when it is THE GENITIVE 3OI thus added, the compound is regularly pronounced, not as nte, but as ndre or nde. Neither the Tamil nor the Malayalam possesses any other method of producing the sound which is indicated by these letters (a peculiarly euphonic ml), but that of conjoining the final n of those languages and the hard r ; which, when pronounced in com- bination, have the sound of ndr, or, as some pronounce it, ndz, or more commonly still, nd. Thus, from en, to say, and du, the regular formative of the preterite participle, the Canarese forms endu, saying, or having said ; and this in Tamil is written emu ; but it would be erroneous to suppose tu to be the sign of the preterite in Tamil instead of du, for envu is intended to be, and is pronounced, endu or endru, nearly as in Canarese. Hence some analogies to the Malayalam re (in reality de), which might be suggested, appear at once to be illusory. The Malayalam re was connected by Dr Stevenson with the Canarese genitive ra. It has been shown that a, not ra, is the genitive suffix of the Canarese, and that the r which precedes it is properly ar, an inflexional incre- ment (like ad and in), which is inserted between the root and the case-signs of three cases, besides the genitive, of certain classes of nouns. The Malayalam re (de), on the other hand, is suffixed ex- clusively to the genitive, and no other suffix of case is ever appended to it. Nevertheless, as I connect de with the Tamil adu, it, and as with this I connect also the Canarese ad and its hardened form ar, it may be admitted that in this modified and remote manner the Malayalam and the Canarese forms are allied. Still more illusory is the apparent resemblance of this Malayalam re or de to the adjectival possessive suffixes of the Hindustani personal pronouns rd and ri {e.g., merd, mens, meri, mea), to the corresponding New Persian inflexion rd {e.g., to-rd, thy, thee), and to ra, the Gothic genitive plural suffix of the personal pronouns {e.g., unsara, our, izvara, your), from which the final r of our English our and your has been derived. The Hindustani r is supposed by Bopp to be derived from d ; merd, mens, being derived from the Sanskrit madiya, my ; but I cannot suppose that the Malayalam form has any connection whatever with the Hindustani and the Persian, except, indeed, on the supposition that the d of the Tamil demonstrative neuter singular, adu, is remotely connected with the formative d of the Sanskrit possessive adjective. The Malayalam de, like the Tamil adu, is used as a genitive suffix of the singular alone, a confirmation of the opinion that it is derived from adu, which in its original signification is the neuter singular of 302 THE NOUN the demonstrative. In the genitive plural, the Malayalam uses ude, answering to the colloquial Tamil udeiya (from udei), belonging to, of. Compare the Malayalam ente, endre, or ende, of me, with the corre- sponding Tamil enadu, of me, that which is mine. The Malayalam possessive noun mine, or that which is mine, is endredu, from en-de, my, and adu, it, corresponding to the Tamil enadu. This latter enadu, however, is not the genitive enadu, my, with which I have compared en-dre, but a possessive noun in the nominative case ; and though I suppose the Malayalam de to be itself a corruption from adu, it, yet the demonstrative suffix would be appended a second time, on the origin and true meaning of de being forgotten. We see illustrations of this repetition of an ancient suffix in many languages — e.g., malei-(y)'in-in, High Tam. from a mountain ; and this very demonstrative adu, it, is twice used in the Tamil negative participial noun illddadu, the thing which is not ; in which the first d, though a representative originally of the neuter singular demon- strative, has lost its proper signification, and become a mere euphonic link of connection, or technical sign, in consequence of which d requires to be repeated. (7.) Auxiliary suffixes of the genitive in Telugu and Tamil. (i.) In Telugu, yokka, or yoka, is sometimes appended to the in- flexion, or natural genitive, as an auxiliary suffix of case — e.g., from the ordinary possessive nd, my, is formed optionally the equivalent form nd-yokka, my, of me. This suffix is rarely used, and seems foreign to the idiom of the language ; no other pure Dra vidian dialect possesses any suffix resembling it. A suffix somewhat resembling yokka is found in the Eajmahal and Uraon languages, which contain an overwhelming preponderance of K61 elements, though formed probably upon a Dravidian basis. The possessive suffix of the Rajmahal is ki, that of the Uraon ghi. If these particles are at all connected with the Telugu yoka, which seems doubtful, we should be warranted in connecting the whole with the ordinary possessive or adjectival suffix of the Hindustani, the feminine of which is ki (masculine kd), and through that suffix with the formative ka of the Sanskrit possessive adjectives mdmaka, my, tdvaka, thy, asmdkam, of us, our, &c. A closer analogy to yoka is that of the dative postfix of the Mikir, which is yok or ay ok. (ii.) In Tamil, udeiya is commonly appended to the inflexion of nouns and pronouns as an auxiliary possessive suffix, udeiya [udei- {y)-a), means belonging to, or, literally, which is the property of, and is derived from the noun udei, property, possession, by the addition THE LOCATIVE ' 303 of a, the sign of the relative participle, on the addition of which to any noun it is converted into an adjective. Thus en-udeiya kei, my hand, means literally the hand which is my property, for en of itself signifies my. Through usage, however, there is no difference in signification, or even in emphasis, between en and en-udei-{y)-a. The Malay alam dispenses with ya or a, the sign of the relative participle, and uses ude (in Tamil udei), the uninflected noun itself, as its auxiliary suffix of the genitive. This suffix is still further mutilated in modern Malayalam into de — e.g., futn-de, of a daughter, udeiya is very largely used as an auxiliary genitival suffix in colloquial Tamil, and in some grammars written by foreigners it is classed with the signs of the genitive ; but, properly speaking, it is not a case-sign, or suffix of case at all, but the relative participle of an appellative verb used adjectivally, and it is to be compared not with our prepo- sition of, but with the phrase, belonging to. Locative or ' Seventh ' Case. — Dravidian grammarians state that any word which signifies ' a place ' may be used to express the locative. In each dialect, however, some words or postpositions are so frequently and systematically used for this purpose that they may be regarded as distinctively locative suffixes. In Tamil, han, an eye, which has also the signification of a place, is given in the grammars as the characteristic suffix of the locative. As a verbal root, han means to see : its secondary signification was, look ! its third, there ; its fourth, a place : and in consequence of the last meaning it came to be used as a sign of the locative. It is very rarely used, and the use of hdl (in Malayalam kal), which stands next in the list in the Nannul, is still more rare. I have no hesitation in saying that the most distinctive sign of the Tamil locative is il, a house, a place — literally, this place, here. In colloquial Tamil the most commonly used sign of the locative is idattil, a compound suffix, which is derived from idam, the ordinary word for a place, attu, the inflexion or basis of the oblique cases (id'-attu), and il, an older, purer word for a place, which is added to id'-attu (id'-att'-il), as the real sign of the locative, with the meaning of our preposition in. The signification of the whole suffix is literally, in the place of, or in the place occupied by ; but it is evident that what really dis- tinguishes the locative in t^is compound is il, in — the suffix of a suffix ; and that the meaning which the entire compound receives in actual use is simply in. In the lowest patois of colloquial Tamil, the locative suffix which is most used is kitta, near, the infinitive of a 304 THE NOUN verb. The higher dialect of the Tamil uses also ul and uri, within, among, as signs of the locative. The ancient Canarese generally used 61, corresponding to the Tamil ul, as its locative suffix ; whilst the modern dialect uses alii or illi, a form which answers to the Tamil il. alii is properly a noun of place, formed from the remote demonstrative a ; and its fellow is illi, formed from i, the proximate demonstrative. These words mean literally that place and this place, or there and here, and their use as locative suffixes appears to betoken a later state of the language than the use of il and ul in Tamil, and of 61 in Canarese. The locative suffix of the Tuda is ulch or orzh, which seems to be simply the Tamil ul rudely pronounced, r and I seem generally to become rzh in this dialect. In Telugu the sign of the locative most commonly used is 16 ; another form frequently employed is andu. 16 is more intensely locative in its signification than andu ; it means within, and is obviously identical with the Canarese 61, and the Tamil ul. andu means simply ' in,' and, like the Canarese alii, is properly a noun of place. I consider andu, the adverbial noun, there, identical with andu, the sign of the locative. It is evidently formed from a, the remote demonstrative, with the addition of a formative d, whilst indu, the correlative adverb of place, is derived from i, the proximate demonstrative. The Canarese also possesses adverbs corresponding to these, viz., anta and inta, antalu and intalu, but uses them chiefly to express comparison, like our adverb than. The Telugu locative suffix andu (meaning on or in) bears some apparent resemblance to the Sanskrit antar, among, but this resemblance is illusory ; for andu is derived from a, that, by the addition of the neuter formative du, which becomes euphonically ndu, and, corresponds not to the Sanskrit, but rather to anda, that, the demonstrative adjective of the Tamil. The Tulu locative suffix is du or d\ tu or t\ which Dr Gundert conjectures may be derived from udu, equivalent to ul, Tam. within, 61, Old Canarese, or from ede, equivalent to Tarn, idei or idam, place. The nature of the initial vowel of the Tulu sufiix seems difficult to ascertain. The d is sometimes preceded by o, some- times by a or e ; and sometimes it is obliterated, as in heit\ in the hand, a form which suggests Telugu analogies. On the whole it seems to me most likely that the Tu]u locative du or tu has sprung from the same origin as the Can. alii and the Tel. andu, viz., the adverb of place there, one form of which in Tulu is ade, thither (corresponding to ide, hither, and ode, whither). THE LOCATIVE 305 In Telugu the postposition na, which becomes ni after i, is used as a locative suffix in connection with neuter nouns, ni (and hence its eq^iivalent na also) is evidently identical with in, the sign of the ablative of motion in High Tamil, which I have supposed to be prop- erly a sign of the locative ; and probably this in is the origin of in, the Tamil, and ni and na, the Telugu, genitival or inflexional suffixes. The genitive is more likely to be derived from the locative than the locative from the genitive. With this Telugu locative na we may compare the Ostiak locative na, ne, the Finnish and Magyar an and en, and especially the Japanese locative ni — e.g., Yedo-ni, in or at Yedo. In Telugu, and in the higher dialect of Tamil, the inflexion or basis of the oblique cases, which has generally the force of a genitive, is sometimes used to denote the locative also. This is the case in Tamil "only in those connections in which it is governed by a verb, expressed or implied. In Tamil the inflexion which is chiefly used in this manner is attu — e.g., nilattu, upon the earth. The Malayalam uses attu in a similar manner ; and in Telugu a corresponding change from ti to ta converts the inflexion or obsolete genitive into a locative — e.g., inti, of a house, inta, in a house. The same inflexion in ta denotes the instrumental in Telugu, as well as the locative — e.g., compare c/ie^i, of a hand, Wiihcheta, byahand ; but this form seems to have been a locative originally. This fusion of the meaning of the genitive and locative suffixes corresponds to a similar fusion of the signs of those cases, which a comparison of the various Indo- European tongues brings to light. The genitive and locative case- signs are often identical in the Finnish family of languages also. Bearing this in mind, we may conclude that in or ni, one of the most common inflexional increments in all the dialects, m, one of the Tamil possessive and adjectival suffixes, in, the sign of the Tamil ablative of motion, and im, the Canarese sign of the instrumental, with the various shapes they take, were all originally locatives, and identical with il, which we have seen is so exceedingly common as a locative suffix, with the original meaning of here. In all the Dra vidian idioms the locative suffixes are used like our than, to express comparison. Sometimes the locative alone is used for this purpose : oftener the conjunctive particle is added to it — e.g., il-um,, in Tamil, l6-nu, in Telugu, which compound has the significa- tion of our even than. The Vocative or ' Eighth ' Case. — In the Dravidian languages therc^ 305 THE NOUN is nothing which properly deserves to be styled a suffix or case-sign of the vocative. The vocative is formed merely by affixing or suffix- ing some sign of emphasis, or in certain instances by suffixing frag- ments of the personal pronouns. The most common vocative in Tamil is the emphatic e, which is simply appended to the noun. Sometimes, also, the vocative is formed by substituting a for the formative of gender — e.g., from kartan, Lord, is formed kartd, Lord, by converting the final vowel into ay (a fragment of the old pronoun of the second person singular) — e.g., from tangei, sister, is formed tangdy, sister ; or by lengthening the vowel of the pluralising particle — e.g., from fdvigal, sinners, is formed 'pdvigdl, sinners. Sometimes, again, especially in poetry, rational plurals are put in the vocative by appending to them ir, a fragment of nir, you — e.g., elUr, literally ell4r, all ye. Both in Tamil and Malayalam the voca- tive is often formed by lengthening the final vowel of the nominative — e.g., tori, female friend, voc. tore. This usage prevails also in Japanese. In the Indo-European languages the nominative is often used for the vocative, and what appears to be a vocative case-ending is often only a weakened form of the final syllable. In the Dravidian lan- guages, in like manner, the crude root, deprived of all increments, is often used as the vocative. In Telugu the vocative singular is ordinarily formed by lengthening the final vowel of the nominative (and all Telugu words end in some vowel), or by changing the final u into a or d. ara or ard, from the same root as the Tamil pronominal fragment tr (viz., ntr, ye), is post- fixed as the vocative of masculine-feminine plurals. In addition to these suffixes, various unimportant vocative particles, or particles of exclamation, are prefixed to nouns ; some to one number only, some to both. In Canarese the vocative is ordinarily formed by appending a, by lengthening the final vowel of the nominative, or by adding e or e. Masculine-feminine plurals form their vocative not only by means of e or e, but also by suffixing ira or ird, from the same source as the Telugu ard — viz., the old ntr or tr, ye. Such being the origin and character of the Dravidian signs of the vocative, it is evident that we cannot expect to find allied forms in any other family of languages. Compound Case-signs. — As in the Hungarian and other Scythian tongues, and in some of the languages of the Eastern islands, so in Dravidian, two or more case-signs are occasionally compounded to- gether into one. We have already noticed the custom of annexing the THE VOCATIVE 307 various signs of the oblique cases to the inflexion or sign of the genitive ; but other combinations of case-signs are also in use. Thus, there is a combination of the dative and locative — e.g., vUtukkul (vUtukk'ul), colloquial Tam. within the house, in which the locative ul is combined with the dative or directive kku, for the purpose of intensifying m, and educing the meaning of ' within.' The higher dialect would in this instance prefer vUtul, the simple locative ; but vUtukkul is also idiomatic. The ablative of motion in each of the Dravidian dialects is generally a compound case, being formed of the locative and a verbal participle, or even of two loca- tives — e.g., mane-(y)-ill-inda, Can. out of the house, from illi or alii, the sign of the locative, and inda, a sign of the instrumental, which is used also as a sign of the ablative, but which was, I conceive, a locative originally, and identical with im, the Canarese form of the Tamil in. Such compounds may indeed be formed in these languages at pleasure, and almost ad infinitum. Another instance of them in Tamil is seen in the addition of the dative to the locative {e.g., il-ku, idattil-ku), to constitute the locative-directive, which is required to be used in such expressions as, I sent to him. The Malayalam inikkulla (in-i-kU and ulla), my, is a compound of the dative of the personal pronoun (which is itself a compound), and a relative participial form of ul, within ; in colloquial Tamil, also, a similar form is used as a possessive. Possessive Compounds. — The Dravidian languages are destitute of that remarkable and very convenient compound of nouns and pro- nomin.al suffixes with a possessive signification which is so character- istic of the Turkish, Finnish, and other Scythian families. See Castren's " Dissertatio de Affixis Personalibus Linguarum Altai- carum." In Hungarian they form the following compounds of ur, master, with the pronominal fragments, used as possessives : — ur-am. my master. ur-aim, my masters. ur-ad, thy master. ur-aid, thy masters. ur-a, his master. ur-ai, his masters. ur-un-k, our master. ur-ain-k, our masters. ur-at-ok, your master. ur-ait-ok, your masters. ur-ok, their master. ur-ai-k, their masters. 308 THE NOUN These compounds are regularly declined like uncompounded nouns, in the usual way : e.g. — uramnah (ur-am-nah), to my master. urunknak (ur-un-Jc-nak), to our master. uraimnak (ur-aim-nak), to my masters. urainknak (ur-ain-k-nak), to our masters. The absence of possessive compounds of this nature in the Dra- vidian languages, notwithstanding their agreement with the Scythian group in so many other points, is remarkable : it is the only point in which any structural difference of a generic or class type appears to exist. In all the Dra vidian languages the possessive pronouns are prefixed to nouns, as in the Indo-European tongues, never postfixed, as in the Scythian. There is a class of words in the Dra vidian lan- guages compounded of a noun and a personal suffix, called con- jugated nouns, or appellative verbs. See the section in which these are explained. That class of words, though it resembles, is not identical with, the Scythian possessive compounds. It is identical, however, with the predicative compounds of the Scythian languages. SECTION III.— ADJECTIVES, OR NOUNS USED ADJECTIVALLY. The difference between the Indo-European languages and those of the Scythian group with respect to the formation and use of adjec- tives, is very considerable. The agreement of adjectives with the substantives which they quahfy, in gender, number, and case, forms an invariable character- istic of the languages of the Indo-European family ; whilst in the Scythian languages adjectives have neither number, gender, nor case, but are mere nouns of relation or quality, which are prefixed without alteration to substantive nouns. In this particular the Dravidian languages present no resemblance to the Sanskrit, or to any other member of the Indo-European stock, but are decidedly Scythian in character. Dravidian adjectives, properly so called, like those of the Scythian tongues, are nouns of quality or relation, which acquire the signification of adjectives merely by being pre- fixed to substantive nouns without declensional change ; and, in virtue of that acquired signification, they are called by Tamil grammarians uri chol, qualitative words. Participles of verbs, and FORMATION OF ADJECTIVES 309 nouns with tlie addition of participial fonnatives, are also largely- used as adjectives in the Dravidian, as in the Scythian, family. Such being the simplicity of the construction of Dravidian adjectives, it will not be necessary to occupy much time in the investigation of this department of grammar. It may suffice to state, seriatim, the various modes in which nouns or verbs are used as adjectives, and the formative or euphonic modifications which they undergo on being prefixed to the substantives which they qualify : nor will it be necessary to state all the modifications which are discoverable in each dialect, but only those which appear to be most character- istic, or which are peculiarly worthy of remark. 1. The majority of adjectives in all the Dravidian dialects are nouns of quality or relation, which become adjectives by position alone, without any structural change whatever, and without ceasing to be, in themselves, nouns of quality. Thus, in the Tamil phrases fon aridu, gold (is) scarce, and pon mudi, a golden crown, pon, gold, is precisely the same in both instances, whether used as a substantive in the first, or as an adjective in the second. In a similar manner, in English and the other modern Indo-European dialects, the same word is often used as a noun in one connection, and as an adjective without addition or change, in another connection — e.g., gold is more ductile than silver ; a gold watch. Whilst adjectival nouns of this class undergo in the Dravidian languages no structural change, their combination with the nouns to which they are prefixed is facilitated in certain instances by unimportant euphonic changes, such as the assimilation of the final consonant of the adjective and the initial consonant of the substantive, in accordance with the requirements of Dravidian phonetics (e.g., pot chilei (for fon silei), a golden image), the softening, hardening, or doubling of the initial of the substantive, or the optional lengthening of the included vowel of the adjectival noun, to compensate for the abandonment of the euphonic final u — e.g., kdr, black, in place of karu, or vice versa. These changes are purely euphonic ; they differ in the different dialects, and they contribute to grammatical expression only in so far as they serve lo indicate the words which are to be construed together as adjective and substantive. It is only on the ground of the repugnance of the Dravidian ear to certain classes of concurrent sounds that the changes referred to are inquired by Dravidian rules ; and in the majority of instances nouns sustain no change whatever on being used adjectivally. In the poetical dialects, adjectival formatives are less used than 3IO THE NOUN even in the colloquial dialects ; and it is generally the crude ultimate form of the noun of quality which performs the functions of the adjective in classical compositions. Thus, whilst nalla, good, and pala, many, are commonly used in spoken Tamil, the higher idiom prefers, and almost invariably uses, the crude nouns of quality and relation nal and 2)al — e.g., nal vari, the good way, and pan (for pal) 7nalar, many flowers. 2. Sanskrit derivatives (neuter nouns'of quality) ending in am in Tamil, and in amw in Telugu, become adjectives when prefixed to other nouns by rejecting the final m or mu — e.g., suham, goodness, and dinam, a day, become suha dinam, a good day. This, however, is in imitation of a Sanskrit rule, and it flows from the circumstance that when two Sanskrit nouns are formed into a compound, the crude form of the first of the two nouns is used instead of the nominative — suhha instead of suhJiam. Pure Dra vidian nouns ending in am or amu rarely become adjec- tives in this manner ; and when they do, it may be suspected that it is through imitation of Sanskrit derivatives. In Telugu, final amu is sometimes hardened into ampu — e.g., from andamu, beauty, is formed andapu or andampu, beautiful. In Tamil, when a noun of this class is used as an adjective, am is generally rejected, and attu, the inflexion, sufiixed instead — e.g., from puvam, externality, is formed purattu, external. Sometimes also Tamil deals in this manner with Sanskrit derivatives, converting them into adjectives by means of the inflexional attu ; but in all instances of nouns ending in am or amu, the most common method of using them adjectivally is that of appending to them the relative participle of the verb to become {ana, Tarn., ayana, Tel., or dda. Can.), without any change, whether structural or euphonic, in the nouns themselves. 3. Many Tamil nouns ending in S-u, d-u, nd-u, or r-u, double their final consonants when they are used as adjectives, or when case-signs are suffixed to them — e.g., compare ndd-u, Tam. the country, with ndtt-u varakkam, the custom of the country, or ndtt-il, in the country. (See the " Inflexional Increments.") From the corresponding Telugu ndd-u, the country, is formed ndti, of the country. In these instances the final consonant of the root is doubled and hardened (or in Telugu hardened only), for the purpose of conveying the significa- tion of an adjective ; but in another class of instances the root remains unchanged, and it is the consonant of the formative addition that is doubled. When Tamil nouns ending in the formative mhu are used adjec- FORMATION OF ADJECTIVES 3 II tivally, iribu changes into f'pu — e.g., from irumbu, iron, and kol, a rod, is formed iruppu {k)k6l, an iron rod. A similar change some- times takes place in Telugu, in which inumu, iron, becomes inwpa — e.g., inufa pette, an iron box. Tamil nouns ending in the formative ndu and du change in the same manner to ttu on being used as adjectives — e.g., compare marundu, medicine, and erudu, an ox, with maruttu {p)pei, a medicine-bag, and eruttu {p)podi, an ox-load. More rarely, nouns ending in the formative ngu change into kku both in Tamil and Malayalam — e.g., kurakku-(p)-padei, a monkey army, from kurangu, a monkey. These changes precisely resemble those which neuter or intransitive verbs ending in d-u or t-u (or with the formative additions of mb-u, ng-u, nd-u, &c.) undergo on becoming active or transitive, and a similar principle is in each instance appar- ent in the change ; for when nouns of quality are prefixed to other nouns adjectivally, th^re is a transition of their signification to the nouns which they are intended to qualify, which is analogous to the transition of the action of a transitive verb to the object which it governs. (See " Roots," and also " The Verb.") 4. Each of the inflexional increments, or petrified case-signs, is used for the conversion of substantives into adjectives. These are in in Tamil and ni in Telugu, attu in Tamil and ti or ti in Telugu. In those instances in which in in Tamil and ni in Telugu are used as adjectival forma tives, their use is optional — e.g., in Telugu we can say either tella, white, or tella-ni ; and in Tamil either niral, shady (literally shade, a noun used adjectivally), or (but in the poetical dialect only) niral-in. So also we may say either mara {k)koppu, Tam. the branch of a tree, or mar'-attu (k)koppu. In Tamil, am, an inflexional increment which is apparently equivalent to in, is often used as an adjectival formative — e.g., panan dSppu {panei-am toppu), a palmyra tope. The same formative is used in Malayalam also — e.g. , malam pdmbu {mala-am pdmbu), a rock-snake. It has been shown that the inflexions or inflexional augments attu and ti are in reality locative or possessive case-signs, and that they are used to convert substantives into adjectives through the relation subsisting between possessives — e.g., of gold — and adjectives — e.g., golden. In consequence of the frequency of their use in this con- nection, they have come to be appended even to adverbial forms for the purpose of giving to 4hem an adjectival meaning. Thus, from monna, Tel. before, is formed the adjective monna-ti {e.g., monna-ti tirpu, the former decision) ; and in Tamil, from vadakku, north (perhaps originally a dative), is formed the adjective vadakk'-att-u. 312 THE NOUN northern {e.g., vadaJcJcaUiydn, a northerner). In these and similar instances it is plain that the so-called adverbs are in reality only nouns used adverbially. 5. Relative participles of verbs, and nouns of quality converted into relative participles by the addition of participial formatives, are largely used as adjectives in all the Dra vidian languages. Much use is made of relative participles as adjectives by the languages of High Asia ; and in Japanese also participial forms of the verb are used as adjectives. It often happens that the same root is used, or at least is capable of being used, both as a verb and as a noun ; and hence, in many instances of this kind in the Dra vidian languages, two methods of forming adjectives are practicable, viz., either by prefixing the noun to the substantive which we wish to qualify, or by using one of the relative participles of the related and equiva- lent verb. The colloquial dialect of Tamil prefers the latter method ; the former is preferred by the poets on account of its greater sim- plicity and brevity. Thus, in Tamil either uyar, height (adjectivally * high '), or the relative participle uyarnda, high, literally * that was high ' (from uyar, considered as a verb signifying ' to be high '), may be used to express high or lofty — e.g., uyar maUi or uyarnda malei, a lofty hill : but uyar would be preferred in poetical compositions, whilst uyarnda is better suited to prose and colloquial purposes, and is consequently the form which is commonly used by the Tamil people. 6. The past verbal participle of Telugu verbs is sometimes used adjectivally in Telugu ; hence, when Sanskrit neuter nouns in am are used as adjectives, ayi, ' having become ' (the verbal participle), is often annexed to them instead of ayi-na (Tarn, ana, Can. dda, that became, that is (the relative participle). It seems evident, therefore, that the final i of many Telugu adjectives may be explained as identical with the i by which the past participles of verbs are formed — e.g., kindi, low, from kinda, below — e.g., kindi illu, the lower part of the house. The addition of the same i (if it be the same) converts substantives also into adjectives — e.g., from kun-u, a hump, is formed kuni, hump-backed. (See " Inflexional Increments," 7, i ; and " The Verb : Nouns of Agency.") 7. A very numerous class of Dra vidian adjectives is formed by the addition to crude nouns of quality of the sufiixes of the relative participles, more or less modified. Uyarnda is a perfectly-formed preterite relative participle, comprising, in addition to the verbal root, nd, the sign of the preterite tense, and a, the sign of the relative ; FORMATION OF ADJECTIVES 313 and though the idea of time is in this connection practically lost sight of, yet that idea is included and expressed. On the other hand, in the class of words now to be considered, the signs of tense are modified or rejected to correspond with their use as adjectives, and the idea of time is entirely merged in that of relation. It is words of this class which are commonly adduced by grammarians as specimens of qualitative words, or adjectives ; and, if the name can correctly be used at all in the Dravidian family of tongues, it is to this class that it is applicable. I am convinced, however, that it is more correct to regard these words simply as relative participles ; and I class them under this head, immediately after the investigation of the noun, because in most instances the root to which the relative signs are suffixed is used by itself, not as a verb, but only as a noun of quality or relation, or as an appellative. (1.) Many Tamil adjectives of this class are formed by the addition of iya to the root — e.g., feriya, great, siiiya, small. The roots of these words are per-u and sir-u ; and as u is merely a help to enun- ciation, I do not say that u is changed into i, but prefer to say that iya is added to the root. I have no doubt that we shall be able to explain each part of this addition grammatically, without having recourse to arbitrary mutations. These adjectives are simply the relative participles of " conjugated nouns." Iya (i-y-a) is com- pounded of i, a sign of the preterite tense, and a, the sign of the relative participle, with the addition of y inserted euphonically. In Telugu, the past participle alone is often used adjectivally without the suffix of the relative as we have already seen ; and the i with which taht participle terminates explains the i which precedes the final a of such Tamil adjectives as peri-{y)-a. i is the sign of the verbal participle, and the addition of a or ya, transforms it into a relative participle. In classical Tamil compositions, iya is generally used instead of ina, as the sign of the preterite relative participle of ordinary verbs — e.g., famiiya, instead of pannina, that made. When the same suffix is added to a noun of quality like fer-u, great, it converts it into a relative participle, which, with the form of the preterite, contains in it no reference to time, and which may there- fore be called an adjective. The suffix iya being somewhat archaic, readily loses the idea of time, whereas that idea is firmly retained by ida, ina, and the otker preterite relative suffixes which are in ordinary use. A good illustration of the adjectival use of iya is furnished by the very roots to which we have referred, viz., peru, great, situ, small. 314 THE NOUN When these roots are regarded as verbs, their preterite relative participles are perutta, that was or became great, sivutta, that was or became small ; in which participles the ideas of time and change are always included : whereas, when peru and siiu are regarded as nouns of quality, they are adapted for general use as adjectives by having iya suffixed to them — e.g., periya, sitiya {per'-iya, Ht^-iya). In this shape they mean simply great and small, without any reference to time ; and in consequence of iya being so purely aoristic, adjectives of this mode of formation are largely used, periya, great, kodiya, cruel, may properly be styled adjectives, seeing that they are used as such ; but it is a mistake to regard feriya-{v)-an, or periya-n, a great man, Jcodiya-n, a cruel man, and similar words, as adjectives. They are compounds of adjectives, and suffixes of gender ; and are properly appellative nouns, as has been shown under the head of " Gender," and as appears from the manner in which they are used. It is remarkable that a or ia is postfixed in K61 also to many adjec- tives ; and that the same participle is a sign of the possessive, as a is in Dra vidian. (2.) Some adjectives are formed by simply suffixing a, the sign of the relative participle, without the preterite i, or any other sign of tense whatever — e.g., nalla, Tam. good ; dodda, Can. great ; pedda, Tel. great. The examples here given may be, and doubtless are, derived from preterite relative participles [nalla from the High Tamil nalgiya, and dodda from the ancient Canarese doddida) ; but in some instances, a, the sign of the relative participle, is appended directly to nouns, without borrowing any portion of the sign of the preterite. We have an instance of this even in colloquial Tamil, viz., udeiya {udei-{y)-a), the ordinary colloquia-l suffix of the genitive, which liter- ally signifies that belongs to, that is the property of, from udei, pro- perty, to which a, the sign of the relative participle, is simply suffixed. This mode of forming adjectives from substantives by directly suffix- ing a is very common in the classical dialect of the Tamil, especially in connection with substantives ending in ei or t — e.g., from tnalei, a hill, comes malei-{y)-a, adj., hilly, or of a hill ; from sunei, a spring, comes sunei-{y)-a, that relates to a spring. So also from ti, evil, is formed ti-{y)-a, adj., evil. The circumstance that inmost of these examples the signification of the genitive is as natural as that of the adjective, shows how intimately the genitive and the adjective are allied. Nevertheless, as used in these examples, I regard a as an adjectival termination, rather than as a sign of the genitive, and as acquiring this force from its being the sign of the relative participle. FORMATION OF ADJECTIVES 315 Indeed, 1 would define these qualitative words {malei-{y)-a, &c.) to be the relative participles of appellative verbs. See that class of words investigated in the section on " The Verb." This usage, perhaps, explains the origin of the Tamil adjectives 'pala, many, and sila, few, viz., from the roots pal and sil (which are used in their crude state in the poets), and a, the sign of the relative participle. It is true that these words are also regarded as neuters plural ; and that in some instances they are correctly so regarded appears from the phrase folavin {fala-v-in) j)dl, the Tamil designation of the neuter plural, literally the gender of the many (things). But when we look also at such phrases as ^ala arasar, many kings — phrases of constant occurrence, not only in the colloquial dialect, but in the classics — the a of this latter fala appears to be used, not as a suffix of the neuter plural, or as a sign of plurality of any sort, but as a sign of the relative participle, by the use of which pal-a becomes an adjective. (3.) Many adjectives of this class are formed by the addition to nouns of quality of the sign of the relative participle of the future or aorist, which is um in Tamil — e.g., per-um, great. Native gram- marians suppose this adjective to be derived from the abstract noun perumei, greatness, by the rejection of the final ei, and to all other adjectives of this class they attribute a similar origin, mei, how- ever, not ei, is the suffix by which abstract nouns are formed (vide " The Verb "), and as such it is one and indivisible. It is much better to derive ferum from fer' , the uneuphonised form of the root "peru, greatness, great, and um, the ordinary relative participle of the aorist, in the same manner as periya has been seen to be derived from per and iya, the relative participle of the preterite, um is ordinarily called the relative participle of the future, but this future will be shown, in the part on " The Verb," to be properly an aorist, and as such to be used very indeterminately with respect to time. Vinnil minang-um sudar, Tam., means, not the stars that will shine in the sky, but the stars that shine in the sky, this tense being especially fitted to denote continued existence ; and in consequence of this looseness of reference to time, wn, the sign of the relative participle of this tense, is better fitted even than iya to be suffixed to nouns of quality as an adjectival formative. Hence perum, literally that is, was, or will be great, is £b more expressive and more classical word for great than periya. It has already been shown, in the part on " Sounds," that peim, Tam. green, is not a distinct form of adjective, but is softened from pa^um (payum) by a dialectic rule, whilst pasum 3l6 THE NOUN is derived regularly from pas-u, greenness, green, and idh, the particle which is now under consideration. 8. Dra vidian nouns of every description may be used adjectivally by appending to them the relative participles of the verb signifying to become, which are in Tamil ana and dgum (also uUa, an equivalent word), in Telugu agu and ayina (pronounced aina), in Canarese dda — e.g., uyarvdna {uyarv'-dna), Tarn, lofty, literally that was or has become high or a height. This mode of forming adjectives is especi- ally used in connection with Sanskrit derivatives, on account of their greater length and foreign origin. Such adjectives, however, are phrases, not words ; birt they were at one time incorrectly classed amongst adjectives by Europeans who treated of Dra vidian grammar. I may here also again remark, that certain words have been styled adjectives by some European writers, which in reality are appellative nouns, not adjectives, and which acquire the force of adjectives merely from the addition of the relative participles of the verb to become, which have been referred to above. Thus, the Tamil words nallavan, a good (man), nallaval, a good (woman), nalladu, a good (thing), are appellative nouns formed by the suffix to a noun of quality of the formatives of the three genders ; and the addition of dna, that has become, to any of these words, though it constitutes them adjectives in effect, leaves them in grammatical form precisely what they were before. Bonus may either qualify another noun — e.g., bonus vir, when it is an adjective, or it may stand alone and act as nominative to a verb, when it is a qualitative noun — e.g., bonus virtutem amat. The Tamil nallavan, a good (man), can only be used in the latter sense, and therefore is not an adjective at all. Comparison of Adjectives. — In all the Dravidian dialects, com- parison is effected, not as in the Indo-European family, by means of comparative or superlative particles suffixed to, and combined w^ith, the positive form of the adjective, but by a method closely resembling that in which adjectives are compared in the Semitic languages, or by the simpler means which are generally used in the languages of the Scythian group. When the first of these methods is adopted, the noun of quality or adjective to be compared is placed in the nomina- tive, and the noun or nouns with which it is to be compared are put in the locative and prefixed. It is generally stated in Tamil grammars that it is the ablative of motion which is thus used ; ])ut I am per- suaded that even when the case-sign is that of the ablative of motion, the signification is purely that of the locative, and that in Tamil ii COMPARISON OF ADJECTIVES 317 and m have in this connection the meaning of in {i.e., are locatives), rather than that oi from — e.g., avattr'-il idu nalladu, Tarn, this is better than those, literally in those things this is good. The conjunctive particle um, and, even, is often added, especially in the colloquial dialect, as an intensitive — e.g., avattr'-il um idu nalladu. Tarn, this is better than those, literally even in those this is good. Very frequently the noun with which comparison is to be made is put in the dative instead of the locative. Sometimes, again, comparison is effected by means of an auxiliary verb. The noun with which comparison is to be made is put in the accusative ; it is followed and governed by the subjunctive or infinitive of a verb signifying to see, to show, or to leave ; and the phrase is concluded by the subject of the proposition, with the adjective to be compared. Thus, in Tamil we may say adei-(f)-'pdrhkilum idu nalladu, literally even though looking at that, this (is) good, or adei vida idu nalladu, quitting that, this (is) good, i.e., this is so good as to induce one to abandon that. Such modes of comparison, however, are stiff, cum- brous, and little used except by Europeans ; and in the Dravidian dialects, as in those of the Scythian group, direct comparison of one thing with another is ordinarily left to be understood, not expressed. The effect which is aimed at is secured in a very simple manner by prefixing to the positive form of the adjective some word signifying much or very, or by appending to the subject of the proposition a sign of emphasis, or a word signifying indeed — e.g., id-e (or idu tan) nalladu, Tam. this indeed is good. In Telugu and Canarese the conjunctive particles u and u are not necessarily required to help forward the former method of comparison, like the Tamil um, nor is this particle generally used in the higher dialect of the Tamil itself. The Canarese makes use also of the particles anta and inta, antalu and intalu (which, in their origin, are compounds of locatives and demonstratives), to assist in effecting comparison. In all these dialects the superlative is generally expressed by means of prefixed adverbs signifying much or very, or by the very primitive plan of doubling of the adjective itself — e.g., periya-periya, very great, literally great-great. If greater explicitness is required, the method by which it is effected is that of putting the objects with which comparison is made in the plural and in the locative case. Thus, the phrase, the tigw is the fiercest animal, would be expressed in Tamil as follows : — vilangugalil vengei kodidu, amongst animals (literally in animals) the tiger is the cruel one. Sometimes, for the purpose of increasing the intensity of the superlative signification, 3l8 THE NOUN the adjectival noun elld, all, is prefixed to the plural noun which denotes the objects compared — e.g., in (i.e., amongst) all animals the tiger is cruel. It is evident that the modes of forming the comparative and super- lative degrees of adjectives which have now been described, differ greatly and essentially from those which characterise the Indo- European family of tongues. If Dravidian adjectives had ever been compared, like those of the Sanskrit, it is inconceivable that so con- venient and expressive a plan should so completely have been abandoned. The Dravidian modes of comparison agree, up to a certain point, with those of the Semitic tongues ; but they are in most perfect accordance with the Turkish method, and with the modes of comparison which are employed in the languages of Tartary generally. Robert de Nobilibus and the Jesuit writers endeavoured to naturalise in Tamil the Sanskrit superlative particle tama, but the Tamil adhered resolutely to its own idiom, and the attempt failed. Postpositions. — It has already been stated that all the Dravidian postpositions are, or have been, nouns. When suffixed to other nouns as postpositions, they are supposed to be in the locative case ; but they are generally suffixed in their uninflected form, or in the nominative ; and the locative case-sign, though understood, is rarely expressed. It seems quite unnecessary to enter into an investigation of the postpositions in a work of this kind, inasmuch as they are sufficiently explained in the ordinary grammars, and are to be regarded simply as nouns of relation. d o -^ — ' c8 o .1 I i S S %* s ??> ^ >*^ • 5J 1 1 S t 5g !?^ s ^ K •^r>« 5>. v" ;s .S I ^S 5^ 4 to 1 50 « 1^ 03 03 o +3 % "A C^ c3 8 ^ .S o .9 -^^ a 45 .a ?:; Ph o o -^ 6 S 4 . l-S 8 o S o 53 ^ r^ « ?^ ?> ?^ ^ s §: ^ ^ ^. o . •^ !3 5» S^ ^ cs' ^ 1 fe.^ ^ ^ ^ ?^ ^ » 2g I&5 ^ I: I ^ i ^ I: ^ i 4 t 1 > (— 1 i 4 ii w •^li) ^ ^ e • '"S^ W •f ^. ^. s^ ^-^ - •. 5g H ^. 1 ^ 1-^ ^^ !Z5 O ^ §: §S S ^e Mlii ^ s^ 5e H-t -— -v^-ii^ — .^^--w- CZ2 ^^ < S 1 Ph S' § 1= p^ '^ ^* A ^ <-< ^o -<=- ,^- o § HH i ^- ^i ^'^ il g i ^ g 1 1 ^§ E 1- 1 I ^s ^^' ■ii* o $s 5i t-H > ^-— ■— -V —— -— ^ .^ P i ^ • 4 t -¥ h4 ^^ 't.5f - .. ^g ?iH s § ^* i .« > CD a.> O H H Ph Ph CZ2 ^ W '^ H 358 PART V. THE PRONOUN. Much light is thrown by the pronouns on the relationship of lan- guages and families of languages ; for the personal pronouns, and especially those of the first and second persons singular, evince more of the quality of permanence than any other parts of speech, and are generally found to change but little in the lapse of ages. They are more permanent even than the numerals, the signs of case, and the verbal inflexions ; and though, like everything else, they are liable to change, yet their connections and ramifications may be traced amongst nearly all the languages of mankind, how widely soever sundered by time or place. In some instances the personal pronouns constitute the only appreciable point of contact or feature of relation- ship between languages which appear io have belonged originally to one and the same family, but which, in the lapse of time and through the progress of mutation, have become generically different. This remark especially applies to the pronouns of the first person, which of all parts of speech appears to be the most persistent. A remarkable peculiarity of the Japanese is the absence of personal pronouns, properly so called. Usage alone determines which of the three persons is denoted ; as in English, it is usage that deter- mines that * your servant ' means I, and ' your honour,' you. SECTION I.— PERSONAL PRONOUNS. 1. Pronoun of the First Person Singular. Comparison of Dialects. — Our first inquiry must be, what appears to have been the primitive form of this pronoun in the Dravidiaji languages ? A comparison of the forms it assumes in the different dialects may be expected to throw much light on this question. It will be well to exhibit the facts of the case first, with only such explanations as seem tc^be necessary, reserving to the end the consideration of the inferences which the facts appear to establish. I must here remind the reader of what I have said in the Intro- duction respecting the relation subsisting between the classical and 359 360 THE PKONOUN colloquial dialects of the principal Dravidian languages. There is a presumption in favour of the antiquity of words and forms found in the literature of those languages, especially when found in the gram- mars and vocabularies, which are at least seven or eight hundred years old, and are regarded as works of authority ; but on the whole it is safer to regard those words and forms, not as necessarily more ancient, but only as probably more ancient, and certainly more classical. In citing those dialects, therefore, I shall cite them, not, as has generally been done, under the names of the ancient and the modern dialects, but as the classical and the colloquial. It will be seen that in all cases I compare, not only the nominatives of the personal pronouns found in the various dialects, but also the inflexional bases of the oblique cases and the pronominal termina- tions of the verbs. The base of the oblique cases of the pronoun of the first person in the Indo-European languages seems altogether un- connected with the base of the nominative. In the Dravidian lan- guages it is evident that the nominative and the inflexions of this and of all pronouns are substantially the same. Differences, it is true, are apparent, but they are comparatively insignificant, and are generally capable of being explained. Where the inflexion differs from the nominative, but agrees with the verbal endings, we may reasonably suppose the inflexion a better representative than the nominative of the oldest shape of the pronoun. In most of the dialects, the included vowel of each of the personal pronouns is long in the nominative, short in the inflexion. In such cases, the inflexion might be supposed to be an abbreviation of the nominative, made for the purpose of enabling the base to bear the weight of the case- signs. On the other hand, as in the Dravidian languages the nomina- tive of the personal pronouns is only used when it is emphatic, the lengthening of the included vowel of the nominative may be regarded merely as the result of emphasis. On the whole, the latter supposi- tion seems preferable. (Compare the lengthening of the vowel of several of the numerals, when used not as adjectives, but as sub- stantives.) It seems desirable also to compare the plural forms of this pronoun with the singular. The mode in which the personal pronouns are pluralised will be explained under a separate head ; but the plural forms themselves will be cited here, for the sake of the light they may be expected to throw on the initial consonant and included vowel of the singular. In all cases it will be found that the ultimate base of the singular and that of the plural are identical. Unlike the Indo-European tongues, as best represented by the FIRST PERSON SINGULAR 36 1 Vedic Sanskrit, in which the plural of the first person has the force of ' I and they,' and that of the second person ' thou and they,' the plurals of the Dravidian languages seem to be simply the singulars with the addition of suffixes denoting plurality. The reader is requested to remember ( see note on Transliteration, preceding Sounds) that in most of the Dravidian dialects y has come to be pronounced before initial e — e.g., in Tamil, en, my, is pronounced yen. This y (and the corresponding v or iv before 0) has frequently made its appearance in the transliteration into the Roman character of words commencing with e, and sometimes even in cases where a comparison of dialects was the object in view. No notice will be taken of this euphonic y of pronunciation in the following analysis. I cite each word as it is written by the best classical waiters, believing that the written form of the word best represents the manner in which it was actually pronounced when the language was first committed to writing. If y appears anywhere in this analysis, it is because in that instance y has a place in the written language, and appears to be radical. In colloquial Tamil the nominative of the pronoun of the first person singular is nan : in classical Tamil it is ydn or nan, more commonly the former. The " Nannul," the most authoritative grammar of this dialect (the date of which cannot, I think, be later than the eleventh century), gives both forms, ydn and nan, but always places ydn first. This proves nothing, I think, respecting the relative antiquity of the two forms ; it only proves that ydn was regarded by the author of the " Nannul," as it is still regarded, as more elegant than nan. The inflexion of this pronoun in both dialects is en. It is here apparent, and will be seen in all the other dialects also, that the included vowel vibrates between a and e. The personal terminations of the verbs are en in the colloquial ; and en and en, and occasionally an, in the classical dialect. (I omit all consideration of those forms of the Tamil verb which, though regarded by native grammarians as belonging to the first person singular and plural, are in reality impersonal.) The corresponding plurals are — nom. colloquial, ndm, ndngal ; classical, ydm or ndm : inflexion, coll. nam, engal ; class. em, nam. The nom. ydm is more common in the classics than ndm ; but in the inflected forms nam is regarded as nearly, if not quite, as elegant as em — e.g., nanmr=emar, our party, nostrates. In the classical compound eldm, all we, corresponding to eltr, all you, the plural nom. is dm. Personal terminations of the verb — coll. 6m ; class, em, em, am, dm, 6m, 362 THE PRONOUN At first sight we might suppose nam and nem to be the pronominal terminations of the class. Tam. nadandanam,nadandanem,we walked, and of many similar verbs and conjugated nouns — nouns with which a pronoun is combined (see " Classification of Dravidian Verbs," " Appellative Verbs or Conjugated Nouns ") ; but the n of these ter- minations is merely euphonic, and is used to prevent hiatus. When it is omitted, the vowels which it had kept separate coalesce — e.g., nadanda-am becomes nadanddm ; nadanda-em, nadandem. The termination 6m is the only one now used in the colloquial dialect. This could not well have been derived from em, but would spring naturally enough from dm. Of this we have an illustration in the fact that dm, contracted from dgum, or d-um, it is so, yes, is some- times written, as well as pronounced, 6m. Moreover, whilst many instances of the change of a into e or ei, and also 0, can be adduced, I do not know any of the converse of this. In Malayalam the nominative is ndn (the initial n of which is the nasal of the palatals, pronounced like ni in onion). The inflexion is ordinarily en, as in Tamil ; but in the dative inikk' is often used, as well as the more regular enahk' and enikh\ en is here altered to in, a form which I do not find in any of the other cultivated Dra- vidian dialects. The verb in ordinary Malayalam is destitute of personal terminations ; but in the poetry an inflected form is fre- quently used, in which the termination representing this pronoun is en, as in Tamil. In conjugated nouns the personal termination, as an or en — e.g., adiyan or adiyen, I (thy) servant ; plural nom. nam, n6m, nammal, nannal, nummal ; inflex. nannal, ennal, em, and also no, n6m, nom, num. Personal terminations of verb (in the poets), 6m. The shortness of the included vowel of nannal, and the ordinary use of this form, rather than of ennal, as the inflexion, are noticeable peculiarities in the Malayalam plural. Another peculiarity is the occasional use of nom instead of ndm, answering to the 6m which forms the personal termination of the verb in poetical Malayalam and colloquial Tamil. In colloquial Canarese the nominative of this pronoun is ndnu, nearly as in Tamil, the inflected form of which, as seen in all the oblique cases, is nan' . The crude form of this pronoun nd is also used as a nominative. This is a peculiarity of Canarese and Telugu ; but the use of ni, the crude form of the pronoun of the second person, instead of ninu, has its counterpart in Tamil, in which ni is the only form of the nominative known. In the classical dialect, or what is commonly called " Old Canarese," the nominative is an, ydn, or dm ; FIRST PERSON SINGULAR 363 the inflexion, en, is identical with that of the Tamil in both its dialects. The pronominal terminations of the first person singular of the verb are enu, enu, and ene in the colloquial dialect, and en in the classical. It is deserving of notice that the final u or nu of the personal terminations, as of the isolated pronouns, is frequently dropped in the colloquial dialect. The