INTRODUCTION TO THE 
 NATURAL HISTORY OF LANGUAGE^ 

 
 / 3Ex LJbria 
 C. K. OGDEN 
 
 INTRODUCTION 
 
 TO THE 
 
 NATURAL HISTORY 
 OF LANGUAGE 
 
 BY 
 
 T. G. TUCKER 
 
 LITT.D. (CAMB.), HON. LITT.D. (dUBLIn) 
 PROFESSOR OF CLASSICAL PHILOLOGY IN THE UNIVERSITY OF MELBOURNE 
 
 LONDON 
 
 BLACKIE & SON Limited, 50 OLD BAILEY E.G. 
 
 GLASGOW DUBLIN BOMBAY 
 
 I 908 
 
 V
 
 Y 
 9lI 
 
 T8 
 
 PREFACE 
 
 Considering the increasing interest which is taken by- 
 persons of intelligence in all branches of science that deal 
 with human development, it is a matter for some surprise, 
 and certainly for no small regret, that so little of an exoteric 
 nature is written upon the science of language. Philologists 
 are aware how vast an amount of linguistic research is being 
 patiently performed, and what excellent treatises of a 
 particular nature are now and again produced ; but for the 
 most part the results hardly spread beyond the sphere of 
 the specially initiated. Professed students of Comparative 
 Philology, to whom the general principles and main facts 
 of the science have become elementary knowledge, are 
 perhaps unaware how close is the corporation to which 
 even that degree of enlightenment is confined. They hardly 
 realise, with Max Muller and Whitney, that there is a world 
 of thoughtful readers who have little time or taste for the 
 minute details or metaphysical speculations of the subject, 
 but who nevertheless desire to obtain some comprehension 
 of the general course of linguistic development and of the 
 chief facts of the distribution and history of language. The 
 age of Volksbiichcr is rather beginning than ending. 
 
 The science of language is perhaps not one of the 
 easiest to handle with this class of readers in view. Even 
 an introductory sketch is apt to involve itself (as does the 
 otherwise excellent work of Sayce) in somewhat abstruse 
 
 V
 
 vi Natural History of Language 
 
 speculations concerning the relations between psychical and 
 physical operations. This rock was successfully shunned 
 by the brilliant expository genius of Max Miiller and by 
 the virile discretion of Whitney. On the other side lies 
 the Charybdis of excessive detail, and of this also both 
 writers steered clear. It happens, however, that, thanks to 
 the progress made since the appearance of the Lectures on 
 the Science of Language, Max Miiller's conceptions or state- 
 ments require frequent correction. On the other hand, 
 while it is hopeless to think of improving upon Whitney's 
 most admirable work in those matters with which he elected 
 to deal, there are sundry other topics which may at this 
 date be included with profit ; moreover, a new writer may 
 be permitted to see various matters in a different perspective 
 as to importance or interest, or he may approach them with 
 somewhat different views. 
 
 The present book is what its title implies ; it is an out- 
 line. So far as it is a student's book, it is meant for nothing 
 more than a first general survey of the field. It does not 
 hope to enlighten materially the professional philologist, but 
 only to bring him disciples. Nevertheless, it is no mere 
 compilation. The facts are necessarily in a considerable 
 measure the same facts which have done duty in other 
 works, whether general or special, but the writer trusts that 
 he has employed sufficient independence of thought in 
 weighing various theories against the material, and in offer- 
 ing, with all due diffidence, conclusions or speculations of 
 his own. 
 
 Where there exist books of a suitable kind dealing with 
 special topics in a sufficiently popular manner, the present 
 work has, in the interests of space, abbreviated the treat- 
 ment of those topics. For this reason comparatively little 
 is said upon the large and extremely interesting themes of
 
 Preface vii 
 
 Semantics and Analogy. Nor has it seemed any more 
 necessary at this date to give a history of linguistic science 
 than it would be to give a history of biological science in 
 an introduction to that study as now established. It has 
 also been thought best not to assume that readers are neces- 
 sarily conversant with Latin or Greek. 
 
 A writer so far removed from his printer, and dealing 
 with detail from so many languages, can hardly expect to 
 have avoided or corrected every error. But it is hoped that 
 any which perchance Jmmana pariini cavit natura will be 
 such as in no way to vitiate the context. 
 
 The author has to thank his friend. Dr. J. P. Wilson, 
 formerly Headmaster of the Presbyterian Ladies' College, 
 Melbourne, for his kindness in reading over the greater part 
 of the work when in manuscript with a view to testing its 
 lucidity and consistency of expression. A similar acknowledg- 
 ment is due to Professor R. T. Elliot, Worcester College, 
 Oxford, who read over the proofs. 
 
 The chief works which have been consulted in the 
 preparation of the book arc the following : — 
 
 Bleek : Compat-ative Grammar of South African Languages. 
 
 Brachet and Toynbee : Historical Frencli Grammar. 
 ^ Breal (and Postgate) : Semantics. 
 
 Brugmann and Delbriick : Grundriss der vergieichenden Gram- 
 matik der indogermanisclien SpracJjen. 
 
 Conway : The Italic Dialects. 
 
 Darmesteter : La Vie des Mots. 
 
 Delbriick : Introduction to the Science of Language. 
 
 Deniker : The Races of Man. 
 
 Diez : Grammaiix des Langues romancs. 
 
 Douse : Introductioji to the Gothic of Ulfilas. 
 
 Gabelenz : Die Spracfnvissenschafl. 
 
 Geiger : Developtnent of the Human Race. 
 X Giles : Alanual of Comparative Philology.
 
 viii Natural History of Language 
 
 Greenough and Kittredge : Words and their Ways in English 
 Speech. 
 
 Grober : Grundriss der r07nanischen Philologie. 
 Henry (Victor) : Comparative Grammar of English and German. 
 „ : Grammaire comparee du Grec et dii Latin (trs. Elliott). 
 X Hovelacque : Families of Language (\.x%. Keane). 
 
 Huxley : Essays. 
 X Jespersen : Progress in Language. 
 Keane : Ethnology. 
 
 „ : Man Fast and Fresent. 
 Kluge and Lutz : English Etymology. 
 
 Kretschmer : Einleitung in die Geschichte der griechischen 
 Sprache. 
 
 Lefevre : Race and Language. 
 
 Misteli (after Steinthal) : Charakteristik der hauptsdchlichsten 
 Typen des Sprachbaues. 
 
 X Miiller (Max) : Lectures on the Science of Laiiguage. 
 Miiller (F.) : Grundriss der Sprachwissenschaft. 
 Murray and Bradley : The Oxford English Dictionary. 
 Oertel : Lectures on the Science of Language. 
 Paul : Friticipien der Sprachgeschichte. 
 X Sayce : Lntroduction to the Science of Language. 
 „ : Frinciples of Comparative Fhilology. 
 Schrader (and Jevons) : Frehistoric Antiquities of the Aryan 
 Feoples. 
 
 Schrader : Reallexikon der indogermanischen Altertumshinde. 
 X Skeat : Frinciples of English Etymology. 
 
 Spencer (H.) : Essays Scientific and Speailative. 
 Strong, Logeman and Wheeler: The JListory of Language. 
 X Sweet : History of English Sounds. 
 Taylor (I.) : The Alphabet. 
 Tylor : Anthropology. 
 
 Wheeler : Analogy and the Scope of its Application in Language. 
 X Whitney : Language and the Study of Language. 
 „ : Life and Gro7vth of Language. 
 
 T. G. T. 
 
 Melbourne,///;/^ 1908.
 
 CONTENTS 
 
 CHAPTER I 
 
 Nature and Scope of the Subject. — Aims of the Science 
 of Language — meaning of ' language ' — utilities of the 
 study — name of the science — meaning of ' Comparative ' 
 Philology — the process scientific — divisions of the subject 
 (Phonology, Morphology, Comparative Syntax, Semasiology) 
 — the material ....•• i 
 
 CHAPTER n 
 
 Phonology : The Production of Speech - Sounds. — 
 Language consists of sounds — which are not uniformly main- 
 tained — physical causes of divergence. The Mechanism 
 OF Speech. — Breathing and speaking — the organs and 
 their employment. SOUNDS IN Detail. — Voice, pitch, 
 stress — production of vowels — sonants, consonants, and 
 semi-vowels — consonants classified — glides . . 17 
 
 CHAPTER HI 
 
 Speech and Writing : Evolution of Alphabets. — 
 Picture-writing and quipu — ideograms — ideograms become 
 phonograms — recognition of syllables^and of single sounds 
 
 Egyptian signs — Phoenician alphabet — Greek alphabets 
 
 — Latin alphabet — provenance of alphabets — summary . 44 
 
 CHAPTER IV 
 
 Classification of Language. — -'Genealogical' and 'Mor- 
 phological' Classification . . . . .70 
 
 ix
 
 X Natural History of Language 
 
 CHAPTER V 
 
 PAGE 
 
 Classification of Language {contimied). — Morphological 
 classification — uncompromising division impossible — the 
 old method of three classes — various types illustrated — • 
 attempted divisions — summary . . . .74 
 
 CHAPTER VI 
 
 Classification of Language {co7itiniied). — Genealogical 
 classification — determination of pedigree — by material 
 (roots and vocabulary) — by grammar — caution concerning 
 vocabulary (phonetic laws)— families discovered — illustra- 
 tions of method . . . . . -95 
 
 CHAPTER Vn 
 
 General Survey of Languages. — L The positional 
 (isolating) languages of Further Asia — especially Chinese 
 — question of common origin . . . -115 
 
 CHAPTER VHI 
 
 General Survey {continued.) — II. Agglutinating-Inflexional 
 tongues : (i.) Suffix-Agglutinating type — the Ural-Altaic 
 family — the Dravidian languages : (ii.) Prefix- Agglutinating 
 type — the Bantu family : (iii.) Prefix-, Infix-, and Suffix- 
 Agglutinating type — the Malayo-Polynesian family : (iv. ) 
 Other families or languages commonly classed as agglutin- 
 ating : (v.) Polysynthetic languages of America . .129 
 
 CHAPTER IX 
 
 General Survey {continued). — III. Amalgamating-Inflexional 
 tongues: A. Semitico-Hamitic Family — question of relation- 
 ship : (i) Semitic division — 'roots' and grammatical 
 method — members of the family : (ii.) Hamitic division — 
 compared with Semitic — members of the family. B. Indo- 
 European Family — the name — characteristics of method
 
 Contents xi 
 
 PAGE 
 
 branches of the family — Aryan (Indian and Eranian) 
 
 Armenian — Hellenic — Illyrian — Italic (and Romance 
 
 tongues) — Celtic — Teutonic — Baltic-Slavic (Slavo-Lithu- 
 anian) — synopsis of Indo-European languages . 158 
 
 CHAPTER X 
 
 Race and Language. — (i.) In General — language no necessary 
 criterion of race — language a social acquirement — ethno- 
 logical classification of peoples — no special type of language 
 identified with an ethnological type, (ii.) The Speakers 
 of Indo-European in Europe — mixed in race — the four 
 stocks of primitive Europe. (iii.) Celts and Celtic — • 
 confusion of the term 'Celt' — the linguistic meaning . 228 
 
 CHAPTER XI 
 
 Original Seat and Diffusion of the Indo-European. — 
 Method of inquiry — the Asiatic theory — European theory 
 of Schrader — historical arguments — linguistic arguments — 
 'wave-theory' of dialectal division — indications from 
 vocabulary — the primitive culture . . . .244 
 
 CHAPTER XH 
 
 Phonetic Change. — Sound and thought — the sound-picture 
 (average of environment) — reproduced in utterance — 
 individual deviations — shifting of the average articulation 
 — counteracting forces inadequate — growth of dialect in a 
 group — interlinking of dialects — phonetic change among 
 cultured peoples — rarely confined to one sound at a time — 
 effects of neighbouring aliens — of complete geographical 
 separation — of climate — of mixture with aliens — summary 268 
 
 CHAPTER Xni 
 
 Phonetic L.wvs. — Instances of phonetic ' laws '—definition of 
 a ' phonetic law ' — question of regularity of law — exceptions 
 — arguments for regularity. Grimm's Law. — Statement-- 
 exceptions — method of application . . . 3°3 
 
 r^.
 
 xii Natural History of Language 
 
 CHAPTER XIV 
 
 PAGE 
 
 General Phonetic Tendencies in Language. — General 
 tendency towards ease — 'ease' is relative — some sounds 
 absolutely less effortful than others — instances of economy 
 of effort — effects of accent . . . .329 
 
 CHAPTER XV 
 
 Application of Phonology to Etymologies. — Sound 
 etymology and knowledge of phonetic principles — special 
 phonetic laws of language concerned — same original : 
 different forms in different languages ; different forms in 
 the same language — words of different origin : the same 
 shape in different languages ; the same shape in the same 
 language . . . ■ ■ • -35 5 
 
 CHAPTER XVI 
 
 Ch.anges of Meaning. — Tendency to change meaning— laws 
 of change not yet discovered- — changes of meaning classified : 
 — changes produced by specialising or narrowing ; by 
 generaUsing or widening ; by shifting or transference — 
 classification of motives for change . -373 
 
 CHAPTER XVn 
 
 The Development of Language. — Evolutionary principle ; 
 easing and economy ; loss of flexion — first devices of lan- 
 guage, not polysynthetic — naming and pointing 'roots' — 
 origin of pointing elements, deictic elements become 
 suffixes — suffixing replaced by preposing — interaction of 
 psychology and phonetic decay ; different notes of phonetic 
 decay — historical phenomena as arguments for the pre- 
 historic — unconnected monosyllables beginnings of speech — 
 sound groups (' roots ') : origins, meanings, theories — word- 
 building from old material inexhaustible : creation of new 
 roots not provable — creation of new words, meaning of — 
 true synonym not created by a language — causes of loss of 
 words — effects of the analogical instinct . . . 397 
 
 Index . ■ • • • • -453
 
 INTRODUCTION 
 
 TO THE 
 
 NATURAL HISTORY OF LANGUAGE 
 
 CHAPTER I 
 
 NATURE AND SCOPE OF THE SUBJECT 
 
 Apart from the study of a special language — whether for 
 practical purposes as a means of communication, or for the 
 comprehension of works of literature in that tongue — there 
 is a science of 'language' in itself Just as there is a 
 science dealing with life, and thence called Biology, or as 
 there is a science dealing with the earth, and thence called 
 Geology, so there is a science dealing with language. Of its 
 proper title something will be said later. 
 
 The methods and scope of this study are analogous to 
 those of the sciences named. Geology, for example, in- 
 vestigates the structure of the globe and its history. It 
 considers the component parts of the earth and their relations 
 to each other, and inquires into " the successive changes to 
 which their present condition and positions are due." 
 Biology, again, takes organised beings, both animals and 
 plants, and inquires into their structure, physiology, develop- 
 ment, and distribution. In brief, the questions which these 
 sciences ask are of the following and similar nature. Of 
 what does the earth consist? How did it come to possess 
 
 , (B 884) B
 
 J- 
 
 2 Natural History of Language 
 
 the formations which it now shows ? What has been the 
 history or development of the various forms of life which we 
 find around us ? And in seeking answers to these and other 
 questions there is applied the scientific method. A science 
 is no science unless its basis is the exact observation of all 
 the ascertainable facts, and unless it draws its conclusions or 
 inferences with scrupulous regard to those facts. 
 
 Similarly, in its ideal conception, the Science of Language 
 should deal with all the material and phenomena of language, 
 or, in other words, with all the facts of speech. With exact 
 observation of these, and with cautious argument therefrom, 
 it should investigate the structure of ascertainable languages, 
 their history and transformations, their relations to each 
 other, their differences, and the causes or laws of those 
 differences. The questions which it should ask are such as 
 the following : — From what beginnings did speech develop 
 itself? By what different processes have various languages 
 sought the means of expressing ideas ? How and why have 
 kindred speeches diverged from each other ? Are there any 
 discoverable laws of change, either for language in general 
 or for an individual tongue ? As Whitney has expressed 
 it, the whole subject of linguistic investigation may be 
 conveniently summed up in the single inquiry, " Why do 
 we speak as we do ? " 
 
 Before proceeding further it is necessary to define 
 ' language.' Language is uttered sound, but not all 
 uttered sounds are language. Tylor's definition, which 
 hardly admits of improvement, is that language is " the 
 expression of ideas by means of articulate sounds habitually 
 ■" allotted to those ideas." Ideas may be communicated also 
 by gesture, but gesture is not language in the sense in 
 which this science treats of it. Language is speech, and 
 speech has always been regarded, except by fable-writers, 
 as the peculiar possession of mankind. If we could gather 
 sufficient trustworthy data concerning any sounds by which 
 animals habitually express ideas, these also might naturally
 
 Nature and Scope of the Subject 3 
 
 come within the scope of the same science ; but at present 
 this matter Hes quite outside practical consideration. Any 
 human utterances, however, which fall within the above 
 definition of language are material for the science. No 
 lanefuasre is too barbarous or obscure. A student who deals 
 with language in general in and for itself, is (says F. Miiller), 
 as compared with the practical linguist or the student of 
 some special language and literature, very much like the 
 botanist as compared with the gardener ; to him wild plants 
 and weeds have their interest as much as useful plants or 
 flowers. To the botanist the facts of docks are as important 
 as the facts of roses ; to the student of language the facts of 
 aboriginal Australian or Eskimo are as important as those 
 of Greek or French. 
 
 Apart from any purpose of direct utility (in the narrower 
 sense of that much-abused word), there must always exist a 
 natural interest in this branch of study. In any case it 
 would be enough to plead with Aristotle that " the use of 
 the intelligence is highly agreeable, not only to philosophers, 
 but to other people as well." Human beings, when properly 
 constituted, have an innate and compelling desire to know 
 as much as possible of the truth, even concerning things of 
 which they cannot determine the fate, or which do not 
 appreciably touch their own destinies. To all such it is a 
 legitimate satisfaction to feel that they are not entertaining 
 erroneous notions as to the interrelations of languages, or 
 the laws of their progress, or the question whether, for 
 example, human speech is improving or deteriorating. An 
 enlightened way of looking at the phenomena of pronuncia- 
 tion, grammar, or the dictionary, is a valuable endowment 
 for any one who loves culture for its own sake. 
 
 To many minds such a reason for the study is alone 
 ample. Others will find a fuller satisfaction in the con- 
 sideration that in language is to be met the most complete 
 record of the expansive efforts of human thinking. W. von 
 Humboldt has well said that language is the outcome of
 
 4 Natural History of Language 
 
 " the eternal striving of the human spirit to make the 
 articulated sound equal to the expression of the thought," 
 It is the manifestation of the human mind through an 
 audible medium, and its history, therefore, throws much 
 light on the history of that mind. The understanding of 
 the origin, history, and development of language must 
 accordingly form a part of the material for arriving at any 
 great synthetical philosophy. 
 
 But, meanwhile, a training in the science of language 
 possesses a more immediate and tangible utility for the 
 student of any given individual tongue. An intelligent 
 grasp of the operations and principles underlying all 
 language is an excellent equipment for one who is brought 
 face to face with the peculiarities of accidence, syntax, and 
 turns of expression which belong to a particular speech. As 
 M. Henry observes, " The grammar of any language, con- 
 sidered by itself, seems a purely empirical collection of 
 arbitrary rules, traversed by exceptions still more arbitrary, 
 which it is satisfied with formulating, without being able to 
 give their raison d'etre.'' For the student of the science of 
 language the sense of arbitrariness or capriciousness is in 
 such cases almost, if not wholly, removed. In many 
 instances he can readily discern the reason for a phenomenon ; 
 sometimes he can prophesy its appearance. Where he 
 cannot actually discover the precise reason, he is sure that 
 a perfectly natural one exists. A student of Greek, and of 
 Greek alone, finds in different writers different forms of the 
 same words. Sometimes he is informed that they represent 
 various dialects ; sometimes they are put down to ' poetic 
 licence.' These statements he may perhaps retain by an 
 effort of memory. Very unlike is the case of the student 
 of Comparative Philology (as the science of language is 
 commonly called in the English academic world). To him 
 the knowledge that the Ionic Greek said Tov<i aypov^ (tous 
 agrous), the Doric Greek tw? aypco^ {tos agros), and the 
 Lesbian Greek roi^ a'ypoi<; {tois agrois) for the plural objective 
 case, ' the fields,' is not merely so much isolated and dead
 
 Nature and Scope of the Stibject 5 
 
 information. He realises that behind these phenomena 
 there is some intelh'gible cause, physical or mental, which 
 will be found to operate again in other matters affecting 
 those dialects. Inquiry reveals the fact that in a more 
 primitive Greek the words were rov^ aypovi (tons agrons), 
 and that the three divergent forms represent the several 
 pronunciations into which tons agrons was slightly corrupted 
 by the three divisions of people concerned. All three 
 abandoned, though gradually and not deliberately, the nasal 
 in the syllable -ons, and the result, which was slightly 
 different in each case, depended on certain peculiar tendencies 
 of each division. Those tendencies we must expect (and we 
 shall find) to repeat themselves on every occasion on which 
 they can come into play. In other words, they are subject 
 to rule, and the cause of the rule is known. Similarly, a 
 student of Latin is no longer puzzled at the statement that 
 a so-called genitive (or 'of case) magnl can be used for 
 the price ' at ' which, nor is he surprised that Romae can 
 mean equally 'of Rome,' 'at Rome,' and 'for Rome.' 
 The empirical teacher of Latin is (or was) wont to speak of 
 the genitive plural demn (' of gods ') as ' a contraction of 
 deoruvi' Our better-equipped student will know, not only 
 that such a contraction is impossible, but also that the two 
 words are distinct formations, and that deum is in reality the 
 older form of the two. In dealing with the English language, 
 he will be prevented from offering confused or baseless 
 explanations of the phenomenon that a man of Somerset 
 or the Lowlands uses a pronunciation and vocabulary which 
 are not those of literary English ; he will also be on his 
 guard against the unsubstantiated etymologies which still too 
 often figure in both literature and conversation. 
 
 Nor, again, is it to be despised that a closer examination 
 of languages reveals analogies and similarities of principle 
 which bring us into more sympathetic understanding of 
 foreign nations. The idioms of foreigners often appear 
 strange or unaccountable, if not infantile, grotesque, or 
 ludicrous. By an illusion, which it is hard to shake off, we
 
 6 Natural History of Language 
 
 are consequently apt to regard their mental processes as 
 peculiar, and occasionally childish. The scientific study of 
 language does more than correct this error. It brings home 
 the truth that the human mind, as disclosed in its efforts 
 towards expression, is, or has been, very similar in its 
 workings, even in the case of nations the most widely re- 
 removed in situation, race, and colour. In Japanese the 
 phrase ' I want water ' is viizu ga hoshiu gozaiinasti, of 
 which the literal rendering is ' water-of desirous am.' To 
 English ears this sounds sufficiently extraordinary, and it 
 might easily be inferred that there was some peculiar ' idio- 
 syncrasy of the Far East ' in the mental operation which 
 evolved the expression. Nevertheless it in no way differs 
 either in structure or order from the Latin aquae ctipidus sunt. 
 If the Japanese for ' a thing to be done ' is literally ' a do- 
 must thing,' there is again no real distinction between this 
 form of phrase and the Greek %/o»}/ia irpaKrkov {cJirema 
 prakteon), the Latin res faeienda, and such English as ' a 
 do-able thing.' 
 
 The fittest name to be given to the science of language 
 is still undetermined. Science of Language explains itself 
 and is open to no objection but that of cumbrousness. 
 What it lacks is the compactness which belongs to such 
 terms as Geology, Physiology, and the like. Comparative 
 Grammar, again, though theoretically correct, is apt to be 
 misleading in consequence of the narrower sense commonly 
 attached in English to the word ' grammar.' It seems 
 better, therefore, to reserve this term for a more limited 
 comparison of structural principles and methods. The 
 German word Sprachzvissenschaft might be translated 
 * speech-science,' but in that form the title is unlikely to 
 meet with any recognition. Linguistic (which as Linguistigue 
 once found some favour in France) has proved abortive ; 
 and Glottolog}', the most satisfactory term in respect both of 
 formation and meaning (' the science of tongues '), has hitherto 
 missed that acceptance which seems to denote ideal aptness.
 
 Nature and Scope of the Subject 7 
 
 The academic term commonly employed in English is Com- 
 parative Philology, and there is perhaps a growing tendency 
 to apply a lax and undesirable use of the single word 
 Philology in the same sense. The matter is somewhat com- 
 plicated by the fact that in its proper acceptation ' Philology ' 
 in English, as in French, means literary and linguistic study 
 and learning, concerned with the writings of a people as 
 representative of its thought, style, culture, and art. 
 ' Philology,' without qualification, has reference to such 
 inquiries in general, while in the case of a special language 
 and its literature we employ the terms ' Greek Philology,' 
 ' German Philology,' and the like. It is to be regretted 
 that in England there should be so strong a disposition to 
 shift the meaning of this useful word to that of the science 
 of language as above defined. It happens, however, that 
 scientific investigation into language in general was first set 
 on foot by ' philologers ' in the older and more correct 
 sense, and in particular by students of Greek and Latin 
 philology. Inquiries into these particular languages led to 
 inquiries into their relationships with others, to comparisons 
 of structures and vocabularies, and thence to researches into 
 the principles of language in general. ' Philology,' it was 
 considered, thus became ' Comparative,' by including in its 
 survey, for purposes of comparison, whole numbers of 
 languages, instead of regarding only a particular speech. 
 Unfortunately, the scope and aim of the new study were 
 not identical with the scope and aim of ' philology ' when 
 applied, as above described, to Greek, Latin, or any other 
 individual tongue and its literature. 
 
 Properly speaking, the epithet 'Comparative ' accompany- 
 ing the name of a science is pleonastic. All truly scientific 
 investigation must be comparative. Says E. Scherer, 
 " Science is made up of general facts ; scientific knowledge 
 is the formation of groups and the establishment of laws ; 
 it elicits the general out of the particular." The study of 
 some special language, such as Latin or Hindustani, can
 
 8 Natural History of Language 
 
 teach us little or nothing concerning the laws and operations 
 of language at large, and, on the other hand, our knowledge 
 of the special language, Latin or Hindustani, itself is but 
 empirical and relatively unintelligent, unless we are in a 
 position to relate its phenomena to those of other tongues, 
 and so refer them with some certainty to their causes. To 
 take a given language — English for example — as it now 
 exists, and register its grammatical forms and conditions, 
 is the business of purely descriptive grammar. To collect its 
 forms and conditions as they are discoverable at different 
 dates (as, for instance, in the days of Spenser, then of 
 Chaucer, and earlier still), and register their successive 
 changes, is the work of historical grammar. The latter pro- 
 cess is manifestly ' comparative ' in a certain sense, inasmuch 
 as it involves the comparison of one stage of a language 
 with another stage of the same. Historical grammar of 
 individual tongues manifestly plays a necessary part in the 
 science of language. Conventionally, however, ' com- 
 parative ' study of language is understood to be rather 
 the study of languages side by side. It is not, indeed, 
 always easy to separate the processes. Thus the historical 
 grammar of French carries us back step by step to Latin, 
 which is in reality the earlier shape, rather than the 
 * parent,' of French. Yet, conventionally and practically, 
 Latin and French are languages so different that they are 
 proper objects of ' comparative ' grammar in the larger sense. 
 This wider comparison is compelled to step in when the 
 ' historical ' grammar of a language has exhausted its 
 material and done its work. 
 
 In studying a particular language we find an idiom or 
 form of such and such a nature. How it came to exist in 
 that shape can only be gathered from a comparison, first, 
 with the language at an earlier stage, then, if that process is 
 insufficient (as it very commonly is), with the forms and 
 idioms of kindred languages. Attempts to explain such 
 phenomena by ingenious guesswork, or by the application 
 of what the explainer is pleased to call logic, are most
 
 Nature and Scope of the Subject 9 
 
 frequently disastrous. In old French the nominative singular 
 of mur was rnurs, while the plural was miir. In modern 
 French the case is reversed. The explanation is to be 
 found, not in a study of French alone, but in a comparison 
 with the Latin from which it came. In Latin the nominative 
 singular is inurus (which became pronounced in France as 
 murs\ and the nominative plural is vniri (pronounced in 
 France as mur). Meanwhile the accusative singular is 
 miirmn, and this was pronounced mur in France, the plural 
 being inferos, which became murs in Franco-Roman mouths. 
 We thus obtain the following scheme for early French, viz. : — 
 
 Nominative Sing. ^;z«rx ( = murus). Plur. tnicr { = Vl\\xx\). 
 
 Accusative Sing, mur ( = murum). Plur. murs ( = muros). 
 
 Next it happened that, through the greater frequency 
 with which the accusative forms were used, the French 
 adopted that one case throughout the singular and also that 
 one case throughout the plural. The nominatives having 
 thus disappeared, we are met by a singular and plural form, 
 strictly accusatives, which wrongly appear to be a merely 
 arbitrary reversal of those previously used. Again, in 
 French the past participle is made to agree with the object 
 when the object precedes, though not when it follows. Thus 
 it is correct to say fai vu la femme, but la femme quefai 
 vue. That this rule is only artificially maintained and often 
 violated is shown by the grammatical concessions authorised 
 in the year 1900 by the French Minister of Education. It 
 might possibly be fancied that the distinction originated in 
 some inconsistent refinement on the part of early literary 
 authorities in France. Unhappily for this theory we find 
 precisely the same phenomenon in the sister language, 
 Italian, which says lio veduto la donna, but la donna che ho 
 veduta. Obviously the French idiom is no mere caprice of 
 the French, but its origin must be sought in the late Latin, 
 which was the common parent of both PVench and Italian. 
 The same cautious comparative method of inquiry is every- 
 where necessary. From modern English we go back step
 
 lo Natural History of Language 
 
 by step to Anglo-Saxon ; this done, our understanding is 
 checked, until we venture afield into a comparison with 
 German and the other Teutonic speeches, and in particular 
 with the oldest extant form of such a speech, the Gothic. 
 These kindred tongues having yielded their quota of informa- 
 tion and enlightenment, we proceed to a further comparison 
 with other languages, still related, but in a less close degree, 
 such as Latin, Greek, and Sanskrit. In this way we are 
 carried nearer and nearer to common sources of form and 
 idiom. Nevertheless what we are so far discovering is but 
 the principles and methods of one particular family of 
 languages, the Indo-European. To deduce principles of 
 language in general, and to speculate rationally upon its 
 primeval origins, demand a yet wider comparison of family 
 with family and of types the most divergent. 
 
 Whatever name may be given to the science (and 
 Glottology is perhaps superior to any other), its processes 
 must vindicate their claim to be called scientific. This 
 truth has been recognised but recently. Though there is 
 not only room, but necessity, for hypotheses and speculation, 
 these must be based upon an ample collection and cautious 
 classification of material. The hypotheses so propounded 
 must also submit to the test of every new example. 
 General principles of language are not to be deduced from 
 a few particular instances ; languages are not to be declared 
 akin in virtue of a few superficial resemblances of vocabulary, 
 however curious or striking ; etymologies or derivations of 
 words are not to be determined by guesswork, however 
 specious. Before accepting an etymology, for example, it 
 is necessary to examine it, not in the light of apparent 
 probability or plausibility, but with critical application of rule 
 and method. Thus the word must be traced back to its 
 earliest ascertainable form and use within the history of the 
 language ; its cognates in other tongues must be found, not 
 from exterior resemblance, but by the application of certain 
 laws of correspondence which have been discovered for the
 
 Nature mid Scope of the Subject 1 1 
 
 languages in question. Few would hesitate to declare that 
 sorry is ' obviously ' the adjective of sorrozv. In actual use 
 it has indeed long served that purpose. Nevertheless the 
 etymologies of the two words are distinct. Sorry is the 
 Anglo-Saxon sdrig ('sore'), while sorrow represents sorg 
 ('care'). In Guatemalan Indian popol means ' public,' and 
 it might easily be fancied that the word is akin to, or 
 borrowed from, the Latin populus (' the people '). Inquiry, 
 however, shows that popol is simply derived from pop, the 
 common ' mat ' of the assembled company. The Greek 
 ^609 {theos), ' God,' is identical in meaning with the very 
 similar word dens in Latin. Nevertheless the scientific 
 etymologists wholly refuse to connect the two. In the 
 multitude of words which form the vocabularies of two 
 languages it is not surprising that there should accidentally 
 occur one or two close coincidences. The marvel would 
 rather be if it were otherwise.^ The Greek for call is 
 Kokelv {kal-ein\ but the English word is neither derived 
 from the Greek nor in any way akin to it. The Aztec 
 Indian for ' temple ' was teocalli, but any one who chose to 
 connect this with the Greek words Oeov koXui {tJieoii kalid), 
 ' hut of a god,' would be finding a mare's nest. This 
 rigorous attitude towards language is comparatively recent. 
 The absurdities once committed by misplaced ingenuity in 
 its attempts to relate English to Hebrew or Hebrew to 
 Greek, as well as in the etymologies of individual words, 
 will be better considered when we come to deal with 
 
 1 Much absurdity is occasionally perpetrated by persons untrained in the science, 
 who are struck by the purely accidental resemblance or even identity which two 
 entirely unconnected languages may happen to exhibit in some word or other. 
 Thus the chance correspondence in both sound and sense between the Polynesian 
 word matt, ' eye,' and the modern Greek tiiaii, is enough to set them speculating 
 upon a probable unity of origin of the languages concerned, although no further 
 support whatever for such an idle notion can be found in either the vocabularies 
 or the structures of those tongues. Instead of reflecting that, among the infinite 
 number of words constructed by human beings out of comparatively few elemental 
 sounds, the wonder is that there are not more numerous examples of accidental 
 coincidence, they are tempted by the siren of the romantic into visionary theories 
 on a level with Anglo-Israelism.
 
 1 2 Natural History of Language 
 
 the principles of etymology. There was ample justifica- 
 tion for Voltaire's caustic remark that " etymology is a 
 science in which the vowels count for nothing and the con- 
 sonants for very little." Among examples of facile, but 
 quite baseless, derivation, Brachet cites such French instances 
 as those of cordonnier (' shoemaker ') from cordon (' twist ' 
 or ' string '), ecuyer (' squire ' or ' rider ') from Latin equus 
 (' horse '). These are cases in which the empiric would 
 and did satisfy himself that the parentage was obvious. 
 Yet in point of fact cordonnier was originally one who 
 worked in Cordovan leather, while ecuyer was the scntarius 
 or ' shield-bearer,' and neither cordon nor equus played any 
 part whatever in the matter. 
 
 In the development of any science there comes an in- 
 evitable partition of the field, and sections of the science, or 
 sub-sciences, are distinguished by a special nomenclature. 
 It is so with the science of language. Language is spoken 
 in sentences ; in these the various parts bear a structural 
 (or grammatical) relation to each other ; the individual 
 words are also formed and modified in themselves on certain 
 principles ; furthermore, each is composed of certain sounds ; 
 and, once again, particular combinations of sounds express 
 certain meanings. It follows that, in the investigation of 
 language, it is necessary to treat (i) sounds, (2) word-forma- 
 tion, (3) syntax, (4) developments of meaning. Thus in 
 Jie sowed the seed in the field, as compared with the 
 Latin semen in agro sevit (where the root se appears in each 
 tongue), we have to consider (i) the exact relations of the 
 sounds sow, see, and se ; (2) the principles of formation shown 
 in the various words soxv-cd, see-d, sc-mcn, se-vit ; (3) the 
 different syntactical methods of the English and Latin as 
 shown in our use of the pronoun and article, in the 
 absence of inflexion from our noun, and in our fixed order of 
 the words ; (4) the development of the meaning ' sow ' from 
 the root sc, which originally meant ' to throw.' Add that 
 the English word etymologically corresponding to agro is
 
 Nature mid Scope of the Subject 1 3 
 
 acre, which has narrowed its meaning to that of a small field 
 of definite extent, and thence to that of a measure of area. 
 If we further take the Greek words ^/xa eV tw urjpM 17/ce 
 {Jiema en tdi agroi hcke), which are etymologically akin to 
 the Latin, viz. r^fia {hcmd) to semen, rjKe (Jie-ke) to se-vit, iv 
 (en) to z'n, and dypS (agroi) to agro, we are met with (i) 
 difference of sound (e.g. /le for se, -ma for -men), (2) difference 
 of formation (e.g. -ke instead of -vit\ (3) difference of syntax 
 (since agro is the ablative case and agrdi the dative), (4) 
 difference of meaning, inasmuch as ^/xa (which is an old and 
 very rare word) does not mean ' seed ' but ' javelin,' and r^/ce 
 does not mean ' sowed ' but ' threw.' In each case the com- 
 pared languages, despite their common origin, have thus 
 come to diverge in four respects. Corresponding respectively 
 to these four heads we have the four divisions of the science, 
 namely. Phonology (Greek phone, ' sound '), MorpJiology (Greek 
 tnorphe, ' shape '), Comparative Syntax, and Semasiology 
 (Greek semasid, ' meaning '). The last is often incorrectly 
 styled Sematology. 
 
 Phonology treats of the sounds used in speech, their 
 physical production, and their modifications, whether singly 
 or in combination. MorpJiology treats of the formation of 
 words by the union of their several constituent parts, and 
 therefore particularly with the modifications involved in what 
 is known as declension or conjugation, or, more comprehen- 
 sively, as accidence. Comparative Syntax next discusses the 
 manner in which such words, being thus formed, are arranged 
 and related to each other in order to form sentences expressive 
 of complete thoughts. Semasiology is concerned with the 
 changes and development of meanings, and should include 
 in its ultimate aim the discussion of the way in which the 
 earliest meanings came to be attached to words. At present 
 no very great progress can be said to have been made in 
 this domain, beyond the collection of the histories of 
 individual words. The subject is psychologically very 
 intricate, the shifting of meaning having often depended 
 on some subtle association of ideas to which it is now hard
 
 1 4 Natural History of Language 
 
 or impossible to find the clue. The evolution is sufficiently 
 clear in cases like those of Latin tectum ('roof'), tegula 
 ('tile'), English thatcJi (a 'roof of a certain material), deck 
 (a word of Low German for the 'roof or 'covering' of a 
 ship) — all derived from the root teg, ' to cover.' But it might 
 easily have been lost in codify or contemplate. The connec- 
 tion of codify with the Latin codex, ' a piece of wood,' could 
 hardly have been divined, and, if in some way divined, it 
 would have been easy to misconceive the steps in the pro- 
 gress of the meaning. We happen to know that wood was 
 used as material to be written upon, and the meanings 
 ' wooden tablet,' ' writing tablet,' ' book,' and then ' book of 
 laws ' in particular, are unequivocally traceable thereto. The 
 connection between library and the Latin liber (' bark ') is 
 known ; but, if not known, it could scarcely now be 
 guessed.^ 
 
 It will be observed that the term Etymology has not been 
 used as the name of any one of these divisions. By some 
 the word is applied to that part of Morphology (as above 
 defined) which is concerned with the derivation and com- 
 position of ' word-stems ' from ' roots,' while the name 
 ' morphology ' is restricted to the study of the terminational 
 modifications or flexions. Thus, in this acceptation, 
 etymology would deal with the construction of ' themes ' 
 or ' bases,' but Morphology with the varying verb and 
 noun terminations. In the Latin sapientissimiis (' wisest ') 
 Etymology would be concerned with the putting together 
 of the elements sap-i-ent-is-s-imu-, while Morphology would 
 deal with the variable inflexions -s, -m, -o, etc., which belong 
 to the declension of that word. It should be obvious that 
 this distinction is quite irrational, and, inasmuch as it is not 
 recognised in English usage, we need make no further refer- 
 ence to it. The New English Dictionary excellently defines 
 etymology as " the process of tracing out and describing 
 the elements of a word, with their modifications of form and 
 sense," while the etymology of a particular word consists of 
 ^ The relationship of book to beech is dubious.
 
 Nature and Scope of the Subject 1 5 
 
 " the facts relating to the formation or derivation of that 
 word." Etymology is thus a process of retracing certain 
 facts which are gathered from phonology, morphology, and 
 semasiology combined. The etymology of a given word is 
 a discovery which results from the knowledge obtained from 
 all these departments. Etymology is thus not a division of 
 Comparative Philology, but an application of its teachings. 
 
 From an ideal point of view the material necessary for 
 the perfect satisfaction of the science would, as Paul remarks, 
 consist of " all the groups of sound ever spoken or repre- 
 sented, together with the associated ideas of which they are 
 the symbols." Such a complete chain of the facts it is, 
 however, impossible to obtain. For perfection of knowledge 
 we must hear the sounds themselves, since language proper is 
 sound, not writing. But sound, once uttered, has no con- 
 tinuance ; its representation in writing is inadequate and 
 often misleading ; and in any case the record only dates 
 from comparatively recent times. It is with the data of 
 language somewhat as with the imperfect data of evolution 
 in biology. But, whereas palaeontology can restore many 
 lost links, such as the mesoJiippus in the history of the horse, 
 or the toothed birds, a palaeontology of language in this 
 sense has practically no existence. The hieroglyphs, 
 cuneiform inscriptions, and the like are, so far as they are 
 strictly decipherable, nothing less than cultivated speeches 
 of no very ancient date, and can therefore only be put side 
 by side with such languages as Sanskrit and the oldest 
 Chinese. The whole field of known languages, living and 
 dead, supplies us indeed with a view of related languages 
 differently modified and in different stages of development, 
 but the origins and rudimentary forms of language are lost 
 for ever, or are only hypothetically recoverable by induction. 
 Moreover, amid the older languages known to us there 
 existed many others which are completely lost, or of which 
 only the faintest traces remain. What light these might 
 have thrown upon the speeches for which we do possess
 
 1 6 Natural History of Language 
 
 data, we cannot now say. Nor even in the case of existing 
 records can we by any means be certain that in all cases we 
 are attaching the right sounds and the exact significations 
 to the material before us. We are, for instance, uncertain 
 of the exact pronunciation of many sounds in the Greek of 
 classical Athens, and its accentuation (which is a matter of 
 great importance in linguistic history) is still something 
 of a puzzle. Again, the written language has never been 
 an exact record of the language of the street, whether in 
 respect of pronunciation, of vocabulary, or of grammar. 
 The material of the glottologist is therefore both recent and 
 otherwise defective. It consists of the sounds, expressive 
 of ideas, uttered throughout the world at the present day, 
 and of the written symbols for such sounds which are any- 
 where discoverable from the past. It is therefore futile to 
 expect a complete solution of all the questions which can be 
 asked concerning language. 
 
 Yet even this material is manifestly too vast for any in- 
 dividual to cope with. A hundredfold Mezzofanti would 
 still fall short of being a universal linguist. Here, as in 
 other sciences, there must be a division of labour. Some 
 students will investigate in one field, such as among the 
 languages of South Africa or of Polynesia ; some in another, 
 such as among the classical tongues or the monosyllabic 
 speeches of the Far East. The specialist workers will thus 
 collect the data, discovering the facts and laws of particular 
 languages ; and upon these another class of students, con- 
 tent to deal with types, will build classifications with a view 
 to deducing general laws and constructing general theories. 
 The man of scientific habit, who generalises upon the 
 collections made from a wide area of languages of different 
 families and structures, is more to be trusted in respect of 
 linguistic principles than the specialist, however deeply 
 learned, who confines himself to the minute study of a 
 special language or of one or two related tongues. The 
 former, in fact, is the glottologist, or ' comparative 
 philologist.'
 
 CHAPTER II 
 
 PHONOLOGY 
 
 TJie Production of Speech- Soimds 
 
 Language proper consists of sounds possessed of meaning. 
 It is true that such sounds have in most languages been 
 reduced to certain symbols which constitute writing, and by 
 which, though they have always been inadequate from a 
 phonetic point of view, a meaning may be communicated 
 almost^ as perfectly to the eye as by actual speech to the 
 ear. Nevertheless writing is in its origin, and for the most 
 part in its actual use, a makeshift for speaking, and no one 
 is likely to suppose that it can claim equality with the spoken 
 sounds as representing ' language.' The older school of 
 philologists were too apt to confine their investigations to 
 written records instead of paying primary attention to the 
 living speech ; and it can hardly be said that the error is 
 even now extinct. Where the dead symbols are all that 
 remains of a language it is for students to base upon them 
 such conclusions as that material warrants ; but it is self- 
 evident that a real and thorough knowledge of either the 
 phonology or the grammar of a language (in the proper 
 sense of that term) at any moment can only be derived from 
 hearing the exact sounds and marking the exact uses cur- 
 rent among the representative generality of its speakers. 
 Literature artificially narrows the grammatical range actually 
 
 ^ But notjquite. The exact tone and the degree of emphasis can hardly be so 
 expressed. 
 
 (B 884) C
 
 1 8 Natttral History of Language 
 
 enjoyed by a living tongue. It gives an appearance of 
 greater uniformity and syntactical consistency than ever 
 exists in the language as practically spoken at the moment. 
 In a still larger measure does it fail to reproduce the language 
 as it is pronounced and heard. In the first place the 
 alphabetical symbols, as hitherto employed, are in every 
 language altogether insufficient to represent the subtle 
 differences and liaisons which are recognised as belonging to 
 the spoken sounds. In the second, no two persons who 
 speak the same language pronounce all the sounds of their 
 words in precisely the same manner. Suppose we admit the 
 possibility of constructing and employing a series of phonetic 
 symbols comprehensive enough to reproduce every shade of 
 vowel-quality, every nicety of consonantal articulation, and 
 every degree of stress or kind of tone ; it would follow that, 
 if each speaker applied those symbols with precision when 
 representing his own speech in writing, there would never be 
 entire agreement between any two writers of a given tongue. 
 No two Englishmen, however equal their educational and 
 social equipment, give an exactly identical articulation to the 
 sound-elements of their conversation. Writing, therefore, 
 according to a rigidly scientific method, they would diverge 
 more or less widely in their ' orthography.' 
 
 We are apt to think of Latin, for example, as spoken by 
 a large community of men who all alike articulated simply 
 and exactly those well-defined sounds which we have 
 learned to attribute to the twenty-three symbols of the Latin 
 alphabet. The student of language must resist that illusion. 
 As in Italian the signs e and o, in English a and u, in 
 German g and ch are possessed of more than one value, so 
 in Latin there are signs which did not mean the same thing 
 in all circumstances. We are not in a position to discover 
 precisely how many or of what description those different 
 values were. Moreover, every speaker of Latin, even from 
 physical reasons alone, differed from his neighbours in this 
 or that respect of his articulation ; and thus, in point of its 
 sounds, the living Latin language, which became the parent of
 
 Phonology 1 9 
 
 the Neo-Latin tongues, was nothing else than the average 
 result of the individual pronunciations naturally employed in 
 everyday life by speakers throughout the linguistic area. 
 Latin orthography, at the time when it was established, no 
 doubt aimed at reproducing the recognised average as nearly 
 as possible by means of its alphabetical symbols. These, 
 however, were insufficient for their purpose, even when first 
 borrowed from Greece. At a later date the orthography, 
 having become fixed, ignored the established changes in 
 pronunciation, as it has done to a much greater extent in 
 French (for example, in the sounds represented by such 
 misleading symbols as oi, an, eu, at, ?/, and in the case of the 
 e mute, the silent final t, s, p, etc.), still more in English, and 
 in some degree in all languages with a literary cultivation. 
 
 As no two individuals utter identical sounds, so in no 
 case do the speakers of two consecutive generations agree in 
 all respects in that ' average ' articulation which forms the 
 orthodox pronunciation for the time being. What is merely 
 an occasional pronunciation of a vowel or consonant in one 
 generation, may become a frequent pronunciation of the 
 second and the customary pronunciation of the third, and 
 may subsequently itself pass gradually away in favour of 
 some new development. From these considerations it results 
 that, inasmuch as variations of this kind are in constant 
 growth, the language as written according to any con- 
 ventionally orthodox spelling can never absolutely correspond 
 to the language as really spoken. If a phonetic spelling 
 fairly represented the average articulation recognised in a 
 particular half-century, it would require some modifications 
 for the next half, others for the next, and so forth. Such 
 a chronological series of spellings would have formed a most 
 useful chain of records for the student, if it had ever existed. 
 But a phonetic spelling could not entirely arrest the changes 
 of pronunciation themselves, inasmuch as those changes are 
 for the most part due to physiological reasons. Psychology, 
 as represented in the working of the analogical instinct, no 
 doubt plays an occasional part, as, for instance, when the
 
 20 Natural History of Language 
 
 accentuation of a word in English tends to shift towards the 
 earHest possible syllable, e.g. in ' blasphemous,' ' balcony,' for 
 * blasphemous,' ' balcony.' No doubt also, with existing 
 orthographies, changes of pronunciation now and then occur 
 over partial areas, and may thence become general, through 
 deliberate, though misguided, attempts at accommodating the 
 sound to the spelling instead of the spelling to the sound. 
 Thus the spelling is responsible for the modern pronunciations 
 of ' oblige,' , gold,' ' tea,' ' Rome,' ' herb,' in place of the old 
 ' obleege,' ' goold,' ' tay,' ' Room,' ' erb.' Similarly the spelling 
 ' again ' is apparently bringing back into wide use a corre- 
 sponding pronunciation, in place of the common 'agen.' It 
 would not be easy to determine the precise extent to which 
 these influences affect articulation ; they are certainly increas- 
 ing with popular education ; but it is quite certain that their 
 effect is, or has been, small compared with the result of causes 
 purely physical. The changes of Latin pronunciation into 
 Italian or French, or of Anglo-Saxon into English, were 
 brought about by speakers who for the most part knew 
 nothing of spelling, and were therefore saved from an influence 
 which may leave its trace here and there upon the utterance 
 of moderns. 
 
 The divergence of individual from individual and of 
 generation from generation, in the treatment of a particular 
 sound, is mainly due to two kindred causes of a physical 
 nature : — 
 
 (i.) The natural difference of formation and flexibility of 
 the articulating organs of different individuals. 
 
 (ii.) The near resemblances between the positions and 
 movements to be adopted by the articulating organs when 
 producing one sound and those to be adopted by them when 
 producing another. 
 
 Add the consideration that pronunciation tends to follow 
 the line of least effort, and that the individual is only 
 prevented from unconsciously indulging his own ease and 
 convenience by the necessity of making himself understood,
 
 Phonology 2 1 
 
 and by the disadvantages of appearing singular, and we 
 deduce the following propositions : — 
 
 (i) Though all the individuals of a linguistic community 
 may deliberately aim at producing a certain sound, many 
 will find it naturally difficult to produce exactly that sound 
 (though they will often be unconscious of their own defect) ; 
 others will find it actually impossible ; and while the latter 
 can but utter the nearest sound of which they are capable, 
 the former also will lapse into the easiest approximate sound, 
 unless there are disadvantages in the proceeding sufficiently 
 serious to compel them to make an effort at each time of 
 utterance. 
 
 (2) Where nicety of pronunciation is not critically 
 considered and remarked upon, the effort to be exact will 
 tend to relax in all cases where any physical difficulty is 
 present. 
 
 (3) Even where there is no physical inability or defect 
 on the part of the speaker, haste, indolence, or inattention 
 will often lead to inexact operations of the articulating organs, 
 and therefore to the substitution of an approximate sound in 
 place of the one which serves as the standard for the time 
 being ; and here again there will be no check upon such 
 deviations, except the necessity of remaining intelligible or 
 the desire to avoid unpleasant results of singularity. 
 
 It is obvious, then, that in order to understand the 
 phonetic metamorphoses which language perpetually and of 
 necessity undergoes, speech must be considered first and 
 foremost on its physical side. 
 
 The Mechanism of Speech 
 
 In ordinary breathing the air which has been inhaled into 
 the lungs is again expelled by the retraction of the muscles 
 of the thorax and abdomen. It rises through the bronchi 
 and the trachea (or windpipe) to the larynx, where it passes 
 between the two membranes known as the vocal chords, but 
 without interference on the part of those membranes. It 
 
 ^A.
 
 2 2 Natural History of Language 
 
 then issues into the pharynx, and passes out by way of the 
 mouth or the nasal passages, without interference through 
 intentional muscular action of throat, tongue, palate, or lips 
 at any point in its progress. Unless the breathing is for 
 some cause or other rendered unusually violent, or unless 
 there exists some abnormal obstruction (as in the case of a 
 cold), the sound which results is inconsiderable, or, so far as 
 it is appreciable, is produced simply by the natural friction 
 of the current of air against the sides of the passage as a 
 whole. The more violent the breathing, or the more 
 obstructed the passage, the greater is the friction and the 
 more distinct the noise. 
 
 To turn breathing into speech is simply to put into 
 exercise various muscular powers at various points of the 
 expiratory apparatus. The departure of the air from the 
 lungs is no longer in the form of slow and effortless streams, 
 but takes the shape of a succession of short currents expelled 
 with more or less force. These pass over the edges of the 
 two membranes above mentioned, which (according to the 
 nature of the sound to be produced) may meanwhile either 
 be drawn towards each other by muscular tension and made 
 to vibrate, or may remain open and flaccid. The first deter- 
 mination of a sound as an element of speech is made at this 
 point and in this way. Next, after leaving the glottis, the 
 current of air is subject to modifications of the finest and 
 most complicated descriptions, by the action of the muscles 
 of the throat, tongue, soft palate, under-jaw, lips, and cheeks. 
 By these agencies the sound-passage during speech is con- 
 tinually shortened, lengthened, closed, or narrowed into a 
 most subtle variety of shapes and with the greatest rapidity. 
 Thanks to the investigations of physiologists like Briicke 
 and Helmholtz, assisted by the laryngoscope, and of phonetic 
 philologists like Ellis, Sweet, Sievers, etc., the operations in- 
 volved in speech and the special characters differentiating 
 the several classes of speech-sounds have been ascertained 
 to a high degree of scientific precision. 
 
 The following diagram may be taken as representing
 
 Phonology 
 
 roughly the positions of the organs of speech in a state of 
 rest : — 
 
 Incisor . 
 Teeth . 
 
 Lower 
 Jaw " 
 
 Glottis and Vocal Chords 
 
 Hard Palate 
 
 Soft Palate 
 (Velum Palati) 
 
 Epiglottis 
 
 In the above figure the head and neck are represented 
 in section from back to front ; the lips are closed, the nasal 
 passage being left open for breathing. The tongue lies in 
 its natural position, the muscles of the pharynx are at rest, 
 while the glottis is left open and without muscular tension 
 
 It may be as well, without entering into unnecessary 
 detail, to say something of the constitution and action of 
 each of the several organs employed in the conversion of the 
 stream of breath into speech. 
 
 («) The Larynx, Glottis, and Vocal Chords. — The 
 larynx is a species of cartilaginous box, composing the 
 upper part of the windpipe (or trachea). Of the enclosing 
 cartilages those known as the thyreoid, or thyroid (the front 
 and side walls, recognisable from the exterior as the ' Adam's 
 apple '), the cricoid (a horizontal ring), and the two arytaenoid 
 (the upper part of the back wall) are employed in varying 
 the form of the hollow of the larynx, which may be widened.
 
 24 Natural History of Language 
 
 narrowed, lifted/ or lowered by muscular action. Within 
 the box of the larynx, somewhat above the middle, is the 
 gflottis, a kind of doorway into the windpipe, formed by 
 two folds of mucous membrane attached to the back of the 
 thyreoid and to the front projections of the arytaenoid 
 cartilages. The cleft between them extends from back to 
 front, not from right to left. These membranes, the VOCal 
 chords, are elastic, and by muscular action may be stretched 
 with various degrees of tension or relaxed. The vulgar 
 notion that they are ' cords ' or strings is altogether incorrect. 
 Rather they resemble two curtains, capable of being brought 
 together or drawn apart, or, as Max Mliller expresses it, they 
 are " like the parchment of a drum split in the middle." 
 The average length of the cleft in men changes from 
 19 millimetres (or nearly £ of an inch) when relaxed to 
 
 25 millimetres (or about an inch) when stretched. In 
 women the corresponding extremes average 14 and 17 milli- 
 metres respectively. 
 
 Above the glottis, and forming a kind of lid to the 
 larynx, is the slender heart-shaped or leaf-shaped cartilage 
 known as the epiglottis. Normally the epiglottis stands 
 almost erect behind and below the back of the tongue, 
 leaving the larynx open for breathing purposes. When the 
 larynx is threatened with the intrusion of foreign matter, the 
 epiglottis falls and closes the aperture. During the act of 
 swallowing it is for the most part sufficient that it should 
 incline far enough to act as a ' shoot.' 
 
 {U) The muscles of the pharynx act in combination 
 with movements of the larynx, tongue, and soft palate 
 to modify the shape of the passage by which the sound is 
 individualised. 
 
 ic) The Hard Palate, Soft Palate, Uvula, and Nasal 
 Passag"es. — The roof of the mouth consists of two parts, the 
 front portion being an arched plate of bone covered with 
 mucous membrane and called the Jiard palate, while the 
 hinder part, called the soft palate or velum palati, is a flexible 
 
 1 This can be felt by placing the finger upon the larynx while speaking.
 
 Phonology 25 
 
 curtain composed of muscular and cellular tissue. At the 
 middle of its rear or free edge the velum tapers into the 
 uvula, which in a state of rest hangs towards the root of the 
 tongue. The soft palate is provided with muscles, by means 
 of which it can be raised or stretched. The uvula also 
 possesses a muscle which lifts or drops it as occasion 
 requires. 
 
 Behind the uvula and the upper fold of the velum is a 
 passage leading to the nasal cavities, and, during speech, it 
 is a function of the soft palate and the uvula, acting by 
 their proper muscles, to regulate the use which shall be made 
 of the nasal channel. They may rise and block it entirely, 
 or narrow it, or fall and leave it free. 
 
 id) The Tongue. — The tongue, which plays the most 
 important part in speech, is fixed only at its root or lower 
 side, where it is attached to the hyoid bone and the floor of 
 the mouth. For the rest it is capable of adopting an alto- 
 gether indefinite number of positions, in virtue of an extra- 
 ordinary endowment of muscles, by which its whole surface 
 is bent upwards or downwards, made broad or convex, or its 
 tip is carried to or towards any point of the teeth, gums, or 
 hard palate, so that the oral passage is widened, narrowed, 
 or wholly closed at various points beneath the roof of the 
 mouth. 
 
 {/) The lips. — The lips are provided with muscles which 
 draw them back, shoot them forward, bring them together, 
 or open them. It is necessary, for instance, in the production 
 of some sounds to close the lips, in others to round them, in 
 others to bring the lower lip into contact with the upper 
 teeth. Sometimes the lips are required to assist in lengthen- 
 ing the sound-passage, sometimes in shortening it. 
 
 (/) The loiver jaiv performs a sufficiently obvious part 
 in widening or narrowing the passage as a whole and pro- 
 viding room for the necessary play of the organs at different 
 degrees of effort.
 
 26 Natural History of Language 
 
 Sounds in Detail 
 
 Voice, Pitch, Stress, and ProdnctioJi of the Vowels. — It has 
 already been said that while the organs, as above described, 
 are at rest (in the ' indifferent ' position) and only- 
 ordinary breathing is in progress, there is no appreciable 
 sound, whether the mouth be open or shut. The sounds of 
 speech are produced by certain efforts and muscular actions 
 at certain points and in certain ways. For the vowels and 
 several of the consonants the first operation of muscular 
 activity consists in the vibration of the vocal chords. 
 
 The vibration in question takes place thus : The two 
 ligaments are drawn closely together from the sides of the 
 glottis, and the ascending stream of air, in forcing its way 
 through the cleft between, pushes them upwards and slightly 
 apart ; they then fall back with a consistent rapidity (many 
 times a second) to the same position, to be as rapidly and 
 regularly driven upward. They act, in short, like the 
 metallic ' tongue ' in certain wind-instruments under a given 
 pressure of air, and in the same way communicate a regular, 
 or ' isochronous,' vibration to the air-stream as it passes 
 through them. It is these vibrations which constitute 
 ' voice,' and a sound is said to possess ' voice ' or ' tone ' 
 only when the vocal chords have been so employed in its 
 production. Some of the consonants, it will be found, are 
 mere ' noises,' produced by friction or ' explosion ' else- 
 where in the vocal passage, and these do not demand a 
 vibration of the vocal chords. They have merely ' breath,' 
 not ' voice,' and are therefore known as ' voiceless.' 
 
 In the case of sounds with ' voice ' the stream of air is 
 modified at the vocal chords as in the following diagram : — ■ 
 
 < 
 o 
 
 Breath from lungs 2_| air with wave-movement 
 
 ^ — 1_ -, -. -^(to oral passage) 
 
 S-l or 'tone ' 
 
 g '
 
 Phonology 2 7 
 
 while for the ' unvoiced ' sounds the corresponding repre- 
 sentation is — 
 
 < 
 
 o 
 
 Breath from lungs p, air without 
 
 > — ' >(to oral passage) 
 
 S-! ' tone ' 
 
 Apart from the question of the duration of a special sound, 
 i.e. the time during which it is prolonged, the vibrations 
 imparted to the stream of air in producing a vowel may be 
 so varied as to characterise it in three distinct ways : — 
 
 (i) They may differ in rate, i.e. in the number of 
 vibrations which may be made to take place in a given time.^ 
 The result is a difference, not of the vowel produced, but of 
 the pitch (or ' height ') at which the sound is delivered. The 
 quicker the vibration, the shorter the wave of air and the 
 shriller the sound. What is called ' sing-song ' speaking is 
 the effect of a strong tendency to variation of pitch, or 
 'pitch-accent.'- Since a shorter chord under the same 
 conditions vibrates more rapidly than a longer one, the 
 different lengths of the vocal chords in men and women 
 produce the familiar difference in the pitch of their voices. 
 
 (2) They may differ in amplitude, the vibration being 
 regarded as a wave, and the amplitude being the relative 
 height of the wave from crest to trough. Thus in the waves 
 A and B— 
 
 their comparative amplitude is expressed by the vertical 
 lines. The result is a difference of loudness, strength, 
 intensity, or stress. Thus a sound is pronounced piano or 
 forte {i.e. without or with stress) in proportion as the energy 
 of expiration creates a smaller or greater swing of the chords, 
 
 1 This is determined by the shortening or lengthening of the chords during 
 utterance. 
 
 * Also called ' tonic,' ' musical,' or ' chromatic ' accent.
 
 28 Natural History of Language 
 
 and therefore amplitude in the vibrations. Stress-accent ^ is 
 of the greatest importance in the history of language. 
 
 (3) They may differ in foinn {i.e. in the shape of the 
 wave-line). With the same pitch and the same stress the 
 sound may assume different ' qualities ' (or timbres) in 
 consequence of different shapes adopted by the resonance- 
 chamber, i.e. the oral passage, through which the vibrating 
 air is made to pass. Thus a sound-wave of the form — 
 
 may, without variation of its rapidity (' pitch ') or its 
 amplitude ('strength'), be altered by the configuration 
 adopted by the mouth-passage into the shape — 
 
 and so become invested with a very different timbre." The 
 whole series of vowels are thus determined. So far as the 
 vocal chords are concerned their production is the same. 
 The difference between a, e, i, 0, u, (i.e. ah, eh, ee, 0, 00), and 
 an entirely indefinite number of others, is simply this 
 difference of quality, caused by the special configurations 
 
 ' Also called ' expiratory ' or ' dynamic ' accent. 
 
 - In music the same note, produced with the same number and amplitude of 
 vibrations, will differ in 'quality' according to the shape, material, etc., of the 
 instrument upon which it is sounded. Thus the note C possesses a special 
 character according as it proceeds from a violin, piano, clarionet, or organ. The 
 reason of the dissimilarity is the difference in the fortu of the vibrations, a 
 difference which causes to be brought out in connection with the ' prime ' or 
 ' fundamental ' note a different set of 'by-tones,' 'overtones,' or ' harmonics ' in 
 the several cases. Any tone (other than that produced by pendulum), though 
 apparently simple, is resolvable into a number of 'partials,' i.e. a predominant 
 combined with accessory tones or 'harmonics.' These accessory tones belong to 
 the vibrations of parts, which go with vibrations of the whole, and are in some 
 degree perceptible to the trained listener. 
 
 In speech the prime tone which is imparted by the vocal chords may have its 
 possible overtones variously brought out and accentuated by various lengths and 
 shapes into which the mouth, tongue, palate, lips, etc., alter the resonance- 
 chamber. Each such shape ' reflects ' different waves, and the reflected waves 
 cause an interference with the shape of the incoming vibration. \
 
 Phonology 
 
 29 
 
 adopted by the resonance -chamber of pharynx, mouth, 
 and nose. 
 
 Theoretically the number of possible vowel-sounds would 
 thus appear to be unlimited. In practice, however, language 
 can make no use of infinitesimal shades of difference which 
 would be too fine for the ear to catch ; and though the fairly 
 distinguishable vowel-sounds, if adequately collected and 
 registered from all existing languages and dialects, would 
 make a formidable list, the philologist will hardly seek to go 
 beyond the "thirty-six elementary vowels" which Mr. Sweet 
 gives as the result of the multiplication of 3 ' heights ' of 
 tongue with 3 points of ' narrowing,' 2 ' widths ' of tongue, 
 and 2 qualities of ' rounding.' 
 
 Thus if, while the vibrating air passes through the oral 
 passage, the tongue lies in its natural position of rest, and the 
 lips are simply held open {i.e. without muscular exertion to 
 push them forward or draw them back), the soft palate being 
 meanwhile raised towards the back wall of the pharynx, — 
 in other words, if the sound is allowed to pass out through a 
 channel shaped as in Fig. 2, the result is the 'indeterminate' 
 vowel d (written variously, e.g. the second a in Marthdi). 
 
 Fig. 
 
 Fig. 3. 
 
 If, on the contrary, the larynx is lifted, the lips and corners 
 of the mouth retracted, and the tongue lifted nearly to the 
 palate as in Fig. 3, so that the shortest possible resonance-
 
 30 Natural History of Langiiage 
 
 chamber is created, combined with the greatest narrowing 
 employed for vowels, the result is i as in sit. 
 
 If, again, the larynx is drawn downwards, and the lips 
 pushed forward and rounded, so that the longest possible 
 chamber is created, and if this is modified in shape by an 
 elevation of the tongue toward the soft palate and by a 
 narrowing of the lips, the result is ti (i.e. oo, like the u in 
 ' rude '). 
 
 If, while the lips are thus pushed forward and rounded, 
 the larynx is raised instead of dropped, the result is il {i.e. 
 the ' modified ' or French u, which is a compromise between 
 •i and 2t as described above). 
 
 *' A full description of the manner of production of each of 
 the more common vowel-sounds is unnecessary, even if it 
 could be made exact. They vary with the degrees of 
 elevation and convexity of the tongue, with the point of 
 narrowing, and with the degree of rounding by means of lips 
 and cheeks. With a slight simplification of the place- 
 element in Mr. Sweet's classification of the thirty -six 
 ' elementary vowels,' we may divide as follows : — 
 
 (A) — Higli, mid, or lozv, according to the degree in 
 which the tongue is raised. (See the figures for d 
 and / above.) 
 (B) — (i) Back (or velar, less properly guttural), when 
 the back of the tongue approaches the soft 
 palate, as in a, 0} u} 
 (2) Front (or palatal), when the surface of the 
 tongue approaches the roof of the mouth, 
 as in c, i, il. 
 (C) — Narroiv or zvide, according as the tongue is 
 kept broad and flat or made narrow and con- 
 vex. Thus a {all) is wide, while u in ' cup ' is 
 narrow. 
 (D) — Rounded or unrounded, according to the action of 
 the lips and cheeks implied in those terms. Thus 
 
 1 It should be remarked, however, that, because of the rounding of the lips in 
 ti and 0, these vowels are also equally classed as labial.
 
 Phonology 3 1 
 
 u shows the greatest degree of rounding, less, 
 a least. 
 The complete technical description of a vowel embraces 
 a reference to each of these divisions. For instance — 
 
 aiu (in ' law ' = am'- ball ') is low-back-narrow-rounded. 
 a (in ' father ') ,, mid-back-wide-unrounded. 
 
 it (in French lune) ,, high-front-narrow-rounded. 
 
 i (in ' sit ') ,, high-front-narrow-unrounded. 
 
 a {ah) being the clear fundamental tone, i.e. the most natural 
 and least modified of the distinct vowels ; / being the 
 ' highest/ with shortest passage ; and ti the most rounded, 
 with longest passage ; we get a, i, u as the three ' prime ' or 
 typical vowels. If we arrange these as a triangle 
 
 a 
 
 (Highest) /Z -^?/ (most rounded) 
 
 we may represent all others, however numerous, as inter- 
 mediate. Thus, to choose only the more salient examples e 
 {ay), o (as in the German sc/ion), 0, il, the arrangement 
 
 will sufficiently indicate their respective degrees of height 
 and rounding. 
 
 The Neutral Vowel. — Besides such distinct vowel-sounds 
 as those recognised, for example, in the English words 
 'father,' 'm^n,' ' brtll,' 'g^t,' ' bzt,' ' g^t,' ' r//de,' ' bz^U,' ' c?qV 
 etc., there exists, as has been seen, an indeterminate, 
 indistinct, or neutral vowel, heard in unaccentuated syllables, 
 e.g. in ' among,' ' at home/ ' atone/ ' attempt/ ' vehement/
 
 •6 
 
 32 Natural History of Language 
 
 * together,' ' phenomena.' The same sound attaches to the 
 ' e mute ' of French, when, as read in verse,^ it ceases to be 
 mute, as well as to the old English final e in the Chaucerian 
 ' fressch^ flowr^s.' It is obvious that the vowel which 
 happens to be written in such cases is not the one which is 
 actually pronounced. Though the exact nature of the 
 sound has not usually been shown in a special symbol, it is 
 generally held to be nearest to e (as in ' get '), and is there- 
 fore now conventionally written by philologists in the shape 
 of that letter reversed, viz. d. 
 
 The Nasalised Vowels. — So far the vowels have been 
 treated as ' pure ' vowels, i.e. as they are ordinarily pro- 
 nounced, with the nasal passage closed. If, however, the 
 velum, or curtain of the soft palate, is allowed to drop, so 
 that the column of air may pass behind it and vibrate in the 
 nasal cavities, the vowel-sound acquires a quality called 
 nasalisation. It is the vibration within the nose-channels, 
 and not simply pronunciation through the nose, which 
 imparts the quality, inasmuch as the nasalisation is stronger 
 when the air is not allowed actually to issue through the 
 nostrils. French supplies the best examples of nasalised 
 vowels in the sounds which it writes, as an {am), en {em), 
 in {im), on {om), un {um). Unfortunately this orthography 
 suggests a combination of two distinct sounds, viz. of a 
 vowel a {e, etc.) with a consonant ;/ {m), and that fact is apt 
 to prove a stumbling-block to foreigners who learn the 
 language otherwise than by ear. Phoneticists prefer to 
 write a, e, i, 6, u, or, less well, a, e, q, etc. 
 
 The Diphthong's. — If the organs of speech, after adopt- 
 ing the position necessary for the production of a certain 
 vowel, pass from it into the position for another vowel, and 
 if ' voice ' is made to sound during the time occupied in the 
 
 1 It is so read because till the sixteenth century it had not become mute, but 
 was pronounced. Verse everywhere retains archaisms of both grammar and 
 phonetics. Compare the English pronunciation of ' wind ' in poetry. The 
 'e mute' in French is historically the general representative of various Latin 
 vowels which equally sank to the indeterminate value.
 
 Phonology -XiZ 
 
 transition, and only during that time, the result is a diphthong. 
 A diphthong such as an, ai, eu, ei, etc., is therefore not a 
 combination of two vowels, but a sound which is neither one 
 nor the other. It may be remarked in passing that the 
 English I (' I,' ' nz'ce,' etc.), though written as a simple vowel, 
 is really a diphthong, viz. ai ; while many sounds in different 
 languages written as diphthongs are really simple vowels, as 
 in the case of ' hvoad,' the French au/r^ ( = otre), /our ( —pUf), 
 pevir ( =p^r), etc. 
 
 Vowels and Consonants ; Sonants and Consonants ; 
 Semi- Vowels. — A very old division of the alphabet into 
 ' vowels,' meaning sounds capable of being pronounced alone, 
 and ' consonants,' or sounds incapable of being so pronounced, 
 is scientifically inexact. The sounds commonly known as 
 consonants are properly distinguished from the vowels above 
 described only by the fact that they are formed mainly, 
 when not wholly, in the oral and nasal passages. Whether 
 * voiced ' or ' breathed,' they are the result of either friction 
 or complete stopping of the air-stream in some part of those 
 passages. While it is true that, as a rule, these latter sounds 
 do not form syllables by themselves, and that some of them 
 cannot, but need the accompaniment of a vowel, in many 
 cases this cannot be said of elements of the alphabet which 
 are usually described as consonants. Thus in English words 
 like ' tab/^,' ' father,' ' evtv/,' ' hosoju,' in French words like 
 ' ardve,' ' table,' and in German words like Hafidel, Vater, 
 ' gut&Q.,' 'g-utem,' the spelling disguises a fact otherwise easily 
 realised, namely, that the sounds actually pronounced are 
 respectively nothing more than ' tabl,' ^ ' fathr,' ^ ' evn,' ^ 
 ' bosm,' ^ ' arbr,' ' gutn,' etc. That is to say, the liquids and 
 nasals /, r, ;;/, n are capable of forming syllables in 
 themselves. In the Slavonic languages and in Sanskrit the 
 liquids, unaccompanied even in writing by a vowel, often 
 constitute distinct syllables, which may even bear the accent. 
 
 ^ This is the case when the words are pronounced in the ordinary quick 
 manner of conversation. Sometimes (or by some speakers) the last syllable is 
 rather ?/, 3 (without audible r), pu, p;;/. 
 
 (b 884) D
 
 34 Natural History of Language 
 
 Similarly, sounds represented by z and v (in which ' voice ' 
 is employed) may be quite independent of vowel assistance. 
 The words was, of, when uttered without stress, are often 
 actually pronounced wz and v.^ The voiced spirant th 
 (pronounced as in ' //^en,' and phonetically represented by 5), 
 may also be made audible in its own right. With the 
 corresponding unvoiced sounds, viz. s, f, th (as in ' ///in,' 
 phonetically written p), i.e. where the hiss takes the place of 
 the buzz, independence of a vowel-sound exists in a smaller 
 degree, but is still a possibility from the continuity which 
 may be imparted to hissing and friction of the mere air- 
 current in the narrow passage. 
 
 It is only the ' mutes,' ' stops,' or ' explosives ' (/, /, k, b, 
 d, g) which are wholly consonantal ^ in the sense implied in 
 the ordinary classification. Practically, however, it may be 
 said that, while sounds which do not partake of voice (but 
 which are simply ' noises ' produced in the vocal passages 
 after the unvoiced air has left the glottis) serve as consonants 
 only, on the other hand those sounds which do partake of 
 voice (except the voiced ' explosives,' b, g, d) are capable, 
 whether they be vowels or not, of performing the function 
 of vowels in forming syllables. Thus, in the English 
 alphabet not only a, e, i, o, u, but also z, v, /, in, n, r — the 
 voiced spirants and voiced liquids and nasals — are capable 
 of syllabic value, whereas the remainder may for practical 
 purposes be considered as entirely consonantal. 
 
 The antithesis of ' vowel ' and ' consonant ' being there- 
 fore unsound, it is more desirable to speak of 'sonant' and 
 ' consonant', the term sonant including all sounds, whether 
 
 ^ Sometimes it is rather av which is heard. 
 
 - It appears, therefore, that Greek observation was remarkably acute, when it 
 divided the alphabet into 
 
 (1) ^wv^evra ('vowels'). 
 
 (2) ffvtx(pu}va ('consonants'), including 
 
 (i.) i}ixi(pu}va ('half-sonant,' or, according to Plato, "not vowels 
 indeed, but nevertheless not without sound"), viz. /, m, n, 
 r, s. 
 
 (ii.) &(pu}va (' mutes '), viz. i, t, p, g, d, h, and the aspirates.
 
 Phonology 
 
 3d 
 
 vowels or not, which are not serving in consonantal function. 
 There is thus a ' sonant ' / or vi, as well as a ' consonant ' 
 / or in. In the word lascar both / and r are consonants ; 
 in the word miserable r and / are really sonants (the 
 phonesis being inizrdbl ^). For the philologist, convenience 
 dictates a distinction of spelling, and the signs I, m, n, r, 
 f, y {i.e. the same letters with diacritical marks) have been 
 adopted for the sonant values, while the ordinary symbols 
 stand for the pure consonants. 
 
 On the other hand the vowels i and u may acquire the 
 function of consonants {y and w) in virtue of a hurried or 
 slurred pronunciation. In this function they are called semi- 
 vowels, or consotiantal vowels, and are written z, 2i. The 
 French out supplies an instance of the consonantising of a 
 vowel by this rapidity of its articulation. 
 
 Consonants and their Classification. — The consonants 
 proper admit of cross-classifications according to several 
 principles. They may be considered from the points of 
 view of ( I ) their kind of articulation, i.e. in respect of voice, 
 breath, or aspiration ; (2) the degree of openjtess of the vocal 
 passage during their production ; (3) \.\\q. place of articulation, 
 i.e. the point in the passage at which stopping, friction, etc., 
 take place. 
 
 (i) Kind of Articulation. — What is meant by 'voice' 
 has been already explained. ' Breath ' implies the absence 
 of vibration on the part of the vocal chords. An * aspirated ' 
 consonant involves either breath or voice, as the case may 
 be, driven with energy against the hollows of the stiffened 
 larynx, so as to experience distinct friction. 
 
 The four possible classes of sounds are generally known 
 respectively as voiced, voiceless, voiced aspirate, and voiceless 
 aspirate} 
 
 ^ The second syllable is more commonly syncopated, and the result is mizribl. 
 
 "^ Other names which are or have been in vogue for voiced sounds are ' sonant,' 
 'mediae,' 'soft'; and for voiceless, 'surd,' ' tenues,' 'hard.' 'Soft 'and 'hard' 
 are misleading terms. ' Sonant ' and ' surd ' are correct enough as meaning 
 respectively that which has tone and that which has no tone ; but it is better to
 
 36 Natural History of Language 
 
 Of the more familiar consonantal sounds the voiced 
 include : — 
 
 (i) g, d, b {stops or explosives'). 
 
 (2) z, z {i.e. the sound of i- in ' pleasure,' or the French 7), 
 
 V, 5, (i.e. th in ' then '), i ( = j), u { = w). These 
 are the voiced spirants. 
 
 (3) ^j ^> ^>^i ^^ (the liquids and nasals, which are voiced in 
 
 most positions at least ^). 
 The corresponding voiceless sounds are : — 
 
 ( 1 ) k, t, p {stops or explosives). 
 
 (2) s, s (the sound of s/i in 'ship'), f, p (i.e. th in 
 
 ' thorn ') — the voiceless spirants. 
 The aspirated consonants, which do not occur in English, 
 but are heard in Irish articulation, once existed in Greek, and 
 belong to general Indo-European philology, are : — 
 
 (i) The voiced aspirates gh, dh, bh (somewhat as in 
 ' lo^/^ouse,' ' dea(^/zead,' ' Ho^^^ouse,' but without 
 the separation between the syllables). 
 (2) The voiceless aspirates kh, th, ph (Greek t^, 6, — 
 somewhat as in ' pen//^ouse,' etc.). 
 
 (2) Degree of Openness. — For the articulation of different 
 consonants the passage may be narrowed at various points 
 in various degrees, until the narrowing agents are at last in 
 actual contact. In the last case the air-stream is sometimes 
 absolutely blocked, sometimes it is allowed an outlet, inas- 
 much as the contact obstructs only a portion of the available 
 channel. Thus the lips may be closed, but the nasal passage 
 allowed to open ; or the tongue may touch the middle of the 
 
 keep the term ' sonant ' for another purpose, viz. for sounds not consonant. 
 'Media' and 'tenuis' are translations of the Greek t^aov ('middle') and y^/CKby 
 ('bare,' 'thin'), as opposed to Satri/ ('thick,' 'rough,' i.e. aspirated). The 
 ' middle ' character of a media lay in the fact that there was something more than 
 the mere breath of the tenuis, and something less of an effortful kind which 
 marked the aspirata. 
 
 ^ That the liquids are commonly voiced is indicated by the change of d (but 
 not /) to / in Latin, as well as by their sonant functions. That the nasals are 
 naturally voiced may be realised from the fact that, when a cold prevents the 
 proper articulation of 'man,' the result is 'bad,' not ' pat.' Nevertheless voiceless 
 nasals and liquids are met with.
 
 Phonology 37 
 
 palate, but may yield a space at each side (as for /), etc. 
 The following operations may be distinguished ^ : — 
 
 (i.) Complete blocking of both the mouth- and the nose- 
 passage, resulting in the sounds known as ' stops,' 
 ' explosives,' or ' mutes,' which are heard only 
 when the contact is broken, e.g. k, t, p, g, d, b, and 
 their aspirates. 
 
 These sounds are known as momentary, because 
 they cannot be prolonged ; the classes which 
 follow are continuotis. 
 (ii.) Complete blocking of the mouth-channel, but subse- 
 quent opening of the nasal passage, so that the 
 stop is evaded. The result is the nasals in, n, ng 
 (as in ' singer '). 
 (iii.) Evasions (in the mouth-channel itself) of a closure 
 made with the tongue. The sounds produced are 
 the 'liquids.' If the air escapes at the sides of 
 the tongue the result is the ' lateral ' dental / ; if 
 the tip is allowed to yield (and generally to 
 vibrate) while the air escapes over it, the result 
 is r. 
 (iv.) Close narrowing of the mouth-passage, a chink 
 being left through which the air either buzzes 
 (with ' voice ') or hisses (without ' voice ') with 
 consequent well-marked friction. The sounds pro- 
 duced in this case are the ' spirants ' or ' fricatives,' 
 e.g. z, s, z, s, V, /, i, u, S, p. Sometimes z, s, z, s 
 are called ' sibilants,' while the term ' spirants ' is 
 restricted to the remainder. 
 (3) Place of Articulation. — Theoretically, the points 
 at which closing or narrowing may take place in the mouth- 
 passage are quite indefinite in number, and the classification 
 of sounds according to the particular organs and particular 
 parts of organs concerned in the narrowing may be made 
 
 1 German phoneticists call the explosives and fricatives Cerduschlaute and the 
 rest (with the vowels) Sotiorlaute.
 
 38 Nahiral History of Language 
 
 more and less elaborate. For practical purposes it is 
 sufficient to distinguish them according as they are formed — 
 (i.) At the soft palate or velum, 
 (ii.) At the hard palate, 
 (iii.) At the teeth or gums, 
 (iv.) At the lips. 
 
 These classes of sounds are known respectively as velars, 
 palatals, dentals, labials. More precise is a further distinction 
 between dentals proper and alveolar dentals (which are pro- 
 duced at the gums). An old terminology by which the first 
 two classes, viz. velars and palatals, are alike designated 
 ' gutturals ' is incorrect. 
 
 (i.) Velar Consonants are those in the production of 
 which the soft palate and back of the tongue are 
 mainly concerned. The Semitic q, the German 
 ch in nach, audi, the Scotch ch in loch, a provincial 
 French and the Northumbrian r are among the 
 most recognisable cases. In Indo-European 
 philology it is usual to assume the existence in 
 the primitive speech of a series of velar stops, now 
 commonly written as q, g, qh, gh, the so-called 
 ' guttural ' being produced at the velum, as is the 
 k in the German Kuh, the g in the French goiit, 
 etc. Velar also is the nasal ng (usually written if) 
 in, eg., the German Zimgc. The English ng in 
 ' tongue,' c in ' come,' and g in ' go ' are velar, but 
 are attracted somewhat forward, 
 (ii.) Palatal Consonants are those formed by narrowing 
 between the tongue and the hard palate. The 
 Semitic k ; the English k in ' king,' g in ' give,' 
 y in ' yes ' ; the German ch in ich ; the Italian^/ in 
 bottiglia, are readily felt to be of this nature. 
 For primitive Indo-European a series of palatal 
 stops is assumed, which, in opposition to the velar, 
 are written k,g, kh, gh. To these is added a nasal 
 n, as in * think.' 
 
 When the tip of the tongue is turned back
 
 Phonology 39 
 
 under the hard palate, as in i ( = sJi in ' ship ') and 
 ( = the s in ' pleasure '), the sound is called 
 cerebral or cacuminal. 
 
 (iii.) Dental Consonants are produced by the tip of the 
 tongue touching or closely approaching the upper 
 teeth, or, in the case of the alveolar dentals, the 
 gums of the upper teeth. The latter is the case 
 with the English dental stops /, d ; the dental 
 spirants s, z ; and the linguals /, r. In German, 
 as in some other languages, the tip of the tongue 
 touches the teeth proper, and the dentals are true 
 dentals. The two lh-so\ir\ds in English,^, S, are 
 interdental. In primitive Indo-European a dental 
 series is assumed, consisting of t, d, th, dh. The 
 corresponding nasal is n (as in ' thin '). 
 
 (iv.) Labial Consonants are formed when the lips play 
 a chief part, e.g. in />, b, zv, and the nasal in. 
 When the lower lip is brought into contact with 
 the upper teeth and not with the upper lip, as in 
 forming f and v, the sound is known as labio- 
 dental. In primitive Indo-European the assumed 
 series of labial stops is p, b, pk, bh, with the nasal w. 
 
 PaPallelism. — It will be observed that, in point of place 
 of articulation, there is a parallelism between the stops, the 
 spirants or fricatives, and the nasals.^ Thus, corresponding 
 to the complete labial block in /, b, come the labial 
 narrowings and friction-sounds w, /, v, and the nasal 
 
 ^ The vowels also have their classification according to place of articulation. 
 Thus a, o, II are velar ; e, o, it, i are palatal ; o, u are lal)ial as well as velar. It is 
 obvious, therefore, that in the process of articulation a given consonantal sound not 
 only goes more easily in combination with one set of consonants than with another, 
 but also goes more easily in combination with certain vowels than with others. 
 It will consequently be found that vowels 'attract' consonants in respect of place 
 of articulation. As it is otherwise expressed, there is a greater ' affinity ' between 
 certain vowels and consonants than between others. For instance, the velar q 
 followed by the palatal e requires much more exertion in adjusting the positions 
 of the articulating apparatus than is needed in the case of k followed by e. In 
 course of time the q may actually be ' attracted ' to k by the vowel.
 
 40 
 
 Natural History of Language 
 
 evasion in. The only difference between vi and b consists 
 in the fact that in the case of the former the nasal passage 
 is opened after the closure of the lips. The respective 
 positions for the two sounds are roughly shown in the 
 following figures : — 
 
 Fig. 
 
 Fio. 5. 
 
 Similarly, corresponding to the complete dental stops /, d^ 
 come the nasal evasion n (the relation between n and d 
 being exactly as between in and b), the spirants p, S, s, s, 
 and the liquids /, r. Answering to the palatal stops k, g, 
 etc., we have a nasal n and spirants ch (in German ic/i), y, 
 etc. With the velar stops q,g^ etc., may be placed the spirant 
 cJi (in ' loch ') and the nasal ?;. 
 
 The Aspirate. — If, the mouth being open, breath without 
 voice is expelled from the lungs with effort and made to 
 experience distinct friction in the cavities of the stiffened 
 larynx itself, the result is the spiritus asper, or so-called 
 ' aspirate ' //. In kind of articulation it is therefore voiceless ; 
 in point of place it is sometimes called ' faucal.' 
 
 The following table may serve as affording a working 
 synopsis of classifications of the more common elements ^ 
 in familiar alphabets : — 
 
 ' (i) Sounds like r/i in 'church' (r) andy' in 'judge' are not simple sounds, 
 but consonantal diphthongs, and are pronounced as /}, di, respectively. 
 
 (2) The symbol r (as in a less degree /) really represents different sounds in
 
 Phonology 
 
 41 
 
 
 Velar. 
 
 Palatal. 
 
 Dental. 
 
 Labial. 
 
 "(3 
 
 1 
 
 c 
 
 ■g 
 
 3 
 U 
 
 Alveolar. 
 
 c 
 
 4) 
 
 Q 
 
 2 
 
 c 
 
 p 
 
 6 
 
 ■.3 
 
 Vowels— 
 
 Pure . 
 Nasalised . 
 
 u, 0, a 
 u, 0, a 
 
 3, 0, ii, e, i 
 5. i 
 
 e, i 
 
 
 
 
 0, U 
 
 5, u 
 
 Nasals— 
 
 Sonant 
 Consonant . 
 
 rp 
 
 9 
 
 n 
 
 
 9 
 n 
 
 
 
 m 
 
 Liquids— 
 
 Sonant 
 Conso7iant . 
 
 Northumbr. r 
 
 gl(Ital.) 
 
 
 rl 
 
 r 1 
 
 
 
 
 Sibilant Spirants— 
 
 Voiced 
 Voiceless 
 
 
 
 2(1) 
 s 
 
 s 
 
 
 
 
 Spirants Proper— 
 
 Voiced 
 Voiceless 
 
 ch {\och) 
 
 ch (k/0 
 
 i(=y) 
 
 
 S 
 & 
 
 V 
 
 f 
 
 «( = w) 
 
 Stops— 
 
 Voiced 
 Voiceless 
 
 Voiced Aspirate . 
 Voiceless Aspirate 
 
 gCfo) 
 
 q (tome) 
 
 gh 
 
 qh 
 
 g C^ve) 
 k (^ey) 
 
 gh 
 kh 
 
 
 d (Eng.) 
 t(Eng.) 
 
 dh 
 
 th 
 
 d (Ger.) 
 t (Ger.) 
 
 
 b 
 
 P 
 
 bh 
 ph 
 
 Glides. — In ordinary speech the various sounds above 
 described are not, of course, articulated independently of 
 each other. In pronouncing the several ' letters of the 
 alphabet' contained in a given breath-group (i.e. a group of 
 sounds uttered with one impulse of the breath-stream), we 
 do not deliver them with just so many separate staccato 
 movements of the vocal apparatus. They are in reality 
 linked to each other by a continuous series of glides or 
 transition-soimds, created while the organs of speech are 
 passing from one position to another. Between any two 
 
 different languages. Thus there are the ' trilled ' r and the ' untrilled ' r, and the 
 narrowing may be made at different points, e.g. at the gums, palate, and back 
 of tongue.
 
 42 Natural History of Language 
 
 consecutive sounds produced in different manners there 
 exists such a ' gUde.' For the most part these transition- 
 sounds receive no representation in writing, inasmuch as 
 the particular gHde is practically inevitable, or implied in 
 the situation. Thus, to take a very obvious case, in the 
 Italian dice it is unavoidable, for a correct pronunciation, 
 that in passing from the u to e there should be created a 
 light t£/-sound. We might in fact write dic''e or du'*e. To 
 write duue itself would be an error, inasmuch as that method 
 would suggest a pronunciation with too protracted a transi- 
 tion (sometimes, it is true, actually heard in Italy), du-we, in 
 which the w is as decisively articulated as the d, ii, or e. 
 The Australian cry ' cooee ! ' is phonetically kt'ii with glide u 
 {ku"t). When the call is given with stress for long distances 
 and in the usual protracted manner, the result is a distinct 
 ku-ui. The difference is readily appreciable between the 
 shapes taken by the lips in pronouncing the word due and 
 those taken in producing an independent du followed by an 
 independent e after a pause. The lips can be seen and felt 
 to round in the act of transition. Similarly in ' piety ' there 
 is a glide between the vowels {pi-ety). The glides belong, 
 however, to consonants as well as vowels. In due there is 
 a glide from d to u and in knl from k to u. Technically 
 that glide is known as the ' off-glide ' o{ d or k and 
 the ' on-glide ' of 7i. Every two consecutive sounds 
 in a breath-group are thus connected by ' on- ' and ' off"- 
 glides.' 
 
 It will be obvious that any deviation in the articulation 
 of a given consonant or vowel will create a consequent 
 variation, however slight, in the character of its on- and off- 
 glides. Furthermore, the glide between two given sounds 
 may be more or less accentuated or prolonged. As a result, 
 between consonants, as between vowels, there may ultimately 
 be developed a new and distinct element in the word, 
 demanding representation in the spelling, The Greek 
 i* avp6<i ianros) becomes av'^po'i {a7t6.ros), the French •(" tenre 
 becomes tenure, the Latin \ enitus becomes em^tus. The
 
 Phonology 43 
 
 modern French oin = oil, pronounced with the ghde o"il, 
 then with the ofF-glide triumphing over its vowel, until it is 
 practically the glide only which is heard. Similarly, the 
 Greek -f* [xpoTo<i {inrotos) passes first into "f" ^j3por6^ {jjibrotos) 
 and thence into ^poro<i {brotos).
 
 CHAPTER III 
 
 SPEECH AND WRITING 
 
 The Evolution of Alphabets 
 
 Very early in the history of mankind there would arise the 
 desire to communicate thought to others to whom there 
 was no opportunity of speaking. Those persons might be 
 absent, or unborn, or unable to understand the language 
 spoken by the man whose thought was to be the subject of 
 communication. For those at a distance, for posterity, or 
 for the foreigner of alien speech, attempts would be made 
 to convey a thought by perception of the eye in default 
 of the power to convey it through perception of the ear. 
 Such attempts would naturally begin with the crudest 
 devices, and, except with the simplest and most unmistakable 
 notions, would fall very far short of their purpose. The 
 natural efforts of mankind, in this as in other matters, to 
 attain to greater precision combined with greater ease, 
 have led to the evolution of %vriting. Far from perfect 
 as modern writing is, it has travelled a long way from its 
 rude origins. 
 
 The first endeavours at the communication of thought to 
 the eye would naturally take the shape of sketches or 
 drawings, such as those executed by the American Indians 
 or the Australian aboriginals. These would represent a 
 whole scene, which each observer would interpret for himself 
 according to his own lights and the greater or less detail 
 and precision of the delineation. Doubtless within a 
 
 44
 
 speech and Writing 45 
 
 particular tribe or community there would grow up certain 
 traditional principles to assist in both the execution and the 
 reading of a drawing ; but even more beyond doubt is the 
 fact that such sketches could only express the concrete 
 elements in very simple notions, leaving undetermined — so 
 far as the drawing was a drawing and nothing else — 
 anything abstract, or any conditions of time, place, fact, or 
 contingency. They therefore necessarily admitted of various 
 and often contradictory interpretations. 
 
 At such a stage there is plainly no connection between 
 symbol and individual sound, or even between symbol and 
 speech at all. It is thought, not language, which puts on 
 the visible shape, a fact which is made clear by the reflection 
 that this method of communication would be just as effective 
 between the deaf and dumb who have never known any 
 language, or between foreigners whose languages are 
 reciprocally unintelligible. Like gesture, it has no reference 
 to the operations of the vocal apparatus, and makes no 
 attempt to indicate them. 
 
 The case is not very different with the ' knot-writing ' 
 which was once practised in a remote period of Chinese 
 history, and which, as quipu-v^x\\AX\^, was found by the 
 Spaniards in regular use among the officials of the Incas of 
 Peru. The quipii was a cord, measuring from two feet 
 upwards, from which hung a parti-coloured fringe of threads. 
 These were knotted and intertwined in various conventional 
 ways, and, partly by the colours and sequence of the threads, 
 partly by the disposition and number of the knots, they 
 could be employed so as to convey official messages and 
 preserve records. This, of course, necessitated the represen- 
 tation not only of material objects, but also of abstract 
 notions. For the latter purpose the colours were employed. 
 Thus white stood for ' silver ' or for ' peace,' red for ' gold ' 
 or for ' war.' Here we take a distinct step beyond the 
 sketch-writing of the North American Indians, inasmuch as 
 we must assume for the Peruvian method the existence of a 
 previous understanding or code of interpretation between
 
 46 Natural History of Language 
 
 the parties employing the device. Whatever the origin of 
 the practice, the connection between thought on the one 
 hand, and colours and knots on the other, became purely 
 artificial and conventional, and instruction for both reader 
 and ' writer ' was as necessary as it is with the modern 
 alphabetical system. We can therefore perhaps hardly deny 
 that the method was a form of ' writing.' That it was 
 utterly inadequate and helplessly dependent on special 
 materials is, no doubt, an immense disadvantage, but hardly 
 touches the essence of the matter. More to the point is the 
 fact that it was purely symbolical of ideas, and not of the 
 sounds in which those ideas were orally conveyed. It was 
 not made up of signs with distinct phonetic values, but of 
 signs with notional values in the rough. Given, therefore, 
 the secret of the meaning of a knot or colour, it would have 
 been possible for persons of entirely different languages to 
 write the same thought identically in quipii. In other words, 
 it has, like the sketch-writing, no relation to speech, but 
 only to thought. 
 
 We are not here concerned with the question whether it 
 might not be a highly desirable thing for the world if it 
 could actually possess certain universal symbols for register- 
 ing thought directly, in spite of the diversities of spoken 
 language. As a fact that rather Utopian aim is not the 
 recognised object of writing, which has so far only striven 
 to impress upon the mind through the eye as exact a 
 realisation as possible of the spoken sounds which would 
 have fallen upon the ear, if it had been the ear which was 
 addressed. 
 
 For this purpose neither scene-drawing nor knot-writing 
 could lead anywhere. As F. Miiller points out, for those 
 who, like the American Indians, possessed languages of the 
 polysynthetic type, and whose mental processes had not 
 arrived at the analysis of the sentence into individual words, 
 much less into individual sounds, no other method of ocular 
 communication of thought would suggest itself than one 
 which expressed a whole conception as a unit. For the
 
 speech and Writing 47 
 
 representation of the component elements, first as far as 
 words, then as far as syllables, and finally as far as sounds, 
 it was necessary to find some new point of departure. 
 
 This lay in what is known as ideography, or writing by 
 pictures, the conception of which — as its name implies — 
 was fundamentally the same as in the more rudimentary 
 scene- drawing above described, while its application was 
 very different, and carried with it the possibilities of an 
 immense development. In the picture-writing adopted by 
 the Chinese in place of the older method of knots, in the 
 earliest hieroglyphics of the Egyptians, in the originals of 
 the cuneiforms of the Euphrates valley, and in the more 
 advanced form of American writing employed by the 
 Aztecs — who had arrived at a livelier consciousness of the 
 individual word, — the pictures do not represent entire 
 sentences. Separate pictures correspond to separate words 
 and stand habitually for those words. Originally, of course, 
 each such sign was a literal representation (' ideogram ') of 
 a material thing, the application to abstract notions being 
 determined by association of ideas. Thus it would be 
 natural to represent the sun by a rayed circle, and a tree, 
 an ibis, a serpent, or a house by rough but recognisable 
 drawings of those objects respectively. The abstract notion 
 of the verb ' hear ' might, as in Chinese, be expressed by the 
 picture of an ear applied to a door, or, in other regions, by 
 some other self-explaining device. In Chinese two hands 
 clasped together denote the general notion of ' friend.ship ' ; 
 in Egyptian ' thirst ' was represented by the picture of a 
 running calf added to the picture for water. Probably 
 within a particular language a number of different devices 
 would at first be attempted for the representation of the 
 same abstract, but ultimately one of them would become 
 universalised, either by right of special fitness (as the most 
 precise or as the easiest to draw) or through some accident 
 of authority. 
 
 In the earliest stage of such writing each sign is a clear
 
 48 Natural History of Language 
 
 picture. The sign for ' sun ' is a sun. The sign for a 
 mountain is an elevation rising to a peak or peaks. If such 
 original distinctness and fulness had been maintained, it is 
 manifest that the ideograms pertaining to one language might 
 have been intelligible — at least in a large measure as regards 
 the mere substance — to speakers of another language The 
 communication is still a direct communication of notions 
 and is not bound up with language. Except for differences 
 in local customs, or in the local shapes of things, or in the 
 special ideogram adopted for an abstract, or in the arrange- 
 ment of the signs, the picture-writing of one people would 
 largely resemble that of another, and each could in a 
 corresponding degree interpret the pictures of the other 
 into thought. 
 
 As a matter of fact, however, such signs, when a series 
 of them has been thoroughly established and accepted over 
 a certain area, tend more and more to lose their frank 
 distinctness and to adopt conventional shapes whichi admit 
 of more easy and fluent writing. Their resemblance to the 
 simple picture is at first obvious enough, then merely 
 suggestive, and at last scarcely, if at all, discoverable. In 
 Chinese, for example, a mountain comes to be represented 
 first adequately enough by the sign /v/Vy , and then 
 simply by three points ,J_,, while the symbol J^^ suffices 
 for a man. In later Egyptian the picture of a lioness is 
 replaced by an amorphous character something like the 
 letter l^ (of which it is in fact the origin) with flourishes 
 in the horizontal stroke (^ ). If, therefore, writing origi- 
 nated in several quarters with pure ideograms, and if each 
 people thus modified the outlines till the picture was 
 wholly lost in a symbol, and if, as is inevitable, the modifica- 
 tions made by one people differed in their evolution from the 
 modifications made by another, the result would be that, 
 whereas the original pictures might have been reciprocally 
 translatable with more or less accuracy into thought, the one 
 set of conventionalised symbols would suggest nothing what- 
 ever to those who employed the other set. Even the people
 
 speech and Writing 49 
 
 which used a particular set of symbols would require to 
 be taught their meaning and use, since they would no 
 longer tell their own story. In Chinese the pictures were 
 abbreviated (in the graphic sense) into conventional signs 
 by at least as early as 2000 B.C., and the Egyptian hieratic 
 symbols, when borrowed by the Semites at almost as early 
 a date, had long ago lost the realistic outlines of the oldest 
 hieroglyphs. In the 'demotic' writing of Egypt the de- 
 parture is still more marked. 
 
 The pictorial representation of individual words, while 
 a distinct advance upon the scene-drawing which corresponds 
 to entire thoughts, still labours under the disadvantage that 
 it can only embody the substance or material of the thought. 
 Even then the finer species of the same genus of objects 
 {e.g. trees or birds) will hardly be adequately differentiated. 
 Setting this last consideration aside, such a method lacks 
 the power to express, except by artificial and awkward 
 devices which are not properly pictorial, any of the necessary 
 conditioning or auxiliary elements, such as those of tense 
 or mood. For instance, an unmistakable and self-explain- 
 ing pictorial representation of the notion 'if may fairly be 
 called inconceivable. Moreover, since each original picture 
 stands for a special concept, the representation of thought 
 in writing will require as many different picture -signs as 
 there are concepts. It is, of course, the same with the 
 merely conventional symbols, when they have ceased to be 
 real pictures. The number of different signs to be mastered 
 by writer and reader will therefore be very large, and the 
 whole system exceedingly difficult and cumbrous. Chinese, 
 which, as we shall see, has ingeniously simplified the system 
 by its method of employing the same sign for all 
 ' homophones ' (or ' homonyms '), still cannot be reduced 
 to less than about five hundred primary symbols. 
 
 Nevertheless it is from this pictorial or conceptual 
 
 writing and its abbreviated symbols that a phonetic, and 
 
 ultimately an alphabetic, writing has been derived. The 
 
 process by which the symbol is transferred from represcnt- 
 
 (b 884) E
 
 5C Natural History of Language 
 
 ing a particular notion to representing a particular sound — 
 that is to say, is converted from an ideogram into a 
 phonogram — is not difficult to trace, and is historically 
 demonstrable in a degree sufficient for practical purposes. 
 
 All languages contain, in a greater or less measure, 
 * homonyms,' or ' homophones,' words which are different 
 in meaning but identical in sound. In English, for 
 example, we have hare and hair ; air and heir ; / and eye ; 
 all and awl; box in its different senses, etc. In French sans 
 (' without '), ce)it (' hundred '), sang (' blood ') are all pro- 
 nounced sa ; ?Zi??// means both ' nine ' and ' new.' In German 
 Tor is either ' fool ' or ' gate ' ; adit is ' eight ' or ' attention ' ; 
 backen is either ' bake ' or ' cheek.' In Italian hanno (' they 
 have ') and anno (' year ') are identical in pronunciation. 
 In Latin pdret means either ' obeys ' or ' appears ' ; est means 
 either * is ' or 'eats.' In ancient Egyptian w^r = 'love' and 
 also ' eye ' ; nefer means ' youth,' ' colt,' ' lute,' etc. In some 
 languages the number of such homophones is very large, and 
 inevitably so where all words have been cut down to very 
 light and simple monosyllables, as in Chinese. In that 
 language the monosyllabic combination of sounds Imd has 
 the meanings ' eagle,' ' prince,' ' cold water,' ' fear,' and a score 
 more. 
 
 According to the logical application of written signs 
 derived from pictures, i.e. of ideograms, each of the different 
 concepts in such a list of homophones should receive a 
 symbol of its own. As a fact, however, both the languages 
 for which pictorial writing was brought into existence, viz. 
 Chinese and ancient Egyptian, fell very early into identify- 
 ing a sign with the sound of its word rather than with its 
 meaning. Manifestly this would occur less readily while 
 the written sign was a literal picture ; but it would occur 
 easily enough when the picture had vanished from the 
 symbol. To the mind of a Chinese a symbol for Imd 
 would first and foremost call up the utterance hud ; to the 
 mind of an Egyptian a symbol for mer would call up the
 
 speech and Writing 51 
 
 utterance mer. From the other side, the utterance hud, 
 heard alone or in a doubtful context, might call up (to the 
 mind's eye) any one of the series of original symbols belong- 
 ing respectively to the several meanings of that utterance ; 
 but it would naturally first call up the most representative of 
 those symbols, i.e. the symbol which was attached to the 
 most frequent (or perhaps the most dignified) sense of the 
 word. Thus for the sound hud the notion which suggested 
 itself par excellence would be that of ' prince.' It is there- 
 fore easy to understand the step by which the symbol 
 for /j;/rt = ' prince ' would come to stand for the sound hud 
 in all its settings. In other words, the symbol would 
 obtain a purely phonetic value, or become a ' phonogram.' 
 Add the consideration that, among the large number of 
 symbols originally standing for the different meanings of 
 the same sound, there would arise (when the true pictorial 
 character had been lost from each) the greatest danger of 
 confounding their use, and so defeating the end of writing 
 altogether. In English the association of sound rather than 
 sense with a symbol causes such frequent misspellings as 
 ' their ' for ' there,' ' here ' for ' hear,' and vice versa. In 
 Chinese the inevitable result in the case of a word like Jmd 
 with, say, twenty -six meanings would be chaos. Hence 
 another inducement to express the sound ]md by its most 
 representative sign, and to define the particular acceptation 
 of that sign by the device of adding ^determinants^ or 'keys' 
 
 According to this expedient, ' cold water ' is expressed 
 by the sign for Jiud ( = ' prince ') added to the sign for 
 * water.' Thus the utterance is to be hud, but the particular 
 sense of hud intended is defined by the determinant as ' cold 
 water.' Similarly, taken alone, the phonetic sign for pe 
 signifies ' white ' ; if we add the sign for ' tree,' the whole 
 is to be read as '' pe, but in the sense of the tree bearing that 
 name. i.e. the cypress." 
 
 The same generalisation of one symbol for a whole 
 series of homonyms took place in Egypt, and a similar 
 expedient was hit upon for differentiating meanings. On
 
 52 Natural History of Language 
 
 the monuments, hovv^ever, the determinants added to the 
 phonogram are actual pictures of the olden type. Thus the 
 sign for the sound nefer along with a lute would mean " the 
 sound nefer^ to be understood in the sense of the lute, and 
 not in any other which might attach to it." 
 
 The change from the conceptual to the phonetic accepta- 
 tion of a symbol is thus brought about by what is practically 
 rebus-writing. The process may be well illustrated by the 
 practice of the Aztecs — who had made considerable advance 
 in picture-writing before the Spanish conquest — in their 
 efforts to represent the Latin Paternoster by means of their 
 existing symbols. In Aztec /«« = ' pennant' and tetl = 
 * stone.' The picture of a pennant combined with that of a 
 stone suggests the sounds pa{n)-te{tl). Similarly jws = 
 ' cactus-fig,' and a picture of the fig combined with the 
 picture of a stone suggests the sounds nos-te. The sight 
 of the symbols ' flag ' + * stone ' + ' cactus-fig ' + * stone ' called 
 up before the Aztec mind an utterance pa{n)te-iwste, or 
 paternoster as nearly as available symbols could give it. 
 Dissociation of the symbol from the sense and its associa- 
 tion with the mere sound are here complete.^ 
 
 Thus far we have arrived at a phonetic value for a 
 symbol, but not necessarily at even a syllabic, much less an 
 alphabetic, value. The symbol represents a certain com- 
 bination or group of articulated sounds, but not the individual 
 components of the group. In the Chinese sign for ta there 
 is no special mark for the / nor for the a. In very ancient 
 Egyptian there would be a sign for the whole word nefer, 
 but none for n, e,f, r, respectively. The oldest monuments 
 do, indeed, show such alphabetical characters already in 
 partial use, but the non-alphabetical stage is necessarily 
 presupposed. In Chinese, it is true, the words being all 
 
 1 For Egyptian Dr. Isaac Taylor gives the amusing instance ol khesteb ( = lapis 
 lazuli), represented by a man trj'ing to hold back a pig. ' Pig ' in Egyptian is teb, 
 ' stop ' is khesf, and each ideogram stands for a sound, regardless of the meaning. 
 Thus the picture reads simply khes{f)teb. 
 
 I
 
 speech and Writing 53 
 
 monosyllables, each character necessarily represents no more 
 than a single syllable, e.g. ta, viin, pab. In Egyptian also 
 the greater part of the root-words were monosyllabic, and the 
 signs, therefore, were largely of the value of single syllables ; 
 but this was by no means invariably the case. 
 
 In Chinese, with its always monosyllabic and entirely 
 invariable words, writing stopped at the stage at which a sign 
 represents the totality of the sounds combined into a syllable. 
 There was nothing to bring home to the Chinese mind the 
 fact that even a monosyllable was compounded of a number 
 of different sounds. The conception of a smaller phonetic 
 unit than the word was either never grasped, or seemed to 
 have no bearing upon the practical needs of a writer. With 
 old Egyptian the case was different. Here the words were 
 not, like the Chinese, unchangeable in either length or sound. 
 For the expression of grammatical relation they might be 
 inflected by the addition of prefixes and suffixes. They 
 might also undergo internal modification, corresponding to a 
 modification of the root-idea. The result is well pointed out 
 by F. Miiller. Since the prefixes and suffixes might be 
 single vowels or consonants as well as whole syllables, the 
 writer would become alive to the oral practice of adding such 
 single-sound modifying elements. They would acquire an 
 independence in his consciousness. Thus with son ('brother') 
 would go sonz. (' my brother '), son^s. (' thy brother '), sonl 
 (' his brother '), sonu (' brothers '), son\> (' sister ')• From this 
 fact the individual sounds a, k, f, u, t would come to claim 
 independent recognition, and thus the resolution of a com- 
 bination of sounds into its component elements became 
 possible. 
 
 The recognition of syllables would necessarily some- 
 what precede the recognition of entirely isolated sounds. 
 When the latter was effected, it would naturally lead 
 to the observation that a series could be made of 
 words all beginning with one particular sound (e.g. /), and 
 another series of words all beginning with another sound 
 (e.g. /), etc.
 
 54 Natural History of Language T^ 
 
 The step next taken was of supreme consequence for 
 all the subsequent history of writing. This consisted in 
 allowing the character standing for a particular syllable 
 or group of sounds to be employed also simply for the 
 initial sound of that group. Thus the character for 
 the sound-group ahoin, 'eagle' — originally a picture of 
 that bird — came to stand for a. The character for the 
 sound-group re (originally the picture of ro, ' the mouth ') 
 was used for the single sound r ; the character for 
 taboi {' lioness ') for /, etc. There w^ould, of course, thus 
 be in existence a larger or smaller number of characters 
 available for the same purpose, inasmuch as different 
 syllables or sound -groups equally began with the same 
 ' letter ' (e.g. I'et, ra, and others, as well as ro, began 
 with r). The question which of these should be chosen to 
 represent the initial r universally, would naturally depend on 
 the greater frequency or the easier shape of one as compared 
 with the others ; or on some other consideration which we are 
 not in a position to discover. For a time there would be 
 room for option, till from amid vacillating use the fittest 
 would survive. Indeed in actual records, while there is a 
 regular representation for each of twenty-five simple sounds, 
 there are nevertheless found in the case of some of these 
 certain occasional, though not very frequent, variant forms. 
 Thus the sound n was represented not only by the regular 
 sign — the conventional derivative from the ideogram for 
 'water' — but also by a crown, and occasionally by a 
 vase. 
 
 By this means a true alphabet comes tentatively into 
 existence. Nevertheless, though purely alphabetical signs 
 appear on very antique monuments, the ancient Egyptians 
 themselves never came to employ them purely and con- 
 sistently. The older methods of ideography and of syllabic 
 phonograms continued still to blend with the incomparably 
 more convenient device for which the way had been shown. 
 The consistent use of alphabetical characters proper triumphed 
 only when an alien people borrowed for the purposes of its
 
 * speech and Writing 55 
 
 own language the best expedients at which Egypt had 
 arrived/ 
 
 Development of Modern Alphabets from Egyptian 
 by way of Phoenicia 
 
 That the Phoenician (or Babylono-Phoenician) alphabet 
 came by borrowing from Egypt was a tradition of antiquity, 
 and has been demonstrated beyond all reasonable doubt by 
 the research of De Rouge. The precise date and manner of 
 the borrowing are perhaps beyond discovery, but, granting 
 the fact, and examining the earliest traceable existence of 
 Semitic writing and its oldest specimens, the most plausible 
 presumption is that an intimacy with the Egyptian characters 
 was first forced on the Semites during that occupation by 
 them of Lower Egypt which is associated with the name of 
 the Hyksos. From that period onwards the art of writing 
 in similar characters would inevitably spread among the 
 other Semites. The Phoenicians in particular, who carried 
 on a large commerce with Egypt and held trading settlements 
 
 1 A word must be said of another form of writing which had meanwhile sprung 
 up and was struggling towards the same goal in the plain of Mesopotamia. The 
 cuneiform ('wedge-shaped ') script was in its origin as ideographic as the Egyptian 
 hieroglyphs. In the method of writing invented by the non-Semitic ' Accadian ' 
 (' Sumerian ' or ' Proto-Chaldean ') inhabitants of that region the concept was first 
 represented by a picture (' fish,' for instance, by the outline of a fish). Such out- 
 lines, as they appear in the very ancient inscriptions known as ' linear Babylonian,' 
 were originally drawn in strokes of an ordinary character, but speedily became 
 conventionalised and unrecognisable as pictures, the process being accelerated by 
 the fact that the symbols were chiefly impressed upon soft bricks rather than in- 
 scribed upon stone. This practice resulted in the use of an implement which 
 imprinted a combination of wedge-like marks. At first such marks corresponded 
 to at least the more salient lines of the ideogram, and recalled it. Later they were 
 simplified both in number and arrangement till they were intrinsically almost 
 meaningless. Like the Egyptian and Chinese symbols, they next became phono- 
 grams. That this conversion would begin with the monosyllables, in which the 
 single ideogram answered to a single sound-group, seems a reasonable assumption. 
 In the hands of the Semites, who were first the neighbours of the Accadians and 
 afterwards occupied their place, the signs became purely syllabic. The Persian 
 conquerors at a later period adopted the signs with a further change to an almost, 
 though not quite, purely alphabetic value.
 
