o 
 
 SB 13 SSb 
 
 3OTB 
 
 y- 
 
 Ik 
 
 
 ;iil 
 
 
 
 
 > "^V -1, 
 
 t^ 
 
 O 
 O 
 
Ctbrarg uf 
 
 ite Hbrrlrr 
 
GIFT OF 
 Ben. 1 ami n Ide Wheeler 
 
*"/ 
 
 The Origin of the Recessive 
 Accent in Greek. 
 
 MAURICE BLOOMFIELD 
 
 JOHNS HOPKINS UNIVERSITY 
 
 Reprinted from The American Journal of Philology, 
 Vol. IX, N(,\. r,, pp. 1-41. 
 
 \\\ LT I M OR K, I 888 
 Publication Agency of the Johns Hopkins University 
 
THE ORIGIN OF THE RECESSIVE ACCENT 
 IN GREEK. 
 
 Jacob Wackernagel, in KZ. XXIII 457 fg., made the important 
 discovery that the so-called ' recessive ' accent in the finite forms 
 of the Greek verb represents a substitute for an older Indo- 
 European fact in sentence-accentuation, to wit, that the finite verb 
 in principal clauses was treated as an enclitic. This enclisis was 
 extended in Greek to the finite verb in both principal and subordi- 
 nate clauses, but was, on the other hand, restricted by a law 
 according to which an enclitic word may not contain more than 
 two syllables and three moras. Therefore only two syllables at 
 the end are allowed to be barytone : fopopfv for enclitic ^. *(epo/u,/ ; 
 three moras at the utmost, and that only in a polysyllabic form, 
 ending in a trochaic cadence : XeXoi7ro>ju,ey for ^ *\e\oura)p.v ; <pep<0ij.<-v 
 for JL *<epo>/xei/. Elsewhere only two moras were left barytone: 
 
 (pep6p.eda for -. <pepop,e()a \ (pfpopcv for JL (pepopfv ', (pepa> for ^. (pepo) } 
 
 olda (i. e. *f6l8d) for ^ *oi'Sa. In words containing altogether but 
 two moras, one was left barytone : XtW, augmentless aorist for 
 ^ *Ai7res ; /3i) (i. e. */3ee)> augmentless aorist for j. *fa. Monosyllabic 
 forms of one mora are accented, so that no mora is left toneless : 
 /3ai/, a-rav, (pddv, augmentless aorists for ^ */3ay, etc. 
 
 We may refrain at present from any attempt at justifying the 
 derivation of these ' recessively ' accented verbal types from the 
 assumed enclisis : we shall return to that question in the end. It 
 is enough to state that these accentual types are one and all deriv- 
 able from the enclitic theory, and that they represent every con- 
 ceivable manifestation of the ' recessive ' mode of accentuation, 
 providing only it is remembered that words of more than three 
 
 411279 
 
syllables are treated in the same way as words of three syllables : 
 
 do0r)o-6pc6a, dwdpfOa like fjfjL0a, etc. 
 
 In an article entitled ' Historical and critical remarks introduc- 
 tory to a comparative study of Greek accentuation,' American 
 Journal of Philology, IV 21 fg., I proposed an extension of this 
 law, so that it would serve as a theory by which all non-etymo- 
 logical accentuation in Greek words could be accounted for. My 
 statements were as follows : 
 
 P. 56 (p. 36 of the reprint). 'The explanation of the Greek 
 recessive accent must start from the finite form of the verb, where 
 alone it is evidently at home.' 
 
 P. 30 (10). ' It is a fact perfectly clear that the recessive accent in 
 Greek, whatever its explanation, started with the finite forms of 
 the verb, and thence succeeded in attacking nominal formations 
 also.' 
 
 P. 50 (30). ' It (the recessive accent) excludes with particular care 
 non-finite forms of the verb in the same tense system and in 
 evident connection with finite forms, exhibiting thus on Greek 
 ground a most outspoken character as a grammatical quality of 
 finite verbs.' 
 
 P. 62 (42). ' No doubt the noun has to a large extent followed 
 the verb in its enclisis.' 
 
 This theory involves, of course, the belief that the extension of 
 the recessive accent from the verb to the noun took place according 
 to processes of analogy, not different in principle from those which 
 elsewhere break in upon the regular line of phonetic facts. I 
 shall show below, in a somewhat detailed fashion, the manner in 
 which this must be imagined to have taken place. 
 
 The only writer, since the publication of my treatise, who has 
 subjected the question of the recessive accentuation in Greek to 
 an independent investigation is B. I. Wheeler, in his book, Der 
 Griechische Nominalaccent, Strassburg, 1885. Wheeler's work has 
 been for me, as for others, one of great interest. He has brought 
 to his work good training and esprit. His method of investi- 
 gation is comprehensive ; he does not draw an arbitrary line which 
 cuts off the domain of his inquiry from adjoining territory open to 
 search and likely, nay certain, to yield information. His study is 
 nothing if not comparative. His methods are rigorously exact, 
 perhaps a little overdrawn in that direction, as I shall endeavor to 
 show in the sequel. He seizes upon, with rather too eager emphasis, 
 the working principle which I formulated in my article, p. 31 (n) 
 
fg., namely, that accent must be investigated with the same funda- 
 mental presumptions, or principles, as other phonetic matter. 
 Phonetic change in accordance with phonetic law and analogy, I 
 urged, loc. cit., are the prominent factors, aside from the influence 
 of foreign words, which are at the bottom of the frequently por- 
 tentous changes on the face of the accentuation of a given language. 
 Wheeler operates with these factors almost entirely, but he narrows 
 the operation of both so as to admit under these heads only such 
 phenomena of change as appear familiarly in extra-accentual 
 phonetics. He fails to do justice to the fact that the centrifugal 
 force of phonetic change and the centripetal force of analogy 
 operate both at a totally different rate in the change of accentuation, 
 and in the change of other phonetic material, simply because the 
 scope of any accentual type is greater than that of any type 
 involving a given mode of vocalization or consonantal treatment. 
 The application of the principles of phonetic law and analogy to 
 the accentuation of the Lettish dialect is a la rigueur justifiable, but 
 it must be done in the spirit of the preceding sentence. The 
 Lithuanian is related so closely to Lettish that the two are pre- 
 ponderatingly convertible, if a certain number of phonetic changes 
 are rigorously observed. The Lithuanian exhibits a free accentua- 
 tion which can be compared and identified with the Vedic accent 
 in spite of many deviations. The Lettish, which is related as 
 closely to the Lithuanian as the language of Herodotus is to that 
 of Thucydides, has abrogated all etymological accentuation and 
 has the summit tone everywhere on the first syllable. The change 
 from the free Baltic accentuation as represented by the Lithuanian 
 to this mechanical accentuation of the Lettish is due, or may be 
 due, to a preponderance of the. analogy of such words as accented 
 the first syllable etymologically, and in this sense the change is 
 analogical. But it would be useless to demand further that every 
 word which obtained this accentuation secondarily must exhibit 
 some formal or functional cause for adopting it. Only in this sense 
 can a levelling accentuation be the result of analogy. Such are the 
 accentuation of the radical syllable in German, the accentuation of 
 the final syllable in the French of the last century, the accentua- 
 tion of the first syllable in Bohemian and Serbian, the accentuation 
 of the penult in Polish and Welsh, the complete 'recessive' 
 accentuation of the Aeolic, the practically complete barytonesis of 
 Latin and its restriction of the accent within three syllables, etc. 
 Analogy with its ordinary scope word influencing word, form 
 
influencing form may carry on its humble working by the side 
 of and in the teeth of a great leveling tendency. eKvpos ' father-in- 
 law ' may exhibit oxy tone accentuation secondarily after the analogy 
 of cKvpd in spite of the ' recessive ' tendency (Wheeler, p. 59). 
 Originally it was *cicvpos ; cf. Sk. gvd$ura-; Gothic swaihra (orig. Ger- 
 man *sweh[u]ro-') ; Lith. sze'ssuras. Such cases barely cause a 
 ripple on the quiet, strong current which carries the accentuation 
 into the opposite direction. 
 
 The foundation upon which Wheeler's book is built is a new theory 
 in explanation of the ' recessive ' accentuation. He denies that the 
 phenomena thus designated were originally a property of the finite 
 verb, and claims that they are due to a phonetic fact which permeated 
 the whole material of the language. He follows a suggestion of 
 OsthofTs, which had been previously indicated by Curtius, and 
 assumes that in words containing a sufficient number of moras a 
 secondary accent was developed, which fell upon the third mora 
 from the end in all words except those of more than two syllables 
 ending in a trochee ; in the latter the secondary accent fell upon 
 the fourth mora from the end. 1 This secondary accent is assumed 
 to have developed upon all spondaic and iambic words and 
 upon all words of three or more syllables. For reasons which it 
 puzzles the reader to find out, he excludes from the effect of this 
 secondary accent trochaic dissyllables (ol^os ' way ' =. Sk. ema- ; 
 fAs ( appearance '= Sk. vedas ; aWos ' fire ' = Sk. edhas], though 
 they possess just as many moras as iambic dissyllables (rplnovs 
 t tripod ' = Sk. tripad'}, and the ' secondary ' accent is palpably 
 represented by the circumflex. About this more below. 
 
 From the benefits of this ' secondary ' accentuation he therefore 
 excludes short monosyllables, long monosyllables, words of two 
 short monosyllables, and trochaic dissyllables. He assumes, 
 moreover, that this secondary accentuation gained the upper hand 
 under certain circumstances, while under others the old etymolo- 
 gical accent survived. Accordingly he divides the whole material 
 of the language into four categories, barring of course the special 
 effects of other minor phonetic laws and analogies, 2 as follows : 
 
 1 Cf. Curtius in Fleckeisen's Jahrbiicher for 1855, p. 342; Osthoff, cited by 
 Wheeler, p. 10, note 2 ; and Wheeler, p. 9 fg. 
 
 2 Wheeler in reality posits five divisions, but his fourth division is one alto- 
 gether independent of the general theory. In it he has collected considerable 
 material which aims towards the establishment of a phonetic law previously 
 hinted at by Bopp and Curtius, according to which words originally oxytone, 
 
I. Monosyllabic forms and dissyllabic ones with short final 
 syllable retain the inherited accent intact. 
 
 II. If the original accent lay nearer to the beginning of the 
 word than the secondary accent, then the secondary accent pre- 
 vailed. 
 
 III. If the original accent coincides with the secondary accent, 
 then it remains undisturbed. 
 
 IV (Wheeler's No. V). If the original accent lay nearer to the 
 end of the word than the secondary accent, there arose a vacilla- 
 tion which was settled later on in favor of one or the other. Some- 
 times the cause of the choice is apparent, sometimes not. 
 
 Wheeler's book was reviewed by Wackernagel in the Deutsche 
 Literaturzeitung for 1886, column 221 fg. (No. 7) ; by Delbriick 
 in the Literarisches Centralblatt for 1886, column 290 (No. 9) ; by 
 Fr. Stolz in the Neue Philologische Rundschau for 1886, column 
 J 37 fe- (No. 9) ; by Walter Prellwitz in the Gottinger Gelehrte 
 Anzeigen for 1886, p. 755 fg. (No. 19 of September 15) ; by Kautz- 
 mann in the Berliner Philologische Wochenschrift for 1886, 
 column 597 fg. (No. 19) ; by Peile in the Classical Review for 
 1887, Vol. I, No. 4, p. 103 fg. ; finally by a writer in The Nation 
 (New York) for 1886, April 8 (No. 1084, p. 304). Moreover, 
 Brugmann has carried this theory bodily into his treatment of 
 Greek accent in his Grundriss der Vergleichenden Grammatik, 
 Vol. I, p. 543 fg. 
 
 There are in the list just mentioned as fair scholarly names as 
 can be mustered from the ranks of the workers in Indo-European 
 philology, and yet I venture to say that Wheeler's book has hitherto 
 not been subjected to the kind of criticism which it deserves. 
 The glamour of his attractive method and the many excellent 
 observations in detail have blinded his readers to the fundamental 
 errors upon which his book is built. 
 
 In the following I shall endeavor to show that his theory is 
 untenable on account of the following misconceptions : 
 
 i. He regards the recessive accent as one which manifests itself 
 only on the penult or antepenult, only on the third or fourth mora 
 from the end, and fails to recognize the fact that dissyllabic words 
 of two moras (XtVey above), and monosyllabic words can also be 
 
 having a dactylic final cadence, become paroxytones in Greek. To a criticism 
 of this thesis we may hope to return at some future time. This is the fourth 
 of Wheeler's five theses, and as we shall not be concerned with it we will omit 
 it in the count. 
 
accented, either etymologically or recessively, precisely as dis- 
 syllabic or polysyllabic words of three or more moras. 
 iLj~) j 2. He fails to recognize the fact that barytone dissyllabic tro- 
 W^t jchaic words are, with a regularity which knows practically no 
 1 ti* exception, accented recessively, not etymologically. 1 
 VJ""*^ fa -A 3. Throughout the treatise the difference between circumflex and 
 -acute accent is practically ignored, while in reality a circumflex 
 upon the same syllable as an acute indicates in the vast majority of 
 cases not only a difference in the quality of the tone, but also a 
 difference of position. The circumflex accent marks an accentua- 
 tion further away from the end of the word than the acute. 
 
 4. He has obliterated the difference which is manifestly exhibited 
 in the scope of the recessive accent in the domain of the finite 
 verb on the one hand, and of the remaining word-forms on the 
 other. 
 
 5. Wheeler was led to his identification of the recessive accent 
 with a secondary accent by a fact hinted at by Wackernagel and 
 expanded in my article, p. 43 (23). My statement is : ' Enclisis and 
 recessive accent are ruled by the same law of three morae.' . . . 
 
 ' If we take the Cases . . . avdpairos ris, noiSes rives, Xoyoi rii/ey, WC 
 
 have in every case an enclisis which is rectified or rather cut short 
 by the law of three morae as exhibited in the general recessive 
 accent.' The identification of the recessive accent of the verb 
 with the secondary accent of a group consisting of an orthotone 
 word plus an enclitic word is the keynote of Wackernagel's and my 
 own theory. In the group avdpanov nva the secondary accent clearly 
 goes hand in hand with the existence of a second word ; the enclitic 
 secondary accent of ^pol^Ba in the group frybv (pepoip.f6a for 
 
 1 The Trpurov -tpevdos which vitiates Wheeler's theory manifests itself very 
 clearly in his statement on p. 2 : " Den ' recessiven 7 accent mussen wir also so 
 aufnehmen wie wir ihn vorfinden : als einen accent, der auf der antepaenultima 
 oder paenultima ruht, je nachdem ob die endsilbe kurz oder lang ist." This 
 statement shows, as does the entire treatise, that the ' recessive ' character of 
 the following accentual types has in reality escaped his notice : (i) olda (^^) ; 
 (2) Awre? (^^}; (3) /?? (-) ; (4) P&v, 3d plur. aor. (w). So also, p. 6 : " Nun 
 ist aber der recessive accent kein specifischer accent, sondern vielmehr ein ac- 
 centprinzip,und fasst in sich paroxytona, proparoxytona, und properispomena." 
 This is true, but it embraces, furthermore, perispomena, and short monosyl- 
 labic oxytona. The two statements are, moreover, inconsistent : the pro- 
 perispomena which are introduced as ' recessive ' on p. 6 are excluded by the 
 statement on p. 2, as also impliedly on p. TO: " Ich gehe so weit und nehme 
 diesen nebenton fur jedes spondaische, iambische oder polysyllabische wort 
 
(=1. ^L.yugom bheroimedhd) is also due to the second 
 word. Wheeler, in claiming that a singfeword in sentence-nexus, 
 e. g. *yeVo/u>off Sk. jdnamdnas, developed the same secondary 
 accent, demonstrably associated only with presence of a second 
 word, advances a hypothesis which is unlikely on the face of it, 
 and whose untenableness will be demonstrated in the course of 
 this essay. 
 