personal termination of this person of the verb, when nu is dropped, becomes e, with which the Tulu termination may be compared. Plurals : nominative, coll. dial. ndvu ; class, dial, dm, dvu ; inflexion, coll. nam ; class, em. Personal terminations of verb : coll. evu, evu, and eve ; class, evu. evu is as clearly a softened form of em as dvu of dm. In colloquial Telugu the nominative of this pronoun is nenu : the crude ne may also be used, like nd in Canarese. In the classical dialect, enu is preferred, and this is sometimes represented by e alone. nenu takes nd for its inflexion in all cases except the accusative {nanu or nannu), in which it is nan', as in colloquial Canarese. It appears from this that the vowel of the pronominal base vibrates between a and e, but that e is probably to be regarded as the more ancient, as well as the more elegant form, in so far as Telugu usage is concerned. The verbal inflexions of the Telugu retain only the final syllable of the nominative of each of the pronouns — viz., nu or ni after i (from nenu, I) ; vu or vi after i (from nivu, thou) ; and ndu (from vdndu, he). Plurals : nominative, coll. memu, manamu ; class, emu ; in- flexions, md, mam, nmna ; personal termination of verbs, mu, or mi after i. The most essential part of the personal pronouns has been dropped, we see, in the verbal inflexions of the Telugu, the fragments which have been retained being probably merely formatives, or at most signs of number and gender. Of the same character is the ru, or ri after i, which forms the personal termination of the second person plural and the third person epicene plural. It represents merely the ar by which epicene nouns are pluralised. The Tulu nominative is ydn' ; inflexion, yen\ This is the only instance in any of these dialects in which y, the initial letter of the nominative, appears in the inflexion in writing. In classical Canarese and Tamil the inflexion is written en, though pronounced yen. The personal termination of the verb is e (compare the colloquial Canarese verbal termination e, and the classical Telugu nomi- native e). This e, Mr §rigel informs us, is pronounced nearly like a in man ; whilst the e which forms the termination of the third person masculine of the verb is pronounced pure. Plurals : nomina- tive, naitm, yenJcalu ; inflexion, nam\ yenhu]'. The included vowel 364 THE PRONOUN of 7iama is short in the nominative, as well as the inflexion. The only- instance of this in the other dialects is nammal, ©ne of the Malayalam nominatives, and its related nannal. Personal termination of the verb, a. The personal terminations of the first person plural and the third person neuter plural (both a) are alike, which is a remarkable peculiarity of this dialect. The Tuda nominative is an {a is pronounced in Tuda like the Eng- lish aw) ; inflexion, en ; personal termination of verb, en, eni, ini ; plural nom. dm or 6m, also em ; inflex. em (the nominative dm is also used, according to Dr Pope, like an inflexion). Mr Metz writes this not dm, but am, which is more in accordance with analogy. Personal terminations, emi, imi. In the dialect of the Kotas, according to Mr Metz, the nominative singular is dne ; inflexion, en ; plural nom. dnfie, eme, and also ndme ; inflex. em, 71am ; personal terminations, singular, e, as in Tuju ; plural, eme and eme. In Gond the nominative is anna ; inflexion, nd ; plural, ammdt ; inflexion, md. Personal terminations of the verb : singular, dn or na ; plural, dm, am, or 6m. In the Ku or Khond the nominative singular is dnu, as in classical Canarese ; inflexion, nd, as in Telugu and Gond (Dr Hunter's lists, dnu ; inflex. ndncle) ; plural nom. dmu ; inflex. md ; also dju ; inflex. ammd. Personal termina- tions of verb : singular, in or in {mdin, I am), or e {mdsse, I was) ; plural, dmu. In the Brahui the nominative is ^ ; but in the oblique cases [e.g., hand, of me ; kane, me, to me) the pronominal base is ka or kan, a root which seems to be totally uncomiected with the Dravidian ndn or ydn, and which is to be compared rather with the Cuneiform- Scythian, Babylonian, and Gujarathi ku, liu, &c. The plural of the first person, nan, is on the whole in accordance with the Dra- vidian pronoun. The verbal inflexion of the plural is en — e.g., aren, we are. In the Kajmahal dialect, I is en ; mine, ongki ; we, nam, om ; our, emki, ndm-ki. tJraon, I,' enan ; mine, englii ; we, em (Dr Hunter, en) ; our, emhi. We have now to determine, if possible, from a consideration of the facts elicited by this comparison, what was the primitive form of the Dravidian pronoun of the first person. In the first edition, I said, " The weight of evidence seemed to be in favour of our regarding ndn, the Tamil nominative, as the best existing representative of the old Dravidian nominative of this pronoun, and nd, the crude form of the Canarese, as the primitive unmodified root." In coming to this con- FIRST PERSON SINGULAR 365 elusion, I was much influenced by the extra-Dravidian relationships of this pronoun, which, as will be seen hereafter, are strongly in favour of nan, as against ydn. Viewing the question, however, from a purely Dravidian point of view, the conclusion I arrived at did not seem to me quite satisfactory ; and the passage cited above had hardly been printed ere I wished I had decided in favour of ydn. I did not suppose, however, that when we arrived at nan (or ydn), the earliest organic development of this pronoun, we had reached a point in its history beyond which we could not go ; for it seemed to me, and still seems, probable that the final n is only a formative, denoting the singular number, and that the initial n (corresponding as it does with the initial n of the pronoun of the second person) is another formative, denoting in some way personality ; whilst it is by means of the included vowels (a and i) alone that the pronoun of the first person is to be differentiated from that of the second. In consequence of this, I thought I could recognise in those included vowels (a and i) the very earliest shape of the Dravidian pronoun. Dr Gundert considers ydn as probably older than nd7i. This is also Dr Pope's view, though in his " Outlines of Tuda Grammar," p. 5, he says, very truly, I think, " The original form of the Dravidian pronoun of the first person is uncertain." The late Mr Gover, in a paper on the " Dravidian Pronoun," of which he was so kind as to send me a privately printed copy, advocated ydn as against ndn, but further on rejected the y also, as probably not primitive, and adopted an or en as the real base. It was necessary to his theory to regard the final n as primitive, being derived, as he supposed, from the m of the Aryan ma (changed first, he thought, to na, and then to an). Dr Pope seems to concur in Mr Gover's view of both of the initial letters and of the final n (though for a different reason), when he says in his " Outlines," p. 5, " I would compare dn with the very ancient Sanskrit aham'' I conclude that both Dr Pope and Mr Gover may be cited, not only in favour of ydn, as against ndn, but also in favour of dn, as against ydn. This latter point may be considered first. Which is to be regarded as the older form, ydn or dn ? A change of ydn into dn seems to me much easier and more natural than a change of dn into ydn. But in this instance we are not left to mere abstract probabilities ; parallel cases can be adduced, an(d that from the list of pronouns and pro- nominals. The Tamil dr, who ? epicene plural, has undoubtedly been softened from ydr, and that from ydvar ; and this is quite certain, because both the changed form and the unchanged are still 3^6 THE PRONOUN in daily use ; the only difference is, that the older form is considered more elegant. We have another instance in dndu, Tam. a year, which is properly ydndu, when ? a year, from the same interrogative base ya. ydndu is the form of this word invariably used in inscriptions of any antiquity. The ease with which ya would change into a may be con- cluded also from the ease with which it has changed into e, an instance of which we have in the change of the interrogative pronoun already cited, ydvar, not only into ydr and dr, but also into evar. It is evident from these facts that yisa, particularly changeable letter, and there- fore that dn may safely be regarded as a softened form of ydn. The next point to be considered is, what is the relationship of ydn to nan ? I refer here to the initial consonant alone, not to the difference between the Tamil ndn, ydn, and the Telugu nenu, enu. That difference consists in the included vowel, and will be considered afterwards. As I have already said, it appears to me now that ydn is probably older than ndn, but ndn also I consider as of great antiquity. It is quite clear that there is a tendency in the Dravidian dialects, especially in Tamil and Malayakm, to convert y into n. Several words which begin with n or w in Tamil begin with a vowel in other dialects. Comp. Tam. nindu, to swim, with Tel. tdu ; Tam. and Mai. nandu or nandu, a crab, with the Tel., Can., and Tulu endi, entri, yandri. In these cases, however, it cannot be determined with certainty whether the initial n of the Tamil may not have been radical. Clearer evidence might perhaps appear to be furnished by the relative participles of the preterite Tamil verb, which may take either y or n — e.g., solliya or sonna (for sollina), that said ; with respect to which it might be concluded that y, being considered more elegant, is also more ancient. This, however, seems to me doubtful, seeing that the use of n, as in this case, to prevent hiatus, is capable of being traced back to a very early period in the history of the language. The only instances of the change of y into n that are quite reliable are those that are seen in Sanskrit tadhJiavas. The Sanskrit yuga, a yoke, is ordinarily in Tamil nugam, sometimes ugam. The Sanskrit Yama, the god of death, though ordinarily yaman, is also found, especially in the poetry, as naman, naman, and eman. Here we have indubitable instances of the changeableness of y. It is evidently liable both to be hardened into n, and also to be softened away into a vowel. We see therefore the possibility of a primitive Dravidian ydn changing on the one hand into ndn, and also on the other into dn or en. What seems to raise the possibility in this case into a probabiHty is the circumstance that the en, which forms the FIRST PERSON SINGULAR 367 only inflexion of this pronoun in the classical dialects of Tamil and Canarese, could much more easily be weakened from ydn than from nan. This is partly in consequence of y being more easily softened away than n ; partly in consequence of the peculiar tendency in the Dravidian languages to pronounce y before e, so that en would naturally be pronounced yen, and would therefore naturally connect itself with ydn. It is curious also that yd seems to have a special tendency of its own to change into e, as we have seen in the case of the interrogatives — ydvar, Tam. who ? which becomes evar ; ydngu, where ? which becomes engu. The change of ya (short) into e in Tamil may also be illustrated from Sanskrit tadbhavas. y antra, a machine, becomes endiram ; yajamdna, a sacrificer, a master, esamdn. There is an ulterior tendency in Tamil to change a into e, which will be illustrated further on, in considering the included vowel of this pronoun. The change of ydn into nan would be facilitated if we should take the Malayalam ndn, as I think we fairly may, as the middle point. If y were usually pronounced with a slightly nasal sound, it would naturally become n ; and this would naturally harden in some instances into the n of the dental series, possibly even into n and m. We have seen in the course of our comparison of the different Dravidian dialects that the initial n ov n oi ndn, nenu, ndn, has entirely disappeared in the verbal inflexions. The final n, whatever its origin, has shown itself more persistent ; though it also, as we shall see, sometimes disappears ; but in none of the dialects has the initial n or n, or any relic of it, been retained in the personal ter- minations of the verb. I think it unsafe, however, to conclude from this, or from any of the facts mentioned, that the initial n of ndn is of modern origin, ndn may have been altered from ydn, as I think it was, and yet the alteration may have taken place at so early a period, and both forms may have continued so generally in use, that the question to be considered is not so much, which is ancient, and which is modern ? as, which is to be regarded as the best repre- sentation of the primitive form of the word ? It would not be correct to say that the initial n is not contained in any of the old forms, or that it has disappeared from every ancient dialect, ndn is repre sented, as we have seen, as alternating with ydn in the most authori- tative grammar of the classical Tamil ; and whilst the singular in- flexion is always en, the plural may be either em or nam. nam is found in Tamil compounds of high antiquity, like namhi (comp. embi), lord, literally, our lord, nd or nan is the inflexion of the 368 THE PRONOUN singular in Telugu, colloquial Canarese, Ku, and Gond. In Malay- alam nan is the most common form of the nominative, though ydn also is known, and the ft of nan is lost in the inflexion. In Tulu the plural is nama. The Telugu plural memu has plainly been derived from nemu. These deep-seated traces of the use at one time of a nominative in nan, contemporaneously with one in ydn, in the dialects of people so long and so widely separated from one another as the Ku and the Tamil, the Gond and the Malayalam, seem to carry us back to an antiquity far greater than that of any of the so- called ancient dialects. The classical compositions commonly called ancient carry us back not much more than a thousand years ; but we must go back perhaps three times that period before we reach the time when the ancestors of the existing Tamilians lived side by side in the plains of Northern India with the ancestors of the existing Gonds. At that time, whenever it was, 7idn may be concluded to have been in use as well as ydn ; but even then ndn appears to have been a secondary form ; ydn, the more characteristic and authori- tative. An excellent illustration of the admissibility of this hypo- thesis may be derived from Sanskrit. It is commonly asserted, and may perhaps be admitted to be a fact, that the Vedic asme, we, is older than vayam, the corresponding word in use in the later litera- ture. The use of asme in the Vedas is one argument for its antiquity ; another and still better is its appearance in Greek in the shape of a/x/x€? (ammes). But we must not too hastily assume that, because vayam appears in the later Sanskrit literature, whilst asm£. is found in the earliest, vayam is therefore a modern corruption ; for we find {va or ve) the base of this form not only in the Zend vaem, but also in the Gothic veis (English, we) ; and this carries us back to the period — a period of unknown antiquity — when the Teutonic tribes had not yet left their early seats in the East. The reappearance in the plural, in the Pali-Prakrit tumhe, you, of the tu out of which the yu of yushme and yuyam was corrupted, after it had wholly disappeared from every other form of Aryan speech, is another case in point, as tending to prove that an old form may be retained in existence, and, to a certain extent, in use, long after another form has supplanted it in popular favour. The antiquity of one form is evidently there- fore no valid argument against the antiquity of another. In a discussion of this kind, it should not be forgotten that the pronouns of the first and second persons in all the Dra vidian dialects are evidently formed on the same plan. They have been exposed to the same influences, and have changed in nearly the same degree. FIRST PERSON SINGULAR 369 Dr Pope (" Outlines of Tuda Grammar "), who considers the initial n of nan, I, a late addition, thinks the initial n of nin (or w^), thou, undoubtedly radical. If, then, n is to be regarded as undoubtedly- radical in m, though it disappears in most of the inflexions, and in the personal terminations of all the verbs, and though even the nominative becomes i in Tulu and ivu in poetical Telugu, may we not conclude that the initial n of nan, I, though not radical (I have never claimed for it that distinction), carries us back to a period in the history of the language beyond which we can do little more than guess our way ? What was the included vowel of the primitive Dravidian pronoun ? We have only to choose, I think, between a and e. 6 is found in the plural in some connections in Tamil and Malayalam, but it is derived, as I think I have shown, from the a of dm. The i which makes its appearance in a solitary instance in Malayalam is quite exceptional, and seems to be the result of attraction, en, which occupies so impor- tant a place in almost all the dialects, both in the inflexion and in the verbal terminations, seems to point to a nominative in en, the best representative of which is the classical Telugu enu. On the other hand, in the greater number of the dialects, including both the culti- vated dialects in Southern India and the uncultivated dialects in the hills in Northern India, the nominative is nan or an. a, I think, is to be preferred, on account of the existence of a tendency in almost all languages, and particularly in the Dravidian, to weaken a into e, whilst I cannot discover any distinct trace of the existence of the contrary tendency. The tendency of the Tamil to weaken a into e may best be illustrated by Sanskrit derivatives, inasmuch as in these cases we know which vowel was the original, and which was the cor- ruption. Some have been quoted already, as showing the tendency of ya in particular to change into e ; but the following examples, in connection with other consonants, may be added — e.g., japa. Sans. prayer, Tam. sebam ; hala, Sans, strength, Tam. belam. This ten- dency shows itself in the pronunciation of many Sanskrit words used in Tamil in which the vowel remains unaltered in writing. I should add that Dr Gundert appears to consider not ya, but ye, euphonised to ye, the primitive form of this pronoun. He admits, however, that e is only another form of a. What is the origin of the final n of ydn, nan, &c. ? Whatever be its origin, it seems to me certain that it is not radical. It is more persistent than the initial n, but in the plural it is uniformly rejected, and m (probably from the copulative urn), the sign of plurality dis- 2 B 370 THE PRONOUN tinctive of the personal pronouns, used instead. This sign of plu- raUty is not added to n, as it would have been if n had been regarded as a part of the root, or even as a help to the expression of the idea of personality, but substituted for it. If we compare nan, I, with nam, we, nin, thou, with nim, you, tan, self, with tdm, selves, it is evident that the final n is a sign of the singular number, and the final m a sign of the plural. The pronominal base is evidently the same in both numbers ; and the certainty of this is not affected by any question that may arise as to the shape of the oldest form of the pronominal base. If we regard ydn as more primitive than nan, the conclusion we come to must be the same, the plural of ydn being yam. This appears to prove that nd (or yd) denotes either I or we, according to the singularity or plurality of the suffixed particle {nd-{-n —1 alone ; nd-\-m =I's (ego que) we) ; and that the final n of nan, no less than the final m of ndm, is a sign, not of personality, but merely of number. Is the final n of ndn a sign of gender as well as of number ? Is it a sign of the masculine singular, and connected with an or n, the ordinary masculine singular suffix of the Tamil ? The pronouns of the first and second persons are naturally epicene, but it is not unusual in the Indo-European languages to find them assuming the grammatical forms of the masculine. Thus in Sanskrit the ter- minations of the oblique cases of the pronouns of the first and second persons, are those which are characteristic of the masculine gender. I am not inclined, however, to adopt this explanation of the origin of the final n of the Dravidian personal pronouns. I am not satisfied, either, with the supposition that this final n is merely euphonic, like the final nasal of the Tartar nmn, I. The explanation which appears to me to suit the facts of the case best is, that this n is identical with the an, alternating with am, which is so largely used, especially in Tamil and Malayalam, as a formative of neuter singular nouns — e.g., ur-an, Tam. strength =wr-am. It would thus accord in use (possibly in part even in origin) with the final am of the nominative of the Sanskrit personal pronouns, ah-am, I, tv-am, thou, svay-am (sva-m), self (compare Greek lywv) (egon), which is evidently a formative, and identical with one of the most common nominative and accusative singular neuter case-signs. (See " The Noun : the Nominative.") Compare the optional use of m instead of n, as the final consonant of the pronoun of the first person in classical Canarese — e.g., dm, I, instead of dn. So also the same dialect has avam for he, instead of avan. FIRST PERSON SINGULAR 371 am, the formative of the nominative of the Sanskrit pronouns, is used not only by the singulars, but, in later Sanskrit at least, by the plurals — e.g., vayam, we, yuyam, you ; but properly these plurals are to be regarded as abstract neuter singulars in form, though plurals in signification. The Dra vidian formative am or an is exclusively singular. Whatever be the origin of the final n in question, it must have had a place in the personal and reflexive pronouns from a very early period, for we find it in the Brahui ten, self (compare Dravidian tan), and in the Ostiak nyn, thou (compare Dravidian nin). This throws light on the probability of the supposition I advanced with regard to the initial n of nan — viz., that though nan was apparently de- rived from ydn, the date of its origin might be far earlier than that of any portion of the literature which is written in what are some- times called the ancient dialects. If, as we have seen, nd or yd is to be regarded as the primitive form of the Dravidian pronoun of the first person, and the final n as merely a sign of number, it might appear extraordinary that in the pro- nominal terminations of the verb the initial n (or y) should have invariably and altogether disappeared, whilst the first person singu- lar should be represented, either by the final 7i alone, or by the fragmentary vowel e alone. Similar anomalies, however, are dis- coverable in other languages. In Hebrew, anachnu, we, from anach (in actual use anoki), I, with the addition of nu, a sign of plurality, is the full form of the plural of the pronoun of the first person ; yet in the verbal terminations anachnu is represented solely by nu, the final fragment, which originally was only a sufiix of number. But we need not go beyond the range of the Dravidian languages themselves for an illustration. We are furnished with a perfectly parallel case by the Telugu. The pronoun of the second person singular in Telugu is nivu, thou, from ni, the radical base, and vu, an euphonic addition. This vu is of so Httle importance to the expression of the idea of personality, that it totally disappears in all the oblique cases. Nevertheless, it forms the regular termination of the second person singular of the Telugu verb, and it has acquired this use precisely like the n which forms the ordinary termination of the first person singular of the Dravidian verb, simply from the accident of position, seeii^ that it is not even a sign of number, like the n of the first person, much less of personality, but is merely an euphonisation. Supposing nd, yd, or a, to be the primitive form of the Dravidian 372 THE PRONOUN pronoun of the first person, and ni, yt, or i (as we shall presently find it to be) the corresponding form of the pronoun of the second person, it seems evident that the only essential difference between the two consists in the difference between the two vowels a and t. We seem to be able also to trace back these pronouns historically to the same two vowels. The initial consonant, whatever be the consonant used, seems to be the common property of both pronouns and the means by which their personality is expressed, whilst the annexed a re- stricts the signification to the first person, or that of the speaker ; t, to the second person, or that of the person addressed. Some resemblance to this arrangement may be noticed in the personal pronouns of the Hebrew, in which I is an-6M ; thou, an-td (corrupted into at-td). The method adopted by the Dra vidian languages of expressing the difference between the first person and the second by means of the vowels a and i, does not appear to be the result of accident. It is probably founded on some ultimate principle, though it may be difficult or impossible now to discover what that principle is. If the pronominal bases, a and i, be considered as identical with a and ^, the demonstrative bases, an idea which would suit the signification, and which is corroborated by the circumstance that u, the next vowel in order, is also a demonstrative, we are met by the apparently insurmountable difficulty that in all the Dravidian tongues, and (as far as the use of these demonstrative vowels extends) in all the tongues of the Indo-European family also, a is not the proximate, but the remote, demonstrative ; and i is not the remote, but the proximate ; whilst u is used in Tamil as an intermediate between these two. If this supposition had been well grounded, we should have expected to find i mean I, and a, thou. But what we actually find is that d means I, and ^, thou. In Tamil, avvidam, literally that place, is occasionally used as a polite periphrasis for you, and ivvidam, literally this place, as a courtly periphrasis for we. So in Malayalam, addeJiam, literally that body, is sometimes used for thou, and iddeham, literally this body, for I. angu, thither, means also, in Malayalam, to thee, to you ; ingu, hither, to me, to us. This use of the demon- strative vowels is exactly the reverse of the use to which we find a and i put in the personal pronouns in all the Dravidian dialects. It seems useless, therefore, to look to the existing demonstrative bases for the origin of the d of nd, I, and the ^ of ni, thou. Is any weight to be attributed to the circumstajice that a, being the easiest and most natural of all vowel sounds, has the first place FIRST PERSON SINGULAR 373 in all lists of vowels, whilst i, being the next easiest vowel sound, stands second ? The first vowel sound would thus be taken to repre- sent the first person, whilst the second person would be represented by the second vowel sound. If this theory had anything to support it beyond its plausibility, it would take us very far back indeed into the history of the origin of human speech. It is remarkable, however, that this theory seems to receive confirmation from the Chinese, which exhibits probably the oldest stage of human speech of which any written records survive. According to Mr Edkins, the oldest forms of the first two pronouns in Chinese were a and i. I may add, that the most peculiar and distinctive, possibly the most ancient, of the Dra vidian demonstratives — the demonstrative which denotes in Tamil, Malayalam, Canarese, something intermediate between a and i — was u. We thus find the whole of the first three simple vowels utiHsed, a=l ; ^=thou ; w=he, she, it. Extra- Dravidian Relationship. — We now enter upon a comparison of a, ya, or na, the Dravidian pronoun of the first person, with the pronouns of the same persons which are contained in other families of tongues, for the purpose of ascertaining its relationship. As nd constitutes the personal element in nam, we, as well as in nan, I (and it is the same with ya and a, the verbal forms), it is evident that our comparison should not be exclusively restricted to the singular, but that we are at liberty to include in the comparison the plurals of this pronoun in the various languages which are compared ; for it is not improbable a priori that some analogies may have dis- appeared from the singular which have been retained in the plural. It is also to be remembered that we are not obliged to restrict our- selves to comparing the pronouns of other families of languages with the Dravidian ya alone, ya may be older than na, na, or a ; yet each of these is old enough for any comparison that can be instituted. All pronouns of the first person singular that have been used at any time in Asia, Europe, or Northern Africa, whether it be in connection with the Indo-European, the Semitic, or Scythian family of tongues, can more or less distinctly be traced back, I believe, to two roots. Each of those roots has been preserved in Sanskrit, and in the more primitive members of the Indo-European family ; one {ah) in the nominative, the other, and by far the more widely prevalent one {ma), in t]je oblique cases. In order, therefore, to investigate the affiliation of the Dravidian pronoun of the first person, it will be necessary to extend our inquiries over a wider area than usual. 374 THE PRONOUN 1. Semitic Analogies. — The Semitic pronoun presents some remarkable analogies to the Dra vidian. This will appear on com- paring the Dravidian nd with the corresponding Hebrew ani, wdth the prefix an of the Hebrew anoki, of the Egyptian anuky and of the Babylonian anaku, dnaka, or anku, and especially with the Jewish- Syriac and, the Christian-Syriac eno, and the ^Ethiopic and Arabic and. The plural of the Aramaic and is formed by suflSixing n (the final consonant of in or dn) : we may therefore compare the Tamil ndm, we, with the Aramaic plural andn, and also with the Egyptian plural anen. Notwithstanding this remarkable resemblance between the Semitic pronoun and the Dravidian, it is doubtful whether the resemblance is not merely accidental. The Semitic initial syllable an, in which the resemblance resides, is not confined to the pronouns of the first person. We find it not only in ana (from anah, and that again from anah), I, but also in the Arabic and Old Hebrew antd and the Aramaic ant, thou (Egyptian, en-tek, en-ta). The prefix being pre- cisely the same in both cases, the pronoun of the second person seems to have as good a claim to it as that of the first. It does not seem, moreover, to be an essential part of either pronoun ; for we find a similar prefix in the third person in some of the Semitic dialects — e.g., in the Egyptian entuf, he, entus, she, and the Chaldaic and Hebrew suffix enhu, he. Moreover, the alliance of the Semitic pronouns of the first and second persons with the Indo-European comes out into more distinct relief when this prefix is laid aside. When the initial an is removed from the pronoun of the first person, we cannot doubt the connection of the remaining syllable {oki, ah, ah, uk, aku, or ak) with the Sanskrit ah, the Gothic ik, and the Greek-Latin eg ; and it is equally evident that when an or en is rejected from the pronouns of the second person [antd, anti, ant, entek, enta), the ta, ti, te, or t, which remains, is allied to the Sanskrit and Latin tu. It has sometimes been supposed that this Semitic prefix an is simply euphonic — a sort of initial nunnation like that which is admitted to exist in the Talmudic' vnhu, he, when compared with the ordinary and undoubtedly more ancient Hebrew M. On this supposition, it is allied, in nature and origin, to the euphonic suffixes or nunnations which may be observed in the Greek ey(ovr) (egone), in the Finnish mi-nd, I, and in the final nasal of the North Indian main, I, and taifl or tun, thou. If this be the origin of the Semitic prefix an, it must certainly be unconnected with the Dravidian nd or and. Sir H. Rawlinson supposes an to be a particle of specification, a sort of definite article ; and he also considers it to be identical with FIRST PERSON SINGULAR 375 am, the termination of the Sanskrit personal pronouns ah-am, I, tv-am, thou, va-y-am, we, yu-y-am, you. The only difference, he says, is that the particle is prefixed in the one family of languages, and suffixed in the other, with a change of m into its equivalent nasal n. I have already stated that I regard the Sanskrit termination am as the ordinary termination of the nominative of the neuter singular, and as used instead of the masculine and feminine, simply because of the intense personality which is inherent in the first and second personal pronouns, especially in their nominatives, and which renders the terminations distinctive of those genders unnecessary. I have also stated that I regard it as probable that the terminal n of the Dra vidian personal pronouns is identical with the formative an or am of many Dravidian neuter singular nouns, and possible that it is identical also with the Sanskrit nominative-accusative neuter case-sign am, which has found its way, as it appears to me, into the nominatives of the Sanskrit pronouns ah-am, &c. If the initial an of the Semitic languages is allied to the final am of the Sanskrit aham, then it may possibly be allied also to the final n or an of the Dra- vidian pronouns nd-n, I, ni-n, thou, td-n, self. On the whole, how- ever, it appears to me more probable that the resemblance between the Semitic and Dravidian languages on this point, though deserving of notice, is altogether accidental. 2. Indo-European Analogies. — It has already been remarked that there appear to be but two pronouns of the first person singular known to the Indo-European family of tongues,' as to the Semitic and Scythian, one of which appears in the nominative of the older Indo-European languages, the other in the oblique cases. The nominative of this pronoun is ah-am in Sanskrit, ad-am in Old Persian, az-em in Zend, eg-o in Latin and Greek {€ywv=aham) (egon), ik in Gothic, ih in the Old German, az in the Old Slavonic, asz in Lithuanian, and gd in Bohemian. We find substantially the same root in the Semitic dh, ah, uh, aku, 6M, &c., and in several languages of the Malayo-Polynesian group — e.g., Malay dku, Tagala aco, Tahitian au. The oblique cases of the pronoun of the first person singular in the Indo-European family are formed from a totally different base from that of the nominative, and of this oblique base perhaps the best representative is the Sanskrit ma. m forms the most prominent and essential portion of ma^ and this m is followed either by a or by some vowel which appears to have been derived from it. In the oblique cases of Sanskrit, this pronoun has the form of 7na, whenever the nature of the succeeding syllable allows a to remain unchanged — 376 THE PRONOUN e.g., ma-yi, in me, ma-ma, of me. In the secondary forms of the d^ative and the genitive it becomes 7nd. In Zend and Old Persian, mxi preponderates, whilst compounded and abbreviated vowels appear in the Zend dative-genitives me, moi ; and a pronominal base in ama is found in some of the Old Persian prepositional compounds. In the Greek /xe, e/xe, /xoi, /zov (me, erne, moi, mx)u), &c., the vowel which is employed librates between e and o, each of which is naturally derived from a ; whilst the initial e of l/xe is in accordance with the tendency of Greek to prefix a vowel to certain words beginning with a consonant — e.g., ovofia (onoma), for vcojua {noma), Latin has me, except in the dative, which is mihi. Gothic has mi and mei (gen. meina). Lithuanian uses man as the basis of its oblique cases ; though possibly the final n of this form belongs properly, like the n of Gothic, to the sign of the genitive. In the pronominal terminations of the verb in the Indo-European languages, the first person singular almost invariably makes use of this oblique pronominal base, in preference to the base of the nominative, with such modifications as euphony may require. The termination of the first person singular is mi or m in Sanskrit and Zend, in all primary and secondary verbs. We have the same ending in Greek verbs in /xt (mi), and in the fxai (mai) of the middle voice ; in the m of the Latin sum and inquam, in the Lithuanian mi, in the Polish am, in the Armenian em, in the New Persian am. It becomes m in the old High German gam, I go ; tuom, I do ; and him or pim (Sans, bhavdmi), I am, converted in modern German to bin. On comparing the pronominal terminations of the Indo-European verb, it is evident that the preponderance of use and authority is in favour of mi, and that m has been derived from mi by abbreviation. It seems equally clear, however, that mi itself has been derived from ma, the normal base of the oblique cases ; for in all languages a evinces a tendency to be converted into some weaker vowel, i, e, or o ; whereas no instance is adducible of the opposite process. Perhaps the best illustration of the regularity of this change from ma to mi is that which is furnished by the Esthonian, a Finnish dialect, in which each of the personal pronouns has two forms, the one primitive, the other euphonised — e.g., ma or minna, I ; sa or sinna, thou. The question of the relative antiquity of the nominative base agh and the inflexional base ma does not appear to me to be one of any great importance, both bases, as we have seen, being of immense antiquity. Still, if any considerable difference in age exists, I am inclined to consider ma as the older. Children learn to say ' mine ' FIRST PERSON SINGULAR 377 long before they discover the meaning and use of I ; and it may have been the same in the childhood of nations, ma, the base of mine, may probably claim to be one of the oldest shapes of the pronoun of the first person now discoverable in the world. We have now to inquire whether any analogy is discoverable between the Dra vidian na, ya, or a, and the ultimate Indo-European base ma. I do not seek for traces of the derivation of the one from the other. The only admissible idea, as it appears to me, is that of analogy, or remote relationship. Before proceeding further in the inquiry, it is desirable that we should ascertain what changes the m of ma sustains in the Indo-European languages themselves. It appears certain that mxi changes into na and va, and probable that it changes also into a. (1.) The m of ma often changes in the Indo-European languages into n. The final m of the first person of Sanskrit and Latin verbs (the abbreviation and representative of mi or ma) has in some instances degenerated into n in Greek — e.g., compare the Sanskrit asam, I was, and the corresponding Latin eram, with the Greek rjv {en) ; and adada-m with e8i-5w-v {edi-do-n). We see a similar change of m into 71, on comparing the modern German bin, I am, with the old High German bim or fim ; and the Persian hastam, I am, with the Beluchi hastjan. Compare also the Laghmani pdkan, I go. The n which constitutes the initial and radical consonant of the plural of the pronoun of the first person in many of the Indo- European languages is evidently, like the final n of the singular terminations referred to above, derived from an older m. One of the oldest forms of the plural of this pronoun, if not the very oldest, is that which is employed in the verbal inflexions, and which in Sanskrit is mas (Vedic-Sanskrit masi), in Latin mus, in Greek fxev {men) (for the more ancient and more correct ^olic fxes) {mes) : the most natural explanation of which pronominal ending is to consider it as derived from ma, the old first person singular, by the addition of s, the sign of plurality. The m of this primeval mxis often becomes n — e.g., in the Latin nos, the Celtic ni, the Greek vwt {noi) ; and also in the San- skrit secondary forms nas and nau, the Zend no, and the Old Slavonic na. This n is evidently a weakening of m>, and represents the personality of the pronouji of the first person, irrespective of the idea of number ; which is expressed, I conceive, by the subsequent portion of the word. It is remarkable that in Welsh, whilst the absolute forms of the personal pronouns I and we, are mi and ni 37^ THE PRONOUN respectively, the personal terminations of the verb m and n are often found to change places, so that the first person singular comes to be represented by n, and the corresponding plural by m — e.g., gwelwn, I saw ; givelem, we saw. Something similar has been observed in the Greek eSlSow {edidoun), compared with the plural of the same, eSiSofxev {edidomen) ; but the use of n in the singular and m in the plural in verbal terminations, is much more systematic in the Welsh and its related dialects than in Greek. The Irish generally differs from those dialects in this particular — e.g., compare Irish cairim, I love, with the Welsh carwn. Welsh verbs of the first person, ending in n in the singular and m in the plural, bear a remarkable resem- blance to the Tamil singular en, plural em or 6m. (2.) The m changes also into v. v alternates with n as the initial and radical consonant of the plural of the first person in several Indo- European languages ; and this v, I conceive, is merely a softened form of m. It was shown in the Part on "Sounds" that, in the Dravidian languages, wherever n and v are found to alternate, we have reason to conclude that both are derived from, or represent, an older m ; and the rule appears to hold equally good in regard to the Indo-European languages. When we find in Sanskrit the nominative plural vayam (from va and the neuter formative am), we, and at the same time nas, which is optionally used for the accusative, genitive, and dative plural of the same pronoun, we cannot avoid coming to the conclusion that both the na of nas and the va of vayam are derived from a more primitive ma. This idea is confirmed by finding n and v in exactly the same connections in Zend. Compare the Old Slavonic plural mes, we, with the Gothic veis, and especially the Old Slavonic dual ve, we two, with the accusative of the same, na, us two. In the Lithuanian dual, v alternates, not with n, but with m — that is, with what appears to be the more primitive con- sonant. The nominative-accusative masculine may be either ve-du or mu-du. In the personal endings of the Old Slavonic verb, ve represents the first person dual ; in Lithuanian, va ; whilst the plural proper ends in mu in the former language, and me in the latter. (3.) The m of the pronoun of the first person disappears sometimes altogether, so that ma changes into a. This is the only reasonable explanation that has been given of the origin of the Vedic asm^, we = a/xju€s {ammes). When this is compared with yushme, you=i!/>t/xes [ummes), it is evident that sme, whatever its origin, is in use simply a sign of the plural, and that as the yu {=tu) of yushme represents the singular thou, so the a of asme must represent the singular I. FIRST PERSON SINGULAR 379 This being the case, a-sme must be equivalent to ma-sme. This seems to be the best explanation also of the a of the Sanskrit dual dvdm, we two, probably derived, some think, from ma, I, and dva, two. We find the a of the plural asme itself similarly lengthened in the Bengali ndmi, modern Bengali ami. (See " Pluralisation of Pronouns.") The same pronominal root m changes also in the Scythian tongues, as will be seen, to n and ng, and even to b ; but at present we have to deal exclusively with the changes that take place in the Indo- European tongues. Can we now infer the existence of any relationship between the Dravidian pronominal base and the Indo-European ? Is the Dra- vidian ya, varying to n or n, on the one hand, and a on the other, connected in any way with the Indo-European ma, varying to na on the one hand, and on the other to va, and possibly also to a ? I think we are warranted in inferring the existence of some connection. It is more difficult, as it appears to me, to suppose that these two series of words, belonging to the earliest requirements of human speech, identical in meaning, and so nearly alike in form, were from the beginning independent of one another, than that an ultimate relationship of some kind existed between them. If we were at liberty to compare the Dravidian na directly with the Indo-European ma, no room for doubt could exist — ma, as we have seen, being proved to change into na. And even though we are obliged to be suspicious of the credentials of the Dravidian na, and to prefer ya as probably a better representative of the very oldest form of the word, yet we are not altogether precluded thereby from making the comparison under consideration, the antiquity of na being almost as great as that of ya, just as the Indo-European na, va, and a must be almost as ancient as ma. ya, it is true, is not one of the shapes the primeval ma is found to have assumed within the circle of the Indo-European tongues ; but as ma is not confined to that family, but is the common property also of the languages of the Scythian group, in which it will be found to have sustained a set of changes peculiar to them, it does not seem unreasonable to suppose that ya, varying to na, may have been the shape it first assumed amongst the early Dra vidians. 3. Scythian Analogies. — When we examine the personal pronouns of the Scythian group of tongues, some independent and very interesting analogies to the Dravidian pronoun are brought to light. The pronominal root which constitutes the basis of the oblique cases in the Indo-European languages, is adopted in the languages of 380 THE PRONOUN the Scythian family, not only in the oblique cases, but also in the nominative itself. Whilst in both families the oblique cases are substantially the same, the Indo-European uses as its nominative the base in ah, the Scythian the base in ma. There are a few lan- guages even in the Indo-European family in which ma has found its way into the nominative — e.g., the Celtic has mi, the New Persian mun, the North Indian vernaculars main. In some cases, also, especially in the later dialects of this family, the accusative has come to be used instead of the nominative, in violation of ordinary grammatical rules. Thus, the Singhalese mama, the Kavi mami, and the Cuneiform Persian m^m, are probably accusatives in their origin, like the Italian mi and the French moi. On the other hand, we are met by one, and only one, exceptional case in the Scythian tongues. The Scythian of the Behistun inscriptions makes use of hu as its nominative ; but in mi, the corresponding possessive suffix, the ordinary Scythian base reappears. (1.) The nominative (as well as the oblique cases) of the first personal pronoun in all existing languages of the Scythian group is derived from a base in Tna ; and it will be shown that this ma not unfrequently comes into perfect accordance with the Dravidian pronoun, by changing into nga and na. In those languages ma is very generally euphonised or nasalised by the addition of a final n, or of an obscure nasal resembling the Sanskrit anusvdra ; in con- sequence of which, not ma, but 7nan, may be stated to be the normal form of the Scythian pronoun, and this bears a closer resemblance than ma to the Dravidian nan. The addition of this euphonic nasal is not unknown even to the Indo-European languages. It may be seen in the Persian man, the Sindhian man, and the Beluchi menik ; and a similar inorganic addition is apparent in the old Greek kyiov-q (egone), as also in rvvr] {tune). This nasal is much more common, how^- ever, and more characteristic in the Scythian tongues. On examining the Turkish family of tongues, we find men in Oriental Turkish ; 77idn in Turkoman ; 7ndm in Khivan ; ben (m degraded to b) in Ottoman Turkish. In the Finnish family, the Finnish proper has mind ; the Lappish mon ; the Esthonian ma or minna ; the Mordvin and Votiak mon ; the Ostiak nm (dual min, plural men) ; the Magyar en. The Samoiede dialects have nmn, mani. In both Mongolian and Manchu the nominative of this pronoun is bi ; but this is evidently cor- rupted from mi (like the Ottoman ben, from the Oriental or Uigur men) ; and it is mi, with a final nasal, which forms the basis of the oblique cases. In both languages the genitive is mi-nu FIRST PERSON SINGULAR 38 1 or mi-ni ; and the dative is men-dou in Mongolian, min-de in Manchu. It is evident from the above comparison that the true and essential representative of this pronoun in the Scythian tongues is ma. In many of those idioms ma still retains its place unchanged, or may optionally be used instead of the later man. The Mingrelian has tna, the Saunian mi, the Lasian ma, the Georgian me. The Finnish has both me or ma and mind, and also mia ; the Ostiak both min and ma. It is found also in those languages in which man constitutes the isolated pronoun that m is used as its equivalent in the personal terminations of the verbs, and generally in all inflexional compounds. We see this usage illustrated in the colloquial languages of Northern India and in Persian. For example, whilst man is the nominative of the Persian pronoun, the basis of the oblique cases is not man, but ma (e.g., ma-rd, me, of me) ; and the pronominal ending of the verb in the first person singular is m. In a similar manner, in the Turkish family of languages, m is used in composition as the equivalent of man or men. Thus, in Oriental Turkish, whilst men is retained in the present tense — e.g., bold-men, 1 am — the preterite is contented with m alone — e.g., holdi-m, I was. The same suffix is used to denote the first person singular in most of the Scythian possessive compounds, a class of words which is peculiar to the Scythian family — e.g., Turkish bdbd-m, my father, from bdbd, father, and m, the representative of the first person singular. In the Magyar also, though the isolated pronoun of the first person singular is en, yet m is used instead of n in the possessive compounds and " objective " inflexional terminations — e.g., from atya, father, is formed the possessive compound atya-m, my father ; and the first person singular of " objective " verbs ends in m — e.g., szeretem, I love (some one). It is also to be noticed, that whilst the Magyar has en as the singular of the isolated pronoun, its plural is mi or mink ; the former of which is evidently pluralised from ma or me, the latter from min. (2.) It was shown that the initial and radical m of the Indo- European pronoun was occasionally converted into n : we have now to show that a similar change from m to w is apparent in the Scythian languages also, and that in some of those languages n has become as distincti-sj^ of the first person as in the Dravidian family itself. In Finnish, though the isolated form of this pronoun is ma or mind, yet in all inflexional additions and compounds m is represented by n — e.gr., from isd, father, is formed isd-ni, my father, 382 THE PRONOUN and from 61, to be, is formed 6l-en, I am. This final n is not derived from the euphonic n of mind ; but from a direct conversion of m into n ; for though we see the same euphonic addition of n in sind (from se or sia), thou, yet we have t alone (the equivalent of s) in 6l-et, thou art. n has, therefore, become in Finnish, as in Dra- vidian, the ordinary sign of the first person singular of the verb ; though there is this difference, that in Dravidian the n is the final w, which is distinctive only of numbers, whereas the Finnish n seems to be derived by conversion from an older m, the initial m of ma. The Magyar en, I, appears to be still more nearly allied to the Dravidian pronoun ; and in this case n is certainly derived from m, for whilst n is found in the nominative, m is used instead in all pos- sessive compounds and verbal inflexions. With the Magyar nomina- tive en, compare the Tamil-Canarese en or en. May we also compare an, I, in the Lar, a Sindhian dialect ? A similar form of this pronoun is found in the Mordvin, another idiom of the Finnish or Ugrian family, in which, whilst mon is the isolated nominative, an is used instead in verbal inflexions — e.g., faz-an, I (am) the Lord. In the Olet or Calmuck dialect of the Mongolian tongue, there are distinct traces of the same change of ma into na ; and in this instance the n appears, not as the final, but as the initial, and is therefore in more perfect accordance with the n of the Dravidian pronominal base. The nominative of this pronoun in Calmuck is hi (from mi), and the same base appears in the genitive mini ; but the rest of the oblique cases are formed, not from hi or mi, but from nad or na — e.g., na-da, to me, na-da-edze, from me, and also na-mai, me. We here discover the existence of a pronominal base in na (probably derived from 7mi), which is in remarkable agreement with one of the forms of the Dravidian base. In a few of the Scythian languages, the isolated pronoun, including its nominative, seems to be almost identical with that of the Dra- vidian family — e.g., na in the Quasi Qumuk, a Caucasian dialect ; and ne in Motor, a dialect of the Samoiede ; na or 7iai in Corean ; ne or ni in Basque. In the East Asian languages, gn or ng (which are pronounced alike) are often found to take the place of n. Some- times n and gn alternate in the same language, like n and n in Tamil-Malayalam. The Canton Chinese is ngo ; the Mandarin, wo. Old Chinese forms, according to Mr Edkins, are nga, ga, go, kan, a. The analogy of the pronoun of the second person would seem to show that a was the oldest form of all. Compare Burman, nd or ngd ; Tibetan, written nd, colloquial gnyd (' mine,' written naki, nayi, FIRST PERSON SINGULAR 383 colloquial gnay) ; Tetenge, an Assam dialect, ne ; Mikir, ne ; Khari Naga, ni. The Burman ngd prevails in the languages of the sub- Himalayan tribes. A very common form among those tribes, and those of the north-eastern frontier, including also the Kols of Central India, ends in ng — e.g., ang, ung, ing, aing. I am not clear, however, as to the nature of the relationship of the latter forms to ma, nga, and na, the High Asian group, with which the Dravidian (and also the Indo-European) pronoun seems to stand in closer connection. I feel, however, on tolerably firm ground in comparing the Tibetan nd, I, colloquial nga, with the Malayalam nd ; and if so, the Chinese ngo, especially when examined in the light of the Chinese ni, thou, may also be allowed to claim kindred. We may here, too, compare the Australian pronouns of the first person — viz., nga, nganya, I ; its dual, ngalee, we two ; and the plurals ngadlu and nadju, we. (3.) A few traces of the softening of na or nga to ya and a, or at least of the use of ya and a instead of nga and na, may also perhaps be discovered in the East Asian languages. Thus the Sgau-Karen is yd, yah ; the Pwo-Karen yer ; the Manyak a. The Pekin Chinese wo may also be compared. On the whole, we seem to have reason to conclude that the various forms which the pronoun of the first person singular assumes in the Scythian group of languages, and which we have now compared, are identical. Possibly, also, we may see reason to conclude that the Scythian forms (ma, na, ha, nga, ya) have had a common origin with the Indo-European (ma, va, na, and a). The Dravidian ya, na, a, bear so close a resemblance to the pronouns of both groups (especi- ally, as we have seen, to the Scythian), that we seem to be justified in regarding them as related to both in common. If this be admitted, we seem to be justified in arriving at the conclusion that one and the same pronoun of the first person, probably ma, was the common property of the whole Japhetic family prior to the separation of the Indo-European tribes from the Scythian. The conclusion arrived at by Professor Hunfalvy (in his paper on the study of the Turanian languages, read at the International Congress of Orientalists, 1874) is substantially similar. He notices the resemblances between the Aryan and Turanian languages with regard to the personal pronouns, and then says that " considering this fact, he is inclined to suppose that a stage of l^guage anterior to both classes must have existed." He thinks he sees also in certain single words, as papa, mama, &c., visible remains of that ancient form of speech. 