 56 Natural History of Language 
 
 in the Delta, could not fail to realise the advantages to be 
 derived from this easy means of record and communication. 
 In virtue of their pre-eminence as traders, the Phoenicians 
 would naturally be the principal agents in the diffusion of 
 writing both eastward and northward of Egypt. Neverthe- 
 less, for the Semitic peoples alone, we need not assume that 
 Phoenicia was a first and indispensable intermediary. 
 
 The style which the Semites might naturally be expected 
 to learn would be that which was regularly practised by 
 Egyptian scribes for ordinary purposes, and not that which 
 is illustrated in the more elaborate and rigid forms of the 
 monumental hieroglyphs. That is to say, they would adopt 
 the ' cursive ' or ' running ' forms of the ' hieratic ' (or priestly) 
 system which was then in vogue ; and it is to the shapes of 
 those characters, and not to those of either the older monu- 
 ments or the latter 'demotic' (or popular) script, that the 
 origin of the Phoenician letters must be referred. Those 
 prototypes are best seen in the Papyrus Prisse, an extremely 
 ancient document transported from Egyptian Thebes in the 
 first half of the nineteenth century to the National Library 
 at Paris. 
 
 In historical times the forms of the characters came to 
 diverge widely in different areas of the Semitic domain. 
 The Aramaean variety, followed by the post-Babylonian 
 Hebrew, is markedly distinguishable from the Phoenician. 
 Nevertheless, if we take the shapes of the letters in the pre- 
 Aramaean ' Siloam inscription ' of Jerusalem (dating from 
 about 700 B.C.), the forms on the Moabite stone (of about 
 900 B.C.), and the writing of what is known as the Phoenician 
 'Baal Lebanon' inscription (looo B.C.), and compare them 
 with the oldest shapes employed in the Aramaean region, 
 we are enabled to arrive at a primitive Semitic type which 
 is directly deducible from that of the Prisse hieratics. Thus 
 the hieroglyphic ideogram for /, a lioness, has become in the 
 hieratics ^. On the Moabite stone the shape is ^, that of 
 Baal Lebanon is ^, of Siloam ^, and of later Phoenician 
 / . The passage of this symbol to the earlier Greek \__
 
 speech and Writing 57 
 
 and Roman L is obvious. For practical purposes the 
 oldest Phoenician characters may be taken as representing 
 the script which the Semites had made out of the Egyptian 
 hieratics, and which became the parent of almost all the 
 alphabets in use to this day throughout the civilised world. 
 
 Casting aside all 'determinant' ideograms and variants, 
 the Phoenicians adopted one sign consistently for each sound, 
 and so constructed an alphabet of twenty-two characters. 
 They abandoned also the arbitrary arrangements possible in 
 Egyptian, and wrote, after the manner preferred in hieratics, 
 invariably from right to left. For the most part they 
 naturally applied a given Egyptian character to the same 
 sound in their own language, or, where they did not possess 
 the exact sound, to the one which was most nearly akin. 
 The sign for b, viz. ^ (originally the ideogram of a crane) 
 became in Phoenician S, with the same value b. The sign 
 for / ,J>- (originally the ideogram of the horned snake), 
 became V and stood for zv. There is, however, one series 
 of exceptions which particularly distinguishes Semitic ortho- 
 graphy. The signs employed in Egyptian for vowels were 
 not so used in Phoenician, but received values of a different 
 kind. Thus alepJi, the sign corresponding to Egyptian a, is 
 transferred to represent the light faucal consonant known to 
 students of Greek as the ' spiritus lenis ' or ' smooth breathing.' 
 Similarly, the sign for / becomes the consonant y iyod). The 
 vowels find no representation at all in old Semitic. The 
 significant elements in an ordinary Semitic root are its three 
 consonants, the vowels being regularly varied according to a 
 system. The consonant was supposed to carry its vowel 
 with it, the k-s\gn standing for ka, ki, ku, etc., according to 
 circumstances. Thus ktb may be read as kataba, kzitaba, 
 kdtabu, katiWu, etc. It was presumed that the reader of such 
 script would possess a lively understanding of the language 
 as spoken, and, in virtue of that understanding, would in- 
 stinctively supply the proper vowels from the context. To 
 some extent this practice was encouraged by the fact that 
 the parent Egyptian script itself frequently attached such a
 
 58 Natural History of Language 
 
 ' syllabic ' value to a character, leaving vowels (at least other 
 than a, i, ii) unrepresented ; but it was indisputably the 
 peculiar importance of their consonants and the entire sub- 
 ordination of their vowels which prevented the Phoenicians 
 from developing a completely alphabetical writing. The 
 inconveniences of the system manifested themselves later, 
 when it was discovered that ambiguity or lack of precision 
 might occasionally arise from the absence of vowel-signs. 
 The earliest invention of diacritical marks for the notation of 
 vowels appears to date from about the fourth century A.D. 
 The Phoenician script, therefore, still lies somewhere between 
 a purely alphabetical and a syllabic writing. Phonetic 
 simplicity is not yet attained. 
 
 It has already been said that the Phoenician alphabet 
 is the parent of almost all the alphabets of modern 
 times. Their pedigree will be given immediately. Mean- 
 while some interest attaches to the names applied by the 
 Phoenicians to the borrowed characters, inasmuch as those 
 names were passed on in a corrupted form to the Greeks, 
 and, in that form, have become familiar to most educated 
 people. The very word alphabet is derived through the 
 Greek alpha beta from the Phoenician names for the first two 
 characters, aleph, beth. The Phoenician names themselves 
 were not borrowed with the symbol, but are pure Semitic, aleph 
 meaning ' ox,' betJi ' house,' giniel ' camel,' daleth ' tent-door,' 
 and so forth. Each such name is not only appropriate as 
 beginning with the sound of the letter concerned, but also 
 possesses, or did originally possess, a more or less appreciable 
 aptness to the shape of the character. Thus aleph (' ox ') 
 not only begins with the sound of the spiritus lenis i^alepli), 
 but is a sufficiently natural description of the sign <t', which 
 suggests the head and horns of an ox viewed from the front. 
 Daleth (' tent-door,' or the flap of such a door) not only has 
 the ^-sound for its initial, but is also no bad description of the 
 triangular symbol A. Similarly, mem (' water ') answers to 
 the idea in its wavy line *^. Gimel (' camel ') passed 
 through the stage ^ (suggestive of the animal kneeling).
 
 speech and Writmg 59 
 
 before it reached the simpler form "]• Apparently the 
 Phoenician, after adopting a symbol in its Egyptian phonetic 
 value or one approximate thereto, cast about for a name 
 which should fulfil both the conditions and so prove helpful 
 for mnemonic purposes. For instance, to the sign V or H 
 (Egyptian for/"), with a value iv in Phoenician, would thus 
 be fitly applied the name Van {waii), ' peg' or 'hook.' It is 
 also altogether probable that, at the time when the names 
 were first applied, the resemblance was much more evident 
 than it would remain after custom had come to permit of an 
 easier calligraphy. 
 
 That the Greek alphabet is simply a modification of the 
 Phoenician is well known. Not only is all ancient tradition 
 in favour of this assertion, but the shapes, order, and names 
 of the letters are sufficient to prove it. AlpJia, Beta, Gainina, 
 Delta,Q\.c., are merely Graecised forms oi Alcph,Beth, Giinel, 
 Daleth, and the rest. For Greek itself the words have no 
 meaning. In comparing the Hellenic characters with the 
 Phoenician we have to remember that Greek was first 
 written after the Semitic manner, from right to left ; that 
 then followed a period in which the lines could proceed 
 either way (mostly boustrophcdon, i.e. in alternate directions, 
 the hand working back across the page as oxen turn in 
 ploughing a field), and that establishment of the rule of 
 writing from left to right was comparatively recent {circ. 5 00 
 B.C.). Moreover, the shapes of the Greek letters themselves 
 varied in different places and at different periods, the oldest 
 inscriptions showing forms which are much nearer to the 
 Phoenician than are those of the standard ' Ionic ' alphabet 
 familiar in Greek books. 
 
 Again, just as the Semite, while retaining a sign in its 
 Egyptian phonetic value, felt himself compelled to alter the 
 value of some few in order the better to express the peculiar 
 sounds of his own language, so the Greek, while he mostly 
 applied a Semitic symbol to the same sound in his own 
 tongue, changed the value of some which he did not require,
 
 6o Natural History of Language 
 
 in order to express other sounds which the Phoenician left 
 without representation. He found, for example, in the 
 Phoenician alphabet a surplusage of signs for sibilants and a 
 lack of signs for vowels. But, to the Greek, vowels played 
 no such subordinate part as in Semitic, and the necessity 
 of giving them equal prominence was manifest from the 
 first. Hence, out of the number of Phoenician characters, 
 four which stood for non-Greek sounds were employed to 
 represent the vowels A, E, I, O, while Vau was soon 
 employed in two shapes, one (T) to represent the vowel ii, the 
 other (F) to represent the consonantal n{='w). 
 
 In this transmission there was thus finally evolved 
 alphabetic writing pure and simple. 
 
 The derivation of the original twenty-three letters of the 
 Greek alphabet may be synoptically stated as follows. An 
 asterisk indicates a letter which changed its value to that of 
 a vowel ; an obelisk betokens that the letter subsequently 
 fell out of use.
 
 speech and Writing 
 
 6i 
 
 Old Phoenician Signs 
 
 Old Greek 
 
 (Typical) 
 
 'Ionic' 
 Alphabet 
 
 I <V Aleph ( ox ) = i, lenis) 
 
 *AA(= a) 
 
 A {alpha) 
 
 2 A Beth ('house') = b 
 
 
 
 B ibcta) 
 
 3 7^^ Gimel ('camel') = g 
 
 ^r 
 
 1 (gatnma) 
 
 4 /\ Daleth ('tent- door') = d 
 
 ^ 
 
 ^ (delta) 
 
 5 3 He (meaning obscure) = h (weak) 
 
 '^yf (=e) 
 
 r /ci, later ^ 
 C ( epsTlon) 
 
 6 ^ T Vau ('peg') = w 
 
 7 1 Zayin ( swords') = z 
 
 I (=dz, zd) 
 
 lost (see 23) 
 (dig-amma) 
 
 8 n Cheth ('hedge') =ch ('loch') 
 
 B (=h) 
 
 H ( e ) {2ta) 
 
 9 vl/ Teth (meaning obscure) =t (strong) 
 
 ©0( = th) 
 
 ® fi'/^t'/aj 
 
 ro Z Yod ('hand') = y ( i ) 
 
 \ (=0 
 
 1 (i5ta) 
 
 ri y Kaph ('bent haad' ) = k 
 
 n 
 
 K (^aj>fia) 
 
 12 U L Lamed ('goad') = 1 
 
 Lh 
 
 A /lambda \ 
 
 13 ^ Mem ('water') = m 
 
 MM 
 
 ^ (';«!?; 
 
 r4 ^ Nun ('fish') = n 
 
 V||V 
 
 |s/ r««; 
 
 15 ^ "TSamech (meaning obscure) = s 
 
 ^ (=ks) 
 
 1— 1 
 
 ^ U'i) 
 
 16 Ayin ('eye') ='h (a click) 
 
 *0 (=0) 
 
 -^ /oK.jatcr^, 
 kj ( omikroiij 
 
 17 "^ Pe ('mouth') = p 
 
 1 r 
 
 TTr//; 
 
 rS Yi- Tsade ('hook'?) = ss 
 
 t/^ (=s) 
 
 _ (--■.'« ') 
 
 19 (p Koph ('back of head') =q 
 
 t? (=q) 
 
 — r^'//'0 
 
 20 A C Resch ('side of head') = r 
 
 ^ P 
 
 P r-/.j; 
 
 21 WShin ('tooth') = sh 
 
 >U=s) 
 
 ^ r^vv;..,,; 
 
 22 + Tau ('cross') = t 
 
 T 
 
 T r;<r»; 
 
 23 (new vowel value given to N0.6 ) 
 
 '"Y ( = u) 
 
 /i. later* 
 I \ upsllonj 
 
 * In the above list the names epsilon (? xpiXSi'), upslloti (D xpiXdv), and omikron 
 (fl fiLKpdv) are new creations of grammarians. The old names were el (i.e. i) from 
 HS and, by analogy thereto, ?2 (5), (od).
 
 62 Natural History of Language 
 
 In the above scheme the Greek alphabet has been treated, 
 for the sake of simphcity, as if it had been adopted from 
 the Phoenicians in one and the same form. As a matter of 
 fact it more naturally found its way to different parts 
 of Greece at different dates, through the agency of more 
 than one contingent of Semitic traders, or by being passed on 
 from one Greek people to another. We possess but scanty 
 material for fixing the earliest acquisition, though there is 
 little probability that it can have been later than the tenth 
 century B.C. What we do know is that numerous local 
 varieties were soon established, and these are found in vogue 
 by the time we first meet with inscriptions. We need not, 
 however, here concern ourselves with more than the two 
 main systems, the Ionic (which, after the archonship of 
 Eucleides in 403 B.C., was officially established at Athens 
 and gradually became universally recognised as the standard 
 Hellenic), and the Chalcidic (or ' Euboean '), which was carried 
 to the Greek colonies of Italy on the Bay of Naples and 
 elsewhere, and was thence taken up by the Latins (in the 
 shape of that alphabet from which the modern scripts of 
 Western Europe are derived), and by other Italic peoples 
 in cognate forms. 
 
 Throughout Greece there had been recognised the 
 necessity of modifying the Phoenician alphabet in other 
 respects than that of giving a representation to the vowels. 
 Peculiar among Greek consonantal sounds or combinations 
 of sound were dz {zd), th, ks, ph, kh, ps. For dz {zd) it was 
 easy to take the sign I (Z) of the non-Greek sibilant. 
 For th the otherwise superfluous /-sound, ©, was adopted. 
 For ks the form H was available. For the rest it was found 
 cumbrous to write TIH, KH, ITS, and the Ionic division 
 applied to this purpose the signs <I>, X, ^ (V), which were 
 thereupon placed in that order after the original twenty-three 
 characters above described. The same division felt it 
 desirable to give a special representation to the frequent 
 vowels, c and o. For the former it altered the value of the sign 
 H from h to c\ and for the latter it modified the shape of O to
 
 speech and Writing 63 
 
 n, and appended that sign to the h'st. We thus arrive for 
 the Ionic (and thence for the post-Eucleidean Attic and 
 Greek in general) at the alphabet — 
 
 A B r A E • Z ( = ^ij,) H ( = ^-) • I K A M N 
 
 Hon-PSTT<j)( =///) x( = kJi) ^Y{ =ps) n 
 
 The symbols f , ^ , and San were discarded. 
 
 Meanwhile the Chalcidic variety retained F and ^ , kept 
 H in its original use ( = /^), in some obscure way abandoned 
 3, but afterwards supplied its value later in the alphabet 
 with X { = ks), accepted <I> {=pJi) and ^, but used the latter 
 for kJi (not ps). For A the form was L. Special signs for 
 e and o were not employed. The Chalcidic alphabet was 
 therefore — 
 
 A B r (C) A (D) E F Z H ( = 70 I K L /i 
 
 M N • O n Q P (R) 2 T T (V) X ( = ^5-) <D V ( ^ .^/^) 
 
 From the latter variety is derived the Latin alphabet, 
 which for the most part remained true to both the forms 
 and phonetic values of the letters. There are, however, 
 certain noteworthy exceptions. The aspirates (Jh), <t> (///), 
 V {kh) were not required for the purposes of Latin, and 
 were accordingly rejected. V was accepted in both the 
 consonantal {ic = w) and the vowel function (//). F became 
 the voiceless spirant. The third letter C at first possessed 
 the value of F, which it continued to retain in the abbrevi- 
 ations Cn. (Gnaeus), C. {Gaius), but for some uncertain 
 reason ^ came to adopt the value of K as well. By this 
 change K was rendered useless, although it still remained in 
 the alphabet and was occasionally employed in certain words. 
 Through C having acquired two values, an ambiguity arose, 
 which was met (perhaps about 300 B.C.) by modifying 
 the shape to G when it stood for the original ^-sound. 
 
 ^ Probably through the inlluence of Etruscan, which possessed no ' voiced mute' 
 g. The influence of Etruria on early Rome is known to have been very great.
 
 64 Natural History of Langtiage 
 
 Meanwhile the Z had been dropped as useless, and a place 
 was thus found in the abecedaria for the new symbol. 
 Much later {circ. 100 B.C.) there was borrowed from Greek 
 the sign T to represent the sound of v in Greek words {i.e. 
 French li), and Z was re-adopted for similar reasons. 
 These two signs T and Z were naturally placed at the end 
 of the alphabet already in vogue. The Latin alphabet 
 thus becomes — 
 
 ^A, B, C { = k\ D, E, F (=/), G (formed from C and 
 put in the place of the discarded Z), H, I, K (hardly 
 used), L, M, N, O, P, O, R, S, T, V ( = ii and w\ X, 
 Y ( = Gk. u), Z (used in Greek words). 
 
 Synoptically the development of the classical Greek and 
 Latin alphabets from the older Greek forms may be re- 
 presented thus : — 
 
 1 The names applied by the Romans were twofold, viz. either the Greek 
 {alpha, beta, etc.) or those which we have inherited. The latter were constructed 
 by combining with each consonant its easiest vowel. "The easiest vowel 
 preceded the continuants {i.e. the fricatives) and followed the explosives " (Taylor). 
 Thus <2,. e, i, o, n ; ef, el, em, en, er, es ; be, ce, de, ge, pe, te ; but ka, ha, qu, ix. 
 Zed retained its Greek name {zetd).
 
 speech and Writing 
 
 65 
 
 Old Greek 
 
 Ionic 
 
 Chalcidic 
 
 Latin 
 
 . A 
 
 = 
 
 a 
 
 A 
 
 A 
 
 A 
 
 2 % 
 
 = 
 
 b 
 
 D 
 
 B 
 
 B 
 
 3 r 
 
 = 
 
 & 
 
 r 
 
 r (c) 
 
 
 4 ^ 
 
 = 
 
 d 
 
 ^ 
 
 A(D) 
 
 D 
 
 5 ^ 
 
 = 
 
 e 
 
 E ( = §) 
 
 E ( = g) 
 
 E ( = 6) 
 
 6 r 
 
 = 
 
 w 
 
 (lost) 
 
 F ( = w) 
 
 F ( = f) 
 
 7 I 
 
 = 
 
 dz.zd 
 
 I (z) 
 
 Z 
 
 lost, re-borrowed 
 later, its place 
 taken by G.(No.j) 
 
 8 8 
 
 = 
 
 h 
 
 H ( = e- ) 
 
 H ( = h) 
 
 H ( = h) 
 
 9 © 
 
 = 
 
 th 
 
 e 
 
 e 
 
 (absent) 
 
 m 2. 
 
 = 
 
 i 
 
 1 
 
 1 
 
 1 
 
 .1 K 
 
 = 
 
 k 
 
 K 
 
 K 
 
 K 
 
 12 ^-A 
 
 = 
 
 1 
 
 A 
 
 L 
 
 L 
 
 >3 M 
 
 = 
 
 m 
 
 M 
 
 M 
 
 M 
 
 14 r^ 
 
 = 
 
 n 
 
 t^ 
 
 N 
 
 N 
 
 15 m 
 
 = 
 
 ks 
 
 IS 
 
 ^ lost, replaced \ 
 Vby X. No. 23 / 
 
 (absent) 
 
 16 
 
 = 
 
 
 
 ( = S) 
 
 C = 5) 
 
 ( - 5) 
 
 .7 nr 
 
 = 
 
 P 
 
 n 
 
 TT r (p) 
 
 P 
 
 18 A 
 
 = 
 
 s 
 
 (discarded) 
 
 (discarded) 
 
 (absent) 
 
 19 9 9 
 
 = 
 
 q 
 
 (discarded) 
 
 Q 
 
 Q 
 
 .0 1^ 
 
 = 
 
 r 
 
 P 
 
 P (R) 
 
 R 
 
 21 c 
 
 = 
 
 s 
 
 I 
 
 I 
 
 S 
 
 22 T 
 
 = 
 
 t 
 
 T 
 
 T 
 
 T 
 
 (b 884)
 
 66 
 
 Natural History of Language 
 
 Old Greek 
 
 Ionic 
 
 Chalcidic 
 
 Latin 
 
 23 Y = u 
 
 reshaping of No. 6 
 
 T 
 
 nr (V) 
 
 V = iu I 
 
 24 \ 
 
 
 ([) = (ph) 
 
 © ( = ph) 
 
 (absent) 
 
 -5 
 
 Later additions 
 for all Greek 
 
 X ( = kh) 
 
 X ( = ks) 
 
 X ( = ks) 
 
 26 
 
 
 HJ ( = psi 
 
 Y V ( = kii) 
 
 (absent) 
 
 3_ f Additions for 
 t Ionic Greek 
 
 n ( = 5) 
 
 ( absent) 
 
 (absent) 
 
 ^8 ( Late 
 
 ■1 re-borrowings 
 29 [ by Latin 
 
 
 
 Y 
 
 Z 
 
 Greek and Latin, like all other writing, inevitably developed 
 from the capitals or 'majuscules,' an easier 'cursive' or 
 running script for the pen ; and the small letters ('minuscules'), 
 whether printed or written, are the result of nothing but 
 progressive adaptations of the alphabetical forms given above. 
 
 The modern symbols J (for consonantal i) and W (for con- 
 sonantal u) have no place in the Latin alphabet. The sign 
 J'arose in the sixteenth century ; that of W was adopted for 
 England and Germany in the later middle ages (thirteenth 
 century). A consistent distinction of sound accompanying a 
 distinction of form between the rounded and unrounded V 
 {ix. U and V) is very recent. 
 
 It would be beyond the scope of the present sketch to 
 deal in any detail with the evolution of the numerous families 
 of alphabets which have been derived from Egypt by way of 
 the Semites. It will be readily understood that transmission 
 from one people to another has been regularly accompanied 
 by modifications in shape, additions of new symbols, and 
 other changes analogous to those already described. The 
 provenance of the more important alphabets, so far as it is 
 satisfactorily ascertained, may be most conveniently shown in 
 the form of a family tree.
 
 -Phoenician 
 type 
 
 -Aramaean 
 type 
 
 Speech and Writing 
 
 -Phoenician — Punic — Berber (Tafinaq). 
 (Carthaginian) 
 
 67 
 
 — Palestinian — 
 
 — Old Israelite. 
 — Moabite. 
 
 — Hellenic- 
 
 — Greek (general). 
 
 — Coptic (Graeco-Egyptian). 
 
 — Ionic Gothic. 
 
 — Runic. 
 
 — Cyrillic (S.E. Slavonic). 
 
 — Glagolitic (S. Slavonic). 
 
 -Chalcidic — Italic- 
 
 — Latin. 
 
 — Umbrian, Oscan, etc. 
 — Etruscan. 
 
 — Later Hebraic ('square'). 
 
 , — Pehlevi — Parsi. 
 
 -Eranian- 
 
 — Indian — 
 
 ' — Armenian. 
 
 — North Indian (Devanagari type). 
 
 — Pali — Burmese, etc. 
 
 — South Indian (Dravidian type). 
 
 I — Persian. 
 — Nabathean — Arabic — ' — Turkish. 
 
 I — Hindustani. 
 
 — Himyaritic — Ethiopic — Amharic. 
 
 ■Neo-Syrian. 
 
 — Nestorian- 
 
 — Uigur — 
 
 — Mongol. 
 — Kalmuk. 
 — Manchu.
 
 68 Natural History of Language 
 
 It should be added that, whereas the Occidental or 
 Graeco-Italic alphabets have assigned an equal representa- 
 tion to the vowels, and are therefore alphabets in the proper 
 sense, the Oriental have more closely followed their teacher, 
 and have accepted the Semitic practice of subordinating, 
 when not entirely ignoring, the vowel. So far as vowels are 
 expressed it is by means of diacritical marks of the nature 
 of points and strokes. Thus the typical method in India of 
 indicating i is by a mark above the consonant, while u is 
 indicated by a mark beneath. The most frequent vowel, «, 
 has no symbol, but is assumed, after the Semitic fashion, as 
 inherent in the consonant. The Oriental alphabets are, 
 therefore, still in a measure syllabic. 
 
 We have thus followed the process by which writing has 
 gradually adapted itself to speech till it assigns a separate 
 symbol to a separate sound instead of a picture to a notion. 
 The stages are summed up by F. Muller in terms which we 
 may somewhat modify in view of the treatment and 
 expressions adopted in this chapter : — 
 
 I. Conceptual script — 
 
 {a) Of thought as a whole (a scene-picture) ; e.g. 
 
 of N. American Indians. 
 {U) Of thought in its components (pure ideograms), 
 
 e.g. of Mexicans, primitive Chinese, and 
 
 primitive Egyptians. 
 
 II. Phonetic script — 
 
 {a) Word- writing by phonograms {i.e. with the 
 same ideogram for a series of homophones), ^.^. 
 in standard Chinese and early Egyptian. 
 
 ib) Syllabic writing {i.e. with restriction of the 
 meaning of a phonogram to part of its 
 phonetic contents, a syllable), e.g. in 
 Japanese,^ the Semitic cuneiforms, etc. 
 
 ^ When the Japanese had borrowed the Chinese signs for the writing of their 
 own language, a new departure was found to be necessary. Their words consisted 
 not only of one, but of two, three, or more syllables. The Chinese phonograms
 
 speech and Writing 69 
 
 ic) Alphabetic writing, not yet clear of syllabism 
 {i.e. with the corrupted symbol of the 
 phonogram standing for its initial only, 
 that is to say, for a letter ; but so that, 
 while each consonant is expressed by such 
 separate symbol, the voivels are left un- 
 expressed), e.g. in Semitic and, in a less 
 degree, the Oriental derivatives of the 
 Semitic script. Egyptian exhibited this stage 
 along with the next. 
 
 {d) Alphabetic writing proper {i.e. with a special 
 sign for each sound), e.g. the Graeco-Italic 
 style, and, less completely, the later Egyptian 
 and the Persian cuneiforms. 
 
 could not, therefore, suffice for whole Japanese words. They could only 
 represent syllables. Hence about the beginning of the ninth century there was 
 introduced the system called the Katia, by which a select number of Chinese 
 characters (or at least abbreviations of them) were employed to represent syllabic 
 instead of word values. Thus there was a sign for each of the vowels a, e, i, o, ii 
 when used as a syllable, and one for a combination of each of these vowels with 
 each of the consonants (i.e. ka, ke, ki, etc., sa, se, si, etc. ). Meanwhile no provision 
 was made for representing k, s, etc., by themselves. Forty-seven such characters 
 thus formed a syllabary, which sufficed for the orthographic representation of the 
 language. As a fact, however, two varieties of the syllabary were either invented 
 or evolved, the one less, the other more popular, called respectively the Katakana 
 and the Hiragana. The characters used in the two styles differ widely, but the 
 syllabic principle is the same. More recently other modifications of orthography 
 have been introduced, and the system as a whole has become very involved and 
 perplexing. The invention of the Kana is commonly attributed to a certain 
 Kobodaishi, a Japanese who had studied Sanskrit while in China and had gathered 
 from that language the hint for distributing his series of characters. The 
 resemblance to the Sanskrit principle cannot be denied.
 
 CHAPTER IV 
 
 THE CLASSIFICATION OF LANGUAGE 
 
 Even a somewhat superficial examination of the languages 
 of the world suggests the possibility of classifying them ac- 
 cording to two different principles. We may consider them, 
 first, from the point of view of their methods of expression, i.e. 
 we may consider how far they agree or differ in their 
 manner of building words and in their grammatical devices ; 
 second, from the point of view of their material of expression, 
 i.e. we may consider what kinship, if any, exists between the 
 actual words which they respectively use for the expression 
 of the same ideas. In other words, languages may agree 
 with each other almost wholly, or in the main, or partially, 
 or not at all, in respect of the manner in which they take the 
 elements of speech and build words and sentences from them ; 
 and they may largely, or partially, or not at all, agree in the 
 actual material which they employ, i.e. in the use of the 
 same ' roots ' and ' formative elements,' in order to express 
 the same ideas. 
 
 Thus a comparison of Chinese with Japanese will reveal 
 the fact that, while Chinese possesses no forms but mono- 
 syllables, which admit of no inflexions or variations whatever, 
 Japanese has no such restriction as to number of syllables, 
 and does build up various word-forms of different shades of 
 meaning upon the common element vulgarly known as the 
 * root.' The term ' root ' is hereafter to be examined and 
 defined ; meanwhile, accepting it for convenience in the sense 
 popularly attached to it, viz. as the element of general 
 
 70
 
 The Classification of Language 71 
 
 meaning left in common to a group of words when each such 
 word has been divested of modifying or formative elements, 
 we may say of Chinese that it consists of nothing but 
 unchangeable and uncompoundable monosyllabic roots, 
 whereas Japanese can add one or several new elements to a 
 root, till different words are formed upon it, possessed of 
 different applications of meaning or different relations to the 
 sentence. Thus the Chinese word pab, with the idea of 
 ' protect,' is absolutely unalterable in form, whether by 
 addition, abbreviation, or internal change, whether it corre- 
 sponds to an English verb, noun, or participle, whether the 
 idea expressed is what we should call active or passive, whether 
 it refers to past, present, or future. There is no possible 
 variation such as we employ in ' protect,' ' protected ' ' pro- 
 tector,' ' protects,' ' protection,' ' protecting ' ; in the French 
 protegex, proteg^, protegQd,\s, ; or in the Latin protego, proteg\%, 
 protegexQ, protesi, etc. Differences of meaning implied by 
 such forms as these must be expressed in Chinese by other 
 devices of the sentence. On the other hand, in Japanese 
 we have such words as kuru = ' (I) come,' kureba = ' if (I) 
 come,' koi = ' come ' (imperative), kitai='{^) want to come.' 
 Here the common element is both modified by addition and 
 also phonetically varied. It is obvious that, in respect of 
 structural method, Chinese and Japanese cannot be classed 
 together. 
 
 A comparison of the same two languages in point of 
 vocabulary will further show that (except for a stock of 
 Chinese words which are historically known to have been 
 borrowed into Japanese, like the Latin vocables borrowed 
 into English) there is no perceptible agreement whatever 
 between the roots which belong to the two languages. 
 Therefore in respect of material also they cannot be classed 
 together. 
 
 With these instances let us proceed to compare the Latin 
 facio, faos,, feci, /edsti, con/idebam, and the English 
 ' come,' ' came,' ' comest,' ' coming,' ' forthcoming,' ' become,' 
 etc. In both instances the root is built into words repre-
 
 72 Natural History of Language 
 
 senting various applications of the general meaning, or 
 various functions in the sentence, by means of added 
 elements, sometimes accompanied by internal phonetic 
 change. So far Latin and English are in a certain agree- 
 ment with Japanese, but are markedly distinct in method 
 from Chinese. Further examination (for which this is not 
 yet the place) will again show that this structural agreement 
 Avith Japanese does not proceed very far, and that we shall 
 necessarily place Latin and English in a different sub-class 
 from that into which Japanese would fall. English, again, 
 would subsequently fall into a minor sub-class apart from 
 Latin. In point of material there is no resemblance what- 
 ever between the specific roots which form the basis of Latin 
 and English on the one side and those of either Japanese or 
 Chinese on the other. 
 
 Thus we discover that, of the languages already dealt 
 with, (i) when looked at with regard to structural principles, 
 Chinese stands in a class distinctly apart, while Japanese, 
 English, and Latin possess certain features which bring them 
 together, though (as has been hinted) it will subsequently 
 prove necessary to separate the two last from the first in 
 consideration of other features sufficiently pronounced. In 
 other words, Chinese is an entirely ' uninflected ' language, 
 while Japanese, English, and Latin are all ' inflectcdl though 
 in different manners or degrees : (2) looked at from the 
 point of view of the roots (or sound-groups) which are used 
 for a given notion by these languages respectively, Chinese 
 is found to stand apart ; Japanese also stands alone ; while 
 Latin and English are related to each other. 
 
 Classification from the point of view of method is called 
 Morphological ; from that of the material, Genealogical. 
 The former title is explained by the definition of morphology 
 previously given. Simply stated, it implies a division of 
 languages into classes according to the forms which their 
 words may take upon themselves for the expression of
 
 The Classification of Language JZ 
 
 grammatical relations and for determining special applica- 
 tions of meaning. The term ' genealogical ' implies that 
 languages which can be shown to possess a common basis 
 of root-material (together with a fundamental identity of 
 method) are historically related by an actual pedigree, the 
 relation of two such languages consisting in the derivation 
 of one from the other, or of both from a common source. 
 Thus, if the Latin expressions for one, two, three, etc., are 
 unus, duo, tres, etc. ; if its words answering to acre, horn, 
 mouse, etc., are ager, cornu, mils, etc. ; and if a scientific 
 examination of the two vocabularies shows that a corre- 
 sponding identity of radical material exists to a very large 
 extent throughout the languages, and also that the principles 
 of their morphology were ultimately the same, it is assumed 
 that the two languages are akin in the sense of possessing 
 historical relationship. We may, for instance, speedily place 
 English, German, Dutch, and Danish in one genealogical 
 class because of such obvious resemblance, a resemblance 
 which grows all the closer when scientifically examined. 
 We may assume as indubitable their community of origin 
 from a primitive Teutonic source. We may then go on to 
 discover that the most significant portion of the word- or 
 root-material which forms the staple of Greek, Latin, and 
 Sanskrit is again fundamentally identical with the material 
 serving the same functions in primitive Teutonic. The 
 genealogical class is thereupon widened to include those 
 languages also. In other words, a genealogical class of 
 languages embraces as many speeches as can be assumed 
 to derive their most significant material, together with the 
 framework of their structure, from a common origin.
 
 CHAPTER V 
 
 THE CLASSIFICATION OF LANGUAGE {continued) 
 
 A. Morphological Classification 
 
 A TRENCHANT and uncompromising division of the languages 
 of the world, as we find them, into clearly distinct morpho- 
 logical classes would be regarded as hopeless by an un- 
 prejudiced inquirer, that is to say, by one who did not 
 inherit the preconceptions and repeat the dicta of a past 
 generation of philologists. Luminous differentiations, 
 however desirable from an ideal point of view, can hardly 
 be made in all cases without outraging scientific truth. 
 
 Possibly, if we could see the various families of languages 
 at their earliest origins, we might (i) find their rudimentary 
 structural methods to have been entirely similar to each 
 other, and it may be that, in the length of time and the 
 course of their several evolutions, they have drifted more or 
 less widely apart. It is, however, quite conceivable that 
 yet further evolution, resulting in the survival of the fittest 
 methods, may once more bring them nearer to each other 
 after they have essayed their various paths. It is, for 
 instance, an obvious remark that, different as Anglo-Saxon 
 was morphologically from Chinese of the same date, yet in 
 many respects (such as in the expression of grammatical 
 relation by ' empty ' words and by mere position in a sent- 
 ence rather than by flexion) the method of modern English 
 is tending to a distinct similarity with that of modern 
 Chinese. 
 
 74
 
 The Classification of Language 75 
 
 Or (2) it may be that all languages have started with 
 the same morphological methods and have progressed along 
 practically the same lines with one another, but some with 
 more rapidity and some with less (a case of such difference 
 being manifest in a comparison of the rapid development of 
 English beside its more conservative sister German). The 
 different morphology of different languages, as we see them, 
 may thus represent different stages on the march of parallel 
 development. 
 
 Or, again, (3) it may be that from various incunabula, or 
 cradles of language, languages started on their career with 
 very dissimilar methods, and that, in feeling their way 
 towards their own perfection, they approach each other at 
 certain points, while at others they remain apart. 
 
 Whatever the truth may be, we have no right to set out 
 with any one of these assumptions, although, after a colla- 
 tion of the evidence, it is the business of the philologist to 
 speculate upon the matter and to decide for the probability 
 of one view as against that of another. 
 
 The assumption has been generally, though not univer- 
 sally, made since the days of Schleicher (more than a 
 generation ago) that the second of the three above-named 
 possibilities is the only reasonable one ; and there has 
 followed a tendency to detect and mark off the supposed 
 stages of that uniform development with more regard to 
 symmetrical simplicity than to the facts of the data. 
 
 For the most part languages have been divided into 
 three morphological classes, which are generally assumed to 
 stand in an ascending order of progress, viz.: — (i) the 
 'monosyllabic,' 'isolating,' or 'radical' languages, such as 
 Chinese; (2) the 'agglutinating' languages, including 
 tongues as unlike in structure as Turkish and Japanese ; (3) 
 the ' inflexio7ial' 'organic^ or 'amalgamating' languages, 
 such as the Semitic and most of the European. 
 
 The first class in this too symmetrical account consists 
 of languages, such as Chinese, in which the words are all
 
 76 Natural History of Language 
 
 nothing but simple ' roots,' incapable of receiving suffixes, 
 prefixes, or any other modification, internal or external. A 
 sentence consists of root following root ; and if ' R ' stands 
 for ' root,' ni for its internal modification, and a for affixed 
 element, the formula for a sentence in these languages, in 
 their ideal conception, may be given as R, R, R, R, with no 
 employment of either ;// or a. 
 
 The second class ideally comprises languages such as 
 Turkish and Magyar, in which the chief ' root,' or element 
 containing the main idea of a word, may receive, in the 
 shape of suffixes or prefixes, additions of other elements for 
 the purpose of defining more closely the application of that 
 main idea, or of expressing its grammatical relation to other 
 w^ords in the sentence. These added elements, though they 
 have no independent existence in actual speech (whatever 
 they may once have had), are, in the ideal conception of 
 such languages, distinctly expressive of the required 
 modifications, and are attachable and detachable at will, the 
 main root meanwhile being left unobscured. The term 
 ' agglutinative ' implies that, theoretically, the elements are 
 ' glued on ' to each other, not dovetailed or amalgamated. 
 The formula for the single word in these speeches may be 
 given as Ra" or a^'R, where n implies that there may be an 
 indefinitely large number of the affixes. Here again vi does 
 not, ideally at least, come into the formula. 
 
 The third class consists of languages in which the element 
 expressing the main idea {i.e. the ' root ') is capable of re- 
 ceiving, in the shape of affixes, accretions of other elements 
 for the same purpose as in the second class. But these 
 accretions may become so welded with each other and with 
 the main root as to be separately inextricable in the 
 consciousness of the speaker, while the main root itself may 
 be indefinitely modified in sound and shape, so as to be 
 often quite unrecognisable for the same in the various words 
 in which it occurs. The formula for a word in these 
 languages is therefore R"'a""'. [Thus, to take a simple, but 
 purely imaginary, case. Suppose (though the supposition is
 
 The Classification of Language yj 
 
 vain) that we could find the same roots for the same ideas 
 existing in a language of each of these classes. Let EI = 
 * go ' and ENT = ' they.' The three types might respectively 
 give (l) EI ENT (or ENT EI), (2) EIENT (or ENTEl), 
 (3) EIENT, EINT, EIN, INT, IN, etc., without any care 
 to distinguish the components of the word or to maintain 
 one form of the root. So if ES = ' be ' and SI = ' thou,' we 
 might have (i) ES SI (or SI ES), (2) ESSI or SIES, (3) 
 ESSI, ESS, ES, E(S)I {i.e. El), SI, S, etc.] 
 
 This triad classification has been much in vogue, and is 
 apt to be repeated in the literature of the subject as if it 
 were satisfying and exhaustive. Of those who so endorse 
 it the greater part, moreover, take it for granted that a 
 language whose place is in the third class has necessarily 
 at some time passed through the second, and, still earlier, 
 belonged to the first. That is to say, it is assumed that a 
 typical inflected word grew out of what was once clear 
 agglutination, in which the main root stood distinct while 
 the affixes v/ere obvious attachments with a separable form 
 and function ; and that this stage was preceded by another, 
 in which the elements destined to become affixes were still 
 independent words, merely juxtaposed to the main root and 
 to each other, in the same way in which the French prudeni- 
 inent derives from a Latin juxtaposition prudeiite (or 
 prudentt) mente. A small minority, however, are disposed 
 not only to assert that the assumption lacks adequate proof, 
 but also to deny that it is a priori any more probable than 
 the reverse development. The discussion of this question, 
 so far as it is arguable upon the data, may be reserved for 
 treatment later in this work. Meanwhile a few typical 
 examples will show the difficulty of bringing the varieties of 
 extant languages under any triple morphological classifica- 
 tion whatsoever. 
 
 I. In most of the native American tongues the ex- 
 pression of a simple sentence regularly takes the form of a 
 single word, often of unwieldy length, which consists of an
 
 78 Natural History of Language 
 
 interweaving or amalgamation of merely the most significant 
 sounds of those different sense-elements which would in most 
 languages stand as separate words. There is thus, normally, 
 no such thing as a transitive verb without object and subject 
 incorporated into it. From Greenland is reported as a stock 
 example a sentence-word aiilisariartorasicarpok, ' he hastens 
 to go fishing,' of which the component factors are mostly 
 truncations, corresponding respectively to aulisar {' to fish'), 
 peartor (' to be engaged in '), and pmtiesuarpok (' he hastens '). 
 Again, from Mexico, nisotsitemoa, ' I seek flowers ' (literally 
 ' I-flowers-seek '), and nikalcliihiia, ' I build a house' (literally 
 I-house-build '). Cherokee gives nadholinin^ ' bring us the 
 boat,' which is analysed into the components naten (' bring '), 
 amokhol (' boat '), and iiin (' us '). The case is usually stated 
 as if the component elements were absolutely incapable of 
 existence as individual words. Sometimes, however, such 
 sense-factors must be ejaculated alone. Moreover, Green- 
 landish can and does form certain sentences in which single 
 words are left clear and are marked as subject or object by 
 means of suffixes. Mexican, again, can and does in- 
 dividualise ' words ' in the sense more familiar to us, 
 e.g. sotsitl, ' flowers.' That is to say, what is known as 
 Polysynthesis or Holophrasis is to some extent varied by 
 methods which are practically flexional. 
 
 II. The Basque language, spoken about the extreme 
 Western Pyrenees, is quite distinct from the American 
 languages in respect of this fusing of a zv/iote series of words 
 into one compound which forms the sentence. It is not 
 polysynthetic in that sense. Nevertheless, just as it is the 
 dominant principle of the American languages to pack all 
 the other elements of a sentence into a framework consisting 
 of a verb and its subject (as in the Mexican ' I-flowers-seek,' 
 or ni-k-tle-watsa, ' I-it-(with)-fire-roast '), or in some way to 
 weld objective and qualifying matter into an amalgam with 
 the verb, so in Basque the same phenomenon occurs, though 
 in a more limited degree, the amalgamated elements being 
 only the direct and indirect objects, the subject, and the verb.
 
 The Classification of Language 79 
 
 The incorporation, moreover, is confined to pronouns. Thus 
 d-{a)-kar-ki-o-t ( = ' it-carry-to-him-I ') forms but one word, 
 dakarkiot, ' I-carry-it-to-him.' Similarly n-{a)-kar-su = ' me- 
 carriest-thou,' /i-(a)-kar-t = ' thee-carry-I,' h-{a)-bil-ki-t = 
 ' thou-goest-to-me.' In Basque the verb proper has no 
 existence apart from its pronominal complements. This fact 
 removes the language from serious comparison with French 
 or Italian, when those tongues exhibit a milder species of 
 pronoun-incorporation in a disguised form. The French 
 je te le disais (' I-to-you-it-said ') and the Italian {io) glielo 
 dissi (' I-to-him-it-told ') are, it is true, but slightly dis- 
 tinguished from a zovCiOvc\-dJi\or\jetle disais (zdthdlze) ox glielodissi. 
 In some cases orthography recognises the partial unification, 
 as in the French donnez-le-lui and the Italian darvelo (' give- 
 to-you-it '). Nevertheless the French or Italian verb enjoys 
 regularly a separate existence, and even partial incorporations 
 of the kind illustrated are the exception. It is worth while, 
 however, to note this point of approximation in languages 
 otherwise so unlike. 
 
 Apart from this degree of polysynthesis Basque is 
 agglutinating-inflectional. For instance, it expresses the 
 grammatical relations of nouns by a declension with suffixes, 
 e.g. gizon, ' man,' gizon{e)z, ' by man,' gizong'axiASis.^ ' from the 
 man.' 
 
 III. In the Bantu (or Zulu- Kafir) family of South 
 African languages we find new phenomena. The words are 
 regularly separate in the sentence, their modifications of 
 sense and their grammatical relations to each other being 
 expressed by elements mostly (though not solely) /r^fixed. 
 For example, in Zulu ng-umu-ntu = ' with (the) man,' ng-aba- 
 ntu = ' with (the) men,' ng- representing the sense ' with,' 
 while the variation ada-, as against umu-, conveys the idea 
 of plurality. Moods and tenses are likewise formed by 
 prefixes. The Bantu speeches meanwhile exhibit a certain 
 amount of the above-mentioned incorporation of the objective 
 pronoun. Sinitatida ( = si-xa-tandd) = ' we-it-love ' ; while 
 i-/-ba-/««^« = ' we-them-love.' Here again we may compare
 
 8o Natural History of Language 
 
 (with a reservation) the French 7ious raimons, nous les 
 aiuions. 
 
 IV. In Hungarian (Magyar), Turkish, and kindred 
 languages the ideal principles of declension, conjugation, 
 and word-formation in general are those described above as 
 * agglutinative.' A typical example is the Turkish 
 
 Sevmels., ' fo love ' 
 
 sevinmek, ' to love oneself 
 
 sev\^inek, ' to love 07ie another ' 
 
 sei/dixniek, ' to make to love ' 
 
 sev\S.viek^ ' to be loved ' 
 
 sevxsneitnek, ' not to love ' 
 
 j'^c'indirilmew^/^, ' Jiot to be made to love oneself 
 
 Similarly in Magyar ;:«'r= ' shut,' £r(irat = ' locks ' (causa- 
 tive), zdr&tgdit — ' he often shuts ' (frequentative causative), 
 jsdraXgaXh&t = ' he can often shut ' (potential frequentative 
 causative). 
 
 Compare this with a declension in Turkish — 
 
 ' house ' Nom. Plural ev\QX 
 
 evlerin 
 evlere 
 evleri 
 evlerA& 
 evlerdi%Vi 
 
 Here we have the root left unchanged throughout, and 
 its grammatical relations or modifications of sense expressed 
 by post-positions or suffixes, which can be easily detached 
 by the mind and recognised as distinct elements with a 
 distinct function. How distinct that function is may be seen 
 from the above declension of ev, where the same sign is used 
 for the same case whether in the plural or the singular. If 
 the main 'root' = ^, the case-suffix =j, and the plural sign 
 = x\ the case for the singular is Ay, while that for the plural 
 is Axy. The number of such additions is limited only by 
 their practical utility. But, though this is the ideal morpho- 
 logical principle of the Ural-Altaic family of languages, it is 
 far from being the case that the suffixes can always be thus 
 
 Nom. 
 
 ev, 
 
 Gen. 
 
 ^■zdn 
 
 Dat. 
 
 evQ 
 
 Ace. 
 
 evi 
 
 Log. 
 
 evdiQ 
 
 Abl. 
 
 ^7'den
 
 The Classification of Language 8r 
 
 detached and conceived of apart from the root as possessed 
 of any original independent existence. In Finnish, for 
 instance, there is almost as complete an amalgamation of 
 flexional elements with the root as in the Latin ag&t, agri, 
 agro. 
 
 Within the Ural-Altaic family also there occurs occa- 
 sional ' incorporation ' of a pronoun, eg. Magyar variak = ' I 
 await thee.' Prefixes occur also, as in Magyar, but these 
 languages have no power of compounding in any other 
 sense. 
 
 It is also noteworthy that in the vernacular speech 
 (which is the real language) of certain peoples whose 
 languages are usually included in this class, there is often 
 a polysynthesis which could not be distinguished from 
 that of the American peoples, if we did not possess 
 information to make us artificially conscious of the separate 
 elements. Thus in Kalmuk dgiingddshi baitiai bi (' I shall 
 soon go ') becomes by mere rapidity or economy of articula- 
 tion dgilngddshdndb, while iisddshi bai?iu tsJii is welded into 
 iisddshdniitsJi. Dealing simply with the facts as they are, 
 and not inquiring into origins, we cannot deny that these 
 instances and their like represent sheer polysynthesis. 
 
 V. In Latin, Greek, Sanskrit, and their kindred 
 languages, in their synthetic stage, grammatical relations 
 and modifications of meaning are expressed, as in the pre- 
 ceding class, by suffixes ; but these are for the most part 
 indistinguishable as anything more than modifications of the 
 stem-termination. They have no independent existence, nor 
 are they recognised as regular and distinct in one definite 
 function. In these tongues the root part of the word is 
 susceptible of complete obscuration, while the suffixes also 
 undergo phonetic corruption to an unlimited extent. The 
 roots and suffixes may be welded together into one inextric- 
 able mass, so that the speaker is unconscious, and science 
 can often hardly discover, how much is root and how much 
 is suffix or suffixes. Nevertheless, the close analogy of 
 principle itself (viz. of modifying meaning and relation by 
 
 (n 884) G
 
 82 Natural History of Language 
 
 suffixes) between this group and the preceding is obvious. 
 As a matter of principle the Latin declension vir, viri, viro, 
 vzram differs in no respect from the Turkish ade?n (' man '), 
 ademia, ademe, ademi ; nor does the system ago, ^^^> ^.^tj 
 agixsiyiiS, differ in principle from the Magyar /^/em, /^/ed, leli, 
 Mjxtk. The distinction lies in the transparence and regularity 
 of the word-building process in the Ural-Altaic group, and 
 its lack of uniformity and its unrestricted obscuration in the 
 Indo-European, The exact original meaning of the re- 
 spective suffixes of case or person may or may not have been 
 different for the two series ; in effect they amount to the 
 same in function. For this reason it is unscientific to use 
 the current phraseology, by which an ' agglutinative ' is 
 distinguished from an ' inflexional ' class. Agglutination is 
 inflexion, and if another term is required to distinguish the 
 more patent and detachable from the obscured and inextric- 
 able flexion, the latter may more properly be styled ' amal- 
 gamating,' while ' inflexional ' should stand for the larger 
 division, which includes both the ' amalgamating ' and the 
 * agglutinating.' In further illustration of the difficulty of 
 definitive demarcation, it may be added that, in the stage of 
 synthetic Greek, Latin, etc., which we call ' classical,' those 
 languages already tend to help out the grammatical relations 
 by means of other devices than simple flexion, such as 
 prepositions and auxiliaries. Thus, in Latin in iirbe 
 manere, ex urde inovere, take the place of an older urb^ 
 manere, urbe inovere ; and in Greek viro, 8id, etc., with the 
 genitive are substituted for a simple case like the instrumental. 
 In this class of languages there is no power of incor- 
 porating, though there is much of compounding, prefixes 
 being in all cases originally non-flexional and the result of 
 composition. 
 
 VI. The lineal descendants of the languages in the 
 preceding group tend to an analytical form, ie. they tend to 
 render unnecessary or to replace the inflexional terminations 
 of the synthetic stage (i) by the use of pronouns, auxiliary 
 verbs, prepositions, etc., and (2) by utilising position and
 
 The Classification of Language 83 
 
 order in the sentence. Generally speaking, they make a 
 practice of expressing by pre-posed ' demonstrative ' words 
 the same relations which were formerly expressed by 
 terminational (post -posed) demonstrative elements. The 
 extent to which analysis has yet been developed differs 
 much in the various languages. English and Persian are 
 examples of the most advanced. For purposes of com- 
 parison of method we may take the French 7non frcre a vu 
 le rot, as contrasted with the Latin vieics frdtcr vldit regevt, 
 or f rater mens vldit regent, or rcgeni vldit frdter mens, etc. 
 In Latin the order of the words admits of numerous varia- 
 tions. In French it is fixed. Similarly in le roi a vu mon 
 frere the nouns are without flexion, and therefore without 
 option of position, whereas the Latin terminations permit of 
 the order rex vldit frdtrem mcnvi, ox frdtrem meiini vldit rex, 
 or vldit rex frdtrem meum without ambiguity. These 
 instances will illustrate the character of the change, as it 
 appears in the most advanced form to which analysis has 
 yet attained in French. English has become even more 
 flexionless. ' We have,' ' you have,' * they have ' are all that 
 is left to correspond to original Teutonic differences which 
 were as wide as are Jiabexims,, habcWi,, kabeut in Latin. 
 French, in respect of its verbs, still combines an inheritance 
 of synthesis with the newer analytical method, a fact which 
 is evident in nous avons, vous avez, ils ont. German, again, 
 has retained, though in a modified form, a case declension, 
 which English and the Romance languages have cast off. 
 If we take English as the most advanced type, we discover 
 the principle of analysis to be the expression of grammatical 
 relations and modifications of the idea by means of separate 
 * demonstrative ' or ' presentative ' words and by means of 
 position in the sentence. 
 
 Thus the almost monosyllabic sentence ' I go to town 
 every day and come back at the same hour from one end of 
 the year to the other,' contains not one mark of flexion. 
 Compare this with the Latin ro cottldie in urbeia. ct inde 
 ekdem horh ab initio ann\ usque ad flnQva. redeo, which,
 
 84 Natural History of Language 
 
 however, itself shows a share of the analytical in respect of 
 its prepositions, 
 
 A kind of reaction in the shape of a limited degree of 
 polysynthesis is met with in the commoner phrases of such 
 analytical tongues, the separate words running into each 
 other in abbreviated forms. Thus the French qu'est-ce que 
 c'est que ga ? is in popular speech a complex of the form 
 kekseksa ? In German zu der becomes zur, bei dem becomes 
 beim. In English don't, shan't, don, doff, you're, are simple 
 examples of the process. Goodbye = God be wH ye. Such 
 forms are, doubtless, the accidental outcome of haste or 
 indolence, but there is nothing to show that the American 
 Indian polysynthesis is not also similarly accidental. 
 
 VII. The group of languages to which belong Chinese, 
 Anamese, Siamese, etc., exhibits a character unusually con- 
 sistent. It is monosyllabic, and possesses no true compounds 
 nor any flexion whatever. Neither by suffixes nor by pre- 
 fixes proper does it show what part a word plays in a 
 sentence. By far the chief, though not indeed the only, 
 grammatical device is the position of the words. The order 
 is subject, predicate, object. The qualifying word precedes 
 the word qualified. Thus in Chinese, ' I beat thee ' is ngb 
 td ni (' I beat thou '), but * thou beatest me ' is ni td ngb 
 (' thou beat I ')• * Great state ' is td kuok, but ' the state is 
 great' is kuok td. Nevertheless, this device, which is 
 theoretically sufficient, is in practice assisted by the use of 
 certain series of words which have come to be virtually 
 ' demonstrative,' i.e. to play the part of determining elements, 
 such as prepositions, signs of number, etc. Certain words 
 (called by the Chinese ' empty ') are employed in a symbolic 
 use to indicate relation. Thus /// {tchi) which originally 
 means ' place,' has come to be used in positions where it 
 practically = ' of.' Thus ' the power of the people ' is viin tsi 
 lik, an expression in which inin = ' people,' tzk = ' power,' and 
 tsz is symbolic. However independent a meaning tsi may 
 once have possessed, it here amounts simply to the 's in 
 * people' j-,' and it has only to be attached more closely to its
 
 The Classification of Language 85 
 
 preceding word to form an actual inflexion. In so far as 
 such empty words follow a noun and are included under one 
 accent with it, we have a crude form of suffixing. Yet the 
 flexionless character of this group of languages gives them 
 an obvious differentiation from any hitherto mentioned, 
 though it must be evident that analytical Enghsh is 
 approaching them in point of syntactical expedients, though 
 not in point of word-content and word-building. English 
 can still construct words by combining certain factors on 
 existing models, while Chinese can build no such words, 
 because it possesses no such patterns. 
 
 VIII. The Semitic languages (Hebrew, Arabic, etc.) 
 exhibit morphological peculiarities not revealed in any speech 
 yet mentioned, while in other respects they agree with other 
 groups. In point of external flexion, i.e. elements external 
 to the ' root,' they possess, or have possessed, suffixed signs 
 of case, number, gender, and person, which, however dis- 
 similar in form and exhaustiveness to those of Latin, Greek, 
 etc., are in no way different from them in principle. 
 
 Thus in classical Arabic there is a noun-declension 
 which may be illustrated as under : — 
 
 (Singular) (Dual) 
 
 (Plural) 
 
 Norn. abdVi abdzXii 
 
 abduna, 
 
 Ace. abddi, abd^^SxiX 
 
 abdina, 
 
 Gen. abdl abddXni 
 
 abdina, 
 
 Similarly in the verb-system : — 
 
 
 1st pers. sing. katabtv,., 
 
 ' I have written ' 
 
 2nd ,, „ (masc.) katabtz. 
 
 
 „ ,, ,, (fern.) katabti 
 
 
 3rd ,, ,, (masc.) katabdi 
 
 
 ,, ,, ,, (fern.) katabdXt etc. 
 It is often impossible to distinguish its grammatical procedure 
 from the agglutinative method of Turkish or Magyar. For 
 example, in the Arabic scheme 
 
 ki/db\, ' my book ' 
 
 /.'/Ar/mka, 'thy (masc.) book' 
 
 kitdb\ji]i.\, 'thy (fern.) book' 
 
 kzidb\xh.Vi, 'his book ' 
 
 kiidb\x\ia., ' her book,' etc.
 
 86 Natural History of Language 
 
 the agglutination of elements is quite as direct and trans- 
 parent. The only difference of practice lies in the very- 
 limited extent to which it can be carried out in Semitic. 
 In this group, again, we find in later times the tendency to 
 analysis, as when Hebrew loses its declension-endings and 
 replaces them by particles. 
 
 In the treatment of the ' roots ' themselves, however, 
 these languages show a remarkable peculiarity. The root 
 arrived at by induction consists regularly of three consonants, 
 into the framework of which any of the vowels may find 
 their way according to modifications of function and meaning. 
 Thus from a ' root ' qtl there are in use in their several 
 functions such forms as qtel, qtol, qtzl, qtal, qtul, qutl, qutel, 
 qutal, qatal, qdtal, qetol, qetol, etc. Thus qutila, ' he was killed,' 
 qatala, ' he killed,' aqtala, ' he caused to kill,' yaqtulu, ' he kills,' 
 qdtil, ' killing.' It is true that there appears to be something 
 of the kind in the Indo-European phenomena technically 
 known as ' Ablaut ' and ' Umlaut,' as in swim, swam, swum ; 
 long, length ; ireiOw, eiridov, ireTroida {j)eitho, epithon, pepoithd) ; 
 fides, fido, foediis. These variations, however, depended 
 originally on accentuation, not on meaning, and they are now 
 only an occasional and irregular phenomenon in extant Indo- 
 European tongues, while in the Semitic languages they are 
 wide in range, systematic, and of the very first importance. 
 They are of the essence of the language, the most prominent 
 device for the modification of sense. In other words, they 
 are ' dynamic,' and serve a necessary morphological purpose, 
 which is not the case with the vowel-changes of Indo- 
 European. 
 
 The Semitic languages further admit of no large accretion 
 of suffixes and of no compounding. 
 
 The eight groups selected above are merely main types, 
 and are by no means exhaustive. They have been chosen 
 simply in order to illustrate the difficulties which lie in the 
 way of any brief classification which is to be both trenchant
 
 The Classification of Language 87 
 
 and true. It may be well to review them briefly for the 
 same purpose. 
 
 Type I. has for its principle polysynthesis or complete 
 incorporation. Such languages have been styled Poly- 
 synthetic, Incorporating, or Holophrastic [i.e. speaking in 
 words which are whole sentences). At the same time it 
 must be recognised that several of them tend towards more 
 resolved forms and to a species of flexion. Frequently also 
 the polysynthetic process might just as well be called 
 agglutinative. The Chilian ein\i (' to give '), ^/«guen (' to 
 give more '), ^/^duamen (' to wish to give '), ^/?^valen (' to 
 be able to give ') represent a system quite analogous in 
 method to that illustrated for Turkish with scvm^^, etc. 
 
 Type II. is ' incorporating' only in so far that pronominal 
 elements are ' included.' In other respects it is practically 
 an inflected language. To call Basque simply ' incorporating ' 
 is to mislead. 
 
 Type III. consists of 'agglutinating' languages which 
 express relation and modification mainly by means of 
 prefixed elements. Pronoun-incorporation is not alien to 
 them. 
 
 Type IV. consists of the typical ' agglutinating ' languages, 
 expressing relation and modification by means of suffixes, 
 which are theoretically unlimited in number and represent in 
 that sense a certain measure of polysynthesis, and which, 
 theoretically again, are easily recognised as detachable from 
 each other and from the main root. It must be granted, 
 however, that this ideal is seldom reached, and that the 
 suffixes are frequently no more than flexional ' signs ' which 
 cannot be conceived as existing apart from the main body of 
 the word. Moreover, whatever their shape and transparency, 
 they do not differ in function or purpose from those 
 employed, though obscured, in the amalgamating languages 
 which next follow. Pronoun-incorporation is found sporadic- 
 ally to a small extent. 
 
 Type V. consists of languages commonly spoken of as 
 ' inflexional,' sometimes as ' organic' We shall prefer to
 
 88 Natural History of Language 
 
 call them ' amalgamating.' These modify the sense and 
 relation of words by variations of the terminating elements, 
 which are either demonstrably suffixes or are generally 
 admitted to have been such in their origin, but which have 
 mostly lost their individual shape and signification and have 
 often become inextricably welded into the main root. 
 Often, however, the elements are still as distinct in their 
 ' agglutination ' as those in the type last treated. At the 
 same time all such languages, so far as we can meet them in 
 history, are partly analytic, like the next type ; i.e. they 
 tend to assist the purely flexional method by means of 
 prefixed pronouns, auxiliaries, prepositions, etc. 
 
 Type VI. comprises languages in which the ' analytic ' 
 method prevails ; i.e. the expression of relation is mostly 
 transferred from undetachable suffixes to separate prefixes. 
 Fldxion still exists in greater or smaller measure, but is no 
 longer the essential and vital principle. The chief syntactical 
 principle is position. 
 
 Type VII. embraces those languages, quite devoid of 
 flexion, which have been named ' radical ' or ' isolating.' 
 The term ' monosyllabic ' is less philosophical, since the 
 ultimate question is not how many syllables a word contains, 
 but whether it contains or does not contain in itself any 
 element which marks off its grammatical position in the 
 sentence. Nevertheless, though flexionless, these languages 
 do employ certain signs of relation in the shape of separate 
 words, which perform in a limited degree the same function 
 as flexional suffixes or analytical prepositions. Their essential 
 principle, however, is, as in the last type, position. 
 
 Type VIII. includes languages which possess suffix- 
 flexion and to some extent prefix-flexion, with a tendency to 
 the analytical ; but which also systematically represent 
 modification of sense by internal variation of the vowels of 
 the 'root.' These are usually spoken of simply as 'inflexional' 
 and are unsatisfactorily classed with type V. 
 
 Thus in these eight types alone, considered in their 
 chief features, we have : —
 
 The Classification of Langtiage 89 
 
 1 . Essentially Polysynthetic, Holophrastic, or Incorporating 
 tongues ; such as the American Indian. 
 
 2. An inflected tongue habitually incorporating pronouns: 
 e.g. the Basque. 
 
 3. Prefix-Agglutinating languages : the Bantu family. 
 
 4. Suffix- Agglutinating languages : e.g. the Ural-Altaic. 
 
 5. Suffix-Amalgamating : the Indo-European tongues in 
 their older stage. 
 
 6. Semi- inflexional, or Analytical (largely positional) \ 
 most Indo-European languages in their modern shape. 
 
 7. Isolating, or Radical-positional : Chinese, etc. 
 
 8. Suffix-and-Prefix- Amalgamating, with dynamic vari- 
 ability of the root : the Semitic tongues. 
 
 It has already been pointed out that even these more 
 salient types touch each other at various points and in some- 
 what complicated ways, as well as that the principle is often 
 virtually the same, even when the results of its application 
 render a convenient difference of terminology possible, as in 
 the expressions ' agglutinative' and 'amalgamating.' When 
 we come to deal with yet other tongues less saliently and 
 consistently characterised, such as Japanese, Malayo-Polynesian, 
 or the Dravidian languages of Southern India, it becomes 
 practically impossible to place them uncompromisingly under 
 any morphological description indicated above. Where the 
 words of a language employ different endings to express 
 different relations, the question whether such endings con- 
 stitute ' amalgamation ' or simple ' agglutination ' is generally 
 a difficult one to determine, the distance from the clearly 
 detachable modifying suffix to the undetachable inflexional 
 termination being marked by every possible gradation. In 
 some cases the evidence has been flouted. Thus the 
 Dravidian languages are usually placed among the agglu- 
 tinative. Yet such a system as kal (' stone '), kallzx (accus.), 
 kallin (gen.), differs in no morphological respect from the 
 Latin lapis, lapideva., lapidis. The endings are flexion-
 
 90 Natural History of Language 
 
 endings ; for they are certainly not conceived by the speakers 
 as having a substantial value in an abstract shape. The 
 same remark applies to, e.g., Finnish viaa (' earth '), maan 
 (gen.), niaaWz. (instrumental case). The Polynesian tongues 
 are commonly classed with the agglutinative (apparently 
 because their cognate Malay is mainly agglutinative), though 
 the cases, persons, tenses, and moods are formed by particles 
 in as purely analytical a way as in English. Thus in Maori 
 te tangata = ' the man,' o U tangata = ' of the man,' ki te 
 tangata = ' to the man,' etc., follow principles identical with 
 those of English. In Japanese, which again is relegated to 
 the same class, we find that the relation of nouns is indicated 
 by the post-position of separate particles, e.g. de, ' by,' m, ' in,' 
 no, 'of,' ye, 'to' or 'at' Thus hasami de kirn (literally 
 ' scissors-with to-cut '), neko no tsume (lit. ' cat-of claws '), 
 gakko ye (lit. ' school-at '). The line of demarcation between 
 forms like these and a fully agglutinated hasami^% nekono, 
 gakkoYQ is so slight that it appears optional to treat the 
 particles as actually affixed or not ; but the line between 
 these and the ' full ' and ' empty ' words of the ' isolating ' 
 Chinese mm tsi (' people-of ') is almost as slight. If, again, 
 it is laid down (as it usually is) that one criterion of an 
 agglutinative language is the distinct phonetic preservation 
 of its ' roots,' as distinguishable from its suffixes, it is 
 obviously impossible to treat Japanese as agglutinative in 
 view of such paradigms as kitai, kureba, koi, from the same 
 element for ' come,' or sureba, shiinasu, from the same element 
 for ' do.' Gozaunasu (' shall be ') is colloquially — and, after 
 all, it is the colloquial which constitutes the essence of 
 language — turned into gasu or gesu, a phenomenon which is 
 directly opposed to the ideal agglutinative principle. The 
 case is the same with Basque, another language usually 
 called agglutinative. Dadiikat (' I have it ') becomes daut or 
 dut, the ' root ' of which Van Eys gives as eduki, while the 
 verbal theme is often simply eu. That the principle upon 
 which the Japanese so-called verb-forms were built up was 
 originally agglutinative is highly probable, but the same
 
 The Classification of Language 91 
 
 may be said of the amalgamating tongues. Certainly the 
 agglutinative principle is not kept clear. Japanese saniasanai 
 (' I do not make cool ') is, according to Mr. Chamberlain, 
 composed of sain (' cool ') + as (connected with suru, ' to do ') 
 + a (negative base-inflexion) + nai (negative verbal form 
 expressing non-existence). No one can maintain that these 
 various accretions are felt in their individual force as 
 detachable, nor do they differ in any classifiable manner from 
 forms of the amalgamating morphology, such as the Gothic 
 aflageinai (' for the remission '), a word composed of af (^ off') 
 + lag ('lie') + ^z (causative suffix) + «/ (verbal noun-sign) 
 -H dative case-ending. 
 
 The fact is that languages overlap each other's morpho- 
 logical borders and invade each other's territory at so many 
 points that a scientific frontier is impossible. We must not 
 limit the number of morphological classes actually existing 
 and then force a given language into one or other and 
 exclude it from participation in any of the rest. 
 
 Among attempts of philologists to divide languages 
 morphologically we find the following supported by great 
 names : — 
 
 F. and A. W. Schlegel— 
 
 (i.) Languages without inflexions, 
 (ii.) Languages with affixes. 
 (iii.) Inflexional languages. 
 
 Grimm — 
 
 (i.) Languages without inflexion, 
 (ii.) Inflexional languages, 
 (iii.) Analytical languages. 
 
 Bopp — 
 
 (i.) Languages with monosyllabic roots without power 
 
 of composition, 
 (ii.) Languages with monosyllabic roots which admit of 
 
 composition into grammar, 
 (iii.) Languages with disyllabic roots capable of interior 
 and exterior modification.
 
 92 Natural History of Language 
 
 Pott— 
 
 (i.) Flexional languages with complete amalgamation of the 
 
 word-elements, 
 (ii.) Agglutinative languages with external union of the 
 
 elements, 
 (iii.) Isolating languages without power of union, 
 (iv.) Incorporating languages in which there is no difference 
 between word and sentence. 
 
 Schleicher — 
 
 (i.) Isolating languages with roots merely juxtaposed, 
 (ii.) Agglutinative languages with external union of uncor- 
 rupted and unobscured main roots with subordinate 
 modifying roots which admit of obscuration, 
 (iii.) Inflexional or amalgamating languages with all the 
 word-elements capable of corruption. 
 
 It will be seen that no two of these methods of division 
 are identical, though there is a little difference between that 
 of the Schlegels and that of Schleicher. It will also be seen 
 that there is a measure of justification for each, while, at the 
 same time, each is unsatisfactory. That of Schleicher, upon 
 which we have laid special emphasis in consequence of its 
 great vogue, is open to particular objection, since, as Whitney 
 says, it classes Indo-European and Semitic as morphologically 
 alike, " while their structural discordance is vastly greater 
 than that which separates Indo-European from the aggluti- 
 native tongues." It is open also to the objection that it 
 denies agglutination to be a form of inflexion. The inevitable 
 conclusion is that each philologist has either divided too 
 much or too little, or rather that he has neglected to make 
 the proper main divisions first and the subordinate divisions 
 afterwards. If we begin afresh and without prejudice we 
 can only divide languages into two main classes : — 
 
 I. Languages which express grammar and modification 
 
 of sense by position, without external or internal 
 
 modification of the ' roots.' These may be called 
 
 Inorganic or Positional. 
 
 II. Languages which do employ the device of external 
 
 i
 
 The Classification of Language 93 
 
 or internal modification in different manners and 
 measures. For convenience these may be called 
 Organic. 
 
 The Positional languages include Chinese, Burmese, 
 Anamese, and their group. The Orgfanic include all the 
 rest, though the extent to which they share in the organic 
 quality varies to such a degree as to admit of much, though 
 by no means clear-cut, or mutually exclusive, subdivision. 
 
 In proceeding to subdivide the organic tongues, we must 
 be content to place a language under that head to which its 
 phenomena seem most prevailingly and consistently to refer 
 it. If, for example, a language with inflecting affixes {i.e. 
 either suffixes or prefixes) appears in a marked degree to 
 preserve root and affix distinguishable from each other, we 
 may refer it to a class called ' agglutinative,' though it may 
 exhibit no few departures from that principle and may 
 appear at times as 'amalgamating' as the amalgamating 
 tongues. So also vice versa. In other words, the classifica- 
 tion is one rather of ideal types, to which, in practice, 
 existing languages conform but partially. The ascription 
 of a given language to a given class is often to be made on 
 the ground that it is less removed from the norm of that 
 particular type than from any other. 
 
 Accepting for convenience the distinction (which is not 
 fundamental, but is worth recording as a phenomenon in 
 extant fact) between ' agglutinating ' and ' amalgamating ' as 
 above defined, while reserving the term ' inflexional ' to 
 include both species, we may divide the organic tongues 
 with approximate correctness as follows : — 
 
 A. Inflected — 
 
 I . Amalgamating", zvithout internal dynamic change — 
 {a) Synthetic (Sanskrit, Greek, Latin, Gothic, 
 
 etc.). 
 {b) Analytic (English, Persian, and most 
 modern Indo-European languages in 
 various measures).
 
 94 Natural History of Language 
 
 2. Amalgramating, with internal dynamic change — 
 
 {a) Synthetic (old Semitic). 
 \b) Analytic (later Hebrew, etc., to a certain 
 extent). 
 
 3. Aggflutinating — 
 
 {a) Sufflx-Agg-lutinating- {e.g. Ural-Altaic). 
 
 \b) Prefix-Ag-g-lutinating- {e.g. Bantu). 
 
 \c) Suffix - and - Prefix - Agglutinating {e.g. 
 
 Malay). 
 {d) Pronoun-Incorporating {e.g. Basque). 
 
 B. POLYSYNTHETIC OR HOLOPHRASTIC (the American 
 Indian languages).
 
 CHAPTER VI 
 
 THE CLASSIFICATION OF LANGUAGE {continued) 
 B. Genealogical Classification 
 
 When a more or less close examination of a number of 
 languages shows that they all participate largely in a 
 common stock of word-material, that is to say, when they 
 contain a large proportion of words, or at least root-elements 
 of words, in common, especially in the names of familiar 
 and ordinary things, creatures, operations, family relations, 
 etc., and in their numbers, pronouns, and other everyday 
 elements of speech, a notion is naturally formed that the 
 languages in question may be historically related to each 
 other, in the sense that they may be all descended ultimately 
 from a common source, to which they could be traced back 
 if we were in possession of each step of their history. In a 
 comparison of the Romance (or Neo-Latin) languages — 
 French, Spanish, Portuguese, Roumanian, Italian, and their 
 cognate dialects — we know from the facts of history, as well 
 as from the data of language, that they represent so many 
 local developments of the popular Latin, Probably no real 
 harm is done by calling them ' sisters ' to each other and 
 * daughters ' of the Latin of ancient Italy. Philosophically 
 considered, however, they are simply so many varied shapes 
 which the popular Latin in the course of its development 
 has taken upon itself through the working of different 
 conditions in different localities ; but, with this fact clearly 
 grasped, there can be no objection to the current figure of 
 
 95
 
 96 Natural History of Language 
 
 speech, according to which they form a family with a 
 common parent. Similarly, the relationship between English 
 and certain continental Teutonic tongues is rendered the 
 more conclusive by the historical fact that the Anglo-Saxons 
 were originally tribes of Angles, Saxons, etc., inhabiting 
 continental Germany before they transported their language, 
 or rather their several dialects, into Great Britain. 
 
 In these two instances the data afforded by the languages 
 themselves are corroborated, and their indications rendered 
 certain, by the further data of history. Yet, even supposing 
 that history had been silent in each case, and that the 
 philologist simply met with the Romance languages, or the 
 Teutonic languages, as they actually exist, without any of 
 the knowledge which we now possess of the former extent 
 of the Roman Empire or of the migration of the Angles, 
 Saxons, and Jutes, he would nevertheless be at once and 
 necessarily impressed with the mere amount of common 
 material contained in French, Spanish, and Italian, and he 
 could not resist the notion that these languages were 
 probably related to each other by actual historical pedigree, 
 or common descent. Nor could he resist a similar impression 
 in regard to English, Dutch, and German. Nevertheless, 
 whereas the empirical observer would be quick to assume 
 that the correctness of such a notion was self-evident, in 
 view of the large proportion of similar words employed in 
 similar meanings by the languages under survey, the 
 scientific student will refrain from asserting their genealogical 
 connection to be by any means certain until he has looked 
 under and beyond the most striking correspondence of 
 vocabulary. To him that connection is as yet entertained 
 merely as a hypothesis with much prima facie likelihood. 
 The philologist is aware that one language may simply 
 borrow a very large proportion of its vocables from another 
 language, with which it may originally have had no material 
 in common. These vocables, whether phonetically adapted 
 or not, may have become so thoroughly assimilated into the 
 current speech of the borrowing language that all sense of
 
 The Classification of Language 97 
 
 their foreign origin is lost, and they may gradually oust 
 from use the native words which once covered much the 
 same ground. Hence a language which has borrowed 
 liberally may contain x-{-y words, of which x are borrowed 
 from foreign sources and y represent the remains of the 
 genuine native diction. It may conceivably happen (though 
 the occurrence is not frequent) that under x there come to be 
 comprised a large proportion even of the words in frequent 
 use. A hasty or superficial examination of the language, 
 particularly in its literature, might therefore easily result in 
 a fancy that it was directly, and from the first, related to the 
 speech from which it had borrowed the x words, whereas in 
 reality its genealogical connection is with some other 
 language, of which the vocables are more in correspondence 
 with those which we have called y. 
 
 Modern Persian, which is an Indo-European tongue in 
 point of origin, has been so freely grafted with Arabic word- 
 material that it might on the surface present the appearance 
 of being an offshoot from the Semitic family. Japanese, 
 which is not cognate to Chinese, long encouraged, though 
 not to so great an extent, the borrowing of Chinese terms ; 
 and of all languages which have been brought for any length 
 of time into active contact with others, the generalisation 
 can be made, that their vocabularies have lent and borrowed 
 in the process. English, indeed, if we go back far enough, 
 is in truth cognate to Latin and therefore to French, as well 
 as being cognate to Teutonic in general and therefore to 
 German ; but the degree of its historical relationship to 
 German is very much more immediate than to Latin and 
 French. Nevertheless English has indulged at different 
 dates in such comprehensive borrowings from French and 
 Latin that a superficial comparison of mere dictionaries 
 might possibly group it in the first instance with these 
 tongues rather than with German. Thus instead of its 
 proper scheme of pedigree 
 
 (b 884) H
 
 98 Natural History of Language 
 
 Indo-European 
 
 Latin Old Teutonic 
 
 French, etc. English German 
 
 we might, from merely counting the words in the dictionary 
 and in the absence of historical information, be led to assume 
 as probable an incorrect affiliation 
 
 Indo-European 
 
 Latin Teutonic 
 
 French English German 
 
 In this instance we should, it is true, chance to be relating 
 languages which are actually relatives, though we should be 
 relating them in the wrong degree. But in other cases, such 
 as that of Persian with Arabic, we should be claiming 
 original community of linguistic descent where there was 
 none, but only contamination of a language of one stock by 
 material derived from a language of another and quite alien 
 stock. 
 
 It might perhaps be argued that word-material is, after 
 all, the most prominent and essential constituent of language, 
 and that, therefore, any speech which went on to borrow the 
 larger portion of its vocables from a foreign tongue would 
 ipso facto throw over its original relationships and become 
 adopted into that family, or branch of a family, to which the 
 foreign tongue belonged. It might, for example, be asked 
 how far English would have to increase and freely use its 
 borrowings from Latin and French before it ceased to be 
 fairly classifiable as Teutonic. K people may continue to 
 be racially much the same as ever, while its language may 
 
 I
 
 The Classification of Language 99 
 
 gradually be replaced by one of alien derivation. Thus it 
 does not follow (it might be urged) that, because the Anglo- 
 Saxon dialects preceded modern English and were spoken 
 by tribes of Germanic extraction, therefore modern English, 
 even with all its additions and alterations, is still at the 
 present day akin to German in a closer degree than it is to 
 French. We must not — it will be said — confound the 
 original and propagated kinship of the English and Germanic 
 peoples with any necessarily permanent kinship of their 
 languages. Old English was distinctly Teutonic, but modern 
 English is by no means characterised with equal distinctness. 
 Language consists of words used, and, if we reject one set 
 of words in favour of another, our language is no longer the 
 same in extraction. If the actual words of the vocabulary 
 are not the criterion of genealogical relationship, where is 
 the criterion to be found ? Given that modern Persian 
 contains a-\-b words, of which a are Semitic and b are 
 Indo-European, how are we to decide whether Persian, as 
 now actually spoken, belongs to Semitic or to Indo-European, 
 except from the numerical proportion of the words contained 
 in a and b respectively ? 
 
 Before answering this question there is one prime 
 qualification to be made in connection with the statistics as 
 to the relative proportions of native and borrowed vocables. 
 A language in the proper sense is " that part of the aggregate 
 of the language of the individuals " (who are considered as 
 speaking the said language) " which is supported by the 
 usage of the majority" (Whitney). The living English or 
 French language, for example, is not the language as 
 represented by the litterateur, the scientist, or the lawyer ; 
 neither is it the language of any special locality, district, or 
 class. It is not a language in which all the words of the 
 dictionary habitually play equally prominent parts, or are 
 equally understood by the majority. In actual fact some 
 members of the vocabulary are used everywhere, every day, 
 and by every one, such as ' I,' ' we,' ' me' ; ' three,' ' four,' 
 ' five ' ; ' go,' ' do,' ' give ' ; ' good,' ' bad,' ' best ' ; ' father,'
 
 lOO Natural History of Language 
 
 ' town,' ' field.' Others are sufficiently frequent over all 
 parts of the linguistic area, and are universally understood, 
 though not equally employed, such as ' courage,' ' sorrow,' 
 ' natural,' ' ornament,' ' forgive,' ' saint,' ' receive,' etc. Others 
 are comparatively rare, used habitually only by the learned, 
 and require explanation for the majority, such as ' com- 
 mutation,' ' philosophical,' ' rhetoric,' ' incipient.' Yet others 
 are of altogether exceptional occurrence, such as the 
 strictly technical words of the sciences and arts. It must 
 be obvious that the character of the vocabulary of the 
 English or French language at any given moment, depending 
 as it does upon the question what words are actually being 
 spoken at that moment by Englishmen and Frenchmen, and 
 not upon what is being written by them, is determined 
 mainly by the words in the first two groups. The ' language ' 
 consists to the philologist of what is habitually spoken and 
 habitually understood by the people at large. Perhaps it is 
 not always possible to discover precisely how much of the 
 vocabulary authorised by the dictionaries does fall under 
 this definition. It is, however, possible in the case of current 
 and well-known tongues to arrive at a reasonable opinion 
 upon the point, and no person competent to deal with 
 French, English, or German philology would feel seriously 
 embarrassed by the question. This being granted, it 
 becomes evident that a mere statistical statement of the 
 total number of words existing in a language, and of the 
 proportion of ' dictionary ' words connected with one family 
 or group to those connected with another family or group, 
 is fallacious as a guide to the character of the vocabulary 
 as actually distinguishing the language. If the English 
 lexicon contains x ' classical ' and y ' Teutonic ' words, 
 the question still remains to be asked how large a proportion 
 of the words daily current in the mouths of average 
 Englishmen is drawn from ,r, and how large a proportion 
 from y. And inasmuch as in the first of the four groups 
 above mentioned, i.e. in those words which are spoken 
 everywhere, every day, and by everybody, the Teutonic
 
 The Classification of Language loi 
 
 element {y) does immensely preponderate, while the adopted 
 element (,r) becomes more and more considerable only just in 
 proportion as we depart from ' English as it is spoken,' it 
 must be manifest that the character even of the vocabulary 
 of English must be put down as unequivocally Teutonic. 
 On the ground, therefore, of word-material alone the 
 affiliation of English would be to a Teutonic and not to a 
 Latin parentage, even if the history of the language were 
 lost. 
 
 But word-material is far from being the only subject of 
 investigation. A language being, as we have observed, 
 made up of sentences, in which the special application of 
 the meaning of a word and its relation to the other words 
 are expressed by those inflexional or other devices known 
 as ' grammatical,' there may naturally be presumed to exist 
 in languages genealogically akin to each other a greater or 
 less measure of similarity in such devices. In proportion as 
 the connection of two languages is more immediate and 
 recent, the devices may be expected a priori to show more 
 obviously a common starting-point. We should, therefore, 
 look not only for a common basis to the vocabulary, but also 
 for a common basis to the grammar. The close correspond- 
 ence which we find in the Romance languages in such parts 
 of the verb-system as the French j'aiirai, tu auras, il aura, 
 and the Italian avrb, avrai, avra, or in the French J 'at vn, 
 Italian ho veduto, etc., is explainable, not from any special 
 similarity of psychological operations in the French and 
 Italian minds (though that may possibly also exist to some 
 extent), but from an equal original participation in the vulgar 
 Latin future system habere-habeo, and the perfect system habeo 
 vidutum. Similarly, the correspondence of flexion between 
 the German possessive case des Mannes and the English 
 ' the man's ' is due to an equal original participation in a 
 Teutonic scheme of declension in which the -cs played a 
 prominent part. There is no trace whatever in current 
 English or German of a compound future analogous to the 

 
 I02 Natural History of Language 
 
 French or Italian ; nor is there any trace in current French 
 or ItaHan of the possessive in s which marks the Teutonic. 
 A detailed examination of the remains of flexion in English 
 in the way of declension and conjugation would, even apart 
 from our historical knowledge of its older forms, leave no 
 doubt that it was grammatically characterised as belonging 
 to the Teutonic rather than to the Romance branch or group 
 of tongues. 
 
 Now, though the subject of mixture in language requires 
 separate treatment, it is almost unanimously admitted by 
 philologists that, however large a borrowing of foreign 
 vocables may take place, there is no satisfactory evidence 
 that foreign grammar is ever borrowed to any appreciable 
 extent by a truly live and spoken language. The utmost 
 that can be effected grammatically by the influence of one 
 language upon another is to assist in breaking down the 
 unessential elements of an old system. The language so 
 influencing does not go on to impose its own grammar, or, if 
 it does so, it is only within a narrow social or literary sphere 
 of conscious imitation and artificiality, which leaves little or 
 no trace upon that which we have defined as the real 
 language of a people. This was the case with the 
 ' Graecisms ' of the Augustan writers of Latin. Norman 
 French no doubt helped to break down the less useful or 
 systematic Teutonic inflexions in English, which were already 
 on their way to decay, but of its own grammar it has left us 
 practically nothing. The prevalence of the English plural 
 in s is, indeed, frequently attributed to French influence, but 
 is so attributed only in ignorance of the facts. In Persian 
 the only distinct case of grammatical borrowing from Arabic 
 is that of certain plurals in -at, which, however, are found 
 almost solely with words of Arabic origin, and are little more 
 characteristic of Persian than plurals like ' phenomena,' 
 * beaux ' are of English. 
 
 On the positive side, similarity of grammar, therefore, 
 affords an even more trustworthy evidence of original relation- 
 ship than similarity of vocables, which may be freely
 
 The Classification of Language 103 
 
 borrowed. On the negative side the matter is not so simple. 
 It is usual to assert that a demonstrable resemblance of 
 grammar must be a sine qua non of genealogical connection. 
 If this means that, where we do not possess the past history 
 of a language, it is unsafe to take even a considerable 
 similarity of vocabulary as a sufficient and sure criterion, 
 when there is no further support of grammatical similarity, 
 the statement may be taken as a salutary caution. But if it 
 means that languages cannot be genealogically related unless 
 they still possess demonstrable grammatical similarity, it is 
 incorrect, as a little consideration will show. English has 
 gone very far in the direction of analysis, and it would take 
 but a few steps to make it so destitute of flexion as to leave 
 no trace of its former connection in that respect with 
 Sanskrit or Greek. It thus appears that though, without the 
 argument of distinct grammatical correspondence, it may be 
 difficult to prove the positive, on the other hand the absence 
 of such correspondence cannot demonstrate the negative. 
 In the case of languages whose separation has been com- 
 paratively recent the evidence of grammar is generally 
 procurable at once along with that supplied by the vocabu- 
 lary. In a comparison of Italian, Spanish, and French, of 
 English, Dutch, and German, of Hebrew and Arabic, it springs 
 readily to the eye, and, when combined with the data of the 
 word-material, renders each linguistic connection absolutely 
 assured. On the other side, inasmuch as no language 
 remains stable in its grammar any more than in its words 
 and their pronunciation, but each passes through a more or 
 less rapid evolution of its own, it is a fact easily compre- 
 hended that languages which are only akin in a less 
 immediate degree (such as that of distant cousins or of a 
 comparatively remote ancestor to its descendants) will 
 perhaps present but few salient points of grammatical 
 resemblance. They may require to be placed in different 
 morphological classes or sub-classes as above described. A 
 wholly synthetical tongue may become partly synthetical, 
 partly analytical, may then pass into one mainly analytical,
 
 1 04 Natural History of Language 
 
 and thence reach an absolutely analytical or positional 
 stage. It requires the acumen of an expert philologist to 
 detect the few traces of Indo-European synthetic grammar 
 which still hold their ground in English. So far as these 
 exist, they would, if required, be of value in establishing the 
 affiliation of English ; but it is quite conceivable that in the 
 further evolution of the language they may cease to exist 
 altogether. 
 
 It is manifest, therefore, that a language which sprang 
 from source A, and carried thence both its original grammar 
 and its original vocabulary, may subsequently borrow a large 
 portion of its word-material from source B, and may also by 
 its own natural development modify its grammar past recog- 
 nition. In such an instance the philologist who met with 
 the language without its history would possess no criterion 
 of its proper affiliation, except that which was derived from 
 discovering which parts of the vocabulary were its bed-rock. 
 Those words would naturally be mainly derived from stock 
 A. That criterion would, no doubt, be scientifically some- 
 thing less than conclusive ; nevertheless, it would happen 
 that the theory based upon it would be correct. The 
 language would be actually descended from A, and yet its 
 supposed sine qua non of grammatical similarity would be 
 conspicuously absent. 
 
 To sum up the considerations urged in the above para- 
 graphs. We may lay it down that, apart from historical 
 data, an entirely conclusive genealogical classification of 
 languages depends upon the mutual support of two criteria, 
 namely, similarity of the elements of what may be called the 
 working part of the vocabulary, and similarity of grammatical 
 forms and devices. Without the latter the connection is not 
 disproved ; but, on the other hand, it cannot be indubitably 
 demonstrated by the former alone, inasmuch as the borrowing 
 of foreign vocables is easy in theory and frequent in practice, 
 and might conceivably affect (though it rarely does) even the 
 commonest words. In a case where the elements of the
 
 The Classification of Language 105 
 
 working part of the vocabulary are fairly divided between 
 resemblance and non-resemblance to some other given speech, 
 the decision will lie with the grammatical criterion, if it 
 exists. 
 
 Here a further remark of much importance must be 
 made concerning the cognate elements in the vocabularies 
 of languages which are genealogically related— that is to say, 
 in regard to those ' roots,' words, and parts of words which 
 are common to such tongues. It is that such common 
 elements will not necessarily, nor even frequently, but, on 
 the contrary, with great rarity, appear under identical forms 
 and with identical sounds in any two of the languages con- 
 cerned. If A, B, C are kindred speeches, each lineally 
 inheriting much of the vocabulary of an original common 
 ancestor X, we must not look to find a given root or word 
 appearing in A exactly as it appears in B or in C. Rather, 
 if it did so appear, we should mostly have good grounds for 
 suspecting that the word, in the state in which we meet it, 
 was actually borrowed by A from B or C, or vice versa, and 
 not inherited directly from the original source X. Every 
 language exhibits ' phonetic laws ' of its own, which operate 
 with great, if not invariable, regularity upon all the material 
 included in the language, so far at least as it is included by 
 direct inheritance and not by borrowing. The phonetic laws 
 of language A will determine an original root or word to one 
 shape, while those of B are determining it to another. For 
 this reason the relationship to each other of words in 
 languages A, B, C, will often be by no means obvious upon 
 the surface. The Greek word /3o{}9 iboiis) is etymologically 
 identical with Sanskrit gdus, despite the existing differences 
 in the aspect and pronunciation of the words ; and the 
 English ' cow,' German Kuh, are no less to be referred to the 
 same original, which it is customary to postulate as *gdns. 
 Each word, bous,gdus, or cow, is the proper phonetic representa- 
 tive in its own language of the original word *gdus. It is the 
 shape which that word would naturally take upon itself
 
 io6 Natural History of Language 
 
 during its transmission from generation to generation in the 
 mouths of a particular line of speakers. If in English or 
 German we had met with \bow or \gow (instead of cow or Kuh) 
 as the name of the animal, we should, with our knowledge 
 of the phonetic laws of those two languages, have been 
 practically certain that such a form was borrowed, from the 
 Greek perhaps (if \bow\ or the Sanskrit perhaps (if \gow). 
 Similarly, the word ' ten ' is in Greek Se/ca {deka\ in Latin 
 decern, in Sanskrit dasa, and in Gothic taihun. None of these 
 representations is identical with another, but their evidence 
 of direct inheritance from a common source is made all the 
 surer thereby, since these are precisely the forms into which 
 each language would respectively transmute the original 
 word *dekm. Had the words been exactly alike, or almost 
 so, there would have been every reason for suspecting that 
 there had been borrowing one way or the other in compara- 
 tively recent times, and it would therefore be unscientific to 
 use the common possession of such an identical word as a 
 proof of the common origin of the languages. The French 
 ph'e and mere correspond to padre and viadre in Italian, 
 These are the natural phonetic representatives in the two 
 tongues respectively of the Latin patreni and mdtrem. Had 
 French shown the form padre or madre, v/e should have had 
 no hesitation in declaring such a word to have been taken 
 directly from Italian, and not to owe its existence in French 
 to direct inheritance. 
 
 What we have to expect, therefore, in comparing the 
 vocabularies of several languages with each other for the 
 purpose of establishing genealogical classification, is not 
 exact identity of shape, but that correspondence of shape 
 which might be in a large measure, if not in every detail, 
 predicted in the light of the phonetic laws peculiar to each 
 language. The form naturally corresponding to 7eVo? 
 {genos) of Greek and genus of Latin would be in Sanskrit 
 janas and in English kin. These forms, indeed, might be 
 predicted, and the philologist is entirely satisfied of the 
 original identity of these words (from Indo-European * genos).
 
 The Classification of Language 107 
 
 whereas he would have been altogether incredulous in the 
 case of an English form \gin, except on the supposition that 
 English had borrowed that form from Greek or Latin and 
 corrupted its vowel. For this reason he rejects from his 
 list of corresponding words such falsely obvious instances as 
 call and the Greek /ca\-w {kal-o), despite the identity of 
 meaning. He refuses to admit whole as the native English 
 word akin to Greek 0X09 {holos). If English has not simply 
 borrowed these two words (and it has not), they must be 
 quite distinct in origin from the Greek. On the other hand 
 the philologist will readily accept sweat as the natural 
 representative in English of the original root whose natural 
 representative in Greek is seen in t'S-pco? ijiid-rds). 
 
 It is indeed true that, in the first instance, while feeling 
 his way, a student, meeting with a group of languages which 
 appear to contain a largely similar vocabulary, will place 
 side by side from the various tongues those words of 
 approximately the same sense which possess at the same 
 time appreciably similar shapes. He will put the Latin 
 meditis beside the Greek /ieVo? {viesos), Sanskrit inadhyas, 
 Gothic viidfis, and he will also — wrongly as it happens — 
 put the Greek 6eo<i {theos) beside the Latin dens, Sanskrit 
 deva. He will set the Greek ^e'po) {phero), Latin fero, 
 English bear, Sanskrit bhardmi side by side, and no less — 
 though wrongly again — the English call side by side with 
 the Greek KoXdv {kalein). When, by the aid of a sufficiently 
 convincing list of such correspondences, he has arrived at 
 the conclusion that a genealogical connection of the tongues 
 concerned is highly probable, he will proceed to apply the 
 other criterion of grammatical correspondence. If this test 
 further establishes the theory and renders the relationship 
 practically certain, it becomes time to work over the list of 
 word and root correspondences again, with a view to dis- 
 covering the laws which in each language determine the 
 particular shape which the root or word will adopt. So 
 soon as these laws are discovered and the regularity of their 
 operation made apparent (as it will soon be), a large number
 
 io8 Natural History of Language 
 
 of other correspondences of a less obvious but no less sure a 
 character will spring to the light and corroborate the previous 
 conclusions. But, on the other hand, it will become equally- 
 apparent that a nunnber of the most obvious-looking cor- 
 respondences are not correspondences at all, and the notion 
 that ^609 {theos) is the same word as dens and call as icaXelv 
 {kalein) becomes quite untenable. The philologist thereupon 
 erases all such pseudo-equivalents from his lists, while he 
 inserts as sound and demonstrable others as unlikely-looking 
 as Sanskrit panca, English five, Latin quingue, Greek irevTe 
 ijente) ; English come, Latin venio, Greek ^aivta {baino) ; 
 English goose, Latin anser, Greek -^-qv {chen). 
 
 It is upon lists thus scientifically assured that is based 
 the ' similarity of vocabulary ' which is one of the two criteria 
 for the classification of languages according to nearer or 
 more remote degrees of historical relationship. 
 
 In the classification of languages according to their 
 genealogy there is still much to be done. It would be 
 eminently desirable, if it were possible, to establish definitive 
 families of languages, and to work out in each family the 
 precise pedigree of the speeches belonging to it, showing the 
 closer and more remote connections of the individual dialects 
 and languages in the several branches of the family-tree. 
 It is an instinctive aim of the human intellect to arrive as 
 far as possible at unity, whether of origin or principle, in the 
 phenomena which it encounters. But true science does not 
 permit itself to be carried away by this desire, and the 
 science of language is at present so far from finding adequate 
 support for the opinion that all languages are ultimately 
 referable to a common parent, that it cannot even begin by 
 referring them to any small number of families. The con- 
 clusions of science must not proceed beyond the evidence 
 of its data, and, in comparative philology, the method of 
 rigorously comparing the vocabularies and grammatical 
 structures of languages as we find them must not be cor- 
 rupted by any preconceptions as to the unity or diversity 
 
 I
 
 The Classification of Language 109 
 
 of origin of language in general. In actual fact the applica- 
 tion of this method to the various speeches of the world 
 does not bring us very far on the road towards unification. 
 Many larger or smaller groups and a few considerable 
 families of languages are discoverable, but even the more 
 sanguine students are unable to reduce existing speeches to 
 fewer than eighty or ninety divisions, between which no 
 genealogical connection is yet warranted by the only criteria 
 which science possesses. It must be confessed that in the 
 case of many of the barbarian and savage dialects, as, for 
 example, among the African negroes and American Indians, 
 the information is very scanty and ill-digested, and a 
 complete examination of these tongues might result in either 
 an increase or a diminution of the reasoned divisions into 
 linguistic groups. On the whole, the estimate of Friedrich 
 Miiller, which places the number of apparently unconnected 
 groups at about one hundred, may be taken as approximately 
 correct in the present state of our knowledge. 
 
 Chief among the families as yet more or less clearly 
 ascertained are the Indo-European, the Semitic, the Ural- 
 Altaic, the Malayo-Polynesian, the Bantu of Southern Africa, 
 and the Dravidian of Southern India. Fuller details of the 
 contents of these families, as well as of less definite or less 
 important groups, will be given in the following chapters. 
 Meanwhile it is expedient to illustrate by a few examples 
 the correspondences of vocabulary and grammar upon which 
 genealogical connection is established in the typical instance 
 of Indo-European. 
 
 In this family are included Sanskrit (the ancient 
 language of Northern India) and its descendants ; the 
 ancient Eranian dialects, old Persian and Zend, with the 
 modern speeches related to them ; Armenian ; Greek ; 
 Albanian ; the old Italian dialects, Oscan, Umbrian, and 
 Latin, with the Neo-Latin or Romance tongues ; Celtic, 
 with its several branches in Ireland, Scotland, Wales, and 
 Brittany ; Teutonic, with all the High and Low German 
 tongues ; Lithuanian ; Lettic ; and all the branches of the
 
 I lO Nat2tral History of Language 
 
 Slavonic languages. It may be difficult to find words and 
 ' roots ' which have remained common to each and all of the 
 speeches included under these designations, but it is an easy- 
 matter to produce long lists of words and ' roots ' which are 
 found distributed over a large number of them, under the 
 several phonetic shapes which are proper to the several 
 languages. One word, or its chief element or elements, will 
 be found represented in languages A, B, C, D, G, but will 
 be absent in E, F, H, etc. ; another in A, C, D, E, H, but 
 absent in B, F, G ; a third in A, C, E, F, G, but absent 
 in the rest, and so on ; until it will be discovered that each 
 possesses much in common with any one of the rest, although 
 for the most part, according to the greater or less immediate- 
 ness of its historical connection, it will be marked by closer 
 correspondence with some than with others in this respect. 
 
 Among words descriptive of familiar creatures, objects, 
 qualities, operations, and relationships, and among pronouns 
 and numerals, may be adduced such examples of widespread 
 agreement as the following : — 
 
 ' Horse ' : Skt. asvas, Zend aspa, Gk. t'77— os {hippos), dialectal 
 
 tKKos {ikkos), Lat. equus, Goth, aihva, Irish ec/t, 
 
 Lith. asziva. 
 ' Pig ' : Skt. su{-kara), Zend ///?, Gk. rs {hiis) and (rrs {sfis), 
 
 Latin sfis, O. H. Germ, sii, (Eng. sow). 
 ' Sheep ' : Skt. avis, Gk. oFts {owis), Lat. ovis, Eng. ewe, Irish 
 
 oi, Lith. awis. 
 'Cattle': Skt. pasu, La.t. pecu, Goth. /aihu. (Germ. Vieh). 
 
 ' Night ' : Skt. naktis, Gk. stem vvkt- {/tuki-), Lat. noct-, Goth. 
 
 nahts, Lith. naktis, O. Slav, iiosti. 
 ' JNIead ' (wine) : Skt. madhu, Zend fuadhu, Gk. fxkOv {inethti), Eng. 
 
 ?nead, Lith. niidus, O. Slav, incdu. 
 ' Yoke ' : Skt. yugam, Gk. (\jyov izttgon), Lat. iiigu7n, Goth. 
 
 ink, Lith. jungas. 
 ' Red ' : Skt. rudhiras, Gk. k-pvQpo% [e-ruthros), Lat. ruber, 
 
 Goth, rduds, Ir. ruad, O. Slav, rudru. 
 ' Father ' : Skt. pitar, Gk. 7rar7;p {pater), Lat. pater, Goth, fadar, 
 
 Ir. athir. 
 'Mother': Skt. mdtd, Armen. mair, Gk. (Doric) fidrrip (mater), 
 
 Lat. mater, O.H.G. inuotar, Ir. mathir, O. Slav. 
 
 mati. 
 ' Brother ' : Skt. bhrdtar, Zend brdtar, Gk. (ftpdrrip {phrdter, ' clan-
 
 The Classification of Language 
 
 III 
 
 brother'), Lat. frdtcr, Goth. bfd\)ar^ Ir. brathir, 
 
 Lith. broils. 
 ' Mother-in-law ' : Skt. svasru, Armen. skesur, Gk. tKvpa {liekura\ Lat. 
 
 socriis, Goth, svaikro, O. Slav, svekry. 
 ' Plough ' : Gk. dpouv {aroun), Lat. arar^, Goth. ar/Vz/z, Eng. 
 
 '^ar,' Ir. airtin, Lith. ar-//. 
 ' Grind ' : Gk. {j-vXt] {?niile, ' mill '), Lat. molerc, Goth, malan, 
 
 Ir. ;;z^/z>;z, O. Slav, melja. 
 ' Bear ' (verb) : Skt. bhardjnt, Armen. berein, Gk. <^€pw {phero)., Lat. 
 
 yt'r^, Goth, baira. 
 ' Know ' : Skt. yw^z-/"/, Gk. yt-yvw-o-Kw {gignosko), Lat. {g)7idscd, 
 
 Eng. know., O. Slav, zna-ti. 
 'See' (know): Skt. w7/a, Gk. {F)L8eiv, {F)oiSa, {(7o)idein, {u')oida), 
 
 Lat. z//^/^^, •c^f^f, Eng. wzV, Germ, ivisseii. 
 'Thou': Skt. /■z/aw, Armen. ^z^, Gk. (Doric) tv {tu\ Lat. //?, 
 
 Goth. ]7zi, Welsh /";/, Lith. hi, O. Slav. /y. 
 'Six': Skt. sas, Gk. e^ (Jiex), Lat. j-^.r, Goth, j-az'/^j, Welsh 
 
 chwech, O. Slav. j-^j--/"^. 
 ' Ten ' : Skt. dasa, Gk. ScKa (rt?'^/^rt:), Lat. decein, Goth, taihuii, 
 
 Welsh ^/^'_^, O. Slav. desei7-te. 
 
 For very close correspondence of grammatical structure 
 we may compare portions of the conjugation of the verb 
 and of the declension of the noun in the synthetic languages 
 of the family. It is not, indeed, to be understood that each 
 form as it actually appears in a given language is in all 
 respects the phonetic equivalent of the form which appears 
 in another language for the same person or case. Certain 
 analogical influences have been at work upon each language 
 during the course of its development, and have prevented 
 such absolute regularity. The correspondence, however, is 
 fairly, or even strictly, preserved in a sufficient number of 
 instances to render indubitable the original identity of the 
 inflexions employed. Thus the present indicative of the 
 verb ' to bear ' appears in the closely parallel shapes — 
 
 Sanskrit 
 
 
 Greek 
 
 Latin 
 
 Gothic 
 
 bhar%{ini) 
 
 
 <^e/5w i^phcro) 
 
 fero 
 
 bairz, 
 
 bJiM-zsS. 
 
 
 <^e/3et5 (///rreis) 
 
 fers 
 
 bairis 
 
 b/uv'dXi 
 
 
 <^e/3ei (^p/iC}-Q\) 
 
 fert 
 
 ba/ri]? 
 
 b/iarsimSbS 
 
 (Doric 
 
 ) c^epo/zfs {p/ierorsXQS) 
 
 /vimus 
 
 ba/?-2im 
 
 b/iamtha, 
 
 
 (fiepere (p/icrete) 
 
 /tvtis 
 
 baii\\-> 
 
 Mar&nti 
 
 (Doric 
 
 ) <f)epovTL (p/wronti) 
 
 /t-mnt 
 
 baird^ndi
 
 112 
 
 Natural History of Language 
 
 The singular of the present indicative of the verb 
 to be ' is — 
 
 Greek Gothic Lith. O. Slav. 
 
 zm ^^mi jesvcxi. 
 
 is es\ jesi 
 
 Sanskrit 
 
 asmi 
 
 asi 
 asti 
 
 €t/x6 (^/mi) 
 
 et {icrcTL) {ez,essi) 
 
 ecTTt (esti) 
 
 z'st 
 
 esti 
 
 jesti 
 
 The singular of the present optative of the same verb- 
 
 Sanskrit Greek 
 
 jyam €t'>;v ( = *k(T-irj-v) (*^5yen) 
 
 jyas e"/s ( = *e<T-ti]-s) (*esyes) 
 
 j-yat el,] ( = Jo--i7;-t) (*^j-yet) 
 
 Old Latin Gothic 
 
 ^iem jjjan 
 
 sies njais 
 
 siet i^ijai 
 
 The noun ' horse ' is declined in the singular — 
 
 Nom. 
 
 Voc. 
 
 Ace. 
 
 Gen. 
 
 Abl. 
 
 Dat. 
 
 Sanskrit 
 as'o'&S 
 
 asvaiva. 
 
 asva.SJBj 
 
 asvSLd 
 
 ai't/aya 
 
 Greek 
 
 tTTTTOS {/l2pJ>0S) 
 
 tWe {JiippQ) 
 
 ITTTTOV {/lippOXl) 
 ITTTTOIO {hippQXO) 
 
 Latin 
 eguus (older equos) 
 equQ 
 
 equMVQ. (older equom) 
 [equi] 
 
 equb (older equoA) 
 equb 
 
 The nominative 
 respectively — 
 
 Sanskrit Greek 
 
 < I ' : aha7n eyw(v) {cgd{)i)) 
 
 ' me ' : ma fJie {me) 
 
 I ' and the accusative ' me ' are 
 
 Latin 
 
 Gothic 
 ik 
 mi-k 
 
 There can be no other reasonable explanation of this close 
 and consistent correspondence of vocabulary and structure 
 in a series of languages than the obvious one, that all these 
 speeches alike are derived from one and the same source — 
 a primitive language of which they are the descendants in 
 various degrees and with various interrelations. There is 
 no necessity, nor would it be true (as will be pointed out 
 later), to assume at the same time that those who speak the 
 related languages are also themselves all alike descended 
 from a common stock. The question is here one of the 
 speeches, not of the speakers. The ancestors of those who
 
 The Classification of Language 113 
 
 now speak Indo-European tongues belonged to various 
 races, which at one time possessed languages equally alien. 
 But this is no more than to say that such races have at 
 some time, or rather during some period, wholly unlearned 
 their primitive speech and wholly adopted the Indo-European, 
 just as a pure Magyar by race may be a pure German in 
 speech, or a South American Indian may speak Spanish as 
 his only tongue. In whatever way it has come about that 
 so many peoples ethnologically alien now speak these 
 languages genealogically akin, the fact remains that the 
 languages themselves did unmistakably flow from a common 
 source. 
 
 The same principles of investigation, applied to Hebrew, 
 Arabic, Aramaic, and other tongues, have resulted in the 
 establishment of a ' Semitic ' family of languages, within 
 which the correspondence is strikingly close ; so much so, 
 indeed, that they are one and all related in quite as intimate 
 a degree as are the various speeches sometimes contained in 
 a single branch of the Indo-European family. In other 
 words, the extreme divergences of the Semitic languages 
 represent no wider a separation than can be found inside 
 the Teutonic or the Slavonic branch alone of Indo-European. 
 It is, for instance, part of the structural method of Semitic 
 to agglutinate the possessive pronoun to the noun in such a 
 way that a paradigm, or kind of declension, is formed of 
 such a system as ' my house,' ' thy (masc.) house,' ' thy (fem.) 
 house,' ' his house,' ' her house,' ' your (masc.) house,' ' your 
 (fem.) house,' etc. Not only is this structural habit common 
 to such languages as Arabic, Aramaic, Assyrian, Hebrew, 
 ' Ethiopic,' but the forms of the agglutinated elements are 
 exceedingly close to each other, and in a large measure 
 identical. Thus the singular endings are respectively — 
 
 Arabic ' Ethiopic ' Hebrew Assyrian Old Aramaic 
 
 ' my ' : 
 
 -i (-ya) 
 
 -ya 
 
 -i 
 
 -i (-ya) 
 
 i 
 
 'thy' (masc.) : 
 
 -ka 
 
 -ka 
 
 -kha 
 
 -ka (-k) 
 
 -k 
 
 'thy' (fem.): 
 
 -ki 
 
 -ki 
 
 -kh 
 
 -ki 
 
 -ki 
 
 ' his ' : 
 
 -hu 
 
 -M 
 
 -hu 
 
 su 
 
 -h 
 
 ' her ' : 
 
 -ha 
 
 -ha 
 
 -ha 
 
 sa 
 
 -h 
 
 (b 884) 
 
 
 
 
 
 I
 
 114 Natural History of Language 
 
 The correspondence of the numerals (which take both a 
 mascuHne and a feminine form) may be judged from such 
 instances as — 
 
 ' Three ' (fern.) : Arab, thalathu, Eth. sa/ds, Heb. sd/os, Assyr. salsu, 
 
 Aram. ^/di/i. 
 ' Five' (fern.) ; Arab. {c)hamstc^ Eth. {c)hames^ Heb. hames, Assyr. 
 
 katnsUj Aram, hanies. 
 
 Similarly, the relationship of the Bantu family of South 
 African languages, which extends from the eastern side of 
 Cape Colony to beyond the Equator and across to the 
 Atlantic, is clearly marked both in the use of forms and in 
 the working elements of the vocabulary. Thus — ■ 
 
 ' Man ' : Zulu umu.'z/w, Kafir uin«/z^, Sesuto mo/^/, Congo 
 
 Gvnykntii, Herero Qimyindu. 
 ' Men ' : (in the same languages respectively) sib&n/u, aib&niu, 
 
 h&iu, oa;//«, oYa^ndii. 
 'With the man' : Kafir ngo?nuntu, Herero udbtnundu, etc. 
 
 To add further instances in illustration of the kinds of 
 similarity demanded for the scientific ascertainment of a 
 linguistic family is needless. We are at present concerned 
 only with the method of investigation. The collection and 
 examination of the exact details of corresponding words and 
 forms belong to the special philology of each family. The 
 general results hitherto arrived at by the united efforts of 
 experts will be summarised in the following chapters.
 
 CHAPTER VII 
 
 GENERAL SURVEY OF LANGUAGES 
 
 The following survey of languages which either exist in 
 actual speech or, being extinct, are known in a greater or 
 smaller degree from their literatures and other records, is 
 not intended to be an exhaustive enumeration. It is not 
 yet the case that the entire field of language has been so 
 thoroughly explored that we can in all instances distinguish 
 a mere dialect from an independent speech, relate all kindred 
 dialects and speeches to each other definitively, or attach to 
 each either its truest morphological or its truest genealogical 
 designation. Particularly upon many of the aboriginal 
 languages of North and South America, of parts of Africa, 
 and of Oceania, as well as upon the speeches of nomad 
 populations of North and Central Asia and upon the isolat- 
 ing idioms of the South-East, there is much to be done 
 before the information necessary to the philologist can be 
 said to be obtained. 
 