 We turn now to a review of the several theses propounded by 
 Wheeler and reported above. 
 
 On pages 13-38 he attempts to prove his first thesis, namely, 
 that monosyllabic words and dissyllabic words with short final 
 syllables retain the original etymological accentuation intact. As 
 far as long monosyllables are concerned this statement is correct 
 in the equation Zeu$-z=Sk. dydtfs = I. E. dieus, but it is incorrect 
 in the equations vavs = Sk. nails = I. E. naus ; p.vs = Sk. mus == 
 I. E. mus; /3oCff~Sk. gdusl. E. jous. 1 The difference 
 between the accentuation of vavs and Zeus-, as far as quality and 
 position are concerned, is clearly the same as that in fifjv II. 13. 
 297, ftco Eur. Ale. 864, when compared with /3a? II. 6. 65 ; S> (from 
 t7//u) Soph. Ph. 816: s II. i. 434; arfv II. n. 744, o-rS> Eur. 
 
 Ale. 864: crra? II. l6. 231; 65) Soph. O. C. 480: 6tis II. 23. 254; 
 dS> Od. 20. 296: dovs Od. 15. 369; <pqv II. l8. 326, ffis II. 5. 473, 
 
 0ij II. 2. 37, <a Find. I 2. n, $& Aesch. Ch. 91 : <j>as II. 9. 35 ; 50 
 (Sua>) II. 17. 210 : dis Xen. Cyr. 5. 5. 9; <C II. 6. 253, 14. 232: $ts 
 Od. 18. 410, Find. Ol. n. 20; y v s>v II. 4. 357, Hes. Th. 551. 3, 
 yi/S II. i. 411 : yvovs Soph. El. 731 ; $&} II. ii. 451, <f>6G) PI. Polit. 
 
 266: $0as Her. 3. 71, 9. 46, (wro-)00aff II. 7. 144; rXiy II. 5. 385: 
 
 rXus- Soph. O. C. 1076 ; *dp< in <J7rdfyxo Ar. Pax, 234 : (mro-^pas Od. 
 17. 516; ^j/, 2 'I was,' II. 2. 96, Soph. Tr. 414, etc. (see Veitch, 
 Greek Verbs Irregular and Defective, 1879, p. 225), rjs ' he was/ 
 Doric e. g. I A. 342. 3, Lesbian, Theocr. 30. 16, Tegeatic, 
 (Gelbke in Curtius' Studien II 40; G. Meyer, Griech. Gramm. 2 
 p. 432), ^, 'he was,' II. 5. 9, Soph. Tr. 9, Thuc. 2. 3, etc. 
 (Veitch, ibid.) : <ai/ Horn. Hymns 19. 32, Soph. Ag. 767, etc., 
 el's Doric, Lesbic (KZ. XXVII 393). We may add the circum- 
 
 ] I shall endeavor to show below (p. 18) that the equations /3wy= Vedic 
 g&m t and /3f Vedic gas, as also Z//P Vedic dyam, are probably correct for 
 accent as well as the sounds. 
 
 *% 'I was,' Aesch. Ag. 1637, Soph. O. C. 973, etc. (Veitch, p. 225), is Attic 
 contraction of the old perfect-form rja = Sk. asa. I. E. esm. The subjunctive 
 Attic u is also a contract form from Epic w. 
 
 c ) / 
 V ^ ' 
 
8 
 
 fleeted monosyllables : ^, ' he said,' II. 6. 390, 22. 77, Od. 3. 337, 
 22. 292, Theocr. 22. 75, Plato Rep. 327 ; r, Vj 'I said,' PI. Rep. 
 328, Luc. Philop. 23. Here also perhaps belongs K*J) ' he scraped,' 
 II. ii. 639 : see Veitch, 379; G. Meyer 2 , p. 47. The accent of the 
 augmentless imperfect xp*i v > Find. Fr. 100, Soph. El. 529, 579, etc. 
 (Veitch, p. 707), can be considered significant only in so far as it 
 may perhaps reflect the accent of ^ (xpi) plus rjv) ; cf. G. Meyer ', 
 p. 430, note 2. Further instances of long monosyllabic oxytone 
 participles are : KTUS in Kara-*? II. 22. 323, Aesch. Sept. 965, Eur. 
 
 I. T. 715; (aico-^o&els HippOCr. 5. 176; *(a7ro-)/7>as in drrovpas, II. 
 
 1.356, etc. ; (a7ro-)KXa?, Anacr. 17 (Bergk) ; (eVi^Ti-ras-, Anth. n. 
 407, (aTTo-Vrus 12. 105; (e7ri->Xa>? H. 6. 291 ; (ay^t-)^Xcos G. Meyer 2 , 
 
 P- 459- 
 
 No one can fail to admit that the difference between the oxytone 
 J accentuation of these long monosyllabic participles and the peri- 
 I spomenon of the finite forms is fundamental : that in fact the accent 
 f of the participles is etymological, and that of the finite forms is 
 recessive. As ffiv l is to /3us, so are XiW II. 10. 406 : \ura>v II. 9. 
 194 ; oiSe : eldws ; ireKovdc i TTfTrovQo)?, etc. Now the circumflex of 
 vavs, pvs, /3ovs, j3ff, as well as the circumflex of Aeolic Zeus, TT, 
 etc., differs from the acute of Zew in the same way : it is recessive. 
 The same difference is to be found in a considerable number of 
 nominatives, consisting of a long monosyllable, for which no 
 etymology, or only a partial one has been found. 2 The following 
 are oxytone, and have presumably preserved the old accentuation 
 of this type. In a number of cases there is a conflict of authorities, 
 which is indicated under the word discussed : 
 
 nipt ' month,' Doric ufa Ionic /xV s : I. E. stems mens-, mes-, Vedic 
 mas (? mdng-catil, Grassm.); Lat mens-is, Goth, mena, Lith. menu, 
 Old Irish mi, Old Bulg. mesecl. 
 
 xQav ' earth ' : Vedic stem ksam-, Zend zem- y Lat. hum-us, Lith. 
 z /me, Old Bulg. zemlja. 
 
 X^v f goose ': Doric xv : Sk. hansd, Lat. anser, QHG.gans, Lith. 
 zas\s, Old Bulg. gasL 
 
 xw 'hedgehog' (Hesych.): Lat. her; cf. Cu. Etym. 5 , p. 200. 
 
 1 fifjv is not Vedic gam, etc., but (Jsev for -^ *{3qv Vedic -^ gfim, etc. 
 
 2 It affords me sincere pleasure to acknowledge that I have been aided very 
 materially and most intelligently in making the following collection of mono- 
 syllabic nouns by a member of my seminary for Greek grammar, Mr. Henry 
 Clarke, A. M., formerly Fellow and now Fellow by Courtesy of the Johns 
 Hopkins University. 
 
 3 /xe/f is wrongly perispomenon in Stob. Eel. 1,27, p. 556; see Chandler 566. 
 
 
'spleen': Ved. pllhdn-, Zd. spereza-, Lat. lien, Old Bulg. 
 slezena. 
 
 Bfo 'wild beast,' Lakon. o-^p, Aeol. (p/?p : Lat.ferus. 
 
 f? s , plur. IMS, up* ' strength ': Lat. vis. 
 
 6a>s, &a>6s, stem 6af- 'jackal ' from root QtF, Sk. dhdv ' to run.' 
 
 |3Xa|, As/ok, ' slack,' ' silly '; cf. dupXawv and the Sk. roots mid 
 1 to wither/ mlech ' to babble.' Cf. Am. Journ. Phil., Vol. VI, p. 48. 
 
 <po>p, (pcopo?, 'thief: <pepo>; cf. \^z.\..far,furis. 
 
 <np;}, o-cprjKos, Doric <r<pa|, oxpa/coff : Lat. vespa, OHG. wefsa, Lith. 
 vapsa (?). 
 
 o-Kvi^j vKvi<$>6s and O-KI/TTTOS- ; also Ki/fy and O-KTX//- ' a kind of ant,' 
 cf. Old Bulg. sknipa 'culex'; cf. Lob. Par. 114, Cu. Etym. 5 694. 
 
 pf^, pTTrdy, 'mat': Lat. scirpus, OHG. sciluf; cf. Cu. 5 352. 
 
 xeip, Dor. x^p, ground form, in Timocreon fr. 9 B., \eps ' hand ': 
 Sk. hdrdmi. 
 
 \is ' smooth ' (cf. XTro?)> stem yXtr : Lat. glittus, Lith. -/zYw.? 
 'smooth'; see Cu. Etym. 5 p. 367. 
 
 TTOVS, Dor. Trds-, Hesychius : TTMS'TTOS. VTTO Acoptewv : Sk. pad, Old 
 Norse /^/r. The accent of TTOVS is in no wise significant for the 
 accentuation of long monosyllables in general, as this form of the 
 nominative is certainly secondary ; see KZ. XXV 14. The writing 
 irovs occurs and is supported in some measure by the grammarians : 
 see Lobeck Paralip. 93, Chandler 566 (p. 163). For Doric nS>s 
 see below, p. 15. 
 
 7rpot', npoiKos ' gift,' Ionic npotg ace. to Etym. Mag. 495, 32. The 
 word is reported as perispomenon by Herodian, but apparently 
 this is incorrect : see Gottling, p. 242, Chandler 566 (p. 163). 
 
 Spw^ avdpcoTTos (Hesychius). Probably a compound = i/ 
 cf. the Vedic stem nr- 'man.' Cf. also i/o>\^ (= ^-cty) 
 o\/m (Hesych.); Lob. Par., p. 118. 
 
 cug, alyos ' goat.' There is some authority for the circumflex in 
 Attic; see Lob. Par. 99; Chandler 566. 
 
 ddX = auXa^ ' furrow.' &X is reported in Orion and Arcadius ; 
 see Lob. Par. in, Gottl. 242, Chandler 566 (p. 163). 
 
 7rro> ' crouching with fear '; cf. TTTCOO-O-OO ; 7rr< is reported by a 
 grammarian, Gottl. 243. 
 
 T|, ace. oca, also T^, nom. plur. facs ' a grub which destroys vines '; 
 Lob. Par. pp. 103. 104; 101. 115; Curtius Etym. 5 461. 
 
 pis (late pfi/), gen. pivos, 'nose,' and Bis (late 6iv), gen. 6iv6s 
 ' heap,' are universally reported as oxytone, 1 but there is good 
 
 1 Cf. also pe/f, pe^of and Oeif, 6eiv6^, Lob. Par. p. 91. 
 
10 
 
 authority for both \is and \ls ' lion' (Cu. 5 366), ids and ids ' wood- 
 worm ' (cf. Sk. kita ?). The authorities are cited and discussed 
 Gottl. 241, Lob. Par. 92, Chandler 566 (p. 162), Misteli, Zur 
 griechischen betonung, p. 116. 
 
 K\LS : Lat. cldvis, ' key.' Here also there is authority for the 
 circumflex : Lob. Par. 92, Chandler 566 (p. 162). The Doric 
 (Theocr. 15. 33) *Xd, K\aKos no doubt belongs here, though the 
 mode of its derivation from K\fis is obscure. 
 
 6evs, Doric = 0e6s, ace. 6evv. The circumflex is reported : 
 Chandler ibid. Likewise Doric \evs = Xaas. 
 
 Furthermore the following are unanimously reported as oxytone : 
 'thrown': root-forms /3eXe-, /3X^- ; nXws, 'swimmer': TrXe'co, 
 'thief: /tXeVro) ; /3i? 'cough,' cf. /3?}orcro; O-KWI//- 'Owl,' cf. 
 
 6f]s ' serf/ cf. ridq/u ; rpa> ' caterpillar,' cf. rpa>ya> ; Sat's-, 6>}s- (Lob. 
 Par. p. 82), &UTOS ' feast ': daia> ' to divide'; 0p/?i/, Dor. <ppa>, ' breast, 
 mind'; 6dp 'louse'; p^v (late) 'lamb,' cf. Curtius, Etym. 5 p. 345; 
 xpvs ' skin '; pa, paydy, and later po)|, pwyo's ' berry'; 1 /c^ 'seagull,' 
 cf. Cu. 5 p. 567 ; &j ' wood-worm '; <9pi^, 6piiros ' wood-worm '; <ppig, 
 (pp$y6s ' ruffling, ripple '; K\<UV ' sprout '; *reiV, Kret/ds- ' comb '; Qaty 
 'flatterer'; 6/xcos-, 5/icody 'slave': Sa/ia-o>; yu\^, -yuTrds- 'vulture'; ypu^, 
 yplTTos ' griffin '; o-^ ' sore,' cf. O^TTO) ; o-jy?, o-ed? (as though from 
 a-evi) later gen. O-TJTOS-, ' moth '; o-cptjv ' wedge '; ^^ ' gall-insect '; 
 \^f, \|ri^dff ' crumb '; ^ap, ^dpds, Ion. ^;p, ^pd? ' starling,' cf. Cu. 6 
 355 5 pty ' brushwood '; irpw ' dew-drop '; TTU^, irvyos, late form 
 of TTuy^ ' buttocks '; yXyv, late form for yXrjvrj ' pUpil '; /oyp, /c/^pdy, 
 'fate': /ce/pco; <pd>9, ^xord? 'man'; TT^, Tri/Tyd? 'suffocation': 7n>fyo>; 
 po)| 'cleft': pr'jywfjii; fog, prjyos, in imitation of Latin rex, regis\ 
 o-wp, o-eip, Chandler 565 ; oty, Et. Mag. 344, 55, gen. WTTOS ' eye,' 
 o-Koty ' ^Mpa (Hesych.), Lob. Par. p. 115. The grammarians posit 
 a nom. Kpas for gen. Kpuros, ace. Kpara ' head '. 
 
 For 8ovg and 5p^ see Lob. Par. p. 102 ; 8us (Cu. 5 p. 237 writes 
 Sdo-is ib. 87 ; ^'p, ^pd? ib. 76 ; KI/W^ *<pXdy (Hesych.) ib. 118 ; 
 ib. 94 ; Xa\^ ib. 1 1 1 ; XO>A//-, \<on6s = Xo)7r/7 (^Xa/iuy Hesych.) ib. 
 118; Trpew ib. 93; ravs (peyas, TTO\VS Hesych.) ib. 91. Here we 
 may mention x pv (G. Meyer, 2 p. 430) if the word is indeed of sub- 
 stantival origin. 
 
 The following proper names consisting of a long monosyllable 
 are oxytone : 
 
 Zeus-, Boeot. Aevs', Zys (owes its rj to the accus. zfjv: Herodian 2. 
 911.9, from Pherecydes), Dor. Zas (ibid.) ; z^, Dor. Zav, Boeot. A^, 
 
 1 Cf. also dpai-, Lob. Phryn. p. 76 ? 
 
II 
 
 hysterogenous nominatives abstracted from gen. Zqi/o?, znvl. All 
 nominatives except Aeolic Zevs agree in their oxytonesis. Cf. G. 
 Meyer, Gr. Gramm. 2 324. The following also are oxytone: 'Pan/, 
 
 XdV, "Hp, Etp, Ncop, T\T)s, K\r)S, Kp?;s, Tpa>s, TXj^y, $pr')S, Tvijs, Neuy, 3>Xeu?, 
 2>7p, "Q\^, Fetp (r/p), 2etp, Kap, Mr^j/, Upon/, Hay, IIap, 'Pap ('Papoy and 
 
 ', &0us (3>6avTos), iipu (or Hpa|). Hesychius has 0po> 'Xt/zos- ; 
 (see Lob. Par. 120). For*Atp see Lobeck Par. p. 74 ; 
 B^X and BciX, ib. 70; B/Jp, ib. 75, note 8 ; rXoW (also T\ovs and rX< ff , 
 ib. 95) ; Avcvs, ib. 92 ; Ilpa/ib. 94 ; Mfc, ib. 82 ; Ma*, ib. 88 ; 'Pat, 
 ib. 99; C P^, ib. 113; 'Pity, ib. 117; Sovp, ib. 77; *^et'p, ib. 74; $f 
 (Boeot. or Doric for 2(pt'-y), ib. 104. 
 
