Vous êtes sur la page 1sur 12

Linguistic Society of America

On the Development of the Greek Intonation Author(s): Jerzy Kurylowicz Reviewed work(s): Source: Language, Vol. 8, No. 3 (Sep., 1932), pp. 200-210 Published by: Linguistic Society of America Stable URL: http://www.jstor.org/stable/409650 . Accessed: 28/01/2012 05:21
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

Linguistic Society of America is collaborating with JSTOR to digitize, preserve and extend access to Language.

http://www.jstor.org

ON THE DEVELOPMENT OF THE GREEK INTONATION


JERZY KURYLOWICZ
UNIVERSITY OF LWoW

[An attempt is made to show that the intonation designated by the Greek circumflex is not inherited from IE, but is produced partly by specifically Greek contractions, and partly by the analogical extension of the intonation thus arising to morphologically similar forms.]

The traditional theory of the Indo-European origin of Greek intonations is founded on the comparison of Vedic with Greek and on the comparison of Lithuanian with Greek. As to Slavonic, it seems to present a development more recent than Lithuanian and therefore its testimony regarding intonations is hardly to be accepted as an independent one (cf. my article, Le probl~me des intonations balto-slaves, in Rocznik Slawistyczny 10. 1-80). Streitberg thought that certain phenomena concerning abbreviation or conservation of final long vowels in Germanic were related to the corresponding phenomena in Lithuanian and were to be explained by original differences of intonation. But if, as has been shown in the article just quoted, these Lithuanian phenomena have nothing to do with intonation, although this language offers a distinction between two intonations, the existence of similar phenomena in Germanic, even if complete parallelism between Germanic and Lithuanian has been proved, does not involve the existence of intonations in Protogermanic. It is possible that the rather complicated treatment of final vowels in Germanic depends not only upon the phonetic circumstances (whether the vowel is short, long, or a diphthong, whether it is final or followed by a consonant, whether this consonant is an occlusive, s, or a sonant), but also upon the functional value of the phonetic elements in question (accounting for secondary arrangements). In his book Sprachkorper und Sprachfunktion, Horn has shown how differences of phonetic treatment are to be explained by different degrees of functional value. This, however, does not interest us for the moment. The chief point is that there are so many principles explaining the differentiation of finalvowels that to introduce a new and a problematic one is to complicate the matter still further. 200

ON THE DEVELOPMENT OF GREEK INTONATION

201

A survey of Vedic dissyllabic vowels (cf. my article, Quelques problImes metriques du Rigvyda, in Rocznik Orjentalistyczny 4) shows that on the whole their dissyllabic character is etymological. But not all vowels proceeding from contraction are dissyllabic in the Rigveda. We find only seven examples (five different words) of the ablative ending -dt scanned as two syllables (Arnold, Vedic Metre 99) and not one case of a dissyllabic -d- in dya (dative sing. of a-stems), which etymologically must be the result of a contraction. There is no sure example of a dissyllabic -ds (nominative pl. of a-stems) or of a dissyllabic -ais (instrumental pl. of a-stems), where prehistoric contractions seem also pretty sure. We find no sure case of the dissyllabic character of the optative suffix -e- (< -oi- < -o + i-), or of the subjunctive suffix -d- (< -j/5- < -e/o- + -e/o-), etc. The only case where a dissyllabic Vedic ending corresponds to a Greek ending with circumflex intonation is the -Jm of the genitive pl. (Greek -cv). It is, however, to be remembered that the Greek ending -cw is never dissyllabic. There is also no trace left of the originally dissyllabic character of the dative ending -w, of the optative suffix -o0- or of the subjunctive suffix -7/w-. All these contractions must have been effected beforethe limitation of the accent to the three last syllables. We have X'yo suE,XMyflrE and not *0 K;V < *XEIoL-E, *XEYr-r < *XEyb-6-~y., *XeY)-e-rE, otKWV and not Even the long vowel of the second member of compounds *oK6-WV. like WA-77oris never dissyllabic, although it is of a more recent origin than the long vowels just mentioned. But the apparently weightiest argument that has been advanced in favour of the Indo-European origin of Greek intonations, was that of Bezzenberger, BB 7.66 ff., who compared systematically the flexional endings of Greek with those of Lithuanian. The Lithuanian difference between algd and algos is, according to Bezzenberger, genetically the same as the Greek difference between aXpi and &Xpis. It is not our task to discuss here the reasons of this fundamental error of Bezzenberger, which has influenced Indo-European linguistics for generations. Suffice it to observe that the possibility of such a comparison between Greek and Lithuanian is swamped by the recent results of Slavonic linguistics (I refer to the above mentioned article in RS 10). In his Litauische Mundarten 2.201, Specht has formulated a law according to which there is no distinction of intonation on the last syllable of Lithuanian words, because the long vowel or diphthong of the final syllable may show only the so-called circumflex intonation. But under the influence of current ideas concerning certain secondary phonetical phenomena