 It appears more profitable, therefore, to avoid doubtful 
 detail and to dispense with the tabulation of unimportant 
 dialects in lists which would be likely to represent so many 
 mere names without adequate notions attached to them. 
 By the word ' unimportant ' it is not implied that any 
 dialectal variation of any language is without a certain value, 
 provided its nature and extent be fully apprehended and 
 therefore made capable of contributing something to the 
 data of linguistic philosophy. Philologically, a language or 
 dialect becomes important just in proportion as its contribu- 
 
 "5
 
 1 1 6 Natural History of Langtcage 
 
 tion to this end is a material one, whether the language or 
 dialect be that of a numerous or civilised people or not. 
 Nor is it implied that, from an ideal point of view, there can 
 be such a thing as too exhaustive a subdivision of speeches 
 from families into groups, from groups into branches, from 
 branches into languages, from languages into dialects, and 
 from dialects into sub-dialects. On the contrary, such con- 
 tinuous distinction is an essential part of the scientific 
 method, so long as the differentiae which supply the motive 
 for subdivision are clearly realised and clearly stated. In 
 that case they all teach something. They may either 
 present new phonetic, structural, or semasiological phenomena, 
 or else they may corroborate generalisations arrived at from 
 previous material. When, however, the points of divergence 
 are not so apprehended and set forth, the mere fact that 
 certain dialects have a local existence and a name may be 
 described as philologically ' unimportant.' We may accord- 
 ingly content ourselves with a brief account of the principal 
 families and groups of languages as yet fairly ascertained, 
 and a briefer indication of such speeches as are but dubiously 
 classified or not classified at all. In so doing it will only 
 be in the case of languages specially instructive by reason of 
 a pronounced typical character that we need enter into 
 descriptive details. 
 
 I. The Positional or Isolating Languages 
 OF Further Asia 
 
 I. Chinese. — The literature of China claims to date 
 from as far back as the year 2500 B.C., through its sacred 
 books and the Shu-kmg, the official and classical history. 
 These works, however, were edited, and their language 
 probably revised, by Confucius in the latter part of the sixth 
 century B.C. The seat of literary production was about the 
 lower part of the course of the Hoang-Ho (Yellow River), in 
 the centre of the region in which the modern imperial 
 dialect, the Mandarin, is actually spoken, though, it may be
 
 General Siirvey of Languages 1 1 7 
 
 remarked, it is precisely that dialect which has since 
 departed farthest from the ancient standard. 
 
 Such information as can be gathered concerning an 
 earlier condition of Chinese is deduced from the literature 
 above mentioned, the ancient phonetics being in some 
 degree recoverable from the old poetry (which was rhymed) ; 
 from some indications of pronunciation as presupposed by 
 the written signs ; and from transcriptions of Sanskrit words 
 by early Buddhists of China. From a comparison of the 
 modern Chinese dialects with one another it is also possible 
 to arrive by induction at earlier forms and pronunciations, 
 as they presumably existed at a stage of the common 
 language before the dialects had developed. 
 
 So far as these data can carry us, we find the language 
 to have been structurally what it is at present, that is to 
 say, positional and monosyllabic, with the difference that the 
 use of ' empty ' or symbolic words has'considerably increased 
 in the interval. This fact is of importance as indicating the 
 line of linguistic ease naturally taken by a language which 
 finds the purely positional structure inadequate to it needs. 
 In point of pronunciation the changes have been great, 
 and in the invariable direction of curtailment or economy of 
 effort. Thus, according to Dr. Edkins, the older forms 
 to which the modern Mandarin yi, ta, ye correspond 
 respectively were tit, dap, tik. Similarly, Mandarin drops 
 from its pronunciation the initial ng of such words as 7iga, 
 ngo, leaving only a, as their representatives. A natural 
 result of such phonetic decay is that words which were once 
 sufficiently differentiated in their sound-elements to serve 
 without further help for the expression of different ideas, 
 have become indistinguishable in outward shape and very 
 often absolutely so in pronunciation. In other words, 
 ' homonyms ' or ' homophones ' are very numerous and 
 remarkable for the variety of their contents. Such 
 homonyms are by no means uncommon in English, when we 
 admit among them (as we obviously should, seeing that 
 language properly appeals to the car) words of which the
 
 1 1 8 Natural History of Language 
 
 pronunciation is the same though the spelling is different, 
 such as * way,' ' weigh ' ; ' so,' ' sew,' ' sow.' These cannot, 
 indeed, compare with those of Chinese either for frequency 
 or for multiplicity of meaning. Yet such ambiguities as 
 ' sound ' ( = A.S. siind, ' narrow of the sea,' and Fr. son, 
 ' noise ') are identical in character, so far as they go, with 
 those existing in that language. The English word ' bay ' 
 may mean a certain colour, a tree, the noise of a dog, or an 
 inlet of the sea. If such words are traced back to their 
 origins, it becomes manifest that they have been brought to 
 their present identity of shape by phonetic decay. Thus 
 ' bay ' in the four senses above mentioned corresponds to 
 dadt'us, bdca, abbayer, baia, respectively. What has occurred 
 here may very well have occurred on a much more 
 comprehensive scale in Chinese. This may not be in every 
 instance the one explanation of the extraordinary number 
 of quite distinct meanings often attaching to a single 
 Chinese word, but it is a demonstrable explanation in many 
 cases and a plausible explanation in many more. However 
 that may be, the material of Chinese now consists of less than 
 five hundred different monosyllables, with which, nevertheless, 
 all conceived ideas manage to find expression, thanks to 
 particular principles and methods of intonation, position, and 
 grouping. By these expedients the Chinese practically 
 attain to a vocabulary of forty thousand words. 
 
 Thus if the syllable tao in itself may, from whatever 
 cause, represent such dissimilar ideas as ' road,' ' flag,' ' cover,' 
 ' corn,' etc., and hi may signify ' road,' ' turn ' ' gem,' ' dew,' 
 etc., it is perhaps a clumsy, but still a fairly effectual, device 
 to combine the two words in order to make certain the only 
 sense which they ordinarily possess in common, and so 
 designate the meaning ' road,' unmistakably by means of the 
 reduplicating combination tao he. This procedure deserves 
 noting as one which may conceivably have occurred in 
 other languages. In other cases such homonyms are 
 differentiated by the use of those ' tones ' which give their 
 sing-song character to the Asiatic positional speeches.
 
 General Stirvey of Languages 1 1 9 
 
 Theoretically there are in Chinese eight such tones, the 
 ' upper ' and ' lower ' series of ' even,' ' rising,' ' departing,' and 
 ' entering,' the difference between the four in each series 
 being similar to that between the European customary- 
 pronunciation (the ' even '), the interrogative (the ' rising '), 
 the surprised interrogative (the ' departing '), and the strongly 
 insistent (the 'entering'). Such differences do not admit of 
 representation in ordinary writing, but something of their 
 effect may be gathered from observation of the various 
 manners in which ' Yes,' ' No,' ' Oh,' ' Ah ' are intoned in 
 English according to the feeling and implication in each 
 instance. In Chinese, however, the tones are utilised to 
 differentiate entirely the meaning of words, not merely the 
 mood of the speaker. Thus hah (' departing ' tone) = ' good,' 
 but had (' rising ' tone) = ' love.' All the eight tones are met 
 with in the Fukien dialect, but the official and literary 
 ' Mandarin ' speech employs only five, the upper series and 
 the ' even ' lower tone. 
 
 The origin of these tones is problematical. It is contrary 
 to what is known of language in general to suppose that the 
 special tones were assigned to special meanings by any 
 deliberate convention or conscious attempt to differentiate. 
 It is true that a differentiation which already exists may be 
 maintained in pronunciation, in spite of those ordinary 
 phonetic tendencies of a language, which, if left to them- 
 selves, would have worked confusion. P'or example, though 
 the English phonetic tendency is to accentuate as early as 
 possible in the word and to shift all accents forward, yet 
 'humane,' 'project,' etc., still defer their accent in order to 
 avoid confusion with ' human ' and project.' But any 
 deliberate effort to create de novo, or to re-create, a useful or 
 significant difference of pronunciation between words which 
 have once become identical in sound could only be made, if 
 made at all, by order of some Draco who absolutely controlled 
 a complete and uniform national system of education, and 
 who managed to force his artificial distinction upon a 
 generation. In actual experience no such event has ever
 
 1 20 Natural History of Language 
 
 occurred. The differences in the tones of Chinese words must 
 manifestly be acquired by Chinese children in infancy, just 
 as they hear them from the lips of their elders. The same 
 must have occurred in each generation. Consequently the 
 tones must have become attached to the words in the first 
 instance from natural causes, and have been thence trans- 
 mitted with such faithfulness as the survival of the fittest 
 demanded. Jespersen {Progress in Language) compares the 
 use of the high and low tones in a dialect of South Jutland, 
 by which such homonyms as nd (' fool '), nd (' fools,' or the 
 verb ' to fool '), Jem (' home '), Jem (' at home ') are 
 distinguished. The low tones there represent earlier 
 monosyllables (nar, hjeni), and the high tones earlier 
 disyllables {iiarre, Jijeme?) The variation in this case is 
 purely involuntary in its evolution. The first syllable in a 
 disyllabic word is not intoned in the same way as it would 
 be if it stood alone, and the continued existence of the 
 tone is due to a natural survival of the necessary, and not 
 to any conscious interference with the spontaneous develop- 
 ment of the language. 
 
 There is something of the same kind, though not an 
 identical phenomenon, in the case of the English ' vowel- 
 mutation ' or Umlaut. Thus originally the difference of vowel- 
 quality in Frz^nce and French was due to an influence 
 exerted on the a of the ' root '-syllable by a following 
 syllable containing i, viz. -isc in Frencisc. The mutation in 
 long, length, was caused by the same following vowel, which 
 has now disappeared. Had we been as ignorant of the 
 history of English as of Chinese it would have been difficult, 
 if not impossible, to account for this change of vowel-quality. 
 It might easily have occurred to the empiric that in Frafik, 
 French, the differentiation had been invented for the sake of 
 the meaning. In the same way as in the Jutland dialect, 
 it may be the case that Chinese tones are an inheritance 
 from a stage of the language when the words to which they 
 are now attached were longer, that is, contained more 
 phonetic material or even more syllables. Of the total
 
 General Survey of Langttages 1 2 1 
 
 original differences the tones would be the only survivals, 
 holding their ground of sheer necessity, and possibly 
 becoming further accentuated, though unconsciously, in virtue 
 of their utility. 
 
 Of Chinese grammar something has been said already. 
 Every word consists of one syllable and one only. Using 
 the term ' root ' in the ordinary acceptation, each word is a 
 ' root ' and each ' root ' a word. Such root-words admit of no 
 modification of form, whether by means of internal change 
 or by addition of other elements before or after the root. 
 Inasmuch, therefore, as a root contains the idea only in its 
 most general sense (e.g. td conveys the unspecialised idea 
 of ' greatness '), there is in Chinese no classification of 
 individual words corresponding to our own division into the 
 various ' parts of speech.' Td may be employed sometimes 
 where we should use a verb (' to be great,' or ' to make great '), 
 sometimes for our adjective ' great,' or our noun ' greatness,' 
 or our adverb ' greatly.' It is regularly with a Chinese 
 word very much as it is occasionally with English words 
 like ' love,' ' walk,' when looked at in isolation from a sentence. 
 A context is required before we can decide whether what is 
 intended is the noun or the verb. But whereas in English 
 such identity of form is rare for even two parts of speech, in 
 Chinese it is constant. There being thus no devices for the 
 expression of grammatical relations — of case, number, person, 
 tense, mood — by means of suffixes, prefixes, or any form 
 of flexion, other expedients have necessarily been developed. 
 A language cannot exist without grammar of some kind, 
 and the grammar of Chinese is mainly one of simple position 
 or word-order. That order is subject, verb, direct object or 
 complement, indirect object or complement. Qualifying 
 words (adjectives, adjectival genitives, adverbs) precede 
 the words qualified. Thus td obtains its function-mark 
 as a part of speech from its position : tdjin, ' great man,' 
 Jin td, ' (the) man (is) great.' Similarly in ngb (' I ') id 
 (' strike ') ni (' thou '), and ni (' thou ') td (' strike ') Jtgb (' I '), the 
 relative positions determine object and subject. In this case
 
 122 Natural History of L anguage 
 
 we happen to be dealing with pronouns, in which English 
 or French retains separate forms for subject and object. 
 Otherwise, had the words meant John strikes James and 
 James strikes John, the device of position would have been the 
 same for each language, though English or French would still 
 indicate person, tense, and mood in the word strikes or Jrappe. 
 So htM pad inin, ' (the) king protect(s) (the) people ' ; but 
 min pah huci, which is the reverse, is only made the reverse 
 by precisely the same expedient which would be employed 
 in English or French. 
 
 Nevertheless position alone has proved insufficient for 
 the necessities of the language. The danger of ambiguity 
 or indefiniteness has evolved its cure in the shape of those 
 symbolical, ' demonstrative,' or ' help ' words, which the 
 Chinese themselves call ' empty,' in opposition to the ' full ' 
 words, i.e. the words which express an integral portion 
 of that sense which the sentence predicates. The 
 difference may be illustrated by the English prepositions ' of,' 
 ' with,' ' to,' and the auxiliaries ' shall,' ' will,' which are 
 ' empty ' words in the sense of expressing no idea in them- 
 selves but only assisting in denoting the relations of other 
 words. To all intents and purposes Chinese has, therefore, 
 evolved prepositions and auxiliaries which precede or follow 
 the words whose reference they assist. Thus tsi (written 
 also tchi, li, tsi by various transcribers) comes to mean ' of 
 Jin idzin) kiun, literally ' man prince,' is, indeed, allowable 
 for ' prince of men,' but that meaning is made much more 
 definite in the form jin tsi kiun. So min lik, literally 
 ' people power,' may stand for ' power of the people,' but 
 that notion is more commonly and with greater precision 
 expressed by min tsi lik. Similarly y practically = ' with ' 
 (instrumental), yu = 'to' (dative), //='at,' 'in' (locative), 
 tsimg, ' from ' (ablative). There is, however, this distinction 
 between the Chinese ' empty ' words and our prepositions. 
 The Chinese ' empty ' words are recognisable predicative 
 roots which have come to play a subordinate part, while our 
 prepositions are felt to be in themselves wholly meaningless. 
 
 •i:
 
 General Survey of Languages 1 2 3 
 
 Thus in Chinese y cang = ' with (a) stick,' but whereas 
 ' with ' conveys no idea in itself, the Chinese y, when 
 independent, is the predicative root 'use.' So ///=' place' 
 and ' possess,' yti = ' give.' This discoverable meaning is 
 not, indeed, present in full force in the minds of the 
 speakers whenever the ' root ' plays this part of a relational 
 word, and it is for that reason that Chinese grammarians 
 have employed the term ' empty ' ; nevertheless the step 
 towards an emptiness as complete as that of our own 
 prepositions is only half made. Though the original 
 meaning is obscured and in a certain measure arbitrarily 
 transferred, it has not wholly faded out of the word, nor 
 has the form of the word suffered. So far as an empty 
 word suffers any loss in value phonetically, it is due only to 
 the fact that in the case of a couple of words, of which one 
 is a ' full ' and the other an ' empty ' word connected with it, 
 the two fall under one main accent, which is laid upon the 
 'full' root. It must be added that the cases of nouns are 
 not the only relationships expressed in this way. Thus, 
 though it is possible for the plural sense to be inferred from 
 the context, it is made more definite by the use of sub- 
 ordinate words, e.g. to (' many '), hi (' number '), of which it 
 cannot be said that they possess no meaning of their own, 
 while, on the other hand, it cannot be pretended that they 
 are thought of in the full value which they would possess 
 when independent. 
 
 The laws of position being unavoidably rigid, the total 
 result of the Chinese devices is no doubt unsatisfactory to the 
 European, who looks in literature for perfect lucidity com- 
 bined with variety and other charms of style ; but there is 
 no evidence that the Chinese themselves experience any 
 appreciable difficulty in the mere communication of thought. 
 Strongly marked dialectal differences, mainly phonetic, 
 exist in various parts of China. The official and literary 
 language of the Empire is that of the Mandarin dialect 
 spoken about Pekin and Nankin. Other local varieties of 
 chief note are the dialects of Fukien and Canton. Thus in
 
 1 24 Natural History of Language 
 
 the province of Fukien are heard the sounds b and g, which 
 do not occur in Mandarin {ng being a mere sign of nasalisa- 
 tion). In Canton ki represents an older form of the ts of 
 the more northerly regions. The divergences, however, are 
 of such a kind that they do not bring into question the 
 unity of Chinese as a language. 
 
 Whether the Chinese language was always monosyllabic, 
 or whether, at some earlier undiscovered stage, it may have 
 possessed longer words and flexion-endings, is an open 
 question. That there has been much phonetic decay, in the 
 uniform direction of abbreviating words, has been already 
 stated. It is not inconceivable that this decay has been the 
 means of eating away a whole system of flexion and of 
 reducing words of two or more syllables to one syllable only. 
 For both results it is natural to compare the French roi (with 
 which we may treat rois as identical, since the pronunciation 
 is the same) as the sole representative of regent, regis, regl, 
 rege, reges, reguin, regibus ; or ange {anges) as the sole repre- 
 sentatives of angelus, angeluin, angclds, etc. The cases are 
 not, of course, identical, since roi properly = only the accusative 
 regent, and rois = only reges, the remaining Latin cases having 
 disappeared. Nevertheless the illustration is instructive. It 
 has also been stated already that the existence of the Chinese 
 ' tones ' may perhaps be best explained from the assumption 
 of a large amount of phonetic decay. Moreover, if the 
 development of help-words is recent as compared with the 
 time during which the language must have existed, i.e. if 
 they have only been gradually coming into use, it seems 
 highly probable that they have been brought into employ- 
 ment in much the same way as prepositions and auxiliaries 
 have been evolved in the analytical Indo-European tongues, 
 namely, as substitutes for worn-out modes of expressing 
 grammatical relations. If so, the old Chinese modes were 
 not simply the mode of word-position, inasmuch as that still 
 exists and is found insufficient. 
 
 The view usually entertained is that Chinese actually
 
 General Szirvey of Lmiguages 1 25 
 
 presents us with an example of language in its most 
 crude and primitive form, namely, that of a series of simple 
 roots at which human utterance has arrived for the ex- 
 pression of certain ideas, but which the speakers have not 
 yet learned to relate organically. It is then argued that the 
 recent development of help-words is a sign that the belated 
 Chinese is making the first step towards the so-called ' higher ' 
 stage of agglutinative inflexion. Closer argument, however, 
 is bringing philologists rather to the opinion that the Chinese 
 language may have actually passed through the synthetic 
 inflexional stage, as English has largely done ; that it may 
 have lost its flexions by phonetic decay, while transferring 
 their functions to the order of the words (as English has also 
 done in a large measure) ; but that its phonetic decay may 
 have been too effectual in its effacement of distinctions both 
 of meaning and relation, and that therefore the modern 
 speech is striving to regain some of its lost precision by 
 means of the help-words and an increasing tendency to new 
 agglutination or combination. Having secured its device of 
 word-order (for which there must apparently be some good 
 psychological basis, seeing that Chinese agrees with the 
 entirely unrelated modern analytical tongues of Europe in 
 fixing it as subject, verb, direct object, indirect object), 
 Chinese is not likely to go the whole round once more. 
 Rather it may work up by its own route to a condition 
 similar to that which will belong to our own tongue in its 
 furthest conceivable analytical completeness. It is probable, 
 not that English is working towards the condition of Chinese, 
 but that both are working towards a common goal. 
 
 It is easy to overstate the resemblance of the structure 
 of an English sentence to that of a sentence in Chinese. Our 
 grammar is indeed largely positional, or dependent on 
 'empty' words, but we still possess certain of the more 
 useful flexions, and above all our vocabulary, thanks to the 
 power of compounding and to the established differentiation 
 of its material, is so distributed over the various parts of 
 speech that there is little chance of a confusion of nouns,
 
 1 2 6 Natural History of L anguage 
 
 adjectives, verbs, and adverbs with each other. We possess, 
 for instance, ' large,' ' largely,' ' largeness,' ' enlarge,' ' enlarge- 
 ment,' and not even a context is required to distinguish the 
 senses which, in the case of the Chinese td, are left to depend 
 wholly upon position or help-word. We need not therefore 
 anticipate that the prosaic and styleless monotony of Chinese 
 writing must become the fate of European literature. The 
 distribution of the vocabulary into definite parts of speech, 
 so that a word, wherever placed, will always be recognised 
 for the part of speech it is, gives scope for much freedom in 
 the building of sentences ; whereas in Chinese the confusion 
 would be utter, unless there were rigidly maintained that 
 one recognised order which alone can show whether 
 the sterile monosyllable is to be treated as a verb or a 
 substantive. 
 
 II. Anamese, the group embracing the dialects of 
 Tonkin, Cochin China, and Cambodia, is in structural method 
 identical with Chinese, being positional and monosyllabic. 
 It possesses six corresponding ' tones.' Its vocabulary, how- 
 ever, is widely different, though in this respect the question 
 is complicated by the fact that Chinese vocables have been 
 borrowed very freely. 
 
 III. Siamese (or Thai) comprises several dialects. 
 Though monosyllabic and positional with ' tones,' its vocabu- 
 lary is again distinct from both Chinese and Anamese ; and 
 in point of word-order it differs from Chinese in making the 
 defined word precede the defining. 
 
 IV. Burmese (comprising various dialects of Burma, 
 Assam, and the neighbourhood) in respect of structure agrees 
 with Chinese, but is peculiar in its vocabulary. 
 
 V. Tibetan. — This group is by some authorities com- 
 bined with the last-named, so that an affinity of the dialects 
 of Tibet, Nepaul, Bhotan, Assam, and Burma, as well as of 
 parts of South-West China, is distinctly claimed. Tibetan is 
 asserted to be less strictly monosyllabic than the other groups 
 in this series.
 
 General Survey of Languages 1 2 7 
 
 It will be remarked that all these tongues, which run 
 into each other geographically, exhibit, apart from their 
 vocabularies, the closest resemblances. Their grammar is 
 one of word-order ; their words are monosyllabic ; they 
 employ, in various manners and measures, the expedient of 
 intonation. Their differences lie chiefly in the dissimilar 
 words which they use for the same notion. Considering 
 how easily vocabulary is changed, and how unlike it may 
 become in a few centuries between languages undeniably 
 related genealogically, there would be no a priori unlikelihood 
 that all these groups of language may have derived the 
 impulse to their structural method and phonetic corruptions, 
 and originally their stock of words also, from a common 
 source ; and that, while they have alike carried out the 
 inherited tendency to positional monosyllabism, they may 
 have worked asunder in their words, until in that respect 
 they are languages practically alien. The English you arc a 
 poet is represented in German by Sie siiid ein Dichtcr, and 
 though a and ein are radically the same word, the fact can 
 only be known from the discoverable history of older English 
 and German, and not from the shape of the words as we 
 now use them. For the rest, you, are, poet, are words quite 
 alien to current German, while Sie, sinci, DicJiter are equally 
 alien to current English. Nevertheless, it is a fact that both 
 the grammar and the words of yoii are a . . . and Sie sind 
 ein . . . come from a stock of material originally possessed 
 in common {i.e. primitive) Teutonic. It is, of course, not im- 
 possible that the same experience has occurred to a far wider 
 extent in the isolating tongues of South-East Asia. Unfor- 
 tunately the question is not yet determinable upon the evidence. 
 We must be content to take the languages as they are, and not 
 seek to go beyond the warrant of the data. Some specialists 
 (E. Kuhn, for example) speak of the several groups as 
 'certainly related.' But it has already been pointed out 
 that the only criteria of the genealogical connection of 
 languages are grammar and vocabulary. The fact of racial 
 and geographical association no doubt encourages an im-
 
 128 Natm'al History of Language 
 
 pression already formed on these data, but in itself it 
 furnishes no trustworthy evidence. Falling back, therefore, 
 on these two tests, we do not find scholars of these groups 
 of tongues prepared to agree that there is sufficient corre- 
 spondence of vocabulary to warrant their affiliation to a 
 common parent, although in respect of grammar such an 
 affiliation would be entirely natural. 
 
 It is quite possible that there should be simply com- 
 municated from one stock of language to a neighbouring 
 unrelated stock a tendency to a sing-song or a lazy pro- 
 nunciation, of which the result would be the cutting-down 
 of words to intoned monosyllables. Such phonetic peculi- 
 arities are undoubtedly highly contagious, as is seen upon 
 all borders between two languages. They would be still 
 more contagious if commerce, wars, lengthened occupations 
 of territory, and intermixture of blood took place from time 
 to time. If two conterminous peoples with quite distinct 
 languages were infected with the phonetic malady above 
 described, the result would be in each case that language 
 would come to consist of monosyllabic remnants of previous 
 longer vocables. As those previous words are quite distinct 
 in the two tongues, so their monosyllabic remains would be 
 quite distinct. The flexions of each tongue being then 
 effaced, each will evolve a compensating device, and that 
 they can hit upon the same expedient of word-position 
 without consulting each other is shown by the fact that the 
 English and French have hit upon it without reference to 
 each other or to the Chinese. It may thus very conceivably 
 come about that, while Chinese, Anamese, Siamese, etc., 
 agree in monosyllabism, intonation, and a certain method of 
 word-order, those languages remain as distinct as ever in the 
 word-material upon which the common tendencies have had 
 to act. The genealogical connection of these speeches is 
 therefore at present unproven.
 
 CHAPTER VIII 
 
 GENERAL SURVEY OF LANGUAGES {continued) 
 
 II. Agglutinating-Inflexional Tongues 
 
 (i.) Suffix- Agglutinating Type 
 
 I. The Ural-Altaic family (which also bears, or has borne, 
 according to the fancy of philologists, the names Finno- 
 Tataric, Scythian, and Turanian) spreads, though not 
 continuously, over a vast region from Turkey, Hungary, and 
 Finland in the West to the sea of Okhotsk in the East, and 
 from the Mediterranean to the Arctic Ocean. The name 
 Ural-Altaic here adopted, though unsatisfactory from a 
 contemporary geographical point of view, is the one most 
 favoured, and is perhaps not inaccurate if taken as roughly 
 indicating two main landmarks and primitive centres in 
 the region from which the various languages probably dif- 
 fused themselves in their several directions, that region 
 being the area about and between the Ural and Altai 
 mountains. 
 
 Whether the languages at present classed under the title 
 are really in all cases genealogically related, is a matter of 
 doubt. The question of racial relationship, as will be 
 shown hereafter, has no necessary connection with the 
 question of linguistic relationship. The languages within 
 each of the special groups (which are to be detailed 
 immediately) are beyond doubt closely akin, but the relation- 
 ship of one such group to another is by no means so distinct. 
 (B 884) 129 K
 
 130 Natural History of Language 
 
 It cannot for the most part be convincingly supported by 
 the criterion of sound-material or vocabulary, inasmuch as 
 any common stock of words or even ' roots ' is sometimes 
 hardly appreciable. Particularly weak is the evidence for 
 connecting the Tunguse and Mongolian divisions with the 
 other three hereafter mentioned. 
 
 In the matter of grammatical method, however, their re- 
 semblance to each other is very obvious and striking. The 
 typical agglutinative device, by which an unlimited number 
 of specially dynamic terminations may be affixed to an 
 unchangeable root, is shared by them all, although the extent 
 to which the operation is kept clear of amalgamation varies 
 considerably ; Finnish, for example, having become in this 
 respect almost a language of the same class as the Indo- 
 European. Another common feature is the attachment of 
 possessive pronouns to the noun. 
 
 Particularly observable is their close agreement in the 
 principle known as ' vowel-harmony,' according to which, 
 when the several suffixes are agglutinated to the main root, 
 the vowels of those suffixes are brought by a regular variation 
 (or Umlaut) into harmony with the vowel of the chief root. 
 Speaking generally, and without striving after a dispro- 
 portionate particularity, the vowels are ' heavy ' or ' light,' and 
 a heavy vowel in a root will be followed by heavy vowels in 
 the suffixes, and a light vowel by light vowels. Thus if a, 
 o, u be regarded as heavy, and a e, i, <?', ii as light, the 
 appearance of the heavy a in the main element, as in the 
 Turkish yas (' write ') or baba (' father '), determines the 
 character of the vowels of the subordinate suffixed elements, 
 which will necessarily be heavy also. Thus yaztnak (' to 
 write '), babalaruvidan (' of our fathers '). On the other hand 
 the appearance of e in the main element will necessitate the 
 ' light ' character in the suffixes. Thus sevinek (' to love '), 
 dedelerinden (' of our grandfathers '). So evler (' houses '), 
 but atlar (' horses ') ; and in Magyar whereas ' to the shepherd ' 
 is yuhassnak, ' to the gardener ' is kertesznek. In Magyar 
 the nominative, accusative, and dative plural of Magyar
 
 General Survey of Languages 131 
 
 are respectively Magyarok, Magyaj'okat, Magyaroknak, while 
 the corresponding cases of Torbk are Torokok, Torokoket, 
 Tdrokohiek. Ideal examples can, it is true, be gathered only 
 from Turkish, Magyar, Finnish, Mongolian, and Manchu. 
 Nevertheless the principle appears, though with less complete 
 application, in the remaining speeches also. 
 
 Such a phenomenon as vowel-harmony might, after all, 
 be merely of interest as showing one natural tendency of 
 languages with the pure form of suffix-agglutination. It 
 might indicate a similar process, partly physical, partly 
 psychological, developed independently by the speakers of 
 languages morphologically similar. Thus the main root or 
 element, being of predominant importance, is distinctly 
 enunciated, while the suffixed elements, so long as they 
 serve their purpose, are not dwelt upon with the same 
 insistence of precise articulation, and their vowel-quality takes 
 its tone from that tone of the predominant vowel which still, 
 as it were, echoes in the mind. That something of the kind 
 has actually occurred, seems to be borne out by observations 
 of Ural-Altaic specialists, that the law of vocalic harmony 
 appears to be of comparatively recent development. In any 
 case the phenomenon cannot be taken as doing more than 
 lend a strong support to other arguments for original com- 
 munity of speech. 
 
 Understanding, then, that any genealogical connection of 
 the more remote groups is still debatable, we may divide 
 the Ural-Altaic * family ' as follows : — 
 
 A. The Finno-Ugfric, otherwise known as the Ugro-Finnic, 
 Finnic, or Ugrian, comprises the Finnish (or Suomi) of 
 Finland and cognate dialects of less account, Esthonian, 
 Livonian, the Lapponic of Lapland, certain dwindling 
 dialects on the Upper and Middle Volga, others to the north- 
 east of these, some dialects about the river Obi in Siberia, 
 and finally the Magyar of Hungary. 
 
 Although there are sometimes considerable divergences 
 exhibited by the languages within this group, their relation-
 
 132 Natural History of L anguage 
 
 ship to each other admits of no doubt. In point of 
 vocabulary we are met by such correspondences as Finnish 
 kdsi (' hand '), Magyar kez^ Mordvinian ked, Ostiak ket ; 
 Finnish kzvz (' stone '), Mag. ko, Mord. kdv, Ost. keu ; Finn. 
 nuotz Q arrow '), Mag. njz't, Mord. nat, Ost. njot ; Finn, vest 
 (' water '), Mag. viz, Mord. vdd ; while in point of grammar 
 the Finnish nominative, genitive, and ablative sz/md (' eye '), 
 silmdn, szlvidltd, etc., are answered in Mordvinian (for 
 example) by selmd, selindn, selniddd. It would be natural, 
 however, to expect that these tongues, and particularly 
 Magyar and Finnish, would contain many words (especially 
 ' culture-words ') borrowed from the Indo-European languages, 
 Teutonic and Slavonic, which adjoin them. To some ex- 
 tent also similar influences may have assisted in breaking 
 down the original Ural-Altaic grammar at certain points in 
 Finnish. 
 
 Specialists are not entirely agreed as to the most 
 accurate division of the Finno-Ugric branch into its sub- 
 groups. The following arrangement is perhaps least open to 
 objection : — 
 
 1. Finnic — 
 
 {a) Finnish (or Sziomi), spoken in the greater part 
 of Finland proper, and possessing a cultivated 
 literature, including the Kalevala, an old epic 
 of twenty-two thousand verses, which has, 
 however, been modernised. 
 
 {b) Tchudic and the cognate dialects Vepsic and 
 Votic, about Lake Onega. 
 
 {c) Karelian, from the White Sea to Lake Ladoga. 
 
 (c/) Livonian (a dialect nearly obsolete) and 
 Crevinian in Courland. 
 
 {e) EsthoTzian, south of the Gulf of Finland. 
 
 {f) Lapponic, of Lapland. 
 
 2. Permian, embracing Permian proper, Szryenzan, and 
 
 Votiak. These are spoken by sparse populations 
 near the Urals in the E.N.E of European Russia.
 
 General Stirvey of Languages 1 2)2) 
 
 3. Volgfa-Finnic (sometimes called Bulgaric, their region 
 
 being that from which the Bulgars emerged in the 
 seventh century). This group includes two divisions — 
 
 {a) Tchej-ejnissian, between Nijni-Novgorod and 
 Kasan. 
 
 {b) Mordvinian on the Middle Volga. 
 
 4. Ugric, which comprises — 
 
 {a) Vogul and Ostiak, dialects of a few thousands 
 scattered over a wide region eastward from 
 the Northern Ural and about the Obi River. 
 
 {b) Magyar (or Hungarian), an important and 
 cultivated language spoken over two separate 
 areas, the larger covering the central and 
 more western portion of Hungary, the smaller 
 lying to the S.E. in Transylvania, where it 
 is surrounded by Roumanian. The language 
 possesses linguistic data from as far back as 
 the end of the twelfth century. 
 
 B. The Samoyedic group, including a number of dialects 
 scattered eastwards from the White Sea and along the western 
 portion of the Arctic shore of Siberia. Perhaps not more 
 than twenty thousand persons speak these tongues. 
 
 C. The Tung'use group, embracing languages spoken by 
 the Tunguses, from the Yenissei in Central Siberia to the 
 coast of the Sea of Okhotsk, and also by the Manchus of 
 Manchuria. These dialects are totally uncultivated. 
 
 D. The Mongolian group, also uncultivated, com- 
 prising — 
 
 {a) Mongolian proper (or Shan-a), spoken in the north 
 of the Chinese Empire westward of the Manchu. 
 
 (J?) Buriat dialects, about Lake Baikal. 
 
 {c) Kalmuk (or Western Mongolian), reaching westward 
 to the Caspian and penetrating into European 
 Russia between the Lower Volga and Don. 
 
 E. The Turko-Tataric group (sometimes known simply 
 as the Tataric or Turkic) is widely diffused, reaching, though 
 without continuity, from European Turkey to the river
 
 134 Natural History of Language 
 
 Lena in East Siberia. The linguistic connection within 
 this group is very close, the languages of the extremes, 
 Turkish and Yakut for instance, being at least as distinctly 
 related as English and German. The sub-groups usually 
 recognised are — 
 
 {a) Yakut, spoken by tribes on the Lena in Siberia, 
 who are surrounded by Tunguses. 
 
 (/;) Kirghiz, comprising the speeches of the Black Kirghiz 
 (or Bu7'uts) in the part of Turkestan bordering on 
 China, and of the Cossack Kirghiz to the north of 
 the Caspian, the Sea of Aral and Lake Balkash. 
 
 {c) Nogair, of the Russian Cossacks, in the Crimea and 
 in parts of Russia from south of the mouth of the 
 Volga towards the Sea of Azov. 
 
 id) Uiguric, including Uigur proper, Jagataic, and 
 Turkoman, the dialects spoken in the parts of 
 Turkestan not occupied by the Kirghiz. 
 
 {e) Turkish (or Osmanli, the latter term, however, being 
 properly used of the official and cultivated Turkish, 
 which contains large Persian and Arabic elements). 
 This language is spoken in the interior of Asia 
 Minor and on its north-east coast (the north-west, 
 west, and south coasts being occupied by Greek), 
 also in Constantinople and other chief centres of 
 European Turkey (the coastward parts being again 
 chiefly occupied by Greek). Turkey possesses a 
 copious literature, the older parts consisting mostly 
 of poetry, story, and proverb. 
 
 TJie Dravidian Languages 
 
 II, The Dravidian family of languages (sometimes called 
 the Tamil, by an illegitimate extension of the name of a 
 portion to the whole) occupies the southern half of the 
 peninsula of India from the rivers Nerbudda and Godavery 
 to Cape Comorin, together with Northern Ceylon. The 
 name 'Dravida' was that applied indiscriminately by the
 
 General Survey of Languages 135 
 
 Aryan {i.e. the Sanskrit-speaking) invaders of North-West 
 Hindustan to the earlier dark-complexioned inhabitants of 
 India both north and south, Ceylon, and Beluchistan. 
 Ethnologically, however, these belong to three distinct races, 
 viz. the Dravida in the narrower sense, the Kol, and the 
 Cingalese. Nor do the languages of these three divisions 
 show any genealogical relationship. 
 
 {a) Kol (or Kolarian) is the name of an uncultivated 
 group of dialects spoken by tribes in the highlands 
 of Chota Nagpur (west from Calcutta). They are 
 classed as agglutinative. Thus apu = ' father,' 
 apiiWng = ' our father,' apin^e = 'your father.' 
 Similarly, the quasi-verb is made active, passive, 
 intensive, causative, etc., by means of suffixes piled 
 upon each other after the manner of Magyar or 
 Turkish. A peculiar feature, however, is the use 
 of infixes like those of some Malayan languages. 
 
 {b) Cingalese (or Ehi) is an exceedingly mixed tongue, 
 having been dominated by Sanskritic influences, 
 till in point of vocabulary it has become practically 
 an Aryan language, while it is impossible to 
 classify its grammar. 
 
 {c) The Dravidians in the narrower sense speak languages 
 which, though now widely divergent, are generally 
 agreed to have been derived from a common 
 stock. Borrowed words from the ' culture 
 language' of the civilising Aryans of the north 
 have naturally found their way in considerable 
 numbers into these speeches, but have not altered 
 their morphological character. 
 
 The chief subdivisions of the Dravidian family are 
 these : — 
 
 I. Telugu (or Telingd), the north-eastern division, 
 extending nearly to Madras along the Coromandel 
 coast, and in Hyderabad between the Lower and 
 Middle Godavery and Krishna.
 
 136 Natural History of Language 
 
 2. Tamil, which from a literary point of view is the 
 
 most important of the Dravidian speeches, and is 
 spoken on the eastern side of the peninsula from 
 a few miles north of Madras (Pulicat) to Cape 
 Comorin, as well as in North Ceylon. 
 
 3. Malaydlain, occupying a comparatively small area of 
 
 the west or Malabar coast south of Mangalore. 
 
 4. Tidu, current in and about Mangalore, employed by 
 
 only a few hundreds of thousands. 
 
 5. Kanarese, extending over the greater part of Mysore 
 
 and the western portion of the Nizam's Dominions. 
 
 But besides these main and more or less cultivated 
 divisions there are many less important dialects, such as the 
 Tuda and Kota in the Nilgherries, the Kudagu (or Kurgi) in 
 Kurg, the Gond in the Vindhya hills, the Kond in the hills of 
 Orissa, and Brahui in Beluchistan. 
 
 In a certain measure prefixing is not unknown in these 
 tongues. Thus in Tamil d,thu = ' that thing,' it/iu = ' this 
 thing,' ef/iu = ' which thing.' Prefixing, however, cannot be 
 called characteristic, any more than the prefixing of the 
 augment in Greek or Sanskrit can be called a characteristic 
 method of flexion in Indo-European. The grammatical 
 device of the Dravidian speeches is that of the Ural-Altaic 
 family, viz. the modification of sense and relation by means 
 of suffixes readily attachable and detachable. It is common 
 to quote from Tulu such forms as mat-puve (' I do '), mal-p&ve 
 (' I cause to do '), ma/]p^ve (' I often do '), with the remark that 
 change of meaning can thus also be expressed by internal 
 change of vowel. It would be more cautious to say that the 
 different forms are so many detachable elements or groups of 
 elements expressive of certain special modifications of mean- 
 ing. The differentiation of those elements in themselves may 
 or may not be due to the same causes which differentiated 
 the Semitic roots by means of internal vowel-change. It is, as 
 a fact, quite improbable. The likelihood is that J>uv, pdv, 
 pev, are combinations and not simple elements. Here, how-
 
 General Survey of Langtcages 1^,7 
 
 ever, we are only concerned with their suffix-use as we find 
 it. This is entirely similar to the method of Ural-Altaic. 
 Thus malpdve is causative, malpdimi is its negative, 
 malpuvedidt, is the conditional-conjunctive, iiialtoxLdLUVpuve is 
 the durative present. 
 
 The declension of a noun according to the ideal 
 agglutinative principle, by which, if ^ = root-element, x = 
 sign of plural, jj/ = sign of case, a case Ay of the singular is 
 answered by Axy in the plural, may be illustrated by the 
 Tamil noun palan, ' reward.' Thus — 
 
 Sing. Nom. palan Plur. palang'sX 
 Voc. pala?j& palangalh 
 
 Ace. pala?ie\ palatjgal^x 
 
 Gen. ;>a/a«udeiya pahmgalyxAAVSZ., etc. 
 
 The differentia (which, as has been before observed, is 
 not always clearly marked) between an agglutinative- 
 inflexional and an amalgamating-inflexional tongue is 
 obvious in the comparison made by Misteli-Steinthal of 
 the treatment in declension of the word scvaka (' servant ') in 
 Sanskrit on the one hand and Kanarese on the other. Thus, 
 in the plural, the instrumental, dative, and locative cases 
 respectively are in Sanskrit scvak^'^, j-rz'rt'/'ebhyas, scvakhhy., 
 where there is no distinctive sign of the plural, but where 
 number and case are combined (no matter how the fact came 
 about) in one general modification of the terminational 
 material. In Kanarese the same cases are expressed by 
 sevaka-x-mdi3., scvaka-x-{))gQ, scvaka-x-{?i))M, where the sign 
 of the plural, r, is consistent as well as independent of the 
 case-suffix. Again, while {d)lli is the locative suffix and ge 
 the dative, it is possible to say sevaka-x-z^\-gQ, ' there to the 
 servant.' 
 
 It may be as well, however, to repeat here that, while we 
 are justified in placing this family with the Ural-Altaic under 
 a common morphological heading of some convenience, we 
 must recognise (i) that there is no fundamental unlikeness 
 between their methods and that of a language which can 
 create the series true, tru-tla., tru-t'h.-iul, /r/^-th-ful-ly, or one
 
 1 38 Natural History of Language 
 
 which can say heUvvfii {deikrviivii), ' I show,' Sei'^w ideik-^-o), 
 ' I shall show,' etc., for though the degree of transparency or 
 of ease of analysis may differ in the two cases, the principle 
 is the same : (2) that it is incorrect to say that the speakers 
 of such a language as Tamil are always alive to a substantive 
 and independent meaning of those affixes which they thus 
 systematically attach. 
 
 (ii.) Prefix- Agglutinating Type 
 
 The Bantu family (otherwise called the Kafir, a term 
 which might be misleading, inasmuch as the title ' Kafir,' or 
 ' Unbeliever,' once applied by the Mohammedans to all the 
 peoples of South Africa, is now restricted to the inhabitants 
 of a particular stretch of country north-east of Cape Colony) 
 is of great geographical extent. The word bantu (Zulu and 
 Kafir abantu, Sesuto batu, etc.) is the form of the word 
 ' men ' which may be taken as most typical for all the 
 speeches included in the family. In point of situation these 
 tongues cover the greater part of South and middle South 
 Africa, between Natal and about 5 ' north latitude. Alien 
 to them in the south-west are the languages of the yellow 
 Hottentots and Bushmen, and to the north they are limited 
 by the domain of the true negroes, the Gallas and the Somali. 
 North of the Hottentots they stretch across to the Atlantic. 
 Specialists enumerate some hundred and fifty languages or 
 dialects belonging to the family, none of these diverging 
 from any other more widely than English from German or 
 Latin from French. They further divide the whole into 
 three main groups, viz. : — 
 
 1. Eastern — including Kafir, Zulu, the speeches of 
 
 Mashonaland, of Mozambique, and those of the 
 Zanzibar coast, known as Kisuahili, Kinika, 
 Kikamba, Kipokomo, etc. 
 
 2. Central — including Sechuana (the speech of Bechuana- 
 
 landj and its dialects Sesuto (of Basutoland) and 
 Serolong ; Tekeza, etc.
 
 General Survey of Languages 1 39 
 
 3. Western — including (as we proceed from the south) 
 Herero (in Damaraland), Bunda (in Angola), 
 Congo, Mpongwe (in Gaboon), Dikele, Isubu, 
 Fernando Po, Dualla (in the Cameroons), etc. 
 
 The distinguishing mark of this family is its employment 
 of prefixes rather than suffixes, though, here again, it is 
 wholly incorrect to speak as if suffixing formed no part of 
 its structural system. The verb, for instance, may be con- 
 jugated with various modifications of sense expressed by 
 the attachment of post-positions. Thus in Kafir, tandz., ' love,' 
 ta7id\^2,, ' make to love ' (causative), /rtwrt'ana, ' love one 
 another ' (reflexive), tand\s,z.XiZ., ' cease to love one another ' 
 (reflexive-causative), tandQ"&3., ' be loved ' (passive). Between 
 a series like this and one of those quoted for Turkish or 
 Tamil there is plainly no distinction to be drawn. Never- 
 theless the method of prefixing is the one which is regular 
 and characteristic for Bantu. Thus, whereas the suffix- 
 agglutinating languages express a dative case by means of a 
 special ending, Bantu employs a prefixed particle, eg., Kafir 
 ku//, ' to us,' kuwz, ' to them,' ku;>, ' to him,' etc. ; and whereas 
 plurality in the typical Turkish or Dravidian is expressed in 
 the formula Ax (the plural case being Axy), in Bantu the 
 plural formula becomes xA (the case being yxA). Thus 
 ^zntu, ' men,' is the Zulu plural of {uinu)ntu, while ' with the 
 men ' is expressed by ngaba;//?/. 
 
 It is necessary, however, to premise that in Bantu the 
 singular itself is denoted by a prefix. In Kafir (which may 
 be taken as typical) one of the forms uin, u, Hi, in, isi, ulu 
 is employed for the singular, e.g. umntu (' man '), -drnwo 
 (' father '), ilis-wz (' word '). For these there are substituted 
 in the plural aba, ama, or ivii (corresponding to uvi), 
 (corresponding to u\ i, ama (to Hi), izin (to in), id (to isi), 
 etc. Thus msLiihi becomes abaw/w, nbawo becomes obaivo, 
 Tlizwi becomes ama,cz£'z'. 
 
 A peculiarity of these tongues is further that they 
 exhibit a most elaborate system of sustained grammatical
 
 1 40 Natural History of Language 
 
 concord, similar in principle, as Jespersen says, to the Latin 
 7nultbrvim, j/iagnomm, virdmm, but pervading the sentence 
 to a more complicated and unnecessary extent. The system 
 may be best appreciated in an example, as given by Bleek. 
 In Kafir 
 
 umMn/u vjetii ovOMchle nyahonakala sim.tanda 
 = ' man our(s) handsome appears, we-him-love,' 
 
 where the ' pronominal ' prefix inu is repeated, or represented 
 by its equivalent {w, u, nt), in every word. Converted into the 
 plural the sentence becomes dibdintu hetu zki^^chk "bd^yabonakala 
 si\idi,tanda. Similarly, 
 
 d?Odbfatia hamz sib&^uhe "b&fanda 
 
 = ' boys my large they-love ' ; 
 
 but 
 
 izinkomo zanii '\zva.kulu zifanda 
 
 = 'cows my large they-love.' 
 
 It may be observed in passing that there is a great 
 resemblance of principle between this and the conversion of 
 the Latin uostQX magnviB homo appdret into nostrl viagiil 
 hojuines app&reul, as compared with the conversion of nostrz. 
 magna, fenimsi appdrcX into nostrz.^ magndiQ feJuiiiduQ 
 appdki'ent. In Bantu, however, the verb, besides being in the 
 plural, is made to share in the special concord of form, while 
 in Latin its termination is the same for any gender. 
 
 (iii.) Prefix-, {Infix-,) and Suffix- Agglutinating Type 
 
 The Malayo-Polynesian family includes languages spoken 
 in the Malay Peninsula, in the East Indian islands, the 
 islands of the China Seas as far north as Formosa, the 
 islands of the Pacific from New Zealand to Hawaii and 
 Easter Island, and also in Madagascar (but without includ- 
 ing the Papuan speeches of New Guinea, sundry Negrito 
 dialects in parts of the Philippines, etc., or the aboriginal 
 languages of Australia). The family falls naturally into 
 three divisions, the Malayan, the Melanesian, and the
 
 General Survey of L anguages 1 4 1 
 
 Polynesian, which, apart from phonetic differences and 
 differences of vocabulary, are so distinguished in structure 
 as to represent three steps in morphological development. 
 To the Polynesian branch belong languages which, if taken 
 alone, could never be classed as agglutinative, inasmuch as 
 they express grammatical relations by means of free particles, 
 after the manner of the most advanced analytical tongues, 
 such as English. The Melanesian branch goes so far as 
 to suffix the possessive pronouns ; while in the Malayan 
 division determining elements are freely attached in the 
 shape of prefixes, suffixes, and also, in some measure, infixes. 
 The prefixes predominate, but not to the same extent as in 
 the Bantu languages. Friedrich Muller strangely assumes 
 that the Polynesian branch represents the lowest stage, from 
 which the Melanesian has advanced one degree, while the 
 Malayan has made considerably greater progress. The 
 natural conclusion is rather the reverse. There can be little 
 doubt that the Polynesian languages were carried over the 
 Pacific by emigrants from the Malay Peninsula or the East 
 Indies, and that in transit their grammatical system was 
 broken down by the influence of Papuans and others, some- 
 what as Persian grammar was largely broken down by 
 Arabic influences, and old English by those of Norman 
 French. Meanwhile, the Malayan, remaining nearer home, 
 would naturally remain more conservative in this respect, 
 because subjected to less external interference. Moreover, 
 the analogy of the modern European tongues, which have 
 all tended in a greater or less degree to abandon inflexional 
 attachments for free particles, lends support to the view that 
 the Malayan represents the older type. 
 
 In any case the fact that the several groups belong 
 genealogically to a common family has been placed beyond 
 doubt,^ and as, despite the changes which have come over 
 its structure, the account of analytical English will be most 
 
 1 Compare, for instance, on the side of vocabulary, the Tongan words (of the 
 Polynesian branch) aka ('root), ato ('roof'), ao ('cloud'), ika ('fish'), laiigi 
 ('heaven') with the Malay akar, atap, awaii, ikaii, Idngit.
 
 142 Natit rat History of L anguage 
 
 naturally given when treating of the amalgamating-inflexional 
 family of Indo-European, so the account of the no longer 
 agglutinative Polynesian must be included here in treating 
 of the family to which it belongs. Judging by Malayan, 
 that family was one whose methods of agglutination differed 
 from those of either Bantu or Ural-Altaic in allowing a 
 choice of three positions for the determining element. 
 The recognised subdivisions are as follows : — 
 
 (i.) The Malayan branch, comprising — 
 
 (a) The Malayo-Javanese group, to which belong — 
 Malay proper, spoken in the Malay Peninsula, in 
 
 the greater part of Sumatra, and on the coasts 
 
 of Borneo, 
 Battak, Achinese, and Lampong in parts of 
 
 Sumatra. 
 Javanese (or Kawi), of three -fourths of the 
 
 population of Java. 
 Sundean, of one -fourth in the same island, in 
 
 the neighbourhood of Batavia. 
 Dajak, in the centre and north of Borneo. 
 Bughi and Macassar, in the Celebes. 
 
 (/3) The Tag"ala group, to which are assigned — 
 
 Tagdla proper, in Luzon (the main island of the 
 
 Philippines) ; and other Philippine dialects. 
 Forviosan (though on the west of Formosa 
 
 Chinese is current). 
 Dialects of the Ladrone Islands. 
 Malagasi (or Hova) of Madagascar. 
 
 (ii.) The Melanesian branch, namely — 
 
 The various Fijian dialects (which are structurally 
 the nearest to the Malayan) and the speeches 
 of New Caledonia, Loyalty Islands, New 
 Hebrides, Solouwn Islands, New Britain, etc. 
 
 (iii.) The Polynesian branch, embracing the languages 
 of New Zealand {Maori), Tonga, Samoa,
 
 General Survey of Languages 143 
 
 TaJiiti, the Sandwich Islands {Hawaiian), the 
 Marquesas} etc. 
 
 Speaking generally, the Malayo - Polynesian family 
 possesses no formal distinction between a root and a 
 word. " The element which corresponds to the roots of 
 other languages appears regularly as a disyllabic, and, where 
 it occurs as a monosyllable, we are justified in believing it 
 to have arisen from a disyllabic by phonetic loss " (F. Miiller). 
 It can be noun, verb, adverb, etc., but is in itself undetermined. 
 Thus wahi ( = Tongan /«/zz = Samoan /rtj-/) may mean either 
 'to break' or 'a piece broken off.' The Fijian rcki = 
 'enjoyment' and also 'to enjoy oneself The M^X^y sakit 
 = ' ill,' ' illness,' ' be ill,' ' become ill.' Use may sometimes 
 bring it about that a particular root-word comes to be only 
 substantive, adjective, or verb, but the result is an accident 
 of habit and not the outcome of anything in the form of the 
 word itself The case may be illustrated, though in a very 
 small measure, from English, where ' love,' 'walk,' 'ride,' etc., 
 are either noun or verb according to position and context, 
 and where ' good ' is either noun or adjective. Much closer 
 is the parallel of Chinese (where the fact of the root-words 
 being monosyllabic is, of course, irrelevant to the principle 
 itself). 
 
 The means by which such material is adapted to the 
 expression of ideas in speech are position, the use of suffixes, 
 prefixes, and (in some of the languages of the Malayan 
 branch) infixes, the use of particles, repetition, and reduplica- 
 tion. The last two processes, though they are not unfamiliar 
 
 ■' According to Whitmee the interrelation of the chief tongues of this branch is 
 probably represented by the scheme 
 
 Marauesan Samoan
 
 144 Nahival History of Language 
 
 elsewhere, are particularly marked in this family. For 
 instance, in the Polynesian branch, Hawaiian /^/^// =' search,' 
 hulihuli = ' search through and through ' (an intensive). 
 Maori Jiacre — ' go,' Jiaerehaere = ' walk up and down ' (a 
 frequentative) ; Rarotongan nui = ' great,' numn = ' very 
 great.' In the Melanesian branch, Fijian /«/« = ' send,' 
 talatala = ' message ' (a frequentative). In the Malayan 
 division, Malay rdja = ' king,' rcijardja = ' kings ' ; Dayak 
 pukul = ' beat,' hapukupukul = ' beat often ' (frequentative). 
 
 Infixes are unknown outside the Malayan division. 
 Prefixes and suffixes differ greatly in the extent to which 
 they can be used in the three divisions respectively. In the 
 Polynesian languages sundry qualifications of sense can, 
 indeed, be expressed by particles loosely appended or 
 prefixed, but in such a way that they neither cause nor 
 experience any phonetic modification through their attach- 
 ment to this or that ' root '-material. These only vary the 
 contents of meaning of the root- word. They do not 
 determine its syntactical relations. Thus faka- (Tongan), 
 faa- (Samoan), zvaka- (Maori) is prefixed in a causative 
 sense, e.g. Maori kite = ' see,' wakakite = (' make to see ') = 
 'show.' Again, particles like ia (Jna,Jia) may be suffixed 
 to form what practically serves as a passive, e.g. Samoan sila 
 (' see '), si/a&a, (' to be seen '). But these in themselves no 
 more make a language ' agglutinative ' in the usual sense 
 than do such prefixes as the English de- in ' hespeak ' or the 
 German 2er-, er-, etc. Truly grammatical relations, such as 
 those of time, mood, person, case, number, are regularly 
 expressed in Polynesian by free particles. Thus in Maori 
 ki te tangata = ' to the man,' while the plural is ki nga tangata, 
 both case and number being expressed by a particle. In 
 the same language it is the free i which imparts a preterite 
 time to a verb, and the free kite which expresses the 
 conditional mood. 
 
 In the Melanesian division the Fijian shows such 
 prefixes as vaka- (the causative above mentioned), iidau- 
 (intensive), etc., and a large number of demonstrative or
 
 General Survey of Languages 145 
 
 'denominative' suffixes attached to verbs. In point of 
 syntactical relations, number, case, tense, and mood are ex- 
 pressed, as in Polynesian, by loose particles, but grammatical 
 agglutination appears in the suffixing of possessive pronouns, 
 e.g. linga = ' hand,' lingaxoM — ' thy hand,' /ingaua, — ' his hand.' 
 In the Malayan branch, as in the others, there are, 
 properly speaking, no plurals or case-forms of nouns, these 
 being determined . by prepositions or by position. Such 
 prepositions, however, are often more closely attached, so 
 that one may write, e.g., diruma/i, ' in the house.' The pos- 
 sessive pronouns are suffixed to the noun as in Melanesian. 
 The person of verbs (or what would be the person of verbs 
 in speeches which possess true verbs) is also expressed by 
 suffixes. The passive can be represented by prefix, suffix, or 
 infix. ' Denominative ' prefixes and suffixes are very 
 numerous. Specially remarkable, however, are the infixes 
 which appear in this branch, and which are used both as 
 ' denominatives ' and also to assist in creating forms which 
 may be regarded as distinguishing voice, mood, and tense. 
 Thus in Tagala, su/a^ = ' writing,' svnau/a^ = ' to write ' ; 
 Formosan sasi = ^ salt,' .nim^j'z = ' to salt'; Battak sasa = 
 ' wash,' sinasa = ' to be washed.' Still further, while Tagala 
 su/ai = ' writing ' and s\na.iclat = ' to write,' .ni-ng-m///^?/ {ng = 
 na, ' already ') is the preterite active, smu/a^am is the preterite 
 passive. The formula for a modified word in this curious 
 * incapsulating ' method is AxyxA. Perhaps it may be 
 allowable to compare with this manner of fusing various 
 elements into a phonetic amalgam the practice of Semitic 
 in its dynamic interior modification of its disyllabic roots. 
 
 (iv.) Other Families or Languages commonly classed as 
 Agglutinative {the agglutinative principle being often 
 impurely or only partially applied). 
 
 (a) The Papuan languages, in New Guinea and some 
 smaller islands, breaking the geographical continuity of the 
 Malayo-Polynesian family. Of these the dialect of the 
 
 (b 834) L
 
 1 46 Natural History of Language 
 
 Mafor of New Guinea seems to be practically the only one 
 tolerably known. In token of their agglutinative character 
 are quoted such forms as s?i2\n = ' man,' smhisi = ' men ' ; dien 
 = ' woman,' bien^x — ' women,' which show pluralising by 
 suffix. Declension is by prepositional prefix, e.g. xosn/ln = 
 ' of the man,' xos)i2hi&\ = ' of the men ' ; loesnun = ' to the man,' 
 hesnunsi — ' to the men.' In the verb, ]a,7;i;iaf (' I hear '), 
 ■wa,wna/ C thou hearest '), imnaf (' he hears'); and (with an 
 object) ]a,}una/'3iU (' I hear thee '), ^xninafi (' they hear him '), 
 illustrate a process which differs from complete analysis only 
 in the degree of separation of the elements. The formula 
 for these tongues is yAx. 
 
 (jS) The Australian languages (which, including the 
 extinct Tasmanian, appear to be derived from a common 
 source,^ so far as can be judged from their structure, and in 
 a certain degree from their vocabulary, particularly their 
 general agreement in the pronouns and the first three 
 numerals) are best known in the southern and south-western 
 portions of the continent. Their chief principle of structure 
 is suffix-agglutination, though they vary in the consistency 
 of its application. The method is clear in a declension like 
 that of punul (' son ') in the Lake Macquarie dialect, viz. 
 Nom. punuHo, Dat. punul)s.o, Gen. punul^sjo^z,, Abl. pzmultva., 
 or in the Kamilaroi mulion (' eagle '), Dat. muliongo, Abl. 
 mutioji6.\, Loc. inutiondidi,, etc. 
 
 (7) The Negro languages of Africa, extending across 
 the continent from the West Coast, north of the Bantu 
 family and south of the Semitic and Hamitic, admit of a 
 certain amount of grouping, but cannot be referred to a 
 common stock. Neither the criterion of vocabulary nor that 
 of grammar supports such a reference. Generally speaking, 
 these languages are agglutinative, though some of them have 
 attained rather the isolating character of Chinese. The 
 principal groups are — 
 
 {a) The Wo/of' dialects of Senegal and Senegambia 
 
 ^ A connection with the Dravidian tongues cannot be upheld. 
 ^ Wolof= ' the black,' while Fu/ie= ' the red.'
 
 General Stirvey of Languages 147 
 
 In these use is made of particles and prepositions 
 as well as of word-position for the expression of 
 syntactical relations. Thus fas u biir = ' horse of 
 king.' It is, of course, only at the point at which 
 the determining particles are actually tacked to 
 the main root and so create a kind of compound 
 that agglutination begins. This process is 
 common in Wolof. Hovelacque quotes inas-nd 
 = ' have-I,' while w«i'-£'«-;^rt = ' having-was-I,' and 
 vias-agu-nu-won-sopa-sopa-lu = ' have-not-yet-we- 
 (imperf sing.) loved.' Forms like these represent 
 agglutination in its rudest shape, viz. the simple 
 juxtaposition of particles, which become more or 
 less symbolic by use, and by pronunciation are 
 brought into an appreciable unity with the main 
 root. 
 
 {b) Mende, of Upper Guinea, including dialects of 
 Mandingo, Bambara, etc. 
 
 {c) Felup dialects, between the Wolof and the Mandingo, 
 on the Gambia, Geba, etc. 
 
 id) Kru dialects of the Grain Coast. 
 
 {e) The Eve {Ife) group, including Yaruba, Ashanti, etc., 
 on and behind the Slave, Gold, and Ivory Coasts. 
 
 (/) Ibo of the Niger Delta. 
 
 ig) Central African speeches, eg. the Songhrai (on the 
 Niger, south-east of Timbuctoo) and the Hausa 
 dialects. These speeches may have become more 
 isolating in structure than most of the Negro 
 tongues from the fact that they have come into 
 more frequent collision with alien influences. 
 Hausa, for example, is in wider use as a lingua 
 franca than any other language in Middle Africa. 
 
 {Ji) Bornu dialects about Lake Tchad. 
 
 (z) The Upper Nile group : Bari, Shilluk, Denka, etc. 
 
 (J) Unclassified dialects existing in Central Africa, 
 south and south-east of Lake Tchad {Bagkinni, 
 etc.).
 
 148 Natural History of Language 
 
 (8) The Nubian-Fula languages. These are distinctly 
 suffix-agglutinating. Without committing ourselves to any 
 opinion as to a genealogical relationship of the speeches 
 involved, we may divide into two groups : — 
 
 (a) The Fit/a (or P^//) dialects, extending from westward 
 of Lake Tchad, north of the Negro languages 
 just described, between 10" and 20° north latitude, 
 and penetrating between certain of those at the 
 Senegal coast. 
 
 (d) The Nubian group, including Nubian proper, 
 
 Dongolawi, Tuniali (of Kordofan), etc. 
 
 (e) Hottentot dialects : viz. Nama (of the Namaqua) to 
 the north-west, Khora (of the Khoraqua) to the north-east, 
 and Cape Hottentot. The last is practically dead. These 
 dialects have no connection whatever with either the Bantu 
 or the Negro speeches. Though generally given as suffix- 
 agglutinating (demonstrative elements and signs of gender, 
 number, and case being terminational affixes), it is difficult 
 to distinguish many Hottentot forms from cases of amalga- 
 ting inflexion or of analysis. 
 
 iX) Bushman dialects, scattered between the latitude 
 of Lake Ngami (20° S.) and the Orange River. These are 
 reported as tending from the suffix-agglutinating to the 
 isolating method. 
 
 (77) Japanese is apparently distinct from any branch of 
 the Ural-Altaic family. So far, again, as it seems to 
 correspond to Chinese in respect of any word-material the 
 phenomenon is simply the result of borrowing. Reduplica- 
 tion appears for the expression of plurality (e.g. janiajania, 
 ' mountains,' kuniguni, ' lands ') in the same way as in Malayo- 
 Polynesian, but there is otherwise no affinity with that 
 family. The regular grammatical method is that of suffixed 
 elements. Thus, with nouns, tsu or no indicate possession, 
 ve and /le direction ; and, if Vedove — ' to Yedo,' a process 
 of combination results in Yedoveno inissi = ' road (of) towards 
 Yedo.' According to the proper agglutinative method, these 
 particles are affixed to singular and plural alike. Thus
 
 General Survey of Languages 149 
 
 fitono, ' of a person,' is equal to a genitive singular, while its 
 plural is Jitoditono ; the locative singular is Ji toni, the plural 
 fitobitorA. The affixes are, however, so far free that some 
 grammarians write them separately. Verbal meanings are 
 more closely defined by ' demonstrative ' suffixes, e.g. nuku = 
 ' warm,' nukuxai = ' become warm,' 7iukii\sx\xv\ = ' make warm '; 
 mi = ' see,' ;;//se = ' make see ' (' show ') ; ;;/?ru = ' to see,' 
 mizaxi = * not see.' Nevertheless, that the language is not 
 strictly agglutinative in the accepted sense of that term has 
 been sufficiently shown already, and in the spoken language 
 the phonetic obscuration of the constituent elements, particu- 
 larly of verbs, is carried to a considerable extent. 
 
 Japanese possesses a literature dating from A.D. 712 in 
 the shape of the Kosiki (the sacred book of Shintoism) and 
 of the national annals. 
 
 {&) Corean is a suffix-agglutinating tongue. Thus from 
 ka-ta (' go ') are formed ka\ta (preterite), kd)LQ\\ta (future), 
 katkeiUa (fut. preterite), /cateni (imperfect), /catteni (plu- 
 perfect). The official speech is Chinese, and a large number 
 of Chinese words have found their way into Corean proper, 
 adapting themselves, however, to the Corean structural 
 method. 
 
 (i) The Hyperborean speeches of Asia, some of which 
 may or may not form a family, include the languages of the 
 Ainus, Ghiliak (of North Saghalien and on the neighbouring 
 coast of the mainland), Kanitchatdale, Koriak, YukagJiir, 
 Tchukchi, and Yenissei-Ostiak (a tongue to be distinguished 
 from the Ural-Altaic Ostiak, with which its agrees neither 
 in its roots nor in the principle of vowel-harmony). The 
 true agglutinative method (with formula Ay for singular, 
 Axy for plural) appears in such a declension as the Yenissei- 
 Ostiak tdfa (' stick '), tdfa^&Q,g (dative), tafaA&ng^x (abl.), 
 td/agei (loc), with corresponding plural td/dn, ^d/dndeng, 
 etc. 
 
 («) The Caucasian speeches fall into two groups — 
 
 (i) North Caucasian (sometimes improperly styled 
 Circassian) with the branches Circassian (west-
 
 1 50 Natural History of Language 
 
 ward to the Straits of Kertch), Kistian (central), 
 Lesghian (of Daghestan on the Caspian), and very 
 numerous sub-dialects. The name of the group 
 is geographical only. There is no sufficient proof 
 that the several languages are derivable from a 
 common parent, although they have many points 
 of resemblance.^ 
 (2) South Caucasian (also called Kartvelian), including 
 Georgian, Snantati, Mitigrelimi, and other dialects. 
 Relationship in this case is distinctly recognisable. 
 The agglutinative principle prevails in all these 
 tongues, but the result often becomes indistinguish- 
 able from amalgamating flexion. Some writers, 
 therefore, speak of them as ' inflexional,' even in 
 the narrower sense of that expression. To take 
 a comparatively mild instance. In Georgian, the 
 present of the verb 'to be ' is conjugated vjar, 
 char, ars, warth, charth., arism, and though ar is 
 no doubt retained throughout, it requires an effort 
 to realise the fact that it is the root. In Khasi- 
 kumuk, a dialect of Lesghian, az is the base or 
 theme of the verb ' to make ' ; but forms like dra, 
 iira, aisara, fiiida, dnda, a, etc., make up its 
 paradigm. Both suffixing and prefixing are 
 employed. The objective pronoun is taken up 
 into the verb as in Basque, and to that extent the 
 languages may be treated as incorporating. 
 (X) Basque is no true agglutinative tongue according to 
 the customary use of the term. Thus Van Eys gives ediiki 
 as the ' verb ' meaning ' to hold.' Of this the ' verbal stem ' 
 is often simply en. Again, for ^.ediiki (' I have it '), dadukat 
 is phonetically substituted, and this, once more, is replaced 
 by daiit or dut. Such an instance shows that it is impossible 
 to speak, in regard to Basque, of a main ' root ' preserved in 
 
 ^ It is highly probable that the Caucasus district has been the refuge of a 
 number of different stocks successively forced into the hills by pressure from both 
 north and south. 
 
 II
 
 General Survey of Languages 1 5 1 
 
 an invariable form throughout a paradigm. It may be true 
 that a Basque verb is made up of combined elements. But 
 so is a Greek or Sanskrit verb. And in Basque, as in 
 Greek or Sanskrit, the elements may evidently obscure each 
 other to a considerable extent. Phonetic obscuration of 
 the root is not allowable in the ideal agglutinative system, 
 yet in Basque it occurs in the instances above given, as, 
 further, in such a formation as neban, ' I had,' beside the 
 theme eu. 
 
 The distinguishing mark of the language is its com- 
 pulsory welding or incorporation into the verb of the 
 pronouns of the subject, direct object, and indirect object. 
 The ' verb ' is a stem or theme plus signs of the personal 
 pronoun, and a transitive verb without an object incorporated 
 is impossible. The subject pronoun is also attached, its 
 position varying from prefix to suffix. Thus, in an intransi- 
 tive verb, n-a-bil=' I go' {a being euphonic), h-a-bil=' thow 
 goest,' d-a-bil='he goes.' With indirect oh]ec\., h-a-bil-ki-t 
 = ' thou goest to me.' In a transitive verb, n-a-kar-su 
 literally = ' me - carriest - thou,' h-a- kar-t = ' thee - carry - I,' 
 d-a-kar-t='\\\m (it)-carry-I,' and, with indirect object, 
 d-a-kar-su-t ' it-carry-to-thee-I.' These particular examples 
 are ' agglutinating ' enough, the formula being j-Rn; but such 
 distinctness of method is not consistently apparent. To a 
 certain extent Basque recalls the polysynthesis of the 
 American tongues, but only in respect of the pronouns. 
 To quote the French je [aime {jTaime) or il te le dit (vul- 
 garly it'Tdit) is merely to show that a close union, affecting 
 both word-order and phonetics, between verb and pronoun 
 is more or less instinctive. In Basque, however, the verb 
 has no existence at all without the pronoun. 
 
 The Basque dialects (properly called Escuara or Euscara, 
 but more familiarly Euskarian) are spoken by nearly half 
 a million people about the Western Pyrenees, in the French 
 arrondissemcnts of Bayonne, Oloron, and Mauleon, and on 
 the Spanish side between Pampeluna and Bilbao. They are 
 very numerous, including Labourdin, Souletin, Bas-Navarrais,
 
 1 5 2 Natural History of Language 
 
 Haut-Navarrais, Biscayan, etc. Unfortunately for a historical 
 examination of the Basque structure or ' roots,' it is not 
 possible to go back by the help of any literature further than 
 the first half of the sixteenth century. The earliest printed 
 work is a collection of poetry of the year 1545, and there is 
 no supply of manuscripts of an earlier date. The intrinsic 
 difficulty of the Basque language was much exaggerated by 
 the earlier French and Spanish investigators, who were looking 
 for the same principles of grammar which prevailed in their 
 own tongues and were perplexed by inevitable failure. 
 
 (yti) The ancient Accadian, Shumerian or Proto- 
 Chaldean, if correctly deciphered from the cuneiforms, is as 
 agglutinative as the Ural-Altaic type, with the power of 
 incorporating objective pronouns. Professor Sayce gives 
 garmn, ' I made,' ^<^;Tiiiimu, ' I made it,' garda,nmvi, ' I caused 
 to make,' gardajummu, ' I did not cause to make.' 
 
 Accadian reveals no distinct affinities with any known 
 family of language. 
 
 (v.) T/ie PolysytitJietic or HolopJirastic Languages of America 
 
 The native American tongues, which are reckoned at 
 over four hundred in number, though they mostly (but by 
 no means universally) agree in the principle of polysynthesis 
 (varying greatly, however, in the extent of its practice), 
 cannot be reduced to less than about thirty groups. Many 
 of them, particularly in South America, are too little 
 ascertained for a theory to be advanced as to their inter- 
 relationships. 
 
 Speaking generally, their distinguishing mark is their 
 power of taking up and incorporating with the verb the 
 subject, object, and elements expressive of time, place, and 
 kind of action (such as negative, causative, reflexive, and 
 other demonstrative tokens), till it often happens that a 
 whole sentence is expressed in one compound word, some- 
 times of formidable dimensions. As Duponceau well states 
 the case, they show a " mode of compounding which is not
 
 General Siirvey of L angtiages 1 5 3 
 
 confined to joining two words together, as in Greek, or 
 varying the inflection or termination of a radical word, as in 
 most European languages, but which interweaves the most 
 significant sounds or syllables of each simple word, so as to 
 form a compound that will awaken in the mind at once all 
 the ideas singly expressed by the words from which they 
 are taken." 
 
 Thus the parts of a simple sentence which in an ordinary 
 analytical tongue of Europe would be represented by a 
 pronoun or noun subject or object, a possessive pronoun, an 
 adverb, and a verb, may, in the more consistently poly- 
 synthetic of the American speeches, be fused into a mass which 
 is as distinctly a single word as the English indefatigableness 
 or the Greek KaTaireiroX.eiJb'qa-ofieOa {katapepolcmesovietJia, ' we 
 shall have been warred-down '). Whereas, however, the 
 long English word in this case is but an abstract noun 
 expressive of a quality (which might conceivably have been 
 equally well expressed by a monosyllable), the polysynthetic 
 word may involve what we should express by several distinct 
 parts of speech. The inflected Greek word, indeed, goes 
 further than the analytical English. It expresses by root 
 and flexion a verb and its subject compounded with a 
 determinant of manner (' down '), and qualified in point of 
 time and voice. Yet the Greek could never incorporate an 
 object, whether pronoun or noun, and the number of 
 determinants is small. It is no characteristic of Greek to 
 weld into its verbs such a complete series of adverbial 
 elements as occur in Greenlandish, for example, where such 
 notions as ' almost,' ' quickly,' ' probably,' ' willingly ' are 
 habitually included in the verbal expression itself From 
 Cherokee we find quoted nadJiolinin, ' bring us the canoe,' 
 conflated from naten, ' bring,' avwkkol, ' canoe,' m'n, ' to us.' 
 Whatever power of compounding elements may have existed 
 in an Indo-European tongue (e.g. in Greek or Old Teutonic), 
 there is no trace whatever of a power to incorporate either a 
 direct or an indirect object. It would only be when we 
 could find e.g. (fjepe ciKarov tj/jLcv fused into some such
 
 154 Natural History of Language 
 
 shape as ^paKadi^ynv that we could fairly compare Greek 
 compounds with the American holophrastic word. In the 
 Indian languages, as in the instance quoted, it is often 
 sufficient if a fraction, sometimes a mere letter, of this or 
 that element remains clear in the composite mass. The 
 portion so remaining must, however, be the most significant. 
 The German words beini ( = bei devi), ziir { = zu der) ; the 
 English sharit ( = shall not), and the vulgar French pro- 
 nunciation kekcekcal (= qiiest-ce que c'est que ^a?) are 
 rare instances in speeches nearer home of a tendency to a 
 similar phonetic abbreviation checked by the same con- 
 sideration. In the American the abridgment of a given 
 element, and its phonetic adaptation to its surroundings, 
 may take place at the beginning, in the middle, or at 
 the end. 
 
 It is easy, however, to overstate the extent to which 
 polysynthetism is carried. It is naturally interesting to 
 quote a word from the Indian of Massachusetts containing 
 eleven syllables, and meaning ' he fell down on his knees 
 and besought him,' or from Mexican a word of twenty-seven 
 letters denoting a priest ; but such citations, which are apt to 
 leave a false impression as to the average length of an 
 American polysyllable, should be treated rather as curiosities. 
 It is more pertinent to our subject to illustrate the principle 
 itself, as shown in the Cherokee kukfisquo, ' I wash my face,' 
 takuteyd, ' I wash dishes,' or in the Mexican nikalcliihua (lit. 
 ' I-house-build '), ' I build a house,' nisotsitemoa (lit. ' I- 
 flowers-seek '), ' I look for flowers,' Jiinomdpopowa (lit. ' I-my- 
 hands-wash '). 
 
 Again, though it is true that incorporation on this large 
 scale is a distinguishing peculiarity shared by nearly all the 
 American tongues and preserved in fair consistency by the 
 North American Indians, it would be an error to suppose 
 that the Greenland language on the one hand, or the 
 Mexican on the other, necessarily does always incorporate 
 the elements which it may incorporate, and that therefore a 
 noun, for instance, can possess no separate existence. The 
 
 II
 
 General Survey of Languages 155 
 
 Greenland speech can on occasion substitute for a long 
 compound individualised words with suffixes of relation. 
 Thus the termination -/ in akhfekhup (' whale ') permits of 
 that word being used as the subject to a sentence. Likewise 
 in Mexican, although, as has been said, ' I seek flowers ' is 
 according to the polysynthetic method expressed by 
 nisotsitcuwa, nevertheless the element sotsi- (' flower ') is 
 capable of a separate existence when accompanied by the 
 termination -//. The answer to the question ' what do you 
 seek ? ' may be given as sotsiH. Similarly, though in the 
 compounds the first personal pronoun singular is expressed 
 by ni (e.g. 7iitcnioa, ' I seek '), the answer to the question 
 ' who seeks flowers ? ' may be given as neivdW, ' I.' Again, 
 beside the noun-incorporating composite ninakakwa, ' I -flesh- 
 eat,' Mexican employs the device ni-']s.-kzva in nakatl, i.e. 
 ' I-it-eat the meat,' where an anticipatory appositive pronoun 
 k is incorporated, while the noun is left free. So 7n-)s.-inati 
 in tedtlatolli, ' I-it-know, the God's word.' The two structures 
 are combined in such an expression as ni'ksdtsitcinoa kzvikatl 
 (I-them-flo\ver-seek songs,' i.e. ' I seek songs like flowers ' (' I 
 cull songs '), a turn not unlike dvdoXoyo) aa/iara. 
 
 If the term 'agglutinative' were applied indiscriminately 
 to any language in which element is tacked on to element 
 (each such element in some way modifying the sense), no 
 matter what the phonetic results may be or what changes 
 the several ' roots ' may undergo in the process, the American 
 tongues might obviously bear the title. The term, however, 
 is not so employed. If we are to separate amalgamating- 
 inflexional from agglutinative on the ground, not of a 
 fundamental principle, but of a difference in its obviousness, 
 extent, or results, we are bound to distinguish polysynthesis 
 equally from either. Unfortunately information concerning 
 the American languages appears to be too scanty or 
 inaccurate or ill-digested to permit of the philologist 
 investigating in a satisfactory manner the exact degrees in 
 which polysynthesis is varied or broken down throughout 
 the two continents. Neither is an accurate grouping and
 
 156 Natural History of Language 
 
 subdivision of the tongues obtainable. So far as attempts 
 have yet gone, the chief classifications (some of which are 
 very roughly made) are as follows : — 
 
 1. The Eskimo (Innuit) speeches, spreading in various 
 
 dialects from Greenland and Labrador along the 
 north coast of North America and reaching into 
 North-East Asia. 
 
 2. The Athabascan group, extending over the north- 
 
 west of the Canadian Dominion from Alaska to 
 Hudson's Bay, and including dialects detached 
 from the main group farther south in the United 
 States, e.g. the Navajo and the Apache (of Nevada, 
 Arizona, and Upper California). 
 
 3. {a) The Algonkin languages, lying partly west, but 
 
 mostly south and south-east, of Hudson's 
 Bay, and including dialects of tribes in the 
 east of the Dominion of Canada and extend- 
 ing into the United States as far as North 
 Carolina. The chief divisions are the Cree^ 
 Mihnak, Ojibway {Chippewa)^, MoJiican (of 
 Connecticut), Abenaki (of Massachusetts), and 
 Delaivare. 
 {U) The speeches of the Iroquois group are sometimes 
 classed with the above and sometimes not. 
 They include the Mohaivk, Oneida, Huron, 
 Tiiscarora, etc., spoken in parts about the St. 
 Lawrence Lake, in the State of New York, and 
 in reservations westward of the Mississippi. 
 
 4. The Dakota group, of the Sioux, etc. 
 
 5. The Pawnee (in Arkansas). 
 
 6. The Appalachian group (of Florida and south-east), 
 
 e.g. Cherokee, Choctaw, Muskogee. 
 
 7. 8, 9. Oreg-onese, Californian, and Yuma (Colorado) 
 
 groups. 
 
 10. Koloche dialects of British Columbia and neighbour- 
 hood. 
 
 I I. Various speeches of Pueblo Indians and of Texas.
 
 General Survey of Languages 1 5 7 
 
 1 2. The Mexican group — 
 
 {a) Shoshone speeches {Comanche, Utah, etc.). 
 
 ib) Sonora speeches (North-West Mexico, South 
 
 Arizona, etc.). 
 {c) Nahuatl (the Mexican of the Montezumas) and 
 
 its derivative the modern Aztec, with which 
 
 are connected many dialects. 
 
 1 3. Other Mexican idioms, independent of the pre- 
 
 ceding and probably the languages of earlier 
 inhabitants of Mexico, e.g. OtJiomi, Mixtek, 
 Tarasca, etc. 
 
 1 4. Maya of Yucatan (with Quiche, Huasteca, etc.). 
 
 15. Unclassified Central American dialects, e.g. the 
 
 Cncva group (Panama). 
 
 16. The Cibuney group of the Antilles or islands of 
 
 the Caribbean Sea. 
 
 17. The Carib (or Galibi), from Panama eastward, and, 
 
 with Araivak, reaching to the Guianas. 
 
 1 8. The Tupi-Guarani group, southward from Guiana 
 
 through Central Brazil to Paraguay and 
 La Plata. 
 
 19. Chibcha in Colombia. 
 
 20. Various languages of the Andes. 
 
 2 1. The Peruvian group {QuicJuia, Ayviara, etc.). 
 
 22. Araucanian dialects of Chili.' 
 
 23. The Guaycuru-Abipone dialects of West Paraguay 
 
 and North Argentina. 
 
 24. Dialects of the Pampas {Picelche), Patagonia 
 
 {Teh2ielche), Tierra del Fuego, etc.
 
 CHAPTER IX 
 
 GENERAL SURVEY OF LANGUAGES {continued') 
 
 III. The Amalgamating-Inflexional Tongues 
 
 (COMMONLY CALLED SIMPLY INFLEXIONAL) 
 
 A. The Seniitico-Hamitic Family 
 
 A VERY important list of languages of South- West Asia and 
 of North and North-East Africa, living and extinct, is known 
 by the name of the Semitico-Hamitic family. In it are 
 included two branches widely divergent from each other. 
 ' Semitic ' and ' Hamitic ' are names which hardly pretend 
 to a more philosophic raison d'etre than that of convenience. 
 Shem (or Sem), the son of Noah, is represented in the 
 Mosaic tradition (Gen. x.) as the ancestor of most of the 
 peoples of South-West Asia with whom the Hebrews felt 
 themselves akin, e.g. the Assyrians, Syrians, and Arabians. 
 Ham, his brother, is the progenitor of the Egyptians and 
 other Africans, the Ethiopians, and the Canaanites (includ- 
 ing the Phoenicians). It will be seen, when we come to 
 arrange the various tongues in detail, that this division of 
 peoples is far from corresponding to a proper division of 
 languages. Thanks, however, to notions based largely on 
 geography and colour rather than upon the exact Biblical 
 text, in the popular conception Ham has come to be re- 
 garded as the father of Africans, and Shem as the father of 
 the Hebrews and the peoples akin to them in Asia. Even 
 
 15S
 
 General Survey of Languages 159 
 
 in this acceptation the Hnguistic application of the terms, if 
 pressed, would be incorrect, inasmuch as several of the 
 languages of Africa which are now to be treated belong, not 
 to the Hamitic branch, but to the other. Nevertheless it 
 would be a needless purism, if not pedantry, to object in the 
 study of language to a convenient term which has acquired 
 a special meaning for that science. 
 
 Roughly speaking, the Semitic group includes the 
 ancient languages of Chaldea, Assyria, Syria, Phoenicia, 
 Palestine, and Arabia, the modern tongues which are derived 
 therefrom, and those which were once carried from them 
 across the Red Sea, or to Carthage, or elsewhere. The 
 Hamitic embraces Ancient Egyptian, Libyan and Ethiopian 
 with their later derivatives. 
 
 The genealogical relationship of the Hamitic to the 
 Semitic tongues is maintained by no few of those most 
 competent to deal with the question. Fr. Muller, for 
 example, speaks of them as ' indubitably ' forming a 
 linguistic unity. It is true that the characteristic formative 
 features of the Semitic languages (such as the tri-consonantal 
 roots and the dynamic vowel-variation) are conspicuously 
 absent from the Hamitic ; but the fact may be accounted for 
 by assuming a very long separation of the two branches, 
 during which the Hamitic speeches at least were probabh' 
 subjected to those alien influences which have already been 
 spoken of as tending to break down the grammatical structure 
 of a language. The extent to which divergence of structure 
 may be carried within the same family has already been 
 seen in the case of the Polynesian tongues as compared with 
 the Malayan. It appears also very appreciably in the 
 difference between such Indo-European languages as modern 
 English and ancient Greek. At the same time there is a 
 very distinct resemblance in the material of the pronouns in 
 the Semitic and Hamitic groups ; the plural suffixes of their 
 nouns appear to be of common origin ; and the use of the 
 affix t to denote the feminine is characteristic of both 
 divisions. It is hardly credible that such resemblances
 
 i6o Natural History of Language 
 
 should be accidental, nor is it in the least likely that the 
 Hamitic borrowed these particular portions of its vocabulary 
 and grammar from the Semitic, or vice versa. Even if we 
 admitted that special grammatical usages might be borrowed, 
 they would certainly not be borrowed without being accom- 
 panied by still larger borrowings of vocabulary. The as- 
 sumption that one whole body of language could be so 
 powerfully affected by another as to give up its own 
 pronouns, which are among the most pertinacious (because 
 among the most universally and perpetually used) elements 
 of a primitive speech, is equivalent to supposing such a 
 penetrating and sustained influence of one language upon 
 the other that it would be impossible for the general 
 vocabulary of the latter not to have been affected in a 
 profound degree. On the other hand, given two languages 
 which start from the same point, provided with the same 
 material both of roots and formatives, and the same methods 
 (perhaps, indeed, rudimentary) of combining such material, 
 it is a matter of experience that, after separation, they will 
 pursue dissimilar developments in both respects. If, then, 
 the interval of separation is made sufficiently long,^ and if 
 alien influences strongly attack one branch or the other (or 
 both), it may well result that little trace of the original unity 
 is eventually discoverable. On the other hand, there may 
 remain some few vestiges of the common inheritance, and it 
 will depend upon the nature of these, rather than upon their 
 number, whether the philologist will argue for a probable 
 relationship or not. In the case of the Semitic and Hamitic 
 tongues the traces above-mentioned, though few, are unusually 
 significant. 
 
 (i.) THE SEMITIC DIVISION 
 
 The most prominent characteristic of the Semitic 
 languages lies in the fact that the ' root,' in the usual 
 
 1 In the present instance the interval must necessarily be reckoned by 
 millennia, in view of the exceedingly remote date to which Egj-ptian records go 
 back. That date is itself obviously far removed from the original establishment 
 of an Egyptian language.
 
 Gene7^al Szirvey of Languages i6i 
 
 acceptation of the word, regularly consists of a group of 
 three consonants, which continue constant elements in all 
 the series of words deducible from that root. These three 
 consonants are in themselves the representatives of the 
 general or undetermined notion. To give to such a tri- 
 consonantal root an actuality as a word or word-stem, it is 
 of course necessary that the consonants should be uttered 
 along with a sonant or more than one sonant. It is a salient 
 peculiarity of the Semitic tongues that on the particular 
 vowel-sounds which are so employed depends the gram- 
 matical or other determination of the word. The consonants 
 are, theoretically at least, unchangeable, while the vowels vary 
 in strength, position, and number ; and each such variation 
 denotes a corresponding variation in meaning or syntactical 
 relation. As Misteli expresses it, " the purely lexical element 
 in the Semitic speech lies in the consonants, the grammatical 
 in the vowel." The vowels thus play in Semitic a part which, 
 in the Indo-European languages, is almost entirely played 
 by terminational flexion or by composition. In other words, 
 Semitic flexion is mainly internal. It is, however, by no 
 means wholly so, since flexion occurs also (as will be seen 
 immediately) in the shape of suffix and prefix. 
 
 The Arabic ' root ' involving the general notion of 
 ' killing ' was a conception as yet undetermined as verb or 
 noun, and devoid of any implication of any relation whatever 
 to other members of a sentence — in fact, an abstraction is 
 given in the form qtl. Similarly, the combination ktb 
 contains the general notion of ' writing,' drb of ' striking,' dbr 
 of ' speaking,' vgd of ' finding,' and so forth. The deter- 
 mination of these roots to special meanings and functions in 
 a sentence begins with the presence of vocalic elements, and 
 as has been stated above, the meaning and function will 
 depend upon the kind and position of the vowels employed. 
 Thus, while ^a/a/a = ' he killed,' ^u/i/a = ' he was killed,' {ya)- 
 gtuln = ' he kills,' ga.t\l= ' killing,' qit/= ' enemy,' qUal= ' a 
 blow,' ^a/a/a = ' seek to slay.' So while (tayc/uba, = ' thou 
 writest,' kd.tz.b{td) = ' thou hast written.' 
 
 (b S84) jyj
 
 1 62 Natural History of Language 
 
 It might seem natural to compare with these vowel- 
 variations the changes which go by the name of Ablaut in 
 Indo-European. Thus, dfink, dr&nk, dinank ; frQQse, froze ; 
 fodio, fodi ; fides,f\do,foQdus ; Xcirelv, Xeiira), XeXocTra, might 
 perhaps be adduced as showing that the Semitic process is 
 not unknown to the Indo-European tongues. It must, 
 however, be observed, first, that such variations in Indo- 
 European are but occasional phenomena, and of an 
 incomparably narrower range in each case ; second, that 
 they altogether lack definiteness of system ; and third, that 
 they originally possessed no grammatical or determining 
 function whatever, but were simply the phonetic result of a 
 shifting of the position of the tonic accent, due to the number 
 or accentual nature of the affixed elements. In Greek it 
 was still the external flexion and not the root-vowel which 
 determined the tense, and if in English the change from 
 ' smg ' to ' sang ' has come to correspond with a change of 
 tense and to serve as its sign, that condition has been 
 brought about by phonetic accident and not by formative 
 principle. It is, of course, a possible theory that even in 
 Semitic the vocalic variation may be ultimately derived 
 from differences of stress and pitch, which originated in the 
 special psychological conditions appertaining to special 
 conceptions. Yet even in that case the Umlaut would still 
 have been at all times formative, whereas in the oldest 
 ascertainable Indo-European no such ' dynamic ' intention 
 is to be discovered. 
 
 It has been said above that the root-elements of Semitic 
 words are regularly tri-consonantal. Nevertheless they are 
 far from invariably so. In many instances it is true that 
 roots which have come to contain but two consonants can 
 be shown to be the results of phonetic condensation or 
 corruption from a radical form with three consonants. Thus 
 in the Arabic qn/ (' speak,' imperative), ^a/a, ' he spoke,' the 
 root is qvt ; and analogously /cit, k&ta,, belong to kj7. In 
 other cases the root is not forthcoming, and not a few 
 bi-consonantal stems, eg. Arabic ai?v- (' father '), t?an- (' son '),
 
 General Survey of Languages 163 
 
 or Hebrew jdd (' hand '), remain without any plausible 
 tri-consonantal etymology. That the comparative grammar 
 of the family tends to bring us back more consistently to 
 tri-consonantal roots in the Semitic Ursprache (or common 
 parent) is so far significant, but it is scarcely scientific to 
 assume that the shorter forms are in all cases derived from 
 longer ones, and to ignore any a priori probability that longer 
 forms may themselves have been constructed out of shorter. 
 It is quite conceivable that both processes have occurred in 
 succession. Bi-consonantal roots may have been amplified 
 by further elements, and these compounds again weathered 
 down by phonetic loss. Whatever may be the truth in 
 regard to these obscure and perhaps unsolvable questions, the 
 fact remains that, taking Semitic both as it is and as far back 
 as it can be hypothetically reconstructed, the tri-consonantal 
 root is an unmistakable characteristic of the family. 
 
 Vowel-variation, however, is not the only device by 
 which modifications of sense are expressed or derivatives 
 formed. Semitic employs particles, position, prefixes, and 
 suffixes in manners analogous to those of other families of 
 language. Composition, in the usual sense of the term 
 (according to which it means the attaching of two or more 
 full stems together), does not exist, but numerous derivatives - 
 may be formed from a root by attaching denominative 
 elements, chiefly as prefixes, but also, to a very small extent, 
 as suffixes. Thus such notions as the causative, reflexive, 
 etc., are expressed by formations from the root qtl in the 
 Hebrew shape \i\qtil, hithqatuV, xaqtal, etc. Arabic may 
 possess as many as fifteen such ' determined ' or specialised 
 conjugations for each verb, and often approaches that 
 number in actual use. 
 
 Apparent compounds, mostly proper names, are often 
 simple juxtapositions of two nouns, one being ' governed ' 
 by the other (or standing to it in the relation of a genitive), 
 e.g. Beth-Sheniesh, Beer-Shelm, Ben-Jauiin} No doubt the 
 
 ^ It may be noted in passing that the Semitic order in Ben-/a)nin, Melekh 
 Israel, etc., is the contrary of that which is usual in Indo-European compounds.
 
 1 64 Natural History of Language 
 
 line between such combinations — in which a certain unity is 
 marked by the closeness with which the two constituents 
 are uttered together — and the simpler forms of true com- 
 position, is slight ; nevertheless it is real. While Indo- 
 European tongues can form such compounds as philopro- 
 genitiveness or Unbcredtsainkeit, it is contrary to Semitic 
 principles to go further than attaching a prefix or a suffix 
 directly to the vocalised root. The accumulation of affix 
 upon affix seen in the Ural-Altaic and Indo-European 
 families is quite unknown. 
 
 The external inflexion of a verb consists in affixing to 
 the noun or adjective stem certain elements which may 
 be regarded as personal pronouns. These do not always 
 occupy the same position. There being but two so-called 
 tenses in the Semitic — the one denoting ' completed ' and the 
 other ' uncompleted ' action — the pronominal elements are 
 prefixed in the incomplete (imperfect-durative) and suffixed 
 in the completed (perfect-aorist) tense. Thus qataldX, ' she 
 killed,' but Xdqtidu, ' she kills ' ; qatalnz,, ' we killed,' but 
 ns^qtulii, ' we kill.' In the second and third persons the sex 
 of the subject is distinguished as masculine or feminine. 
 Thus qatala, ' he killed,' qataldX, ' she killed ' ; ya^tuht, ' he 
 kills,' ta.qtulu, ' she kills,' kaladta,, ' thou (masc.) writest,' but 
 katabXi, ' thou (fem.) writest.' 
 
 In noun-flexion the primitive Semitic employed suffixes 
 to form the singular, dual, and plural numbers, as well as the 
 three cases, nominative, genitive, and accusative (those terms, 
 however, being scarcely applicable in the senses which they 
 bear in Indo-European). In classical Arabic the declension 
 is maintained, as in abd^, abdi, abdd^, a system analogous 
 to the Latin serwiQ, servi, ser^vum (although, of course, no 
 genealogical relationship is suggested between the two sets 
 of terminations). Elsewhere, as the Semitic languages tend 
 to become analytical, the systematic declension disappears, 
 
 The English (and German) arrsingement yacod-son would be in Semitic son Jacob. 
 Beth-Shemesh (lit. ' house sun ') would represent a Greek 'HXfou o?)cos (cf. 'HXt'ou 
 woXij, 'sun's city,') and Samuel shows the reverse order to 9eatT7;ros.
 
 General Stirvey of Languages 1 65 
 
 and in Hebrew, for example, is replaced by the use of pre- 
 positional particles. 
 
 The possessive pronouns are suffixed to their nouns, as 
 in Ell, ' my God,' Arab. katdb\ ' my book,' and the objective 
 pronoun to the verb, e.g. SabachthdsoX, ' thou hast forsaken 
 me.' The article is prefixed in the Arabic form al, Hebrew 
 hal, the final / in each case being brought into phonetic 
 accord with, or assimilated by, the following consonant, e.g. 
 Hebr. \\axciinelekh, ' the king,' Arab, abira/undn, ' the merciful.' 
 In Aramaic the function of the definite article is performed 
 by a suffix -a, e.g. malkd, ' the king.' 
 
 The distinction of feminine from masculine was in all 
 the Semitic tongues expressed by the suffix -/ (or -ai), as in 
 the Assyrian nialak (' king '), malkdX (' queen '), Arabic ibn- 
 (' son '), bin\, (' daughter '). Phonetic departures have occurred 
 in the later separate developments, Hebrew having substi- 
 tuted -th and -/?, neo-Arabic in some cases -//, etc., but the 
 original form is undisputed. 
 
 It would be disproportionate to give' further details of 
 Semitic grammar. The salient facts to be gathered are 
 (i) that these speeches supply us with no new structural 
 phenomena beyond the systematic use of a dynamic vowel- 
 variation ; their prefixes and suffixes in no way differ in 
 principle from those of Bantu, or Malay, or Indo-European ; 
 (2) that, like the modern Indo-European, the Malayo- 
 Polynesian, and other tongues, they have shown a marked 
 progress towards analysis. 
 
 From the genealogical point of view the connection 
 between the several speeches of the Semitic family is rather 
 closer than that existing between the languages even of a 
 single branch of the Indo-European. The Romance or the 
 Slavonic tongues are felt at once to diverge more widely 
 among themselves than do Assyrian, Arabic, and Hebrew. 
 These stand to each other rather in the relation of accentu- 
 ated dialects, much as Lowland Scotch stands to literary 
 English. The general agreement in phonetics and material, 
 both substantive and formative, as well as in the grammatical
 
 1 66 Natural History of Language 
 
 outfit, is very striking. In some measure this comparative 
 homogeneity is due, on the one hand, to the characteristic 
 inabihty of Semitic to form new compounds, and, on the 
 other, to the simplicity of the primitive Semitic grammar. 
 Thus each of the separate dialects was restricted in the 
 creation of new words unknown to the rest, and at the same 
 time there were fewer grammatical forms from which the 
 individual idioms could deviate in various directions. Given 
 a profusion of grammatical forms in an Ursprache, one 
 derived speech will abandon some of these and generalise 
 others ; a second may abandon some of those which the first 
 has generalised, and generalise some which it has abandoned. 
 Primitive Semitic afforded comparatively little scope for these 
 rapidly disintegrating processes. 
 
 The recognised divisions of the family, dead and living, 
 may perhaps be seen best in the following table. The 
 languages which are wholly extinct are marked with an 
 asterisk, while those in italics are presumed stages without 
 actual historical records. 
 
 I
 
 General Survey of Languages 
 
 167 
 
 
 S 
 
 u 
 
 < 
 < 
 
 
 K.y 
 
 
 — ^ 
 
 s 
 
 'W .2 
 
 
 t/3 
 
 
 — ^ Oh 
 
 6 -^ CQ W ' < 
 
 2 cih 
 * 
 
 PQ 
 
 Zl 
 
 < § 
 
 "Pi '>> 
 
 ^ -2 
 
 en ^ 
 
 rt 
 
 - a 
 
 
 ^ 
 
 C/2 "^ 
 
 u 
 
 < Q h 
 U J < 
 
 -pq K ■> 
 
 -H u «
 
 1 68 Natural History of Language 
 
 Upon this tabulation the following notes may be made : — 
 
 Aramaic. — The examination of the cuneiforms and of 
 other data leads to the belief that Aramaic was originally the 
 language of the Semitic highlanders to the north of Meso- 
 potamia. From the important centre Carchemish (Jerablus) 
 it spread, as the language of commerce, over Syria, Meso- 
 potamia, and Chaldaea, until it became by about 800 B.C. 
 the prevailing (though not the only) tongue in those regions. 
 Therefore the Chaldee current from that date in Babylon is 
 different from the earlier Chaldaic of the Assyro-Babylonian 
 inscriptions (see the table given above). It was this later, 
 or Aramaean, Chaldaic which was brought back by the Jews 
 after their Babylonian captivity as the 'language acquired in 
 place of Hebrew, which it entirely supplanted in practical use 
 by the time of the Maccabees. From the fact that this 
 tongue is the original language of a few short passages of 
 some of the canonical books of the prophets (Ezra, Daniel, 
 and Jeremiah) it is commonly known as Biblical Chaldee. 
 Erected into the literary language of the Jews from Maccabean 
 times, Chaldee remained in that position till after the tenth 
 century a.d. 
 
 Meanwhile the Syriac, or Aramaic of Syria and Meso- 
 potamia, came to differ from the Eastern Aramaic, or Chaldee, 
 in appreciable respects, and especially in the position of the 
 accent. Its earliest written appearance is in the inscriptions 
 of Palmyra (first century A.D.), but it is particularly with 
 Christian documents that its history is associated. Syriac 
 may thus be styled the ' Christian Aramaic,' while Chaldee 
 may be distinguished as the ' Jewish.' The literary con- 
 nection of Syriac with Christianity begins with the ' Peshito ' 
 {i.e. ' simple,' ' true ') translation of both Testaments, which 
 is usually assigned to the second century A.D. Syriac 
 flourished till about A.D. 1000, and played an important part 
 in the spread of science and philosophy by means of trans- 
 lations, as, for example, of Aristotle. Ultimately it gave way 
 before the inroads of Arabic, and it is now represented only
 
 General Survey oj Languages 169 
 
 by some scanty dialects about Lakes Van and Urumiyah. 
 It is regularly used, however, in the Nestorian Liturgy. 
 
 Mendaean was the dialect of a religious sect of Meso- 
 potamia, and, as preserved in writings of the fourth and fifth 
 centuries, is a derivative of the general Aramaic, not of 
 Syriac. 
 
 Assyrian (with Babylonian) has been read and digested 
 into grammars from thousands of cuneiform inscriptions upon 
 bricks, cylinders, and monuments of Assyria and Babylonia. 
 " It presents us, on the whole, with an archaic form of 
 Semitic speech. In fact, Assyrian may justly be described 
 as the Sanskrit of the Semitic idioms " (Sayce). In the 
 rock-inscriptions of the Great Kings it forms the third text. 
 The difference between Assyrian proper and the Babylonian 
 variety is no greater than that between two neighbouring 
 English or German local dialects. 
 
 Canaanitic. — The languages of the Canaanitic group are 
 simply so many local or provincial dialects of the same 
 speech. The difference, for instance, between Phoenician 
 and Hebrew is chiefly that the former employs certain forms 
 which are archaic for Hebrew, and that the contents of 
 meaning of particular words do not always correspond in the 
 two tongues. 
 
 The history of Hebrew is somewhat complicated. Its 
 purest form — that which dates from the pre-Babylonian 
 period of national independence — is known to us only from 
 the older portions of the Old Testament. The Pentateuch, 
 the Books of Joshua, Judges, Samuel, Kings, the greater part 
 of the Psalms, Proverbs, the Song of Solomon, Job, and the 
 older prophets (Joel, Amos, Hosea, Isaiah, etc.) represent, 
 with allowance for a certain amount of later ' editing,' the 
 ' golden age ' of the language. 
 
 After the Captivity (as has been stated above) the 
 Eastern Aramaic prevailed more and more in the popular 
 speech, till ultimately it substituted itself for Hebrew in 
 literature also. The period during which this was taking 
 place may be put down as 536-160 B.C. Priests and
 
 1 70 Natural History of Language 
 
 scholars had meanwhile retained the use of Hebrew (more 
 or less modified by Chaldee influences) for religious and 
 educational purposes. It is this ' Chaldaised ' Hebrew which 
 is found in the Books of Ezra, Nehemiah, Chronicles, Esther, 
 the prophets Jonah, Daniel, Haggai, Zechariah, and Malachi, 
 and in a number of the Psalms. After the complete substi- 
 tution of Aramaic in practical use even for literature,^ Hebrew 
 was preserved in the rabbinical schools till the twelfth 
 century A.D., in a contaminated form known as ' Modern ' 
 (or ' Neo-') Hebrew, which has blended Aramaic, Greek, and 
 Latin elements with the true Hebrew. To this period 
 belongs the composition of the Mishna. Since the twelfth 
 century there has been a literary revival of the language, 
 with attempts at a more antique colouring. 
 
 The Moabite dialect appears on the ' Moabite Stone ' of 
 about 900 B.C. It is closely similar to Hebrew. 
 
 Phoenician is represented by a number of inscriptions 
 on tombs, votive tablets, coins, seals, etc., and by a passage 
 in the Poaiulus of Plautus. The most important in- 
 scription is that upon the sarcophagus of King Eshmunazar 
 of Sidon (of the sixth century B.C.). The remainder are 
 mostly from the Phoenician colonies in Africa, Cyprus, Malta, 
 Sardinia, Marseilles, and Spain. Those from Tunis and 
 East Algeria are Carthaginian (Punic), and the oldest of 
 them are identical with Phoenician ; but the Neo-Punic 
 (which lasted in some measure till the fifth century A.D.) 
 exhibits important deviations. St. Augustine (fourth century 
 A.D.), himself practically a Carthaginian, remarks that Punic 
 differed little from Hebrew. 
 
 Arabic. — It is necessary to distinguish between the 
 Classical Arabic, which belongs to literature (except in so 
 far as some of the purer Bedouin tribes of interior Arabia 
 
 ^ The change of language among the Jews necessitated the translation, or 
 rewriting, of the Hebrew books of the Old Testament. These translations (the 
 'Targums') are, however, not pure specimens of the Eastern Aramaic of the 
 Captivity, the Western or Syriac influences of a later date being very strong. 
 
 The Syriac influence was naturally more powerful in the north of Palestine, 
 and the language of the Galilean Jesus Christ was therefore Syro-Chaldaic.
 
 General Survey of Languages 1 7 1 
 
 have retained the same language in current speech), and the 
 Neo- Arabic, as actually spoken in modern times from 
 Barbary to the Persian borders. The latter differs from the 
 former in its advanced analytical character (particularly 
 shown in the rejection of case-inflection, in which it has 
 at length arrived at the same point as Hebrew), in a 
 measure of phonetic corruption, and in certain changes of 
 vocabulary (chiefly borrowings from foreign sources). 
 Classical Arabic is of the old Koreish dialect of Central 
 Arabia, which was raised to the position of the standard 
 language by Mahomet.^ Its golden period was that of the 
 poems (the Moallakdt, etc.), which came into existence 
 shortly before Islam, and of the Koran as revised and 
 authorised by the Caliph Othman {circ. A.D. 650). As com- 
 pared with Canaanitic or Aramaic the classical Arabic is 
 evidently nearer to the primitive Semitic in point of structure. 
 Modern Arabic has ousted most of the other Semitic as well 
 as Hamitic languages, and is now spoken with slight differ- 
 ences in each of the four chief regions — Barbary, Egypt, 
 Syria, and Arabia. Maltese is a corrupt jargon, while 
 Mosarabic, once spoken by the Moors in Spain, is extinct. 
 The influence of Arabic has been very marked upon the 
 vocabulary of Persian, and thence of Turkish on the one 
 side and of Hindustani on the other ; and the language 
 itself has penetrated deeper and deeper into the heart of 
 Africa. A number of terms connected with science or 
 commerce have found their way into European tongues 
 (e.g. algebra, cipher, zero, magazine, etc.). 
 
 Himyaritic. — The Himyarites of South -West Arabia 
 (Yemen) formed along with, or as successors to, the Sabaeans 
 a powerful kingdom, including Aden, Mocha, and the 
 neighbourhood. Subsequently they were overcome by the 
 more northerly Arabic-speaking Mohammedans. Linguisti- 
 cally the term Himyaritic is applied to the large number of 
 both Sabaean and Himyaritic inscriptions found in their 
 
 ^ Compare the effecl of Luther's Bible in Germany, of Attic Hterature in 
 Greece, and of ' Sicilian Tuscan ' in Italy.
 
 172 
 
 Natural History of Langtiage 
 
 region, some of which date from before the Christian era. 
 The modern Ehkili (or Hakili) is probably a direct 
 descendant. 
 
 Abyssinian Semitic. — In prehistoric times Semites 
 crossed the Red Sea from South-West Arabia into Abyssinia, 
 carrying with them a dialect or dialects closely akin to the 
 Himyaritic. The oldest representative of these speeches is 
 the Ghees (sometimes unfortunately called Ethiopic, a term 
 which leads to some confusion with a group of the Hamitic 
 languages). It possesses a Christian literature, with a 
 translation of the Bible dating from about A.D. 400. Two 
 inscriptions at Axum belong to about A.D. 500. Till the 
 fourteenth century Ghe'ez was the Court language, but at that 
 period it began to yield to the younger Amharic, and is now 
 a liturgical language only. Along with Amharic (of South- 
 West Abyssinia) go the contaminated dialects Tigre (of the 
 north), Tigrine (of the centre), and Harari (of Harar, east of 
 Gallaland). 
 
 (ii.) THE HAMITIC DIVISION 
 
 If the Hamitic languages are related by genealogy 
 to the Semitic the separation has been an exceedingly 
 long one, and the very remote original of the two 
 divisions must presumably have undergone large develop- 
 ments before it took, in the one domain, the shape which 
 may be called Primitive Semitic and, in the other, that of 
 such a language as the old Egyptian. It does not, of 
 course, follow from the greater antiquity of discoverable 
 Egyptian that the hypothetical original was structurally 
 nearer to this than, say, to Assyrian. On the whole, indeed, 
 argument points to a more rapid breaking-down of original 
 methods in the Hamitic region. While the Egyptian noun 
 and adjective possess a dual (in id) and a plural flexion (in 
 «, e.g. son, ' brother,' soum, ' brothers '), and both adjective and 
 noun have a feminine (in t, e.g. sonX, ' sister '), there is no 
 sign of declension by case-suffixing in Hamitic. In Egyptian
 
 General Survey of Languages 173 
 
 the person-signs are, indeed, suffixed in the inflexion of the 
 verb — with both masculine and feminine expressed in the 
 second and third persons, as in Semitic, e.g. uon)s., ' thou 
 (masc.) art,' uont, ' thou (fem.) art,' uonf, ' he is,' nons, ' she is ' 
 — but there is very little corresponding to the treatment of 
 the Semitic verb in respect of the 'determined ' conjugations. 
 So far, however, as these conjugations do appear they are 
 significant. Thus, according to Professor Sayce, ked may 
 give keied, keneb^ akeb, sekeb in different senses.^ With the 
 prefixing in the latter cases we may compare the method 
 in the Tamashek (Berber) causative, er/nn, ' to be ill,' ser/iiu, 
 '•to cause to be ill,' and in Beja (Ethiopic) edliib, ' to sell,' 
 esdeliib, ' to cause to sell.' These undoubtedly recall the 
 Semitic devices. Of the characteristic Semitic vowel- 
 variation, however, there is no trace ; and, while Semitic 
 roots are triconsonantal, the Egyptian are generally mono- 
 syllabic and uniform. For the most part, as in Malayo- 
 Polynesian, there scarcely exists a distinction between root 
 and word. 
 
 The Hamitic languages, in the three main divisions 
 Berber, Egyptian, and Ethiopian, show considerable 
 divergences among themselves. They agree, however, 
 in the use of I as the feminine sign. Thus, while Egyptian 
 has son, sont ; nuter (' God '), nutert ; nofei- (' }'outh '), nofert, 
 Tamashek (Berber), both prefixing and suffixing the element, 
 has akli (' negro '), taklit ; ekahi (' cock '), tekahit (' hen ') ; 
 elu (' elephant '), tehil. The Ethiopic Beja has tak (' man '), 
 takaX (' woman '), and vestiges of the same -t remain in 
 Galla, etc. The pronouns exhibit a marked relationship. 
 Thus old Egyptian ajinuk (Coptic anok) = ' I,' for which the 
 Berber Tamashek has ?iek and the Ethiopic languages some 
 modification of ani or anu. ' He ' is in old Egyptian enhif, 
 in Tamcshek it is etita, in Galla ini. In the suffix-position 
 
 1 Outside Hamitico-Semitic, however, we might compare (aUhough with 
 great reservations) such a formation as keneb with the Indo-European \a.^i^a.vui 
 (root \a^), Jm.do (rooted), and seied with Enghsh verbs like bespea/e or German 
 compounds with eji(-, zer-, ve?--. Again, ketcb might be compared with the 
 Malayan and Dravidian infixed roots.
 
 1 74 Natural History of Language 
 
 the possessive pronoun of the second person singular is in 
 old Egyptian -k, in Tameshek -/', in Beja -k, and in Galla 
 -ke, while in some of the other Ethiopia tongues it appears 
 in the prefix-position as kic-. 
 
 The customary divisions and subdivisions of the Hamitic 
 languages are : — 
 
 1. The Eg-yptian branch, comprising — 
 
 (a) Ancient Egyptian, a language which can be traced 
 back in the hieroglyphs for 6000 years, and of which some- 
 thing has been already said. It may be added that, as 
 is usual with cultivated speeches {e.g. Latin or Sanskrit), 
 there was considerable divergence of the popular, or demotic, 
 from the ' sacred,' or hieratic, idiom. 
 
 {^) Coptic, the lineal descendant of popular Egyptian, 
 with a Christian literature, written in Greek characters ^ 
 and dating between the second and eighth centuries 
 A.D. This neo-Egyptian tongue survived, though latterly 
 only in the cloister, till some three centuries ago, when 
 it was finally supplanted by the Arabic which had been 
 gradually ousting it since the inroad of Islam. Coptic 
 was greatly influenced by Greek in point of vocabulary ; 
 neither was it homogeneous, inasmuch as three dialects are 
 recognisable, a northern {Bashnmric), a central {Memphitic), 
 and a Southern {Theban). 
 
