UC-NRLF 
 
 ' S, 
 
 NL 
 
 
 wen 
 
GIFT OF 
 
 Q / ~ftt> / 
 tA^f^&Lz /0^e^x.j 
 
HISTORICAL AND CRITICAL REMARKS 
 
 INTRODUCTORY TO A 
 
 COMPARATIVE STUDY OF GREEK ACCENT 
 
 MAURICE BLOOMFIELD, PH. D., 
 
 Associate Professor for Sanskrit and Comparative Philology, 
 JOHNS HOPKINS UNIVERSITY. 
 
 REPRINTED FROM AMERICAN JouRNjn/6 PHILOLOGY, "VOL. iV, WHOLE No. 13. 
 
 BALTIMORE, 1883 
 
 Press of Isaac Friedenwald, Baltimore. 
 
HISTORICAL AND CRITICAL REMARKS INTRO- 
 DUCTORY TO A COMPARATIVE STUDY OF 
 GREEK ACCENT. 
 
 I. 
 
 Accent is a universal phenomenon in language, and one which 
 is in close union with what is treated by grammar under the head 
 of sound or phonology. 
 
 The sounds of a word without accent are merely separate stones 
 which accent cements into a linguistic entity, either a word or a 
 sentence. W. v. Humboldt says : ' The unity of the word is pro- 
 duced by the accent. This, by itself, is of a more spiritual nature 
 than the sounds, and it is therefore called the soul of speech, not 
 only because it is really the element which carries intelligibility 
 into speech, but because it is, more than other factors in speech, the 
 immediate expression of feeling ' (cited by Gottling, Accent der 
 griechischen Sprache, p. 8). 
 
 The word accent in modern terminology is unfortunately com- 
 pelled to do duty for more than one linguistic fact. First, in the 
 case of the word, it signifies the relative stress and pitch character- 
 istics of its various syllables, with no restriction to that syllable 
 which has the strongest stress or the highest pitch. This is the 
 most scientific function of the word. A closer study of the life of 
 the word cannot be satisfied with a theoretical analysis of its sounds 
 and syllables and a superficial recognition as to which of the syl- 
 lables has the highest pitch or strongest stress, but it must be known 
 also in what way or to what extent this syllable is elevated above 
 those surrounding it. Furthermore, the relations of the remaining 
 syllables to one another will always show that the same character- 
 istics which distinguish the tone-syllable *ar' e^o^v attach themselves 
 in a lesser degree to some one or more of the remaining syllables ; 
 in short, I would define word-accent in this wider sense as the history 
 of stress and pitch in the immediate practical subdivisions of the 
 word, its syllables. This definition of accent has necessarily to be 
 kept apart from that other more familiar one by which, in the cur- 
 rent parlance of grammar, the pitch or stress of the most accented 
 syllable is designated. This, of course, is not all. For just as the 
 
.** J 
 *** 
 
 word has its history of pitch and stress, so has the sentence. The 
 members of the sentence stand in a relation to the sentence as a 
 whole which is not unlike that in which the syllables stand to the 
 word. Here, of course, the word ' accent ' has again to do double 
 duty : first, it indicates the relative characteristics of the words 
 which make up the sentence, and, secondly, the word is also em- 
 ployed to mark that favored member of the sentence which holds 
 the most prominent position, i. e. the one which corresponds to the 
 ' tone-syllable ' in the word. 
 
 In the sentence ' he did it, not she,' we may speak of accent in 
 its most pregnant sense and refer merely to the two summits 
 ' he ' and ' she/ or on the other hand we may call before our minds 
 a picture of the exact relation of each of the words in pitch and 
 stress, not giving our attention merely to the summits, but watching 
 the undulation of the tone-line in which the sentence moves all 
 along, from the beginning to the end. This is the study of accent 
 in its scientific sense. 
 
 That the accent of a sentence is as much under the influence of 
 an organic law of some kind as the accent of the word is seen as 
 soon as one attempts to disturb the natural cadence of a sentence 
 such as the one cited above. By transferring the summit pitch 
 and ictus to the second word of the sentence we destroy the organic 
 life of the sentence fully as much as though we change the summit 
 pitch and stress in a single word. ' He did it, not she ' is as much 
 not an English sentence as ' development ' is not an English word. 
 Frequently the change of relation in pitch and stress does not go so 
 far as to destroy the sentence, it simply makes another sentence 
 out of it, as for instance when the summit tone is shifted successively 
 from one word to another in the group of words ' give me that 
 book.' We obtain four different sentences corresponding to the 
 four different positions of the summit tone. 
 
 With this last case may be compared the way in which, e. g. in 
 Greek, the change of accent changes entirely the character of certain 
 words otherwise the same, and in fact enters as a considerably 
 fruitful factor into word-formation. For instance, rpoxos is an agent- 
 noun or participial formation meaning ' running,' ' a runner '; rpoxos 
 is an action-noun or abstract, ' a running,' ' a course '; (popos means 
 ' bearing '; (popos ' a bearing,' a c tribute ' ; both couplets are forma- 
 tions identical in every respect but their accent ; the accent makes 
 the same phonetic groups into two words as distinctly differentiated 
 in function as two primary noun-formations from the same root can 
 
be. And, lest it be suspected that it was merely the superfine 
 linguistic genius of the Greeks which brought in so delicate a factor 
 as a power in word-formation, it may be stated at once that this 
 difference is prehistoric, and Indo-European ; the couplet <f>op6s and . 
 (popos makes a perfect proportion with Sanskrit bhards ' bearing ' 
 and bhdras ' a bearing,' ' a burden.' In the same manner cf. in 
 Greek ^rpoKrovos ' killing his mother ' as epithet of Orestes, and . 
 wrpoKTovoi * slain by a mother ' as epithet of the children of Medea, 
 the accent alone is the factor which has produced two distinct 
 categories in noun-composition, also prehistoric and Indo-Euro- 
 pean, and up to date not understood by the familiar guides for the 
 study of Greek. 1 
 
 The chapter on sentence-accent is one of the most difficult and 
 obscure in the study of grammar, and has been brought within the 
 range of scientific discussion only very lately. Of course certain 
 obtrusive phenomena which belong under this head had been 
 noticed and discussed long ago ; as for instance the fact that certain 
 words lose their independent accent in the sentence, namely, the 
 
 means literally ' mother-slaying '; it is the kind of compound 
 which is called tatpurusa by the Hindu grammarians, that is, a simple compound 
 in which the first member stands to the second in the relation of a case depen- 
 dent upon it. MyTpo-KTovot, is a secondary adjective compound, what is called 
 in Hindu grammar a bahuvrlhi compound, one upon which the idea of possession 
 and the like is secondarily engrafted ; the meaning is strictly speaking ' pos- 
 sessing,' i. e. being affected by a mother-slaying. The stem KTOVO- in the two 
 compounds is not the same ; in the first instance it is the nomen agentis urovoq 
 1 slaying,' in the second it is the nomen actionis Krovog ' a slaying.' The differ- 
 ence of tone in the two compounds represents one of the most noteworthy 
 archaisms in Greek nominal accentuation. Simple dependent compounds like 
 firj-r po-KTovoc were originally accented on the second member of the entire 
 compound ; this law is so strongly alive in the Greek compounds of this class, 
 whose second member is a noun of agency in -6-, that the law for recessive 
 accentuation is observed only so far as it does not annul the older law according 
 to which the tone must be on the second member, therefore / a;/rpo-/crdvof is 
 against the recessive tendency. On the other hand, possessive compounds 
 were originally accented on the first member, and in accordance with that, such 
 compounds follow freely the laws of recessive accentuation, as /uqrpo-KTovot . 
 The same law reveals itself in such accentual difference as is contained in Sk. 
 yajfiakamds 'desire of sacrifice,' and yajnd-kdmas 'having desire of sacrifice'; 
 the former is a simple dependent, the latter a secondary possessive compound. 
 The Sanskrit regularly differentiates such compounds by varying accentuation, 
 while in Greek the archaic differentiation of accent is preserved only sporadi- 
 cally. See L. v. Schroeder in Kuhn's Zeitschrift, 101 fg., esp. pp. 106, no and 
 116 ; Whitney, Sanskrit Grammar, 1247, 1264 fg. and 1293 fg. 
 
enclitics and proclitics ;' certain words change their accent accord- 
 ing to their position in the sentence : the so-called anastrophe 2 of 
 
 1 That the proclitics do not lack an etymological accent (cf. below, p. 36), but 
 that they lose their accent from syntactical causes, i. e. from their relation to 
 other words in the sentence, can often be shown easily, either by pursuing their 
 history within the language itself, or by comparison with corresponding words 
 in other languages. For instance, ov proclitic appears at the end of a sentence 
 and in some other cases as ov ; wf and ef when they follow the governed word 
 appear as fog and If (Qebg wf, /ca/c.uv If ). That the proclisis of 6, rj is not due to 
 some etymological peculiarity of these words is shown by the Sanskrit corres- 
 pondents sd, sa" ol, at the special Greek new formations for older roi, rai (Sk. 
 masc. //= rot) are made analogically after 6, rj, and borrow from them their 
 proclisis. In the same manner no doubt all proclitics lose their accent owing 
 to syntactical relations, i. e. their lack of accent is due to Greek laws of sentence 
 accentuation. About enclisis we will have much more to say below. 
 
 [It is almost needless to add that the word ' proclitic ' is a modern invention 
 brought into currency by G. Hermann (Gottling, p. 387). That does not militate 
 against the existence of the thing ; only there seems to have been no recogni- 
 tion of it in antiquity, and the omission of the accent in the cursive MSS was 
 due to differentiation, to the desire of distinguishing not only between 6 and o, 
 rj and rj, ol and ol, al and al, but also between ov and ov, etf and elf, iv and ev, 
 f.ef and If, the spiritus asfiernot being heard at that time. See G. Uhlig, Zur 
 ' Wiederherstellung des altesten Compendiums der Grammatik, Festschrift zur 
 j^Begrussung der XXXVI Philologenversammlung, p. 80. B. L. G.] 
 
 2 The true explanation of anastrophe is as follows: Originally 'prepositions' 
 were oftener or as often ' postpositions,' i. e. the position of these small words 
 in the sentence was a free one. This is clear, especially from the Vedic San- 
 skrit, where some of the most common ones occur oftener after their nouns than 
 before them (e.g. a 'to' occurs in the Rig- Veda 186 times after its case and 
 only 13 times before it). The mere fact that in later periods of language (e. g. 
 Greek and classical Sanskrit) the tendency is to place them before their cases 
 in itself proves nothing against this natural assumption. The case of a mono- 
 syllabic preposition like If, which receives its natural accent after the word it 
 governs, but is proclitic when it precedes it, points to the probability that the 
 true accent of these Greek particles must be looked for in their postpositive 
 position. Indeed, just as If (orthotone), so do all bisyllabic prepositions appear 
 with their true accent when they follow their cases, and just like ef (proclitic) 
 do all bisyllabic prepositions exhibit a substitute for proclisis when they accent 
 their ultimate. The grammars which regard the oxytonesis as the original 
 accentuation, of course explain it as due to a desire on the part of the language 
 to point to the word governed by means of the accent, but such an explanation 
 needs hardly to be refuted. 
 
 The originality of the tone of bisyllabic prepositions in anastrophe is proved 
 in addition by the fact that this accent is demanded by the corresponding 
 Sanskrit words whenever the etymology is clear. So Sanskrit dpa is not to be 
 compared with Greek and but with OTTO ; Sk. dpi not with kiri but with em ; in 
 the same manner the archaic character of the accentuation in irept, frdpa and 
 
oxytone bisyllabic prepositions, which, as is now generally believed, 
 preserves the original accentuation of these prepositions. The 
 change of an acute to a grave on an oxytone before another word, 
 though a phenomenon totally unexplained, 1 contains no doubt a 
 
 VTTO is warranted by Vedic ^>dn,pdra and lipa; the etymology of pera and Kara 
 is obscure, but they probably, like those preceding, have preserved their original 
 form in paroxytonesis ; vTrep is not to be directly compared with Sk. updri, 
 which is reflected exactly in the oxytone vneip virep may have preserved an 
 originally different accentuation, or it may have followed secondarily the accent 
 of the other prepositions which suffer anastrophe, aided perhaps by the accent 
 of virepoe = Sk. tipara. On the other hand d/^t/which does not suffer anastrophe, 
 is borne out in its oxytonesis by Sk. abhi ; avri to be sure is oxytone after the 
 case which it governs, against the accenTofSanskrit dnti ; but it may have left 
 the company of the prepositions with anastrophe, because it differs from all of 
 them in having its first syllable long (by position). In fact it appears to be a 
 law, unnoticed even by Benfey, the author of this explanation of anastrophe, 
 that only prepositions of two short syllables are affected by it (virsip always 
 oxytone, but virep virep with anastrophe). The etymology of avd and Aid is 
 obscure, but there is again no reason to doubt that their oxytonesis is based 
 on good etymological grounds. The fact that these prepositions were originally 
 paroxytone is proved also by the fact that they are so accented in adverbial 
 function. Prepositions were originally adverbs, which have become attached 
 to certain cases secondarily and in relatively later periods of language. Many 
 common prepositions in Greek are still adverbs in Vedic Sanskrit: dpa,prd, 
 pdrd (dTro, 7rp6, Trdpa), while pdri (Trepi) does function for both ; conversely the 
 Vedic dti (eri) is both adverb and preposition, while in Greek it has remained 
 adverb only. 
 