 The following particles consisting of a long monosyllable are 
 oxytone : M, Boeot. /m', Elian pa, =.Vedicma, Zd. and Achem. ma =2 
 I. E. me: f}, Boeot. el' = Vedic vd (enclitic) =. I. E. ve\ l rco? 'so/ 
 perhaps zz: to an I. E. ablative tod plus a later s-, &s 'so,' a corre- 
 sponding form of an I. E. stem io- (also a>? ; cf. Chandler, 934) ; 
 ty ' now, already '; dai ' then ' (ri Sat * what then ? ') ; wj (j/) r6/ Am), 
 i/at ' verily ' : Lat. nae ; nai ' and/ Cu 5 . 138 ; ^v, Doric pav 'certainly, 
 truly '; 7rXr> (Doric rrXav), Cu. 5 281 ; b^v, Hyper-Doric dav 'long, for 
 a long while'; \ai (eVl atVxpoupytW, Hesych.) ; at 'O that, would 
 that '; oi', interjection of pain ; o>, interjection of pleasure and pain ; 
 Boeotian TOW, rou ' thou/ are oxytone ; not is the Argive form for 
 Trport, Trpos- ; pa, poetic for pddiov (cf. Lob. Par. 119), is probably con- 
 tracted from a dissyllabic form : see Osthoff, Perfect, p. 447, note ; 
 for /Spa see ibid. ; <p? ' as, like/ Cu 5 . 394, occurs both as $77 and <p^. 
 
 The scope of the circumflex in long monosyllabic nouns is as 
 follows : 
 
 vavs, Epic and Ion. VTJVS, Dor. vdi>$ : Sk. ndtis t Lat. ndv-is, Old 
 Pers. ndvi, Old Irish nau. 
 
 /3oOs, Dor. j3a>y, accusatives jSovi/ and ps>v : Vedic gdiis, ace. sg. 
 -w, ace. plur. gas, Zend Yz0, OHG. ^?^, Lettish guwis, Old 
 Bulg. govedo, (Lat. to). 
 
 ypaCs, Ionic yp^vy, 'old woman' : ytpw 'old man/ cf Sk.jarant- 
 'old man.' 
 
 Ti-avs, TroCis, ' boy, girl/ on old inscriptions on vases, cf. Trat? ; see 
 Benfey Wurzellexicon II 73, Cu 5 . 287, Gust. Meyer Gr. Gramm.' 
 p. 312, note. The circumflex may be assumed upon the basis of 
 the proportion : ypavs : ypats (ypafLbi) = navs : Trals (Tra/tfis) , see 
 Meister, Zur Griechischen Dialektologie, p. 2. 
 
 1 A very different view is advanced by Froehde in Bezz. Beitr. VII 327 fg. 
 and supported by Osthoff, Zur Geschichte des Perfects, p. 128-9. 
 
13 
 
 y\at 'owl'; cf. Vedic gldiis 'tumor' (?). For the oxytonesis 
 of the word in Doric see below. 1 
 
 ous, Cretan and Laconic avs, Ionic $ (inscription from Delos), 
 Doric &>ff 'ear' : Lat. aus-culto, aur-is, OHG. drd, Lith. aus-is, 
 Old Bulg. uch-o (Gen. us-es-e^), Old Irish 6. The declension is 
 heteroclitic : the stem of the oblique cases is ova-r = *OVO-V-T 
 I. E. ous-n-, contained in Goth, stem ausin-, nom. auso, gen. ausms. 
 See De Saussure, M6inoire, p. 224.* 
 
 P.VS, ace. /jivi>, 'mouse' : Sk. mus, Lat. mas, OHG. mus, Old 
 Bulg. mys-i. Cf. also a-p,vs ' 6 p.vs and o-^is ' pv?, o-p-tvQa, both in 
 Hesychius. 
 
 <rvs and us- 'swine': Zend /iu, Lat. sil-s, OHG. su,NHG.sau; 
 cf. also Sk. su-kards, Old Bulg. sw-inija, Goth, sv-ein. Accusatives 
 (rz/ and \iv* 
 
 dpvs ' tree, oak ' : Sk. dru-s, Zd. dru, Goth, tow, AS. treow, 
 Engl. /r<?<?. 
 
 /3a{)ff, a word of unknown meaning, Joann. Alex. TOVLK. napayy. 
 pp. 7, 35; see Lobeck Paralip. p. 91; Chandler 566 (p. 162). Cf. 
 also /3aO ' aSos- uvdovs (Hesych.) 
 
 Hesychius has K&S' elpKTrj 'enclosure'; cf. KolXos, Lat. cav-us, etc. 
 Cu. 6 , p. 157. 
 
 yfj ( Ionic-Attic) ya ( Doric), 3 ' earth.' The contraction from 
 Ionic yea or from *yaa (^yr}o- in Attic yeo)-) is unproved. The 
 etymology seems unknown. See Cu. Etym. 5 , p. 177 ; G. Meyer 2 , 
 p. 200, note 2. 
 
 /3A?7p fitXeap ' bait ' is Aeolic ; the circumflex therefore proves 
 nothing. 4 For iras ' all,' /cJ/p ' heart,' $<s- ' light,' els ' one,' p.va = 
 mina, see below. 
 
 o-Koip, O-KUTOS ' dung ' : Vedic $dkrt, $aknds. There is some evi- 
 dence in favor of oxytonesis ; see Lobeck Par. 77, Chandler 564 ; 
 Liddell and Scott, sub voce. The Dorians are reported to have 
 accented o-*&>p ; see below. 
 
 o-rals, crrairoff, ' dough from wheaten flour.' There is authority 
 for trrais also : Lob. Par. 88. For Kpavg and *a (?) see Lob. Par., 
 p. 100; Tras ibid. 78; o-rpovs (Hesych) ibid. 93. Hesychius has 
 
 also 
 
 1 For traces of oxytone y/lai>f outside of the Doric dialect see Lobeck Paralip. 
 109; Chandler 566 ; R. Meister, loc. cit., p. 3 ; Liddell and Scott, sub voce. 
 
 " G. Meyer's explanation of ovg as a contract form from *oi>ao<;, *ovo, *6of 
 does not seem to me a likely one, see Gramm. 2 , p. 326. 
 
 3 For Doric da see Ahrens, Dial. Dor. p. 80 ; Cu. 5 p. 492 ; for Cypriote C 
 G. Meyer 2 , p. 200, note 2. 4 Etym. Mag. and Hesychius report fi^fo. 
 
12 
 
 Neuter nouns consisting of a long monosyllable are regularly 
 perispomenon. 1 In addition to ovs, aK&p, arals there are : nvp ' fire ': 
 Umbr./zV, OHG. fuir, fiur. Herodian 2, 919, cites a form nvip 
 from Simonides of Amorgus, which leaves room for the suspicion 
 that nvp is contracted. But the genitive irvpos (with gradation 
 of stem and shift of accent), as well as Umbrian pir (cf. sim 
 and sify probably equal to Gr. vv and vs), points to the inde- 
 pendent origin of v in the word ; <pou, probably the Pontic name 
 of the plant valerian ; pa, ' rha barbara '; 8&, Epic for d&pa ' house '; f 
 
 Kpl, Epic form for KpiQr) 'barley'; (pap ' garment ' (cf. (papoi) is re- ^C / 
 ported by Arcad. 124 as perispomenon: Liddell and Scott write 
 (pdp ; 63> is reported as an apocop. form for 6&pa, Anth. P. 6, 
 85 ; ppl according to Strabo was used by Hesiod for ftpiapov, see 
 Liddell and Scott sub fBpi ii. For aav (and a-dv) see Lob. Par. 77 ; 
 for arrow ibid. 1 20. The names of the letters are of neuter gender 
 and perispomena : /ML, vv, gv, nd (Trl), 2 p&>, rau, (pel (<t)i X e * OtO> tyh & 
 
 Monosyllabic accusatives singular, long in quantity, are peri- 
 spomenon : zi/i/, Doric ASj/ = Vedic dyam ; Dor. p&v = Vedic gam 
 (/3o)j> is analogical after nom. (Bovs) ', ypavv, vavv, dpvv, oi>v, Iv, p.vv t 
 Xtv, KIV, K\flv, Qtvv (Doric, ace. of 6evs = ^edy). 
 
 The corresponding plurals are also circumflected : Dor. /3o>s 
 (Theocr. 8, 47) = Vedic gas, Zend tf0. The primary character 
 of Attic /3ovy is doubtful (G. Meyer 2 , 362). Further vavs, ypavs, aOs, 
 
 The following proper names are perispomenon: Q&V, Tpfjs, 2i)p, 
 , Tav, QfvO, Qvs (Qvv, Lob. Par. 86), K5)s, T\S)s (also rXoC? and 
 ws), Xws>, TXws-, Kpws-, ASs-, Aay, $v>v Q Chandler, p. 162, note i), 
 a, Ionic Qp?ji, Homer and the tragedians Qpjg (contracted?), 
 . For^HX see Lob. Par. pp. 70, 116; liav (?) ib. 71. The 
 remaining ones are of the first declension : Tpas, Bay, Xi/as, lias-, eSs, 
 ; $Xa (Herod. 4, 178: our editions read $Xa). For npSs 
 the acute is also reported : Chandler 566 (p. 163). 
 The following particles consisting of a long monosyllable are 
 perispomena: i/Cj/'now': Sk. nu (and nu) t Zend nu, Old Bulg. 
 ^j/w^ ; Ionic, Aeolic, Boeotian and Doric o>i>, Attic olv ' then '; av 
 1 again ': Lat. aut, autem ; ^ ' truly ' (cf. fj above) ; &>, vocative 
 participle (cf. ^ above) ; a, interjection of astonishment and pity ; 
 at, interjection of wonder, blame, etc. : Lat. ai t a loan-word (cf. al 
 above) ; <peO ' ah, alas '; <pO ' fie,' cf. Lat. fue, fu ; 8a (<eG fia, 
 
 1 Cf. Chandler, 563 ; Phil. Anzeiger for 1883, Vol. XIII, p. 580. 
 2 Cf. Meisterhans, Gramxnatik der Attischen Inschriflen, pp. i, 24. 
 
14 
 
 da) J fipvv (7re0 ' tO Cry fipvv '; rav, 1 Cu. 5 686 ; ypv (ovde ypv). A 
 number of particles, representing frozen case-forms of pronominal 
 stems, are perispomena : ?}, rf s TTO, but also TH?, rf enclitics with 
 supplementary accent ; -n}, ^, etc. (old instrumentals : G. Meyer 2 , 
 p. 365) ; wot, of (locatives) ; TTOV, ov, genitives ; 7r$-, 2 *$-, &>$ (also cos : 
 ablatives), etc. The circumflex of these particles is no doubt in 
 many cases old, antedating the period of the recessive accent, 
 as in 7701, ol (cf. Hanssen in KZ. XXVII, p. 614), or a genuine 
 rhetorical circumflex, as in interrogative particles and interjections. 
 They are given here for the sake of completeness. Similarly na, 
 @a, p.a are hypocoristic vocatives, and have vocative accent. 
 
 The report that the Aeolians circumflected every long monosyl- 
 lable : Zws, p&g, TJTCO, Span//-, xn* e * c '> * s universally accepted, and 
 accords with the remaining facts of the Aeolian system of accentu- 
 ation ; see Ahrens, Dial. Aeol., p. n ; Anton Ftihrer, Ueber den 
 lesbischen Dialekt, p. viii. There is, however, a report of 
 Choeroboscus, somewhat doubtful as to its meaning and scope, 
 to the effect that the Aeolians treated monosyllables as oxytones, 
 and there are also special reports to the effect that individual long 
 monosyllables in Aeolic received the acute. Cf. above sub vocc. 
 rfv and jSX^p, and Chandler 567. I see no good reason to doubt 
 the universally accepted perispasis of long monosyllables in this 
 dialect. 
 
 The Dorians are reported to have accented y\av and o-Ko>p in 
 distinction from Attic y\av and o-*6>p ; see Gottling, p. 243 ; Ahrens, 
 p. 27 ; Johannes Schmidt, KZ. XXV 14 ; R. Meister, Zur Griech- 
 ischen Dialektologie, p. 3; Hanssen, Philologischer Anzeiger, 
 XIII, p. 580. The temptation to see in this an instance of 
 vacillation between etymological and recessive accentuation must 
 be resisted. R. Meister (ibid.) believes that Doric y\avg and o-*a>p 
 have ' den alterthiimlichen accent (i. e. no doubt what we here call 
 etymological accent) gegeniiber der im aeolischen dialekt regel- 
 m'assig, im ionisch-attischen hier und da eingetretenen perispom- 
 enierung bewahrt.' 
 
 We must consider, however, that the Doric dialect exhibits 
 many cases of suspended perispasis, as in nroxes, TTTCOKU?, Traces, 
 
 (pares, *A.\Kp.av for A\Kfj.av (fr. ' A\*cfiacoi/) , irav for TTO.V, in the aorist 
 
 infinitives o-rdo-cu, Xuo-cu, deipai and dp.vvai, and that in general there 
 is to be observed something like a ' processive ' reaction against 
 the ' recessive ' tendency, the latter being probably Pan-Hellenic, 
 
 1 Also written rav. Cf. also rav ' cv. 'Arr^/cwf (Hesych.), G. Meyer 2 , p. 382. 
 
 2 Cf. 7T<y TTOV. bdev. nodev, &uptel$, Hesych. 
 
15 
 
 but certainly Attic-Ionic and especially Aeolic. This is exhibited 
 
 in Cases like dvdpwnoi, yvvaiKai, yvvaiKes, opvtdes, eVraaav, e'Aa/3oi>, in the 
 
 accentuation of navTwv (Gottl. 246), and (ppurrjp for Attic cppur^p, etc. 
 These are certainly secondary whether they mark a secondary 
 phonetic charige~oir analogical transformation (dvdpwnoL after dvOpw- ! 
 
 - , r , -\ i T rir 
 
 Treat/, dvOpamois, etc. ', (j)paTrjp alter Trarjjp). 1 preier therefore to 
 regard y\a\> and o-/c%> as the oldest forms on Greek ground, and to 
 consider the coincidence of the Doric accentuation yXavg, a-nup with 
 the etymological accent as accidental. 
 