 2. The BerbeP (also known as the Libyan) group. 
 Originally the Berbers occupied all Northern Africa from 
 Egypt to the Atlantic and from the Mediterranean to the 
 Neo-ro country. They included such peoples as the 
 Mauretani, Numidae, Gaetuli, and Libyans of history, who 
 were at various times dislodged from various parts by Greeks 
 (in Cyrenaica), Carthaginians, Romans, and subsequently by 
 Arabs. Berber languages are still spoken in the Western 
 Sahara (where is to be found the Tamashek, which is 
 usually treated as the most representative dialect), in parts 
 
 1 The influence of Greek on Egyptian was, of course, due in the first instance 
 to the Greek occupation under Alexander, with the consequent Greek civilisation 
 and culture of the Ptolemies and of the later Alexandrine Hellenes and Hellenists.
 
 General Survey of Languages 1 75 
 
 of Morocco, of Algeria (by the Kabyles, etc.), and of Tunis 
 and Tripoli. Up till a century ago the Guanches of the 
 Canary Islands employed a cognate tongue. The old forms 
 of the Berber idioms are in some measure discoverable from 
 a considerable collection of inscriptions. 
 
 3. The Ethiopian group, southward of Egypt. The 
 chief speeches are the Beja (between the Nile and the Red 
 Sea, north of Abyssinia), Galla (south of Abyssinia to the 
 Equator), Somali (from the Straits of Bab-el-Mandeb to 
 Cape Guardafui and southward), Deiikali, and other dialects 
 in West Abyssinia and neighbourhood. 
 
 B. The Indo-European Family 
 
 A family of languages stretching from Northern India 
 through Persia and Armenia, and thence westward and 
 northward, embracing all the speeches of Europe except 
 those previously mentioned (namely, Basque, sundry Ural- 
 Altaic tongues, and speeches of the Caucasus) is not inaptly 
 styled the Indo- European. The term is not, indeed, 
 altogether ideal, inasmuch as it is perhaps suggestive of a 
 combination of Indian and European, excluding, among 
 others, Eranian and Armenian, rather than of a family which 
 finds its eastern and western bounds in Europe and India 
 respectively. Apart from a certain cumbrousness, however, 
 it is open to less objection than any other title yet proposed, 
 unless it be ' Aryo-European,' which also lacks comprehensive- 
 ness. The fact that Europe no longer bounds the European 
 tongues, which are spoken to a greater or less extent in 
 every continent, particularly in America, is of little moment 
 in this connection, since the English, Spanish, and French 
 of North America, the Spanish, Portuguese, etc., of South 
 America, the English, Dutch, etc., of Africa, and the English 
 of Australasia are, as languages, logically included in the 
 term ' European.' 
 
 An older title, still habitually retained in Germany, is 
 that of ' Indo-Germanic,' a name which is said to have
 
 1 76 Natural History of La7iguage 
 
 originated in the notion that the two geographical extremes 
 (in the old world) were occupied by languages of the Indian 
 and Germanic branches respectively. Since, however, the 
 family equally includes the Celtic tongues in Brittany and 
 the western parts of the British Islands, as well as Spanish 
 and Portuguese, the appellation is manifestly incorrect. 
 ' Indo-Celtic ' has accordingly been suggested, but has found 
 no support, the truth apparently being that we instinctively 
 look in the name for something which will express as fully 
 as possible the components of the family rather than specify 
 two particular termini. 
 
 To call the family ' Aryan,' with some writers, is to ex- 
 tend the name of a part to the whole. Arya was the 
 Sanskrit term (frequent in the Rig- Veda) for one who be- 
 longed to the good and proper stock, and, as an adjective, 
 signified ' noble,' the non-Aryas {Dasyus = ' foes ') being the 
 Dravidian and other alien and inferior peoples of India. 
 The expression was employed also in ancient Erania, and 
 therefore dates from a time when the Indian and Eranian 
 divisions of the Asiatic branch of the family were still one. 
 Beyond these divisions the word has no special aptness, and 
 it is therefore more properly restricted to them. 
 
 Other attempts at a brief title have proved failures for 
 obvious reasons. ' Sanskritic ' conveys, as will be seen, an 
 erroneous impression as to the position of Sanskrit in the 
 family. ' Caucasian ' follows an old but quite unscientific 
 ethnological term, and is particularly unfortunate when 
 applied to language, inasmuch as the speeches of the 
 Caucasus (except Ossetian) have no connection with the 
 Indo-European tongues. 'Japhetic' sprang from a desire 
 to match ' Semitic ' and ' Hamitic ' symmetrically with 
 another Biblical label, but several of the peoples traditionally 
 deriving from Japhet (Gen. x.) spoke languages altogether 
 alien to Indo-European. 
 
 In the Indo-European family are included a large 
 number of languages and a still larger number of dialects, 
 dead and living, and these afford unusually ample data for
 
 General Sttrvey of Languages 177 
 
 observinf^ the process of linguistic evolution. At present it will 
 suffice to note the salient features of the family for the purposes 
 of comparison with the tongues already similarly discussed. 
 
 Since a stage as far back as philological study can 
 penetrate, the speeches of this stock have been amalgamating- 
 inflexional. Nouns and pronouns were inflected in case and 
 number ; verbs in number, person, tense, mood, voice. The 
 inflexional modifications were in all cases terminational, i.e. 
 the flexion is by suffixes. These cannot be credited, even 
 at the earliest discoverable stage, with any capacity for 
 independent existence. Nor can it be supposed that those 
 who then spoke the primitive Indo-European tongue were 
 either conscious of any special meaning attaching to a 
 special suffix-element, or, in most cases at least, of the fact 
 that they were combining a root with a suffix. The philo- 
 logist who attempts to dissect the complete words of 
 primitive Indo-European may arrive at a long list of ' stem- 
 forming ' or ' case-forming ' suffixes ; he may discover 
 ' primary ' and ' secondary ' person-endings, and so on ; but 
 he almost entirely fails to discover any independent meaning 
 attaching to those suffixes as separate entities, or any ' pro- 
 nominal,' ' deictic,' or ' demonstrative ' roots from which they 
 may have been corruptions or curtailments. He may feel 
 compelled to believe that originally the said suffixes did 
 possess, in some shape or other, a separate existence and a 
 meaning of their own, and that they were then attached to 
 other ' roots,' in the same way as the affixes in the agglu- 
 tinating tongues, for the purpose of determining the function 
 or relation of those roots in a sentence. Nevertheless, he 
 cannot actually demonstrate the fact. He can point out 
 that this or that element serves this or that purpose in the 
 building of a word, but it is extremely seldom that he can 
 show with any plausibility why it does so, or what shape of 
 its own it possessed before it came into such a position,^ 
 
 ^ Only one or two probabilities can be quoted in qualitication of this remark. 
 The locative case in some declensions might apparently either end in -i or dispense 
 (b 884) N
 
 lyS Nahtral History of Language 
 
 So far as a reasonable theory of origin can be advanced in 
 individual cases, the evidence tends, as a priori argument 
 would also tend, towards the belief that simple agglutination 
 must be presupposed ; but it must be borne in mind that the 
 oldest forms actually deducible for Indo-European are 
 already completely amalgamating. If *^/("//^j' (' horse ') was 
 the nominative, and *ek7josio the genitive, the philologist 
 may perhaps, though without certainty, analyse the former 
 into ek-7i-o-s and the latter into ek-u-o-sw. In ek-u-o-s he 
 may speak of j as a sign of the nominative, and of <? as a 
 * stem-forming ' suffix ; but he is in no position to say what 
 was the independent shape or value of the element here 
 represented by s or o. Nor can he say why a change to the 
 feminine corresponds with a change of form to *ekij.d. In 
 *ekiwsio he cannot tell exactly how much is genitive-suffix 
 and how much is stem-suffix ; nor does he imagine that 
 those who habitually used the form conceived of -osw or 
 -sio as an element attachable and detachable at will. 
 Similarly, if the presumable third person plural of the 
 present optative passive from the root *M^r (' carry ') was 
 *bherointo, it is possible to break it up theoretically into 
 bher-o-i-nto and to call o a ' stem-forming ' suffix, i a suffix 
 of the optative, and nto a combination of signs for the third 
 person plural and the medio-passive voice. Even then the 
 'secondary' ending of the passive differs from the 'primary' 
 ending -ntai. How o comes to form a stem, and i the 
 optative ; precisely how much of nto represents the person, 
 and why it does represent it, are matters not yet discovered 
 and perhaps not discoverable. 
 
 If, leaving the primitive or ' pro-ethnic ' words, we pass 
 on to the separate Indo-European languages in their older 
 forms, we find the amalgamation of root with suffix tending to 
 become (chiefly through phonetic corruption) still more com- 
 
 with that sound. It sometimes appears in Sanskrit and sometimes not, and the 
 Greek words aiVs {aies) and aXei [aiei) are the direct phonetic survivals of two 
 original forms ai%es and aiues-'x respectively. These considerations point to a 
 certain independence of a ' deictic ' ?', employed something after the fashion of ci 
 and /^ in the French celui-ci, celui-la.
 
 General Survey of Languages 179 
 
 plete. It may not be easy in all cases, even in the recon- 
 structed original, to find the ' root ' maintained in a 
 distinct and consistent shape throughout a series of cognate 
 words. Nevertheless the root is much less disguised there 
 than it is apt to be in the later separated speeches, such as 
 Greek, Latin, or Gothic. At the same time, with the 
 phonetic disguising of the root goes a still greater and 
 progressive obscuration of the suffixes. Root and suffix are 
 more and more welded into a phonetic unity. To take a 
 comparatively simple case: if the Indo-European root '^es 
 (' be ') exhibited in the singular of the present indicative the 
 forms esmx, es{%)\, esXi, and in the singular of the present 
 optative the forms j'iem, jies, ,$iet, we find in classical Greek 
 ei'/ii, el, eVrt {eimi, ei^ esti), and elr\v, eir]<;, etrj {eien, eies, eie) 
 respectively. In et'/xi {eiini), el (ei) phonetic corruption has 
 destroyed the symmetry which existed in the earlier esviz, 
 es{s)i, and the root is altogether hidden. In the case of 
 €4 {ei) the suffix also is non-apparent. In the case oietrj (eie), 
 for which a new formation by analogy {esiet instead of stei) 
 has to be presupposed, the said esiet (ficrcTjr), by losing both 
 the s of the root and the t of the person-suffix, has effectu- 
 ally disguised both elements. In primitive Latin the suffix 
 -se formed infinitives, and from roots ve/, /er, such infinitives 
 naturally took the shape of fve/se, f/erse. Phonetic cor- 
 ruption, however, produced ve//e, ferre, combinations in 
 which the identity of the suffix is wholly lost, A similar 
 process of obscuring suffixes by phonetic corruption had 
 probably taken place in pro-ethnic Indo-European, before 
 the earliest times at which we can arrive, and for this 
 reason it may remain impossible to discover the original 
 shape or meaning of a large number of suffixed elements. 
 
 The later stages of the separate Indo-European lan- 
 guages become in all cases more and more analytical. 
 This is largely due to the fact that the original flexions had 
 become so corrupted and confused phonetically as to be devoid 
 of recognisable system, Flexional terminations are therefore 
 replaced by prepositions, auxiliary verbs, and other help-
 
 i8o Natural History of Language 
 
 words. The growth of these devices can, to a large extent, 
 be traced historically. In the primitive stage analysis was 
 very sparingly employed. The derived languages, such as 
 Sanskrit and the ' classical ' tongues, begin to show in their 
 ' synthetic ' stage an increasing use of prepositions, free 
 pronouns, and other resolutions. Later still the modern 
 tongues tend to make their structure depend mainly on word- 
 position, aided by pre-posed auxiliaries and particles. 
 Where the primitive noun possessed at least eight cases, the 
 separate speeches have everywhere tended to confuse the use 
 and shape of these, to reduce their number, and ultimately 
 to dispense with them in favour of a single form qualified 
 by a series of prepositions. Of three original numbers 
 (singular, dual, and plural) belonging to the noun, pronoun, 
 and verb, the dual is rejected ; while English, for example, 
 hardly regards number at all in the conjugation of the verb. 
 No Indo-European tongue indeed has wholly lost the vestiges 
 of the inflexional system. English and Persian have gone 
 furthest in that direction. Nevertheless, English still pre- 
 serves a suffix-sign for the plural (-j-, sometimes -e7i) ; a pos- 
 sessive case, e.g. man's, men's (which, however, shows signs 
 of passing out of use) ; an objective and a subjective form 
 for the pronoun {thou, thee, etc.), and a remnant of flexion- 
 endings in verbs, e.g. lovest, loves, loveA.. Traces of a 
 suffix-flexional stage survive also in the case of the ' modified ' 
 plurals, e.g. foot, feet, and of the ' strong ' verbs, e.g. see, saw ; 
 drink, drank, drunk. The vanished suffix has left its mark 
 behind, and the language is still flexional in virtue of using 
 such variations of form to express variations of meaning, 
 instead of employing separate help-words in every instance 
 to express number, tense, or other relation. When it says 
 
 * I will go,' or ' of a man,' English is analytic ; when it says 
 
 * I went,' ' a man's,' it is both flexional and analytic ; if it 
 further combined the sense of the pronoun ' I ' with the verb 
 ' went ' (as in the Latin ivi), it would be flexional in the full 
 sense. In saying /i? dirai French has flexion where English 
 has none (' I shall say '). In substituting, as it commonly
 
 General Survey of Languages 1 8 1 
 
 does,/*? vais dire^ French makes a further step towards analysis ; 
 but it still carries the burden of flexion in the special form 
 vais^ since it is not permitted to say equally nous vais dire 
 or voiis vais dire. English, except for the practically obsolete 
 ' thou shalt say,' has reached the stage of complete analysis 
 by leaving the auxiliary uninflected. A comparison of 
 primitive Indo-European (so far as it is ascertainable) with 
 classical Latin, then with popular Latin, and next with 
 French, or with Gothic and next with English or German, 
 and so throughout, would show in what various respects and 
 measures the several divisions and subdivisions of the family 
 all make progress in the direction of analysis. 
 
 Primitive Indo-European roots, as induction arrives at 
 them, are monosyllabic. They have, however, no inde- 
 pendent existence apart from stem-forming or word-forming 
 suffixes. A person-ending or a case-ending may indeed 
 occasionally be found attached directly to the root without 
 the intervention of a stem -forming element, as in *es-ini 
 (' I am '), or *iidq-s (' voice '), but a simple es or noq has 
 no place in speech proper. Usually the root is built into 
 a stem or base by some element or elements out of a 
 numerous list (-(?-, -e-, -rid)-, -n{d)-, -u{p)-, etc.), and thereto 
 may be attached further elements one upon another, until 
 the various relations of the complete word (in point of case, 
 number, person, mood, tense, etc.) are fully expressed. Thus 
 ho-6r}-(7-6-[xe6a {do-the-s-o-i)iethd) contains elements represent- 
 ing ' give ' -|- passive aorist sign -t- future sign -|- stem-sufiix + 
 sign of first person plural blended with sign of the passive. 
 So 8ovvai {doimai, ' to give ') = ho-Fev-ac (do-?jeii-ai), and 
 contains elements representing ' give,' noun-stem-suffix, and 
 dative ending, though no Greek was aware of the fact. 
 Obviously these elements do not possess the distinct in- 
 dividuality of those employed in the typical agglutinating 
 languages. The Turkish sev-in-dir-il-ine-iiick combines to 
 a Turk the immediately recognisable series of notions ' love- 
 oneself-make-be-not-to ' (' not to be made to love oneself), 
 and he possesses a ready consciousness as to which element
 
 1 8 2 Natural History of Language 
 
 conveys which notion. On the other hand, the Greek riaav 
 ( = fjSa-av, Hdsan), ' they knew,' would, to a Greek, admit 
 of no distinct analysis. He would feel that the word was in 
 some way related to olSa {oida), ' I know,' and he might, if 
 questioned, be aware that he was in the habit of employing 
 the ending -aav {-san) in certain past connections ; but he 
 would be unable to dissect the word into its constituent 
 portions. The Roman would have had equal difficulty with 
 the corresponding vlderant. 
 
 Prefixing for flexional purposes forms no part of the 
 Indo-European method. Where prefixes occur they are 
 instances of mere compounding. The Greek and Sanskrit 
 ' augment ' of past tenses, as in eSetfa (edeiksa), odiksam, re- 
 presents a demonstrative word which was originally separate 
 and optional. It could, in fact, still be omitted both in 
 Sanskrit and early Greek. Other prefixed ' determinants ' 
 belong to the same category. The negative *n- (Lat. in-, Eng. 
 nn-, Gk. a-, av-) represents as clear a case of compounding as 
 the verbs with prefixed prepositions in Latin or Greek, or 
 those with be-, ivith- in English, or those with ver-, e7it-, zer- 
 in German. 
 
 The form of composition called ' reduplication,' which is 
 so pronounced a feature of Malayo- Polynesian and is found 
 in many languages, was familiar in primitive Indo-European, 
 to the extent at least of adding before the root a syllable 
 either identical with, or accentually modified from, the first 
 sounds of the root itself. The Latin spopo7idi (for -fspo-spondi), 
 the Gothic skaiskaid (from skaida, ' cut '), the Greek tlOtj/xl 
 {tithemi, for '\tJiithevii), the Sanskrit babJiuva and their like 
 are remnants of a process which was by no means confined 
 to the forming of ' perfects,' ' presents,' or any other special 
 tense, and which therefore was not originally inflexional. 
 A liberal faculty for compounding, indeed, was one of the 
 chief features of the family, although different branches have 
 exercised the power in very different degrees. Sanskrit and 
 Greek compound whole word-stems freely ; Latin and its 
 daughter-speeches only slightly ; German is fond of com- 
 
 I
 
 General Survey of Languages 183 
 
 pounding and re-compounding such stems, whereas EngHsh 
 is very sparing of the device. 
 
 Vowel-variation of the root itself was a regular pheno- 
 menon of the primitive Indo-European, but had originally 
 none of the effect upon meaning or function which was exer- 
 cised by the much more extensive and systematic vowel- 
 mutation in Semitic. Thus in Indo-European the root* bher 
 might also appear as bhor, b/ir, and bhr-, the root *es as s ; 
 the root ^seni as soin, sm, sm, the root *ez as z or oi ; but 
 these divergences were apparently brought about only by 
 differences in the position of the main and secondary accents 
 of a completed word, and by a consequent modification of 
 the enunciation given to its ' root ' portion. They in no 
 way served to express causative, intensive, or similar modi- 
 fications of meaning, nor to mark off verbs from nouns, nor 
 (originally, at all events) to distinguish tenses. There is 
 accordingly not the same reason as in Semitic to write a 
 ' root ' with consonants only, but it is convenient to give as 
 M(? root that grade of vowel which would occur when the 
 chief tonic accent fell upon it.^ 
 
 The flexions of both noun and verb in Indo-European 
 were very numerous. There were various declensions of 
 nouns (each possessing its peculiarities in the way of case- 
 endings), in singular, dual, and plural. The exhaustive 
 denotation of number, person, tense, mood, and voice made 
 the complete Indo-European verb an elaborate and compli- 
 cated structure. This fact would naturally give scope for a 
 large and rapidly increasing discrepancy between the sepa- 
 rate languages when they came to pursue their individual 
 developments. The special phonetic tendencies of a par- 
 ticular tongue, and also its idiosyncrasies in respect of the 
 direction in which the analogical instinct would lead it, 
 combined to break down the old paradigms and to cause 
 the substitution of others, which perhaps corresponded little 
 
 ^ Technically the Indo-European vowel-variation is known by the name of 
 Ablaut. The formative matter (or suffix-material) was itself subject to Ablaut. 
 Thus a suffix might appear as -ei- or -i-, as -eu- or -«-, as -er-, -r-, etc.
 
 184 Nahiral History of Language 
 
 to those which another of the tongues had meanwhile evolved. 
 Moreover, the inherited power of composition made the con- 
 struction of new words habitual, and the new constructions 
 in one division could hardly coincide with those invented in 
 another. Nevertheless the correspondence of the languages 
 now to be mentioned, both in their grammar and vocabulary, 
 is sufficient to leave no doubt as to the boundaries and 
 contents of the family. We may proceed to detail its 
 several branches, groups, sub-groups, languages, and in some 
 cases dialects, adding such comments as appear most likely 
 to interest the general student. 
 
 A. The Aryan Branch 
 
 (a) Indian (or Indie) Division 
 
 (i.) Vedic Sanskrit represents the oldest discoverable 
 form of Aryan spoken in India, and stands to the later 
 classical Sanskrit somewhat in the same relation as the 
 Greek of the Homeric age to that of the fifth century B.C. 
 It is principally distinguished by its more numerous verb- 
 forms, as well as by a difference of vocabulary and phonetics. 
 Veda (' knowledge,' and particularly sacred lore) is the name 
 applied to the religious literature of hymns, proverbs, ritual, 
 exegesis, etc., included in various compilations known 
 collectively as ' the Vedas.' The oldest portions of these, 
 the Mantra or SaviJiitd of each, contain hymns and proverbs, 
 and the part of the Vedas most important to the philologist 
 is the very ancient Rig- Veda (' hymn-lore '), a collection of 
 over 1000 hymns in 10,580 highly-finished verses, dating 
 from as far back as the year 1500 B.C. or earlier, when the 
 Sanskrit-speaking people do not appear to have advanced 
 beyond the Punjab. Other portions of the Vedic literature, 
 attached to the Samhita, are the prose ' Brahmanas ' 
 (theological and ritualistic), of which a special and more 
 philosophic part is familiar as the ' Upanishads ' (' sessions ' or
 
 General Survey of Lmiguages 185 
 
 ' lectures '), and the ' Sutras,' a kind of schoolman's supplement 
 to the preceding. 
 
 This huge literature must for centuries have been 
 transmitted orally, and it is not known at what date it was 
 reduced to writing in an ' authorised version.' As in the 
 case of the Homeric poems, we cannot be certain that we 
 possess the hymns of the Rig- Veda exactly in their original 
 phonetic character. The text exists in two forms, the one 
 known as the ' Pada '-text, in which the words forming a 
 sentence are written separately in the same shape which they 
 would possess if they stood alone, and the ' Samhita '-text, 
 in which a whole sentence is linked or woven together 
 phonetically into a unity, the constituent words being made 
 to conform to certain euphonic conditions, according to the 
 ' combination-laws ' (' Sandhi ') established by grammarians 
 like Panini (fifth century B.C.). Thus Monier Williams 
 remarks that, if rara avis in terris were Sanskrit and were 
 written as 'Sandhi' demands, it would take the shape of 
 raravirinsterrih. The original form of the Vedic verse has 
 been interfered with by the application of the rules in 
 question. 
 
 (ii.) Classical Sanskrit, or Sanskrit in a narrower sense, 
 is mainly a literary or artificially cultivated issue of the Vedic 
 Sanskrit, from which it differs chiefly in the loss of verbal 
 forms and in the substitution of large compounds where the 
 older language would have constructed clauses. Its begin- 
 nings (as distinct from the older stage) may be placed about 
 500 B.C. Sanskrit literature embraced epic poetry (of myth, 
 theogony, cosmogony, etc.), e.g. that of the Mahdbharata} 
 drama in verse and prose (the gods, kings, Brahmans, etc., 
 speaking Sanskrit, while women and persons of the lower 
 orders use the ' Prakrit ' or popular dialects), eg. the Sakuntald 
 of Kalidasa ; lyric poetry (mostly erotic) ; beast-fables ; 
 tales ; grammatical works ; philosophy, science, etc. Sanskrit 
 
 1 The epic called Rdmayana is also of this period. It consists of 48,000 lines, 
 and the Mahdbharata of 200,000. The latter is known to have been in existence 
 by 150 B.C.
 
 1 86 Nahiral History of Language 
 
 is still employed both as a spoken and a written language 
 by the learned throughout India, and serves much the same 
 purpose as was served by Latin in the Middle Ages in 
 Europe. Being a language of the cloister, it has undergone 
 no modern development, but remains in the form and obeys 
 the rules established by Panini. 
 
 Though Sanskrit, in its Vedic stage, is ascertainable at a 
 period considerably earlier than the date at which any other 
 Indo-European language is known, it does not follow that 
 its forms are in all cases nearer to those of the primitive 
 Indo-European tongue. It sometimes happens, for instance, 
 that the original is better preserved in a Greek form as well 
 as in a Greek pronunciation. Nevertheless it is the case 
 that the several elements which go to the building up of a 
 word are mostly much more transparent in Sanskrit than in 
 any of the kindred speeches. Hence the special value which 
 attaches to it in the elucidation of Indo-European structure. 
 Sanskrit, again, possesses eight cases and three numbers, 
 while Latin retains but two numbers and six cases, and 
 Greek but five cases with its three numbers. 
 
 (iii.) The Prakrits. — Side by side with the classical 
 ' Sanskrita ' (the ' correctly formed ' or ' perfect ' language), 
 which was employed for purposes of religion and learning, 
 there went in popular use the ' Prakrita ' ('derived,' ' secondary,' 
 or ' vulgar ') speeches, just as, beside the classical Latin of 
 literature, there went, in the various provinces, a popular 
 Latin, which tended to diverge more and more from the 
 conventional literary type. That these dialects were 
 recognised by the third century B.C. is manifest from the 
 rock-inscriptions of the Buddhist king Asoka, which are in a 
 tongue as distinct from Sanskrit as Italian is from Latin. 
 The Prakrits were natural local developments of the older 
 Sanskrit, and resulted from the loss or phonetic decay of 
 earlier forms and from new creations.^ In their turn the 
 Prakrits became recognised literary languages in their several 
 quarters, and were thus to some extent fixed, while once 
 
 ^ The Prakrits are classified as Eastern, Western, and North-Western.
 
 General Survey of L anguages 187 
 
 more the popular speech passed on with further changes, 
 till the result is a number of modern Hindu tongues. The 
 Prakrits are chiefly known from the above-mentioned parts 
 of the old drama, in which women, and persons of the lower 
 orders, are the personae. One speech, however, in particular, 
 is known from an earlier date, and is frequently distinguished 
 from the general list of Prakrits. This is Pali (' writing '), 
 which was carried by the Buddhist missionaries into Ceylon 
 and Further India, and became established there as the 
 sacred language of that religion. It is in Pali that the 
 Buddhist scriptures of the Tripitaka (' three baskets ') are 
 preserved. 
 
 (iv.) The Modern Indian Speeches. — As the Romance 
 tongues sprang from various modifications of popular Latin, 
 so from the Prakrit idioms came the languages now spoken 
 by some 200,000,000 of the inhabitants of Hindustan. No 
 precise date can, of course, be fixed as that at which they 
 ceased to be Prakrit and became ' modern.' Approximately, 
 however, they may be said to have deserved a new name 
 from about the tenth century A.D. Besides the customary 
 changes caused by phonetic corruption and the reduction of 
 earlier forms, the tongues in question have been affected by 
 influences from the Arabic, Persian, and Mongolian speeches 
 of the various conquerors of India, as well as from those of 
 the earlier Dravidian population. As usual, the effect of 
 both causes has been to substitute to a large degree the 
 analytical for the amalgamating-inflexional method. 
 
 The chief modern tongues belonging to this class are 
 Bengali and Uriya of the north-east (reaching from Assam 
 into the north of the province of Madras) ; Hindi (with 
 Hindustani and Urdu) in the centre and west of Northern 
 India {Hmdustani being Hindi with large Arabic and 
 Persian ingredients, and used as a lingua fraiica for all India, 
 while Urdu is the literary form of the same) ; Marathi in 
 the north-west Deccan ; Gujarati, Sindhi, Punjabi, Kashmiri, 
 and Nepali in the regions which their names imply. It is 
 in question whether the Pushtu for Pashtu) of Afghanistan
 
 1 88 Natural History of Language 
 
 belongs to the Indian or to the Eranian branch, although 
 opinion appears to favour the latter. With these must be 
 associated also the Gipsy (or Romany) dialects, which came 
 into Europe apparently in the twelfth century, bearing traces 
 in their vocabulary of having passed through Persia and 
 Armenia. They have since borrowed from Roumanian, 
 Hungarian, Bohemian, German, etc. 
 
 (/3) Eranian (or Iranian) Division 
 
 (i.) Avestic (otherwise loosely styled Zend and often 
 Old Bactrian), the oldest {i.e. in point of records) of the 
 Eranian tongues, is the language of the older extant portions 
 of the Avesta or sacred books of Zoroastrianism, as they 
 have been preserved by the Parsi adherents of that religion 
 in India and Persia, with whom they form a Bible or religious 
 manual. Properly speaking, this book is in no homogeneous 
 language. Apart from contaminations which it has suffered 
 in transmission from age to age, the language falls into at 
 least two dialects, an older and a younger (although it is an 
 open question whether some of the differences may not be 
 due to place as much as to time). A portion marked by a 
 specially antique character is that containing the Gdthas 
 (' hymns '), a series of compositions in a form of verse which 
 recalls the Vedas, and in a language from which Vedic 
 Sanskrit differs but as a pronounced dialect. The full 
 Avesta ('fixed' or 'canonical text') embraces the sections 
 known as the Vendiddd (the anti-demoniacal law), the 
 Vispered (dealing with sacrificial ceremonies), the Yasna 
 (liturgy to various deities), the Yashts (mythological-liturgical 
 hymns), and the Smaller Avesta (containing devotional 
 exercises from priests, etc.). The Gdthas belong to the 
 Yasna, and their antiquity is proved by the fact that they 
 are assumed in the rest of the Avesta as pre-existing. The 
 term Zend means ' commentary,' and was transferred from 
 the commentary (which is in Pehlevi) on the book to its 
 language. The book itself is frequently known by the
 
 General Survey of Languages 189 
 
 compound name Zend-Avesta^ though the Parsi writers more 
 correctly speak of the ' Avesta and Zend.' It might be of 
 advantage to speak of the older Gatha-dialect as the Avestic, 
 and to reserve the term Zend for the newer Avestic of the 
 other portions. Such a distinction is not, however, agreed 
 upon in philological works. Old Bactrian is a title which 
 has been meant to indicate the geographical position of these 
 dialects, as spoken in the eastern part of Iran, while an 
 archaic stage of the Old Persian (to be mentioned immedi- 
 ately) was being spoken in the west. It has, however, been 
 argued strongly that the Avestic language is in reality Old 
 Median, and that Bactria is but the region to which the 
 Zoroastrian doctrines were transported as a more congenial 
 home. 
 
 (ii.) Old Persian. — The language of the cuneiform 
 inscriptions of certain of the Achaemenian kings, viz. from 
 Darius I. (later sixth century B.C.) till 330 B.C., is that of 
 South-Western Iran. These inscriptions, which have been 
 discovered at Persepolis, Behistun, etc., may be regarded as 
 representing the chief speech of ancient Persia proper. Its 
 forms are much worn down as compared with those in 
 Avestic, and the decay in this respect increases appreciably 
 in the later inscriptions, a fact probably due to much collision 
 with the adjoining Semitic tongues. It should be added 
 that little more than 400 different words are represented in 
 these inscriptions. 
 
 Neither Old Persian nor Avestic can be considered the 
 direct ancestor of the later languages of Erania. These are 
 rather the outcome of such developments among the various 
 Eranian dialects as can no longer be traced. 
 
 (iii.) Middle Persian, Pehlevi (or Pahlawi ^), also, as a 
 peculiarly written language, called Huzvaresh, first shows 
 itself on Parthian coins (third century B.C.), and continues 
 in Persian use till nearly 700 A.D. It was the official 
 
 ^ Explained by Persian scholars as a natural corruption for ' Parthian ' {i.e. 
 Parthav becomes Pahlav).
 
 1 90 Natural History of Language 
 
 tongue of the Sassanid dynasty (a.d. 226-652), and appears 
 upon its coins as well as in its inscriptions. Into this 
 tongue was modernised the Avesta, and in it were written a 
 cosmogony (the Bundehesh) and a number of semi-religious 
 works. Whether it was ever more than a conventional and 
 artificial language is not certain, since, in literature at least, 
 not only in its vocabulary and phonetics but also in its 
 grammar it was much interlarded with Semitic (Aramaean) 
 elements. Semitic words, indeed, amount to nearly one-third 
 of the text. 
 
 The question of its popular currency depends, however, 
 upon how it was spoken rather than upon how it was written. 
 As a matter of fact, words which were written as Semitic were 
 read as Persian (just as they are by the Parsis), in much the 
 same way as we pronounce ' for instance ' when reading ' e.g.,' 
 or ' namely ' when reading ' viz.' Thus what is written as 
 malkan inalka (' king of kings ') was read as shahan shah. 
 In any case Pehlevi shows a very marked decay in the Eranian 
 grammatical forms, the declension of nouns and conjugation 
 of verbs having become largely analytical. 
 
 (iv.) Parsi (or Pazend^) was, according to some authorities, 
 meanwhile developed more to the east. According to others 
 it is a modified Pehlevi and an archaising modern Persian. 
 This speech, while comparatively free from Semitic elements, 
 shows the decay of flexion already remarked in Pehlevi. 
 The Parsis of Bombay, Gujarat, etc., now numbering about 
 100,000, carried this language with them when they fled 
 from Mohammedan pressure in the tenth century. 
 
 (v.) Modern Persian (which differs from Pehlevi chiefly 
 in the numerous incorporated Arabic words and in the 
 script) is usually dated from the national epic, the Shdhndniek 
 ('Book of Kings') written by Firdusi (A.D. 10 10). Firdusi, 
 however, was a deliberate purist, and wrote in a Persian even 
 more free from Semitic elements than the Parsi, whereas, in 
 general, Persian borrowed a multitude of Arabic words and 
 
 1 Pd-zend properly means ' re-explanation,' i.e. Pehlevi ' explained ' in its turn. 
 The illogical Huzvaresh style of writing needed such recasting.
 
 General Survey of L anguages 191 
 
 even phrases, until its vocabulary has become almost as much 
 Semitic as Indo-European, Some of its dialects, however, 
 still remain comparatively pure. Modern Persian possesses 
 no grammatical distinction of gender, and practically no 
 case-endings (except a genitive in -z', with which we may 
 compare the solitary English case in 's), while its verb-tenses 
 are expressed by auxiliaries. Along with English it 
 represents the most analytical stage yet reached by Indo- 
 European tongues. 
 
 (vi.) Among Eranian speeches are further included the 
 Kurdic dialects, the Ossetian and a few other dialects of the 
 south-east Caucasus, and the Beluchi of Beluchistan. The 
 language of Afghanistan is also commonly referred to Eranian 
 rather than to Indian. 
 
 Again, some prefer to reckon the Ossetian with the next 
 branch of the family, the Armenian. 
 
 B. The Armenian Branch 
 
 The question whether Armenian should be included in 
 the Eranian division of the Aryan branch, or treated as an 
 independent member of the Indo-European family, is hardly 
 left open. The latter view is much better supported, since, 
 on the one hand, many resemblances with Eranian vocabulary 
 are naturally due to borrowing by Armenian, while, on the 
 other, Armenian flexion and phonetics show peculiarities 
 which distinguish them markedly from the Eranian. The 
 ancient tradition (known to Herodotus and Eudoxus) con- 
 nected the Armenians with the Phrygians both in descent 
 and language, and the Phrygians certainly entered Asia 
 Minor from Thrace. 
 
 Our earliest knowledge of the language begins with 
 cuneiform inscriptions in the neighbourhood of Van. Its 
 literary records proper date from the Christianising of the 
 country, and in the Old Armenian shape extend from 
 the fourth century to the eleventh. This ' classical ' 
 Armenian was originally the dialect of the Ararat region.
 
 192 
 
 Natural History of Language 
 
 It is still the conventional learned and religious tongue, 
 while a more popular speech, the Neo-Armenian, has sup- 
 planted it in current use. The latter is marked by differences 
 of pronunciation as well as by an admixture of Persian, 
 Turkish, and other foreign words, and is varied by at least 
 four dialects. The Armenians of Europe (Constantinople, 
 etc.) have developed phonetic novelties of their own. 
 Armenian still retains a fair measure of both declension and 
 conjugation flexion. 
 
 C. The Hellenic (or Greek) Branch 
 
 Dialects of Greek, differing comparatively little from 
 each other except in phonetics, were spoken in early 
 historical times not only in the modern European Greece, but 
 in ' Hellas ' generally, including the south of Italy (Magna 
 Graecia), Sicily, all the Aegean Islands, Crete, Cyprus, 
 Cyrenaica, much of the western part of Asia Minor, and 
 various localities round the Black Sea in which Greek 
 colonies had been planted. The earliest extant documents 
 of the language are the Homeric poems, the Iliad and 
 Odyssey, which probably date from at least as early as the 
 tenth century B.C., but which presuppose the existence of a 
 considerable body of verse already in vogue. Owing to the 
 necessity of oral transmission for several centuries over 
 regions differing in dialect, the Homeric poems have not 
 been preserved with the precise forms and pronunciations 
 in which they were composed. Nevertheless the archaisms 
 are ascertained with sufficient accuracy to carry far back the 
 history of the tongue. Even in the Homeric time dialects 
 must have been numerous, and the epic of that day was 
 already somewhat artificialised from older models in point 
 of vocabulary, phonetics, and flexion. 
 
 In the post-Homeric and historical period the chief 
 dialects may be grouped as follows. In their purest form 
 they are chiefly ascertainable from inscriptions, inasmuch as 
 their literature is not only often very slight, or even non-
 
 General Survey of Languages 193 
 
 existent, but also mostly aims at a style in which the more 
 pronounced local peculiarities are replaced by forms of a 
 wider standard. 
 
 (a) Doric, with its varieties in Laconia (with Tarentum and 
 Heracleia in South Italy). 
 Messenia. 
 
 Argos (and Aegina). 
 Corinth (and Corcyra). 
 Megara (and Byzantium). 
 Crete. 
 
 Sicilian colonies from the Peloponnese. 
 Rhodes. 
 
 Various Aegean islands {e.g. Thera, with which goes 
 Cyrene). 
 (/3) North-West Greek. — The dialects of Phocis, Locris, 
 Phthiotis, Aetolia, Acarnania, Epirus, Achaia, and 
 perhaps Elis. (These are often loosely and errone- 
 ously classed with the Aeolic.) 
 
 (7) Aeolic, with its branches in 
 
 Lesbos and Asiatic Aeolis. 
 North Thessaly. 
 Boeotia. 
 
 (8) Arcadian (with Cyprian). 
 
 (e) lono-Attic, with its well-marked divisions 
 
 Ionic. 
 
 Attic (in an older and a younger shape). 
 The predominance of Attic literature and culture, 
 established by 400 B.C., and the fact that subsequently Attic 
 became the recognised Court language of the Macedonian 
 conquerors, combined with the loss of independence on the 
 part of the several Doric, Ionic, Aeolic, and other states, 
 caused the establishment, about 300 B.C., of a 'common' 
 literary and official dialect for the Greek world. This was 
 known as the Koin6 (77 Koivr] 8ui\€KTo<i), and was based on 
 Attic. Naturally, however, it departed from strict Attic in 
 respect of certain of the finer grammatical usages of that 
 standard, as well as in the addition of novelties from various 
 
 (b 884) O
 
 1 94 Natural History of Language 
 
 sources. Spread thus over the Macedonian Greek world of 
 Asia Minor, Syria, Egypt, etc., the Koine came to degenerate 
 with time and to possess its local peculiarities ; and, if 
 writers like Lucian, and other ' Atticists ' before him, 
 attempted in the first and second centuries a.d. a literature 
 which should be pure Attic of the classical age, they did so 
 with difficulty and not always with success.^ Meanwhile 
 — as specimens of marked divergence — there arose in 
 Macedonia a mixed Macedonian- Greek dialect, and in 
 Alexandria an Alexandrian or ' ecclesiastical ' Greek, mixed 
 with Jewish and Egyptian elements. The latter is the 
 language of the Septuagint, and was the parent of the 
 Hellenistic Greek of the New Testament. Both these local 
 dialects differ widely from the recognised Koine. 
 
 Later, during the Byzantine times, the vocabulary of the 
 Koine became more and more corrupted with foreign words ; 
 terms shifted their meanings ; the grammar was much 
 confused or forgotten, both in its flexion and syntax. From 
 this corrupted Byzantine stage comes the Modern Greek. 
 (The title Romaic, which is still in occasional use, properly 
 belongs to the popular language during all the time that the 
 Greeks of the Eastern ' Roman ' Empire called themselves 
 Romaioi). As actually spoken, this language was tending 
 to become a collection of local patois, mixed with Italian, 
 Turkish, Slavonic, and other foreign elements ; but for the 
 last century attempts, by no means ineffectual, have been 
 made to reinstate in literature and education a Greek which 
 aims at least at the classicism of the old Koine. The 
 consequence is that both literature and conversation now 
 differ on different occasions according to the cultivation of 
 
 ■' Jannaris {Historical Greek Gratumar, § 013) says of the Graeco-Roman 
 period (150 B.C. -A.D. 300) that four strata can be recognised in its Greek : (i. ) the 
 Atticists, "who adhere rigidly to classical Attic," e.g. Dionysius Halicarnassensis, 
 Lucian, Aristides, Pausanias, etc. ; (ii.) the common or conventional school, who 
 follow Attic "with many concessions to the spirit of the times," ^.^. Polybius, 
 Diodorus, and Plutarch; (iii.) the Levantine group, including "Asiatic Greeks 
 and Hellenised foreigners " ; (iv.) the colloquial. These, of course, can only 
 represent distinguishable planes connected by many intermediate degrees. 
 
 I 
 
 1
 
 General Survey of Languages 195 
 
 the user or of the persons to whom they are addressed. 
 The ' best ' literary language, while much closer to ancient 
 Greek than any Romance tongue stands to Latin, is, on the 
 other hand, far more widely severed from the average spoken 
 language than is the case with any other national literary 
 language in Europe. Taking the mean of Neo-Greek in 
 practical vogue, we find the chief differences between it 
 and ancient Greek to consist in (i.) pronunciation ^ : for 
 example, the sounds v, t, 7;, vi, oc, ec, 77, which were 
 originally distinct, have become merged in one sound / 
 ( = Eng. ee) ; in the combinations av, eu, the v is pronounced 
 as z; or/; 8 as //; in 'then'; (ii.) accent: the old vowel- 
 quantity is ignored and the pitch-accent has become a 
 quantitative or ' stress ' accent ; (iii.) flexion and syntax : 
 e.g. the dual has disappeared ; the classical dative is 
 abandoned in favour of prepositional phrases ; the other 
 cases show confusion of use ; the optative mood, and the 
 future, perfect and pluperfect tenses, are lost and replaced 
 by auxiliaries ; a clause with particle is used instead of the 
 infinitive, etc. ; (iv.) vocabulary : new words have been coined, 
 old words phonetically corrupted, and foreign words imported. 
 Modern Greek is spoken in Greece proper, in some 
 parts of European Turkey, on the coasts and islands of the 
 Aegean Sea and the Sea of Marmora, in Crete and Cyprus, 
 and along the southern, western, and north-western shores 
 of Asia Minor. 
 
 D. The Albanian (or Illyrian) Branch 
 
 A number of dialects, usually grouped under the two 
 heads Ghegh (to the north) and Tosk (to the south), spoken 
 by the Albanians or Arnauts of Albania proper, as well as 
 by those resident in Greece, are admitted to belong to the 
 
 ^ Classical Attic pronunciation is in many points highly disputable. The view 
 held here is the Erasmian (in the proper sense of that term). The argumentation 
 of the Reuchlinians, supported by Jannaris, is frequently turbid and inclined to 
 proceed in a circle. More satisfactory for the most part is the work of Blass on 
 the Pronunciation of Ancient Greek (trans. Purton).
 
 196 Natural History of Language 
 
 Indo-European stock, while it is perhaps impossible to 
 determine their exact position in the family. The records 
 for the history of the language begin only with the 
 seventeenth century, and are rather scanty. Moreover, 
 thanks to a series of political circumstances, the vocabulary 
 is largely mixed with Latin, Slavonic, Roumanian, Turkish, 
 and Greek words. It can, again, hardly be an accident that 
 the article should be suffixed in Albanian as it is in the 
 contiguous Bulgarian (a Slavonic tongue) and the Roumanian 
 (an Italic). Though much debased, the flexion is distinctly 
 Indo-European. 
 
 Albanian dialects are spoken by more than a million 
 and half of people in a region south of the Schar mountains 
 and the river Drin, bounded by the Slavonic languages of 
 Montenegro, Servia and Bulgaria, the Macedo-Roumanian 
 (Vlach), and Greek. 
 
 Albanian is apparently the sole survivor of a wider 
 group of dialects once spoken all along the east coast of 
 the Adriatic as far as the Corinthian gulf, and thence 
 carried across to the south-east of Italy, where they were 
 employed by the Messapians, lapygians, Daunians, and 
 others. The whole branch may be styled Epirot-Illyrian, 
 and its divisions may be thus arranged : — 
 
 , ,.. . J Venetian (i.e. of the Veneti). 
 
 . . . ^ \ Liburnian. 
 
 Epirot-IUyrian ■ / » ,. • 
 
 ^ ' It-- I Albanian. 
 
 I iLpirot < -, . . . 1 
 
 ^ ( Messapian-Iapygian.^ 
 
 E. The Italic Branch 
 
 Apart from the (unknown) Ligurian, the Etruscan (which 
 is almost certainly not Indo-European), the Messapian or 
 lapygian (of which very little is known, except that it was 
 Illyrian in origin), and the Greek dialects of Magna Graecia, 
 
 ^ It is believed that the name Graeci for the Hellenes was carried into Italy 
 by these immigrants, whose knowledge of the Greeks was mostly connected with 
 the neighbouring tribe TpaiKol of Epirus.
 
 General Survey of L anguages 197 
 
 the peninsula of Italy was occupied in the earliest historical 
 period by closely cognate Indo-European languages, which 
 fall into two main divisions, known to philologists as (a) the 
 Umbro-Samnitic, (/S) the Latin. 
 
 (a) In the Umbro-Samnitic group are included in 
 particular the Umbrian, Picentine, Sabine^ Marsian, Volscian 
 and Oscan speeches, extending along each side of the middle 
 Apennines from Northern Umbria as far south as Apulia 
 and Lucania. Umbrian had formerly occupied a wider 
 territory, including the plain of the Po and much of Etruria. 
 
 Only the Umbrian and Oscan afford sufficient material 
 to concern us, and these two, whatever their divergences, so 
 far agree in certain phonetic and grammatical characteristics 
 which (apart from vocabulary) mark them off from the Latin 
 division, that we are justified in coupling them in classifica- 
 tion, even while we treat them as distinct languages.^ 
 
 Oscan was the language of the Osci ( = Opsci), the Opici 
 of the Greeks, a people originally settled in the neighbour- 
 hood of Beneventum and Capua. By the Romans the name 
 was applied to both the Oscans proper and the Samnites (or 
 Sabellians) in general, their speech being the same. As a 
 language, Oscan was for some centuries of great importance 
 in southern Italy, occupying as it did Campania, Samnium, 
 and a portion of Apulia. It was still the popular speech of 
 those parts in the first century B.C., and probably maintained 
 itself in the mountains till some centuries after the Christian 
 era. Its chief records, besides legends on coins and words 
 quoted by Latin writers, consist of some 200 inscriptions. 
 Of these the most important are (i.) the ' Tabula Bantlna,' a 
 
 ^ Thus for the qu in such Latin words as qiiis, quant, Umbro-Samnitic 
 exhibits p (Umbr. pis, Osc. pani) ; for d {kt) it shows ht (the h implying a pro- 
 nunciation like ch\xv 'loch'), as in Umbr. rehte, Osc. Ohtavis, representing the 
 Latin recfe, Octaviiis respectively ; where Latin has an interior b (derived from an 
 original explosive aspirate), Umbro-Samnitic has/, e.g. Umbr. iifi, ife, Osc. sifei, 
 for Latin tibi, ibi, sibi. In respect of flexion we find that, whereas Latin made 
 the genitive of nouns in -a end with -ae (except in an odd survival YCinQ paterfamilias), 
 Umbrian and Oscan both retain the old -as. Similarly their genitive of nouns of 
 the second declension in -us takes the shape -eis {-cs), while in Latin a peculiar 
 genitive in -i was developed.
 
 1 98 Natural History of Language 
 
 bronze tablet, discovered in 1793, referring to the administra- 
 tion of the town of Bantia, and bearing the Oscan on one 
 side and the Latin on the other ; (ii.) the ' Cippus Abellanus,' 
 a monumental stone (which was serving as a door-step when 
 found in 1740), inscribed with a contract made between the 
 towns of Nola and Abella ; (iii.) a bronze tablet of Agnone, 
 containing the inventory of a temple. The remainder, which 
 were mostly found in Campania, are of little individual 
 account. 
 
 Umbrian, which became confined comparatively early to 
 the mountains on the left bank of the Upper Tiber, is known 
 from the ' Eugubine Tables' {Tabulae Iguvinae). These 
 seven large bronze tablets, having reference to the liturgical 
 acts of a corporation of priests, were discovered in a cellar 
 at Gubbio ( = Eugubium or Iguvium) in 1444. Of the 
 seven, five are written in the Umbrian characters and two in 
 the Latin. Their dates are uncertain, but the earlier parts 
 are not likely to have been more recent than 200 B.C. In 
 any case they supply us with the knowledge of more than a 
 thousand Umbrian w^ords, and therewith of the characteristic 
 Umbrian grammatical forms. 
 
 (/3) The Latin division of Italic demands a fuller treat- 
 ment. Ancient Latin was properly the language of the 
 plain of Latium, and, as both a literary and a spoken tongue, 
 grew in importance with the growth of the Roman empire, 
 until, from a local idiom, it became the recognised speech 
 of those parts of Europe now occupied by the Romance 
 languages, and the official speech of an area more extensive 
 still. In a large measure its spread within the Italian 
 peninsula itself was facilitated by the fact that the chief 
 elements of its grammar and vocabulary were closely akin 
 to those of the Umbro-Samnitic dialects above mentioned, 
 while beyond Italy it met (in the West at least) with a 
 people peculiarly adaptable and receptive in matters social 
 and linguistic. 
 
 Meanwhile Latin itself, both in early and later times, 
 was affected in sundry ways by the cognate dialects, first
 
 General Survey of Languages 199 
 
 those in the neighbourhood of Rome, and then those farther 
 afield. For that reason the ' phonetic laws ' of Latin proper 
 are often difficult to formulate with the same precision which 
 may generally be applied in Greek for example. Soldiers 
 and others speaking Umbrian or Oscan dialects would 
 naturally form a large proportion of a Roman army or 
 colony, and it might appear that insufficient weight is 
 usually allowed to this consideration, when the growth of 
 the ' popular ' Latin, as distinct from the literary, is in 
 question. 
 
 The history of Latin falls most naturally into three 
 periods. The first, or Old Latin, is mainly represented by 
 inscriptions dating from the fifth century B.C., and descending 
 to the latter part of the third. The differences between this 
 and the second or ' classical ' period consist in a heavier 
 pronunciation, sundry more archaic forms, and words which 
 became obsolete. The classical period is that of the well- 
 known Latin writers from the last decades of the third 
 century and onward for some 400 years. Meanwhile the 
 vulgar speech had come to diverge more and more widely 
 from the language of literature, until the latter was as 
 artificial or conventional as classical Sanskrit compared with 
 the Prakrit dialects. It is not, however, solely upon a decay 
 of the popular speech that the blame must be laid. The 
 literary language had itself courted distinction by means of 
 certain affectations, particularly of Graecisms. The serino 
 plebeius (or rfisticus) had, indeed, preserved many forms and 
 expressions ignored in the fashionable world of letters, a 
 fact which is clear to those who compare the language of 
 Petronius, for instance, with the language of Plautus. Even 
 in Ciceronian times the colloquial epistolary Latinity of the 
 Romans differs from the learned or rhetorical in much 
 the same respects, if in a smaller degree. Nevertheless the 
 vulgar speech had been anything but stable. Everywhere 
 it tended to phonetic decay and to grammatical analysis 
 {e.g. to the substitution of prepositional phrases for noun- 
 cases, and of an auxiliary with the infinitive or participle for
 
 200 Natural History of Language 
 
 a future or a perfect). It is this current language of the 
 people which formed the Neo-Latin diffused throughout the 
 provinces of the Roman empire, and destined to become the 
 parent of the Romance tongues. It is worth while to remark 
 that Neo-Latin is not to be confounded with ' Low Latin.' 
 The latter is a late and artificial hybrid, a jargon used with 
 a literary intention and containing words and forms more or 
 less invented pro re nata. 
 
 Spread over Italy, Sicily, Spain, Gaul, Dacia, and in a 
 considerable measure over other parts of the Roman world, 
 the Neo-Latin naturally acquired various peculiarities in 
 various regions, and this for two reasons. The imposing of 
 a language of conquerors on the conquered implies intimate 
 and continuous social intercourse. It was impossible for the 
 Latin to be thus gradually imposed on the earlier popula- 
 tions — Iberians, Celts, Dacians, etc. — of the several provinces, 
 without being itself acted upon in turn by their phonetic 
 peculiarities and occasionally by their vocabulary. In the 
 second place, inasmuch as the provinces were settled by 
 soldiers, traders, and others, coming from various parts of an 
 Italy full of dialects, there would frequently be a noticeable 
 difference in the Latin which was imported in the first 
 instance into two given districts. From these two causes 
 (though, no doubt, chiefly from the former), the Ling'ua 
 Romana soon possessed a special and characteristic Gallic, 
 Spanish, or other colouring. Nevertheless the ground- 
 material of the vulgar Latin of Gaul, Spain, etc., was the 
 same, and was differentiated from the classical Latin in the 
 same salient respects. A comparison of the Romance 
 tongues will show that, even before the diffusion, and before 
 the local and non-Latin influences had coloured the Latin of 
 each province, the popular Latin of the empire at large had 
 established an ' unclassical ' vocabulary and a more analytical 
 grammar of its own. It is clear, for instance, that while the 
 higher literature said equus (* horse '), iter (' journey '), os 
 (' mouth '), ignis (' fire '), coviedere (' eat '), the popular Latin 
 said caballus, vidtiemn, bucea, focus, ina?idficdre. Hence the
 
 General Survey of Languages 201 
 
 French cheval, voyage, boiichc, feu, manger, and the Italian 
 cavallo, viaggio, bocca, ftcoco, mangiare. Similarly the vulgar 
 Latin had developed the demonstrative pronoun ille, ilia into 
 a definite article (Italian il or to, la ; French le, la), and the 
 numeral adjective funis into an indefinite article. It had 
 confused the noun-cases both in use and pronunciation, till 
 it became necessary to replace them by prepositional phrases 
 {de, ad, etc.) ; it had substituted such expressions as habeo 
 habutum (French fai eu, Italian ho aviito) for the perfect 
 habul, and habere habeo (Fr.faurai, It. avrb) for the future 
 habebo ; and had otherwise corrupted or transformed the 
 grammar in an analytical direction. 
 
 Starting with this common material, the grammar and 
 pronunciation became increasingly broken down in each 
 province, as the language came in contact with such foreign 
 rivals as happened to exist in the particular locality. Thence 
 (as well as from special additions to the vocabulary derived 
 from special historical experiences) came those ever-increasing 
 local divergences which eventually developed into mutually 
 unintelligible languages — the Spanish, Portuguese, Provencal, 
 Italian, Roumanian, and other dialects. Thus the Latin 
 which was spoken by Roman colonists of all grades in 
 Spain among Iberians and Celt-Iberians, and which, after 
 some generations, came to be spoken by the Iberians and 
 Celt-Iberians themselves and by the mixed race of their 
 descendants, naturally acquired a more or less Ibericised 
 pronunciation, widely dissimilar from the Gallicised pronuncia- 
 tion which grew up among the ' Romans ' of Gaul. The 
 political and social (and in this case also literary) influence 
 subsequently exerted by the Moors led to a still further 
 separation from the Gallic rovidna, which, in its turn, was 
 subjected to influences from the Teutons. 
 
 Having thus led up to the formation of the several 
 Romance tongues, we may proceed to speak briefly of each 
 of these. 
 
 {a) Italian. — The sermo rilsticus of Italy, while pursuing
 
 202 Natural History of Language 
 
 the spontaneous development which would lead it away 
 from literary Latin, received a certain assistance in that 
 direction from the effects of the Gothic conquest. The 
 grammatical disintegration was in some measure quickened 
 by contact with the foreign tongue ; a series of Teutonic 
 words was adopted ; and, withal, the literary Latin was 
 partially reduced from its dignity as a conventional 
 standard. By the tenth century the speech of Italy 
 is distinctly modern. Nevertheless, however far the lingua 
 vulgaris may have advanced both in development and 
 vogue, it is not till the twelfth century that we meet 
 with a literature which dared to appear in the tongue 
 which was alive rather than in the one which was 
 petrified. One reason for this phenomenon is to be 
 found in the existence of numerous dialects, none of 
 which was pre-eminent in perfection of form or in material 
 importance. Dante enumerates fourteen of these, rejects 
 them all as literary standards, and seeks to erect into this 
 position an eclectic lingua aulica (or illtistris), an Italian 
 as artificial as the Ciceronian style was artificial Latin. 
 The language which has actually taken the place of standard 
 is ' Sicilian ' Tuscan. This form of Italian, less remote 
 from Latin than most of the local dialects, was first cultivated 
 at the mixed Court of Frederick II. in Sicily, and, after 
 being thence adopted by Italian poets generally, found 
 its most distinguished treatment (though with Tuscan 
 modifications) in the hands of such men as Dante, 
 Petrarca, and Boccaccio. Meanwhile the dialects of Italy 
 have retained a somewhat unusual independence of the 
 standard language of the grammarians. Not only are 
 they employed in the speech of educated Italians, but each 
 has produced its own literature. The chief dialects still 
 extant are : — 
 
 (i.) The Northern : Genoese, Piedmontese, Lombard, 
 Emilian, and Venetian ; (ii.) the Central : Corsican, Tuscan, 
 and Roman ; (iii.) the Southern : Sardinian, Neapolitan, 
 Calabrian, and Sicilian. Of these the central dialects are
 
 Gejieral Siirvey of Lmigtiages 203 
 
 the nearest to the antique type. At the present day Italian 
 is spoken in Italy proper, Corsica, the Swiss canton of 
 Ticino, the Southern Tyrol, Trieste, and along the coasts 
 of Istria and parts of Dalmatia. 
 
 ib) RhaetO- Romanic (otherwise called Rhaetian, Roic- 
 mansch, or Ladin) is a Neo-Latin language (not a mere variety 
 of Italian), spoken in the Swiss Grisons about the Upper Rhine 
 and Inn, in a part of the Austrian Tyrol about the Adige, 
 and in Italy about the river Tagliamento and thence east- 
 ward into the Austrian province of Goritz. The number 
 of persons speaking this tongue is about half a million, of 
 whom more than three-fourths are in the eastern (Friuli) 
 division. The name Ladin is sometimes more particularly 
 applied to the dialects of the Engadine Protestants, while 
 Roumansch is used of those of the Grisons Oberland. The 
 eastern division of Rhaeto-Romanic possesses inscriptions 
 of the twelfth century, while the oldest literary documents 
 proper are religious works of the sixteenth century. Three- 
 fourths at least of Rhaeto-Romanic words are of Latin 
 origin ; the rest are mostly derived either from Teutonic 
 or from old Rhaetian sources. 
 
 {c) Roumanian (less correctly Wallachian, a purely 
 relative title applied by the Germans in the sense of 
 ' foreign,' walk-), is the outcome of the popular Latin first 
 carried into Dacia (the modern Wallachia and neighbourhood) 
 by the colonies of Roman legionaries planted by Trajan at 
 the beginning of the second century A.D. The history of 
 the speakers of this liniba I'onianesca, whose territor}' has 
 varied at times between the north and the south of the 
 Danube, has been entangled with that of Teutons, Slavs, 
 Greeks, and Bulgars in ways which have left obvious traces 
 upon the language. In vocabulary it has been calculated 
 that over 40 per cent of words in literary use are Slavonic, 
 less than 3 5 per cent popular Latin, and the rest derived 
 from Turkish, Greek, and other sources. Some words are 
 probably old Dacian. A peculiarity of the language as 
 compared with the other Romance tongues is that the article
 
 204 Natural History of Language 
 
 is suffixed, as in Bulgarian. Thus omul ( = Latin Jio^no ille), 
 ' the man,' oc/iml ( — oailus ille), ' the eye.' 
 
 Roumanian is spoken by about ten miUions of persons, 
 of whom some nine millions form the northern or Daco- 
 Roiinianian division, in Roumania proper {i.e. Moldavia and 
 Wallachia), Bessarabia (in Russia), and, within the Austrian 
 empire, in a large part of Transylvania, Bukovina, the Banat 
 of Temesvar, and S.E. Hungary (to Arad), as well as in a 
 portion of N.E. Servia. Within this area dialectical differ- 
 ences are of little account. A southern division, which 
 possesses no literature, is formed by the Macedo- Roumanian 
 (or Vlacli) of parts of Roumelia, Macedonia, and Thessaly. 
 This has been much influenced by Greek and Albanian. 
 
 id) Provencal, the earliest of the Romance languages 
 to take a literary shape, was formerly known as the Langue 
 d'oc, while that of Northern France was the Langue d'otl, oc 
 (Lat. hoc) and oil later oui (Lat. Jioc illud) being their 
 respective words for ' yes.' Properly the speech of Provence 
 was but one dialect of a langue doc spoken over most of 
 Southern France, in Provence, Auvergne, Limousin, Gascony, 
 etc., as well as in N.E. Spain (Catalonia, with Aragon, 
 Valencia, and the Balearic Islands). The importance, 
 however, of the literature which was fostered by the Court 
 of Provence, under its Barcelonian counts, and which 
 flourished in the twelfth and thirteenth centuries as the 
 model for the troubadours and other poets of Western 
 Europe, caused the language itself to assume the name of 
 Provengal. Meanwhile the French of the north had been 
 growing in estimation ; the regions of the langue d'oc of 
 France became politically united to those of the langue d'oil 
 in the thirteenth century ; the Provence country itself was 
 devastated in the wars against the Albigenses ; and the 
 Provencal language thenceforward sank into a patois. In 
 a modern form it is still current in the far south-east of 
 France, and has recently received a share of literary 
 cultivation from patriotic litterateurs like Mistral. 
 
 That Provencal forms a natural connecting link between
 
 General Survey of Langtiages 205 
 
 Italian and French may be seen from such a sentence as 
 that which begins the Lord's Prayer, viz., /«z>^ nostre que 
 etz el eel, sanctificats sia vostre noms. On the surface the 
 language appears nearer to Italian. The fact is not so much 
 due to closer geographical connection as to the greater 
 rapidity with which the lingua roinana of Northern France 
 developed phonetic corruptions and grammatical changes 
 under pressure from various quarters. 
 
 Perhaps the earliest period at which Provencal may be 
 regarded as an established language is the ninth century, 
 beyond which its first productions — a popular poetry — 
 cannot be traced. 
 
 {e) Spanish. — The popular Latin, carried into Spain, 
 came into collision, in the first instance, with existing 
 speeches of the Iberians and Celt-Iberians, a remnant of 
 the former still holding its ground in the shape of Basque. 
 Certain provincialisms were the necessary consequence. 
 The subsequent conquest by the Visigoths produced little 
 effect, but the occupation of Spain by the Arabic-speaking 
 Moors, who were numerous and remained in more or less 
 influential intercourse with the Spaniards for nearly eight 
 centuries, led to a considerable modification of both phonetics 
 and vocabulary. In these respects, indeed, Spanish is 
 further removed from its Latin parent than any other 
 Romance tongue. Thus the pronunciation of j and g 
 (before e and i) is that of a strong so-called 'aspirate' 
 spirant, like the German ch, and the general tendency of 
 the language has been to harden rather than ease the 
 consonants. 
 
 The oldest traces of the language occur in glosses in 
 the Origines of Isidore of Seville, dating from the seventh 
 century, but there is nothing in the shape of Spanish 
 writing proper until the latter part of the twelfth. 
 
 Of the several dialects which occupy or have occupied 
 Spain, the recognised type for literature and society is that 
 of Castile. This is the official tongue for Spain itself, as 
 well as for those parts of South and Central America and
 
 2o6 Natural History of Language 
 
 the West Indies which were colonised by Spaniards 
 (Argentina and the La Plata countries, Chili, Peru, Bolivia, 
 Ecuador, Venezuela, Colombia, Nicaragua, Guatemala, 
 Mexico, Cuba, etc.). It is naturally heard in parts of the 
 Philippines. In the peninsula itself the speech actually 
 current in the east (Valencia) and north-east is Catalonian 
 (which once occupied Aragon also), and in the north-west 
 Galician, a tongue which goes with Portuguese. 
 
 {/) Portuguese (with which goes Galician), in view of 
 its peculiarities of grammar, phonetics, and vocabulary, is 
 not to be regarded as a mere dialect of Spanish, but as a 
 distinct language. Apart from the spontaneous divergences 
 in its natural development, it on the one hand received an 
 admixture of French words through the accession of a 
 French dynasty, and, on the other, was less affected by 
 Moorish influences. Its earliest records date from the 
 latter part of the twelfth century. 
 
 Portuguese is now spoken in Portugal, Brazil, the Azores, 
 Madeira, on parts of the African coast (east and west), and 
 at Goa in India. 
 
 {g) French. — Over the area now occupied by the French 
 language in Europe there were spoken, before the introduc- 
 tion of the lingua rustica of Latin, at least two altogether 
 different tongues, viz., the non-Indo-European dialects of the 
 ' Iberian ' or ' Euskarian ' type, of which the Basque of the 
 Hautes and Basses Pyrenees is a remnant, and also varieties 
 of the Celtic branch of Indo-European. The Celtic was 
 much the more widespread and important, being spoken by 
 both the Belgae and the Celtae of Caesar, Teutonic dialects 
 also enjoyed an irregular existence in the neighbourhood of 
 the Rhine. 
 
 How much modification the popular Latin may have 
 undergone through contact with these earlier denizens it is 
 impossible to tell. Philologists have little or nothing to say 
 of the Euskarian element. The Celtic left remarkably 
 few external traces upon the language of the Gallo-Romans, 
 the principal being a few contributions to the vocabulary. 
 
 I 
 
 4
 
 General Stirvey of Langiiages 207 
 
 e.g. (in their modern shape) lieue, cruche, pot, ruche, harnais. 
 Teutonic influences appear in a very sHght degree in the 
 fourth century, when Germans were permitted to settle in 
 Northern Gaul and many of them were serving with the 
 Gallo-Romans under the eagles. Early in the fifth century, 
 however, the invasion of German tribes (Franks, Burgundians, 
 Visigoths, etc.), and the subsequent domination of the Franks 
 in Northern France, led to the incorporation into the Neo- 
 Latin of a series of Teutonic words, chiefly connected with 
 war and feudalism. A long list of words, which in modern 
 French appear as senechal, inarechal, auberge, guerre, etc. 
 (from siniscalh, marahscaVi, heriberg, werra, etc.), came in at 
 this period in a form phonetically and flexionally adapted to 
 the ' Roman tongue ' as spoken in Gaul. As usual, these 
 collisions of languages caused no mixture of grammar, but 
 simply hastened the confusion of systematic flexion and the 
 tendency to analysis which were already proceeding spontane- 
 ously. By the seventh century a lingua romana widely 
 distinct from Latin prevailed in Northern France. Preachers 
 under the Franks use Latin, Romana, or German according 
 to circumstances, and in A.D. 8 1 3 the Council of Tours 
 orders that exposition shall be regularly made in the 
 ' Romance tongue.' 
 
 The earliest extant document for the determination of 
 the lingua romana, which was irresistibly gaining recognition, 
 is a list of glosses of Latin words found at Reichenau and 
 dating from A.D. j6^. Thus tugurium ('hut ') is glossed by 
 cabaujia (modern French cabane), galea (' helmet ') by helnio 
 (modern French heauine). A Romance poem on the 
 martyrdom of St. Eulalia belongs to the ninth century, and, 
 by the tenth, French (as the langue d'oil) may claim to be 
 an established language. It was not, indeed, homogeneous. 
 Northern France was politically divided under dukes or 
 counts into the four provinces of Normandy, Picardy, 
 Burgundy, and the He de France, each with its own dialect, 
 known respectively as Norman, Picard, Burgundian, and 
 ' French' The fact that the lie de France lay in the centre
 
 2o8 Natural History of Language 
 
 or key position, and that its dukes, the Capets, organised the 
 feudal centres into a monarchy with the capital at Paris, 
 caused the ' French ' of the lie de France to become the 
 special type of the Langue d'oU. By the thirteenth century 
 it was not only predominant in the north, but was rapidly 
 replacing the langue d'oc in the south. Indeed it was already 
 gaining something of that international currency which until 
 lately belonged to modern French, if we are to judge from 
 the significant fact that the Italians Marco Polo and Brunetto 
 Latini chose to write in it. None the less the Norman, 
 Picard, and Burgundian dialects continued to be in reputable 
 literary vogue down to the fifteenth century. They still 
 maintain themselves, but as patois only. For the philologist 
 their influence on the phonetics of standard French cannot 
 be ignored. The Wallon of Belgium, in its two varieties of 
 Liege and Namur, must be added to existing dialects of 
 French. 
 
 The title Old French may be conveniently applied to 
 the language between the ninth and thirteenth centuries. 
 Since that period there has been considerable manufacture 
 of learned words on Latin models, and many borrowings 
 have been made from Italian, particularly in the sixteenth 
 century, in consequence of much social and military inter- 
 course, Spanish, English, and other contributions are 
 obvious in the vocabulary. The pronunciation has been 
 continually cut down and lightened, and analysis, though 
 not carried so far as in English (especially in the verb), is 
 very far advanced. 
 
 At the present day French is spoken in all France 
 proper (with the exception of the lower Breton and the 
 Basque countr}^, and of the extreme south-east near the 
 Italian border), in a large part of Belgium, in the Swiss 
 cantons of Neuchatel, Vaud, Geneva, as well as parts of 
 Berne, Valais, and Fribourg, in part of Canada, in the west 
 of Hayti and in some other West Indian islands 
 (Guadeloupe, Martinique, etc.), in the French possessions in 
 Guiana, Senegal, in the islands of Bourbon and Mauritius, in
 
 General Survey of Languages 209 
 
 New Caledonia and a few spots in India {e.g. Pondicherry). 
 It may be expected to spread further in Algeria, Madagascar, 
 and Tonkin. 
 
 F. The Celtic Branch 
 
 The Celtic languages have continually lost ground in 
 historical times. From being spoken in portions of Central 
 Europe, Northern Italy, the greater part of Gaul, much of 
 Spain, in the British Islands, and in the centre of Asia 
 Minor (Galatia), they have become restricted to some three 
 millions of people in the north-west of Scotland, the west of 
 Ireland, Wales, the Isle of Man, and Lower Brittany. 
 Obsolete are the old language of the Gauls and, since 1770, 
 the dialect of Cornwall. 
 
 The division of the Celtic tongues till recently in vogue 
 was into two groups : (a) the GAELIC {Goidel or Gaedhelic), 
 including Irish -Gaelic (or Irish -Erse), Scotch -Gaelic (or 
 Scotch-Erse, also called simply either Erse or Gaelic in the 
 narrower sense), and Manx ; (/3) the CYMRIC, including 
 Welsh (or Cymric in the narrower sense), Copnish, Breton 
 (or Armorican), and Old Gaulish (or Gallic). 
 
 Of this list the Old Gaulish {Gallic) is known only 
 from personal and geographical names, words quoted by 
 ancient authors, coins, and about two dozen inscriptions, 
 most of which have been discovered about the middle of the 
 course of the Saone. A more exact examination of these, 
 though not very fruitful in positive results, leads to a 
 separation of Gallic from either of the two groups given 
 above. With it goes the Celtic of the Trocmi, Tectosages 
 and Tolistoboii of Galatia, which was carried thither by an 
 eastward wave of invasion from Gaul in 280 B.C. Jerome, 
 in the later part of the fourth century A.D., remarks that in 
 his day the popular speech of Galatia was similar to that of 
 the Treveri (at Treves or Trier) in Gaul, a district with 
 which he was intimately acquainted. This observation, 
 combined with one made by Sidonius Apollinaris late in the 
 
 (b 884) p
 
 2 1 o Natural History of L anguage 
 
 fifth century, to the effect that his brother-in-law Ecdicius 
 was to be congratulated on securing the substitution of Latin 
 for Celtic among the Arvernian nobility, shows that Gaulish 
 took several centuries before it was effaced by the Neo- 
 Latin speech. 
 
 The later and more satisfactory arrangement of the 
 Celtic languages is therefore : — 
 
 (a) Gallic 
 
 [i. Cymric (Welsh) 
 
 (/8) Britannic \ 2. Cornish 
 
 I 3. Breton (Armorican) 
 
 1 1. Irish-G-aelic 
 
 (7) Gaelic - 2. Scotch-Gaelic 
 
 [3. Manx. 
 
 Of the Britannic languages Welsh is the most im- 
 portant, being still spoken by about a million persons, 
 who show much zeal in the conservation and cultivation of 
 the tongue. Its earliest fragmentary records are glosses and 
 inscriptions from the eighth century, while from the eleventh 
 to the thirteenth century it possessed a flourishing literature 
 of chronicles, poems, the sententious ' triads,' etc. Of 
 Cornish a glossary dates from the twelfth century, and there 
 exists a mystery-play of the fifteenth. Its last speaker died 
 in 1770. The position of Breton (or Armorican) is 
 frequently misunderstood. No doubt the Celtic-speaking 
 people of old Gaul retained its original language most easily 
 in the Armorican corner of the country. Nevertheless the 
 Breton, as actually spoken, is another dialect, transported 
 from Cornwall and South Wales by Britons who emigrated 
 under pressure of the Saxons, from the fifth century onwards. 
 Its earliest documents are to be found in monastic registers 
 of the tenth and eleventh centuries. A literature proper 
 appears in the fourteenth. At the present day the language 
 is spoken in the department of Finistere and in the west of 
 two others, C6tes-du-Nord and Morbihan. 
 
 Of the Gaelic group Irish is the most important. Its 
 earliest traces appear in ' Ogham ' inscriptions of the fifth 
 
 II
 
 General Survey of Languages 211 
 
 century A.D. In Latin MSS. of the eighth century (now 
 preserved at Milan, Turin, Wiirzburg, etc.) marginal or 
 interlinear glosses have been made by Irish (then called 
 Scotch) monks. The ' Book of Armagh ' (ninth century) 
 contains similar notes. From the Middle Ages there comes 
 a copious literature of chronicles, legends, poems, and 
 ecclesiastical works, Ireland having already attained to a 
 special reputation for religious learning. Despite sundry 
 patriotic efforts, Irish is declining, but is still spoken by 
 over half a million people in the west and south-west of 
 the island. 
 
 Scotch-Gaelic (or Gaelic simply) is the language of over 
 two hundred thousand inhabitants of the north and north- 
 west of Scotland, with the Hebrides, and is regarded as 
 the furthest corrupted of the Celtic languages. Its chief 
 documents are the ' Fingal ' and some probably authentic 
 ' Ossian ' poems, but these were only put together at a 
 comparatively recent date {circ. A.D, 1530). 
 
 Manx, which once occupied the Isle of Man, is very nearly 
 extinct. 
 
 The Celtic languages are marked (like the French, over 
 the phonetic development of which the Gallic necessarily 
 exercised much influence) by the contraction and syncopation 
 of words, and by the usual progressive substitution of 
 prepositional analysis for the flexional noun -declension 
 which had become hopelessly corrupted. 
 
 G. The Teutonic Branch 
 
 In order to spare words, the inter-relations of the 
 Teutonic (or Germanic) tongues, as ascertained by the most 
 recent research, may be represented most simply in the 
 following synoptic table ^ : — 
 
 ^ An asterisk denotes a stage without extant records, but to be postulated.
 
 212 
 
 Natural History of Language 
 
 
 
 ^cn 
 
 
 C I' 
 
 
 ^-^^ 
 
 
 a 
 
 CP-, 
 3 O 
 
 o •- 
 
 
 a
 
 General Survey of Languages 213 
 
 It is impossible to say how remote in prehistoric times 
 may He a stage of the Teutonic speeches at which they were 
 homogeneous enough to form a single language (the hypo- 
 thetical Primitive Teutonic). That they once did so, or at 
 least were sufficiently homogeneous for us to regard them in 
 that light for theoretical purposes, is manifest from the facts 
 that (i) the farther back we investigate the several languages, 
 the nearer they converge to a common type in respect of 
 vocabulary, grammar, and phonetics ; (2) this common type 
 is distinguished from all other branches of Indo-European 
 by the phenomenon of ' sound-shifting ' [Lautversdiiebung) 
 discovered by Rask and Grimm, and made familiar under 
 the title of ' Grimm's Law.' All the Teutonic speeches, and 
 these alone, share in this systematic displacement of the 
 explosive consonants. 
 
 It remains to comment upon the above table in detail. 
 
 Gothic (or Gotic), as known to us, is the Moesogothic of 
 the Visigoths, when situated south and south-east of the 
 Carpathians. The language of the Ostrogoths of Southern 
 Russia is without records. There can, however, have been 
 little difference between the two, since the Bible of Ulfilas, 
 bishop of the Goths in Dacia and Moesia (Roumania and 
 Bulgaria), was in use among the Ostrogoths when they 
 were subsequently in occupation of Italy. The question 
 of Gothic history is not pertinent here, except in so far as 
 the language, like the tradition, points to a close neighbour- 
 hood of the Goths with the early Scandinavians, before the 
 former migrated from the Baltic and East Prussian regions 
 somewhere about 200 A.D. 
 
 Except for one or two unimportant Gothic deeds of the 
 time of Theodoric and a fragment of a calendar, preserved 
 in Italy, practically the sole record of the language consists 
 of the portions of the Bible of Ulfilas extant in the Codex 
 Argenteus (a manuscript copied about A.D. 500), which 
 found its way via Prague to Upsala. These include con-
 
 2 1 4 Natural History of L anguage 
 
 siderable parts of the four Gospels, fragments of the 
 Pauline Epistles, and also of the books of Ezra and 
 Nehemiah. Ulfilas (or, in Gothic, Wulfila) was apparently 
 born in A.D. 311 and died in the year 381, and his Bible 
 is thus, by at least three centuries, the earliest appreciable 
 document for the history of Teutonic. 
 
 Gothic, carried by conquest into Italy and Spain, soon 
 disappeared entirely from both countries. It lingered about 
 Kustendje, on the Roumanian coast of the Black Sea, till 
 the ninth century, and then became extinct, except for a 
 dialect of Crimean Goths, which was noticed as late as the 
 sixteenth century, but of which no document remains. 
 Gothic supplies us with a Teutonic language at a stage 
 still largely synthetical. It still preserves even a dual (at 
 least in the verb and pronoun), has a declension of several 
 cases for noun, pronoun, and adjective, and is rich in those 
 distinct formative elements which have become phonetically 
 much reduced and confounded by the time when we first 
 make acquaintance with Anglo-Saxon or Old High German. 
 The light thus thrown upon the morphology of Primitive 
 Teutonic is of the greatest value ; but in point of syntax 
 allowance has to be made for an imitation of Greek idiom, 
 which was induced both by the language of Wulfila's original 
 and also by the influence of the standard tongue of the 
 Church of Constantinople. 
 
 It is probable, but not provable, that the Norse (or 
 Scandinavian) languages were carried into the northern 
 peninsula partly in the easterly direction round the Gulfs 
 of Finland and Bothnia and partly in the westerly through 
 Denmark, as well as across the Baltic. The existence of 
 any homogeneous primitive Norse cannot be absolutely 
 demonstrated, but there can have been little divergence 
 during the earlier epoch of the Runic inscriptions, some of 
 which are not later than the fifth century A.D. Up till the 
 Wiking epoch (A.D. 800-1000) the Scandinavian languages 
 may fairly be regarded as one. They share in a number 
 of peculiar phonetic developments, in the sufHxed article,
 
 General Survey of Languages 2 1 5 
 
 and in such formations as a medio-passive with suffix -sk 
 { — sik, Germ, sick, 'oneself'). The language at this period 
 is the Old Norse or Dansk {donsk timgd) of the Skalds, and 
 has been best preserved in documents of Iceland, whither 
 it was carried from Norway at the emigration of the year 
 874. It is in Iceland also that there have been preserved 
 the old mythological popular lays (the verse ' Edda '), and 
 the more artificial panegyric poetry of the Skalds. Whether 
 these compositions ever belonged in anything like the same 
 measure to the mainland is doubtful. They began to be 
 collected early in the twelfth century, but naturally represent 
 a considerably earlier condition of the language, in some 
 cases probably reaching back three hundred years. The 
 younger prose Edda, or ' Snorra-Edda,' the mythico- 
 historical and historical saga-literature, dates from early 
 in the thirteenth century. 
 
 From about icoo A.D. begins a division of the 
 Danish- Swedish from the Norwegian - Icelandic branch. 
 Danish proper possesses no records beyond the law-books 
 of the end of the thirteenth century, and an orthodox 
 national speech, founded on the Zeeland dialect, is first met 
 with in the sixteenth. Foreign influences have introduced 
 many new words into the vocabulary, there has been great 
 phonetic weakening, and the simplification of flexion has 
 followed far in the direction in which English has led the 
 way. Besides being the language of Denmark, Danish is 
 spoken in North Schleswig and, in consequence of the 
 former political union, has been much employed (though with 
 local differences) in the literature and higher society of 
 Norway, while Norweg'ian retains its independence in the 
 country and in popular use. Swedish, which has maintained 
 a more archaic Norse character than the Danish (including 
 a ' pitch ' or ' musical ' accent), though it exhibits the same 
 tendency to the unification of its declensions and of its 
 conjugations, is spoken in Sweden proper, and also along 
 the Finland coast both of the Gulf of Bothnia and of the 
 Gulf of Finfend (at Helsingfors and elsewhere!
 
 2 1 6 Nahiral History of Language 
 
 The West Teutonic languages would now be naturally- 
 divided in the first instance into the Hig"!! German and Low 
 German branches, distinguished from each other by a special 
 development on the part of the former in regard to the 
 general Teutonic consonantal sound-shifting {LautverscJiie- 
 btmg) described under ' Grimm's Law.' This second sound- 
 change, however, on the part of the Upper (or South) Germans 
 was not accomplished before the seventh century, when the 
 Anglo-Saxons had already been for two hundred years in 
 occupation of Britain. Up till the date of this salient 
 differentia it is best to speak of the existence side by side 
 of a number of closely cognate West German speeches, the 
 chief of which can be conveniently grouped under the 
 headings Prankish, Frisian, Saxonie, and Upper German. 
 It is not pretended that each of these divisions was 
 altogether homogeneous in itself, but merely that a close 
 relation subsisted between the (probably numerous) dialects 
 included in each. 
 
 (a) The Prankish stock extended from the Rhine 
 Palatinate and the neighbourhood of the Main as far as 
 the Netherlands, and, in respect of linguistic development, 
 fell in course of time into three sections, viz. the North 
 {Netherlandish), the Middle {Lozver Rhenish or Middle 
 Frankish), and the South {Upper FrankisJi). Of these the 
 Upper portion came to share in the second sound-shifting 
 of its southern neighbours, and is now part of the High 
 German. The Middle, reaching from the Moselle to 
 Diisseldorf, is a dialect which is properly neither High nor 
 Low, but partakes of the characteristics of each. Farther 
 down the Rhine the Old Low FrankisJi (of which a trace 
 is left in a fragment of a translation of the Psalms from 
 the ninth century) was unaffected by the Upper German 
 influences, and was left to develop itself into the speeches 
 of the Netherlands. The Netherlandish literature begins 
 in the thirteenth century and for a time shows three 
 varieties, the Dutch, Flemish, and Brabant. The first- 
 named, however, prevailed from the sixteenth century,
 
 General Survey of Languages 217 
 
 and at the present day the difference between the speech 
 of Holland and that of Flanders is mostly one of pronuncia- 
 tion only. The population of Belgium speaking Flemish 
 numbers about two millions and a half. Dutch is the 
 language of most of Holland, of a large proportion of the 
 European inhabitants of South Africa, and is naturally to 
 be met with in the Dutch East Indies and in Guiana. 
 
 (yS) Frisian is not discoverable before the legal records 
 of the thirteenth or fourteenth century. It possessed three 
 dialects, the West Frisian in Northern Holland, the East 
 Frisian between the mouths of the Ems and Weser, and 
 the North Frisian northward of the mouth of the Elbe. 
 At the present time Frisian lingers in the north-west of the 
 Dutch province of Friesland, in a few spots in the German 
 Oldenburg from which it has not yet been driven by 
 Platt-Deutsch, and (in a more or less mixed state) on the 
 west coast of Schleswig and in the North Frisian Islands. 
 
 (7) The Old Saxonic dialects lay to the south of the 
 Frisian and east of the lower and middle Frankish, over 
 an area roughly corresponding to the modern Westphalia, 
 Hanover, and as far as Schleswig-Holstein. Among peoples 
 speaking dialects of this division were the tribes of Saxons, 
 Angles, and Jutes who invaded Britain in different bodies 
 from about the middle of the fifth century A.D. and for a 
 hundred years and more thereafter. The Saxons settled 
 in Wessex, Sussex, Middlesex, and Essex, the Jutes in 
 Kent and the Isle of Wight, the Angles in East Anglia 
 (Suffolk, Norfolk, and Lincolnshire), Mercia, and East 
 Northumbria, including the eastern half of Southern 
 Scotland. 
 
 On British soil flourished their various dialects, generally 
 comprehended in the term Anglo-Saxon {i.e. Saxon of 
 England), of which, however, there developed three main 
 divisions, the Northern, Midland, and Southern, otherwise 
 known as the Northumbrian, Mercian, and W^essex ^Anglo- 
 Saxon ' in the narrower sense. The extant records belong 
 almost wholly to the last, the dialect of Wessex, whereas
 
 2 1 8 Natural History of Language 
 
 modern English is more directly derived from the Anglian 
 of the East Midland (or East Mercian) division, where 
 the capital came to be located, where the chief literature 
 was afterwards produced (as by Chaucer), and where the 
 current speech formed a connecting link between its 
 neighbours on either hand, so that it naturally acquired 
 a wider vogue. Since Old Mercian literature, like Old 
 Northumbrian, has preserved hardly any material in its 
 original shape, it is customary to ' derive ' English words from 
 the only Anglo-Saxon which is extant, namely, that of the 
 Wessex documents, into which the Northern compositions also 
 were rewritten after the Danish conquests had driven all 
 Anglo-Saxon culture to the south. But, strictly speaking, 
 this ' derivation ' is for the most part only a process of 
 comparing modern words which are Neo-Mercian, not with 
 their ancestral Mercian, but with words in the Wessex 
 dialect, which was indeed closely akin to Old Mercian 
 but not identical with it. That the ' derivation ' is not 
 strictly exact may be seen from the following diagram : — 
 
 Anglo-Saxon 
 
 Wessex Mercian Northumbrian 
 
 ('Anglo-Saxon') | | 
 
 as ordinarily quoted Modern Scottish 
 
 English Dialect 
 
 Many words from the Wessex division were doubtless 
 imported into the Mercian both in the pre-literary and also 
 in the literary period of Midland predominance, and many 
 Wessex words would in any case be precisely the same 
 as those of Mercian. Nevertheless there are important 
 differences, particularly of pronunciation, and in these 
 matters it is the Midland speech which has determined the 
 later English. 
 
 In view of the confusion which results from the use of 
 ' Anglo-Saxon ' thus applied, first to a number of dialects 
 and then especially to one, it is not in idle pedantry that
 
 General Survey of Languages 2 1 9 
 
 the title Englisc has been proposed for the period preceding 
 that of a recognised Enghsh proper, so as to embrace the 
 several dialects in question. The expression ' Old English ' 
 is better reserved for another application. ' First English ' 
 has been suggested with some support. 
 
 The earlier Celtic-speaking occupants of Britain left little 
 trace upon the language of their conquerors, the few Celtic 
 words now existing in English being mostly of later adoption 
 from Welsh, Irish, or Gaelic-Scotch. Of more effect upon 
 the vocabulary were the new elements introduced from Latin 
 through ecclesiastical channels after the Christianising of the 
 country. More considerable still were the contributions of 
 words and the modifications of phonetics made by the 
 ' Danes ' from the ninth to the eleventh century, their 
 influence being strongest in the north. Most important was 
 the effect of the Norman-French conquest of A.D. 1066, 
 after which French became the language of the Court, 
 Parliament, the higher society, law, literature, education, and 
 largely of business. The number of French words intro- 
 duced was very great, and, though there was no borrowing 
 of French grammar, one result of the conflict of the tongues 
 was that the old English flexion — which had already begun 
 to give way, partly through natural development, partly 
 through the collision of the dialects with one another and of 
 all alike with the speech of the Danes — was rapidly broken 
 down. 
 
 Engflish in the proper sense may be dated from about 
 the year iioo, when the flexional terminations of words 
 began a phonetic weakening by which they were doomed. 
 An Old English period may for convenience be said to end 
 in A.D. 1350 — by which time the Midland dialect had 
 established itself as English par excellence — and a Middle 
 English about the year 1460. These dates, however, or 
 any others, are so far from representing clear-cut lines in the 
 history of the language that different divisions and titles are 
 favoured by diff'erent philologists. Modern English may 
 fairly be considered to have begun by the time {citr. A.D.
 
 2 20 Natural History of L anguage 
 
 1460) when the adjectival flexion had disappeared, when the 
 noun flexion had reduced itself to one optional form for a 
 possessive case and to a single prevailing form for the plural 
 (in j), when the flexion of verbs in respect of persons and 
 moods was almost lost, and when the French words borrowed 
 since the Conquest had taken their place in habitual popular use. 
 Since that epoch a large number of Latin words have been 
 added with the ' Revival of Learning,' Greek technical terms 
 and foreign vocables from many sources have been and are 
 being adopted into the vocabulary, much phonetic change 
 has occurred, and many words have greatly shifted their 
 meaning. The consequence of its history is that English is 
 one of the most composite as well as one of the most 
 analytical of tongues. Nevertheless the few remnants of its 
 grammar, like the bulk of its everyday vocabulary, are still 
 distinctly Teutonic. 
 
 The modern English dialects fall into much the same 
 three groups as the Englisc. The most pronounced is the 
 ' Northumbrian ' of the Lowlands of Scotland, a region in 
 which the local peculiarities were long encouraged by 
 political separation. The dialect still possesses a certain 
 independence, although — despite a revival through the genius 
 of Burns — it no longer enjoys a literary standing. The 
 ' broken ' vowels of Anglo-Saxon proper are still heard in 
 Wessex. 
 
 The Old Saxon, which was meanwhile left to be spoken 
 upon the Continent, is chiefly represented by the Heliand 
 ('Saviour'), a Christian poem preserved in two MSS. of the 
 ninth century and composed apparently in Westphalia. By 
 the thirteenth century Saxon began to develop a literature 
 (the Middle Low German), and books were printed in it till 
 the seventeenth. A Bible in Low German is dated 1621. 
 Since that time the High German has become the sole 
 literary language, although the Saxonic dialects, under the 
 name of Nieder-Deutsch or Piatt- DeiitscJi are still in 
 regular use among the populace of North Germany, not 
 only in the older Saxon districts of Westphalia, Hanover,
 
 General Survey of Languages 22 i 
 
 and Lower Saxony, but eastward in Brandenburg, Mecklen- 
 burg, Pomerania, etc, from which regions they have ousted 
 Slavonic and Lettic speeches. 
 
 (S) Hig"!! German, which is now the literary and cultivated 
 language throughout the German empire and in a large 
 part of the Austrian, is the outcome of various dialects 
 spoken in Upper or South Germany. The Southern Frankisli, 
 which was current in the Palatinate and along the Main, 
 has already been dealt with. Others were the Alamajinic, 
 the Swabian, and the Bavarian. All these still survive : the 
 Alaniannic in Alsace, the South of Baden, and the German- 
 speaking parts of Switzerland ; the Sivabian in Wiirtemberg 
 and Western Bavaria ; the Bavarian in Bavaria, the German- 
 speaking Tyrol, Salzburg, Austria proper, North Styria, North 
 Carinthia, and various other districts of the Austrian empire. 
 These dialects were all affected by the second consonantal 
 ' shifting ' of the seventh century A.D. 
 
 The history of the established High German is somewhat 
 complicated. It is conventionally divided into three periods : 
 (i.) the Old Hig-h German (eighth to twelfth century), as 
 represented by the earliest texts, such as glossaries, charters, 
 Biblical fragments, epic fragments, and the Strasburg Oaths 
 (a.D. 842). There is properly no common German of the 
 period, the dialects above mentioned being independently 
 recognised ; (ii.) the Middle High German, in which the 
 varieties of vowel in the noun and verb flexion are reduced 
 to a uniform c, and an orthodox literary speech is being 
 formed, mostly — as it would appear, though the fact cannot 
 be demonstrated — on the basis of the Swabian dialect as 
 modified by the German Chancellors and in Court circles. 
 To this period belongs the redaction of the popular epic of 
 the Nibelungen and a quantity of Minnesinger verse ; (iii.) 
 the Modern High German, which no longer recognises 
 dialects, nor represents the natural development of any 
 special dialect, but is in its origin a more or less eclectic 
 language, intended to secure currency everywhere among the 
 varying speeches of Germany. This object it has in a large
 
 2 2 2 Natural History of L anguage 
 
 measure attained. Great assistance was lent to the spread 
 of this form of German by Luther's translation of the Bible. 
 The notion that Luther created the language itself is as 
 erroneous as the fancy that Chaucer created English or 
 Dante Italian. He expressly states that his translation is 
 in no precise dialect of German, but is the speech of the 
 " Saxon Chancery," a conventional form " for all the princes 
 and kings in Germany." 
 
 Some Low German words have been borrowed into 
 High German ; loan-words from the Roman civilisation are 
 frequent in a phonetically adapted form ; French also has 
 exercised a certain effect on the vocabulary. For the most 
 part, however, despite Romance influences, High German 
 has preserved a strictly Teutonic character and material. 
 Its grammatical flexion still remains more elaborate than in 
 Low German, than in English most of all. Its word-order 
 exhibits a more archaic character, rendered possible only by 
 the existence of its flexional terminations. Its phonetics 
 exhibit less decay. It may be added that High German is 
 particularly marked by a liking for the combination of 
 whole words or word-stems into compounds, often of un- 
 wieldy length. 
 
 H. The Baltic-Slavic (of Slavo-Lithuanian) Branch 
 
 (a) The Slavonic languages. 
 
 Eastward of the East Teutons in the earliest historical 
 times are to be found the Veneti or Venedae (Wends). 
 Without pretending to any very precise determination of 
 the ethnological and linguistic application of the term on 
 each occasion of its use in early writers, we may take it as 
 representing approximately the contemporary communities 
 of Slavs. The westward and southward migrations of the 
 Teutonic peoples — the Goths, Vandals, Burgundians and 
 others — made room for an inroad by the Wends into the 
 country thus left comparatively empty. When the move- 
 ment had once begun, it was carried on so vigorously on
 
 General Survey of Languages 223 
 
 the part of the Slavic peoples that by the fifth century they 
 were in possession of Moravia, Bohemia, and of Germany 
 as far as the Elbe and Saale, In the sixth they overran 
 Moesia, Thrace, Macedonia and Greece as far as the 
 Peloponnese. Of these regions Moesia (Bulgaria) was 
 made completely Slavic. From Moravia Slavs advanced in 
 the same century into Pannonia (Austria, Styria, Carinthia, 
 etc.) under the name of ' Slovaks.' In the seventh 
 century the * Croats ' and ' Serbs ' crossed the Danube 
 from the Carpathians and occupied the Dalmatian parts 
 of Illyricum, the modern Servia, Croatia, Bosnia and neigh- 
 bourhood. 
 
 It thus happens that in the eighth and ninth centuries 
 Slavonic speeches occupied Pomerania, Mecklenburg, Bran- 
 denburg, Saxony, and the greater part of Austria, as well as 
 the territories — Russia, Poland, Galicia, Bohemia, Moravia, 
 S. Carinthia, S. Styria, Carniola, Slavonia, Croatia, Bosnia, 
 Herzegovina, Montenegro, Servia, Bulgaria — which they still 
 hold. The cities of Berlin and Vienna both stand where 
 Slavonic was then spoken, and their present position and 
 environment afford conspicuous testimony to the fact that 
 for several centuries German has been steadily pressing 
 back and eating into the frontiers of both the Slavic and 
 Lithuanian tongues. 
 
 A statement of the exact inter-relations of the Slavonic 
 languages is not yet agreed upon. The oldest recorded 
 forms are those of the Old Bulg-arian (or Church Slavonic), 
 a language in which were written the translation of the Bible 
 and sundry liturgical pieces by Cyril and Methodius, who 
 were " Apostles to the Slavs " in the middle of the ninth 
 century. The tongue employed by these missionaries would 
 naturally be that which was most typical for their Slavic 
 disciples in Moravia and subsequently in Bulgaria. Cyril 
 died in A.D. 868, and a hundred years later a copy from his 
 MSS., containing the Gospels and Epistles for the Sundays 
 after Easter (' Ostromir Gospels '), was made for Russian use. 
 This again was transcribed in 1056, with certain Russian
 
 2 24 Natural History of Language 
 
 elements introduced, and it is from the latter MS. that our 
 knowledge of an old typical Slavonic is mainly derived. 
 The title ' Church Slavonic ' is applied to the language from 
 its use in the church service ; that of ' Old Bulgarian ' is 
 misleading, inasmuch as the language was spoken much 
 more widely than in Bulgaria, especially to the West. 
 Modern Bulgfarian, thanks to much collision with other 
 languages, is the farthest removed of all the Slavonic 
 speeches from the stage represented in Church Slavonic. It 
 has, for instance, lost almost all the declension-flexion and 
 many forms of the verb, suffixes the article (like its non- 
 Slavic neighbours, Roumanian and Albanian), and possesses 
 a vocabulary greatly mixed with Greek, Turkish, Roumanian 
 and Albanian elements. 
 
 The classification of the Slavonic tongues which appears 
 to find most favour with students in that branch is as 
 follows : — 
 
 ( -D • \ Great Eussian 
 
 a. Russian' -„.. „ 
 (Wnite Russian 
 
 b. Little Russian 
 
 c. Bulgarian 
 
 , Til • (Servo-Croatian 
 ./. niyrian|g^^^^^.^^ 
 
 ie Tzech /Tzech proper 
 r- ^^^^^ Isiovakian 
 (ii.) West Group ^ g^^^.^^ ^^^ Sorabian) 
 
 . , fPolish 
 1^^. Lecnisn |poiabisli (Elbe-Slavonic) 
 
 It will suffice to make a note upon each of these in 
 order. 
 
 i. {a) Great Russian (or Russian proper) is a language 
 which received no authoritative formulation till the beginning 
 of the eighteenth century, though it possesses records from 
 as far back as the eleventh. Although on the way to 
 analysis, it still possesses a declension of seven cases, and 
 also a liberal amount of verb-flexion. Great Russian is 
 spoken by perhaps sixty millions of persons in Central 
 Russia and is met with in its purest form in the neighbour- 
 
 (i.) East and South 
 Group
 
 General Survey of L angziages 225 
 
 hood of Moscow. White Russian is a dialect of some three 
 millions in Western Russia to the east of Lithuania and 
 Poland, chiefly in Vitebsk, Mohilev, and Minsk. 
 
 {b) Little Russian (or Ruthenian, a name particularly 
 applied to the language as spoken in Galicia) is spoken in 
 the southern parts of Russia (including, and south of, the 
 provinces of Volhynia, Kiev, and Kharkov) as well as in 
 the Austrian province of Galicia. 
 
 {c) Bulg"arian has already been touched upon. Beyond 
 Bulgaria proper it is spoken in parts of European Turkey, 
 in the places where Albanian, Greek, Vlach or Turkish 
 are not used. 
 
 {d) Servo-Croatian, of which there are many, but not 
 widely divergent, dialects, is the speech used in Servia and 
 South Hungary (Zombor), Slavonia, Croatia, Bosnia, Herze- 
 govina and Montenegro, as well as in Dalmatia and Istria 
 (except for the parts, mostly on the coast, where Italian is 
 spoken). It possesses records from the twelfth century 
 and is philologically of much importance. Slovenian 
 is the language of Carniola, South Carinthia, and South 
 Styria. It possesses linguistic records as early as the tenth 
 century. 
 
 ii. {e) Tzech, which was distinguished by a flourishing 
 literature down to the time of the Hussite wars of the 
 fifteenth century, is the language of Bohemia (except on 
 its western and northern margins) and Moravia. Its earliest 
 traces are to be found in the ninth century, although it is 
 not till the thirteenth that any important document is forth- 
 coming. The revival of the Tzech language and literature 
 belongs to the last hundred years. Slovakian may be 
 regarded as a dialect of Tzech, and is the tongue of about 
 two millions of the population of Northern Hungary, south 
 of the Moravian and Galician borders, between Pressburg 
 and the Carpathians. 
 
 {/) Sorbian {Sorabian or Wendic), closely allied to the 
 Tzech, is gradually becoming extinct. It is still spoken by 
 perhaps 100,000 German subjects in Prussia and Saxony 
 
 (b 884) Q
 
 2 26 Natural History of Language 
 
 along the course of the river Spree. Its earliest document 
 is a prayer-book of i 5 i 2. 
 
 {g) Polish is the tongue of a population of ten millions, 
 divided between Russia, Austria (West Galicia), and Germany 
 (Posen and parts of Silesia). From Germany it is being 
 ousted by the national language. The earliest document of 
 Polish is a translation of the Psalms of the year 1290. 
 Polabish, once spoken by Slavs on the lower Elbe, is now 
 extinct. 
 
 (;8) The Baltic (sometimes called Lettic) languages. 
 
 Three languages, the Old Prussian (extinct since the 
 seventeenth century), Lithuanian, and Lettish, form a 
 special Baltic division of the larger Slavo-Lettic branch. 
 Old Prussian was once in use along the Baltic coast between 
 the Vistula and the Niemen. A German-Prussian glossary 
 of the early fifteenth century and three catechism.s of the 
 sixteenth, represent practically all that is known of this 
 tongue. Lithuanian is spoken by perhaps 150,000 persons 
 in the extreme north-east of Prussia (about Tilsit and the 
 river Niemen), and by a million and a quarter in the adjoin- 
 ing Russian provinces of Kovno and Vilna. In its forms it 
 is the most archaic of the living Indo-European speeches, 
 and is therefore of special importance to the philologist. 
 Its documents, unfortunately, go back no further than the 
 sixteenth century, and the only important specimen of 
 Lithuanian literature is the poem of the ' Seasons,' by 
 Donaleitis, dating from the middle of the eighteenth. 
 Lettish, which has corrupted its forms more than Lithuanian, 
 is the language of about a million people in Courland, South 
 Livonia (about Riga), and along the lower Dvina.
 
 SYNOPTICAL SCHEME OF THE INDO-EUROPEAN LANGUAGES 
 
 The following synoptical scheme of the Indo-European tongues may be of use for convenient reference. 
 Italics imply that the groups so marked are hypothetical steps in the pedigree, warranted by induction! 
 An asterisk only is prefixed to dead languages, while an asterisk combined with an obelisk signifies that, if the 
 speech in question can be regarded as still alive, it is only as one artificially cultivated for literary or liturgical 
 purposes. An obelisk alone indicates that the speech is now regarded as merely a dialect of the recognised 
 national language. 
 
 r^,r^,^^r *\7^A- c I •» ,-*|Classical Sanskrit. 
 -Indian — VedicSansknt — j.' ., ., ,^,,,.,. 
 
 l-*Prakrits (*tPali, etc. 
 
 Aryan^ — 
 
 -Eranian- 
 
 -*01dBactrian(Zend). 
 
 -*Pehlevi (Huzvaresh). 
 -*tParsi (Pazend). 
 
 Assami, Bengali, Uriya. 
 Hindi, Hindustani. 
 Marathi, Gujarati, Punjabi. 
 Kashmiri, Nepali. 
 Gipsy (Romany). 
 Pushtu. 
 
 Armenian- 
 
 Hellenic 
 
 III yrian 
 
 -Modern Persian. 
 
 Ossetian. 
 
 Kurdic. 
 -Belilchi. 
 
 Modern Armenian. 
 
 Old Persian. 
 
 fold Armenian 
 
 -* Doric. 
 * A colic. jt„ 
 
 *Iono-Attic *The Koine r Byzantine-(Romaic)-Modern Greek 
 
 I- Alexandrian — Hellenistic. 
 
 I TALI c- 
 
 -*N. W. Greek. 
 
 -Northern. 
 
 -Southern — Albanian. 
 
 |-*Oscan. 
 
 -Umbro-Samnitic — -*Umbrian. 
 
 '-*Volscian, *Marsian, etc 
 r-*tClassical Latin. 
 "L*Popular Latin 
 
 -Itahan 
 dialects 
 
 'Old Latin- 
 
 Cel tic- 
 
 -*Gallic. 
 -Britannic- 
 
 -Gaelic- 
 
 Teutonic- 
 
 -East Teutonic-- 
 
 ■ West Teutonic - 
 
 ''Gothic. 
 
 -Norse - 
 
 i-East Norse- 
 
 ■ IVestNorse- 
 
 i-fSicilian. 
 — I-Tuscan — Italian. 
 -fNorthern. 
 
 -*Langue d'oc tProven9al. 
 
 -fCatalonian. 
 
 -Spanish dialects — Spanish 
 
 (Castilian). 
 -Portuguese. 
 
 -Langue d'oil dialects-French. 
 -Rhaeto-romanic. 
 -Roumanian. 
 
 I Welsh (Cymric). 
 -Breton. 
 
 L*Cornish. 
 
 i-Irish. 
 -Manx. 
 
 '-Scotch-Gaelic. 
 
 [-Swedish. 
 
 '-Danish. 
 
 i-lcelandic. 
 
 '-Norwegian. 
 Frisian. 
 
 .- Old Frisia n 
 
 i-*Anglo-Sa.xon-(Mercian)-English. 
 \-Old Saxon — dialects. 
 
 - Continental Saxon- 
 |-0/^/-ra«/62V^|-i;t*;^herlandish 
 
 i-tUpper Prankish. 
 '-tSouth German dialects 
 
 SlA yo- LlTHUA NIA N- 
 (BALTlC-SLAyiC) 
 
 -Slavonic — 
 
 - Baltic - 
 
 S. &=■£._ 
 Slavonic 
 
 -W. Slavonic - 
 
 -*i-Church Slavonic. 
 
 -*01d Prussian. 
 
 -Lettic. 
 
 -Lithuanian. 
 
 — tPlatt-Deutsch. 
 —Dutch. 
 -Flemish. 
 —German. 
 
 —Russian. 
 l-fWhite Russian. 
 — Ruthenian 
 
 (Little Russian). 
 -Bulgarian. 
 i-Servo-Croatian. 
 I-Slovenian. 
 -Tzech. 
 -Slovakian. 
 -Serbian. 
 -Polish. 
 -*Polabish. 
 
 There are also various dialects of the Pamir region not here enumerated.
 
 CHAPTER X 
 
 RACE AND LANGUAGE 
 (WITH SPECIAL APPLICATION TO THE INDO-EUROPEAN 
 
 TONGUES) 
 
 (i.) In Gejteral 
 
 The foregoing classification of languages into their fannilies, 
 branches, groups, and sub-groups, as hitherto ascertained, has 
 been accompanied by a more or less detailed statement of 
 the localities in which they are or have been spoken re- 
 spectively. It should, therefore, already have become 
 apparent that we cannot regard kinship of language as any 
 necessary criterion of kinship of race. In other words, 
 peoples which now speak languages genealogically alien to 
 each other may be, in no inconsiderable measure, ethnologi- 
 cally akin, while peoples which now speak cognate or even 
 identical tongues may be descended from very distinct races. 
 If such lax expressions as 'the Indo-European peoples,' 
 ' the Aryan race,' ' the Latin race,' ' the Celtic peoples,' do 
 not come of a confusion of thinking, they undoubtedly lead 
 to such a confusion. On the one hand, they confound 
 ethnology, and on the other, they prevent the student of 
 philology from readily apprehending one of the chief causes 
 of local variation in speech. 
 
 It must be evident from what has been said of the 
 spread of a given tongue and the decline of all others within 
 a certain area, that the prevalence of a language within 
 that area is principally due to political and social relations. 
 
 22S
 
 Race and Language 229 
 
 Experience shows that the poHtical and social organisation 
 called a nation, if it retains the same geographical borders 
 for a period sufficiently long, will develop a common national 
 speech. This will ultimately exterminate any others, cognate 
 or alien, which happen to be in existence within the political 
 union. If, as is often the case, the supposed area is occupied 
 by a population composed of two or more distinct races, 
 which have been brought into close national association by 
 conquest or other circumstances, one or other of these 
 stocks will find its own language decline and finally dis- 
 appear in the course of generations. Nor will it be 
 absolutely necessary for this result that the races themselves 
 should become blended in any large degree. The racial 
 types may remain tolerably distinct while the language grows 
 homogeneous. Generally, no doubt, when there is no pro- 
 nounced bar of colour or religion — and sometimes even then — 
 a considerable amount of racial intermixture does take place, 
 but very frequently certain districts of such a country will 
 continue to show almost pure types of one race or the other, 
 whereas the language which originally belonged to that type 
 has become entirely obsolete. The language which prevails 
 is regularly that of the element which proves superior in 
 social power or culture ; it need not be that of the more 
 numerous. Mixture of the speeches themselves seldom 
 extends beyond the borrowing of certain vocables and some 
 modification of phonetics. 
 
 Among the more extreme and obvious instances of a 
 race losing its own speech in favour of one belonging to an 
 ethnologically unrelated stock is that of the Jews, who, 
 while for the most part maintaining a scrupulous purity of 
 descent, generally recognise as their mother-tongue the 
 language of the country in which they happen to be born 
 and domiciled. The Negroes of the United States speak 
 English ; those of Hayti French. Spanish is spoken by large 
 numbers of persons of American-Indian or very mixed extrac- 
 tion in Central and South America. The Melanesians are of a 
 race unmistakably distinct from that of the Polynesians or
 
 2 30 Natural History of Language 
 
 of the Malays, whereas their languages belong as unmistak- 
 ably to the Malayo-Polynesian family. A Slavonic speech is 
 steadily replacing the Ural-Altaic tongues of the Mongoloid 
 peoples in Russia. The Arabic of the Semites continually 
 penetrates into Africa among the Ethiopic and Negro popu- 
 lations. The Sanskritic branch of Indo-European imposed 
 itself on the Dravidian inhabitants of the larger part of India. 
 In the peninsula of Italy the Indo-European Latin became 
 the language of the alien Ligurian, Etruscan, and * Iberian ' 
 stocks. Except for a remnant in the shape of Basque, the 
 speeches of the old Iberian race in Spain have similarly 
 disappeared in favour of Latin, whereas the race itself still 
 flourishes. The non-Indo-European languages of the swart 
 brachycephalic occupants of Alpine and middle Europe have 
 given place, first to Celtic, then to Latin or Teutonic, but 
 the swart and short-skulled people spread through those 
 regions to this day. 
 
 Instances in which one language has been replaced by 
 another genealogically related to it are sufficiently familiar. 
 Thus Latin ousted Celtic from Gaul, and Greek from 
 Southern Italy and Sicily. A Teutonic speech is continu- 
 ally narrowing the range of Celtic in the British Islands, and 
 in some quarters, as for instance in Cornwall, it has already 
 completed the displacement. In the eastern parts of the 
 German empire one Slavonic language (Polabish) has been 
 entirely dispossessed by German, and others {e.g. Sorbian) 
 are doomed. Among the Semitic tongues Arabic is now 
 spoken by peoples whose ancestors spoke Aramaic, Chaldaic, 
 etc. In these cases the language which supplants is, indeed, 
 cognate to the language supplanted. But it by no means 
 follows that the speakers of the supplanting tongue are 
 themselves, at the time of supplanting, correspondingly 
 related by race to the speakers of the other.^ The Celtae 
 of Central France, who spoke the Celtic which was replaced 
 
 ^ It may be remarked that at no historical date has there been a pure ' Semitic 
 race ' co-extensive with the Semitic family of languages. The Arabian and 
 Assyrian racial types always diverged widely.
 
 Race and Language 2 3 1 
 
 by the kindred Latin, were mainly of a stock which had 
 originally little in common with the speakers of the latter 
 tongue. Latin was closely akin to Celtic, but not there- 
 fore the Celtae to the ' Latins.' The Celtic of the British 
 Islands is now the tongue of the short dark Silurian so- 
 called ' Celt,' of the tall rufous Belgic Gaul,^ also dubbed a 
 ' Celt,' and of a blend of the two. The English language 
 which is supplanting the Celtic was therefore brought by a 
 race only partially akin. 
 
 It requires little knowledge of history to bring home the 
 composite character of almost every nation which possesses 
 a national speech. The modern French in France is the 
 tongue of the descendants of Belgic Gauls (Belgae), 
 Rhaeto- Celts (Celtae), Iberians (Aquitani), a layer of 
 Italians (so-called Romans), with a certain Teutonic ad- 
 mixture from Franks, Burgundians, and Scandinavians 
 (Normans). The area of Italian is occupied by descendants 
 of Rhaetians, Ligurians, Gauls, Etruscans, Greeks, and 
 * Iberian ' Sicanians, as well as by those of the mixed Latin, 
 Umbrian, and Oscan tribes who were the original speakers 
 of the Italic variety of Indo-European. We must further 
 add a tincture of Gothic, Lombardic, and other Teutonic 
 blood, and also of Spanish. The meaninglessness of the 
 expression ' the Latin race,' as applied generally to speakers 
 of the Romance languages, needs no further demonstration. 
 Nor is the term ' Anglo-Saxon,' as applied to the population 
 of Great Britain or the United States, anything but a con- 
 venience. The language indeed is English Saxonic, but those 
 who speak it in Great Britain are racially the result of greater 
 or less blending of Ibero-Silurians, Belgic Gauls, possibly 
 some Roman strain, various tribes of Angles, Saxons, Jutes, and 
 Scandinavians, and the Norman French. The Teutonic racial 
 element by no means preponderates in every part in which 
 the language is now habitually employed. The English of the 
 United States is the tongue of a population still more mixed. 
 