 The assumption that such accentuation as euro, Trapd, etc., contains a substi- 
 tute for proclisis is easily vindicated. As a matter of fact only monosyllables 
 are toneless in proclisis ; the treatment of bisyllabic words in the same position 
 is in perfect accord with the treatment of enclitics when these contain a too 
 great number of morae. Just as enclisis is restricted to three morae and two 
 syllables (therefore /Idyof T7f, but %6yoi rtve^, cf. below, p. 22), so proclisis is 
 restricted to one syllable and two morae (therefore /c iravruv, but Kepi iravruv). 
 The author of this ingenious explanation of anastrophe is Benfey (' Die eigent- 
 liche Accentuation des Indicativ Praesentis von ? sein und tya sprechen 
 sowie einiger griechischen Praepositionen,' Gottinger Gelehrte Nachrichten, \ 
 Febr. 27, 1878, p. 165 fg., reprinted in Vedica und Linguistica, p. 90 fg.); he / 
 closes his article with the following remark : " . . . es ist nicht besonders / 
 rlihmlich fiir die griechische Philologie, dass, nachdem sie mehr als zwei Jahr- 
 tausende mit verhaltnissmassig geringer Unterbrechung geiibt ist, noch in ihren 
 jungsten Lexicis und Grammatiken, die Formen and, eni, Trapd, Trepi, V7r6, Kara, 
 [terd aufgestellt werden, welche in der Sprache weder je vorkommen noch vor- 
 kommen konnten.' 
 
 ] An elaborate discussion of this difficult question, which space forbids us to 
 reproduce even in a condensed form, is contained in the essay of Leonhard 
 Masing : Die Hauptformen des Serbisch-Chorwatischen Accents, nebst einleit- 
 
difficulty whose solution will depend upon further investigation in 
 sentence-accent. The difference between interrogative and indefi- 
 nite pronouns (interrogatives, orthotone ; indefinites, enclitics) is a 
 case where sentence-accent^ apparently, has given the language a 
 method for differentiating an originally single category into two ; 
 this also is not understood, but the archaic character of this phe- 
 nomenon is warranted by similar methods in other languages. 1 
 And it has been urged lately that two different word-forms which 
 perform the same function, may owe their difference in form to 
 different intonation in sentence nexus. 2 
 
 enden Bemerkungen zur Accentlehre des Griechischen und des Sanskrit, St. 
 Petersburg, 1876, p. 19 fg. 
 
 1 The relation of rig, orthotone and interrogative, to rig, enclitic and indefinite, 
 is evidently the same as that of the German interrogative ' wer ' to the indefinite 
 ' wer ' in such sentences as the following : ' Wfr ist gekbmmen ? ' and ' Es ist 
 wer gekommen.' We recognize at once that the enclisis of the indefinite is 
 due to its peculiarly subordinate position in the sentence and not to any etymo- 
 logical deficiency, it is therefore a feature of sentence-accent. Cf. the still less 
 clear method of the Sanskrit for differentiating interrogatives from indefinites. 
 By various particles (some enclitic and others orthotone : ca, cand, at, etc.) the 
 interrogative without losing its own tone becomes indefinite, thus kds ' who ? ' 
 kd$ ca ' any one ' ; cf. Lat. quis and quisqtie, identical in form and meaning. 
 Whitney, Sk. Gram. 507 ; Delbrtick, Die Grundlagen der griechischen Syntax, 
 pp. 138, 145. 
 
 2 The most striking instance of this kind is an attempt to account for the 
 different forms of the third person plural of the copula. It is true that the 
 various forms of it, Doric fart, Attic eiffi, Ionic Zaoi, cannot be carried back to 
 any one origin by any phonetic jugglery. Accordingly complicated processes 
 of analogy have been resorted to generally in order to harmonize these forms. 
 Gustav Meyer's view, e. g. is that a-avn is the Greek ' ground-form.' From 
 this form he derives kaai by assuming that the e was added secondarily from 
 the strong forms of the root (e. g. karC) to *aai for *avn , i. e. *a-avrt ; while 
 Doric fan, Attic elm, are also to be derived from *avri by assuming that the 
 initial vowel was assimilated to the e of the strong forms. Others employ other 
 processes of analogy in order to harmonize these forms with one another. But 
 Joh. Schmidt has taught for some years past that Doric-Attic fan elai is to be 
 referred to a form *a-fan (= Germ, s-ind, Zend. h-enti) % while earn is to be 
 referred to *a-avTt. in the manner exhibited above. The two forms *a-evrt and 
 *o-avri are explained as, originally, respectively the orthotone and the enclitic 
 forms of the word in accordance with the ideas of Wackernagel as laid down 
 in Kuhn's Zeitschrift^iXXIV, p. 457 fg., cf. below, p. 36 fg. Of these two forms 
 *a-ivTi, the orthotone form, crowded out *a-avri in Doric and Attic, while vice 
 versa *a-avri, the enclitic form, gained the supremacy among the lonians. 
 This explanation is laid down with a very slight modification in the doctor- 
 dissertation of his pupil, Felix Hartmann : ' De Aoristo Secundo,' p. 68, while 
 
From the first opening out of the accented Vedic texts, a very 
 important fact bearing upon sentence-accentuation had been noticed. 
 In Sanskrit the finite verb in principal clauses is enclitic, while in 
 subordinate clauses it is orthotone ; this fact lay fallow until Jacob 
 Wackernagel, in the 23d volume of Kuhn's Zeitschrift, p. 457 fg., 
 showed that the Greek verbal recessive accent is nothing more 
 than this enclisis of the finite verb extended to all kinds of sentences, 
 subordinate as well as principal, but at the same time modified by 
 that peculiar law of Greek according to which enclisis cannot 
 extend beyond three morae. Wackernagel's ingenious discovery 
 we will discuss in full further on ; the point which is to be recognized 
 here is the fact that the study of sentence-accentuation is destined 
 to a prominent place in the grammars of the future, and that the 
 present generation of scholars will, beyond a doubt, see this develop 
 into a science ; the delicacy of the subject will call for the keenest 
 penetration, but this will be rewarded by the importance of the 
 results ; results of comparative grammar alike valuable to the 
 phonetist, the morphologist, and above all perhaps the student of 
 syntax. 
 
 The study of accent in these two forms (sentence and word- 
 accent) has then gained a distinct place in grammar. It may be 
 mentioned also that the phonetist recognizes phenomena closely 
 parallel to these in the structure of the syllable. The syllable also 
 has a relative accentuation, i. e. its various parts exhibit different 
 degrees of pitch and stress, and like the word the syllable has 
 usually one summit, which is a sonorous element, most frequently 
 a vowel, as e. g. in hand ; often a lingual or nasal as in the second 
 syllable of an^rrrite, waging, handsmmmsf. That the summit 
 accent is variable in position, according to the character of the 
 syllable, can be readily observed in taking a set of pairs of syllables 
 which vary from one another in their final consonants, these being 
 in the one case surd and in the other sonant : seed and seat, pease 
 and piece, brogue and broke ; the syllable tone of seed, pease and 
 brogue is upon a part of the vowel nearer to the final consonant 
 than in seat, piece and broke. Further, there may, just as in word- 
 accent, be more than one summit-accent, especially in long syllables. 
 
 Schmidt himself has returned to the expedient of analogy in KZ. XXV, 591. 
 Hartmann also employs Wackernagel's ideas on sentence-accent in order to 
 explain the various forms of the second aorist, ibid. p. 66. And Wackernagel 
 himself (KZ. XXIV, p. 470) accounts for the loss of augment in preterits by 
 assuming different accentuation in subordinate and principal sentences. 
 
8 
 
 If the syllable ' yes ' is pronounced in a contemplative way, e. g. in 
 the sentence ' yes, that may be so,' it receives two summits with 
 a decided fall between them. In general it can be noticed that in 
 isolated syllables the relative accentuation of the various sounds 
 gains especially clear expression ; so e. g. in the various uses of 
 the word ' well ' in such connections as ' well, let's go then,' and 
 ' well, are you ready ? ' The first ' well ' has falling tone, the second 
 rising tone. 
 
 The subject of syllable-accentuation so far has not gained a very 
 important place in grammar, and still belongs to the phonetist 
 rather than to the grammarian. But taken in connection with 
 word and sentence-accentuation, syllable-accentuation serves to 
 show that accent has been and still is a constant factor at work 
 upon every infinitesimal subdivision of human speech. If we im- 
 agine the course of human speech represented by a line, this line 
 will be a constantly undulating one when we wish to mark the 
 varying pitch of the sounds ; if we wish at the same time to convey 
 a picture of the varying stress or ictus the line would constantly 
 and gradually vary in thickness. Add to this the fact that this 
 variation in pitch and stress is not the effect of one single kind of 
 accentuation, but of a threefold one, and it will be understood how 
 delicate a subject for investigation it becomes even in living speech. 
 In dead languages the difficulties are increased so as to make it 
 hopeless that all the bearings of accentuation will ever be understood. 
 The discussion must restrict itself almost entirely to accent in its 
 pregnant sense, i. e. what we have termed summit-accent; only 
 rarely will the stations for lower pitch or minor stress play a part 
 in the discussion. For all the tradition on the subject, preserved 
 either in accent marks or in the description of contemporaneous 
 grammarians, is restricted to that, and is very fragmentary, as well 
 as vague in its terminology. 
 
 The general phonetic bearings of this subject can at present be 
 studied most conveniently in Sievers's' Handbuch der Phonetik 
 (Manual of Phonetics), especially 32-6, pp. 177-95 (word and 
 sentence-accent) and 29 and 30 (on syllable tone). 
 
 II. 
 
 It seems to-day almost a truism to state that a discussion of 
 Greek accent must start from whatever knowledge there is on Indo- 
 European accent ; in other words, that the study of Greek accent 
 must be comparative. This is true precisely as much in this division 
 
of Greek phonetics as in any other, as for instance the study of Greek 
 consonants, where one would not now-a-days presume to say 
 much without bringing in the related languages. This, however, 
 does not exclude the fact that accent is, more than other factors in 
 speech, subject to those forces in language which produce change. 
 The Greek and Latin three-syllable accentuations present so fixed 
 and peculiar a physiognomy even in their earliest phases that one 
 would suspect that this restriction to the last three syllables of the 
 word is something that was inherent in these languages from their 
 origin, yet it has been proved for the Greek that this extremely 
 peculiar accentuation is a development out of a system of accentu- 
 ation to which such a restriction was originally totally unknown. 
 
 The German language to-day exhibits a seemingly fixed law of 
 accentuation, namely, that of the root-syllable. This seems a 
 reasonable accentuation, for of all parts of a word the root would 
 seem to be the most prominent and therefore entitled to superior 
 stress and pitch. Yet no fact in linguistic history is at present so 
 clear as this, that the original German accentuation was not re- 
 stricted to the root-syllable, but was a free movable accent, often 
 upon the root, but hardly less often upon some sufExal element. 
 This is proved by Verner's law, and the accentuation of the root- 
 syllable in the German of to-day cannot be due to anything else 
 than the analogy of those words which, under the old free tone-law, 
 exhibited the accent on the root; an analogy carried out with 
 almost flawless consistency. 
 
 This does not exhaust the variety of accentual methods to which 
 Indo-European languages have arrived by various processes, often 
 very obscure. The Lithuanian division of the Lithu-Slavic family 
 consists of Lithuanian proper, Lettish and old Prussian. The last 
 branch has died out without leaving any tradition as to its accentu- 
 ation ; the first, the Lithuanian, exhibits a free accentuation which 
 can be compared and identified with that of the Vedic Sanskrit, 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 abnegated all historic accentuation and accents 
 everywhere the first syllable. 
 
 We need not go so far as the Lithuanian and Lettish to find an 
 equally striking and equally difficult phenomenon. The Aeolic 
 dialect in Greece is differentiated from the other dialects in that it 
 has given up almost entirely the accentuation of the ultimate. 
 Excepting the oxytone prepositions of two syllables and a few 
 
IO 
 
 conjunctions like avrdp, arap, there can be no accentuation except 
 that of the penultimate and the antepenultimate (Gottling, p. 29). 
 This is one of the main elements in the fabled special resemblance 
 between the Aeolic and Latin, and has been the cause of much 
 nonsense, 1 and this resemblance with the Latin has also given birth 
 to the equally erroneous idea that the Aeolic accent is older than 
 that of the remaining Greek dialects. On the contrary, no one 
 fact in Greek accentuation is clearer than this, that the oxytone 
 words in Greek are generally archaic, that they have more than 
 all others resisted the recessive accent. 2 
 
 To this tendency on the part of accentual systems to change in 
 such a way as to lose its original complexion entirely, the fact is 
 due that the comparative treatment of accent was, until very 
 recently, a method which had not gained a firm hold upon the 
 
 1 AII these do not exhaust the varieties of seemingly fixed systems which have 
 been built up upon the debris of the old I. E. accentuation in the various 
 families. In the Slavic languages, the Russian has still preserved noteworthy 
 points of contact with the accented Vedic Sanskrit, but the Bohemian has 
 adopted the same system as the Lettish mentioned above, namely, the accentu- 
 ation of the first syllable, while the Polish has worked out for itself a still more 
 peculiar system. All its words, excepting those borrowed from adjoining dia- 
 lects, are paroxytone, and here we are again led to the only reasonable expla- 
 nation, namely, that the frequent paroxytone accent of I. E. times was here 
 extended into a law. 
 
 We can pick a case from the modern Romance dialects which will show the 
 same complete change of accentuation, and which will at the same time carry 
 the solution of the change with it. The words which are the representatives 
 of the old abstract suffix tat (Lat. nom. fas, fraternitas) are oxytone : French 
 fraternitf, l\.a\. fraternitd ; oxytone accent is a most non- Latin quality. A solu- 
 tion for this case which is altogether probable is that the modern oxytonesis 
 has preserved the accentuation of the oblique cases : fraternitutis , etc. The 
 English on the other hand holds to the accent of the nominative. In the same 
 way the French conscription has the accent of the oblique case, -onis. In a case 
 like French parlor over against Italian parldre the accentuation of the ultima 
 carries its own solution with it still more clearly. 
 