384 THE PRONOUN 2. Pronoun of the Second Person Singular. Comparison of Dialects. — Our first inquiry, as with respect to the pronoun of the first person singular, must be what appears to have been the primitive form of this pronoun. In Tamil, nt, which is properly the crude base, is invariably used as the isolated nominative, instead of ntn — the form which would correspond by rule to nan, the nominative of the first person singular. That nm originally constituted the nominative even in Tamil, appears from this, that the oblique cases in the higher dialect agree in using nin as the base to which the case-suffixes are attached, un is occasionally used as the inflexion in the classics, always in the colloquial dialect. Another form which is occasionally used in the classics is nty, in which the final y appears to bear the same relation to n as the initial n of ydn or nan of the first person — that is, it has either been softened from n, or is the primitive letter from which n was hardened. This final y appears also in ay and 6y, two of the personal terminations of verbs and conjugated nouns. The final n of this pronoun, though it is generally lost altogether in the nomina- tive, and is only represented occasionally by y, is invariably retained in the inflexional base, in which it is the initial n that becomes liable to alteration. When the initial vowel is retained, the included vowel is either i or u (nin or nun), generally the former, but when it is dis- carded, u (un) is the only vowel in use. The inflexions now described are nin, min,un. In the personal terminations of the Tamil verb, this pronoun is represented by the suffixes dy, 6y, ei, or i ; from each of which suffixes the final n, as well as the initial, has disappeared. In the poetical dialect of the language, the initial n at first sight appears to have retained its place in such forms as nadandanei, thou didst walk, and in the corresponding plural nadandanir, ye walked ; but the n of these pronominal terminations [nei and ntr) is merely euphonic (as in similar terminations of the first person of the verb already mentioned), and is inserted for the purpose of keeping separate the contiguous vowels of nadanda-ei and nadanda-ir. The root of the verb is regularly used in Tamil as the second person singular of the imperative, without any pronominal suffix, and even without any euphonic addition ; but the second person plural of the imperative in the colloquial dialect is formed by the addition of urn, which is probably identical with the um or m which constitutes the normal sign of plurality in Dravidian pronouns, and is probably in itself the copulative ' and ' or also. (See the pronoun of the first SECOND PERSON SINGULAR 385 person.) Compare this with the optional addition of mu to the root in Telugu to form the imperative singular. Properly mu forms an honorific singular, and is therefore to be regarded, like the Tamil um, as a plural in original signification. In the higher dialect of Tamil, ay and ir, the ordinary representatives of these pronouns in the verbal inflexions, are often added to the root to form the singular and plural imperative — e.g., heldy, hear thou, helir, hear ye. These forms appear at first sight to be identical with heldy, thou hearest not, and helir, ye hear not ; but they are not really identical, as Beschi supposed, for it will be shown in the section on the " Negative Verb " that a, probably a relic of al, not, is an element in all negative forms ; though in these, and in some other instances, it has been absorbed in the succeeding long vowels. The plural forms of this pronoun in Tamil are as follows : — nom. nir, niyir, nivir, ningal ; inflexion, num, um, ungal. nin, the singular poetical inflexion, does not become nim in the plural, as might be expected, and as we find it in Canarese, but only num. Personal terminations of the verb, ir, ir. Tamil grammarians give min {e.g., henmin, hear ye) as one of the signs of the second person plural in the imperative. The nature of this form will be considered in the section on the " Pluralisation of the Personal Pronouns." In Malayalam the nominative is ni, as in Tamil in both dialects ; the inflexion nin, as in classical Tamil — e.g., ninahk\ to thee ; plural nom. ninnal, ninnal ; inflexion ninnal, also in the poets nim (e.g., nimmodu, with you), from the obsolete nom. nim. The Tulu nominative singular is t (comp. Tel. ivu, from an obsolete t) ; inflexion nin' — e.g., nina, thy. In nikF, to thee, the inflexion is ni. Verbal termination a ; plurals, nom. ir (chiefly used as an honorific singular, like nir in colloquial Tamil), also nihulu ; inflex- ions ir* and nihuV ; verbal ending ar. In Canarese, the nominative of this pronoun in the colloquial dialect is nin-u, classical nin ; but the crude form ni is often used instead of nin-u, as is always the case in Tamil. In both dialects the inflexion is nin — e.g., ninna, thy. In the personal terminations of the verb this pronoun is much changed in all the Dravidian dialects. It not only loses its initial n, like the pronoun of the first person, but its final n also disappears. Generally nothing remains in the verbal inflexions but the included vowel (probably the primitive pro- nominal base), and that also is more or less modified by use. In the colloquial Canarese verb it appears as i, i, iye, and e ; in classical Canarese ay only, closely resembling the Tamil ay. Plurals, nom. 386 THE PRONOUN coll. ntvu ; class, nim ; inflexion in both nim — e.g., nimma, your. Verbal terminations, coll. iri, iri, ari ; class, ir. This ir is identical with one of the classical Tamil terminations. The Telugu nominative is nivu, expanded from ni by the addition of the euphonic particle vu. nivu, Tel. thou, is identical in form, though not in meaning, with the modern Canarese plural of the same pronoun — viz., nivu, you. ni, the crude form, is also used, as in the other dialects. In the oblique cases, Telugu rejects the euphonic addition of vu, and uses ni as its inflexional base, and also as its possessive. The objective alone follows the example of the other dialects in abbreviating the included vowel, and appending a final nasal. That case is nin-u or nin-nu, and is evidently formed from a nominative nin-u. In the higher dialect of Telugu, ivu, from an obsolete nominative i, identical with the Tulu, is occasionally used instead of nivu. The Telugu plural of this pronoun has miru as the nominative, mi as the inflexion, and mimu as the accusative. Both miru and mimu indicate a base in mi, from which they have been formed by the addition of signs of plurality ; and mi bears the same relation to the ni of the other dialects that md, the Telugu plural of the first person, does to the ordinary Dra vidian nd. How this change from nio m has taken place will be inquired into under the head of " The Plurals." The plural in the higher dialect is iru. In the personal terminations of the verb, Telugu rejects every portion of the pronominal root, and employs only the euphonic addition vu or vi. The Tuda nominative is ni, inflexion nin, personal termination of verb i or e. Plural nominative nimxt, inflexion ni^n, personal termina- tion of verb i or e, as in singular. In the dialect of the Kotas, the nominative is ni, inflexion nin, personal termination of verb i. Plural nominative nime (also nive), inflexion nim, personal termination of verb iri, iri. In Gond, the nominative singular is immd, which is evidently an older form of the plural used as the honorific singular. The inflexion is ni (niiva, thy), personal termination of verb ni or i. Plural nominative immdt, inflexion mi, as in Telugu ; personal termination of verb it. The personal terminations of the first and second persons singular in Gond require a little consideration. In both persons the initial n of the isolated pronoun seems to hold its ground in some of the tenses in a manner which is not observed in any other dialect — e.g., dydtond, I am becoming, dydtoni, thou art becoming. In some other tenses {e.g., imperfect dnddn, I became, perfect dttdn, I have become), the termination of the first person resembles that in use in I SECOND PERSON SINGULAR 387 most of the other dialects. In the second person {dndi, dttt), the n, whatever its origin, disappears altogether, and is replaced by the ordinary Dravidian i. I prefer, therefore, to regard the n of the first and second persons, in these tenses, as the n of the pronoun of the third person singular, on, he, forming, when added to the root, a participial noun. dydt-6n-d would then mean, I am one who becomes ; dydt-on-t, thou art one who becomes. If this view is correct, nothing can be observed in these forms differing in reality from those in the other dialects. The Ku pronoun corresponds on the whole to the Telugu. Nom- inative singular tnu, inflexion nt, personal termination of verb i ; plural nominative ir-u, inflexion mt, personal termination of verb eru, dru. The Rajmahal nominative singular is nin, inflexion nin ; plural nina, inflexion nim. Uraon nominative singular nien, inflexion nien ; plural nominative dsu, inflexion dss* . The Brahui nominative is ni, as in most of the Dravidian dialects, inflexion nd ; plural nominative num, one of the inflexions of the plural in classical Tamil ; inflexion num (numd, your) ; verbal termination n, as in many of the Dravidian dialects (compare aren^ we are, areri, you are). See the " Table of Pronouns " of the second person for the forms found in the minor dialects of Central India. We have now to consider the conclusion to be drawn from the comparison made above. We found three forms of the pronoun of the first person singular, ndn, ydn, an, each of which claimed to be the best representative of the original form ; and of these, ydn seemed to carry with it most authority, and to be probably the source from which ndn on the one hand, and an on the other, were derived. With regard to the pronoun of the second person singular, there are only two forms (nin, in) whose relative antiquity we are called upon to decide. No claim can be set up on behalf of yin as a pronoun of the second person to correspond with the ydn of the first person. If such a form ever existed, I can find no trace of it now left;. The final n of nin or in (as of ndn, ydn, an) has already been ascertained to be merely a sign of the singular number. In the plural it is replaced by m, the sign of plurality, or r, ir, a relic of ivar, they (prox.). This final n of the singular m*y, therefore, be dismissed from our con- sideration at once. On comparing ni and i with nd and d, it seems evident that if the initial n oindn did not belong to the root, but was a product of nasalisation, the initial n of nin cannot safely be regarded 388 THE PRONOUN as radical. If nd was derived from a more primitive yd or d, it seems evident that ni must have been derived from a more primitive i. The initial n of ni must be identical with the initial n oi nd. Whatever the origin of the one may be, the origin of the other must be the same. Just as the initial n of nd disappears from all the verbal terminations of the first person, so the initial n of m disappears from all the verbal terminations of the second. If this initial n had been radical, it would have retained its place more or less firmly in the verbal inflexions, like the m of the Indo-European first person, and the t or s of the second person of the same. As the initial n has disappeared so com- pletely from the Dravidian verbal inflexions, though it sometimes retains its place as the inflexional base of the oblique cases, I conclude that it is not radical, and that we are to consider ^ more primitive than ni. Still the antiquity of the initial n of ni must be enormously great — almost equal to that of i itself, seeing that we find it, as we shall presently see, in the Scythian of Behistun, and even in Chinese, in both of which the pronoun of the second person is ni. It is ni also in Bornu, a language of Central Africa. Even when looking at the Dravidian dialects alone, we cannot sup- pose nt much later in origin than 1 Whatever be the relative antiquity of ni and t, I consider the vowel, not the consonant, as the real pro- nominal base. The only question that remains, therefore, is, what is to be regarded as the oldest shape of this vowel ? We find i, u, and also, but more rarely, a and e. The last two may be left out of account. The vowels most generally used are i and u. In the verbal terminations i has driven u out of the field altogether. On the whole, there seems to be more in favour of the antiquity of i than of that of u, though it must be admitted that u changes more readily in Dravi- dian speech to i than i to u — e.g., fuli, Tam. a tiger, becomes in the pronunciation of the vulgar fili ; mun, before, becomes min, &c. It will be seen that generally in the Indo-European languages the vowel of the pronoun of the first person is u, whilst in the Scythian lan- guages it is i. Possibly at the outset there was no very sharp line of distinction between these two sounds. At all events, we cannot safely venture to draw any such sharp line of distinction now between the i and u of the pronoun of the second person in the Dravidian tongues, both vowels being retained, in some connection or another, in most of the dialects. Thus in poetical Tamil we find both nin and nun as the singular inflexion of the pronoun ; in the plural we find num and unga], but not nim, though the nominative ningal must be considered as the representative of an older nim. SECOND PERSON SINGULAR 389 Extra- Dravidian Relationship. — It has been shown that the Dravi- dian pronoun of the first person has afiinities with each of the great Japhetic groups, with some special Scythian affinities. It will be found that the relationship of the pronoun of the second person is less extensive, but more distinctive ; it is more specifically Scythian, or at least non- Aryan. Throughout the Scythian, as well as the Indo-European group, the most prevalent form of the pronoun of the second person singular is that which is formed from the consonant t {e.g., tu), or its euphonised equivalent s {e.g. a-v) {su) ; and the only other form found in any family of either of those groups is that which is built upon the consonant n, and of which the Cuneiform Scythian, the Chinese, and the Dravidian ni is the best representative. These roots appear to have been always independent of one another. I cannot discover any reliable trace of a connection between them, or of a gradual change in any instance of the one form into the other. In order to place this point in a clear light, it is desirable, in the first place, to trace out the connections and alliances of the pro- nominal root tu. It has been conjectured that this pronoun had its origin in the demonstrative base t ; but the investigation of this point is beyond our purpose, which is merely that of tracing its relationship. In Sanskrit the pronoun of the second person singular is tva-m ; in Zend tu-m, and also thw\ as included in the accusative thivd, thee. Connected with the Sanskrit tva, there is a simpler form, ta, which is apparent in tava, thy ; and we have analogies to this in the Kavi ta and the Semitic ta (included in antd, thou). The Semitic td is changed in the inflexions to M, a change which resembles that of the Kavi, which has ta as its nominative and ho as its possessive. Bopp supposes that yu, the base of the most common form of the plural of this pronoun, is derived from tu, and that va, the base of the Sanskrit secondary plural vas and of the Latin vos, is derived from tva. V, however, is more frequently derived from m than from any other letter, of which we have seen an instance in the change of the nm of the first person into va in vayam. It is not very easy to explain how t became v and y. tva-m becomes tuva-m in Old Persian ; and from tu (itself derived from tv) proceeds the Sanskrit dative tu- hhyam, the base of which is aUied to, or identical with, the Latin, Armenian, and Pehlvi twf the MoMq and Doric tv {tu) ; the Persian, Afghan, and Singhalese tu ; and the Gothic thu. The th of the Gothic and Zend seems to point out the path by which the Old Greek rv {tu) was converted into av {su). Mr Edkins, in his " China's Place in Phil- 390 THE PRONOUN ology," has suggested another origin for yu. He supposes it may be connected with ni or nu, the Chinese pronoun of the second person, of which i or u was, he thinks, the primitive form. If this supposition should be correct, yu will then be the Indo-European equivalent, not only of the Chinese ne or nu, but of the Dravidian, which also is ni or nu — ni in the nominative, nu (nu-n) in the oblique. In the personal terminations of the verbs, in Sanskrit and most other languages of the same family, the earlier t of the ordinary form of this pronoun has very generally been weakened into s in the singu- lar, whilst in most of the plural terminations, t, with some trivial modifications, and with a sign of plurality annexed, has succeeded in retaining its place. In our investigation of the pronoun of the first person, it was found that ma was converted in the personal termina- tions of the verb into mi, and still further weakened into m : so also su (for tu) generally becomes si in the verbal terminations ; and si in Hke manner afterwards becomes s. In the Scythian group of tongues, the pronoun of the second person in general use is substantially the same as in the Indo-European — another evidence of the primeval identity of both groups ; but in the Scythian tongues the weaker s has obtained wider prevalence than the older t ; and the vowel by which s is enunciated is more fre- quently i or e than u or a. The Magyar has te in the singular, ti or tik in the plural, with which we may compare the Armenian tu, thou, and tuh, you. The MongoHan tchi or dzi, thou, exhibits the progress of ti towards softening into si. In Finnish proper, the isolated pronoun of the second person singular is se or sina ; but t retains its place in the plural, and the personal termination of the verb even in the singular is t. The chief peculiarity apparent in the Scythian form of this pro- noun is, that it has generally been euphonised by the addition of a final nasal, the consonant n, precisely in the same manner as the pro- noun of the first person singular. In the older Greek, rvvr) [tune) and rovv (toun) correspond to kywr] {egone) and lytov (egon) ; and in like manner, in the languages which belong to the Scythian group, or which have been subject to Scythian influences, where the pronoun of the first person is found to be nasahsed, the pronoun of the second person generally exhibits the same feature. In the vernaculars of Northern India we see this euphonic addition to the pronoun of the second person in the Hindi, Panjabi, and Sindhi tun, and in the Marathi and Gujarathi tun. In some of those idioms, especially in the Gujarathi and Panjabi, the euphonic nasal appears in the SECOND PERSON SINGULAR 391 oblique cases as well as in the nominative, but more commonly it is found in the nominative alone. In the Turkish family of tongues, sin or sen is the usual form of the pronoun of the second person singular. The n retains its place in the oblique cases, but is lost in siz, the plural. Compare also the Georgian sJien, the Samoiede tan, tani, the Lappish don, the Votiak and Mordvin ton (pluraUm), and the Finnish sind, which alternates with se, sia, and sie. The euphonic origin of this n is most evident in the Esthonian dialect of the Finnish, which uses indifferently sa or sinna for the second person, and ma or minna for the first. In the Mongolian and Manchu, n appears in the oblique cases only. In Mongol the nominative is tchi, in Manchu si ; but the genitive in the former is tchini, in the latter sini, and the corresponding datives are tchim-dou and sin-de. In Calmuck the nominative is dzi or dzima, genitive dzini, dative dzimadou, accusative dzimm. In the pronouns of this language we may observe several instances of m being used as an euphonic, instead of n. It is evident that there is no resemblance whatever between any of the pronouns compared above and the Dravidian nt. The final nd of the Finnish sind, and its equivalent, the final vrj (ne) of the Greek Tvvr] (tune), are separable, euphonic, inorganic additions, and can have no real connection with nt, which is an ultimate root. It will be necessary for us therefore to go further in search of a really trustworthy analogy. We have seen that the Indo-European and Scythian m — the initial of the pronoun of the first person — was probably the origin of the n of the Dravidian nd. Is it possible that the radical t of the pronoun of the second person in both those families of tongues was changed in like manner into n, so as that tu or ti was the origin of the Dravidian ni ? I think not. This is supposed by Castren, a very high authority, to be the history of the n by which the second person singular is often represented in the personal affixes of the Finnish and Turkish families. It may also be mentioned here, that a change of t into n is not quite unknown even in the Indo-European languages. It is somewhat frequently found to take place in Pali — e.g., te, they, masculine, becomes optionally ne ; td, they, feminine, becomes nd ; and tdni, they, neuter, becomes ndni. In Sanskrit also, etam, him, is sometimes changed intg enam. There is no evidence, it is true, that the n now under consideration — the initial n of the Dravidian ni — arose from any such process of change. That it proceeded from an older t would be a wholly gratuitous assumption, in so far as the 392 THE PRONOUN internal history of the Dravidian languages is concerned. It would be more in accordance with precedent, indeed, to regard it as a mere nasalisation. Yet when we carry our inquiries a step further, and bring to view a pronoun with n, not t, in some of the oldest languages of the Scythian group, whilst on the one hand we shall find that the resemblance of this Scythian pronoun to the Dravidian amounts to identity, on the other hand we shall possibly find it allied, by a deep- seated, underground relationship, to the ordinary pronoun with t, so that it must always remain doubtful whether these are not two Japhetic bases of the pronoun of the second person, tu and ni, originally independent, like ah and ma of the first, or whether tu did not change into nu, and that to ni, at some early period, now un- known, before the isolation of the Dravidians, and even before the isolation of the Chinese, from the rest of the Japhetic race. I must first endeavour to establish the first point now mentioned, viz., that traces will be found in various languages of the Scythian group of the existence of a pronoun of the second person, apparently identical with, and certainly allied to, the Dravidian nt. I begin with the most ancient analogy which is capable of direct proof, viz., the pronoun of the second person in Chinese. This is ni, precisely as in the Dravidian idioms. The plural is ni-men (compare wo-men, we, fa-men, they) ; Old Chinese ngi, nu, yu, u. Mr Edkins thinks the oldest form of all was i, to which n was prefixed. The same ni appears in some of the dialects of the nomad tribes of the western frontier of China, towards Tibet — e.g., Gyami and Horpa. The plurals in Gyami are ni-me ; in Horpa, ni-ni. The Tibetan itself, though agreeing so closely as regards the first person, seems to present no analogy in the second. In the dialects of Burma, the prevaiHng form of the word is nang ; in the Karen dialects nah, ner, nd. The Manyak, a dialect of the same stock, which has a for the first person, has no for the second. All the analogous forms of Eastern Asia rest upon the Chinese ; and the antiquity of the Chinese language and literature is so great, that the identity of the Chinese pronoun of the second person with the Dravidian is a point of great interest and importance. The next analogy I adduce is one which I regard as almost equally remarkable and decisive, viz., the pronoun of the second person in the Scythian tablets at Behistun. This is ni, precisely as in the Dravidian idioms ; and the possessive which is used in compounds is ni, which is identical with the similarly abbreviated basis of the Dravidian obHque cases of this pronoun. The plural of this pronoun is, unfortunately, unknown. The personal SECOND PERSON SINGULAR 393 termination of the verb is not ni, but nti, which I suspect to be a compound of ni and ti, like the antd, anti, of the Semitic languages. I have given the Brahui a place amongst the Dra vidian dialects, but I refer to it here again on account of its centrical geographical position. The Brahui pronoun, as we have seen, is ni (plural num), the identity of which, both with the Dra vidian, properly so called, and with the Behistun and Chinese, cannot, I think, be doubted. It is a remark- able circumstance, and very difficult to explain, that in the Kanuri, a language of Bornu, in Central Africa, together with several other Scythian peculiarities, the pronoun of the second person is ni. The antiquity of the Dravidian pronoun of the second person is thus clearly proved, and this proof of its antiquity entitles us to regard as real certain resemblances to it which otherwise might be thought to be accidental. In the Ostiak, the most Dravidian of the Finnish dialects, in that compound of nouns with possessive suffixes which is so characteristic of the Scythian group, the first personal pronoun is represented by m, the second by n — e.g., ime-m, my wife ; inie-n, thy wife. In the Syrianian, another Finnish idiom, the second person of the verb, both singular and plural, is formed by annexing a pronoun of which n is the initial and radical — e.g., kery-n, thou hast done (from kery, to do), kery{n)nyd, you have done. In nyd, you, we see indications of a singular ny, thou, which has been plural- ised, as is usual in these languages, by suffixing to it d or t. In addition to the allied forms discoverable in these compounds, we find in the Ugrian tongues several instances in which the isolated pronoun of the second person, which is used as a nominative, is plainly allied to the Dravidian. In the Ugro-Ostiak, or that dialect of the Ostiak which is treated of in Castren's Grammar, thou is nen ; you two, nin ; you (indefinitely plural), nen. Here ne or ni constitutes the pronominal base, and the final n of the singular nen is a formative or euphonic addition like that which has converted the Dravidian ni into nin. The strong pronunciation of this Ostiak final n reappears, as we shall see, in Turkish. In other Ostiak dialects we find num and ma, and also (which is more deserving of notice) nyn, with a plural nynt. In Vogul we find analogies which are no less remarkable than the above — e.g., nei, ny, nan, nyngi, and nank. Compare also the Vogul plurals nen and non. In the Finnish proper, 4he only trace of this pronoun which we observe is one which, but for the existence of such express analogies in other members of the family, we should probably have over- looked. In the plural of the second person of the Finnish verb {e.g., 394 THE PRONOUN olette, ye are, pluralised from olet, tliou art), the suffixed pronoun corresponds to that of which i or s is the initial ; but in the possessive compounds, in which w^e should expect to find precisely the same form, we find instead of it a plural possessive of which the initial and radical is n. Thus, the expression thy hand, being kdtes, we should expect to find your hand, kdtesse, or, more primitively, kdtette, like the corresponding Magyar kezetek (from tek, you, another form of te), whereas the form actually used in Finnish is kdtenne. It thus appears that two pronouns of the second person retain their place in the Finnish ; one, the singular of which is si, or more properly ti, the plural te ; and another, hidden in the ancient compounds, the plural of which is we, and of which, by dialectic rules, the singular must have been ni. Even in Turkish, we shall find traces of the existence of a similar pronoun. In the possessive compounds, the second person singular is not represented, as we should have expected it to be, by sen, as the first person singular is by m ; but n or ng is used instead (a nasal which corresponds to that of the Ostiak nen) — e.g., hdha-n, thy father ; and as the final m of hdhd-m is derived from mi or me, I, we seem to be obliged to deduce also the final n of bdhd-n from an obsolete ni or 7ie, thou, which is allied to the corresponding forms that have been pointed out in other Scythian tongues. We find this possessive noing not only in the Osmanli Turkish, but even in the Yakute, the Turkish of Siberia. The same n makes its appearance in the personal terminations of the Turkish verb, sen is more commonly used than n ; but n is found as the representative of the second person in those verbal forms which must be considered as of greatest antiquity — e.g., in the pre- terite of the auxiliary substantive verbs, tdum, I was, tdun, thou wast, idi, he was. In the Oriental Turkish the forms corresponding to these are boldim, boldun, holdi.; and the same termination of the second person singular — the nasal n — appears in all the preterites of that language. We may compare also the plural forms of this pronominal suffix. The Turkish pronouns are pluralised by changing the final formative n into z, or rather by adding 2; to the crude base. Thus, we is biz (for niiz), and you is siz. In possessive compounds i changes into u ; and hence our father is bdbd-muz. In the same manner, your father is bdbd-nuz, indicating a supposititious, isolated pronoun, niz, you, corresponding to miz, we. Whilst u is used instead of i in OsmanH Turkish, the older and more regular i retains its place in the Oriental Turkish — e.g., uzu-niz, you yourselves ; in THE REFLEXIVE PRONOUN 395 which you is niz or nqiz, and from which, when z, the sign of plurality, is rejected, we deduce the singular in or wfi. The same mode of forming the plural termination of the second person appears in all regular Turkish verbs — e.g.^ compare korkdu-nuz, ye feared, with korhdu-n, thou fearedst. We see it also in the imperative korkdu-nuz, fear ye. In all these instances, I consider the Turkish n or ng to be dialectically equivalent to the Finnish n ; and the pronominal root which is thus found to underlie so many Turkish and Ugrian com- pounds of the second person looks as if it might be regarded as identical with the Dravidian, Chinese, and Behistun-Scythian pro- noun. Even the libra tion between i and u, which we noticed in considering the Dravidian forms of this pronoun, meets us again in Turkish. In the Himalayan dialects, we -can scarcely fail to see Dravidian analogies in the Dhimal nd, in the Miri no, in the Garo ndd ; and in the n which forms the first and most essential radical of the pronoun of the second person in all the rest of the Lohitic dialects. Compare also the pronouns of the second person in various Australian dialects — e.g., ninna, nginnee, nginte ; the duals, niwa, nura ; and the plural nimedoo. On a comparison of the various forms of this pronoun which have been adduced above, it must be evident that the affinities of the Dra- vidian ni are almost wholly Scythian ; and this important circum- stance, taken in conjunction with the predominance of Scythian influences over Indo-European in the formation of the first personal pronoun, tends to show that the Dravidian languages stand in closer relationship to the Scythian class of tongues than to the Indo- European. 3. The Eeflexive Pronoun ' Self.' The Dravidian pronouns of the third person are, properly speaking, demonstratives, not personal pronouns ; and they will, therefore, be investigated under a subsequent and separate head. The pronoun which is now under consideration is entitled to a place amongst per- sonal pronouns, because it possesses all their characteristics, and is decKned precisely in the same manner. It corresponds in meaning to the Sanskrit svayam, to the defective Greek e {he) and the Latin sui, sihi, se ; with a range o^ application which is more extensive than theirs. It may almost, indeed, be regarded as a pronoun of the third person, seeing that, when it stands alone as the nominative of a verb, the verb with which it agrees must always be in the third person. 39^ THE PRONOUN In Tamil the nominative singular of this pronoun is tan : the plural of which (by the usual pronominal change of n into m) is idm {tdhgal) ; and the inflexion, or basis of the oblique cases (which, taken by itself, has the force of a possessive), is formed, as in the case of the other personal pronouns, by simply shortening the included vowel — e.g., tan, of self, sui, or (adjectivally) suus, sua, suum. In all its cases and connections tan is found to be more regular and persistent than any other pronoun. The Canarese nominative is tan in the ancient, tdn-u in the modern dialect : the inflexion is formed, as usual, by the shortening of the included vowel ; and the crude root td (without the formative n) is sometimes used instead of tdn-u, just as nd, of the first person, and ni, of the second, are occasionally used instead of ndn-u and nm-u. In Telugu the reflexive pronoun is more regularly declined, and is more in accordance with the Tamil- Canarese, than any other pronoun of the personal class. The nominative is tdn-u, the inflexion and possessive tdn-a, the plural nominative ^am-w. ^ar-wmay be used instead of ^am-w. This appears to be a contracted form of tamar-u, a form also used in poetical Tamil, and meaning they who belong to one's-self. td may be used at pleasure, as in Canarese, for tdn-u. A similar regularity of formation and of declension is apparent in all the Dravidian dialects, so that further comparison of the forms of this pronoun seems to be un- necessary. The root or base is evidently td or ta, self. The final n of the singular, though only a sign of the singular number (like the final n of wa-w,I,and m-w,thou),is one of great antiquity, for we find it even in the Brahui — e.g., the nominative singular is tenat (compare with this the inorganic t, which is suffixed to the personal pronouns in Gond) ; genitive tena, dative tene. tdn, self (like ndn, I, and ntn, thou), is of no gender. The use of this pronoun agrees, on the whole, with the use of the corresponding Indo-European reflexive. When not itself used as the nominative of a sentence, it always agrees with the principal nomina- tive and with the governing verb, that is, with that verb which is in agreement with the principal nominative. It is also used as an emphatic addition to each of the personal and demonstrative pro- nouns, like the Latin ipse, the Sanskrit svayam, or the English self, in the compounds myself, yourself, &c. — e.g., we say in Tamil ndn- tdn, I myself ; ni-tdn, thou thyself ; avan-tdn, he himself ; aval-tdn, she herself ; adu-tdn, itself or that itself ; and tdm,, the plural of tdn, is in like manner appended to the plurals of each of those pronouns and demonstratives. The reduphcated form of the inflexion, tat-tam, THE REFLEXIVE PRONOUN 397 for tam-tam, is used to mean 'theirs respectively.' The Sanskrit svayam is indeclinable ; the Dravidian tan is regularly declined, which is a difference worthy of notice, tan acquires also an adverbial signification by the addition of the usual adverbial formatives — e.g., tdndy (for tdn-dgi). Tarn., of myself, of yourself, or spontaneously ; and when appended to nouns of quality or relation its use corresponds to that of our adverbs really, quite, &c. — e.g., mey tdn, Tarn, it is really true, sari tdn, quite right. In most of the above instances t is a sonant, and is pronounced like soft th or d. One use to which the reflexive is put is peculiar to these languages — viz., as an honorific substitute for the pronoun of the second person ; and in this connection either the singular, the plural, or the double plural may be used, according to the amount of respect intended to be shown. When used in this manner, it is not annexed to, or compounded with, the pronoun of the second person, but is used alone : and though, when it stands alone, it generally and naturally denotes the third person, yet when thus used honorifically for the second person, the verb with which it is connected receives the pronominal terminations, not of the third person, but of the second. This use of tdn as an honorific pronoun of the second person, illustrates the possibility, if not the probability, of the ultimate origin of the Indo-European pronoun tu, thou, from a demonstrative base. A very interesting class of Dravidian words, the nature of which has generally been overlooked, has originated from the honorific use of the reflexive pronoun. Its inflexion, or possessive, has been prefixed honorifically to most of the pure Dravidian words which denote parents and other near relations, in a manner which somewhat resembles our modern periphrasis. Her Majesty, your worship, &c. In general the plural tarn has been used in this connection instead of the singular tan, as a prefix of greater honour. In some instances also the crude base ta has been used as the first member of the compound instead of the regularly organised tam. This class of compounds especially abounds in Tamil, in which also em and nam, our, and um, your, are optionally used in poetry instead of tam or ta, with the same honorific signification. The following illustrations are from Tamil alone. In the other dialects (except Malayalam, which here is in agreement with Tamil), some of the most interesting of these compounds are unknown, or the different numbers of the compound have become so corrupted that it is more difficult to identify them than in Tamil. 39^ THE PRONOUN tambirdn (Mai. tamburdn), God, lord, the abbot of a Saiva monastery : the nearest English is his lordship ; from tarn, used honorifi- cally, and firmi, lord (probably a derivative from the Sans. pra, before), ernhirdn, our lord, and umbirdn, your lord, are also used, firdtti, tambirdtti, lady. Comp. enherumdn {em, our, fermndn, great person), our lord, literally our great one, a title common in poetry and in inscriptions (fern, ferumdtti, lady). tagafpan, father ; from tarn, used honorifically, and appan, father. This word is sometimes pronounced by Brahmans in the ancient manner, tamappan ; in Malayalam it is both tagappan and tammappan : nearest English, his father- hood. tandei, father, his fatherhood ; a more classical word than tagappan, yet almost as common (Can. tande, Tel. tandri, Mai. tanda). There can be no doubt that the first portion of this word is the honorific reflexive tarn, seeing that we find also in the Tamil poets endei {em), nandei {nam), our father ; and undei {um), nundei {num), your father. Comp. also mundei, ancestor, first father, from mun, before. It is difficult to explain tei {dei), the second member of the compound. It is plain that it means father ; but the only word for father at all resembling it in Tamil is attan, father (also dttan, a superior person ; comp. attei, dttdl, mother). If the tei of tandei, Sec, is connected with this word, it must have come from an older abstract form, attei, meaning either father or mother, according to the connection (as tannei, mother, elder sister, is also used in the poets for elder brother) ; and this word attei we might possibly derive from the verbal root attu, to join, to lean upon. (See " Glossarial Affinities, Sanskrit and Scythian.") * tdy, mother, her maternity ; from td, the base of ta^n, used honorifi- cally, and dyi, mother {ta-dyi) ; Can. tdyi. dyi, mother, matron, lady, is a more classical word than tdy, though retained in many compounds in daily use. Another form is ay (Tam.) This is identical in sound with a verbal root signifying to select ; but it is difficult to suppose that select, pretty, can have been the original meaning of one of the most ancient patriarchal Dra vidian words for matron, mother. Another and perhaps more probable derivation is from a, ancient Tam., cow, from which dyi, fem., would THE REFLEXIVE PRONOUN 399 naturally be formed, with the meaning of mistress of the cows. Comp. duhitri, Sans., a daughter, literally a milk- maid, dchchi, matron, is a South Malayalam form for dyi. dyar, Tam.-Mal., the epicene plural of this word, is a common poetical epithet for cowherds. tammei, mother ; from ta, honorific for tarn, and ammei, an honorific word for mother, matron (also annnan, amtnd, ammdl). tannei, mother ; from ta, honorific, and annei, an honorific word for mother, probably identical in origin with ammei. This word means not only mother, but also both elder sister and elder brother. tameiyan, elder brother, his eldership ; from tarn, used honorifically, and eiyan (sometimes ay an), a senior or elder, and therefore meaning also father, elder brother, or guru. Another very common word for elder brother is annan, annul, from annu, to resort to, to lean upon (Tel. anna, Can. antia). Comp. tammun (poetical), an elder brother, from tam and mun, before, his precedence-ship. tamuJcJcei, elder sister, her eldership ; from tam and akJcei, elder sister (also mother). The ordinary Tamil forms are akkd and akkdl. tamhi, younger brother ; from tam, honorific, and pi, a word or portion of a word of doubtful origin and meaning. The Telugu tammudu and the Canarese tamma throw no light on the meaning of pi (Mai. both tamhi and tamhdn). Comp. with pi, peidal, Tam. and Mai., a boy, literally that which is fresh and green. The most probable explanation, though one which is not free from difficulty, is that pi is for pin, after. Comp. tammun, Tam., from tam and mun, before, a poetical word for elder brother, tambi is explained by the native lexicographers as meaning pin-pitandon, he who has been born afterwards. They also give pinnon, he who is after, as a synonym for tamhi, and pinnei, the corresponding feminine or neuter abstract, as a synonym for tangei, younger sister. Probably pi was the primitive shape of pin, as mu was certainly the primitive form of mun ; still it is difficult to see how the formative n (changing to r in pitagu, after), which wa« retained in mun when used as the final member of a compound, happened to be omitted altogether from pin. Equivalent forms of this word in poetical Tamil are emhi, our younger brother, uynhi and numhi, your 400 THE PRONOUN younger brother ; probably also nambi (which see) is to be regarded as another form of the same word. tangei, younger sister ; from tarn, used honorifically, and kei, a word of doubtful origin (Mai. tanga, Can. tangi, Coorg tange). It would seem from the Tamil poetical word nangei, a lady, that kei does not mean one that is young, or one that comes afterwards, as I have supposed the fi of tamhi to mean, but must have had a meaning in some way suitable to be applied to women in general (mangei, a girl, looks as if it included the same kei) ; yet, on the other hand, we find in the Tamil poets this very word kei, in the shape of keiyei, an abstract noun, used as a synonym for tangei, a younger sister. This appears to settle the question as regards the meaning of kei ; but the origin of the word continues doubtful. It cannot be connected with keimmei, keimben, Tam., a widow, that w^ord being most naturally derived from kei (another shape of which is kasu), to be bitter ; hence also the noun kei, adversity. We seem, therefore, to be obliged to fall back on kei, a hand, in the sense of a help, a handmaid, and to explain tangei as meaning her handmaidenship* — a meaning which suits well the position a younger sister would natu- rally have assigned to her. The corresponding Telugu word chellelu, younger sister, includes the meaning of playful, petted. nambi, a title of inferior priests, meaning probably, like tamhi, younger brother (which see). Comp. namhuri, properly nambutiri, the title of a class of Malaya lam Brahmans. Comp. also Telugu tammali, a petty priest. I notice in Coorg two instances of tam used honorifically, which are not in Tamil — viz., tammdvu, father-in-law, from tam and mdvu (Tam. mdman), the same, and tammdvi, mother-in-law, from tam and mdvi (Tam. mdmi), the same. Another remarkable use of the reflexive pronoun is the adoption of its possessive, or inflexional base, tan, of self, or self's, as the base of the abstract noun tan-mei or tanam, quality or nature, literally selfness. tanam is the form of this word used in Telugu. Tamil uses both tanam and tanmei ; but the latter can stand alone, whilst * Compare with this meaning of a younger sister the name of spinster, which is applied by ourselves to unmarried females ; and also the derivation attributed to duhitri {duhitar), Sans, daughter, viz., a milkmaid, the milkmaid of the family. THE REFLEXIVE PRONOUN 4OI tanam is used only in compounds, mei is the regular formative of Tamil abstracts ; like our English ness, the Latin tas, or the Sanskrit twam. tanmei is identical in meaning with the Sanskrit tatvam, nature, property, which is derived from tadov tat, that, and is possibly allied to it in origin, though indirectly. td or to, the base of the Dra vidian reflexive pronoun, has no connec- tion with, or resemblance to, any other pronoun of this family of languages, though it is unquestionably a pure Dravidian root. If we look at its meaning and range of application, it must, I think, have originated from some emphatic demonstrative base ; and it will be found that there is no lack, either in the Indo-European or in the Scythian family, of demonstratives closely resembling ta or ta-n. We see examples of this resemblance in the Sanskrit tat, that (from ta, the demonstrative base, and t, the sign of the neuter singular) ; in tadd, then, at that time ; and also (with the t weakened into s) in sah, he, sd, she. The reflexive pronouns of this family, sva, se, &c., are probably derived from the same base, though considerably altered. Compare also the old Greek article, which is properly a demonstrative pronoun, tos (tos), ri] (te), to (to), and the corresponding German der, die, das. We find the same or a similar demonstrative (with an an- nexed nasal, as in the Dravidian tan) in the Doric rrjv-o^ (ten'Os), he, that, which is the form from which the iEolian Krjv-o ^ 5S . • • g S g "^e -^e ^ ^ ^ s g S g IS Qj ^ o 5= ^ § ^ f _: •^e ^e § s: S g s s? se H §1 n3 c« O f3 r^ "^ cS S O o3 M OHOOMHha!:HPHp • rO ■3 ^ S> . 4J a> cv>. 'S la -5 « g •g ^ S CO K %!. C3 . •» •j^ §1 •S . . ■fe "1 1 ,^ 1 .Is "S U-i r--s. .2 e 'w d &:> < J=y 1 i Oo. « p h) C *l "^g 1 1 Ph p> •! 1 si 5£ i Si Si Si § Si Si si Si ^ >- *li ^* CV.. ■* '1 5J ov.. > 1 r-^ § f^ si t^. 1 Si • Si 'a g ;§ c ^ 1 1 ^ e i 1 S 1 ^^ i ^^ • ,>j • • «<> •«>> •<>> 525 § g si Si si si si si Si Si .4<^ — ' s 53 . ■z> ^ -M q; <»^ •<>i ^ eS +3 ^ Cc^ • t^ ii <;j k •> > ^«^ Si si g 1 .i^ Si '^'^ ^. '1 1 ." 5£ S ^ ^ "A ^ , ?> sf "si • . , gj sf 5 ^iS> ^<:i> ^5 r -.S" i ^^ "CS- •"S" • ^•7* ^ SS Si Si ^^ Si Si si si ^ Si • a • . • • • • ^ • • • c3 % . . . . H Canare Tulu Coorg 1(3 c3 Kota Raj ma Uraon Brahui A O < P 1— 1 r^ ft ft ^^ M H ;> W P H ^ w O Zfl O ^ il y-i xn J^ o H rn '^ Ph ti P^ w PlH tf H p w « 5&5 l-"i &5 ?S ^ i '^ i e e fl ^e ^ %- ?s s tft ^ 5^ ^' .g .$?i • i e 53 g g g g -^ g g § g -< . ^s ^-^ bo ....... . ce • - r 1 1 s ^ -§ .-^ ^ ^ -^ -g .-^ ^ ^ g g •I S C C ^ ""S* • •«*. g g §= 5: •£ g 5Si . -pi . . — r -g v§ g ^, -^ .-^ ^ ^ ^ -g r^ ^ r§ g g OoS^^c3?^cSO-a3vajpiaJ,^^aJ_g 420 THE PRONOUN SECTION II.—DEMONSTRATIVE AND INTERROGATIVE PRONOUNS. It is very difficult to treat the demonstrative and interrogative pronouns of the Dra vidian family separately. The bases are different, but they are built up on those bases in precisely the same manner, and obey one and the same law, so that what is said about the one class may be regarded as said about the other also. I shall discuss them separately as far as possible, but it will often be necessary to treat them together. 1. Demonstrative and Interrogative Bases. 1. Demonstrative Bases. — The Dra vidian languages, like most other primitive uncompounded tongues, are destitute of pronouns (properly so called) of the third person, and use instead demonstra- tives signifying this or that, with the addition of suffixes of gender and number. In these languages ' he,' means literally that man ; ' she,' that woman ; and ' they,' those persons or things. The interroga- tives are formed in the same manner by the addition of suffixes of gender and number to an interrogative base signifying ' what.' The words which signify man and woman have gradually lost the definiteness of their original signification, and shrunk into the posi- tion of masculine and feminine terminations. They are no longer substantives, but mere suffixes or signs of gender ; and are so closely incorporated with the demonstrative bases that it requires some knowledge of the principles of the language to enable us to separate them. In comparison, therefore, with the Turkish and Ugrian lan- guages, in which there is but one pronoun of the third person, the Dravidian languages, which possess a great variety, appear to con- siderable advantage. Nevertheless, the speech of the Dra vidians appears to have been originally no richer than the other Scythian idioms, and to have at length surpassed them only by the Aryan- istic device of fusing that-man, that-woman, that-thing, into single euphonious words. The signification of man and woman still shines through in the masculine and feminine terminations ; but no trace remains of the words by which a thing and things were originally expressed, and which are now represented only by d, the sign of the neuter singular, and a, that of the neuter plural. Four demonstrative bases are recognised by one or another of the Dravidian dialects, each of which is a pure vowel — viz., a, the remote DEMONSTRATIVES AND INTERllOGATIVES 42 1 i, the proximate, and u, the medial demonstrative ; together with e, which is the suffix of emphasis in most of the dialects, but is a demon- strative in Ku. The first two — viz., a, the remote, and i, the proxi- mate demonstrative — are the most widely and frequently used. The medial u is occasionally used by the Tamil poets, more frequently in classical Canarese and in Tulu, to denote a person or object which is intermediate between the remote and the proximate ; and it will be found that it has ulterior affinities of its own. e, the ordinary Dra vidian suffix of emphasis, is used as a demonstrative in Ku alone — in addition however to a and i — e.g., evdru, they. It appears also in the tJraon edah, this, the correlative of hudah, that. The use of e being chiefly emphatic, I refer the reader, for an account of it, to a subsequent head. The ordinary demonstratives of the Dra vidian dialects are the simple, short vowels a, i, and u ; and it will be found that every other form which they assume is derived from this by some euphonic process. 2. Interrogative Bases. — There are two classes of interrogatives in the Dra vidian languages — viz., interrogative pronouns or ad- jectives, such as, who ? which? what? and syntactic interrogatives, such as, is it ? is there ? Interrogative pronouns and adjectives resolve themselves in the Dravidian tongues into interrogative pre- fixes, resembling the demonstrative prefixes already considered, by suffixing to which the formatives of number and gender we form interrogative pronouns. The interrogative particle itself, when simply prefixed to a substantive, constitutes the interrogative adjective what ? {a.) The most common interrogative prefix is the vowel e. In all the Dravidian dialects this prefix is used in the formation of pro- nominals, in precisely the same manner as the demonstrative bases a and i. It forms one of a set of vocalic prefixes (a, i, u, and e), which occupy one and the same position, obey one and the same law, and differ only in the particular signification which is expressed by each. The unity of principle pervading these prefixes will be clearly apparent from the subjoined comparative view. The forms which are here exhibited are those of the Tamil alone ; but in this ■ particular all the dialects agree on the whole so perfectly with the Tamil, and with one another, that it is unnecessary to multiply examples. I exhibit here 2^ alternative (probably an older) interrog- ative base in yd, which will be inquired into further on. 422 THE PRONOUN Mas. sing. Fern. do. Neut. do. Epic. plu. Neut. do. rroximate Demonstrative ivan, hie. ival, lisec. idu, hoe. ivar, hi, hse. ivei, hsee. Remote Demonstrative avan, ille. aval, ilia. adu, illud. avar, illi, illso. avei, ilia. Intermediate Demonstrative uvan udu. uvar. uvei. Interrogative e or ya. evan or yavan, quis ? e^aZ or ydval, quae ? ecZt* or yddu, quid ? CTar or ydvar, qui ? qu8D ? evei or ?/ai;e*, quae ? I need not eall attention to the beautiful and philosophical regular- ity of this quadruple set of remote, proximate, and intermediate demonstratives and interrogatives. In no other language or family of languages in the world shall we find its equal, or even its second. In addition to which, the circumstance that the demonstrative vowels are not only used in these languages with an invariable and exact discrimination of meaning which is not found in the Indo-European tongues (with the solitary and partial exception of the New Persian), but are also associated with a corresponding interrogative vowel of which the Indo-European tongues are totally ignorant, tends to confirm the supposition which I have already expressed, that the Dravidian family has retained some prse-Sanskrit elements of im- mense antiquity ; and, in particular, that its demonstratives, instead of being borrowed from Sanskrit, represent those old Japhetic bases from which the demonstratives of Sanskrit itself, as well as of various other members of the Indo-European family, were derived. (h.) The other interrogative base of the Dravidian languages is yd. yd is not used at all in Telugu, but is largely used in Canarese, and somewhat more rarely in Tamil. Probably there was originally only one interrogative base, and if so, it must have been yd, and e must have been corrupted from it. The process by which yd became e is tolerably clear, a evinces a tendency to be weakened into e. (See " Part I., Sounds.") We have seen an illustration of this in the cir- cumstance that the Sanskrit yama, the name of the god of death, becomes in Tamil ema(n), pronounced yema{n). In Tulu, ydr, who, becomes yer\ This is a considerable step towards e. Then, also, e is DEMONSTRATIVES AND INTERROGATIVES 423 commonly pronounced as ye, and e as ye ; and in Telugu this y is frequently written, as well as heard. This would facilitate the omission of the y in writing, when yd came generally to be weakened into ye. e alone would in time have the same force as ye, and would come to be regarded as its equivalent. The long form e still survives . in the Malayalam evan, eval, he, she, for evan, eval ; and in the Tamil and Malayalam edu, and the Telugu edi. In Telugu e sometimes directly corresponds to the Tamil yd — e.g., compare ydrdu, Tam. where, when, a year (nasalised from yddu), with the Telugu edu, where, edi, a year. We see also this long interrogative e in the Telugu ela, how, in what manner, compared with dla, ila, in that manner, in this manner. " There is a remarkable change in Canarese of the interrogative yd into da. We may say either ydvan-u or ddvan-u, what man ? ydval-u or ddval-u, what woman ? ydvadu or ddvadu, what thing ? So also the crude interrogative is ydva or ddva, who, which, what ? In Tulu we find the same dd, which ? alternating with vd and vova ; also ddne, what ? ddye, why ? In these instances the analogy of the other dialects leads me to conclude yd to be the older and more correct form of the interrogative base. In yer, who ? yd appears as ye, which is a very trifling change. The Gond interrogative hd and ho appear to be hardened from yd, like the Tulu vd. In High Tamil, yd is not only prefixed adjectivally to substantives (like a, e, and e) — e.g., yd-{k)kdlam, what time ? but it is even used by itself as a pronoun — e.g., yd-(s)seyddy, what hast thou done ? It forms the basis of only one adverbial noun — viz., ydiidu, Tam. when ? a year, a correlative of dndu, then, and tndu, now. The only use to which yd is put in the colloquial dialect of Tamil, is that of forming the basis of interrogative pronouns ; a complete set of which, in Tamil as well as in Canarese, are formed from yd — e.g., ydvan, quis ? ydval, quw ? yddu, quid ? ydvar, qui, quce ? ydvei, quce ? The Canarese interrogative pronouns accord with these, with a single unimportant exception. The neuters, singular and plural, of the Canarese are formed from ydva, instead of yd — e.g., ydvadu, quid ? (for yddu,) and ydvavu, quce ? (for ydva.) This additional va is evidently derived by imitation from the euphonic v of ydvanu, he, and its related forms ; but it is out of place in connection with the neuter, and is to be regarded as a corruption. In Tamil, a peculiar usage with respect to the application of the epicene plural ydvar, qui, quce, has obtained ground. It is largely used in the colloquial dialect, with the signification of the singular as well as that of the plural, 424 THE PRONOUN though itself a plural only and epicene ; and when thus used, ydvar is abbreviated into ydr — e.g., avan ydr, who is he ? (literally he who ;) aval ydr, who is she ? ydr has also been still further corrupted into dr, especially in compounds. 