 Hence I cannot subscribe to Wheeler's first comparison in sup- 
 port of his thesis that monosyllabic words have retained their 
 etymological accent unchanged. He writes Doric TTWS- (!) 2 ~ Sk. 
 pad-=. I. ^L.pots. As far as I know the only source upon which 
 this 7ro>y is based is the gloss of Hesychius : VMS ' TTOS. vn6 Avpuwv. 
 In writing Trews- Wheeler is probably guided he does not say so 
 by the consideration advanced by Joh. Schmidt in KZ. XXV 14. 
 The latter judges from the reported oxytonesis in Doric of y\av and 
 <TK<ap that Hesychius has misreported this accent, as he frequently 
 does. Whether this be so or not, even the hypothetical Doric 
 *7ro>r is not to be compared directly (as Schmidt and Wheeler do) 
 
 1 R. Meister in his very thoughtful tract, Zur Griechischen Dialektologie (I. 
 Bemerkungen zur dorischen Accentuation) endeavors everywhere to explain 
 these cases of ' procession ' as due to one of two causes : either some analogy 
 within the paradigm of the word in question, or to a suspension of the ^tpia^aotq 
 KarrfvajnacfjiEvr] of the other dialects. I do not believe that he is on the right track, 
 as he does not point out any reason why the manifold special phenomena of 
 accent in Doric agree in promoting the accent towards the end of the word. 
 If looked at in detail, special causes may be found readily enough for every 
 instance of Doric procession : avOpuiroi might well be accented after dvdpcjTruv ; 
 Trrw/cef after Trrwf ; slidpov after kAafiopev ; nav might have preserved the old 
 participial accent, as in Pan-Hellenic GV/J.TTUV, -xpoirav, Trdfnruv ; fyparrjp might 
 be oxytone after the analogy of Trar^p (Attic ^parrjp') ; ontip and y^avt; might 
 represent instances of preserved I. E. oxytonesis in Doric, etc., etc. Yet each 
 one of these explanations quite reasonable when considered singly is 
 rendered improbable because they all operate in the same direction. Why 
 do not some of these Doric accentual modifications operate in the other direc- 
 tion, i. e. ' recessively,' if they are merely the results of individual effects? 
 Unless we wish to burden the Dorians with an apparently Ideological choice 
 of such analogies as tend to 'procession,' we must assume that the reported 
 Doric instances of ' processive ' accent they are not actually quotable in the 
 language are due to some single fact in accentual phonetics whose scope and 
 cause we are unable to determine owing to ftie deficiency of the tradition. We 
 will encounter later on the same difficulty in Wheeler's (and Prellwitz's) 
 attempts to explain the phenomena of Aeolic accentuation. 
 
 ' J See p. 13, as also Prellwitz, loc. cit. p. 764. 
 
16 
 
 with Sk.pad, I. E.pots, but may be as well regarded as a second- 
 ary Doric product out of Pan-Hellenic *7ra>s, as long as an Attic- 
 Ionic TTO>S is not discovered. 1 
 
 Tray and neuter nav are of especial interest, as illustrating the 
 existence of recessive accentuation in monosyllabic nominal stems. 
 They represent an I. E. non-thematic participle = Sk. *$vdnt-, 
 I. E. Tcuni-. This conclusion may be derived from Benfey's old 
 discovery that Sk. gdqvant-, for *sd-gvant, is Greek a-Travr-', see 
 Wurzellexicon II 167 ; Orient und Occident I 573 ; " Das indo- 
 germanische Thema des Zahlworts 'zwei' ist DU," Abhandlungen 
 der Gesellschaft der Wissenschaften zu Gottingen, XXI 7. The 
 word *nas lost its connection with any finite verbal system very 
 early : hence the recessive accentuation, avoided by other parti- 
 ciples, lodged within a finite system, forms like (pas, /Say, rXoy, etc.; 
 see above. On the other hand the Attic genitive plur. Trdj/rcoj/, 
 dative plur. na<ri are still participial like /3dW<i>, o-rai>ra>j/, etc. The 
 neuter participle irav, which is still preserved in anav, ird^Trav, o-vp-nav 
 and npoTrav, in changing to irav, adopted both the quantity and 
 accentuation of Tra?. 2 
 
 I have left out of question the perispasis of the vocatives of long 
 monosyllabic nouns. At first sight this is one of the strongest 
 proofs of the recessive character of the circumflex in cases like vavs, 
 etc. The classical example Zeu, circumflexed vocative : Zevs, oxy- 
 tone nominative, is reflected in Vedic dydus, vocative with inde- 
 pendent svarita : dydiis, nominative with udatta. Each pair goes 
 back to I. E. couplet dieu(s), vocative with I. E. independent 
 svarita: dieus, nominative with acute accent. But the very an- 
 tiquity of the recessive accent in the vocative makes it chronolog- 
 ically unfit as an argument for the recessive character of the type 
 The recessive accentuation of the vocatives Zfv, a-wrep, av- 
 , etc., is proethnic; that of vavs at best Pan-Hellenic. That 
 does not exclude the fact that both processes, the old recession of 
 the tone in the vocatives to the first mora (Zeu), and the substitute 
 for enclisis which is contained in the last resort in vavs, have finally 
 worked to the same end, but the recessive accent in Ze does not 
 directly prove that the circumflex in vavs is also recessive. Cf. 
 also the note on p. 17. 
 
 1 Of course the oxytonesis of Trove proves nothing for TTW^, as the entire end- 
 ing (-oi>f) is secondary ; cf. above s. v. Trovg, and G. Meyer, 2 77, 313. The 
 latter also writes TTW^ in both places. 
 
 2 7rav is explained differently by J. Schmidt, KZ. XXV 14. rcav occurs even 
 outside of composition proper, e. g. Find. Ol. 2. 93. 
 
A theory which ignores in its consideration of long monosyl- 
 lables so important and far-reaching a difference as that exhibited 
 in the collection above is not calculated to inspire confidence. 
 Wheeler does not anywhere allude to it ; much less does he make 
 an effort to explain it away. Though such an attempt seems in/ 
 any case an after-thought, I have nevertheless surveyed the ground 
 as carefully as possible for some expedient by which the circumflex 
 accent of the type vav$ and Aeolic Zfvs, 7rro>, etc., might be ex- 
 plained without the assumption of recessive accent. But I cannot 
 say that I have succeeded in finding one. Hanssen in KZ. XXVII 
 612 f., by employing successfully Leskien's important little article 
 ' Die Quantitatsverschiedenheiten im Auslaut des Litauischen,' 
 Archiv fiir slavische Philologie, 1881, Vol. V, p. 188 fg., has proved 
 that a kind of circumflex accent 1 existed quite extensively on final 
 syllables of words in I. E. times, but his proof does not include asingle 
 case of a monosyllabic noun-stem, nor indeed a single monosyllable. 
 I hold myself ready to accept the original character of the circum- 
 flex wherever there is good ground to accept it. Hanssen may 
 perhaps not have gone far enough in his assumption, or at least 
 in the express statement of such accentuation, inasmuch as he 
 gives only examples in which at least two of the languages com- 
 pared (Greek, Lithuanian, and Gothic) testify directly to the exist- 
 
 1 At least it appears as the ordinary circumflex in Greek. It is in reality that 
 accentuation of a long syllable in which the summit-tone either permeates the 
 two moras of which the syllable consists, or in which there is'double summit 
 accentuation (Sievers' Phonetik 3 , 203; Bloomfield, Historical and Critical 
 Remarks, 27-8). In Lithuanian grammar this mode of accentuation is called 
 ' geschliffener ton,' after Kurschat's precedent. ' Geschliffener ton' is a mis- 
 nomer for ' geschleifter tcTn ' (Leskien orally: Brugmann, Grundriss, I, p. 562, 
 suggests ' schleifende ' sc. betonung). We may designate in English this 
 mode of accentuation by the term 'drawled tone,' or 'slurring tone' (in the 
 acceptation of the word in music). This I. E. drawled tone probably has a 
 very definite scope (see Hanssen above), and we must for the present keep it 
 differentiated from the genuine I. E. circumflex (svaritd) of Zev Vedic dyaus 
 which is the result of the fusion of an acute plus a grave ( A ). In the ordinary 
 recessive and contract circumflex : olda oIo, rpei -pee^ (see below) we must 
 recognize a third type of circumflex, phonetically very similar to the second, 
 but chronologically very different, inasmuch as the earliest date which we can 
 assign to it is the Pan-Hellenic period. Even that is probably too early for 
 the contract circumflex; see G. Meyer, 2 p. 140. Heterogeneous accentual 
 materials of the Greek have become fused by one mode of designation. In 
 Sanskrit the first two kinds are designated respectively by the genuine svarita 
 and the riddtta ; the enclitic s-uarita (Whitney, 85 fg., Historical and Critical 
 Remarks, p. 45) generally takes the sign of the genuine svarita. 
 
i8 
 
 ence of the circumflex. For Trodwv =. Sk.padam we may suppose 
 an I. E. pedom with 'drawled' or 'slurring' tone (seethe footnote 
 on p. 17), from the testimony of KWWV = Lith. szunu = I. E. Ttunvm 
 (despite pindm, Atharva-Veda III 9, 4). Accordingly it seems to 
 me not unlikely that the perispasis of long monosyllabic accusatives 
 sg. and pi. may be founded upon this I. E. ' drawled ' (' slurring ') 
 tone, although the nature of the case is such that we may perhaps 
 never be in the position to prove it. If we survey the list given 
 above : sg. z^, D6ric &av ; Doric puv, Attic fioZv, ypavv, vavv, 8pw, 
 
 (TVV, VV, flVV) \IV, K.IV, K\lv, 6fVV ', pi. DOHC fi&S, Attic /3oij, VOVS, (TVS, VS, 
 
 dpvs, pvs, we are struck by the solidarity of the perispasis. Yet 
 much of it is palpably secondary : Doric Aav has a Hyper-Doric a, 
 as zfjv = I. E. diem (leaving the accent aside). Attic /Sow, pi. /3os ; 
 vavv, plur. vavs ', ypavv, pi. ypavs, are secondary formations : povv, 
 vavv and ypavv after the pattern of the nominatives sg. ftovs, vavs, and 
 ypavs ; the ace. plural fiovs, vavs, and ypavs in their turn after the pattern 
 of the ace. sg. povv, vavv, and ypavv. The accusatives p.vv sg., pvs pi. 
 are also secondary issue of the nom. pvs, inasmuch as the stem is 
 p,vs = I. E. mils- ; the proper accusatives are */^o for */i{W and pW 
 for*/iuCTaff (cf. G. Meyer 2 , pp. 321, 346). K\V and /cXels are pretty 
 certainly secondary to K\f1da, Horn. <\r)lda, pi. K\el8as. Doric 6evv 
 ace. to 6evs 6e6s can only be the product of a later propagation 
 of the type, as the word is certainly originally a dissyllable. I do 
 not venture to decide the question of the originality of the accusa- 
 tives o-vv, vv, 8pvv, 1 \lv, K.IV] pi. avs, vs, 8pvs. G. Meyer 2 , 331 says: 
 ' Den i- und u-st'ammen kommt v zu : idpiv, noXiv, KIV, \lv, o-vv,' etc. In 
 361 he places the ace. plur. ovs, dpvs, vs, among dissyllabic v-stems 
 like yevvs, VKVS, etc. It seems from one point of view that we ought 
 to expect for long monosyllabic stems accusatives of the type sg. 
 o<pva 2 = Sk. bhnivam, atapathabrahmana III 2. i. 29, and Ho- 
 meric ocppvas = Sk. bkruvas. Cf. also Lat. suem. On the other hand 
 Umbrian sim and sif point to vv and vs. Be this as it may there re- 
 main in any case the accusatives z^v, @S>v, @S>s, where I believe that 
 both form and accent are original. I venture the following recon- 
 struction: zf)v Vedic dyam'=. I. E. die(u}m ; p&v Ved'\cg-am r= 
 I. E. 3<?(VJ7# ; p&s =1 Vedic g&s l. E. ^d(ujs ; i. e. I suppose that a 
 slurring or drawled mode of summit accentuation accompanied the 
 utterance of these monosyllables. Direct unequivocal proof of this 
 
 1 There is in the Rig-Veda no single case of an ace. sg. in -urn. 
 
 2 The form occurs late, Oppian. Kyn. 4. 405, but can certainly make no claim 
 towards direct identification with Sk. bhriivam despite the formal equality. 
 
19 
 
 assumption is impossible, as 'drawled' tone and udatta have the 
 same designation in the Veda (see the footnote on p. 17), but the 
 morphology of the forms renders it likely enough. This is the 
 only instance in which the circumflex may be even approximately 
 proved original in the case of long monosyllables, and even if we 
 consider it not impossible that it may yet be proved proethnic in 
 other monosyllabic words, 1 it is nevertheless quite incredible that 
 types in every other respect so perfectly parallel to Zevs = dydiis 
 as vavs-ndiis, and fiovs-gdiis, should have differed in this singular 
 manner in their original accentuation. 
 
 I have thought of the analogy of contract forms. So fa (gen. 
 rjpos) is the lyric form for eap (Alcman 24) ; Attic ol$ =z &'$ for of is 
 = Lat. ovis =. Sk. avis\ Attic (p6ols = $0diV, * cake, pill'; $,{ = 
 
 j3oa|, ' a kind of fish '; o-rrjp, (TT^TOS ^^ ffre'ap, oWaror, ' tallow '; wp ^i: 
 
 oap 'consort'; Kpfjs, Doric for Kpeas 'flesh' Ved. krdvis; vovs =: 
 voos 'mind'; x v s (wovs, Lob. Par. p. 93) =. xyoos 'surface'; Attic 
 0povs = Opoos ' noise.' Attic xs is declined as though it were 
 a monosyllabic stem (gen. x oo/ff > dat. ^ot, ace. plur. ^ovs ; cf. /Sons-, 
 jSot, etc.), yet it is in reality a dissyllabic stem like <pop6s (xoFos, XQOS, 
 contracted x^0 5 third declension forms of povs pofos also occur, 
 e. g. genitive po6$ like ^ody ; see Lob. Phryn. 454. 
 
 The epic word x^p, Krjpos 'heart' is explained by Gottling, p. 425, 
 Leo. Meyer KZ. V 69 ; Misteli, Uber griechische Betonung, p. 118, 
 as contracted from the later Ktap (tragic, lyric), in the same way as 
 ^>, rjps fr m */> Curtius, Etymologic 6 , p. 143, points out the 
 independent character of the stem (*KJ?p?), and Brugmann, in 
 Curtius's Studien IX 296, note, explains the accent as an imitation 
 of the contract accent of ^>, rjpos. The persistence of the cir- 
 cumflex in the declension of the word renders this explanation 
 fairly plausible (gen. Kijpos, dat. Krjpi) ; cf. also Wackernagel in KZ. 
 XXV 280. 
 
 The explanation of Attic <ps>s as equal to Homeric, etc., <pdo$ is 
 the current one; see e. g. Brugmann in Curtius' Stud. IV 173; 
 G. Meyer 2 , p. 326. No one as far as I know has, however, offered 
 any explanation which bridges over the difference in the stems of 
 the two words: ^a/W- but (par-. The stem $o>r- seems to me 
 better comparable with Vedic bhds, neuter in the oldest language, 
 the T being ' adscititious ' (Brugmann, ibid.) ; cf. (pao-cpopos and 
 bhds-kara ' shining.' But the circumflex may be due to the fact 
 
 1 The regularity with which long monosyllabic neuters are circumflected is 
 worthy of attention ; Chandler 563 and above. 
 
20 
 
 k" 
 
 f n 
 
 that the form was felt to be associated with faios as its contract 
 form. 
 
 The difference between ef? (Doric fc : Heraclean tablets i, 136) 
 and ovdeis, fj,r]8fLs (lacon. ovfys IA. 79, 4) is as yet unexplained ; cf. 
 Gottling, 246; Misteli 118; Hanssen, Philologischer Anzeiger 
 XIII, p. 580. The evidence of Attic TOVS = Kretic TOW, etc., points 
 to the fact that the oxytone accent is the fundamental one, and that 
 the perispasis of efs- is secondary. I venture to propose the analogy 
 of the contract circumflex of rpels (Homeric and Attic) = rpces- 
 (Inscr. of Gortyna, 9, 48) Sk. trdyas = I. E. treies. The vig- 
 orous analogical influence of numerals upon one another has long 
 been noticed ; see Osthoff, Morphologische Untersuchungen I 92 
 fg. ; Baunack, KZ. XXV 225 fg. One can understand easily how 
 *i? might be influenced by rpeis, so as to become a?, while ovdeis, 
 firjdfis would be preserved from this contamination by being ' out 
 of the count/ and possibly by the blurred consciousness of the 
 origin of the words. 
 