 2 Almost all the important categories of noun-formation which are oxytone 
 appear in their original accentuation, as can be seen even from superficial 
 comparison. Thus nouns of agency in -6$, 4>op6f = Sk. bhards; but the nouns 
 of action are paroxytone, <j>6poc = Sk. bhdras; adjectives in -vg, r)6v$ = Sk. 
 svadds, k-"kaxv<; = Sk. laghiis, UKVC = Sk. a$iis', adjectives in -p(5f, kpv6p6g = Sk. 
 rudhirds ; verbal adjectives in -rof, K^vrog = Sk. frutds = O. H. G. hlut = 
 Eng. bud (KZ. XXIII 123), ireirros = Sk. paktds; the word for father, Trar^p 
 = Sk./*Vtf = Goth. faa*ar(ibid. 117); the perfect active participle eid&s = Sk. 
 vidvans (cf. idvla = Sk. vidust). In declension trove ' ^o86g = Sk. pad : padds ; 
 = Sk. dyaus : divds. 
 
II 
 
 minds of investigators. Parallelisms and resemblances between 
 individual facts of Greek and Sanskrit tone-laws were noted very 
 soon ; even large collections of words and word-categories which 
 exhibited identical accentuation were made, yet this did not seem 
 to impress investigators with the fact that, unless these resemblances 
 were accidental and that theory was not advanced the two lan- 
 guages were committed to the same original accentuation in every 
 part, and that it must be shown why and how they present such 
 important differences in historical times. On the contrary, investi- 
 gators were content to call in, for Greek as well as Latin, the 
 recessive principle (which after all is not recessive, inasmuch as it 
 stops at the third syllable) as a something gotten no one knows 
 where, perhaps as Bopp has it ' because the greatest recession of 
 tone expresses the greatest dignity and energy.' 1 
 
 To-day any one who wishes for a hearing on the subject of the 
 accentuation of any Indo-European language must operate with 
 the following principles : 
 
 1. The accentuation of any I. E. language is a development out 
 of the common I. E. accentuation, precisely as much so as the 
 sounds and forms of that language, be they ever so changed, and 
 be their analysis ever so difficult or even impossible. 
 
 2. The principle which changes accent is precisely the same as 
 that which changes other language matter, regular phonetic change 
 based upon phonetic law. Just as an I. E. consonant is changed 
 in German according to Grimm's law, so it is possible" that, e. g. 
 originally oxytone word-categories may become paroxytone in 
 some one language, 2 only this must be shown to take place accord- 
 
 1 Vergleichendes Accentuationssystem des Sanskrit und Griechischen, p. 16. 
 
 2 Or we will recognize below (p. 22) as important another Greek phonetic 
 law of accent, namely this, that enclisis cannot extend beyond three morae and 
 two syllables. Enclisis in general is an Indo-European quality (e. g. Greek 
 re = Sk. ca = Lat. que, etc., are all of them enclitic), but the Greek restriction 
 as to morae and syllables is a Greek phonetic law in exactly the same sense as, 
 e. g. the loss of f or I. E. v. The Vedic Sanskrit knows no restriction of this 
 kind ; a word of any length may be enclitic, as e. g. the stem sama ' any one * 
 (Greek stem dpo- in d/j.6-6ev) is enclitic, not only in forms containing two 
 syllables, but in all its forms, e. g. ace. samatn, abl. samastndt, gen. samasya. 
 And several enclitic words may follow one another, so several vocatives, or 
 vocatives with cases depending upon them, as e. g. Rig-Veda, VII 64, 2 : d 
 rajana maha rtasya gopa . . . yatam: 'O ye kings, guardians of great right 
 come hither.' Here four successive words are enclitic, cf. Journal of the 
 American Oriental Society, Vol. XI, p. 59. 
 
12 
 
 ing to a law, and this law must like all other phonetic laws be based 
 pon the results of observations exercised upon extensive material. 
 
 3. Where no phonetic law can be adduced, the influence of 
 analogy must be the changing factor. So e. g. the modern German 
 with its prevailing accentuation of the root-syllable, the significant 
 syllable has been explained above ; the influence of analogy in the 
 Greek ' recessive ' accent will be discussed further on ; it is perhaps 
 the most striking and convincing case of the workings of analogy. 
 
 4. The influence of foreign languages and adopted words cannot 
 be left out of account. These usually carry their tone with them 
 from home. So e. g. large categories of words in German betoken 
 by their accentuation what is also known otherwise, namely, that 
 they are of foreign descent, e. g. nouns in -tat, -ton, etc., universitat, 
 institution, which exhibit foreign accent ; the entire class of verbs 
 in ieren, studieren, marschieren, in the same manner exhibit French 
 suffix and French accent ; according to Grimm words like reiterei, 
 malerei) etc., have suffixal accentuation, although they are in their 
 root good German words, because they were formed on the analogy 
 of melodey (/ueXca&a), abtei (abbatia), so that this is an example 
 where a distinct category of German words received both suffix 
 and accent from abroad. 1 
 
 The question which arises next is : What was the character of 
 this Indo-European accentuation from which the various peculiar 
 accentuations of the several languages have developed ? Of course 
 the question can be answered only for the smallest part ; almost all 
 that is known is restricted to the summit-accent, and even here 
 nothing is absolutely and completely clear. We will here consider 
 only the one fact which, above all others, has gained an unimpugned 
 position, namely, the freedom of position of the summit-tone of the 
 I. E. word ; other qualities both of word and sentence-accentuation, 
 which are probably Indo-European, will be discussed further on 
 in connection with the Greek itself. 
 
 The fact that the I. E. parent-language knew none of those 
 restrictions as to the position of the tone which we see in almost 
 all the languages that are still alive, and also in Greek and Latin, 
 especially the latter, is seen by a comparison of the accented Vedic 
 
 1 The influence of foreign languages upon accentuation is still more strikingly 
 exhibited in the threefold tone of the German word grammatik^ namely, grdm- 
 matik, grammdtik and grammatik. The last contains the French accent (gram- 
 matique], the one preceding the Latin (grammdtica}, while the first represents 
 the genuine German pronunciation with the tone on the root. 
 
13 
 
 Sanskrit with the Greek and German. This comparison yields the 
 result that the Vedic accent has preserved very closely the old 
 word-accent of the I. E. parent-speech. Of course this result was 
 obtained by the usual methods of comparison. Whatever in Greek 
 and German accent has, upon investigation, proved itself to be 
 archaic, is not only to be found freely in the Vedas, but is usually 
 seen there in the form of a principle of wider scope. So e. g. the 
 seemingly irregular accent of the participles and infinitives of the 
 thematic or second aorist in Greek is an archaism on Greek ground. 
 In the Veda this entire tense-system is accented on the same place, 
 the thematic vowel, except in the augment forms, where the augment 
 always takes the tone, cf. below, p. 38. In the same manner it will 
 be observed repeatedly that the Greek cases of oxytonesis are 
 usually of a somewhat disjunct and fragmentary character. Not 
 clear in themselves, they do not yield up any principle until we see 
 them in their full bearings in the accent of Vedic word-categories 
 which accent the ultima. And again in German, Verner's law has 
 shown that the more salient principles of Vedic accentuation, such 
 as the shifting of the accent from the root to the flexional element 
 in the non-thematic conjugations, belong to the oldest property of 
 I. E. speech, cf. below, p. 15, note ; it has also shown that appa- 
 rently irregular accentuations, such as the Vedic accent of the nouns 
 of relationship, pitdr but mdtar, must be carried back to the primi- 
 tive Indo-European language. * 
 
 No syllable, then, of an I. E. or Vedic word was, on account of its 
 position or on account of its quantity, unable to bear the summit- 
 tone ; no restriction, such as is seen in the three-syllable accents of 
 Latin and Greek, or in the root-accent of the German, is to be 
 found. Thus indra, indrena, dnapacyuta, dnabkimldtavarna, ag- 
 ninam, abhimati$dhd, parjdnyajinvita, etc. (Whitney, Sk. Gram. 
 95) present instances of Vedic accentuation. As far as the meaning 
 and value of this free accentuation is concerned, it must be confessed 
 that little or nothing is known. Indeed, it may be fairly said that, 
 in accordance with the more modest spirit in which linguistic 
 investigation is carried oh to-day, no very ardent search is made at 
 present for a cause which distributes the accents over these various 
 syllables. It is felt generally and justly that final explanations of 
 such delicate questions are not in order. The energy of accent- 
 investigators must be directed to an investigation of the simple 
 details of accentuation, and the causes of these variations in the 
 separate languages, before it can be hoped at all that the original 
 
H 
 
 cause of these phenomena will be understood. As long as e. g. 
 the restriction of Latin accent to penult and antepenult is a mystery, 
 so long there can be no hope of actually penetrating into the inner 
 life of the accentuation which preceded it. 
 
 Yet a noteworthy attempt to explain the I. E. accentuation dates 
 back to 1847. The first one and almost the last one who undertook 
 to describe, systematically, the accent in its historical development 
 in the I. E. languages, and at the same time to assign a cause for its 
 original character, was a French scholar, Louis Benloew, in a work 
 entitled ' De 1'accentuation dans les langues indo-europeennes tant 
 anciennes tant modernes.' According to Benloew the summit- 
 accent was originally an accent purely of pitch, a musical accent 
 without stress or ictus. In each word which consisted of more 
 than one syllable, some one syllable was pronounced musically 
 higher than all the others ; the syllable which was thus distinguished 
 from the others was, according to Benloew, the chronologically last 
 denning element in the word (le dernier determinant). That is, 
 according to the theory of word-construction which ruled in Ben- 
 loew's day without opposition, and which is accepted to-day also 
 to a very considerable extent, a word is made up of root, suffix, 
 personal inflexion, case-ending, augment, reduplication and so forth, 
 and whichever one of these various elements in the word had been 
 joined to the word last, that was entitled to this higher musical 
 pitch. So e. g. in an augment-tense the augment, in a noun in the 
 genitive the genitive ending ; when a word was compounded with 
 a preposition, the preposition. As long as this principle was still 
 in existence, the unity of the word in our sense had not as yet 
 developed; the marked emphasis of the 'dernier determinant' 
 directed the attention of both speaker and hearer so strongly to 
 some part of the whole, to some special element in what afterwards 
 became a unit, that it must be supposed that this accentuation was 
 in force in a period previous to that of word-formation in its strictest 
 sense. The cementing of the word as we have it now was pro- 
 duced by an additional force. By the side of the principle of the 
 last determinant there was developed slowly and gradually a logi- 
 cal principle of accentuation whose purpose it was to act without 
 reference, and in fact in opposition to the specializing tendency of 
 the ' last determinant.' This logical accent, it is assumed, affected 
 the root-syllable, which, in the word as a whole, is the ruling 
 syllable. The further history of accentuation in the separate I. E. 
 languages exhibits, then, a gradual process by which this logical 
 
15 
 
 accentuation gains the ascendancy in the word. This in turn is 
 gradually counteracted and affected by the influence of quantity, 
 which Benloew, with true instinct, regards as the last factor which 
 entered the arena. In Sanskrit, as far as is known, the accent is 
 totally independent of any considerations of quantity ; in Greek, 
 quantity, especially of the final syllable, begins to exercise an influ- 
 ence on accent ; still truer is this of the Latin, where quantity and 
 accent balance each other almost entirely. 
 
 The boldness and the esprit of Benloew's thoughts on this subject 
 are quite out of proportion with their sobriety, with the extent of the 
 material upon which they were based. In fact they are in all 
 important respects hardly more than ingenious assumptions. Yet 
 his theories deserve even to-day a certain degree of consideration, 
 for they gained such wide adherence that certain of his thoughts 
 are even now silently accepted. So, above all, the musical character 
 of the early I. E. summit-accent, which has never been proved, and 
 which, if separated from stress/is certainly to our ears an extremely 
 peculiar accentuation. Verner, in his explanation of the Old German 
 accent and its influence upon the mute consonants, starts with this 
 statement : ' The I. E. accent was, in its nature, chromatic (z. e. 
 musical), and, in its use, of unlimited freedom of position ' (KZ. 
 XXIII, p. 128). He then proceeds to explain his exceptions to 
 Grimm's law, by the assumption that the accent became an accent 
 of stress (expiratory) in primitive German, or possibly a combina- 
 tion of musical and stress accent. Benloew's other important idea, 
 namely, that of the ' last determinant,' has also been revived in our 
 day to explain a phenomenon of the widest extent and of great 
 importance, namely, the variation of stem and accentuation in the 
 non-thematic verbal conjugations. 1 
 
 1 In Greek this variation of stem is preserved intact only in a few cases, and 
 its immediate cause, the shift of accent from the stem to the root, is lost to 
 sight, owing to the leveling force of the recessive accent in verbal accentuation. 
 But the variation of stem-form as well as the accompanying shift of accent is 
 easily established as archaic by comparison with the Vedic Sanskrit, so in the 
 following cases: 
 
 ei-// el(*i-((j)i) 
 
 e-mi /-si 
 
 fold-a fola-6a 
 
 ved-a ve't-tha 
 
 The duals, though they agree in both languages in having weak root-form 
 (and accordingly are accented on the personal endings in Sanskrit), are left out 
 
 el-ai(*el-Ti) 
 
 : l-fiev 
 
 t-re 
 
 i-Cicri(*i-avTt) 
 
 d-ti 
 
 : i-mdsi 
 
 i-thds 
 
 i-dnti 
 
 folS-e 
 
 : fid-/Ltev 
 
 
 
 ve'd-a 
 
 : vid-md 
 
 
 
i6 
 
 Benloew's work represents the first and also the last attempt on 
 so pretentious a scale to inquire into the original character, develop- 
 ment and history of I. E. accentuation. The next somewhat com- 
 prehensive work we owe to the founder of comparative philology, 
 Fr. Bopp, in a book entitled ' Vergleichendes Accentuationssystem 
 des Griechischen und des Sanskrit,' Berlin, 1854. This work has 
 really a much narrower scope, it does not profess to deal with 
 general questions in any way, it merely attempts to give an exhaus- 
 tive list of those words in Greek which have still preserved the 
 accentuation of the Sanskrit and, therefore, in all probability the 
 I. E. accent. 
 