1. Demonstrative and Interrogative Pronouns. — The original char- acter of the demonstrative bases, like that of the interrogative, is best exhibited by the neuter singular, the formative of which does not commence with a vowel, like an and al (Tamil), the masculine and feminine suffixes, but consists in a single consonant, d, followed by an enunciative vowel — that is, a vowel intended merely as a help to enunciation. This vowel is i in Telugu, a very short u in the other languages. The remote and proximate neuter singulars are in Telugu adi, idi, that (thing), this (thing) ; the interrogative edi, what (thing) ; in Tamil, Malay alam, and Canarese they are adu, idu (with the inter- mediate udu), and edu. In Gond the demonstratives are ad, id. The anomalous forms of the Tulu and the Tuda will be considered further on. d having already been shown to be the sign of the neuter singular used by pronominals and appellatives, and there being no hiatus between a, i, or u and d, and therefore no necessity for euphonic insertions, it is evident that the a, i, and u of the neuter singulars cited above constitute the purest form of the demonstrative bases. The suffixes which are annexed to the demonstrative bases a, i, and u, for the purpose of forming the masculine and feminine singulars and the epicene and neuter plurals, commence with a vowel. Those suffixes are in Tamil an for the masculine, al for the feminine, ar for the epicene plural, and ei or a for the neuter plural ; and v is the consonant which is most commonly used to prevent hiatus. The following, therefore, are the demonstrative pronouns of Tamil — viz., avan^ ille ; ivan, hie ; aval, ilia ; ival, hsec ; avar, illi ; ivar^ hi ; avei, ilia ; ivei, hsec. To these must be added the intermediates uvan, uval, udu, uvar, uvei, which do not admit of being translated by a single word. I quote examples from Tamil alone, because, though different formatives of number and gender are sometimes annexed in the other dialects, those differences do not affect the demon- strative bases. The anomaly which w^ill be noticed in the case of Tulu will be found, when examined, to be only apparent. All the above suffixes of gender have already been investigated in the section on " The Noun." The mode in which they are annexed to the demonstrative bases is the only point which requires to be examined here. DEMONSTRATIVES AND INTERROGATIVES 425 The demonstrative bases being vocalic, and all the suffixes, with the exception of the neuter singular, commencing with a vowel, some euphonic consonants had to be used to keep the concurrent vowels separate and pure, v, though most frequently used to prevent hiatus, is not the only consonant employed for this purpose. The Ku being but little attentive to euphony, it sometimes dispenses altogether with the euphonic v, and leaves the contiguous vowels uncombined — e.g., ddnju, he ; ddlu, she. Even Tamil sometimes combines those vowels instead of euphonically separating them — e.g., ydvar, who ? is commonly abbreviated into ydr ; and this is still further softened to dr in the colloquial dialect. In the higher dialect of Tamil, n is often used euphonically in- stead of V, especially in the personal terminations of the verbs. Thus, instead of irunddn (for irundavan), he was, the poets some- times say irundanan ; and for irundava, they (neuter) were, the form which we should expect to find used, irundana is universally used instead. This euphonic v has in some instances come to be regarded as an integral part of the demonstrative itself. In the nominative plural of the Gond neuter demonstrative, the final and characteristic vowel a has disappeared altogether, without leaving any repre- sentative — e.g., av, those (things) ; iv, these (things). In the oblique cases a is represented by e. In Telugu, though the nominatives of the neuter plural demonstratives avi and ivi use v merely as an euphonic, yet in the oblique cases, the bases of which are vd and vt, the demonstrative vowels have got displaced, and v stands at the beginning of the word, as if it were a demonstrative, and had a right per se to be represented. In the masculine singulars vddu, ille ; vidu, hie ; and in the epicene plurals vdru, illi ; viru, hi, v euphonic has advanced a step further, and assumed the position of a demon- strative in the nominative as well as in the inflexion. That this v, however, is not a demonstrative, and that the use to which it is put in Telugu is abnormal, is shown by the fact that in da and di, the inflexions of adi and idi, illud and hoc, the neuter singular demon- stratives of the Telugu, d, though certainly not a demonstrative, nor even euphonic, but simply a sign or suffix of neuter singularity, has been advanced to as prominent a position (by a similar euphonic displacement) as if it belonged to the root. Compare especially the corresponding Telugu^interrogative. In Tulu the proximate neuter singular demonstrative is indu or undu, the remote avu. indu and undu correspond to the Tamil proximate idu and intermediate udu : the only difference consists 426 THE PRONOUN in the nasalisation of the d. avu, the remote demonstrative, though a neuter singular, is identical in form with the Canarese avu, they (neuter). The v of avu seems to be merely euphonic, as it disappears altogether in the plural, which is not avukulu, but eikulu {avu =ayu=ei). The corresponding masculine pronoun is dye, he, in which y is used euphonically where v would have been used in Tamil. In the feminine dV, she (Tam. aval), even the y has disappeared, and the two contiguous vowels have coalesced. The proximate pronouns of the Tulu masculine and feminine singular and plural present several peculiarities, imbe, he {hie), corresponds to the Tamil ivan, the Old Canarese ivam. The euphonic v of those languages seems to have been hardened into m, and this m to have become mh. The plural of the same is mer' (the remote is dr, for avar). The feminine proximate she (ha3c) is moV, the plural of which is mokulu. mer stands for ivar=imar, and moV for ival—imal. Compare the appar- ent disappearance of the demonstrative bases i and a in the Telugu viru and vdru, they, proximate and remote, for ivar and avar. See also " The Noun," epicene plural, in mar. The same peculiarity appears in the Tulu demonstrative adverbs, avulu, there, corre- sponds with similar words in the other dialects (Can. alii) ; but mulu, here, presents the same peculiarity as mol, hsec. In the Tuda dialect the pronoun of the third person is the same for both numbers and for all three persons, Hke the Sanskrit reflexive pronoun svayam. atham represents everything of which ' that ' can be predicated ; itham is the equivalent for this. With atham, itham, compare the Telugu atadu, atanu, dtandu, dtadu, itadu, itanu, Uadu, Uanu ; the Old Canarese singular masculines dtam, itam-, utam. The final am of the Tuda is occasionally dropped. Tamil possesses a complete set of abstract demonstrative and inter- rogative nouns of perfect regularity and great beauty. I class them here (for convenience of comparison) with demonstrative and inter- rogative pronouns ; but they are in reality nouns, expressing abstractly the ideas that are embodied in the pronouns in a concrete shape. They consist of the demonstrative and interrog- ative vowel bases {a, i, u, e), with the addition of 7nei, the ordinary formative of abstract nouns, which we have already noticed in tan-mei, nature, literally selfness, in the section on the reflexive pronoun tdn. The initial consonant of mei is doubled by rule after the demonstrative and interrogative vowels. The words referred to are immei, thisness ; ammei, thatness ; ummei, an intermediate position between that-ness and this-ness ; emmei, what-ness. In use, DEMONSTRATIVES AND INTERROGATIVES 427 the words chiefly denote the different states of being or births. immei, the present state or birth, is the only word of the set in com- mon use ; the rest are found only in the poets, ammei (common equivalent mavumei, otherness) denotes the future birth ; ummei, the birth before the present ; emmei, what birth ? generally found with the addition of um, and so as to give the meaning ' in what- soever birth.' We have seen that the neuter singular of the demonstrative and interrogative pronouns, properly so called, is formed by the addition of the neuter formative d to the vowel bases a, i, u ; e or yd. There are traces also of the existence of two classes of pronouns formed by means of the addition to the same vowel bases of m, the equivalent of which is n, or of I. Pronominals ending in I are used chiefly as adverbs of place and mode. There are exceptions, however — e.g., alia, Tel. that, has the force of an adjective {alladi, that thing). See Adverbs : formative I, V . The demonstrative pronouns and pronominals ending in m or n are not free from doubt. I shall, therefore, adduce first the interrogatives belonging to this class, about which no doubt can be entertained. Each of the dialects possesses a neuter interrogative pronoun, formed from the interrogative base e or e, and the neuter formative n or m. This formative is more abstract than d, but less so than mei. ed-u means which ? en, what ? In Tamil we find en, what ? from which is formed the singular appellative ennadu, what thing ? and the plural enna, what things ? en is also lengthened into en, the ordinary meaning of which is why ? Though enna is properly a plural neuter, it has come to be used also as a singular, and is even turned colloquially into a singular neuter noun, ennam — e.g., ennamdy, how ? Malayalam uses en, like Tamil, meaning what ? rather than why ? but does not use en ; instead of this we have endu, what ? which, however, is probably the Malayalam shape of the Tamil ennadu=en-du. In Canarese enu is not a mere interrogative particle, but a regularly declined interrogative pronoun, like the vulgar Tamil ennam. We have substantially the same word in the Telugu emi, w^hat ? why ? emi bears the same relation to edi, Tel. what (thing) ? that en in Tamil bears to edu. The only difference is in the use of the more abstract rt or m as a neuter formative, instead of d, which gives more distinctly the sense of the neuter singular. In the compound word emo, Tel. I know not what (Tam. Mai. Can. end), from em and 6, the particle of doubt, we see that emi is a secondary form of em ; and by the help of Tamil we are able to trace 428 THE PRONOUN this em back to the shorter form em. eni, which I consider the equivalent of emi, is used in the conjugation of Telugu verbs as a conditional particle ; properly it implies a question. We now return to the demonstratives which appear to be formed from the demonstrative vowels a, i, u, with the addition of m or n. am, that, appears to survive in the am which is used so largely as a formative by neuter nouns in Tamil and Malayalam ; and possibly also in am, which seems to be the oldest sign of the Dra\ddian accusative case. In each of these instances an is often used instead of am. See the sections treating on these formatives and case-signs in Part III., " The Noun." im shows itself in the Canarese sign of the ablative case, originally a locative, and in the corresponding Tamil in, wath w^hich il corresponds. The primitive meaning seems to be this place, here, and hence, a place, a house. Both at and il appear also in verbal derivatives, especially in Tamil, in which, e.g., the number of nouns derived from verbal roots which take al or il as their formative, is almost as large as those which take am or an. Dr Gundert derives from am, or im the Tamil demonstrative adjec- tives anda, that, inda, this ; and I presume would attribute the same origin to the Telugu and Canarese adjectives anta, inta, &c., which are more or less demonstratives in meaning. On the whole, however, I still prefer to regard these forms as nasalised from ad\ that, id\ this. We had an instance of this nasalisation before us just now in the Tulu pronoun indu, undu, this (thing), which must be identified with the idu, udu of the other dialects. On the other hand, I have no doubt of the origin of inda, the Canarese sign of the ablative, from im ; and the Tamil adverbial nouns andru, indru, endru, that day, to-day, what day, seem to be formed either from am, im, em, or from al, il, el. See the Demonstrative and Interrogative Adverbs. A very interesting inquiry remains. Is um, the Tamil-Malayalam particle of conjunction, and, even (Tel. u, classical Can. um, am ; coll. Can. u), to be regarded as a demonstrative pronoun, formed from u, the intermediate demonstrative base, and the formative m, corresponding in origin to the demonstrative am. and im, and also to the interrogative em, considered above ? That this is the origin of um is one of the many ingenious suggestions contained in Dr Gundert's communication. In his Malayalam dictionary he prefers to derive um from u, the supposed root of the verbal noun uyar, height, with the meaning of above. In classical Canarese am is sometimes used as the equivalent of um ; and this seems to connect the particle at once with the demonstratives. In Tamil poetry we DEMONSTRATIVES AND INTERROGATIVES 429 find an adverbial demonstrative of place, umhar, with the meaning of the intermediate demonstrative u, the correlatives of which are amhar, that place, imbar, this place, and emhar, which place ? umhar means literally a place intermediate between two other places ; but it is remarkable that it is also used in a secondary sense to signify on, upon, above, and even uyar, height. We thus get for urn, the conjunctive particle, the meaning above, which is one that suits it exceedingly well, without any inconsistency with its ultimately demonstrative origin, um at the end of verbs changes occasionally in the Tamil poets to undu, which reminds one of the undu, this (thing), and also yes, of the Tulu. 2. Demonstrative and Interrogative Adjectives. — When the demon- strative bases a and i are simply prefixed to substantives, they convey the signification of the demonstrative adjectives that and this. When prefixed, they are indeclinable ; but on thus prefixing them to sub- stantives, either the initial consonant of the substantive is euphoni- cally doubled — e.g., anndl {a-{n)-ndl), Tam. that day ; or if this euphonic doubling is not resorted to, the demonstrative vowels are lengthened. Tamil invariably adopts the former plan : the latter is more common in Malayalam and Canarese. When the sub- stantive commences with a vowel, and v is inserted as usual to prevent hiatus, Tamil, by a dialectic rule of sound, doubles this v, as if it were regarded as an initial consonant — e.g., when ur, Tam. a village, receives this prefix, it becomes not avur {a-(v)-ur), but avvur. The origin of this doubling of the initial consonant of the word to which the demonstrative vowel is prefixed, is to be ascribed to the emphasis which is necessarily included in the signification of the demonstrative. Through this emphasis a and i assume the character, not of ordinary formatives, but of qualifying words ; and the energy which they acquire influences the initial consonant of the following substantive, which is no longer an isolated word, but the second member of a compound. In the same manner and from a similar cause, when Sanskrit words which commence with a privative are borrowed by Tamil, the consonant to which a is prefixed is often doubled, at least in the colloquial dialect — e.g., anndnam {a-{n)-ndnam), ignorance. The occasional lengthening of the demonstrative vowels, when used adjectivally, in Malayalam, Canarese, and the other dialects (without the doubling of the succeeding consonant), is merely another method of effecting the same result. The emphasis which is imparted in this manner to the demonstrative, is equivalent to that which the doubled consonant gives ; and hence when the demonstrative vowels are 430 THE PRONOUN lengthened, from a and I to a and i, the succeeding consonant always remains single. The fact that the demonstrative vowels are short in the pronouns of the third person in each of the Dra vidian dialects without exception, shows that those vowels could not originally have been long, and that the use of long a and i as adjectival prefixes, instead of a and ^, is owing to emphasis. Some curious illustrations of the lengthening of a vowel through emphasis alone, are furnished by the common speech of the Tamil people — e.g., adigam, much, large — a word which is borrowed by Tamil from Sanskrit — when it is intended to signify very much, is colloquially pronounced adigam. Similar instances might be adduced from each of the colloquial dialects. The. only peculiarity which requires notice in the use of the inter- rogative prefix e, is the circumstance that it is occasionally lengthened to e, precisely as a and i are lengthened to a and i. In Tamil this emphatic lengthening is very rare. It is found only in the neuter singular interrogative pronoun edu, what or which (thing ?), quid ? which sometimes, especially in composition, becomes edu ; and in the interrogative en, what, why ? which is ordinarily lengthened to en. In Malayalam edu and en have entirely displaced edu and en. In Telugu also this increase of quantity is common. It appears not only in emi and ela, why ? but is often used as the interrogative prefix, where Tamil invariably has short e. Thus, whilst Tamil has evvidam, what manner ? how ? Telugu says either evvidhamu or evidhamu. So also, whilst Tamil occasionally only uses edu, quid, instead of the more classical edu, the corresponding interrogative of Telugu is invariably edi, and its plural evi. On the other hand, the Telugu masculine interrogative pronoun evvadu, quis ? preserves the same quantity as the Tamil evan ; and even when the prefix is used adjec- tivally, it is sometimes e (not e) as in Tamil — e.g., eppudu, what time ? when ? and epudu, epdu, in poetry, but not epudu. In the Tulu interrogatives of time, e is the interrogative base ; in those of place — e.g., olu, where (pronounced ivSlu) — e is replaced by 6. In addition to the use of the simple vowels a, i, and e, and their equivalents a, i, and e, as demonstrative and interrogative adjectives, much use is also made in Tamil of a triplet of adjectives derived from the above. The simple vowels may be styled merely demonstrative prefixes. The adjectives referred to may be called by right demon- strative adjectives. They are anda, that, inda, this, enda, which ? or what ? — e.g., anda maram, that tree, inda nilam, this land, enda dl, which person ? These demonstrative and interrogative DEMONSTRATIVES AND INTERROGATIVES 43 1 adjectives are unknown to the other dialects of the family. They are unknown even in Malayalam, and in the higher dialect of Tamil itself they are unused. They appear to have been developed in Tamil subsequently to the separation from it of Malayalam, and subsequently to the first beginnings of its literary cultivation. We find demonstrative and interrogative adjectives similar to these in form, and probably in origin, but differing some- what in meaning, both in Telugu and in Canarese. The Tamil anda, inda, enda, mean simply that, this, which ? the parallel Telugu and Canarese words have the meaning of such, like that or this, so much, &c., and are used more like adverbs than like adjectives. They are in both languages anta, inta, enta, with a few dialectic differences of no importance. Connected with these is the Tam.-Mal. adjective inna, such and such — e.g., inna ur, such and such a town. There is no corresponding adjective derived from a. The final a of all these adjectives is clearly identical with the a which is one of the most common formatives of the relative participle, and the most common case-sign of the possessive, by means of which also so many adjectives are formed. The first part of these words (and', anf, &c.) has been considered above under the head of " Demonstrative Pronouns." I should here add the Telugu triplet of adjectives itti, atti, etti, this like, that like, what like ? Also the Canarese triplet, with a signification partly adjectival, partly adverbial, initu, anitu, enitu, this much, that much, how much ? With this is connected the Telugu set of secondary pronouns, indaru, so many people, inni, so many things, with their corresponding remote and interrogative forms, andaru, anni ; endaru, enni. The demonstrative and interrogative bases il, al, el are used, as has been mentioned, almost exclusively as adverbs. One of them makes its appearance in Telugu as an adjective, viz., alia, that [e.g., alladi, that thing). Both in Tamil and Malayalam the demonstrative pro- nouns adu, idu are often used instead of the demonstrative adjectives a, i, anda, inda, in Tamil, and a, t in Malayalam — e.g., adu hdriyam, Tam. that matter, adu porudu, Mai. that time. This usage illustrates the manner in which I suppose anda, &c., to have been derived from adti, &c. • 3. Demonstrative and Interrogative Adverbs. — All Dravidian ad- verbs, properly speakings are either nouns or verbs. Adverbs of manner and degree are mostly infinitives or gerunds of verbs. Adverbs of place, time, cause, and other relations are mostly nouns. Some of those adverbial nouns are indeclinable, and those of them 432 THE PRONOUN which are capable of being declined are rarely declined. Whether declined or not declined, they have generally the signification either of the dative or of the locative case. The latter is the more usual, so that words literally signifying that time, what time ? really signify at or in that time, at or in what time ? Any noun whatever, conveying the idea of relation, may be converted into a demon- strative or interrogative adverb by simply prefixing to it the demon- strative or interrogative vowels. There is a class of words, however, more nearly resembling our adverbs, formed by annexing to the demonstrative and interrogative vowels certain formative suffixes. The suffix is not of itself a noun, like the second member of the class of words mentioned above. It is merely a formative particle. But the compound formed from the union of the vowel base with the suffixed particle is regarded as having become a noun, and is treated as such, though in signification it has become what we are accustomed to call an adverb. A comparison of the demonstrative and interrogative adverbs of the various dia- lects shows that the same, or substantially the same, word is an adverb of place in one dialect, an adverb of time in another, an adverb either of place or of tiine, as occasion may require, in a third, and an adverb of mode or of cause in a fourth. It seems best therefore to arrange them, not in the order of their meanings, but in the order of the different suffixes by means of which they are formed. (1.) Formative Jc, g, n. — Tam. ingu, tngu ; angu, dngu ; engu, ydngu, here, there, where ? Can. tga, dga, ydrdga, now, then, when ? Mge, hdge, hydge, in this manner, in that manner, in what manner ? ydke, why ? Cond, hoke, thither, hike, thither, haga, aga, there, iga, here, baga, where ? inga, now. I consider the Tamil angu, &c., nasalised from agu. The primitive unnasalised form is seen in the Canarese and Gond. The change of the gu of the other dialects into ngu in Tamil is exceedingly common. The resemblance between the Gond iga, here, and the Sanskrit iha, here, is remarkably close ; yet there is no appearance of the Gond word having been borrowed from the Sanskrit one. The demon- strative base i is, as we have seen, the common property of the Indo- European and the Dravidian languages ; but though iga seems to bear the same relation to iha that eg-o bears to ah-am, yet the Dravidian formative k, g, ng, by suffixing which demonstrative vowels become adverbs of place and time, and so many nouns are formed from verbs, does not seem to have any connection with the merely euphonic h of iha. Comp. Mongolian yago, what ? DEMONSTRATIVES AND INTERROGATIVES 433 (2.) Formative ch, /, n. The only instances of this are in Tulu. inchi, anchi, onchi, hither, thither, whither ? incha, ancha, encha, in this, that, what manner ? In Tinnevelly, in the southern Tamil country, inge, here, is vulgarly pronounced inje. (3.) Fonnative t, d, n. Tamil (classical dial.), %ndu, here, in this present life, in this manner; dndu, there (vulgarly, but ■ erroneously used for ydndu, a year) ; ydndu, where ? when ? a time, a year, dttei, annual, should be ydttei. wan, avan, evan, here, there, where ? Telugu, ita, ata, eta, here, there, where ? itu, atu, etu, in this, that, what manner ? tda, dda, eda, here, there, where ? From eda, with the secondary meaning ' when,' comes edu, a year. Tulu, ide, ade, ode, hither, thither, whither ? We see now that the primitive, unnasalised form of the Tamil ydndu must have been yddu, formed regularly from yd-\-du, like edu, which ? from e-\-du. (4.) Formative t, d, n, also ndr. Tamil, indru, andru, endru (secondary forms, ittrei, attrei, ettrei) ; Canarese, indu, andu, endu ; Malayalam, inn' , ann\ enn' ; Tulu, ini, dni, eni. In each case the meaning is the same — viz., this day, that day, what day ? or now, then, when ? In the Telugu, indu, andu, endu, we have evidently the same triplet of words. The only difference is that they are used as adverbs of place, not, as in the other dialects, as adverbs of time. They are used to mean, in this, that, what place — i.e., here, there, where ? indu and andu have acquired the special meaning of, this life and the next, here and hereafter, like the Tamil immei, ammei ; and andu, there, is commonly used as the sign of the locative case, like the Canarese alii. In all the dialects these adverbs are declinable. In form they are simply nouns. It appears on the whole most probable that these words have been nasalised from the pronouns idu, adu, edu. There is a peculiarity in the Tamil form of these words, consisting in this, that ndr suggests the idea that andru is formed from al, that, like the corresponding andru, not, it is not (from al, not-\-du), or endru, classical Tam. the sun (from el, the sun, tim.e-\-du) ; but the testimony of the other dialects does not confirm this idea. As, however, in Tamil endru (the sun) is formed from el, so another endru is formed from en — viz., endru, having said, which is from en-\-du. ^ (5.) Formative mb. Tamil-Malayalam, imbar, amhar, embar, here, there, where ? The formative mb is as commonly used in the formation of deriva- 2 F 434 THE PRONOUN tive nouns as ng, but the demonstrative adverbial nouns formed from mb are now obsolete. They survive in poetry alone. The final ar is the equivalent of al. Strange to say, there is an interrogative in Mongolian which looks almost identical with this, yambar, what ? This might be supposed to be a mere accident were it not that the Mongolian yambar is formed from the interrogative base ya, which is also the true, primitive Dravidian base. This base appears also in the Mongolian yage, what ? (6.) Formative I, I. Canarese, illi, alii, elli, here, there, where ? In Telugu il, the proximate, is not used as a demonstrative, but survives in ilu, illu, a house, the root-meaning of which appears to be this place, here. The longer form oithis word, however, is used demonstratively — e,g.^ lid, in this manner ; ala, there, did, in that manner ; elli, where ? elli is used also to mean to-morrow (in Tulu elle is to-morrow) ; ela, eld, in what way ? These words show that I holds an important place amongst the demonstrative and interrogative formatives. In some Tulu adverbs I is replaced by the lingual / — e.g., mulu, avalu, olu, here, there, where ? The existence in Tamil of demonstratives and interrogatives formed from I, like those we find in Telugu and Canarese, is by no means certain, but traces of them, particularly of the interrogative el, may, I think, be discovered, el is not now used directly as an in- terrogative, but there are many words formed from el, the meanings of which seem to me to presuppose the existence of a primary in- terrogative sense. Compare ydndu, Tam. a year, primarily where ? when ? also Tel. edu, a year, primarily where (eda) ? I shall here set down the various meanings of the Tamil el in what appears to me to be the order of their growth. It will be found, I think, that they include the words for ' a boundary,' and for ' all,' not only in Tamil, but in all the Dravidian dialects. (1.) What, where, when ? as in Canarese and Telugu (supposititious meaning). (2.) A period of time, a day, to-morrow (compare Telugu and Tu|u), the sun (the cause of day), night (that being also a period of time). Other forms of this word are elvei, elvei, time, a day ; elli, ellavan, endru (el-\-du), endravan, the sun. The meaning of the sun appears in erpddu, properly el-'pddu, sunset, elli means night, as well as the sun. (3.) A boundary. This in Tamil is ellei, old Tamil elgei {gei, a formative of verbal nouns). This word means in Tamil, not only a boundary, but also a term, time, the sun, end, the last. There DEMONSTRATIVES AND INTERROGATIVES 435 appears to me no doubt of the identity of this word with meaning No. 2. The meaning of boundary is derived from that of termination. Compare the poetical compound ellei-(t)-ti, the last fire, the fire by which the world is to be consumed. (4.) All. This stage of development is more doubtful, but I find that Dr Gundert agrees with me here, at least as to el, the first part and base of the word meaning a boundary. I explain el to mean ' whatever is included within the boundary,' everything up to the last. Dr Gundert thinks ell-d a negative, meaning boundless. This would be a very natural derivation for a word signifying all, but I am obliged to dissent, as I find no trace of this a of negation in any of the older poetical forms of this word in Tamil — e.g., el-dm, all we, el-ir, all ye. The colloquial word elldm (properly elldvum) is not to be confounded with the classical word eldm, all we. It does not contain the meaning of ' we.' The d of el{l)-d-{v)um is the abbreviated relative participle of dgu, commonly used as a connective or continuative link, and meaning properly ' that which is.' um is added in Tamil to give the word a universal application. This use of um confirms me in the idea that el, all, is identical not only with el, a boundary, but with el, what ? The latter and primitive meaning seems to me to shine through that of a boundary, and to throw light on that of all. Just as evan-um, who-and, means whosoever, so if el were originally an interrogative, el(l)-d-{v)mn would naturally be used to mean whatsoever, all. The Tamil ellavan, the sun, from el, when ? time, is a singular noun. Pluralise it, and we get ellavar, which is a classical Tamil form of the word all. We may safely, therefore, I think, conclude that these words are identical. The traces we find in Tamil of the existence of demonstratives in il and al are more indistinct than those of the interrogative el ; but if an interrogative en, en, pointed to the existence of the corresponding demonstratives in, im, an, am, we may reasonably regard the existence of il and al as testified to by the existence of el. We find il in the locative case-sign alternating with in, and mean- ing also ' house ' ; also, I think, in verbal nouns ending in il, such as katt-il, a cot, vand-il, a wheel, a cart, al we find in a still larger class of verbal nouns, such as kad-al, the sea, in which al seems to be equivalent to am and an {e.g., dr-am, depth, Jcad-an, debt). The most conclusive illustrations o# the use in Tamil of il and al as demonstra- tives, and of el as an interrogative, would be furnished by indru, andru, endru, this day, that day, what day ? if we could be sure that they are formed from a base in /, and not from one in n or m. The 436 THE PRONOUN peculiar combination ndr could be derived from either. Thus, en-\- du, having said, becomes endru, and equally also el-{-du, the sun, becomes endru. Considering the identity of endru, the sun, with el, the sun, time, a day, to-morrow, it seems to me probable that endru, what day ? must be the same word, and if so, indru and andru, this day, and that day, will become representatives, not of m and an, but of il and al, and the original existence of demonstratives in il and al will then be placed beyond the reach of doubt, andru in Tamil, though derived from al, might possibly become andu, annu, in the other dialects. On the whole, however, the evidence of those dialects is unfavourable to this supposition. The Dravidian negatives il and al bear a strong apparent resem- blance to demonstratives, il negatives existence (there is not such a thing) ; al negatives attributes (it is not so and so), al, Tam. as a verbal root, means to diminish, and as a noun, means night {alii, night, a night flower). No similar extension of the idea of negation seems to proceed from il. il and al resemble demonstratives not only in sound, but in the structure of the derivatives formed from them. Compare andru, it is not, with andru, that day ; indru, there is not, with indru, this day. I am unable, however, in this matter, to go beyond resemblance and conjecture. No connection between the demonstrative and negative meanings of il and al seems capable of being historically traced. Affiliation of Demonstrative Bases : Extra- Dravidian Affinities. — There is only a partial and indistinct resemblance between the remote a, proximate i, and medial u, which constitute the bases of the Dra- vidian demonstratives, and the demonstratives which are used by the languages of Northern India. In Bengali and Singhalese, e is used as a demonstrative ; in Marathi M, hi, hen : in Hindustani we find vuh, that, yih, this ; but in the oblique cases the resemblance increases — e.g., is-Jco, to this, i is used as the proximate demon- strative in the North Indian languages more systematically than a or any corresponding vowel is used as the remote — e.g., Marathi iJcade, here ; Hindi idhar, hither ; Mar. itake, so much. The Sindhi proximate is M or he. In the Lar dialect, h is commonly dropped, and the base is seen to be i, as in the Dravidian tongues. The remote in Sindhi is hu or ho ; in Lar u or 6. A general resemblance to the Dravidian demonstrative bases is apparent in several of the Himalayan languages — e.g., Bodo imhe, this, hohe, that ; Dhimal i, u ; t^raon edah, hudah. The Rajmahal eh and ah are perfectly identical with the Dravidian demonstratives, DEMONSTRATIVES AND INTERROGATIVES 437 and form another evidence of the Dravidian character of a portion of that idiom. The connection which appears to subsist between the Dravidian medial demonstrative u and the u of the "Oraon and Dhimal is deserving of notice. Perhaps the Dravidian medial u (Dhimal u, IJraon hudah) may be compared with the Old Hebrew masculine- feminine pronoun of the third person, M ; and thus with the Old Persian remote demonstrative hauva, of which the first portion appears to be hu, and the second ava — which ava forms the base of the oblique cases. It may also be compared with the u or which forms the remote demonstrative in some of the Scythian languages — e.g., Finnish tuo, that, tdma, this ; Ostiak toma, that, tema, this. Compare also the Hind, vuh, that ; Bodo hole. The Magyar demon- stratives are more in accordance with the Dravidian a and i — e.g., az, that, ez, this. The demonstratives of the other languages of the Scythian family {e.g., the Turkish bou, that, ol, this) are altogether destitute of resemblance. When we turn to the languages of the Indo-European family, they appear in this particular to be closely allied to the Dravidian. Throughout that family both a and i are used as demonstratives ; though not to so large an extent, nor with so perfect and constant a discrimination between the remote and the proximate, as in the Dravidian family. In Sanskrit a is used instead of the more regular i in most of the oblique cases of idam, this ; and the correlative of this word, adas, means not only that, but also this. Nevertheless, a is more generally a remote than a proximate demonstrative, and i more generally a proximate than a remote. In derived adverbial words i has always a proximate force ; but ta, the consonantal demonstrative, is more generally used than a. The following are examples of each vowel : — i-ha, here ; i-ddnim, now ; ta-ddnim, then : also i-ti, so, this much ; a-tha, so, thus, in that manner, i, the proximate demonstrative root, is in all probability identical with i, the sign of the locative in such words as hrid-i, heart. Probably, also, we see the same root in the preposition in. We may compare the Old Persian avadd, thither, in that direction ; and the corre- sponding proximate i-dd, hither, in this direction. The resemblance between the bases of these forms, notwithstanding the irregularity of their application, and the Dravidian remote and proximate demonstrative bases, s«ems to amount to identity. All irregularity disappears in the New Persian, which in this point accords as per- fectly with the Dravidian languages as if it were itself a Dravidian idiom. Its demonstratives are an, that, in, this. These demon- 43^ THE PRONOUN strati ves are adjectival prefixes, and naturally destitute of number ; but when plural terminations are suffixed, they acquire a plural signification — e.g., dndn, those (persons), tndn, these (persons). The same demonstratives are largely used in modern Turkish, by which they have been borrowed from Persian, an and in are undoubtedly Aryan demonstratives. This is apparent when we compare an with the Zend aem, that, and that again with the Sanskrit ayam ; but in is still more clearly identical with the Zend infi, this. This same im constitutes the accusative in Vedic Sanskrit (and is also identical with iyam, the masculine-feminine singular of the Old Persian, and the feminine of Sanskrit) ; but in Zend ^m is the nominative, not the accusative, and it is to this form that the New Persian is most closely allied. The demonstrative base i (without being restricted, however, to a proximate signification) appears in the Latin is and id, and in the Gothic is ; and the Dra vidian and New Persian distinction between the signification of a and that of i, has been re-developed in our English that and this. Whilst the New Persian an and in are closely connected with Sanskrit and Zend demonstratives, it does not follow that they are directly derived from either the one tongue or the other. On the contrary, the exactness with which the Persian discriminates between the remote and the proximate, leads me to conclude that it has retained more faithfully than either of those languages the primitive characteristics of the prse-Sanskritic speech. If so, instead of supposing the Dra vidian dialects to have borrowed their demonstratives, which are still purer than the Persian, from Sanskrit (which are irregular and greatly corrupted), it is more reasonable to suppose that the Dravidian demonstrative vowels retain and exhibit the primeval bases from Mdiich the demonstratives of the Sanskrit and of all other European tongues have been derived. Affiliation of Interrogative Bases : Extra- Dravidian Relationsldp. — There seems to be no analogy between either e or yd and any of the interrogative bases of the Indo-European family. Both in that family and in the Scythian group, the ordinary base of the inter- rogative is the guttural k — e.g., Sanskrit, kim, what ? The same base appears in the Sanskrit interrogative initial syllables ka-, ki-, ku-, which correspond to the Latin qu-, the Gothic hva-, and the English wh-. We find the same base again in the Turkish kim or kim, who ? what ? in the Magyar ki, who ? plural kik ; and in the Finnish kuka (root ku). I am unable to suppose the Dravidian yd derived from the Sanskrit and Indo-European ka. I see nowhere else any trace of a Sanskrit k changing into a Dravidian y. It would be DEMONSTRATIVES AND INTERROGATIVES 439 tempting, but unsafe, to connect ka-t (Sans.) with yd-du (Tarn.) which ? In the absence of a real relative pronoun, the interrogative is used as a relative in many of the Scythian languages. The base of the Sanskrit relative pronoun ya (yas, yd, yat) bears a close apparent resemblance to the Dravidian interrogative yd. The Sanskrit ya, however, like the derived North Indian jo, and the Finnish yo, is exclusively used as a relative, whereas the Dravidian yd is ex- clusively and distinctively an interrogative. It has been conjectured that the Sanskrit ya, though now a relative, was a demonstrative originally ; and if (as we shall see that there is some reason for supposing) the Dravidian interrogatives e and a were originally demonstratives, it may be supposed that yd was also a demonstrative, though of this no direct evidence whatever now remains. If yd were originally a demonstrative, the connection which would then appear to exist between it and the Sanskrit relative would require to be removed a step further back ; for it is not in Sanskrit that the relative ya has the force of a demonstrative, but in other and more distant tongues — viz., in the Lithuanian yis, he ; and in the Slavonian yam, and the Zend yim, him. Emphatic e. — It has been seen that in Ku e is used as a demon- strative — e.g., evdru (e-{v)-dr), they ; and this may be compared with the demonstrative e of the Sanskrit etat, this (neuter), and the corresponding Zend aetat. In the other Dravidian dialects, however, e is not used as a demonstrative, but is postfixed to words for the purpose of rendering them emphatic. The manner in which e is annexed, and the different shades of emphasis which it communi- cates, are precisely the same in the various dialects, and will be sufficiently illustrated by the following examples from Tamil. When e is postfixed to the subject of a proposition, it sets it forth as the sole depositary of the quality predicated — e.g., kalvi-{y)-e selvam, learning (alone is) wealth ; when postfixed to the predicate, it intensifies its signification — e.g., kalvi selvam-e, learning is wealth (indeed). When postfixed to a verb or verbal derivative, it is equivalent to the addition of the adverb truly, certainly — e.g., alla-(v)-e (certainly) not. In the colloquial dialect, it has often been annexed to the case terminations of nouns without necessity, so that it has sometimes becoHte in that connection a mere euphonic ex- pletive ; in consequence of which, in such instances, when emphasis is really required by a sign of case, the e has to be doubled — e.g., enndleye {enndl-e-{y)-e), through me (alone). In Tulu, emphatic e 440 THE PRONOUN becomes euphonically, not only y{e) and v{e), as in Tamil, after certain vowels, but also w(e). e, however, is always to be regarded as the sign of emphasis. The same sign of emphasis forms the most common vocative case-sign in the various Dravidian dialects, the vocative being nothing more than an emphatic enunciation of the nominative. Compare with this the use of the nominative, with the addition of the definite article, as the vocative in Hebrew and in Attic Greek. The Persian e of supplication may also be compared with it. Some resemblance to the use of e as a particle of emphasis may be discovered in the Hebrew ' he paragogic,' which is supposed to intensify the signification of the words to which it is annexed. The * he directive ' of the same language is also, and not without reason, supposed to be a mark of emphasis. A still closer resemblance to the emphatic e of the Dravidian languages is apparent in Chaldee, in which a suffixed to nouns constitutes their emphatic state, and is equivalent to the definite article of many other languages. The Persian e of particularity, the e of ascription of greatness, &c., in addition to the e of supplication, which has already been referred to, probably spring from a Chaldaic and Cuthite origin, though each of them bears a remarkable resemblance to the Dravidian emphatic e. Honorific Demonstrative Pronouns. — I have deferred till now the consideration of a peculiar class of honorific demonstratives, which are found only in Telugu and Canarese, and in which, I think, Aryan influences or affinities may be detected. In all the Dravidian dialects the plural is used as an honorific singular when the highest degree of respect is meant to be expressed ; but when a somewhat inferior degree of respect is intended, the pronouns which are used by the Telugu are dyana, he, ille, and dme, she, ilia ; with their correspond- ing proximates iyana, hie, and inie, hsec. These pronouns are destitute of plurals. When a little less respect is meant to be shown than is implied in the use of dyana and iyana, and of dme and ime, Telugu makes use of atadu, ille, dse, ilia, with their corresponding proximates itadu and ise ; atanu and itanu are also used, also the longer forms dtanu dtadu, &c. Here Canarese agrees with Telugu — e.g., dtanu, ille, Uanu, hie (class. Can. dtam, itayn). The Canarese feminines dke, ilia, ike, hsec, do not appear so perfectly to accord with the Telugu dse, ise. Both the above sets of Telugu pronouns are destitute of plurals, but both are pluralised in Canarese — c.fj., dtagalu, itagalu, those and these men ; dheyar, ikeyar, those and these (women). The Tuda atham, he, she, it, appears to be allied to the pronouns now referred to. I consider it to be a neuter singular, DEMONSTRATIVES AND INTERROGATIVES 44 1 synonymous with adu, the neuter singular of the Tamil-Canarese, and used corruptly for the masculine and feminine, as well as for the neuter. An Aryan origin may possibly be attributed to some of these words, especially to dyana, tyana, dme, ime ; and this supposition would account for the circumstance that they are found in Telugu only, and not in any other dialect of the family (except the Tulu dye, he, is to be regarded as a connected form) : it would also harmonise with their use as honorifics. Compare dyana with the Sanskrit masculine ay am, ille, and lyaria with the Sanskrit feminine, and the Old Persian mas. fem. yam, hie, haec. dm^, ilia, and ime, hsec, the corresponding feminine pronouns of the Telugu, may be compared not only with the plurals of the Sanskrit pronoun of the third person {ime, mas., imdh, fem., imdni, neut.), but also with amum and imam, him, which are accusative singulars, and from which it is evident that the yn of the plural forms is not a sign of plurality, but is either a part of the pronominal base, or an euphonic or formative addition. Bopp considers it to be the former, but Dravidian analogies incline me to adopt the latter view, and the m of these forms I conceive to be the ordinary neuter formative of Dravidian, and especially of Tamil, nouns, whilst the v seems to be merely a softening of m. me is a common suffix of Telugu neuter nouns. When the Telugu masculine of respect dtadu, dtanu, and the corre- sponding Canarese honorific dta-nu, are scrutinised, it is evident that in addition to the vocalic demonstrative bases, a and i, which are found in Dravidian demonstratives of every kind, the ta which is subjoined to a and i, possesses also somewhat of a demonstrative or pronominal signification. It cannot be regarded like v, as merely euphonic ; and its restriction to masculines shows that it is not merely an abstract formative, as the k of the feminine dke may be presumed to be. It can scarcely be doubted, I think, that the affinities of this ta are Aryan ; for we find in all the Aryan languages much use made of a similar ta, both as an independent demon- strative, and as an auxiliary to the vocalic demonstrative, ta-d. Sans, that, is an instance of the former ; whilst the secondary or auxiliary place which ta or da occupies in the Sanskrit etad (e-ta-d), this, and adam, adas {a-da-m, a-da-s), this, or that, is in perfect agreement with the Telugu and Canarese d-ta-nu, d-ta-du. The final e of dse, ise, dme, tm£, dice, ike, is equivalent to the Tamil ei. e or ei is an ordinary termination of abstracts in these languages, and a suitable one, according to Dravidian notions, for feminine honorific pronouns. 442 THE PRONOUN Syntactic Interrogatives, a and 6. — The interrogative prefixes e and yd are equivalent to the interrogative pronouns and adjectives, who ? which ? what ? &c. Another interrogative is required for the pur- pose of putting such inquiries as are expressed in EngHsh by a change of construction — e.g., is there ? is it ? by transposition from there is, it is. This species of interrogation is effected in all the Dravidian languages in one and the same manner, viz., by suffixing an open vowel to the noun, verb, or sentence which forms the principal subject of interrogation ; and in almost all these languages it is by the suffix of a or d alone, without any syntactic change, or change in the collocation of words, that an interrogative verb or sentence differs from an affirmative one — e.g., compare the affirmative avan tanddn, Tam. he gave, with avan tanddn-d ? did he give ? and avan d tanddn ? was it he that gave ? compare also adu ur, that is a village, with adu ur-d ? is that a village ? This interrogative is never prefixed to nouns or pronominals, or used adjectivally ; but is invariably postfixed, like an enunciated or audible note of interrogation. 6 is used instead of d in Malay alam, in which the interrogative use of d is almost unknown, d seems to survive only in idd (Tam. ido) lo, literally what is this ? o is used occasionally in Tamil also as a simple interrogative ; but its special and distinctive use is as a particle expressive of doubt. Thus, whilst avan-d means is it he ? avan-6 means can it be he ? or, I am doubtful whether it is he or not. 6 is postfixed to words in precisely the same manner as a, and is probably only a weakened form of it, in which, by usage, the in- terrogation has been softened into the expression of doubt. It has acquired, however, as a suffix of doubt a position and force of its own, quite independent of d ; in consequence of which it is often annexed even to interrogative pronouns — e.g., evan-6, Tam. I wonder who he can be ; enna{v)-6, what it may be I know not — compound forms which are not double interrogatives, but which consist of a question evan, who ? or enna, what ? and an answer 6, I am doubtful, I know not, there is room for further inquiry. In Tulu, in addition to the use of d and 6, as in the other dialects, e (euphonically {v)e or {n)e) is used syntactically as an interrogative. This e is doubtless identical with the e of emphasis in origin. The use of a or 6 as an interrogative suffix does not seem to have any counterpart in any language either of the Scythian or of the Indo-European family. It is alto- gether unknown to Sanskrit ; and Cashmirian is the only non- Dra vidian tongue in which it is found. I am inclined to consider a, the ordinary Dravidian interrogative. DEMONSTRATIVES AND INTERROGATIVES 443 as derived from, or at least allied to, a or a, the remote demonstrative of the same family. The quantity of that demonstrative a is long or short, as euphonic considerations may determine ; and though the interrogative a is always long in Tamil, yet in consequence of its being used as a postfix, it is pronounced long by necessity of position, whatever it may have been originally. In Telugu it is generally short ; always so in poetry. Hence the question of quantity may, in this inquiry, be left altogether out of account. The only real difference between them is the difference in location ; a demonstrative being invariably placed at the beginning of a word, a interrogative at the end of it. If the interrogative a were really connected with a the demonstrative we should expect to find a similar connection sub- sisting between e or e, the adjectival interrogative, and some demonstrative particle, with a similar interchange of places ; accord- ingly this is found to be the case, for e is not only the ordinary sign of emphasis in all the Dravidian tongues, but it is used in Ku as an adjectival demonstrative ; and it is curious that in this instance also there is a change of location, e emphatic being placed at the end of a word, e interrogative at the beginning. 6 would naturally be derived from a, as in the change of yam, we, Tarn., into 6m, in the pronominal terminations of the Tamil verb. A similar change in the position of particles, to denote or corre- spond with some change in signification, is not unknown in other tongues. Thus in Danish, the article en has a definite sense in one position, and an indefinite in another — e.g., en konge, a king, kongen, the king. But it is still more remarkable, and more corroborative of the supposition now advanced, that in Hebrew, one and the same particle, he (for it must be regarded as one and the same, and any difference that exists seems to be merely euphonic), imparts emphasis to a word when postfixed to it, and constitutes an inter- rogative when prefixed. Even in English the interrogative is founded upon the demonstrative. ' That ? ' differs from ' that ' only in the tone of voice with which it is pronounced. Distributive Pronouns. — In all the Dravidian tongues distributive pronouns are formed by simply annexing the conjunctive particle to any of the interrogative pronouns. Thus, from evan, who ? by the addition of um, and, the conjunctive or copulative particle of the Tamil is formed, viz., evamim, every one, whosoever (literally who ?- and) ; and from epporudu, when ? is formed in the same manner epporudum, always (literally when ?-and). In Canarese similar forms are found, though not so largely used as in Tamil — e.g., 444 THE PRONOUN ydvdgalu {yd-dgal-u), always ; and in Telugu u (the copulative par- ticle which answers to the Tamil um and the Canarese u) is used in the same manner in the formation of distributives — e.g., evvadunu (evvadu-(nn)-u), every one, eppudunnu {ej)pudu-(nn)-u), always. SECTION III.— RELATIVE PRONOUNS. I give this heading a place in the book solely for the purpose of drawing attention to the remarkable fact that the Dravidian languages have no relative pronoun, a participial form of the verb being used instead. Instead of relative pronouns, they use verbal forms which are called by English grammarians relative participles ; which see in the part on " The Verb." All other words which correspond either in meaning or in use to the pronouns of other languages will be found on examination to be nouns, regularly formed and dechned. PART VJ. THE VERB. The object in view in this part of the work is to investigate the nature, inflexions, and relations of the Dra vidian verb. It seems desirable to commence with some general preliminary remarks upon its structure. 1. A large proportion of Dra vidian roots are used indiscriminately, either as verbs or as nouns. When case-signs are attached to a root, or when, without the addition of case-signs, it is used as the nomina- tive of a verb, it is regarded as a noun ; the same root becomes a verb without any internal change or formative addition, when the signs of tense (or time) and the pronouns or their terminal fragments are suffixed to it. Though, abstractly speaking, every Dravidian root is capable of this twofold use, it depends upon circumstances whether any particular root is actually thus used ; and it often happens, as in other languages, that of three given roots one shall be used solely or generally as a verbal theme, another solely or generally as the theme of a noun, and the third alone shall be used indiscriminately either as a noun or as a verb. Herein also the usus loquendi of the various dialects is found to differ ; and not unfrequently a root which is used solely as a verbal theme in one dialect, is used solely as a noun in another. 2. The inflexional theme of a Dravidian verb or noun is not always identical with the crude root or ultimate base. In many instances formative or euphonic particles (such as vu, ku, gu or ngu, du or ndUy hu or mhu) are annexed to the root, — not added on like isolated post- positions, but so annexed as to be incorporated with it. (See Part II., " Roots.") But the addition of one of those formative suffixes does not necessarily constitute the root to which it is sufiixed a verb : it is still capable of being used as a noun, though it may be admitted that some of the roots i^ which those suffixes have been annexed are more frequently used as verbs than as nouns. 3. The structure of the Dravidian verb is strictly agglutinative. The particles which express the ideas of mood and tense, transition, 445 446 THE VERB intransition, causation, and negation, together with the pronominal fragments by which person, number, and gender are denoted, are annexed or agglutinated to the root in so regular a series and by so quiet a process, that generally no change whatever, or at most only a slight euphonic change, is effected either in the root or in any of the suffixed particles. (See this illustrated in " Roots.") 4. The second person singular of the imperative may perhaps be considered as an exception to the foregoing rule. The crude theme of the verb, or the shortest form which the root assumes, and which is capable of being used also as the theme of a noun, is used in the Dravidian languages, as in many others, as the second person singular of the imperative ; and the ideas of number and person and of the conveyance of a command, which are included in that part of speech, are not expressed by the addition of any particles, but are generally left to be inferred from the context alone. Thus, in the Tamil, sentences adi virundadu, the stroke fell ; ennei adi-ttdn, he struck me ; and idei adi, strike thou this ; the theme, adi, strike, or a stroke, is the same in each instance, and in the third illustration it is used without any addition, and in its crude state, as the second person singular of the imperative. 