 The circumflex of p.va may possibly be due to its assumed deri- 
 vation as a contract form from Ionic /Wa (Hdt. 2, 180) ; fj.vd is 
 probably more original than /uWa. It is Hebrew-Phoenician nw ; 
 cf. Lat. mina, Sk. mana (also a loan-word) ; cf. Zimmer, Altin- 
 disches Leben, p. 50. 
 
 The circumflexed form e* ' thou art ' I have explained previously 
 in a totally different connection and from a different point of view, 
 as owing its orthotonesis in distinction from the enclisis of /u 
 , etc. to the analogy of verbal forms with the circumflex due 
 to contraction : see Historical and Critical Remarks, p. 59 (39). 
 I see no reason for retracting this view, and I am not aware that 
 any authority has objected to it, or that a more plausible one has 
 been advanced hitherto. 1 
 
 Much of the perispasis of long monosyllables in Attic-Ionic may 
 be in this way ultimately exhibited as secondary. The small 
 investigation given above may serve rather as a guide in the matter, 
 than lay claim to an exhaustive examination of the possibilities in 
 that direction. Yet it seems to me that no one will be found will- 
 ing to undertake the thankless task of explaining away all the 
 j instances of the circumflex on long monosyllables recounted above, 
 K without calling in the aid of that retraction of the accent which is 
 \ an infallible law in the monosyllabic forms of the finite verb. 
 
 
 1 Osthoff 's explanation of el, as equal to ef for I. E. esl, with 'nehentonig- 
 tieftonigem vocalismus des personal-suffixes ' rests upon too slender a basis of 
 fact to inspire confidence. See Zur Geschichte des Perfects, p. 18, note. 
 
21 
 
 And even the sturdiest determination in that direction would be 
 of no avail on account of the Aeolic dialect. Wheeler has taken 
 no account of the constant perispasis of long monosyllables in 
 Aeolic, ZeCv, 7rr<, xw fyw^ etc -> etc. : we may assume that he has 
 consistently placed circumflex and acute upon the same level here 
 as throughout his work. Prellwitz, in his review, loc. cit. p. 757, 
 recognizes this deficiency and proceeds to remedy it. But his 
 processes do not in my opinion redound to the advantage of the 
 cause. Prellwitz would explain the perispasis of Zevs and TJTCO as 
 due to the analogy of z/?i/, Zev, nTwua. This is well possible when 
 taken by itself: it would be simple paradigmatic analogy. If we 
 consider, however, that Aeolic vaos, j/ai'~ non- Aeolic vaos, vat; ft6os, 
 non-Aeol. /Sods-, KWOS, furthermore if we consider Aeolic 
 , a6(f)os, Trorapos, etc., the fallacy of the assumption of such 
 analogy becomes apparent. It is of the same sort as that criticized 
 above in R. Meister's explanation of suspended perispasis in 
 Doric : there is no reason provided for the infallible motion of 
 these supposed analogies in one direction. I shall return to this 
 point more fully later on in connection with Wheeler's explanation 
 of the accent of Aeolic Ovpos and o-d^os. I fail to see how the 
 assumption can be avoided, that certain accentual types, namely 
 the ' recessive ' ones, have propagated themselves in Aeolic with- 
 out reference to the function of the forms involved. 
 
 Nor will it do to assume that the difference between acute and 
 circumflex became indistinguishable at a period so early that the 
 Homeric difference between Zeus and vavs may be accounted as non- 
 significant. Deutschmann, in his treatise De poe'sis Graecorum 
 rhythmicae primordiis, Malmedy, 1883, p. 3, assumes this state of 
 things for the first century A. D., but his assumption is fitly 
 refuted by Hanssen in the Phil. Anz. XIII, p. 422. As late as 
 Babrius the difference between acute and circumflex must have 
 existed, for he categorizes words like Kdpva> and rouro together, and 
 differentiates KO.\LV<* and p^rrjp ; cf. Hanssen, Rheinisches Museum 
 XXXVIII, p. 239 fg. He could not have put the accentuation of 
 Kd/ifo) and roOro upon the same level without recognizing that the 
 fundamental difference between acute and circumflex is rather a 
 topical than a qualitative one. The acute accent on a long syllable 
 means in reality that the second mora has the acute, the first one 
 being grave; /^r;/p /^eeV/jp ; the circumflex on a long syllable 
 means the accentuation of the first mora ; TO^TO = rdvro. Hanssen, 
 Phil. Anz. XIII, p. 422, without offering anything new, well describes 
 
ll 
 
 22 
 
 the ordinary (not 'drawled') circumflex as follows: ' There is in 
 fact in Greek but one {grammatical} accent, the acute ; but this can 
 Jill but one mora. Short syllables therefore admit of but one kind of 
 accentuation : their vowel carries the acute ; syllables with a short 
 vowel, long by position, also admit of only one kind of accentu- 
 ation : their vowel bears the acute, they also have the tone on the 
 first mora; syllables containing a long vowel (or a diphthong) 
 admit of a twofold method of accentuation ; the acute may stand 
 on the first mora of the vowel (circumflex), or the acute may stand 
 on the second mora of the vowel (acute on a long syllable). The 
 designation of accent is deficient in marking an acute upon a short 
 vowel and an acute upon the second mora of a long vowel by the 
 same sign, and it is an unlucky circumstance that a special name 
 and a special mark was not constructed for the acute on a long 
 syllable, but for the circumflex. In reality the words a\s and irovs, 
 which carry the same accent-mark, are not accented alike, and 
 they are not both accented differently from ovs, but a\s and ovs are 
 accented alike on the first mora, while TTOVS is accented on the 
 second mora. The difference between acute and circumflex in all 
 probability was given up along with the differentiation of short and 
 long vowels, at a time when the difference in the pronunciation of o 
 and o> was given up ; at that time the difference in the accentuation 
 of r6, rw, and ro> was no longer felt.' 
 
 This applied to vavs and Ztvs means that the accent of vavs is vdi>s, 
 
 that of Zeus- is Zevs. 
 
 The difference is a topical one, not one of quality only, and our 
 previous considerations have made it probable that no other source 
 than the recessive accent of the formally corresponding verbal 
 forms will be found for accent of vavs, which deviates from the I. E. 
 ndiis. The sporadic or unsettled character of the recession in the 
 noun-types (vavs, pous but Zeus, /3a?) is, as far as I can see, well ex- 
 plained by the statements on pages 30, 50, 57 and 62 of my 
 treatise, quoted at the beginning of this paper. The recessive 
 accent in the verbal forms is enclisis, or rather a substitute for it, 
 therefore a grammatical quality, which covers the entire ground ; 
 in the noun it is secondary, no doubt analogical, apparently on the 
 way towards absorbing it. This process of absorption is complete 
 in the Aeolic dialect. The manner in which this analogy has 
 operated I shall endeavor to delineate below. 
 
 The considerations given thus far are in themselves quite suffi- 
 cient to unsettle one's belief in Wheeler's hypothesis, with its 
 
23 
 
 fundamental idea of a subsidiary tone. In his assumption of a 
 subsidiary tone on the third or fourth inora from the end, there is 
 no provision made for the change from original */3^(Y) i. e. */3ee 
 =. Vedic gat, to ffi, i. e. * for j. */3q ; from original *vavs, i. e. *i/aus- 
 to vavs, i. e. vdvs. Here it would be necessary to assume a subsidiary 
 tone on the second mora from the end. Will any one be found 
 willing to believe that a single long syllable was burdened with a - 
 summit tone and a subsidiary tone, and that at a certain time, to 
 use Wheeler's own terminology, 'trat ein Schwanken ein, das 
 Spater zu gunsten einer der beiden Accentuation ausfallt ' ? 
 
 I believe that enough has been said to show that that part of 
 Wheeler's first thesis which refers to monosyllables is not tenable. 
 Still less do I find myself in the position to adhere to the second 
 part of it. The claim that dissyllabic forms with short final syllable 
 retain the inherited (I. E.) accent seems to me quite groundless. 
 
 At all times comparisons like the following have been considered 
 legitimate : ol8a, olo-da, oiSe = Ved. veda, vtftha, veda =. I. E. 
 uoidm (?), uoistha, uoide ; ei/u, n = Ved. e'mi, eti I. E. Simi, 
 e'iti ; r,a = Vedic asa = I. E. ism (?), perfect ind. act. first sing. ; 
 ^e (v) Vedic asa, I. E. ese, perfect third sing. ; dual and plural 
 forms of the imperfect of the copula: ^OTOI/, ^i/, ^ore = Vedic 
 astaniy asma, asta ; /terrai =. Vedic gete ; ^o-rai = Vedic aste. 
 
 Comparative grammarians are usually pleased to speak of such 
 cases as being equal sound for sound. But is it true that any 
 respectable authority has ever ignored the thoroughgoing differ- 
 ence in the accent ? There is absolutely no reason for doubting 
 that the Vedic udatta of veda y e'mi, etc., represents the I. E. acute 
 or * cut ' (' gestossen ') tone on the second mora of the first syllable, i 
 To my knowledge no one has ever hinted at a similar accentual, 
 condition in the cases above (foiSe, et/u, etc.). The circumflex on 
 verbal forms of this type: r w, i. e. dissyllabic trochaic forms, 
 has always been understood to be ' recessive,' utterly independent 
 of any accentuation prior to Greek period, olda =. Vedic veda is 
 ' recessive ' precisely in the same sense as /3J/ = Vedic gat. No 
 one will be found so bold as to assume a proethnic ' slurring ' tone 
 or a proethnic svarita (cf. p. 17, note) for all dissyllabic trochaic 
 verbal forms, whatever their connection. The very fact that the 
 Trept'o-Trao-ts in such cases is Kar^vayKua-pevr] shows that all etymological 
 accentuation is superseded by the ' recessive ' law. 
 
 In the case of long monosyllables, the verb is recessive, without 
 exception : the noun, according to our discussion above, has 
 
 i 
 

 
 followed the verb only to a certain extent. On the other hand the 
 entire body of trochaic dissyllables : verbs, nouns, pronouns and 
 other parts of speech are properispomena, 1 aside from the Doric 
 
 instances like 7n-&>Kes, irTUKas, Traides, (frwTes, orao-ai, \vo-ai, etc., in which 
 
 the perispasis is suspended. I have indicated above that these 
 cases are in my opinion due to a genuine ' processive ' reaction 
 against the recessive tendency, and that Meister is of a different 
 opinion. Whatever their explanation may be none that is abso- 
 lutely convincing has to my knowledge been advanced it must 
 not be forgotten that they rest almost entirely upon reports of 
 native grammarians, and not upon good ' quotable ' material. 
 The reported cases of suspended perispasis in trochaic dissyllables 
 are no way fit for testimony against the assumption that this peri- 
 spasis is Pan-Hellenic, and practically without exception. 
 
 That the circumflex here, as in the case of the monosyllabic 
 perispomena, indicates a difference of position, as well as one of 
 quality, we may learn, aside from the general description of the 
 value of the circumflex given above, from a single example of the 
 type w, namely 01*01 'houses,' if we compare it with 01*01 'at 
 home.' It has been known for a long time that the syllable -KOI 
 of olKot counted for one mora and that the accent was therefore 
 driven forward to the first mora of ol- (i. e. 01*01 =. OJKOI), while in 
 oucoi the second syllable counted for two moras, and the summit 
 tone was therefore placed upon the second mora of ol- (i. e. oi'/coi 
 OIKOI). Misteli, Ueber griechische Betonung, p. 128, came very 
 near to an explanation of this difference in his sentence : ' Wenn 
 OIKOI von 01*01 absticht und at ot des Optativs iiberall seine Lange 
 wahrt, so ist im ersten Falle i eigentlich Casussuffix des Locativs, 
 das mit dem Stammvocale o regelrecht zusammengezogen wurde, 
 und die Folge davon ist eben die Lange, wahrend im Nomin. Plur. 
 jedenfalls die Rede nicht von Zusammenziehung sein kann, so 
 wenig als beim altindischen Pronominalausgang z. B. #=roi,j// 
 
 = 01 U. S. W.' 
 
 Leskien, in the article quoted above, broke the way toward a 
 full explanation of this extraordinary difference by showing that 
 
 Joh. Alexandr., p. 5, 17: Trdaa tyvati [ncmpa Trpo j3pa^eiaf XIJKTIKJJC, <]>' 
 
 rdvov, TcepiaTrarat ; Gottling, p. 42. So unfailing is this law that forms 
 which really ought to have an etymological accent upon the second mora of 
 the first syllable are absorbed by the type ^. So Sovvai ~ *6ofevai Vedic 
 ddvdne (*(^ofsvai is hypothetical because Cypr. dofevac has no accent) ; 
 for irafid^ ace. to Meister, Zur Gr. Dial., p. 2, cf. above, p. n, etc. 
 
25 
 
 the Lithuanian exhibited two kinds of syllable-tone (in distinction 
 from word-tone): 'cut' tone ('gestossener' accent) and 'drawled' 
 tone (' geschliffener ' accent) ; cf. the footnote on p. 17. Syllables 
 which have the summit tone, as well as syllables without 
 the summit tone (grave syllables), exhibit this difference in the 
 different treatment of the vowels. Hanssen in KZ. XXVII 612 
 fg. successfully applied Leskien's discovery to Greek. Accord- 
 ing to this theory, ot in a final grave syllable is long if the same 
 syllable with the summit tone has the circumflex ; on the other 
 hand, if the same syllable with the summit tone has the acute, ; 
 then it has the value of a short syllable. If we compare the 
 two nominatives plural 01*01 and KaXoi with the two locatives 
 singular oiW and 'lo-fycot, we can see that the syllable 01 of 
 OIKOI is counted short because it would have the acute when 
 accented (cf. *aXoi), while the 01 of o'Uoi is counted long, because it 
 would be circumflexed if it were accented (cf. 'io-0/uor). We may 
 say that the second syllable of 01*01 has the * sub-acute ' accent, or 
 ' sub-cut ' tone, while the second syllable of O'LKOL has the ' sub- 
 circumflex' or 'sub-drawled' tone. Cf. also Brugmann, Grundriss 
 
 ! PP- 533, 539- 
 
 Nothing could show more directly the fact that the circumflex 
 of oucot really represents an acute on the first mora of/cot = dl/coi, the 
 acute of OIKOI an acute on the second mora : the result arrived at 
 independently in the case of the long monosyllables is repeated 
 here from a new point of view for trochaic dissyllables ; the circum- 
 flex of these represents a summit accentuation of the first mora. We fl / jJ 
 must therefore pronounce as incorrect the following of Wheeler's \ _ 
 
 comparisons (p. 20 fg.) as far as the accent is concerned: 90? ~ flt-^ ^- 
 Sk. yavat ; rfjos = Sk. tavat ; aWos Sk. e'dhas ; ei6\>y = Sk. vedas ; 
 *avyos (from tpiawyqi) = Sk. QJLS\ irlos = Sk. pivas ; olfjios Sk. 
 
 t 
 
 emas\ oi/uos- = Sk. dhsas ; d5>ns (Hesych.) z=Sk. dati-\ d^a = Sk. ? t i^J 
 
 daman- ; ff/xa r= Sk. vdsman- ; 0v/xa = Sk. b human- ; x.^^ a (x f ^^ a ^ s F ( 
 
 misprinted) = Sk. homan- ; olpa zn Sk. e'man- ; ^et/ua : Sk. adverbial 
 
 locative heman ; olQap : Sk. udhar, etc. 
 