 Yet, incidentally, Bopp does express himself on general matters, 
 and in a way that cannot be called happy, either in its method of 
 treating the question or in the result reached. He recognizes as 
 the principle of Sanskrit as well as Greek accentuation ' the greatest 
 possible recession of the tone to the beginning of the word,' p. 16- 
 17. This mode of accentuation possesses the greatest dignity and 
 strength. The limitation of the summit-tone in Greek to the last 
 three syllables he looks upon as a degradation or enervation of 
 
 of consideration owing to the problematic character of the endings. In this 
 variation of stem and accent one fact seems clear beyond all doubt, namely, 
 that the weakening of the root is due to the shift of accent to the personal 
 ending ; but the question arises, what may be the cause of this varying position 
 of the accent ? There has been, as far as is known, but one answer to this 
 question, that of F. De Saussure in his Memoire sur le Systeme Primitif des 
 Voyelles dans les Langues Indo-Europeennes, p. 189, and that is distinctly in 
 the spirit of Benloew's theory of the * last determinant.' Saussure assumes 
 with Friedrich Mliller (cf. now also Fick in the ' Gottinger Gelehrte Anzeigen ' 
 for 1881, Vol. II, part 45, 46, p. 1462) that the so-called secondary personal 
 endings of the verb are more original than the primary, not that the secondary 
 are the result of weakening from the primary, as has been generally held from 
 Bopp's day down. The primary endings often differ from the secondary by an 
 additional *, and it is thought that this i is the same deiktic particle which 
 appears, e. g. in Greek TOVTOVI. Thus 
 
 I sg. 2 sg. 3 sg. iplur. 
 Primary : mi si ti nti 
 
 Secondary: m s t nt. 
 
 By assuming that the secondary endings first entered into verbal formation 
 and that these personal endings received the tone, whenever they could, a reasonable 
 ground is gained for the exceptional position of the three persons of the singular ; 
 here the endings are only m, s, t, which are not fitted for carrying the tone of 
 the word ; therefore the tone remains on the root and preserves in it a stronger 
 vocalization. 
 
17 
 
 language. The accentuation of final syllables or syllables near the 
 end is due to the ' sinking ' of the accent from a position nearer 
 the beginning of the word, etc. Nowhere, however, does he indicate 
 in any manner by what process of investigation he came to this 
 result, though these ideas permeate the entire book and are urged 
 upon the reader with an evident fondness on the part of the 
 author. They do not seem to be the result of investigation as to 
 the nature and quality of the accent of these languages ; they are 
 in fact not offered as such. They are given merely as the ex 
 cathedra opinion of the master who, if any one, has a right to 
 speak ex cathedra. 
 
 Since Bopp's book, no comprehensive treatise on I. E. accent 
 has appeared, nor is it likely that any such pretentious attempt will 
 be made until investigation in the separate languages has established 
 a better insight into the special accentuations ; there is reason to 
 hope that the now recognized importance of the study of sentence- 
 accent will shed much light both upon the original history of accent 
 in primitive times, as well as upon the ways in which the historical 
 accentuations of the several languages developed out of the single 
 Indo-European language. 
 
 What we have gained from this discussion of Indo-European 
 accent is, first, the knowledge that the word-accent was a free one, 
 restricted to no special syllable or syllables of the word, and 
 untrammelled by quantity ; secondly, that the I. E. language knew 
 certain well-defined laws of sentence-accentuation, the traces of 
 which may be fairly looked for in the separate descendants of it. 
 Thirdly, that the elements which may be supposed to have changed 
 this original accentuation can scarcely be different from those at 
 work elsewhere in the formal life of language, regular phonetic 
 change and analogy. As will be seen, what knowledge we have 
 of Greek accent calls for no other factor and no other principle, nor 
 is it likely that any new principle, as yet unknown, will ever exer- 
 cise any important function in the progress of this difficult study. 
 
 III. 
 
 We turn now to the Greek itself. The literature of the subject, 
 both ancient and modern, up to the year 1875 is carefully collected 
 in the first paragraph of the book of Franz Misteli : ' Uber griech- 
 ische Betonung: Sprachvergleichend-philologische Abhandlungen,' 
 
i8 
 
 Paderborn, 1875.' Among the ancients the subject is scarcely 
 touched upon in classical times. The first mention of it is in Plato's 
 Cratylus, p. 399, where the terms 6vs and papvs first turn up ; next 
 in order is Aristotle, Poetica, chap. 20, where, in addition to the 
 O^VTTJS and [BapvTrjs of Plato, a /ztVoi/ is mentioned, i. e. a middle-tone, 
 which has been by some exalted to a most important position in 
 the theory of Greek accent, as we shall see soon. Aristarchus in 
 Alexandria is the next authority in chronological order ; but above 
 all other works of the ancients, the source for information is Hero- 
 dian: Herodiani technici reliquiae, collegit, disposuit, emendavit, 
 explicavit, praefatus est Augustus Lentz ; especially the first volume 
 containing Lentz's famous preface and the book ircpl KadoXiKrjs Kpo- 
 o-wS/as-, to which Misteli gives the first place among his authorities. 
 In the study of modern writers on this subject one need not go 
 back behind Gottling, Carl Gottling: Allgemeine Lehre vom 
 Accent der griechischen Sprache, Jena, 1835 ; a book valuable for 
 its digest of the opinions of the Greek grammarians, containing 
 rich collections of material, but of course to-day almost worthless 
 as far as theory and explanation of phenomena are concerned. 
 Next in order are the books of Benloew and Bopp, which have 
 been discussed in the preceding chapter. It may be added that 
 Bopp's book, while almost worthless as far as its general theories 
 
 1 The literature which is given there is more than full enough up to 1875. 
 He omits one book which is practical and valuable for accent of nouns, namely, 
 Chandler, 'A practical introduction to Greek accentuation,' which has appeared 
 lately in a second revised edition, Oxford, 1881. Since Misteli there have 
 appeared in addition to the many and often extremely valuable incidental 
 remarks and minor investigations of comparative grammarians, a few important 
 monographs bearing upon the subject : 
 
 Leonhard Masing : Die Hauptformen des Serbisch-Chorwatischen Accents, 
 nebst einleitenden Bemerkungen zur Accentlehre des Griechischen und des 
 Sanskrit, St. Petersburg, 1876, valuable for Greek accent in its first half, pp. 
 1-49, containing especially an exhaustive criticism of all opinions on the 
 grave accent, 44 fg., p. 19 fg. 
 
 Jacob Wackernagel : Der griechische Verbal-accent (KZ. XXIII, p. 457 fg.), 
 of the greatest importance for the general theory of the so-called recessive 
 accent. 
 
 Theodor Benfey : Die eigentliche Accentuation des Indicativ Praesentis von 
 ef und <p(i, etc., cited in the note on p. 5. Important for its solution of ana- 
 strophe, and its valuable remarks upon enclisis and proclisis. 
 
 Leopold von Schroeder: Die Accent-gesetze der homerischen Nominal-corn - 
 posita dargestellt und mit denen des Veda verglichen, KZ. XXIV, 101-28; 
 the first systematic attempt to establish Indo-European laws for the accentuation 
 of compounds. 
 
19 
 
 are concerned, is valuable as a clear and comprehensive exhibition 
 of the facts which it treats, namely, the coincidences in the 
 accentuation of Greek and Sanskrit words. Next, the subject owes 
 some noteworthy and ingenious essays to Franz Misteli and James 
 Hadley ; Franz Misteli : Uber die Accentuation des Griechischen, 
 KZ. XVII, p. 81 fg., p. 161 fg.; XIX, p. 81 fg. ; XXI, p. 16 fg. 
 After the appearance of Hadley's article these essays were rewritten 
 in book-form : Uber griechische Betonung : Sprachvergleichend- 
 philologische Abhandlungen, Paderborn, 1875. Hadley's brilliant 
 paper was published no less than three times : On the nature and 
 theory of Greek accent, by James Hadley, from the transactions 
 of the American Philological Association, 1869-70; translated in 
 Curtius Studien, V 407-28, reprinted in Hadley's collected essays, 
 edited by Whitney. Hadley's as well as Misteli's theories, which 
 are closely implicated with one another, will be discussed below. 
 Finally, much important material is contained in the four monographs 
 cited in the foot-note on p. iS. 1 
 
 If we now attempt to give a short general statement with regard 
 to the position of the summit-tone as it appears in Greek, compar- 
 ing it with that of the free I. E. summit-accent which we have seen 
 established, we may best formulate the facts as follows, under two 
 heads : 
 
 1. This free I. E. accentuation has been allowed to continue in 
 Greek in all kinds of formations, excepting finite forms of the verb, 
 when the free accent did not go beyond the antepenultima, e. g. 
 K\VTOS : K\efos = Qrutds : grdvas, cf. the Germ, hint (Ags. hlud} ; 
 irovs : Trodos = pad : padds ; XITTCOI/ : eXwroj/ = ricdn : dricam, vidvan : 
 vidusi = fldvs (for older *i'8o>s-) : *8wa, etc. See Bopp, Vgl. Accen- 
 tuationssystem, pp. 178-84. 
 
 2. In all the finite forms of the verb and in all those formations, 
 verbal, nominal, or otherwise, in which the old accentuation stood 
 before the antepenult, a new principle of accentuation has established 
 itself to the exclusion of the old free accent. The chief trait in this 
 new law is that it does not allow the accent to remain on any 
 syllable beyond the antepenult, but restricts it to the last three 
 syllables of the w.ord. To this law there is scarcely an exception 
 in the entire tradition of Greek ; the grammarians have fixed the 
 accent of two Aeolic words which contain diaeresis on the syllable 
 
 1 Misteli, in his list of authorities, mentions also the most important treatises 
 on Latin and Sanskrit accentuation, which do not, however, concern us so 
 directly. 
 
2O 
 
 before the antepenult, utjSeia in Sappho and the Lesbic c 
 which are not of enough importance for a general discussion. 
 Gottling, p. 20, note 2, and especially Misteli, p. 19, discuss them 
 fully. There are, of course, some words in which the theoretical 
 analysis of forms would lead to seeming exceptions to this law of 
 three syllables, . g. p&cuva if we carry it back to its */aeXawa, or Ovya- 
 rpes if it is derived from *0uyarepey ; but this is prehistoric ; at the 
 time when the pronunciation was pe\atva, all reminiscence of an 
 earlier */zeXawa was gone. Within these three last syllables the 
 position of the tone evidently stands in relation to the finer measure 
 of mora, as appears clearly in the law that the accent cannot pass 
 beyond the penultimate when the ultimate is long, so that the 
 Greek accent is, to a considerable extent, restricted to the last three 
 morae, ^. g. in such types as r)8!.Kovv, SiSotei/, eXeyofju-v. To this there 
 is in fact only one seeming exception and one real one : 
 
 i. A seeming exception to this restriction to three morae is offered 
 by such cases as . g. the genitive K^TTOV, where the acute is appa- 
 rently four morae from the end of the word, but where in reality 
 the second mora of the long penultima has the tone, so that if we 
 analyze into morae and write */c6eVoo, it becomes clear that the 
 exception is only apparent. That the acute on a long vowel means 
 the accentuation of the last mora is not a mere assumption, as is 
 shown by such cases as farms contracted from co-raw?. 1 In such cases 
 a contraction has taken place, and if the tone had been on any 
 other than the last mora the result would have been a circumflex ; 
 the reason for the absence of the circumflex is to be found in the 
 fact that the last vowel contains two morae (*eVra-ooy), with the first 
 of which, the toneless mora, the a contracts ; it thus leaves the 
 accent untouched in the result, e 
 
 1 That eara-6f is the old type of this perfect participle can be seen from the 
 Sanskrit equivalent tasthi-van; here the Sk. i equals the Greek a, as in sthi-tds 
 = cra-rds. 
 
 2 The circumflex cannot display itself upon less than two morae (' ^), therefore 
 also this projected *rra-8df results in oxytone eorwf. A case where this law of 
 circumflex is clearly exhibited is the vocative of the word Zevj. Zei>c (for *Ae6f) 
 is an old oxytone = Sanskrit dyaiis. By an Indo-European law the accent in 
 the vocative recedes to the first syllable of the noun, that is, the tone is as near 
 the beginning of the word as possible. The result for this stem is the vocative 
 Zev (i. *. 'Lev) dyaus (i. *. dtaks). The recession has taken place, but as the 
 word contains but one long vowel, the tone has passed from the last mora to 
 the first, exhibiting at least for diphthongs the actual divisibility of long vowels 
 into morae. 
 
21 
 
 2. The second exception to the law of three morae is much less 
 easily disposed of. When the tone is on the antepenult and the 
 last syllable is therefore short, but the penult is long, then it stands 
 at least on the fourth mora from the end, as e. g. in ufao-ros ; and 
 when both the penult and antepenult are long, apparently on the 
 fifth mora from the end in a case like fJTrcipos. 1 In both of these 
 cases there is, of course, no a priori reason why the law of three 
 morae should not have been kept in force by making both words 
 properispomena. 2 The only explanation that the authorities have 
 been able to bring forward is the rather unsatisfactory one which 
 assumes that in such cases the long penultima received a more 
 hurried pronunciation and suffered a loss in quantity. So Gottling, 
 p. 27 : ' the penultimate loses a part of its quantitative value because 
 the strength of the tone of antepenult outweighs the following long 
 syllable,' and in the same tone other writers down to Ku'hner. 
 The difficulty in the way of such an assumption lies, of course, in 
 the metrical value of such toneless long penultimates ; they are 
 just as inviolably long as any other long syllables ; the of fjneipos 
 differs in no way metrically from the 77 of the same word, and 
 the explanation given has quite the appearance of having been 
 constructed ad hoc without any sufficient ground. It is not 
 uninteresting that there are quite a number of cases in the lan- 
 guage in which both accentuations occur in the same word, one 
 having the tone farther back from the end than the third mora, 
 and the other having it on the third mora. In every case the 
 one which follows the rule of three morae is the older one, e. g. 
 fpfjfj.os Epic and in Herodotus, but Attic usually eprjpos; o/zoto? 
 Homeric, Ionic, and Old Attic, later fyoioy; rpoTralov Ionic and 
 Old Attic, common TpoVaio^; in the same way of eVor/ioy and 
 eroifjios the first is the more archaic form. In opolos : fyotos the 
 historical precedence of o>oloy is easily proven etymologically ; 
 6fj.olos is a secondary derivative from the oxytone stem 6^6- = 
 Sk. samd- with the secondary suffix -to- = Sk. ya- (Vedic -id}. 
 By an accentual law, which perhaps dates back to the common 
 
 'Apparently only if we assume that the tone is on the last mora of fjTreLpog 
 (*ETretpo(;) as in K^TTOV (*K.ee7roo). 
 