5. As the normal Dravidian noun has properly but one declension, so the normal Dravidian verb has properly only one conjugation and but very few irregular forms. It is true that grammarians have arranged the Dravidian verbs in classes, and have sometimes styled those classes conjugations ; but the differences on which this classi- fication is founded are generally of a trivial and superficial character. The structure of the verb, its signs of tense, and the mode in which the pronouns are suffixed, remain invariably the same, with such changes only as euphony appears to have dictated. Consequently, though class differences exist, they are hardly of sufficient importance to constitute different conjugations. When I speak of the normal Dravidian nouns and verbs I mean those of the more highly culti- vated dialects, Tamil, Malayalam, Canarese, and Telugu. The Tulu and Gond verbs will be found exceptionally rich in moods and tenses. Such is the simplicity of the structure of the normal Dravidian verb, that the only moods it has are the indicative, the infinitive, the imperative, and the negative, and that it has only three tenses, the past, the present, and the aorist or indefinite future. There is reason to suspect, also, that originally it had no present tense, but only a future and a past. The ideas which are expressed in other families of languages by the subjunctive and optative moods, are expressed in STRUCTURE 44/ all the members of the Dra vidian family, except in Tuju and Gond, by means of suffixed particles ; and the imperfect, perfect, pluper- fect, future perfect, and other compound tenses, are expressed by means of auxiliary verbs. In these respects the normal Dra vidian verb imitates, though it does not equal, the simplicity of the ancient Scythian verb. The modern Turkish has, it is true, an extraordinary number of moods — conditionals, potentials, reciprocals, inceptives, negatives, impossibles, &c., together with their passives, and also a large array of compound tenses ; but this complexity of structure appears to be a refinement of a comparatively modern age, and is not in accordance with the genius of the Oriental Turkish, or Tartar properly so called. Eemusat conjectures that intercourse with nations of the Indo-European race, some time after the Christian era, was the occasion of introducing into the Turkish language the use of auxiliary verbs and of compound tenses. " From the ex- tremity of Asia," he says, " the art of conjugating verbs is unknown. The Oriental Turks first offer some traces of this ; but the very sparing use which they make of it seems to attest the pre-existence of a more simple method." All the Dra vidian idioms conjugate their verbs, with the partial exception of Malayalam, which has retained the use of the signs of tense, but has rejected the pronominal terminations, except in the ancient poetry. Nevertheless, the system of conjugation on which most of the Dra vidian idioms proceed is one of primitive and remark- able simplicity. Tulu and Gond verbs possess more complicated systems of con- jugational forms, almost rivalling those of the Turkish in abundance. Tulu has a perfect tense, as well as an imperfect or indefinite past. It has conditional and potential moods, as well as a subjunctive. Tamil has but one verbal participle, which is properly a participle of the past tense, whilst Tulu has also a present and a future participle. All these moods, tenses, and participles have regularly formed nega- tives. I do not refer here to the pluperfect and second future, or future perfect tense, of Tulu, these tenses being formed, as in the other dialects, by means of the substantive verb used as an auxiliary. Gond has all the moods, tenses, and participles of Tulu, and in addition some of its own. It has an inceptive mood. Its imperfect branches into two distinct»tenses, an imperfect, properly so called (I was going), and a past indefinite (I went). It has also a desiderative form of the indicative — that is, a tense which, when preceded by the future, is a subjunctive, but which when standing alone implies a wish. 448 THE VERB On comparing the complicated conjugational system of the Gond with the extreme and almost naked simplicity of the Tamil, I conclude that we have here a proof, not of the superiority of the Gond mind to the Tamihan, but simply of the greater antiquity of Tamilian literary culture. The development of the conjugational system of Tamil seems to have been arrested at a very early period (as in the parallel, but still more remarkable, instance of the Chinese) by the invention of writing, by which the verbal forms existing at the time were fossilised, whilst the uncultured Gonds, and their still ruder neighbours the Kols, w^ent on age after age, as before, com- pounding with their verbs auxiliary words of time and relation, and fusing them into conjugational forms by rapid and careless pro- nunciation, without allowing any record of the various steps of the process to survive. The Dravidian languages do not make a distinction, as the Hun- garian does, between subjective and objective verbs. In Hungarian, ' I know ' is considered a subjective verb ; I know (it, them, some- thing), an objective verb. A like distinction is made by the Bornu or Kanuri, an African language, bufc not by any of the Dravidian dialects. 6. The Dravidian verb is as frequently compounded with a noun as the Indo-European one ; but the compound of a verb with a prepo- sition is unknown. An inexhaustible variety of shades of meaning is secured in Sanskrit and Greek by the facility with which, in those languages, verbs are compounded with prepositions ; and the beauty of many of those compounds is as remarkable as the facility with which they are made. In the Scythian tongues, properly so called, there is no trace of compounds of this kind ; and though at first sight we seem to discover traces of them in the Dravidian family, yet when the component elements of such compounds are carefully scrutinised, it is found that the principle on which they are com- pounded differs widely from that of Indo-European compounds. The Dravidian preposition-like words which are most frequently compounded with verbs are those which signify over and under, the use of which is illustrated by the common Tamil verbs met-Jcol, to overcome, and hir-{f)'padi, to obey. Dravidian prepositions, how- ever (or rather, postpositions), are properly nouns — e.g., mel (from mi-{y)-al), over, literally means overness, superiority ; and mel-kol (euphonically mei-kol), to overcome, literally signifies to take the superiority. These and similar verbal themes, therefore, though compounds, are not, after all, compounds of a preposition and a verb. TRANSITIVES AND INTRANSITIVES 449 but are compounds of a noun and a verb ; and the Greek verbs with which they are to be compared are not those which commence with Trept iferi), Kara (kata), dvd (ana), Sec, but such compounds as irokLopKeo) {foliorheo), to besiege a city, hterally to city-besiege ; vavTT'qykii) (naupegeo), to build a ship, literally to ship-build. In such cases, whether in Greek or in Tamil, the first member of the compound (the noun) does not modify the signification of the second (the verb), but simply denotes the object to which the action of the verb applies. It is merely a crude noun, which is used objectively without any signs of case, and is intimately combined with a governing verb. Dravidian verbs acquire new shades of meaning, and an increase or diminution in the intensity of their signification, not by prefixing or combining prepositions, but by means of auxiliary gerunds, or verbal participles and infinitives — parts of speech which in this family of languages have an adverbial force — e.g., mundi {p)'p6ndn, Tam. he went before, literally having-got-before he went ; sutti [suttri) (f)f6ndn, he went round, literally rounding he went ; tdra {k)kudittdn, he leaped down, literally so-as-to-get-down he leaped. A great variety of compounds of this nature exists in each of the Dravidian dialects. They are as easily made, and many of them are as beautiful, as the Greek and Sanskrit compounds of prepositions with verbs. See especially Dr Gundert's " Malay ajam Grammar." SECTION I.— CLASSIFICATION. 1. Transitives and Intransitives. Dravidian grammarians divide all verbs into two classes, which are called in Tamil fiva vinei and tan vinei, transitives and intransitives, literally outward-action words and self-action words. These classes correspond rather to the parasmai-padam and dtmane-fadam, or transitive and reflexive voices, of the Sanskrit, than to the active and passive voices of the other Indo-European languages. The Dravidian fita vinei and tan vinei, or transitive and intransi- tive verbs, differ from the parasmai-padam and dtmane-padam of the Sanskrit in this, that instead of each being conjugated differently, they are both conjugated in precisely the same mode. They differ, not in their mode of conjugation, but in the formative additions made to their themes. Moreover, all pita vinei, or transitive verbs, are really, as well as formally, transitives, inasmuch as they neces- 2g 450 THE VERB sarily govern the accusative, through the transition of their action to some object ; whilst the tan vinei, or intransitive verbs, are all necessarily, as well as formally, intransitives. The Dravidian transitives and intransitives closely resemble in force and use, though not in shape, the objective and subjective verbs of the Hungarian. The Hungarian objective verbs, like the Dravidian transitives, imply an object — an accusative expressed or implied — e.g., szeretem, I love (some person or thing) ; whilst the Hungarian subjective verbs, like the Dravidian intransitives, neither express nor imply an object — e.g., szereteh, I love — i.e., I am in love. In a large number of instances in each of the Dravidian dialects, including entire classes of verbs, there is no difference between tran- sitives and intransitives, either in formative additions to the theme, or in any structural peculiarity, the only difference is that which consists in the signification. Thus, in Tamil, all verbs of the class which take i as the sign of the past participle are conjugated alike, whether they are transitives or intransitives — e.g., from pann-u, trans, to make, are formed the three tenses (first person singular) pannu-gii-en, I make, fann-i-(n)-en, I made, and fannu-v-en, I will make ; and in like manner from fes-u, intrans. to talk, are formed, precisely in the same manner, the corresponding tenses 'pesu-gir-en, I talk, 'pes-i-(n)-en, I talked, and pesu-v-en, I will talk. In a still larger number of cases, however, transitive verbs differ from intransitives, not only in signification and force, but also in form, notwithstanding that they are conjugated alike. The nature of the difference that exists and its rationale are more clearly apparent in Tamil than in any other Dravidian dialect ; my illustrations will, therefore, chiefly be drawn from the Tamil. There are three modes in which intransitive Tamil verbs are con- verted into transitives. 1. Intransitive themes become transitive by the hardening and doubling of the consonant of the appended formative — e.g., peru-gu, to abound, by this process becomes 'peru-kku, to increase (actively), to cause to abound. Transitives of this kind, which are formed from intransitives in actual use, are often called causals, and they are as well entitled to be called by that name as many causal verbs in the Indo-European tongues ; but as there is a class of Dravidian verbs which are distinctly causal (and which are formed by the annexing to the transitive theme of a causal particle — e.g., pannuvi {v-i), Tam. to cause to make, from pannu, to make), it will contribute to perspicuity to regard the whole of the verbs of which we are now TRANSITIVES AND INTRANSITIVES 45 I treating simply as transitives, and to reserve the name of causal verbs for the double transitives referred to. When transitives are formed from intransitives by doubling the consonant of the forma- tive, there is no change in any of the signs of tense, or in the mode in which those signs are added ; and the hardened formative appears in the imperative, as well as in the other parts of the verb. The nature of these formatives has already been investigated in Part II., on " Roots " ; and it has been shown that they are generally either euphonic accretions, or particles of specialisation, which, though permanently annexed to the base, are not to be confounded with it. I subjoin a few illustrations of this mode of forming transitives by the doubling and hardening of the consonant of the formative. (1.) gu, or its nasalised equivalent ngu, becomes kku — e.g., from fo-gu, to go (in the imperative softened into fo), comes 'po-kku, to drive away ; from ada-ngu, to be restrained, comes ada-kku, to restrain. (2.) su becomes chchu — e.g., from adei-su, to be stuffed in, comes adei-chchu, to stuff in, to stick on. (3.) du, euphonised into ndu, becomes ttu — e.g., from tiru-ndu, to become correct, comes tiru-ttu, to correct. (4.) hu, euphonised into mhu, becomes ffu — e.g., from nira-mbu, to be full, comes iiira-pfu, to fill. When intransitives are converted into transitives in this manner in Telugu, gu or ngu becomes, not kku as in Tamil, but chu — a difference which is in accordance with dialectic rules of sound. Thus from tu-gu, or euphonically tu-ngu, to hang, to sleep, comes tu-chu,ov euphonically tu-nchu, to weigh, to cause to hang (Tam. tu-kku). Telugu also occasionally changes the intransitive formative gu, not into chu, the equivalent of kku, but into fu — e.g., from mey, to graze, comes me-fu, to feed ; and as ffu in Tamil is invariably hardened from hu or mhu, the corresponding Telugu fu indicates that hu originally alternated with gu ; for the direct hardening of gu into fu is not in accordance with Dravidian laws of sound. This view is confirmed by the circumstances that in Telugu the use of fu instead of chu (and of m'pu instead of nchu) is in most instances optional, and that in the higher dialect of Tamil the formative fjt sometimes supersedes kk — e.g., the infinitive of the verb ' to walk ' may in that dialect be either nada-k^a or na(la-ppa. It is obvious, therefore, that these formative terminations are mutual equivalents. If the transitive or causal p of such verbs as nira-ppu, Tam. to fill, me-pu, Tel. to feed, were not known to be derived from the hardening 452 THE VERB of an intransitive formative, we might be inclined to affiliate it with the f, which is characteristic of a certain class of causal verbs in Sanskrit — e.g., pvd-p-aydmi, I cause to live, jnd-p-aydmi, I make to know. It is evident, however, that the resemblance is merely accidental, for etymologically there is nothing of a causal nature in the Dravidian formatives ; it is not the formative itself, but the hardening of the formative which conveys the force of transition ; and on the other hand, the real sign of the causal in Sanskrit is aya, and the p which precedes it is considered to be only an euphonic fulcrum. It has already been shown (in " Eoots ") that the various verbal formatives now referred to are used also as formatives of nouns, and that when such nouns are used adjectivally, the consonant of the formative is doubled and hardened, precisely as in the transitives of verbs — e.g., maruttu, medicinal, from marundu, medicine ; pdppu, serpentine, from pdmhu, a snake. When nouns are used to qualify other nouns, as well as in the case of transitive verbs, there is a transition in the application of the meaning of the theme to some other object ; and the idea of transition is expressed by the doubling and hardening of the consonant of the formative, or rather by the forcible and emphatic enunciation of the verb of which that harden- ing of the formative is the sign. There is something resembling this in Hebrew. The doubling of a consonant by Dagesh forte is some- times resorted to in Hebrew for the purpose of converting an intransi- tive verb into a transitive — e.g., compare Idmad, he learned, with limmed, he caused to learn, he taught. 2. The second class of intransitive verbs become transitives by tlie doubling and hardening of the initial consonant of the signs of tense. Verbs of this class are generally destitute of formatives, properly so called ; or, if they have any, they are such as are incapable of change. The sign of the present tense in colloquial Tamil is git ; that of the preterite d, ordinarily euphonised into nd ; and that of the future, h or v. These are the signs of tense which are used by intransitive verbs of this class ; and it will be shown hereafter that they are the normal tense-signs of the Dravidian verb. When verbs of this class become transitives, gii is changed into kkii, d or nd into it, and hoiv into pp. Thus, the root Ur, to join, is capable both of an intransitive sense — e.g., to join (a society) — and of a transitive sense — e.g., to join (things that were separate). The tense-signs of the intransitive remain in their natural condition — e.g., ^er-gir-en, I join, Mr-nd-en, I joined, ser-v-en, I will join ; but when the significa- TRANSITIVES AND INTRANSITIVES 453 tion is active or transitive — e.g., to join (planks), the corresponding parts of the verb are ser-kkii-en, I join, ser-tt-en, I joined, ser-pp-en, I will join. The rationale of this doubling of the first consonant of the sign of tense appears to be exactly the same as that of the doubling of the first consonant of the formative. It is an emphasised, hard- ened enunciation of the intransitive or natural form of the verb ; and the forcible enunciation thus produced is symbolical of the force of transition by which the meaning of the transitive theme overflows and passes on to the object indicated by the accusative. In verbs of this class the imperative remains always unchanged ; and it is the connection alone that determines it to a transitive rather than an intransitive signification. It should here be mentioned, that a few intransitive verbs double the initial consonant of the tense-sign, and that a few transitive verbs leave the tense-sign in its original, unemphasised condition. Thus, iru, to sit, to be, is necessarily an intransitive verb ; nevertheless, in the present tense iru-kkk-en, I am. and in the future iru-pp-en, I shall be, it has made use of the ordinary characteristics of the transitive. So also padu, to lie, though an intransitive, doubles the initial consonant of all the tenses — e.g., padu-kkiv-en, I lie, padu-tt-en, I lay, padu-pp-en, I shall lie. On the other hand, ^, to give, to bestow, though necessarily transitive, uses the simple, unhardened, unemphatic tense-signs which are ordinarily characteristic of the intransitive — e.g., i-giv-en, I give, i-nd-en, I gave, i-v-en, I will give. These instances are the result of dialectic rules of sound, and they are not in reality exceptions to the method described above of distinguishing transitive and intransitive verbs by means of the hardening or softening of the initial consonant of the tense-signs. Besides, this anomalous use of the transitive form of the signs of tense for the intransitive is peculiar to Tamil. It is not found in Telugu or Canarese. 3. A third mode of converting intransitives into transitives is by adding a particle of transition to the theme or root. This particle is du in Canarese, and ttu (in composition tu or du) in Tamil, and may be regarded as a real transitive suffix, or sign of activity. We have an instance of the use of this particle in the Canarese tdl-du, to lower, from tdl-u, to be low, and the corresponding Tamil tdr-ttu, to lower, from tar or tditu, to be low. When the intransitive Tamil theme ends in a vowel which is radical and cannot be elided, the transitive particle is invariably ttu — e.g., padu-ttu, to lay down, from padu, to lie. It might, therefore, be supposed that ttu is the 454 THE VERB primitive shape of this particle ; but on examining those instances in which it is compounded with the final consonant of the intransitive theme, it appears to resolve itself, as in Canarese, into du. It is always thus compounded when the final consonant of the theme is I or I, d OT T ; and in such cases the d of du is not merely placed in juxtaposition with the consonant to which it is attached, but is assimilated to it, or both consonants are euphonically changed, according to the phonetic rules of the language. Thus I and du become tt-u (pronounced ttr-u) — e.g., from sural, intrans., to be whirled, comes suraii-u (surattr-u), trans., to whirl. I and du become ttu — e.g., from mil, to return, comes mitt-u, to cause to return, to redeem. From these instances it is clear that du, not ttu, is to be regarded as the primitive form of this transitive suffix. What is the origin of this transitive particle, or sign of activity, ttu or du ? I believe it to be identical with the inflexion or adjectival formative attu or ttu, which was fully investigated in Part III., " The Noun," and of which the Canarese form is ad\ the Telugu ti or ti. There is a transition of meaning when a noun is used adjectivally {i.e., to qualify another noun), as Well as when a verb is used transitively {i.e., to govern an object expressed by some noun in the accusative) ; and in both cases the Dravidian languages use (with respect to this class of verbs) one and the same means of expressing transition, viz., a particle which appears to have been originally a neuter demonstra- tive. Nor is this the only case in which the Tamil transitive verb ex- hibits the characteristics of the noun used adjectivally, for it has been shown also that the doubling and hardening of the consonant of the formative of the first class of transitive verbs is in exact accordance with the manner in which nouns terminating in those formatives double and harden the initial consonant when they are used to qualify other nouns. Another illustration of this principle follows. 4. The fourth (a distinctively Tamil) mode of converting intransi- tive verbs into transitives consists in doubling and hardening the final consonant, if d or r. This rule applies generally, though not invari- ably, to verbs which terminate in those consonants ; and it applies to a final nd-u (euphonised from d-u), as well as to d-u itself. The operation of this rule will appear on comparing vdd-u, to wither, with vdtt-u, to cause to wither ; 6d-u, to run, with 6tt-u, to drive ; tind-u, to touch, with tUt-u, to whet ; mat-u, to become changed, with mdiT-u (pronounced mdttr-u), to change. The corresponding transi- tives in Telugu are formed in the more usual way by adding chu to the intransitive theme^e.r/., mdiu-chu, to cause to change, vddu-chu. CAUSALS 455 to cause to wither. Tamil nouns which end in d-u, nd-u, or r-w, double and harden the final consonant in precisely the same manner when they are placed in an adjectival relation to a succeeding noun — e.g., compare kdd-u, a jungle, with kdtt-u vari, a jungle-path ; irand-u, two, with irattu nul, double thread ; dt-u, a river, with dnu (pro- nounced dttru) manal, river sand. Thus we are furnished by words of this class with another and remarkable illustration of the analogy which subsists in the Dravidian languages between transitive verbs and nouns used adjectivally. 2. Causal Verbs. There is a class of verbs in the Dravidian languages which, though generally included under the head of transitives, claims to be regarded distinctively as causals. These verbs have been classed with transi- tives both by native grammarians and by Europeans. Beschi alone places them in a class by themselves, and calls them eval vinei, verbs of command — i.e., verbs which imply that a thing is commanded by one person to be done by another. Causals differ from transitives of the ordinary character, as well as from intransitives, both in signification and in form. The signification of intransitive verbs is confined to the person or thing which constitutes the nominative, and does not pass outward or onward to any extrinsic object — e.g., fo-gii-en, I go. The signification of transitive or active verbs, or, as they are called in Tamil, outward action-words, passes outward, to some object exterior to the nominative, and which is generally put in the accusative — e.g., unnei anuppu-gir-en, I send thee : and as to send is to cause to go, verbs of this class, when formed from intransi- tives, are in some languages, appropriately enough, termed causals. Hitherto the Indo-European languages proceed pari passu with the Dravidian, but at this point they fail and fall behind ; for if we take a verb which is transitive of necessity, like this one, to send, and endeavour to express the idea of causing to send, i.e., causing one person to send another, we cannot by any modification of structure get any Indo-European verb to express by itself the full force of this idea : we must be content to make use of a phrase instead of a single verb ; whereas in the Dravidian languages, as in Turkish and other languages of the Scyth^^n stock, there is a form of the verb which will express the entire idea, viz., the causal — e.g., anuppu-vi, Tarn, to cause to send, which is formed from anuppu, to send, by the addition of the particle vi to the theme. Transitives are in a similar manner 456 THE VERB converted in Turkish into causals by suffixing a particle to the theme — e.g., sev-dur, to cause to love, from sev, to love ; and dtch-u, to cause to work, from dtch, to work. There is a peculiarity in the signification and use of Dravidian causal verbs which should here be noticed. Indo-European causals govern two accusatives, that of the person and that of the object — e.g., I caused him (ace.) to build the house (ace.) ; whereas Dravidian causals generally govern the object alone, and either leave the person to be understood — e.g., vittei {k)kaUuvitten, Tam., I caused to build the house (or, as we should prefer to say, I caused the house to be built) ; or else the person is put in the instrumental — e.g., I caused to build the house, avanei {h)kondu, through him, or employing him ; that is, I caused the house to be built by him. Double accusatives are occasionally met with in classical compositions in Tamil, and are not uncommon in Malay alam. Dr Gundert quotes the Malay alam phrase avane Yama-loham fugichchu, he caused him to enter the world of Yama — to die ; but in all such instances, I think, Sanskrit influences are to be suspected. Though the Dravidian languages are in possession of a true causal — formed by the addition of a causal particle — yet they sometimes resort to the less convenient Indo-European method of annexing an auxiliary verb which signifies to make or to do, such as sey and 'pann-u in Tamil, mdd-u in Canarese, and cJiey-u in Telugu. These auxiHaries, however, are chiefly used in connection with Sanskrit derivatives. The auxiliary is annexed to the infinitive of the principal verb. Tamil idiom and the analogy of the other dialects require that causals should be formed, not from neuter or intransitive verbs, but from transitives alone ; but sometimes this rule is found to be neglected. Even in Tamil, vi, the sign of the causal, is in some instances found to be annexed to intransitive verbs. This usage is not only at variance with theory, but it is unclassical. In each of those cases a true transitive, derived from the intransitive in the ordinary manner, is in existence, and ought to be used instead. Thus, varu-vi, Tam., to cause to come, is less elegant than varu-ttu ; and nada-ppi, to cause to walk, to guide, than nada-ttu. The use of the causal, instead of the active, where both forms exist, is not so ruuch opposed to the genius of the other dialects as to that of Tamil. The use of one form rather than another is optional in Telugu and Canarese ; and in some instances the active has dis- appeared, and the causal alone is used. Thus ra-{p)-pinchu, or CAUSALS 457 rd-vinchu, to cause to come, the equivalent of the Tamil varu-vi, is preferred by Telugu to a form which would correspond to varu-ttu ; and instead of dkk-u, Tarn., to cause to become, to make, which is the active of dg-u, and is formed by the process of doubling and hardening which has already been described, Telugu uses the causal kdv-inchu, and the Canarese the corresponding causal dg-isu. One and the same causal particle seems to me to be used in all the Dravidian dialects, with the exception of Tulu and Gond. It assumes in Tamil the shapes of vi, hi, and fpi ; in Telugu, inchu and finchu ; in classical Canarese, ichu ; in the colloquial dialect, Uu. It seems difficult at first sight to suppose these forms identical ; but it will be found, I think, in every case that the real form of the causal particle is i alone, and that whatever precedes or follows it pertains to the forma tives of the verb. I begin with Telugu, w^hich, in regard to this point, will be found to throw light on the rest of the dialects. In Telugu, causal verbs end either in inchu or pinchu — e.g., chey-inchu, to cause to do, from chey-u, to do ; fili-f inchu, to cause to call, to invite, from filu-chu, to call, nchu, the final portion of inchu or pinchu, has first to be explained, nchu (pronounced ntsu) is a nasalised form of chu, which is a very common formative of Telugu verbs. When chu follows i — i.e., when the base to which it is attached ends in i, it is invariably euphonised or nasalised into nchu — e.g., jayi, a Sanskrit derivative, though not a causal, ends in i ; hence the Telugu verb formed from it is jayi-nchu, to conquer ; and hence also, as the causal verb in Telugu is formed by affixing the particle i to an ordinary verbal root, all such causal verbs end in inchu. ichu is to be regarded as the original form, and ichu is compounded of the causal particle and the affix chu. What is this chu ? We have already shown, in the section on " Formative Additions to Koots," that the Telugu chu is a verbal formative, identical in origin with the Tamil kku. The formative kku of Tamil is affixed to the verbal base of causals, as to various other classes of verbal bases, before adding the a which forms the sign of the infinitive. It is also affixed to the base before adding um, the sign of the indefinite future ; and the identity of this Tamil kku with the Telugu nchu will appear as soon as the Tamil infinitive is compared with the Telugu — e.g., comp. Seyvi-kka, Tam. infinitive, to cause to do, with the Telugu cheyi-ncha ; areippi-kka, Tam. in- finitive, to cause to call, with the Telugu pilipi-ncha. Comp. also an ordinary transitive verb in two languages — e.g., nmta-kka, Tam. 458 THE VERB infinitive to forget, with the Telugu mata-cha. It thus appears that the ch or nch of the Telugu is as certainly a formative as the kk of the Tamil. Even in the vulgar colloquial Tamil of the extreme southern portion of the Tamil country kk systematically becomes ch. Thus matakka, the word just mentioned, is mavaclia in the southern patois, precisely as in Telugu. The chief difference between Tamil and Telugu with respect to the use of this formative is, that it is used by two parts of the Tamil verb alone (the infinitive and the neuter future), whereas in Telugu it adheres so closely to the base that it makes its appearance in every part of the verb. What is the origin of the 'p which often appears in Telugu causal verbs before inchu ? The causal formed from viduchu, Tel. to quit, is not vidinchu, but vidipincJm, to release. This p shows itself, not in all causals, but only in those of verbs ending in the formative c/m, and it is a peculiarity of that class of verbs that ch changes optionally into p. Their infinitives may be formed by adding either pa or cha to the base. On the causal particle i being affixed to such verbs, ch changes by rule into p : thus, not pili-ch-inchu, to cause to call, but pili-p-inchu. This preference for p to ch before another ch looks as if it had arisen from considerations of euphony. But however this may be, p is frequently used in Telugu in the formation of verbal nouns, where such considerations could hardly exist — e.g.^ mavap-u, forgetfulness, from mava-chu, to forget (Tam. matappu) ; teta-pa, an opening, from teta-chu, to open (Tam. titappu). This formative is sometimes doubled in Telugu — e.g.^ tepp-inchu, to cause to bring, from techch-u, to bring. In Tamil p is always doubled, except after nasals or r. Though the use of this hardened form of p is rare in Telugu, yet its existence tends still further to identify the Telugu causal with the Tamil. Certain verbs in Telugu, ordinarily called causals (ending in c/m, nchu, pu, mpu, &c., without a preceding i), are to be regarded not as causals, but simply as transitives — e.g., viduchu, vidupu, to cause to quit ; vanchu, to bend ; lepu, to rouse. They are formed, not by annexing vi or i, but by the doubling and hardening of the final consonant of the formative {e.g., compare lepu, to rouse, with the corresponding Tamil eruppu, the transitive of erumhu) ; and the verbs from which they are so formed are not actives, but neuters. Instead, therefore, of saying that tir-u, to end, forms its causal either in tir-chti or tir-pinchu, it would be more in accordance with Tamil analogies to represent tir-u as the neuter, tir-chu as the transitive, and tir-pinchu as the causal. It is of the essence of what I regard as the CAUSALS 459 true causal that its theme is a transitive verb — e.g., katt-inchu, to cause to build, from katt-u, to build. In Canarese, causal verbs are formed by suffixing isu, or rather i-su, to the transitive theme — e.g., from mddu, to do, is formed mdd-i-su, to cause to do. This causal particle i-su (in the classical dialect i-chu) is annexed to the theme itself before the addition of the signs of tense, so that it is found in every part of the causal verb, like the corresponding Telugu particle i-nchu, with which it is evidently identical. It has been shown that the Telugu i-nchu has been nasalised from i-chu (the phonetic equivalent of the Tamil i-kku) ; and now we find this very i-chu in classical Canarese. The change in colloquial Canarese from i-chu to i-su is easy and natural, S being phonetically equivalent to ch, and chu being pronounced like tsu in Telugu. An additional proof, if proof were wanting, of the identity of the Canarese i-su with the Telugu i-nchu, is furnished by the class of derivative verbs, or verbs borrowed from Sanskrit. Sanskrit deriva- tive verbs are made to end in i in all the Dravidian dialects {e.g., jay-i, to conquer) ; and those verbs invariably take in Telugu, as has been said, the formative termination nchu — e.g., jayi-nchu. The same verbs invariably take i-su, or yi-su, in Canarese. Thus from the Sanskrit derivative theme, dhari, to assume, Telugu forms the verb dhari-nchu, the Canarese equivalent of which is dhari-su, Tamil infinitive tari-kka. These verbs are not causals ; but the use which they make of the formative nchu or su, preceded by i, illustrates the original identity of the Canarese causal particle i-su with the Telugu i-nchu, and of both with the Tamil i-kku. Generally the older and harsher sounds of Canarese have been softened by Tamil ; and in particular, the Canarese k has often been softened by Tamil into s or ch ; but in the instance of the formative annexed to the causative particle, exactly the reverse of this has happened ; the Tamil kk having been softened by the Canarese into L Canarese, like Telugu, does not so carefully discriminate between transitive and causal verbs as Tamil. The true causal of Tamil is restricted to transitive themes ; but Canarese, notwithstanding its possession of transitive particles {e.g., compare nera-hu, to fill, with neri, to be full, and tiru- pu, to turn (actively) with tiru-gu, to turn (of itself), often annexes the causal particle i-su ifo intransitive themes — e.g., 6d-i-su, to cause to run (Tam. 6tt-u), from 6d-u, to run. In Japanese, causative verbs are formed by affixing si to the root, si means to do. We now return to consider the causal particle of Tamil, instead of 460 THE VERB beginning with it. vi is generally supposed to be the causal particle of Tamil, hardening in certain connections into bi or ppi. In the first edition I adopted this view in substance, though regarding i alone as the causal particle in Telugu and Canarese, but preferred to consider bi, rather than vi, the primitive form, seeing that v does not readily change into b in Tamil (though v in Tamil often becomes b in Canarese — e.g., vd, Tam. to come=Can. bd), whilst b would readily soften into v on the one hand, or harden into pp on the other. On reconsideration, however, it seems to me better to regard i alone as the causal particle of Tamil, as of Telugu and Canarese, provided only the v, b, or pp, by which it is always preceded, be found capable of some satisfactory explanation. A clue to the right explanation seems to be furnished by the use of p instead of ch in Telugu. kJc in Tamil answers to ch in Telugu, and we find the Tamil kk changing optionally in classical Tamil into pp, precisely in accordance with Telugu usage. Instead of the infinitive nada-kka, to walk, nada-ppa may also be used. On com- paring the Tamil nadakka, to walk, with the Telugu naducha, and the Tamil nadappikka, bo cause to walk, with the Telugu nadipindm, we find them substantially identical. No difference exists but such as can be perfectly explained either by the change of kk into ch, nasalised into nch after i as already mentioned, or by the " har- monic sequence of vowels " explained in " Sounds." The p preceding i has clearly the same origin, and is used for the same purpose in both dialects. As it is certainly a formative in Telugu, it must be the same in Tamil ; and accordingly we find it actually used as a verbal formative in the classical Tamil infinitive nadap2)a, to walk, as mentioned above. It will be seen hereafter that a alone is the sign of the infinitive, and that whatever precedes it belongs to the verbal theme, or its formative. This circumstance might explain the pp of the Tamil causals ; but it is necessary to go a little further in order to be able to explain the v oib which alternates with pp. The most common formative of Tamil causals is vi — e.g., varu-vi, to cause to come ; the next is ppi — e.g., padi-ppi, to cause to learn. The remaining form is bi, used only after nasals — e.g., en-bi, to cause to say, to prove, from en, to say, kdn-bi, to show, from kdn, to see. There is no doubt that neither the b of bi nor the pp of ppi can have been inserted merely for euphony, v before i (as in vi) might be merely euphonic ; but this is rendered improbable by the circum- stance that vi is added, not only to verbs ending in vowels, but also to certain verbs ending in consonants (y and r) — e.g., sey-vi, to cause CAUSALS 461 to do, from sey, to do. Telugu and Canarese add i nakedly to the base {e.g., chey-inchu, from cMy-u, gey-isu, from gey-u). We have an instance of the use of vi after the soft, deep r in Tamil, as well as after y, in vdr-vi, to cause to flourish, from vdr, to flourish, vi is almost always used after u {e.g., hattu-vi, to cause to build), but in some instances 'pfi is used by rule after u — viz., where u is preceded by a short vowel and a single consonant — e.g., edu-fpi, to cause to take up, to erect, from edu, to take up. The Tamil future tense-signs seem to throw light on the formatives to which the causal particle i is afiixed. It is remarkable, at all events, that those three signs, v, h, fp, are identical with the for- matives of the causal verb, in what way soever this identity may be accounted for, so that if we know which of those three signs is used by any verb in the formation of its future tense, we know at once how the causal of the same verb is formed. Compare varu-v-en, I will come, with varu-v-i, to cause to come ; edu-pp-en, I will take up, with edu-pp-i, to cause to take up, to erect ; padi-pp-en, I will learn, with padi-pp-i, to cause to learn, to teach. This rule applies also to verbal roots ending in consonants — e.g., compare vdr-pp-en, I will pour, with vdr-pp-i, to cause to pour, to cast ; vdr-v-en, I will flourish, with vdr-v-i, to cause to flourish ; Mn-b-en, I will see, with Mn-h-i, to cause to se6, to show. Tamil admits of the use of a double causal — that is, of a verb denoting that one person is to cause another to cause a third person to do a thing. In this case also the new causal agrees with the future of the first causal, on which it seems to be built. Compare varu-vi-pp-en, I will cause to cause to come, with varu-vi-pp-i, to cause to cause to come. The explanation of this curious coincidence seems to be that the Tamil future was originally a sort of abstract verbal noun, which came to be used as a future by the addition of pronominal signs, whilst the same abstract neuter noun was converted into a causal (as we have seen was probably the case also with Telugu causals in p-i-nchu) by the addition to it of the causal particle. The addition of the causal particle in all cases in Canarese to the verbal root would seem to indicate an older and simpler period of Dravidian speech. Tulu forms its causal verbs in a somewhat different manner from the other Dravidian dialects — viz., by suffixing a instead of i to the verbal theme, or i^metimes du, and then adding the signs of tense — e.g., from malp-u, to make, is formed malp-d-vu, to cause to make, from nadapu, to walk, nadapudu, to cause to walk. This a of the Tulu resembles the Hindustani causal — e.g., chal-wd-nd, to 462 THE VERB cause to go, from chal-nd, to go ; and as the Hindustani causative particle wd has probably been derived from the Sanskrit aya or p-aya, the Tulu a might possibly be supposed to proceed from the same or a similar source. In Gond ha or h is the causal particle, and is added to the present participle of transitive verbs, not to the theme. Origin of the Dravidian Causal Particle ' *.' — The oldest form of the Indo-European causative particle is supposed to be the San- skrit aya (with p prefixed after a root in a), aya becomes i in old Slavonic, and the apparent identity between this i and the Dravidian i is noteworthy. Notwithstanding this, it does not seem to me either necessary or desirable to seek for the origin of Dravidian particles out of the range of the Dravidian languages, if those languages themselves provide us with a tolerably satisfactory explanation. The Dravidian causative particle i may be supposed to have been derived from t, to give. This i is short in various portions of the Telugu verb. The crude base is i-chch-u, the infinitive i-va or i-vva. The Canarese Uu also, the causal of t, seems to be formed, not from t, but from i {i-isu=tsu). In nearly all cases in the Dravidian lan- guages the short vowel seems to be older than the long one. The meaning of ' give ' seems tolerably suitable for a causal particle ; but we find it developing into a still more appropriate shape in Telugu, in which t is used after an infinitive to mean to let, permit, &c. — e.g., p6(n)4, let it go, from 'po, to go, literally give it to go. In Canarese also t-su, the causal of t, is used in the same sense of to let, permit, &c., as the original verb itself in Telugu — e.g., p6gal4sn, permit to go. It is remarkable also that in Canarese the correspond- ing and more common word kodu, give, is used in the same manner as a permissive or causal — e.g., mdda kodu, permit (him) to do. 3. Frequentative Verbs. There is a class of verbs in all the Dravidian languages that have sometimes been called iterative or frequentative. The following are Tamil examples : minuminu-kku, to glitter, from min, to shine ; velu- velu-kku, to whiten, from velu-kku, to be white, root vel, white ; muta-muta-kku, to murmur, munamuna-kku, to mutter, kirukiiu-kka, to be giddy. It does not seem to me, however, necessary to enter into the examination of these and similar words, seeing that there is no peculiarity whatever in the mode in which they are conjugated, the iterative meaning resides in the root alone, and is expressed THE PASSIVE VOICE 463 by the device, in common use in all languages, of doubling the root. Compare Latin murmuro, tintinno, &c. In Tulu, however, there is a form of the verb rightly called frequentative. It is formed by insert- ing e (probably the particle of emphasis) between the base and the personal signs, whereupon a new verbal base is formed, which is regularly conjugated — e.g., malpeve {mal])u-\-e-{-{v)e), I make again and again. 4. Intensive Verb. This form of verb is also found only in Tulu. Compare malfuve, I make, with maltruve, I make energetically ; kenuve, I hear, with kendruve, I hear intensely ; huruve, I fall, with hurduve, I fall heavily. 5. Inceptive Verb. We find a fully developed inceptive or inchoative form of the verb in Gond alone. It is formed by annexing the signs of person and tense, not to the base, as in the case of the ordinary verb, but to the infinitive. 6. The Passive Voice. Each of the primitive Indo-European languages has a regular passive voice, regularly conjugated. The Sanskrit passive is formed by annexing the particle ya (supposed to be derived from yd, to go) to the verbal theme, and adding the personal terminations peculiar to the middle voice. Most of the languages of the Scythian family also form their passives by means of annexed particles. In order to form the passive, the Turkish suffixes to the verbal theme il or il ; the Finnish et ; the Hungarian at, et, tet ; and to these particles the pronominal terminations are appended in the usual manner. Japanese has a passive voice, the form of which is active. The Dra- vidian verb is entirely destitute of a passive voice, properly so called, nor is there any reason to suppose that it ever had a passive. None of the Dravidian dialects possesses any passive particle or suffix, or any means of expressing passivity by direct inflexional changes ; the signification of the passive voice is, nevertheless, capable of being expressed in a variety of ways. We have now to inquire into the means adopted by the Dravidian languages for conveying a passive signification ; and it will be found that they correspond in a considerable degree to the means used for 464 THE VERB this purpose by the Gaurian vernaculars of Northern India, which also are destitute of a regular passive voice. In the particulars that follow, all the Dravidian dialects (with the exception of the Gond) agree : what is said of one holds true of all. (1.) The place of a passive voice is to a large extent supplied by the use of the neuter or intransitive form of the verb, somewhat as in Japanese. This is in every dialect of the family the most idiomatic and characteristic mode of expressing the passive ; and wherever it can be used, it is always preferred by classical writers. Thus, it was broken, is ordinarily expressed in Tamil by udeindadu, the pre- terite (third person singular neuter) of udei, intransitive, to become broken ; and though this is a neuter, rather than a passive properly so called, and might literally be rendered, ' it has come into a broken condition,' yet it is evident that, for all practical purposes, nothing more than this is required to express the force of the passive. The passivity of the expression may be increased by prefixing the instrumental case of the agent — e.g., enndl udeindadu, it was broken by me, literally it came into a broken condition through me. (2.) A very common mode of forming the passive is by means of the preterite verbal participle of any neuter or active verb, followed by the preterite (third person singular neuter) of the verbs to become, to be, to go, or (occasionally) to end. Thus, we may say either mudindadu, it is finished, or mudind' dyittru, literally, having finished it is become. This form adds the idea of completion to that of passivity : not only is the thing done, but the doing of it is completed. Transitive or active verbs which are destitute of intransitive forms may in this manner acquire a passive signification. Thus hatt-u, to bind or build, is necessarily a transitive verb, and is without a corresponding intransitive ; but in the phrase hovil hatti dyittru, the temple is built, literally, the temple having built has become, a passive signification is acquired by the active voice, without the assistance of any passive-forming particle, p dyittru, it has gone, may generally be used in such phrases instead of dyittru, it is become. Verbal nouns, especially the verbal in dal or al, are often used in Tamil instead of the preterite verbal participle in the formation of this constructive passage — e.g., instead of seyd' dyittru, it is done, literally, having done it has become, we may say seydal dyittru, which, though it is used to express the same meaning, literally signifies the doing of it has become — i.e., it has become a fact, the doing of it is completed. The Dravidian constructive passives now referred to require the THE PASSIVE VOICE 465 third person neuter of the auxiliary verb. The force of the passive voice will not be brought out by the use of the mascuHne or feminine, or by the epicene plural. If those persons of the verb were employed, the activity inherent in the idea of personality would necessitate an active signification ; it would tie down the transitive theme to a transitive meaning ; whereas the intransitive relation is naturally implied in the use of the actionless neuter gender, and therefore the expression of the signification of the passive (viz., by the intransitive doing duty for the passive) is facilitated by the use of the third person neuter. A somewhat similar mode of forming the passive has been pointed out in the Hindustani and Bengali — e.g., jdnd ydy, Beng. it is known, literally, it goes to be known, jdnd is represented by some to be a verbal noun, by others to be a passive participle ; but, whatever it be, there is some difference between this idiom and the Dra vidian one ; for in the corresponding Tamil phrase terind' dyittru, it is known, terind-u is unquestionably the preterite verbal participle of an intransitive verb, and the phrase literally means ' having known it is become.' terindu p dyittru, literally, having known it is gone, conveys the same signification. It is remarkable, however, that a verb signifying to go should be used in the Dra vidian languages as a passive-making auxiliary, as well as in the languages of Northern India. Occasionally Dravidian active or transitive verbs themselves are used with a passive signification, without the addition of any in- transitive auxiliary whatever. Relative participles and relative participial nouns are the parts of the verb which are most frequently used in this manner — e.g., erudina suvadi undu ; acJicJi aditta fustagam vendum, Tam. I have a written book ; I want a printed one. In this phrase both erudina, written, and achch' -aditta, printed, are the preterite relative participles of transitive themes. The former means literally ' that wrote,' yet it is used passively to signify ' written' ; and the latter means literally ' that printed or struck off,' but is used passively as equivalent to ' that is printed.' The relative participial noun, especially the preterite neuter, is oftentimes used in the same manner — e.g., in sonnadu podum, Tam. what was said is sufficient, sonnadu, literally means ' that which said ' ; but the connectioif and the usage of the language determine it to signify passively that which ivas said ; and so distinctively in this case is the passive sense expressed by the connection alone, that the use of the more formal modern passive, solla-(f)fattadu, would sound 2 H 466 THE VERB awkward and foreign, endra, Tarn., anede, Tel., that is called, literally that spoke, is another very common instance of the same rule. lyesu enhavar, Tam., signifies literally, Jesus who speaks ; but usage determines it to mean he who is called Jesus. The mode of expressing the passive adopted by Tulu is on the whole similar to this. The perfect active participle is used for the passive in this manner, but the pronoun is repeated at the end — e.g., dye nindis'dindye dye, he is one who has despised, meaning, he is one who has been despised. (The corresponding Tamil would be avan nindittavan avan.) (3.) The passive is formed in Gond in a manner peculiar to that language, viz., by the addition of the substantive verb I am to the participle of the active voice. In the other Dra vidian dialects this is the usual mode in which the perfect tense is formed. In Tamil, ndn aditt' iruhkiven, I am having beaten, means I have beaten. The corresponding Gond expression ana jisi aidtona means I am beaten. This corresponds to the modern English mode of forming the passive, as in this very expression, I am beaten ; but still more closely to the mode adopted by New Persian, in which the same form of the verb has an active meaning when it stands alone, and a passive meaning when followed by the substantive verb. (4.) The verb un, to eat, is occasionally used in the Dra vidian languages as an auxiliary in the formation of passives. It is invari- ably appended to nouns (substantives or verbal nouns), and is never compounded with any part of the verb — e.g., adi unddn, he was beaten, or got a beating, literally he ate a beating ; fadeiff' widen, I was created, literally I ate a creating. The same singular idiom prevails also in the Gaurian or North Indian vernaculars. The par- ticular verb signifying to eat used in those languages differs indeed from the Dra vidian un ; but the idiom is identical, and the existence of so singular an idiom in both the northern and the southern family is deserving of notice. It is remarkable that the same peculiar contrivance for expressing the passive is found in Chinese, in which also to eat a beating means to be beaten. (5.) Another mode of forming the passive used in each of the modern cultivated colloquial dialects of the Dra vidian family, except Tulu, is by means of the auxihary verb fad-u, to suffer, to experience, which is annexed to the infinitive of the verb signifying the action suffered — e.g., kolla-{j))j)attdn, Tam. he was killed, literally, he suffered a killing. It is also annexed to nouns denoting quality or condition — e.g., vetka-(f)fdttdn, he was ashamed, literally, he THE MIDDLE VOICE 467 suffered or experienced shame. The ultimate base of a verb is sometimes used instead of the infinitive or verbal noun in con- struction with this auxiliary, in which case the base is regarded as a noun — e.g., instead of adikka-{'p)faUdn, we may say adi pattdn, he was beaten, or literally he suffered a beating ; and where this form can be used, it is considered more idiomatic than the use of the infinitive. It is evident that this compound of pad-u, to suffer, with an infinitive or noun of quality, is rather a phrase than a passive voice. It is rarely found in the classics ; and idiomatic speakers prefer the other modes of forming the passive, fad-u is often added, not only to active, but also to neuter or intransitive verbs ; but as the intransitive expresses by itself as much of a passive signification as is ordinarily necessary, the addition of the passive auxiliary does not alter the signification — e.g., there is no difference in Tamil between the intransitive teriyum, it appears, or will appear, and teriya {'p)padum ; or in Telugu between telusunu and teliya hadunu, the corresponding forms. In ordinary use, pad-u conveys the mean- ing of continuous action or being, rather than that of passivity — e.g., irukka-{'p)'patta (Tam.) is vulgarly used for irukkita, that is ; and I have heard a Tamilian say, nan nandrdy sdffida-(f)fattavan (Tam.), meaning thereby, not I have been well eaten, but I have been accustomed to eat well. The Dravidian languages, indeed, are destitute of passives properly so called, and, therefore, resist every effort to bring fad-u into general use. Such efforts are constantly being made by foreigners, who are accustomed to passives in their own tongues, and fancy that they cannot get on without them ; but nothing sounds more barbarous to the Dravidian ear than the unnecessary use of fadu as a passive auxiliary. It is only when combined with nouns that its use is thoroughly allowable. 