 1 The related Goidels and Brythons are here considered as but two waves of 
 practically the same stock.
 
 232 Natural History of L anguage 
 
 Facts such as these, which are discoverable within the 
 period of history, or which are placed sufficiently beyond 
 doubt by the combined conclusions of ethnologists, historians, 
 and philologists, must be few compared with those which are 
 shrouded in the obscurity of prehistoric ages, or in the 
 comparatively recent, but quite recordless, past of barbarian 
 peoples. If history were entirely silent on the subject, the 
 position of the Hungarian tongue in Europe would be a 
 riddle. Nor could it ever be guessed that a wave of Gauls 
 had once passed into the centre of Asia Minor and deposited 
 there an island of language (Galatian) akin to those now 
 spoken on the extreme verge of North- Western Europe. That 
 island of language, after an existence of several centuries, 
 entirely disappeared, and the descendants of its speakers, so 
 far as they exist, are part of an exceedingly mixed popula- 
 tion whose ' national ' language is Turkish. No one, again, 
 would imagine the conquests, migrations, and colonisings of 
 Phoenicians, Macedonian Greeks, Gauls, Romans, Goths, 
 Slavs, Huns, Saracens, Turks, Mongols, Spaniards, or English. 
 Who then can say (to use an expression of Whitney) what 
 " encroachments, superpositions, mixtures, displacements, and 
 destructions " may have occurred among both races and 
 languages in times and places upon which history throws no 
 light ? 
 
 While, therefore, it is true that a race left to itself will 
 naturally speak a tongue developed directly from that of its 
 ancestors, and that two peoples of the same race will, under 
 these conditions, speak cognate dialects, nevertheless, inas- 
 much as a language can be altogether displaced through 
 circumstances such as those above illustrated, it is clear that 
 (i) only when physical as well as linguistic similarities are 
 forthcoming can we attach much weight to the argument of 
 language in support of a theory of community of race ; 
 (2) when the physical resemblances between two peoples are 
 very pronounced, a lack of kinship in their languages is no 
 very strong argument for their racial separation. The lin- 
 guistic test is therefore rated by the student of the science of
 
 Race and Langtiage 233 
 
 language at a value much lower than that which appears to 
 be often put upon it by the ethnologist. 
 
 Not only does historical experience prove that peoples 
 may be forced to change their language. There is, further- 
 more, no contravention of nature in the process. It is an 
 obvious remark that language is not a physical attribute, but 
 a social acquirement. The colour of the skin, the shape of 
 the skull, the proportions of stature, and the nature of the 
 hair, are normally matters of inheritance, and are determined 
 before birth. On the other hand, the language which the 
 child is to speak will depend on its associations and en- 
 vironment — that is to say, on the sounds habitually used for 
 the expression of ideas by those from whom it learns to 
 speak. An English child born and bred in France may 
 speak French as perfectly as the child of native parents. 
 And not only will this occur where the two languages are 
 akin, but the case would be the same with Chinese, Japanese, 
 or Tamil. It may be difficult, if not impossible, for an adult, 
 already habituated to the expression of his ideas by means 
 of certain sounds and certain grammatical methods only, to 
 acquire a perfect control of novel methods and novel sounds ; 
 and such a person might possibly conceive that he was racially 
 incapacitated for some particular process of phrase-building or 
 of articulation. But there is no adequate evidence that a 
 child, when first learning to speak, will experience any 
 difficulty either in the articulation of a language alien to its 
 ancestors or in the expression of thought by its means. 
 Psychologically the child of alien race appears to labour 
 under no disadvantage. Physically, indeed, it is possible 
 that there may sometimes — as with negroes speaking English 
 — exist a certain difference of timbre, due to some congenital 
 character of the vocal organs ; but, for the rest, such 
 idiosyncrasies as distinguish the English of the North 
 American negroes from that of the whites are mostly of the 
 ordinary nature of dialect, and are due to the more intimate 
 association of the negroes with one another, and to their
 
 2 34 Natural History of L anguage 
 
 consequent propagation of peculiarities of ' broken ' English 
 as spoken by the early slaves. For the most part, however, 
 idiosyncrasies are non-existent when the child of one stock 
 is brought up entirely, and from the first, in the language of 
 another. As regards the psychological element in the case 
 it is well to observe (i) that different races adopt similar 
 structural devices (the principle of agglutination, for example, 
 belongs to Magyars as it does to Australian aborigines) ; 
 (2) that the same language, at different stages, employs 
 dissimilar devices (as in the instance of analytical English 
 compared with the synthetic Old Teutonic) ; (3) that in all 
 probability the same (or at least analogous) structural devices 
 He at the bases of all languages alike. 
 
 It is, of course, possible that, if we could get back to a 
 period of unmixed typical races, we might find the language 
 of each to be more peculiar to itself Each such language 
 might possess some distinctive feature or tendency. In other 
 words, there might be one strongly marked tongue to one 
 strongly marked race. As things actually are, however, 
 while the totally distinct families of languages are far more 
 numerous than the races, and while nearly a hundred groups 
 of tongues may be discovered which exhibit no genealogical 
 connection whatever, these tongues display the most compli- 
 cated resemblances and divergences in respect both of sounds 
 and morphology, and in no way admit of such classification 
 as to make them correspond with any arrangement by races. 
 Nor does it seem very profitable to speculate whether, on the 
 one hand, many of these apparently unrelated speeches may 
 have gradually diverged from each other to such a degree as 
 to retain no recognisable traces of a connection which once 
 existed, or whether, on the other, they may be but survivals out 
 of a number much larger still. Whatever may be the truth 
 in this respect, we cannot fit the classifications of language, 
 either genealogical or morphological, to any of the racial 
 classifications into which various ethnologists divide mankind. 
 
 This is not the place for a discussion of the soundest
 
 Race and Langit age 235 
 
 ethnological system. But whether we adopt Peschel's distri- 
 bution into seven classes, viz. (i) Australians, (2) Papuans 
 (with the Negritoes, Melanesians, etc.), (3) Mongoloids (with 
 Malays, Polynesians, and American Indians), (4) Dravidas, 
 (5) Hottentots and Bushmen, (6) Negroes, (7) Mediterranean 
 peoples (Indo-Europeans, Semites, and Hamites), or Flower's 
 distribution into three classes (Black, Yellow, White), it will 
 remain impossible to discover any principle according to 
 which the languages of the peoples concerned can be simi- 
 larly divided. Simply by way of illustration we may take 
 (with a slight modification) an arrangement by Huxley, 
 which combines Haeckel's classification according to the hair 
 with a familiar distribution according to the ' index ' of the 
 skull,^ and these with yet another according to the colour of 
 the hair and skin : — 
 
 (a) Leiotrichi (smooth-haired ^) — 
 
 I. Lejicous {i.e. blond) dolichocephalic {i.e. long- 
 headed) — the Xanthochroi. 
 H. Leucomelanous {i.e. with dark hair and white 
 skins) — the Melanochroi. 
 {a) dolichocephalic — the Iberians, Semites, and 
 
 Berbers. 
 {b) brachycephalic — central Europeans (Rhae- 
 tians). 
 
 III. Xanthomelattous {i.e. with black hair and yellow 
 
 skins). 
 
 {a) dolicho- (to vieso-) cephalic — the Eskimo, 
 Amphinesians {i.e. Polynesians, Indo- 
 Nesians, etc.), American Indians. 
 
 {b) brachycepJialic — the Mongoloids. 
 
 IV. Melanous {i.e. black) dolichocephalic — the Aus- 
 
 tralians, Dravidians. 
 
 1 The most approved denominations, according to the proportion of the 
 extreme width of skulls to their extreme length, are — ' dolichocephalic,' when 
 width : length : : 75 (or under) : lOO ; ' mesocephalic,' when the * index ' is between 
 75 and 80 ; 'brachycephalic,' when the width is over 80 per cent of the length. 
 
 ^ This hair in section is cylindrical, while the woolly hair is flat.
 
 236 Natural History of L anguage 
 
 (yS) Ulotrichi (woolly-haired) — 
 
 I. Xanthoinclanous dolichocephalic — the Hottentots 
 
 and Bushmen. 
 II. Melanous dolichocephalic — the Negroes, Negritoes, 
 Papuans. 
 
 Linguistically there is no special speech-type attaching 
 to ' xanthomelanous brachycephalic Mongoloids.' In their 
 method of structure, in phonetics, and in vocabulary, the 
 Chinese are as widely sundered from the Tartars or Japanese 
 as from the Hindus or the Teutons. Their monosyllabic, 
 intoned, positional language may or may not, at some ex- 
 ceedingly remote period, have borne a closer resemblance to 
 the tongues of other Mongoloids, whose speeches are now 
 agglutinating. Philology, however, cannot prove the point. 
 Similarly the philologist can establish no community of type 
 between the polysynthetic speeches of the American Indians 
 and the partly agglutinating, partly analytical, tongues of 
 the Amphinesians. If, with Huxley, we class the ' leuco- 
 melanous dolichocephalic ' Iberians with the ' leucomelanous 
 dolichocephalic ' Semites, it is nevertheless a fact that the 
 Basque language remains as remote from Arabic in structure, 
 phonetics, and root-material as it is from American Indian 
 or Bantu. If, with other ethnologists, we class the ' Semites ' 
 with the ' Europeans ' as ' Mediterranean ' peoples, we get no 
 nearer to a common type of speech. The differentia which 
 divides Semitic structure from Indo-European is at least as 
 palpable as that which divides either from Ural-Altaic. The 
 difference of their phonetics is not less great, and the stocks 
 of root-material are entirely distinct. The principle of 
 agglutination is to be found equally among woolly-haired 
 long- skulled blacks or smooth-haired short -skulled yellow 
 men. The development of analysis has occurred among 
 ' xanthomelanous ' Polynesians as among the ' leucous ' 
 English. The tendency to a purely positional grammar is 
 both English and Chinese. Reduplication belongs to the 
 Xanthochroi as to the Red Indian. Similarly the pro-
 
 Race and Language 237 
 
 nunciation of s as // appears in Greece, in Persia, and in 
 New Zealand ; the confusion of / with b, or of t with d, 
 belongs to Polynesia as to ancient Germany. Nor would it 
 be difficult to illustrate much more amply the facts that 
 (i) peoples racially unlike may possess or develop like modes 
 of structure and phonesis, (2) peoples racially like may 
 possess or develop unlike modes of structure and phonesis. 
 
 To sum up. Given that frequent and complicated 
 migrations of whole peoples or of conquering bodies have 
 taken place in all parts of the world both during and prior 
 to historical times ; that the language of one people has often 
 been imposed upon another, with sometimes complete, some- 
 times considerable, and sometimes but little, racial blending 
 of the peoples themselves ; and that no type of man 
 possesses a congenitally necessary or indefeasible type of 
 language, — it becomes manifest that language must be treated 
 with more confidence as one of the data for history than as 
 one of the data for ethnology. Community of language 
 necessarily implies close and continued social relations at 
 some period or other. It need not imply more. 
 
 The application to the Indo-European tongues and the 
 diffusion of these from Europe (with a discussion of the 
 cradle of the race) forms a longer and separate chapter. 
 
 (ii.) The Speakers of Indo-European in Europe 
 
 Inasmuch as the error involved in the expression ' the 
 Indo-European peoples ' or ' the Aryan peoples,' as applied 
 to the bulk of the modern Europeans, is very common, and 
 has been encouraged by a laxity on the part of philologists, 
 it is perhaps worth while to examine in some detail the 
 relations of race and language in Europe. The examination, 
 moreover, will serve to illustrate one of the chief reasons why 
 a once homogeneous language becomes diversified as it 
 spreads further afield.
 
 238 Natural History of Language 
 
 At the present day Indo-European languages are spoken 
 in every part of Europe except the regions occupied by 
 Basque and by the Finno-Ugric and Turko-Tataric branches 
 of the Ural-Altaic family. The arrival of the Turks and of 
 the Magyars, and therewith of their languages, in the territories 
 now occupied by them, is a matter of history. The Finnic 
 tongues of Russia are rightly understood to be spoken 
 normally by detachments of the Mongoloid stocks of Eastern 
 Europe. Basque is naturally regarded as the language of a 
 remnant of the Iberian or Euskarian race which once 
 occupied the west. In these cases the difference of race is 
 brought home to the student more distinctly by the difference 
 of language. Otherwise it is commonly assumed, or at least 
 writers express themselves as if it were assumed, that the 
 populations of Italy, Great Britain, Germany, etc., belong 
 equally to an Indo-European or Aryan 'race.' The fallacy 
 at once becomes obvious if we imagine a time when the 
 Basque, Magyar, Turkish, and Finnic idioms shall have 
 become obsolete. In the case of Basque and some of the 
 Finnish speeches that contingency is anything but remote. 
 In the supposed case will the Basques, Magyars, or Finno- 
 Ugrians become ethnologically any the more ' Aryan ' or 
 ' Indo-European ' because their language has been sup- 
 planted ? 
 
 What is obvious in this example is, after all, but very 
 slightly obscured in the case of the population of Europe at 
 large. The Indo-European languages now occupy ground 
 which was formerly divided between several alien tongues 
 spoken by various ethnological stocks. The Rhaetian, 
 Ligurian, and Etruscan languages have wholly disappeared, 
 the Iberian almost. But the peoples did not disappear with 
 their languages, and, whether mixed or not, they must still 
 form a large, and sometimes the largest, element in the 
 population of the regions which belonged to their ancestors. 
 This fact is sufficient to account for the greater or smaller 
 degrees of physical unlikeness existing between the Italians, 
 Spaniards, Teutons, Celts, Slavs, etc., who agree in the
 
 Race and Language 239 
 
 possession of Indo-European tongues. A homogeneous race 
 did not come into occupation of an empty Europe and then, 
 thanks to setthng north or south, develop a prevaiHng tall 
 blond type and a prevailing short brunet type in those quarters 
 respectively. Such metamorphoses appear to be a physical 
 impossibility. The obvious truth is that, though the cognate 
 Indo-European tongues necessarily imply that a fairly 
 homogeneous Indo-European language must have been 
 spoken somewhere by some considerable number of persons 
 from whom it was diffused, nevertheless, as now spoken, 
 these tongues are the languages of a number of heterogeneous 
 peoples linked together in complicated ways, but in some 
 cases probably sharing little in the blood (whatever it may 
 have been) of those from whom the said common language 
 was so distributed. 
 
 The conclusions of ethnologists (Broca, Huxley, Isaac 
 Taylor, etc.), based partly upon the examination of large 
 numbers of prehistoric skulls and skeletons exhumed from 
 various regions of Europe, and partly upon the prevalence of 
 corresponding types over much the same areas at the present 
 day, are in sufficient agreement to permit of a tolerably 
 confident summary. 
 
 At a primitive (apparently neolithic) period Europe was 
 occupied by at least four distinguishable racial types. These 
 were no doubt mixed in various proportions on their respective 
 borders, but, speaking generally, they were situated and 
 characterised as follows : — 
 
 A. — A short dark dolichocephalic stock, averaging 
 5 ft. 4 in. in height, and with an index of 72. 
 These inhabited the west of the British Islands, 
 the west of France, most of Spain, and also of 
 Southern Italy, with Sicily, Sardinia, and Corsica. 
 Among the various names assigned to them 
 ' Iberians ' is perhaps most in vogue. Others are 
 ' Euskarian,' ' Silurian,' ' Basque.' 
 B. — A short dark brachycephalic stock, averaging 5 ft, 
 3 in. in height, and with an index of 84, occupying
 
 240 Natural History of Language 
 
 Central France, Northern Italy, and the central 
 
 highlands of Europe, and variously denominated 
 
 ' Auvergnat,' ' Savoyard,' ' Rhaetian,' and ' Rhaeto- 
 
 Ligurian.' 
 
 C. — A short dark bpachycephalic stock, with Mong^oloid 
 
 characteristics distinguishing them from B, and 
 
 inhabiting the E. and N.E. of Europe as well as 
 
 Central Asia. 
 
 D. — A tall light dolichocephalic stock, averaging 5 ft. 
 
 9 in. in height, and with an index of 72. In their 
 
 purest form these occur in Scandinavia, North 
 
 Germany, and in parts of the British Isles. The 
 
 name ' Xanthochroi ' (' blond-complexioned ') has 
 
 been suggested by Huxley. 
 
 The natural result of greater or less blending along the 
 
 racial frontiers, and often, thanks to migrations and inroads, 
 
 deeper into a given racial area, is that we may meet with 
 
 tall light people who are brachycephalic (as among the 
 
 Slavs, and frequently among the tall rufous ' Celts ' of 
 
 Ireland), or with light dolichocephalic people who are short, 
 
 or with dark dolichocephalic people who are tall, and so 
 
 forth. That, in so far as groups of individuals deviate from 
 
 one stock or another, they are the result of a mixture of the 
 
 stocks, is a necessary conclusion. For the most part the 
 
 Slavs apparently afford an instance of a blending of types 
 
 D and C. The Belgic ' Celts ' of Britain, who are ' moderately 
 
 brachycephalic ' (Taylor), while in other respects they belong 
 
 to type D, are probably the outcome of D somewhat crossed 
 
 with B. The Spaniards are apparently mainly of type A, 
 
 with a comparatively small proportion of type D. It is 
 
 needless to pursue the matter further, inasmuch as enough 
 
 has been said to make clear how little identity of race exists 
 
 among the peoples whose present languages are of common 
 
 origin. 
 
 To which stock the primitive Indo-European language 
 belonged is a question which, while not absolutely settled by 
 the data, admits of no reasonable doubt. Stock C, where it
 
 Race and Language 241 
 
 is purest, employs the Ural-Altaic speeches. Stock A (at 
 least in Spain and S.W. France) has adopted an Indo- 
 European speech within historical times, and, so far as traces 
 of the old Iberian exist, they agree with the surviving Basque 
 in exhibiting an entirely alien tongue, the Euskarian. At 
 the original language or languages of stock B, however, we 
 can only guess ; but such indications as are forthcoming 
 point to non-Indo-European idioms displaced at an early 
 date by Celtic. In the case of stock D, however, there is no 
 vestige of any other languages than Indo-European having 
 ever been spoken by the peoples who best represent it. 
 Moreover, all tradition and historical evidence connected 
 with the characteristics of the Greeks, Thracians, Aryans, etc., 
 who migrated into the quarters severally occupied by them, 
 point to the languages having been carried by tribes who 
 were conspicuously of type D. Probably, therefore, no 
 philologist or ethnologist feels any hesitation whatever in 
 placing among a tall blond dolichocephalic stock the source 
 from which the Indo-European languages were diffused. 
 
 (iii.) Celts and Celtic 
 
 Perhaps greater confusion exists nowhere than in the 
 application of the term ' Celt,' and it is certain that the 
 confusion is largely, though not solely, due to false impres- 
 sions derived from language. In Caesar's day Gaul was 
 tenanted by three races, the Aqidtani of the south-west (who 
 represent the Iberian or Euskarian stock A), the Celtae (or 
 Galli) of the centre and the Savoy hills (who belong to the 
 ' Auvergnat ' type, called B in the classification given above), 
 and the Bclgae of the north-cast (who are of type D). The 
 ethnological distinction is in itself sound and clear. It 
 happens, however, that by the Roman time the Celtae and 
 the Belgae, though different in race, spoke varieties of the 
 same language, which had apparently been imposed upon the 
 Celtae (of stock B) by the kindred of the Belgac (of stock D) 
 while sweeping over Central Europe. This language, being 
 
 (B 884) R
 
 242 Natural History of Language 
 
 first recognised among the Celtae as the Romans advanced 
 northward, was naturally called ' Celtic' The fact that it was 
 likewise spoken by the Belgic Gauls led to the latter also 
 being denominated (though not by Caesar) ' Celts.' 
 
 Furthermore, congeners of the Belgic Gauls occupied 
 from an early period the south and east of Great Britain, 
 whence they spread continually westward, imposing their 
 language meanwhile upon the earlier Siluro-Iberian inhabitants 
 of stock A. The only language discoverable in Great Britain 
 in Roman times is therefore ' Celtic,' although, in respect of 
 race, Romans do not fail to observe the existence of a 
 ' Silurian ' type in South Wales resembling the Iberian of 
 Spain. 
 
 Thus far we find the following racial types speaking a 
 common language, or at least cognate dialects, viz, (a) the 
 Central Gaulish, (/3) the Belgic Gaulish and Belgic British, 
 (7) the Siluro-Iberian. To (a) first belongs the name Celtae ; 
 among the Celtae was first met the ' Celtic ' tongue ; that 
 tongue was also spoken by (y5) and (7) ; and hence has 
 grown up a form of expression which confuses (a), (yS), and (7) 
 under one ethnological head as ' Celts.' The tall rufous 
 Irishman and his short swart compatriot are popularly linked 
 together as equally ' Celts,' endowed with a certain ' Celtic ' 
 quality of mind ; and with these are frequently combined 
 the Auvergnat French in some entirely imaginary psycho- 
 logical generalisation. 
 
 There is, perhaps, among philologists (so far as they 
 speak in terms of ethnology) a growing tendency to narrow 
 the use of the word ' Celt ' to the stock (tall and rufous) 
 which is primarily responsible for the diffusion of the Celtic 
 languages. Though this involves the transference of the 
 ethnical name from the central people to which it originally 
 belonged, to a people which originally had no claim to 
 it, no practical harm would result if the name were used 
 consistently, and if another title were no less consistently 
 applied to the ' Auvergnat ' or ' Savoyard ' race. For 
 the latter Rhaeto - Ligurian may perhaps be open to
 
 Race and Language 243 
 
 least objection ; ' Rhaeto-Celtic ' might only propagate the 
 misconception. 
 
 That the Belgae, or 'Celts' in the philological and 
 transferred sense, were racially very closely akin to, and at 
 first sieht identical with, the Teutons, is manifest from the 
 descriptions in early writers. Strabo, Tacitus, Lucan, Dio 
 Cassius, Silius Italicus, Ausonius, and others apply descriptive 
 terms, some to the Britons or Belgic Gauls, some to the 
 Germans, in such a way as to show that the Southern 
 European drew no distinction between their build, the colour 
 of their hair, or the colour of their eyes. Both are flavi 
 (' yellow-haired ') or rutili (' red-haired ') ; both have eyes 
 which are caerulei (' blue ') ; both have the appearance 
 indicated by the epithet truces (' fierce ') ; both have huge 
 frames. Tacitus feels a difficulty in deciding whether 
 the Treveri and Nervii are Germans or Gauls. Strabo 
 (vii. I. 2. 290) speaks of the Transrhenane Germans as differ- 
 ing only slightly from the Gauls, in possessing a greater 
 degree of the same characteristic fierceness, size, and tawni- 
 ness. It may be true that the ' Celtic ' individuals incline 
 more to brachycephaly, and that there actually exist certain 
 finer differences of eyes and complexion than the Romans 
 observed. These, however, are of little importance in 
 view of the very striking resemblances. So far as the 
 differences exist they in all probability betoken a certain 
 amount of racial modification through mixture, undergone 
 by the ' Celtic,' and not by the German, contingent.
 
 CHAPTER XI 
 
 THE ORIGINAL SEAT AND THE DIFFUSION OF THE INDO- 
 EUROPEAN FAMILY OF LANGUAGES 
 
 It is beyond doubt that at some period, however remote, 
 there existed somewhere a practically homogeneous people 
 speaking a practically homogeneous language, which we 
 may call the Primitive or Proetlmic Indo-European. The 
 fact that the existing Indo-European tongues, as we 
 trace back their phonetics, forms, and meanings, steadily 
 converge towards identity, is not explainable on any 
 other hypothesis. Everything goes to show that the 
 said people consisted of the tall Xanthochroi, and was 
 therefore sharply distinguished from any other race in Europe 
 or Asia. That the derived languages are now spoken by 
 peoples in whom the Xanthochroic element is often small 
 and sometimes even inconsiderable, is a phenomenon of 
 which an easy, if not immediately obvious, explanation has 
 been given above. 
 
 It is not unnatural that inquiry should be made as to 
 the situation of the primitive home of the Indo-European 
 stock and the lines of diffusion of its speech. Though 
 the matter is not one of the first importance in a general 
 work upon language, the intrinsic interest of the subject 
 perhaps justifies a note upon the progress of the investigation 
 up to the present. The opportunity may also be taken to 
 illustrate the caution with which arguments from language 
 must be applied to the determination of prehistoric questions. 
 The search for the primitive centre and the manner of 
 
 244
 
 Diffusion of Indo-European 245 
 
 distribution may be made more and less scientific in method, 
 and the discussion of the geographical starting-place of the 
 Indo-European languages has afforded exceptional scope for 
 the play of that poetical, patriotic, or other fancy, which 
 must end in corrupting the evidence. True science will 
 proceed only upon a collection and digestion of the data, 
 and these are manifestly to be sought from history (in which 
 well-sifted tradition may play its part), from antiquarian 
 research, from the indications of ethnology, and from the 
 testimony of facts which appear in the vocabularies of the 
 derived languages and in the reconstructed primitive tongue. 
 History supplies definite knowledge concerning the spread 
 of the Indo-European speeches in certain regions, as in the 
 obvious cases of Spain and Italy. Tradition (as in the case 
 of the ' Dorian Invasion ' and other early Greek movements, 
 and in that of the Eranian migrations of the Vendidda) is by 
 no means to be disregarded. Antignaj'ian research often assists 
 in ' placing ' a people at a date long prior to that at which it 
 is met in history proper, by means of such investigations as 
 those of barrows, ' kitchen-middens,' lake-dwellings, and other 
 prehistoric vestiges. Ethnology can distinguish the greater 
 and less purity of type prevailing in given localities, and so 
 help to determine the probabilities of racial movement and 
 the degree of superposition. Add to these that, if we can 
 arrive adequately at the primitive Indo-European vocabulary 
 and its meanings, we are in a fair position to name the fauna 
 and flora with which the early people was most familiar, the 
 most distinctly marked features of its climate and its natural 
 surroundings, and the salient elements in its food, occupations, 
 and external civilisation. Provided always that we can 
 assume a given word to have actually belonged to the 
 primitive people in a given meaning, we are thus supplied 
 with no inconsiderable data as to the probable latitude 
 and longitude of its prehistoric abode. This ' Linguistic 
 Palaeontology,' as it has been called, has of late been 
 cultivated with much assiduity, and, at the same time, with 
 increasing caution. Next, by following up the development
 
 246 Natural History of Language 
 
 of each of the derived speeches and comparing it with the 
 development of each of the rest, we may in some measure 
 secure two results : (i) we may discover in a given language 
 the growth of novelties ; new names of new objects, signs of 
 changes of habit, of new experiences and foreign contact ; 
 and these may afford some indication of the geographical 
 direction in which that particular wave of the primitive 
 language has advanced ; (2) we may perhaps, from a detailed 
 examination of both vocabulary and grammar, detect various 
 degrees of closeness in the interrelation of branch with branch, 
 especially in respect of special elements in their cultus ; and 
 hence we may gather a clue to their earlier situation in some 
 closer or less close juxtaposition, as the case may be, to each 
 other. 
 
 It is manifest that any serious attempt to determine the 
 original home and the lines of diffusion of the Indo-European 
 tongue will be based upon data drawn from all these com- 
 plementary departments of investigation, and not upon those 
 of one alone. Yet this fact has often been forgotten or 
 ignored, and, even where it has been remembered, the con- 
 clusions derived from the examination of the languages 
 themselves have been vitiated by many dubious assump- 
 tions as to what the vocabularies of those languages actually 
 prove or disprove. 
 
 Within the generation after the discovery that the Indo- 
 European languages constituted a distinct family, the notion 
 universally prevailed that the primitive abode of the primitive 
 stock was to be sought in Asia. To a large extent this 
 notion was a pure assumption. There were, however, two 
 causes which combined to maintain its vogue. 
 
 In the first place, the surprising discovery in Asia of a 
 very ancient Sanskrit literature, and of an only less ancient 
 Zend, led to an exaggerated estimate of the relative antiquity 
 of those languages. The European tongues, even Greek 
 and Latin, appeared suddenly to have grown young by 
 comparison, and there were not wanting reputable students
 
 Diffusion of Indo-European 247 
 
 who, like F. Schlegel (1808), actually derived the whole 
 family from one or other of these more venerable Asiatic 
 representatives. It was not yet realised that, if it had 
 been possible to acquire documents of Latin or Gothic at as 
 early a date as those of the Vedic Sanskrit or the Avestic, 
 the former languages at that date would have been found 
 to possess a character equally archaic. The research of 
 historical and comparative grammar leaves any other opinion 
 barely conceivable. If the step from the Gothic of the 
 fourth century A.D. to even the least advanced Teutonic 
 speech of the present day is so great, what appearance would 
 a primitive Teutonic of, say, lOOO B.C. have borne as com- 
 pared with the Gothic which we know? If all the records 
 of the dead languages of the Indo-European family had 
 vanished, and no data existed but those afforded by the 
 living speeches of to-day, by far the more archaic types 
 would now be extant in Europe, Lithuanian standing dis- 
 tinctly foremost in this respect. What is the case in the 
 twentieth century A.D. may well have been the case in 
 1500 B.C., but, unfortunately, the records for that date 
 are only found in one quarter, namely, in Asia. 
 
 In the second place, both Biblical and profane history 
 had made inveterate a habit of looking to the East for the 
 origin of all civilisation found existing in the West. But 
 here we are concerned with the possession of a mere language ; 
 and this is altogether distinct from the acquiring of a social, 
 religious, or artistic culture. As late as 1 840 the dis- 
 tinguished F. A. Pott was satisfied with this deduction from 
 irrelevant analogy, and he failed to perceive that his assertion 
 ex oriente lux belonged to the province of fancy, not of 
 argument. 
 
 Discussions of an ostensibly more reasoned character 
 were not, indeed, absent. Rhode (1820), taking his stand 
 upon the tradition of a series of Eranian migrations recorded 
 in the Vendiddd, and Lassen (1847) and others, supporting 
 him by a number of mostly superficial or wholly futile 
 arguments which it is scarcely profitable at this date to
 
 248 Natural History of Language 
 
 recount, placed the primitive centre on the Asiatic plateau 
 about the upper Oxus and Jaxartes. Pictet (1859) similarly 
 argues for Bactria and its vicinity, in a work which aims at 
 combining what are considered to be the geographical 
 probabilities of migration with indications afforded by Indo- 
 European words. By a process sufficiently familiar when the 
 conclusion is — however unconsciously — prejudged, he forces 
 from the vocabulary relating to the fauna, flora, natural 
 features and products of the primitive abode, all the evidence 
 calculated to support his theory. It has, however, been left 
 for later and more philosophical inquirers, such as Schrader, 
 to demonstrate the rigorous principles by which the linguistic 
 data should be sifted. 
 
 Though the Asiatic origin of the Indo-European 
 languages is still not wholly surrendered by the older genera- 
 tion of students, it now reckons few supporters among those 
 who were not, so to speak, born into the notion. As early 
 as 1 8 5 1 Latham had preferred a European cradle of the 
 family, and in 1862 he elaborated the position that the 
 Asiatic, being the lesser, contingent was the more likely to 
 have left the neighbourhood of the European, which is the 
 greater, than vice versa, there being nothing but a tacit 
 assumption to the contrary. Since that date Benfey has 
 decided for Southern Russia, Geiger (largely from the indica- 
 tions of the primitive flora) for Central Germany, Cuno for 
 the long northern plain of Europe, Posche (from a peculiar 
 fancy as to topographical causes of depigmentation ' resulting 
 in a blond type) for the middle of Western Russia in the 
 marshy neighbourhood of Pinsk, Penka (mainly on supposed 
 geological grounds) for Scandinavia. The divergence is 
 considerable, but the net result indicates at least that the 
 Southern and Western parts of Europe are virtually out of 
 the question. If, without disrespect to the labours of the 
 writers named, we may be permitted to pass upon them, in 
 the shape of brief dicta, verdicts which have been arrived 
 at after proper consideration, it will suffice here to remark
 
 Diffusion of Indo-European 249 
 
 that, (i) Penka's arguments are somewhat fanciful and far- 
 drawn, although he has meanwhile rendered good service in 
 directing attention to the fact that one cause of divergence 
 between the speeches derived from the proethnic Indo- 
 European must lie in the influence of the various non-Indo- 
 European peoples over whom the advancing waves of 
 Indo-European speech have swept. He urges, for example, 
 that special phenomena of the Lithuanian and Slavic branches 
 may be explained by the influence of sundry Mongoloid 
 speakers of Ural-Altaic tongues, who have been blended 
 into asocial and linguistic unity with ' Indo-Europeans.' (2) 
 The position of Posche, that the ' Indo-European ' ethno- 
 logical type is the result of albinism produced by the peculiar 
 marshy environment of Pinsk, is not supported by any 
 considerable authority in physiology. (3) Cuno's claim is 
 based upon the unjustified assumption that the primitive 
 stock must have been extremely numerous, and in any 
 case it ignores the consideration that, before a people is 
 extremely numerous, it must once have been small, and that 
 therefore our original inctcnahula are still to seek. (4) 
 Geiger's data are inadequate to the subject, and the reasoning 
 upon them is not convincing. (5) Benfey pretends to no 
 fulness of treatment. 
 
 By far the most satisfactory, if not in all respects 
 satisfying, treatment of the question is that of Schrader, who 
 is led by exhaustive arguments of an eminently judicial 
 character to place the most probable seat of the undivided 
 Xanthochroi on the Middle Volga, and to postulate a subse- 
 quent period of common life for the European division in an 
 area approximately bounded on the south by the Danube, 
 on the east by the Dnieper, on the north by the forests and 
 swamps of Volhynia, and on the west by the Carpathians. 
 
 That the Sanskrit-speaking people entered India from 
 the north-west seems clear from the fact that the topography 
 of the Vedic literature does not extend east or south of the 
 Punjab. The evidence of language, particularly as relating
 
 250 Natural History of Language 
 
 to culture and religion, makes it tolerably certain that there 
 was a period of joint ' Aryan ' {i.e. Indo-Eranian) life. We 
 may assign a fair value to the tradition of Eranian migration 
 recorded in the Vendiddd, according to which the starting- 
 point {Airy ana VaejanJi) of that branch lay in the neighbour- 
 hood of Sogdiana ; and we may combine this hint with our 
 historical knowledge that the West Eranian speech was forced 
 along from the east in Medo-Persian times. From all these 
 considerations we may reasonably arrive at the modern 
 Bokhara as a highly probable landmark in the march of the 
 Indo-Eranian contingent. Though, through lack of similar 
 data, we can take no further step north or west, the 
 undoubted racial affinities of the high-caste Aryans in ancient 
 times are to be sought in Europe, and, as Latham fairly 
 urged, it is a priori far more probable that the comparatively 
 few Asiatic members of the Indo-European family had 
 severed themselves from the comparatively numerous members 
 in Europe than that the reverse should have happened. Of 
 the European languages the Baltic-Slavic branch is recognised 
 by philologists as being phonetically the most immediately 
 related to the Indo-Eranian, and if at the earliest discoverable 
 period the line of affinity runs from Russia to the vicinity of 
 Bokhara, it is no rash speculation, especially in the light of 
 modern analogy, to conjecture that the Indo-Eranian division 
 found its way to the latter neighbourhood from the north of 
 the Caspian. 
 
 In view of its phonetics and its culture-vocabulary the 
 Armenian language is more naturally associated with the 
 European than with the above-named Asiatic members of 
 the family. There appears to be truth in the remark of 
 Herodotus that the Armenians were " a colony of the 
 Phrygians," and of Eudoxus that they were " Phrygian by 
 race and much like the Phrygians in language." ^ But the 
 Phrygians in their turn appear, on ancient evidence (which 
 is certainly better than mere modern guesswork, and in this 
 
 ^ The same association is made in Gen. x. 3, where Ashkenaz and Togarmah 
 (Phrygians and Armenians) are ' brothers.'
 
 Diffusio7i of Indo-European 2 5 i 
 
 instance is supported by a particularly strong consensus), to 
 have passed over to Asia Minor from Thrace. It is natural, 
 therefore, to suppose a line of advance from the European 
 side of the Propontis along the northern half of Asia Minor 
 almost to the Caspian, the linguistic links in the chain being 
 Thracian — Phrygian — Armenian. 
 
 Meanwhile, of the situation of the speakers of Indo- 
 European languages in Eastern Europe it is difficult to find 
 satisfactory evidence. These were much further afield from 
 the Greek observers than were the Phrygians and Armenians. 
 ' Scythian ' and ' Sarmatian ' are terms very loosely applied 
 by ancient authors to peoples northward of the Black Sea, 
 and, considering the generally nomadic and shifting character 
 of the hordes concerned, and the absence of any sound 
 chartographical, ethnological, or philological science among 
 the ancients, it is necessary to regard with much suspicion 
 the conflicting statements as to the race, language, or geo- 
 graphical extension of tribes in these regions. The prob- 
 abilities appear to be strongly against regarding the titles 
 ' Scythian ' and ' Sarmatian ' as properly borne either by 
 Slavs or by peoples closely related to Slavs, although it is 
 very probable that true Slavs, situated more to the north 
 and north-west, were often loosely included in the com- 
 prehensive expression. Though the question is an open 
 one, the more probable theory is that the region north of 
 the Black Sea, from the Danube and Pruth through South 
 Russia and the Cossack country into Asia, was occupied in 
 ' classical ' times by tribes whose languages would mostly 
 come under the Ural-Altaic classification. The alternative, 
 but less likely, hypothesis is that Scythians and Sarmatians 
 both spoke Indo-European tongues, which would form links 
 between the Slavonic and the Eranian. 
 
 North of these, however, we meet in Herodotus with the 
 mention of the Neuri,^ who are separable from the Scythians 
 and employ " a distinct speech." With these are associated 
 
 ^ The name may still exist in those of the river Nurefl' and the Nurska country 
 on the Bug (the tributary of the Vistula).
 
 252 Natural History of L angiiage 
 
 the Budini (north of the Sarmatians), " a great and numerous 
 race, strongly marked all through by grey-blue eyes and 
 ruddy hair." For these the geographical indications of 
 Herodotus point to a tract extending from the provinces of 
 West Russia (Volhynia, Grodno, Minsk, Mohilev) eastwards, 
 while the ethnological traits undoubtedly suggest the 
 Xanthochroic Slavs, with whom they are now generally 
 identified. Early in our era the Slavs are met with under 
 the names Veneti, Venedi, or Venedae, the ' Wends ' of later 
 German history. According to Tacitus the Venedi " roam 
 through the woods and mountains " between the Peucini {i.e. 
 the Teuton Bastarnae) and the Fenni {i.e. the Finns). Pliny 
 and Ptolemy place them in the same region, which is 
 practically that above described. It would not, of course, 
 be easy to separate them in all cases from their Lettic and 
 Lithuanian neighbours and linguistic congeners, but the 
 latter (or Baltic) division would seem to be represented by 
 tribes known to Tacitus as the Aestii and by the Galindae 
 and Sudini of Ptolemy. All the data, therefore, combine to 
 place the Slavs, as first known to us, in a tract of Western 
 and Central Russia, with the Letto-Lithuanians separating 
 them from the Baltic on their north-west, with Finnic 
 peoples bordering them on the north, and with non-Indo- 
 European hordes to the south-east. 
 
 Meanwhile, Teutonic languages were spoken from the 
 Vistula to the Rhine (where Pytheas in 325 B.C. found a 
 nation distinct from the Celts), and it was Germans who 
 limited the Venedi to the west. 
 
 That the central highlands of Europe, the basin of the 
 Upper Danube and of the Main, and Bohemia were once 
 occupied by Celtic-speaking peoples, is a conclusion not 
 only deducible from the intimations of history, but apparently 
 proved by the Celtic place-names surviving in the districts 
 concerned. This fact in itself would not, of course, decide 
 the question whether the speakers of Celtic had originally 
 passed from the east through that region or had overrun it 
 from the west, but argument is all in favour of a migration
 
 Diffusion of Indo-European 253 
 
 through steady pressure from the east, which forced them 
 into Gaul, and thence into Britain on the one hand, and 
 towards Italy or Spain on the other. The ethnological and 
 linguistic application of the term ' Celt ' has already been 
 discussed, and the remark of Herodotus that Celts are to be 
 found " outside the pillars of Hercules," and are " after the 
 Cynetes the most westerly people in Europe," is perhaps not 
 of much value as establishing the contemporary situation of 
 a definite racial or linguistic stock. It does, however, fairly 
 correspond to what is otherwise known or conjectured of the 
 situation of Celtic-speaking peoples, who came " next west " 
 to an Iberian or Euskarian population. 
 
 For the Italic division of the Indo-European family 
 little more can be said than that (i) the advance of the 
 language in the earliest historical times is southward, (2) the 
 earliest situation of the Umbrians, if tradition and archaeology 
 can be trusted, was farther to the north, where they occupied 
 the Po valley, from which they were subsequently forced 
 south by Gauls. Meanwhile, in the north-west and west, 
 the Ligurian coast and Etruria were occupied by non-Indo- 
 European tongues, while the islands (including Sicily) and 
 the south of the peninsula were inhabited by ' Iberians.' The 
 line of entrance to the peninsula would thus appear to be from 
 the north-east and along the eastern side. From the linguistic 
 standpoint a comparatively close connection of the Celtic and 
 Italic branches is practically certain, and it would not be 
 unnatural to suppose that the two had at some period formed 
 but one division somewhere nearer the centre of Europe. 
 
 The inroad of the Hellenes into Greece, it is no longer 
 doubted, took place from the north, and rather along the 
 western than from the eastern side of the Strymon. 
 
 The probabilities resulting from the material thus 
 collected may perhaps be best embodied in the form of an 
 outline map, in which the hypothetical movements of the 
 several branches of the language from the hypothetical 
 primitive centre are indicated by an arrow.
 
 Diffusion of Indo-Europeati 255 
 
 While history and tradition thus point to an East 
 European centre, the same hypothesis is supported, or, to 
 say the least, in no way challenged, by the linguistic argu- 
 ments. In reconstructing the primitive Indo-European 
 speech from the data afforded by its several branches, we 
 arrive at a language already marked by slight dialectal 
 variations. It may not be difficult to postulate a still 
 earlier stage of that language, in which such local diver- 
 gences did not yet exist ; but, as a fact, on reaching the 
 common linguistic terrain to which the branches converge, 
 we find at least phonetic differences already in operation. 
 The manner in which disintegrating tendencies arise and 
 become established till they result in dialects, is fully dis- 
 cussed elsewhere. Everything goes to show that the great 
 divisions of the Indo-European family of languages represent 
 so many local dialects, or amalgamations of local dialects, 
 of the common original. Doubtless many others, originally 
 interlinking with these in a complicated chain-work, and 
 serving as continuous transition-steps, have died out or been 
 absorbed, until the surviving divisions are marked by com- 
 paratively wide divergences. In any case, philologists no 
 longer hold the view (largely due to the assumption of an 
 Asiatic starting-place) that, from time to time, a large 
 contingent (A, B, etc.) of the hitherto linguistically homo- 
 geneous people broke off in a certain direction from the 
 original home, developed certain characteristics of its own in 
 consequence, and then split again in the same manner into 
 minor branches {b, c, etc.), each of which after a time 
 developed new features in virtue of its isolation. In other 
 words, the probable course of linguistic separation of the 
 Indo-European branches is not to be represented by 
 such a diagram as the following : —
 
 256 Natural History of Language 
 
 Such a history is, doubtless, readily conceivable, and 
 may actually have occurred in the case of this or that 
 family of languages, but it is contrary to the evidence in 
 the case of Indo-European. It could be upheld only by 
 a distinct gradation in the affinities of the several branches, 
 and that gradation is not to be found in fact. We cannot 
 (except with Italic and Celtic) take any two branches and, 
 from the total argument of their vocabularies, grammar, 
 and phonetics, relate them more closely as against a third. 
 If, for example, we discover an apparently special affinity 
 between Greek and Latin in respect of vocabulary, we 
 immediately appear to see no less special an affinity 
 between Greek and Indo-Eranian. If the Baltic- Slavic 
 tongues exhibit a marked phonetic characteristic in which 
 they are at one with the Aryan branch, in another respect 
 of phonetics they will disagree with the Aryan and agree 
 with the Greek or Teutonic. The notion that the ' classical ' 
 tongues, Greek and Latin, were related in a special degree 
 of intimacy, was chiefly due to the fact that those languages 
 were habitually studied together, and their relationship of 
 grammar and vocabulary therefore brought into strong 
 relief, while the other branches of the family had not 
 received the same detailed attention. More accurate re- 
 search shows that a resemblance of Greek to Aryan is at 
 least as distinct, if not more so. In the vocabulary relating 
 to arms both offensive and defensive, and also to agriculture, 
 that agreement is as striking as is the almost total unlike- 
 ness of Latin to Greek in the same respects. In point of 
 grammar — as in the case of the ' augment,' the infinitives, 
 and many other elements of their morphology — the same 
 apparent nearness of Greek to Indo-Eranian reveals itself 
 Nevertheless it is none the more possible to commit 
 ourselves to the hypothesis of a special Graeco- Aryan 
 connection, to be set in sharp antagonism to a Graeco- 
 Italic affinity. That Greek and Latin not only agree in 
 a genitive plural ^-dsoin (Gk. -dwv, Lat. -drum), which is 
 found nowhere else with nouns, but also in a quite peculiar
 
 Diffusion of Indo-European 257 
 
 plural imperative form in ^-ontd{n) (Gk. (pepovroyv, Lat. 
 ferunto), can hardly be set down to mere chance coincidence. 
 Such considerations bring us more naturally to a view of 
 the relationship between the three branches which we may 
 embody in the diagram 
 
 Once more, the original palatal and velar explosives of 
 Indo-European are treated by the Aryan and Baltic-Slavic 
 branches in a way which markedly distinguishes these from 
 nearly all the European divisions of the family. The 
 palatals, for example, are regularly converted by them 
 into spirants, mostly sibilants, while Greek, Latin, Celtic, 
 and Teutonic no less regularly retain them as explosives. 
 Thus the *i which remains k (whether so written or as c) 
 in the latter tongues, takes the shape of s in Sanskrit, s or 
 s in Zend, and s in both Lithuanian and Slavonic. The 
 consistency with which this rule is carried out might point 
 to a prolonged juxtaposition, if not community, of the 
 Baltic-Slavic and Indo-Eranian divisions, and it might seem 
 a reasonable theory that they developed this phonetic 
 characteristic in common, while separated from the remaining 
 branches. Yet, on the other hand, the Baltic-Slavic tongues 
 retain, with the European, the *^ and *^, both of which the 
 Aryan branch invariably converts into a ; and, whereas 
 a case-ending containing -bli belongs to Aryan, Latin {-bus), 
 Celtic {-i-b), and is found in old Greek (-^t). Baltic-Slavic 
 agrees with Teutonic in rejecting this inflexion for a form 
 in -in (Gothic -;//, Lith. -viiis, Old Slav. -mu). It is mani- 
 festly safer, therefore, to suppose an earlier relation 
 
 (b 384)
 
 258 NattL ral History of L anguage 
 
 with, perhaps, a later situation 
 
 -^eutor 
 
 than to postulate for any period a connection represented 
 by 
 
 With Italic and Celtic the case is different. Un- 
 equivocal morphological affinities of a special character 
 occur in the shape of a passive in -r (Lat. sequi-tu-i% Old 
 Irish -sedie-tha-r) and in the creation of a compound future 
 tense, Lat. -bo^ O. Ir. -b (Lat. amabo, cf. O. Ir. c]iarii\i). 
 The evidence of phonetics and of general resemblance, after 
 all allowance has been made for borrowing, points in the 
 same direction. We need not, indeed, assume an entire 
 identity for the linguistic ancestors of the two branches, 
 but we are compelled to regard them as having been in 
 unusually close relations with each other in the original 
 home and probably afterwards. 
 
 There are few inquirers who do not now accept upon 
 the evidence the rational ' wave - theory ' or theory of 
 ' continuous transition ' advanced by J. Schmidt, according 
 to which the several divisions of the cognate languages are 
 the development of dialects which originated in situ. The 
 process of such development, which is not elaborated by 
 Schmidt, is examined elsewhere in this work, and explained 
 as the inevitable experience of any uncultivated language 
 extending over a wide area. 
 
 Taking the relative positions of the branches of the 
 Indo-European family as we find them at the date to 
 which the map above given refers, we obtain the following 
 scheme : —
 
 Diffusion of Indo-European 
 
 259 
 
 ^acian ^ (Sarmatian)'"-., 
 
 ^hracian | (Scythian) 
 
 lado-Iranian 
 
 A search for the interrelations of the groups on linguistic 
 grounds leads, as we have seen, to the view that, except in 
 the case of Italo-Celtic, those relations are practically equal. 
 The natural conclusion from this fact, combined with the 
 indications of geography, is that, at a period just prior to the 
 actual break-up of the family, the dialectal areas were inter- 
 linked to each other somewhat after the following fashion ^ : — 
 
 ^ Schmidt points out that, in the matter of vocabulary, there are special 
 correspondences interrelating the branches with each other in ways which make 
 against any other theory of diffusion. Thus, while Greek has 132 vioxAs peculiar 
 to its connection with Latin, it has at the same time 99 peculiar to a connection 
 with the Aryan tongues. The figures given for Aryan and Teutonic are 15, 
 for Aryan and Italic 20, for Aryan and Baltic-Slavic 90, for Teutonic and Baltic- 
 Slavic 143. In some cases, of course (as between Slavs and Teutons, or Greeks 
 and Italians), simple borrowing of words may have taken place at so early a date 
 that the borrowed words have gone through the same phonetic experiences as 
 the native vocables, and so escape detection as borrowings. Nevertheless, had 
 there been a specially intimate connection between two branches we should have 
 looked for more proof in the vocables.
 
 26o Natural History of Language 
 
 If this diagram approximately represents the truth, the 
 primitive centre can hardly be looked for elsewhere than 
 in the same region of Eastern Europe to which we were 
 brought above by history and tradition. 
 
 In the handling of the Indo-European vocabularies for 
 the purpose of drawing conclusions as to the original home 
 or the original civilisation, great caution must be employed. 
 Before there can be any certainty that the undivided (or 
 ' proethnic ') people was familiar with a given plant, animal, 
 natural feature or phenomenon, social custom, or implement, 
 it is necessary to ascertain that a word unmistakably 
 expressing the concept in question is or has been shared 
 by a conclusively representative number of the derived 
 speeches. It is unsatisfactory to quote the agreement of 
 even a considerable number of languages whose speakers 
 have always lived in close vicinity, or have exercised more 
 or less close relations of trade or culture. Such agreement 
 is not representative. If we found that the Celtic, Italic, 
 Teutonic, and Slavonic branches possessed a common name 
 for the same object, while in other branches the word was 
 non-existent, w^e might, according to circumstances, argue 
 for either of two views : (i.) that the object was not familiar 
 to all the stock, but was only known ia) in that part of 
 its area occupied by the linguistic ancestors of the said 
 four branches, or {U) in some new region subsequently 
 occupied by those ancestors while still associated ; (ii.) that 
 the name either {a) originated in post-ethnic times wuth 
 one of the four branches, and was transmitted to the rest, 
 often along with the object itself, or else {b) was borrowed 
 by all alike from some alien source. No doubt any late 
 transmission would naturally betray itself in the shape of 
 the word, which would ordinarily contravene some phonetic 
 law of the borrowing language. A very early borrowing, 
 however, might be anterior to the operation of such phonetic 
 laws, which would therefore act upon the loan-word as 
 much as upon the native material. It is clear that the
 
 Diffusion of Indo-European 261 
 
 probabilities in the case of each word require to be estimated 
 separately. 
 
 As against either of these views, however, it might be 
 urged (iii.) that the name might easily have belonged at 
 one time to all the branches, but that several may have 
 dropped it out of use, while four have chanced to retain it. 
 Thus the fact that for the Greek dvydrrjp {tJiugater), Skt. 
 duhitar, Eng. daughter, the Latin has no etymologically 
 corresponding word, but only the distinct filia or ndta, 
 would hardly be taken as proof either that daughters were 
 unrecognised or that they enjoyed no distinctive name. It 
 would be no sufficient answer to this third view if one 
 remarked upon the apparently significant fact that the 
 word was retained only in four geographically neighbouring 
 divisions. Even among the primitive people dialects sprang 
 up, and differences of vocabulary arise in dialects as much as 
 differences of phonetics. Experience shows that a particular 
 dialectal section often completely drops a term w^hich was 
 once in current use over the whole speech-domain. The 
 linguistic forefathers of the four branches in question might 
 have formed a dialectal section in respect of retaining a 
 particular word, however much they may have differed 
 among themselves in other respects. 
 
 Any one of these three positions is manifestly reasonable, 
 if not carried to the point of deliberate resistance to the 
 probabilities of the case. The unprejudiced inquirer will 
 hold one view in one instance and another in another, 
 according to the aggregate of the evidence. 
 
 Finally (iv.) it might be argued that the object was 
 perhaps entirely unknown to the primitive people, that it was 
 met with by the separate branches after the dispersion, that 
 each either invented a name out of its own resources or 
 else applied some existing term in a new (or transferred) 
 sense, and that a certain number of them chanced to invent 
 or to transfer the same word. For this position little can 
 generally be said. The laws of probability are against 
 such fortuitous agreements on any large scale, and the
 
 262 Natural History of L anguage 
 
 improbability is immensely increased when those who 
 merely ' chanced ' to invent or transfer the same terms 
 ' chance ' also to be the nearest neighbours. 
 
 To take a concrete example. The sea bears in Latin, 
 Gaulish-Celtic, Gothic, and Old Slavonic respectively the 
 cognate names mare, more, marei, morje. No such word 
 appears in Greek,^ Armenian, or Indo-Eranian. According 
 to the above four lines of argument we may ask : (i.) Was 
 the sea known only to one half of the stock, com- 
 prising the linguistic ancestors of the branches first 
 mentioned ? (ii.) Was the sea known only at a still later 
 time to a single branch, which passed on the name to its 
 neighbours ; or, as an alternative to this, was the term 
 borrowed by all alike ? (iii.) Was the sea known to all the 
 stock, but its name changed by one or other of the 
 divisions, much as the name for ' horse ' (Germ. Pferd, Lat. 
 egiius. Popular Lat. caballus) has changed within historical 
 knowledge ? (iv.) W' as the sea known to none, but did the 
 word represented by mare, etc., originally mean, say, a lake 
 or pool, or is it some new invention expressing, say, ' bitter ' 
 or ' shining ' or ' restless,' or some other notion upon which 
 four branches chanced to light equally ? 
 
 Few would doubt that, in this instance, the probabilities 
 are altogether against both the second and the fourth views. 
 On the one hand, the geographical situation at any known 
 or imaginable stage of the history of the four branches makes 
 it extremely unlikely that the Italic division would borrow 
 its word for the sea from the Teutons or Slavs or vice 
 versa, or that all these alike would borrow from some alien 
 source. On the other, the entire agreement in both root 
 and suffix-element puts out of the question the notion that 
 each people, by an identical psychological process both in 
 respect of the appellation itself and its morphological deter- 
 mination, separately invented precisely that term for the 
 sea ; while, before we could treat as better than any other 
 
 ^ Possibly, however, MaFpa, the name of one of the sea-nymphs in Horn. //. 
 18. 39 sqq., may conceal a trace of the cognate. M.a.l.pa.=.\ixap-ia. { = mar-ya).
 
 Difftisioii of Indo-European 263 
 
 guess the notion that viari is simply a general primitive 
 word transferred from another sense to the sense ' sea,' we 
 must find some traces of the existence of the word at all 
 beyond the branches in question. The first and third views 
 would, perhaps, have been almost equally tenable if it did 
 not appear, on closer examination, that, whereas the four 
 branches in question agree in their word "fnmri, and may 
 represent an original dialectal division of the family in that 
 respect, the branches which do not possess the word f7;iarz 
 do not show any agreement in an alternative term. Hence 
 it might rather be gathered (though it is not, of course, 
 proved) that the ancestors of the latter branches were not 
 together acquainted with the sea in the same degree as were 
 the former. 
 
 Hitherto we have dealt with the case of a word contained 
 in a group of languages which, from their position, cannot 
 fairly be considered representative of the whole family. 
 Adequate representation begins when the hypothesis of a 
 later borrowing, or that of a common development in one 
 dialectal section of the stock, is precluded. This would 
 manifestly be the case if the four branches, Aryan, Armenian, 
 Greek, and Celtic, should agree in the use of a particular name 
 for a particular thing. Though the word would still — as in 
 the instance already supposed — be found in but four divisions, 
 the argument for its original existence throughout the entire 
 stock would be on a very different footing. It would be 
 difficult to maintain any other than the third view of the 
 four above enumerated. The more widespread the word, the 
 more entirely convincing its ascription to the earliest stage ; 
 but the agreement of very few ' representative ' branches is 
 practically sufficient. The Sanskrit, Armenian, Greek, and 
 Irish words for 'a bear' are respectively 7-ksa, arj, apKro^ 
 {arktos), art. It is impossible not to believe that the bear 
 was known under a common name to all the stock. The 
 fact that the Latin is also iirsus ( = ■\orcs2is) increases the 
 certainty, if increase were desired. The name for copper is 
 in Sanskrit ayuxs, in Latin acs, in Gothic aiz. The geo-
 
 264 Natural History of Langtiage 
 
 graphical situations of the three branches make any late 
 transmission of the word from the European to the Asiatic 
 language, or vice versa, inconceivable, however readily we 
 might conceive of a borrowing between the Latin and the 
 Teutonic. That a word with stem *aies- existed in the 
 primitive tongue in the sense of ' copper,' is an inference 
 practically inevitable. 
 
 In pursuing this line of investigation the philologist may 
 easily travel beyond the evidence. Before ascribing to the 
 primitive people an acquaintance with an object, a custom, 
 or an operation, it is necessary to make sure, not only 
 that they possessed the word, but that to the word there 
 was attached the precise sense in question. That the 
 original ' Indo-Europeans ' were familiar with the horse, ox, 
 sheep, pig, dog, wolf, bear, beaver, otter, goose, duck, and 
 mouse ; with the birch and willow ; with winter, snow, and 
 spring ; with wheeled waggons and yokes ; with honey and 
 mead ; is placed beyond doubt by the widespread identity 
 of the word and its meaning in each case. That they knew 
 the oak is probable, but not proved, inasmuch as, though 
 the Greek hpv^ {drfis) and the Irish daiir both mean 
 ' oak,' the Gothic tnu, Sanskrit dru- have the meaning 
 ' tree,' while the Old Norse tyr (another form of the same 
 word) is ' fir.' The oak is pretty obviously ' the tree ' par 
 excellence, and it is easy to understand how in Scandinavia 
 it was the fir which became ' the tree.' ^ 
 
 Often, as Schrader fully demonstrates, a word in a certain 
 distinct meaning is confined to the European languages, a 
 fact which may be due either to a joint European period 
 after separation from the Indo-Eranian, or to the loss of the 
 word by the Asiatic division, thanks to new circumstances or 
 to the loss of the thing itself Thus the beech indisputably 
 possessed a name common to the linguistic ancestors of the 
 Teutonic, Slavic, Latin, and Greek divisions (although, since 
 there is no beech in Southern Greece, the name was there 
 
 ^ The same change of meaning occurs in the equation, Latin quercus ('oak ') 
 = 0.H.G.>r/2a ('fir').
 
 Indo-European Primitive Culture 265 
 
 transferred to the tree nearest in appearance^). Whether 
 it was ever possessed by the Hnguistic forefathers of the 
 Aryans also, it is impossible to tell, since they could have no 
 occasion to use the word when they came into a country 
 which produced no beeches, while they might not, like the 
 Greeks, happen to transfer the name to any other tree. 
 From a collection of examples, however, Schrader arrives at 
 the conclusion that a joint European period is to be assumed, 
 during which at least that contingent of the stock was 
 acquainted with the sea, with the oak, beech, and pine, with 
 barley, wheat, sowing and ploughing. When we have put 
 together these data, they cannot, indeed, be called very 
 decisive as to the primitive seat of the linguistic stock ; for, 
 though certain latitudes are necessarily excluded, and merely 
 mountainous regions rendered highly improbable, there is 
 still a vast space left to satisfy the conditions. Highly 
 significant, however, is the fact that the beech is not actually 
 found eastward of a line from Konigsberg on the Baltic to 
 the Caucasus. All that can be said on the strength of this 
 description of evidence afforded by the Indo-European 
 vocabulary, is that the primitive centre already arrived at by 
 the arguments from history and dialectal interconnection is 
 not disturbed by it, but further supported. 
 
 It is hardly germane to this discussion to enter into 
 an examination of the primitive culture of the Indo-Euro- 
 peans. It may, however, be of some profit for the general 
 student of language to be reminded of the diffidence with 
 which conclusions touching this matter should be drawn from 
 vocabulary. Thus, that the primitive people constructed 
 habitations of some kind, with doors of some description, is 
 demonstrated by the series of words Skt. dama. Arm. tun, 
 Gk. hoi^Q^ {doinos), Lat. domus, O. Slav, domu (connected 
 with the root *dem, ' build,' from which comes timber, Germ. 
 Zimnier), all meaning 'house,' and in the same languages 
 respectively dvdr-, duru, Ovpa {thiird), fores, dviri, all mean- 
 
 1 Phonetically Gk. <pa.-y6^ (/>/ia^os) = U. /J^ns =S]ziV. duly=Ex\g. beech.
 
 266 Nattival History of Language 
 
 ing ' door,' Yet these words in themselves prove nothing as 
 to the material and shape of the house or door. Nor does 
 the absence of a common word for ' window ' absolutely 
 prove that windows were unknown. It is only the collection 
 of other data which leads to a natural inference that the 
 houses were built of mud, wood, and straw, that they were 
 thatched, and that they were without windows. Similarly 
 the words ndu-, ndvi, vav<; {iians), ndvis, noi, in Sanskrit, Zend, 
 Greek, Latin, and Old Irish respectively, may be translated 
 ' ship,' but the object originally so named may have been 
 simply a ' dug-out ' or other boat. The Greek fMvWeiv 
 {inullebi), Lat. violere, Goth, inalan, O. Slav, vialja, are used 
 of grinding corn in a mill, but the original Indo-European 
 sense may have been simply that of grinding or pounding 
 between two stones. That there was a root "^nebh in some 
 sense of ' weaving ' is clear from Sanskrit vabJi, Ossetian 
 wafun, Eng. zveave, Gk. ix^i) (JuipJic), but of what kind was 
 the operation, and with what appliances ? Much sug- 
 gestion and some positive information can, of course, be 
 gathered in this way ; but the earlier inquirers were apt 
 to be over precise in their pictures of the civilisation of the 
 prehistoric undivided people. 
 
 Particularly fascinating was the quest for the primitive 
 religion, and theories which connected many of the larger 
 and smaller names in the Greek mythology with others in 
 the Indian branch were advanced in violation not only of 
 general probabilities, but also of the known phonetic laws. 
 As a fact, the only pertinent equation which ultimately 
 emerges is that of Skt. Dymis-pitar, Lat. Jn-pitcr, Gk. Zeu? 
 iraWjp {Zeus pater), Teutonic Tzu. The primitive word 
 "^dicns might, no doubt, have been used simply of the sky, 
 but the peculiar agreement of several branches in the frequent 
 addition of *pdter (' father '), and the complete characterisation 
 of the word as a proper name (while other terms were sub- 
 stituted for 'sky'), practically demonstrate that the primitive 
 stock already recognised a deity bearing that name. That the 
 notion of deity itself existed, is fairly indicated by the words
 
 Indo-European Primitive Culture 267 
 
 deva^ dens} tivar (plural), dcvas, occurring in Sanskrit, Latin, 
 Old Norse, and Lithuanian respectively. For the rest, it 
 proves nothing to quote a common name for ' dawn,' ' fire,' and 
 * sun,' and to show that in several branches these entities are 
 poetically personified and in some cases worshipped. Nor is 
 it worth the trouble, even when we have allowed for possible 
 deviations from phonetic regularity on the ground of ' popular 
 etymology ' and other disturbing psychological causes, to 
 force a word like 'E^ti^y? {Erlnus) into identity with Sanskrit 
 Saranyu. 
 
 ^ The Greek ^eos {theos) is a different word.
 
 CHAPTER XII 
 
 PHONETIC CHANGE 
 
 Sound and thought : leaj'ning to speak 
 
 Language is primarily the communication of ideas by 
 means of articulate sounds. The attempt to make such 
 communication definite naturally grows in the human being 
 with the growth of ideas, or at least with the consciousness 
 of possessing them. Before the development of its reason 
 the infant expresses its discomforts and its imperfectly 
 realised desires by natural cries, which are inarticulate, and 
 which do not belong to language in the proper sense. It 
 may be reasonable to suppose that these were the original 
 germ of language ; nevertheless they are not yet language. 
 Language begins when the sounds uttered are articulate, 
 habitually attached to particular ideas, and recognisable by 
 others as expressing those ideas. Such sounds do not come 
 to a child by nature and intuition, but depend upon its social 
 environment. They are acquired from example by imitation. 
 The English child attaches English, and the French child 
 French, sounds to the same notions. Within the borders of 
 the same language the child of the peasant, the gamin of the 
 city, and the child of refined nurture, acquire a different 
 articulation for what is nominally the same group of sounds. 
 The human being, in the words of Aristotle, is fii/xrjTiKcoTarov 
 TMV ^axov, " the most imitative of creatures," and " acquires all 
 his first lessons by means of imitation." 
 
 268
 
 Phonetic Change 269 
 
 The sound-picture^ derived from the average articulation 
 of the environment 
 
 The first efforts of a child at speech are attempts of its 
 vocal apparatus to reproduce for a certain purpose the 
 sounds which it hears habitually applied to that purpose. 
 Impressed upon its mind through the ear there comes to 
 exist a certain sound-picture. The more often it hears a 
 sound, the more distinctly is the picture impressed. If, as 
 sometimes, though rarely, happens, the sound (or sound- 
 group) is heard only from the lips of one person, such as a 
 mother or a nurse, it will have been consistently articulated, 
 and the mental picture of the child will correspond almost 
 absolutely to the special articulation of that person. If, as 
 is more commonly the case, the sound (or sound-group) is 
 heard from the lips of a number of speakers, such as the 
 two parents, elder children in the family, and playmates, the 
 articulation will not have been consistent, and the sound- 
 picture impressed on the mind of the child will mostly be a 
 compromise, the result of a series of superpositions.^ It will 
 correspond to the average articulation of the environment. 
 That average is determined not solely by the average of the 
 manners of articulation of the several persons concerned, but 
 also by the degree of frequency with which one such 
 articulation is heard as compared with another. Thus if the 
 child is surrounded by varying articulations a, a^, a,^, <?,, it 
 may happen that a^, being that of a mother, nurse, or 
 constant playmate, is heard as often as all the rest put 
 together. The sound-picture resulting in the mind of the 
 child before and during its own attempts at utterance might 
 therefore, roughly speaking, be put down as the average of 
 a-{-a.^-\-a -\- 31^. We are, however, not dealing with material 
 quantities, but with subtle psychological processes, and the 
 fact appears to be that the much greater frequency with 
 which one articulation (a^) is heard tends to make its share 
 in the sound-picture greatly outweigh the share contributed 
 ^ That is to say, the sountl-picture is a kind of composite photograph.
 
 2 yo Nattiral History of L anguage 
 
 by the others. The result is that the sound-picture in such 
 a case is really a^ somewhat qualified by the interference of 
 the more occasional articulations. With this understanding 
 it may be laid down that, given a normal trueness of ear, 
 the mind of the infant will develop a certain sound-picture, 
 which, for practical purposes, may be said to correspond to 
 the ' average articulation of its environment' 
 
 We must, however, remember in this connection to speak 
 only of children while they are acquiring their speech in a 
 purely natural manner,/.^, by the ear, without the interference 
 of formal instruction in purely theoretical ' correct,' as opposed 
 to ' incorrect,' pronunciations. In cases where such instruc- 
 tion is given, and where a particular articulation is held up 
 as alone to be imitated, while others are checked and 
 ridiculed, the tendency to the average is in a large measure 
 thwarted. In early and uneducated societies, or among un- 
 educated or uncritical portions of a society, this disturbing 
 artificial factor is almost wholly absent, and in the general 
 treatment of phonetic change in language in the past it may 
 be theoretically ignored. 
 
 The sound-picture in later life 
 
 Modifications in the sound-picture are not confined to the 
 period of infancy. They proceed, though in a diminishing 
 degree, throughout life. In educated societies, indeed, it is 
 usual for the individual to keep a watch upon articulation, 
 and possible innovations in his sound-picture are checked by 
 conscious mental acts. Apart from this consideration the 
 mind, after maturity, is less plastic and receptive ; impressions 
 are less vivid, and the sound-picture formed in youth grows 
 less liable to interference from influences of later environment. 
 Nevertheless some degree of shifting does actually occur, and 
 in the earlier portion of life may be very considerable, 
 whether brought about unconsciously or by effort. Its 
 effects are frequently seen in the case of persons who have 
 changed their abode from one district to another, or have
 
 Phonetic Change 271 
 
 long resided abroad, or have risen to a more cultivated plane 
 of society. In later life endeavours to adjust the articulation 
 to the new sound-picture are often as unsuccessful as similar 
 attempts to pronounce a foreign tongue. They do, how- 
 ever, indicate that the sound-picture itself has undergone 
 alteration. 
 
 Relation of the sound-picture to utterance 
 
 For the endeavours of an infant at articulation the 
 existence of a mental picture of the sound is necessarily 
 presupposed. How that psychical condition comes to relate 
 itself to physical operations of the organs of speech is an 
 obscure question which lies beyond our scope. There is, 
 of course, little or no consciousness in the first attempts. 
 Equally of course any recognisable articulation has been 
 preceded by a series of abortive efforts, which have striven 
 to bring their effect into closer correspondence with the 
 mental sound-picture. It is somewhat as with a person 
 who possesses a fairly correct mental picture of a tune or 
 an accent, and yet requires much practice to reproduce the 
 tune in singing or to mimic the accent in speech. In the 
 latter case, it is true, the process is conscious, while with 
 speech it is not yet so. Nevertheless there is sufficient 
 analogy to permit of the illustration. 
 
 When at last the efforts have resulted in actual produc- 
 tion of the sound, the series of movements through which 
 the vocal apparatus has passed become related in the brain 
 (by a species of nerve-memory) with the sound-picture, and 
 are recalled with it when the next occasion arises for their 
 use. It is not here assumed that one successful reproduction 
 of the mental picture by means of the voice will absolutely 
 fix the necessary series of vocal movements in the brain, nor 
 that the brain, in the next attempt at reproduction, will 
 absolutely control the vocal apparatus into repeating the 
 exact movements. It is only claimed that, if the adequate 
 reproduction is attained with sufficient frequency, the associa-
 
 2 72 Natu ral History of L angiiage 
 
 tion of the necessary movements with the sound -picture 
 becomes so intimate that the power of repeating them at 
 will is mastered. 
 
 Individual deviations in the sotind-picture and in utterance 
 through physical causes 
 
 It has, so far, been presumed that the child possesses the 
 trueness of hearing requisite for an accurate sound-picture, 
 and also that the vocal apparatus, being normal in its power 
 and flexibility, is capable of adequately reproducing, and 
 makes the necessary effort to reproduce, a given sound after 
 the ordinary amount of practice. If this were always so, 
 and if there were no disturbing psychological idiosyncrasies 
 at work, the child w^ould simply register the average articula- 
 tion of its environment. It would itself contribute nothing 
 towards progressive phonetic change, when it came in its 
 turn to influence others in its own neighbourhood. Indeed, 
 if all human beings had been so constituted, by far the chief 
 cause of phonetic variation would have had no existence. 
 In actual experience, however, the case is otherwise. On 
 the one hand, different minds are not impressed with equally 
 vivid or true sound-pictures. On the other, dissimilarities 
 as great as those which are manifested in external feature 
 and physique, or in character and intellectual power, exist 
 in the forms and capacities of the speech-apparatus. 
 
 Short of absolute dumbness, extreme departures from 
 the normal are at once recognised in the absence of ' roof 
 to the mouth,' in the stammer, the lisp, and that thickness 
 of speech which is vulgarly known as ' talking with a plum 
 in the mouth.' When a cold blocks the nasal passages 
 certain sounds are unpronounceable. The loss of teeth 
 causes a ' mumbling ' of certain others. In such instances 
 the physical defect is obvious. Sometimes, again, where a 
 peculiarity of the kind is not sufficiently marked to admit 
 of being named or described, the listener nevertheless feels 
 its presence. In the immense majority of cases, however.
 
 Phonetic Change 273 
 
 physical idiosyncrasies, if they ever existed, appear to 
 have been overcome. It seems as if the average speaker 
 experiences no more difficulty with one sound than with 
 another, and it is not suspected that, however unconsciously, 
 he may have been slower than his neighbour to acquire the 
 requisite flexibility for a particular articulation, or that he 
 has been compelled to counteract certain natural tendencies 
 with which his neighbour was never embarrassed. Even less 
 is it suspected that, though his deviations from the required 
 sound have been reduced to the imperceptible, they may 
 nevertheless be still in existence in some minute degree. 
 Certain it is that every individual is naturally better equipped 
 for certain articulations than for others, that he requires 
 longer practice for the adequate reproduction of those others, 
 and that, even if he attains to adequacy, he will seldom in such 
 circumstances arrive at an absolute correspondence with the 
 sound aimed at. The deviation of his articulation, however 
 slight, becomes a factor in the ' average ' to which he con- 
 tributes, and lends just so much impulse to a sound-shifting 
 within his sphere of influence. He is not only acted upon 
 by his environment ; he reacts upon it. 
 
 PJionetic predilections of the individual illustrated 
 
 Phonetic predilections of individuals are often clearly 
 manifested in infants when they are at the stage called 
 'learning to talk.' The word 'little,' attempted in exactly 
 the same environment, takes in the mouths of different 
 children the shapes ' lickle,' ' ickle,' ' ittle.' For ' this ' we 
 get ' dis ' and ' zis ' ; for ' spoon ' the forms ' poon ' and 
 ' foon ' ; for ' very ' the form ' vewy ' ; for ' you,' ' 00 ' ; for 
 * with,' ' wif,' ' wiz ' ; and so forth. In some instances the diver- 
 gence is the result of defects in the sound-picture gathered 
 through the ear. In others the ear may be at fault along 
 with the tongue. More frequently, however, the defect lies 
 in the control of the articulating organs. 
 
 The same divergences are shown in the case of persons 
 (b 884) T
 
 2 74 Natural History of Language 
 
 learning a foreign language. Two pupils under the same 
 teacher and belonging to the same environment may exhibit 
 very different degrees of aptness in seizing and reproducing 
 particular sounds. In this case imperfect apprehension by 
 the ear is frequently as much to blame as imperfect adapta- 
 tion of the mechanism of utterance. It is the same with 
 vocal mimicry. To reproduce a peculiar pronunciation, a 
 * burr,' an ' accent,' or a ' drawl,' is a rare gift, and implies an 
 unusually vivid sound-picture directing unusually flexible 
 speech-organs. 
 
 Further illustration might be adduced, but enough has 
 been said to substantiate the present point, which is that, in 
 the acquirement of speech by each new speaker, the average 
 articulation of his environment as above described, is (for 
 physical reasons alone and without regard to psychological 
 idiosyncrasies) not reproduced by him with the faithfulness 
 of a phonograph, but with modifications, derived from the 
 peculiar nature and interworking of his own hearing and 
 utterance. For the most part those modifications are 
 greatest in infancy and become minimised with increased 
 practice. They are, however, never wholly eliminated and 
 often remain considerable. 
 
 Reaction of the individual, causing a shifting in the average 
 articidation of Ids environment 
 
 So far as phonetic stability is possible in a language it is 
 only through the entire elimination of individual peculiarities. 
 But, in actual fact, these are never eliminated. In some 
 cases they are wholly invincible, in other cases only partially 
 overcome, in most cases rendered inappreciable, but never 
 entirely abolished. Thus the sound which should be w {u) 
 may be heard occasionally as a frank v, in other cases as oo 
 {u\ in others as w but with a certain colouring which, though 
 palpable, does not admit of description. If in most cases it 
 is apparently 2V proper, yet in the production of what appears
 
 Phonetic Change 275 
 
 to be the same sound, there are always infinitely subtle 
 differences. 
 
 It follows, therefore, that there are properly as many 
 dialects of a language as there are speakers of it. Absolute 
 identity of speech never existed, and would not exist even 
 if only two persons spoke a language. If, for theoretical 
 purposes, we take the environment of two persons A and B 
 to be practically identical and state its average articulation 
 at a given time as x, the articulation of A will be x modified 
 by the a which (however small) represents the uneliminated 
 element of his peculiarities, while correspondingly that of B 
 will be X modified by b. Hence it follows that, with the 
 continual flux of the units and the interpenetration of the 
 generations forming any society, and with the ever-varying 
 degrees in which their reciprocal influences are felt, the 
 average articulation cannot remain constant at x. The 
 average of x (a), x {b), x (c), etc., shifts from x to x (j), and 
 this, in turn, is soon replaced by the x (s) which is the 
 average of newly developing constituent units, namely, 
 X (j/) modified by a, x (j/) modified by /3, etc. Individually 
 the elements a, b, c, etc., are mostly inconsiderable, and the 
 displacement of the average (represented by y) very slight. 
 It is, of course, the same with the further individual displace- 
 ments a, ;8, 7, etc., and their average displacement z. But a 
 number of inconsiderable displacements at length become 
 very considerable in the aggregate, and the language is then 
 found to have undergone a distinct ' phonetic change.' 
 
 Counteraction of individual influences never complete 
 
 It might perhaps be imagined that, in spite of the slight 
 physiological deviations of the individual, it will result by 
 the ' law of averages ' that, as disturbing elements in the 
 aggregate speech of the community, they will cancel each 
 other. In practice, however, no such complete counteraction 
 takes place ; nor is it a priori likely. If we suppose, for
 
 2 76 Natural History of Language 
 
 example, that there is a sporadic tendency to pronounce w 
 {u) as V, this will not be corrected by any different tendency 
 to pronounce iv in some other manner. A substitution by 
 some individuals of / for r will not be counteracted by a 
 substitution of, say, w for r on the part of others. A speaker 
 is guided by a sound-picture, which he aims at reproducing. 
 This sound-picture is derived from a series of impressions. 
 A child hearing w regularly pronounced as such, but 
 occasionally as v, will not have the sound-picture restored 
 to consistency by occasionally, on the other hand, hearing w 
 pronounced as 00 (u). The result would not be correction, 
 but simply a further disturbance. If he hears r habitually 
 articulated as such, but sometimes as /, no counteraction of 
 the interference with the sound-picture caused by the latter 
 will result from hearing the sound now and then pro- 
 nounced as w. Nature does not work in such a way that a 
 sound, sporadically mispronounced in one direction, is also 
 sporadically mispronounced in exactly the opposite direction 
 in just as many and as influential individuals, or just as 
 frequently. A stable average would perhaps ensue if this 
 were the case. Thus, if the sound e were articulated by 
 some persons with too much ' rounding,' by the same number 
 with just as much too little, by some too far ' back ' and, 
 by an equal number, too far forward, there might be a 
 mathematically precise cancelling of differences. In experi- 
 ence, of course, such a state of things does not occur. The 
 * average ' in that sense is a fantastical abstraction. There 
 may, as a fact, be very few with a tendency to ' narrow ' 
 too much, while many may have a tendency to * widen.' 
 Many, again, may tend to make w labio-dental instead of 
 pure dental, but a counteracting degree of ultra-labialism 
 is not likely, even if conceivable. 
 
 The counte7'acting forces in primitive societies 
 
 In pre-literary or pre-educational conditions of society 
 the vocabulary is comparatively small, the recurrence of
 
 Phonetic Change 277 
 
 any particular word is therefore more frequent, the sound- 
 picture and its reproduction are more likely to be brought 
 into closer relation with the average articulation. Moreover, 
 the language being acquired entirely by the sounds through 
 the ear, with no interference from any impression of a word 
 as read by the eye, there is so much the less danger to 
 homogeneity of pronunciation. Important, too, is the con- 
 sideration that at such a period there are no such things as 
 various strata of society distinguished from each other in 
 their pronunciations through differences of culture. The 
 members of the speech-community are on a level in that 
 respect. Their articulation is so far homogeneous, and the 
 resulting sound-pictures are impressed with a more distinct 
 consistency. There are thus practically no starting-points 
 for phonetic change, except in the slight aberrations, physical 
 and psychical, of the individual. Nor is it to be supposed 
 that any large deviation in pronunciation on the part of an 
 individual would escape criticism and correction in primitive 
 communities. The absence of the schoolmaster would be 
 more than compensated by the well-known fondness of 
 barbarians for mockery. Excess of individuality and 
 departures from the norm are more strongly resented in 
 early clan-life than in the complex civilisation of populous 
 and extensive societies. The striving to attain to the normal 
 would be even greater within a group of close association 
 than under modern conditions, where the normal is less 
 easily ascertainable, thanks to the wider range of intercourse, 
 and to the social, educational, and local differences marking 
 the individuals within that range. We may therefore lay 
 it down that, in the earliest periods, phonetic change in a 
 language over a certain area would proceed only by im- 
 perceptible stages, through a series of minute individual 
 deviations of which neither speaker nor hearer was made 
 conscious, but which would nevertheless ultimately involve a 
 shifting of the average articulation in the manner above 
 described. Such a shifting, however gradual, is inevitable.
 
 278 Natural History of Language 
 
 Growth of dialects and spread of phonetic change : 
 from ifidividual to group 
 
 In tracing the ordinary course of phonetic change we 
 may be allowed to conceive, in the first instance, of a small 
 aggregation of persons whose speech is as nearly as possible 
 homogeneous. There are as yet no dialects of either locality 
 or class. We may suppose the individuals forming the 
 speech-community over a small geographical area to be in 
 continual and equal communication with each other. The 
 extremes of phonetic divergence are quite inconsiderable. 
 Moreover, in such a society there naturally exists a close 
 degree of blood-relationship, and therewith, it is to be pre- 
 sumed, a nearer physiological resemblance of the individuals. 
 Their vocal capacities are likely to differ less than in large 
 populations of mixed descent. Those conditions of climate 
 and life which exert some influence upon articulation are 
 the same for all, and may be disregarded in our present 
 consideration. In these ideal circumstances the shifting 
 of an average articulation, though it must occur, would 
 proceed only with the slowness of glacier motion. 
 
 The circumstances, however, are too purely ideal to 
 have ever actually existed. A community in which there 
 is equal linguistic intercourse between all the individuals 
 belongs only to theory. If once the area of the common 
 speech extends beyond a single village or camp, it will 
 comprise a number of districts, within each of which com- 
 munication between the individuals will be close and frequent, 
 while the communication of any of them with the members 
 of another district must be intermittent and may be very 
 rare. Even in a single district special knots and groupings 
 of the units will form themselves, through nearer kinship or 
 neighbourhood or some common sentiment or interest. 
 
 The smallest instance of such a minor group is the 
 household, between whose members linguistic intercourse 
 is doubly effective, inasmuch as it is both earliest and most 
 continuous. It is still no rare thing to find all the children
 
 Phonetic Change 279 
 
 of the same family employing with peculiar uniformity some 
 distinguishing articulation, for which they received the hint 
 in infancy, which they have accentuated by acting and 
 reacting on one another, and for which their common 
 parentage supplies a common predisposition, on either the 
 physical or the mental side, or both. The average articula- 
 tion of the household thus forms a rudimentary dialect, 
 kept in check by the counteracting influence of the average 
 articulation of the wider group, but itself reacting on that 
 wider average. The larger the household, the greater its 
 effect upon the group, an effect made all the more telling 
 by its solidarity. But in the household itself the effect of an 
 individual is often disproportionately large. The ' average ' 
 of the house may be mainly determined by the father, the 
 mother, the eldest child, or whoever most sets the pattern 
 for the younger family in general, whether by having most 
 occasion for imparting the lesson of speech or the greatest 
 personal impressiveness or ' magnetism ' in imparting it. 
 Hence a peculiarity which was individual may come to 
 colour the speech of a whole house, and, if the house be 
 large or its members personally impressive or even loquacious, 
 some degree of that colour may still further force itself upon 
 the next larger group in which they play their parts. 
 
 If, to take a crude example, the sound e{ = a\x\ fate) tends 
 in the mouth of A to be corrupted into di (i.e. nearly the i 
 in bite), and if A linguistically dominates a family, the sound 
 di may be so imposed upon the house that its correction by 
 the wider average is never effected. The household of A 
 therefore contributes this same di to the aggregate articula- 
 tion of the larger group. Should the family be numerous 
 and its members linguistically much in evidence, the sound 
 di is rendered unusually familiar to the group, and must 
 exercise some influence upon the sound-picture of all in 
 that environment. Moreover, diseases of language are 
 highly contagious. The same physical or mental causes 
 which produced the change of e to di in the mouth of A 
 are nearly always present in greater or less strength in, say
 
 28o Natural History of Language 
 
 D, G, K, M, S, or other individuals who come into the 
 group. One or other of them may be already distinguished 
 by a similar deviation. Upon these the effect of hearing 
 the sound with appreciable frequency is to encourage them 
 in their own natural tendency in the same direction. The 
 mental sound-picture impressed by the environment is thus 
 infected with the .?z-sound, and is therefore not ready to 
 restrain the vocal organs from the corrupt articulation to 
 which they naturally incline. Hence the pronunciation di 
 grows in frequency within the group. If hypothetically we 
 isolate that group, the path of a phonetic change is clear. 
 It may chance at last that the sound di exceeds in frequency 
 the original sound e. If it is intrinsically an ' easier ' sound, 
 i.e. one actually requiring less muscular effort in its produc- 
 tion, the contagion will rapidly bring about this result. If 
 not, its progress will be slower and less certain. In 
 proportion, however, as the variation comes into greater 
 use, the older articulation loses its prominence in the sound- 
 picture, both of the persons already in the environment 
 and still more of the new generation who come into it. 
 Ultimately it is lost altogether. The phonetic change 
 from e to si is thus accomplished, only, in all probability, to 
 be itself succeeded in due time by another displacement, which 
 will be gradually effected in the same way for similar causes. 
 It thus appears that a peculiarity of individuals may 
 possibly be established as a phonetic peculiarity of an entire 
 linguistic group, such as a hamlet or tribal camp. If the group 
 is in a large measure isolated from others, the new development 
 will, of course, be more rapidly accomplished than if the group 
 is in frequent association with other groups. For the effect of 
 group on group is similar to, though less than, that of indi- 
 vidual on individual. Where it does not wholly check, it may 
 retard or minimise incipient deviations. Thus if the figure
 
 Phonetic Change 281 
 
 be taken as a crude representation of the linguistic relations 
 within a group (a) of households A, B, C, D, the intercourse 
 of a with other groups /S, 7, 8 may be correspondingly 
 represented by 
 
 There will be, once more, a certain amount of levelling and 
 averaging within this wider combination, due to the same 
 sort, though not degree, of interaction of its component 
 groups as that which occurred in the case of individuals. 
 As with the individual, if the intercourse is comparatively 
 frequent, we shall find on the one hand a counteraction of 
 some of the sectional tendencies, on the other a wider 
 currency gradually acquired by some particularly strong 
 deviation for which the conditions are favourable. Often, 
 of course, it is the case that a deviation which originated 
 from natural causes in a has, from the same natural causes, 
 meanwhile shown itself to some extent in /3, 7, or S also, 
 and though any one of these groups, if left to itself, would 
 exhibit the peculiarity less markedly than a, or even suppress 
 it, there will obviously exist so much wider predisposition 
 to the sound which a contributes. Thus if the pronunciation 
 di for e has become peculiarly strong in group a in the 
 manner above detailed, then, if group a counts for much in 
 respect of influence within the sphere of intercourse, and if 
 predisposition favours the easy acquirement (or encourages 
 the contagiousness) of the sound di among the members 
 of another group, /3 or 7, or both, the result will be 
 a widening use of that sound, till it becomes general and 
 finally ousts the other from the sound-picture of a new 
 generation. 
 
 We may in this way, from a peculiarity marking (i) the 
 speech of the individual (individual dialect) and (2) the 
 speech of his everyday group (minor local dialect), arrive at
 
 282 Natural History of La7iguage 
 
 (3) a peculiarity marking the speech of associated groups 
 (local dialect proper). 
 
 A further step on the same lines is readily conceivable. 
 One district or association of groups may be in linguistic 
 intercourse with other districts, and action and reaction will 
 proceed as before. The intercourse, however, naturally 
 grows less close and continuous as we widen the field, and 
 the forces which either cancel or generalise a sectional 
 peculiarity are slower and feebler in their operation. 
 
 Distinction between co7iditions of dialectal growth among 
 primitive and cultured peoples. In primitive peoples the 
 groups mostly isolated 
 
 Here we must draw a distinction between the ordinary 
 condition of primitive and barbarian peoples and that of 
 modern civilised communities. In primitive times it 
 generally happens that groups, either of the larger or 
 smaller kind, are both geographically and socially isolated 
 in a degree unknown to advanced civilisations, A village 
 or tribe, or at least a local federation of tribes, leads its own 
 life, and exercises little communication with others, though 
 linguistically akin. Often it is in actual hostility to them. 
 It possesses its own valley and keeps a mountain between 
 itself and its neighbours, or segregates itself behind rivers 
 or other natural barriers. The result is the accentuation of 
 local peculiarities, and the absence of any levelling influences 
 beyond the group or association of groups which happens to 
 be included in the tribal union. Occasional assemblages of 
 a wider kind for religious purposes, military leagues, etc., 
 may be ignored. A language of which the speakers are 
 thus divided soon splits into a number of dialects, which 
 work further and further away from each other, as very 
 different series of new phenomena rise successively within 
 each one and become universalised among its speakers. In 
 a rough figure we may give the situation as
 
 Phonetic Change 
 
 283 
 
 where the divisions I. II. III. IV. represent dialects formed 
 by isolation as above described, and where the single division 
 may comprise one or several minor groups. 
 
 When groups are not isolated there is contiftuous transition 
 and interlinking of dialects 
 
 The above is not, of course, invariably the case. In 
 some instances, over a considerable tract of country, there 
 may be, even among barbarians, no distinct severance either 
 by natural barriers or by tribal unfriendliness. There will, 
 indeed, always be a series of groupings of individuals with 
 closer intercourse internally, but it is possible that each such 
 group may be in more or less frequent communication with 
 neighbour groups on either side. Thus we may picture a 
 language spoken by divisions intercommunicating, and to a 
 certain extent phonetically interacting, across the border 
 lines indicated in the fio^ure 
 
 I 
 
 T 
 
 II 
 
 
 
 III 
 
 VII 
 
 
 \ 
 
 V 
 
 VI 
 
 
 <) 
 
 
 I 
 
 VIII 
 
 
 where each of the sections, I. II. III. etc., may be taken 
 to represent an area over which internal intercourse is 
 habitual, thanks to the use of some common centre for 
 industrial, religious, or other purposes, or to closer tribal
 
 284 
 
 Natural History of Language 
 
 affinity, or merely to the ' lie of the land.' Over each such 
 region speech becomes comparatively homogeneous ; the 
 district possesses a common dialect. Nevertheless, if inter- 
 course is free along the borders, a particular section will not 
 develop its peculiar phonetics in entire independence of the 
 effect of the adjoining divisions. The speech of section V. 
 must reckon with such interferences with its average articula- 
 tion as proceed from the influences of sections I. II. IV. VI. 
 VIII. across its borders. Obviously the strength of such 
 influences will be slight as compared with the strength of a 
 tendency operating in the heart of the individual section. 
 So far as it is exerted, it will naturally tend to give to each 
 division a phonetic colouring somewhat nearer to that of an 
 immediate neighbour than to that of a remoter district {eg. 
 Ill, or VII.). It is further obvious that, if the country were 
 inhabited with fair equality throughout, those who dwelt 
 farthest from the centre of each district and nearest to the 
 borders of the next would probably have a somewhat less 
 everyday intercourse with their own centre, and would 
 therefore attain less nearly to the ' average ' of their district. 
 On the other hand, they would be in somewhat more frequent 
 contact with the borderers of the neighbouring district, and 
 would so much the more approximate to them. We should 
 thus be compelled to readjust the dialectal relations as in 
 the following scheme 
 
 - T 
 
 1 II ; 
 
 1 
 
 i III 
 
 1 L 
 
 H-- _ ^-^. __ -^ 
 
 
 1 
 
 
 .1 
 
 ; > 
 
 
 VII 
 
 \^ ~--K- 
 
 \. IV 
 
 r T T 
 
 
 ' ['. 
 
 
 \l 
 
 i VIII ; 
 
 
 where the dotted lines indicate the continuous overlapping 
 of dialects through districts which are transitional both 
 geographically and linguistically.
 
 Phonetic Change 285 
 
 Continuous transition in phonetic phenomena 
 
 Treating the matter from the point of view of the total 
 range of a phonetic phenomenon, rather than from that of the 
 district of intercourse, we may express the same interpenetra- 
 tion of dialects in the following form : — 
 
 If the three circles represent the areas over which are spread 
 the phonetic phenomena a, b, c respectively, it is implied 
 that the phenomenon a will extend over portions of the area 
 occupied also by the phenomena b and r, and that the case 
 will be the same with b and c. Thus the area I. will be 
 marked by a and c, II. by a, b, and c,\\l.hy a and b, IV. by 
 b and c. It thus results that, in the ideal circumstances of 
 a speech-domain socially and geographically unbroken, each 
 area of speech is phonetically linked in an equal measure 
 with its neighbours on either side, somewhat after the fashion 
 of the arrangement in chain-mail. Such a domain will there- 
 fore normally consist of a series of dialectal areas between 
 which no strong lines of cleavage can be drawn, and which 
 find their closest affinities in their nearest neighbours. In 
 actual experience this is almost exactly the case with the 
 High and Low German-spoken dialects, as they pass from 
 the limits of the South German peoples to the sea-border on
 
 286 Natural History of Language 
 
 the North, and also with the dialects which extend over the 
 Romance areas of Spain, France, and Italy. 
 
 Conditions of phonetic cJiange among cultured peoples : 
 chajige inevitable 
 
 Under modern civilised conditions dialectal variations 
 beginning with the peculiarities of individuals are continually 
 springing into existence, and either (as happens in most 
 cases) being suppressed, or else winning their way over small 
 areas, large areas, or the whole linguistic domain. Ease of 
 locomotion and communication, the spread of literature and 
 of education (largely carried on by those who have been 
 born in one district, trained in at least one other, and sent 
 to teach in a third), tend to thwart local developments and 
 to produce a general levelling throughout the community. 
 Where the standard articulation of a people is adequately 
 represented in writing, possible deviations in the sound- 
 picture, due to the idiosyncrasies of local environment, are 
 strongly counteracted by the effect of the word-picture as 
 impressed through the eye by frequent reading. To attain 
 and preserve absolute homogeneity is out of the question, 
 but the forces which make for that result are certainly more 
 potent than in times when travel was difficult and education 
 rare or absent. 
 
 On the other hand, however, there arises, in some 
 languages, and particularly in one with an orthography so 
 inconsistent as the English, a new danger to phonetic 
 stability, springing not from physical but from psychological 
 causes. Education directs the attention of the individual to 
 his own articulation and that of others, and conscious 
 attempts are made to amend pronunciation according to 
 some other standard than that of the average of the 
 environment. In some cases the pronunciation of a certain 
 social class is deliberately imitated, and not always with 
 success. In others there is a deliberate adaptation of
 
 Phonetic Change 287 
 
 pronunciation to spelling, either because a word-picture has 
 first been impressed on the mind from reading and not from 
 hearing, or else because it is fancied that the usual pronuncia- 
 tion must be wrong because it does not correspond to the 
 orthography, or because it does not seem in logical agree- 
 ment with cognate words. Thus the modern pronunciations 
 of tea, oblige, Rome, yellozv, instead of te {tay), oblige {obleege), 
 Rfiin {roovi), yalh, are the outcome of what was originally 
 either ignorance, affectation, or some misguided mental 
 operation. The frequent pronunciation of ' again ' as agane 
 {agen) is due to the spelling, as is that of conkwer for 
 ' conquer.' A particularly obvious case of the spelling affect- 
 ing pronunciation is that of such words as herb, Jiospital, 
 in which the ' //-mute ' has come to be //-aspirate. In the 
 first instance, no doubt, the aspiration began with those who, 
 having the orthographic picture of the word before their 
 minds, were afraid of appearing to ' drop ' the //. The 
 occasional pronunciation knoivledge (as in ' knoiv ') is 
 probably due to a mistaken sense of logical consistency. 
 Languages in which orthography is more strictly phonetic, 
 such as Italian, are not liable to the same interference in 
 any palpable degree, but the effects upon English are already 
 considerable. 
 
 At a given time, therefore, in an educated country 
 possessed of a common language, we have at work these 
 disintegrating phonetic forces : (i.) the perpetual and 
 inevitable cropping-out, now here, now there, of individual 
 peculiarities, mainly of a physical nature, which affect dif- 
 ferent sounds in different places, and which may chance to 
 exhibit themselves with especial strength at a given period in 
 some particular district, or in more than one district at the 
 same time. These may be for the most part successfully 
 checked, or at least confined to the special district or districts 
 as dialectal phenomena, but in some cases they will force 
 their way into wide, then general, then universal acceptance : 
 (ii.) psychological operations, under which head may be in- 
 cluded phonetic affectations, workings of the analogical
 
 2 88 Natural History of L anguage 
 
 instinct, and adaptations of the sound to the spelling. These 
 arise as sporadically as the physical deviations, but are less 
 likely to be confined to a special area. Rather they spread 
 first through a social class. These are more deliberate, and 
 therefore persisted in, and, thanks to the similar workings of 
 men's minds, are more highly contagious. 
 
 It is impossible that these forces should in all cases 
 fail to make abiding impressions on the general lan- 
 guage, that, in every single instance and form in which 
 they are exerted, the ' average articulation ' should still 
 remain unaffected by them. There seldom arises strongly 
 in any one quarter a peculiarity which does not appear 
 sporadically in other quarters, or for which there is not a 
 frequent predisposition. Pronunciation is therefore necessarily, 
 to use the expression of Heracleitus, " in a state of flux," 
 however gradual and indefinable may be the displacement 
 in a given generation. 
 
 Phonetic change rarely cotifined to one sound at a time 
 
 Hitherto we have treated the inevitable growth of dialect 
 and the spread of phonetic change as proceeding in regard 
 to one particular sound {e.g. of e to 9i). It must, however, 
 be obvious that not one, but a number of different sounds 
 may be affected, either simultaneously in the same quarter, or 
 simultaneously in different quarters, or at successive periods 
 in either the same or different quarters. Thus a ' Cockney ' 
 change of e to di may chance to be simultaneous with a 
 shifting of a (as in bank) to e (i.e. benk), or oi a { = ah as in 
 laugh) to the sound represented in law/. Some such further 
 deviations may easily occur within one and the same region, 
 since it is very improbable that a marked displacement of 
 one vowel-sound (no matter whether it be due to physical 
 difficulty or to indolence or to affectation) should be accom- 
 panied in the same mouths by rigorous exactness in the 
 articulation of every one of the rest. The old English
 
 Phonetic Change 289 
 
 vowel a ( = ah) has become in the modern speech ; but 
 similarly e, I, o, ft (i.e. ay, ee, 0, 06) have become respectively 
 f, di, tl, an (i.e. ce, J, 00, ozv as pronounced in present-day 
 English). 
 
 Few of the sounds in a given language will be left with- 
 out occasional deviations of the physical kind above discussed ; 
 most of them will tend to some sort of corruption at some 
 time or times in some place or places. The probabilities are 
 that, for this reason alone, a dialect will be marked by more 
 than one peculiarity, and that several phonetic changes will 
 be making progress in a language at the same time, though 
 in different stages of their development and with different 
 degrees of rapidity. 
 
 Moreover, sounds are often so related in place or kind of 
 articulation that a physical difficulty in regard to one implies 
 a certain amount of difficulty in regard to others in the same 
 class. The deviation may be concerned with a particular 
 place in the vocal passage at which the articulation is to be 
 effected. Thus in the series of dentals including the stops 
 t, d, the interdental spirants >, 5, and the dental spirants s, s, 
 z, ^, it is not likely that a shifting of one of these, due to 
 some embarrassment at the place, will be altogether dis- 
 sociated from conditions influencing the production of one or 
 other of the rest. Or the difficulty may be concerned with 
 a particular kind of articulation, for instance, with that of 
 a liquid, of a spirant, or of an aspirate. The causes which 
 tend to a deviation in the case of one in the series are 
 naturally such as will in some way affect another. If th 
 (i.e. /-aspirate, not ]> or 6) is a sound uncongenial to a certain 
 organisation, it is probable that other aspirated stops, ph, kh, 
 dh, etc., will also be uncongenial, though not necessarily in 
 the same degree. In Greek the sounds 6 {th), (f) (ph), x (i^'h) 
 have all undergone modification from aspirated stops to 
 spirants. 
 
 Again, it must be remembered that in language proper 
 single sounds are not articulated independently, but in com- 
 bination with others, and that the sounds which are uttered 
 
 (B 884) U
 
 290 Nahiral History of Language 
 
 in one breath-group {i.e. with one impulse of air) are Hnked 
 to each other in very subtle ways by ' glides,' which vary in 
 their nature with every change in the sounds which they link. 
 Hence a change of one sound in such a breath-group may 
 induce a change in another sound of that group, not because 
 of any difficulty in regard to the latter sound in itself, 
 but because in this particular conjunction it becomes less 
 congenial than some other. If, for example, a vowel 
 between two consonants is so much slurred in a hasty or 
 careless pronunciation that it finally disappears, the synco- 
 pation of the syllable may bring together two consonants 
 which are phonetically incompatible, and a certain amount 
 of assimilation may take place in one or both. Similarly 
 the nonarticulation of a consonant may bring together 
 two vowels which may therefore need a certain mutual 
 readjustment. 
 
 Finally, it seems possible to detect the influence of 
 unconscious psychological operations, which cause one dis- 
 placement to be followed by others throughout a series. It 
 seems hardly probable that the wholesale shifting of the 
 long vowels in English, bat to boat, fit Vo feet, niys to mice, 
 bot to boot, mils to mouse, should have been brought about 
 simply by physical causes, inasmuch as there is no avoidance 
 of the several vowel-sounds as such, but only a progressive 
 transposition of them. It would appear that when, among 
 the series of sound-pictures existing in the mind, one was 
 displaced, there followed a quite unconscious rearrangement 
 of the others relating to the new sound. The shifting of 
 Teutonic stop-sounds, known as the working of Grimm's Law 
 {Lautverschiebiing^, appears to owe something at least to the 
 same psychological influence. 
 
 That dialectal variation or general phonetic change should 
 be confined to one particular sound is therefore not merely 
 contrary to experience, but also unlikely a priori.
 
 Phonetic Change 291 
 
 Fortuitous agreement of widely separated dialects in some 
 particular sound-change 
 
 Since the sound-changes occurring within one dialectal 
 area may come to be more or less numerous, while at the 
 same time other series of more or less numerous changes are 
 proceeding in other areas, it may easily be the case that two 
 such areas, between which for geographical reasons there is 
 little or no direct communication, may accidentally and in- 
 dependently agree in one or more of the changes which they 
 produce, however widely they may differ in the rest. We 
 may imagine, for example, that over a district (I.) at the 
 Western extremity of the whole speech-domain there is 
 among other physical tendencies which become habits, a 
 habit of turning the palatal guttural k into a spirant c before 
 certain vowel-sounds. It is readily conceivable that the 
 same tendency may, for identical reasons, develop itself and 
 become an established habit in a district (VIII.) on the 
 Eastern borders. That particular sound-change might, there- 
 fore, be a mark of both dialects, while their other changes 
 might show no such coincidences. Similarly a dialect (II.) 
 on one side of the speech-domain may regularly change o 
 into a, while precisely the same change is in progress in a 
 dialect (VIII.) on the other side. Meanwhile, dialect II. may 
 preserve e as /, whereas dialect VIII. transforms this also to 
 a. Generally speaking, of course, any two areas which join 
 or interpenetrate each other will exhibit more points of 
 agreement than two which are more remote, simply because 
 of their interaction as already described. Nevertheless we 
 must not overlook the possibility of even a considerable 
 number of fortuitous coincidences between two dialects in 
 the case of which there is no longer phonetic interaction. If, 
 therefore, the list of deviations occurring in various parts of 
 a whole speech-domain be represented by a, b, c, d, e,f, etc., 
 and the several dialects by A, B, C, D, E, etc., it may be the 
 case that A is marked by the peculiarities a, c, d, //, B by b,
 
 292 Natural History of Language 
 
 ^> <^> yi C by c, d, e, k, but D by a, c, k, I, and E by b, c, g, h, 
 as in the figure. 
 
 It is therefore not safe to conclude that, when two lan- 
 guages derived from a common ancestor coincide in a given 
 phonetic practice, they are necessarily descended from dialects 
 which once adjoined each other in the original speech-domain. 
 Such an inference must be based upon more ample data, and 
 can grow into a fair hypothesis only in proportion as the 
 coincidences become either specially numerous or specially 
 striking. 
 
 Suviniary of tJie above considerations 
 
 It has thus far been shown how phonetic changes begin 
 with the individual, how they spread or are checked, and 
 how, if not checked, they may establish themselves as 
 dialectal varieties. The forces producing and counteracting 
 change have been examined both for educated and un- 
 educated societies. The a priori impossibility of a complete 
 counteraction in all instances has been demonstrated. The 
 ordinary case of a geographically or socially interrupted 
 speech-domain has been set against the ideal case of one in 
 which intercourse is evenly distributed. In the former, lines
 
 Phonetic Change 
 
 293 
 
 of cleavage soon grow distinct. In the latter, it has been 
 shown how dialects are naturally interlinked, so that each 
 is transitional to its neighbours on either side. It has been 
 observed that, not one, but a series of sound-changes must 
 be expected in a dialect, and that there may occur entirely 
 chance agreements in special phenomena between dialects 
 occupying areas remote from each other. 
 
 It remains to contemplate the further phonetic diver- 
 gences which may cause the dialects to become mutually 
 unintelligible, or, in other words, distinct languages. Diver- 
 gences in grammar and vocabulary, of course, proceed along 
 with these, but for theoretical purposes must be treated 
 separately. 
 
 Effects of neighbouring aliens 
 
 We have hitherto spoken of the total linguistic area as if 
 it were isolated from any influence of entirely alien speeches 
 surrounding it. Its dialects have been treated simply in 
 relation to each other, and some degree of cohesion and 
 geographical continuity of the groups has been assumed. 
 In historical experience, however, the state of things is 
 ordinarily otherwise. To begin with, the speakers along the 
 frontiers of any extensive speech -area usually come into 
 some sort of contact or collision, and frequently into some 
 measure of intermixture, with speakers belonging to adjoining 
 domains occupied by alien tongues. The latter often possess 
 phonetics of a character very different from those prevailing 
 in the domain with which we are specially concerned. Thus 
 it may happen that, at and about the points A, B, C, D on 
 the borders of the area 
 
 A
 
 3l- 
 
 294 Natural History of L anguage 
 
 there may be situated neighbours whose languages are 
 altogether unrelated either with the tongue enclosed or with 
 each other. One of these (A) may be distinguished by an 
 abundance of palatal sounds, another (B) by a fondness for 
 strong aspirations, the third (C) by a dislike of spirants, the 
 fourth (D) by an absence of aspirates. Add to these the 
 possible peculiarities of stress and pitch accent, of rapid or 
 drawling pronunciation, or of nasalisation, which may be vari- 
 ously represented in these various quarters. Any continued 
 intercourse, such as usually takes place where natural barriers 
 do not exist, between the people within and those beyond 
 the speech frontier, will result in a certain interference with 
 the phonetics of both parties, and the outlying dialects will 
 tend to move away from those in the centre for reasons un- 
 felt by the latter. Where English borders on Welsh it 
 takes a kindred phonetic colouring, and in South Africa 
 the Hottentot ' clicks ' have in some cases proved con- 
 tagious to adjoining speakers of the Bantu languages. 
 
 Effects of complete geographical separation, by natural 
 expansion or by specific migration, etc. 
 
 We may next suppose the geographical, and consequently 
 the social, continuity of the dialects to be broken up. This 
 may occur through a natural expansion of population and an 
 outward movement (which proceeds either like a wave 
 spreading gradually on all sides, or else by specific migrations 
 of particular portions of the linguistic body), or else through 
 the invasion of aliens, who may wedge themselves in between 
 larger or smaller sections. In the first case, that of simple 
 expansion, territories will come to be occupied which are 
 separated from each other by mountain ranges, rivers, seas, 
 or other natural partitions, and the peoples occupying such 
 territories will be practically cut off from communication 
 with the rest, and will be left to develop their phonetic and 
 other idiosyncrasies without further appreciable counteraction 
 from cognate dialects.
 
 Phonetic Change 295 
 
 When special bodies migrate, whether for conquest or 
 under external pressure, or simply in a nomadic quest for 
 pastures new, the same result will occur, but the more readily, 
 inasmuch as such bodies are likely to be carried into a com- 
 plete isolation from their linguistic kin. It must, however, 
 be added that such a specific migration need not be made 
 by a population speaking merely a single dialect. It may 
 consist of a number of detachments drawn from various 
 dialects, which may be brought into a closer intercourse in the 
 new quarters and so produce a new phonetic ' averaging ' 
 there. This was the case with the tribes of Angles, Saxons, 
 Jutes, and Frisians in England. Nevertheless, generally 
 speaking, such migrations will naturally be those of a popula- 
 tion occupying a particular tract of territory, intercom- 
 municating and therefore mostly speaking dialects which are 
 closely cognate, if not the same. 
 
 Where the inwedging of an alien stock takes place, the 
 consequence will obviously be much the same. Such inter- 
 wedging and splitting apart appear historically in the experi- 
 ence undergone by the Celtic languages in Great Britain and 
 by the Finno-Ugric tongues in Russia, in the position of 
 Magyar in Central Europe, and in the case of the earlier or 
 Dravidian tongues of Hindustan, among which Brahui is 
 sundered from the languages of the Deccan by speeches 
 which are Indo-European. Each of the other forms of dis- 
 placement is also historically demonstrable. Among striking 
 instances of specific migration it is easy to recall that of the 
 Celts from Gaul into Asia Minor in 280 B.C., which left a 
 Celtic language in Galatia for at least seven hundred years ; 
 that of the Goths into Moesia ; and that of the Anglo- 
 Saxon dialects into Great Britain and subsequently into 
 North America and Australasia. 
 
 In whatever way geographical dissociation is effected, the 
 result is the same in kind. The cessation of intercourse 
 with cognate dialects causes an accentuation of the phonetic 
 peculiarities with which a given division begins its separate 
 existence. Each such separated division will also continue
 
 296 Natural Histoiy of Lmtguage 
 
 to develop new dialectal peculiarities within itself, in pre- 
 cisely the same way in which such phenomena arose in 
 the primitive linguistic community. These will contribute 
 perpetually to the formation of the ever-shifting ' average 
 articulation ' of that division. In the multiplicity of phonetic 
 innovations which continually struggle for existence the 
 chances are that, while one set is establishing itself in this 
 division, an almost, if not entirely, different set will be 
 establishing itself in the other cognate divisions. Add these 
 accumulating deviations to the now unfettered play of the 
 original phonetic tendency which marked each division while 
 it was but a local dialect of the older area, and the com- 
 plete phonetic divergence of the several divisions to the 
 point of mutual unintelligibility is only a matter of time. 
 
 Effects of climate and manner of life 
 
 It is easy to overstate the effects which climate and 
 manner of life exert upon speech. Nevertheless these supply a 
 factor which cannot be ignored. Mr. Sweet observes that in 
 cold countries there is less disposition to open the mouth 
 wide, and he offers this fact as an explanation of the Germanic 
 tendency to substitute for a. The extent to which the 
 front sounds, labials and dentals, may prevail over the back 
 sounds, such as the gutturals, or the open sounds over the 
 close sounds, or vice versa, may be in no inconsiderable 
 measure determined by atmospheric conditions. These affect 
 the speaker in his employment of the breath-stream which 
 forms the raw material of speech. They affect also to some 
 extent the constitution, or at least the flexibility, of the 
 speech-organs themselves. It follows therefore that partialities 
 for certain sounds, or certain degrees of ease and precision 
 in delivering them, may (in the case of the posterity of the 
 migrating bodies) depend on such considerations as whether 
 displacement has been northwards or southwards, towards 
 the sea or the mountains. They may also depend in some 
 slight measure, which is not ascertainable, and which can
 
 Phonetic Change 297 
 
 only be hinted at, upon the question of diet and of the 
 physical practices and material environment of life. Further 
 elaboration on these points might become fanciful. 
 