 2 The extent to which such accentuation is favored in Greek may be best 
 seen in the rendering of such Latin names as Dentdtus, Mocttstus, Ahenobdrbus , 
 etc., by Afvrarof, Mofctrrof, Am$/3ap/3of, etc. Nothing, except the predilection 
 of the language, is in the way of such an accentuation as Aevrdrof, etc. Hadley 
 in Curtius's Studien, V, p. 413. 
 
22 
 
 Indo-European period, 1 such a combination as 6^6 + to yields 
 ofiolo-, i. e. opolo-, cf. the case of ZeD (i. e. Ze't>) discussed above 
 on p. 20, note. We might then see in such cases the trace of a 
 still more stringent law in favor of the three morae ; possibly the 
 principle whidi underlies the recessive accent started strictly from 
 that point. 
 
 Whatever this law of three morae is, it may be noticed right here 
 that it is also the Greek law for enclisis, i. e. a Greek word can 
 incline upon the preceding word only in such a way that the result 
 does not produce conditions which are in conflict with the law of 
 three morae as laid down above. So e. g. Zevs /MU offers the condi- 
 tions which are apparent in KTJTTOV ; KO\OS m, a the same conditions 
 as afao-Tos. When, however, it is desired to incline fow, falv, ^as, 
 or vfj,>v, vp.lv, iip-as, the result is Zevs fjfia>v, Zfvs fjfjuv or rjfjLiv (with a 
 shortening of the last vowel which may stand in connection with 
 the removal of the tone from the ultima), etc. That is to say, owing 
 to the fact that these words contain at least four morae they cannot 
 become entirely enclitic, but become so as much as possible. The 
 grammars 3 (e.g. Hadley, 232) do not understand this phenomenon, 
 when they describe fj^v, etc., merely as optional weaker forms, 
 and not as enclitic forms. 4 Aside from the testimony in favor of 
 
 1 The circumflex in such cases is probably Indo-European, for in Sanskrit 
 also the acute vowel on the a of samd- would be followed by the so-called 
 enclitic svarita on the next syllable (id), which seems to imply that the voice 
 instead of sinking from the acute to lowest pitch without mediation, passes 
 down gradually, and this amounts evidently to the same phonetic result as the 
 circumflex in 6//o2o-. See Whitney in the Proceedings of the American Philo- 
 logical Association, 1870, p. 9 ; Sk. Gramm. 85. 
 
 2 The grammars falsely set up the paradigm elfii, kari, kcpiv, etc. The wurds 
 are enclitics and receive this acute only when enclisis of the entire word is 
 made impossible because the result would leave too many morae unaccented. 
 The accent is therefore due to sentence-law and is not etymological. The 
 true accent of ten is preserved in orthotone eari, see below, p. 41. The reason 
 why these words as well as <j>q/u, etc., are enclitic will be discussed in full 
 below, p. 37. 
 
 3 Kiihner calls it ' eine ganz eigenthiimliche Art der Deklination,' I, p. 264. 
 
 4 The assumption of enclisis in the shorter forms (/not, /uov), but of orthotonesis 
 or a merely changed accent in the longer forms ($/m>, ijfJ-as), apparently receives 
 a certain kind of support from the Sanskrit, where the enclisis of the personal 
 pronouns of the first and second persons, being evidently of a piece with the 
 enclisis of the same persons in Greek, is also restricted to monosyllabic forms. 
 The pronouns of the third person, i. e. the various demonstrative stems which 
 perform that function, do, however, incline forms of more than one syllable 
 freely, e. g. asmai ' to him,' asya ' of him,' are used both orthotonically and 
 
23 
 
 enclisis that is afforded by the parallelism of, e. g. poi and /zov, 
 when compared with e>oi and e>oO, we have most interesting native 
 authority to the effect that in Greek pronouns, the recession of the 
 accent in accordance with the law of three morae was the substitute 
 of enclisis when the word inclined possessed itself at least four 
 morae. Wackernagel, in KZ. XXIII 458, cites from Apollon. 
 Synt. p. 130, a passage, also treated by Lehrs, Quaestiones epicae, 
 p. 123, which bears upon this question: rjpKea-Qr] 17 eyK\io-is dia rrjs 
 
 peradecreas TOV TQVOV, fJKOva" fjp.(dv . . rrjs racrecos p.TaTi6efJ.i>r]s Kara rrjv 
 apxovcrav ' rjSuvarfi yap eVl TO TrpoK^ip-cvov TrpocreXtfeu/. 1 his passage, 
 
 from excellent ancient authority, proves almost beyond a doubt 
 what seems in every other way also probable, namely, that wav, 
 vfwv, etc., are cases of enclisis, and that, therefore, enclisis and 
 recessive accent are ruled by the same law of three morae. The 
 same principle is, of course, patent in other well-known attempts 
 to observe the same law ; in fact if we take the cases which Hadley 
 gives in 107 : avdpunos ns, Trails Tives,ioyoi nvts, we 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 ; it is to 
 be noted that the position of the tone on the fourth mora from the 
 end is also exhibited here, when the penult has a long vowel and 
 the ultima is short, ov $770-1 like Teapoy, Xoyoi Ti(yti) like a<ooro?. It 
 will be seen below, p. 36, of what importance it has been thought, 
 that the laws which govern the scope of enclisis, and of recession 
 of the Greek accent, are identical. Wackernagel's theory about 
 the recessive accent, which has commended itself to the acceptance 
 of most modern grammarians, is in the main based upon this 
 coincidence. 
 
 IV. 
 
 If, in stating the most prominent views with regard to the peculiar 
 character of Greek accentuation, we were to begin with Gottling, 
 this would be done in deference to a book which must still be kept at 
 one's elbow in the study of this subject. In some respects it might 
 
 enclitically, cf. above, p. II, note 2. It may be further said that the Sanskrit 
 proves nothing against the enclitic character of such forms as ijfiuv by the side 
 of r}fj.uv, because it happens to possess different polysyllabic forms made from 
 different stems by the side of the monosyllabic ones. It is not surprising that 
 a language which can choose between asmdbhyam and nas for the dative plural 
 of the personal pronoun of the first person, should choose asmdbhyam when it 
 required an orthotone form, but nas when it desired enclisis. The Greek has 
 no such choice in the cases involved. 
 
24 
 
 still be necessary to warn against it, while in others it might be 
 mentioned profitably as a scientific curiosum of efforts in this 
 direction, not as yet fifty years old. Guttling might also perhaps 
 deserve a mention because he represents the last attempt to account 
 for Greek accentuation, entirely out of itself, though even he 
 occasionally takes a glance at the incipient work of comparative 
 philology he often refers to Humboldt and Bopp or brings 
 on some real or seeming parallelism from some other language. 
 Occasionally again he sees farther than some of his successors, as 
 when he recognizes the fact that the recessive accentuation began 
 in the non-Aeolic dialects with the finite verb. The neglect of 
 this fact is one of the weakest spots in the theory of Misteli-Hadley, 
 which will be discussed immediately. Yet the limited space of an 
 article forbids any systematic mention of Gottling's views, and as 
 the views of Benloew and Bopp are already disposed of, we can at 
 once turn to the Misteli-Hadley theory. Misteli's theory on the 
 peculiar form of Greek accentuation was based upon comparative 
 studies as well as ' philological ' investigations in the Greek gram- 
 marians. It was first laid down in Vols. XVII, XIX and XXI of 
 Kuhn's Zeitschrift, and afterwards embodied in the form of a book, 
 whose title was given above, p. 17. In the period between Misteli's 
 articles and Misteli's book there appeared Hadley's article in the 
 Proceedings of the American Philological Association (cited ibid?), 
 an article which aimed to rectify Misteli's theory, and which ex- 
 tended it by bringing in the Latin within the framework of the 
 theory. Therefore the name Misteli-Hadley theory. 
 
 The key to the explanation of the three-syllable or three-morae 
 accent according to this theory is the assumption of a middle-tone 
 (mittel-ton) which, already in the parent-language, followed imme- 
 diately upon every summit-tone, as a kind of intermediate step 
 which served to bring the voice gradually from the musical height 
 of the summit to the lowest depth (the toneless syllable). Nowhere 
 was the passage from the summit-tone to tonelessness in the same 
 word one which did not involve this middle-tone. If there were 
 syllables left in the word after the two which are bespoken for the 
 summit-tone and the middle-tone, these and their number is left 
 indefinite are toneless, or according to the preferable terminology 
 of the German receive the * tief-ton.' This theory of a middle-tone 
 is suggested in the first place by the Vedic Sanskrit. This possesses 
 a mode of accentuation which distinguishes three kinds of tone, I. 
 a higher (uddtta 'raised') or acute; 2. a lower (anudddtta 'not 
 
25 
 
 raised'), *' * toneless or ' tief-tonig ' ; 3. a third, which is called 
 svarita, according to Whitney 81 is always of secondary origin, 
 being the result of actual combination of an acute vowel and a 
 following toneless vowel into one syllable. This is uniformly defined 
 by the natives as compound in pitch, a union of higher and lower 
 tone within the limits of a single syllable. It is thus identical, as 
 far as can be seen, with the Greek and Latin circumflex, and in all 
 probability goes back with the circumflex to the common I. E. 
 period, as e. g. in the case voc. Ze5 : dydhs = nom. Zeus : dydtis> 
 discussed on p. 20, note 2. 
 
 So far everything is in reasonable accord with Greek notions of 
 accent. But there is a further element. ' The Hindu grammarians 
 agree in declaring the (naturally toneless) syllable following an 
 acute, whether in the same or in another word, to be svarita or 
 circumflex, unless indeed it be itself followed by an acute or cir- 
 cumflex, in which case it retains its grave tone. This is called by 
 European scholars the enclitic or dependent circumflex,' Whitney, 
 85. Misteli and Hadley then impugn the statement of the native 
 grammarians that this was a circumflex, and regard it as incompa- 
 rably more probable that this svarita is a middle-tone. And 
 Whitney, who is the first authority in matters of native Vedic 
 grammar, says (85) ' This seems to mean that the voice, which is 
 borne up at the higher pitch to the end of the acute syllable, does 
 not ordinarily drop to grave pitch by an instantaneous movement, 
 but descends by a more or less perceptible slide in the course of 
 the following syllable. No Hindu authority suggests the theory 
 of a middle or intermediate tone for the enclitic, any more than for 
 the independent circumflex. For the most part, the two are iden- 
 tified with one another in treatment and designation.' Whitney's 
 opinion with regard to the enclitic svarita, while it denies it the 
 name of middle-tone, does, we can see, nevertheless support a kind 
 of tone which does not lie very far removed in its nature from that 
 middle-tone in favor of which Misteli and Hadley argue. 
 
 But on the other hand the testimony for a middle-tone in Greek 
 which attaches itself immediately to the summit-tone in the manner 
 of the enclitic svarita is extremely weak, in fact may be said not 
 to exist at all. Not that there is not mention made by the ancients 
 of other accents than the three familiar ones. Aristotle, Poetica, 
 ch. 20, and Rhet. 3, i, 4 mentions a ^a-ov in addition to the o^rq? 
 and Papvrrjs of Plato, and this, according to Misteli, p. 44, note, and 
 Hadley, Cu. Stud. V 417, is probably a middle-tone, though both 
 
26 
 
 admit the possibility that the circumflex is indicated by it. The 
 Greek grammarian, Tyrannic from Amisus, who was captured by 
 Lucullus and brought to Rome, reports four accents according to 
 Varro (in Servius de accentibus, cf. A. Wilmans de M. Terenti 
 Varronis libris grammaticis, p. 187). Varro mentions other Greek 
 grammarians who report more than three accents ; there are in fact 
 those who report six accents altogether. Misteli seeks further (7, 
 p. 50) to fasten this middle-accent immediately after the summit- 
 tone, in a manner parallel with the enclitic svarita, by the aid of a 
 well-known passage of Dionysius of Halicarnassus de comp. verbo- 
 rum liber, section XI, but in this attempt he positively fails. The 
 
 passage reads SiaXe'/crou p,ev ovv p,e\os fvi /nerpemu Siacrr^/zari r<w Xe-yo/xeWi) 
 
 dia ire'vTc, i. e. the two limits of tone in spoken speech (between 
 summit-tone and low-tone) are said to be a fifth. Now Misteli 
 argues that this interval must have been mediated by the middle- 
 tone in passing from an accented syllable to an unaccented one, 
 because the unmediated skip of the voice through a fifth would 
 give to the language ' einen schneidenden und widerwartigen 
 character,' and because Greek ' speech would move in extremes ' in 
 such a case. But as Masing, loc. cit. p. 23, points out, another 
 passage in the same author makes this construction impossible. 
 
 I 1 1 For Dionysius continues, not many lines beyond this passage, with 
 the antithesis to the pe\os SiaXeVrov in the following manner : 17 Se 
 
 opyaviKrj re KOI coSix;) p-ovcra diaa'TTj/Jiaa'i re ^prjrai ir\cio<rw } ov dia irtvre fj.6vov, 
 aXX drrb rou dia TTCHTWV dp^cifjLevr], KOI TO 8ia nevTf /ueXoodei, Kal TO 8ia re<rcra- 
 
 pav, K. T. X. ' Music, however, instrumental as well as vocal, employs 
 several intervals ; not only fifths, but, to begin with octaves, next 
 fifths, fourths, etc.' It is evident from this passage that Dionysius 
 recognizes a plurality of intervals only for music and not for common 
 speech, and it appears that according to this author there is but 
 one interval, the fifth, in use in speech. 
 