7. The Middle Voice. In none of the Dravidian dialects is there a middle voice, properly so called. The force of the middle or reflexive voice is expressed constructively by the use of an auxiliary verb — viz., by kol, Tam. to take (Tel. kon-u ; Tulu, konu and onu) — e.g., 'panni-{k)kond^'i^, I made it for myself, literally, I made and took it. This auxiliary sometimes conveys a re(?iprocal force rather than that of the middle voice — e.g., fesi-{k)konddrgal, Tam. they talked together ; adittu-{k) konddrgal, they beat one another. The same usage appears in the other dialects also. 468 THE VERB 8. The Negative Voice. Properly speaking, the Dravidian negative is rather a mood or voice than a conjugation. All verbal themes are naturally affirm- ative, and the negative -signification is expressed by means of additions or changes. Nevertheless, it will conduce to perspicuity to inquire now into the negative mood or voice, before entering upon the consideration of the pronominal terminations and tenses. The regular combination of a negative particle with a verbal theme is a peculiarity of the Scythian family of tongues. Negation is generally expressed in the Indo-European family by means of a separate particle used adverbially ; and instances of combination like the Sanskrit ndsti, it is not, the negative of asti, it is, are very rare ; whereas, in the Scythian languages, every verb has a negative voice or mood as well as an affirmative. This is the case also in Japanese. The Scythian negative voice is generally formed by the insertion of a particle of negation between the theme and the pro- nominal suffixes ; and this is as distinctive of the Dravidian as of the Turkish and Finnish languages. Different particles are, it is true, used in the different languages to express negation ; but the mode in which such particles are used is substantially the same in all. In general, the Dravidian negative verb has but one tense, which is an aorist, or is indeterminate in point of time — e.g., fogen, Tarn. (povanu, Tel., pogenu, Can.), I go not, means either I did not, I do not, or I will not go. The time is generally determined by the context. Ku, Gond, and Tulu use the negative more freely. In Ku there is a negative preterite as well as a negative aorist ; and in Tulu and Gond every tense of every mood has its appropriate negative verb. Malayalam has three negative tenses — the present, the past, and the future — e.g., f6gd-{y)-unnu, I go not ; 'pogd-nnu, went not ; f6gd-(y)-um, will not go. In the other dialects there is only one mood of the negative in ordinary use, viz., the indicative. If an infinitive and imperative exist, it is only in classical compositions that they appear ; and they are ordinarily formed by the help of the infinitive and imperative of the substantive verb, which are suffixed as auxiliaries to the negative verbal participle — e.g., Seyydd'-iru, Tarn, do not thou, literally, be thou not doing. In Telugu a prohibitive or negative imperative is in ordinary use even in the colloquial dialect. In the Dravidian negative voice, as in the affirmative, the verbal theme remains unchanged ; and in both voices the pronominal THE NEGATIVE VOICE 469 terminations are precisely the same. The only point, therefore, which it is necessary to investigate here is the means whereby the idea of negation is expressed. The Tamil-Telugu-Canarese negative is altogether destitute of signs of tense ; it is destitute, not only of the signs of present, past, and future time, but even of the sign of the aorist ; and in Tamil and Canarese the pronominal suffixes are annexed directly to the verbal theme. Thus, whilst the present, past, and future tenses (first person singular) of the affirmative voice of the Tamil verb vdr, to flourish, are vdr-giv-en, vdr-nd-en, vdr-v-en ; the corresponding negative is simply vdr-en, I flourish not — literally, as appears, flourish-I, — without the insertion of any sign of time between the theme and the pronoun. What is the rationale of this negative ? The absence of signs of tense appears to contribute to the expression of the idea of negation : it may at least be said that it precludes the signification of the affirm- ative. In consequence of the absence of tense-signs the idea ex- pressed by the verb is abstracted from the realities of the past, the present, and the future : it leaves the region of actual events, and passes into that of abstractions. Hence, this abstract form of the verb may be supposed to have become a negative mood, not by a positive, but by a negative process, — by the absence of a predicate of time, not by the aid of a negative particle. Is this to be accepted as the rationale ? If we examined only Tamil and Canarese, we might be satisfied with this explanation ; for in the various persons of the negative voice in both languages there is no trace of the insertion of any negative particle ; and though the vowel a has acquired a predominant and permanent place in the verbal and relative participles, we should not feel ourselves warranted in con- sidering that vowel as a particle of negation, without distinct, trustworthy evidence from some other source. The only peculiarity in the personal forms of the Tamil negative is the invariable length of the initial vowel of the pronominal terminations. Thus the initial a of the neuter singular demonstrative being short, we should expect the Tamil of ' it flourishes not ' to be vdr-adu ; whereas it is vdr-ddu or vdr-d. This increase of quantity might arise from the incorporation and assimilation of some inserted vowel ; but we might alio naturally suppose it to be merely length- ened euphonically for the sake of emphasis. The corresponding vowel is short in Telugu. In the Canarese negative we miss even this lengthening of the initial vowel of the pronominal terminations — e.g., 470 THE VERB we find invariably hdl-adu, instead of the Tamil vdr-ddu. In the verbal and relative participles in both languages the vowel a is inserted between the theme and the formative, and this a is invariably short in Canarese and long in Tamil — e.g., hdl-a-de, Can. not having lived, or without living ; Tam. vdr-ddu or vdr-d-mal, without living. The verbal noun in Tamil is vdr-d-mei, the not living. The relative participle that lived or lives not, is in Canarese hdl-a-da, in Tamil vdr-d-da. In these instances, if euphony alone had been considered, u, the ordinary enunciative vowel, would have appeared where we find a : it may, therefore, be concluded that a (euphonically d in Tamil and Malayalam, has intentionally been inserted, and that it contributes in some manner to grammatical expression. It will be found that light is thrown upon this subject by Telugu. The pronominal terminations of the negative voice of the Telugu are identical with those of the present tense of the afiirmative. In Tamil and Canarese the pronominal terminations of the verb com- mence with a vowel ; but in Telugu verbs the pronoun is represented by the final syllable alone, and that syllable invariably commences with a consonant. Hence, if no particle of negation were used in the conjugation of the Telugu negative voice, the pronominal suffix would be appended directly to the verbal theme, and as every Telugu theme terminates in the enunciative u, that u would not be elided, but would invariably remain. What then is the fact ? On examining the Telugu negative, it is found that the vowel a invariably intervenes between the theme and the pronominal suffix ; and as the final enunciative u of the theme has been elided to make way for this a, it is evident that a is not an euphonic insertion, but is a particle of negation. Compare chey-a-nu, Tel. I do not, with Tamil sey(y)en ; chey-a-vu, thou dost not, with Tamil sey{y)-dy ; chey-a-mu, we do not, with Tamil sey(y)-6m ; cJiey-a-ru, you do not, with Tamil Sey{y)ir. From this comparison it cannot be doubted that a is regularly used in Telugu as a particle of negation. We find the same a used in Telugu, as in Canarese and Tamil, in the negative verbal participle — e.g., chey-a-ka, without doing ; in the relative participle — e.g., chey-a-ni, that does not ; and in the verbal noun — e.g., chey-a-mi, the not doing. In each of these participials a is used in the same manner by the Canarese, and d by the Tamil : and that those vowels are not euphonies or conjunctives, but signs of negation, even in Tamil-Canarese, is now proved by the evidence of Telugu, in which a similar a is used, not only by the participles, but by all the personal forms of the verb. THE NEGATIVE VOICE 47 1 The Telugu verb to go forms its ordinary negative, it is true, without any trace of this vowel of negation — e.g., fonu, I go not, povu, thou goest not. This, however, is only an apparent irregularity, for the classical forms are fov-a-nu and fov-a-vu. The lengthening of the included a of Mnu, I become not, is in accordance with the Telugu law of displacement, kdnu being instead of ak-a-nu or ag-a-nu, the equivalent of the Tamil dgen. We have thus arrived at the conclusion that a is the sign of negation which is most systematically used by the Dravidian languages in the formation of the negative voice of the verb. It has, it is true, disappeared from the conju- gated forms of Tamil and Canarese ; but the analogy, not only of the Telugu personal forms, but also of the Tamil and Canarese participles and participial nouns, shows that it must originally have been the common property of all the dialects. The negative a, being succeeded in Tamil and Canarese by the initial vowel of the pronominal suffix, appears gradually to have got incorporated with it ; and an evidence of this incorporation survives in the euphonic lengthening of the pronominal vowel in Tamil. The negative particle of the Tulu is ijji, answering to Tarn, illei, Mai. and Can. ilia. Most of the tenses of the Tulu negative verb are formed by annexing to the temporal particles of the verb ;, the abbreviation of this ijji, with such enunciative vowels as euphony is supposed to require. The negative of the future tense appears to be formed from a, the particle used in the other dialects. Comp. mal- fuji, I do not make, malfdiji, I have not made, with malpaye, I shall not make, and the conditional form tnalt'dvaye, I should not make. Gond inserts the negative particles hille or halle (Drav. ille or alle) between the pronoun and the verb, without abbreviation. This crude use of the form has doubtless come down from a high antiquity, as we shall find that al is sometimes used in a somewhat similar manner by the Tamil poets. It is desirable now to inquire into the participial and imperative formatives of the negative verb. The negative verbal participle of Tamil is formed by suffixing d-du or d-mal — e.g., sey(y)-d-du or sey{y)' d-mal, not doing, or without doing. In the highest and lowest Tamil mei is used as the formative of this participle instead of mal — e.g., varuv-d-mei, without slipping, mei constitutes the ordinary termina- tion of abstract nouns, «ind is added both to crude roots and to the relative participles of verbs — e.g., tdr-mei, lowness, humility ; iru- Jckindr-a-mei, a being or the being. The formative termination of negative verbal nouns is identical with this abstract mei ; and mal, 47^ THE VERB the participial formative, is evidently equivalent to it. Probably also it is the original form ; for, on the whole, it is more likely that a final I should have been softened away than added. The verbal noun of the Telugu negative verb ends in mi, which is virtually the same as mei. The other Tamil termination of negative verbal participles, du, is an ordinary formative of neuter nouns of quality. The corre- sponding Canarese termination is de ; and in Tamil du, with a subse- quent emphatic e, is commonly used as a negative imperative or prohibitive — e.g., sey{y)-d-d-e, do not thou, — a proof that the nega- tive verbal participle in du or de is properly a verbal noun. The relative participle of the negative verb in each of the dialects, except Telugu, is formed by suffixing a, the sign of the relative, to the verbal participle in d-u, eliding as usual the enunciative u — e.g., sey{y)-d-da, Tam., gey-a-da. Can., that does or did not. Many additional forms are constructed by the addition of the various tenses and participles of the substantive verb, and it is by the help of that verb that the negative imperative and negative infinitive in both Canarese and Tamil are ordinarily formed. The negative relative participle of Telugu is formed by adding ni, instead of the usual relative a, to the negative particle — e.g., chey-a-ni, that does or did not. This ni is one of the Telugu inflexional increments, and is also used as a particle of conjunction, as will be seen under the head of the " Eelative Parti- ciples." Mr A. D. Campbell, in his " Telugu Grammar," states that the negative verbal particle of the Telugu is formed by suffixing Jca to the infinitive of the affirmative voice ; and that the prohibitive is formed in like manner by suffixing ku or ka to the infinitive [ka is not so used], with the ordinary addition of mu or mo. In consequence of this representation, Dr Stevenson was led to consider ku as a Telugu sign of negation, and to search for allied or equivalent particles in other Indian languages. The comparison of the negative verbs in the various Dra vidian dialects which has just been made proves that this representation is inaccurate, and that the a to which the ka and ku aforesaid are suffixed is not the a which forms the sign of the infinitive, but the negative particle a. The suffixes of the forms in question, therefore, are not ku or ka, but a-ku and a-ka ; and thus chey-a-ka, without doing, or not having done, and chey-a-ku, do not, come into harmony with the other Telugu forms, viz., chey-a-ni, that does not, chey-a-mi, the not doing ; and also with the negative parti- ciples and verbals of the other dialects. The a of the Telugu imperative and negative verbal participle being THE NEGATIVE VOICE 473 undoubtedly the sign of negation, it only remains to inquire into the origin of the ka or hu which is suffixed to it. The participial suffix ka is evidently used in Telugu for the same purposes as the Tamil suffixes du, mat, and tywi, and the Canarese de. Those suffixes, though used by verbal participles, are undoubtedly to be regarded as forma tives of verbal nouns. I consider ka also as proceeding from a similar origin ; for in Telugu many verbal nouns are formed in this very manner by adding ka to the root — e.g., nammi-ka, confidence, from nammu, to confide ; and kori-ka, hope, from koru, to hope. This ka is kkei, in Tamil {e.g., nambi-kkei, confidence), and ge or ke in Canarese : it is a very common formative of verbal nouns, and is equivalent in use to the formatives of which d oi t, b or 'p, is the initial. When we compare Telugu derivative nouns ending in ka (e.g., teliyi-ka, semblance, from teliyu, to appear) with the negative verbal participles of the same language, which invariably end in ka {e.g., teliy-a-ka, not seeming), it is evident that the particle ka is not that by which the difference in meaning is expressed. The a which precedes ka is evidently the seat of the difference. In those cases in which the derivative noun and the negative participle are absolutely identical in sound and appearance, the negative a has been absorbed by the preceding long a of the root. This is the cause of the similarity between rdka, a coming, and rdka, not or without coming, the latter of which is for ra-a-ka. In the dialect of the Kotas of the Nilgherry Hills, f appears to be used as the formative suffix of the negative verbal participle instead of the Telugu k and the Tamil-Canarese d — e.g., hogd-fe, without going, corresponding to the Canarese hogade, and the Telugu povaka. This is in accordance with a rule often already noticed, viz., the interchangeableness of k and p in the formatives of verbs and nouns. The Telugu prohibitive suffix ku is, I conceive, substantially identical with ka, the suffix of the verbal participle, just as de, the colloquial Tamil prohibitive, is identical with du, the negative verbal participle in the same dialect. Dravidian imperatives are in general nothing but verbal nouns pronounced emphatically. Hence, the Tamil sey{y)- d-de, do not thou, is simply sey{y)-d-du, doing not, with the addition of the emphatic e ; and the Telugu chey-a-ku, do not thou, is in like manner, I conceive, identical with the verbal participle chey-a-ka, doing not, or without doing, with an emphasis understood. There is in classical Tamil a prohibitive particle which nearly cor- responds to this Telugu prohibitive, viz., avka — e.g., sey{y)-atka, do not. It is used in connection with both numbers and every gender ; 474 THE VERB and I believe that it is by usage only that the corresponding Telugu form is restricted to the second person singular ; for when we com- pare the Tamil sey(y)-atka and the Telugu cJiey-a-ku, we can scarcely doubt that they are substantially identical. What is the origin of this Tamil prohibitive suffix avha ? It is derived from at (pronounced at before h), the particle of negation, the origin of which from the negative base a will presently be shown, and ka, which is identical with ka or ga, a sign of the Tamil infinitive, optative, or polite impera- tive, apparent in such words as vdr-ga, may (he, thou, you, they, &c.) flourish. This infinitival, participial, or imperative form appears to have been originally a verbal noun. We should here notice the prohibitive particle of Gond, viz., manni or minni. This is not sufiixed to the verb, but prefixed, like the Latin noli, manni closely resembles the Tamil suffix min, in such words as sey{y)an-min, do not ye ; but the resemblance is purely accidental, for the prohibitive particle of sey{y)an-min is an euphonised from al, and min is not, as Beschi supposed, a prohibitive particle at all, but is a sign of the second person plural of the imperative, and as such is systematically used in the higher dialect by the imperative of the affirmative voice, as well as by the prohibitive — e.g., fovu-min, bear ye. This in Malayalam is vin, fin (see the imper. of the affirmative). In poetical Tamil also atfin {al-fin) is occasionally used instead of an-min. There is also a plural form of this, avpir. Possibly the Gond prohibitive, manni, may be connected with the Hindustani mat and the Sanskrit md, or, but very remotely, with the Turkish particle of negation me or ma, which is used like the Dravidian a in the for- mation of the negative voice of the verb, manni resembles inni, the prohibitive particle of the Scythian tablets of Behistun. Origin of 'a,' the Dravidian Negative Particle. — We have seen that a is the sign of negation in Dravidian negative verbs, and that it is inserted between the theme and the signs of personality and other suffixes to form the negative voice. Has this a any connection with the alpha privative of the Indo-European tongues ? I think not, though this would seem a more natural use of the alpha privative than that of forming the temporal augment in Sanskrit and Greek, according to Bopp's theory. There is no trace of alpha privative or any equivalent privative pre-^x in the Dravidian languages ; and its place is supplied by some post-^xed relative participle or verbal noun formed from il or al — e.g., from ner, Tam. straight or straightness, is formed ner-inmei {il-mei euphonised), crookedness, want of straight- THE NEGATIVE VOICE 475 The negative a of the Dra vidian negative verb is, I have no doubt, , equivalent to al or it, the ordinary isolated particle of negation. This very sign of negation is sometimes used by the Tamil classics instead of a in verbal combinations — e.g., ari{g)-il4r, you know not, takes the place of the more common an-(y)4r : compare also nmei-{y)-al-d, not considering ; sey(g)-al-dddr, they will not do, or they who will not do. In all these examples the al is evidently the isolated negative particle. There cannot be any doubt whatever of the negative force of al in the negative appellatives, which are formed from al-an or il-an, he is not, combined with verbal roots — e.g., pes-al-em, we speak not, und-il-ei, thou eatest not or hast not eaten. Compare also mdttralan (mdxtalan), Tam. and Mai. an enemy, from mdttru-\-al-\-an, he who cannot be changed. Dr Gundert derives this from 7nattrd-\-ul-\-an, he who is -[-unchangeable. In the ordinary negative form, mdttrdn, Tam. and Mai. an enemy, the idea of negation is expressed by d ; but in mdttra-alan I have no doubt we have the negative particle al. Gond regularly forms its negative voice by suffixing Jialle or Mile, a barbarous euphonisation of the more correct alle or ille ; and the dialect of the Kotas makes a similar use of the particle ilia. This particle is also systematically used in forming the prohibitive, or negative imperative, of poetical Tamil, in which connection al is ordinarily lengthened to dl or el — e.g., sel-el, go not, muni-{y)-el, be not angry. But it is also, as we have seen, often retained unchanged — e.g., sey(y)-av-ha (at for al), do not, and sey{y)-an-min {an for al), do not ye. In modern colloquial Tamil, illei (for ilia) is commonly sub- joined to the infinitive of the affirmative verb to form an aoristic negative — e.g., vara-(v)-illei (I, thou, he, &c.), did not, do not, or will not come. This form, though very common, is not classical, and has arisen from the tendency which compounds evince to break up in process of time into their component elements. It is evident that a, the sign of negation in the Dra vidian negative verb, and al, the isolated negative particle, are substantially identical. The use of al instead of a in various verbal combinations in classical Tamil seems to me to prove this point. It remains, however, to endeavour to ascertain which is the older form. Has a been softened from al ? or is al a secondary form of a ? There are several parallel instances of the apparent disappearance of a final I — e.g., dal, the formative of many veibal nouns in Tamil, is represented by ta in Canarese and Telugu. Thus muti-dal, Tam. a breaking, is in Can. mura-ta ; sey-dal, Tam. a doing, is in Tel. cJie-ta. The infinitive is al or a in Canarese, a alone in Tamil. We have seen also that the Tamil 476 THE VERB suffix of the negative verbal noun may be either mal or met. None of these instances, however, is decisive ; as it may be supposed, and is I think probable, that a final I, answering to a final m, n, or r, was annexed to many verbal nouns in process of time for the purpose of making them more distinctive. In those instances, therefore, a may be the primitive shape, al the secondary. The same explanation seems to be the most satisfactory mode of accounting for the double form of the negative particle. I regard a as the original shape of that particle — the primitive negative base — answering to a, the primitive demon- strative base, and al as the more fully developed form of the negative — a negative noun — answering to the demonstrative nouns am, ad, al, &c. I refer in this only to the resemblance in form between the demonstrative and the negative bases and nouns ; but perhaps we may now venture to go a step further, with Dr Gundert, and derive the negative meaning itself from the interrogative, and ultimately from the demonstrative. He says (in his private communication to me), " I believe the [remote demonstrative] pronoun a forms the [particle of negation in the] negative verb ; just as this a in its interjectional [syntactic] form has the signification of a question. From the meaning of a question comes the meaning of negation. adu varum-d ? will it happen ? ^it will never happen." In the collo- quial dialect of the Tamil, at least, it is certain that the idea of negation is very often expressed by putting a question. It is at once a poetical and a vulgar usage. I am unable, however, to agree with Dr. Gundert when he pro- ceeds to say that he does not consider al a negative in itself, but only a negative when followed by the negative particle a, as in the words alia, &c. Whether al may or may not have been a demon- strative in origin, as I think it probably was, yet, when used as a particle of negation, it seems to me certain that it is a negative of itself without any addition, and that the added vowels a, &c., are merely enunciative. This applies with equal force to the correspond- ing negative particle il. The following words in Tamil seem to me to prove that al and il have of themselves the full force of negatives. Al : — andru (al-du), it is not ; (class. Can. altu, Tulu, atf) ; amnei {al-mei), notness, negation ; al-gu, to become less, al, darkness. al-vari, a grammatical term, absence of inflexion. Il : — indru, it is not ; inmei (il-mei), notness, non-existence. Hi, one who has nothing ; il-forul (forul, thing), non-existence, &c., the thing that is not. Whatever opinion we entertain respecting the derivation of al APPELLATIVES 477 from a, the widely extended affinities of al, dl, or el, the prohibitive or negative imperative particle, are deserving of notice. Compare the Sanskrit prohibitive particle alam, no, not, which looks as if it were derived from the Dra vidian al. The prohibitive particle of the Santal, a K61 dialect, is did ; the Finnish prohibitive also is did ; the Ostiak ild ; and we find a similar prohibitive particle even in Hebrew — viz., al : Chaldee, Id. 9. Appellative Verbs, or Conjugated Nouns. In some languages of the Ugrian group the pronominal termina- tions of the verbs, or those pronominal fragments in which verbs commonly terminate, are suffixed directly to nouns ; which nouns become by that addition denominative or appellative verbs, and are regularly conjugated through every number and person — e.g., from the noun faz, the Lord, the Mordvin forms faz-dn, I am the Lord ; and from the possessive paz-an. Lord's, it forms paz-an-dn, I am the Lord's. Adjectives being merely nouns of quality in the Scythian languages, every rule which applies to nouns applies to adjectives also. In the New Persian, possibly through the influence of the conterminous Scythian languages, there is a similar compound of a noun or an adjective with the verbal terminations — e.g., merd-em, I am a man, from merd, a man, and em, the contracted form of the substantive verb I am. This class of compounds resembles, but is not identical with, the class of possessive compounds described in p. 307 ; that class is not found in the Dra vidian languages. The agreement between the Dravidian languages and those of the Ugrian family with respect to the formation of appellative verbs of the character referred to is very remarkable, and has been admitted to be very remarkable by Professor Hunfalvy, though in other par- ticulars he fails to see much resemblance between the Finno-Ugrian and the Dravidian languages. Any Dravidian noun and any adjec- tive may be converted into a verb in the more ancient dialects of each of the Dravidian languages, and in some connections even in the colloquial dialects, by simply suffixing to it the usual pro- nominal fragments ; and not only may nouns in the nominative case be thus conjugated as verbs, but even the oblique case-basis, or virtual genitive, may in classical Tamil, as in Mordvin, be adopted as a verbal theme. Tamil grammarians call the verbs here described vmei-{k)kuiiffu, literally verbal signs ; and they have, not inappro- priately, been styled conjugated nouns by an English writer on 478 THE VERB Tamil grammar : but I think the best name is that which was given them by Beschi — viz., appellative verbs or conjugated appellatives. Appellative verbs are conjugated through every number and person, but they are restricted to the present tense ; or rather, they are of no tense, for the idea of time is excluded from them. Thus, from Mn, Tam. a shepherd or king, may be formed kon-en, I am a king, kon-ei, thou art a king, kon-em, we are kings, kon-ir, ye are kings. So also we may annex to the crude base the oblique or genitival formative in, and then from the new constructive base kon-in, of the king, or the king's, we may not only form the appella- tive nouns, kon-in-an, he who is the king's, kon-in-ar, they who are the king's (each of which may be used also as an appellative verb, which signifies he is the king's, or they are the king's), but we may also form the more distinctively verbal appellatives, kon-in-en, I am the king's, kon-in-em, we are the king's, &c. This use of the oblique or inflexion as the basis of appellative verbs is a peculiarity of clas- sical Tamil ; but the formation of appellative verbs from the nominative or crude base of nouns is common to the whole Dra vidian family. Thus, in Telugu (in which the vowel of the pronominal termination varies by rule in accordance with the preceding vowel), from sevakudu, a servant, or kavi, a poet, we form the appellative verbs sevakuda-nu, I am a servant, kavi-ni, I am a poet ; seva- kuda-vu, thou art a servant, kavi-vi, thou arb a poet. In the plural, Telugu has allowed the base of the noun (to which the pronominal terminations are affixed) to be pluralised, apparently from having forgotten that the plural sign of the pronominal termination was sufficient of itself — e.g., it says sevakula-mu, we are ser- vants ; whereas in Tamil the difference between adi-(y)-en, I am (your) servant, and adi-{y)-em, we are (your) servants, appears in the pronominal terminations alone ; and the plan of denoting the plural which the Tamil has adopted is evidently more in accordance with the true theory of the appellative verb. The Malayalam singular adiyan or adiyen agrees with the Tamil, but the plural adiyannal bears marks of corruption. The classical Tamil words el-dm, all we, el4r, all ye, belong to this class. The Telugu appellative verb is destitute of a third person except in the neuter singular. It is obliged to be content with placing the isolated pronoun of the third person and the substantive noun in apposition, with a substantive verb understood — e.g., vddu kavi, he (is) a poet. Tamil is in this particular more highly developed, for its appellative verbs are freely conjugated in the third person in APPELLATIVES 479 each gender and number, by suffixing the final fragment of the pro- noun — e.g., ixom'nal, goodness or good, is formed nal{l)-an, he is good, nal{l)-al, she is good ; nal{l)-adu or nan-dru (for nal-du), it is good, nal{l)-ar, they (epicene) are good, nal{l)-ana, or nal{l)-a, they (neuter) are good. The neuter singular in Tamil may appear to take a variety of forms ; but on examination those various forms will be found to be identical, and the apparent differences which exist are owing either to the euphonic union of the final du with some previous consonant, or to its euphonic reduplication. The third person neuter, singular and plural (and occasionally the third person masculine and feminine also), of every species of Dravidian verb, is often used not only as a verb, but also as a verbal or participial noun. Its primary use may have been that of a participial noun, and its use as a verb may be a secondary one ; but at all events, the two uses are found to be interchangeable — e.g., iruhkitadu means either it is, or that which is, or the being, according to the context. It is especially with relation to appellatives that this twofold use of the forms of the third person must be borne in mind ; for in the third person (singular and plural, masculine, feminine, and neuter) there is no difference whatever in spelling or pronunciation between appellative verbs and appellative nouns, and it is the context alone that determines which meaning is the correct one. Generally the appellative verb is more commonly used in the classical dialect, and the noun in the colloquial dialect ; but to this there are exceptions, and (e.g.) nalladu more frequently signifies in the colloquial dialect ' it is well ' than ' that which is good ' — that is, it is used more frequently as an appellative verb than as an appellative noun. It is certain, however, that the appellative verb, whatever person or gender it takes, is used more largely in the higher dialect of the Tamil than in the lower ; and its brevity and compression render it pecu- liarly adapted for metaphorical use. Adjectives are formed into appellative verbs as well as nouns ; but as the Dravidian adjective is merely a noun of quality used adjectiv- ally, the difference is more in terms than in reality — e.g., oli-{y)-ei. Tarn, thou art bright, is literally thou art brightness ; and ini-(y)-ei, thou art sweet, is thou art sweetness. Apellative verbs are formed from adjectives, or nouns of quality, not only in the cultivated Dravidian dialects, but ev^ in Ku, which is spoken by an uncultured race — e.g., negg-dnu, Ku, I am good, negg-dmu, we are good. When nouns of quality are used as the bases of appellative verbs or nouns, they are generally adopted in their crude shape, as in the 480 THE VERB instances which have just been cited ; but in many cases we find the particle iya intervening between the -crude b^tse and the pro- nominal termination or sign of gender — e.g., kod-iya-n (as a verb), he is cruel ; (as a noun) one who is cruel, or a cruel man ; val-iya-n, a strong man, or he is strong, &c. This is the same particle which we have already seen to be used as an adjectival formative — e.g., val-iya, strong, fer-iya, great, siv-iya, little, &c., and I have stated that I conceive words like these to be relative participles, i is iden- tical with the i of the past verbal particle, which is often used in Telugu as an adjectival formative without any addition ; and the final a is the sign of the relative, which is kept separate from i by an euphonic y. iya is therefore the formative of the relative preterite participle, and val-i-{y)-a, strong, means properly that which was strong. But though the form of the preterite tense is em- ployed, the signification (as often happens, especially in the case of relative participles) is aoristic, or without reference to time. This being the origin, as I conceive, of such forms as val-iya, an appellative noun like val-iya-n, a strong man, is in reality a participial noun, signifying he who is strong, and so of the other genders ; and this explanation brings such forms into perfect harmony with other parts of the Dra vidian conjugational system, for participial nouns are regularly used in these languages as verbs. In some instances a, the sign of the relative participle, is dispensed with, and the pronominal signs or signs of gender are elegantly suffixed to i, the sign of the verbal participial — e.g., peri-du, Tam. it is great, or that which is great, instead of j)eri-(y)-a-du. On the other hand, in another class of instances, i disappears, and a alone remains. Words of this class, when deprived of their signs of gender, are commonly called adjectives, and undoubtedly it is as adjectives that they are used ; but, looking at their construction and force, I should term them relative participles of appellative verbs. In the words referred to, a, the sign of the relative participle, is directly annexed to crude substantive roots — e.g., udei-{y)-a, belonging to, more literally which is the property of. malei-{y)-a, hilly, literally which is a hill ; ti-y-a, evil, literally which is evil. As udei-[y)-an, considered as a noun, is certainly an appellative, signifying he who owns, a proprietor ; and as the same word is used poetically as an appellative verb when it signifies he is the owner, it seems evident that the proper light in which to regard udei-{y)-a (and every similar word) is to consider it as the relative participle of an appellative verb used adjectivally. PRONOMINAL SIGNS 48 1 SECTION II.— CONJUGATIONAL SYSTEM. Mode of Annexing Pronominal Signs. -^The persons of the Dra vidian verb, including the related ideas of gender and number, are formed by suffixing the personal or demonstrative pronouns, or their fragmentary terminations, to the signs of tense. The change which the pronouns undergo when they are appended to verbs as. signs of personality have already been exhibited in the section on " The Pronoun." They consist chiefly in the softening away of the initial consonant ; but in a few instances the final consonant has also been softened away, and nothing left but the included vowel. In Telugu, ni-vu, the pronoun of the second person singular, has lost both its radical initial and its formative final ; and in the personal terminations of the verb it is represented only by vu, an euphonic addition. In the Indo-European languages the personal signs of the verb are formed by suffixing pronominal fragments to the root ; and those fragments are disguised in a still greater degree than in the Dra vidian languages, not only by frequency of use and rapidity of enunciation, but also by the love of fusing words and particles together, and form- ing them into euphonious compounds, which distinguishes that family of tongues. Sometimes one dialect alone furnishes the key to the explanation of the inflexional forms which are apparent in all. Thus the origin of unt or ant, the sign of the third person plural in the various Indo-European languages {e.g., fer-unt, ^f/o-ovrt (pJier- onti), bharanti, &c.), is found in Welsh alone, in which hwynt is a pronoun of the third person plural. The various changes which the Dravidian pronouns undergo on being used as the pronominal signs of verbs have already been stated in order. In Telugu, and partly also in Canarese, the pronominal terminations vary according to the tense ; but this arises from the operation of the law of harmonic sequences (see " Sounds "), by which a vowel is affected by a preceding vowel, and changed so as to harmonise with it. What requires here to be investigated is simply the mode in which the pronominal signs are attached to the Dravidian verb. 1. The pronominal signs of the Dravidian verb are suffixed, not prefixed. The primitive luranian verb seems to have been destitute of pronominal terminations altogether. The pronoun was neither prefixed nor affixed, but had a position of its own as a separate word. This continues to be the case with the most distinctively Turanian 2 I 4o2 THE VERB languages ; but in the Buriat dialect of the Mongolian, and in the Tungusian idiom, spoken near Njertschinsk in Siberia, personal terminations have recently been added to the verb. In Turkish, Finnish, and Hungarian, as in the primitive Indo-European lan- guages, the pronouns have been compounded with the verb, and have dwindled down to pronominal terminations. In the modern Indo-European vernaculars, most of the verbs have lost their old pronominal terminations, and the pronouns which are used as nominatives to verbs are usually isolated and placed first. Thus, instead of love-I, in accordance with the ancient am-o, we have learnt to say I love, — an alteration of position which produces no change in meaning. In the Semitic languages a change in the position of the pronoun from the termination of the verb to its commencement produces an important change in grammatical signification : the position of the pronouns or pronominal fragments determines the tense. When the pronominal fragments are prefixed, the tense of the verb is regarded as future or aoristic : it is regarded as past when they are suffixed. Prefixing the pronominal fragments appears to denote that the action of the verb has, as yet, only a subjective existence in the mind of the speaker or agent — i.e., it is future ; suffixing them may denote that the action of the verb has already acquired an objective existence, apart from the will or wish of the speaker or agent — i.e., it is past. No peculiarity of this kind characterises the Dra vidian languages. The tenses are formed, not by means of the position of the pronouns, but by particles or signs of present, past, and future time suffixed to the theme ; and the personal signs, as in the Turkish and Finnish families, are suffixed to the signs of tense. The only exception to this rule is that which forms the most characteristic feature of Malayalam — a language which appears to have been originally identical with Tamil, but which, in so far as its conjugational system is concerned, has fallen back from the inflexional development reached by both tongues whilst they were still one, to what appears to have been the primitive condition of both — a condition nearly resembling that of the Mongolian, the Manchu, and the other rude primitive tongues of High Asia. In ancient times, as may be gathered from Malayalam poetry, and especially from the inscriptions pre- served by the Syrian Christians and the Jews, the pronouns were suffixed to the Malayalam verb, precisely as they still are in Tamil. At present, the verb is entirely divested, at least in the colloquial dialect, of signs of personality ; and with the pronouns the signs of PRONOMINAL SIGNS 483 number and gender also have necessarily disappeared ; so that the pronoun or nominative must in every instance be separately prefixed to the verb to complete the signification, and it is chiefly by means oi this prefixed pronoun that a verb, properly so called, is distinguished from a verbal participle. Though the personal signs have been abandoned by the Malayalam verb, the signs of tense or time have been retained, and are annexed directly to the root as in the other dialects. Even in modern English some persons of the verb retain archaic fragments of the pronominal signs (e.g., lovest, loveth.) ; but in modern Malayalam every trace of those signs has disappeared. Thus, whilst we should say in Tamil aditten, I beat ; adittdy, thou didst beat ; adittdn, he beat ; Malayalam uses in these and all similar cases the verbal participle adichu (for adittu), having beaten, with the prefixed pronouns I, thou, he, &c. — e.g., nan adichu, I beat ; nt adichu, thou didst beat ; avan adichu, he beat. Though the pro- nominal signs have been lost by the Malayalam verb, they have been retained even by the Tuda ; and notwithstanding the comparative barbarity of the Gonds and Kus, their conjugational system is peculiarly elaborate and complete. 2. Another peculiarity in the manner in which the personal signs are suffixed in the Dravidian languages consists in their annexation, not directly to the root, as in the Indo-European family, but to the temporal participles. The first suffix to the root in the affirmative voice is that of the sign of tense, then follows the suffix of personality. Every pure Dravidian affirmative verb is compounded of three elements, which are thus arranged and named by Tamil grammarians, viz. (1) the pagudi (prakriti. Sans.), or root ; (2) the idei nilei, or medial particle, i.e., the sign of tense ; and (3) the vigudi {vikriti, Sans.), the variation or differentia, i.e., the pronominal termination. When the signs of tense are attached to the theme, some euphonic changes take place (not in the theme, but in the signs themselves), which serve, as has been shown, to distinguish transitive verbs from intransitives. Other euphonic changes also take place in accordance with Dravidian laws of sound, which will be inquired into when those signs of tense are one by one examined. The changes which take place in the pronominal signs when they are annexed to the signs of tense have already been stated in the section on " The Pronoun." In the Indo-Europeaif languages we meet, I think, with no instance of the annexation of the pronominal signs to the participles, i.e., to the combination of the root with the signs of tense. I know of no instance of the use of any form like amant-o, instead of am-o, to 484 THE VERB signify I love. This, however, is the method which is invariably employed in the Dravidian languages, and which constitutes an essential element in the family likeness by which they are pervaded. It is also distinctive of Turkish. Thus, the Turkish olursen, thou art, is formed from olur, being, the present participle of the verb 61, to be, with the addition of the pronoun sen, thou. So also the Oriental Turkish boldmen, I am, is formed from hold, being (theme, hoi, to be), and the pronominal suffix rmn, I. An important difference generally found to exist between the Dravidian languages and the Gaurian vernaculars should here be stated. In the languages of Northern India the present tense of a verb is ordinarily formed by annexing the substantive verb to its present participle — e.g., karitechi, Beng. (karite-dchi), I am doing, instead of I do. In Telugu, perhaps through the influence of the North Indian vernaculars, a similar usage prevails ; but it is found in the present tense only ; it may readily be dispensed with ; and the simpler usage, which accords with that of all the other Dravidian dialects, is undoubtedly the more ancient. In Tamil and Canarese this use of the substantive verb, as an auxiliary in the formation of the present tense, is unknown : it is used as an auxiliary only in the formation of the compound preterite and future tenses. Malay- alam occasionally uses the substantive verb in a similar manner to Telugu, but with a somewhat different signification. In Telugu naduchutunndnu, I walk (from naduchu-tu, walking, and unndnu, I am), has simply the meaning of the present tense, and is equivalent to the simpler form naduchutdnu, answering to the Tamil nadakkiten, and the Canarese nadeyuttene ; but in Malayalam, whilst nan nadak- kunnu means I walk, nan nadakkunnunda has generally an emphatic sense — e.g., I am really walking. Tamil has a form precisely re- sembling this. 3. It is a peculiarity of Telugu that the third person of the pre- terite is sometimes left altogether destitute of the signs of time, person, number, and gender ; and this peculiarity applies also to the third person of the aorist. Thus, whilst unditini, I was, and unditivi, thou wast, are supplied with the usual signs of tense and person, the third person of the same tense is simply unde-nu, he, she, or it was, or they were, without distinction of number or gender, and without even the particle ti, which con- stitutes the usual sign of the preterite. The aorist third person, with a similar absence of distinction, is undu-nu ; and in both cases the final nu is merely a conjunctive suffix, like the corresponding PRONOMINAL SIGNS 485 Tamil um. Sometimes even the aorist formative nu is discarded, and the root alone is used as the third person singular. Thus (he, she, or it) falls or will fall, may either be fadu-nu, or simply fadu. The usage of poetical Tamil occasionally agrees with that of the Telugu with respect to the neuter gender, both singular and plural, especially in connection with the negative voice of the verb — e.g., sey(y)-d, it will not do, is often used for sey(y)-ddu. A usage similar to this prevails in many languages which are widely different one from the other. Thus, the New Persian uses for the third person singular of the preterite the contracted infinitive, as grammarians style it — an abstract verbal noun, which may be regarded as the theme of the verb. The Hebrew third person mas- culine of the preterite tense is also a verbal noun, without pronominal addition. We see a similar peculiarity in the third person of the present tense of the verb in some languages — e.g., compare the three persons of the present tense of the Turkish substantive verb, olurum, I am ; olursen, thou art ; olur, he is. Compare also the Armorican kanann, I sing ; kanez, thou singest ; kan, he sings. Compare with these examples the Hungarian ismerek, I know ; ismersz, thou knowest ; and ismer, he knows. 4. There are traces in ancient Tamil and Canarese of the existence of a very primitive system of conjugation. A form of the verb is occasionally used by the poets, which must have come down from a period of great antiquity. In High Tamil, seydu (sey-du), which is now the preterite verbal participle, may be used for the preterite tense of the finite verb in all persons in the singular, and seydum {sey-d'-um) (the same form with the addition of the conjunctive um, used as a pluralising particle), for all persons in the plural. A some what similar form may be used for the future, by means of the addition of ku or gu to the root, instead of the sign of the preterite, du. sey-gu is used to mean I will do ; sey-g'-um, we will do. The use of this form is not extended to the other persons so widely as that of seydu, an irregularity which shows that it had become nearly obsolete when it received a place in written compositions. The um of the aoristic future in modern Tamil is restricted to the neuter gender, but it is used for both numbers indiscriminately. The gu and gmn of poetical Tamil is found also in classical Canarese in the form of gu77i or kum, in whidfi it has a wider range of application than in Tamil. In classical Tamil its use is confined to the first person ; in classical Canarese it is used indiscriminately for all persons — e.g., avar mddugum, they do. ku also survives in Canarese — e.g., ke-ku 486 THE VERB (Tarn, vend-um), must. It would appear, therefore, that the Dra vi- dian verb was originally uninflected ; and this may partly account for the circumstance that Malayalam so readily lost the inflexions which, in common with Tamil, it had acquired. The period when the Dravidian verb was uninflected must have been long prior to the reparation of the present tongue into dialects, in all which, even in the rudest, a system of inflexions has been developed. The retention of traces of the ancient verb in Tamil and Canarese, and partly also, as noticed in the previous paragraph, in Telugu, seems to prove the great antiquity of the literary culture of the Dravidian languages. 5. The Dravidian verb, as now inflected, like the verb of many other languages, does not distinguish the genders of either the first person or the second, whether singular or plural ; but in the third person it marks all existing distinctions of gender with peculiar explicitness and minuteness. Thus, without the use of isolated pronouns, and employing the inflexions of the verb alone, we can say in Tamil varugivdn, he comes ; varugirdl, she comes ; varugiradu, it comes ; varugivdr, they (men and women) come, or honorifically he comes ; varugixdrgal, they (men and women) come ; varugin- drana, they (things) come. Formation op the Tenses. — Most of the Dravidian tenses are formed from participial forms of the verb : an inquiry into the parti- ciples is, therefore, a necessary preliminary to an inquiry into the tenses. Dravidian verbs have two species of participles, one of which (called relative participles, because they include the significa- tion of the relative pronoun) will be inquired into in a subsequent part of this section ; the other, commonly called verbal participles or gerunds, and which are now to be considered, constitute the bases on which the tenses are formed. The forms which are assumed by the verbal participles will be inquired into in connection with the signs of tense, from the consideration of which they cannot be severed. I content myself here with some general remarks on the signification and force of this class of words. Verbal Participles, their Signification and Force. — In ordinary collo- quial Tamil there is but one verbal participle, that of the past tense. In Malayajam and in classical Tamil there is a verbal participle of the future tense as well as of the past. In Canarese and Telugu there is a verbal participle of the present and of the past. In Tulu there are three verbal participles, that of the present (or future), that of the imperfect past, and that of the perfect. In this particular, therefore. FORMATION OF TENSES 487 colloquial Tamil may be considered as the poorest of the Dravidian dialects. Properly speaking, the words which are called verbal farticiples are not participles at all, seeing that they do not fariici- pate in the nature of adjectives, as all the Indo-European participles do. They have somewhat of the signification of gerunds, inasmuch as in addition to the idea of time, they include more or less of the idea of cause. Nevertheless, as each of the Indo-European participles is commonly used also as a gerund, without losing the name of a participle, and as the gerund in do (to which alone, amongst Latin gerunds, the Dravidian participles have any resemblance) has a very restricted application, it appears advisable, after all, to style these words participles instead of gerunds, — or more fully verbal parti- ciples, to distinguish them from what are called relative participles. The following sentences will illustrate the force of the Dravidian verbal participles : — 1. Present Verbal Participle, — This verbal participle, though unknown in Tamil and Malayalam, is commonly used both in Canarese and in Telugu. I quote the illustration which follows from Canarese. " Vikramarka, punishing the wicked and protecting the good, reigned over the kingdom." Here the English words ' punish- ing ' and ' protecting ' are participles of the present tense, used gerundially ; and the Dravidian words which they represent (in Canarese, sikshisuttd and rakshisuttd) have precisely the same force. In this respect only there is a difference between them, viz., that the English participles are capable of being used also as adjectives whereas the Dravidian words, though called participles, cannot be used adjectivally, or in any other way than that here exemplified. 2. Preterite Verbal Participle. — " Salivahana, having killed Vikra- marka, assumed supreme power." Though the English participle ' having killed,' which is here used, is a compound one (being formed from the present participle having, and the passive participle killed), its signification is that of a simple, uncompounded participle of the past tense, and the Dravidian word which it represents {kondru, Tam., kondu, Can.) is also a preterite active verbal participle. In this instance, neither the English participle nor the Dravidian one is capable of being used as an adjective. In reality, they are both preterite gerunds or gerundials, though they retain the name of participles as a matter 9i convenience. In those Dravidian dialects in which there is a present, as well as a preterite, verbal participle (as in Canarese and Telugu), the present is used to express subordinate actions which are contemporaneous with 488 THE VERB that which is denoted by the principal and finite verb ; whilst the preterite expresses subordinate actions which are antecedent in point of time to the principal action. In Tamil, the preterite participle is used to express all subordinate actions, whether simultaneous with the main action or antecedent to it ; but though that participle is always a preterite in form, it possesses the force of a participle of the present tense when the connection requires it. In each of the dialects and in every connection, the nominative of the final govern- ing verb is the nominative of all the subordinate verbal participles. The Dra vidian verbal participles may be compared with the San- skrit indeterminate past participle in tvd — e.g., kritvd, having done. Like that participle they are indeclinable and indeterminate. One of the chief peculiarities, however, of these verbal participles is, that they have a continuative force, dispensing altogether with the use of conjunctions. In the Dravidian languages, though nouns and pronouns are united by means of conjunctions, finite verbs are never so united. In every sentence there is but one finite verb, which is the last word in the sentence, and the seat of government ; and all the verbs which express subordinate actions or circumstances, whether antecedent or contemporaneous, assume an indeterminate, con- tinuative character, as verbal participles or gerundials, without the need of conjunctions or copulatives of any kind ; so that the sense (and more or less the time also) waits in suspense for the authoritative decision of the final governing verb. Hence those participles might properly be called continuative gerundials. Tamilian grammarians class them, together with infinitives and subjunctives, as vinei echcham, verb defects, or verbal complements — i.e., words which require a verb to complete the sense. It is a peculiarity of these languages that when a series of verbal participles constitutes a relative clause in a sentence, antecedent to a noun to which the relative clause relates, the last of the verbal par- ticiples alone is converted into a relative participle. All the rest remain in form verbal participles or gerunds. So also in the Scythian languages. " The Turanian," says Mr Ed kins, " in describing a succession of events gives to his verbs the form of gerunds, and adds to them, when needed, the case suffixes," — converting the gerund thereby into a relative participle, as in Tamil, &c. The rationale of the process seems to be that in both families of tongues the gerund is treated as a noun, and must have been a verbal noun in origin. 