 Effects of mixture with alien peoples 
 
 Of much more importance is it to observe that a popula- 
 tion which either expands or migrates is not likely to find 
 unoccupied the ground upon which it trespasses. Other 
 peoples with other languages are already in possession of 
 the field. Sometimes these are gradually driven forwards, 
 either to infringe on yet others or else to make a last stand 
 in the corners, the promontories or mountains, beyond which 
 further movement is barred. The British ' Iberians ' have 
 fared in this way at the hands of the Celts, and the latter at 
 the hands of the x'\nglo-Saxons. At other times, when their 
 numbers are very great, and an onward movement is im- 
 possible, they remain as the staple population of the territory 
 in question, perhaps admitting the newcomers as their 
 conquerors, but gradually absorbing them more or less 
 completely into their own stock. In the latter case, 
 the newcomers may either succeed in imposing their 
 own language, as with the Aryans in Northern India, or 
 they may be insufficiently numerous or otherwise influential 
 to do so. The lingua romana of Gaul was but little affected 
 by the invasion and reign of the German Franks, nor did 
 the lingua romana of Spain yield appreciably to the Teutonic 
 of the Visigoths and Vandals. The failure of Norman 
 French in England was less utter, but sufficiently decisive. 
 Even the Arabic of the powerful and cultivated Moors, which 
 had been imposed upon all Northern Africa, failed in Spain. 
 Obviously if the language of the intruders loses itself in the 
 pre-existing local speech, we are here no further concerned 
 with it. If, on the other hand, it overcomes the native 
 tongue and establishes itself throughout the mixed popula- 
 tion, it will not accomplish its conquest without marked 
 effects upon itself So long as the invaders keep to them-
 
 298 Natural History of Language 
 
 selves, in the sense of forming a pure caste in the midst of 
 an aHen people and language, their speech will be affected 
 but slightly by their new conditions. Yet it will be affected 
 somewhat. Certain novel articulations used by the sur- 
 rounding natives will gradually become familiar to the ear 
 and will leave some effect on the sound-picture, particularly 
 of children born into the new environment. The passage 
 from frequent hearing of a vocal trick, especially when of a 
 peculiar description, to some measure of unconscious 
 imitation of it, is readily conceivable. Moreover, a certain 
 amount of intercourse of an official and industrial kind is 
 inevitable, even where there is none of a social character. 
 Some practical familiarity with the native vernacular must 
 therefore become frequent, and will contribute something to 
 new phonetic tendencies in the invading tongue. We must 
 not, of course, exaggerate the effect of such conditions ; on 
 the other hand, they cannot scientifically be ignored. 
 
 Impulses to phonetic change are supplied much more 
 emphatically when the two peoples have settled down to a 
 common life, with intermarriage and free social intercourse 
 in all or most departments. While the invading language is 
 asserting its supremacy there is obviously a transition-period, 
 which may occupy several generations or even centuries, 
 and during which both adults and (what is much more 
 important) children are of necessity subjected to the phonetic 
 influences of the native speech, which is still widely prevalent 
 in every .settlement. Many persons will speak both tongues, 
 often with a disregard of niceties of articulation. Others 
 will speak one of the languages fluently, as being that of 
 closer association, while the other they will use in a more or 
 less broken manner, employing its words with phonetic 
 peculiarities of their own. Often in the same house the 
 elders will be familiar only with the one language, while the 
 younger members may be rather more familiar with the 
 other. Apart from the intermixture of races, which produces 
 by way of heredity physiological effects which we cannot 
 exactly specify, a considerable modification of phonetics
 
 Phonetic Change 299 
 
 must manifestly result during this period. The invading 
 language cannot fail to borrow during the struggle some- 
 thing from the more salient idiosyncrasies of the invaded. 
 The struggle between the native Celtic and the Latin lasted 
 in Gaul till at least the fifth century A.D., and the phonetic 
 character of the resulting lingua romana is largely the out- 
 come of the Celtic habit and tendency. In the same way 
 the distinctive palatal and cacuminal sounds of the Sanskrit 
 and modern Hindu languages are the result of the transition 
 period of conflict with the Dravidian speeches and of 
 mixture with the Dravidian races. English as spoken by 
 Irishmen or by Highlanders frequently bears manifest 
 traces of the phonetic modification undergone during the 
 collision with the Celtic tongues in Ireland and Scotland 
 respectively. 
 
 We have so far spoken of one imaginary separated dialect 
 of our imaginary primitive tongue as if it were the invader. 
 It may, on the contrary, be the invaded, in the sense in 
 which Anglo-Saxon English was invaded by Norman French, 
 or Persian by Arabic. In this case any phonetic modifica- 
 tion which it may undergo will be due to the same social 
 causes as those above described, and the extent of the 
 modification will depend upon the numbers and other 
 conditions of influence of the speakers of the invading 
 tongue. Even when the latter is routed, a long and well- 
 sustained conflict will leave its mark upon the language 
 which has successfully resisted. Chiefly, no doubt, the 
 evidence will lie in the vocabulary, but it will also exist, 
 even if less clearly, in the phonetics. 
 
 There is, it must be added, a linguistic invasion of yet 
 another kind. When a language is in full occupation of a 
 certain area, the population speaking that language may (as 
 in the case of the United States of America) receive large 
 influxes of persons speaking alien tongues. The several 
 speeches of the immigrants will doubtless generally disappear 
 with the second generation, but, inasmuch as the ' average
 
 300 Natural History of Language 
 
 articulation ' is determined by the contributions of the 
 constituent individuals, it will be inevitable that the large 
 amount of misarticulated or broken or totally foreign 
 utterance which will make itself heard, will exercise some 
 effect upon the average articulation of the whole. 
 
 Recapitu la tion 
 
 In order that we may now take a survey of the wood 
 without having our view impeded by the trees, we may 
 briefly recapitulate the phenomena of phonetic change, and 
 the steps and conditions of the phonetic disintegration of an 
 imaginary homogeneous speech. 
 
 I. Phonetic aberrations of the individual 
 
 In non-educated societies due mostly to physical 
 causes, resulting in 
 Inaccuracy of the sound-picture through the ear, 
 
 and 
 Inaccuracy of its reproduction by the voice. 
 (In educated societies often due also to psycho- 
 logical causes, e.g. adaptation to spelling, real 
 or fancied etymological association, etc.) 
 Mostly counteracted by the influences of immediate 
 environment 
 Through spontaneous correction of the sound- 
 picture ; 
 Through ridicule ; 
 Through education. 
 Sometimes incapable of correction ; seldom wholly 
 
 eliminated. 
 The persistent cases and the residuum of individual 
 tendencies contribute to the average articulation 
 of the environment and insensibly modify it. 
 (The influence of the individual is particularly potent 
 When the same tendency exists in others ; 
 When his personality counts for much.)
 
 Phonetic Change 301 
 
 Phonetic aberrations of a group of close inter- 
 course 
 
 The result of the interaction of the individual 
 
 aberrations and the estabHshment of some of 
 
 them. 
 Largely counteracted by the influence of other groups 
 
 which intersect ; seldom wholly cancelled. 
 The persistent cases contribute to the average 
 
 articulation of the larger environment and 
 
 insensibly modify it. 
 (The influence of such a circle is specially potent 
 When the group for any reason counts for much 
 
 relatively to the environment ; 
 When (in educated societies) it is a learned 
 
 group or one in high social place.) 
 
 Phonetic aberrations of a special area or district, 
 with internal intercourse habitual and external 
 only occasional ( = a dialect) 
 
 The result of the interaction of the deviations of 
 the groups comprised. 
 (Sometimes also of phonetic influences of neigh- 
 bours of alien speech.) 
 
 In non-educated societies only slightly checked by 
 the occasional influences of neighbouring dia- 
 lectal areas. 
 
 (In educated societies artificially checked to some 
 extent by the same influences and also by 
 instruction.) 
 
 The distinctive peculiarities are more marked at 
 the centre of each district, less so on the 
 borders. 
 
 When there is no geographical break, the dialects 
 are phonetically interlinked according to 
 actual proximity. 
 
 (Yet remote dialects may accidentally agree in 
 special phenomena.)
 
 302 Natural History of Language 
 
 When a dialect is geographically isolated the 
 peculiarities become fixed and accentuated. 
 
 (The changes are seldom confined to one sound at 
 a time.) 
 
 4. Further phonetic aberrations of a dialect (or 
 association of dialects) when separated from 
 the rest ( = growth to a distinct language) 
 
 The result of a now entirely fi-ee play of inherited 
 tendencies and a free development of new 
 ones ; together with 
 
 Influence of intermixture with people of alien 
 tongues in the new home, whether by 
 ' invading ' them or being ' invaded ' by them ; 
 
 Influence of climatic conditions and conditions 
 of life.
 
 CHAPTER XIII 
 
 PHONETIC LAWS : ' GKIMM'S LAW ' 
 
 The last chapter has discussed phonetic change and its 
 natural causes. It should now be evident that no nation, or 
 large area of a nation, can show at any moment an absolutely 
 uniform and consistent articulation of the sounds which 
 compose its language. There must be at all times a state 
 of instability. Theoretically, no doubt, an educated linguistic 
 community possesses, at a given date, a standard articula- 
 tion of each sonant and consonant in each word, although 
 experience shows that the said standard is by no means 
 always discoverable with ease. For example, even professed 
 phoneticists will sometimes allege that in English words 
 like hard or cork the r is entirely unheard, and that the pro- 
 nunciation is hahd, cawk. Others are no less certain that 
 the r, though not trilled, is distinctly audible from speakers 
 of ordinary care. The tongue, they maintain, does adopt 
 the usual English r-position as part of the articulation of 
 hard or cork. Such disagreement is due partly to difference 
 of ear, partly to actual difference of practice. The only 
 point of certainty in the matter is that the articulation of 
 the r is weak. Nevertheless, despite occasional diversity of 
 opinion, we may accept as approximately true the statement 
 that, at a particular date, there is a standard or orthodox 
 articulation such as is registered in the dictionaries. Yet we 
 have also to admit that there are few persons who do not 
 deviate more or less widely from the standard, and that there 
 is probably no one who adheres to it in all particulars. 
 
 303
 
 304 Natural History of Language 
 
 Moreover, " all things are in a condition of flux," said 
 Heracleitus ; and a language is anything but an exception. 
 It never stands still, whether in grammar, vocabulary, or 
 articulation. Even the theoretical standard of one generation 
 is not precisely the standard of the next. 
 
 This being the case, it becomes necessary to employ with 
 great caution the much-abused expression ' Phonetic Law.' 
 The term first requires some explanation. In philological 
 works much is inevitably said of ' the phonetic laws ' of 
 Latin, Greek, English, or other languages. We hear, for 
 example, of the following phonetic law of Greek. Whenever 
 a simple s stood between two vowels in a word of the parent 
 Indo-European tongue (as in *genesos), if such a word 
 passed by direct inheritance into the Greek language, the s 
 in question first became // and then disappeared (the result 
 thus being geneos, <yeveo'q). So far as purely phonetic opera- 
 tions go, the statement, thus worded, is correct. Any 
 apparent violation of this particular ' law ' is for the most 
 part easily explainable from considerations which are psycho- 
 logical, and have nothing to do with vocal tendencies. The 
 meaning of this will be made clear when we deal with 
 ' Analogy.' 
 
 Meanwhile it is stated as a phonetic law of Latin that 
 the same s in the same circumstances will, if the word passes 
 by direct inheritance into the Latin language, become r. 
 Thus *genesos will properly become, not geneos, as in Greek, 
 but \generiis} Again, any apparent breach of this law is 
 either not an exception or is explained by certain mental, 
 and not vocal, tendencies of the Latin speech. 
 
 For English a phonetic law is laid down that, where an 
 original Indo-European word or root began with the sound 
 /, the same word or root, if directly inherited by the English 
 language, will begin with the sound th. Thus for the original 
 *treies English will properly offer three. 
 
 The above phonetic laws have been selected for their 
 simplicity and regularity. It will be observed that each 
 ^ The actual Latin genitive generis is from an alternative original *geneses.
 
 Phonetic Laws : ' GHjjwis Law ' 305 
 
 refers to a whole language, Greek, Latin, or English, as the 
 case may be, and comprehends all its dialects. Each is a 
 statement concerning the treatment of a particular sound in 
 a particular situation, and each is confined to a change which 
 has occurred during the direct passing on of a word from 
 father to son in a given branch of the Ursprache. 
 
 Now at the date when we first meet with the Greek 
 language the loss of s between two vowels has already been 
 fully effected. Similarly the Teutonic substitution of fk for 
 / had been accomplished centuries before English is met with 
 as English. But there must have been an epoch, probably 
 extending over several generations, during which the change 
 in either instance was gradually prevailing. In the case of 
 Old Latin we are able to see something of such an epoch ; 
 for, when we get our first glimpses of that language, a 
 number of words which were afterwards pronounced with 
 r (as arborem, lares) are still written with s {arbosem, lases), 
 although doubtless the pronunciation intended was s, the 
 intermediate step. 
 
 If we could have watched these several periods of transi- 
 tion, we should have seen certain early speakers of Greek 
 differing among themselves in their treatment of the inter- 
 vocalic s, and certain early speakers of Teutonic differing in 
 their treatment of initial t. At those periods we might 
 speak of a growing phonetic tendency, but not of a phonetic 
 law, in these respects. How long it took for the Indo- 
 European ^genesos to become in the whole Hellenic branch 
 (first \genehos and then) geneos we cannot ascertain. The 
 phonetic law is simply a statement of the change as a com- 
 pleted phenomenon. 
 
 Within any language, as has been shown above, dialects 
 are continually endeavouring to form themselves from special 
 phonetic tendencies. Sometimes these threatened develop- 
 ments are entirely thwarted or rendered inconsiderable. 
 Sometimes they persist and become conspicuous over an 
 area more or less extensive. Sometimes one of them will 
 gradually prevail over the whole area of the language. It is 
 
 (B 834) X
 
 3o6 Natural History of La7iguage 
 
 obvious therefore that there are abortive phonetic tendencies, 
 locally successful tendencies, and complete phonetic victories. 
 A phenomenon arises, and either perishes, or wins a district, 
 or conquers the whole language. The history of a given 
 tongue on its phonetic side is a history of these unceasing 
 efforts and partial or complete victories. A locally success- 
 ful tendency becomes a ' phonetic law ' for the local dialect ; 
 a ' phonetic law of the language ' is the wider result of a 
 tendency which has won the day over the entire field. 
 Manifestly, however, such phonetic results are liable to 
 further variation in process of time. We may register the 
 phenomena of a language at a particular date, and record 
 the phonetic laws which are valid for that date, but we must 
 not expect these, when once established, to remain unaffected 
 for ever. It may be a Teutonic phonetic law that Indo- 
 European g passes to k, and all the Teutonic languages may 
 have obeyed that rule of change, but there is nothing to 
 prevent the k itself from passing on subsequently to ch (ts), 
 as, for example, when *Ug becomes t/iak but then passes 
 outward to tJiatch. 
 
 As a definition of a phonetic law of a language we 
 may attempt the following. A phonetic law of a language is 
 a statenioit of the regnlar practice of that language at a 
 particular time in regard to the treatniott of a particular sound 
 or group of sounds in a particular setting. The several parts 
 of this definition demand separate consideration. 
 
 A phonetic law, it will be seen, is simply the statement 
 of an observed phenomenon. This is a different thing from 
 asserting that some indefeasible natural condition has made 
 such and such a change inevitable with such and such a 
 people. Nothing is asserted as to any ' law of nature ' 
 absolutely compelling a certain set of speakers to follow a 
 certain articulation and no other. A strong propensity of a 
 considerable number at a certain date is often sufficient to 
 determine the fate of a whole language ; there may also be 
 psychological and social conditions, possibly transitory, but
 
 Phonetic Laws : ' Griuwi s Law ' 307 
 
 in any case much too complicated to unravel. An examina- 
 tion of the history of a language simply happens to prove 
 that it has shown a conspicuous regularity in the treatment 
 of a certain sound or group of sounds. It might therefore 
 be conveniently stated that such regular treatment is a 
 phonetic ' rule ' of the language. ' Law,' however, properly 
 implies stringency of application, and the word has been 
 deliberately employed with that intention. That school of 
 philologists which most uses the term maintains that there 
 are no real exceptions to the phonetic laws of a given 
 language ; that, if a language comes habitually to pronounce 
 a certain sound in some new way, it will always and invari- 
 ably so affect that sound ; that exceptions are only apparent, 
 being borrowed words, dialectal words, or archaic survivals 
 and revivals, or words diverted from their expected shape by 
 some instinct of analogy, or otherwise to be accounted for. 
 With this position we shall deal briefly in a later paragraph. 
 Meanwhile all philologists alike mean by ' a phonetic law ' 
 simply the statement of a regular phonetic practice, however 
 their opinions may differ as to the exact degree of the 
 regularity. 
 
 In the next place, a phonetic law properly refers only to 
 a particular time. A language which has established a 
 regular phonetic practice at a certain epoch of its develop- 
 ment, may proceed to a modification of that practice by 
 reason of a new tendency. Or, again, a language in which, 
 at a certain epoch, a certain phonetic tendency has been 
 conspicuous and has triumphed, may at a later epoch show 
 no such tendency. 
 
 Thus, to take the former case, it is a phonetic law of 
 Teutonic, including early English, that an original Indo- 
 European g shall become k. For a primitive *tcg Old 
 English therefore gives thak, for *ncg it gives zvak. But, 
 since the period of Old English, the language has developed 
 a new tendency — though not consistently exhibited — to 
 ' palatalise ' a final k into the sound phonetically written as 
 //, but currently as ck or tch. Hence modern English says
 
 io8 
 
 Natural History of Language 
 
 thatch and watch. It is not to the purpose at this point to 
 discuss with what dialect this change began or what causes 
 have prevented its regularity. The fact remains that when 
 we speak of ' a phonetic law of English ' affecting the 
 primitive g in question, we must be careful to distinguish the 
 period to which the law strictly applies from the period for 
 which it must be stated with a subsequent modification. 
 
 Again, to take the second case. In an early stage of the 
 Greek language, before the date at which we meet with it in 
 history, the tendency of the speakers of that tongue had been 
 to convert s between two vowels, first into )i, and then to 
 drop that sound completely out of use. Thus *gcnesos 
 became \ge71ehos and thence geneos. This tendency so com- 
 pletely triumphed that it had become a phonetic law for the 
 Greeks in an epoch long before Homer. Nevertheless when, 
 in after generations, the other corruptions of sounds in Greek 
 words had accidentally produced new cases of the sound s 
 standing between two vowels {eg. in vioiisa, which is not an 
 early Indo-European form but a corruption of *})iontia^ or in 
 basis, which is not an early Indo-European form but a 
 corruption of Abatis'), there no longer existed any particular 
 tendency to convert these into Ji and then to drop them. 
 Words containing an intervocalic s thus became very 
 numerous, but this s — called ' hysterogen,' as being of later 
 birth — remained undisturbed. The case may be stated 
 thus : — 
 
 Epoch I 
 (I.-E. original) 
 
 *genesos > 
 
 When I.-E. [*mon/ui > 
 did notJ 
 contain j [*med/iws > 
 
 Epoch 2 I Epoch 3 
 
 (Period of phonetic (Later Greek period 
 tendency affecting inter- (tendency expired)) 
 vocahc s in Greek) 
 
 jgenehos, geneos > geneos 
 
 t 
 
 \motisa "I s does not yet > jnousii\ tendency 
 - stand be- - lose s 1 
 
 7>iessos] tween vowels > 7;/£'5'<7J' J expired 
 
 to 
 has 
 
 So in the instance of the English ivatch, thatch, and the like, 
 the history may be thus expressed : —
 
 Phonetic Laws : ' Grimm s Law ' 309 
 
 I I 2 i 3 
 
 Indo-European Early Teutonic \ Modern English 
 
 (including Old English) ' (further development) 
 
 *ues: 
 
 thak 
 ivak 
 
 thatch 
 watch 
 
 A phonetic law further apph'es to a sound or group of 
 sounds only when situated in particular settings. The fact 
 that a language at a given epoch habitually converts, say a t 
 into t/i, in most positions, does not necessarily imply that it 
 so converts a / in each and every position. It may, for 
 instance, retain the t unchanged when it is combined with a 
 preceding s. That is to say, a rule which applies to ^ in 
 most, or even all other, positions, may not apply to / in the 
 special combination st. Thus in English the Indo-European 
 *trcies becomes three, and the Indo-European *pet- produces 
 feath-{cr). Yet the Indo-European *std- becomes sta-{jid). If 
 these were (as they are not) the only facts to be considered 
 concerning /, the phonetic law might be stated thus : ' Indo- 
 European / becomes in English ///, except in the combination 
 St, where t is unchanged.' But it would manifestly be in- 
 accurate to lay down as a phonetic law simply that ' t 
 becomes th! 
 
 A phonetic law, then, must be understood to belong to a 
 particular period and to a sound particularly placed. Is 
 it to be assumed that, for such period and such sound, the 
 law will be without exceptions in the language concerned ? 
 Before answering, it is necessary to consider again what 
 naturally constitutes 'the language.' The point has already 
 been elaborated that, in the case of any living tongue spoken 
 naturally over an appreciable area, there have been dialects. 
 Doubtless the language, conventionally recognised as such, is 
 generally the outcome of one of these, and, for the most part, 
 of one which is fairly central and the nearest approach to a 
 compromise between the rest. Thus English is mainly 
 developed from the language of the East and South-East
 
 3 1 o Natural History of L angziage 
 
 Midlands, French from that of the lie de France, Spanish 
 from Castilian, Italian from Tuscan, Greek from Attic. 
 Gradually such dialect absorbs into its sphere the speakers of 
 the other dialects, while these drop into patois or slowly 
 disappear. Yet, even if the conventional language is mainly 
 the outcome of a single dialect quite consistently spoken 
 (which, in truth, it seldom is), it will be impossible but that 
 words and pronunciations of the subordinate dialects should 
 gain some foothold in it. Later English thus contained 
 words and sounds of the South and the North which were 
 alien to the tendency of the Midlands, even if we assume 
 that tendency itself to have been homogeneous. Similarly 
 the basis of Latin was the dialect of a portion of the 
 plain of Latium ; but as the surrounding tribes, particularly 
 the Sabines, were drawn into the social and linguistic area, 
 the language came to contain many words in which the 
 sounds conflicted with the purely Latin tendency. 
 
 If, therefore, we take such a language as it is actually 
 spoken at a given time, we shall for this reason seldom 
 succeed in proving that all its phonetic phenomena are strictly 
 obedient to the rules. There will be inconsistencies and re- 
 calcitrant words. The philological specialist may perhaps be 
 able to disentangle the ' pure ' Mercian or ' pure ' Latin words 
 from the dialectal, but, truth to tell, his method is apt to be 
 the somewhat illogical one of laying down a phonetic law 
 for Mercian or for Latin and then relegating to the subsidiary 
 dialects all such words as contain exceptions. It is, doubt- 
 less, highly probable that the words so relegated were for 
 the most part actually introduced from the dialects in 
 question ; but it is by no means disproved that a real 
 vacillation and inconsistency of pronunciation, no longer to 
 be accounted for, may have occurred in the case of the same 
 sound in different words. Much might depend on the 
 frequency with which one word was used as compared with 
 another, or on the classes of persons by whom it was used. 
 The fact must always be remembered that new tendencies 
 are continually arising, and that, though one of these may not
 
 Phonetic Laws : ' Grimm's Law' 311 
 
 prevail in the case of all words, it may so prevail in the case 
 of some special word, perhaps a word of the populace taken up 
 by the orthodox speech. Thus an original Indo-European 
 diphthong ai becomes in orthodox Latin ae. The popular 
 tendency, long resisted by the orthodox style, substituted c. 
 Caecilius and Cecilius were thus heard side by side. On the 
 other hand, levir (not laever) is the only form recorded for 
 that word. Until the c finally triumphed it is therefore 
 hardly scientific to assert an 'invariable' phonetic law of 
 Latin in respect of the primitive *ai. Again, it is a phonetic 
 law of Latin that an original *gh (initial) becomes //. Thus 
 the primitive ^ghayis (Greek ^^i^. Germ. Gayis, Eng. goose) 
 should be represented in Latin by hansier). In reality only 
 anser is found. The tendency to omit h from pronuncia- 
 tion occurred early among the Latin populace, but literature 
 also says only anser. If we call this a ' dialect word,' we are 
 using the term ' dialect ' somewhat freely. 
 
 There is no intention here to impugn the view that 
 phonetic laws do operate with very great and remarkable 
 regularity ; but it seems well to utter a warning against a too 
 stubborn a priori position that, when borrowed words — 
 whether taken from a dialect or from a foreign language — 
 are eliminated, the rest will be found to offer no exceptions 
 whatever of a purely phonetic kind. This is to assume 
 greater homogeneity on the part of a multitude than is 
 known to be the case with an individual. Experience proves 
 that it is quite unwarrantable to assume (as is commonly 
 assumed) that an individual who habitually pronounces, say, 
 broad as brode, will always and inevitably so pronounce an 
 aw- sound as o. He will not necessarily say cose for cause. 
 The human being is more than a machine, and is subject to 
 many casual influences, partly from changing environment, 
 partly from subtle operations of his own mind. The 
 Elizabethans spelt their own names in various ways at 
 various moments, and at the present day the same documents 
 will often show in the same handwriting different shapes of 
 the same letters. There is something of the kind in articula-
 
 3 1 2 Natural History of Language 
 
 tion, as anyone may discern who carefully observes the speech 
 of an acquaintance. Add to these considerations the subtle 
 psychological influences which we can never hope to track. 
 Of absolute inability to produce a given sound there is com- 
 paratively little ; the chief cause of corruption is unchecked 
 laxity or indifference. But it is by no means a fact that 
 one who generally follows the line of least resistance will 
 necessarily do so in every instance. 
 
 And, if this is the case with the individual, it is assuredly 
 not less so with that shifting aggregate of classes and 
 individuals whose averaged speech is the language of a 
 country or a linguistic area. It has already been shown how 
 complex are the interworkings of individuals and dialects, 
 and it appears an excess of boldness to maintain that all 
 these interworkings will always end in full consistency of 
 phonetic result. 
 
 It is doubtless often possible to prove that the apparent 
 exception to the law in a certain word is due to the fact that 
 the word is borrowed from another language, or from a 
 special dialect which does not obey the law. Deck (properly 
 ' covering ') is borrowed from the shipping folk of the opposite 
 Teutonic coast ; the English form is tJiak. Popma, the Latin 
 for ' cook-shop,' is an Oscan word from Campania ; the proper 
 Latin form is coqiiina. Often such borrowing may be sus- 
 pected, where it cannot be proved. 
 
 Sometimes, again, it can be shown, or plausibly 
 maintained, that the phonetic law is counteracted by the 
 influence of an ' association of ideas ' or ' analogy.' Thus in 
 Greek earrjcra (estesa) should have become earrja {ested), since 
 the s stood from earliest times between two vowels, and 
 should therefore have disappeared in Greek. But ecrrTjaa is 
 a verb-tense (' I placed '), in which s regularly played a part 
 similar to that of the English -ed. Thus eSei^a {edeiksd) ' I 
 showed,' eTvyjra {ehipsci) ' I struck.' In these words s is not 
 lost, because it does not stand between two vowels. Hence, 
 from a natural instinct for uniformity, the mind of the listener 
 awaited, and that of the speaker supplied, the sound s in the
 
 Phonetic Laws : 'GrivDusLaw' 313 
 
 same tense (' I placed '), whether phonetically it would have 
 disappeared or not. Thus ea-ji^aa held its ground, and con- 
 tinued to be rebuilt, in spite of the phonetic rule which 
 should have corrupted the word into earrja. 
 
 There are also other psychological influences which it is 
 possible to comprehend. In Latin an original diphthong oi 
 becomes by phonetic law first oe and then ft. Thus funis = 
 oenus = oinos. Nevertheless foidos (' treaty ') never gets 
 beyond the form foedus ; there is no ^ffidus. But it is easy 
 to understand that foedus is a word of official and legal 
 documents and smacks of the formalities of the Senate. In 
 such circumstances the old-fashioned pronunciation is apt to 
 be retained in spite of all phonetic tendencies to the contrary. 
 Nevertheless we cannot pretend to discover all such instances 
 nor to be sure of our clue. 
 
 Be this as it may, when a mental operation checks a 
 phonetic change, there is in point of fact an exception to the 
 phonetic law. That it may be explained, does not m.ake it 
 less an exception. It is therefore, after all, meant by the 
 severely rigid school that, if there were no borrowing, and if 
 psychology did not interfere in articulation — if there were 
 no instinct of grammatical analogy or deliberate archaism or 
 the like — a sound in a certain environment would always be 
 treated by the same language in precisely the same way. 
 But this is to deal with a hypothetical or sublimated 
 language, not with a real language. If, for instance, modern 
 English is cleared of all words not of direct Mercian 
 descent ; if all the words of Wessex and Northumbria, all 
 Scandinavian forms, as well as the Latin, French, Dutch, 
 and other borrowings which it contains are eliminated ; if all 
 forms due to association of ideas or instinct for uniformity of 
 system are also taken out ; then we have English as it might 
 have been, but not as it is. The same observations are 
 applicable to Latin or to Attic Greek or to any other 
 civilised tongue. 
 
 Though it has seemed worth while to make these
 
 314 Natitral History of L anguage 
 
 qualifications, it is not to be denied that, if we agree to 
 treat a language in this fashion, it will in point of fact show 
 remarkably little violation of its phonetic laws in what is 
 left. When foreign words, dialectal words, archaic words, 
 and words influenced by analogy have thus been subtracted, 
 and when the phonetic laws themselves have been stated 
 with their proper limitations, the rigid view appears 
 justified by experience. 
 
 Absolutely to prove the inviolability of phonetic laws 
 would require immense material. It would be necessary 
 to obtain a panoramic view of the history of every word, to 
 know how it came into the language — whether by direct 
 inheritance or by borrowing — to discern by what classes 
 of persons and how frequently it was used, and to possess 
 a complete comprehension of the subtle workings of analogy. 
 All this being out of the question, we fall back on logical 
 probabilities. 
 
 The a priori argument has been that with a people, as 
 with an individual, if a certain sound habitually commends 
 itself to the organs of articulation, it will do so whenever the 
 circumstances are the same. A people which commonly 
 pronounces an initial k before i and e z.% c is), does so for 
 some strong reason, all the more strong for being unconscious. 
 It may therefore be expected to pronounce such a k, not 
 sometimes, but always, in that way. A people like the 
 Maoris or the Zulus, whose own words invariably end in 
 a vowel, as the result of a peculiar manner of articulation, 
 will always pronounce even borrowed words with a vowel 
 at the end, simply because of the power of that habit. The 
 Maori, moreover, thanks to some peculiar trick of articulat- 
 ing, has so habitually separated his consonants by means of 
 a vowel, that he unconsciously does the same with the 
 consonants of a European name. Thus Thomson in Maori 
 is Tamahana and Croft is Korowhata. Similarly in South 
 Africa gold is igolide and sugar is isugile. 
 
 The argument is practically that the thing is done 
 because it is not thought about, but ' does itself.' This
 
 Phonetic Laws : ' Grimm s Law ' 
 
 31:) 
 
 argument is no doubt stronger for small and homogeneous 
 barbarian peoples than for extensive nations of mixed 
 racial and linguistic origin. Nevertheless, though not 
 conclusive, it possesses considerable force even for the latter. 
 It forms at least a good working hypothesis for the student, 
 and there would have been little exactness of etymology 
 without it. 
 
 There is, however, a more substantial argument for 
 phonetic regularity than general reasoning and a priori 
 likelihood. Philologists have found that, the more closely 
 and cautiously they examine a language, the fewer are 
 the apparent exceptions to its phonetic rules. What 
 were once called ' sporadic exceptions ' tend to disappear. 
 Words once empirically assumed to have a certain deriva- 
 tion are now shown to have had an origin altogether 
 distinct. The Greek ^eo? {theos), assumed to be identical 
 with the Latin dens, was once regarded as a ' sporadic 
 exception ' to a law. The Greek initial 6 {th), it was recog- 
 nised, ought not to answer to the Latin initial d. One or 
 other language must have broken its phonetic laws ; either 
 the Greek initial ought to be 8 {d), or the Latin initial ought 
 to be f. Yet the words possessed so exactly the same 
 meaning and were so much alike in shape that there was 
 a natural reluctance to separate them. The more modern 
 and more expert philologist denies the exception in denying 
 the identity. It may take some searching to discover that 
 the Greek word descends from ^dlmes- and the Latin 
 from "^deiij-y and that the true congeners of the former 
 are to be found in the Middle High German {ge)twas 
 (' ghost ') and the Lithuanian dvase (' spirit '), but it is 
 through such searching that sound etymologies are to be 
 ascertained. 
 
 In other cases investigation has shown that phonetic 
 laws are apt to be stated too comprehensively. Part of 
 the work of the modern philologist has been to re-state the 
 laws, limiting them to their true application and discovering 
 the qualifying laws or sub-laws. It had, for example, been
 
 J 
 
 1 6 Natural History of Language 
 
 laid down by 'Grimm's Law' that an Indo-European t 
 (represented commonly by r in Greek and t in Latin) 
 became in Old English ///, as in three = *treies, that = *tod. 
 One qualification of this law has been already stated, viz. 
 that st remained unchanged (as in stand). To the over- 
 whelming majority of instances the law duly applied. 
 Nevertheless there were exceptional words which refused 
 to be brought under the law. The Old English /<^7^^r, inoder 
 must manifestly represent the original words which appeared 
 in Latin as pater, viater, and in Greek as irary^p, /jLrjrrjp 
 (pater, meter). Those primitive words must be assumed 
 to have been ^p9ter, *vidter. Brotlier, which is the repre- 
 sentative oi'^bhrdter (Latin /rater, Gk. (^parrjp), was obedient 
 to the Teutonic rule concerning t ; and why were fader, 
 nioder not equally so ? Yet Anglo-Saxon and Old English 
 (including the English of Chaucer) had no such forms as 
 father, mother. It is true that we now actually use these 
 latter forms, but that is simply because we have assimilated, 
 by mental association, the shape ' mo//^er ' and then ' fa/>^er ' 
 to the shape ' bro//^er.' The question is, ' Why do Anglo- 
 Saxon and Old English offer brotlier according to the law, 
 but fader, inoder in violation of the law ? ' (The notion that 
 these were accidental (or ' sporadic ') exceptions was over- 
 thrown by IkLeiner, who showed by an exhaustive process 
 that the phonetic law in question must be qualified by a 
 sub-law. The phonetic law was true for an initial /, and 
 for any t which folloived the accent in the Indo-European 
 word (e.g. ^bhrater), but it was not true for a t (other than 
 initial) which preceded the accent (as in ^p9ter, ^vidte'r). 
 In the latter circumstances it passed on to d. The point 
 may appear somewhat subtle, but it only the better ex- 
 emplifies the care with which both laws must be laid down 
 and apparent exceptions must be treated. When once 
 discovered, ' Verner's Law ' did away with other ' exceptions ' 
 of the same kind. The I.-E. *klut6s (' heard ') becomes in 
 Greek k\vt6<; (klutos) and in Latin (in)clutiis. Noticing the 
 accent, and asking ourselves what should be expected in
 
 Phonetic Laws : ' Gri7mn s Law' 317 
 
 Old English, we shall now say, not that by Grimnn's Law 
 it should become hliith, but that by Verner's Law it should 
 be hlud ( = /oud^).'J What were apparently examples of 
 the laxity of law thus turn out to be illustrations of the 
 contrary. 
 
 All languages have their special phonetic tendencies, 
 which may result in laws and sub -laws. Some of these 
 are obvious and far-reaching, and are among the chief 
 differentiators of related speeches. Thus the French and 
 Italian sister -tongues diverge conspicuously in their treat- 
 ment of those parent Latin words in which a single con- 
 sonant stood between two vowels. French in such positions 
 loses all trace of the consonants g, t, d. The following 
 examples of the law may suffice : — 
 
 l^^^.Iigare > Yr. Her \^^^ ^^ 
 
 regina > reine j 
 
 natlvus > 'f^/ I loss of ^ 
 
 dotattis > aoite ) 
 
 crudelis > ^'-"'^ \ loss oi d 
 
 Cauda > queue] 
 
 From this tendency (which appears to be due to Celtic 
 racial qualities) Italian is free, its words being ligare, regina, 
 nativo, dotato, crudcle, coda. Should the etymologist meet in 
 French with a word which appears to violate the phonetic 
 rule above stated, he generally finds it easy to show that 
 the word in question is not part of the vocabulary inherited 
 directly from Latin by the speakers of the lingua roviana of 
 France, but is a later importation, borrowed from Latin 
 after the French language had already become by phonetic 
 operations practically what it now is. Thus inaitrc, re- 
 presenting magister, is French of the older stock, to which 
 alone the law applies, while viagistrat is but French by 
 adoption. 
 
 1 In reality hltid=*Uut6s {i.e. with a long quantity of the root-vowel).
 
 3 1 8 Natural History of Language 
 
 GRIMM'S LAW 
 
 An example of a phonetic law, highly interesting in 
 itself, and especially so to students of the English language, 
 is that known by the name of Grimm. The actual 
 discovery of the phonetic correspondence concerned had 
 been (in part at least) anticipated by Ihre and Rask, but 
 the definite formulation of the law — though incorrect, be- 
 cause too comprehensive — was first made by Jacob Grimm 
 in his German Grammar. It would be disproportionate 
 here to give in detail the history of all the modifications 
 which have been found necessary in the too symmetrical 
 statement of Grimm ; nor need we be concerned to set forth 
 minutely every sub-law which investigation has discovered. 
 The following must be taken as but a working account 
 of the main facts of the law and a rough guide to its 
 practical application. 
 
 Grimm's Law is concerned with the Indo-European 
 explosives or ' mutes,' and their treatment in the Teutonic 
 languages. It will be sufficient if we write these original 
 consonants as the Aspirates gh, dk, bh, the voiced Explosives 
 g, d, b, and the voiceless Explosives (or Tenues) k, t, p. 
 We have thus three classes of sounds, to which we may 
 refer respectively as A, M, T {i.e. Aspiratae, Mediae, 
 Tenues). On passing into the ' classical ' languages, i.e. 
 into Greek, Latin, and Sanskrit, these sounds commonly 
 (though by no means always) adhere to their original 
 class. A Tenuis becomes regularly a Tenuis, a Media 
 is still a Media, and an Aspirata is either represented 
 by an Aspirata or by some sound which may be taken 
 as the easy development from an Aspirata. Exceptions 
 to this regular representation are, indeed, not few, and 
 are duly classified by special philologists in their proper 
 place under the phonetic laws of Greek, Latin, or 
 Sanskrit as the case may be. Yet, broadly speaking, 
 the classical tongues remain faithful to the original
 
 Phonetic Laws : ' Grimm s Law ' 
 
 19 
 
 kind of sound. Thus for Greek and Latin the ordinary 
 correspondence is : — 
 
 
 I.-E. 
 
 Greek 
 
 Latin 
 
 1. 
 
 A 
 
 A 
 
 A (or spirant therefrom) 
 
 viz. 
 
 ^^ 
 
 X m 
 
 h (and/) 
 
 
 dh 
 
 e {th) 
 
 / (and d ) 
 
 
 bh 
 
 ^ {Ph) 
 
 f (and b) 
 
 2. 
 
 T 
 
 T 
 
 T 
 
 viz. 
 
 k 
 
 K 
 
 C 
 
 
 t 
 
 T 
 
 t 
 
 
 P 
 
 7r(/) 
 
 P 
 
 3. 
 
 M 
 
 M 
 
 M 
 
 viz. 
 
 g 
 
 y C?) 
 
 g 
 
 
 d 
 
 8(^) 
 
 d 
 
 
 b 
 
 i^{b) 
 
 b 
 
 On the other hand, the Teutonic languages show a 
 pecuHar and systematic ' shifting ' in each of these classes. 
 To whichever class a sound belonged in primitive Indo- 
 European, we may be sure that (except in special circum- 
 stances to be stated immediately) in Teutonic it will have 
 changed to one of the others. An Aspirata will not appear 
 as an Aspirata, nor a Media as a Media, nor a Tenuis as a 
 Tenuis. At some period in the joint (or Ur-) Teutonic 
 epoch there was a progressive shifting all along the line. 
 An original Aspirata became a Media ('A passes to M') ; an 
 original Media became a Tenuis (' M passes to T ') ; an 
 original Tenuis became a Spirant (' T passes to S '). Thus, 
 in English, *bhu- becomes be, *d?jo becomes two, ^trcics 
 becomes three. This sound-change, which was shared by all 
 the Teutonic speeches, and is therefore early in date, is 
 known as the 'First Teutonic Sound-shifting.'^ Its cause, 
 whether racial admixture or not, is unknown. We may 
 
 ^ In German, Laidvenchidning.
 
 120 
 
 Natural History of Language 
 
 roughly represent as follows the correspondence which 
 resulted : — 
 
 I.-E. 
 
 
 Teutonic 
 
 
 I. A 
 
 -= 
 
 M 
 
 
 'iz. *^/i: 
 *dh 
 
 *bh 
 
 = 
 
 d 
 b 
 
 (e.g. *ghans — Y.ng. goose) 
 {*dhe-- =Eng. dee{d)) 
 {*bhii- - = Eng. be) 
 
 2. M 
 
 = 
 
 T 
 
 
 viz. *g 
 *d 
 *b 
 
 \ 
 
 t 
 
 P 
 
 {*genos = Eng. kin) 
 {^diio = Eng. two) 
 {*slub- = Eng. slip) 
 
 3. T 
 
 = 
 
 S 
 
 
 viz. *k 
 *t 
 *P 
 
 ^ 
 
 h 
 th 
 f 
 
 {*krd- = Eng. heart) 
 {*treies = Eng. three) 
 <*pod- — Eng. foot) 
 
 The difference between the shifted Teutonic and the 
 faithful ' classical ' tongues may now be perceived by com- 
 paring the Greek and Latin words with the English words 
 in the above list, viz: — 
 
 I.-E. 
 
 Greek 
 
 Latin 
 
 English 
 
 '*ghans 
 A- *dhe- 
 *bhu- 
 
 Xr/v {then) 
 dyj- {the-) 
 (f)v- {phu-) 
 
 {h)anser 
 fe-{ci) 
 fu-{i) 
 
 [goose 
 
 ]sVdee{d) 
 \be 
 
 { *ge7ios 
 u\*duo 
 y*slub- 
 
 yevos {genos) 
 8vo {duo) 
 
 genus 
 duo 
 
 {s)lrib-{ricus) 
 
 [kin 
 
 T- txvo 
 
 \slip 
 
 '*krd- 
 Ta "^treies 
 
 y^pod- 
 
 Kap^La {kardia) 
 rpei's {treis) 
 TToS- {pod-) 
 
 c-ord-{is) 
 
 tres 
 
 ped-{ein) 
 
 [heart 
 S' three 
 \foot 
 
 One modification of this simple and symmetrical scheme, 
 in the shape of Verner's Law, has already been mentioned 
 and will require further reference. Certain other qualifica- 
 tions will be stated in due course. For the present we shall 
 do best to keep to the main points in the description of 
 Grimm's Law.
 
 Phonetic Laws : ' Gri^nms Law ' 321 
 
 At the ' First Shifting,' as described above, the majority 
 of the Teutonic speeches stopped. The formula first given 
 remains generally applicable to the East and North Teutonic 
 tongues (Gothic and Scandinavian) and to those dialects of 
 North-West Germany which have since developed into 
 English, Netherlandish, and Platt-Deutsch. With South 
 Germany and its dialects the case was different, and the 
 previous formula is only partially true for German proper, 
 as it is now spoken. 
 
 In the seventh century A.D. we find the southern dialects 
 experiencing a ' Second ' or ' High German ' shifting. The 
 cause of this divergence is again obscure, but is probably to 
 be sought in racial influences (Celtic and Rhaeto-Celtic) more 
 peculiar to the south. The extent of this ' second shifting ' 
 is apt to be overstated. As enunciated by Grimm, it 
 appeared to affect all the explosive consonants alike, so that 
 it was necessary to add a new column to the formula. We 
 may express this by the following : — 
 
 Incorrect Statement of Grimm's Law 
 
 - -^ Primitive Teutonic and „ , , ... , ,, 
 
 I--E- Teutonic Tongues other than ^^^ ]}}S^, V-^";™^" 
 
 O.H.G. (O.H.G.) 
 
 A = M = T 
 
 M = T = S 
 
 T = S = M 
 
 Such a scheme is doubtless attractively symmetrical, but 
 the spurious symmetry thus imparted by Grimm was only 
 partially obtainable, and then only by extremely violent 
 handling of the South or Old High German dialects. Not 
 only did these show little consistency or agreement with each 
 other, but even after hunting for his examples first in one 
 dialect and then in another, however obscure, Grimm found 
 it difficult to collect a sufficiently respectable array to bolster 
 up his third column. For any comparison of Modern High 
 German with the Low German speeches the scheme is in the 
 highest degree misleading. Thus the initial consonants in 
 Gafis and goose, Bruder and brotJier, Fuss and foot are the 
 
 (B 884) ^
 
 32 2 Natu ral History of L anguage 
 
 same in High German as in English, whereas the formula 
 would demand ^Kans, ^Pj'uder, \Buss. Grimm's Law, it is 
 true, only postulates such forms for some dialect or other of 
 High German in its ' Old ' stage, not for the modern German 
 language. It is perhaps not his fault if the scheme has been 
 too commonly stated as if it were of use for the comparison 
 of the languages as they actually exist. For that purpose 
 it will be found, though not absolutely sufficient, much nearer 
 the mark to state the law thus : — 
 
 Teutonic Languages Modern German 
 
 I.-E. (including German, except in (In regard to the 
 
 regard to the dentals) dentals) 
 
 A {*dhc-) = M {do) = T {tun) 
 
 M {"^duo) = T {two) = S^ {swez) 
 
 T {*treies) = S^ {three) = M {drei) 
 
 It will now be perceived how strictly the following words 
 obey the law, there being but one form for all Teutonic in 
 the case of other consonants than dentals, whereas on the 
 appearance of a dental the High German at once shows the 
 ' second shifting-.' 
 
 I.-E. 
 
 English 
 
 German 
 
 *bhrdter 
 
 brother 
 
 Bruder 
 
 *krd- 
 
 heart 
 
 Herz 
 
 *pod- 
 
 foot 
 
 Fuss 
 
 *dekm 
 
 ten (Goth, taihiin) 
 
 zehn 
 
 *pet- 
 
 feath{er) 
 
 Fed{er) 
 
 It is true that High German, like English, has to reckon 
 with a number of modifications which have occurred since 
 the general sound-shifting. Just as English has come to 
 say thatch for thak, so German now says Dach for Dak and 
 Schaf for Schap, the final tenuis being weakened into a 
 spirant in each case. Nevertheless to enter into these 
 details would be but to blur the general view with which 
 we are here concerned as sufficient. 
 
 Of sub-laws affecting the application of the formula 
 
 1 S (Spirant) includes/ -v, th, s, i (u), h.
 
 Phonetic Laws : 'Grimms Law' 323 
 
 to the English language it is enough to mention the 
 following : — 
 
 (i.) The law does not apply to onomatopoetic words, 
 which are directly imitative of sounds heard and mimicked 
 alike in each generation, e.g. cuckoo^ bow-wow. A Greek, 
 Roman, or English child would naturally reproduce these 
 in the same way. 
 
 (ii.) In the original combinations of consonants sk^ st, sp, 
 kt, pt, the second consonant is protected from the Teutonic 
 shifting. Thus stand (Lt. stare), guest (Lt. hostis, literally 
 'stranger'), spy (Lt. -specio), right (Lt. rectus). Also in the 
 combination ght both consonants adhere to their class, e.g. 
 in '^dkughter, Eng. daughter, Germ. TocJitcr. 
 
 (iii.) Verner's Law (as mentioned above) applies not only 
 to t but also to the other tenues k and /. When the Indo- 
 European accent followed these in other syllables than the 
 first, the three tenues are represented, not by spirants, but 
 by mediae {J..e. T becomes M). The instances of fader, 
 7noder, loud have been given. Similarly A.S. swegor=QtV. 
 kKvp6<i (root *s?jek-, with accent on the suffix). 
 
 In addition to these statements of actual phonetic laws 
 peculiar to the Teutonic branch, and therefore affecting the 
 English tongue, the following important considerations must 
 not be forgotten : — 
 
 {a) Grimm's Law and its sub-laws have no concern with 
 English words which have been borrowed from Latin, 
 French, or any other foreign speech. These words naturally 
 exhibit the consonants of the language from which they 
 were borrowed. Thus pain contains the same p as the 
 Greek ttolvi^ {poind) and Latin poena, simply because it is 
 the French word peine, and French, being the outcome of 
 Latin, is a language with which the sound-shifting of 
 Grimm's Law has nothing to do. Similarly/// is an Anglo- 
 Saxon borrowing from the Latin puteus, dish from discus, 
 candle from candcla. 
 
 {b) As has been stated already, English and other 
 Teutonic speeches, after the old period in which they
 
 324 Nat II ral History of L anguage 
 
 illustrated more closely the simple shifting of the formula, 
 have proceeded to various further corruptions in certain 
 respects. English, for example, in its oldest period, used 
 the forms thak and wak, and these are the exact representa- 
 tions by Grimm's Law of the original *tcg, ^ncg. In more 
 recent English, however, they have become thatch and watch, 
 or, as it is technically expressed, k has become ts (or 
 ' palatalised '). Similarly, corresponding by Grimm's Law 
 to the Latin ge?ia and the Greek YeVi"? {genus) the Anglo- 
 Saxon word is cin (pronounced khi). But in this case also 
 palatalisation has since taken place and the result is cliin 
 {tsiii). Answering to Latin piscis (original *pisk-) we meet 
 in A.S. with the correct form fisc, but later Jish. In the 
 same way we find an A.S. initial g before a palatal vowel 
 subsequently turning into y. Thus A.S. geard, geolu, gestra- 
 became yard, yellow, yestei'iday). Such palatalisations as 
 these must be allowed for in dealing with modern English. 
 
 {c) Sounds once heard in an older stage of the language 
 are sometimes entirely lost from modern English. Thus 
 hliid i^klutos) becomes loud, hraiv (Lt. ci'i'i-dus) becomes 
 raw, stigel becomes stile, ic (Lt. ego^ becomes /. Ten once 
 contained the h (seen in Gothic taihun, Germ, zehn) which 
 should answer to the k in *dckiu. 
 
 It follows from what has now been said, that in applying 
 Grimm's Law to the English language we must always — 
 (i) Go back to the forms of oldest English. 
 
 (2) Disregard all words borrowed from foreign tongues. 
 
 (3) Remember the exceptions which come under the 
 sub-laws, 
 
 (4) Understand that the comparison is strictly with the 
 original Indo-European and not primarily with the * classical ' 
 tongues. 
 
 The last point is of much importance. The habit of 
 treating Grimm's formula, even when amended in the 
 Teutonic columns, as if it stated the relations of English 
 and German directly to Greek and Latin is an erroneous
 
 Phonetic Laws : 'Grimm s Law' 325 
 
 one. The relations which are stated are those of English 
 and German to the primitive original and to each other. It 
 is true that the classical tongues are on the whole faithful 
 to the original. Nevertheless these also have their peculiar 
 phonetic laws, which prevent an equal adherence in all 
 cases to the same class of consonant as occurred in the 
 original. Thus Latin turns the original aspirates, when 
 initial, into spirants. There are also special circumstances 
 in which it represents dk by b or by <:/, gh by g, bh by b. 
 A discussion of the rules for these and other phenomena 
 belongs to the special phonology of Latin. The reason of 
 their mention here is that an English word like red, which 
 correctly answers to the Latin ruber, or like mid, which 
 correctly answers to viedius, might easily be suspected of 
 violating Grimm's Law, on the ground that the English 
 Media (M) should, according to the formula, be answered 
 by Latin Aspirate (A). Yet what the formula really 
 enunciates is simply that the English Media (M) is a 
 derivative from an Indo-European Aspirate (A). In other 
 words, the d in red and mid represents original *dh. And 
 this is true. The Indo-European forms are to be postulated 
 as *rudhros and *inedhips. In these cases it happens that 
 Latin, by a rule of its own, does not remain true to 
 the class of the consonant, but represents an Aspirate 
 by a Media. In order therefore to know whether an 
 English and a Latin word correspond by phonetic law to 
 each other, we must know not only Grimm's Law and its 
 qualifications as applied to English, but also the phonetic laws 
 which have meanwhile operated upon the Latin language. 
 
 The proper method of testing the supposed identity of an 
 English word zvith a classical ivord is to find the I.-E. original 
 of the English word according to English phonetic laws 
 {including Grinivis) and then to ask zvhctfu-r the said original 
 could become the classical ivord in question. Are mid and 
 mcdius, for example, the same word, without borrowing on 
 either side ? This is really to ask whether, since mid pre- 
 supposes an original *medh-, the said ^rnedh- could result
 
 326 
 
 Natural History of Language 
 
 in the med- of the Latin iiiedtus. Of course the process 
 may be reversed by beginning with the Latin word, 
 postulating the original from which it would or might come, 
 and then asking what the shape in English ought to be by 
 the phonetic laws of that language. Given the Latin fdgus, 
 Greek (f)r}y6<i (^pJiegos), as the name of a tree, what shape 
 should we expect for the same word in English ? The 
 answer is that, since fdgus and pJugos point to an original 
 *bhdgos, the Old English form should by Grimm's Law 
 contain the consonants b—k. In point of fact the Old 
 English was boece (with c as k), and if this now appears as 
 beech (though a by-form of the original is possibly retained 
 in book), the ch is but another instance of the palatalising 
 tendency already observed in thatch, watch. On the other 
 hand, given the English tooth (once toiitJi), we postulate by 
 Grimm's Law a primitive Indo-European form containing 
 the consonants d-nt. Looking thence to Greek and Latin, 
 and applying the phonetic rules of those languages (which 
 in this instance do not change either consonant), we are 
 satisfied with Greek o-hovr- {o-doiit-), Latin dent-. 
 
 For the sake of clearness we may tabulate a few 
 examples of the process indicated : — 
 
 (I) 
 
 English 
 
 (2) 
 
 Original I. -E consonants 
 according to Grimm's 
 Law and sub-laws 
 
 (3) 
 
 Greek or Latin natural 
 
 derivative from (2) 
 
 yoke 
 
 *i---g 
 
 JLat. iugU7n 
 
 (Gk. {'vyo'v {zugon) 
 
 guest 
 
 * gh - - st 
 
 Lat. hostis 
 
 gall 
 
 *gh--l 
 
 ( Lat. fel 
 
 \Gk. x^^V {chole) 
 
 thatch {thak) 
 
 *t---g 
 
 JLt. tego 
 
 \ Gk. Tcyo? itegos) 
 
 harvest (A.S. haerfest) 
 
 * k - rp 
 
 tLt carpo 
 Gk. KapTTos {karpos) 
 
 feather 
 
 *p--t 
 
 /Lt. peto ('fly to') 
 \Gk. TTiTOfiaL (petomat)
 
 Phonetic Laws ; ' Grimm s Law ' 
 Or conversely : — 
 
 327 
 
 If for the English words in each case we seek the German, 
 it will be remembered that the dentals only are in question. 
 Thus for the consonants affected by Grimm's Law in tkak 
 
 (t/iaJc/i),/eat/ier, ten, sweat, we shall expect in German d k 
 
 (later d--ch, Dach), f--d {Feder), z--h {zeJin), sw - - z 
 {schwizzen). 
 
 With inquiry directed in this way, exceedingly interest- 
 ing correspondences of words are to be discovered, as may 
 be judged from the few examples already given. 
 
 In conclusion, as some practical assistance to the testing 
 of the original relationship of English, German, and classical 
 words, wherever Grimm's Law is in point, we may make 
 the following fuller tabulation : —
 
 328 
 
 Natural History of Language 
 
 Greek Latin 
 
 Original 
 I.-E.i 
 
 Early Teutonic (and English) 
 
 High German. 
 
 ' Second Shifting ' 
 
 in dentals 
 
 [k c 
 [tt P 
 
 *k 
 *P 
 
 {a) 
 
 ih 
 ^-.th 
 
 { (b) {c) \ 
 
 with unchanged 
 Verner's Law in 
 
 g sk 
 d st, pt, hi 
 \ b sp ) 
 
 d (but sf) 
 
 fy g 
 m4s d 
 
 \^ b 
 
 *d 
 *b 
 
 ik {ch when palatalised) 
 T-v / 2 (initial), jj, sz 
 \p (/at end) 
 
 [X /2-/(^ interior) 
 A- Q /(interior <^, d) 
 [^/(interior b) 
 
 *gli 
 *dh 
 *bli 
 
 yi\d 
 
 b 
 
 t 
 
 ^ Strict science would demand that we should add certain sounds which have 
 been omitted with a view to simplicity. These are the velars, for which the corre- 
 spondences are — 
 
 I.-E. 
 
 Teutonic 
 
 Greek 
 
 (varying according to special 
 
 phonetic laws) 
 
 Latin 
 
 (varying according to special 
 
 phonetic laws) 
 
 *g^ 
 
 [h)w, wh 
 kw{ — gu), c 
 
 K.T.n. 
 fi.o.y. 
 
 4>.e.x- 
 
 gu.c 
 g.v. 
 f.v. 
 
 Thus English warm may answer to Greek ^ep/i6s (thermos), 'La.i.formus, from 
 *gkit - - rtn - ; come = §alvui {baind) — venio from *gu - • m -. But these are matters 
 of the special treatise.
 
 CHAPTER XIV 
 
 GENERAL PHONETIC TENDENCIES IN LANGUAGE 
 
 While each language thus exhibits its own phonetic laws, 
 it is not easy to discover phonetic laws which are of 
 universal application. A sound or combination of sounds 
 favoured by one language is for some reason rejected, or 
 never developed, by another, although the two speeches may 
 be closely related. Thus English has entirely abandoned 
 the so-called ' guttural ' ch (really either palatal, as in the 
 German ich, or velar, as in ancJi). Our spelling still offers a 
 token of its previous existence in \\\& gJi of night, 7'ight, eight, 
 bought, but the pronunciation is attempted only in a word 
 like the Scottish loch, which itself proves the rule by the very 
 difficulty which the average Englishman finds in articulating 
 it. Meanwhile the sound is familiar and easy to a German. 
 On the other hand, while English retains the old pro- 
 nunciation of %v, as in ivork, zvine, ivit, German has rejected 
 it for the sound v. 
 
 In few such cases of national preference can we speak of 
 the greater and less intrinsic ease of the respective sounds, 
 as if the one were inherently more difficult of articulation 
 than the other. Occasionally, it is true, we may make that 
 assertion. Doubtless some appreciable muscular effort is 
 required to utter the aspirate h. Its omission is distinctly 
 * easier,' and therefore nearly all languages in which it has 
 existed have shown a tendency to drop it, and many — such 
 as Greek and the Neo-Latin tongues — have dropped it 
 altogether. Doubtless also the entire omission of other 
 
 329
 
 330 Natural History of Language 
 
 sounds, as in the English night and eight, may fairly be said 
 to represent an absolutely easier articulation than would be 
 the case if they were retained. All languages therefore tend 
 more or less to the omission of sounds, although they may 
 not be the same sounds. Again, an explosive, such as k, 
 demands a completeness of contact which there is a common 
 tendency to relax, and the result may be a spirant, e.g. ch 
 (ts) or s (.z). 
 
 Nevertheless, in any comparison of the sounds affected 
 and rejected by one speech with those affected and rejected 
 by another, it is generally futile to discuss their relative 
 difficulty from any absolute standpoint. For the most part 
 it can only be said that a familiar sound is easy, while 
 an unfamiliar one is hard. If not attempted until a certain 
 age, the English th generally baffles the Frenchman, just as 
 the French u or en generally baffles the Englishman. But 
 this cannot be put down to any absolute and inherent 
 difficulty in the English th or in the French u or en. It 
 simply means that certain actions of the vocal apparatus — 
 a certain adjustment of the tongue and teeth — necessary for 
 th, are actions of a kind to which a Frenchman has not 
 been habituated, whereas an Englishman, on his side, has not 
 been accustomed to raise, but on the contrary to drop, his 
 larynx while shooting out his lips for the z^-sound. The 
 position for th is in no sense harder or more effortful than 
 for / or V, which are sounds entirely familiar to French ; 
 nor is the position for the French u in itself any harder 
 than for the common English sound of ee in g^een. The 
 relative intrinsic difficulty of sounds may therefore be here 
 set aside as a complicated question of a physiological nature. 
 
 It is impossible to trace the various conditions, racial, 
 climatic, and social, which may have determined the 
 peculiarities and limitations of articulation of a particular 
 people. The chapter of mere accidents would almost 
 certainly be a considerable one. But against one or two 
 misconceptions it is necessary to guard.
 
 General Phonetic Tendencies in Language 331 
 
 Experience proves that any actual physical difference of 
 the articulating organs of different peoples plays little or no 
 part in the matter. An English child born and bred in 
 France or in Hindustan is as fully able to articulate the 
 language of his adopted country as the native child. A 
 Chinese born and educated in Australia has no perceptible 
 difficulty with English sounds which his parents could not 
 achieve. 
 
 Again, a very little observation will show that there is 
 small truth in the common notion that the languages of 
 warm regions are necessarily ' softer ' in their sounds than 
 those of colder and more strenuous zones. In actual fact 
 some of the hardest and harshest sounds (from our point of 
 view) are to be met with in the gutturals of the Semitic 
 languages of Arabia and Northern Africa, and in the modern 
 Spanish with its J and g as influenced by the Moors. The 
 careful philologist will not at present sanction any broad 
 statements as to the necessary effects of this or that strain 
 or climate. He will be satisfied with observing the facts of 
 each language. From these he will derive the conviction that 
 the ' difficult ' sound, like the ' ugly ' sound, generally means 
 the sound to which the articulating organs and the ear are 
 unaccustomed. The fallacy of ' ugliness ' is indeed an obvious 
 one. Latinists of the old school are prone to speak of the 
 now recognised pronunciation of e.g. neu patriae validas in 
 viscera vertite vires (with v sounded as w) as ' ugly ' in itself. 
 In other words, w is an uglier sound than v. Yet the same 
 judges would equally describe as ' ugly ' Mr. Weller's pro- 
 nunciation of his own name and of vidows. It turns out that, 
 here at least, the ugly sound is simply that which the ear 
 does not expect. 
 
 The conditions and influences, then, which have 
 determined the various schemes of articulation developed 
 by various peoples, are beyond discovery. The linguals 
 / and r, for example, are (so far as we can tell) equally 
 ' easy.' The tongue is brought to the roof of the mouth
 
 332 Natural History of Language 
 
 above the teeth. If its sides are allowed to relax, so that 
 the air-stream may issue, the result is /; if it is the tip 
 which is relaxed, the result is r. Why therefore, while 
 very many languages possess both r and /, should / be 
 unknown in Maori but r be unknown in Chinese? In 
 Chinese Christ becomes Kilissetu ; on the contrary, in 
 Maori Babylon becomes Paparona. The Maori or Chinese 
 child who is taught English from the cradle can use r and / 
 indifferently. Moreover, the fact that a language now 
 contains no r cannot be taken as proof that it never did 
 contain that sound. There is apt to be an interchange of r 
 and /, as in the Italian albero for Latin arbor or French 
 rossignol for lusciniola, and frequently in Sanskrit. Primi- 
 tive Greek certainly once possessed the sounds zu and y 
 (//, i), and the former at least was still vigorous in the day 
 of epic poetry, and died but slowly in certain of the dialects. 
 Yet the ordinary classical Greek, like the modern, possessed 
 neither w nor j. Similarly the zv of Old Latin is unknown 
 to its Romance descendants, which have substituted v. 
 
 If all the sounds in all the languages of the world were 
 collected and classified, it would be found that each 
 individual speech employed but a portion, greater or smaller, 
 of the total. Max Muller observes that in Hindustani 
 there are forty-eight consonants, while Polynesian is satisfied 
 with ten. Sanskrit had no short vowels e and o, although 
 these had been possessed by the original Indo-European, and 
 are faithfully retained in Greek and duly represented else- 
 where. These sounds cannot have been ' harder,' even to the 
 mouths of primitive speakers of Sanskrit, than short a, i, and u. 
 With any theory as to the influences which caused the 
 transformation of e and o into a in that branch of the family 
 we are not here concerned. The present point is that neither 
 inherent difficulty of the sound, nor peculiar physical 
 formation of the vocal apparatus, nor peculiarity of climate, 
 must be urged as an efficient cause of this phonetic 
 aberration. All we know is that a tendency must somehow 
 have begun, spread, become general and finally universal.
 
 General Phojietic Tendencies in Language 333 
 
 The process by which such spreading took place has been 
 sufficiently outlined in a previous chapter. 
 
 It has been necessary to safeguard with this brief 
 introduction the usual statement that languages all alike 
 modify their sounds in the direction of greater ease. This 
 is true, but the greater ease is generally not that of some 
 absolute standard, but is relative to the habits of the 
 speakers respectively concerned. A sound which manifestly 
 demands considerable muscular exertion may be so common 
 in a particular language that it may even attract towards 
 likeness with itself another sound which is intrinsically 
 easier. Thus the Irish articulation of t and / as t\ p\ i.e. 
 as t x>x p followed by a sound approaching the aspirate, may 
 be said to demand more effort than the English / or p. 
 Yet an Irishman accustomed to t' and p' will find it easier, 
 from force of habit, to say T'oinies than Times. There is 
 actually more work on the part of the articulating organs in 
 pronouncing Koroivhata than in pronouncing Croft. Yet 
 to add that work is easier for the Maori than to break an 
 inveterate habit and articulate two consonants together. 
 
 That there are, indeed, certain sounds which do 
 demonstrably call for less muscular effort than certain others 
 has already been acknowledged. Thus in a stop-sound (or 
 explosive) the air-stream must be unequivocally blocked. 
 With or without the knowledge of the speaker there is 
 decision on the part of the articulating apparatus. Thus 
 in the case of k the check made by the back of the tongue 
 against the palate is complete ; for b the lips are entirely 
 closed. Such resistance to the air-pressure, such actions of 
 the tongue and lips, are muscular and involve effort, even 
 though that effort be unconscious. It may be assumed that 
 in general the effort would be less if the complete contact 
 were not insisted upon. If the closure of the lips were less 
 decisive, or the block with the back of the tongue less 
 determined, there would be somewhat less demand upon 
 muscular energy. In other words, a more languid
 
 334 Natural History of Language 
 
 articulation tends to leave a fissure or chink in the oral 
 passage. The result is a spirant, and ch (Scotch or German) 
 takes the place of k, and v of b. Such a change might occur 
 particularly at the end of words, by reason of the relaxing 
 effort. Thus in German Dach is an easing of \Dak, and the 
 English thatch, batch, and the like, though they involve more 
 change than the German (including a change of place to the 
 front palate, as represented in the phonetic symbol ts), illus- 
 trate the same operation. The modern pronunciation oicalf, 
 as compared with an older \kalb-, is another example of the 
 tendency to drop from the muscular standard of a stop to 
 the less exacting standard of a spirant. 
 
 It may be accepted that such changes from stop to 
 spirant are changes in the direction of absolutely greater 
 ease, that is to say, less effort. Also the omission to 
 articulate final consonants, so observable in the French /, s, p 
 {est, fois, drap, etc.), is an evident relaxing of exertion. The 
 same phenomenon occurred in very early Greek with nearly 
 all consonants except s, r, n. Thus tr/S {tid) became rl (tz), 
 'fyvvaiK (gunaz'k) became jvvat {gunai). In the spoken 
 Latin also a final m or s was scarcely audible, if at all, and 
 locus, locum were pronounced locii. Chinese also has clipped 
 a word like dap into ta, and in Polynesian the Tongan ato is 
 all that answers to the Malay atap. Final vowels have been 
 lost for the same reason. This has been conspicuously the 
 case with English. Large numbers of Anglo-Saxon words 
 ended in various vowels, e.g. a, o, u. These first became all 
 alike slurred into e (the indeterminate d), and then even that 
 sound gradually disappeared from use. We still often write 
 its sign (though inconsistently) without any consciousness of 
 its origin. The history of the French ' ^-mute ' is the same. 
 Peine was once a word of two syllables, like the Latin poena, 
 but after the pronunciation had become, by the first laxity, 
 ^pend, it passed, by a second laxity, vc\\.o pen. 
 
 Nor is it merely final sounds, whether vowel or consonant, 
 which have disappeared through remissness of effort. In- 
 terior syllables, when they bore no accent, have been first
 
 General Phonetic Tendencies in Language 335 
 
 slurred over and then totally omitted. A striking example 
 of progressive curtailment is that of the English via'ain. 
 The pronunciation is now monosyllabic {inani), and in the 
 mouth of the uneducated it is often simply 'in ( = 7;i), as in 
 Yesm. The chief steps in the abridgment are niea domina, 
 m{e)a doinna, madam, mdam {inam), in. In this, as in 
 countless other instances, economy of effort is self-evident. 
 Languages differ much in the extent to which they have 
 omitted their unaccentuated syllables, but the tendency is 
 natural and is very widely spread. Old Latin thus synco- 
 pated such words as ^subrego {swgo), ^usiirapo {usurpd). 
 With French such syncopation is a special characteristic. 
 Thus septimdna becomes scmaine^ Sequana becomes Seine, 
 decima becomes dime. In English Thursday = Thunresdaeg, 
 York = Eoforwic, England — Aenglaland \ forecastle is i^xo- 
 nounced/i?^'j/(?, lark = laverock. It is through the operation 
 of the same tendency that in German bei dem is more 
 commonly beim and ::ic der shortened to ziir. 
 
 Of distinct easing or economy of effort we have thus 
 enumerated the following cases : — 
 
 (i.) Relaxings of closure and casings of a stop to a spirant. 
 
 (ii.) Omissions of final consonants and vowels. 
 
 (iii.) Omissions of unaccented syllables in the interior of 
 words (or syncopation). 
 
 These, however, are by no means all the classes of sound- 
 change which are plainly due to the same cause. There 
 may be omissions from the beginning as well as from the 
 end. Greek thus entirely discarded the initial z£/-sound, as 
 in o\ko^ ( = ^{?i)oikos), exactly as it tends to be omitted in 
 dialectal English or Scottish before the sound 00 (u) in 
 'ooman, '00' ( = 7vool) and the like. Old Irish similarly 
 dropped the initial / in such words as at/itr (Lat. pater), 
 ore (Lat. porcus). In English mend = (a)mcnd, fend = 
 (de)fend. 
 
 Less obviously, but quite as certainly, to be placed among
 
 336 Natural History of Language 
 
 yieldings to ' absolute ' ease are the assimilations of one 
 sound to another when the two come to be juxtaposed. Such 
 assimilation may occur in respect of either the place of 
 articulation in the mouth, or of the kind of articulation. 
 While it cannot be shown that, of the two sounds in and n^ 
 one is more or less effortful than the other, and while there- 
 fore the change of the labial ni to the dental n cannot itself 
 be taken as an example of economy of effort, yet when n is 
 substituted for vi in the combination nit or mdy we are 
 justified in regarding it as such an example. The Latin 
 centum represents an older ^cevitum, just as English hundred 
 represents a more ancient ^Jmmd-. The modification of 
 Latin verumtamen and scpteni triones into veruntamen and 
 septentrio is actually seen in process of occurrence. It 
 requires little discernment to realise that the successive 
 articulation of two different consonants, one at the lips (;«), 
 the other at the teeth (/), requires more muscular exertion 
 than that of two consonants at the same place (viz. n and /). 
 Similarly, when a consonant with one kind of articulation 
 (voiceless, voiced, or aspirate) is brought into immediate 
 contact with a consonant of another kind, it is absolutely 
 more easy for the vocal machinery to adapt or assimilate the 
 two to each other so that both are of the same kind. Thus 
 in the three Greek words Kpv<^Qoi {krupJitho), icpvirro^ {kruptos)^ 
 Kpv^ha ikrubdd) the variation in the final letter of the root 
 (viz. Kpv^- KpuTT- Kpv^-) is caused by the difference in the 
 character of the succeeding letters {-9 -r -B). The aspirate 
 6{t/i), the voiceless stop T(t), and the voiced stop 8{d) of the 
 suffixes assimilate to themselves the final sound of the root, 
 making it aspirate (f)(p/i), voiceless stop 7r(p), or voiced stop 
 I3{b) respectively. The sounds in speech, it has already been 
 observed, are not pronounced singly and staccato. They link 
 themselves together very much as writing links together the 
 letters in a word. Just as the pen adopts the easiest or most 
 fluent method of ' running on ' letter into letter, so the organs 
 of articulation follow the flowing course of least effort in 
 ' running on ' sounds. One result is assimilation, as above
 
 General PJionetic Tende7icies in Language 
 
 00. 
 
 described. Another is the development of the glides into 
 fully audible sounds. The glide from n to r develops into 
 d in the French gendre, tendre, cendre (w^hence English tender, 
 cinder), from 'fge/ire, "^tenre, fcenre (l^atin gen(e)rujH, ten{e)ruin, 
 cin{e)reni). So in Greek taVpe? {anres) becomes avhpe'^ 
 {andres). Although this development appears at first sight 
 to be adding to the amount of articulating work, it is in 
 truth a saving of muscular energy to pronounce ndr in place 
 of nj', just as it is easier for the pen to link the two letters in 
 writing than to lift itself after « and begin afresh with r. In the 
 same way it is easier to pronounce mbl or ;;/// than ml alone. 
 Hence humble instead of ^kwnle (Lat. hum{i)lis), and the 
 Latin exempluin (' a thing picked out ') for ^ex-em-hwt. The 
 old spelling HaDiblet represents a natural easing of Hamlet, 
 as does the modern Ambleside for Hammel-set. 
 
 To the three forms of easing first specified we have 
 therefore to add — 
 
 (iv.) Assimilation of neighbouring sounds, 
 (v.) Development of glide-sounds. 
 
 Of all these five forms — relaxings of closure, omissions of 
 final sounds, syncopation through loss of unaccented syllables, 
 assimilation, and development of glides — we are entitled to 
 say that they result in a demonstrable saving of effort and in 
 the creation of an absolutely easier articulation in place of a 
 harder. But of many other changes adopted by this or that 
 language we cannot say the same. The new sounds are not 
 necessarily always easier in the sense that less precision or 
 energy is required, and a phonetic law of a given tongue 
 may in reality run counter to any supposed rule that change 
 is always and inevitably in the direction of diminished work 
 on the part of the articulating organs. Such a rule might 
 perhaps exist if the mind, or artificial social influences, played 
 no part in determining sounds ; if, for example, the shape 
 of one word did not influence the shaping of another. 
 Doubtless the effortful English h, which was disappearing 
 
 (B 884) Z
 
 2,;^S Natu ral History of L anguage 
 
 from the language, has been brought back mainly by the 
 conscious reaction of education ; but the result is that 
 the sound has been imported into a number of words in 
 which it has no historical right to exist. In this instance 
 we may be aware of the influence which runs counter to 
 phonetic easing, but the counter-running is none the less a 
 reality. In many cases we are not in a position to discover 
 the cause, but we cannot deny the fact. 
 
 Notwithstanding the qualifications which it has seemed 
 necessary to make, it remains now to reaffirm the doctrine 
 universally held, that tJie car dhial principle of sound-change is 
 economy of exertion. That each and every change must be 
 in that direction is indeed a statement by no means to be 
 accepted. Nevertheless the principle itself is one of over- 
 whelming power, and the exceptions count for comparatively 
 little. Omissions, casings, and assimilations are tendencies 
 of language everywhere met with. The popular Latin 
 jnetipsissiuiiun now appears simply as the French nicine or 
 the Spanish inisnio, and, if inea doniina appears in English 
 as the monosyllable indam, it is because the vocal apparatus 
 has step by step stinted its task of full and exact enuncia- 
 tion. Of the slurring and abbreviating which have taken 
 place at each step the speakers have nearly always been 
 unconscious. The series of generations which gradually 
 curtailed inetipsissiniu{m) into victissiniu, this into metesvie 
 and medes7ne, thence into mesme and so to meme, were not 
 alive to what was taking place in their own articulation. If 
 the word had proved unintelligible when uttered by the 
 numbers, small or large, who first dropped into the laxity 
 of omitting a portion of its sounds, the remissness would 
 have been checked. The same result would doubtless have 
 occurred if there had existed a universal education alert to 
 criticise such laxity and to brand it with reproof or ridicule. 
 But, in the first place, each step in the change was so slight 
 and gradual that there could be no mistake as to the 
 meaning, and, in the second, there existed no such general
 
 General Phonetic Tende^tcies in Language 
 
 OJ. 
 
 educational criticism. Metissimu is easily understood in 
 the intended sense of the earlier word. When this is 
 established, the slurring of a further syllable {inetisum) 
 causes no perplexity to the hearer. And the case is the 
 same with each succeeding loss of sound. Again, the 
 infinitive ending -an of Anglo-Saxon, bearing as it did no 
 accent, became lightened into -en{jdn) in early or semi-Saxon 
 English. From this, again, the -n was dropped, and -^(-^) 
 alone remained. Later still the e also disappeared from 
 pronunciation, and an infinitive such as bear, tuii becomes a 
 word of one syllable instead of two. The method of 
 grammar which came to prevail in English made Jielpe quite 
 as intelligible an infinitive as kelpen, and when subsequently 
 help took the place of helpe the sense was still conveyed 
 equally well. Meanwhile it was to but an inconsiderable 
 extent that the schoolmaster was abroad to insist upon a 
 full articulation. Perhaps in the educated present day 
 speakers would hardly be permitted to drop into a slurring 
 of place-names like Cirencester, Worcester, or of person- 
 names like Marjoribanks and Cholmondeley, if those names 
 had hitherto been articulated as they are written. But the 
 case was otherwise when that gradual and tentative abbrevia- 
 tion was taking place which has ultimately resulted in the 
 pronunciations Cister, Wooster, MarcJibanks, and CJiuinley. 
 At each stage of degeneration the pronunciation still served 
 its purpose to the satisfaction of all concerned, and there 
 was little or no educated consciousness to offer an aesthetic 
 protest. 
 
 Such is the history of easing and omission of sounds in 
 all known tongues. Languages differ remarkably, it is true, 
 in the extent to which they have carried the process. The 
 causes of this difference are probably, in some measure at 
 least, racial and aesthetic. The quick social instinct and 
 the tendency to rapid thought and a volubility or nimble- 
 ness of speech which are attributed to the ' Celts ' may be 
 some explanation of the fact that French is particularly 
 distinguished by the syncopation and curtailment which
 
 340 Natural History of LangiLage 
 
 Latin words have undergone in the language. While Italian 
 tarries at inedesivto and sicuro as representatives of metipsis- 
 siimis and sealrtis, French has reached mime and stir. 
 Facere has become the monosyllable faire, while the Italian 
 fare is still normally of two syllables. In Greek losses were 
 comparatively slight, consisting chiefly, though not solely, of 
 an early disappearance of iv and y {ii and z'), omission of 
 final consonants, and contraction of vowels. Early Latin 
 tended to lose unaccented syllables, as in prudens for 
 providens, motiis for jmovitus, final s, m, and e, and often 
 one or more consonants from a group, e.g. in scdtae, which 
 stands for jscand-slae. 
 
 But a more tangible cause of difference in the degree of 
 loss is the difference in the nature and history of the accent 
 in a given language. The meaning of the two kinds of 
 accent, stress-accent and pitch-accent, has been explained in 
 the chapter on the production of speech-sounds. A stress- 
 accent is due to greater vehemence of utterance, causing a 
 deeper amplitude in the vibrations of the vocal chords ; a 
 pitch-accent is the effect of greater rapidity in the vibrations. 
 The one produces greater loudness, the other a higher note. 
 In every word which is more than a grammatical sign there 
 is at least one syllable which bears an accent. In the 
 English human there is a stress-accent on the first, in 
 iiumdne on the second. In the Greek Qka {thed, 'spectacle') 
 there is a pitch-accent on the e, in 6ed {thed, ' goddess ') on 
 the a. In Latin vionere and rcgere differ in the syllable 
 upon which a stress-accent falls. 
 
 Now, had the principle of accentuation been alike for all 
 languages, they would have been much nearer than they 
 are in the degree in which they have come to abbreviate 
 their words by the omission of interior or final sounds. 
 Had they all been characterised by a stress-accent of equal 
 intensity and of similar position in the word, or by a pitch- 
 accent of equal height and similar position, then we might 
 (racial and aesthetic considerations apart) have looked for
 
 General Phonetic Tendencies in Language 341 
 
 much the same procedure in the matter of dropping syllables 
 which were unaccentuated, or of easing sounds which were 
 not adjacent to the ' chief tone ' of the word. Thus primitive 
 Latin regularly placed a strong stress -accent on the first 
 syllable of a word, however long. Both in the nature of the 
 accent (stress) and in its position (on the first syllable) it 
 agreed with the present regular practice of English. Thus 
 primitive Latin said \n'tetuli, \c6nfacio, just as English says 
 obligatory, comparable. The consequence is that there is a 
 very close likeness between certain characteristic phonetic 
 phenomena of early Latin and of English. The strong 
 stress on the first syllable tended either to dull the vowels 
 in the weaker syllables which followed or else to cause the 
 entire loss of some such syllable. As the English vowels in 
 the unaccented syllables commonly became dulled to the 
 indeterminate 9 (as in compdrsbV)^ or a syllable disappears 
 (as in obligat'ry = Sbligdtri)^ so in the aforesaid old stage of 
 Latin there was a dulling from e.g. ^cSnfacio to conficio, 
 \nebela to Jicbula, or the loss of a syllable, as in rettiili for 
 \retetiili, dmpictdre for \dnibiputdre. It is true that at a 
 later date the Latin accent, for obscure reasons, shifted its 
 position and receded, either to the last syllable but one, if 
 that vowel was long in quantity, or otherwise to the last 
 syllable but two. Thus the accent is then either as in 
 respondere, bonitdteni, or as in conficcre, proconsillejii. Still 
 the accent was one of stress, and both its nature as such 
 and its position in the word were retained by the lingua 
 roinana which was carried into Gaul and destined to become 
 French. The history of slurring, syncopation, and curtail- 
 ment of the unaccented portions of words subsequently 
 repeated itself. Bonitdtcm, conpniare, viridis, populus became 
 bonte^ conter, vert, peuple. This process is identical with that 
 by which in English Eoforwic became York or eleeviosyne 
 became first aelmesse and then alms. 
 
 (Since, however, at a later time French has tended to 
 minimise the stress-accent or to replace it by a slight pitch- 
 accent on the last syllable, the result is a wide breach
 
 542 Natural History of Language 
 
 between the modern English and French treatment of 
 unaccented syllables. The tendency to slur them no longer 
 exists in French, except in the case of the accentless e 
 (which is really the indeterminate d), as in de, cheiniti, 
 popularly pronounced d\ cliinin. The force laid on one 
 particular syllable having been removed, its value is, as it 
 were, spread over the rest. In contemporary French the 
 vowels a, e, i,o, 21. are therefore in every syllable duly pronounced 
 and audible as such ; whereas in English, which persists in 
 its strong accentual method, they all tend to merge in a 
 common indistinct ^ when they do not happen to be in the 
 accentuated syllable. In the English Airady the second 
 vowel is 3, in French Arcadic it is a distinct a. 
 
 English, early Latin, and the older French, agreeing 
 as they do in the nature of the accent, also agree largely 
 in its phonetic consequences. Meanwhile the accent of 
 classical Greek was not one of stress but of pitch. It 
 was 'musical,' not 'emphatic' The acute accent -mark 
 upon a classical Greek vowel does not imply that the vowel 
 was delivered with more force, but that it was delivered on 
 a higher note. The terms ' accented ' and ' unaccented ' 
 syllable therefore mean for classical Greek something 
 different from that which they mean for Latin or English. 
 It follows that the history of the Greek unaccented syllable 
 is also different. Such collapsing of a word or curtailment 
 of a syllable as is frequent in Latin or English is scarcely 
 to be met with. Where it does occur it is rather due to the 
 close similarity, if not absolute identity, of two adjacent 
 syllables, which causes the speaker to omit one of them 
 as if it had been already said. Thus a/x^opeu? {avipJioretis) 
 and Ke\aLve(f)7]^ {kelainepJies), though manifest syncopations of 
 jafKJjLcfiopev^; {amphipJioreus, ' with a handle on either side ') 
 and i K€\aLvove(f>y]<i {kelainonephcs, ' black-clouded '), do not 
 primarily owe their abbreviation to the position of the 
 accent, but to the obvious cause just stated.^ One of the 
 
 ^ This omission of one of two like syllables, found in Latin also, as in 
 tru{ci)cidare, is technically called Haplolog}'.
 
 General Phonetic Tendencies in Language 343 
 
 distinguishing features of ancient Greek, as opposed to 
 Latin, is its faithful retention not only of the accentless 
 syllables, but also of the original vowels in them, and this 
 difference of habit is chiefly, though not solely, due to the 
 different character and effect of the respective accentuations 
 of the languages. Thus a hypothetical original *bheroinos 
 would appear in Greek, with simply a musical accent, 
 as i(f)epofio<; (p/ieromos), while in Latin it would sink into 
 ferimus, the accent being one of stress and of such effort 
 that the subsequent syllables suffer from a proportionate 
 weakness. 
 
 It is not, indeed, quite true that even words with a 
 pitch or musical accent tend to experience no loss at all 
 in the accentless syllables. In Greek comedy the fish- 
 monger, who superciliously grunts out the price of his fish, 
 pronounces oKT(ti o^okwv [pkto obolon) as '/ctoj 'jSoXmv (^kto 
 'bolon). This is not, indeed, the accepted language, and it 
 is actually given as an instance of slovenly curtness. What 
 it shows is that the phonetic tendency exists, though in a 
 much less powerful degree, even with the tonic-accent of the 
 Greeks. How far the counteracting of the tendency was 
 due to a subtle ear and an acute aesthetic sense can hardly 
 now be ascertained. The truth is that the difference of 
 note which a pitch-accent represents is apt to become 
 increasingly associated with greater energy also. The other 
 accentless syllables are pronounced evenly, but when we 
 come to the syllable in which a peculiar note has to be 
 struck, there is an impulse to deliver it with emphasis at 
 the same time. This impulse grew in later Greek. Hence 
 it has come about that in modern Greek the accent is 
 no longer pitch, but stress. In a word like av6pw'7ro<; 
 {iinthropos), the accent mark now represents what it would 
 represent if we placed it on a syllable in an English word. 
 As a consequence modern Greek, as actually spoken, has in 
 its turn learned to abbreviate and curtail the older words, and 
 to shorten or otherwise modify many of the older vowel 
 sounds. Stainboiil, Stancho represent (e)? Tap^ ttoKlv (e)? rav
 
 344 Natural History of Language 
 
 Kw {es tdm polin, es tan Co). Similarly the word o^^idriov 
 {ommdtion) has been cut down to yuan {nidti). This could 
 never have occurred while the accent was purely chromatic. 
 
 Stress-accent and pitch-accent thus vary in their results. 
 Yet if languages had throughout their history remained con- 
 sistent in their use and placing of either one or the other, a 
 survey of the phenomena resulting from ' economy of effort ' 
 would have been far more simple than it is. In reality a 
 language may possess the one kind of accent at one stage of 
 its history and the other kind at a later stage. As in 
 Greek, a tonic-accent may be replaced by a stress-accent. 
 As in French, stress-accent may gradually become so weak 
 that it is difficult to recognise it as such. Again, a language 
 may at one period affect a particular position of the accent 
 in every word, but may subsequently shift the position to a 
 different syllable. Latin has already been cited as an 
 example. Had the accentuation which prevailed in historical 
 times (as in conficid) been always in existence, the history 
 of Latin words would have been very different from that 
 which we find. Thus, at the date when facio was first joined 
 to tc<?;;/, the total \c6nfacio took the vigorous stress-accent 
 on the first syllable. The unaccented a was in consequence 
 weakened to /. But, if the later system had then been in 
 use, \confdcio would before all things have retained the a, 
 whatever other vowel might suffer. Similarly, if the present 
 French accentuation (or non -accentuation) had been in 
 vogue when the lingua roniana commenced with viride' and 
 securic , the cutting-down into vert and si\r could not have 
 occurred. 
 
 Such vacillation in the nature and position of the accent 
 is a common phenomenon in linguistic history. It is practi- 
 cally certain that at one stage the primitive Indo-European 
 employed a strong stress-accent. It is in no less degree to 
 be assumed that in its later stages, just before the various 
 branches of the family spread apart, the accent was mainly 
 one of pitch. Probably the two methods existed side by 
 side. In modern Swedish we meet with such a combination,
 
 General Phonetic Tendencies in Language 345 
 
 the high pitch without stress and the stress without high 
 pitch being quite familiar in Swedish words. In the end the 
 pitch-accent in the primitive speech appears to have prevailed, 
 its position in the word being variable. The changes in 
 Greek and in French accentuation, as described above, are 
 sufficient evidence of the possibility of alteration in kind 
 within the same language. Among the branches derived 
 from the parent original we find Sanskrit and Greek con- 
 tinuing to employ the pitch-accent, while Armenian, Latin, 
 Teutonic, and Celtic substitute the accent of stress. In 
 point of position the Greek accent cannot be placed earlier 
 in any word than the third syllable from the end, but within 
 that limit it varies, sometimes according to rule, sometimes 
 not. In Sanskrit its range is more free. In all Old Teutonic, 
 in Old Latin, and in Old Irish its place is on the first syllable. 
 The Latin accent subsequently shifted, as we have already 
 seen. Among modern tongues of the Slavonic branch Polish 
 lays the accent on the last syllable but one, while Tzech lays 
 it on the first. In modern English or German the old 
 Teutonic principle still holds. It is not, indeed, invariably 
 observed, particularly in words compounded with particles, 
 e.g. besides, without, forlorn, although of these a special 
 account can readily be given. Nevertheless the principle 
 still holds so strongly that foreign words borrowed into 
 English commonly shift their accent as they become 
 naturalised. Already in Chaucer the accent varies, as 
 between fdrtnne and fortune, although in general a com- 
 parison of his common use of the Latin-French accent in 
 Anglo-French words, e.g. honour, cordge, with the modern 
 hojiour, courage, is an object-lesson in the history of English 
 accentuation. The principle is well illustrated by blasphemous, 
 balcony, which were borrowed as blasphemous, balcony. If 
 there are instances to the contrary they are mostly explain- 
 able with ease. Humane, project, frequc'nt, refuse are 
 deliberately given an un-English position of the accent in 
 order to distinguish them from human, project, frequent, 
 refuse. The latter have adapted their accent to the rule.
 
 346 N-atural History of Language 
 
 The former would have done the same if the question had 
 been simply one of phonetic, as distinct from psychological, 
 tendency. 
 
 While, then, accent plays the most important part in 
 that weakening or abbreviation of the phonetic contents of a 
 word which is attributed to economy of effort, and while all 
 languages share to a larger or smaller extent in such abbrevia- 
 tion, it is clear that they do not agree, and could not be 
 expected to agree, either in the extent to which they 
 abbreviate or in the particular portions of the words in which 
 the abbreviations occur. The pitch-accent languages will 
 curtail or syncopate less than the stress-accent languages. 
 The languages which accentuate at the beginning will lose 
 or weaken other syllables than those affected by the 
 languages which accentuate near the end. Even those 
 which have accentuated on the same syllable, as in the case 
 of the Romance languages on their path from Latin, will 
 vary in the amount and nature of their abbreviation. 
 The loss in Italian, as has been remarked, is slight as com- 
 pared with that in French. We may therefore lay special 
 weight upon the fact of a universal tendency to weaken 
 unaccented syllables, but we cannot lay down any universal 
 laws as to the degree or place of the weakening. Philologists 
 may collect and tabulate the syncopations, curtailments and 
 weakenings in a special language, and in some cases 
 (particularly in French) they may establish regular phonetic 
 laws for such occurrences in the language in question, but 
 it is not possible to establish phonetic laws which shall be 
 universal. 
 
 By way of summary we may take a few further examples 
 of some of the various forms of absolute easing and economy 
 of energy which have now been enumerated : — 
 
 (c?) Relaxings of Closure. 
 
 k >ch { — ts) 
 
 KvpiaKov A.S. cyrice (i.e. kirike, cf kirk) 
 
 > church.
 
 General Phonetic Tendencies in Language 347 
 
 A.S. cild> child. 
 
 Lat. Cicero {^ = Kiker6):>\\.d\. cicerone {tsits-). 
 k>ch { = s) L. campus >¥r. champ. 
 „ caballus> „ cJieval. 
 „ vacca > ,, vache. 
 k> c { = s), s {p) ,, Cicero > „ Czccron. 
 „ vicinus > „ voisin. 
 Gk. /cv/cXo? L. gr/?/'jr>Eng. 9'c/^. 
 
 sk>sh (i) A.S. y?i"t: >/i-//. 
 
 „ scacafi> shake. 
 
 g>j {dz) A.S. brycg > bridge { = bridz). 
 
 g>z (' soft g') L. gauderoYv. Joiiir. 
 
 „ gener > „ gendre. 
 g>y A.S. ^^(7/i9 ^yellozv. 
 
 „ geard>yard. 
 
 L. ^^c >It. zV' (from i{y)o). 
 
 And in modern Greek. 
 
 /(z) >^ L. /(?;■/(? > Fr, poison. 
 t >s „ platea> „ place. 
 
 Older Greek tStScort > StScocri. 
 t >s L.. potio >Kng. po/io7i (=pdsn). 
 
 d >l l^. olere (verb) but odor (noun). 
 
 O.L. dacruma > lacriima. 
 d >th A.S. to-gaedre > together. 
 „ fader > father. 
 And so in modern Greek. 
 
 di and di^>j (5) L. diurnum>Vx. jour. 
 I.E. ^dieus >Gk. Zeu?. 
 
 / >/ I.E. */'(?«r/- >Eng. /('^/. 
 
 O. Teut. tj-//^/>Germ. Schaf. 
 p >v (through b). 
 
 L. ripa >Fr. rive. 
 
 „ sapo{nejn)> „ savon.
 
 348 Natural History of Language 
 
 b >/ O. Teut. \kalb- >Eng. calf, 
 b >v \vl early French and regularly in later Greek 
 {e.g. /3a(TtXeu9, now pronounced vasilefs). 
 
 (b) Omissions of fijial consonants. 
 
 In older Latin verse a line like tuin lateraliis) dolor 
 certissiniuis) nuntiiiij) mortis is scanned as if the -s 
 were absent in each case. It is obvious that in 
 current speech it was not audible. The same applies 
 to -in. Hence in the pedigree of Romance words it 
 is necessary to begin with forms (accusative) like 
 secuni, a7nicu, actione (It. sicuro, amico, azione), not with 
 secur?im, amicnm, actionem. Before this stage Latin 
 had also lost sundry other final consonants, as in lac 
 for ^lact, agcr for \agers, esto for estod. Very early 
 Greek dropped all final consonants except -;;/ (which it 
 turned to -«), r, and s. Modern Greek tends to drop 
 even these {a\o^o = aXoyov). French regularly leaves 
 final f b, p, d, s, t, s unpronounced, except when 
 there is liaison with a following vowel. English also 
 has lost final consonants which were once heard, as 
 in the adverbial termination -ly (formerly -lie), in I{c), 
 plou{gh), hi{gh), day ( = daeg), woodbine ( = -bind), and 
 in the -e{n) of old infinitiv-es and other flexions. 
 Similarly hussy = /ms-wi(f). 
 
 (c) Omissions of final vowels (normally only in syllables 
 
 not bearing the accent ^). 
 In Old Latin such loss was occasional, e.g. cdlcdr{e), 
 dnimdl{e), v6lnp{e), sdti{s)n{e), dtqu{e) (whence ac). In 
 French the final Latin a passed into e (<?), and this was 
 quite audible in the older stage of the language. It is 
 now mute. Thus 'L. porta > port{e), femina > femm{e). 
 Other Latin vowels in the same situation have either 
 left no trace at all (e.g. F. mer— L. mare) or have also 
 passed through ? to ^-mute. In Middle English the 
 Anglo-Saxon vowels, used as flexional terminations, all 
 
 ^ Technically called apocope.
 
 General Phonetic Tendencies in La^igitage 349 
 
 became e{p), which was heard as a syllable and must 
 be so pronounced, though not with strict regularity, 
 in Chaucer. Thus A.S. steorra, Mid. Eng. sterr'e 
 becomes star ; Jieortc becomes heart ; siuiu becomes 
 son. So\n?ieyi\on/icai-d=/icj-di\fed—feddc\ German 
 has been subject to similar loss in its nouns, e.g. Here, 
 Solin. After the English infinitive and the plural 
 verb in -en had first passed to -^'that termination went 
 on to leave no trace at all in the spoken language. 
 • Thus coni{e) = conie~ couieji. In modern Italian a 
 
 tendency, not merely colloquial, is to drop a final e in 
 the infinitive and to pronounce far for fare and the 
 like. So tin = ?ino, del = deiio, etc. 
 
 (d) Omission of syllables not bearing the accent. 
 The initial syllable disappears ^ in e.g. — 
 
 L. Jiistoria >It. storia, Eng. story. 
 „ episcopus >Eng. bishop. 
 ,, hydropsis > „ dropsy. 
 
 So {di)spend, {de)spite, {e)gipsy, {a)vdn{f)guard, {en)gin, 
 {ap)peal, {e)squire, {op)p6sal (i.e. puzzle). 
 
 In the interior of the word syncopation has already been 
 shown to be exceedingly frequent. Further examples 
 for English are monk ( = A.S. munec from monachus), 
 fortnight ( =fourteennight), kerchief ( = coverchief from 
 F. couvrechef), palsy {= paralysis'), proxy {= procuracy), 
 lord ( = A.S. hlafweard, 'loaf-keeper'), sheriff (A.S. 
 scirgerefa), Friday ( = Frigedaeg) ; for French /;v//t' (L. 
 frdxinus), Loire (L. Ligeris), age { = edge = eddge = 
 aetaticum), dne ( = dsinus) ; for Italian donna ( = L. 
 domina), vcrde ( = viridem), vcnti ( = viginti), cittd 
 { = civitdteni), fare {^fdcere); for Latin nundinae { = 
 \n6vendinae), princeps ( = \primiceps). 
 
 {e) Assimilation of 7icighbouring sounds is too common a 
 
 * Initial omission is called aphesis by Dr. Murray when a vowel only is lost, 
 but aphaeresis when the syllable contained more than such a vowel.
 
 350 Natural History of Language 
 
 phenomenon to call for much illustration. The 
 process may be seen in English ant ( = \amt from 
 A.S. aemete), ivomcn (i.e. \winiinen = wifmeti), gossip 
 ( = jgodsib, ' relative in God '), Lammas ( = A.S. 
 hlafmaesse), cupboard (pronounced cubbrd) ; in French 
 nourrir (L. nutrire), assez (^ = ad-satis) \ in Italian 
 freddo ( =frig{i)dtim), doima ( = dom{i)na). 
 
 (/) Development of glide-sounds. 
 
 We have already cited the word humble, the French 
 cendrc, gendre, tendre, the Greek avhp£<;, and the Latin 
 exemplum. We may add the Latin siim{f)siy em{p)tuvi 
 (from sumo, emo), the English em{p)ty ( = A.S. aemtig), 
 ga)n{b)le, crum{b)le, tim{U)er, the French sembler 
 ( = \semler from L. simulare), and the Italian scmbrare 
 ( = \semrare). 
 
 These classes of ' corruptions ' are here given as the chief, 
 but by no means as the only, forms of easing to which all 
 languages alike are prone. It would be beyond the scope 
 of the present work to proceed further into details or to 
 accumulate a larger mass of illustrations. 
 
 It has been a fashion to speak of this gradual contraction, 
 erosion, attrition, or ' weathering-away ' ( Verwitterung) of 
 words as the effect of ' laziness.' The term is ill-considered. 
 ' Laziness ' implies a moral reproach, which by no means 
 attaches to the gradual and unconscious abbreviation and 
 easing of language. ' Laziness ' implies an inclination to do 
 as little work as possible ; but a little observation will show 
 that speaking is not one of the forms of work from which 
 human beings ordinarily shrink. The nation which in modern 
 times has cut down its words beyond all others is that of the 
 French, but it can hardly be held that the French are ' lazy ' 
 speakers. Their vocal organs are conspicuously anything 
 but indolent. Moreover, persons who are fond of using the 
 longest possible words and of hearing themselves speak are
 
 General Phonetic Tendencies in Language 351 
 
 often among the least exact of articulators. They must be 
 charged with something else than laziness. 
 
 That there is such a thing as laziness of speech is, of 
 course, not to be denied. / diinno for / dont know, when 
 delivered with a slow drawl, can hardly be described other- 
 wise. On the other hand, the shortening of cabriolet to cab 
 and of omnibus to bus was partly the outcome of humour — 
 a sort of Cockney smartness, like that which uses phone for 
 telephone. Underlying such abbreviation is the assumption 
 that the one syllable is sufficient for any ready apprehension. 
 There is a sort of compliment to the respective intelligences 
 of speaker and hearer : ' As between you and me there is no 
 need to give the full and pedantic name ; pJione and bus are 
 much handier, and also show that we are on the most familiar 
 terms with the thing in question.' We need not commend 
 this practice, but it is clearly a misuse of terms to call it 
 laziness. 
 
 There is in language, as elsewhere, a survival of the 
 fittest. This does not mean necessarily an aesthetic fittest, 
 but only a phonetic and grammatical easiest. The aim of 
 language is to communicate thought. That style of language 
 which can be used to this end with the quickest effect and 
 with the least strain upon speaker and hearer is the style 
 which will develop itself everywhere, unless some artificial 
 check be put upon the natural evolution. A people of 
 nimble thought and vehement desire to impart it, such as 
 the ' Celtic ' French, will unconsciously economise the sounds 
 in a given word, not through any unreadiness to exert the 
 vocal apparatus in producing it, but through eagerness to 
 reach the next word and the full sense. The condensation 
 of the eleven syllables of eccistn{in) JtomineijiC) nietipsis- 
 sinm(in) into the three of cet Jiouune Diane is in reality a 
 most distinct progress in language. The main agent in the 
 change has been, not slovenliness, but rapidity. When 
 human beings sharpen their tools or take short cuts to avoid 
 roundabout roads we do not brand their ' economy of effort '
 
 352 Natural History of L anguage 
 
 as laziness. The man with the sharper tools can perform 
 the greater amount of work, and the man who takes short 
 cuts can accomplish the greater distance. The parallel is 
 not, of course, complete, since the sharpenings and short cuts 
 of language are less conscious, if conscious at all. Neverthe- 
 less the influences and the effects are the same. Language 
 works out its evolution in phonetics as well as grammar. 
 So long as the apprehension can fairly meet the speaker, 
 the shortest and simplest means are gradually selected as 
 the best. The stop to abbreviation comes when the word 
 threatens to be unintelligible. 
 
 Street cries, such as ' O' Clo',' have sometimes been 
 quoted as illustrations of the tendency to laziness in language. 
 But what is the case with these ? ' Sir,' said the crier of 
 ' O' Clo',' when reproved for his indolence, ' I can say " Old 
 Clothes " as well as you, but if you had to cry " Old 
 Clothes " a hundred times an hour, you also would say 
 " O' Clo'." ' And, if he had done so, it would have served 
 equally well. The hearer would have apprehended the 
 meaning, and nothing would have been gained for either 
 party by the very considerable labour undertaken by an 
 over-conscientious tongue. In point of fact ' O' Clo' ' in this 
 connection ceases to be language ; it is merely a symbolic 
 cry, and is treated as such by all concerned. To make a 
 point of clearly articulating ' Old Clothes ' at each delivery 
 would be conscientiousness of the kind displayed by the 
 actor of Othello who blacked himself all over. In this case 
 ' laziness ' is the wrong word. The proper term is ' con- 
 venience,' and the economy of effort commends itself to 
 common sense. 
 
 This is, indeed, an extreme case ; but, returning to 
 language proper, we have again to consider two parties — the 
 speaker and the listener. The speaker is naturally disposed 
 to deliver a word according to the sound-picture which exists 
 in his own mind and which he believes to exist in the mind 
 of the hearer. In ordinary circumstances he is thinking of 
 what he is saying, not examining the precision of his own
 
 General Phonetic Tendencies in Language 353 
 
 articulation. Meanwhile the listener also is primarily 
 attending to what is said, not to the precision of the speaker's 
 utterance. Only in cases of exceptional aberration will he 
 comment, either inwardly or openly, on the deviation from 
 the normal. But all those aberrations in the direction of 
 weakening and shortening which are eventually to win the 
 day begin faintly, and occur in those places in the sentence 
 where tney are most natural and where they will be least 
 marked. Sometimes, it is true, they arise from mere in- 
 dolence or a sullen slovenliness of the vocal operations, as 
 in the famous Scottish a! ae '00' for all one wool ; more 
 commonly, however, they are the effect of quickness and 
 fluency. The same aberrations will occur for the same 
 reasons with numbers of speakers. Gradually the sound - 
 picture in the mind comes to adapt itself to these cases, and 
 to the next generation of both speakers and listeners this 
 represents the normal. Developments of the same kind 
 follow each other, little by little, line upon line, and no one 
 is really conscious of the change which is taking place. 
 In ordinary conversation Old could not at once be con- 
 tracted to 0\ nor clothes to clo\ though each is the kind of 
 contraction to which the French language would have 
 brought such an expression if it had inherited it from Latin. 
 The speaker is not, like the street-crier, repeating the ex- 
 pression till the same muscles of the oral passage are wearied. 
 He is not, like the street-crier, using a merely symbolic 
 utterance explained by his situation. He has no temptation 
 to shorten these words more than any others ; while, if he 
 were to shorten all others in an equal degree, he would 
 become hopelessly unintelligible. Yet in rapid speech, 
 entirely intelligible to the listener, he may perhaps glide 
 rather easily over the d of old, and the th of clothes may, 
 without the knowledge of either party, be somewhat slurred. 
 The physiological reason for these glidings and slurrings on 
 his part will at the same time exist in most other persons, 
 and the ear will gradually become accustomed to something 
 like ol' close. Such has been, in fact, the vulgar pronuncia- 
 (B884) 2 A
 
 354 Natural History of Language 
 
 tion of England, until the artificial reactionary influence of 
 education was brought to bear. Yet ' laziness ' is an unfair 
 name for the cause, as it would be for that of our having 
 reduced the capital letters of print to the cursive shapes 
 which run so easily off the pen. These were not deliberately 
 invented to save trouble ; they have simply developed them- 
 selves by right of their fitness in the way of rapid efficacy. 
 
 Hitherto the practical necessity for clearness of under- 
 standing has been almost the only check upon the full play 
 of this convenience. In more recent and educated times the 
 existing words and uses of a language have come to be 
 regarded as objects calling in themselves for an alert pro- 
 tection, of which the motive is merely aesthetic or senti- 
 mental. This spirit, though it can never completely arrest 
 change in language, will necessarily act as a drag upon its 
 natural tendencies. It would be futile to prophesy further 
 as to its effects.
 
 CHAPTER XV 
 
 APPLICATION OF PHONOLOGY TO ETYMOLOGIES 
 
 If the foregoing chapters have been written to any purpose, 
 it will be sufficiently manifest that a word as it now exists 
 may have travelled a long and eventful journey from its 
 primitive shape. Enough has been said to show how both 
 change and loss may occur repeatedly, until a polysyllabic 
 word becomes the most curt of monosyllables, and may 
 contain, as actually pronounced (for the spelling is obviously 
 not the first question), not a single letter of the original form. 
 Though attention has been directed mainly to the change in 
 consonants and to entire loss of sounds, it is to be understood 
 that the vowels suffer complicated change, as well as loss, 
 equally with the consonants. The English word five (pro- 
 nounced faiv) is to be derived from an Indo-European 
 *penge. But faiv contains not one vowel or consonant found 
 in *penqe. The word civc is by derivation the same as the 
 Latin ovis. The original form was ^oivis {puis), and, of the 
 various sounds in this word, the English eive ( = iii) contains 
 none. There is no sign whatever in car that it is etymo- 
 logically related to the Greek ou9 ipus), nor in be to show its 
 connection with the Latin fii^i)., fuiturus), or with the phy- 
 in physics, nor is there any sign in Jiundircd) to relate it 
 at once to kicarov {hekaton). 
 
 Again, no one would naturally suspect that the de- 
 monstrable equivalent of come was to be found in the -vcn- 
 of invention and the ba- of basis. In these instances the 
 
 355
 
 356 Natural History of Language 
 
 original was *^m, and it is the Latin and Greek which have 
 made the wider departure. 
 
 Very frequent are the instances in which, as between two 
 languages derived from a common source, one letter may 
 perhaps still be common to both, whereas there is no other 
 likeness in the general body of the words. Thus F. larine, 
 Eng. tear (i.e. tir) agree only in the retention of r. The 
 original was *dakru, which took an Anglo-Saxon shape tear, 
 but a primitive Latin shape \dacni. To this a later stage 
 of Latin added a suffix -ma and subsequently converted (as 
 it sometimes does) d into /. From the classical Latin 
 lacriina is derived larvie by the regular French process of 
 syncopation. The English /c//r. Germ. •z7zVr show the same r 
 as the Greek TeTTape<i (tettares), but bear no further re- 
 semblance to that word. Quick (originally ' alive ') has one 
 letter in common with the Greek /3to9 (' life,' as in biology^). 
 Greek '^vvt] {gune), Eng. gzieen agree in « only. For Latin 
 sol, Greek 7;A,to9 {Jielios) a common basis lay in *sduel-, but / 
 is the only sound of the five which both languages have 
 preserved. That sol{ar) and liel{iac) should ultimately 
 converge in their history is not one of those facts which 
 could readily be guessed. / also is the only letter left 
 in common to Eng. hasel =^l^d.t. corulus, or ^rt//=Lat. 
 
 In other cases, while each may retam one sound or more 
 of the original, those which happen to be retained or changed 
 may be precisely the contrary in the two languages, and 
 there may be nothing left to indicate to eye or ear that we 
 have two derivatives from the same original. Thus the word 
 sit appears to have no affinity whatever with {cat)hed{ral). 
 Yet sit and -hed- are the respective English and Greek shapes 
 acquired by an original *sed. While English keeps s and 
 changes the rest, Greek keeps the rest and changes j to a 
 light h {ehpa = hed-ra, ' seat '). For the English goose and 
 Greek yr]^ {chen) the original was ^ghans. Of this, English 
 shows only the s, Greek only the w, while the other sounds 
 have been modified in each. Meanwhile ans{er) of Latin
 
 Application of Phonology to Etymologies 357 
 
 agrees with the Engh'sh in only one letter, and with the 
 Greek in only one other. 
 
 Though these are not some few extraordinary instances 
 culled for a purpose, and though it would be very easy to 
 add to their number, it is of course more common to find a 
 divergence somewhat less extreme, but striking nevertheless. 
 Thus the Greek SapT6<; (dartos, ' flayed ') is the cognate of 
 tor{n) ; \jaX, fracijiis) of brok{eti) ; sudipr) oi siveat ; sord{idiis) 
 of swart ; lingua of tongue. So far as the consonants are 
 concerned (in the roots) the words are originally identical ; 
 the vowels, where they are not absolutely identical, are 
 recognised gradations of each other. 
 
 The general influences which have been at work in thus 
 disguising words are those previously described, viz. omissions, 
 curtailments, assimilations, and other casings. But it has 
 already been insisted that each language possesses phonetic 
 laws of its own, and that, though it cannot be absolutely 
 demonstrated that these are invariable for that language, 
 they do operate with an extremely high degree of regularity. 
 In all the instances already given in the present chapter 
 (though they have by no means been selected with that 
 consideration chiefly in view) the philologist is in a position 
 either to state the phonetic law which each vowel and 
 consonant obeys, or else to render an account of such 
 departure as may occur therefrom. So far, indeed, does the 
 regularity of phonetic law extend, that in most cases the 
 exact forms in the several languages might be prophesied by 
 one who had studied the laws but had never met the 
 particular words. The analogical instinct — a mental inter- 
 ference to be treated of hereafter — has indeed modified the 
 Greek form ^771/ {c/icn), for which uninterrupted phonetic law 
 would have produced *%«? {c/ids). With our present 
 knowledge we should also have been doubtful whether to 
 expect *dacrunia or lacrunia, *dingua or lingua in Latin ; 
 but, given the form in the first column of the following list, 
 we should (dealing with the consonants only) at once
 
 358 Natural History of L anguage 
 
 postulate the form which follows in the second for the 
 language there concerned : — 
 
 I. 2. 
 
 Eng. s{i)t Gk. h{e)d 
 
 „ br{p)k- Lat. fr{a)g (but c before / by 
 
 assimilation) 
 Lat. {h){a)ns{er) Eng. g{oo)s{e) 
 
 I.-E. *su{e)rd- Lat. s{o)rd-; Eng. s-u>{a)rt 
 
 „ *suoid- „ Slid- ; Eng. .ra/^-a/ 
 
 „ -dengua " or t^!;^ [ E"^" ^^''^"^^^ 
 
 Eng. ^(0 „ f{u)-; Gk. (^(v)- 
 
 The detailed rules, together with those for the vowel 
 correspondences, do not belong to a treatise of this kind, but 
 to more specialised works. It is important here merely to 
 observe that there can be no sound etymological method 
 without a knowledge of the phonetic laws of the languages 
 concerned and a strict working assumption that their 
 operation is regular. The theory of an irregularity is only 
 to be attempted when it can be reasonably accounted for, 
 on the ground that analogy, tolerably self-evident, has 
 diverted the form in question, or else that the word is 
 borrowed from another dialect or language, to whose phonetic 
 laws it can be shown to conform, or from which it might 
 naturally be taken with a characteristic mispronunciation. 
 
 Where it is urged that the pertinent phonetic laws may 
 not yet be fully known and that the suggested etymology is 
 provisionally good, it must at least be demonstrated that any 
 supposed change (or series of changes) which has brought 
 the word to its present shape is physiologically probable — 
 that is to say, is in keeping with what we know of the 
 processes of sound-change. That p may become in certain 
 cases b, f, in, or z^ is a known fact which may be utilised, 
 but an interchange of p and r or I or s is out of the 
 question. 
 