 One may be fairly surprised that Wheeler discriminated against 
 these forms and shut them out from his theory of a secondary 
 accent. They could have been well enough provided for under its 
 shelter. He allows the secondary accent in iambic dissyllables 
 (three moras) : why should it not also have developed upon trochaic 
 dissyllables (of the same number of moras)? On p. 16 he says: 
 1 The only cases of monosyllabic stems like novs, nodus, cty, OTTOS, K.T.A. 
 
26 
 
 which were fit to receive the subsidiary tone were the genitives 
 and datives plur., and it is worthy of note that the accentual ex- 
 ceptions which are almost unanimously reported by the gram- 
 marians appeared in just these cases (naidcov, etc.).' It seems 
 unlikely that he, whose methods are most rigorous, should have 
 allowed himself to override such considerations, because the as- 
 sumption of recessively accented trochaic dissyllables would intro- 
 duce exceptions into almost all his categories of dissyllabic words, 
 and thus prevent the clean-cut arrangement of the words under 
 his category I. Was he prevented from making the assumption 
 of a secondary accent by the unlikely result : a principal and a 
 secondary accent upon the same syllable, <u/i, i. e. <j)fv^a with a 
 secondary accent developed upon the first v (cf. above)? His 
 attitude here again seems to me, however, best described by saying 
 that he has not regarded the difference between circumflex and 
 acute as an expression of topical difference any more here than 
 above in the case of long monosyllables. 
 
 We may, I think, take this for granted from the way in which 
 he explains certain points in the recessive accentuation of the 
 Aeolic (Lesbian). 
 
 It has long been customary to regard the Aeolic accentuation as 
 an extreme carrying out of the recessive tendency, without refer- 
 ence to the number of moras involved in a given word. Excepting 
 a few uninflected, therefore solitary words, the prepositions and 
 conjunctions dvd, Kara, 8id, /zera, ardp, avrdp, for whose oxytonesis 
 there is good grammatical authority they have the secondary 
 proclitic accent the whole mass of word-material has assumed 
 the recessive accentuation, exactly as it holds in the Pan-Hellenic 
 personal forms of the verb. Wheeler, p. 24, attempts to show that 
 his theory of a middle tone needs to be applied also when one is 
 face to face with the over-emphatic application of the recessive 
 principle in the Aeolic. The Aeolic declension of Attic 0vp.6s 
 is 0vp.os, BtfjLoi (dupoio), Buna, GCfjiov, dupe. He is not willing to 
 recognize independent recession in every case, but applies his 
 theory rigorously. Accordingly it is possible that the genitive 
 and dative should have changed their accent, because the former 
 contained three syllables and the latter is spondaic : there was 
 room for the secondary accent. On the other hand, 6vp.os and 
 6vfj.ov must have obtained their accent analogically from such forms 
 as BifjiU) (!). But if the process is simply analogical, why not Qvpos 
 and dt^ov ? When Meister, Zur griechischen Dialektologie, makes 
 
27 
 
 the assumption that Doric TTTUKCS, yW*-, etc., owe their acute to 
 the analogy of 7rro>, yuwj, etc., the assumption of analogy, whether 
 made correctly or not, is a reasonable one. But if 6i>p.& shall affect 
 forms like 0i>p6s, Ovpov by ' blind analogy,' why the change to the 
 circumflex? Thence I conclude that Wheeler did not bear in 
 mind the difference expressed by circumflex when compared with 
 acute : neither the topical difference nor the difference in quality. 
 Brugmann, who has adopted for his ' Grundriss ' Wheeler's 
 theory without expressed reserve, treats the matter corresponding 
 to Wheeler's first thesis in 676, i. Do I err in believing that 
 although he adopts this thesis in his statement, he ' hedges ' in 
 the choice of his examples? The paragraph in question is as 
 follows: 'Zweisilbige Worter mit kurzer Endsilbe lagen ausser- 
 halb der Wirksamkeit des Secundaraccentes und hielten im allge- 
 meinen den ererbten Worton fest. 7r68a Tro'Sey, 7ro86s noo-l : &\padam 
 padas, padds, patsii. rpcls aus *Tpe(t), rpia-i : trdyas trisii, trevre 
 #e/ca, eTrrd: pdilca ddgct saptd. Trepi: pdri. apKros: rksas, etc. 
 s : jdmbhas. ITTTTOS : dgvas. dyos : ajds. 6pB6s : urdkvds. 
 : midhdm. Qvpos : dkumds. ayvos : yajnds. K\VTOS : grutds. 
 os : jndtds. fiapvs : gunis. f)8vs : svddiis. peQv : mddhu. ytcpos : 
 ndbhas. avdos : dndhas. e!)ua : vdsma. With the exception of the 
 single example ef/^a : vdsma, there is no word with a circumflex 
 mentioned in the passage, and Brugmann may have admitted 
 fljj.0. because the diphthong is not Pan-Hellenic (Aeolic e/^a and 
 
 yeor/Ltara iz: */ r eo-/zara, Doric yjj/m, yecrrpa . crroX^, Hesych.) It does 
 
 not seem to me to be without significance that Brugmann has 
 failed to put his signature to accentual equations like efSoy 
 ve'das, etc., even while adopting the theory which would render 
 them legitimate. 
 
 I believe that I have thus far shown that two types of mono- 
 syllabic and dissyllabic words do not respond to Wheeler's theory, 
 inasmuch as their explanation from his own point of view demands 
 the assumption of secondary accent under circumstances not pro- 
 vided for by the theory, and under circumstances intrinsically 
 thoroughly improbable. Neither can I give in my adhesion to 
 that part of thesis I which is left after deducting the long mono- 
 syllables and the trochaic dissyllables, namely the pyrrhic dissyl- 
 lables. To begin with, one will naturally be less trustful towards 
 Wheeler's attempt to derive the recessive accent on pyrrhic dis- 
 syllables of Aeolic words, by the analogy of forms within the same 
 paradigm, after the fallacy of such a derivation of forms like 
 
28 
 
 BofAos, Qupov has been exhibited above. One naturally asks here as 
 several times before : Why this untiring consistency in these cases 
 of ' blind analogy '; why is the tendency always forward, why not 
 sometimes the other way ? He explains the change from <ro(f)6s to 
 o-o^os- as follows (p. 24) : ' Obwohl der Nominativ eines zweisilbigen 
 Nomens dem secundiiren Accent nicht zuganglich sein mag, werden 
 doch einige der andern Casusformen es immer sein konnen ; z. B. 
 Cronos- aber o-bfov (with secondary accent). Und die Neigung zur 
 Gleichmassigkeit in der Flexion vermochte dann wohl den einmal 
 in die iambischen oder spondaischen Formen aufgenommenen 
 recessiven Accent durch Analogic auf die trochaischen resp. pyr- 
 rhichischen hiniiber zu fiihren. Solches war durchweg der Fall 
 im aeolischen (lesb.) Dialekte.' But we may fairly ask : Why did 
 not the ' Neigung zur Gleichmassigkeit,' if untrammelled by any 
 other tendency, occasionally equalize in the other direction, if the 
 recession of the accent is due simply to paradigmatic analogy, and 
 not to some other cause ? Wheeler assumes in the fourth category 
 (his No. V) that a vacillation between the etymological and reces- 
 sive accentuation took place in the case when the etymological 
 or inherited accent was nearer to the end of the word than the 
 recessive accent. What is it that deprived the Aeolic from the 
 benefit of this choice? Wackernagel, in his review, recognized 
 the improbability of this view, without refuting it : ' Wol aber hatte 
 er die Consequenzen seines Satzes, dass die Tieftonigkeit der 
 aeolischen Mundart unmittelbar mit dem Secundaraccente zu- 
 sammenhange, besser erwagen sollen. Der Secundaraccent ist den 
 zweisilbigen Wortern trochaischer und pyrrhichischer Messung 
 fremd, und doch sind diese im aeolischen ebenso barytonetisch 
 als die andern. Die par Ausnahmen von der Barytonese auf die 
 sich der Verfasser S. 25 beruft (avd, ardp u. s. w.) sind lauter Worter 
 mit dem von ihm auch sonst anerkannten proklitischen Accent auf 
 der Endsilbe.' (Deutsche Literaturzeitung 1886, Nr. 7, Col. 221.) 
 If the genuine retraction of the accent in Aeolic pyrrhic dissyl- 
 lables is due to the recessive ' principle,' then the existence of the 
 same in the other dialects is a priori probable, for the Aeolic does 
 not anywhere do more than exaggerate the accentual facts of the 
 sister-dialects. As far as oxytone pyrrhic dissyllables are con- 
 cerned they do indeed largely retain their accent, but so do all 
 kinds of I. E. oxytones which are not finite verbal forms. That is 
 the one fortress which has never been scaled by the enclitic 
 accentuation : it is ' the last ditch.' But exceptions to the retention 
 
29 
 
 of this oxytonesis are not wanting in the language any more in the 
 case of the pyrrhic oxytone type (^ 6) than in the case of trochaic, 
 iambic and spondaic oxytones (_ 6, ^ ^, and - ^), and polysyl- 
 labic oxytones. Examples are : vdpos ( water-serpent ' (Homer) =: 
 Sk. udrds ' water-animal ' ; KOJXOS ' shell ' Sk. qaiikhds ; TTO/IS "=. 
 Sk. pur is ; oyKos ' bend, hook ' = Sk. ankds ; ZKTOS ' sixth ' ~ Sk. 
 sast/ia's ; irep-Trros Sk. pafiQathds. Much the simplest explanation 
 of the paroxytonesis of KVK\OS is to regard it as the recessively 
 accented accentual equivalent of Sk. cakrds. KVK^OS : *KfeK\os : 
 cakrds = VTTVOS : *affirvos = Sk. svdpnas, Zend qafno = Old 
 Norse svefn (orig. Germ. *svtfnos). Kluge KZ. XXVI 100, and 
 Wheeler, p. 23, prefer to derive KVK\OS direct from the paroxytone 
 German forms, Anglo-Saxon hveohl, Old Norse hvel (for *hvehl) ; 
 both from a German stem hvehlo-. 
 
 Singular is the explanation which is proposed on p. 33 for the 
 accentuation of abstract nouns in -ns- (-<m)> which are recessive 
 without exception : cf. Historical and Critical Remarks, p. 50 (30). 
 These were originally oxytones, as is shown by their vocalism and 
 the prevailingly reported oxytonesis in Vedic and German. 1 But 
 in Greek the whole type is completely in the bonds of the ' reces- 
 sive ' accent '. ^e'cris, pv^is and pevcris, TTIHTTIS and Trevcris, yevcriS) yvoxris, 
 atipoiais, aio-^o-ty, aXcocriy, ^TTjtm, /na^cris, etc. Wheeler, p. 34, States 
 
 categorically that this accentuation of the type is due to the fact 
 that the abstracts in -m (-o-is) compounded with prepositions were 
 originally accented on the preposition : Sk. dpaciti ' reverence,' cf. 
 aTTono-is ; ut-krdnti ' ascent '; prd-drpti ' haughtiness'; prd-niti ( guid- 
 ance,' etc. They therefore had room for the development of the 
 secondary accent, thus : IK/SOO-I?, di/a/SA^cm, avairtvcriS) dcpatpeo-ts, etc., 
 and from these the accent of the uncompounded abstracts was 
 derived by transfer. This explanation is subject to suspicion to 
 begin with, because the parallel formation of the perfect passive 
 participles in -ids (verbals in -TO?) when compounded with prepo- 
 sitions also accents the preposition, and yet never makes the least 
 attempt to encroach upon the uncompounded forms. Thus we 
 have Sk. vi-cyuta ' fallen apart ': cyutd l moved, fallen '; dva- 
 naddha ( bound down '; naddhd ' bound '; prd-vista ' entered into ': 
 vistd ' entered '; dva-ruddha ' enclosed ': ruddhd ' obstructed,' etc. : 
 
 1 For a few cases in which the accent of abstracts in -ti-s seems to have left 
 the final syllable and passed to the radical syllable in proethnic times, see 
 Bloomfield, Am. Journ. Phil. I 296, and Wheeler, p. 33 (where other refer- 
 ences may be found). 
 
30 
 
 see Bruno Lindner, Altindische Nominalbildung, p. 71. In the 
 same way regularly in Greek e/cSoroy : Soros ; d/i<pi/3X 177-05- : /SAqros ; 
 
 fJL7T\r)KTOS '. 7r\TJKTOS ', aTTOTfXfVTIJTOS '. T\VTT)TOS, CtC. And CVCn if it 
 
 were granted that the absorption of the accent of the uncom- 
 pounded nouns in -TIS (-cuy) by the compounded might have been 
 accomplished, although the old relations in the verbals in TOS were 
 left undisturbed 1 a freaky choice of analogy it does not appear 
 clear in what way the accent of aTroYio-is could have affected the 
 prehistoric *o-i's so as to render it nW. Wheeler's explanation 
 approaches within dangerous proximity of a method which he else- 
 where takes especial pains to deprecate. The recessive accent, he 
 says (p. 6), is ' a principle of accentuation, and embraces paroxy- 
 tones, proparoxytones and properispomena. 2 It would therefore 
 be necessary that the various accentual types arising in the inflec- 
 tion of airoTivis should have affected severally the corresponding 
 cases and numbers of *no-i'r. In other words : 
 
 Nom. sing. dnoTKns changes *TUTIS tO TLO-IS. 
 
 Gen. sing, anorio-ios " *TIO-IOS " TiVios. 
 
 Ace. sing. dnoTto-Lv *ri<riv " rtViv, etc. 
 
 For my part I do not understand in what sense the historical 
 paroxytones TLO-IS and rtW can be imagined due to the proparoxy- 
 tones d-rroTio-is and oTrorto-tj/, unless the principle of recession in Mrta-is 
 and dnoTio-iv, and that too in the most abstract version imaginable, 
 is supposed to repeat itself in TICKS and TLO-IV the very assumption 
 which he contends against. The explanation of the recession in 
 the action-nouns in -TIS (-o-ts) is I think as follows : The large mass 
 of abstracts in the language are recessive or barytone : they are not 
 accented upon the ultimate. So the three most prominent types : 
 neuters in -os (ycvos) ; neuters in -pa (^eu/ia) ; masculines in -o- (Xo^oy, 
 (pdpos). The inherited contrast between oxytone nomina agentis and 
 barytone nomina actionis (<popds : <opos, \jsevdrjs : ^e^fios) is kept alive 
 and even extended beyond its old limits (d6\ixos 'a long race- 
 course,' based upon do\ix6s = Zend daregha, Sk. dirghd 'long,' 
 Lat. largus, in imitation of rpo^os : rpo^os). The forms in *- T ls (*-o-i's) 
 can easily have followed the accentuation of these abstract types, 
 
 1 Nay even there is a vigorous tendency on the part of the accentual type of 
 the uncompounded oxytones to encroach upon the recessive compounds, e. g. 
 dm/zerp^rdf, Trapappjjrdg, KaradvrjTog, ovfKpeproc, etc. ; see L. v. Schroeder, KZ. 
 XXIV 122. For the difference in the function of the compounded oxytones 
 and barytones see Gottling p. 313, Kuhner I, p. 415. 
 
 2 Compare the foot-note on p. 6. 
 
and if the Greek language really brought with it from the common 
 stock sporadic instances of paroxy tones (pdais == Sk. gdtis == 
 Goth, gaqiimps ; cf. above, p. 29, note), these may have helped on 
 that analogy. 
 