1. The Present Tense. — It may be stated generally that the present tense of the Dravidian verb is formed by suffixing the pronominal THE PRESENT TENSE 489 as signs to the present verbal participle, with such trivial changes only as euphony requires. The exceptions to this general rule are as follows : — (1.) In poetical Tamil the tenses are sometimes formed by suffixing the pronominal terminations to the relative participles, instead of the gerunds or verbal participles — e.g., nadanda{n)an (equivalent to the colloquial nadanda{v)an), he walked, literally a man who walked. In such instances a verbal or participial noun is used with the force of a verb. This is not an uncommon usage in other languages also ; and in colloquial Tamil the third person neuter of the verb, both singular and plural, is certainly a verbal noun in its origin, though used with the force of a verb — e.g., nadandadu, it walked, literally means a thing which walked ; and the plural, nadanda{n)a, means literally things which walked. A peculiarity of the poetical dialect is the extension of this usage to each person of the verb — e.g., nadanda{n)en, I walked, literally I who walked ; nadanda{n)am or nadanda(n)em, we walked, literally we who walked. This mode of forming the tenses has been developed from the Dravidian custom of using participial and verbal nouns as the conjugational bases of verbs, and, so far, is in accordance with the genius of the language ; but it has a constructive, artificial look, and it ig an exception to the mode which prevails throughout all the other dialects of the family, whether colloquial or classical. (2.) Tamil and Malayalam have, properly speaking, no present verbal participle, but only a particle denoting present time, which is suffixed to the theme of the verb, and to which, in Tamil, the pro- nominal signs are then suffixed for the purpose of forming the present tense. The combination, however, of the root and the particle of present time, forms virtually a present participle. I think it may, therefore, be assumed that the Tamil-Malayalam had a verbal participle of the present tense at a former period, which has now become obsolete, except in combination with the personal termina- tions, when it constitutes the present tense of the verb. (3.) In the ancient or classical dialect of Canarese there is another exception to the general rule. In the colloquial dialect tJie present tense is formed regularly from the present participle ; but the present tense in the classical dialect is altogether unconnected with that participle, or at le^t is only very distantly related to it. The sign of the present participle is ute, &c., whilst that of the present tense is dap — e.g., hdlute, living, bdldafem, I live. (4.) The Telugu usage of employing the substantive verb in a 490 THE VERB modified form (viz., unndnu, I am, unndvu, thou art, &c.) ay an auxiliary in the formation of the present tense, can scarcely be called an exception to the general rule specified above ; for this auxiliary is annexed to the present verbal participle, which is closely allied to that of Canarese ; and its use in this connection is only a refinement, not a necessary element in the formation of the present tense. These real or apparent exceptions being disposed of, it remains to inquire into the formation of the present verbal participles in the various dialects. Formation of the Present. — In both the classical and colloquial dialects of Canarese the verbal participle of the present tense is formed by adding to the verbal root a particle, of which ut is the most essential portion — e.g., coll. Can. bdl-uta, living ; 07id-utta, joining ; mdd-uttd, making : class. Can. 6d-ute, reading ; ili-(y)-uUe, descend- ing ; katt-uttu, tying ; geyutum {geyuttum, geyyutum), doing. The final vowel of this particle ut assumes various shapes, and is elided before the initial vowel of the pronominal signs in the formation of the present tense in the colloquial dialect {e.g., comp. mdd-utta, doing, with mdd-utfene, I do). It may, therefore, be concluded that it is simply enunciative ; and as u is the vowel most commonly used as a help to enunciation in all the dialects, the primitive shape of this particle must have been utu. I have no doubt that Mr Kittel is correct in identifying this utu with udu, the intermediate demonstra- tive pronoun of the Tamil and Canarese, used as a proximate demon- strative in Tulu. Another form of udu in classical Canarese is utam. utu, with the meaning of ' this,' would very naturally come to be used as a sign of present time in the formation of a participle of the present tense. It will be seen, in considering the preterite tense, that the d which constitutes the sign of past time is probably a relic of adu, the remote demonstrative ' that.' Probably the U7n of utum is the ordinary conjunctive um, used for the purpose of more distinctly emphasising the time. It is more difficult to explain the origin of the sign of present time used in the formation of the present tense in Old Canarese. The present tense in that dialect is not formed from the present participle. That participle is, as we have seen, substantially the same in both dialects ; and in the colloquial dialect the present tense is formed by affixing to this participle the personal terminations. The ancient dialect, on the other hand, makes no use of its present participle in forming its present tense, but forms that tense by inserting the THE PRESENT TENSE 49 1 particle daf between the verbal root and the pronominal fragments. The colloquial Canarese bdluttene, I live, is formed from the colloquial and classical present participle hdlutte ; but the corresponding form in classical Canarese is hdldafem, in which present time is represented by daf. What is the origin of this particle ? Mr Kittel (in a private communication with which he has favoured me) regards daf as being properly dafa, and dapa as consisting of da-\-a'pa. This afa he considers identical with alia, the future participle of ahu (in coll. Can. dgu), to become ; da he regards as the sign of the past tense. Hence mddi -}-da -{-apa -\-em (rmdidapem) would mean ' having made I shall be.' This form, therefore, was properly a second future. He traces its origin to the custom of replying to a command by an answer in the past tense — e.g., you say to some one. Come ! and the reply is, I came — i.e., I come. The fact that this form was originally a second future accounts, he thinks, for the introduction at length into the modern or colloquial dialect of a present tense distinctively denoting the present, being formed from the present participle in use in both dialects. This explanation is certainly very ingenious, and seems to me satisfactory. It will be shown further on that one of the forms of the present in Tamil makes use of a participle of the verb a {dgu), to become, and that most of the Dravidian presents were formed from futures. It will also be shown that the use of d, the ordinary sign of the preterite in all the dialects, was not originally restricted to that tense so absolutely as it is now. The present verbal participle of Telugu is ordinarily formed by adding chu (pronounced tsu) to the theme of the verb. In the colloquial dialect tu is used instead of chu ; and though it is possible that chu may be the original, and tu (from tsu) the corruption, yet it would be more in accordance with analogy to derive chu from tu ; and this tu so nearly resembles the Canarese uta or ute, that we may safely conclude both forms to have been originally identical. Probably also du, the particle which in most instances is inserted as a sign of tense between the verbal theme and the pronominal terminations of the Telugu aorist, springs from the same origin as tu, chunnu or tunnu, the ordinary termination of the participle of the present tense in grammatical Telugu, is a compound form derived from chu or tu, the real and only sign of present time in this language, and unnu, a participle of the* substantive verb undu, to be, used as an auxiliary. The Tulu participle of the present tense is also used for the future as well as for the present, and was probably a future originally. The 492 THE VERB sign of the present used in the present tense of the verb is v, which is identical with the Tamil-Canarese sign of the future. The sign of present time used by the Tamil and Malayajam, differs considerably from that of the Telugu-Canarese. The present tense in Tamil is formed by suffixing git-u, gindr-u, or d-nindr-u, to the verbal theme, to one or other of which particles the pronominal signs are annexed, d-nindr-u is a compound form, which is rarely used even by the poets, and is derived, I conceive, from a, the ultimate base of d-gu, to become (and which is not unfrequently used in this shape in the poets), and nindr-u, standing, abiding, continuing ; root, nil, to stand. The meaning of the compound seems to be continuing to become — e.g., tdrdnindrdn (tdr-d-nindr'-dn), he is low, he is humble, literally, he continues to become low or humble. Documentary evidence is forthcoming of the accuracy of my supposition that the a of dnindru was a representative of dgi. In an Old Tamil inscription (774 A.D.) in the possession of the Syrian Christians on the Malabar coast, I find dyi-nindru instead of the d-nindru which has been universally used in later times, dgi is often softened into dyi even in modern Tamil, then into dy, and then into d. The other particles of present time, git-u and gindr-u, are in more common use, especially the former — e.g., varu-git-dn, or varu-gindr'- dn, he comes. The only difference between them is that gindr-u is considered more euphonious and elegant than gix-u, and more suitable, in consequence, for poetry and elevated prose. I have no doubt that they are identical in origin, and that the one is either an euphonised or a corrupted form of the other. In some connections gii-u and gindr-u are changed by dialectic rules of euphony to kkii-u and Tckindr-u — viz., when they are attached to roots consisting of two short syllables (like fadu, to lie ; iru, to be ; nada, to walk), the final vowel of which is regarded as a part of the root, and is incapable of being elided. It is a rule of the language that if in such cases the sonants g, d, h, immediately follow, they shall be hardened, that is, converted into the corresponding surds h, t, and p ; and in Tamil the only method of hardening sonants is by doubling them, — for it has already been shown that in this language the same con- sonant is a sonant when single and a surd when doubled. Hence we say in Tamil not iru-gii-en, 1 am, but iru-kkii-en. A similar result follows in another and more numerous class of instances from a differ- ent cause. It has been shown in a former part of this section that transitive or active verbs are in many instances made to differ from intransitives by the hardening and doubling of the initial consonant THE PRESENT TENSE 493 of the sign of tense. In such cases git-u and gindr-u become (not for the sake of euphony merely, but as a means of grammatical expression) Ickit-u and kkindr-u. Malayalam uses the same sign of tense somewhat modified : the sign of present time in Malayalam is unnu or kkunnu, suffixed to the verbal theme. The older dialect of Malayalam has generally innu, especially in connection with the negative verb — e.g., vard-{y)-innu, comes not. Where Tamil would use gindru, Malayalam omits the g. When Tamil doubles the g and says kk, Malayalam uses kk also. The Malayalam innu is clearly a softened and euphonised form of the Tamil particle. The Tamil compound sound ndr is constantly con- verted into nn in Malayalam — e.g., ondru, Tam. one, is in Malayajam onn\ and mundru, Tam. three, is in Malayalam munn\ Even in vulgar colloquial Tamil the same or a similar tendency appears : ondru, one, being commonly pronounced onnu, and mundru, three, munu. The Tamil gindru and kkindru would, therefore, naturally and dialectically be converted in Malayalam to ginnu and kkinnu. The next point is the softening away of the g of ginnu. This has arisen from the circumstance that in Tamil g is pronounced in the middle of a word so softly as to be little more than an indistinct guttural breath- ing ; in consequence of which, it is used to represent the h of San- skrit, and in the colloquial dialect it is often discarded altogether — e.g., fogiten, I go, is commonly pronounced fo-ven ; and varugitdn, he comes, varu-tdn or vd-tdn. Hence ginnu (from gindru) would natur- ally become in Malayalam innu. The only remaining difference is be- tween the i of innu and the u of unnu ; but this presents no difficulty, for even in Tamil i is very often pronounced as u by the vulgar, and we have seen that in Malayalam also innu is more classical than unnu. The identity of the Malayalam sign of the present tense with that of Tamil, cannot be doubted. Sometimes in Malayalam poetry the pronominal signs are suffixed to the signs of tense, as in Tamil ; and in that connection the identity of the signs of tense is clearly apparent — e.g., compare adikkindrdn (adi-kkindr-dn), Tam. he beats, with the corresponding form in poetic Malayalam adikkunndn {adi- kkunn'-dn). A priori it might have been supposed that the Malaya- lam unnu or kkunnu was related to chunnu or tunnu, the sign of the present participle in Telugu. The resemblance, however, is alto- gether illusory ; for the* Malayalam particle is derived from the Tamil gindru or kkindru, whilst the Telugu chunnu is compounded of cJiu, the real sign of present time, and unnu, a participle of undu, to be ; which participle is in Malayajam und/. 494 THE VERB I have said that I believe the Tamil giv-u and gindr-u were identical in origin, and that the one is merely an euphonised or corrupted form of the other. Which is the original form ? and which the euphonised or corrupted ? There are many instances of r being euphonised in Tamil into 7idr — e.g., kundru, as a verb, ' to become small,' as a noun, ' a small hill,' must be a secondary form of kut-u, small, a form of the root which constitutes the basis of a large number of words, such as kuTTam, a fault. The change of ndr into r is not so easy, nor can I find any instances of it which are free from doubt. Still such a change may be suspected to have taken place in several instances, one of which is indru, now, to-day. A secondary form of indru in Tamil is iviei (pronounced ittrei), and this seems to point to il-tei. l-\-t, sometimes became ndr in Tamil, and sometimes in the poets shrinks into r. Thus sel-tal, the verbal noun of sel, to go, is changed to setal in the " Nannul," the Tamil classical grammar. In this case, how- ever, there is also a lengthening of the preceding vowel. If we may suppose il-tu to have become, on the one hand, indru, and on the other, perhaps at a later period, ivu, we arrive at the best explanation which has been given of gindru or gitu, the Tamil sign of present time. Dr Graul, I believe (in his " Outlines of Tamil Grammar "), was the first to suggest the origination of this sign from A;=^gr, a sign of the future in poetical Tamil, and indru, now. His idea appears to have been that Tamil was originally without a present tense, and that the present was a new secondary tense, formed from the future by the addition of a sign of present time, kindru was thus = k-indru (then kiiu). The same view seems to have been adopted, or independently arrived at, by Dr Gundert. The fact that the form of this particle retained in Malayalam is unnu (in older compositions often innu) would seem to prove that kindru, not kitu, was the form in use in Tamil prior to the final separation from Tamil of the Malayalam, and, there- fore, not only the more classical form in Tamil, as it is admitted to be, but also the more ancient. This fact, though it does not prove the derivation of kindru from k-indru, yet favours that supposition. The present tense is seldom used in Tamil poetry, and I have never found it in inscriptions, though the past and future and combinations of both abound. In the talk of the common people, though the pres- ent tense is freely used, yet the grammatical signs of the present, givu, &c., are generally omitted. They say vegudu, it burns, instead of vegu-{giv)-adu. It would seem, consequently, that the inflexional forms of the present tense are not very deeply rooted in the language. In the language of the Tudas the present and future seem to be THE PRETERITE TENSE 495 identical, and the sign of time seems to be k or g — e.g., pokeni, I go, pokemi, we go ; ershken (ersh-k-en), I am, ershkimi, we are {rsh for r). In the second and third persons the k seems to be softened into ch — e.g., ershchi, he or they are. In the language of the Kotas, p seems to replace k — e.g., hogape, I go, hogapenie, we go. In the third person, however, singular and plural, k asserts its place — e.g., hogako, he or they go. The Tuda A; of the first person and the Kota k of the third seem naturally to connect themselves, not only with the gu of the Old Tamil, but with the kum or gum of the Old Canarese aorist — e.g., Old Tamil sey-gu, I do or will do ; Old Canarese rmdu-gum (I, he, they, &c.), do. The f of the Kota present is evidently connected with dap, the Old Canarese sign of the present tense, but still more nearly related to the v, h, or p of the Tamil-Malayalam-Canarese future. In some Kota verbs k is the sign of the present tense, as in Tuda — e.g., vindkene (vind-k-ene), I ask, vindkeme, we ask. In some, both letters seem to be mixed, as in ettakepe, I raise up, I build, of which the past tense is ettape. 2. The Preterite Tense. — The mode in which a language forms its preterite constitutes one of the most distinctive features in its grammatical character, and one which materially contributes to the determination of the question of its relationship. In the Semitic languages past time, or the objective reality of past events, is denoted by placing the verbal theme first, and suffixing to it the sign of the personal agent. In the primitive Indo-European languages, the preterite appears to have been most commonly formed by means of the reduplication of the root or verbal theme ; but this reduplication has in many instances been so softened and euphonised, that it has dwindled into the mere use of a different vowel in the preterite from that which appears in the root. The Indo-European preterite was also frequently formed by means of a prefixed temporal augment ; a prefix which Bopp considers to be identical with ' alpha privative,' but which is supposed by Meyer to be identical with a, a relic of the auxiliary verb to have, which is still prefixed to verbs in the Celtic languages as a temporal augment — i.e., as a sign of past time. In a large proportion of the verbs in the modern Teutonic tongues, in the modern Persian, in the Turkish and Finnish families of languages, in the vernacular language* of Northern India, and, with a few excep- tions, in the Dra vidian languages, the preterite is formed by sufiixing to the verbal theme a particle, sometimes a consonant, sometimes a vowel, which is significant of past time. 496 THE VERB The Dravidian preterite tense is ordinarily formed, like the present, by annexing the pronominal signs to the preterite verbal participle. It is in that participle that the idea of past time resides : by it alone that idea is expressed. The changes that are made when the pro- nominal signs are added will be shown to be euphonic merely, not structural ; and in Malayalam (in which the pronominal signs have ceased to be annexed), that part of speech which corresponds to the Tamil preterite verbal participle expresses by itself the past tense of the verb. Consequently, an inquiry into the Dravidian preterite tense resolves itself into an inquiry into the formation of the preterite verbal participle. The preterite verbal participle is used in Tamil with a wider range of signification than in any other dialect, though its proper and inherent meaning is that of the preterite alone. Tamil, being destitute of a present verbal participle, uses the preterite verbal participle instead, in consequence of which, in a Tamil sentence, the question of time is in abeyance till it is determined by the tense of the final governing verb. This is more or less the case in all the dialects. Where there is a present participle as well as a preterite, the present is used to denote simultaneous actions, the preterite successive actions ; but it is the final verb which determines whether those actions, whether simultaneous or successive, belong to the present, the past, or the future. This indeterminateness of time in Tamil applies to the verbal participle alone, not also to the preterite tense of the finite verb, which is restricted in Tamil to the expression of past time, precisely as in other languages. We have now to inquire particularly into the Dravidian methods of forming the preterite. They divide themselves into two — (1.) by reduplication of the final consonant ; and (2.) by suffixing a sign of past time. 1. The Formation of the Preterite by Eeduplication op THE Final Consonant. — This mode of forming the preterite is adopted by a very small number of verbs in each of the Dravidian dialects ; but its existence cannot be doubted, and it is a mode which is as interesting as it is remarkable. In the Indo-European languages, when the preterite is formed by means of reduplication, it is the root which is doubled, or at least the first syllable of the root ; but in the Dravidian dialects the reduplication is that of the final consonant alone. The verbal themes which form their preterites in this manner are those which end in d-u, g-u, or r-u, preceded by a single short vowel — e.g., in classical Tamil pad-u, to suffer ; fug-u, THE PRETERITE TENSE 497 to enter ; and j)eT-u, to obtain — the preterites of which are patt-en I suffered ; 'pukk-en, I entered ; and pettr-en, I obtained. In each of the above examples the final consonants — d, g, and r — are doubled, and being thus doubled, are converted by rule into the corresponding surds tt, kk, and rr (pronounced ttr). Whilst the above and similar verbs form their preterites in this manner in the classical dialect of Tamil, in the modern colloquial dialect some of those very verbs have adopted the more ordinary method of denoting past time by means of a suffixed particle or consonant. Thus pukk-en, I entered, has been superseded in the modern dialect by pugu-nd-en, and nakk-en, I laughed, by nagei-tt-en. Canarese forms the preterites of this class of verbs in exact agreement with classical Tamil — e.g., nakk-anu, he laughed, from nag-u, to laugh ; and Telugu, though less systematic in this point, exhibits the operation of the same rule, especially in the relative participles of the preterite. This Dra vidian reduplica- tion differs materially in form from that of the Indo-European languages, but it appears to proceed from a similar principle, and it constitutes, so far as it goes, an interesting point of resemblance between the two families. 2. The Formation of the Preterite by Suffixing some Particle or Sign of Past Time. — This, with the exception of the very few verbs included in the previous class, is the method of form- ing the preterite which is invariably adopted by the Dravidian languages, and which may be regarded as their characteristic mode. For the purpose of thoroughly investigating this subject, it will be desirable to inquire into the practice of each dialect seriatifn. (1.) The Canarese Preterite. — The most characteristic Canarese pre- terite is formed by annexing d (euphonically d-u) to the verbal theme. This addition constitutes the preterite verbal participle — e.g., ili-d-u, having descended, nudi-d-u, having spoken ; to which the pronominal terminations are suffixed to form the preterite tense — e.g., ili-d-enu, I descended, nudi-d-i, thou saidst. All verbal themes (both in the classical and in the colloquial dialect, and whether transitive or intransitive) which end in i or e, form their preterites in this manner, together with many themes ending in u. All the apparent irregu- larities that exist are merely modifications of the d in question. Thus, sometimes t is substituted for d — e.g., aritanu, he knew, instead of aridanu (corresponding»to the Tamil atinddn) ; sometimes the d of the preterite combines with the final consonant of the root, and converts it into dd or tt — e.g., iddanu, he was, instead of irudanu (Tam. irunddn) ; eddu, having risen, instead of eludu (Tam. erundu) ; 2 K 49^ ' THE VERB uttu, having ploughed, instead of uludu (Tarn, urudu) ; nintu, having stood, instead of niludu (Tam. nindru). Another Canarese preterite is formed by suffixing i to the crude verbal theme — e.g., mdd-i, having done, from mdd-u, to do. Between this i and the pronominal terminations, d is inserted in the formation of the preterite tense — e.g., mdd-i- {d)-enu, I did, bdl-i-(d)-anu, he lived. This mode of forming the preterite characterises most verbs ending in u in the modern dialect. The final u of such verbs is merely- euphonic, not radical, and is elided on i being annexed ; and the d which is inserted between i and the pronominal signs, though possibly identical in origin with the d which constitutes a sign of the preterite, is merely euphonic in so far as the use to which it is now put is concerned. In a considerable number of instances the formation of the preterite in i appears to be a modern corruption. Intransitive verbal themes ending in u form their preterite in d in the classical dialect ; and it is in the colloquial dialect alone that i forms their preterite — e.g., instead of bdl-i (coll.), having lived, the classical dialect has hdl-d-u ; and as the classical dialect is undoubtedly more authoritative and probably also more ancient than the colloquial, d or d-u may be considered as the legitimate form of the preterite of this class of verbs. This conclusion is confirmed by the analogy of Tamil, in which the corresponding verbal theme forms its preterite verbal participle by suffixing nd, an euphonised form of d — e.g., vdr-nd-u, having flourished, which is the equivalent, not of the modern Canarese hdl-i, but of the ancient hdl-d-u. How is this diversity in the formation of the preterite to be accounted for ? Can i have been derived in any mannei* from d ? An argument in favour of this supposition may be deduced from the circumstance that the classical bdl-d-en, 1 lived, which is in perfect dialectic agreement with the Tamil vdr-nd-en, has in the colloquial dialect become hdl-i-d-enu. Even in the ancient dialect itself, though this i is generally unknown, it makes its appearance in the preterite relative participle, which may be hdl-i-d-a, that lived, as well as hdl- d-a, though the corresponding Tamil is always vdr-nd-a. If we could form a judgment, therefore, from these instances alone, i would seem to have come into existence as a vocalic bond of connection between the root and the sign of the preterite. The future, both in Canarese and in Tamil, often makes use of u as a bond of union between the verbal root and v, the sign of tense — e.g., hdl-u-v-enu, coll. Can., and vdr-u-v-en, coll. Tam. I shall live, THE PRETERITE TENSE 499 instead of the ancient and more correct bdl-v-cn, Can., and vdr-v-en, Tarn. In this case the u is certainly euphonic, though it has not come to be used, as i has, to express grammatical relation, or in lieu of the sign of tense which it is employed to euphonise. If we had to account for the insertion of i before d in such instances only as have been mentioned, we might be content with the supposition of its euphonic origin ; but the use of i as a sign of the preterite has a much wider range. All transitive verbs ending in u, both in the classical dialect of Canarese and in the colloquial, form their preterite verbal participles by suffixing i ; and there is nothing to show that those verbs ever formed their preterites in any other manner. A very large number of verbs of this class form their preterites in Tamil also by suffixing i ; and in Telugu the preterite is formed by suffixing i to the root, not of one class of verbs only, but of all, with the exception of the small class of reduplicative verbs. This statement applies, it is true, to the preterite verbal participle of Telugu, not to the preterite tense of the verb, which generally suffixes or inserts, as a tense-sign, some additional consonant or par- ticle ; but in Malayalam the preterite verbal participle constitutes by itself the preterite tense, without the addition of any pronominal signs ; and in that dialect i is the only sign of past time which is used by a large number of verbs. Thus fddi, which means having sung in the other dialects, signifies in Malayalam (he, she, or it) sang ; i is, therefore, in that dialect a distinctive sign of the preterite in the class of verbs referred to ; and it is to be remembered that the addition of the pronominal terminations, though the means of expressing personality, effects no change in the means whereby time is expressed. The extent and prevalence, therefore, of the use of i as a sign of the preterite seems to forbid our supposing it to have been in all cases derived from an euphonisation of d ; and as d, on the other hand, cannot have been derived from i, it appears probable that d and i are distinct and independent signs of past time. Of these two signs of past time d is to be considered, if not the older, yet at least the more prevalent and more characteristic. We have seen that in many instances in which the colloquial Canarese has i, the classical dialect and Tamil have d. Not in those instances only, but universally, Telugu uses i as the sign of the preterite ; but the great antiquity of the grammatical forms of Tamil and Old Canarese precludes the supposition that their most characteristic sign of past time has been borrowed from that of Telugu. In addition to which, it will be shown that in Telugu itself there are traces of the existence 500 THE VEKB of an old sign of the preterite agreeing with that of Tamil and classical Canarese. It would, therefore, appear that two modes of forming the preterite being in existence, one in d, another in i, the latter form has in many instances, particularly in Telugu, superseded the former ; and the prevalence of i in Telugu and Gond would .seem to prove that this form must be one of great antiquity. In the Indo-European family of languages we find similar inter- changes amongst the signs of past time ; and though in some instances one form or mode may have been derived from another, yet this cannot have been the case uniformly — e.g., the iveah Germanic conjugations cannot have been corrupted from the strong, or vice versa ; though it seems certain that the strong method of forming the preterite was more ancient than the weak, and though it is also certain that the former mode has in very many instances been super- seded by the latter. What is the origin of the d which is inserted in Canarese between i and the pronominal terminations, and also between i and the sign of the relative participle ? It appears to be used (whatever be its origin) merely for the purpose of preventing hiatus between con- current vowels — e.g., mddi-(d)-enu, I did, mdd-i-(d)-a, that did. Hiatus is generally prevented in the Dravidian languages by the insertion of a nasal, or of one of the semi- vowels y and v ; and it seems extraordinary that d should be used for this purpose. It is true that in some of the inflexions of Canarese nouns — e.g., mara-d-a, of a tree, d might seem to be used euphonically ; but it has been shown in the section on " The Noun " that xhat d is the remnant of a neuter demonstrative, and is used as an inflexional increment ; it is not, therefore, a precedent for the use of d for the prevention of hiatus merely. Possibly the use of this d by the Canarese verb may thus be accounted for : a consonant for preventing hiatus between the sign of the preterite and the subsequent signs of personality and relation being required, Canarese preferred using for this purpose a sign of the preterite which still survived. Thus d was not a new invention, but an old particle used for a new purpose, and placed in a position in which it would not have appeared but for the use to which it had already been put. (2.) The Tamil Preterite. — The preterite is ordinarily formed in Tamil, as in Canarese, in two ways — viz., by suffixing either d or i to the verbal theme. In the former case, d itself is more rarely used than some euphonisation of it or related consonant ; but such second- ary forms invariably resolve themselves into d. Thus, when a THE PRETERITE TENSE 5OI theme with I as its final letter is followed by d as the sign of the preterite, the compound becomes ndr — e.g., the preterite verbal participle of fol, like, is not 'pol-d-u, but fon-dr-u. Sometimes, however, when d follows I, the compound becomes rr, pronounced ttr — e.g., from kal, to learn, comes, not kal-d-u, but kavt-u (kattr-u), having learned (Can. kali-d-u). I followed by d becomes nd — e.g., from mdl, to die, comes mdnd-u, having died. Sometimes, however, when d follows I, the compound becomes tt — e.g., from kel, to hear, comes kett-u, having heard. These and similar combinations are merely instances of euphonisation, in accordance with the fixed phonetic rules of the language ; and in each case it is in reality d alone which constitutes the sign of past time. In some verbs the primitive d still remains unchanged and pure — e.g., uru-d-u, having ploughed, from uru, to plough ; or with a conversion of the dental d into the cerebral d — e.g., kan-d-u, having seen, from kdn, to see. The euphonisation of d which occurs most frequently, and is most characteristic of Tamil, is its conversion into nd. This conversion takes place without phonetic necessity, and solely through that fond- ness for nasalisation which is so deeply inherent in Tamil and Telugu, especially in Tamil, and by means of which the formatives gu, du, and bu have so generally been changed to ngu, ndu, and rribu. In the majority of cases in Tamil in which d (preceded by a vowel or semi- vowel) once formed the sign of the preterite, it has been nasal- ised into nd ; whilst Canarese, wherever it has preserved the primi- tive d, has preserved it unnasalised and pure. Thus whilst the Tamil preterite of iru, to be, is iru-nd-en, I was, the corresponding Canarese is iddenu (for iru-d-enu) ; and whilst the preterite of the Tamil verb vdr, to flourish, is vdr-nd-dn, he flourished, the equivalent in classical Canarese is hdl-d-am. The higher dialect of Tamil retains some traces of the primitive unnasalised purity of this sign of the preterite — e.g., viru-nd-u, having fallen, from viru, to fall, is occasion- ally written by the poets vir-d-u. {vir is phonetically equivalent to viru.) It is curious to notice the progress of nasalisation which is apparent in this verb on comparing the Canarese biddu (for bil-du), the High Tamil virdu, the modern Tamil virundu, and the Malayalam vinu. Another change which d undergoes in Tamil consists in its being hardened and doubled m certain cases, so as to become it. This happens to nd as well as to d, — a clear proof of the development of the former from the latter ; and when the d of nd is doubled, the nasal entirely disappears. Just as the doubled form of iig is kk, and 502 THE VERfi^ that of mb, pp, so the doubled form of nd is tt. In some instances this change is merely euphonic — e.g., padu, to lie, an intransitive verb, takes for its preterite, not padu-d-en or padu-nd-en, but padu- tt-en, I lay. Such cases, however, are rare, and in general the use of tt as a sign of the preterite instead of d or nd, is a means of dis- tinguishing transitives or active verbs from intransitive — e.g., the tt of tdr-tt-en, I lowered, is formed by the doubling and hardening of the nd (the equivalent of d) of the corresponding intransitive tdr-nd-en, I became low. See the furtl^er explanation of this subject under the head of " The Classification of Verbs." The second mode of forming the preterite in Tamil, as in Canarese, is by suffixing i to the verbal theme. The themes which form their preterite in this manner are those which terminate in u euphonic, and of which the radical portion consists either in one long syllable or in two syllables, whether short or long. In this connecfcion, as in prosody, a vowel which is long by position is equivalent to one which is naturally long. The following are examples of the classes of verbs which take i for their preterite : — (long syllable) pddu, to sing ; (long by position) pann-u, to make ; (two short syllables) erud-u, to write ; (one syllable short, and one long by position) tirupp-u, to turn. All verbs of which the final consonant is a liquid semi-vowel {I, I, r, r, not V or r), whatever number of syllables they may contain, form their preterite by means of d or some of its modifications : such verbs are therefore exceptions to the above rule. Even in the class of Tamil verbs which take i as their preterite suffix, there are traces of the prevalence of c? at a more ancient period. Thus, whilst ' thou didst go ' is in the ordinary dialect p6-{n)-dy (properly p6g-i-(n)-dy, from p6, or p6-gu, to go), in the poets p6-d-i is sometimes used instead ; so instead of d-{n)-dy (for dg-i-{n)-dy, from d-gu, to become), thou becamest, the poets sometimes use d-d-i. In these instances Canarese also, even in the colloquial dialect, says podi and ddi. Even nd is sometimes d only in Tamil poetry — e.g., varu-d-i, thou camest, is found instead of the more modern va-nd-dy (for varu-nd-dy) ; and it is evident that this form, varu-d~i, exactly corresponds to the forms quoted above, p6-d-i and d-d-i. Notwithstanding, therefore, the prevalence of * as a sign of the preterite in Tamil, as in Canarese (though in a less degree than in Canarese), there seems to be some reason for regarding it as an inno- vation, or at least as a less characteristic and less widely used sign than d. n is inserted in Tamil (as d in Canarese) between the i which constitutes the sign of the preterite of certain classes of verbs and the THE PRETERITE TENSE 503 pronominal terminations, and also between the sign of the preterite and the sign of the relative participle — e.g., from fdd-i, having sung (the preterite verbal participle of pdd-u, to sing), is formed 2)dd-i-{n)- en, I sang ; fdd-i-(n)-dy, thou didst sing ; fdd-i-(n)-dn, he sang : so also 'pdd-i-{n)-a, the relative participle, that sang. Whatever be the origin of this n, it cannot be doubted that its use in Tamil is at present wholly euphonic ; and this statement applies also to the use of the same n in the preterite relative participle of Telugu. It, in no respect, contributes to the expression of grammatical relation ; and when used by the relative participle in Tamil, it may optionally and elegantly be changed into y, which is one of the semi- vowels that are systematically used for the prevention of hiatus — e.g., instead of 'pddi{n)a, that sang, we may write with still more perfect propriety fddi{y)a. Probably y is in this connection older than n. (See " Sounds.") We see a parallel use of n in the Turkish verb, in the frequent insertion of an euphonic n between the theme and the infinitival particle, and also between the theme and the sign of the passive. The most weighty argument in confirmation of the euphonic origin of the Tamilian n in question is derived from the use of n as an euphonic fulcrum, or means of preventing hiatus in the Dra vidian languages generally, and even in connection with another part of the Tamil verb. Thus, in the classical plural neuter of the present tense, varugindrana (varu-gindr-ana), they (things) come, the n of the pro- nominal termination ana is undoubtedly equivalent to the v of the isolated plural neuter avei (for ava) ; and is used merely for the euphonic prevention of hiatus between the first a, or the demon- strative vowel, and the final a, or the sign of the neuter plural. {a{n)a or a{v)a is equivalent to a-a.) Native Tamil grammarians consider in, not i, the sign of the preterite ; but as i, never m, is the form used by the preterite verbal participle, it is evident that they have given too important a place to what is at present at least a merely euphonic letter. If Tamil and Telugu alone were concerned, we should perhaps be justified in considering the purely euphonic origin of the n in question to be a settled point ; but a difficulty arises on comparing those lan- guages with Canarese. Wherever Tamil and Telugu use n in the formation of the preterite tense and the preterite relative participle, there Canarese, as hasjbeen observed, uses d — e.g., mddi-{dyenu, I did, not mddi-(n)-enu ; and mddi-(d)-a, that did, not mddi-{n)-a. Now, though this d of the Canarese is certainly euphonic in its present use, it has been shown that there is reason for suspecting it 504 THE VERB to be derived from d, the old sign of the preterite ; and if this sup- position be correct, it would follow that the Tamilian n, which corre- sponds so perfectly to the Canarese d, may be derived from the same source as d, and euphonically altered from it. The n of the Tamil preterite, therefore, as well as the d of the Canarese, may testify to the primitive universality of the use of c? as a sign of past time. Whether d {==n) was originally a sign of the preterite or not, the conversion of d into n in this connection, viz., in the preterite tense, and especially in the preterite relative participle, is analogous to the change of ta or da to na in the past participle of the Indo-European tongues, especially in German, from which the final n of our own past participles (such as ' fallen ') has been derived. (3.) The Malaydlam Preterite. — The Malayalam preterite is sub- stantially the same as the Tamil ; the only real difference consists in the disuse in Malayalam of the pronominal terminations. The sign of past time is invariably the same in each Dra vidian language, with only such modifications of sound as are dialectic and regular. That which constitutes the preterite verbal participle in Tamil is in Malayalam the preterite tense of the verb — e.g., nadandu in Tamil signifies having walked ; the corresponding Malayalam word nadannu, means (he, she, it, or they) walked. Some confusion has been introduced in Malayalam books by writing the preterite verbal participle nadanna, having walked, as if it were identical with the preterite relative participle nadanna, that walked. The rendering of the sound of the latter word is correct, the final a being the sign of the relative participle in all the Dravidian languages, and, as I con- ceive, identical in origin with a, the sign of the genitive, nadanna, that walked, is therefore identical with the Tamil nadanda. On the other hand, the final a of the preterite verbal participle ought either to have been u, corresponding to the Tamil nadandu, having walked, or, being a very short vowel, merely enunciative and euphonic, it should have been elided (as it is when followed by another vowel), after the fashion employed in North Malabar, in which this word is written nadann\ In Dr Gundert's Malayalam Grammar and Dictionary, the short u is denoted by m, in accordance with Lepsius's system of transliteration. This mode of rendering the letter has also been adopted in Brigel's " Grammar of the Tulu," in which language the short enunciative u has acquired a very prominent place. It is to be hoped, therefore, that this blemish in Malayalam orthography, as Dr Gundert terms it, will now disappear. (4.) The Telugu Preterite. — In Telugu all preterite verbal parti- THE PKETERITE TENSE 505 ciples, without exception, are formed by adding i to the theme. Even those verbs which form their preterites by suffixing d or some modification of it in Tamil, Canarese, and Malayalam, form their preterites in Teiugu by suffixing i — e.g., kon-du, Tarn, and Can., having bought, is in Teiugu kon-i, and kan-du, Tam., and Can., having seen, is kan-i. Notwithstanding the universality of this rule, there are traces even in Teiugu of the use of a particle corresponding to the d of the other dialects as a sign of past time. Though the preterite verbal participle never takes any suffix but that of i, some parts of the preterite tense of the verb in the higher idiom of the language (viz., the first and second persons both singular and plural) insert the particle ti between the i of the verbal participle and the pronominal terminations. It cannot be doubted, I think, that this ti, which is found nowhere but in the preterite, is allied to the d which is inserted in the same place in the Canarese preterite. Thus, whilst both in Canarese and in Teiugu the preterite verbal participle of dd-u, to play, is dd-i, having played, in both dialects ti or d is suffixed to i before adding the personal terminations — e.g., compare Can. dd-i-d- enu, I played, Tel. dd-i-ti-ni. It has already been shown to be probable that the d thus inserted by the Canarese, though now used to so large an extent euphonically, was originally a sign of the preter- ite, identical with the d which is still used for that purpose by many verbs. This view derives confirmation from Teiugu, in which the corresponding ti does not appear to be used euphonically at all, and certainly is not used for the prevention of hiatus ; for there is no hiatus and no necessity for an euphonic insertion between the afore- said ddi and ni, the pronominal fragment, or in the second person between ddi and vi. It therefore follows that we must regard ti as a sign of past time, subordinate indeed to i, and unused in the third person of the preterite, but immediately alHed to d, the past tense-sign of Tamil and Canarese, and testifying to the existence of a time when d, or its equivalent t, was one of the signs of the preterite in Teiugu as in the other dialects. In some Teiugu verbs, ti is combined in such a manner with the final consonant of the theme, as to prove beyond doubt its identity in origin and force with the Tamil d — e.g., cMs-ti- ni, Tel. I did (for chesi-ti-ni), is evidently equivalent to the Tamil sey-d-en ; and kon-ti-ni, I bought (for koni-ti-ni), is equivalent to kon-d-en. So also when ejithe Teiugu conditional particle, answering to the Tamil dl, is suffixed to the preterite tense of a verb for the purpose of giving to it the meaning of the subjunctive, it appears evident that the ancient sign of the preterite of the Teiugu must 506 THE VERB have been, not i, but ti or t — e.g., compare the Telugu chest-e, if (I, thou, he, &c.) did or do (abbreviated from chesi-t-e), with the Tamil Uyd-dl. It may be mentioned as a singular coincidence that in Mon- golian the gerund du has been modernised into ju, and that again has been changed colloquially into ji. We have seen that Tamil inserts n between the preterite verbal participle and the pronominal terminations in many instances in which d is used for this purpose in Canarese. The colloquial dialect of Telugu makes much use of na in the same connection — e.g.^ dd-i- (n)-dnu, I played (answering to the Tamil dd-i-{n)-en), instead of the more elegant and probably more ancient dd-i-ti-ni. Compare ay-i-{n)-dnu, Tel. I became, d-{n)-en, Tam. (for dg-i-{n)-en), and d-{d)-enu, Can. (for dg-i-(d)-enu) . On the whole, it may be concluded that the Telugu agrees with the other dialects in exhibiting distinct and deep-seated traces of the ancient use of cZ or i as a sign of the preterite, notwithstanding the universal prevalence in Telugu at present of the use of i, as the sign of the preterite verbal participle. I may here take occasion to guard against an illusory resemblance to which my attention was once called, viz., the resemblance which subsists between the Telugu preterite verbal participle veichi, having placed, and the corresponding Tamil participle veittu, which is vulgarly pronounced veichi. The tt of the Tamil vei-tt-u, being simply the hardened and doubled form of d, is the ordinary sign of the preterite ; and if there were any real alliance between tt-u, through its provincial pronunciation, and the Telugu ch-i, we should un- doubtedly have here an instance of the use of tt — i.e., of d — in modern Telugu as well as in Tamil, as a sign of the preterite verbal participle, and consequently of past time. The resemblance, how- ever, is illusory. The ch of the Telugu veicJii corresponds, not to the tt of the Tamil veittu, but to the Jck which constitutes the formative of so many verbs and nouns in Tamil, kk makes its appearance in the infinitive of this very verb, viz., vei-kk-a, to place, the Telugu of which is vei-ch-a. kk is vulgarly pronounced ch in the southern part of the Tamil country, and the same pronunciation universally obtains in Telugu. The imperative or theme of this verb in Telugu is not vei, as in Tamil, but veich-u (with the addition to vei of the formative ch-u, which is equivalent to the Tamil kk-u) ; and from this veich-u, the preterite verbal participle veich-i is regularly formed, in this as in all other cases, by the addition of i. If the corresponding Tamil verb formed its preterite in the same manner, its verbal participle would be vei-kk-i, not vei-it-u. A case in point in illustration of this is the THE PRETEEITE TENSE 507 Tamil tu-kk-u, to lift, to weigh (Tel. tu-ch-u), the preterite verbal participle of which is tu-kk-i (Tel. tu-ch-i). (5.) TJie Tulu Preterite. — The Tulu preterite, like that of Gond, divides itself into two tenses, an imperfect and a perfect, each regu larly inflected. The tm'perfect tense is that which corresponds to the ordinary preterite of the other dialects, and is formed in substantially the same manner by suffixing to the root either the ordinary Dra vi- dian t or d, or the i, which is still more commonly used in several dialects. Compare Tulu itte, I was, with iddenu, Can. ; irunden, Tam. : Tulu kende {ken' for kel) with ketten (kel-ten), Tam. : kelidenu, Can. i appears in buriye, I fell, from burn, to fall (Tam. vim, vir). The perfect tense seems to be formed by suffixing an additional d, with such euphonic changes as the dialect requires. Compare itte, I was, with itf de, I have been. (6.) Preterites of Minor Dialects. — It is difficult to make out the Tuda preterite, th appears to be the sign of the past, corresponding to the Tamil and Canarese d — e.g., compare dd-k-en, I dance, with dd-th-b-ini, I danced. This th is written cJi by Mr Metz — e.g., bindch- pini, I asked ; and, according to him, the same ch appears alike in the present and the past, in each person except the first. Dr Pope inserts th before ch in the past — e.g., dd-th-chi, danced. In the Kota dialect the past seems to be represented by si — e.g., compare hogape, I go, with hosipe, I went. In this it does not stand alone, as will be seen. In Gond, si or ji, apparently softened from ti, forms the verbal participle of the preterite ; but the perfect tense is formed by sufiixing tt — e.g., kei-tt-dn, I have called ; kei-si, having called. In Seoni Gond, also, the preterite or conjunctive participle suffixes si — e.g., wunk-si, having spoken ; but the past participle is formed by suffixing tur — e.g., wunk-tur, spoken ; and the past tense simply suffixes t — e.g., wunk-t-an, I spoke, wunk-t-i, thou didst speak. An imperfect or progressive tense is formed by inserting und or nd, apparently the substantive verb, between the root and the pronominal terminations. These instances tend to confirm the supposition that d, or some modification of it, is, if not the only, yet at least the most ancient and characteristic sign of the Dra vidian preterite. Origin of the*Dravidian Signs of Past Time. 1. The most probable conjecture I can offer respecting the origin of i, is one which would confirm the supposition of its secondary 508 THE VERB character. I conceive it to have been originally a vowel of con- junction, employed for the purpose of euphonically connecting the verbal theme and the true sign of past time, d or d-u. Where the theme terminated in a hard consonant, euphony would require some such vocalic bond of connection — e.g., the Old Canarese hdl-d-en, I lived, is undoubtedly somewhat harsh to an ear that is attuned toDra- vidian phonetics ; and it was natural that it should be softened, as it has been in modern Canarese, into hdl-i-d-enu. We see a precisely similar euphonic insertion of i in the Latin dom-i-tus (instead of dom- tusy, tamed, and the Sanskrit pid-i-tah (instead of pid-tah), pressed. Subsequently we may suppose the true preterite d to have gradually dropped off ; whilst i remained, as being the easier sound, with the adventitious signification of the preterite. There are many instances in all languages of euphonic additions coming to be used instead of the parts of speech to which they were attached — e.g., in the Telugu verb, vu is used to represent the second person singular of the pro- noun instead of nt, thou, though vu was originally only an euphonic addition to ni, by which it was converted into ntvu. It deserves notice that wherever i is used in Canarese or in Tamil, instead of d, as a sign of the preterite, the use of d would in that instance be harsh and uncouth ; and that on comparing the Tamil verbs which form their preterite in i with those that suffix d, no reason but euphony can be alleged why the one suffix should be employed rather than the other ; consequently euphonic causes must at least have helped the development of i. This supposition of the origin of i from the vocalic conjunction of d with the verbal theme, would also account for the circumstance that wherever i is followed by a vowel (whether the initial vowel of the pronominal terminations, or the a which constitutes the sign of the relative participle) it picks up again the d which it had gradually lost, and uses it as an euphonic bond of conjunction, either in its original shape of d, as in Canarese, or in its nasalised shape of n, as in Tamil and Telugu. The manner in which ti is separated from the theme in some Telugu preterites — e.g., hon-i-ti-ni {kon-ti-ni), I bought, confirms this supposition of the euphonic origin of i. 2. d, the more characteristic sign of the Dra vidian preterite, presents many interesting resemblances to corresponding signs of past time in various Indo-European and Scythian languages. It may have an ulterior, though remote, connection with t or ta (alternating with na), the ordinary suffix of the Indo-European pas- sive participle- — e.g.,jnd-ta-h, Sans, known ; Greek yi'0)T6-u tti ndu ttu ndei du dei foU-^ei, interest. fada-di, chaff. kata-ndi, a spoon, a trowel. kuia-du, pincers. kuru-du, blindness. uru-di, strength. ondi, a lizard. paru-tti, cotton. paru-du, defect. maru-ndu, medicine. kuru-ttu, young shoot of palm. iru-dei, a lie. kura-ndei, an infant. poll, to increase. =padar, chaff, the same. =kaTa-riei, the same. kuta-ndu,to be crooked; from huT-u, short. kur-u, tender. ur-u, to be strong. = odi, the same. par-u, to expand. par-u, old. mar-u, sweet-smelling, kuru, tender. ir-u, to swerve. kur-a, young. [Euphonic changes of the formatives dei, after consonants.] i, du, and kdt-chi (kdn-di), a spectacle. ter-chchi {ter-di), intelligence ural-chchi{ural-d'i),a,wh.u:\ing nan-dri (nal-di), a benefit. vet-ii (vel-di), victory. pugar-chchi{pugar-di),i^ia>ise. dt-chi (dl-di), possession. dt-ti (dl-di), a woman. Um-du (tol-du), ton-dru {tol-du), pet-tei (pen-dei), a hen. parat-tei (parand-dei,) shag gin^ss. ton-dei (tol-dei), the throat. ut-tei (un-dei), uncleanness. antiquity. kdn, to see. ter, to ascertain. ural, to whirl. nal, good. vel, to conquer. pugar, to praise. dl, to possess. dl, a person. ' tol, old. pen, female. parandu, to scratch. tol, to perforate. un, flesh. 550 THE VERB Formative. Noun. Root. n Jcad-an, debt {=kad-am). kad-u, harsh ? kad-a, to pass over. at-an, virtue (=ar-am). ar-u, to cut, to define. hu m^ra-hu, usage. m 0^ S 5§ 5gv i ^ ^ 1-S16 ^ ri ,v . 5S ?> 5> 1. ^ ^ f oiS ■^ -S. i 1 i I, ^. ?3 «> ?Si ^ ?^ ^ 5f » ,:> jg 5sj JSj -^ f^ 4 a. o o > ?i. 5is i N>3 ■I 1 N>3 5f^ v>5 r5^ ^ •i .s I 5S5 ^ CSi &s •I ^ V>3 C3:> I? xc ^ ,j >. .^ J, ' >^ 4^ tn CO 4 • (V 4-S -§ $ 6 f ^ ^. I i I O I p^ '=^ < § > H Ph o o 5S I t ??5 I I ^ >i ?Si ^ O ^ <» EH < p^ o'S^ O • ^H t-( J-) r— t t\i M a-' ^4-t f^ in ^-^ (73 O I? _C e ^*-< ^ c/-i y^ >*# o -1^ ^ c/j o ft ^ p o o P3 i .1 4 -I >5 S i ?»5 -a 8 -a -« ai ai 6^ ^ to^ I i CO s: . S to 'S .^ ^s •c^ f< 1^ f^ CQ ^ ^ rS -^^ '^ lit n3 to --d i 1. ^00 ^ ^ <4i ■^ vS^ '^ i NV3 vco 1, ^00 i <4i MX) i e ^ 1 * ^ ;si 5S1 ■^ v^ • S^ ^co ^e ^ ^ 5?:, - M s ^^§^ 1*^ "^i 5?i c3^ ?5l ^ ^ ^ N>3 • H P^ P4 c- r-i >-• >H ^ =5 ■« CO "T? t>^ ID =^ S ?i5 rC3 a) o S g ft O o ?^ ?»i g . ^ ^ 5^ 5g 53 53 5> 5i 5^ 5s>, c5i JSj ?S5 JSj <» ^ ?3i ^ Crs ?55 Crs S . S 5> . S 5g • § • '^ * g Jg g ^ 5*5 JSj 5s:. .^ ^^ r(4j 8 s 5fi 55^ S 5s^ I? >5 o o :3 o H ^ 6 o rrrj --d 1— 1 ^ . 1-4 ^ o ^ <1> 4^ 1— 1 ^ § »4i , kejfnd present; i II - r—^ — V i o 5r5 P-i ^^ ?35 I 5= !3 <» -co -co <» >» -CO -CO I i i i « H PR H I. I I? o ^ _, 6 «= s: r? § i^ -^ -^ ^ — , — ' ^ ^ 6 02 =} ^1 4-= o ^ ^ I, ^ ... -^^ ' ^ '\ s^ ^ . I ,>^ :^ ^^ .Is ^ ^^ >5 Ss >^ *- X5 '1 55i 5*5 ^ ^ c*:, <;i) (^ !» f< r< r< <;j o o g 5s^ sS ^. "■ ^s ■^^ ^cs ■>:© ' 55^ ?&5 ^ Sh 5s^ 5^ 5?^ Sr> «» <;i ^■> <» N>5 vco ^co vcc ' 4^ 2 o ti r^ nS -p ^ § O O O ■+^ ' — ! O ^ rt ^ J3 O o y r^ O rd M >—i H rjl \ i ,— « — , >:, ^ !S^ ^ 5S ?> f^ !S p-l ^ ^ ^ ^ 55^ :^ 55^ JSi ;^ i 5S5 -^ -^ O W ' > H m zo <{ ?H ci a s 1 H ^ ■ — ' ^— ^ ^e ^^ ^^ ^53 >i ;^ ^ S^ s^ ^ 5^ ^ <» •^ <4i o lij ^ S g ^^ ^i j&j •5^ ^^ § ^ %^ S^ ^^j ■^o ?5^ '■^r^ ^e ?5^ ^ 5s^ ;5^ ?5i 5Si Ss^ ■^ <» -03 -co vvo -cc -+^ 6 nii -, K^ ^ -P rt nj r^ H r-H 1^ g^ i° ^ 8i 4 !> ^;<;i -§ I? ^1, ■^ ■« -s 4 i i e e ?3 ?*:) ?S5 55:) ^^ ^4) '^'i) r< r< r« 5s^ ^ ^ -a 5?^ 03 2 o I. 'is ?Sl QJ ^^ 5S I ^1 o o O '-^ ?si «§£ CO ■^!S ■^e P. ^ ^ •^^ JS Jg ?s » V ^ i; 1.1 o o o ^ -2 -P jj o . <^ o q ^ o fl, S a -a o §^^ -^ ^ .3 ^ Z t^ ^^. a. H g'^ 1. O O o +3 +5 O -s O rt ^ T-! n3 !=! ^ '-C O r^ w rS^ H w. ^ :^ ^- ^ .• o o A 02 0? ce o3 y -^ . ^^"^ . 4^ S -t-> S 4^ >i a S^ ^ ^ ^ ^^ ^ ^ - 9^ fl >5 ^ Ej P^ g -^ -^ ^ I ^^^ >■- ^ II O a «i5 e <;3 '^* ^ 5si 5si >i ^ 5S e ?3 e ?s e ^Sh ?iH ?iH ?5h ?^ 1,1 5i. aa a % " 1 ^^ ^-^ 1 6 02 a . w s S -e § ^ «§£ Jsi Jg «§ «§ ?S *§ »g ^e 555 ?5^ 5S5 "i^ ^ r< ^ M P4 g g IS ^_^ t 1, '"SS ;^ t I N>0 O '7J o •73 I I 000000 r^ i-d r^ r^ -T^ n3 ■73 b 1 +3 ■73 -73 H .• ^1 ^ •rf^ ?i fl ^ a» PI r^ ^ ^ & H 6 .0 6 ' 7=i nc^ rt if O ^ - 40 4^ 2 fl -^ (_> (-H ^ w ^ 4^ ^ ^ ^ '% ^ F| ^ -^ £' -§ .^ .