 Moreover, this passage by no means certainly describes word- 
 accent ; so <?. g. Gottling, who by the way denies that Greek word- 
 accent was musical at all, construes this diaXe'm /icXos as a rhetori- 
 cal sentence-accent. Certainly it cannot be brought in as testimony 
 in favor of that special kind of middle-tone which follows every 
 summit-accent. Hadley does indeed recognize that the testimony 
 of the ancients for it, or for that matter any middle-tone, leaves 
 much to be desired ; but argues that the peculiar effectiveness of 
 it in the theory which he defends and extends is the surer testimony 
 in favor of its actual existence. 
 
27 
 
 The theory is then completed by the following assumption, which 
 is to account for both Greek and Latin accentuation : There was 
 developed in the Graeco-Italic division of the family, after they 
 had separated from the common stock, a disinclination to allow 
 more than one toneless syllable to follow upon the middle -tone ; 
 this disinclination caused a moving forward of the summit-accent 
 to such a position that there was room after it, and after the middle- 
 tone which necessarily followed it, for only one toneless syllable. 
 Thus originated the Graeco-Italic law by which the summit-accent 
 is restricted to one of the last three syllables of a word. The im- 
 mediate ancestors of the Greeks and Romans, the ' Graeco-Italians,' 
 before their separation from one another, accented their words 
 alike according to this simple law, e. g. ^e'XeiVo/i^i/, *avBpa>Trov, *gati- 
 dcres, *le'gendus, i. e. all words which originally had the summit- 
 tone before the antepenult simply shifted it to the antepenult, thus 
 producing a very special cadence agreeable to the Graeco-Italic 
 ear, summit- tone, middle -tone, tonelessness (low tone). In words 
 which did not have the tone anterior to the antepenult, words like 
 XeXu/ieW, x^ C7r <k> the accent remained undisturbed ; for here there 
 was no room for the violation of the law that the middle- tone should 
 not be followed by more than one toneless syllable. But as Greeks 
 and Italians divided off they developed their common three-syllable 
 tone-law in a manner which led to pretty sharp differences. The 
 point of departure from the Graeco-Italic law was the toneless 
 syllable in the cadence for the Greeks, the middle-tone for the 
 Italians. 
 
 Let us first remain a while with the Greeks. They developed 
 a dislike for a long toneless, i. e. final, syllable, so that the Graeco- 
 Italic cadence of summit-tone, middle-tone, toneless syllable, was 
 modified for the Greek into summit-tone, middle-.tone, and short 
 toneless syllable, whenever the accent had originally, in I. E. times, 
 stood before the antepenult. In order to exhibit the application 
 of this law, Hadley divides the phenomena of the Greek recessive 
 accent into four divisions, and one need but remember in addition 
 that he regards the circumflex as a compound accent containing 
 both summit and middle-tone, in order to understand his reasoning. 
 
 1. The simplest case. The acute cannot stand on any syllable 
 before the antepenult, therefore I. E. *?Xei7reTo becomes Greek 
 
 eXeiTrero. 
 
 2. The antepenult must, if it takes the accent, take the acute ; 
 ^eXetTrero (/. e. *eXtl7rero) is impossible, because it leaves two toneless 
 syllables at the end. 
 
28 
 
 3. When the penultimate carries the accent and the ultimate 
 contains a long vowel, then this must be the acute, TOIO.VTT), not TOIO.VTT) 
 (= *ToidvTTj), because this would result in a long toneless syllable. 
 
 4. A long vowel in the penultimate must take the circumflex if 
 the ultimate is short, roioOros, not *TOIOVTOS, because there would be 
 no room for the toneless syllable. 1 
 
 This method of accentuation in the separate life of the Greek 
 also did not gain ground when it was necessary to draw the sum- 
 mit-tone back from the end in order to gain the desired cadence. 
 Therefore types like AtAv/ieW, XMTCOI/, remained undisturbed. 2 
 
 Only one division of the Greek people, the Aeolians of Asia 
 Minor, took also this step completely, that is they subjected their 
 entire accentuation to the law of cadence, summit-tone, middle-tone, 
 low tone, therefore \e\vpfvos, ^aXe'Trw?. Where the entire cadence 
 was not to be procured, as in o-o<ds, they drew the accent back at 
 least as far as possible, ad<os>. 
 
 The theory then proceeds to explain the Latin accent by assum- 
 ing that the Graeco-Italic cadence-accentuation there also received 
 a modification, namely, that there developed with the Italians a 
 disinclination against a long middle-tone, so that the Latin cadence 
 became summit-tone, short middle-tone, low tone. We will return 
 to the Latin further on and see that this theory accounts for the 
 Latin system about as well as for the Greek. At present the Greek 
 will be dealt with alone. 
 
 i. In the first place it has been shown that the assumption of 
 this middle-tone following every summit-tone is a purely theoretic 
 one, and that the testimony of the grammarians in favor of such a 
 middle-tone amounts to nothing at all. Not that it is to be supposed 
 that the Greek word did not possess subsidiary tones just as much 
 as words of to-day ; but the assumption of a special middle-tone 
 which must follow the summit, implying that the pitch of the sum- 
 mit was especially high, so as to stand in need of a mediator between 
 it and the low tone, is warranted by no fact of Greek grammar or 
 
 1 This is the weak spot in the arrangement. The theory by which the expla- 
 nation of the Greek accent is here attempted does not in reality claim that the 
 cadence, summit-tone, middle-tone, low tone, must be established in every case ; 
 it makes only the negative claim that after summit-tone and middle-tone no 
 more than one low tone should follow. This condition would be satisfied as 
 well by *roLovrog as by Totovrof. 
 
 2 This rule knows exceptions from the earliest times. So e. g. nouns in -r/.f 
 (-CYf), pvaic, riatf, are originally oxytone formations, Sk. srutis, citfs, and yet 
 appear in all periods of the language with recessive accent, cf. below, p. 30. 
 
2 9 
 
 tradition. The passage of Dionysius not only proves nothing, but 
 if it speaks of word-accent at all, disproves the existence of any 
 interval in the SiaXexrou /ueAor, except the fifth. 
 
 2. The assumption of a Graeco-Italian accentuation (eAeiVo^i/, 
 legendus} stands entirely in the air. Not one historical fact is in 
 its favor ; it is solely based upon the fact of the restriction of the 
 accent to the last three syllables. At the time when Misteli and 
 Hadley wrote, the assumption of a Graeco-Italic period was very 
 generally, though even then not universally, accepted. It is to-day 
 a theory of the past. In just that particular factor of form which 
 stands in especially close relation to accent, namely, vocalism, these 
 families are about as far removed from one another as possible. 
 Further, it will be urged below that the Greek recessive, or, to speak 
 with Hadley, cadence-accent, began with the verb ; it is precisely 
 in the verb that Greek and Latin have diverged so extensively 
 that mere fragments of the older system of formations are left in ' 
 the latter, and it is altogether improbable that the Latin should j 
 have saved an old system of verbal accentuation for a new and j 
 obscure set of formations. 
 
 3. The assumption of the sequence, summit-tone, middle-tone, 
 and short toneless syllable, is after all nothing more than the for- 
 mulation into a more complicated shape of the simple law that the 
 recessive accent does not recede beyond three, or in one case (forms 
 like afrjoros and fjneipos') four morae. The theory does not find it 
 possible to free itself from the count by mora any more than the 
 formulation by which the accent was described above. While it 
 appears to dispose of the case of a^ooros- better (for here it was 
 necessary above to assume recession to four morae), it is deficient 
 in cases like TOIOVTOS, because it does not account for the constant 
 circumflex, cf. p. 28, note i, which on the other hand is accounted 
 for perfectly within the theory of the three morae. 
 
 4. Finally, the last objection is one which more than any other 
 undermines the middle-tone theory. The original I. E. succession 
 of summit-tone, middle-tone, low tone, it is claimed was in Graeco- 
 Italian times moved down a place or two or even more in order to 
 pander to a dislike on the part of the Graeco- Italians to allow more 
 than one toneless syllable after the middle-tone. An aesthetic 
 dislike which is powerful enough to reform the accent of an entire 
 language in a thoughtful, laborious manner, is a sufficiently doubtful 
 factor in modern linguistic explanation. It cannot exactly be called 
 a phonetic law, because a phonetic law acts spontaneously, and 
 
30 
 
 would not be likely to count the syllables of a certain word, and 
 then, upon rinding that the summit- tone upon a certain syllable 
 would leave too many toneless syllables at the end, move it down 
 a sufficient number of morae to ward off such an event. At least 
 so complicated a process must seem highly improbable when it is 
 compared with the workings of such a law in other quarters. Yet 
 the explanation as a phonetic law might, for lack of a better one, 
 be accepted with reserve, but for the fact that the theory fails to 
 account for a strictly grammatical, and not aesthetic, fact connected 
 with it ; namely this, that the recessive accent has most certainly 
 in Greek begun with the finite verb, where there is practically no 
 exception to it ; that it excludes, with particular care, non-finite 
 forms of the verb in the same tense-system and in evident connec- 
 tion with finite forms, exhibiting thus on Greek ground a most 
 outspoken character as a grammatical quality of finite verbs : eXwrov, 
 eXiTrofiT/j/, XiVa>, etc., but XITTCOZ/, Xwrea>, XwreV&u, etc. Of course noun- 
 formations are not spared in historical times. But here the tendency 
 is not regulated by any traceable law. Certain noun-categories 
 become recessive ; others, with apparently the same claim to favor, 
 do not; so adjectives in -vs versus nouns in -ns (W)- 1 It is in 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 ; it cannot, therefore, 
 have been due to the disinclination of the Graeco- Italians to allow 
 two toneless syllables after the middle-tone. Such a cause cannot 
 have differentiated between noun and verb. 
 
 V. 
 
 The strength of Misteli's system as completed by Hadley seems 
 at first sight to lie in the fact that it includes the Latin, which shares 
 with the Greek the sufficiently remarkable quality of restricting the 
 summit -tone to the last three syllables of a word. This coincidence 
 Hadley explains by the assumption of a Graeco- Italic accent which 
 knew no restriction except this, that the assumed I. E. cadence 
 of summit-tone, middle-tone and low tone, when it began before 
 
 ! Both are originally oxytone noun-formations; the adjectives in -vg have 
 remained so, 6pacv<; = Sk. dhrstis, fipadvg = Sk. mrdtis, irharvg Sk. prthzis, 
 &axv<; = Sk. raghris, iraxvc; =: Sk. bahiis, fiapvc; rz Sk. gurris, etc'.; the nouns in 
 -rif have without exception become recessive, as in the cases of pvaig and 
 cited above, p. 28, note 2. 
 
 
the antepenult, was moved down to avoid more than one low tone 
 at the end of the word. After the separation of the Greeks from 
 the Italians, the two peoples refined the common Graeco-Italic 
 accent; the Greek by insisting upon summit-tone, middle-tone, 
 short loiv tone, the Lat. by developing a fondness for summit-tone, 
 short middle-tone, and indifferent low-tone. Accordingly the 
 Graeco-Italic accentuation, which still permitted forms like legen- 
 dus, gaildcres, etc., was modified; and this modification again 
 becomes at least superficially easy if the definition and description 
 of the Latin circumflex, as given by the Latin grammarians, is 
 remembered, cf. Corssen, Ueber Aussprache, Vocalismus und Be- 
 tonung der lateinischen Sprache, II, p. 800 fg. According to them 
 the Lat. circumflex was employed upon long monosyllables (ex- 
 cepting ne with the imperative), and on penultimas with long vowels 
 (not, however, by position) when the ultimate was short. Every- 
 where else the acute was employed according to the remaining 
 well-known rules. How much value is to be attached to the state- 
 ment that in Latin gander e had the circumflex, made as that state- 
 ment is by grammarians who were under the influence of Greek 
 grammar down to the minutest particulars, is after all an open 
 question ; even Curtius, a strong supporter of the Graeco-Italic 
 accentuation, has said in my hearing that " der Circumflex im 
 lateinischen bedeutet iiberhaupt nicht viel, ist mehr auf Theorie 
 gegriindet." 1 
 
 But the assumption of the existence of the circumflex, and the 
 cadence projected for the Latin, summit-tone, short middle-tone, 
 and low-tone, seemingly procure a satisfactory arrangement of the 
 historical phenomena. 
 
 The simplest case is that of types like leg ere and legeret ; here 
 the cadence, summit-tone, short middle-tone, and low-tone, is easily 
 procured. In the type gaudere, the same result is procured by 
 dividing the circumflexed e between summit-tone and middle-tone, 
 quasi *gaudei.re. Greater is the difficulty in the type gaude~res> 
 for the first e is not circumflexed, therefore the syllable res must 
 furnish the place for both middle-tone and low-tone, *gauderees ; 
 but who will after all believe that there was so thoroughgoing a 
 
 1 Petrus Lange is the strongest assailant of the Latin circumflex, in three 
 treatises : De grammaticorum latinorum praeceptis quae ad accentum spectant, 
 Bonn, 1857 ; in a critique of Weil and Benloew's Theorie generale de 1'accen- 
 tuation latine, in Fleckeisen's Jahrbiicher, Vol. 79, 1859, p. 44-71 ; Untersuch- 
 ungen liber den lateinischen Accent, in Philologus 31, p. 98-121. 
 
32 
 
 difference in the accentuation of the two words gaudere and gau- 
 deres, or upon what tangible fact in the life of the language is this 
 differentiation based ? And in the type legendus we are left without 
 a place for the low-tone, because gen cannot take the circumflex, 
 *legend%s, while the type legendl again divides its final long syllable 
 between middle and low-tone, *legendii. Here the arrangement 
 is weakest ; it institutes a complicated difference between the accent 
 of gaud&re (gaudeere) and legendus (legendus ^.), which is devoid 
 of all foundation in the actual and not hypothetical life and history 
 of the language. 
 