 Still more peculiar is the use which Wheeler makes of the unim- 
 paired etymological oxytonesis of several substantival types in 
 order to prove his theory that dissyllabic forms are not recessive. 
 So notably the verbals in -rdy. On page 27 we have a list of 
 dissyllabic forms of that category : faros : Sk. gatds ; K\VTOS = Sk. 
 pnf&& = OHG. hint and Anglo-Saxon hlud y both according to 
 Verner's law from Germanic hlifods. But he nowhere tells us upon 
 what ground he makes the undisturbed oxytonesis of these forms 
 dependent upon their dissyllabic and dichronic character. Are there 
 no trisyllabic verbal adjectives in -r6s with the same claim to origi- 
 nality as the dissyllabic ones ? The types ri^a-ro's- Lat. ama-tus ; 
 (poprj-ros = Lat. dele-tus ; purGa-ros = Lat. aegrd-tus ; /ze^-rds-, o-iceXe- 
 TOS, fpire-Tos = mOflt-tuS, habi-tus ; cpa-ros (epa-cr-ros) '. Sk. ari-S 
 1 friendly ' (? Pick) ; *dapa-Tog in dddparos = Sk. dami-tds =. Lat. 
 
 domi-tus ; yeXa-a-ros *. ye\do> (cf. yeXd-er-(rai) ; Gr. e/xe-ros ^^ Sk. 
 
 vami-tds = Lat. vomi-tus ; further dvi/a-rdf ; SiSaK-rds ; dy^-ro's ; 
 afoOtf-ras, dptdfjirj-Tos ', fidffiXev-Tos are either directly inherited from 
 the Indo-European or are more or less modified ' continuators ' of 
 I. E. types. 1 Their numerical representation is probably more 
 extensive than that of the dissyllabic forms. He does not even 
 employ the ordinary domestic remedial expedient to which we all 
 of us resort for good and for bad of explaining the undisturbed 
 oxytonesis of the polysyllabic forms as due to the analogy of the 
 dissyllabic forms. He simply mentions, 78 pages later (p. 105), two 
 polysyllabic verbals in - T 6s, a^iros and eparor under his fifth 
 division : ' when the inherited accent lay nearer to the end of the 
 word than the place of the secondary accent a vacillation took 
 place, which was decided in favor of one or the other.' They 
 appear here quite accidentally, as it it were, as representatives of 
 those who chose to retain the old accent. Surely the undisturbed 
 oxytonesis in all uncompounded verbals 4 is a definite property of 
 the entire category, inherited from the common period, unaffected 
 
 1 Cf. now especially Karl Ferdinand Johansson, De Derivatis Verbis Con- 
 tractis Linguae Graecae (Upsala Universitets Arsskrift, 1886), pp. 96,97, 100. 
 
 2 This oxytonesis makes inroads even upon the compounded forms : dta/nerprj- 
 76$, TrapappvTos, KaTaOvrrrk, av^eprdr, etc. See Leopold v. Schroeder in KZ. 
 XXIV 122, and note 2 on p. 30. 
 
32 
 
 by all later vicissitudes of Greek accentuation, whether we call 
 them secondary, enclitic, or recessive accentuation. This oxy- 
 tonesis is moreover no doubt to be considered along with the 
 oxytonesis of the very numerous active participles, e. g. e&v Sk. 
 sdn ; la>v Sk. ydn ; ora?; &*{$', 8ovs', upvvsi Sk. Tnvdn\ \nr<av: Sk. 
 
 ricdn ; XeXoiTra* : Sk. ririkvahs-, etc. This brings us to another 
 even more serious manifestation of the same error. Wheeler, in 
 deference to his theory, has divided artificially in his presentation 
 the accentuation of just those oxytone active participles : palpably 
 the most single in origin and treatment. On p. 38 we have the undis- 
 turbed oxytonesis oi\nro)v'=.ricdn ; ewv : uqdnt- ; a-ropvts : strnvdn ; 
 l<ov \ydnt\ eo)v : sdn ascribed to their dissyllabic character. 1 The 
 perfect active participle appears again on page 105 (XeXotTrcos-, Trrfcv- 
 y<bs, TTf(pv<as, etc. : Sk. ririkvahs, bubhujvahs-, babhuvahs-, etc.), as 
 \ though there were no link which binds together the accentuation of 
 
 the dissyllabic and trisyllabic types in their common functional 
 properties. He ought to have been deterred from this error by his 
 own statement (p. 67), that the old middle participles degap.evrj 
 ' cistern,' da^vfj * lowland,' most capable of developing the ' sec- 
 ondary ' accent, remain oxytone. All this is of one piece, and in 
 my treatise these facts were presented in peculiarly strong relief: 
 cf. especially Historical and Critical Remarks, p. 50 (30). 
 
 For the same reason I cannot grant in any sense that examples 
 based upon the undisturbed oxytonesis of dissyllabic adjectives 
 in -p6s prove that dissyllabic forms, as such, preserve the old 
 accentuation. It is true that naKpos=. OHG. magar; frpos 'dry' 
 == Sk. ksdrds 'caustic'; vp6s ksurds 'razor'; further, ^v8p6s 
 Xi/3pos-, (m<pdff, (TaTTpos, etc., are oxytone and dissyllabic, but this 
 fact is rendered pointless as far as the theory is concerned by 
 icpos, lapos = Sk. isirds ; <me-p6s = Sk. piva-rds ; cpvOpos Sk. 
 rudhirds ; raxe-pos 1 , Traye-poy, (pavc-pos, p.ia-pos ', Ka6a-p6s "=. githl-rds 
 (Wheeler, p. 6l), /3pia-p6s, ^aXa-pcs-, 0o/3e-poy, TrXoKe-pos 1 , rpo/ze-poff, 
 
 TTwOe-pos, irovrj-pos, /Mo^^-po?, \iyvp6s, and a large number of others ; 
 cf. Johansson, loc. cit. p. 97. 
 
 Surely any consideration which puts these facts into any other 
 light than that into which the verbals in -roy are put above is mis- 
 
 1 The motive by which Wheeler is induced to account in this way for forms 
 like ALTTUV, e/cwv, aropvijg escapes my comprehension. They are iambic forms, and 
 according to his theory are capable of taking the secondary accent. They 
 would belong along with /IfAofTrwf, Tre^evywf, Tre^vwf, etc., to his category V, 
 aside from the criticism given above. The same difficulty I find below in the 
 case of the agent-nouns in -T^p, -/^", -futv, etc. 
 
33 
 
 leading. These adjectives have preserved the original oxytonesis 
 as a category, and have barring singular cases come in no way 
 under the influence of the ' recessive ' accent, whatever this be. 
 I cannot for my part comprehend at all why we find (p. 36) 
 the nomina agentis in -rtjp introduced in illustration of the first 
 thesis. All the examples given on this very page go to disprove 
 the theory: they are either iambic dissyllables, as dorrjp, Trorfp, 
 
 0er;;p, orarqp, Or Spondaic dissyllables, as Swrjyp, {evKTrjp, yva>(rrJ)p, 
 
 dwrfp. All these are forms which ought to have the * secondary ' 
 accent, and Wheeler's reason for cataloguing them on p. 36 rather 
 than on p. 105 is obscure in the extreme (fy^p occurs in both 
 places). All of these examples, as also yeverfp = Sk. jdni-tar, 
 Lat. geni-tor, aXe^njp : Sk. raksitdr, Aar^p, aXeiTrr^p, x a P aKT W> 
 
 Treucmyp, revKTrjp, TreiOT^p, z/euorqp and vevrrjp, etc. (cf. p. 105), again 
 
 simply exhibit an old oxytone category, left with its inherited 
 accentuation undisturbed by the later vicissitudes of Greek accen- 
 tuation. All of them ought to have the recessive accent in order 
 to prove anything for the theory of a * secondary ' tone. 
 
 I will continue no longer to point out perfectly parallel errors in 
 the rest of the material arrayed in support of the first thesis. Pages 
 13-38 are in my opinion honeycombed with erroneous presenta- 
 tion : the essential difficulties I have pointed out thus far. It will 
 not prove difficult to apply the same adverse criticism to the use 
 which is made of masculines in -/*//? and -/uo>i/ on p. 36 : they are 
 capable of taking the ' secondary ' accent, and belong to category 
 V ; to the treatment of the adjectives in -vs on p. 32 (cf. eXa^vs- on 
 p. 105) ; of the nomina agentis in -6s on p. 29 (cf. doiSoy, dpoifios, 
 dpoyyoff, afjLopyos, etc.). 
 
 Only one other point needs, I think, especial mention, as it 
 involves a view, held as far as I know, universally and yet methinks 
 incorrectly: It has been the custom up to date in comparative 
 grammar to regard equations like K\efos = Sk. grdvas ; avdos "= 
 Sk. dndhas ; Kpef as = Sk. krdvis ; ^6v = Sk. mddhu ; yews = Sk. 
 hdnus ; Tep/ua = Sk. tdrman ; ftiuris = Sk. gdtis =. Goth, gaqiimps ; 
 aKpis = Sk. dgris, etc., as expressions of accentual equality. 1 This 
 may be regarded as true in so far as the acute on a short syllable 
 
 1 An attempt to cite all the literature in support of this statement would 
 involve references to every author who has considered Greek accent from the 
 day of Bopp's Vergleichendes Accentuationssystem down to our own day. 
 I will therefore merely refer to Bopp's work, especially pp. 25-35 > Bloomfield, 
 Historical and Critical Remarks, p. 39 (19) ; Wheeler, p. 26 fg. ; Brugmann, 
 Grundriss I, 676, i. 
 
34 
 
 in historical times in Greek, the udatta on a short syllable in his- 
 torical times in Sanskrit, and finally the I. E. acute (' cut tone ') on 
 a short do not differ from another in quality, barring perhaps such 
 difference as is involved in the question whether they represent 
 chromatic or expiratory accentuation, or a combination of both. 
 It is also true that the summit-tone has never been anywhere else 
 in these types. And yet there is nothing to show that they do 
 not after all really exhibit the recessive accentuation. It must be 
 considered wrong from the point of view of any theory to regard 
 the accent of finite verbal types like fa'pov (II. 3. 245), an augment- 
 less imperfect 3d plur., as preserving in reality the same accentu- 
 ation as Vedic bhdran, I. E. bheron(f], in spite of the perfect 
 coincidence externally. The form cpcpov is recessive, for enclitic 
 f- *(pcpov ; it coincides therefore with Vedic enclitic bharan. This 
 example is typical for every finite verbal form consisting of a paroxy- 
 tone pyrrhic dissyllable. What right have we to assume that it is 
 otherwise in the corresponding types of the noun ? What right have 
 we to regard the 'recessive' accent as suspended in K\efos, etc., 
 while finding it in full force in ^eCSo?, eiSos, epe/3os, etc., and in certain 
 cases of the inflection of nXefo? itself: KKef<(<i)<Qv t etc.? I venture 
 therefore to assert that these types also are in reality recessive. 
 There is in fact nothing but the ' recessive principle ' to be found 
 in all dissyllabic word -forms, aside from the retention of proethnic 
 oxytonesis, and some special deflections in every direction, due no 
 doubt every time to singular analogies between word and word, or 
 words and words. 
 
 If the preceding exposition is at all correct, then the barrier 
 which Wheeler has erected about dissyllabic word-forms must be 
 considered as broken. To begin with, there is no difference 
 between category I and V, aside again from special deflections 
 arising from individual analogical effects. I have shown above 
 sufficiently that he pursues an artificial and misleading method 
 when he categorizes dissyllabic forms with the suffixes -TO?, 
 -p6s, -6s, as well as monosyllabic and dissyllabic active parti- 
 ciples, under thesis I, as not being able to retract their accent on 
 account of their dissyllabic (or monosyllabic) character, while at 
 the same time reporting in a different part of the book that poly- 
 syllabic forms of the same categories likewise refuse to retract the 
 accent. I would only add to the details given above that there is 
 no indication in the language that the recessive forms in category 
 V are even numerically better represented than in category I. 
 
35 
 
 There is no boundary line of any sort between nominal monosyl- 
 lables, dissyllables and trisyllables any more than between finite 
 verbal monosyllables, dissyllables and trisyllables : pav, fifj, XiW, 
 olda, (pepu, (pepofjicv, (pepcopev. All these types are repeated in the 
 noun, but they have not become an infallible rule there : they have 
 absorbed some types, while sparing others in a measure. 
 
 The third thesis contains material which is absolutely otiose. 
 Forms like a>eKa'=Sk.dvada$a; a0^iroff=Sk. dksitas; eWepoi/nrSk. 
 dntaram ; vo-rcpos = Sk. iittaras ; Voc. Qvyarep = Sk. diihitar; o/cra>- 
 novs = Sk. astapdd, etc., prove nothing for any theory. Etymo- 
 logical accentuation, enclitic accentuation (if that be true), or the 
 assumed secondary accentuation, all meet on neutral ground. I \ 
 would only add here a statement parallel to that made above (p. 33 
 fg.) in connection with words like K\COS, av6os, Kpeas, pe'Qv, etc. : I see 
 no reason for regarding the accentuation of the type represented 
 by dadtKo, evrepov, etc., as anything else than recessive. The coin- 
 cidence of the accent with that of I. E. duodekm, enterom, etc., is 
 indeed merely a coincidence, as is shown by eWpou, eWpow, etc. 
 
 We are thus left with those longer word-forms, in which the ety- 
 mological accent lay so near the beginning of the word as to leave 
 more than three moras (or four in trochaic polysyllables) unac- 
 cented : *rj8ia>v = Sk. svadiyan ; *<pepofjLvos = Sk. bhdramdnas. As 
 willing as I should be on a priori grounds to admit the effects of 
 a secondary accent in the apparent shift forward of the summit- 
 tone in rjSiuv and cpcpopevos, I find myself constrained to judge that 
 there is no more real evidence of its quondam existence in such 
 cases than in any of the preceding, barring of course combinations 
 consisting of an orthotone word plus an enclitic, where the second 
 accent has always been characterized as ' secondary.' In the first 
 place it is to be considered that the terminus a quo from which the 
 count of moras starts in this category is evidently the same as in 
 the preceding categories : the final mora of the word. It seems 
 therefore very unlikely that an essentially different principle was 
 in operation here than in the preceding cases. It would be totally V 
 against the spirit of Wheeler's theory to accept this secondary accent 
 as anything else than a phonetic law, pervading the entire material 
 of the language ; cf. his remarks on p. 8 and 9, bottom. But we found 
 the assumption of a secondary accent in the types vavs (Aeolic 
 Zev?) and eiSo? an impossible one: therefore it is also unlikely 
 here. Secondly, the mode of manifestation of the secondary accent 
 in the polysyllabic forms would be an extraordinary one. It strikes 
 
 
me that a sober definition of a secondary accent means that the 
 utterance starting with the summit tone has a tendency to repetition 
 of the stress or pitch at regular intervals, or bars from the primary 
 accent, and it is difficult to see how the end of the word, already 
 accented, can be introduced as the guide for the deposit of the 
 secondary accent. If the I. E. word bheromenos =. Sk. bhdramd- 
 nas developed a secondary accent upon the second syllable in 
 Greek, so as to produce fapopevos, we ought to have the secondary 
 accentuation in every case upon the syllable following the primary 
 accent. Instead of this it appears two syllables from the primary 
 accent in aveirlBeros = Sk. dnapihitas = I. E. nnepidhdtos ; three 
 syllables in aveniOtrov = Sk. dnapihitasya = I. E. hnepidhdtosio, etc. 
 To put the case still more strongly by bringing an example of an 
 enclitic finite verb, which inclines upon the preceding orthotone 
 word : Is there any likelihood that the phonetic need which did 
 not allow even one unaccented Qjrave) syllable to intervene 
 between primary and secondary accentuation in ^epd/za/os- = I. E. 
 bheromenos, later fapopwos, would be content to allow four grave 
 syllables to intervene in I. E. meghisthom bhoreiomedha =: ptyiarov 
 (popcofjLfda ? One asks further : If this accentuation has nothing to 
 do with the history of the word as such (the reverse is assumed if 
 we adopt the theory of enclisis), what right has one to disregard 
 the grave syllables in the word following the one for which the 
 accent is being determined ? To illustrate, why are the first two 
 grave syllables of AaftpoV in such a sentence as fieyio-rov cpopeontda 
 0r)o-avp6v left out of account in fixing the accent of (popf6p.0a if this 
 accent is not determined by the character of (popeopeda as a word, 
 but by the number of unaccented syllables preceding and following ? 
 Wheeler's own words on p. 7 can be brought up against him : ' In 
 der Phonetik des Satzes sind aber die vier letzten Silben in Sk. 
 tttiksamahdi ebenso sehr enclitisch wie die Verbform in vt'gvd ekasya 
 vimidas titiksate, RV. II 13. 3; die zwei letzten Silben in fapo- 
 peQa sowol wie die zwei letzten in dem Lautcomplex &v6p&- 
 TTovriva.' If this is merely a question of sentence-phonetics, in 
 which the individuality of the word, which is maintained particu- 
 larly by its word accent, is given up, we must consider any group 
 of unaccented syllables in the same light, and the development of 
 the secondary accent on the unaccented syllables of ^(yurrov 
 (j>opcofjic0a Qr](rav)p6v would yield one secondary tone on the syl- 
 lable Ba of (popeopeQa, a second one on the syllable e of (o/>eo/ie0a, 
 and a third one on the syllable rov of peyio-rov in addition to the 
 
 w* 
 
37 
 
 accents actually written. Cf. also the statement on p. 119 : ' Die 
 Eintheilung des Satzes in Worter ist immer mehr oder weniger 
 kiinstlich. Dieselbe miisste sich in jedem verschiedenen Satz 
 je nach dem Character des betreffenden Wortes und seinen Ge- 
 brauch in dem betreffenden Satze verschieden gestalten.' 
 