 " A sound etymology," it has been epigrammatically said, 
 " has nothing to do with sound." It has, however, everything 
 to do with the sounds. That one word does not now sound
 
 Application of Phonology to Etymologies 359 
 
 in the least like another (as in the instance of goose and 
 miser, five and cinq) is, indeed, no disproof of their common 
 etymology. Neither, on the contrary, is it any proof of a 
 common etymology when two words of similar meaning 
 sound alike, as in the case of English call and Greek koK-, 
 whole and oX- (Jiol-), Greek Qeo^ {theos) and Lat. deus. In 
 this sense it is correct to say that sound-etymology has 
 nothing to do with the sound. But, while this is true, the 
 identity of five and cijiq is only to be accepted in the light 
 of the discoverable (or at least scientifically assumable) past 
 history of their several sounds ; and the gradual divergence 
 of the original ^penqe into these two strikingly unlike forms 
 must be traced convincingly through a series of steps in 
 sound-change, each natural and intelligible in itself, first on 
 general principles, and second according to the phonetic 
 laws of the speeches in question. If a definite phonetic law 
 for a particular step is not yet forthcoming, the reasonable- 
 ness of the assumed shifting must be proved by other 
 examples of the same phenomenon or of one closely 
 analogous. Thus, to go backwards in W'c^q, five {faiii) is a 
 modification of f'lf, flf of fimf, fimf of finf, finf of "^penqe. 
 Beginning with *penqe, the change of the first vowel from e 
 to Teutonic i, and of the consonants from / to / and q to /, 
 is in accordance with law. The labial / further assimilates 
 the dental n and makes it labial jh (as in Greek Tre/xTrro?, 
 pemptos, for \pcnptos). The loss of the final syllable e is the 
 common effect of the Teutonic stress-accent on the previous 
 syllable. The subsequent loss of the nasal in in Anglo- 
 Saxon and the consequent lengthening of the vowel is 
 illustrated by e.g. gos (goose) = igons, toth {tooth) = \tontJi. 
 The further modification of/ to v {fiv (or flf) is common 
 (as in of pronounced ov). If we deal in the same way with 
 cinq, our steps backward are cinque, quinque, \pinque, \penque. 
 By a phonetic law of Latin c before nq, as before some other 
 consonantal combinations, becomes i. The step ^piyique 
 having occurred, or while it is occurring, the labial element 
 in qu attracts (or assimilates) the pronunciation of the /, and
 
 360 . Natural History of Language 
 
 quinque is substituted. (Tiie same thing takes place in 
 ^peqzio, ' cook,' which passes to \quequo and thence to coquo.) 
 Possibly also the qu of the previous numeral quattuor 
 (' four ') influenced the change of form. In the transition 
 from Latin to the Romance tongues the initial qii here 
 becomes c {cinque)^ and this in French, before i, is pronounced 
 s. The final syllable disappears in consequence of the stress- 
 accent which formerly existed in French as in English. 
 
 This example is much less simple than many which 
 might have been chosen. Take, for instance, the word loud 
 as compared with the Sanskrit srutds. Though these 
 possess no sound in common, they are (except for the 
 particular grade of the root vowel) identical words according 
 to the perfect operation of the phonetic laws of their re- 
 spective languages. Through loitd, Jdfid, \hlud{a), ^hlnthd{s) 
 we arrive at an original *klut6s. In Teutonic the changes 
 to a; the sound -shifting of the stops ('Grimm's Law') 
 produces \hluthds ; the position of the accent (' Verner's 
 Law ') changes th to d ; the adoption of a strong stress- 
 accent on the first syllable causes the loss of the second. 
 From Anglo-Saxon hind the initial h before a liquid 
 disappears, as in {h)razv, {Ji)riddle ( = ' sieve ') ; the Anglo- 
 Saxon n regularly becomes in modern English on, as in cozv 
 ( = cii)^ mouse ( = mus). Meanwhile the same original 
 *klut6s (with a weaker root vowel) will in Sanskrit become 
 letter by letter srutds. Every change in these words could 
 be foretold by the comparative philologist. Ask him what 
 would naturally become in Sanskrit of a word *klnt6s, and 
 he will reply srutds ; for ^klutSs he would predict the exact 
 English form loud. To express it otherwise, in Anglo-Saxon 
 k will have become h. In will remain //?, t before the accent 
 will have become d, the accent will have shifted to the first 
 syllable, and the second syllable will be lost. Hence Jdfid 
 must be expected. For modern English Ji in such a position 
 will have vanished, I'l will have become ou, and loud is 
 practically the inevitable shape of the word.
 
 Application of Phonology to Etymologies 361 
 
 It is somewhat misleading to say, as is often said, that 
 sound etymology must not assume any intermediate step of 
 which there is no historical proof Doubtless an etymology 
 is made the more certain by a complete series of discoverable 
 connecting-links. The relation of loiid to srutds, as to the 
 Greek kXvto^ {klutos) and the Latin {in)clHtus, is undeniably 
 made surer by the existence of the Anglo-Saxon hind. If 
 all record of an Anglo-Saxon hind had been lost, the 
 theoretical history of loud might have been somewhat 
 less convincing, especially if it were pleaded that the mean- 
 ing of srutds is not ' loud,' but ' heard,' while of kKvto'^ and 
 {in)clutus it is ' famous.' The intermediate step Idud might 
 indeed have been postulated with the greatest plausibility as 
 strictly in keeping with phonetic laws, but it would after all 
 have been regarded as but a highly probable assumption. 
 But, with kind actually before us, all reasonable doubt is 
 removed, and, in respect of the meaning, loud signifies that 
 which is distinctly ' heard,' while ' famous ' is that which is 
 much ' heard of.' It is manifest here that, even if we had 
 lost hh'id and were compelled to assume it, the etymology 
 would still in fact be correct ; it would, however, hardly be 
 regarded as proven. 
 
 While acknowledging this, it is nevertheless necessary to 
 amend the statement that no purely theoretical intermediate 
 step must be assumed, and we must rather say that any 
 assumed step must be in itself of a nature which might have 
 been anticipated, and that no considerable number of such steps 
 must be assumed together. There must be no jumping of 
 wide chasms. We should, for example, not be justified in 
 deriving mcme from nietipsissimus if we did not possess either 
 certain steps in French itself, e.g. mesme, meesme, medesme or 
 the cognate Provencal metessme and Italian medesimo. But 
 he would be a timorous etymologist who, having before him 
 inane, metessme, and medesimo, would shrink from advancing a 
 hypothetical medesme as an intermediate step in French. 
 His justification would be that a medesme would necessarily 
 become mane, and that the existence of the obviously
 
 362 Natural History of Lang2iage 
 
 cognate Provencal and Italian words fully warrant the as- 
 sumption of the supposed Old French form closely resem- 
 bling them. In other words, medcsnie is virtually proved 
 by metessme and inedesimo, and, if medesme did exist, then its 
 modern shape would be mane. Breal rightly says " scientific 
 etymology does not consist in a vague statement of the 
 affinity which may exist between two words ; it must track 
 out, letter by letter, the history of the formation of a word, 
 and show all the intermediate stages through which it has 
 passed." But this must not be taken to mean that none of 
 the stages whatever shall be hypothetical, and that we must 
 have before us documentary evidence of each and every 
 successive phenomenon in the history of the word in question. 
 It means that a complete series of the steps, either docu- 
 mentary or assumed, between the termini of the history 
 must be set out, and that each step must commend itself to 
 the trained investigator as entirely probable. When Menage 
 derived the French rat from Latin vms (' mouse ') by the 
 facile process of forming from the stem inur- the participle 
 jmftrdtus and cutting this down to \rdtiis and thence to rat^ 
 he managed to violate every canon of etymology. Without 
 treating this amusing effort too seriously, we may employ it 
 in illustration of method. jRdtus could not become rat, 
 ^mfirdtiis could not become jrdtus, and \mnrdtus itself is 
 an absurd formation which existed only in an untaught 
 imagination. In French \mnrdtus would have become 
 \nnirc. Had there been such a thing as a French \iimri 
 Menage might have postulated a late Latin jDifirdtus, but 
 there was no such word in French and no equivalent of it in 
 any cognate language. Guesswork, pure and simple, has no 
 place in etymology. 
 
 It is somewhat difficult to define the exact limits of the 
 use of hypothesis. We may perhaps state them thus. 
 There must be no assumption of a hypothetical intermediate 
 form ivJiich is not actually pointed to by some cognate language, 
 or which is not at least precisely the form indicated (by the 
 phonetic lazvs and formative practices of the language') as a
 
 Application of Phonology to Etymologies 363 
 
 practically inevitable stage if the suggested derivatioii of the 
 word is correct. In the case of \imlratus we have an 
 imaginary form not only pointed to by no cognate speech, 
 but also so far from being indicated by phonetic laws or 
 formative practice that it is actually in conflict with them. 
 The French ecurie means * stable ' and ecuyer ' squire,' and a 
 connection with the Latin eqiius ' horse ' would be one of the 
 most natural of empirical guesses. Unfortunately the Lingua 
 Romana used caballu and not eqiiiis for ' horse,' and, though 
 this objection might be overcome, the intermediate forma- 
 tions linking ecurie or ecuyer to equus are absent from either 
 French or its sister speeches. If equus had formed any 
 derivatives in the sense of ' place for the horse,' ' man to 
 tend the horse,' those derivatives could never have been of 
 such a shape that ecurie and ecuyer could develop from them. 
 In reality ecurie is derived from a Teutonic source, represented 
 in Low Latin as scuria, while ecuyer is from the late Latin 
 scutarius. If it be asked how far hypothesis of intermediate 
 forms is required even by these derivations, the answer is 
 that the older French escurie is actually extant, and that this 
 is on the one hand the necessary shape given to scuria in 
 old French, and, on the other, can only be represented by 
 ecurie in the modern tongue. There is no assumption in 
 the case. But in deriving ecuyer from scjitarius there must 
 come in the hypothesis of a non -extant early French 
 \escudier. By what is this warranted ? First by the 
 existence of a Provencal escudier and Italian scudiej-e, second 
 by the appearance in older French of the form escuyer, and 
 third by the fact that scutarius would regularly or almost 
 inevitably produce ^escudier, while this would as regularly 
 proceed to ecuyer. If we suppose a word scutarius to be 
 known, but no Provencal escudier, Italian scudiere or old 
 PVench escuyer, it would be a highly plausible derivation, 
 based purely on a consideration of the meaning and of 
 phonetic laws, to suggest the series ecuyer, \ escuyer, \ escudier, 
 scutarius^ with the intermediate steps thus hypothetically 
 assumed. Each step would be phonologically accurate, but
 
 364 Natural History of Language 
 
 the etymology, though extremely probable, would lack 
 something of absolute convincingness. 
 
 Utterly unconvincing must be the etymologies in which 
 a mere general and superficial resemblance of two words 
 and some approximation in sense are treated as sufficient, 
 while the connecting links, if suggested at all, are without 
 evidence from cognate tongues, from phonetic laws, or from 
 general probabilities as to their formation. Of this kind 
 were the usual efforts of the old etymologists, from the time 
 of Greek philosophers like Plato or Roman grammarians 
 like Varro down to a very recent date. A Greek, having 
 no conception of the kinship of tongues, naturally derived 
 his words from Greek sources. Possessing no knowledge 
 of phonetic laws and little regard for even the commonest 
 phenomena of sound-change, he was attracted by any 
 slight correspondence of form which seemed to go with an 
 imaginable connection of meaning. Plato's Cratylus is full 
 of amazing assumptions, sometimes amusing and often 
 ironical. Aristotle, with a more cautious temperament, 
 nevertheless finds in yaip&iv {c/iairein), ' rejoice,' the natural 
 source of fiaKcipio^ (yinakar'ios), ' happy,' simply because of a 
 measure of likeness in the syllables icap- and %a(/c)- and 
 an association of sense. Again, he derives hiKaio<i {dikaios), 
 ' just,' from S/%a {dicha), ' apart,' assuming that justice is 
 the act of due partition. In reality S/-%a is a formation 
 from Sl-, which is akin to St<?, ' twice,' and to the tzvi- of 
 be-twi-xt. The primitive source is "^dui-. But SiKaio<i is 
 the adjective of BUr] (dike),' justice,' its original sense being 
 that of a proceeding ' shown ' or taught. The primitive 
 source is *c///i'-, the root from which is derived our own 
 word Uac/i. 
 
 Although the discovery of etymologies became a 
 recognised part of Alexandrian linguistic study, its pro- 
 cedure was still hopelessly unscientific. In legend the ark 
 of Deucalion rested on Mount Parnassus, Now an ark 
 is Xdpva^ {larnax) ; hence, it was said, came the name of 
 the mountain. That the sound / cannot pass into the sound
 
 Application of Phonology to Etymologies 365 
 
 / was a truth unknown and therefore offering no difficulty. 
 It is true that we once meet with such a remark as that " i 
 is never turned into a" but this rational proviso is 
 exceptional for the times. 
 
 After the Revival of Learning in Europe the case was no 
 better. On the one hand, theological study revived the 
 notion of deriving modern European words from Hebrew. 
 This practice had already prevailed in the days of Jerome 
 or Origen, when absorption in the new religion and the 
 Biblical view of the history of mankind invested Hebrew 
 with a false character as the primitive language. It was 
 renewed under the same impulse when the Reformation 
 had turned theological reading back to the Hebrew Old 
 Testament. The derivations of the seventeenth century 
 attempted under this impression are amazing in their 
 ingenious absurdity. On the other hand, the revived study 
 of the classics brought into prominence the affinities of 
 Latin and Greek, and etymological guesswork ran its course 
 in the connecting of these languages with one another and 
 of both with Hebrew or modern tongues. Curtius quotes 
 Scaliger's equation of Latin pidcher, ' beautiful,' with Greek 
 TToXvx^i'P {poliicheir), which happens, however, to mean ' many- 
 handed.' Voss again identified the Latin similis, ' like,' with 
 the Greek ixLixrj\6<i (///Jw^/c'j-), ' imitative,' and ev/^, ' pluck,' 
 with the Greek riWoy {tilld) of the same meaning. It is 
 unfortunately probable that even at the present day the 
 impossibility of these equations will not be realised by those 
 who have made no study, first of general phonology and 
 then of Greek and Latin phonology in particular. But one 
 who has made the first steps in such study will know that if 
 the true equivalent of TiKXay {tilld) appeared in Latin it would 
 be \tilio or possibly \tillo, and that / and v can in no case 
 have anything to do with each other. On the other hand, if the 
 true equivalent of vello appeared in Greek it would be ^XXw 
 ieild) or possibly eXKw {elld). In the other instance, fxi^i^Xo^ 
 would in Latin be -\jiiii/iclns, and in no case could the jt answer 
 to the m or the vowels to each other. It is little wonder that
 
 366 Natural History of Language 
 
 Voltaire described the etymology of his day as a science 
 in which " the vowels count for nothing and the consonants 
 for very little." In modern etymology every vowel and 
 consonant counts as an element of separate importance. 
 
 Sound etymology, then, is based on a knowledge of 
 phonetic principles, and in especial of the phonetic laws of 
 the languages concerned. (A connected chain of meaning, 
 it will be understood, must be ascertainable or rendered 
 probable by similar examples ; but that point is here 
 assumed, and it is comparatively little that etymologists 
 have sinned, like Scaliger with his pulchcr, against this 
 obvious canon.) Intermediate steps, when hypothetical and 
 assumed, must justify their assumption, if not by positive 
 evidence from cognate languages, by their convincingness 
 from both a phonetic and a formative point of view. Sound 
 etymology, therefore, has everything to do with sounds. But 
 when it is remembered how various are the phonetic laws 
 of different tongues in respect of both vowels and 
 consonants, and how variously also they have abbreviated 
 their words at beginning, middle, and end by reason of 
 differences in the nature and position of their accent, it must 
 be evident that a word inherited by one language from its 
 primitive ancestor will in all probability become extremely 
 unlike the same word as inherited and changed by another 
 branch. When a word of similar meaning in native English 
 looks strikingly like a word of native Greek or Sanskrit, it is 
 well to look upon the etymological identification with doubt. 
 If the one language can be shown to have borrowed the 
 word from the other, the case is of course simple, as \\\\q.x\ pain 
 and age are borrowed from French peine, age. But when a 
 search into the history of the respective languages shows 
 that the words are to be supposed native to each, a very 
 close similarity is generally a note of warning. The French 
 feu and German Feuer both mean ' fire.' Superficially they 
 would appear to be the same word. But when investigation 
 shows that neither language has borrowed from the other.
 
 Application of Phonology to Etymologies 367 
 
 the etymologist will take their very agreement m feu- as in 
 itself a proof that the words were originally very different, 
 and that they have only reached an accidental similarity of 
 shape as well as of meaning through the operation of different 
 phonetic laws. That is to say, a primitive word which 
 could produce feu in French could not produce Feu- in 
 German ; but one primitive word might result in Feu- in 
 German and a different primitive word might result m fcic 
 in French, while at the same time the meanings might 
 chance also to meet. In reality the French /^/^ represents 
 the Latin focus, with the meaning of the ' hearth ' or ' fire- 
 place,' and thence the ' fire,' whereas for the German Feucr 
 we must according to Grimm's Law assume a primitive 
 initial */, as found in the Greek irvp {pfir). The develop- 
 ment of the Teutonic word is \piMr, jfnr, feuer ; of the 
 French word it is focus, jfuocu' {It. ftwco), feu. We cannot 
 at present get back to the Indo-European primitive for the 
 Latin focus, but it certainly did not begin with *///- nor 
 contain that syllable. 
 
 It is not indeed to be flatly asserted that there can be 
 in no case such a thing as identity of form between two 
 languages in regard to the same inherited word, after all 
 their respective phonetic laws have operated upon it. The 
 English star happens to correspond to the Sanskrit star-, 
 inasmuch as an original '^ster- has eventually come to take 
 that shape in each tongue. Yet in Anglo-Saxon star was 
 steorra. But such accidental coincidences are of the greatest 
 rarity, and for working purposes the rule holds good that 
 identity or very close similarity of pronunciation is 
 commonly either a sign of borrowing or else indicates 
 (despite all apparent connection of meaning) that the words 
 in question are of different origin. The English wliole (i.e. 
 hot) has the meaning of the Greek 0X09 {Jiolos). But whole 
 is in Anglo-Saxon Jidl and o\o<i represents an earlier *sobios. 
 At these stages the resemblance vanishes. 
 
 Perhaps enough has now been said to guard against hasty
 
 368 Natural History of Language 
 
 or unscientific etymologising. It seems advisable, however, in 
 this connection to follow the example of Max Miiller in 
 summarising four chief phenomena as affecting etymology on 
 its phonetic side. They are these : — 
 
 (i) Under the operation of phonetic laws the same 
 original word may take different forms in different 
 languages. 
 
 As has been shown above, it is truer to say, not that it 
 may, but that it nearly always will and must, when the 
 languages have followed a separate existence for any 
 considerable period. The Latin fociis becomes French /^w, 
 Italian fiwco, Spanish fuego ; Jilius becomes French fits, 
 Italian j'^cr/Zc, Spanish hijo ; and such differentiation must in 
 most cases almost necessarily occur, so long as phonetic 
 operations alone are involved. Even where two languages 
 appear to possess the same form, it is more commonly the case 
 that the identity is only an identity to the eye, not to the ear. 
 The words look the same without being the same. Thus it is 
 incorrect to speak of the German Hand as identical with the 
 English hand. Neither the a nor the d are pronounced in the 
 same way. Phonetically written the English word is Jicend, the 
 German is Jia7it. If we go beyond languages of the same 
 branch {i.e. beyond a comparison of a Romance tongue with 
 a Romance tongue, or of a Teutonic with a Teutonic), and 
 compare a speech of one branch with a speech of another, 
 we must expect to find a still wider divergence, and we may 
 even arrive at such complete unlikeness as that already 
 exemplified in goose, anser, '^I'jv. 
 
 (2) The same word may possess different forms in one 
 and the same language. 
 
 This, it is true, could not occur through the simple 
 operation of the phonetic laws of the one dialect. If the 
 language in question were developed consistently from one 
 dialect without interference from any other, there could 
 arise no diversity of form unless some mental influence 
 stepped in, (Mental operation is seen, for example, in the 
 second part of shamefaced as compared with steadfast.
 
 Application of Phonology to Etymologies 369 
 
 The termination should be -fast in both instances, but 
 popular etymology connected the former with face) But 
 few, if any, languages can show such purity of descent and 
 simplicity of operation. English, though based on the 
 dialect of the East or South-East Midlands, has been com- 
 pelled to reckon with the dialects of the South and North, 
 particularly the latter, which again has been affected by 
 Scandinavian tendencies. Moreover, it has borrowed freely, 
 especially from Latin, either directly or through the French. 
 It therefore possesses many doublet-forms, although it may 
 be remarked in passing that in all such cases an appreciable 
 distinction or specialising has been developed between the 
 use of the two words in each pair. Thus ivise, bench, the 
 native English forms, and gnise, bank, the same words, but 
 adopted from French in their French shape, are by no means 
 identical in use. Similarly gard{en) and yard are doublets 
 within the dialects of England itself. The same is the case 
 with draw and df'ag, scale and shell, hale and whole, shirt 
 and skirt, clench and clijich. No is Anglo-Saxon, nay is 
 Scandinavian. Again, count is the French compter, from 
 Latin coniputare, while compute is the form taken directly 
 at a later date from Latin by English. 
 
 French, again, possesses words inherited from Latin 
 through the lingua romana and also words borrowed or 
 adapted from Latin by the learned in more recent times. 
 The same Latin word may therefore occur in two forms, 
 the one influenced by phonetic law through all the history 
 of French as a language, the other influenced only by such 
 phonetic laws as have had full play since the time at which 
 it was borrowed. Thus fabricare as inherited becomes 
 forger, as borrowed or adapted it is fabriquer. Similarly 
 serment and sacrement are an older and a younger form of 
 sacravientum ; sembler and simuler of simulare. Moreover, 
 French, like English, had its dialects. The ca- of Latin 
 became cha- in the French proper of the lie de France, but 
 remained ca- in Picard. If modern French possesses both 
 champ and camp (though in distinct senses), the variety is due 
 
 (B884) 2 13
 
 370 Natural History of Language 
 
 to the taking up of a form from the Picard dialect. The 
 Central French gives us leal, the Northern French loyal. 
 There are not, in fact, two forms in a single dialect. When 
 on and hoinme are declared to be two representatives of the 
 same word in the same dialect the remark is incorrect. 
 On is the nominative ( = Jiovio), hoimne is the accusative ( = 
 hominein). The //, of course, is unheard, and its presence or 
 absence is merely a matter of the writing, not of the sound. 
 
 (3) Words different in origin may take the same shape 
 in different languages. 
 
 If we press these words and their intention strictly, this 
 phenomenon is one which will trouble the etymologist but 
 little. When the different words which have been 
 ' weathered down ' to the same form have altogether uncon- 
 nected meanings, there is no danger of derivation being 
 tempted from the scientific path. The German viel, ' many,' 
 and the French fil, ' wire,' may be said to possess the same 
 form as the English y^W, since the pronunciation of all alike 
 is /il. But there is no imaginable association of sense, and 
 no one is likely to attempt to relate such words. On the 
 other hand, it is with the greatest rarity, if ever, that two 
 words of different primitive form come to develop both 
 absolutely identical shapes and also approximately the 
 same meaning. ' Call ' and Greek koX- (^al-), though the 
 same in sense, are in pronunciation only approximate, not 
 identical. The same is true of the Polynesian wala, ' eye,' 
 and the modern Greek fidrL ijiiati). Accepting, however, 
 such approximations as instances of agreement in the 
 ' same ' form, it is only to be wondered at that they are so 
 few. Amid the possible permutations and combinations of 
 sound which make up the various words of language, and 
 amid all the hundreds of languages which exist, it would not 
 be surprising if purely accidental coincidences were some- 
 what numerous. There is little similarity between an 
 earlier Greek o/A/iartoi/ {pmmdtion) and the Polynesian mata, 
 and perhaps if we knew the fuller history of the Polynesian 
 word the original resemblance would prove smaller still.
 
 Application of Phonology to Etymologies 371 
 
 ofiixdrLov {ovundtion) itself is a late diminutive of o/i/xa 
 {oinma)^ and, when the word is brought back to this form, 
 the similarity practically vanishes. It would scarcely have 
 been necessary to deal with this topic but for the fact that 
 even one solitary case of such chance resemblance has some- 
 times been enough to evoke the most baseless speculations 
 on the part of those whose imagination is stronger than 
 their judgment. 
 
 (4) Words of different origin take the same form in one 
 and the same language. 
 
 Thus in English bay is used of a tree, a colour, an inlet 
 of the sea, and the deep bark of a dog. The origins are 
 respectively the vulgar Latin bacca, badius, baia, abbaubare. 
 In bear and bare we may disregard the spelling and recognise 
 one form with three distinct senses and derivations. Sound, 
 as ' strong,' is from Anglo-Saxon sund, as a noise it is the 
 French son (Lat. sonus), as an act of testing depths it is the 
 French sonder (Lat. siibnndare). Bound is a participle of bind, 
 or it is the Anglo-Saxon {ge)bnn, ' ready.' Seal is the Latin 
 sigillum or the Teutonic seol, according as it denotes the 
 instrument or the animal. Dole, if ' grief,' is the French 
 deuil ; if a ' portion,' it is Teutonic, related to deal and the 
 German Teil. In French louer, ' to let,' represents locare, as 
 ' to praise ' it represents laudare. It is vain to multiply 
 examples, which are extremely numerous. The lesson to 
 be drawn from them is that in etymologies it is futile to 
 attempt ingenious tours de force and to concoct for words 
 like these some sort of association of ideas with a view to 
 identity of derivation. Had the various origins of sound 
 been lost, it is quite conceivable, in the light of experience, 
 that the empirical etymologist would have invented some 
 device for connecting the sense of ' strong ' with the sense 
 of ' noise ' or of ' testing depth.' 
 
 A first principle of sound etymology is to trace the 
 actually discoverable history of the word and to indulge in 
 no speculation until that material is exhausted. Nothing
 
 372 Natural History of Language 
 
 must be taken for granted. Belfry has nothing to do with 
 bell, but is old French berfray. In cordwainer the word 
 cord plays no part, but Cordova does. Restive and restless 
 are not from the same main root. 
 
 A second principle is to assume rigid regularity of 
 phonetic laws unless good motive is ascertainable to counter- 
 act or traverse them. 
 
 A third is to postulate lost intermediate steps with 
 great caution and by a process of strictly scientific induction. 
 
 The part played in etymologies by laws — if there can 
 be said to be laws — of meaning-change belongs to semasiology. 
 Here we are concerned simply with the phonetic side of 
 derivation.
 
 CHAPTER XVI 
 
 CHANGES OF MEANING 
 
 Along with the changes in the pronunciation of a language 
 there proceed changes in its vocabulary. Many of its words 
 gradually drop out of use and are replaced by others ; 
 many new words are either borrowed from foreign tongues 
 or coined from the materials supplied by the language itself. 
 Of these losses and additions something will be said in 
 the next chapter. Meanwhile, in the case of those words 
 which are retained in use, there is a constant tendency to 
 widen, narrow, or shift their meanings. Even in the almost 
 unimaginable event of two cognate languages retaining 
 substantially the same sounds in such words as they keep, 
 they would not only come to differ in the extent of their 
 vocabularies, but also the same words might in the one 
 language eventually possess senses unknown to the other, 
 until, for that reason alone, their respective speakers would 
 be perpetually at cross-purposes.^ 
 
 ' Laws ' of meaning-change are not yet discovered and 
 are probably undiscoverable. Interesting observations can 
 be made of certain tendencies and phenomena, but beyond 
 this we can hardly go. In the case of phonetic change 
 the mind plays comparatively little part ; in changes of 
 meaning it is, of course, the one factor. We can perceive 
 the physiological operations which will naturally transform 
 one sound into another ; we cannot hope to reduce to 
 
 ^ Certain English words as used in America might exemplify this statement. 
 Bug and pie are instances. 
 
 373
 
 374 
 
 Natural History of Language 
 
 rules the complex psychological operations which transform 
 one meaning into another. We can say that some words 
 widen their meaning, that some narrow it, and that some 
 shift it, and we may often be able to trace the association 
 of ideas which brought these occurrences to pass. But 
 we cannot get to fundamental principles which determine 
 that a certain class of words shall necessarily widen 
 rather than narrow their meaning or vice versa. There 
 is no ascertainable principle which makes the English 
 7nan retain its wider sense, while wife, which originally 
 meant ' woman,' has become narrowed. If any such 
 principle be imagined for the moment, it will be found to 
 break down when applied to German, in which IVeib (the 
 same word as luife) means ' woman ' and not ' wife,' while 
 Mann often has the narrower sense (which has disappeared 
 from standard English) of ' husband.' In provincial English, 
 it is true, one may still hear a woman speak of her ' man,' 
 and the older and wider sense of wife (as ' woman ') per- 
 sists in fis/nvife, ho?csewife, but the following scheme 
 represents the development of the meanings in standard 
 English and German respectively : — 
 
 Old 
 English 
 
 Old 
 German 
 
 Modern 
 English 
 
 Modern 
 German 
 
 inan = man (wider) 
 and husband (nar- 
 rower) 
 
 wife = woman (wider) 
 and wife (occasional 
 narrowing) 
 
 ^4/a««( senses as 
 in O. English) 
 
 Weib (as in 0. 
 English) 
 
 7nan (wide sense 
 only) 
 
 wife (narrower 
 sense only) 
 
 Mann (both senses 
 of Old Germ, re- 
 tained) 
 
 Weib (wide sense 
 only). 
 
 It is doubtless sufficiently easy, after the event, to follow 
 the process of narrowing in the case of wife. But why did 
 not the same process occur with the German Weib ? The 
 Latin fcmina (' woman,' but not ' wife ') in its French form 
 fcuime means both ' woman ' and ' wife ' ; meanwhile the 
 Latin inulier ('woman') in its Italian form moglie means
 
 Changes of Meaning 375 
 
 ' wife ' only. Why should there be this divergence of 
 history ? The English deer is the German Tier (' animal '). 
 A sufficiently intelligible reason may be found, after the 
 event, why deer became restricted in England to the cervine 
 tribe of animals ; but why did not Tier follow the same 
 course ? Or why did not deer retain as broad a sense as 
 the German ? This is a question which could only be 
 answered from the whole history of the association of 
 each people with the fauna of its surroundings, and from 
 mental processes which are often started by chance or 
 transient circumstance, 
 
 A primitive language, such as the Indo-European 
 in its hypothetical homogeneous state, possesses a certain 
 stock of words. Some of these are names of definite 
 single objects, such as ' sun,' ' moon,' ' sky,' or of definite 
 classes of material objects, e.g. ' tree,' ' ox,' ' star,' ' tooth.' 
 Others express relationships, eg. ' father,' ' brother,* ' son.' 
 Some are pronouns, e.g: ' I,' ' we,' ' you ' ; some numerals, 
 ' two,' ' three,' ' ten,' etc. There are definite physical 
 actions, e.g. ' sit,' ' stand,' ' speak.' Within this region of 
 names for simple and definite notions it might seem that 
 there is no room for change of meaning, that the speakers 
 of the primitive tongue itself could not fail to attach 
 to one of these words exactly the same sense, and that 
 the derived languages of a later date, if they retained the 
 words, could hardly be subject to any unconscious prompt- 
 ing to shift their meaning. Yet a little consideration will 
 reveal the contrary. In the first place, in all languages, 
 even the most uncultivated, the metaphorical or figurative 
 use of words is one of the most natural efforts towards 
 expression. The word for ' tooth ' may speedily lend 
 itself to the wider sense of any ' peg,' the word ' sit ' to 
 any settling down. In the second place, experience will 
 show that if the name of a class of objects includes more 
 than one species, the name of the class may easily become 
 more narrowly applied to the species which is most
 
 376 Natural History of L anguage 
 
 prominent. Thus the word for ' ox ' may come to be 
 appHed specially to the female, or cow. This has actually 
 occurred in the English cow and German Kuh, as compared 
 with the wider sense of the cognate bos in Latin and /SoO? 
 {bous) in Greek, these words being used, like the common 
 original ^gous, of both sexes. The same narrowing has 
 befallen the English eive, which is the same word as the 
 Latin ovis and a primitive *oins, used of any sheep. 
 
 Again, as the once homogeneous and homogeneously 
 situated people is dispersed, the various branches, in their 
 later homes and amid new natural surroundings and social 
 conditions, may be led to make a somewhat different 
 application even of the common words above described. 
 If, for example, a special kind of tree prevails in a particular 
 region, it is conceivable that the word for ' tree ' might 
 come to be applied in that region, first particularly and 
 then solely, to such kind. At first the special sort of 
 tree — the commonest tree, or the tree — would suggest 
 itself to the mind when ' tree ' was mentioned. Thence 
 ' tree ' would gradually be accepted and employed commonly 
 in that narrower sense. This was apparently the case with 
 the Greek use of Ipv^ {drus\ ' oak,' a word originally identical 
 with our tree. Sometimes the name properly applied to 
 one object may be transferred to another closely resembling it. 
 *Bhdgos in primitive Indo-European meant beech, and that 
 meaning is retained in English and in the Latin fagus. But 
 in Southern Greece the beech is unknown, and the word 
 4)r}^o<; {phcgos) is transferred to an oak with edible acorn. 
 
 If, turning from natural environment, we look to social 
 conditions and their effects, we find meaning-change occur- 
 ring in cases where it might have been least expected. Our 
 word brother, like the Latin /r^/«?r, unless used figuratively, 
 implies actual community of parentage on one side at least. 
 But in Greek ^pdryjp (phrdter) is a ' clan-brother,' and 
 possesses only a special politico-religious application. To 
 the present discussion it would not matter which sense was 
 the more primitive ; there has been change in one language
 
 Changes of Meaning 2>77 
 
 or the other. In point of fact it is tolerably certain that it is 
 Greek which has deflected the meaning. Something similar 
 has occurred in the history of the Romance tongues. The 
 Latin frdter is, like brother, a term of definite blood-relation- 
 ship. That sense is retained in the French frere. But in 
 Spanish frate has come to signify a religious brother or 
 friar, while the blood -brother is known as hennano (from 
 the Latin germdnus). Even pronouns are not secured 
 against sense -shifting. The plural you is now used as 
 singular in English and French ; in German the plural 
 third person Sie (' they ') is used as the polite singular 
 second person ('you'); in Italian the third feminine ella 
 (' she ') is the formal pronoun for the second masculine 
 (' you '). A satisfactory account can doubtless be given of 
 the mental processes which have caused a substitution of 
 these plural and indirect modes of address for the singular 
 and direct. Nevertheless the substitution could not have 
 been foreseen as the necessary outcome of a clear principle, 
 nor can any fundamental explanation be given of the 
 reasons why French, German, and Italian have not agreed in 
 the forms of substitution at which they have severally 
 arrived. 
 
 If changes like these occur in words of so common 
 and definite a type, it is easy to understand how wide 
 and complex may be the shiftings in words more abstract. 
 Except in naming the commonest material objects or opera- 
 tions, no two individuals use their words in precisely the 
 same sense. Names of abstract notions, terms of ethical 
 description and the like, are employed with different values, 
 depending on the conceptions of the speaker. If a speaker 
 describes a thing as ' good ' it is very seldom that a hearer 
 can realise precisely how much or how little the word is 
 meant to convey. The notion of what is ' good ' or other- 
 wise differs in different minds, according to their moral or 
 aesthetic constitution. The term is of course accepted on 
 all hands as one of commendation ; in a particular connec-
 
 ^"/S Natural History of Language 
 
 tion the particular point of commendation may be fairly 
 appreciable ; a thing is ' good ' to the moral sense, or it is 
 good for its purpose, or it is pleasing to the palate. Yet 
 the moral sense of A is not in all respects the moral sense 
 of B, nor are their palates identical. A word is a coin or 
 token of language ; a speaker may intend his token to 
 represent sixpence, while to the listener its current value 
 may either be only fourpence or it may be ninepence. As 
 compared with the speaker's use of the word, his own may 
 mean less or more ; it may leave out one element in the 
 conception while perhaps adding another which was not 
 in the speaker's mind. There is, of course, a sort of 
 averaging in the meaning as accepted by the whole 
 community, but, as with articulation, there are in reality 
 as many subtly graduated meanings as there are speakers 
 of the language. At the present day to say merely that 
 a man is a ' gentleman,' or a woman a ' lady,' is to convey 
 little specific information. What does the speaker, as 
 opposed to the hearer, mean by a ' gentleman ' or a ' lady ' .-' 
 What pictures have their respective education, environment, 
 ethical tone, sympathies and tastes built up in their minds ? 
 The terms are commendatory, but it would be difficult to 
 define the particulars which the whole speech -community 
 agrees in attaching to them. 
 
 Such width or vagueness of application must inevitably 
 exist already in many words of a primitive tongue. In the 
 effort to express abstract or general notions a greater strain 
 is put upon the simple resources of such a language than 
 is necessitated at a higher stage of development, when the 
 vocabulary has been immensely enlarged and when words 
 have been in a large measure specialised to correspond to 
 more specific concepts. It is true that the distinctions 
 of thought have meanwhile been increased by distinctions 
 of language, and that, in the attempt to cope with thought, 
 language is always struggling in the rear. Nevertheless it 
 is to be assumed that the degree in which speech falls short 
 of thought was much greater in a primitive language than
 
 Changes of Meaning 379 
 
 it is at present. Abstract notions were habitually expressed 
 by figurative terms drawn from the physical world, and such 
 terms are after all only partially applicable. Some, it is 
 true, are more conspicuously apt than others. The English 
 right, etymologically identical with Latin rectus, properly 
 means ' drawn out,' ' stretched taut,' and so ' straight.' 
 Wrong is that which is 'wrung' or 'twisted,' 'crooked,' and 
 the same is the sense of the Latin pnnms (whence de-praved). 
 Moral or intellectual rightness has thus been expressed 
 imaginatively by words drawn from the concrete and visible. 
 In this instance the aptness is of the first order. Yet it 
 must be evident that under one primitive word ^rektds 
 there would be included a large group of associated notions, 
 physical, moral, and intellectual, not all of which would be 
 in the mind of the speaker every time the word was used. 
 There was room for various wider and narrower acceptations. 
 That this must be the case is plainly indicated by the 
 history of *rekt6s in English. Right has practically lost 
 the physical sense of ' straight ' altogether, and is now 
 applied only in the moral or intellectual sphere. Wrong 
 has undergone the same narrowing. And, even so, the 
 meaning intended by ' right ' and ' wrong,' as by ' gentleman,' 
 will still depend on the mental constitution of the speaker. 
 
 We must allow therefore for the existence, in both the 
 present and also a very early stage of every language, of 
 numerous words of vague or complex signification, in the 
 use of which there must be much diversity. Wherever 
 there can be diversity, it is impossible to tell what ' average ' 
 meaning will eventually prevail as the language develops. 
 
 But apart from the words whose meanings never can 
 possess a distinct outline, we may observe several tendencies 
 which lead to thft narrowing, widening, or shifting of meaning, 
 even in cases where the application of a word is originally 
 clear enough. Widening, it is true, is comparatively rare. 
 Since the aim of language is to express thought more and 
 more adequately, it follows that language tends rather to a 
 more precise than to a less precise application of a word.
 
 380 Natural History of Language 
 
 Many examples of so-called widening are in reality rather 
 cases of transference. The Latin villa is a country-seat of 
 a certain type, commonly consisting of a mansion with farm- 
 buildings. If in French ville is a town, and in Italian villa 
 is a special kind of residence, these uses can hardly be called 
 widenings of meaning. The sense is different from that of 
 the original, but it is no less specific in its new domain. 
 Nevertheless widenings, though less frequent, are occasionally 
 found, as, for instance, in the case of jjiiscreant, originally a 
 ' misbelieving ' heathen or heretic, and then a discreditable 
 character in general. The Latin domina means ' mistress.' 
 Being used as a title of courtesy, first for any lady and then 
 for any woman, it has come, in the shape of the Italian donna, 
 to mean simply ' woman,' and to serve as the current word 
 in that general sense. Unkind once signified specifically 
 ' unnatural ' to relatives ; its present sense is both wider and 
 weaker. Place (Latin platen, from Greek ifXaTela, i.e. ' broad ') 
 was first a broad street, then a comparatively open place in 
 a city, such as a ' square.' Its present indefinite English use 
 is a remarkable instance of broadening and attenuation. 
 
 We may now examine somewhat more closely the manner 
 in which meaning- change takes place, confining ourselves 
 generally to examples from the English language, as likely 
 to be more readily appreciated. We shall find that such 
 changes are reducible to — 
 
 1. Specialising or narrowing ; 
 
 2. Generalising or widening ; 
 
 3. Shifting or transference ; 
 
 and that the chief motives of change may be classed as — 
 
 1. Indefinite width of meaning in a word as originally 
 
 applied, causing diversity of use. 
 
 2. Predominance of one element in a thing named, 
 
 causing a more special application of the name to 
 that element. 
 
 3. Unconscious inclusion of a secondary meaning, due 
 
 to a natural association of ideas, and thence either
 
 Changes of Meaning 381 
 
 a gradual transference to that secondary meaning, or 
 else a widening to comprise it. 
 
 4. Effort at force or clearness of expression, or at liveli- 
 
 ness, causing a figurative application of words and 
 hence a broadening of these words, 
 
 5. Emotional emphasis, leading to a misuse of a term 
 
 in a wider or weaker sense than the true one. 
 
 6. Euphemism and irony, or a desire to avoid the un- 
 
 pleasant or less courteous term, and thence a new 
 sense acquired by the euphemistic substitute. 
 
 7. Other laxity in the use of words, through ignorance 
 
 or misapprehension. 
 
 Of the changes caused by diverse treatments of words 
 (mostly abstract) in which the outline of meaning was 
 necessarily vague, we have already spoken in the case of 
 good, right. For the rest it is most profitable to proceed 
 by the examination of examples. 
 
 When a thing is named or thought of, a picture of that 
 thing is called up before the mind. In such a picture certain 
 elements or features will be more distinct than others, certain 
 characteristics will be foremost in suggesting themselves. 
 Which elements, features, or characteristics these will be, will 
 depend on the experience which the speaker or listener has 
 had of the thing in question and on the point of view from 
 which he commonly regards it. The most obvious instances 
 are those in which the name belonging to a class of objects 
 becomes restricted to one species of them. If, for example, 
 deer means a ' wild animal,' the picture which is called up 
 will depend partly on the sort of wild animals to which the 
 speakers are accustomed and the point of view from which 
 they mostly concern him, and partly on the connection in 
 which the wild animal is immediately spoken of. In England, 
 when wild animals of any size had become rare, and when 
 they were mainly considered as edible game, it was not 
 unnatural that the expression to ' go hunting deer ' summoned 
 up at once a picture of that wild animal which was par 
 excellence the object of the chase. At first a narrower use of
 
 382 Natural History of L anguage 
 
 deer for this particular animal would be a laxity, but it would 
 be a frequent and contagious laxity, and, inasmuch as no 
 misapprehension ensued, it gradually passed into the current 
 sense. Again, in a country in which tame cattle, as kept on 
 farms, are mostly of the female sex, a term which once 
 applied to both sexes may in process of time come to be 
 accepted as denoting the commoner sex only. Thus the 
 Indo-European *gdHS ('ox'), '*si'is ('pig'), *oij.is ('sheep') 
 are represented in English respectively by cow, soiu, ewe, 
 words which apply only to the female. Yet neither Greek 
 (^oG?, t)?, 0I9) nor Latin {bos, sfis, ovis) has made this limitation. 
 The psychological process in the case of the English is 
 sufficiently clear. At the mention of sow or ezve there was 
 conjured up, at first unconsciously, the assumption of the 
 female animal rather than the male. This occurred so 
 generally that (although again the step is unconsciously 
 taken) it was not found necessary to add any further term 
 denoting the sex. Alaid and girl were originally of common 
 gender. The limitation to the female would doubtless first 
 occur through the greater frequency with which the allusion, 
 in conversation, was to children of that sex. In the case of 
 wife ( = ' woman ') the process is more easy to follow. There 
 was in each house a ' man ' and a ' wife ' {i.e. ' woman ') ; the 
 husband would habitually speak of ' my wife ' {i.e. ' my 
 woman '), and whenever in Old English the common phrases 
 ' my wife,' ' the wife,' ' man and wife ' were heard, ' wife ' 
 would naturally be taken to mean, not simply ' w^oman,' but 
 also wedded woman, mate, or spouse. That aspect of the 
 woman would come to be most prominent, until eventually 
 it would be necessary to add some qualification to the 
 word ' wife ' when it was not to be accepted in this more 
 special sense. Hence a compound, * woman ' (i.e. wif-man), 
 won its way for that purpose. 
 
 The meaning of the late Latin capitate is ' principal ' or 
 ' sum of movable property.' From the Norman French the 
 word came into English as catel, and from the Parisian 
 French of the law courts as chatel. The notion called up by
 
 Changes of Meaning '^Z'^ 
 
 the mention of catel {cattle) was chiefly that of property in 
 animals. (We may compare the frequent modern narrowing 
 of ' stock ' to Hve stock.) Since its general acceptance in this 
 sense, there has happened in more recent times a further 
 tendency to limit the notion to oxen, as forming the more 
 conspicuous or valuable portion of such catel. At a date 
 later than the popularising of catel in this meaning of 
 animal property, the form chatel (as in ' goods and chattels ') 
 also emerged from the phraseology of the law into the 
 phraseology of ordinary life, and in this shape the word has 
 been narrowed in the contrary direction, to signify inanimate 
 but indefinite movable articles. 
 
 It should be remarked, however, that the narrowing of 
 such original senses will hardly proceed far unless there either 
 exists already some word capable of representing the older 
 and wider meaning, or unless some new word comes into 
 existence for the purpose. At the least it will be necessary 
 that some word should either exist or come into existence to 
 express that part of the wider meaning which has been 
 abandoned by the narrowed word. Thus if a word originally 
 including the area of meaning aArb becomes restricted to «, 
 there must either be or become some other word for a + b, or 
 at least some word to represent b. The narrowing of cozv 
 i^goHs) is made possible by the existence of the wider term 
 ox, as well as by the use of bull and bullock ; the narrowing of 
 ewe i^ouis), by the existence of sheep, as well as by the use of 
 ram and zuether ; the narrowing o{ sow (^sus), by the existence 
 oi pig and the use of boar. The limiting of corpse to a dead 
 body could scarcely have occurred if body or its equivalent had 
 not been ready to fill the area from which corpse was to shrink. 
 
 Of many other English words which have specialised 
 their meanings under influences of the kind in question a few 
 may be mentioned, eg. 
 
 
 Earlier Sense 
 
 Narrowed Sense 
 
 yard 
 
 Stick or pole 
 
 Stick of a certain length : 
 pole in a ship's rigging 
 
 voyage 
 
 journey 
 
 journey by sea
 
 384 
 
 Natural History of Language 
 
 
 Earlier Sense 
 
 meat 
 
 food 
 
 acre 
 
 field 
 
 starve 
 
 die 
 
 parson 
 
 legally representative 
 
 
 ' person ' 
 
 purchase 
 
 procure 
 
 fame 
 
 good or bad report 
 
 censure 
 
 judgment 
 
 thatch 
 
 covering (roof) 
 
 punctual 
 
 careful in any point 
 
 fowl 
 
 bird 
 
 undertaker 
 
 (of any business) 
 
 ghost 
 
 spirit 
 
 hound 
 
 dog 
 
 disease 
 
 discomfort 
 
 tide 
 
 time 
 
 raisin 
 
 grape 
 
 Narrowed Sense 
 the pihe de resistance in food 
 
 (flesh meat) 
 field of a certain area 
 die of hunger (provincially, of 
 
 cold) 
 ' person ' of a church 
 
 procure by money 
 good report 
 unfavourable judgment 
 (usual) roof of straw 
 careful in point of time : 
 domestic bird (gallinaceous) 
 (of funerals) 
 spirit of dead 
 dog of chase 
 discomfort in health 
 time of ebb and flow of sea 
 dried grape 
 
 The same phenomenon occurs in all languages. 
 Latin we have, e.g;. 
 
 In 
 
 actor performer 
 
 auctor one who pledges his word 
 
 studium pursuit 
 
 princeps a first man 
 
 > pleader 
 
 > author of book 
 
 > pursuit of literature, study 
 
 > emperor 
 
 In the later Greek aXoyov (a/ogon), properly ' unreasoning' 
 and thence ' brute,' became the current term for the horse, 
 just as deast in agricultural language commonly means a 
 bullock. 
 
 The process is in tentative operation at all times, and, 
 precisely like phonetic change, it may win a partial (or 
 dialectal) success in a certain region or among a certain 
 class, or it may conquer the whole speech-community, or it 
 may be counteracted and suppressed. Anyone who con- 
 templates the narrower use of such expressions as t/ie paper 
 or a paper {i.e. newspaper), the gas or gas {i.e. coal-gas), engine 
 {i.e. steam-engine), liquor {i.e. alcoholic liquor), rail {i.e. railway), 
 wire {i.e. telegraph wire), will perceive that they originate 
 indeed in connections or circumstances which explain them,
 
 Changes of Meaning 385 
 
 or are first so used among- those who are specially concerned 
 with then:i, but he will also perceive that the numbers who 
 readily apprehend the narrower sense without explanation 
 are very large, and that at least in the case o{ gas, engine, and 
 liquor that sense has practically come to be accepted by the 
 whole community. In what particular way a wider sense 
 may become specialised depends again on conditions too 
 complex to follow. The Latin Jirinare is ' to make firm ' (or 
 * fast '). In YxQ:x\c)\ fermer is only to ' shut ' fast {eg. a door), 
 in Italian fermare is only to ' make to stand ' fast {i.e. to 
 ' stop '). 
 
 Closely cognate to this cause of change — because due, 
 like it, to the mental picture which a word evokes — is 
 another in the list above given. This is the addition 
 (originally unconscious) of a secondary notion to the proper 
 notion of the word, prompted by an irresistible association of 
 ideas. This secondary notion may be so regularly connoted 
 and understood by speaker and hearer that it becomes as 
 prominent as the primary ; gradually it may grow to be 
 regarded as itself the primary notion, and eventually as the 
 only one. That is to say, a word of which the proper notion 
 is X may call up — as a result of some common experience 
 in connection with the thing named — the notion y. In- 
 sidiously the sense comes to be x-\-y, and finally y alone. 
 This has been the history of villain, knave, churl, boor, idiot, 
 wit, gossip, and a large number more. Attached to the 
 Roman villa (' country mansion with farm ') there were 
 villdni, labouring slaves. In its older French shape villain 
 the word denotes a feudal serf These uncultivated rustics 
 lacked the polish of the gentleman, and were supposed also 
 to lack his sense of honour. Hence with the mention of 
 villain (in the proper sense x) there was evoked the notion of 
 a 'low fellow' (sense y), and gradually that sense became 
 predominant. The change would naturally be assisted by a 
 deliberate metaphorical application of the term to persons 
 who were not serfs. For example, one might frequently say 
 of a man who nominally belonged to the gentry ' He is a 
 (b 884) 2 c
 
 o 
 
 86 Natural History of Language 
 
 villain,' in the figurative sense that ' he is (as low as) a serf.' 
 The transference would be completed at once when feudal 
 villenage no longer existed and when the strict sense was 
 therefore obsolete. 
 
 In the earlier Anglo-Saxon a churl is a ' man,' then a 
 * common man ' without social standing, especially a rustic, 
 then a serf, then a mean and surly person. How difficult it 
 is to prognosticate the exact mental process of such change 
 is shown by the difference between the new sense of villain 
 and that of churl. The notion in villain is of wickedness, in 
 churl it is of surly meanness. It is possible that the un- 
 amiable characteristics of the Romance villanus and the 
 Saxon ceorla actually differed most saliently in these ways. 
 IMeanwhile a boor was a peasant farmer, then any peasant, 
 then anyone marked by the rude manners of the peasant. 
 Here again the modern sense, though akin to that of churl, 
 is not identical with it. The churl is mean and surly, the 
 boor is rude and rough. 
 
 A knave was simply a boy (German Knabe). But 
 ' knave ' {Le. ' boy ') was a common term for household servant 
 (as in the Greek -rraU, the Latin puer, and the French garqon, 
 ' waiter '). A chief characteristic of the household ' boy,' 
 whether in Greece, Rome, France, or England, was supposed 
 to be dishonest cunning and trickery. Hence the special 
 modern sense attaching to knave. Nothing could be more 
 instructive than these four examples in revealing, first, the 
 manner in which the associated idea usurps the place of the 
 primary, second, the variety of such associated ideas which 
 may be presented. 
 
 An idiot was originally one who led a private or lay, as 
 opposed to a public or professional, existence. Such a 
 person was non-expert ; he was commonly ignorant. From 
 conspicuous ignorance to mental incapacity is an easy step 
 in the transference of meaning. Wit is properly knowledge 
 and enlightenment. In the early eighteenth century it is 
 culture and intellectual brightness. The manifestation of 
 this gift in bright and vivacious conversation has gradually
 
 Changes of Meaning ^i^'] 
 
 led to the specialised sense of the present day. A gossip 
 {god-sib) was a ' relation in God,' a co-sponsor at the baptism 
 of a child. Such godfathers and godmothers were naturally 
 intimates and cronies. The godmothers in particular were 
 ' gossips ' to each other, and their conversation was of a more 
 familiar and personal kind than is even usual with elderly 
 females. Hence — though doubtless figuratively in the first 
 instance — the later application of the word. It is interesting 
 to note that the French coniincre has gone through a similar 
 history. A coinmcre is a ^ feninie bavarde,' and coninierage is 
 ' p rapes de fennncs bavardcs' 
 
 In all the above circumstances, when referring to persons, 
 the change has been in the direction of disparagement. The 
 same is the case with the epithets artful, crafty, simple, silly, 
 daft, cunning, and others. But it is an error to assume that 
 the depreciation in the meaning of these words was necessarily 
 due to any contempt or dislike of art, craft, simplicity, 
 blessedness, innocence, or knowingness. Nor is it well to 
 philosophise to the effect that art, craft, and knowingness 
 are commonly turned to injurious uses. On the other 
 hand, it is tolerably certain that many of these terms were 
 frequently applied euphemistically or ironically. These 
 figures of speech, and also litotes (or under-statement), play 
 no little part in the development of meanings, and to call a 
 mentally feeble person silly (i.e. selig, ' blessed '), or simple (i.e. 
 ' open' ), or innocent, was an exhibition of considerate courtesy 
 in some mouths, of irony in others. Similarly, to style a 
 schemer ' knowing ' {cimtmig) or full of deftness {artfur) 
 may be called litotes, euphemism, or irony, according to 
 circumstances. 
 
 The generalisation of the courtesy titles Air. (i.e. Master), 
 of Italian donna, ' woman ' ( = domina, ' mistress '), and the 
 frequency in vulgar speech of lady and gentleman for 
 ' woman ' and ' man ' are due to the same euphemistic 
 principle. Another occasion of its employment is in speaking 
 of things which are in themselves unpleasant. Certain plain 
 words bring the disagreeable thing more crudely and coarsely
 
 388 Natural History of Language 
 
 before the mind ; another word throws a kind of veil over it. 
 In the effort to express otherwise unnameable things a wider 
 term is commonly applied. This gradually contracts to the 
 special meaning alone, and then must itself give place to 
 something less distinct or uncompromising. 
 
 The effort towards force and vivacit)- leads to two 
 departures from the normal use of words. There is first the 
 emphasis which substitutes a stronger word for the weaker 
 but truer one, as when extremely, Jiorribly, atrociously , delight- 
 ful, ivinieuse, and the like are used in place of ' very,' ' dis- 
 agreeably,' ' much,' ' pleasing,' ' large.' A frequent use of such 
 intensifying or exaggerating terms leads to a discounting 
 of their meaning. They become weak and colourless. It is 
 this habit — not to enter into any question of the habitual 
 dilatoriness of mankind — which has corrupted the meanings 
 of presently and by and by, so that they now no longer 
 express the sense of ' now, without delay,' but the exact 
 contrary, ' not now, but in a little while.' Immediately has 
 gone far upon the same path. Often emphasis is humorous 
 and deliberate ; at other times it is the effect of strong 
 emotion ; at others the result of a constitutional fondness 
 for fine language. 
 
 Far more important in this kind is the figurative or 
 metaphorical application of a term. It has already been 
 mentioned that, in order to express abstract notions or 
 mental operations, a primitive language must draw upon 
 names of things and operations which are physical. The 
 ' poetic ' faculty in mankind — which, as Aristotle puts it, per- 
 ceives the resemblances between things — expresses the unseen 
 in terms of the seen, the intangible in terms of the tangible, 
 the processes of thought in terms of the action of the senses. 
 To guess or judge correctly is to Jiit ; that which is un- 
 sympathetic and emotionless is cold; a face which suggests 
 the geniality of the sunshine beams ; a poem bears the stamp 
 of genius ; a statesman steers a country's course ; a vague 
 notion is ha::y. A primitive language acts thus, perhaps in
 
 Changes of Meaning 389 
 
 crude forms, of necessity ; but such enterprise of expression 
 is in progress at all times. It is a particular function of the 
 poet to present ideas vividly in this apprehensible form of 
 the concrete ; the man of original mind, when his vocabulary 
 is deficient, constantly falls back on metaphor, and in this 
 sense a backwoodsman is often endowed with the poetic 
 faculty ; he re-creates metaphor in its original simple and 
 homecoming terms, unconscious that the language already 
 possesses the same figure of speech in a faded form. The 
 humorist and the ingenious inventor of slang add their quota 
 of metaphor to the possible meanings of words. Every such 
 new application must at first widen the sense-contents of 
 a word, and this wider scope may be long maintained ; at 
 other times it happens that a special metaphorical use 
 increasingly prevails, until it drives other meanings, including 
 the primary sense, out of the field. The speakers of vulgar 
 Latin, whether humorously or not, applied the word testa 
 (' pot ' or ' potsherd ') to the skull. This was also loosely used 
 for the head in general. In modern Italian testa and French 
 tcte the sense of ' head ' is the only one current. A woman 
 of vicious temper is called a vixen. The word was originally 
 applied metaphorically, from the notion of her resemblance 
 to a she-fox (^h.'S). fyxe)i). It is true that vixe7i may still 
 mean literally a she-fox, but to many speakers of English 
 that meaning is quite unknown, and in the mouths of the 
 majority the metaphorical sense of vixen is the only one. 
 
 Since metaphor, alive or fossilised, is so obvious in 
 almost every sentence of language, there is no need to dwell 
 further upon examples. What chiefly concerns us here is 
 the fact that there is no uniform and necessary course of 
 development in figurative language, that the words which it 
 will affect and the extent to which it will go cannot be 
 calculated, and that, while all languages employ words 
 metaphorically, they by no means always turn the same 
 terms to the same metaphorical use, and that consequently 
 two related languages may find the senses of their cognate 
 words differing widely through this cause alone. We may
 
 390 Natural History of Language 
 
 harp upon a thing in English, but though German has the 
 cognate word Harfe for ' harp,' we must not look for 
 German to have applied that particular word in any phrase 
 expressive of the same notion. Head and Haiipt were once 
 the same in both form and meaning ; not only has the 
 articulation changed in both English and German, but the 
 metaphorical senses which we have come to attach to head 
 are only partially represented by the metaphorical meanings 
 which have been developed by German. If .r stands for the 
 original meaning ' head,' the present English use may be 
 represented by x-\-y and the German by x-\-z, and in 
 these formula; the elements in y are only partially those in 
 z. The modifying process of metaphor goes on for ever in 
 the meaning of words, and success or failure of innovation 
 depends on the same conditions as those affecting phonetic 
 change, with the difference that here the mind plays a part 
 much greater and more complex. 
 
 A last cause of meaning-change is to be found in mere 
 laxity or sheer ignorance. A word is misused by a speaker 
 who has never properly understood it, but who, on hearing 
 it from others, has in larger or smaller measure misconceived 
 the sense. Such misconception is not likely to be diminished 
 when a superficial education is general, and when the 
 vocabulary of the language, particularly in polysyllables, is 
 vast. In the first instance, there is generally an easily 
 appreciable affinity between the correct sense and the sense 
 as wrongly conceived, but the latter is certain to be some- 
 what vague and the confusion which results may eventually 
 lead to a very wide departure. The true meaning of 
 preposterous was equivalent to that of the phrase 'putting 
 the cart before the horse.' That it should have been 
 accepted in a wider sense of ' absurd ' was very natural with 
 those who were unaware of its etymology and had never 
 been warned of the particularity of its meaning. In its 
 legal sense /;r;«zj-^i' signifies the 'matters previously stated' 
 or described in a document. The layman took the term
 
 Changes of Meaning 391 
 
 to be simply equivalent to ' the property.' Thence the 
 modern use, which has become even more restricted. Secure, 
 from the Latin secfirus, is still used in Shakespeare in the 
 sense of ' free from care,' 'untroubled.' That the meaning 
 was intended by careful speakers to be distinct from ' safe ' 
 is best shown in the line — 
 
 Men may securely sin, but safely never. 
 
 Yet the context so often implied, or seemed to impl\-, actual 
 safety, that the word was commonly understood in that sense 
 by those who were ignorant of its history. At the present day 
 it is only too common an experience that many who use the 
 words infer, transpire, problematical are altogether mistaken 
 as to their proper values. They use infer for ' imply,' 
 transpire for 'happen,' and problematical for 'doubtful.' 
 They treat equivocation as simply a longer word for ' lie.' 
 
 In many cases it is scarcely possible to separate the 
 misunderstanding of pure ignorance from laxity through 
 association of ideas. The widening of miscreant may very 
 probably have been as much due to the misuse of ignorance 
 as to the reason already given. Without a live conscious- 
 ness of etymology a word like caitiff (originally identical in 
 form and meaning with captive^ might easily be misunder- 
 stood, from the tone of the speaker, as simply equivalent 
 to a miserable wretch. So puny ( =puis-tie, ' later born ') 
 must have been much misused before it could be so far 
 shifted as to bear its present meaning. Complexion is 
 properly the composition or constitution affecting the whole 
 body, as determined by the proportions of the ' humours ' 
 which played so great a part in medireval medicine. To some 
 extent a man's complexio?i could be gauged by the colouring 
 of his face. Yet it is natural to suppose that the first 
 persons to call that colour itself the complexion must have 
 been those whose notions of the true meaning of the term 
 were hazy. If we consider the change in the sense of 
 officious, which has passed from ' scrupulous in the perform- 
 ance of duties ' to ' meddlesome,' we must remain in some
 
 392 Natural History of L anguage 
 
 doubt whether the early deviations were not due as much to 
 a wrong notion of the intended meaning as to a less 
 complimentary suggestion, which often arose in the mind 
 when picturing the conduct of persons ' scrupulous in the 
 performance of duties.' In point of fact, the misunder- 
 standing and the association of the secondary idea operate 
 together. It is not customary in conversation for the 
 hearer, if perchance he is not certain of the precise sense of 
 a word, to ask the speaker for an explanation or definition 
 of it. He usually interprets it as best he can, and, after 
 some familiarity with it, uses it himself without taking the 
 trouble to correct his lax notion. If now we suppose that 
 A, in speaking to B, styled C an ' officious ' person in the 
 older and favourable sense, and if B had not himself been 
 accustomed to hear and use the term, one of two things 
 might happen. If B were actually acquainted with the 
 manner of the ' officious ' person and had noted that such 
 zeal generally went with meddlesomeness, that unpleasant 
 connotation would link itself to the word in his comprehen- 
 sion, and would be contained when he himself repeated it. 
 Or if A spoke with a tinge of sarcasm, or in such a tone 
 that, while employing the complimentary term, he implied 
 a dislike, B would naturally imagine the word to be dis- 
 paraging. In either case we have ignorance combining 
 with a disagreeable association to produce a shifting of the 
 meaning. 
 
 In general it may be said of the causes of meaning- 
 change, as above illustrated, that they seldom act singly. 
 A congregation is strictly a gathering together into a flock 
 or throng. All animate things may congregate, and for all 
 manner of purposes. Though the wider sense of congregation 
 cannot be called dead, the usual, and with many persons the 
 only, meaning is that of a gathering for religious service 
 or to hear preaching. Doubtless the chief cause of this 
 narrowing is to be found in the principle dealt with in the 
 case of ' deer,' ' cattle,' and other specialisations. The form
 
 Changes of Meaning 393 
 
 of gathering or congregation which, as being the commonest 
 or most important, was first called up in the mind by the 
 mention of the word was the religious or church gathering. 
 This would be so prevailingly the case that a powerful 
 movement would be given to the specialisation. But there is 
 also a time in each speaker's life at which he makes his first 
 conscious meeting with a word and especially a long word. 
 He is ignorant of its exact meaning, but gathers it as best 
 he may from the context. It would be almost inevitable 
 that a child should first habitually hear the word congrega- 
 tion in its narrower reference, and to that narrow sphere he 
 would naturally thenceforward attach it. We thus find not 
 merely specialisation, but ignorance assisting specialisation 
 in determining the fate of a meaning. The same history 
 belongs to censure, retaliate, resent, and other words, which 
 were originally neutral in meaning, but which have become 
 restricted to the unfavourable side. 
 
 It cannot be too carefully borne in mind that there are 
 always these two parties to speech — the speaker more or less 
 familiar with the word, and the listener who is first meeting 
 with and acquiring it. Into whatever laxity an individual 
 speaker falls, change might be more easily corrected if it 
 were not for the circle of younger and imitative speakers 
 who come more or less under his influence. As in the 
 corruption of sound, so in the corruption of meaning, each 
 individual radiates a certain force, which may be very small 
 or very great, but which is necessarily greatest upon those 
 who are brought very young within its sphere. A sarcastic 
 or ironical use of a word, a specialised use, a lax use, a 
 mistaken use, or a figurative use will in such instances 
 spread in some degree from the individual to his environ- 
 ment. And, as with change in articulation, change in 
 meaning acquires a conquering force when it is one which 
 occurs in many quarters at once and is backed by natural 
 psychological reasons. 
 
 It remains to point out that the meaning of certain
 
 394 Natural History of Language 
 
 classes of words may vary from epoch to epoch, or in 
 different regions of the same language, through a diversity 
 in social circumstances or in material practices and appli- 
 ances. A volume is properly a roll, and the term was 
 applied to a book, or portion of a book, forming such a roll. 
 The modern use is, of course, simply due to the substitution 
 of a new shaping or construction of books. Book itself 
 originally meant a wooden tablet or slip, probably of beech, 
 but the name changed its meaning with the change of 
 material. The same is true of paper, once the sheet of 
 papyrus pith. The Roman scnatus was a gathering of the 
 ' old men ' isenes), but in process of time it referred only to 
 a certain order in the body politic, without any consideration 
 whatever of the age of its members. In early Athens the 
 /Sao-tXeu? {hasileus), or king, was also the head of the State 
 religion and the chief officiator in its services. After the 
 abolition of kingship the name was retained for the annually 
 elected officer who performed these religious functions. 
 The political sense of the word minister (properly ' servant ') 
 is another instance in point. Until the introduction of 
 steam a ship sailed in the proper sense, but in modern 
 parlance a vessel sails although it is propelled only by its 
 screw. A town was a homestead surrounded by an enclosure 
 (compare the German Zaun, ' hedge ' or ' fence '). As the 
 Anglo-Saxon settlements, which were of this nature, grew 
 into villages and then into towns, the name went with the 
 change. Provincially, in the North, a ' town ' or ' township ' is 
 still often but a farm. The development in this word is 
 peculiarly like that from villa to the French ville. 
 
 How meanings may shift, either widely or slightly, and 
 how indeterminable antecedently the shiftings are, may be 
 most clearly seen from a comparison of ' doublets,' or words 
 of the same etymology, which are to be met with in most 
 modern tongues, but which are especially common in 
 English. When two forms of the same word exist in a 
 language, they either represent — as we have said already — 
 
 ?5P
 
 Changes of Meaning 395 
 
 different dialects, or one is borrowed from a cognate 
 language (such as by English from Dutch or German), or 
 both are borrowed, but through different channels at different 
 dates. It will be found that such doublets differ materially 
 in sense. There would indeed otherwise be no room for 
 both. The reason of their existence side by side is that 
 they do so differ, and therefore fulfil different parts in the 
 economy of speech. Examples are story and history (an 
 earlier and a later form borrowed from Greek through Latin), 
 drag and draw (from different dialects), lawn and land 
 (Celtic through Old French and also direct), tJirill and drill, 
 thatch and deck, ship and skiff, skirt and skirt, scale and shell. 
 In these cases there has been no deliberate differentiation. 
 The language has not consciously distinguished the two 
 meanings because it possessed two forms. On the contrary, 
 the side-by-side existence of the two is due to the fact that 
 they were already differentiated. Each dialect had given to 
 its word a particular shade of application and thence each 
 supplemented the other. 
 
 If now we suppose that a number of languages, such as 
 the Indo-European tongues, were at one time practically 
 identical, and if we confine ourselves to their gradual 
 divergence in respect of the meaning of the terms once 
 common to both, leaving other changes {eg. of phonetics, 
 syntax, and novel creations) out of account, we may represent 
 the main possibilities by the following scheme : — 
 
 Let a be the common word and the original sense, 
 a^ a specialised sense of that word, 
 cr a euphemistic or ironical application, 
 a^ a metaphorical application, 
 a^ an exaggerated emotional use, 
 b an associated notion, 
 c a sense misconceived through ignorance. 
 
 Then perhaps, when they do not actually retain the simple 
 sense a,
 
 396 Natural History of Language 
 
 Language A develops a into cv or a^. 
 
 „ B „ ,, b (through the step a-\-b'). 
 
 „ C „ „ a^ (through a + a^). 
 
 „ D ] agree in developing a into a^, but the special- 
 
 „ EJ isation may be different in the two cases. 
 
 „ F develops a into c. 
 
 Or, since a word may possess several applications simultane- 
 ously, we may have something like the following : — 
 
 A B 
 
 C 
 
 D 
 
 E 
 
 F 
 
 a and a^ d' a + d 
 
 a + d b 
 
 a a^ 
 
 c ai 
 
 j etc. 
 b 
 
 ■ — , • 
 
 ' .' 
 
 
 v 
 
 
 (generally a^) 
 
 (with b most 
 common) 
 
 
 (with c most 
 common) 
 
 (only) 
 
 It is easy to perceive that, after such varied development, 
 the speeches in question, though using etymologically the 
 same words, may be as foreign or unintelligible to each 
 other as if their words had been without any such original 
 connection.
 
 CHAPTER XVII 
 
 THE DEVELOPMENT OF LANGUAGE 
 
 I. The evolutionary principle in language; easing and 
 economy ; loss of flexion. — It is natural to assume that human 
 language, like other characteristic human functions and 
 attributes, has undergone a long and gradual process of 
 evolution. It is also natural to assume that it will continue 
 the process indefinitely. If there is a continuous adaptation 
 of man as an organism to his environment, and a progressive 
 emergence of the type and qualities which are best fitted to 
 the conditions of human existence, so there is a continuous 
 adaptation of language to the needs of expression. New 
 developments arise, for the most part unconsciously, and 
 also for the most part called forth by some actual defect in 
 the instrument. Sometimes a speech positively lacks an 
 adequate means of presenting a conception of the mind. 
 Sometimes, though the means is forthcoming, it is too re- 
 condite for either ready use or ready apprehension. There 
 is not only a perpetual struggle of language to keep pace, 
 by means of its vocabulary and its phrases, with thought in 
 all its subtleties and discriminations. There is an equally 
 perpetual tendency to economy of effort, both on the part of 
 the speaker and on that of the hearer. The evolution of 
 language means the development not only of the necessary 
 material but also of the most facile devices for its use. These 
 include phonetic easing and also structural easing. In other 
 words, such tendency to increased facility affects both the 
 phonology and the morphology of speech. Of phonetic 
 
 397
 
 398 NaUiral History of Language 
 
 easing and its phenomena enough has been said already. 
 The motives for morphological variation are the same ; they 
 are unconscious motives, which do not deserve to be called 
 indolent, but which invariably tend towards a recognisable 
 sparing of effort. To present thought not merely with as 
 complete as possible an approach to " the truth, the whole 
 truth, and nothing but the truth," but to present it in such 
 a manner as will make the least demand upon either the 
 time or the mental labour of the speaker or listener — this is 
 the ideal spontaneously pursued by language. 
 
 Except with men of letters (whose influence, however, 
 may be of considerable potency) there is no deliberate or 
 organised cultivation of these virtues. So far as they 
 advance, they are the outcome of the struggles of multitudes 
 of individuals against the limitations of their own linguistic 
 resources. New ' pointers ' to the precise meaning, in the 
 shape of novel devices of phrase, novel formations of vocables, 
 novel arrangements or positions of words in a sentence, 
 are begotten spontaneously when the speaker is fighting 
 against his own unreadiness, or against possible ambiguity or 
 vagueness, or against a false or insufficient understanding. 
 " Struggle is the begetter of all things," according to 
 Heracleitus, and of nothing is the dictum more true than of 
 language. Among the new phenomena which are thus at 
 all times tentatively offering themselves, some are without 
 further life. They belong to an individual here and there, 
 or to special occasions ; they do not indicate a widespread 
 need, or, if they indicate it, they fail for some reason to 
 satisfy it. At other times they are such as many speakers 
 are driven to hit upon, and many more, from the same sense 
 of a want gratified and a convenience gained, are prompt to 
 borrow. In such cases there follows a " survival of the 
 fittest." 
 
 Meanwhile, along with such new development, there 
 proceeds, as naturally as among organisms, a decline and 
 gradual disappearance of the unfit. In the case of language 
 the unfit is the inadequate and unserviceable, or the
 
 The Development of Language 399 
 
 cumbrous and superfluous. If speech, struggling to remedy 
 the lack of precision in a phrase or to facilitate its compre- 
 hension, is led to invent something new, it is obvious that — 
 if the evolutionary principle is true — in proportion as the 
 new device prevails, the less distinct or more awkward will 
 drop into neglect, desuetude, and gradually into complete 
 oblivion. The process may be slow, but it will be certain, 
 unless there arise some powerful artificial check, such as that 
 of a universal, or at least a very wide, scholastic education 
 in the language. Such a check may operate for an in- 
 definite period ; but should any social cataclysm occur 
 within the particular speech-area to interrupt the educational 
 tradition (as for instance through the invasion and domina- 
 tion of a barbarian or quite alien people), the elimination of 
 the useless and ' ornamental ' elements would speedily be 
 renewed. It is, for example, nothing but such an educational 
 tradition which keeps alive the use of the objective whom in 
 English. The vulgar speaker says Who did yotc see P, and 
 even to the educated it is something of an effort to begin a 
 sentence with the objective form in this instance. In such 
 a setting it serves no adequate purpose, while it puts upon 
 the speaker the strain of foreseeing or providing the exact 
 shape of the sentence into which it is to come. It is pre- 
 cisely against this and the like strain that language regularly 
 contends in its evolution, and it is obvious that the removal 
 of the mere educational precept would mean the total dis- 
 appearance of whom as the first word in the sentence. At 
 the same time it is easily felt that, in combination with a 
 preposition, e.g. in to zuhom, by whom, no effort of foresight 
 is required, and such phrases might retain the objective sign 
 for many generations. Even educational direction is some- 
 times compelled to yield to the evolutionary force in this 
 respect, and to abandon a rigid demand for the retention of 
 that which experience has proved to be nugator\'. In 
 France, where bureaucratic regulation of the language is 
 practised with at least as much stringency as in any other 
 country, the Department of Public Instruction has recently
 
 400 Natural History of Language 
 
 declared — among other concessions — that no penalty attaches 
 to the use of the masculine past participle in such sentence 
 as la ville que fai vu in place of vue. 
 
 Universal, or even very general, education is, however, 
 a factor but recently brought into play. Whatever may be 
 the strength of the break which it may exercise in the 
 future, it counts for little in a survey of the experiences of 
 language in the past. While the men of letters of the later 
 Roman Empire were keeping alive their array of noun- 
 inflexions, the people at large were dropping them. They 
 had evolved, by the use of prepositions, a greater precision, 
 and by means of a recognised word -order a more facile 
 manner, than pertained to the standard Latin declension in 
 the shape to which it had become reduced in historical 
 times. Since ad, de, in and the like expressed all that 
 needed expression, it was supererogation to retain therewith 
 such terminational distinctions as those in ad locum, in loco, 
 ad Rdmam, de Roma. Consequently these distinctions fell 
 into a secondary place in the consciousness of the speaker. 
 There being no necessity for their distinct and unconfused 
 articulation, the tendency already described to economise 
 articulating effort was allowed free play. With this went 
 also the tendency to economise mental effort, by allowing 
 one and the same form to stand when the work of distinc- 
 tion was already performed by the preposition. Hence the 
 Romance languages provide us with nothing more than lieu 
 and Rome, or luogo and Roma. 
 
 All this of course occurred gradually and spontaneously, 
 the speakers exercising no conscious discretion in the 
 matter. In the absence of any concerted action, the 
 process of elimination of the useless is seldom very rapid, 
 and within the Indo-European languages it is nowhere com- 
 plete. English has dropped the signs of genders from 
 its adjectives, while French and German have retained 
 them. The flexion of French and Italian verbs is still 
 copious, while that of English verbs is of the slightest. 
 Meanwhile, English has allowed to persist one person-
 
 The Development of Language 401 
 
 termination which cannot offer any greater practical claim 
 to existence than those which have been lost. It says / 
 have, we have, you have, they have, but he has. There is 
 manifestly no more need of a special form for the third 
 person singular than for the first, or than for the second or 
 third plural. In the past tense had suffices for all. When 
 in Eng-lish dialect he have is used, the sense is as unmistak- 
 able as in he has. 
 
 If we inquire into the reasons for these different rates 
 of speech at which words have travelled towards the dis- 
 carding of formative lumber, they will be found to lie mainly 
 in phonetic, not in psychological, conditions. One inflected 
 word has farther to travel than another, when it is to lose 
 its flexion and pass to the simple uniform shape. It was 
 a natural tendency for the Romance languages to drop a 
 final VI or s in pronunciation, and it was easy also to reduce 
 a final vowel or diphthong. The pronunciation of locum as 
 locu, and of this as loco, might occur from purely phonetic 
 causes. The substitution of the one form for several would 
 in such circumstances be rapid ; ad locum and de loco would 
 soon show no difference except in the preposition. In 
 English the gradual phonetic corruption of a first person 
 plural -en to -e, and thence to -e mute, soon causes the word 
 to coincide with an original singular -e, which also passes to 
 e mute. But no phonetic corruption would bring the first 
 person plural habhnus so close to the first person singular 
 hdbeo, and, if one form is ultimately to be used for both 
 persons, it will be the outcome of that instinct for uniformity 
 (sometimes called the ' system-instinct,' and commonly the 
 instinct of ' analogy ') which is to be mentioned later. This 
 phonetic consideration accounts in a very large measure for 
 the inconsistencies in the dropping or retaining of flexional 
 signs which are no longer essential to clearness and which 
 are still some bar to ease. It is evident that the tendency 
 to drop elements which no longer possess a use, and to 
 reduce several different forms to one shape, has been assisted 
 by a purely physical tendency to corrupt the final syllables 
 (B8S4) 2 D
 
 402 Natural History of Language 
 
 by lax articulation. Even the latter, if permitted free play, 
 would have gone far towards providing a common shape for 
 words once different. But, first, the phonetic corruption 
 alone could never produce the same complete identity ; 
 second, the phonetic corruption itself would never have been 
 permitted to proceed so far, if the very life of the meaning 
 depended on the information given by the inflexional signs. 
 Speech could not neglect the elements which were essential 
 to its intelligibility. The interaction of phonetic corruption 
 and grammatical reconstruction is, however, soon to be dealt 
 with again in another connection. 
 
 2. First devices of language : not polysynthetic. — The 
 question of the origin of speech may be for the moment 
 postponed. But it may be taken for granted that, when 
 certain articulate sounds or combinations of sounds had 
 become identified with certain notions, the earliest devices of 
 language were exceedingly crude and inadequate. Not only 
 would the number of such differentiated combinations of 
 sound be insufficient to distinguish all the objects and 
 actions distinguishable by the mind ; but it would prove a 
 difficult task to indicate the relations of such objects and 
 actions to each other, to the person speaking, to the person 
 addressed, or to some third person. How was it to be 
 possible to refer an action to past time, as opposed to the 
 present or future? Still harder, apparently, would be the 
 problem of stating a thing as conditional, as something 
 which 'would' occur 'if such and such another thing pre- 
 ceded. The evolution of the necessary expedients must 
 have been extremely slow in its first steps, and those which 
 established themselves must have been the fittest survivals 
 out of numerous endeavours. Their fitness would be 
 determined by their greater explicitness, their greater ease 
 of handling, or by both virtues combined. 
 
 It is self-evident that in its most primitive stages 
 language must have been almost helpless without abundant 
 use of two auxiliaries which still play some part — one of 
 them a considerable part — in the spoken language even of
 
 The Development of Language 403 
 
 the most civilised peoples. These were gesture (including 
 facial expression) and intonation. Gesture would be the 
 readiest means of indicating the relation between two notions. 
 For example, the speaker would point from this to that with 
 a motion of the hand to signify the direction in which the 
 action travelled. If a sign of certain numbers might be 
 that of so many fingers held up, a sign for large and in- 
 definite numbers might be the rapid opening and closing of 
 all the fingers, or a wide sweeping of the hands. Many such 
 gestures are readily conceivable. Intonation, again, would 
 serve for question or command. The notion ' if you kill 
 him, I will kill you ' might be expressed by the use, first, of 
 the sounds forjj'cw, kill, him (not necessarily, nor even prob- 
 ably, in that order) accompanied by pointing and a look of 
 forbidding or disapproval, and then of the sounds for /, you, 
 kill, accompanied by pointing and a look of threat. The 
 gestures would indicate subject and object, the look and 
 tone would express — so far as they were expressed — the 
 condition and the futurity. Gestures, however, are useless 
 in the dark, or when for other reasons the action of the 
 speaker is invisible to the listener. The inadequacy of 
 gesture itself, and the effort to express when gesture failed, 
 called forth all the ingenuity of primitive man, and the 
 gradual development of recognised ' demonstrative ' or 
 ' deictic ' elements in speech was the consequence. Once in 
 existence, such elements were combined in various more 
 or less ingenious ways, until it became possible to express 
 all the relations known as numbers, cases, tenses, moods, 
 persons, etc. 
 
 Theoretically the natural evolution of language was after 
 the kind described. The recognition of an individual object, 
 such as ' dog,' ' man,' ' food,' and of an individual action, e.g. 
 ' bark,' ' move,' ' feed,' must precede the recognition of the 
 relation between the object and the action which it performs 
 or which is performed upon it, eg. ' dog barks,' 'man moves,' 
 ' eats food.' That is to say, any combination of sounds ex- 
 pressing the whole thought that a dog barks or food is eaten,
 
 404 Natural History of Langttage 
 
 presupposes sounds which represent respectively ' dog ' and 
 'barks,' 'food' and 'eat.' Some hold (as Jespersenj that 
 language began with inseparable irregular conglomerations 
 (similar to those in the holophrastic tongues) and that the 
 component elements subsequently disentangled themselves, 
 the evolution of language giving to each element a free 
 existence, and providing devices by which such elements 
 enter into grammatical relations with each other. This may 
 be true in so far as it means that the expression of a com- 
 plete sentence — containing, for example, subject, predicate, 
 and object — may have been originally confused, without 
 definite grammatical principle, and may have consisted 
 simply of more or less close juxtaposition of the components 
 accompanied by gesture. But it entirely lacks plausibility 
 if it means that separate sounds, or sound-groups, for separate 
 concepts were only evolved from a complex of sounds or 
 sound -groups for complex concepts. This is surely to 
 reverse the chronological order. It is impossible to com- 
 prehend a genesis of compounds or conglomerations forming 
 such predications as ' dog barks,' if the components were not 
 previously in existence and ready to combine for the 
 purpose. Even in the polysynthetic languages, which 
 express a whole sentence in one compound utterance, such 
 as ' bring us the boat,' there is one element which signifies 
 ' bring ' and one which signifies ' boat.' When other 
 sentences involve the mention of ' boat,' that element appears, 
 although it may be again in combination and not isolated. 
 Such sentence-words imply the previous existence of a sound- 
 group for ' boat.' By what methods that element is to be 
 brought into relation with other sound-groups expressing 
 other notions, so as to form a sentence, is another question. 
 It might be the method of polysynthesis, or of agglutination, 
 or some other, but, whatever it was to be, the special sounds 
 expressing the special notion are to be presupposed. Apart 
 from the facts, first, that polysynthesis is a quite exceptional 
 linguistic method, and second, that what may possibly be 
 true for the history of the American Indian languages is not
 
 The Development of Language 405 
 
 therefore true for the history of any other family, there is 
 a palpable error of assumption concerning those languages in 
 particular. We have no right whatever to assume, as 
 appears to be sometimes assumed, that the American Indian 
 languages necessarily represent a primitive stage of language, 
 or are an example of arrested development. The packing 
 of a number of curtailed words into a compound might very 
 conceivably arise subsequently, and not prior, to a stage in 
 which the words had been as separate as in Indo-European, 
 and in which their sentence-relations had been expressed by 
 some not dissimilar device. 
 
 In Greek a compound heairoTT]^ {despotes) = ' vc\2iSX.&r of 
 the house,' and Zd-rrehov {dapedoii) = ' house-floor.' These 
 words represent respectively heyu-a-izorri^ {dem-s-potcs) and 
 Bd-ireBov ( = dm-pedon). But these formations follow, and 
 do not precede, the stage of separate words possessed of 
 flection. A whole sentence may sometimes be formed in 
 Greek by one word, e.g. Sacfjvrjcjiopeo) {daphnephoreo) = ' I bear 
 laurel,' ^vXoTOfxeo) {xidotouieo) = ' I cut wood ' ; but these 
 follow, and do not precede, the existence of a separate and 
 inflectible Bd(f)vrj {daphne), ^vXov {.xuloii), and </)e/>- or ^op- 
 {pher- or phor-). What has here occurred in Greek may 
 very well have occurred on a much larger scale in the holo- 
 phrastic languages, proceeding indeed so far that the earlier 
 method fell out of existence. Latin, again, has s7irgo, ' rise,' 
 pergo, ' proceed,' and these are compoundings and contractions 
 of sub-rego, per-rego. Navigo = ' I drive a ship,' rcitiigo = ' I 
 drive an oar,' and these are respectively compounds of furz'/- 
 and refno- with ag-. But no one doubts the previous inde- 
 pendent existence of ?idvi{s), rcinu{s), or of the verb ag{d). 
 Here the compounding of the 'most significant elements' of 
 certain words into a sentence-word is so far polysynthesis. 
 And, again, it is not unimaginable that some other language 
 might have done what Latin has not done, namely, carried 
 the practice so far that it ultimately became the regular, if 
 not the sole, method of sentence-forming. In English the 
 vulgar durino is meant for don't know, or ultimately do Jiot
 
 4o6 Natural History of Language 
 
 knoiv, while 7iot itself is compounded of na and the old 
 English form of whit. Don, doff, and the vulgar dout (a 
 candle) are compounds — do-\- on, do + off, do-{- out. I'd and 
 /'// are the colloquial forms of / zvonld and / zvill. To the 
 eye there is an apostrophe ; to the ear there is none, but 
 only the monosyllables did, dil. If in formal literature such 
 conglomerations are discountenanced, they are continually 
 being created in popular speech. For brevity or from 
 indolence such speech often combines the ' most significant 
 elements' in the separate words, i.e. the minimum of each 
 element necessary for the sense, into a compound. If Paris 
 converts Boulevard St Michel into Boule-Miche, London con- 
 verted 'Colonial and Indian Exhibition' into ' Colinderies.' 
 Doubtless the latter was deliberate, but it was no deliberation 
 which converted God be ivith ye or God be by ye into Good-bye. 
 The German zur ( = zu der) or beini ( = bei dem) are but 
 short steps in the same direction, but they illustrate the 
 same tendency. In Kalmuk usadsi baimitsi passes into 
 usads'dnutsi, and to distinguish this result from polysynthesis 
 is somewhat futile. By an unlettered people, engaged in 
 conversation, the line between separate word and whole 
 phrase is not drawn with the conciseness which belongs to 
 those who have learned to read and write. The tendency of 
 such a people is to run the several words into some sort of 
 amalgam, and that amalgam is obviously later in date than 
 the several words themselves. There is no good reason why 
 we should not hold for the polysynthetic languages the view 
 which we are led to hold for other families, viz. that a 
 separate sound-group for each of the separate notions con- 
 tained in a sentence was at one time in existence and in use. 
 The polysynthetic structure now prevailing is quite probably 
 the outcome of an unchecked habit of phraseological com- 
 pression and curtailment of the kind just illustrated for 
 languages nearer home. In these languages, if we possessed 
 their history, we should probabh' arrive at the same sort of 
 primitive material as in the amalgamating-inflectional or 
 agglutinating-inflectional tongues.
 
 The DcvelopJiicnt of Language 407 
 
 3. Primitive material: naming and pointing ^ roots.' — 
 We may proceed to consider the probable nature of this 
 material. The premises upon which a theory must be based 
 are derived partly from an examination and dissection of the 
 material which languages offer, partly from a priori xediSomng. 
 The scientific inquirer will not permit conclusions of the 
 latter kind to prejudice his treatment of the former. 
 
 In the chapter upon Morphological Classification it has 
 been shown that (apart from the polysynthetic American 
 tongues), whatever structural differences may distinguish 
 languages into convenient classes, they share (in various 
 degrees) in a common principle, viz. that of attaching, or 
 juxtaposing, to certain main or ' predicative ' elements certain 
 other elements whose function is ' formative.' These may 
 take the shape of prefixes, suffixes, infixes, or separate 
 particles, whether pre-posed, as in our own prepositions, or 
 in postposition, as with certain Japanese elements and certain 
 Chinese 'empty words.' The modern analytic tongues are 
 known to be the outcome of older synthetic languages, and 
 we may therefore go back to their synthetic stage before we 
 enter upon speculation in place of assertion. 
 
 Whether now we examine the amalgamating or the 
 agglutinating speeches, it will be found that two classes of 
 elements are readily distinguishable. The one is a ' naming- 
 element,' whether of object or action, and is commonly styled 
 ' predicative ' (sometimes ' nominal ') ; the other is a forma- 
 tive element, or a grammatical sign, and is frequently styled 
 ' demonstrative ' or ' deictic ' (sometimes ' symbolic,' and less 
 appositely ' pronominal '). A single naming-element in its 
 simplest form is called a predicative ' root ' ; a single 
 formative element in its simplest form is also frequently 
 called a demonstrative ' root.' Since the term ' root ' is 
 often loosely and inaccurately applied, sometimes to whole 
 stems and sometimes even to whole words, it is necessary 
 to define its scientific use. A predicative ' root ' is that 
 naming-element which is present (whether in an obvious 
 form or phonetically disguised) in one and all of a series of
 
 4o8 Natural History of Language 
 
 cognate words, each of these being formed upon it, as upon 
 a common core or nucleus, by means of other elements which 
 have their special purposes. Thus the Latin tenuis and tendo^ 
 the English thin, the Greek Telvw {teino), ravao'^ {tanaos), 
 together with their parts, derivatives, and compounds, are 
 all formed from the same Indo-European root *ten, of which 
 the sense was ' stretch ' or ' pull.' According to various 
 degrees and natures of the accentuation (pitch or stress) 
 which it might bear in different circumstances, this root was 
 pronounced in Indo-European times as ten- or ton- or tn-, 
 and, if we describe it as ten-, the practice must be regarded 
 as one merely of convenience, this form being chosen because 
 it is the one which was apparently employed when the accent 
 fell in the most clear-cut manner upon the root itself. In 
 the English be, the Latin fiii and super-bus, the Greek ^vw 
 {phuo) and vTrepcpvi]^ {Jiuper-phucs), with their cognates, we 
 arrive at a common core in the shape of the Indo-European 
 *bheu- (which again has by- forms ^bhu- and *bhu-). The 
 application of the special phonetic laws of the several 
 languages, as ascertained by scientific inquiry, leads us to 
 postulate, with a fair degree of certainty, the presence of a 
 * root ' *ten- in the one series of words and of a ' root ' '^bhen- 
 in the other series. Such a root, as far as we can go back 
 in the family of languages, is almost never found as an 
 independent word (although the Latin imperative I ' go, 
 and the Greek el, may very well represent the pure root 
 *ei), but it regularly carries with it some other element 
 of the non- predicative nature. This fact has led some 
 writers to speak of such roots as ' abstractions.' Abstrac- 
 tions they are, in so far as their separation or enucleation 
 from the words which contain them is an act of thought ; 
 but the term must not be applied to them as denying that 
 they ever possessed a separate existence. The absence of 
 positive ocular or auricular proof can hardly weaken the 
 overwhelming probability derived from common reason 
 If as the core of each of a list of words, whose meanings 
 literal or figurative, are related, there is to be found a common
 
 The Development of Language 409 
 
 element ten- (varied only by grades of pitch, stress, or dura- 
 tion of utterance), how are we to account for the fact, except 
 by supposing that a sound -group ten- had acquired an in- 
 dividual shape and a ' nanning ' power of its own in the minds 
 of primitive speakers, before any words could be formed 
 involving it ? How, in an agglutinative tongue like Turkish, 
 did the clear and distinct syllable sev come to be a base 
 upon which was built a whole series of words connected 
 with 'love,' unless a value 'love' had been previously 
 assigned to the sound-group sevl In Semitic the three 
 consonants qtl are the constant elements in a long series 
 of words connected with ' killing.' The differences of the 
 words lie mainly in the vowels enclosed or attached to this 
 framework. Yet how came this framework to exist in each, 
 and to be always associated with this meaning, unless its 
 naming value had been first recognised ? No pronounceable 
 form of a ' root ' can indeed have consisted merely of the 
 three consonants (any more than the Indo-European ^bhu 
 can have had an absolutely separate existence like *bhH or 
 *bheu\ but it is not claimed for Semitic that the consonantal 
 roots at which we arrive by the enucleating process represent 
 the earliest shape of the independent utterance expressing 
 the single notion. In all families of language we arrive 
 at a list of predicative roots for that family ; and it is hard 
 to doubt that at some primitive stage these had been 
 recognised as separate sound -groups possessing a naming 
 power of their own. We cannot indeed be sure, nor is it 
 asserted, that when we have arrived at an Indo-European 
 *ten we have discovered a syllable which sprang fully 
 formed from the lips of remote ancestors of that people. 
 It is not denied that ten itself may have been gradually 
 built in some miocene period of language out of still 
 simpler constituents, such as a stop -sound t and a sonant 
 nasal n. An attempt to vocalise the act of ' pulling ' may 
 possibly have been made in the poor efforts of one portion 
 of very primitive humanity, by first pressing the tongue 
 against the teeth while drawing the arm back with a strain,
 
 4IO N^ at m'Lil History of Language 
 
 and then, as the strain increases, letting the tongue press 
 back along the gums with the pull of the body, while the 
 air is allowed to escape through the nose. Anyone who 
 pronounces tn in a protracted manner while making these 
 motions will realise that such a development is by no means 
 unnatural. But, though such a suggestion may be thrown 
 out in passing, it leads to speculation beyond the immediate 
 point. That point is that families of languages provide us 
 with predicative 'roots,' which, for all that we can discover 
 of those families, are irreducible. For the most part (though 
 Semitic is an exception) the roots reveal themselves as mono- 
 syllables, and, the more sagacious and penetrating etymology 
 becomes, the more simple is apt to be the form of root 
 arrived at. Thus the Latin fujido, with perfect fftdi, is 
 not referred to a root already containing <-/, still less n, but is 
 first simplified mto fu-, which relates itself to the Greek ')(y- 
 (c/m-) in ^ucTi'? (chiisis). Similarly y^a'*?, with perfect y?a, is 
 not referred to a root containing c, but is first reduced to 
 fa-, which finds its entirely certain congeners in the remote- 
 looking English do, Greek 9ri- itJtc-). A priori it would be 
 supposed that a primitive utterance expressive of a single 
 concept would consist of sounds uttered at one effort or 
 impulse of the breath — that is to say, of a single syllable — and 
 the theory is in general borne out b\' etymological investiga- 
 tion. The roots arrived at are regularly monosyllabic. 
 The Semitic dissyllables may themselves very well represent 
 a second stage in development, being already composites at 
 the date when we first discover them. 
 
 The independent existence of predicative or naming roots 
 of a monosyllabic shape being thus extremely probable, it 
 remains to consider the subordinate elements which act as 
 ' pointers ' to the relations or special applications of such 
 roots. These are the formative, demonstrative (' deictic '), 
 or ' symbolic ' elements which serve as flexion-signs. Given 
 a naming root *ten (' pull '), by what process does ^tenio- 
 come to be formed, and to possess the sense of a verb 
 applied to the first person singular (' I stretch ') ? If the
 
 The Development of Language 41 1 
 
 Latin fac- (' make ' or ' do ') represents fa-c-, and if this is 
 reducible to *dhd-k-, what and whence is the -k-} U*Hoq- 
 denoted the general sense of ' utterance ' or cry, how does 
 the addition of *-os or *-es {^uoqos, *ijoqes) determine the 
 meaning to be that of a noun in the genitive singular ( = ' of 
 an utterance '), while the addition of *-m (^rwqm) makes it 
 an accusative ? If in Turkish -lar (or by vowel-harmony 
 -/^r), added to a singular, formed a plural, how did it come 
 to have this effect ? Why in Zulu should umu;z//^ mean 
 'man' and aba;///^ 'men'; and why should the prefixing 
 of nsr- denote the grammatical relation ' with the men ' ? 
 The hypothesis which would naturally arise is that these 
 added elements also once possessed an independent form 
 and meaning. It is difficult to comprehend how a factor 
 like -k- or -os or -m or iig- or -lar ever came to perform 
 the work which it does perform, unless at some earlier 
 stage it meant something which, by juxtaposition (or actual 
 union) with the naming- root, determined to some special 
 application the general sense inherent in that root. That 
 is to say, it pointed out (or ' demonstrated ') a certain bear- 
 ing of the 'named' sense. If the bearing was not made 
 sufficiently precise by one such addition, it was an obvious 
 device to employ yet another, or a number. A great part 
 of the evolution of language would be achieved by tentative 
 efforts to that end, and by a process of natural selection of 
 the fittest. No objection to this view should be felt from 
 the apparently unpronounceable character of -k- or 7ig- in 
 isolation. It should be manifest that the original independent 
 shape was not merely -k-, but something more (e.g. kd), and 
 not merely ng, but something more. In the English mans 
 we write an apostrophe, but we do not hear it. The word is 
 now nothing more nor less than mans. To one who argues 
 that -s could never have had an independent existence, the 
 answer would be that, as a historical fact, it was once -es. 
 (Moreover in this particular instance the assumption itself 
 would be incorrect, since s is quite capable of a sonant value.) 
 4. Origin of the poi?iting or deictic elements. — The fact
 
 4 1 2 Natural History of Langtiage 
 
 that these demonstrating elements were secondary in 
 importance would cause them gradually to assume the 
 lightest possible form compatible with the serving of their 
 purpose. Their phonetic abbreviation would proceed accord- 
 ing to the principle of economy already sufficiently illustrated. 
 In English don't (heard as donf), not is no longer a separate 
 word ; it has passed from the enclitic stage of do not, become 
 a negativing suffix, and in that capacity has lost much of 
 its original phonetic contents. Similarly, as a prefix, ne 
 became ;/, and in old English nill=ne will, nas = ne was. 
 In Roumanian domnul is the sufficient outcome of dominum 
 ilium (the subjoined article), while in French le or even /' 
 (which, though written separately, becomes practically part of 
 its noun) is all that remains of Latin ilk {illd^. Whether 
 as prefix or suffix, therefore, the demonstrative element 
 would in most instances go through considerable curtailment, 
 and possibly other phonetic corruption. This fact renders 
 much more difficult than it might otherwise be the attempt 
 to discover the original independent meaning of these 
 elements. It is impossible to speak with any confidence 
 of their ' root ' sense, when we cannot discover their ' root ' 
 form. If the history of the English suffix -ly were lost, no 
 speculation would be likely to hit upon the truth that it was 
 once lice and meant ' with the body,' just as the French 
 corresponding termination -inent is the Latin inente, ' with the 
 mind.' If some bold spirit did make a happy guess at the 
 origin of -ly, he would find it hard, in the absence of the 
 direct evidence, to prove his case. It might be strengthened 
 by quoting, first, the vulgar use of -like {zviselike), second, the 
 Homeric expression Se/jLa<i Trvpo^ aWojievoLo (literally ' with 
 the body of blazing fire ' = ' like blazing fire ') ; but it would 
 remain at most a plausible suggestion. 
 
 In dealing with the demonstrative or formative elements 
 as we find them clearly in use at the earliest stage of Indo- 
 European, Ural-Altaic, or Semitic actually known to us, 
 we lack all data as to their past career. It must be 
 remembered that even the pro -ethnic (or Ur-) Indo-
 
 The Dcvelopjucnt of Language 413 
 
 European or Semitic, at which we arrive by induction, is a 
 comparatively late stage in the history of those primitive 
 tongues. There must have been millennia of previous 
 development in each case. To trace back the formative 
 elements to their original independent shape and meaning 
 is practically a hopeless task. Doubtless some were once 
 predicative and named an action. Others were probably 
 natural sounds of an exclamatory nature, which accompanied 
 pointing and other gesture, and which came to signify what 
 the gesture itself signified, viz. such notions as ' here,' ' there,' 
 ' to,' ' from.' In some instances we may guess, so long as we 
 are satisfied to acknowledge that we are but guessing. If 
 Sanskrit has -s-ya- as formatives of a future stem of verbs ; 
 if Greek appears also to have -ct-lo- {-s-io) ; and if the 
 sum of evidence leads us to believe that the primitive 
 Indo-European used a future formation in -s-i-o- (e.g. 
 *deik-s-i-o-\ we may speculate as to the meaning once 
 possessed in independence by these particles or by the roots 
 from which they have been curtailed. It would be no 
 absurd conjecture that -s- is the root *es- (' be ') while -i- is an 
 abbreviation of the root *ei or *z (' go '). Here we should 
 have two curtailed roots originally predicative, employed 
 as pointers to another root, that to which they point being 
 the sense of futurity. The total sense would thus be 
 ' show-be-go.' Strange as this may look, it is not far 
 removed from our own ' I am go'm^ to show.' Colloquial 
 French substitutes for its future jc dirai (which itself means 
 ' I have to say ') je vais dire, '\ go to say.' Similarly Greek 
 has the idiom epxa/Jt^ai ipecov (' I come about-to-say '). In 
 English ' I am to say ' is a familiar turn of phrase. In 
 Latin the futures in -do (e.g. amabd) are formed from the 
 root ^bheii- (' be ') already mentioned, and are therefore 
 equivalent to ' I am (to) love,' etc. Since ' I go to-' and 
 ' I am to-' are thus well-substantiated manners of express- 
 ing future action, it is not improbable that primitive Indo- 
 European employed both these methods, the added -s- 
 (i.e. *es) representing the one, and the added -{- (i.e. *ei)
 
 4 1 4 Natu ral History of L anguage 
 
 the other. The combination of the two was for the greater 
 explicitness. This is, however, conjecture, and though it 
 may be plausible in this instance, there are few of the 
 formative elements to which conjecture can be applied with 
 any satisfactory result. It was once assumed, for example, 
 that the person-endings of the Indo-European verb, most 
 familiar in Sanskrit, Greek, and Latin, were identical with 
 the pronouns which are actually found in independent use 
 for those persons. Thus the first person singular ending 
 Greek -\ii or -v (for older -in\ Latin -m, was identified, at 
 least to the extent of the consonant, with the in in /xe, vie. 
 The third person singular found in Greek in -t-o (and also 
 formerly as -r, which was regularly dropped when absolutely 
 the last sound in the word, e.g. te^e/jer), Latin -/ (e.g. est, 
 habet\ English -th {hath), was similarly identified, to the 
 extent of the consonant itself, with the t in the demonstrative 
 ro-v (Js-)te, ovTO'i (hou-to-s), found in English as th in that. 
 There may be truth in these identifications, but unfortunately 
 the probability ends with these two persons. The attempt 
 to connect the second person in -s (Greek €^rj<;, Latin eras, 
 etc.) with a pronoun su (Greek av) is at once put out of 
 court by the fact that the pronoun in question was *tu 
 (as in Doric Greek, the Latin /;/, and the English thou), 
 while av (su) is only a corruption of Greek itself, and had 
 no existence in Indo-European times. For the plural 
 person-endings no rational account can be given of derivation 
 from the recognised extant pronouns. It is not indeed 
 out of the question that they may be derived from pronouns 
 no longer extant ; the fact remains that we are in no 
 position to prove that hypothesis. 
 
 Nevertheless, while we must decide that the original 
 shape and meaning of the once independent demonstrative 
 ' roots ' are for the most part beyond hope of ascertainment, 
 there is little room for doubt that they did once enjoy such 
 independence. 
 
 It has been argued already that the recognition of an 
 object or an action must precede the recognition of the
 
 The Development of L anguage 4 1 5 
 
 relations of objects to actions or to each other. The root- 
 name of the action or object must have existed before it 
 could be uttered in connection with such relations. The 
 relations would be various, and would demand various 
 expedients, which, apart from gesture, would naturally take 
 the shape of sounds appropriate to such gesture, or else of 
 root-sounds used in a subordinate manner as helps to define 
 the bearing of the main root-sound in the particular case 
 (as in *dcik-s-i-\ 
 
 When we have passed the period of dependence upon 
 gesture, the pointing element is indispensable. Along with 
 the chief notion, as expressed in a root-word, would go such 
 deictic or defining element or elements as the case seemed 
 to require. Whatever the source or sources of these, they 
 would be developed gradually and tentatively from the efforts 
 of various individuals and generations. Once found effective 
 in use, such an element would be repeated for the same 
 purpose when the next need arose. Its use would spread 
 from individual to individual until it became part of the 
 inherited tradition of a speech-community. 
 
 5. Deictic elevients becovie suffixes. — In the first instance 
 the ' deictics ' would be simply juxtaposed. As a purely 
 hypothetical case we might assume a demonstrative element 
 -/, originally accompanied by a gesture, and meaning ' here ' 
 or ' there ' (viz. ' where I point '). Such a particle might 
 or might not be identical with a naming- root 'go,' that 
 root being itself nothing but an exclamation accom- 
 panying a gesture to indicate a direction. The derivation, 
 however, is beside our immediate point. An expression 
 equal to ' at the house ' (' at home ') might be constructed 
 by means of a juxtaposition, ' house here,' say *Hoik i. 
 The regular use of this same juxtaposed clement, wher- 
 ever occasion called for it, would, as in the case of the 
 English -ty { = lfcc) and the Chinese 'empty words' (e.g. ///), 
 cause it to be treated as a mere sign. Articulated rapidly 
 in connection with the root-word to which it referred, it 
 would become virtually one with it (i.e. ^iwiki). That is to
 
 4 1 6 N at lira I History of Language 
 
 sa\', it would become a suffix, and would be destined, in the 
 far-off days when language became a matter for grammatical 
 examination, to be called a ' locative ' ending. When in 
 Japanese ' at school ' may be written as either gakko ne or 
 gakkone (' school-at '), we are able to see the step taking 
 place before our eyes. When French writes celui-ci, celm-ld, 
 we have a purely scholastic hyphen, which means nothing in 
 speech, while in ceci, cela not even the written sign of separa- 
 tion is left. 
 
 The same secondariness of the suffix (illustrated in an 
 easily recognised form in the Ural-Altaic adaptation known 
 as ' vowel-harmony ') commonly causes it to be phonetically 
 slurred or abbreviated. The reasons which have shortened 
 the pre-posed English of and on into a {aboard, etc.), or o, 
 as in the vulgar pronunciation usually represented by d {the 
 Pope d Rome, etc.), may operate no less when the pointing 
 element follows. The docked articulation serves its purpose, 
 and economy of time and effort is once more achieved. 
 
 6. Suffixing replaced by pre-posing: the psychological reason. 
 — So far we have spoken as if the added demonstrative 
 element necessarily followed the predicative root ; and this, 
 speaking generally, was the rule of primitive speeches. Their 
 flexion, derived from juxtaposition, was terminational. 
 This is the rule of early Indo-European, Ural-Altaic, and 
 almost all other ' agglutinative ' speeches. It is also the 
 rule of early Semitic (in regard to which family we are not 
 here considering the very obscure question of the further 
 phenomenon of internal vowel-change). It is not, however, 
 the rule of the Bantu family, which is prefix-agglutinating, 
 and it is a principle from which the modern analytic tongues 
 are steadily withdrawing. Terminational flexions or post- 
 positions are giving way to pre-posed pointing words, known 
 as prepositions and auxiliaries. For Bantu ngumuntu, i.e. 
 ng-mnu-ntu = ' with (the) man,' while ng-aba-ntu — ' with (the) 
 men.' In the absence of records we are not in a position to 
 decide through what evolution the Bantu languages have 
 passed. When English invariably puts zvith and the like
 
 The Development of Language 417 
 
 in the same position as the Zulu ng-, and makes these pre- 
 positions serve for flexion, we are aware, from our more 
 extensive knowledge of the history of Indo-European, that 
 this grammatical device is of late growth. It is the ' analytic ' 
 method, as opposed to the synthetic. We cannot tell that 
 the Bantu tongues have not made a somewhat similar 
 reversal of order from an older system. There is doubtless 
 no need to assume it, and there are no data to prove it, but, 
 considering the general similarity of human evolution and 
 the fundamental likeness of mental processes among all 
 races of the world, it is a theory far from improbable. It 
 would appear natural that primitive man should utter first 
 the sounds which denoted the concept upon which his mind 
 was primarily set. The relating elements would follow^ 
 That the Bantu languages should form an exception is not 
 impossible, but is open to legitimate doubt. 
 
 The tendency to pre-pose qualifying words, which is so 
 marked in the analytic stages of languages that have once 
 been synthetic, and which appears in Chinese, reveals itself 
 also in the pre-posing of the demonstrative signs. This 
 reversal of the primitive principle cannot but be due to some 
 good psychological reason, however difficult to discover. 
 
 To the primitive speaker the post-posing of the qualifica- 
 tion may have been natural for the reasons already given. 
 But it does not follow that it was the ideal method of expres- 
 sion for the listener. The aim of language in its evolution is 
 to communicate thought in the manner most sure of ready 
 and complete apprehension by the hearer, not primarily in that 
 which first suggests itself to the speaker. From the results 
 of ages of experience it becomes possible to realise the 
 normal workings of the human mind, and to adapt the 
 arrangement of words — pointers and predicatives — thereto. 
 This is not, of course, done deliberately, nor after any 
 psychological study. Speakers in all ages have perceived 
 difficulties in making themselves precisely understood ; they 
 have been conscious of slow apprehension on the part of 
 the hearer, and innumerable varieties of effort have been put 
 (B884) 2 E
 
 4 1 8 Nahiral History of L anguage 
 
 forth to secure greater readiness and fulness of comprehension. 
 From the varied effects of these efforts, intuitively recognised 
 as successful or unsuccessful, there emerges by natural 
 selection a certain ordering of the words which is psycho- 
 logically preferable. 
 
 Herbert Spencer, in his essay on " The Philosophy of 
 Style," remarks : " We have a priori reasons for believing that 
 there is some one order of words by which every proposition 
 may be more effectively expressed than by any other ; and 
 that this order is the one which presents the elements of the 
 proposition in the succession in which they may be most 
 readily put together. As, in a narrative, the events should 
 be stated in such sequence that the mind may not have to 
 go backwards and forwards in order to rightly connect them 
 ... so in every sentence, the sequence of words should be 
 that which suggests the constituents of the thought in the 
 order most convenient for building it up." He illustrates by 
 the position of the adjective in ' a black horse ' as opposed 
 to that in the French un cheval noir, arguing that, if ' horse ' 
 is pronounced first, the hearer's mind conjures up some 
 picture of a horse (probably brown) which must be corrected 
 to ' black ' by the next word ; whereas the prior mention of 
 * black ' calls up nothing but the notion of a colour, which 
 waits for the mention of the object to which it is to be fitted. 
 The mental acts may be incalculably rapid, nevertheless they 
 have to be performed. 
 
 If this is psychologically true, it is true also of the 
 evolved tendency to pre-pose deictic or ' empty ' words 
 instead of suffixing elements of the same value. The 
 mention of a thing, followed by the specification that we 
 are considering that thing as possessor (genitive) or recipient 
 (dative) or source (ablative), is open to the same objection 
 as the expression ' a horse black.' To the speaker the pro- 
 position is clear from the beginning. The problem is to 
 prevent the need of any travelling back or correction of 
 thought evoked in the hearer. With the Latin mcciivi 
 (literally vie tvitJi) there might arise in the listener some
 
 The Development of Language 419 
 
 false anticipation or preconception formed at nic, to be 
 corrected — although in an infinitesimal fraction of a moment 
 — by cum [zvitli). Thought may at the word inc have 
 entered a step, or glanced, into a side-path, from which it 
 must again withdraw. On the other hand, cum me helps 
 the mind forward on the only possible track. With the 
 utterance of cuniy or ivith, no picture is conjured up, but 
 only a sense of a certain relationship, and the mind is led 
 onward to look for the word to which that sense attaches. 
 
 7. Development of the nezver method: interaction of 
 psychology and phonetic decay ; different rates of the latter. — 
 That language has followed, or is following, this evolution 
 is highly probable. Some speeches have proceeded much 
 farther than others, but the explanation of such difference 
 of progress is not to be hastily assigned. The extent to 
 which various languages have advanced in this direction of 
 psychological adaptation is not necessarily an index to the 
 mental development of the speakers respectively. It does 
 not follow, because a language has reached the pre-posing 
 analytic stage, that those who use it are therefore more 
 advanced in thought or social civilisation than others who 
 use a terminational flexion or synthesis. The average 
 Athenian intellect of twenty-four centuries ago was at least 
 equal to the average intellect of the present-day Englishman 
 or Frenchman. If the Bantus are found to employ the 
 psychologically more easy method of pre-posing (or prothesis), 
 it is not therefore to be assumed that they would require to 
 have reached a higher stage of mentality than a Dravidian, 
 who employs post-positions. There are other considerations. 
 The rapidity of a change in linguistic method may depend 
 upon various circumstances. If the post- positional or in- 
 flexional method has for some reason — such as phonetic 
 corruption — become complicated, confused, difficult or in- 
 effective, the unconscious motives to the newer device are 
 strengthened. This has been the case with the Indo- 
 European tongues, as will be illustrated immediately. On 
 the other hand, if the post-positional method has been kept
 
 420 Natural History of Language 
 
 clear and simple, as in the most typical of the Ural-Altaic 
 speeches, inducement to change is minimised. The motives 
 which lead to change are to be found in difficulties ex- 
 perienced, not in any mere desire of change. Again, peoples 
 speaking different languages sometimes become blended over 
 the same area, as through conquest of one by the other. If 
 in such instances there is a struggle of each to understand, 
 or to make itself understood by, the other, the result will 
 be a quickening of the efforts above described. The old 
 manner is rapidly broken down in favour of one more 
 adapted to easy comprehension. Such a case has occurred 
 in Persia, where the conquering Arab, with his Semitic 
 speech, was perpetually endeavouring to comprehend the 
 native, with his Indo-European speech. The outcome of 
 the struggle was that the native Persian language met the 
 Arab half-way in vocabulary, while in respect of structure 
 it gradually abandoned synthesis for analysis. Though it 
 is easy to overrate the influence of the Norman invasion 
 upon the native English speech, there is no question that 
 it greatly accelerated the decay of English flexion and the 
 acceptance of its alternative. The speedier progress of 
 certain languages towards pre-posing and analysis is mainly 
 due to these two causes : in some cases a conflict of tongues ; 
 in other cases the difficulty or ineffectiveness into which the 
 flexional system happens to have fallen. 
 