 Of the four main objections which were urged above against this 
 theory when applied to the Greek, three hold good against Latin 
 also ; others can be added from the point of view of the Latin itself. 
 
 1. The still more complete absence of testimony in favor of a 
 middle-tone which regularly followed the summit-tone. There is 
 no such testimony at all to be obtained from the Latin. 
 
 2. The assumption of the Graeco- Italic accent, against which 
 what was said above, p. 29, is to be compared. 
 
 3. The combination with Greek recessive accent, which has origi- 
 nated with the verb, and will be shown below to be due to an I. E. 
 law pertaining to the verb, which therefore separates that method 
 of accentuation incontrovertibly from the Latin, where the special 
 influence of the verb is not to be thought of, and has not, as far 
 as is known, ever been suggested. 
 
 4. The very similarity of the Latin accent to the Greek becomes, 
 if we look more narrowly, reduced to the restriction of the tone to 
 the last three syllables. In every other respect the accentuations 
 of the two languages stand in the sharpest opposition to one another. 
 
 a. In Greek the summit-tone is not excluded from the last 
 syllable, in Latin it is so entirely. 
 
 b. In Greek the penult is absolutely without influence as far as 
 deciding the position of the summit-tone is concerned ; in Latin 
 the penult is the pivot around which everything revolves, its quan- 
 tity decides the position of the accent. 
 
 c. Just as indifferent as the penult is in Greek, so in Latin the 
 ultima has no influence upon the position of the accent, while in 
 Greek it is the main factor in determining the position of the reces- 
 sive accent. 
 
 5. A fifth reason against the assumption of the Graeco-Italic 
 accent is presented by the fact that there are distinct traces in Latin 
 of an accentuation which was not restricted to the last three syllables. 
 
33 
 
 The law of three syllables was preceded in an archaic period by a 
 freer accentuation, the vestiges of which are not sufficiently numer- 
 ous to make it possible to describe its exact character, though 
 enough can be seen to render it probable that it did not know this 
 restriction, at least not in the form of an inviolable law. 
 
 a. Very strong indications of a different regime in matters of 
 accentuation are contained in the vowel changes which attend 
 reduplication and composition. The reduplication and prepositional 
 prefixes in Latin exercise an influence upon the vocalism of Latin 
 roots which would remain unexplained, unless it be assumed that 
 they once regularly received the accent. Thus, when juro becomes 
 in composition p-jero,facio becomes *con-jicio, gnotus (with very 
 old vocalism = Greek yvuros = Sk. jndtds] becomes co-gnitus ; it 
 is necessary to assume that the accent stood originally upon the 
 preposition at a time when the root-vowel was not as yet weakened 
 (*pe-jiiro, *cd-gndtus], and therefore accented in a manner 
 thoroughly different from the laws of accent in historical times ; 
 for it would be incredible that this weakening of the root-vowel 
 should take place under the summit-tone (^pe-juro, etc). This 
 accentuation of the preposition with the finite forms of the verb 
 inclining upon them is Indo-European, and at any rate an accen- 
 tual condition which must be admitted for the Latin at some remote 
 period. On the same principle con-ficio must have originated from 
 a prehistoric *cdn-facio, with the accent on a syllable anterior to 
 the antepenult. And, further, in the perfects, tetigi, pepigi, cecini, 
 fefelli, cecidi (: cado), cecldi (: caedo), the weakening of the root- 
 vowels is due to the accentuation of the reduplicating syllable ; 
 this leads to forms like *tetigimus, etc., which again have the tone 
 before the antepenult. Moreover, certain Italian forms not Latin 
 support this view. E. g. the Oscan forms fe-f de-id (perfect opta- 
 tive third singular), or fe-fdc-ust (future perfect third singular), 
 when compared with Latin con-fic-io y or with an ideal reduplicated 
 *fe-jic-i from *f<!-fdc-i, show that this regular weakening of the 
 root-syllables is a special Latin phenomenon; so also Umbrian 
 Jupater is probably the common Italian predecessor of Latin Jupi- 
 ter. If this weakening of the vowels, as would appear from such 
 examples, is not common to all Italian dialects, but belongs espe- 
 cially to the Latin branch, and if it is assumed correctly that these 
 weakenings would be impossible under an accent like fefaciist, 
 fefacid, we have an historical corroboration in actually occurring 
 Italian words of the assumption that the three-syllable accent is a 
 
34 
 
 special Latin feature, not common even to all the Italian dialects ; 
 this disproves a Graeco-Italic accentuation directly on Italian 
 ground, aside from the general considerations which have been 
 brought on above. It must be assumed, then, that f^-f de-id was 
 the old accentuation ; this accent weakened the root-syllable fdc 
 in Latin alone, and after that pattern the same process is assumed 
 necessarily for forms of more than three syllables, cdnficio, displiceo, 
 displicemus, displicebamus, etc. 1 
 
 b. Other isolated forms, not within verbal paradigms, lend also 
 a certain support to the assumption of a freer position of the Latin 
 accent previous to the purely quantitative one of historical times. 
 Thus, when early opitumo- becomes later optumo-, it seems very 
 improbable that the i should have been lost under the accent ; an 
 original accentuation, *dpitumo- seems much more probable. 
 Samnium is for *Sabinium, (cf. Oscan Safinim) ; the accent of 
 this *Sabinium seems again to have stood on the first syllable, this 
 alone accounts satisfactorily for the loss of the i. Again, when the 
 Greek no\v8evKrjs passes through Polouces, Polluces, into Pollux, it 
 seems also improbable that these weakenings at the end should 
 have taken place under the accent Poludeuces, though in a proper 
 name it is not certain but what popular etymology may have 
 contributed to the corruption. Other forms favorable to this older 
 accentuation are cited by Corssen, Aussprache, Vocalismus, Beto- 
 nung, etc., IP, 902 fg.; Kritische Beitraege, 577 fg. 
 
 c. Less important is the statement of the grammarian Nigidius 
 Figulus, as reported by Gellius, that proper names in io in the 
 vocative accent the first syllable, e. g. Vdleri, fr. Valerius. Gellius 
 
 ! A great part of this argument against the assumed Graeco-Italian three- 
 syllable law was made as early as Kuhn's Zeitschrift, Vol. IX, p. 77, by Lottner. 
 Curtius, in an article in the same volume, p. 321 fg., attempts to refute Lottner's 
 assumption that forms like confide prove an accent before the antepenult (*cdn- 
 fido), by assuming that this weakening process started in forms which contained 
 but three syllables, e. g. cdn-fadt, cdnfidt, and was thence generalized for the 
 entire paradigm ; accordingly the paradigm of this tense was originally *confd- 
 cio, cdnfids, ctfnfidt, *confdtimus, *confddtis, *confddunt, and only later the 
 long forms assumed secondarily the vowel of the short ones, confidmus, etc. 
 This explanation is insufficient, first, because there is no trace of any such 
 forms as *confacio to be found anywhere in Latin ; secondly, because it is 
 impossible to assign any reason why the numerically stronger forms like *con- 
 fddo should have always succumbed to those with the weaker vowels, and why 
 the reverse process did not occasionally take place. There is too much con- 
 sistency and regularity in the use of the weak vowels to render such an expla- 
 nation at all probable. 
 
35 
 
 himself, it appears, remarks that such a pronunciation would have 
 been laughed at in his day, and Nigidius Figulus was also in other 
 respects a peculiar scholar, who set up a special terminology and 
 indulged in other idiosyncrasies for accentual words. The possi- 
 bility that he merely theorizes on Greek vocative patterns is not 
 to be denied. But Benfey, Der indo-germanische Vocativ, p. 51, 
 enthusiastically defends his view, saying that even if in Gellius's 
 time such a pronunciation was laughable, it may nevertheless be 
 a learned possession of Nigidius, a quasi ' lectio doctissima,' and 
 employs it to establish by its aid the Indo-European accentual law, 
 otherwise also secure enough, namely, that vocatives were accented 
 on the first syllable, regardless of the accent of their themes under 
 other circumstances. 
 
 Against all this stands then the single fact of the restriction within 
 three syllables, a fact striking enough on the outside, but yet not 
 very significant if we remember certain other facts. The possibility 
 that two I. E. languages, starting from the common stock and from 
 a common point of departure, should leave that point behind them, 
 work out their accentual destiny separately and very differently, 
 and should yet arrive in the end at a certain similarity in historical 
 times, can be proved in more striking cases than that of Greek and 
 Latin. Polish, we saw above, regularly accents the penult ; the 
 same thing is true of the Cymric (Welsh) ; while their respective 
 closest sisters, the Bohemian and the Irish, accent the first syllable. 
 No one would therefore presume to hint at any kinship either 
 between Polish and Cymric, or between Bohemian and Irish, 
 closer than that warranted by the general fact that they all belong 
 to the I. E. family. On the other hand the Lithuanian and Lettish 
 both belong to the Lithuanian branch of the Lithu-Slavic family, 
 and they are so closely related to one another in sound and form 
 that Lithuanian may be changed into Lettish by the observance 
 of a moderate number of phonetic laws ; yet they have gone so 
 far apart in their accentuations that the Lithuanian still preserves, 
 to a fair extent, the free I. E. accentuation, while the Lettish 
 regularly accents the first syllables of all words. And it may not 
 be improper to point out that the Arabic shows a striking resem- 
 blance in its laws of accent with the Latin, as may be seen in 
 Caspari's Arabic Grammar, fourth edition, Halle, 1876, p. 22. It 
 will appear that Arabic accentuation is identical with Latin, (i) 
 in never accenting the last syllable of words of more than one 
 syllable ; (2) in always accenting a long penult ; (3) in never accent- 
 
36 
 
 ing a short penult. In fact it possesses every law of the Latin 
 excepting its restriction of the tone within the last three syllables. 
 Who would on that account alone attempt to establish kinship 
 between the accentual methods of the two languages ? Therefore 
 the supposed common origin of the. Greek and Latin systems of 
 accentuation stands upon the weakest possible ground, and an 
 explanation of the Greek recessive accent which ignores the exter- 
 nal similarity of the Latin may now be approached with reasonable 
 confidence. 
 
 VI. 
 
 The explanation of the Greek recessive accent must start from 
 the finite forms of the verb, where alone it is evidently at home. 
 This special nexus between the verb and the accent is not noticed 
 by Misteli and Hadley, and has been pointed out above as the 
 weakest point in their system. Yet the fact had been noticed and 
 utilized to a certain extent even by Gottling, who puts the verb on 
 the same level with the Aeolic accent in this respect. It is Wack- 
 ernagel's merit and the reason of his success that he began his 
 investigation with this fact as the basis. And he has succeeded, as 
 will be now shown, in explaining the Greek accent, as far as the 
 verb is concerned, by a series of qualities or laws of treatment to 
 which the finite verb was subjected in sentence-nexus in I. E. times, 
 so that the Greek recessive accent appears to be a development of 
 tone-laws pertaining to sentence-accent in distinction from word- 
 accent. We must from the start let the etymological accent of 
 the individual word lie latent, or better, keep in mind that the ety- 
 mological accent of a word may under certain circumstances vanish 
 under the influence of sentence-accentuation. 
 
 Wackernagel starts with the observation that both in Greek and 
 in Sanskrit the finite verb is occasionally subjected to enclisis, of 
 course with the greatest possible differences in other respects. In 
 Sanskrit, every finite verb becomes enclitic under certain conditions 
 and according to certain laws (see Whitney, Sk. Gramm. 592 fg.). 
 In Greek only two verbs in the present indicative, /u and $?y/u, 
 are enclitic. The old explanation, according to which this enclisis 
 was due to paleness of meaning, he rejects justly, because <j?/u is 
 no paler than Xc'yco. 1 He assumes, then, that this restriction of the 
 
 1 So far is this from being true that $HMI is, and continues to be, the 
 strongest of the verbs of saying, often meaning ' aver,' ' asseverate,' and some- 
 times actually taking prj as if a verb of swearing, /leyw in Homer is not yet a 
 full verb of saying. B. L. G. 
 
37 
 
 enclisis to these two indicatives is due to the Greek law of enclisis, 
 according to which an enclitic word may not contain more than 
 two syllables and three morae. This, it will be remembered, was 
 exhibited in detail above, p. 22, where the examples Zevs rjpiv, etc., 
 with enclisis of the orthotone fnuv t was shown to be the substitute 
 of the enclisis which is exhibited in Ztvs pot. Of course these are 
 not the only individual Greek finite verbal forms which, in spite of 
 this restriction to three morae, could be inclined, but here Wack- 
 ernagel recognizes with consummate acuteness that the present 
 indicatives of these roots represent the only cases in the language 
 where the entire paradigm of the tense or mood would allow the 
 enclisis throughout. A form like Xe-yw, Treitfe, rja-av would by itself 
 be capable of enclisis, but not Xeyo^ev, Xtyere, Treitfere, fj<rrr)v ; therefore 
 enclisis could not sustain itself in the paradigms to which these 
 words belong; on the other hand, the undisturbed capacity for 
 enclisis of ei/u, (ei), eon, eoroi/, ecrp-ev, core, <n ; (frrj/jii, (^y), 1 (frycri, (JHITOV, 
 $a/zei/, $are, $a<ri, without a single interloper that would be debarred 
 from enclisis by containing too many morae, is the secret of the 
 preservation of their enclisis. The test for other tense or mode- 
 systems is easily made and will always bring up some form con- 
 taining either more than two syllables or three morae. The enclisis 
 of these two present indicatives is then identical with the enclisis 
 
 in Zeus /zot. 
 