 Further, he who puts the grave syllables in a single word 
 upon the same level as the enclitic syllables in a combination 
 of an orthotone word plus an enclitic, *av<-7ri0fTos like avBpwirovnva, 
 ought to point out some reason why both of the accents in the 
 latter type are retained (av6pam6vri.va), while one is given up in the 
 former (avtirlOtroi). In the nexus of the sentence there is no more 
 reason for one than the other. And if one were to assume that 
 the first accent of dv0pa>7r6vTiva is due to an analogical restoration 
 after the single word avdpanov, in other words that the falling aside \ 
 of one accent in the early types *$epd/zevoff, *avm6eTos, *dve7ri0fTov, ^. 
 etc., was due to a law according to which a single word could /, 
 bear but one summit tone, he would still have to point out the * 
 reason why the first and original accent always succumbed in the 
 struggle for existence ? I do not believe that this could be ac- 
 counted for without calling in the aid of some external analogy. 
 And that would necessarily be the analogy of the finite forms of 
 the verb. One zs absolutely driven to recognize the possibility that 
 the analogy of accentual types is capable of being extended with- 
 out reference to the function of the words involved. I shall present 
 this view more systematically below, and would submit and empha- 
 size that accentual investigations which exclude this point of view 
 will ever tend to violent and complicated assumptions, such as shall 
 carry their own refutation with them. 
 
 I believe that nothing has as yet appeared which is calculated 
 to weaken my theory that the recessive accentuation in Greek is a 
 modification" of a special Greek law of enclisis. which has spread 
 from the finite verb until it has absorbed many quantitative word- 
 types in general in the Pan-Hellenic speech and all in the Aeolic. 
 After our renewed survey of the ground, and after having demon- 
 strated the untenableness of Wheeler's theory, the feeling of 
 security in entertaining the theory of enclisis must be enhanced 
 materially. There is as far as can be seen no other nov or for 
 the recessive accent, and I shall endeavor to show below that such 
 objections as have been advanced are either not well taken, or 
 are to be set aside by modifications which do not affect the main 
 current of the theory. I will for the sake of clearness state the 
 theory point for point : 
 
38 
 
 1. The Greek language exhibits distinctly in its treatment of 
 enclitics an aversion against a limitless enclisis. The normal 
 restriction of this enclisis is executed by repeating the summit tone 
 or by supplying with a secondary tone the difference does not 
 appear in writing after a certain number of syllables, or rather 
 moras, as in the following examples : QiXovrivos, avdpanos ns, av&panrov 
 nva, av0pa7ros (prjvi, etc. This secondary accentuation is therefore 
 the syntactical property of a combination of two words , the first of 
 which is accented in such a way that more than the permitted 
 number of unaccented moras would follow. A single word of any 
 number of syllables does not carry more than one accent unless 
 
 followed by such an enclitic. There is no reason to believe that 
 such a secondary accent develops independently from these syn- 
 tactical conditions upon any number of grave syllables however 
 great : witness e. g. the interval of five grave syllables in such a 
 
 Combination as peyio-Tov dij^aycoyov. ~ 
 
 2. The finite verb in principal clauses was enclitic in I. E. times 
 
 fand is so in Greek, when the number of syllables in all the forms 
 of a given paradigm does not exceed the legal number of moras. 
 Such cases are preserved in the inflection of /u and fopi. Elsewhere 
 /, A the enclisis of the finite verb is checked by the excess of moras in 
 the word to be inclined. Wherever some word or words in a given 
 ' /;W paradigm exceed the number of syllables which are allowed to be 
 / inclined, the entire paradigm is orthotone : those forms which have 
 three or more moras take the enclitic tone on the third or fourth 
 mora from the end; when a member of the paradigm does not 
 . n^. ^ ~ contain so many moras it places the accent as near to the theo- 
 i+ -it-retically correct place as possible, i. e. on the first mora of the 
 
 ~~ j C word. Thus, j^8o6r)(T6p.da t ^-(pfpofifda, -^-opvvtri) --oi6V, -^XtTrcs, -^-/SJ/, 
 
 / <h,d {'**+' ^<pepero>; ^<pe'pa>, -c-arav make up a representative group of enclitic 
 ./, ^ o verbal types, some of which are identical with accentual types 
 
 arising out of a combination of an orthotone word plus a full enclitic, 
 
 
 lk ft 
 \> 
 
 6. g. -t-cpepofjifda '. (ivQpamoi/Ti.va ] ^-opvvai : ScoKpern/y ns ,* -e-fapera) : Trarfjp 
 
 A IOU > av6pa>K6s TTOU, etc., while others seem not to have any parallel 
 L) among the ordinary combinations of orthotone word plus a full 
 ' enclitic. Cf. for this Chandler, 935 fg., Wheeler, pp. 119 fg. 1 
 
 1 I would not, however, as Wheeler has done, go so far as to deny the 
 originality of all combinations which do not coincide with the verbal law : to 
 assume, e. g., that avdpuTroc; Ttg and avdpuirtiv TIVUV are combinations in which 
 the position of the secondary accent is regulated by the analogy of avdpuTrdv 
 riva seems to me very unlikely. Indeed, I consider that the assemblage of 
 
39 
 
 3- The question as to the manner in which the enclitic accentu- 
 ation passed from the verb to the noun has been surrounded with 
 unwarrantable difficulties. I do not hesitate to retract my own 
 surmise, that the I. E. enclisis of vocatives and their qualifying 
 words in the middle of a sentence formed the bridge for the transfer 
 of the enclitic accentuation from verb to noun. 1 We have no enclisis 
 of vocatives reported in Greek at all (no cases of vocative enclisis 
 as in dpi and $7/u), and it is perhaps not unlikely that the treatment 
 of the vocative in the middle of a clause became identical with 
 its treatment at the beginning (with I. E. accent on the first mora) 
 before the transition of the enclitic recessive accent from verb to 
 noun took place. So that the vocatives probably were all orthotone 
 before the extension of the verbal enclisis and its substitute, reces- 
 sion, into the noun began. I believe that the transition from verb 
 to noun took place by a kind of analogy, which must be supposed 
 to be largely in operation in 11 the movements of accent. This 
 differs from the kind of analogy ordinarily discussed, in substituting 
 for the two terms form and function the two terms form and accent. 
 I can make my meaning clear very easily. Supposing we have a 
 number of long monosyllables consisting of consonant + long vowel 
 -j- consonant pronounced with rising-falling inflection (A), i. e. with 
 circumflex accent. Let us designate this type by xay. Let us 
 suppose that by its side there exists a single instance of a long 
 monosyllable consisting of consonant + long vowel -j- consonant 
 with falling-rising inflection (\/), i. e. with cut tone (acute) on the 
 syllable. Let us designate this type by xay. Will any one be 
 found willing to doubt that this single case, no matter how great 
 its functional distance from the type xay, might be attracted by 
 the latter so as to conform to them in inflection of voice, as well as 
 in number, arrangement and quantity of its consonants and vowels ? 
 We may call this the term is not a new one analogy of sound 
 or phonic analogy (lautliche analogic). The principle involved 
 in the single example is~ofte without which the rapid permutations 
 of accentuation will never be explained. Wheeler labors strenuously 
 with the doctrine advocated in my essay, that change in accent 
 
 cases given on pp. 125-132 teaches rather, that the verbal treatment of enclisis 
 is but one of many other which are possible in the language, so that the law of 
 verbal enclisis is not even binding for all combinations of orthotone plus 
 enclitic, much less for every bit of unaccented territory as in the forms *rj6iuv 
 and *av7r/6 l rof, which Wheeler operates with. Evidently we do not as yet 
 understand all the minutiae of Greek sentence-accentuation. 
 
 1 Cf. Historical and Critical Remarks, 62 (42) ; Wheeler, p. 7, 49. 
 
 
4 o 
 
 can only be due to regular phonetic change or analogy. But the 
 possibility of this kind of analogy does not seem to suggest itself 
 to him. What straits and improbable assumptions he is led to by 
 operating 1 only with functional analogy we saw best above in his 
 explanation of the completed Aeolic recession. Instead of grant- 
 ing that phonic types equal in the number and arrangement of 
 their consonants, equal in the number arrangement and quantity of 
 their vowels, but differing in their accentuation, would tend to extend 
 the similarity by allowing the accent of the less numerous instances 
 (or for that matter even the more numerous instances) to follow that 
 of the prevailing ones, he prefers the assumption that all the thous- 
 andfold instances of recession in the Aeolic, over and above the 
 Pan- Hellenic, were due to assimilation within the paradigm. But 
 he cannot tell us what mysterious force always drove the simple 
 paradigmatic assimilation into the arms of that kind of analogy 
 which wound up with the accent either on the third (fourth) mora 
 from the end, or as near to it as the number of moras contained 
 in the word would allow. 
 
 Accordingly I fail to see any other possibility of explaining the 
 ciigcumflexed trochaic dissyllables (type ~ w, olfe, cldos, rovro^Bevpo), 
 a type which is Pan- Hellenic, without any exception worth remark- 
 ing; cf. above, p."24.7 "IF we assume that it belonged originally to 
 finite verbal representatives of the type, due to recession, we have the 
 only explanation with a genuine historical background which has 
 been advanced since the days of the Misteli-Hadley theory. It would 
 be interesting but unessential to see statistics as to the relative 
 frequency of the verbal and non-verbal forms. I do not venture 
 to assert which would turn out more numerous. In the same 
 manner all the various enclitic verbal types of more than three 
 syllables which exhibit the accent upon the antepenult when the 
 ultima is short (<popeV f "> <opeoyiei>, fopeopeOa) must have proved a 
 phonetic type of such prevalence and attractiveness that all other 
 accentuation before the antepenult was given up for it. All the 
 various verbal types of more than two syllables, which exhibit the 
 accent upon the penult when the ultima is long ((popeo>, ecpopeo/^, 
 etc.) in the same way attracted to themselves the non-verbal types 
 corresponding. We cannot escape the assumption of purely 
 phonic analogy in this question, and though this kind of transfer 
 seems to call forth our sympathy less readily, though the motive 
 at the bottom of it is less easily apprehended than in the kind of 
 analogy in which form is influenced by similarity of function, it is 
 
4' 
 
 undoubtedly at work in the development of accentual systems. 
 We may add of course that many nouns had the etymological accent 
 upon the same place as the corresponding phonic verbal types, 
 and this may have helped the process of transfer. All the words 
 assembled in Wheeler's third category (p. 56 fg.) are of this sort. 
 This transfer of the enclitic and recessive accentuation to the noun, 
 etc., has been so complete that only a few phonetic types have 
 resisted it. They are j- (/3us, Zcvi), ^ C (/3aros, Bpaarvi) ; ^ v6 
 
 ((piXrjTos, yfverr/p, lapos) ', ~ (KVVWV, Trora/iwv, 'icr^/uot, Tip.rjs I Hanssen 
 KZ. XXVII 614)5 ^ ^ [ofioios, yeXoIo?, epfj/Jios, erot/zos 1 I Bloom- 
 
 field, p. 41 [21] ; Wheeler, p. 113). A few old polysyllabic paroxy- 
 tones may also be mentioned : Uvai, el&vai : Ved. ddvdne ; OKTUTTOVS : 
 Ved. astapad-; yeyeV&u : Ved. gamddhydi, etc., although in the 
 last two cases the paroxytonesis is identical with the 'recessive' 
 accent. 
 
 But of these also the majority have made the resistance only in 
 part, and the question as to whether a given form gives up its 
 etymological accentuation is a matter which is usually determined 
 by the category to which it belongs. It is not a question of the 
 number of moras or syllables of which the word consists. It is 
 one of the gravest errors of Wheeler's presentation that he gave 
 to this fact a different coloring. Dissyllabic, trisyllabic, and partly 
 even quadrisyllable (compounded) verbal adjectives in -rds ; adjec- 
 tives in -p6s and -vs ; monosyllabic, dissyllabic, and trisyllabic active 
 participles (<ui/, AITTQ>I/, 7re0uo>s), etc., are oxytone throughout the 
 language. On the other hand, dissyllabic and trisyllabic nouns in 
 TIS (-o-isO, or the ordinals in /*o-, fo- (e^So/xoy, oydofos : Ved. saptamd-, 
 astamd-^), etc., are recessive. The cause of the transfer, while no 
 doubt many times based upon some attraction within the language 
 (cf. the explanation of the recessive accent of the abstracts in -TIS 
 above, p. 30), is in most cases simply a tribute to the more preva- 
 lent accentual types, as TreXoo;? = Ved. paragus ; TTO'XI? ~ pun's, 
 and many others. 
 
 MAURICE BLOOMFIELD. 
 
 BALTIMORE, March, 1888. 
 
RETURN CIRCULATION DEPARTMENT 
 TO""^ 202 Main Library 
 
 LOAN PERIOD 1 
 HOME USE 
 
 2 
 
 3 
 
 4 
 
 5 
 
 6 
 
 ALL BOOKS MAY BE RECALLED AFTER 7 DAYS 
 1 -month loans may be renewed by calling 642-3405 
 
 6-month loans may be recharged by bringing books to Circulation Desk 
 Renewals and recharges may be made 4 days prior to due date 
 
 DUE AS STAMPED BELOW 
 
 SfP 17OT 
 
 
 
 IN STACKS 
 
 
 
 AUG171977 
 
 ftFP Pfff WIT 1 T E 77 
 
 
 
 ntv uln,AUu if ti 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 = 
 
 
 . 
 
 
 
 
 FORM NO. DD 6, 
 
 UNIVERSITY OF CALIFORNIA, BERKELEY 
 BERKELEY, CA 94720 
 
YC 00370 
 
 * I I 
 
 J 79 
 
 
 
 UNIVERSITY OF CALIFORNIA LIBRARY 
 
^Vv\"\ ;vH-^\.^*w:; -;^.tv ^-jr'~ 
 
 
 X --->..