 The latter requires some illustration. If we assume— a 
 pure assumption for the sake of such illustration — that 
 primitive Indo-European expressed a dative case (the 
 recipient) by means of the termination *-«/, an ablative 
 (the source from which) by *-^^/, and an instrumental (the 
 means by which) by *-«, the distinction implied would be 
 clear (though it might make an occasional strain upon 
 attention) so long as these terminations were fully and 
 unequivocally articulated, '^ekiw + ai, ^ekiw + ad, ekiw -f a 
 serve their respective purposes, although even at this stage 
 it is palpable that the method is scarcely ideal, if the aim 
 is ease to the ear and mind of the listener. But if in
 
 The Development of Language 421 
 
 process of time the three words, true to the amalgamating 
 tendency (which is ultimately little else than a tendency to 
 quick or economical utterance), are compressed in shape and 
 result into *ek?jdi, *ekudd, "^ekijo respectively, the discrimina- 
 tion between the first and third becomes difficult and un- 
 certain. If, in further process of time, a language like 
 Latin, with a phonetic tendency to drop certain final letters 
 — a practice habitual also to the Greeks and well known in 
 modern French — reduces all the three forms to ^ekud {equo), 
 there is no longer any distinction between dative, ablative, 
 and instrumental case. The hearer is compelled to gather 
 the intention from the context with some labour of mind 
 (even if the labour be unconscious) and with no little chance 
 of error. Endeavour to secure greater ease and precision 
 would inevitably follow corruption such as this. The 
 endeavour would be unconscious, but it would be widely 
 spread, and many speakers would light upon practically the 
 same device, which would pass (by the notorious contagion 
 of language) into the mouths of those who might not have 
 evolved so convenient an expedient for themselves. 
 
 The example here given is scarcely exceptional. In 
 both Old English and Old French the tendency of articula- 
 tion was to slur all final vowels (-<? -0 -u -e) into one and 
 the same, written as -c, but pronounced as -3. Such a 
 tendency, if permitted full play, necessarily ends in the 
 destruction of systematic declension and conjugation. 
 
 We must not, however, attribute the free use of the pre- 
 posed elements solely to such phonetic confusion. The 
 phonetic corruption would itself have received some check 
 from the practical necessity of remaining intelligible. The 
 truth is that the two tendencies interwork and mutually 
 support each other. On the one side, the psychological 
 motive was prompting to the substitution of prepositions, 
 pre-posed personal pronouns, and auxiliaries ; on the other, 
 the phonetic confusion of ending was increasing that motive 
 and accelerating its operations. These pre-posed elements 
 did not require invention or new creation. The means for
 
 42 2 Natural History of Language 
 
 a new system were within reach. Adverbs with meanings 
 such as ' under,' ' withal,' ' down,' ' out,' ' away ' (repre- 
 sented in Greek by, e.g., viro, avv, Kara, e/c(?), airo and in 
 Latin by, e.g., sul?, cum, de, ex, ab) were already in use in 
 synthetic sentences. They ' governed ' no nouns, but 
 qualified the whole predication. Thus, in an early form 
 of such sentences, one might say * He came, and they came 
 zvith{al)! The position of with was free {e.g. ' and, with, 
 they came '). Very early Greek, for example, might say oi 
 Se r)\dov ^vv, or ol 8e ^vv rfkOov, or ^vv he rfkOov eKetvot. 
 Similarly, ' I threw away,' or ' I away threw,' or ' away I 
 threw.' A sentence in a primitive stage of Greek might be 
 formed after this pattern : ' as-regards-a-horse (or for-the- 
 horse), they came with,' or ' they came with, as-regards-a- 
 horse,' or ' they with came, etc' Put into Greek words 
 the sentence is ittttg) (a dative case of the thing advantaged 
 or affected) ^vv ^\6ov {hippoi ksun elthofi), or rfkOov ^vv 
 'iTTTTw, or ^vv TjXdov iTTTTQ). So * I threw away, starting- 
 from-a-bag,' or ' I away threw, etc.,' might be represented 
 by, e.g., uTTo e^akov atiKKov [apo ebalori sakkou), or ctukkov 
 CLTTO e/3aXov, or ejBaXov inro ctukkov, where ctukkov is inflected 
 so as to signify the point of departure. Originally, in an 
 example like the last, it is the inflexion of the noun which 
 indicates departure, and e/3aXov ctukkov {ebalon sakkou) would 
 be sufficient in that sense. But the demonstrative or point- 
 ing word (here an adverb ' away ') assisted the compre- 
 hension of the general meaning ; it helped and determined 
 the sense intended in the inflexion of the noun. Meanwhile, 
 in point of position the adverb very often immediately 
 followed or preceded the noun, although originally it in no 
 way determined the inflexion itself Thus ^vv Xirirui or 
 'iTTircp ^vv (ksun Jiippoi or hippoi ksuti), utto ctukkov or ctukkov 
 uTTo, and the like, came to be frequent combinations. The 
 closeness of such position, and the fact that the separate 
 pointing-word was assisting the meaning of the inflexion, 
 caused the adverb to be regarded as actually an appurten- 
 ance of the noun. Since, again, the psychological principle
 
 The Development of Language 423 
 
 above described was leading to the utterance of the relational 
 or pointing word before, rather than after, the utterance 
 of the word whose relations it demonstrated, it became 
 customary to say ^vv lirirw rather than Xttitw ^vv, airo 
 (TUKKov rather than auKKov utto. The same phenomenon, 
 and for the same reasons, occurred in Latin, and if we find 
 a solitary example of post-position surviving in the case of 
 personal pronouns with cum {mecum, tecuniy vobiscuvt, etc.), 
 that instance merely serves to indicate the former freedom 
 of the help -word and also the subsequent almost total 
 restriction of that word to one position. These steps have 
 converted what were once adverbs into prepositions, which 
 have come to be regarded as ' governing ' certain cases of 
 nouns. The greater clearness, and the superior psychological 
 fitness, secured by this development naturally led to an in- 
 creasing use of such 'adverbs turned prepositions,' till 
 inflected cases — which were once treated as sufficient in 
 themselves — were regularly introduced by them. 
 
 Some noun-cases, as has been stated, had become so 
 obscured or confused in pronunciation that the help was 
 necessary, if due discrimination of sense was to be maintained. 
 On the other hand, when prepositions came to be thus 
 regularly used in introducing the nouns, it is obvious that 
 the different inflexions were of secondary importance, or, 
 from a practical point of view, of no importance at all. 
 Their presence or absence in most cases made no difference 
 to the clearness of expression, and there was little motive 
 for checking the universal tendency to phonetic economy. 
 If in ad Roinam, and de Roma the preposition told all that 
 was required, there was a removal of the cause which had 
 once kept -am and -a distinct in articulation. It thus 
 appears that we cannot judicially speak either of phonetic 
 decay as having produced the analytical method as a remedy, 
 nor yet of the growth of the analytical method as having 
 produced the phonetic decay. The truth is that there is an 
 interaction of the two processes, which began independently 
 of each other.
 
 424 Natural History of Language 
 
 What has been said of noun-flexion and its decay in 
 favour of the prepositional method, may be said also of the 
 decay of verb -flexion and the substitution of pre -posed 
 pronouns and auxiliaries. If the purely analytical develop- 
 ment has not been so complete in this case, it is not because 
 the psychological motive thereto is wanting or inoperative, 
 but because it has a larger field to cover and more material 
 in the shape of a damnosa hereditas to dispose of The 
 same phonetic corruption which brought the noun-flexions 
 so near to each other and made it easy to ignore such 
 comparatively small differences as were left, has played its 
 part upon the verbs also ; but the endings of person and 
 tense were more widely differentiated than those of case, 
 and phonetic erosion pure and simple would require a 
 considerably longer period to bring the various forms within 
 the same danger of confusion. Consequently the psycho- 
 logical tendency has not received the same impetus from 
 this quarter. In the English past tense had, the same form 
 (with the exception of the almost disused second person 
 singular hadsf) is used for all persons. The pre-posing of 
 the pronouns has made this possible. The same preposing 
 might suffice for French, yet French has brought only the 
 three persons of the singular and the third plural to identity. 
 In f avals, tu avals, II avalt, lis avalent (in which, apart from 
 liaison, we may ignore the spelling) there is one and the 
 same articulation of the verb ( = cbvc). But in nous avians, 
 vous avlez the flexions remain distinct. Despite their 
 practical uselessness, they persist. The explanation of 
 these facts is that which has been given. Take, for instance, 
 the English had. If we go back to Gothic (which is not a 
 direct ancestor of English, but which shows earlier flexion 
 in a peculiarly distinct form for Teutonic) we find the tense 
 proceeding thus — 
 
 Jiabalda 
 
 habaidcs 
 
 habalda 
 
 habaldedum
 
 The Development of Language 425 
 
 habaidedutJi 
 habaidediin 
 
 In Old English the forms analogous to these have been 
 phonetically compressed and their terminations slurred into 
 
 haefd'e 
 
 haefdesit) 
 
 haefd'e 
 
 haefdon 
 
 liaefdon 
 
 haefdon 
 
 The subsequent conversion of -on into -en, thence into -r, 
 thence into -e mute, and then into the non-existent, represents 
 a number of easy phonetic steps, which there was nothing to 
 retard, since the pronouns /, he, etc., were found entirely 
 adequate for differentiation. But haefdest, according to 
 English phonetic tendencies, was not liable to the same 
 rapid corruption. The phonetic slurring of a final -st is no 
 necessary tendency of those who slur a final -n. Had there 
 been any such tendency, the second person singular would 
 have taken the shape of had like the rest. For French we 
 have to begin with the Latin 
 
 habebam 
 
 habebas 
 
 habebat 
 
 habebdmus 
 
 habcbdtis 
 
 habebant 
 
 It will be noted that in four of these the accent falls 
 upon the e, in two of them it is upon the next syllable. 
 Any purely phonetic corruption which takes place will 
 affect terminal letters and unaccented syllables ; it is the 
 accented syllable which is the last to be influenced, and it is 
 never so influenced that it disappears. Hence the identity 
 of pronunciation which has resulted in French in four of the
 
 426 Natjiral History of Language 
 
 persons, as against the different aspect of the two others. 
 Had the accent fallen throughout upon the e, the French of 
 to-day would probably {liaison apart) have been employing 
 one pronunciation (whatever they might write) for all the 
 persons of the tense. It is no feeling of the need of 
 differentiation which retains the varied forms ; nor does their 
 present existence mean that the unconscious psychological 
 tendency to remove them is dead. Their retention, if it is 
 permanent, will be due to the artificial action of literary 
 education. 
 
 8. Historical pJienoniena as arguments for the prehistoric. — 
 If we once accept the doctrine, now sufficiently expounded, 
 that for ease of comprehension the pointing element is better 
 pre-posed than suffixed, it will become obvious why the use 
 of auxiliary verbs, ' will,' ' may,' ' do,' ' have,' tends to supplant 
 the use of those flexions which once expressed the same 
 sense. But it will be observed that, whereas such elements 
 as ' will ' or ' have ' once possessed independent predicative 
 values (' to be minded,' ' to possess '), they have now, as 
 auxiliaries, lost those values, and have become mere pointers, 
 help-words, the ' empty ' words of the Chinese. The same 
 is the case with ' I am going to do ' {je vais /aire). It is the 
 case also with the adverbs which have passed into preposi- 
 tions. No intrinsic meaning now attaches to ' by,' ' with,' 
 or ' at.' 
 
 In these phenomena we have palpable facts and linguistic 
 history to draw upon. What more natural than to use them 
 as arguments for the view already set forth as to the origin 
 of the older flexional forms ? It was argued that in all 
 probability a naming-root, e.g. *uoik, was first followed by 
 an independent element, e.g. z, with a meaning of its own ; 
 that gradually this substantial meaning disappeared (as in 
 the Chinese mtn tsi lik, ' people s power,' where tsi originally 
 meant ' place ') and the element i became merely demonstra- 
 tive ; and that finally it was amalgamated with the previous 
 word (like faurai = ego liabere habed) and so passed into 
 flexion. The only difference between that process in
 
 The Developmoit of Langtiage 427 
 
 linguistic history and the process called analysis — -with 
 its prepositions and pre- posed auxiliaries — is that in the 
 former case the pointers follow, in the latter they precede. 
 And it has been contended that, while the former manner 
 may be the more ready to the speaker, it is the latter which 
 is the more facile and effective for the hearer, and therefore, 
 since the evolution of speech is in the direction of easier and 
 fuller comprehension, the latter method is the natural out- 
 come of such evolution. 
 
 9. The begmnings of speech are unconnected Dionosyllables. — 
 The view that language began with unconnected mono- 
 syllables is not only to be deduced from the indications of 
 linguistic history itself It is also the almost inevitable a 
 priori hypothesis. According to a cognate science the 
 biological history of the individual from the earliest embryo 
 is an epitome of the evolutionary history of the race. His 
 ' learning to walk ' is an epitome of similar character. His 
 linguistic history should also afford some clue to the evolu- 
 tion of human speech. It happens, however, that after the 
 first efforts at articulation, when command of the organs 
 has been acquired, the child steps almost at once into the 
 full inheritance of both the vocabulary and the grammar 
 which countless generations have evolved. Nevertheless, so 
 far as the evidence goes, it supports the theory that the 
 utterance of isolated monosyllables was followed by their 
 simple juxtaposition to each other. The child passes from 
 inarticulate ' gooing,' crowing, and crying to the simplest 
 combinations of vowel and consonant, e.g. via, da, ta, mil, 
 and the like. With increasing control of the articulating 
 apparatus it achieves monosyllables with fuller contents, eg. 
 cat, tat, pufif), pJis{s), div ( = giv), dis ( = this). For a time 
 there is a noticeable want of flexibility and a failure to 
 articulate even monosyllables, when there are more than 
 three constituent sounds. Thus spoon may be uttered 
 as poon or foon, still as till, floor as foor, flowr as foivr. 
 It is found easier to repeat a simple monosyllable {papa) 
 or to pronounce two monosyllables {baby, naught)>) in
 
 428 Natural History of Language 
 
 quick succession. The earliest attempts at sentence-making 
 are juxtapositions with a grammarless shape, e.g. me go 'way ; 
 not want tea ; bad man hurt baby. It is difficult to resist 
 the belief that these gradual advances in articulation and 
 rudimentary attempts at conveying thought reproduce the 
 advances and attempts of prehistoric mankind. 
 
 It is worth while also to observe the manner in which 
 persons who are foreigners to each other, starting with 
 reciprocally unintelligible languages, make their first attempts 
 at communicating thought. The ' pidgin ' speech which is 
 first developed regularly consists of the chief naming-words, 
 accompanied by some demonstrative words and by gesture, 
 but without grammar. It distinctly recalls a ' baby language.' 
 The speakers in such circumstances are not ignorant of 
 grammatical devices of their own tongues respectively. 
 Those devices have, however, been inherited by them, while, 
 in respect of the developing ' pidgin,' they are placed in the 
 position of inventors. That is to say, they are in a position 
 somewhat analogous to that of early mankind, when language 
 was first struggling to express a thought by means of its 
 very limited material. It is not unreasonable to suppose 
 that the analogy applies also in general to the devices 
 adopted, and, so far as the argument is valid, it supports 
 the conclusions at which we have already arrived. 
 
 10. Origin of sound - groups (^roots') zvith meanings: 
 theories. — If primitive speech consisted thus of independent 
 naming-words (apparently monosyllabic), to which inde- 
 pendent deictic words were juxtaposed, the question still 
 remains, ' How did these words come to exist ? ' We may 
 suppose, for example, that *Hoq expressed the general sense 
 of utterance, while a following demonstrative or ' pointer ' 
 (or several such) determined its special application to a 
 noun or verb sense, and also its relations to the other words 
 in a sentence. But how did the group of sounds *uoq come 
 to attach itself to the notion of utterance ? And how did 
 the sound / (for example) come to express some such notion 
 as 'there'? In other words, what was the origin of
 
 The Development of Language 429 
 
 language, regarded as " articulate sounds which are habitu- 
 ally appropriated to certain notions " ? 
 
 It must be confessed that upon this subject nothing 
 very satisfactory has yet been said or is likely soon to be 
 said. The question is exceedingly far-reaching, and its 
 solution is necessarily devoid of adequate data. What has 
 been the development of human mind ? What have been 
 the relations between the progress of mental operations and 
 the development of the muscular powers and apparatus 
 employed in speech ? How did the psychological stimuli 
 find physiological response ? It is manifest that any sound 
 theorising on the subject of the origin of language must 
 wait its time until the sciences dealing with the bodily and 
 mental history of man have laid a surer basis than they 
 have done at present. So far the theories advanced have 
 been pure guesswork, although the guesses possess various 
 degrees of plausibility. 
 
 According to one suggestion speech began with the 
 imitation of non-human sounds. Primitive man is supposed 
 to have exercised in a fuller measure the faculty which is 
 found in parrots, cockatoos, lyrebirds, or magpies, and to 
 have reproduced to the best of his ability the sounds which 
 he heard around him. These would not necessarily be the 
 calls and cries of animals, and therefore ' the Bow-wow 
 theory ' is scarcely a fair or happy title to appl)-. Imitation 
 would not end with the lowing of oxen, the bleating of 
 sheep, and the quacking of ducks, nor with the snarling, 
 snoring, puffing and hissing of man or other creatures. It 
 would include such sounds as the soughing of the wind, the 
 various noises of crashing, splashing, plunging, breaking, 
 tearing, pattering, and so forth. Such ' onomatopoetic ' 
 words are still numerous in ever}- vocabulary — a number of 
 obvious examples being actually contained in the list here 
 incidentally given. 
 
 It is quite conceivable that in the first linguistic efforts 
 of early man the ox-species might be represented by a sound 
 like viu or bu, and that the same sound might stand for the
 
 430 Natural History of Language 
 
 general notion 'to low' or 'bellow.' In the latter sense 
 it would first be used in thinking of the ox itself, but it 
 would speedily pass to any other form of noise which 
 resembled that of the ox. It would be easy to offer a 
 long list of similar probabilities. Thus bd might represent 
 a sheep, or the cry of a sheep, and thence possibly be 
 transferred to any other plaintive cry. The r-r-r (pro- 
 longed r) of an angry dog might serve to denote anger, 
 not only when felt by the speaker, but when (with gestures) 
 he attempted to indicate its presence in another. It might 
 become, in short, the root signifying 'anger.' It should be 
 further remembered in this connection that, like the present- 
 day savage, primitive man, from his way of life and the 
 needs of the chase and the like, would be in the habit of 
 keenly differentiating the various sounds made by animals, 
 the differentiation extending to the cries of the same animal 
 in different moods. The bark, howl, snarl, growl and yap 
 of a dog, the mewing and purring of a cat, are readily dis- 
 tinguished by ourselves, because of our familiarity with those 
 domesticated animals. Primitive man would carry his dis- 
 crimination much further than civilised man has either the 
 need or the opportunity to carry his. The list of sounds 
 which he might thus distinguish and imitate would be 
 considerable. If used both in their primary sense and also 
 in transferred senses (such as those hinted for mu or bd or 
 r), a still more considerable range of notions might be 
 covered, however crudely. Let us suppose again that rapid 
 motion, whether of wings or of running, was associated in 
 the mind of prehistoric man with a sound of skirring or 
 scurrying or the like. To express such rapid motion he 
 might — the instance is invented solely for illustration — 
 employ the sound skr or kr (probably prolonging the r). 
 From such a use would arise the ' predicate root ' skr or kr 
 ' move fast,' 'run.' 
 
 Though this theory is not here being pressed, it is 
 no answer to it to urge that, since the sounds of 
 animals and of inanimate nature are much the same all
 
 The Development of Language 431 
 
 the world over, the roots of human speech would also be 
 much the same, and therefore the different families of 
 laneuasres should show a closer fundamental resemblance of 
 vocabulary than is actually the case. There are several 
 considerations which weaken that argument. In the first 
 place, though the sounds of animals and of nature may be 
 much the same, it is not true that they are apprehended in 
 precisely the same way by different ears, or that imitations 
 of them take the same shape in different mouths. Neither 
 the animal cries nor the sounds of nature are articulate, and 
 there is abundant room for diversity in reproducing them. 
 To some ears the cry of a lamb is bd, to others it is mix. It 
 is in reality neither, since sheep do not articulate with either 
 a human in or a human b. To the Roman poet a frog 
 seemed to say something like sub aqua, to the Greek 
 comedian he said koo^ {ko-dsh). The same comedian 
 makes a young pig cry kol (i.e. ko-ee or kwee). Most of 
 those who imitate the p>g would use no k, whatever else 
 they might substitute. The inarticulate sound commonly 
 represented in English as puff-puff is heard by many as 
 }w-Jl9. To some ears the sound of rapid motion would 
 not be skr, but nr or something equally remote. In the 
 second place it is not strictly true that the sounds taken 
 from animals would be the same throughout the world. 
 Different families of language doubtless developed in 
 different quarters of the globe, and there was great dis- 
 similarity in the fauna of the various regions. The bow- 
 wow of the dog (which, it ought to be observed, is regarded 
 by some as wozv-ivotv) might be tolerably general, but the 
 birds and beasts from which sounds were imitated were by 
 no means commonly the same in species or in cry. In the 
 third place, there was extensive variety of choice as to the 
 particular onomatopoetic ' root ' which should be employed 
 in a particular sense. If, for instance, ;;/;/ (derived from the 
 ox) is one possible root for ' bellow,' and thence for roaring 
 or reverberating noise, an equally possible root might be 
 derived from the roaring of some other animal, and ru, hii,
 
 432 Natural History of Language 
 
 Tir are among easily conceivable shapes. If one family 
 chanced to derive its word for anger from r-r-r (r) of a dog, 
 another might have chanced meanwhile upon the s-s of the 
 snake. Fourth, even if the roots had been once in a large 
 measure phonetically similar (which we obviously need not 
 grant), the various phonetic tendencies of the various families 
 of speech, acting through countless generations, would in- 
 evitably corrupt them to such an extent that their re- 
 semblance would in most cases become entirely obliterated. 
 A root which began as kr in several families of language 
 alike, might easily appear, when we first meet with it in 
 family A, as s/, while in family B it might have taken the 
 shape of gr. 
 
 It has seemed fair to treat the onomatopoetic theory at 
 this length, partly on its merits as being at least as plausible 
 as any other, and partly because it is a palpable fact that 
 many words have been, and are constantly being, formed in 
 this manner. A process which may be observed at all 
 discernible times of linguistic history, and which instinctively 
 repeats itself in unlettered speech, has assuredly a claim to 
 be considered as a highly probable process for the times 
 which are beyond discovery. 
 
 Yet, even if we allow that numerous primitive roots may 
 have found birth in this manner, it would be absurd to 
 suppose that the human species uttered no sounds of its 
 own until it began to imitate those of other creatures. 
 That dogs, oxen, and birds should produce cries and calls, 
 such as they are, while man was incapable of the same 
 action for similar purposes or from similar motives until 
 taught by these, is contrary to reason. The ' gooing,' wail- 
 ing, and crowing of infants are once more nature's reproduc- 
 tion in the individual of a stage in the infancy of the race. 
 It is rather to be assumed that, from a superiority of 
 physical conformation in respect of vocal potentialities, man 
 would produce a greater variety of such utterances and in 
 a more articulate shape. 
 
 The genesis of such sounds, and their relations with
 
 The Development of Language 433 
 
 mental operations, are — as has been already maintained — 
 scarcely capable as yet of discussion upon a scientific basis. 
 In all probability the earliest phase to which we can reduce 
 them is that of physical reflex of a physical sensation. 
 According to Darwin {Expression of the Emotions) there is, 
 for example, a purely physiological reason for the production 
 of a certain sound as indicative of disgust. From the same 
 causes which beget other instinctive actions in the cause of 
 self-preservation, there would occur in this case an expulsive 
 or rejecting movement by the breath. That is to say, the 
 tendency would be to blow out of the mouth or nostrils, the 
 result being such sounds as are commonly represented by 
 pooh, pish. Similarly, a shudder is accompanied by hti/i. 
 Pain produces a contraction of the muscles, and such con- 
 traction about the oral passage affects the breath-current in 
 such a way that ah ! results. The ludicrous produces an 
 obvious physical effect upon the facial and other muscles, 
 not only without our volition but often in spite of it, and 
 there commonly follows an involuntary ejaculation of the 
 nature of ha ! ha ! or he ! he ! The notion that language 
 arose from the gradual converting of such sounds into 
 significant ' roots ' has been dubbed ' the Pooh-pooh theory." 
 Though some contributions from this quarter should not 
 be denied, it cannot be said to possess the plausibility 
 of the onomatopoetic theory. Such ejaculatory sounds 
 would be so few that it is hard to see how they could 
 ever produce from themselves even the most rudimentary 
 vocabulary. 
 
 There is, however, another theory which deserves more 
 consideration. In this case we are not concerned with the 
 mere reflex from physical sensations, but with a more 
 obvious interplay of psychological and physical action. It 
 assumes deliberate effort on the part of the vocal muscles to 
 adapt their action to some picture in the mind. It has been 
 already illustrated incidentally in the case of an imaginary 
 root tn (' pull tight '). Those who have set themselves to 
 pull with all their strength have perhaps, along with 
 (B884) 2 F
 
 434 Natural History of Language 
 
 muscular tension, set their mouths in a particular manner. 
 They have first placed the tongue against the teeth ; then, 
 as all the muscles pull backward, the tongue also draws 
 back along the palate, but the nasal passage is compelled 
 to open. Experiment, it has been claimed, will show 
 that this spontaneous operation is eminently natural. 
 If any emission of breath occurred during this strain its 
 phonetic value would be tn. So far the production of the 
 sound -group would possess no psychological value. It 
 would simply be a physical concomitant or outcome of 
 the action. But through frequent repetition of the process, 
 and as an impression left by it, such action of the tongue 
 might come to be associated in the mind — however un- 
 consciously — with the notion of putting forth strength in 
 pulling. Thence, whenever the notion itself arose in the 
 mind, even without the action, there might be a tendency 
 for the tongue, by an act of reminiscence, to respond in the 
 manner which belonged to the action. Any endeavour to 
 express the notion of ' pulling ' would in the first instance 
 be by appropriate gesture accompanied by the tongue-action, 
 and generally by the audible sound-group, described. If 
 this procedure — begotten by identical natural causes — was 
 shared by a considerable number of individuals, the sound 
 tn would come to be, in itself, associated with pulling. 
 Linguistic uses being highly contagious, that sound would 
 become general in connection with the gesture, and next, in 
 the absence of gesture, would arrive at independent signifi- 
 cance, as a full predicative root conveying the sense of ' pull.' 
 Standing firm, with tense muscular effort, might similarly 
 be accompanied by a planting of the tongue against the 
 upper gums in a manner from which st might result. 
 Although this, we may acknowledge, is guesswork, it at 
 least deserves to be considered as a not unlikely source of 
 many elemental roots. 
 
 1 1 . Word - biiildirig inexhaustible from old material : 
 creation of new roots not provable. — Language, then, first 
 reached the stage of possessing both predicative elements
 
 The Development of Language 435 
 
 and deictic elements ; second, that of combining these. 
 But not only might the deictic element, or a number of 
 such elements, become attached to the naming-root ; it was 
 also possible for two or more naming-roots, or for two or 
 more fully formed and inflected words, to combine with each 
 other into a compound, as in Se(/u.)(r7ror7;?, landscape, birthday, 
 calefacio. It is manifest that the possibilities of creating 
 new words, and so enlarging the vocabulary, were practically 
 unlimited. Such new creation has gone on at all times, and 
 is proceeding every day ; but in the overwhelming majority 
 of cases, it is simply creation out of old material. Doubtless 
 at all times the stock of words in a given language was 
 increased by borrowings from other tongues, whether akin 
 or alien. Some words of very early Indo-European were 
 almost certainly taken from Ural-Altaic neighbours, whether 
 through trade with them, or through conquest over (or by) 
 them. The Sanskrit vocabulary accepted a number of 
 Dravidian vocables, and vice vej^sa. The proceeding is still 
 in as full life as ever it enjoyed. Yet in all such circum- 
 stances the material itself needs no creating ; it exists 
 already. All that is done by the borrower is to adapt the 
 borrowed word to his own pronunciation, grammar, and 
 special denotation. If in modern times we ' invent ' 
 scientific or philosophic terms, we do so either out of Greek 
 and Latin materials, or, as is the custom in German, by 
 means of new compounds made from words already in the 
 language. 
 
 Is there, however, no such thing as absolutely original 
 invention of a root or word itself? That is to say, have we 
 ceased to exercise the creative faculty which must have 
 been employed by primitive man ? It is customary in 
 answer to quote the word ' gas ' as the single example of a 
 word known to have been deliberately invented. It may 
 be remarked, in passing, that the case is not parallel, since 
 primitive man did not deliberately ' invent' a root to express 
 a distinct idea. The root evolved itself by natural steps in 
 some such manner as those already described. Do we
 
 43 6 Natural History of Language 
 
 evolve, or have we in comparatively recent times evolved, 
 roots or words in the same manner ? The evidence points 
 to the negative. Spontaneous creation appears to be out 
 of the question. Necessity is the mother of invention here 
 as elsewhere, and there is no necessity to invent, when the 
 scope for analogical creation is infinite. The vulgar and 
 ignorant have no new conceptions to express, while the 
 cultured know where to look for the most adequate materials 
 of expression. It is an error to cite the inventions of 
 contemporary slang. These are never pure ' creations.' 
 It will be found that they always depend upon combinations 
 of sound already existing. They may be curtailments of 
 standard words (as in mob, bus, cute), or applications of 
 proper names (e.g. boycott), or humorous distortions, sometimes 
 deliberate, sometimes accidental, of standard words or 
 phrases (e.g. 7iary = never a), or they may be conflations of 
 two words of cognate sense, due to the mind vacillating and 
 ending in confusion (sometimes wilfully) between them. 
 Thus from boovi and hoist may be produced boost. A 
 word like gertyjjiander is the outcome of the proper name 
 Gerry conflated with salamander by a distinct effort in 
 humorous manufacture. Some (e.g. vamose) are foreign 
 terms taken over and generally corrupted beyond recognition. 
 For the most part such ' words ' have but a transient vogue ; 
 but a few are persistent, and ultimately become part of the 
 accepted language. If we are not always able to trace their 
 origin, it need nevertheless scarcely be doubted that, if dis- 
 covered, it would prove to be of some such nature as in the 
 instances of which we know the history. 
 
 Apart from words which we know or believe to have begun 
 in such slang, there does exist, as Paul remarks, a large 
 number of words " which can neither be referred to roots of 
 the original language, nor yet proved to be borrowed from 
 foreign tongues." But this fact need not, as he argues, 
 prove the persistence of spontaneous invention. The word 
 dog apparently arose in England, whence it has been to 
 some extent borrowed by Continental peoples. It has
 
 The Developynent of Language 437 
 
 supplanted the earlier hound in all but one or two specialised 
 uses. Yet no trace can be discovered of the source from 
 which it sprang into use. Since, however, it appears to be 
 the case that the word was first applied to a particular 
 breed of the animal, its genesis would naturally be local, 
 and probably due absolutely to an individual. The name 
 would spread with the spread of the breed, which carried 
 the title with it. Who can be sure that, in such circum- 
 stances, the word was anything more than the corrupted 
 name of a place, or of the first conspicuous owner, or the 
 actual ' calling-name,' applied by some abbreviation or some 
 childish mispronunciation, to the first recognised animal of 
 the kind. This suggestion is, of course, the merest guess, 
 but there is nothing in it contrary to experience. It would 
 be impossible to guess, if we did not actually know, why a 
 certain bird which was once a pie is called a magpie, and 
 thence by children niaggy or mag. A little parrot is often 
 known in the Australian country districts as a keet, and few 
 would guess that a word which once meant ' Little Peter ' 
 had passed through perroqnet and parrakeet to the mere 
 termination of the latter. On the whole, therefore, though 
 we are in no position to deny positively the continuance of 
 spontaneous invention of words or roots, the evidence in its 
 favour is so slight as to render it extremely improbable. 
 The only words (if they are so accepted) which appear to be 
 undeniably created with new circumstances are the imitations 
 of sounds, as with the tuff-tuff o{ the automobile. But for 
 these there are few occasions, and the results seldom grow 
 into standard words. 
 
 I 2. Meaning of ' creation ' of new words. — ' Creation ' of 
 words is, and in all discoverable linguistic history has been, 
 simply building. The existing material of roots and suffixes 
 or prefixes is capable of an incalculable number of new com- 
 binations. The aim of language in its evolution is to 
 express every possible notion in distinction from every other. 
 Different individuals, feeling the lack of a satisfactory word 
 — whether the lack be due to their own limited vocabulary
 
 438 Natural History of Language 
 
 or to the absolute non-existence of such a word in the 
 language — attempt different manners of supplying it. But 
 in each case they will take existing materials and construct 
 their word from them. The difference of the results will 
 depend upon the difference of the materials chosen, both in 
 respect of the predicative elements and of the defining or 
 demonstrative elements, which in the Indo-European tongues 
 are commonly suffixes. This difference may occur in naming 
 an object, an action, or an abstract notion. It may be con- 
 ditioned by the various aspects under which a thing or 
 phenomenon happens to be regarded by various persons, or 
 by the various suggestions which it calls up in their minds. 
 A goldfincJi was so called by those who were struck by its 
 colour ; while in French it was styled a chardonneret, or 
 ' little thistle-bird,' by those who were particularly struck 
 by its fondness for that plant. To the Latins the first 
 aspect of a dcsertuin was its emptiness or abandonment by 
 mankind. To the German inventor of words it presented 
 itself as a ' sand waste.' A certain domineering attitude of 
 mind might be regarded as ' demanding too much ' {arrogans), 
 or as ' lofty ' {jiautain), or as betrayed in uplifted ' eyebrows ' 
 {superciliosus). Agreeable manners might be described as 
 those of a person ' to whom one might speak ' iaffabilis), 
 or as those of a person accustomed to the Court {courteous). 
 The notion of ' opposite ' might be expressed by ' set against ' 
 {pppositiis) or ' face to face ' {yis-a-vis). The ' master of the 
 house ' might be so expressed, as in Se(yLt)cr7roT7/?, or as ' the 
 father of the menage ' {pater familids), or as 'the keeper of 
 the loaf {hlaf-weard = lord). 
 
 To express assent one might say ' indeed ' {etiani), or ' so ' 
 {sic), or ' this ' {hoc), or ' that is it ' {hoc illud, tovt €K€ivo), or 
 ' yea ' or ' yea so ' ( =^ycs), or, according to an American turn, 
 ' that's so.' A ' kind ' person is one who has natural feeling, a 
 heiiignus was one who was ' well-natured ' {bene, gen-). There 
 is no easily conceivable limit to the possible combinations of 
 the available material, when human minds are so subtly 
 various in their processes, and amid such diversity in the
 
 The Development of Language 439 
 
 circumstances under which their respective impressions are 
 formed. 
 
 13. A language does not create true synonyms. — The 
 coinage of words being thus in easy and perpetual practice, 
 it might appear to follow almost inevitably that words are 
 sometimes created for which there is no actual need, the 
 language already containing some vocable which serves the 
 purpose, if the coiner of the new word could readily recall 
 it. Yet, upon investigation, few examples of such creation 
 can be found. The slang of the vulgar does, it is true, often 
 proceed from ignorance of the resources of the language or 
 from unreadiness in their use. But slang of this kind seldom 
 secures a permanent footing. Technical terms excepted, the 
 coinages destined to survive are those of which the want is 
 felt, not by an individual here and there, but by a number 
 approaching the majority. They represent a previous 
 deficiency in the language. The words its and selfish are 
 of comparatively recent appearance in English. Their 
 utility is self-evident. They displaced no other words, since 
 no previous words expressed precisely and only those senses. 
 
 It is a principle in the evolution of a language that it 
 has no place for absolute synonyms. Between words 
 commonly classed as synonyms there is always some 
 difference in suggestion or in area of meaning, however 
 subtle, slight, or difficult to define. The 'differentiation of 
 synonyms ' is no self-contradictory expression. Any speaker 
 who is alive to the niceties of English realises that trust- 
 worthy has taken upon itself a certain colouring which does 
 not attach to reliable, and that reliable is capable of wider 
 uses than are permitted to trustworthy. The invention of 
 reliable was due to the feeling that trustworthy scarcely 
 expressed the truth, the whole truth, and nothing but the 
 truth. The same reason has secured for reliable an increasing 
 acceptance, though objection was, and is still, taken by the 
 precisian to its formation. The same precisian has probably 
 accepted equally bold formations in the shape of laughable, 
 available, or indispensable, acknowledging that these words
 
 440 Natural History of Language 
 
 possess shades of meaning which make him welcome their 
 existence. Cicero was compelled to invent Latin words in 
 order to express the terms or concepts of Greek philosophy, 
 while Greek philosophy itself had coined words with the 
 greatest freedom, ttoctott;? {posotes) was a convenient term 
 whereby to express ' how-much-ness,' and to Latin it was a 
 distinct gain to possess a similar word quantitds. German 
 thought is perpetually active in the way of such creations. 
 But it is difficult, if not impossible, to find any such mere 
 coinage which has existed beside, much less ousted, a pre- 
 existing word of precisely the same value. 
 
 1 4. Loss of zvords not due to ' coinage ' of true synonytns : 
 causes of loss. — Nevertheless, although new coinages do not 
 evict true synonyms of older standing, and although the 
 effort of language to express all its conceptions with the 
 nicest differentiation perpetually increases the vocabulary, it 
 is a fact of elementary knowledge that in all languages words 
 have continually dropped out of use and disappeared. To- 
 gether with phonetic change, meaning change, and new 
 creation, the loss of words is one of the chief causes of 
 linguistic instability. This is apparent from the most casual 
 comparison of Anglo-Saxon with the English of Chaucer, of 
 this with the English of Shakespeare, and of the last with 
 the current English of the twentieth century ; or of the 
 Italian of Dante with the Italian of Carducci ; or of the 
 Greek of Homer with that of the fifth century B.C. The 
 process is visible in all languages which can exhibit records 
 covering any considerable length of time. It is easy, how- 
 ever, to misjudge the process by which this has occurred. 
 It is only among tongues which, like English and Persian, 
 have been in a considerable degree bilingual (whether from 
 accident of circumstances or more or less by design), that 
 one word has simply ousted an existing equivalent. But 
 this is not a case of new coinage ; it is a case of borrowing. 
 After the Norman conquest of England and the spread of 
 French culture, as after the Arab conquest of Persia and the 
 spread of Saracen culture, a writer or speaker might possess
 
 The Develop?nent of Language 441 
 
 two words — a French word and an English word, or an 
 Eranian word and an Arabic word — between which to 
 choose, or he might, thanks to his social situation, be more 
 familiar with the word introduced than with the native term. 
 In point of fact the tendency was generally to select the 
 word with the higher social connections. In many cases 
 both words were retained (as in uncoicth and strange, or roays 
 and means), with a differentiation of meaning which pro- 
 duced an enrichment of the language. In many other 
 cases the older native term was discarded in favour of the 
 imported word. Whitney quotes the substitution of despair 
 for ivanhope ; reniorse for ayenbite ; conscience for inwit. 
 Again, in the ultra-Latinising period of English or French 
 after the Revival of Learning, writers and speakers with 
 pretensions to scholarship introduced into those languages 
 many Latin words for which it can scarcely be said that 
 there was any practical need. ' Coinage ' of such words 
 consisted only in adapting the terminations to forms already 
 familiar in the borrowing languages. They are in no sense 
 original creations. Their borrowing is but a form of 
 deliberate bilingualism. Some of the words adopted were 
 valuable acquisitions ; some others were made useful by 
 differentiation of meaning side by side with the already 
 existing synonyms ; others proved abortive ; but no few of 
 them flourished at the expense of the native terms which 
 they superseded. 
 
 Losses of words from these causes are of an obvious 
 nature and require no further explanation. More complex 
 is the question why, when there has been no such bilingualism, 
 words once in full use have nevertheless disappeared and 
 been supplanted by others. The current Latin for ' fire ' 
 was once ignis, but the word had entirely ceased to be 
 employed in the sermo plebeius before the Romance tongues 
 sprang from it. The same fate befell equus (' horse '), urbs 
 (' city '), ensis (' sword '). Sometimes a word, otherwise en- 
 tirely obsolete, lingers only in some one special phrase or 
 compound ; as in the English Jinck-ster, chap-man, nec-dle,
 
 442 
 
 Nahtral History of Language 
 
 woe worth the day. The area once occupied by such words 
 is now occupied by others. 
 
 If we begin with the common Indo-European vocabulary 
 as possessed by the ancestors of the various branches, we 
 may denote its contents by x. At this vocabulary {pc) we 
 arrive by induction, i.e. from the evidence gathered from a 
 collection and comparison of all the material discoverable 
 in the earliest stages of the several branches. If, next, we 
 turn to those branches themselves, we shall find that each 
 has already lost a proportion of the original stock of words 
 and substituted others. One branch possesses x — a-\-w, 
 another x — b-\-y, another x—c-\-z. Take, for example, the 
 followinsr series : — 
 
 
 SAN;SKRIT 
 
 GREEK 
 
 LATIN 
 
 TEUTONIC 
 
 reg- 
 
 rdjan- 
 
 (^acriAevs) 
 
 reg-is 
 
 reiks 
 
 (only in the Celtic 
 shape) 
 
 *su-nu- or 
 
 siinu- 
 
 I'l0>j 
 
 (filius) 
 
 son 
 
 su-ip- 
 
 
 ( = su-ios) 
 
 
 
 *dhugh- 
 
 diihitar 
 
 dvydryjp 
 
 (filia) 
 
 daughter 
 
 *ghi- 
 
 himd 
 
 XdfJ-a 
 
 hiemps 
 
 (winter) 
 
 *peku 
 
 pasti 
 
 (KTvyvr/, 
 irpofiaTa) 
 
 pecu 
 
 Goth./a//w 
 
 *dru-s 
 
 drfis 
 
 Spis 
 
 (arbos) 
 
 tree 
 
 The absence from extant Greek of any form of the 
 nouns from reg- or peku, from Latin of those from su- and 
 dru-, and in Teutonic of that from gJii-, is an illustration of 
 the loss of material. At a later date, the same process con- 
 tinues within each of the branches. The word ^ckiws (Lat. 
 equus, Gk. tTTTro?) was carried into Teutonic (Gothic aihva), 
 but has left no representative in English, which uses ' horse.' 
 
 Now since in none of these and numerous other cases 
 has there been borrowing or bilingualism, and since language 
 does not create new words unless from a desire to find ex- 
 pression for a notion which no existing word expresses, how 
 comes the existing word to have been evicted ? The answer 
 is to be found in what has been said in a previous chapter 
 concerning the shifting of meanings. If a word a is long 
 current in a language, but becomes ousted by /3, which is no
 
 The Develop7iicnt of Language 443 
 
 borrowed word, but a growth of the said language itself, 
 then either a or /3 or both will be found to have undergone 
 some shifting of meaning before the supplanting could take 
 place. If the French foz'son (' fleece '), for example, represents 
 the Latin tonsidn{eni), and this has supplanted the Latin word 
 formerly current, viz. vdlus, there was first a shifting of mean- 
 ing either in tofisio (strictly ' shearing ') or in vellus or in both. 
 That is to say, the word /3 was not a gratuitous creation ex- 
 pressing the same sense which was already covered by a. 
 Sometimes (i) a and /3 existed side by side, with kindred but 
 not identical meanings, e.g. the Latin ignis (' fire ') and focus 
 ('hearth'), and the one (focus) gradually came to connote 
 and then to denote the other {ignis), until there was no 
 place for that other in use. So viagnus (' great ') was 
 supplanted hy grandis ('grown up' and so 'big') from the 
 widening of meaning in the latter word. (2) Sometimes 
 a new formation /Q may be invented to express some shade 
 of meaning as yet unexpressed by a single term, and may 
 at first threaten no danger to a, which has its own sphere ; 
 but, by a shifting in the meaning of /3, the words become 
 synonymous, and a disappears. Thus a late Latin diminutive 
 (whether purely descriptive or affectionate) of apis (' bee ') 
 took the shape of apicula (French abeille). In course of time 
 the frequent and rather undiscriminating use of the diminutive 
 caused it to lose its particular shade of meaning and to 
 become identical in sense with apis itself. The Latin word 
 filius, which has supplanted the cognate of son (Skt. sunus, 
 Gk. u£09), apparently meant ' suckling,' and therefore was no 
 idly created synonym. The loss of the earlier term is 
 assuredly due to a frequent and increasing application or 
 misapplication of the name ' suckling,' that is to say, to a 
 shifting in the sense oi fllius. 
 
 The doctrine that a language creates no true synonyms, 
 is not contradicted by the existence side by side of, e.g., Latin 
 serpens and unguis, Greek eyjb<i and o^l<^, for ' snake.' The 
 self-same snake may be regarded from various points of 
 view ; he may be the ' crawler,' the ' coiler,' or the ' gleaming-
 
 444 Natural History of Language 
 
 eyed.' When each of the several words is coined, its 
 special signification is felt and intended. If at a later date 
 the special colouring is lost from consciousness, and nothing 
 more is thought of than the general notion ' snake,' we arrive 
 at synonyms. Nevertheless the language did not create 
 them to serve as such. So an early boat of the self-same 
 formation and purpose may be the ' floater ' (a sense which 
 seems to underlie Latin ndvis and Gk. vav<i) or the ' dug- 
 out ' (as in Gk. aKd(f)o<i). It is probable that in primitive 
 Indo-European times both names were applied. The earth 
 may be regarded as the ' dry ' element in contrast with the 
 water, or as the ' firm ' in contrast with the fluid. A snail 
 is either a 'crawler' or a ' house -bearer ' (Gk. ^epe'ot/co?). 
 The house may be regarded as a ' covering ' (the probable 
 meaning of the English term), or an ' abode,' or a ' build- 
 ing.' Though a house might be called by any one of these 
 titles, the words were not synonyms so long as their original 
 meanings were realised. Yet it is manifest that such a series 
 of expressions habitually applied to the same object would 
 often end in two or more of them ceasing to be discriminated 
 in actual use. The result is a pair (or larger number) of 
 real synonyms, and, as a further consequence, one of these will 
 prevail, and the other, or others, will disappear. The true 
 statement, however, to meet such instances is not that an 
 original word a was thrust out by the invention of a 
 synonymous /3, but that, through shifting (whether widening 
 or narrowing or decoloration) of the sense of one or both, 
 the two words became synonymous, and one thrust out the 
 other. 
 
 The further obsolescence of words through the obso- 
 lescence of the thing or process which they denoted is 
 a phenomenon easy to understand. Whitney illustrates by 
 the loss of terms (e.g. jess) pertaining to hawking, when 
 hawking itself ceased to be practised. Not infrequently 
 some of the words belonging to an obsolete art or custom 
 may survive in a figurative application. Thus jovial and 
 saturnine are survivals from astrology, the lists are properly
 
 The Development of Lang2iage 445 
 
 those of the tournament. But for the most part the terms 
 disappear with their occasion. Such losses, however, count 
 for comparatively little in the development of language in 
 general. 
 
 15. Effects of the analogical instinct. — Of the greatest 
 importance in the history of language has been the force of 
 analogy. The subject is so large and is capable of such 
 extensive illustration that it deserves to be, and has been, 
 treated in monographs. Fortunately it is among the most 
 easy to comprehend ; and in a first survey it may suffice 
 to indicate the general character of the force in question. 
 Under the name of ' analogy ' are appropriately classed all 
 those phenomena of change or formation which are due to the 
 simplifying or systematising instinct. It is, for example, a 
 natural tendency of the human mind to reduce the flexion 
 for numbers, cases, tenses, and the like to one form for each 
 purpose. Varieties tend to disappear in favour of uniformity. 
 Words belonging to the same class, such as verbal abstracts, 
 incline to follow one principle of formation in place of 
 several. The position of the accent in a word also tends 
 to become regular and consistent. 
 
 The changes of language in this direction have not, 
 however, been deliberate. The minds of the speakers with 
 whom they first begin are not suffering any conscious dis- 
 pleasure at the apparently arbitrary diversity of the existing 
 signs for the plural or the genitive, or of the existing 
 manners of forming adverbs or verbal nouns, or of the exist- 
 ing incidences of the accent. If the plural of day in old 
 English ended in s (days) ; if that of ox and eye ended in n 
 (whence oxen, the poetical eyn, and the proverbial hazvks 
 een) ; if that of sheep or head took no plural ending at all 
 (whence sheep, and such expressions as tiventy head of 
 cattle') ; and if plurals formed in these and other ways (e.g. 
 cildrii,fet, i.e. feet) displayed far more variety than they dis- 
 play in modern English (in which eyes, heads, etc., have been 
 substituted), it is not to be supposed that such changes as 
 have occurred were in any way the outcome of vexation at
 
 44^ Natural History of Langiiage 
 
 such apparent absence of system, or even of conscious 
 comment upon it. If blasphemous, balcony, satellites were 
 once the recognised accentuations, while modern English 
 says bldsphenioiis, balcony, satellites, it is not because speakers 
 said to themselves, " English regularly places its accent upon 
 the first syllable, and it is right that these words should be 
 accommodated to the rule." It is rather that the speakers 
 were, for the time being at least, oblivious, if not altogether 
 ignorant, of the diversity in question. Those who first said 
 eyes or heads (and there must have been many who so spoke 
 before the use became on a fair way toward establishment) 
 were not at the moment alive to the fact that the standard 
 form was eyeji or head. To them the notion of a plural was 
 commonly associated with the sign s, and eyes, heads sprang 
 to their lips instinctively. Similarly those who first said 
 bldsphenious were led to do so, not by any desire to be con- 
 sistent, but from a long-established habit of placing the 
 stress-accent upon the earliest syllable. The psychological 
 process is precisely that which is so easily discernible in 
 children. Though they may hear around them the standard 
 forms feet, men, along with the standard forms chairs, cups, 
 cakes, etc., such ' irregular ' plurals are comparatively in- 
 frequent — that infrequency being, indeed, the only reasons 
 why they are styled ' irregular.' That, in speaking of more 
 than one thing of a kind, the sound s is commonly appended, 
 is one of the earliest facts borne in upon the consciousness 
 of the child. The exceptions, which meet the ear more 
 seldom, are realised but slowly, and sometimes only after 
 special instruction. Hence the child naturally speaks of 
 foots, mans, oxes. The same is commonly the case with a 
 foreigner, when he is acquiring a language without constant 
 reference to his grammar-book. 
 
 Similarly the child comes at a very early stage of speak- 
 ing to apprehend that in referring to past time the sound d 
 is attached to the verb. Though there are no inconsiderable 
 number of words which in this respect are ' irregular,' and 
 which the child cannot fail to have heard (e.g. ate, drank,
 
 The Develop7ne7tt of Language 447 
 
 went), yet the preponderance is altogether in favour of such 
 preterites as walked, talked, asked, told, in which d plays its 
 part. Instinctively therefore he uses goed, coined, bited, seed, 
 until he is taught better, either by further observation or by 
 direct admonition. Such innovations as these must have 
 sprung up, with or without permanent result, at all periods 
 of a language, when that language afforded similar occasions 
 for them. We now live in an age of education, which 
 deliberately protects peculiar or abnormal forms. But it is 
 manifest that, in ages and languages in which the natural 
 tendencies are not thus artificially checked, far-reaching 
 effects must have followed the systematising instinct. 
 
 So far as actual linguistic evidence goes, simplifying 
 through analogy has been in perpetual operation. Nor 
 should anything else be expected, if we concede the claim 
 already made for the evolution of language, namely, that it 
 is in the direction of least effort to both the speaker and the 
 hearer. In maturity, and after abundant observation and 
 practice, neither party to the act of conversation may be in 
 the least aware of any effort to employ or recognise a less 
 usual form of past tense or plural sign or case-ending. Yet 
 every form of correction or adjustment involves some effort, 
 however little realised. Since, as has been shown, the natural 
 tendency of childhood is towards entire uniformity, there 
 must have occurred certain acts of mental readjustment — 
 however inappreciable to consciousness — during the time 
 when seed, goed, foots \v&x& being replaced by the standard 
 saxv, went, feet. With an unfettered development of speech 
 along natural lines there would cease to be any obligation 
 to make such efforts at all. In other words, abnormalities 
 would disappear. Meanwhile, from the point of view of the 
 hearer, it is self-evident that a greater amount of mental 
 exertion is required in instantaneously recognising any one 
 of several different suffixes as denoting plurality than in 
 recognising only one of them in that function. We do not, 
 and cannot, in actual speech, follow the minute psychological 
 operations which occur when we hear the word days. We
 
 448 Natural Histo7'y of Language 
 
 are blissfully unaware of the operations at all, and are wont 
 to call them automatic. Yet it is manifest that the added 
 s converts the mental conception of a day into that of 
 several, and that one operation follows another, with whatever 
 incalculable rapidity. 
 
 It is as certain as anything can be in psychology that it 
 would be an economising of the work to be done by the 
 brain-tissues, if the invariable s were the only sound ever 
 employed to produce this particular effect. The same 
 remark is true of all other ' irregularities ' in grammar. At 
 all times, therefore, and in all languages in which perfect 
 regularity does not exist, steps towards greater uniformity 
 have been made. 
 
 Where there is great variety of form in a special function 
 (e.g. many forms of plural), and where no one form is 
 palpably predominant, the process will be slow : there is as 
 yet none so strong as specially to attract the rest. But 
 when once a particular form unmistakably preponderates 
 among common words, it tends more and more to assimilate 
 the others. This has been the case with the English plural 
 s. A similar phenomenon is conspicuous in French. In 
 the transition from Latin to modern French it happened (for 
 causes which might be assigned) to be the accusative which 
 prevailed over the other cases as the accepted form of a 
 word. But in Latin, whatever differences there may be in 
 the nominative plural, the accusative plurals (except in 
 neuter nouns) all ended in s (-as, -os, -es, -us). It followed 
 that s was the French plural-sign in the great majority of 
 words, to wit, all that were Latin masculines and feminines. 
 Against this preponderance the neuters (whose Latin plurals 
 ended in -a, and thence their old French plurals in -e) could 
 not hold their ground, and hence to them also the -s passed 
 by analogy. 
 
 We cannot, indeed, always trace the reasons for which 
 one form gained its first preponderance amid an earlier 
 variety ; but, the predominance once gained, its effect is 
 sufficiently easy to follow. Its gains are at first slow, but
 
 The Development of Language 449 
 
 their increase is apt to be somewhat like that of the pro- 
 verbial snowball. The termination -Titus in past participles 
 passive would not be regarded by a Latin scholar as par- 
 ticularly characteristic of that language. It occurs, indeed, in 
 the sufficiently common tocfitus, sccfitus, sotfttus, indfitus. But 
 it is not very clear why this particular form, rather than others 
 more common and used in at least equally familiar words, 
 acquired an assimilative power and at last converted habitus, 
 venditus, riiptus and the like into habutus, vendutiis, rumputus. 
 The beginnings of the assimilation must have been gradual, 
 but the gain of any one word obviously lends just so much 
 more strength to the influence put forth by the group, and, 
 given no counteracting principle, natural or artificial, the 
 end would be the absorption of all other forms. Brugmann 
 points out that in modern Servian and Slovenian all verbs 
 end alike in the first person singular (-w), the original 
 difference having been entirely wiped out by the influence 
 of four old Slavonic MQ.x\is,jesnii, vevii, daim,jami. 
 
 The same force of analogy which operates upon flexions, 
 operates upon the formation of words. In the spontaneous 
 creation of new words out of old material the mind is 
 influenced by existing words of similar character or function. 
 A certain word, or type of word, possesses certain associa- 
 tions. In inventing a new word with similar or analogous 
 associations we incline to give it a shape recalling that 
 already in existence. It is, in fact, this shape which 
 commonly offers itself before we think of any other. Strict 
 etymology or technically 'correct' formations are sought 
 only in the creation of technical or abstract terms by 
 persons of learning. For the most part no such discrimina- 
 tion is made, and it would be irrational to suppose that any 
 was made in pre-educational times. The English further is 
 the comparative oi forth, and the th is etymologically correct. 
 The formation of farther from far is a striking example of 
 the eff'ect of association of ideas, or analogy, although in this 
 instance a general similarity in the shape of the words must 
 have lent no inconsiderable help. The Tatin noun for 
 
 (n884) 2 G
 
 450 Natural History of Language 
 
 ' north ' was septentrio, with stem septentrioti-. The adjective 
 was formed by adding -dlis, i.e. scptentriondlis. Meanwhile 
 the noun for ' south ' was merldies. Those who created the 
 adjective meridi-ondlis were as manifestly influenced by 
 septentriondlis as the English inventor of ivitti-cisni was 
 influenced by the existence of critic-ism. In French, as 
 Henry observes, the suffix -ier added to, e.g., lait, sabot, 
 produces laitier, sabotier. Since the final t of lait, sabot is 
 inaudible, while in the interior of laitier, sabotier it is pro- 
 nounced, the mind naturally conceives of -tier as the suffix. 
 Hence from bijou the result is not bijoii-ier but bijoutier. 
 Contemporary inventors of patent drugs or washes commonly 
 attach a suffix -ine or -line, of which they could render no 
 account, except that it occurs in certain words belonging to 
 articles of the same kind. The English -ling, which serves 
 as diminutive, is an accidental instance of the attachment 
 of a more original -ing to certain words ending in -/, just 
 as the German abstract suffix -keit is due to certain words 
 in which a true element -heit was suffixed to a termination 
 in -c (e.g. viiltec-Jieit, whence iniltekcit). The same non- 
 etymological, but highly natural, manner of word-building 
 is to be found at all stages of language, so far as we are in 
 a position to analyse their morphology. In Greek a feminine 
 suffix was -la. This, attached to a stem in -n, produced 
 -avia {-a}iia) and thence -aiva {-ai?ia). Such words as 
 TeKraiva, \eacva, are thus correctly formed from the same 
 stem as the masculine reKTwv [tektoti), Xicov (leon). These 
 forms developed themselves without conscious effort of 
 thought and, of course, without etymological reflection. 
 But, when once they existed, the notion of the feminine 
 animal became associated with a termination -aiva, and in 
 consequence the feminine of \vKo<i (luko-) or u? (ku-) also 
 becomes XvKatva, vaiva, although there never was any -n 
 whatever in the stem. 
 
 This process is so entirely natural, not to say inevitable, 
 that it must have been in operation long before we possess 
 the linguistic records. We have to assume that Indo-
 
 The Development of Language 451 
 
 European, for example, had passed through a long develop- 
 ment before the date at which we can approximately 
 discern its words and forms ; we must also assume that the 
 analytical influence had been perpetually at work upon it 
 for an unknown number of generations. This consideration 
 will show how impossible is the task of reducing the entirely 
 primitive suffixes or demonstrative elements of that language 
 to a simple and definite list. It would be easy to fall into 
 the error of assuming that -ling in English was a simple 
 suffix, or that -acva was such in Greek. We may as easily 
 make the same mistake with apparently simple suffixes in 
 the Indo-European or other Ursprache. 
 
 But, while we must guard against such error, it is beyond 
 question that the general outcome of the study of analytical 
 working is to reduce primitive suffixes to simpler and 
 simpler forms — a result which is in keeping with the theory 
 that they were once mainly ejaculations accompanied by 
 deictic fjesture. 
 
 (B884) 2 G 2
 
 INDEX 
 
 a, changes of the vowel, 288 
 "a ac '00 ," 353 
 
 Abbreviation in words, instances of, 346- 
 350 
 
 Ablaut, examples of, 162, \%2,notc 
 
 Abyssinian Semitic and its branches, 172 
 
 Accadian (Shumerian or Proto-Chaldean) 
 language, 152 
 
 Accent, pitch or tonic and stress or 
 dj'namic, explained, 27, 28, 340 ; 
 accent in primitive Latin and its effects, 
 341 ; in French, 341 ; in classical and 
 modern Greek, 342, 343 ; results of, 
 344 ; in Indo-European, 344 ; in 
 English, 345 ; phonetic effects of, 
 340-346 
 
 Accentuation, Greek, 16 ; English, 20 
 
 acre, 13, 384 
 
 Aeolic dialect, where spoken, 193 
 
 aes, 263 
 
 African, South, languages and their divi- 
 sions, 138 ; Central African languages, 
 147 
 
 again, pronunciation of, 20, 287 
 
 Agglutinating languages, 76, 87, 129- 
 152 ; criterion of, 90 ; subdivisions, 
 94 
 
 Agnone, tablet of, 198 
 
 Albanian (or Illyrian) language, 195 
 
 albero (Ital. ), 332 
 
 Algonkin languages and their chief 
 divisions, 136 
 
 alms, 341 
 
 0X0701', 384 
 
 Alphabet, Greek division of, 34 nofc ; 
 classification of letters, 41 ; evolution 
 of, traced, 47-54 ; modern, whence 
 derived, 55 ; Semitic and Phoenician, 
 55-58 ; origin of the word and of the 
 naines of the letters, 58 ; Greek, 59 ; 
 table of Old Phoenician, Old Greek and 
 " Ionic " alphabets, 61 ; modifications 
 introduced by the Greeks, 62 ; Ionic, 
 
 Chalcidic and Attic alphabets, 62 ; 
 
 Latin, 63 ; table of Old Greek, Ionic, 
 
 Chalcidic and Latin alphabets, 65 ; 
 
 alphabetic family tree, 67 ; Occidental 
 
 and Oriental alphabets, 68 ; stages 
 
 in the development of writing and 
 
 alphabets, 68 
 Amalgamating (inflexional or organic) 
 
 languages, 76, 87, 158 - 227 ; two 
 
 divisions of, 93, 94 
 Ambleside, 337 
 American languages, characteristics of 
 
 native, 77, 152-157 
 Amharic language, 172 
 afxcpopeiis, 342 
 amputdre, 341 
 Analogy, its force in linguistic growth 
 
 illustrated, 445-451 
 Analytic languages, 83, 87, 93 ; pre- 
 
 posing analytic stage of language, 
 
 419-426 
 Anamese group of languages, 126 
 Anglo-Saxon and its subdivisions, 217 ; 
 
 as a racial term, 231 
 anguis, 443 
 Animals known to the original Indo- 
 
 Europeans, 264 
 dvpos, dvdpds, 42, 337 
 anser, 108, 311, 356, 368 
 ant, 350 
 
 Aphaeresis and aphesis defined, 349 note 
 apicula and apis, 443 
 Appalachian group, Cherokee, Choctaw, 
 
 Muskogee, 156 
 Arabic, peculiarities of its grammar, 85, 
 
 161 ; notes on classical and neo- 
 Arabic, 170 
 Aramaic language, notes on, 168 
 Araucanian dialects of Chili, 157 
 Aristotle quoted, 3 ; his etymologies, 364 
 dpKTOi, 263 
 
 Armenian language, notes on, 191 ; 
 diffusion of, 250 
 
 453
 
 454 
 
 Natiiral History of Language 
 
 Armorican (or Breton) language, 210 
 
 arrogans, 438 
 
 artful, 387 
 
 Articles, def. and indef. in French and 
 
 Italian, 201 
 Aryan, improper use of, 176 ; as a 
 
 branch of the Indo-European family of 
 
 languages, 184-191 
 Asoka, King, his rock-inscriptions, 186 
 Aspirate, voiced and voiceless, 35, 36, 40 
 Assimilation of neighbouring words, 
 
 instances of, 335, 349 
 Assyrian language, notes on, 169 
 Athabascan group of languages, 156 
 auberge (Teutonic), 207 
 Augment in Greek and Sanskrit, 182 
 Australian languages, 146 
 Avestic (Zend or Old Bactrian), 188 
 ayenbite, ousted by remorse, 441 
 Aztec rebus-writing, example of, 52 
 
 " Baal Lebanon" inscription, referred to, 
 
 56 
 Bactrian, Old, language, 188 
 balcony, accentuation of, 20, 345 
 Baltic (or Lettic) languages, 226 
 Baltic-Slavic and its subdivisions, 222- 
 
 226 
 bank, 369 
 
 " Bantine Table" {Tabula Dantina), 197 
 Bantu (or Kafir) languages, peculiarities 
 
 of, 79; similarities in, 114; main 
 
 groups, 138 ; peculiarities of grammar, 
 
 139 ^ 
 ^actXei's, 394 
 Basque language, peculiarities of, 78, 90, 
 
 150-152 
 bay, different meanings of, 118, 371 
 be, 355, 408 
 beast, 384 
 beech, 326, 376 
 belfry, 372 
 
 Beluchi language, 191 
 bench, 369 
 
 Bengali, where spoken, 187 
 Berber, where spoken, 174 
 bijou tier, 450 
 bishop, 349 
 blasphemous, accentuation of, 20, 345, 
 
 446 
 body, 383 
 book, 394 
 boor, 385, 386 
 boost, 436 
 Bopp, his morphological classification 
 
 of languages, 91 
 Bornu dialects about Lake Tchad, 147 
 
 Boule-Miche, 406 
 
 bound, 371 
 
 |3oOs, 105, 376 
 
 povcrTpofpTjddu, method of writing, 59 
 
 Brachycephalic stock (European), 239, 
 
 240 
 Brahui language, 136 
 Br^al on etymology, quoted, 362 
 Breathing, process of, 21 ; conversion of, 
 
 into speech, 22 
 Breton (or Armorican), 210 
 broken, 357 
 brother, 316, 322, 376 
 Ppords, 43 
 
 bucca, popular Latin for os, 200 
 Bulgarian, Old, 223 ; modern, 224 ; 
 
 where spoken, 225 
 Buriat dialects, 133 
 Burmese dialects, 126 
 bus, 351 
 
 Bushman dialects, 148 
 by and by, 388 
 
 cab, 351 
 
 caballus, popular Latin for equus, 200, 
 
 262 
 cabanne, 207 
 Caecilius, 311 
 caitiff, 391 
 call, not akin to KaXeiv, 11, 107, 359, 
 
 370 
 
 camp, 369 
 
 Canaanitic group of languages, 169 
 
 candle, 323 
 
 capital, 382 
 
 captive, 391 
 
 Case-endings, confusion of, 420 
 
 Catalonian dialect, 206 
 
 cathedral, 356 
 
 cattle, 382 
 
 Caucasian (North and South) languages. 
 149 
 
 Cecilius, 311 
 
 Celtic languages, usual division of,|209 ; 
 a better arrangement, 210 ; notes on 
 the different subdivisions, 210 ; small 
 trace of, in English, 219 ; Celtic and 
 Italic, 258 oY.^n? 
 
 Celts and Celtic discussed, 241-243 ; 
 Celtic migration into Asia Minor, 295 
 
 censure, 384 
 
 Chaldee language, 168 
 
 champ, 369 
 
 chapman, 441 
 
 chardonneret, 438 
 
 chattels, 382 
 
 Xh^i 108
 
 Index 
 
 455 
 
 Children's attempts at articulation and 
 speaking, 427 
 
 chin, 324 
 
 Chinese picture-writing, 47, 48, 49; 
 characteristics of language, 50, 52, 
 70, 84; ancient literature and language, 
 116; numerous homonyms, 117 ; how- 
 differentiated, 117; "tones" and their 
 explanation, 118-120; " roots," gram- 
 mar and position of words, 121 ; 
 "empty" or relational words, 122; 
 dialectical differences, 123 ; original 
 character of the language, 124-126 
 
 Cholmoiideley, pronunciation of, 339 
 
 church, its etymology, 346 
 
 churl, 385, 386 
 
 Cingalese (or Elu) language, 135 
 
 "ng^ 359 
 
 " Cippus Abellanus," with Oscan in- 
 scription, 198 
 
 Cirencester, pronunciation of, 339 
 
 Classification of vowel-sounds, 30 ; table 
 of classified sounds, 41 ; of languages 
 genealogical, 27, 95-114; morpho- 
 logical, 72, 74-94 ; of races, 235 
 
 Climate, influence of, on speech, 296, 
 
 331 
 Codex Argenteus, 213 
 codify, 14 
 
 " cooee," Australian cry referred to, 42 
 Colinderies, 406 
 
 come, akin to fSaivu}, vcnio, 108, 355 
 comvidrage, 387 
 commere, 387 
 "Comparative," remarks on the word 
 
 as applied to a science, 7 ; as applied 
 
 to language, 8 
 
 complexion, 391 
 congregation, 392 
 Consonants, how produced, 26 ; old 
 
 definition not accurate, 33, 35 ; their 
 
 classification, 35-41 ; omission of final, 
 
 348 
 Coptic and its dialects, 174 
 coquina, 312 
 cog no, 360 
 
 cordonnicr, etymology of, 12 
 cordwaincr, 372 
 Corean language, 149 
 Cornish language, 209, 210 
 corpse, 383 
 Corruption, linguistic, causes of, 333-354. 
 
 See under Phonetic tendencies 
 count, 369 
 coiv, 105, 376 
 
 '■''<^/0'. 387 
 Crevinian dialect, 132 
 
 criticism, 450 
 
 Cuneiform writing, origin of, 55 note ; 
 
 language, 189 
 cunning, 387 
 Cymric language, 209, 210 
 
 daft, 387 
 
 Dakota group of the Sioux, etc., 156 
 
 8dKpv, 356 
 
 Danish language, 215 
 
 Dansk (or old Norse), 215 
 
 odweSov, 405 
 
 5a(pvr](pop4(i}, 405 
 
 Baprds, 357 
 
 daughter, 261 
 
 deal, 371 
 
 deck, 14, 312 
 
 deer, 375, 381 
 
 Deictic or demonstrative elements of 
 
 language, 407-411, 415 
 Sefxas, Homeric use of, 412 
 Dentals, 38 
 deawdTrii, 405 
 deuil, 371 
 
 deum, an older form than deorum, 5 
 deus, not identical with ^e6s, 11 107, 
 
 267, 315. 359 
 Dialects, growth of, 278-282 
 5t/c7?, 364 
 dime, 327 
 
 Diphthongs, how produced, 32 
 disease, 384 
 dish, 323 
 dojff, 406 
 
 dog, origin of word unknown, 436 
 dole, 371 
 Dolichocephalic stock (European), 239, 
 
 240 
 domina, 380 
 doiiius, 265 
 don, 406 
 donna, 380, 387 
 door, 265 
 
 Doric dialect, where spoken, 193 
 dod-qa6jj.fda, elements in, 181 
 " Doublets," instances of in English, 394 
 BovvaL, analysed, 181 
 dout, 406 
 Dravidian languages, 89 ; divisions and 
 
 subdivisions, 134-136 ; peculiarities of 
 
 grammar, 136-138 
 dropsy, 349 
 8pvs, 264, 327, 376 
 dun no, 351, 405 
 Dutch language, 216 
 
 '•'7''. 355
 
 456 
 
 Natural History of Language 
 
 Economy of exertion the cardinal principle 
 of phonetic change, 338 
 
 dciirie, 363 
 
 dci/yer, etymology of, 12, 263 
 
 Egyptian language, ancient, notes on, 
 174 ; methods of writing, 56 ; table 
 of families of alphabets from Egyptian 
 source, 67 
 
 eiytit, an instance of phonetic corruption, 
 179 
 
 emius, emptus, 42 
 
 English language, how classified, 97, 
 212 ; how formed, 218 ; periods of, 
 219 
 
 Epiglottis, described, 24 
 
 equivocation, 391 
 
 equies, 200, 262, 442 
 
 Eranian (or Iranian) and its subdivisions, 
 188-191 
 
 Erse (Irish-Gaelic), 210 
 
 Eskimo (or Innuit) language, 156 
 
 ^(TT-qaa, 312 
 
 Esthonian language, 132 
 
 Ethiopian language and its divisions, 175 
 
 Etymology, not a division of comparative 
 philology, 14 ; Voltaire's sarcastic 
 definition of, 12, 366 ; as defined in 
 the New English Dictionary, 14 ; a 
 sound etymology deals with sounds, 
 not with sound, 358, 366-368 
 
 Etymologies, examples of mistaken, 11, 
 364; application of phonology to, 355- 
 372 : great changes in words from 
 their primitive shapes, with instances, 
 355-357 ; regularity of phonetic laws, 
 with instances, 357-360 ; intermediate 
 steps and use of hypothesis in, 361- 
 363 ; superficial resemblance un- 
 scientific and erroneous, 364-366 ; 
 must be based on phonetic principles, 
 366-368 ; phonetic phenomena affect- 
 ing, 368-372 
 
 " Eugubine Tables," 198 
 
 Eulalia, St., Romance poem on, 207 
 
 Euskarian (or Basque) dialects, 150-152 
 
 ewe, 355, 376 
 
 Exaggerated use of words, 388 
 
 exemplum, 337 
 
 fabnquer, 369 
 
 fagus, 264, 265 note, 326, 376 
 fame, 384 
 faiiher, 449 
 father, 316 
 feather, 326 
 
 Felup dialects on the Gambia, etc., 147 
 fern me, 374 
 
 fend, 335 
 
 fermarc, fenncr, 385 
 ferre=fer-se, 179 
 feu, 367 
 
 Fciter, 367 
 
 Fijian dialects, 142, 144 
 filius, 443 
 
 Finnio-Ugric group of languages, and 
 
 its divisions, 131-133 
 fir, 264 
 frmdre, 385 
 fsh, 324 
 fishwife, 374 
 
 fi-'e, 355. 359 
 
 Flower, his distribution of races, 235 
 foc'sle, 335 
 focus, popular Latin for ignis, 367, 
 
 443 
 foedus, foidos, 313 
 foot, 320, 322 
 fores, 265 
 forger, 369 
 four, 356 
 fowl, 384 
 fr actus, 357 
 
 Frankish language, 216 
 frate, 2,77 
 f rater, 316 
 
 French language, historical account of, 
 
 206 ; Celtic and Teutonic words in, 
 
 207 ; early dialects of the langiie d'oil, 
 207 ; dialect of the lie de France 
 becomes the standard, 208 ; old 
 French, 208 ; plural in French words, 
 448 
 
 fraie, 349 
 frcre, 377 
 
 Fricatives or spirants, 37 
 Frisian language, 217 
 fuiido, 410 
 further, 449 
 
 Future tense, elements of its formation, 
 413 ; in French, 181, 201, 413 ; in 
 Latin, 258, 413 ; in English, loi 
 
 Gaelic, Scotch, 211 ; Irish, 210 
 
 Galician (Spanish dialect), 206 
 
 gall, 326, 356 
 
 Gallic (Old Gaulish), 209 
 
 Gans, 311, 320, 356 
 
 gas, 435 
 
 Gathas, 188 
 
 Genealogical classification of languages, 
 
 72, 95-114 
 German language, 212 ; West German 
 
 languages (Low), 216-221 ; High, old, 
 
 middle, and modern, 221 
 
 /|^*
 
 Index 
 
 457 
 
 Gerrymander, a humorous formation, 
 
 436 
 Gesture, use of, as an auxiliary of 
 
 language, 403 
 Ghe'ez language, 172 
 ghost, 384 
 
 ,:?2«. 349 
 
 Gipsy (or Romany) dialects, 188 
 
 girl, 382 
 
 Glides, or transition-sounds, 41 ; in- 
 stances, 42, 337, 350 
 
 Glottis, the, described, 24 
 
 Glottology, a suitable name for the 
 science of language, 6, 10 
 
 gold, pronunciation of, 20 
 
 goldfinch, 438 
 
 Gond language, 136 
 
 goodbye, 84, 406 
 
 goose, 108, 311, 320, 356 
 
 gossip, 350, 385, 387 
 
 Gothic language, notes on, 213; im- 
 perfect tense in, 424 
 
 Grammar, comparative, use of the term, 
 6 ; object of grammar, descriptive and 
 historical, 8 ; as a criterion for genea- 
 logical classification, 101-105 
 
 grandis, 443 
 
 Greek, see Hellenic 
 
 Grimm, his morphological classification 
 of languages, 9 
 
 Grimm's Law, 318 ; applied to Latin 
 and Greek, 319 ; to Teutonic languages 
 (First Teutonic Sound-Shifting), 319 ; 
 illustrative examples, in Latin, Greek, 
 and English, 320 ; ' ' Second" or ' ' High 
 German" shifting, 321 ; modified state- 
 ment of the law, wuh examples, 322 ; 
 sub-laws and other modifications as 
 applied to English, 323 ; Verner's law 
 for the tenues, 323 ; considerations in 
 applying the law to English, 324 ; 
 method of applying, with examples, 
 325-327 ; full tabular statement of, 
 328 
 
 giterre, 207 
 
 guest, 326 
 
 guise, 369 
 
 Gujarati, where spoken, 187 
 
 yvvT], 356 
 
 Hamblet, 337 
 
 Hamitic languages, remarks on peculiar- 
 ities, 172 ; their divisions, 173-175 
 hand, 368 
 harp, 390 
 Harfe, 390 
 hamais, 207 
 
 harvest, 326 
 
 Haiipt, 390 
 
 hazel, 356 
 
 head, 390 
 
 heart, 320 
 
 Hebrew language, notes on the history 
 of, 169 
 
 f;\60S, 356 
 
 Hellenic (Greek) branch of the Indo- 
 European languages, 192 ; dialects, 
 193; the " conmion " dialect, 193 I 
 modern Greek (Romaic) and its differ- 
 entiae from ancient Greek, 195 ; 
 diffusion of, 253 
 
 helvio, 207 
 
 Henry, M. , his remarks on grammar, 4 
 
 Heracleitus quoted, 398 
 
 herb, pronunciation of, 20, 287 
 
 hermano, yjl 
 
 Herodotus referred to, 250, 251 
 
 High German dialects, 221 ; modern 
 High German, 221 
 
 hi jo, 368 
 
 Himyaritic inscriptions, notes on, 171 
 
 Hindi language, 187 
 
 Hindustani language, 187 
 
 history, 349, 395 
 
 Holophrasis (polysynthesis), 78, 87 
 
 Homeric poems, 192 
 
 Homonyms, or "homophones," examples 
 of, 50 ; how differentiated in Chinese 
 and Egyptian, 51, 117 
 
 horse, 262 
 
 hospital, 287 
 
 Hottentot dialects, 148 ; "clicks," 294 
 
 hound, 384 
 
 housewife, 374 
 
 vaivo., 450 
 
 huckster, 441 
 
 humble, 337 
 
 Humboldt, W. von, his remark on 
 language, 4 
 
 Hungarian (iMagyar) language, 80, 133 
 
 v-trep(j)vq%, 408 
 
 i'0r;, 266 
 
 hussy, 348 
 
 Huxley, his classification of races, 235 
 
 Huzvaresh (or I'ehlevi), notes on, 189 
 
 Hyperborean languages of Asia, 149 
 
 Hypothesis, limits of its use, 362 
 
 Icelandic language, 215 
 
 Ideography (picture-writing) and ideo- 
 grams, 47-54 
 
 Idioms, explained by the study of 
 languages, 518 
 
 idiot, 385, 386
 
 458 
 
 Natural History of Language 
 
 Imitative theory of the origin of language, 
 
 429-432 
 immediately, 388 
 inclutiis, 361 
 
 Incorporating languages, 87 
 Indian languages, ancient and modern, 
 
 184-188 
 Indo-European family of languages, 
 various names for, 175 ; salient char- 
 acteristics, 177-184; synoptical table 
 of the, 227; primitive Indo-Kuropean 
 stock, 240, 244 ; their original home, 
 244 ; data for the discussion, 245 ; 
 reasons for supposing it to have been 
 in Asia, 246 ; now almost given up, 
 247; various opinions, 247; Schrader's 
 treatment of the question, 249 ; diffu- 
 sion of Indo-European races, 249-254 ; 
 wrong ideas as to this diffusion and 
 interrelationship, 255-259 ; linguistic 
 evidence from vocabularies as to 
 original home, civilisation and religion, 
 260-267 
 -ine or -line as a suffix, 450 
 infer, 391 
 
 Infinitive, Anglo-Saxon in -an, 339 
 Infixes in Malayan languages, 144, 145 
 Inflexional (organic or amalgamating) 
 
 languages, 76 ; term unscientific, 82 
 innocent, 387 
 Intonation, use of, 403 
 inwit, ousted by conscience, 441 
 Iranian (or Eranian) division of the 
 Aryan branch and its subdivisions, 
 188-191 
 Irish-Gaelic (or Erse), 210 
 Iroquois group of languages, 156 
 Irregularities of grammar, etc., remarks 
 
 on, 445-448 
 Isolating languages, 76, 88 
 Italian, gradual evolution of, from the 
 sermo rusticus, 201 ; several dialects. 
 202 ; "Sicilian" Tuscan becomes the 
 standard, 202 ; chief extant dialects, 
 202 
 Italic branch of the Indo-European 
 family and its two divisions — Umbro- 
 Samnitic and Latin, 196-198 ; Neo- 
 Latin and the Lingua Ronidna, 200 ; 
 diffusion of, 253; relation to Celtic, 258 
 its, 439 
 
 /, the letter, modern, 66 
 
 fai eu, origin of, 201 
 
 Japanese system of writing, 68 note ; 
 
 language different from Chinese, 70 ; 
 
 grammatical peculiarities, 90, 148 
 
 j'aiirai, origin of, 201 
 
 je vais dire, a colloquial future, 413 
 
 jess, obsolete term, 444 
 
 jovial, 444 
 
 Jiipitcr, 266 
 
 Kafir, see Bantu 
 Kalmuck language, 81, 133 
 Kanarese, where spoken, 136 
 Karelian language, where spoken, 132 
 Kartvelian (or South Caucasian), and 
 
 dialects of, 150 
 Kashmiri language, 187 
 keet, 437 
 kerchief, 349 
 kin, 106, 320 
 Kirghiz, where spoken, 134 
 kirk, 346 
 
 kXvtos, 36, 317 note, 360 
 knave, 385, 386 
 " Knot-writing " of the ancient Chinese, 
 
 45 
 knowledge, pronunciation of, 287 
 Kol (or Kolarian) group of dialects, 135 
 Koloche dialects of British Columbia, 
 
 156 
 Kond dialect, 136 
 Kota dialect, 136 
 
 Kru dialects of the Grain Coast, 147 
 Kudagu (or Kurgi), an Indian dialect, 
 
 136 
 Kurdic dialects, 191 
 
 Labials, 38 
 
 Ladin (Rhaeto-Ronianic) language, 203 
 
 Lammas, 350 
 
 Language, science of, its nature and 
 scope, I ; definition of language, 2 ; 
 interest of the study, 3 ; its use for the 
 study of a special language, 4-6 ; 
 various names proposed for the study, 
 6-8 ; methods of investigation must be 
 scientific, 10 ; divisions of the science, 
 12-14 ; material and data for the study, 
 15 ; specialised study necessary, 16 ; 
 development of, 397-451 ; the evolu- 
 tionary principle in language, ease and 
 economy, loss of flexion, 397-402 ; 
 first devices of language — not poly- 
 synthetic, 402-406 ; primitive material 
 — naming and pointing " roots," 407- 
 411 ; origin of the pointing or deictic 
 elements, 411-415 ; deictic elements 
 become suffixes, 415 ; sufifixing re- 
 placed by pre-posing, the psychological 
 reason, 416-419 ; development of the 
 newer method, interaction of psy-
 
 Index 
 
 459 
 
 Language — 
 
 chology and phonetic decay ; different 
 rates of the latter, 419-426 ; historical 
 phenomena as arguments for the pre- 
 historic, 426 ; the beginnings uncon- 
 nected monosyllables of speech, 427 ; 
 origin of sound -groups ("roots"), 
 with meanings : theories, 428 - 434 ; 
 word-building inexhaustible from old 
 materials : creation of new roots not 
 provable, 434 - 437 ; meaning of 
 "creation" of new words, 437-439 I 
 a language does not create true 
 synonyms, 439 ; loss of words not due 
 to coinage of true synonyms : cause of 
 loss, 440-445 ; effects of the analogical 
 instinct, 445-451 ; race and language 
 not necessarily identical, 228 ; illus- 
 trative instances of this, 229 ; one 
 cognate language may supplant an- 
 other, while the races who speak them 
 are alien, 230 ; modern nations are of 
 mixed races, though speaking the same 
 language, 231 ; so in prehistoric times, 
 232 ; linguistic not to be pressed by 
 the ethnologist, 232 ; is a social, not a 
 physical attribute, 233 ; psychological 
 element, 234 ; classification of races 
 and its bearings on language, 235-237 ; 
 relations of race and language in 
 Europe, 237 ; early European racial 
 types, 239 ; the stock who spoke the 
 primitive Indo-European language, 
 240 ; Celts and Celtic discussed, 241- 
 
 243 
 Languages, basis of classification (mor- 
 phological and genealogical), 70-73 ; 
 morphological classes — isolating, 
 agglutinating, amalgamating — illus- 
 trated, 75-77 ; difficulties connected 
 with this classification, as shown in 
 various types of languages — American, 
 78 ; Basque, 78 ; Bantu, 79 ; Hun- 
 garian and Turkish, 80 ; Latin, Greek, 
 and Sanskrit, 81 ; descendants of 
 these three, 82-84 '< Chinese, Anamese, 
 Siamese, etc., 84; Semitic (Hebrew, 
 Arabic, etc. ), 85 ; general review of 
 these eight tyjies, and modified classi- 
 fication, 87-89 ; languages overlap 
 morphologically, 91 ; various attempts 
 at morphological classification by 
 Schlegel, Grimm, Bopp, Pott and 
 Schleicher, 91 ; only two main classes 
 possible — inorganic and organic, 92 ; 
 genealogical classification, 95-98; 
 basis twofold — similarity of vocabulary 
 
 and of grammar, 98-103 ; grammatical 
 similarity may disappear, 103; in what 
 sense it holds good, 105-108 ; unity of 
 language as yet incapable of proof, 
 
 108 ; Mailer's estimate of the number 
 of unconnected groups, 109 ; Indo- 
 
 • European family and its subdivisions, 
 
 109 ; examples of similarity in vocab- 
 ulary and grammar in this family, iio- 
 112 ; kinship in language does not 
 imply kinship in race, 112; Semitic 
 and Bantu families of languages 
 similarly akin, 113; general survey of 
 languages, 115; positional or isolating 
 languages of Further Asia — Chinese, 
 116-126; Anamese, Siamese, Burmese, 
 Tibetan, 126; genealogical connection 
 of these languages unproven, 127 ; 
 agglutinating - infiexional languages : 
 suffix -agglutinating type, 129-138; 
 prefix -agglutinating type, 138-140; 
 prefix- and suffix-agglutinating type, 
 140-145; other agglutinating lan- 
 guages, 145-152; polysynthetic or 
 holophrastic languages of America, 
 152 - 157 ; amalgamating - inflexional 
 languages : Semitico-Hamitic family, 
 158-160 ; Semitic division, 160-172 ; 
 Hamitic division, 172-175; Indo- 
 European family, 175-227. {For 
 
 further details see under special entries) 
 
 Langue d'oc, 204, 207 
 
 Langue d'oil, 204, 207 
 
 Lapponic, 132 
 
 I'lrk, 335 
 
 larme, 356 
 
 Xdpva^, 364 
 
 Larynx, the, described, 23 
 
 Latham, his opinion as to the original 
 home of the Indo-European family, 248 
 
 Latin, ancient, 198 ; its three periods, 
 199, 200 ; the sermo plebeius (or 
 rusticus), 199 ; Low Latin, 200 ; 
 examples of popular Latin, 200, and 
 of its analytical tendencies, 201 ; for- 
 mation of Romance languages, 201 ; 
 wrong ideas as to relationship ot 
 classical Latin and Greek, 256 
 
 Lautverschiebung, or sound-shifting in 
 Teutonic languages, 213, 290, 319 ; 
 examples of, 320 ; second or High 
 German shifting, 321 
 
 Laziness, not a suitable term to account 
 for phonetic decay, 350-354 
 
 leal, ^70 
 
 Leiotrichi, proposed racial designation, 
 and its subdivisions, 235
 
 460 
 
 Natural History of Language 
 
 Letters of the alphabet, in Phoenician, 
 
 57 ; meaning of the Hebrew names, 
 
 58 ; Greek names, 59 ; remarks on 
 some of the Latin letters, 63, 66 
 
 Lettish, where spoken, 226 
 
 levir, 311 
 
 liber, 14 
 
 lieiie, 207 
 
 •ling, Enghsh diminutive, 450 
 
 lingua, 2,57 
 
 Linguistic, French name for the science 
 
 of language, 6 
 Lips, their function in producing sounds, 
 
 25 
 
 Liquids, 37 
 
 lists, 444 
 
 Lithuanian, the most archaic of living 
 
 Indo-FAiropean speeches, 226 
 Livonian language, 132 
 lord, 349, 438 
 Loss of inflections, 398-402 ; of words, 
 
 440-442 
 loud, 360 
 loner, 371 
 Low German, 220 
 loyal, 370 
 \vKaiva, 450 
 
 Luther's German Bible, 222 
 -ly, adverbial suffix, origin of, 412 
 
 ma'am, mam=^madam, 335 
 
 magistral, 317 
 
 mag, magpie, 437 
 
 maginis, 443 
 
 Magyar (Hungarian) language, 80, 133 
 
 maid, 382 
 
 maitre, 2^7 
 
 Majuscules, or capital letters, 66 
 
 Malayalam language, 136 
 
 Malayan group of languages, 142, 145 
 
 Malayo-Polynesian family of languages, 
 its divisions, 140-142, and sub- 
 divisions, 142 ; peculiarities of gram- 
 mar, etc., 143-145 
 
 Maltese, a corrupt Arabic jargon, 171 
 
 man, 374 
 
 ?n and lad re, popular Latin for comedire, 
 200 
 
 manger, 200 
 
 Mann, 374 
 
 Manx language, 211 
 
 Maori language, 142 ; peculiarities of 
 articulation, 314, 332 
 
 Marathi, where spoken, 187 
 
 mare, 261 
 
 mai-^chal, 207 
 
 Marjonbanks, pronunciation of, 339 
 
 mata ( Polynesian = e3'e), 11 ?iote, 370 
 
 fioiTi, modern Greek = eye, 11 note, 370 
 
 Meaning, changes of, a constant tendency 
 in language, 373 ; laws of meaning- 
 change not physiological, but psycho- 
 logical and difficult to formulate, 373 ; 
 instances, 374 ; changes due to 
 metaphor, specialisation, change of 
 environment and social conditions, 
 375-377; changes in names for 
 abstract notions, 377-379 ; tendencies 
 leading to narrowing, widening or 
 shifting of meaning, with illustrative 
 instances, 379-390 ; changes due to 
 laxity or ignorance, 390-393 ; due to 
 change of circumstance, social or 
 material, 393; "doublets," 394; 
 scheme to represent possibilities of 
 change, 395 
 
 meat, 384 
 
 Melanesian group of languages, 142 
 
 meme, etymology of, 338, 361 
 
 Manage, his derivation of rat from mns, 
 362 
 
 me?id, 335 
 
 Mendaean dialect, 169 
 
 Mende dialect of Upper Guinea, 147 
 
 -ment, French adverbial termination, 
 412 
 
 Mercian dialect, progenitor of modern 
 English, 218 
 
 meridionalis, 450 
 
 Metaphor, its effect on the meaning of 
 words, 375, 388 
 
 /i-!jT7?p, 316 
 
 Mexican group of languages, 157 
 
 Migration, linguistic effects of, 294, 295 
 
 minister, 394 
 
 Minuscules, or small letters, 66 
 
 miscreant, 380, 391 
 
 Misteli, F. , his remark on the Semitic 
 elements of speech, 161 
 
 Mixture of races, its linguistic effects, 
 296-300 
 
 Moabite dialect, 170 
 
 fnoglie, 374 
 
 mole re, 266 
 
 Mongolian group of languages, and their 
 divisions, 133 
 
 Monosyllabic (or isolating) languages, 
 75, 88 
 
 Monosyllables, the beginnings of speech, 
 427 
 
 Morphological classification of languages 
 72, 74-94. .S^e ////fl'^/' Languages 
 
 Morphology, 13 
 
 mother, 316
 
 Index 
 
 461 
 
 mulier, 374 
 
 \xvKK(.Lv, 266 
 
 Mulier, F. , his stages of alphabet-develop- 
 ment, 68 ; his estimate of number of 
 unconnected languages, 109 ; his 
 phonetic principles of etymology, 368- 
 
 372 
 Mulier, Max, on the principles of sound 
 
 etymolog)', 368-372 
 mio-, its declension in Old French 
 
 explained, 9 
 Mutes, or explosives, 37 
 
 naif, 317 
 
 nary, 436 
 
 nas, 412 
 
 Nasals, 37 
 
 //aOj, 266 
 
 iidvigo, 404 
 
 navis, 266, 444 
 
 nay and no, 369 
 
 ne, 412 
 
 nebula, 341 
 
 needle, 441 
 
 Negro languages, principal groups of, 
 
 146 
 Nepali language, 187 
 Nieder-Deutsch (or Platt-Deutsch), 220 
 nill, 412 
 
 Nogair, where spoken, 134 
 Norman invasion, its effects on English 
 
 speech, 420 
 Norse (or Scandinavian) languages, 214 
 Northumberland dialect, progenitor of 
 
 the Scottish dialect, 218 
 Norwegian language, 215 
 not, 406 
 Nubian-Fula languages, 148 
 
 oblige, pronunciation of, 20, 287 
 
 " O'Clo' ," Coleridge's story of, 352 
 
 officious, 391 
 
 oil, 43 
 
 onixdriov, 370 
 
 on, 270 
 
 Onomatopoetic theory of the origin of 
 language, 429-432 
 
 Order of words (pre- and post-posing), 
 414-426 
 
 Organic languages, and their sub- 
 divisions, 93 
 
 Origin of language, Jespersen's opinion 
 as to, 404 ; began with unconnected 
 monosyllables, 427 ; theories as to 
 origin of naming, and deictic words, 
 428-434 
 
 5(pis. 443 
 
 Oscan language, remains of, 197 ; 
 
 granmiatical peculiarities, 197 nole 
 Ossetian language, 191 
 Ostiak language, 133 
 oui, 43 
 
 pai^/, 323 
 
 Palatalisation, examples of in English, 
 
 324 
 
 Palatals, 38 
 
 Palate, hard and soft, 24 
 
 Pali, sacred language of Buddhism, 187 
 
 palsy, 349 
 
 paper, 384, 394 
 
 Papuan languages, 145 
 
 Parallelism between mutes, spirants, and 
 nasals, 39 
 
 parrakeet , 437 
 
 Parsi (or Pazend) language, 190 
 
 parson, 384 
 
 Participle, past, agreement of with pre- 
 ceding object, 9, 399 ; in Latin, 449 
 
 pater, 316 
 
 Paul, H., quoted, 15 
 
 Pawnee group (in Arkansas), 156 
 
 Pehlevi (or Pahlavi) language, 189 
 
 pergo, 405 
 
 Permian language, 132 
 
 perroquet, 437 
 
 Persian language, 189 ; modern, 97, 420 
 
 Person -endings of the Indo-European 
 verb, their origin, 414 
 
 Peruvian group of languages, 157 
 
 Peschell, his classification of races, 235 
 
 Peshito, Syriac translation of Old and 
 New Testament, 168 
 
 Pharynx, its agency in producing sound, 
 24 
 
 ^t;7(5s, 265 note, 326, 376 
 
 Philology, comparative remarks on the 
 term, 7 
 
 Phoenician alphabet, 57, 58, 6i ; 
 language, 170 
 
 phone, 351 
 
 Phonetic change : sound and thought ; 
 learning to speak, 268 ; the sound- 
 picture, whence derived, 269 ; the 
 sound-picture in later life, 270 ; its 
 relation to utterance, 271 ; individual 
 deviations in the sound-picture and in 
 utterance, 272 ; phonetic predilections 
 of the individual illustrated, 273 ; re- 
 action of the individual causing a 
 shifting in the average articulation of 
 his environment, 274 ; counteraction 
 of individual influences never complete,
 
 462 
 
 Natural History of Language 
 
 Phonetic — 
 
 275 ; the counteracting forces in 
 primitive societies, 276 ; growth of 
 dialects and spread of phonetic change : 
 from individual to group, 278-282 ; 
 distinction between conditions of dia- 
 lectical growth among primitive and 
 cultured people : in primitive peoples 
 the groups mostly isolated, 282 ; when 
 groups are not isolated there is con- 
 tinuous transition and interlinking of 
 dialects, 283 ; continuous transition 
 in phonetic phenomena, 285 ; con- 
 ditions of phonetic change among 
 cultured peoples : change inevitable, 
 286 ; phonetic change rarely confined 
 to one sound at a time, 288-290 ; 
 fortuitous agreement of widely separ- 
 ated dialects in some particular sound- 
 change, 291 ; summary of the above 
 considerations, 292 ; effects of neigh- 
 bouring aliens, 293 ; effects of com- 
 plete geographical separation, by 
 natural expansion, or by specific 
 migration, etc., 294-296; effects of 
 climate and manner of life, 296 ; 
 effects of mixture with alien peoples, 
 297-300 ; recapitulation of the phe- 
 nomena of, 300-302 ; examples and 
 illustrations of phonetic corruption in 
 noun and verb-flexion, 420-426 
 
 Phonetic laws : illustrative examples of, 
 in Greek, Latin, and English, 304 ; 
 definition of, 306 ; refers only to a 
 particular time, 307 ; and a sound 
 particularly placed, 309 ; exceptions 
 discussed, 309-315 ; apt to be stated 
 too comprehensively, 315 ; Verner's 
 law, 316 ; Latin intervocalic con- 
 sonants, g, t, d, in French and Italian, 
 
 317 
 Phonetic tendencies, general, due to 
 various conditions, racial, climatic, and 
 social, 329-331; some erroneous 
 notions regarding these, 331 ; some 
 special phonetic peculiarities, 331-333; 
 cases and examples due to desire for 
 greater ease or less effort, 333-335 ; 
 assimilation and glide-sounds, 335- 
 337 ; changes not always due to these 
 reasons, 337; sound -change due to 
 economy of exertion generally holds 
 good, 338 ; very common in French, 
 339 ; nature and history of the 
 accent and its phonetic effect in 
 various languages illustrated, 340- 
 346 ; examples due to economy of 
 
 effort, etc., 346-350; "laziness" not 
 a suitable term for this, 350-354 
 
 Phonology, scope of, 13 ; remarks on 
 spoken and written language and 
 their divergences, 17-20 ; causes of 
 this divergence, 20 ; mechanism of 
 speech, 21-25 '< sounds in detail, 26- 
 43 ; application of, to etymologies, 
 355-372. [See also tinder Etymology 
 and other headings) 
 
 (ppdrrip, 376 
 
 Picture-writing, gradual evolution of, 
 into alphabetic, 47-54 
 
 " Pidgin " speech, 428 
 /?7, 323 
 
 Pitch, or height of a sound, 27 
 place, 380 
 
 Plato's ideas on etymology, 364 
 
 Plural in Latin and French, 448 
 
 Polabish language (extinct), 226 
 
 Polish language, 226 
 
 Polynesian languages, 141, 142 
 
 Polysynthesis, examples of, 78 
 
 Polysynthetic (or holophrastic) languages 
 of America, 152-156 ; classification of, 
 156 ; Jespersen's opinion on the, 404 
 
 "Pooh-pooh" theory of the origin of 
 language, 433 
 
 popitia, 312 
 
 Portuguese language, notes on, 206 
 
 Positional, or organic languages, 92 
 
 put, 207 
 
 Pott, his morphological classification of 
 languages, 92 
 
 Prakrit dialects, notes on, 186 
 
 premises, 390 
 
 Prepositions originally with adverbial 
 force, 422 
 
 preposterous, 390 
 
 presently, 388 
 
 problematical, 391 
 
 Pronouns, sense-shifting in the use of, 377 
 
 Pronunciation, remarks on, 18-21 
 
 Proven9al (langue d'oc), notes on, 204 
 
 proxy, 349 
 
 pr It dens, 340 
 
 Prussian, Old, 226 
 
 piilcher, 365 
 
 puncttial, 384 
 
 Punjabi language, 187 
 
 ptiny, 391 
 
 ptirchase, 384 
 
 Pushtu (or Pashtu) language, 187 
 
 puzzle, etymology of, 349 
 
 (/ t I, the root in Semitic languages, 86, 
 161, 164
 
 Index 
 
 463 
 
 qiiantitas, a Latin equivalent for irocroTrjs, 
 
 440 
 queen, 356 
 quercus, 264 note 
 queue, 317 
 quick, 356 
 Quipu-writing of the Peruvians, 45 
 
 Race, not necessarily convertible with 
 language, 228, races may mix, 
 languages do not, 229 ; proposed 
 classification of, and its bearing on 
 language, 235-237 
 
 raisin, 384 
 
 Reduplication in Indo - European lan- 
 guages, 182 
 
 Reichenau glosses, the, 207 
 
 reiiie, 317 
 
 reliable, 439 
 
 remigo, 405 
 
 resent, 393 
 
 restive, 372 
 
 restless, 372 
 
 retaliate, 393 
 
 rdttuli, 341 
 
 Rhaeto- Romanic (Ladin, or Rhaetian 
 Rouniansch), notes on, 203 
 
 right, 279 
 
 Rig- Veda, 184, 185 
 
 Romaic, or popular modern Greek, 194 
 
 Romance (or Neo- Latin) languages, 
 development of, from popular Latin, 
 
 95 
 Romany (or Gipsy) dialects, 186 
 Home, old pronunciation of, 20, 287 
 Roots, primitive Indo - European, are 
 monosyllabic, 181 ; their vowel varia- 
 tion, 183 ; predicative and demon- 
 strative, 407 ; definition of predicative 
 roots or naming-element, with illus- 
 trations, etc., 407-411 ; origin of the 
 formative or deictic element, 411 -41 5 ; 
 these become suffixes, 415 ; origin of, 
 with meanings, 428-435 ; spontaneous 
 creation of new roots not provable, 
 
 435-437 
 rossignol, 332 
 Roumanian (or Wallachian) language, 
 
 origin of, 203 ; its vocabulary, etc. , 
 
 203 
 ruche, 207 
 Russian, Great, 224 ; White, 225 ; 
 
 Little (or Ruthenian), 225 
 
 S, intervocalic, in Latin and Greek, 304, 
 
 305. 308 
 sail, 394 
 
 Samoyedic group of languages, 133 
 Sanskrit, Vedic, literature, 184 ; classical 
 
 Sanskrit, 185 ; diffusion of, 249 
 Sarmatian, 251 
 satellite, 446 
 Saturnine, 444 
 
 Saxonic, Old, dialects, 217 ; Old Saxon 
 remains, 220 
 
 scalae, 340 
 
 Sandinavian (or Norse) languages, 214 
 
 Scherer, E., his remark on science, 7 
 
 Schlegel, his morphological classification 
 of languages, 91 ; Whitney's objection 
 to it, 92 
 
 Schleicher, his morphological classifica- 
 tion, 92 
 
 Schmidt, J., his " wave- theory," 258 
 
 Schrader, his opinion as to the home of 
 the primitive Indo-European languages, 
 249 
 
 Scottish dialect, its origin, 218, 220 
 
 Scythian, 251 
 
 seal, 27^ 
 
 secure, 391 
 
 Seine, 335 
 
 selfish, 439 
 
 semaine, 335 
 
 Semasiology, scope of, 13 
 
 Semitic languages, morphological peculi- 
 arities of, 85 ; examples of similarity, 
 113; peculiarities of grammar, 160- 
 166; table of subdivisions, 167 
 
 Semitico-Hamitic family of languages, 
 remarks on their relationship, 158-160. 
 {See Hamitic and Semitic) 
 
 Semi-vowels, 35 
 
 senate, 394 
 
 siinc'chal, 207 
 
 septentrio, 336 
 
 septe}itriondlis, 450 
 
 serpens, 443 
 
 Servo-Croatian, where spoken, 225 
 
 Sev-in-(iir-il-nie->nek (Turkish), 181 
 
 Shahnameh, national epic of Persia, 190 
 
 shamefaced, 368 
 
 sherif, 349 
 
 Shifting-sound, see Lautverschiebung 
 
 shirt, 369, 395 
 
 Shumerian (or Accadian) language, 152 
 
 Siamese (or Thai) dialects, 126 
 
 Sibilants, 37 
 
 silly, 387 
 
 " Siloam inscription" referred to, 56 
 
 similis, 365 
 
 simple, 2,^7 
 
 Sindhi language, 187 
 
 skirt, 369, 395
 
 464 
 
 Natural History of Lang2iage 
 
 Slavonic languages, where spoken, 222 ; 
 
 classification, 224 
 Slovakian, a dialect of Tzech, 225 
 Slovenian, where spoken, 225 
 
 soh 356 
 
 Sonants and con-sonants, 33-35 
 
 Serbian (Sorabian or Wendic) language, 
 225 
 
 sorrow, distinct from sorry, 11 
 
 sound, 371 
 
 Sounds, how produced, 26 ; in what 
 respects they differ, 27-30 ; Sweet's 
 classification of vowel-sounds, 30 ; 
 vowels and consonants, 31-40 ; table 
 of sounds, 41 
 
 Spanish language, its characteristics, 
 205 ; dialects, 206 
 
 Specialisation, or narrowing of meaning 
 in words, instances of, 383 
 
 Speech, mechanism of, 21-25 
 
 Spencer, H., quoted on the order of 
 words, 418 
 
 Stamboul, 343 
 
 Stancho, 343 
 
 star, 367 
 
 starve, 384 
 
 story, 349, 395 
 
 Stress of sounds, 27 ; accent, 27 
 
 Suffixes, origin of, 415 ; suffixing re- 
 placed by pre-posing from psycho- 
 logical reasons, 416-419 
 
 superbus, 408 
 
 superciliosus, 438 
 
 s^r, 340 
 
 sjirgo, 405 
 
 sweat, 107, 327, 357 
 
 Swedish language, 215 
 
 Sweet, his classification of vowel-sounds, 
 30 ; observation on the modifying 
 influence of climate, 296 
 
 Syllables, omission of unaccented, 349 
 
 Syncopation, instances of, 335 
 
 Synonyms, 439-444 
 
 Syntax, comparative, scope of, 13 
 
 Synthetic languages, 93 
 
 Syriac language, notes on, 168 
 
 Tagala group of languages, 142 
 
 Tamil, where spoken, 136 
 
 Tchudic language, 132 
 
 tea, pronunciation of, 20, 287 
 
 tear, 356 
 
 tectum, 14 
 
 tegula, 14 
 
 Teil, 371 
 
 reivu}, 408 
 
 Telugu (orTelinga), where spoken, 135 
 
 ten, 106 
 
 tendo, 408 
 
 tendre, 42, 337 
 
 tenuis, 408 
 
 testa, XkXs., 389 
 
 rh'Tapes, 356 
 
 Teutonic branch of the Indo-European 
 languages, table of subdivisions, 212 ; 
 diffusion of, 252 ; relation to Baltic- 
 Slavic and Aryan, 257 
 
 thatch, 14, 306, 307, 309, 322, 324, 384 
 
 ^e6r, not identical with deus, 11, 107, 
 267, 315 
 
 thin, 408 
 
 three, 320 
 
 dvyarrip, 261 
 
 diipa, 265 
 
 Thursday, 335 
 
 Tibetan group of languages, 126 
 
 tide, 384 
 
 Tier, 375 
 
 timber, 265 
 
 Timbre, or quality of a sound, 28 
 
 toison, 443 
 
 "Tones" in Chinese, 11 8- 120 
 
 Tongue, the, as an organ of speech, 25 
 
 tongue, 357 
 
 toiisio, 443 
 
 tooth, 326 
 
 torn, 357 
 
 town, 394 
 
 transpire, 391 
 
 tree, 264, 327 
 
 truciddre, 342 note 
 
 Tuda, a dialect of the Nilgherries, 136 
 
 Tulu language, 136 
 
 Tunguse group of languages, 133 
 
 Turkish (or Osmanli) language, 134; 
 peculiarities of declension and conjuga- 
 tion, 80 
 
 Turko-Tataric group and its divisions, 
 
 133 
 Tylor, his definition of language, 2 
 Tzech, language of Bohemia, 225 
 
 Ugric and its divisions, 133 
 
 Uignric dialects, 134 
 
 Ulfilas, his translation of the Bible into 
 
 Gothic, 213 
 Ulotrichi, proposed racial designation, 
 
 and its subdivisions, 235 
 Umbrian dialect, 198 
 Umbro-Samnitic group and its divisions, 
 
 197 
 Umlaut, or "vowel-mutation," 120; in 
 
 the Ural-Altaic languages, 130 ; in 
 
 Indo-European, 183
 
 Index 
 
 46: 
 
 tindertake, 384 \ 
 
 Unity of languages, as yet unproved, 108 
 unkind, 380 
 
 unus = oeniis = oinos, 313 
 Ural-Altaic family of languages, general 
 remarks on, 129- 131 ; its divisions, 
 
 131-134 
 Urdu langtxage, 187 
 Uriya, where spoken, 187 
 itrsHS, 263 
 fisurpo, 335 
 -iitus in Latin past participle, assimilative 
 
 power of, 449 
 Uvula, 25 
 
 vamose, 436 
 
 van, 349 
 
 Velars, 38 
 
 velle = vel-se, 179 
 
 vello, 365 
 
 velliis, 443 
 
 Velum palati, or soft palate, 24 
 
 Venedi, Veneti, Venedae, the Wends, 252 
 
 Vepsic dialect, 132 
 
 Verner's law, 316 
 
 vert, 344 
 
 veru)i[in)tamen, 336 
 
 viaticum, popular Latin for iter. 200 
 
 vier, 356 
 
 villa, ville, 395 
 
 villani, villain, 385 
 
 vixen, 389 
 
 Vlach, a Roumanian dialect, 204 
 
 Vocabulary, similarity of, one of the 
 criteria for genealogical classification 
 of languages, 97-101 ; what this 
 similarity means, 105-108 ; examples 
 of, in Indo-European, no 
 
 Vocal chords, 24, 26 
 
 Vogul dialect, 133 
 
 Voice, how produced, 26 
 
 Volga-Finnic (Bulgaric) language and its 
 divisions, 133 
 
 Voltaire, his sarcastic definition of ety- 
 mology, 12, 366 
 
 volume, 394 
 
 Votiak dialect, 132 
 
 Votic dialect, 132 
 
 " Vowel -harmony," principle of, illus- 
 trated, 130 
 
 Vowels, how produced, 26 ; Sweet's 
 classification, 30 ; the neutral vowel, 
 31 ; nasalised vowels, 32 ; old defini- 
 tion of vowels and consonants in- 
 correct, 33 ; change of, 288 ; omission 
 of final, 348 
 voyage, 383 
 
 W, a modern letter, 66 
 V^^'allachian (or Roumanian )_ language, 
 
 203 
 wanbope, ousted by despair, 441 
 ■watch, 307, 309 
 
 "Wave-theory" of J. Schmidt, 258 
 weave, 266 
 Weib, 374 
 
 Welsh language, notes on, 210 
 Wendic (or Serbian) language, 225 
 W'ends (X'eneti), 252 
 Whitney's definition of a language 
 
 quoted, 99 
 zohole, not akin to oXos, 107, 359, 
 
 367 
 
 7uife, 374, 382 
 
 wise, 369 
 
 7mt, 385, 386 
 
 witticism, 450 
 
 Wolof dialects of Senegal and Sene- 
 gambia, 146 
 
 woman, 374, 382 
 
 Worcester, pronunciation of, 339 
 
 Words, identical with roots in mono- 
 syllabic languages, 71 ; the part they 
 play in the vocabulary of a language, 
 99-101; their ethnographic value, 260- 
 267 ; phonetic changes in words and 
 their causes, 268-302 ; origin (ety- 
 mologies) of words, 355-372; changes 
 in meaning of words, 373-396 ; word- 
 building from old material, 431-440 ; 
 loss of, 440-445 
 
 worth, 442 
 
 Writing, evolution of, 44-54 : stages in 
 the process, 68. {See also under 
 Alphabet) 
 
 wrong, 379 
 
 Xanthochroi, 240 
 ^vXoTO/jLeu), 405 
 
 Yakut language, where spoken, 134 
 
 yard, 369, 383 
 
 yellow, pronunciation of, 287 
 
 yes'm, 335 
 
 yoke, 326 
 
 Vork. 335 
 
 Yuma, spoken in Colorado, 156 
 
 Zaun, 394 
 
 Zend and Zend-Avesta, 188, 199 
 
 Zei^s, 266 
 
 Ziminer, 265 
 
 Zulu language, 138
 
 tcwti^
 
 THE LIBRARY 
 UNIVERSITY OF CALIFORNIA 
 
 Santa Barbara 
 
 THIS BOOK IS DUE ON THE LAST DATE 
 STAMPED BELOW. 
 
 jjgiS NOV & ^ • 
 RErD0GT23l997fi(,^ 
 
 lOOM 11/86 Series 9482
 
 ^ "|"ii"iiliM(ti|i[|:ji|itl|l|ill|llll|||||| III 
 
 3 1205 01173 9420 
 
 UC SOUTHERN REGIONAL LIBRARY FACILITY 
 
 AA 000 815 177 1