 The question now arises : What has happened to the other verbs 
 which were debarred from enclisis by containing too many morae ? 
 Precisely the same treatment that has happened to an enclitic pro- 
 noun of too many morae. They were inclined as much as possible, 
 in accordance with the principle exhibited in the change of ortho- 
 tone rjplv to enclitic fain, and orthotone foav to enclitic fjfitov ; just as 
 Zevff rjfj.o)v contains orthotone fj^v changed to fjpiv, just so does Zeus- 
 8oir) contain the prehistoric Son? = Sk. deyat ; however, not in its 
 orthotone, but in its enclitic form, for dofy is the enclitic to *8oifj just 
 as much as fj^v is the enclitic of r^w. This may be formulated 
 in the following proportion : 
 
 Zevs p.01 '. Zevy rjfjuov = Zeus ecrri : Zevs doir]. 
 
 The recession of the Greek accent in the finite verb is accordingly 
 everywhere not due to a process of accentual change within the 
 word, but to a secondary accentuation which is a substitute for 
 enclisis. It is false, therefore, to compare directly the accent of 
 
 1 El and $tf<; will be discussed further on. 
 
38 
 
 finite forms with corresponding accent, e. g. in Sanskrit. Thus, 
 the accent of Sofy is not to be compared with that of Sk. deyat, but 
 with * deydt, i. e. the enclitic form ; ^^(n^a^v not with babhuvimd, 
 but with -^ babhuvima; Ztvs opwa-t is Sk. dydils rnoti (the verb 
 enclitic). 
 
 But in one respect the Greek enclisis of the finite verb has over- 
 stepped what was no doubt an old law accompanying it, a law which 
 appears in the Vedic Sanskrit, cf. Whitney, Sk. Gramm. 591 fg. 
 The Sanskrit verb is inclined in independent paratactic clauses^ 
 except when it stands at the beginning of a clause ; the verb in 
 hypotactic clauses, or at the beginning of a paratactic clause, etc., 
 is orthotone. The Greek, it must be supposed, has forgotten and 
 given up this original distribution of orthotonesis and enclisis, and 
 has spread the analogy of the inclined forms over the entire finite 
 verb, 1 
 
 If the recessive accent of the Greek finite verb is regarded as a 
 substitute for enclisis, then we can understand why the participles 
 and infinitives are exempt from this accent with such perfect regu- 
 larity. These forms were never subject to enclisis and have there- 
 fore retained their etymological accent in Greek undisturbed, even 
 more so than noun-categories which stand in much looser relation 
 to the finite verb ; for these have often adopted the recessive accent. 
 This result is obtained by comparing Sk. bhdran with <epo>j>, ricdn 
 with XITTCBV, rnvdn with opvvs^ydn with lav, babhiivan with nc^vus ; so 
 also XiTreo/, \i7Tfo-6ai, etc., which exhibit the same accent of the thematic 
 vowel as in \nr<av, have remained undisturbedly in the possession 
 of their prehistoric etymological accent. In the same manner the 
 accentuation of verbs compounded with prepositions is explained. 
 The finite verb is inclined upon its preposition, sdm bhara (written 
 with tmesis in the Vedas) = o-^epe, dpi' asti = escort ; on the 
 
 1 It is interesting in this connection to mention that the enclisis of the Sans- 
 krit verb had been regarded in the light of a prehistoric quality of I. E. speech 
 much before Wackernagel ; to be sure only in a casual mention. In a programme 
 of the gymnasium at Wismar, 1869, there appeared a paper by Sonne entitled 
 ' Zur ethnologischen Stellung der Griechen,' in which he writes : When we see 
 that in Sanskrit the verb of the principal sentence is inclined upon every 
 preceding 'objectiv-bestimmung,' we believe that we must recognize in this 
 phenomenon, as strange as it is to our European conceptions, a remnant of 
 proethnic accentuation (p. 3, cited by Delbriick, Sprachstudium, p. 132, note). 
 He has in mind the coinciding enclisis of Greek elfii and Qqfu in making his 
 statement, but he never extended his idea in any way beyond this mere sugges- 
 tion. 
 
39 
 
 other hand, here again the forms which are not enclitic when un- 
 compounded retain their accent, and the preposition loses its accent 
 both in Greek and Sanskrit, vTroXaftw, CTTIUV ; in the same manner 
 
 KadrjTai and <araKrcu, but Kadr)<r8ai, KaraKeur&u, cf. Whitney, Sk. 
 
 Gramm. 1083. 
 
 Wackernagel turns next to the second persons el and c^?, which j 
 are orthotone, and would endanger his entire explanation unless 
 their orthotonesis is explained. The explanation which is proposed 
 is a totally different one in each case. 
 
 For ef an etymological explanation is attempted. This word is * 
 Attic and Ionic, but post- Homeric ; it is a form, then, which is j 
 later than the period in which the enclisis of the verb was fixed. 
 Possibly it may be restricted even to Attic alone, inasmuch as it 
 has been removed by Stein from Herodotus. 1 In order to explain 
 this late and contracted et, Wackernagel assumes that it is a middle 
 form *e<reo-ai to ecroptu. Such a word, containing as it does three 
 syllables, would, owing to the limitations of enclisis, not become 
 toneless, but would appear with recessive accent as a substitute for 
 enclisis in the usual way, and this *e(<rX<r)at, *?e<u would then con- 
 tract to el, as *7roie(jj/)e(ai, iroucai becomes TToiet. But there are at 
 least two objections to this explanation. First, the natural expla- 
 nation of ei, which seems to be almost unimpugnable, is a totally 
 different one. The word, whether restricted to Attic or not, is f 
 evidently old ; it is *n = Sk. dsi = Zd. ahi = Goth, is = Lithu- 
 anian esi and Old Bulgarian j-esi / the assumption of a ground- 
 form *eo-eo-at is therefore unnecessary and improbable. Secondly, 
 Wackernagel has assumed with indubitable success that within one 
 tense-system, forms which by themselves could have been enclitic 
 became recessive by the attraction of the rest of the system ; why 
 has not the analogy of the enclitics in the paradigm of '/u succeeded 
 in overcoming this single recessive example in its turn ? It seems 
 therefore much more probable that the lack of enclisis in ef is due 
 to the influence of contracted forms in general. At the time when 
 *e(o-) contracted to et, other contractions taking place at the same 
 
 'According to Veitch, Greek Verbs Irregular and Defective, Stein and Abicht 
 read efc , while Becker and Dindorf read el. 
 
 2 The Indo-European form of the second person singular was *esz, e. g. the 
 two j's coming together from the root es plus the -si of the second person singular 
 were simplified into a single s by some I. E. law of sound, before the separate 
 existence of the languages of the family. Neither in Greek nor in Sanskrit 
 would the theoretical *essi lose one of its s j s. For the Sanskrit, see Whitney, 
 Sk. Gramm. 166. 
 
40 
 
 time received the circumflex so generally that this form received it 
 also, and was protected from the attracting influence of the enclitic 
 paradigm, which it belongs to, by that same contract character. 
 
 The case of $ys is quite a different one. Long monosyllabic 
 finite verbal forms in Greek are regularly perispomena, that is, 
 barytone or recessive ; so /3?}, ^ (or analyzed into morae, */3, *0ee); 
 if this word does not receive the circumflex, it is therefore not 
 accented recessively or quasi-enclitically, but is orthotone. Wack- 
 ernagel ingeniously finds the cause of this orthotonesis in its pecu- 
 liar function. 'A speaker never gets into the situation in which he 
 speaks to another person about his (the other person's) (frdvat in the 
 present tense, without bringing the (f>dvat of that person into relation 
 to something else, or otherwise when the <j)dvat of that person is 
 still unknown to the speaker ; in other words, grammatically 
 expressed, fyys occurs regularly in subordinate and interrogative 
 sentences,' loc. cit. p. 461. That this rule is actually and not only 
 theoretically true, Wackernagel then proceeds to show by bringing 
 on all the passages in Homer and the tragedians in which the word 
 occurs, p. 461-2, and his statistics bear him out completely. 
 
 The orthotonesis of this word in subordinate clauses is then 
 identified with the regular orthotonesis of the verb in Vedic Sanskrit 
 when it occurs in subordinate clauses (Whitney, Sk. Gramm. 595) 
 and the orthotonesis in interrogative sentences, with a very similar 
 rule for the Veda, according to which verbs in interrogative sen- 
 tences retain their tone, or perhaps rather have their natural ety- 
 mological tone, heightened still further by the rhetorical tone, 
 natural in questions, cf. Whitney in Kuhn and Schleicher's Beitrage, 
 I 200. It is clear that the criticism made above against the 
 assumption that an old *eVeo-ai which functionally was not different 
 from ta-fu should remain orthotone and resist the analogy of the 
 rest of the forms of the paradigm does not hold good here, because 
 there is a thoroughgoing functional difference in 0^? which might 
 well hold it above the forces of assimilation ; especially true might 
 this be in the case of the peculiar interrogative tone, which this 
 word is subjected to with especial frequency. In the later literature, 
 as representatives of which Wackernagel brings on Plato and the 
 comedians, this occurrence of fys in interrogative and subordinate 
 clauses is not so strictly adhered to ; he finds in 140 passages 18 
 not interrogative and not subordinate, but these passages are made 
 to yield strong support to the correctness of his method in bringing 
 on the Vedic accent for constant comparison ; they also are explained 
 
by Vedic analogies. The word ^g occurs in these 18 passages in 
 the first one of two paratactic clauses, e. g. Plato Gorg. 491 B, < 
 pep yap <f)ys . . . tya* Se vov rovvaiviov. Compare with this Whitney, 
 596 : ' The verb of a prior (principal) clause is not infrequently 
 accented in antithetical construction. Sometimes the relation of 
 the two clauses is readily capable of being regarded as that of 
 protasis and apodosis ; but often such a relation is very indistinct.' 1 
 Of course the Greek example comes under the head of antithetical 
 construction ; in the same manner the other 17 examples of Plato, 
 etc., are readily disposed of. It seems that Wackernagel has 
 beyond peradventure pointed out the correct reason for the pecu- 
 liarly isolated position of the word ^rjs in accordance with the rules 
 of Vedic and Indo-European accentuation. 
 
 He turns further to various minor specialties of the recessive 
 and enclitic accent, and explains them again in accordance with 
 well-known laws of Vedic accent. Only the most interesting of 
 these, the orthotonesis of ?OTI, will be mentioned. The older Greek 
 grammarians, according to Lehrs, Quaestiones epicae, p. 126, know 
 of no functional difference between eon and * e<m, but teach that 
 the orthotone word stands at the beginning of the sentence and 
 where certain particles, etc., immediately precede the word. Ac- 
 cording to some, only ou has this effect ; according to others ou, /cat 
 and is ; ef, dXXa, and TOVTO are also added by a few. With the 
 exception of rovro these words are either too weak to allow inclina- 
 tion upon them, or, like K<U, are not real members of the sentence 
 which they introduce, so that the m which follows stands in reality 
 at the beginning of the sentence. This peculiarity is again explained 
 by a rule in Whitney's Sk. Grammar, 593, 'The verb of a 
 principal clause is accented when it stands at the beginning of the 
 clause,' e. g. syama id indrasya Qdrmani, ' may we be in Indra's 
 protection.' 
 
 Other details of Greek accentuation, which need not be repeated 
 here, are successfully explained, and everywhere Wackernagel's 
 results are strictly in accordance with the principles which have 
 been stated above for all kinds of phonetic investigation, and they 
 are themselves new proof of the success of such investigations when 
 carried on with these principles. In the first place every line of 
 his investigation is permeated with the thought that it is not allow- 
 able to discuss the accent of the separate I. E. language without 
 
 1 E, g. prd-pra 'nyJ ydnti, pary anyd asate : 'some go on and on, others sit 
 about.' 
 
42 
 
 taking for a basis the reconstructed I. E. accent. Further, this 
 
 I. E. accent could only change by regular phonetic law or by 
 
 -x analogy. Both factors are shown to have been at work. The 
 
 | phonetic law is the Greek law of enclisis by which real historical 
 
 I enclitics appear accented, though in manner clearly enough a mere 
 
 I compensation for enclisis; the reason for this phonetic law lies 
 
 ^Uwithin the province of phonetics just as, e. g. the rhotacism which 
 
 A^changes in so many languages an s to an r. 
 
 The workings of analogy we saw in many ways ; above all this, 
 that the enclitic character of the verb in principal clauses has been 
 extended to the verb in subordinate clauses. It would be interest- 
 ing in this connection to count the number of principal and 
 subordinate clauses in Homer; no doubt the principal clauses 
 would preponderate, as they most certainly do in the Rig- Veda. 
 Wackernagel is the first one who has clearly established any kind 
 of law as regards the sentence-accent of the I. E. languages, the 
 leading fact being the enclisis of the finite verb in principal clauses. 
 His results prove completely the fact that the study of accent 
 cannot be carried on from the point of view of the word alone, but 
 that it must also consider the larger speech unit, the sentence, and 
 perhaps ultimately also the smaller, the syllable. 
 
 Wackernagel does not carry his results beyond the finite verb, 
 but he leaves no room for doubt that the nominal accent in Greek, 
 so far as it is not archaic and etymological, is enclitic and recessive. 
 No doubt the noun has to a large extent followed the verb in its 
 enclisis ; the Vedic accent leaves us here almost entirely, but yet not 
 altogether. In the Veda the vocatives are accented only when they 
 stand at the beginning of a sentence, or clause, or verse, elsewhere 
 they also are enclitic ; see Whitney, Sk. Gramm. 314, and Haskell 
 in the Journal of the American Oriental Society, Vol. XI, p. 57 fg. 
 Further, an adjective or genitive qualifying a noun in the vocative 
 constitutes as far as accent is concerned a unity with it. Thus there 
 arises in the case of a vocative in the middle of a clause a group 
 of two or three, sometimes even more, unaccented nouns, cf. above, 
 p. u, note 2. The quantity of enclitic vocative material cannot 
 have been very great at any period in any language of the family, 
 yet it may have at least helped on the analogy of the verb in its 
 inroads upon the noun. Possibly future investigations may 
 succeed in pointing out the details of this process in an acceptable 
 manner. 
 
 MAURICE BLOOMFIELD. 
 

 411292 
 
 
 UNIVERSITY OF CALIFORNIA LIBRARY 
 
S*i*4t