^ 2 o P3 ^ -^ ° § as, 4f O SI U hi ^ r— I rri .2 s .2 ^ 1 I ■§ I I w Q 2 ft 1^ <1 PQ O > >; H ■ J^ p^ iz; P • . • • iJ 1— 1 W H H 3 P^ — ~ S 4^ > 4-3 +3 cn CQ OQ ■§ §^ •§ ■g 1— • < ^ 1 § <1 bb «si <1 bb '8 ^e 1 9. ^ -^ o o ^ •• 584 GLOSSARIAL AFFINITIES or distinctively Aryan ; for it will be shown hereafter, under the head- of Scythian affinities, that this same pair of words is found in the Tartar and Finnish languages as well as in the Aryan and Dra vidian, and in particular that the Dra vidian word for ' hand ' reproduces itself in all those languages with an almost perfect exactness. Jcur-al, Tarn, noise, voice ; root Jcur, to make a noise. Comp. Sans. kar, to shout ; gar, to sound. Possibly the Tam. hori, the gallus gallinaceous, is connected with kur ; and if so, the word gallus itself will appear to be related to kori, gallus being in- stead of garrus ; comp. garrulus. The ultimate root of the Tam. kur appears to be ku, to sound (probably a mimetic word), as in ku-y-il, the Indian cuckoo. kudirei, a horse ; Can. kudure, probably from kudi, to leap. Comp. Sans. gJiota, a horse. The Dravidian languages have borrowed ghota from Sans, (in Tamil goram, godagam), said to be from ghut, to retaliate ; but kudirei is regarded as an underived, indigenous Dravidian word. It is probable, however, that the two words are ultimately related. kir-i, to tear. Comp. khur, Sans, to cut, to scratch. ked~u, to spoil or destroy, or (intransitively) to be spoiled or de- stroyed ; verbal noun ked-u, ruin ; relative participle ketta [tt for ^d), bad ; Tel. ched-u. Comp. Sans, khid, to suffer pain or misery, and its verbal noun kheda, sorrow, distress. Comp. also khit, to terrify, and its derivative khet, bad, low ; Greek X^ySos {chedos), sorrow. If these words are allied to the Dra- vidian one, as they appear to be, it must be in virtue of a common origin, for there is not a more distinctively Dravidian word in existence than ked-u. kod-u, Tam. -Mai. fierce, extreme, rough, literally crooked — e.g., kod- ukku, Tam. the claws of the crab ; kod-il, Mai. pincers. Comp. Sans, kut, crooked. Hl-ir, to tremble, to have the hair standing on end. Comp. cJiel, Sans, to shake, to tremble. — See also subsequent list under kulir, cold. se, to be red ; Can. kena, ken ; chem, clien. This root forms the basis of many adjectives and nouns {e.g., sen, red), but is not used anywhere in its primitive, unformed shape. Comp. sona, Sans, to be red. sevi, Tam. ; clievi. Mai. -Tel., the ear ; Can. kivi, Tulu ke'pfi. Comp. kava, the ear, Sans., from sru, to hear. SANSKRIT 585 tad-i, a stick, a club ; verbal tlieme, tadi, to be thick or heavy ; tatt-u, to hit. Comp. tad, Sans, to strike, to beat. ti, fire. Comp. Sans, di, the base of dip, to shine. tuv-u, to sprinkle gently (as dust). tuT-u, to drizzle, to scatter, to spread abroad (as a report). The transitive of tuT-u is tutt-u (pronounced tuttru), to winnow. The ultimate root of all these words evidently is tu, which is also a Tamil form of the root. Comp. dhu Sans, to shake, to agitate ; a derivative from which is dhuli, dust. Comp. also tusta, dust (derivative tus, to sprinkle), with which our own word dust is evidently identical. From Sans. dhuli, Tamil has borrowed tuli, tul, dust, and also tus-i ; but there cannot be any doubt of the Tamil verbs tuv-u and tut-u being underived Dra vidian themes, dhu or tu appears, there- fore, to be the common property of both families of languages ; whilst it is in the Dravidian family that the original meaning of this root appears to have been most faithfully preserved. nad-a, to walk. Comp. the Sanskrit theme nat (said to be from nrit), to dance, to act ; derivatives from which are nata, dancing, ndtaka, a drama, a play. It seems improbable that the Sans- krit word has been borrowed from the Dravidian tongues ; and yet it seems certain that the Dravidian word has not been borrowed from Sanskrit ; for Telugu and Canarese make a broad distinction between the Sans, derivative natinchu or natisu, to dance, and their own theme naduchu or nadi, to walk ; and whilst Sanskrit has many words signifying walking, the Dravidian languages have nad alone, nad-u, to plant, means also in Malayalam to enter, to walk, probably to plant the foot ; nader (nadei), a way, a derivative from this root, is one of the words adduced by Kumarila-hhatta as speci- mens of the non-Sanskritic words contained in the Dravidian languages. Probably, therefore, the Sanskrit nat, nrit, and the Dravidian nad, have been derived from a common source. ney, to weave. Comp. Sans, nah, to spin, originally to join together. Comp. also Greek vr/6^a) (netho) ; German nahen, to sew ; Latin necto. nut. Tarn, thread, to spin, seems to be a verbal noun from a lost root nu, which must have meant to join together, like thelBans. nah. fat, many ; as a verb, to be multiplied. Comp. Sans, fulu, much, more commonly furu. fdd-u, Drav. to sing. Comp. Sans, fath, to read, to recite. The 586 GLOSS ARIAL AFFINITIES Sans. fatJi, is, I have no doubt, the theme from which the corresponding Tel. path-i, and the Tamil 'pad-i, to read, have been borrowed ; and the Tamil fdda-m, a lesson, is clearly derived from the Sans, fdtha, reading, pdd-u, to sing, how- ever, and pdtt-u, a song (Tel. pdta, Can. pdt-u, Gond pdtd)y do not seem to be derivatives from Sanskrit ; but I suspect them to be ultimately related to patJi-a and pdtha, as descend- ed from some ancient source common to both. The ideas expressed are nearly related ; for the reading of all Hindus (and all Orientals) is a sort of cantilena ; and even the San- skrit derivative padi, to read, often receives in colloquial Tamil the meaning to sing. pdl, Tam. Mai. and Can. a portion, a part, a half. Comp. Sans, phal^ to divide ; also Latin pars, a portion, pal appears to be identical with pagal, Tam. a division (also daylight), from pag-u, to divide. The medial g was softened away, as in pddi, half, originally pagudi, from the same root pag-u. See Semitic affinities of this word. piva, other — e.g., pim-n, another man. Comp. para, Sans, in the sense of other, different, foreign, a sense which it often bears — e.g., para-desa, a foreign country. It is with this preposition, and not with pra, before, forward, that I think the Tamil pira, other, should be compared. The use of the Tam. piva, and that of the Sans, para (in the signification adduced above), are identical ; and we might naturally suppose the Tamil word to have been derived from the Sanskrit. Tamil, however, whilst it admits that para was borrowed from Sanskrit, regards piia as an indigenous theme. The r of pira is unknown to Sanskrit, and is considered to be a distinctive mark of Dra vidian words. Tamil has another word, piv-a-gu, after (ultimate base pit), which is generally considered to be independent of, and un- connected with, pita, other ; and yet this very meaning, after, is one of the many significations which are attributed to para in Sanskrit. Possibly, both in Tamil and in Sanskrit, after may have been the first meaning ; other, the secon- dary one. Comp. also pita, Tam. to be born=to come after. It may be concluded, I think, that para and pita are radically allied ; and yet the supposition that the one is derived from the other is inadmissible. Each is too deeply seated in its own family of tongues to allow of this supposition. SANSKRIT 587 and we seem, therefore, to be driven to conclude that both have been derived from a common source. fot-u, to bear. Comp. Sans, hhri (bhar), to bear. It is impossible to suppose that either of these words has been borrowed by the one language from the other ; yet they appear to be , nearly related. See next section. pal, milk. The Dra vidian languages do not seem to contain the verbal theme from which this word is derived. We may compare it with the Sanskrit pdyasa, milk, and also with pdya, water, Zend peo, Afghan po'i ; all of which words are derived from fd, Sans, to drink — a root which runs through almost all the . Indo-European languages. Possibly the Dra vidian pal, milk, may be a verbal noun formed from this very theme ; for a large number of verbal nouns are formed in Tamil by simply adding al or I to the root. Notwithstanding this, the purely Dra vidian character and connections of this word pal, preclude the supposition of its direct derivation from the Sanskrit pd. If pdl, milk, could be considered as identical with pal, a portion, its root would be pag-u, to divide. It is difficult, how- ever, to see why milk should have been called a portion, a share. A poetical, but very common, name for arisi, unboiled rice, in Tamil is amudu-padi, the ambrosial portion or allow- ance. Was it in some such sense that milk was called pdl ? pes-u, to speak ; Can. pel-u. Comp. bhdsh, Sans, to speak. pu, a flower, or to blossom, Tam., Tel., and Can. Comp. phull-a. Sans, to blossom, and pusJipa, a flower. Looking, however, at the Marathi phul, a flower, from phulla, the Dravidian pu seems likely to have been derived from the Sanskrit after all. Tamil has an ancient word of its own for flowers, malar. val, strong ; val-mei, strength. Comp. Sans, bal-a, strength. See also next section. 3. Extra Sanskritic or West Indo-European Affinities. Dravidian words which appear to be specially allied to, or specially to resemble, words that are contained in tlie languages of the Western or N on- Sanskritic branches of the Indo-European family. Some of the words contained in the following list have Sanskrit as well as West-Aryan angjiogies ; but they have been placed in this, rather than in the preceding list, because the West-Aryan affinities appear to be clearer and more direct than the Sanskrit ones. The greater number, however, of the words that follow, though apparently 588 GLOSSAHIAL AFFINITIES connected with the Western tongues, and especially with Greek and Latin, exhibit little or no analogy to any words contained in Sanskrit. If the existence of this class of analogies can be established, it may be concluded either that the Dravidians were at an early period near neighbours of the West- Aryan tribes, subsequently to the separation of those tribes from the Sanskrit-speaking people ; or, more probably, that both races were descended from a common source. The majority of the Dra vidian words which exhibit West- Aryan resem- blances, do not belong to that primary, rudimental class to which the words that the Dravidian languages have in common with the Scythian are to be referred. Nevertheless, they are so numerous, many of them are so interesting, and, when all are viewed together, the analogy which they bring to light is so remarkable, that an ultimate relation of some kind between the Dravidian and the Indo- European families may be regarded as probable. As before, the Dravidian words are to be regarded as Tamil, except it is stated that they are taken from some other dialect. as-ei, to shake. Comp. o-ei-w {sei-6), to shake, to move to and fro. aru-vi, a waterfall ; from ar-u, to trickle down. Comp. riv-us, Lat. a brook, Eng. river ; also the verbal theme of those words, pe-oi {rlie-6) or pvin (rhuo) (as in pmj) (rhue), to flow ; Sans, sru or ru, to run. al-ei, to wander, to be unsteady : alei, as a noun, means a wave. Comp. aAa-o/xat {ola-omai), to wander, aAr; {ale) ; Germ, welle, Armen. alik\ a wave. av-d, desire ; also dv-al, a verbal noun, derived from an obsolete root dv-u, to desire. Comp. Sans, av, of which one of the rarer meanings is to desire. The affinity between avd and the Latin ave-o, to desire, is still more complete, inasmuch as this is the only meaning of the word in Latin, as in Tamil. See also Semitic Affinities. avv-a, Tel. a grandmother. In Tuda av means a mother. In Canarese avva or awe means either a mother or grandmother, or gener- ally, an old woman. The ordinary Tamil form of this word is auv-ei, an honorific term for a matron, an elderly lady, but avv-ei is also used. Comp. the Latin av-us, a grandfather ; avi-a, a grandmother ; av-unculus, a maternal uncle. dvi, a spirit, literally vapour, breath ; then life, and also a spirit : verbal theme dvi, to yawn, to breathe. Comp. the Greek aw (ao), to blow ; also Sans, vd, to blow. WEST INDO-EUROPEAN FAMILY 589 ir-u, to draw, to pull. Comp. ipv-oi {eru-6), to draw. Comp. also ele, the Canarese equivalent of iru, with cAk-w {helk-6), to drag, a word which is probably related to epv-o) {eru-o), through that alliance of r to i^ which is apparent in all languages. iru-mhu, iron ; from iru, ir, the ultimate root, and mbu, a formative euphonised from hu or vu : Tel. inumu. Comp. the Lat. ces, cBT-is, Saxon iren, Danish iern, Old Ger. er, Armen. erJceth. The r of these words appears to have been hardened from s, as may be concluded from comparing them with the German eisen and the Sanskrit ayas. Though I compare this word with Sanskrit, I do not suppose the one to be derived from the other. The root of the Tamil word {ir) appears to mean dark ; and irul, darkness, appears to be another derivative from it ; perhaps also ird, night. tn-u, to bring forth young. Tarn. Can., said of cattle especially. Comp. Eng. to ean or yean (Shakespeare ean), Anglo-Sax. eanian. uyar, high ; when used as a verb, to raise (ultimate root probably u, that). Comp. detp-M (aeir-S), to raise up ; also dep (aer) in dep-d-qv (aer-then) (Aor. pass.), and in the adverb dep-Sr^v (aer- den), lifted up. Comp. also drjp (aer), the air ; Armenian war, high ; Ossete arw, heaven. ur-i, Can. Tulu, to burn ; Tam. er-i. Comp. ur-o, Lat. to burn ; Armenian or, fire ; Afghan or, ivur. There are also some Semitic analogies — e.g., Hebrew ur, fire, and or, light. ul-ei, mire. Comp. e'A-os (el-os), a marsh. ul-ei, howling. Lat. ululo, to howl ; Greek oAoXvfw (olo-luzo) ; English Tioivl (mimetic words). ey, to shoot (an arrow), to cast (a dart). Comp. los {ios), an arrow, iw {hid), to shoot, to cast, ew (heS), to send. er-u, to rise, to get up. Comp. Lat. or-ior, to rise, to get up. eru- ndyitu, Tam. the rising sun, may be compared with the Latin ori-ens sol. ell-a, all. The classical Tamil forms el-dm, all we, elAr, all ye, show that the root is el. Tel. -Can. ella, all ; colloquial Tam. elld. The meaning of el, from which that of ' all ' was derived, appears to ha\e been * boundary.' The primitive meaning seems to have been ' where ? ' Comp. Ossete al,^li, all ; Saxon eal ; old High German al ; English all. Probably the Greek oA-09 (hol-os) and the Hebrew hoi are allied rather to our own ' whole,' Lat. sal-viis, Sans, sarva, than to the Dra vidian and Germanic el, all. 590 GLOSS ARIAL AFFINITIES er, a plough ; also in classical Tam. the work of ploughing ; Can.- Tulu er-u, a plough. We find also in Canarese dt-u, a plough, which appears to be a different form of the same word. Judging by analogy er must be a lengthened form of er, a verbal root, of which the meaning must have been to plough. This verbal root no longer exists in a separate shape, but it seems to survive in erud-u, the ox, erumei, the buffalo, literally as appears, the ploughing animals. In Tulu both animals are called er' (er'). In Tam. er-u, in Tel. er-u, means also manure, especially cow-dung. There is in Tamil a secondary verb, er-ukJcu, to hew, to cut, which seems to contain a reminiscence of the primitive meaning of er. This meaning appears more distinctly in the classical Canarese dr-u, a plough, which seems to be a lengthened secondary form of ar-u, to cut, to sever, a root still in common use in each dialect (comp. ar-u, dr-u, six). The verb meaning 'to plough' in actual use at present is Tam.-Mal. iir-u ; Can. ul-u ; Tulu ur. Comp. Lat. ar-o ; Greek d/)o-co [aro-6), apo-Tos (aro4os); Lith. aru. I do not feel sure of the existence of any relationship be- tween these and the Dravidian words, but the resemblance is worthy of notice. Dr Schlegel in his " Sinico-Aryaca " (Bata- via 1872) connects all the Indo-European words which desig- nate the plough and its uses, and which contain the root ar, er, ir, or, with the Sanskrit ar (ri or ri), signifying to cut or hurt, and ultimately with the Chinese li, sharp, ground to a point, whence are derived various compounds, also pronounced ?i, with the signification of to plough, &c. From this basis he derives the renowned designation of Aryas or Aryas, as meaning ' the ploughing people.' If the Sanskrit ri or ri ordinarily or naturally meant to cut, or even if it could be clearly proved to have ever meant to cut at all, this explanation of the meaning of the words used for plough and ploughing in the Indo-European languages would have carried more weight. I should then also have felt surer of the relationship of the Dra- vidian words with the Indo-European ar', to cut, being an undoubtedly Dravidian- root, and probably the origin of at. Can. a plough. The radical meaning of ri, however, seems rather to be to run. 6r-am, border, brim, margin, coast. Comp. Lat. ora, border, margin. 6r-am has no connection with any Dravidian word signifying mouth ; probably therefore its resemblance to the Latin ora is WEST INDO-EUROPEAN FAMILY 59 1 accidental. The corresponding word in Gujarathi, Marathi, and Hindi, is hor. kad-% to cut, to rend, to reprove ; hatti, a knife, a sword. Comp. Sans, krit, to cut, but especially the English cut ; Norman- French cotu ; Welsh cateia, to cut ; Lat. caed-o. The Persian and Ossete hard, a knife, and probably also the Dravidian hatti, a knife, hattiri, scissors, is from barton, Sans., a derivative of ^nY. han, the eye ; Mn (in the preterite Jcan-du), to see ; also secondarily, to mark, to consider, to think. In the latter sense it becomes hannu in Tamil, but the base remains unchanged. In (kanu, hannu) Telugu, the ordinary n, the nasal of the dental row, is used instead of w, the cerebral nasal. Comp. the Welsh ceniaiv, to see ; English hen, view, power or reach of vision, to ken, to know by sight. In Webster's " English Dictionary " hanna was said to be ' an eye ' in Sanskrit ; whereas it is exclusively a Dravidian word. This error may be compared with Klap- roth's representing huruta, blind, as a Sanskrit word, instead of referring it to the Dravidian languages, to which alone it belongs. There is a curious word in Sanskrit, hdna, one-eyed, which seems to have some Dravidian relationship. It becomes in Bengali hand, blind, which, in form at least, is identical with the Dravidian negative hand, that sees not. Possibly the Dra- vidian hdn, to see, hannu, to consider, may have some ulterior connection with the Gothic hunn-an, to know ; Greek yi/w-vat (gno-nai) ; Sans, nd ; Latin gna (gnarus) ; Old High German chann. The different shades of meaning which are attributed in Greek to yvto-vat (gno-nai) and elSe-vat [eide-nai), seem to corroborate this supposition ; for the latter is represented as meaning to know by reflection, to know absolutely, whereas the former means to perceive, to mark, and may therefore have an ulterior connection with the Dravidian root. haradi, a bear, from haradu, rough, knotty, uneven, the ultimate base of which must be hara or har. The Tuda word for ' a bear ' is harsh (har). Comp. the Persian chars, Kurd harj, and even the Latin urs-us. Comp. also the Samoiede hor go, and the Tungusian huti. harug-u, an eagle. Comp. Persian hergish ; Ossete hartziga ; also Sans, garuda, the mythical eagle ; gridhra, a vulture. hal-a. Can. to steal ; Tam. halavu, Mai. hall-am, a theft. Comp. Lat. clep-o, to steal ; Greek KXair-eis {hlap-eis). See also Scythian Affinities. 592 GLOSSAEIAL AFIINITJES •^ gaV'i, Can. a cave, a cell, Jcapp-u, Can. a pit-fall ; Jceb-i, Tarn, a cave. The equivalent Sanskrit words are guhd, a cavern, from guh, to conceal, and gaha, a cave, from gah, to be impervious. guhd has become in Tamil kugei. It seems doubtful whether the Canarese gavi and the Tamil kahi are not both tadbhavas of guhd. On the supposition that they are independent words, comp. the Lat. cave-a, a cavity, a den, from cav-us, hollow ; theme cav-o, to hollow out ; and with this the Telugu kapp-u, to cover over, probably the origin of the Tamil kaff-al, a ship. I See also Scythian Affinities. kdy, to be hot, to burn. The Tel. kd-gu (also kd-lu), Can. kdy-u, to burn, and the Can. kdge, heat, compared with the Tamil kdn- gei, show that the ultimate root is kd, to which y or gu is added dialectically as a formative. The only Sans, word which seems to be related to this Drav. one is kdm-a, to desire ; and we should not, perhaps, have suspected this to be related, were it not for its connection with the Hebrew hdm-ad, to desire, and the derivation of that word from hdm-am (base Mm), to be warm. Comp. with the Dra vidian kd or kdy, the Greek Kat-w {kai-6) (Attic Ka-w) {ka-o), to burn, to be hot. The words seem identical. Liddell and Scott represent /catw {kaio) to be connected with the Sanskrit such, to dry. How much more nearly it appears to be connected with the Dra vidian kdy. Besides, the Dravidian languages have another word which seems to have a real relation to sush-a — viz., sud-u, to burn. kind-u, to stir, to search, to turn up the ground. Comp. Kevr-ew {kent-eo), to prick, to goad, to spur on. kira, old (not by use, but with respect to length of life) ; ultimate root kir, beneath, that which has gone down. Comp. Sans. jar-as, age, but especially the Greek words signifying age, aged^ — viz. yyjpa-s, [gera-s), yr]paL-6<5 [gerai-os), yepat-os {gerai-os), yep-cov (ger-Sn) . See also the Scythian affinities of this word. kira-mei (base kira), a week, literally property, possession, each portion of a week being astrologically regarded as the property or inheritance of some planet. Comp. Ossete kuri, kore, a week ; Georgian kuire. Possibly these words are derived from the Greek Kvpt-aiaj ikuri-ake), Sunday, the Lord's day; but whence is the Greek word derived? From KvpL-o (ep-6) (for FeTrw) {gepo), to' speak. sel, to go, to proceed. This is unquestionably a Dra vidian root, and abounds in derivatives — e.g., sel, the white ant ; sel-avu, ex- penditure ; sel-vam, prosperity. It forms its preterite also in a manner which is peculiar to pure Dra vidian verbs. It is obviously allied to the Sanskrit sal, to go or move ; sel, to move, to tremble ; ehal and char, to go, to shake, to totter ; and also to the Hindustani derivative chal, to go. Close as these analogies are, sel appears to bear an equally close resem- blance to eel, the obsolete Latin root, signifying to go, from which are formed celer, and also ex-cell-o and prce-cell-o. The same root is in Greek kcA (kel) — e.g., KeX-rjs {hel-es), a runner ; and KeAAw {kello), to urge on. tag-u, fit, proper, worthy. Comp. German taug-en, to be fit for ; Gothic dug-an ; German tiigend, virtue ; tiichtig, fit, able ; English doughty. tayir, curds. Comp. Greek Tvp-6<; {tur-os), cheese ; Sans, dhayi, drinking, sipping, sucking. tin, to. eat light food, to eat away; tindri, eatables. Comp. rei'S-o) {tend-6), to nibble, to eat daintily ; r^vO-q^ (tenthes), a gourmand. tit-a (pronounced nearly like tota), to open; tita-vu, an opening, away, a means. Comp. Greek Ovpa {thura), a door ; German thur ; Old High German tor ; Gothic daur ; Sans, dvdra ; Vedic- Sans. dur. These words are commonly derived from the Sans. theme dvri, to obstruct, to cover ; but as they all mean not the door-leaf, but the door-way, and metaphorically a way, or means, this derivation of them from a root meaning to close seems less natural than that of the Dra vidian tita or tota (Can. teta), to open. Comp. tri, Sans, to pass ; tiras, through. tind-u, to touch, to kindle. Comp. Gothic tandya, I kindle. Possibly WEST INDO-EUROPEAN FAMILY 595 there may be a remote connection also with the Sans, danh, to burn, the intensitive of which is dandah. On the other hand, the n of the Tam. tmd-u is probably euphonic, for it disappears in the Can. tid-u, and in the Tam. transitive form of the verb tUt-u, to whet. te]^ clear. Comp. 5^A-os (del-os), clear, manifest. tol-ei (base tol), distance, limit, end. As a verb, tol-ei signifies to end, or come to an end. Comp. Greek tcA-os (tel-os), an end, and rrjX-€ {tel-e), far off, which Buttmann derives from reA-og {tel-os). tripp-u, Tel. to turn ; from which, by corruption, tippu, the ordinary form of the word, has been derived ; Can. tiru-pu ; Tamil tiru-ppu. These are causal or active verbs, and the corre- sponding neuter or intransitive verb signifying to turn, is in Tel. tiru-gu, in Tam. tiru-mhu. Canarese has tiru-hu, tiru-vu, and tiru-gu. There are also a few related themes — e.g., tiru-gu, Tam. to twist or turn, and tiri, the same ; from which is derived tirigei, a mill. From a comparison of these words, it is manifest that their common base is tir, to turn, to which various formative additions have been made, for the purpose of expressing modifications of meaning. Comp. Greek Tpeir-io (trep-6), to turn ; which bears a remarkable likeness to the Tel. tripp-u, and the initial portion of which (with that of our English turn) seems closely allied to the Dra vidian base. Possibly also the Sarus. tarku, a spindle, is either a collateral word or one which has been directly borrowed by Sanskrit from the Dra vidian tongues. nas-u, to crush, to squash. Comp. vda-cr-u} (nass-d), to squeeze close, to stamp down. nar-a, Can. a tendon, a sinew ; sometimes, but improperly, a vein or artery ; adjectivally wiry, stringy. Tel. naramu, Tam. nar- ambu, Rajmahal ndru. A secondary form of the word is ndr, fibre, from the base nar ; with which compare the Latin nerv-us and the Greek vevp-ov {neur-on), a tendon, a ligament. nin-etf to think, to remember ; Can. nen-e, nen-a. This word is un- doubtedly a Tamil primitive, and is probably the basis of nenj-u, Tam. -Mai. the mind, the heart ; also the diaphragm. [With respect to this double signification of the word nenj-u, compare the twof^d meaning of ^pi]v (phren), in Greek, viz., the diaphragm or chest (supposed to be the seat of the mental faculties), and also the mental faculties themselves.] If there is any analogy between the Dravidian nin or nen, and the 596 GLOSS ARIAL AFFINITIES Sanskrit man, to think, it comes to light only by comparing it with the corresponding Greek word vo-ew [no-eo), by reduplica- tion vevo>^-/xat [neuoe-mai). fivd-ofxaL (mna-omai), to think on, to remember, and /xey-os {men-os), wish, are in perfect accordance with the Sanskrit mana, and are probably more ancient than vo-€ii) (no-eo) ; of which the initial v (n) has been changed either from /x (m) or from yv (gn) (yvw-vai) (gno-nai). The Dravidian nen or nin may in like manner, I conceive, have been changed from an older men or min, allied to man-a and fj.€v-o.r. ivan. ivanu. ivdnu. (4.) In the Dravidian languages contingency is expressed by the addition of a particle to any verbal tense, person or number. This subjunctive suffix is in Telugu eni, or i ; in Canarese re, rd, or dgyu. One of the suffixes employed in the Tamil is kdl, which in the speech of the vulgar becomes kd ; and this very particle kd, added, as in Tamil, to the preterite, is the suffix by which the Ku also forms conditional or contingent verbs — e.g., If I do is in Telugu nSnu cheyuduneni ; in Canarese ndiiu geyidare ; in colloquial Tamil this is ndnchey- ddkkd ; in Ku also (from the root gi, to do), it is dnu gitekkd. On the other hand, in the following particulars Ku agrees more closely with Telugu than with Tamil or Canarese. RUDER DRA VIDIAN TONGUES 631 (1.) It uses the neuter singular to denote the feminine singular. (2.) The oblique cases or "inflexions " of the pronouns of the first and second persons, singular and plural, are identical with those of Telugu. (3.) The case-terminations of Ku are nearly in accordance with those of Telugu. (4.) The pronominal signs suffixed to the Ku verbs accord on the whole better with Telugu than with any other dialect— e.g'., in Tamil the second and third persons plural end differently, the one ir, the other dr ; in Telugu they end alike — both generally in am ; in Ku also both these persons end alike in eru. (5.) In Canarese all relative participles, including that of the relative verb, end in a ; in Tamil all relative participles, with the exception of that of the future, have the same ending. In Telugu the relative participle of the indefinite or aoristic tense ends in edi or ei ; and in the Ku also the relative past participle exhibits this ending. Thus, ana, Tamil, that became ; in Canarese dda ; in Telugu (indefinite tense), ayyeti ; in Ku the same form is dti. The various particulars now mentioned prove Ku to be distinct from Gond, and though it is allied to it, it is allied only in the same manner as to the other Dravidian languages. In some points this language differs from all the other dialects of the family ; for example, it forms its past verbal participles not by means of the suffixes du, i, or si, the only suffixes known in the other dialects, but by suffixing to the root d, sometimes sd or jd, after the manner of some of the languages of Northern India. In the other dialects of this family, with the exception of the Tulu, the negative verb possesses only one tense, an aorist ; the Ku, in addition to this negative aorist, has also^ like the Tulu, a negative pre- terite — a decided advantage over the other dialects. The Ku suffixes of the present verbal participles are also different from those which are found in the other Dravidian dialects. The formative suffix of the present verbal participle is in Telugu c/iu or tit ; in the Canarese utaov utc ; in the Ku it is i or pi. 5. Rajmahal. — The lists of words hitherto published do not go a great way towards proving this language distinctively Dravidian. The evidence of the pronouns and lowest numeral is clear ; but the other distinctively Dravidian words found in the lists are not numerous. For the present, perhaps, all that can be said with certainty is that the Rajmahfd contains a distinctively Dravidian element. When it has been examined as carefully as the Ordon, it may be possible to speak of its relationship in stronger terms. It is commonly stated that it is almost the same as the Oraon ; but this o]nnion, though ])robably correct, requires confirmation. The principal and most essentially Dravidian words I have noticed are as follows :— I, en. eye. kdne. Drav '. kan. we, en. {nam. om). nose. muna. do. mUkkii. thou, nina. tooth, pdla. do. pal. yoi-i. nina {nima in nimki , yours). ear. kedu. do. kddu. he, she, it. dth. hand, kekha. do. kei. they, owar. hair. tale. do. ' head.' this, ih. tree, man. do. mdn-u. that, dh. • flower. phitp. do. pH. here. ino. fish. m in. do. mm. there, ano. dove, pur ah. do. purd. one, art, ort scorpion. , tUah. do. tsi. 632 APPENDIX why, endhar. dative suffix, ku. Drav, do. do. do. do. do. do. ay I. maq-al. dl." har-a. eg-a. pain, ndgi. Drav, nogu. above, mekthi. do. mSl. motiier, aya. daughter, ?wo^'*. man, al, alia. come, bar-en. go, ek-cn. Unfortunately the inflexions of the Rajmahal noun and verb are not given in any of the lists, so that with the exception of a very few incidental particulars the grammatical construction of this language remains unknown. In the parti- culars that follow^ the construction is Dra vidian. The dative postposition is ku; m is the sign of the plural of the pronouns of the first and second persons, replac- ing n of the singular ; ar is the sign of the plural of pronouns of the third person. 6. *Oraon. Much light has been thrown on the construction and vocabulary of the Oraon by an article on that language in the Journal of the Bengal Asiatic Society, vol. xxxv. , by the Rev. F. Batsch, a missionary who has laboured amongst the Oraons in Chfitia Nagpur. The personal pronouns, which are very regular and distinctively Dravidian, are as follows : — I. We. Thou. You. em. emhai. emage. em an. emgu&tin. emanti. emanii, emanum. emim. (1.) em, we, is the ordinary plural of the first person, used when we means more than two. nam, which is equally Dravidian, means, it is said, ' we two,* This use of nam as a dual may throw light on the origin of the plural inclusive of the other Dravidian languages. (2.) The third person is represented only by ds, he, ad, it or she, dr, they. Notice the Dravidian epicene plural in r. What ? is end. (3.) Postpositions, ge, to, for ; nu, upon ; mund, before ; mcclila, above ; kuti, besides; katha, beyond ; menya, up ; kinya, beneath. These are purely Dravidian words, gusti, from, I cannot identify. (4.) Numerals. One, onta ; two, enr; three, mund ; four, 7idck. Adjectival numerals— or^ dlao, one man, irih alar, two men. The rest of the numerals are borrowed from the northern vernaculais, (5 ) Words certainly Dravidian are «/, man, pal, tooth, khan eye, hoi, mouth, raoy, nose, har, come. (6.) With the exception of the words cited above, the rest of the Oraon nouns, adjectives, and verbs present scarcely any point of resemblance to Dravidian words. The mass of the words in the Oraon vocabulary may be Kolarian, but do not seem to be Dravidian. This instance tends to show that languages may be * Since tlie publication of tlio 2nd Edition of this Grammar much light has been thrown upon the connection of the ORAON or KURUKH with the Dravidian languages by the Rev. O. Flex's "Introduction into the Orfton Grammar" and a "Kurukh Grammar," by the Rev. Fred. Hahn, German Evangl. Lnth., Mission, Chota Ndapur, 1905.— Eds. t nom. en. gen. engha. dat. engage. ace. engan. abl. engusti. instru. enganti loo. engmi. agent. enim.. nin. mm. ninghai. nimhai. ningdge. nimdge. ninin. niTnin, nimanuvi. ningusli. nimgustim. ninante. nimanti. ninganu. nimganu. ninim. nimim. THE BRAHtri 633 cognate, whilst yet the proof may survive only in the pronouns, the first few numerals, and the structure. 7. Dravidian Element in Brahui. — In many of the particulars in which the Brahui is found to he allied to the Dravidian tongues, it is equally allied to each of the families of tongues included in the Scythian group, so that to that extent it would he safest to content ourselves with saying that the non- Aryan element contained in Brahui — the element which is incapable of being affiliated to the Indo-Persic — appears to be Scythian, using the term Scythian in its widest sense. Thus in Brahui, as in the Dravidian dialects, and in the whole of the Scythian tongues, the cases of nouns are denoted by postpositions. The gender of nouns is expressed, not by their inflexions, but by prefixed separate Avords. The number of nouns is ordinarily denoted by the use of separate particles of plurali- sation, such as many, several, &c. When a noun stands alone without any such sign of plurality, its number is considered to be indefinite, and it is then regarded as singular or plural according to the context, or the number of the verb with which it agrees. This rule is more characteristic of Tamil than of the other Dravidian idioms. Adjectives are destitute of comparatives and superlatives. On the other hand, there are certain particulars in which the Brahui appears to me to present traces of the existence of a distinctively Dravidian element. The observations I made on the Brahui in the first edition of this work were founded on a brief grammar and vocabulary of the language contained in vol. vii. of the Journal of the Bengal Asiatic Society. A fuller grammar and vocabulary have now been supplied by Dr Bellew, in his book entitled "From the Indu.s to the Tigris" (Triibner, 1873), and it appears to me that the theory I advocated — (not that the Brahui was a Dravidian language, but that " it evidently contained a Dravidian element, an element which was probably derived from the remnant of some ancient Dravidian race incorporated with the Brahuis") — has been con- firmed. (1.) The Brahui pronoun of the second person singular is ni, thou, precisely as in all the Dravidian tongues. The plural of this pronoun — viz., num, you {numd, of you), is also wonderfully in accordance with old Dravidian forms. The Canar- ese is nim^ you ; the Oraon tiim ; the old Tamil possessive in num-a, you (in which we see traces of an obsolete base num or nUniy you) ; and the ordinary base of the oblique cases of this pronoun in Tamil is um. It has been objected that there is nothing distinctively Dravidian in these forms, seeing that ni, thou, appears in some shape in the Australian dialects, in Chinese, and in many of the languages of High Asia. This pronoun of the second person has undoubtedly a very wide range, as has been shown in loco, but it is remarkable that throughout India and the countries adjacent to India it is found only in the Dravidian languages and the Brahui. The change from ni in the singular to num in the plural appears to me still more distinctively Dravidian. (2.) Whilst ntm or nUm is to be considered as the most classical form of the plural of the Dravidian pronoun of the second person, nir is the form ordinarily used in a separate shape in Tamil, niiru in Telugu ; and in consequence of this plural termination in r, in nearly all the Dravidian idioms the second person plural of the verb in the indicative mood ends, not in im or um, but in ir, eru, dru, iri, &c. The same pecfiliarity reappears in the Brahui. Whilst the separate pronoun ends in m, r is the pronominal sign of the second person of the verb — e.g., areri, ye are, arer, they are ; with which compare the Canarese iru{tt)iri, ye are, iru{tt)dre, they are. 634 APPENDIX (3.) A remarkable analogy between the Brahui and the Dravidian languages is apparent in the reflexive pronoun ten^ self, se. In the Dravidian languages this pronoun is tan or tdn, and is regularly declined, whilst the nominative is also used adverbially in the sense of ' indeed.' In Brahui ten is similarly used, not as a particle, nor only as an adverb, but as a pronoun, and is declined as regularly as the other pronouns. (4.) Nouns form their plurals by adding ^■, as in Gond — e.g., huli, a horse, hiilik, horses. (5.) The root of the substantive verb in Brahui is ar, in Tamil and Canar- ese ir. (6.) Bopp remarks that the three lowest numerals could never be introduced into any country by foreigners. The truth of this remark is illustrated by several circumstances of which Bopp could scarcely have been aware. From five upwards the numerals of the Oraon are foreign. From four upwards the Brahui numerals are of Indo-European origin {e.g., char, four, panj, five, shash, six) ; and in the compound numerals twenty-one and twenty-two, the words for one and two are also Indo-European, but the separate numerals one, two, three, are totally uncon- nected with the Sanskrit family, and two of them are identical with Dravidian numerals. In Brahuil two is i7mt; compare Can. eradu, two ; Tam. irat{tu), twofold or double. InTBrahui, three is musit; compare Can. mUr-u ; Tel. mUd-u ; Tulu miiji. The Dravidian bases of these numerals are ir, two, mu, three ; and if we notice the terminations of the Brahui numerals (one, asit, two, irat, three, musit), it is obvious that the second syllable of each of these words, it or at, is merely a neuter formative, like that which we find in the Dravidian languages {e.g., compare ir, the base and numeral adjective * two,' with irachc, the abstract neuter noun) : consequently the agreement of the Brahui -with the Dravidian numerals, both in the base and in the formative, is complete. If we remember the interchangeable relation of s and r, and if w^e regard the Canarese miir, three, and the Brahui mus, as an instance of this interchange, as I think we may safely do (illustrated as it is by the Tulu mUji), we may also venture to connect the Dravidian numeral base or, one, with the Brahui as. This connection, how- ever, is doubtful, whereas there cannot be any doubt respecting two and three. It is worthy of notice that one is achat in Pehlevi. (7.) In the class of auxiliary words (prepositions, conjunctions, &c. ) compare the Brahui 7no7ii, opposite, with the Tamil munne, before. The number of nouns and verbs in Brahui which can with certainty be identi- fied with Dravidian roots is not considerable, but it is equal to the number found in the Oraon vocabulary. Brahui. DllAVIDIAN. Brahui. Dravidian. eye. khan, Jchan, kan. stone. khal, kal. mouth, ha. My, vdy, hoi. bow, hil, hil, bir, vil. ear, khaf. kivi, kdd-u, kdi(. saw. ara, ara-m. face. mon, mun, before. scorpion, telt, te,. brain, milt, muei. to cut. hare. aril, ari. son, mar, marri (G6nd). to beat. khal. kol (to kill). mother, di. dyi. to do, ke, kar. ke, ge, chey. water. dtr, nir. to come, har. har, var. milk. pdhlt, pal. to be, ar, ir. THE BRAHUI 635 The analogies between the Brahui and the Diavidian languages which have now been pointed out, are much closer than any analogy which subsists between the Dravidian languages and the Bodo, the Dhimal, and the languages of the other tribes on the north-eastern frontier of India which were termed "Tannilian" by Mr Hodgson. Those analogies appear to me to be almost as remote as those of the Tibetan family ; and are not only less numerous, but also of a less essential character and less distinctive than the analogies which are discoverable between the Kolarian tongues and the Dravidian. Compare the following list of Dravidian words of primary importance with analogous words in the Brahui, and with the words in the Bodo and Dhimal which correspond in signification : — Dravidian. Brahui. thou, ni, ni, you, num, num, we, nam, nan, self. tan, ten, one, or. as-it, two, irad-u, irat, three, mur-u. mus-\ eye, lean, Tchan^ ear, Mvi, khaf. water. nir, dir. stone, kal, khal. It seems unnecessary to give a larsrer Bodo. Dhimal. nayig, nd. nangchur, nyel. Jong, kyel. goui, tdi. che, e. gne, gne. tlidm, Slim. mogon. mi. khomd, ndhdlhong. dOi, chi. onthdi. unthur. :r of instances ; for whilst the Brahui does appear to a certain extent to contain Dravidian forms and words, the Bodo and Dhimal, and to them may be added most of the other dialects of the north-eastern forests, present no special analogies whatever ; and contain only a few of those structural affinities which they have in common, not only with the Dravidian, but with the Tibetan, and with every language and family of languages of the Scythian group. INDEX Abaseni, 90 Abhiras. 109. Accadian, 607. Ethiopians, 108. Afghan, 389. 600. Agamas, 83. Agastes'vara. 115. Agastya. 99. 114. 115. "Aioi. Aioi. 87. 95, 96. Akrida. 14, 18. Albanian. 622. Altaic, 61. Ambalakkadu, 8. Amravati. 124. Andarae, 12. Andhra, 2, 4. 8. 9, 10. 26. 27 _ 107. Andhra-Dravida-bhasha, 4. 27. Andre Indi, 10, 26. 28. Anglo-Saxon. 46, 300. Aornis, 101. Aravam, 14, 15. 16. Arcot. 93. Argaric Gulf. 100, 101. 'ApicLKT), AriakS. 10. 94. Arjuna. 11. 12. 111. "ApKUTos, Arkatos. 93. Armenian, 246. 332, 342, 389. 390, 598, 616, 622. Armorican, 485. Arrian, 280. Irya-bhatta, 87. Arya-vartta, 108, 148. Asoka's inscriptions, 10, 13, 124, 128. Asam, 39. Asseniani. 23 note. Ati-vlra-Rama Pandya. 99. Augustus. 11. 103. Australian. 75. 76. 77. 383. 395. Avar, 622. Ayodhya. 110. B Babylon. 65, 89. 364. 374. 597. 609. Badaga, 29, 30, 34, 157, 276. 626. Baghistan. 65. Barace. 95. Barbosa. 22, Basle missionaries, 31. Basque, 382. 614. Bdrot. Batoi. 402. Batsch. Rev. F.. 36. 032. Beames. Mr. 58, 60, 124, 128, 159, 164, 185, 186, 280, 281, 403. 410. Behistun, 49, 56, 57, 65, 68. 77, 104, 108, 184. 212. 274, 282. 295. 296. 322. 327. 357. 393. 395. 406. 474. 524. 610. Bellew. Dr. 633. Beluchf. 877, 409. Baluchistan, 2, 67. Benfey. Professor. 575. 576. Bengal. 133. Bengali. 41, 42, 53. 56. 57, 147. 155. 158. 272, 297, 327. 379. 410, 412. 436. 465. 509. 619. 591. Beschi, 130. 215. 224. 474. 478. BrjTTiyJ), Bettigo. 93. 102. Bhlls. 107. BhojpurJ Hindi, 519. Bhotas, 5. Bhotiya, 606. Bhiimig, 418. Bhutan, 107, 281. Bleek, Dr, 63, 72, 77, 159, 185. Blunt, Captain, 630. Bodos, 39, 281. 294. 634. 635. Bohemian. 375. Bohtlingk. 63. Bolingae. 28. Boiler, 63. Bopp. 247, 333, 389, 413, 441, 474. 495, 509. 570, 571, 634. Bornu, 77, 240, 388, 393, 448. Bo^iTTa, Boutta. 102. Brachme, 102. Brahma, 190, 343. Brahmadesam, 102. Brahmans. 2. 41. 45, 47, 48. 53.84. 102. 110. 112, vassim. Brahui. 39, 40, 57. 66. 77. 105. 108. 148. 150. 160. 164. 232. 239. 246. 272. 274. 296. 300. 326. 331, 333, 335, 364, 371. 387, 393. 409. 417. 571. 632, 633, 634. 635. Brigel. Mr. 32. 123. 279. 363, 504, 526. Brown, Mr C. P., 22. 27. 29. 30. 175, 233. Buchanan, Dr. 95. Buddha, 15, 79. Buddhists, 9, 15. 79. Buhler, Dr, 150, 568. Bunsen, 62. Buriat, 482, 620. Burghers, 30, 34. Burma, 392. Burmese, 124. Burnell. Dr. 33. 94, 95, 96, 116, 117, 126, 127. 128, 578. Burton, 14. Buttmann, 595. Caki., 99. Cselobothras, 92. Caldwell, Mr R. C, 70, 351. Cahcut, 18, 95, 570. Calingae, 12. Calingam, 29. Calingas, 102. Calingon, 100. Calmuck, 246, 272, 296, 382, 391, 616, 620. Calymere Point, 100, 101. Campbell, Mr A. D., 27. 28. 472. Campbell. Sir George. 35. 37, 38. Canara, 30. Canarese, 4, passim. Cannanore, 7. Canopus, 115. CareT, 96. Carey 42. Camaticre, 31, 93, 102. Carrical, 3. Cashmirian, 442. Castren. 63. 243. 391, 393. 409. Caucassian, 228, 327, 331, 332. Celobotras, 18. Celtic, 69, 70, 71, 105, 148. 184, 294, 356, 377. 380. Celts. 68. Central India, 415. Xal3ripos, Chaberos, 23, 102. Chaldean. 54, 206, 374, 409. 440, 477, 609. Chalukya, 13. Chandragiri, 16, 20, 31. ('handraiiiri, P.aja, 7. Chennapattanara, 7. Chennappa Nayakka, 7. Chentsu. 418. Chera Dynasty. 94. 128. Cheralam. 18. Ch^ras. 11. 14. 128. Cheremiss. 281, 295. 335. 615, 617, 621. Chicacole. 25. Chinas, 4, 115. Chinese, 56, 61, 65. 71, 75. 77. 82. 104, 157, 185, 191. 332, 382. 383, 389, 390, 392, 393, 395, 406, 448, 466, 590, 617, 623, 633. Chinglepat, 7. Chintamani, 48, 84, 265, 299. Chittar, 98. Chola, 5, 8, 10, 11, 13, 14, 21. 25, 93, 107, 115, 127. Choliya, 14. Chijtia Nagpijr, 36, 37, 632. Clay, Mr, 286, 345. 527. Clemens Alexandrinus, 10, 102. Cochin. 3, 16, 86, 126, 134. Coimbatoor, 18, 94, 98. Colchic Gulf, 99. Colcis Indorum, 99. Cole, Major, 33. Colebrooke, Mr. 8. Colis, Coliacum, 101. Colombo, 7. Comorin, Cape, 1, 6. 17, 96, 97, 98. 99, 101, 102, 114. Conti, N., 102. Coorg, 6, passim. Coptic, 614. Corean, 382. Coromandel, 20, 21, 22, 100, 115, 573. Cosmas Indicopleustes, 18, 23, 103. Cottara, 95. Cottonara, 95. Ctesias, 89, 90, 91. Cunningham. General. 29. 93 102. Curzon. Mr. 108. 109. Cuthite. 440. Dalton. Colonel .34. 35. 3 37. Damilo, 9. 109. Damirici. 10. 94. Dandakaranya. 115. Danish. 10. 443. 589. 602. Daradas. 4. Darius Hystaspes. 65, 357. 636 INDEX 637 Dasyus, 53. 106. Dawson, Rev. J., 34. 626. De Nobilibus. Eobt, 318. Deva-Nagart, 124. 125. 129, 130. 131. 132. 138. 151. Dharwar, 25. Dhimai. 39, 75. 245. 436, 437. 634. 635. Dimirica. 10. Dionysius Periegetes. 101. 104. Doms, 107. Doric. 152. 242. 247. 389. 401. 411. Dowson. Professor, 18. Dravidas. 2. 4. 5. 6, 9. 10, Dravidian. 1. passim. Dravidi Prakrit. 5. Draviras, 42. Driberg, Rev. J. G., 34, 626. Duff, SirM. E. Grant. 11 note. Durga. 96, 613. Dutch. 21. 593. E Edkins, Dr, 52, 64, 65, 373. 382. 389. 392. 488. 512. 524, 545, 623. Eggeling. Dr. 18. 25, 128. Egypt, 103. Egyptian, 89, 127, 295. 374. Ellis, Mr. 32, 125, 572. English, 46, passim. Esthonian, 185, 376. 380, 391, 542 Ethiopic, 127, 374. Eugubian Tables, 242. Eusebius. 104. Eastatbius. 104. Festus Avienus, 104. Finnish, 42. 49, 55, 61, 63. 64, 67. 71. 82. 104, 105, 140, 149. 161. 162, 180. 184, 185, 217. 246, 254. 274. 277. 281, 282, 285. 295. 305. 307, 327, 332, 335, 338, 341. 342. 354, 380. 381. 382. 390. 391. 393. 401, 406, 409, 437. 438, 439. 463. 477. 482. 495, 510, 520. 525. 542. 569. 610, 612, 613, 614, 615, 616, 617, 620, 621, 622. Finno-ugrian, 357, 409, 477. Fluellen, 59. Frater Paulinus a St Bar- tholomfeo, 21. French, 46, 184, 591. Frisian, 300. Gaeiic, 620. Gangarides Calingse, 29. Ganges, 28, 29. 59. 95. Ganjam, 102. Garo, 395. Gauras, 42. Gaurian, 56, 57, 280, 409, 410. 411. 412. 464, 466, 484, 509. Gentoos, 25. Georgian. 327. 381. 391, 592, 593, 616. Georgius Syncellus, 104. German, 52, 135, 173, 211, 247. 294. 375. 376, 377, 401, 500. 568, 585, 589. 591. 593. 594. 598. 622. Gesenius, 533, 609. Ghauts, 6, 16. 19. 31, 32, 33, 575. Goa, 3. GodavarJ, 29. Gond, 6, passim. Gondvana, 25, 34, 35. Gothic, 69, 70, 71, 148, 251, 295, 299, 300, 323. 338. 342, 356. 368, 374, 376, 389, 413, 438, 509, 510. 591 594 Gover, Mr,' 365, 410. Gr.TBCO-Scythian, 40. Graul, Dr. 494. 511. Grantha, 19. 86, 124. 125. 129. 131, 133. 146. Grater, Rev. B., 33. Greek, 10, 52, 68, 69, 70. 71. 73. 97, 98, 99. 101, 103, 134, 148, 152, 164, 173, 175. 179. 180, 182, 192, 206, 211, 233, 247, 252, 253, 273. 276, 285, 294, 356, 368, 374, 375, 376, 377, 378, 390, 401, 410, 411. 448. 449. 474. 508. 542. 566. 571. 572. 580, 582, 584, 588. 589. 591. 592, 593. 594, 595, 596, 597, 598, 599, 610, 616, 621. Growse, Mr F. S. (M.A.), 54. Guiarati, 1, 2, 3, 56, 57, 158, 272, 274, 364, 390, 394, 409, 412, 414. 591, 616. Gulf of Manaar, 99. Gumsur, 35. Gundert, Dr, 9, 15, 19, 20. 21. 24. 30, 58. 86, 95, 97. 109, 134, 136, 153. 155. 163, 168, 171, 174. 178. 187. 188. 190. 194, 197, 209, 213, 217, 230. 242. 247. 250. 251. 256. 261. 262. 263. 268. 272, 276. 279. 282. 283, 304, 305. 325. 327. 331. 332, 347. 351. 352. 354. 356. 365, 369, 428, 435, 449. 456. 475, 476, 494. 504. 524. 525, 530, 531, 547. 575. 578, 598, 623, 629. Guntur, 30. Gyami, 392. H Hahn, Rev. F., 520 note. Hala Kannada. 30, 126, 128. Hassoun, Mr, 90. Hebrew, 3, 41, 71, 74, 75, 88. 134. 139, 172, 204. 205, 206, 217, 259. 371, 374, 401. 437. 440. 443, 452, 477, 485, 538, 545. 609. 614. Herodotus, 89, 90, 107. Hieronymus. de S. Stephano, 22. Himalayan, 294, 395, 436. Himyaritic, 127. Hindi, 2, 50, 53, 56, 57. 58. 147. 158, 272, 280. 281. 390. 410. 412. 436, 534, 535, 591, 612. 613, 627. 628. Hindu. 26, 93, 586. Hindustani, 2. 7. 38, 155, 280, 299, 301. 302. 436. 461. 462, 465, 474. 569. 594. Hislop, Rev. S.. 34. Ho (Hoi). 38. 299. 418. Hodgson, air. 35. 36. 38. 39. 52, 635. Horpa, 392. Hottentot. 185. Hunfalvy, Professor. 63. 68 nole, 357, 383, 477. •Hungarian, 63. 64, 67, 77. 161. 184, 185, 197, 235. 307. 308. 448. 450, 463. 482. 510. 574, 610, 612. 613. 614. 615. 616. 617. 620. 621, 622, 623. Hunter. Dr. 364. 415, Hwen Thsang. 10. 13. 26. Hyderabad. 2. 25. Ibn Batutv, 21. 24. 93. 103. Icelandic. 242. 300. 568. Indus. 148. 160. Iravars. 109. Irish. 144. 243. 378. 582. 600. 593. Irtish. 281. Irulars. 34. Italian, 41, 53, 81, Jaffna, 82. Jaina, 84. 85. 117. 190. Jambulus. 128. Japanese. 64. 157, 255. 281. 296. 305, 306, 312, 359. 459. 463. 464. 468. 512. 523. 616. 623. Japhetic. 389. 392. 422. Java. 26. 124. Jenesei. 613, 016, 617. Jesajabus, Patriarch. 23 note. Jews. 86. 87. 126. Jones. Sir W.. 54. Jurazen. 616. K Kafik, 185. Kali, 613. Kalingapura, 31. Kalingas, 7. 8. 26. 27. 29. 107, 111. Ka\\i7t\'6j'. Kalligikon. 100. Kalyanapuri. 31. Kambar. 48, Kambojas. 4. Kamtschadale. 616. Kandh. 629. Kangazian. 616. 622. Kanettri. 95. Kanuri. 77, 393, 448. Kapur Di Giri, 17. Karen, 392. Kapeoi. Kareoi. 96. Karnataka. 2. 9. 30. 31. 85, 86, Karnul, 25. Kdpovpa. Karur. 94. 97. Karu-manal, 22. Kattywar. 25. Katyayana, 12, K.lveri, 13, 98, 102. 114. 569. Kavi. 380. 389 Kayal. 99 Keikardi. 418. 419. Kelan Kelu. 17. Kelat. 39 Kera, 94. Keralamputra. Keralaputra. 17. 92. Keralas. 8. 14. 18. 87. 97. 107. Kerala Utpatti. 143. KrjpojSddpos, Kfirobothros, 18. 94. Kern. Dr. 9, 29, Kesava. 80. Khasas. 4. Khiwan. 380. Khond. 6. passim. Kiratas, 4. Kistna. 124. Kit chin's Atlas. 22. Kittel. Mr. 325. 331. 335. 339. 340, 343, 490. 491. 578. 579, Klaprotb. 591. Klings. 7. 25. 615. Koccb. 37. Kodagu. 6. 32. 38, Kodun-Damir. 6. 63S INDEX Koelle, Mr, 77. Koi, Koitor. 35. Koibal, 323. Kolarai, 418. Kolarian, 36, 39. 107, 281, 299 629. Kolatta-nadn. 9.5. KoXxot, Kolchoi, 12, 96. 97 98, 99. KwXta/fot, Koliakoi. 101. KtDXiS, Kolis. 101. 102 Kolkei. 14, 92, 97, 99, 100, 114, 116. Kols, 14, 38, 39, 59, 60. 107, 243, 302, 314. 383. 448. 477. K61 (Singbhum), 418 Kd/iiap^KOfxdpia, 96, 102. Kongu, 18. Konkari, 36. Konkaiii, 2, 30. Korean, 617. K6ri. 13, Korkei, 99. KQpu, 102. Kota, 6, passim. Kotaur, 95. KoTTLapa, Kottiara, 95. KoTTovapcKT], Kottonarike. 95. Krishna, 12. Kshatriyas, 4, 5, 53, 111, 112. Ku. See Khond. Kuki, 327. 622. Kulasekhara, 114. Kumari, 117. Kuniarilabhatta, 4, 10. 116, 133. Kuralian, 48, 84. Kurd, 581, 594, 606. Kurinnbas, 32, 34, 93. Kynda, 95. Laccadives, 24. Laghmani, 377, 581. Lambadis, 38. Lappish. 56, 140, 149. 166, 185. 246, 295. 327, 332, 335. 341, 342, 346, 380, 391, 401, 612, 614, 615, 616, 617, 620, 621, 622. Lar. 382, 436, 609. Lares. 567. Lasian. 381. 618. Lassen, Professor, 23, 98, 109. Latchmaji, Mr, 35 Latin, 45, 46, 47, 52, 69. 73, 74, 148, 151, 152, 173, 182, 184. 206, 211, 231, 237, 242. 249. 251. 252. 262, 273. 276, 278, 285, 323, 342. 374. 375. 376, 377. 389. 401, 407. 408, 413 438, 463, 474, 508, 520, 534. 580. 582. 588. 589. 591, 593. 594, 595. 597, 598. 599. 600. 609. 610. 621. Lazian. 327. 332. Lekhani. 125. Lepsius. Professor. 338. 504. Lesghian. 614. 616. 617. 620. Lettish. 622. Lithuanian, 69, 148, 295, 300, 375. 376, 378, 439, 593. Lohitic. 395. Lubbies. 2. Avfj^LpiKT]. Lyniirice. 10. M M'abar, 21, 10]. Maccocalingae, 28, 29. MAdi, 343. 418. Madras, 3, 22, 93, 158, 267. Madras city, 8, 22, 32, 144. Madura, 12. 13, 19, 92, 102, 117, 122. Madura Sangam or College, 84. Magadha, 109, 115. 116. Magadht, 5. Magyar, 68, 104, 105, 110, 162, 184, 244, 246, 270, 277, 282, 285, 295, 305, 327. 332. 335, 338. 341. 342. 380, 381, 382, 390. 394, 409, 437, 438. Mahades-Gond, 341. Maharashtri, 5. Mah^, 3. Mah§ndra Male, 102. Maiso-Gothic, 338. Malabar, 8, 13, 20, 23, 24, 87, 88, 91, 94, 103, 117, 127, 143, 492, 504, 573, 576, 597, 615. Malay, 108, 272, 282, 613. Malaya, 575. Malayalam, 2. vassim. Mala-(y)-arasers, 38. Malayo-Polynesian, 375. Mal-dives, 24, 622. MaXe, Male, 23, 24, 103, 575. 622. Malers, 14, 35, 36. Manchu, 50, 65, 160, 180, 219, 232, 253, 281. 295, 296, 323, 332, 349, 350, 380, 381, 391, 409, 414, 482, 523. 615, 620, 621, 622. Mangalore, 31, 33. Manger, Dr, 401, 626. Manu, 4, 54, 112, 113, 115. Manyak, 392. Mappiilas, 2, 127. Marathi, 1, 2, 3, 25, 53, 56, 57, 60, 61, 141, 147, 155. 157. 158. 282. 390, 409, 414, 436, 509, 582, 591. 611. 612. Marcian, 104. Marias, 34. Marshall, Colonel, 33, 625. 626. Marwar, 25. Mathura, 12, 117. Mauritius, 7. Mech, 606. Media, 65. Medo-Persian, 65, 68. Medo-Seythian, 56. 610. Megasthenes, 12, 13. 26. 28. 98, 104. Medopa, 12. Metz, Eev. F., 33. 364, 507. 625, 626. Meyer, 495. Mikir, 245, 383. Minayeff, 89. Mingrelian, 327. 332. 381. Mirasi, 32. Miri, 245, 395. Mlechohas, 53. 116. Modogahngani, 28, 94. Mddovpa, Modura, 10. 13. 92. Mogling, Dr, 33. Mongolian, 50, 61, 65, 68, 76. 104. 160. 180. 219. 243. 246. 272, 274. 281. 295, 296, 332, 380, 381. 390. 391, 414. 432, 434. 482. 506. 512. 523. 524, 612, 614, 616, 617. 619. 620. 622, 623. Mons Maleus, 102. Mordivin, 67, 185, 281, 295, 335, 380, 382, 391, 477, 612, 61.3, 614, 615, 617, 621, 622. Moresby, Mr, 24. Motor, 382, 614. Muir, Dr, 57, 575. MUller. Professor rriedri, 69. Paulino, Fra, 21. Pauiidrakas, 4, Pegu, 7. Pehlvi. 389. 634. Penang, 7. WepL-yKapei, Perigkaerl, 97. Permian, 617, 619, 622. Persian. 24, 25, 55, 65. 67, 70. 71. 73. 148, 160. 220. 237. 247, 251, 263, 282, 294, 295, 327. 375. 376, 380. 381. 389. 401. 409. 410. 412. 422. 437. 438. 440. 441, 466. 477, 485, 495. 509. 510. 512, 525, 580. 581. .582. 591. 593. 594. 598, 600, 619, 620. Persian (Cuneiform). 89. Peut.nger Tables, 10, 26. 28. 94. 95. 96. 99, 101. Pho-nicia. 88. 90. 117. 118. 127. Pliny. 13. 26. 28, 29. 91. 92. 94. 95. 100. 101. 102. Pofligei.^ 99, 102. Wodoirepovpa, Podoperom-a, 103. Poligars, 111. 626. Polo. Marco. 21. 22, 99. Polynesian, 185, 247. 415. Pomponius Mela. 101. Pondicherry. 3, 144. Pope, Dr. 33, 34. 70. 122. 216, 222, 276, 364. 365, 369. 410, 507, 625. Portuguese. 8. 8. 21. 22, 25, 100. 141, 593. Porunei. 98, 115. Pott. Professor, 63. 184. Prakrit. 5. 9, 25, 60, 70, 78, 79, 80, 84, 85, 126, 147, 155. 412, 572, 575, 582, 597. Pritchard, Dr, 62. Prussian, 599. Ptolemy, 10. 13. 22. 28. 91, 92. 93. 94. 95. 96. 97. 98, 99, 100, 101, 102, 108. Puducottah, 111. Pulicat, 6, 21. 25. Pulindas. 5. Pundras, 5. Purali, 97. Pushtu, 282. Pyrard de la Val. 24. Q Quasi Qumuk, 382, 622. Quilon, 21, 23, 96. EajSndra Lal Mitra, Babu. 5. Eajmahai. 2. 6, passim. Eama. 21. 114. Eama Charita. 19. Kama's bow, 100, Eama's bridge, 21. 101. Eamanuja Acharya. 83. Eamayana. 48. 54. 88. 115. Eamfevaram. 100, 101. 114. Eamnad. 111. Eamusio, 22. Eamusis, 38. Eask, Professor, 61, 62. Kavenna, Geographer, 10. Eawlinson. Sir H.. 374. Eeddis. 25. Eed Sea. 127. Eeinaurd. 24. Eemusat. 447. Eenouf. M. Le Page, 89. Eheede. 90. Eoman. 10. 101. 103. 184. 361, 566. Eomance. 162. Eowlandson. 22. Eussian. 593. 619. Eutluk, 418. S Sabaras, 5. Sadras, 7. 8. S'akas, 4. Salem, 18. S'aiivahana, 487. Sa/xaj'ttiot. Samanaioi. 10, Samoiede, 246, 327, 332, 380. 391,613,616,619,621.622. Samoied-Ostiak, 246. S'ankara Acharya, 83. Sanskrit, 2, passim. Santai. 107. 418. 477. 614, 629. Satyavrata. 5. Sauraseni. 5. S'avara. 35. 418. Saxon. 593, 617. Scandinavian, 149. Schlegel, Dr Aug.. 71. 590, 622 Schott, 63. Scotch, 144. Scythian, 49, passim. Scythian (Cuneiform), 364. Scytia Dymirice, 10. 94. Secunderabad. 7. 'ZejivoL, Semnoi. 102. S'en-Damir. 6. Seoni-Gond, 329, 401, 507. S'era. 17. S'eralam, 17. Servian. 593. Setshuana, 159. Shen-Tamil. 45, 77. 78, 80, 82. Siamese, 124. Siberian, 394, 482, 614. Sindhi, 40, 56, 57, 158, 272, 280, 390, 436. Singapore, 7. Singhalese, 6, 7, 24, 109, 124. 380, 389, 436. Slavonian, 68, 69, 148, 375, 377, 378. 462. 574, 597, 616. Sojoten. 616. '^^(liKrjV, Solen. 98. Solomon, 88, 117. S'onagas, 3. Swpa, Sora, 93, Soran. 99. S'oramandalam. 141. Spanish. 41. .53, 164. S'ramana. 9. Stanley ofAlderley. T.ord. 22. Stephen of Byzantium, 90. 104. Stevenson. Eev. Dr. 52. 54, 56, 57. 247. 263. 301. 472. Strabo. 101. Suanian. 332. 381. S'udras, 83. 106, 107, 108. 109, 110, 112. Sumatra, 26. Sundara Pandya. 127. Surgutish. 281. Swedish. 621. Syrian Christians, 14, 16, 19, • 86. 87. 126. 128. 134. 164 281, 374. 482, 492. 606. 609. 619. Syiianian. 393. 615. Syro- Arabian. 172. TAmi.. 3. passim. Tamraparnt, 12, 14. 92. 97, 98. 102. 114. 115. 116. Tanjore, 13. 144. Tairpo^dvr], Taprobane, 98, 101. Taranatha, 9, 28. 29, 30, Tarshish, 88. Tatar. 60. 61. 149. 253, 318, 447. Tatar-Turkish. 66. 232. Telingana. 25. 54. 85. 86. Tellicberry, 95. Telugu. 4. passim. Temple. Sir E.. 34. Ten Commandments, 46, 47. Teutonic, 68, 70, 104, 242. 243. 300, 495, 510, 569. Thomas, Mr Edward, 117, 128, 128. Tibetans, 5, 89, 59, 75. 2451 281, 282, 296, 383, 612, 614. 635. Tinnevelly. 14. 86. 96. 97. 115, 127. 433. Tiruvalluvar. 48, 248. Ttyars, 108. Tratar, 295, 296. Travancore. 6. 14. 16, 86, 95, 96. 97. 127, 155. Trench. Archbishoo. 46. Tributary Mahals. 35. Trichendoor. 99. Trichinopoly. 13, 569, 572. 599. TpiyXvTTTov, TpiKiyyou, Triglupton, Trilingon, 28 Trilinga. 27. 28, 29. Trivandrum. 6, 20, 104. Trumpp, Dr Ernest, 25, 60. 147. 280. 287. Tuda. 33. passim. Tulu. 6. passim. Tungusian. 39. 61. 67. 323. 482. 591. 612. 616. 620. Turanian. 52. 61, 63, 64, 65. 126, 185, 385. 481. 488. 524. Turkish. 39. 42, 49, 55, 57, 61, 64, 67, 68, 76, 104, 172, 180, 197, 219, 243, 246, 272, 274, 281, 282. 295, 300, 307. 318, 327, 328. 332, 342, 381, 391, 394, 395, 403, 420, 447, 455, 463. 482. 484. 495. 503. 510. 512. 523. 542, 611, 612. 613, 616, 619. 620. Turks, 104, 447. Turvasu, 14. Tuticorin, 100. U Ugrian, 65, 66. 67, 68, 104, 162, 335. 338. 356, 382, 393. 395, 420, 477. 510, 520, 615, 620, 621. Uigur, 104, 612. TJmbrian, 155. Uranius, 90. (Jraon. 302. 323. 330. 332, 334, 335. 364, 387. 415. 417. 421, 436, 437. [Jreiyur. 13. 93. Vaisyas. 53. 112. 113. Vaithema's Travels, 22. Vaimiki. 48. Vanji. 97. Varaha-mihira. 9. 10, 30, 94. Vararuchi. 158 640 INDEX Veigei, 92. Veliaias. 32. 113. Vijaya. 11. 115, 116. Vijayanagara. 21. Vikramarka, 487. Vindhyas, 1, 108. Visvftmitra, 5. Vogul. 63. 327. 335. 338. 615. 616. 617, 620. 621. 622. Volgian, 617, 621, 622. Von Humboldt, W.. 73. Votiak. 327. 357, 380, 391. 621. Vrisbalas. 4. Vuller. 25. W W.\HL, 90. Wales. 107, Weber, Professor, 89. Welsh, 166, 377, 378. 591, 593, 620. Wilkms, Sir C, 42. Wilson, Professor H. H., 48. 53. 372. Xavikr, St Francis. 30. Xerxes. 107. Yadavas. 117. Yakute. 394. 614. Yama. 158, 456. Yarkand, 342, 621. Yaranas, 4. Yedo, 305. Yemen, 127. Yerkesian. 616. Yerukala. 418, 419. . Yule, Colonel, 16 note, 22. 23. 93. 99. 103. Zanzibar, 21. Zend, 69. 148. 160, 173, 243. 246. 247. 251, 295. 300, 368, 375, 376, 377, 378. 389. 412, 413, 438. 439. 582, 583. 597. 600. Ziegenbalg, 141. WILLIAM BRENDON AND SON, LTD. PRINTERS, PI.VMOUTH RETURN TO the circulation desk of any ~ I RE rp: University of California Library I or to the ^ NORTHERN REGIONAL LIBRARY FACILITY BIdg. 400, Richmond Field Station University of California Richmond, CA 94804-4698 ^ ALL BOOKS MAY BE RECALLED AFTER 7 DAYS 2-month loans may be renewed by calling — (510)642-6753 n 1-year loans may be recharged by bringing books to NRLF Renewals and recharges may be made 4 days CI prior to due date DUE AS STAMPED BELOW i/l/\R 1 9 1MB m 2 p. my n - - — F 20,000 (4/94) A GENERAL LIBRARY -U.C. BERKELEY BoooaimaM