202

JERZY KURYLOWICZ

(as accent-shift and special abbreviation of final vowels) Specht admits that a previously existing distinction of intonation on final syllables has been eliminated in a prehistoric period of Lithuanian. In reality the comparison between Lithuanian and Greek in the matter of intonations is as specious as etymologies based on similarity of sounds. Bezzenberger had been the victim of an illusion consisting in the using of the term intonation for two functionally' different things (as one and the same sound may be functionally different in two different languages); even the usage of the same symbols for both languages (' ' ') may have contributed to this illusion. Moreover comparisons established between Lithuanian finals and Greek finals do not yield homoris, then geneous results in all cases. If algos, tos correspond to aX,?ps, to is not the case. If which algq, til should correspond *daXpi,i*rv, the to ancient locative (now adverb) anksti aX'pi, algd corresponds in to instead of -ol. The final Greek locatives should correspond *-ol nasal is of in Lithuanian, intonation combination vowel + susceptible insist on detail. The principal we etc. But will not but not in Greek of is but that not details systems, systems must thing to be remembered But as Meillet. as has ever and ever been repeated by be compared, the till now to understand has know no made been far as we attempt in to their i.e. the Greek intonations relation system of Greek intonation, functional value. There are grammatical categories in Greek where such functional value is apparent (see below), but they have not received the attention they deserve. We shall not lose any words on the strange constructions based upon the simultaneous admission of the Indo-European character of Greek, Lithuanian, Germanic, etc. phenomena. Compare, e.g., Brugmann, KVglGr.1 54: the circumflex intonation in Indo-European is due either to contraction (as in Greek r7tA < -d + ei), or to loss of the second member of a long diphthong (as in Lithuanian akmu6 compared with Greek aK.' qa < -d + so). cV), or to loss of a syllable (as in Greek 7r if in It seems to us that we succeed understanding the intonation find it impossible not only to trace system of the Greek language we will that system back to the Indo-European period, but even to any period much older than the oldest Greek documents (Homer). For the determination of the function of Greek intonation it is sufficient to explain the role of the circumflex intonation; the acute intonation is nothing else than a lack of circumflex intonation, i.e., it indicates the normally
1 The function of the intonation in Lithuanian is described in the article referred to (RS 10), for Greek intonation cf. below.

OF GREEK INTONATION ON THE DEVELOPMENT

203

accented vowel not subjected to special phonetic or morphological conditions. The question of the physical description of both intonations is, for our purposes, an irrelevant one, because it is clear that by the fundamental linguistic (or rather psychological) law of contrast a relation such as, e.g., lack of intonation descending intonation may at any moment, by internal opposition, develop into rising intonation "descending intonation. The Greek' circumflex is conditioned in the penult and does not exist in the antepenult. Distinction of intonation in unaccented syllables occurs only in final syllables containing diphthongs, but not in final syllables containing simple long vowels, and not in the penult. But it is chiefly in final accented syllables and in monosyllabic words that the Greek circumflex is autonomous. A first superficial analysis of the whole available material referring to autonomous circumflex, leads to the following tripartition3: > os > (pcas, (1) Circumflex due to a historic contraction: do Ebyevos etc. El'ryov-s, (2) Circumflex due to a morphological reason (see below): 0P (aorist of faLvw), Kp 'heart', etc. (3) Circumflex ascribed to prehistoric phonetical reasons: KaX, 7r~SI, fois, etc. Compared with the groups (2) and (3) the group (1) representing historic contractions is by far the largest. A great part of Greek morphology presents the phenomena of historic contraction, paradigms of declension and conjugation as well as processes of derivation. We have contracted paradigms in all the three declensions (types OvKI, 'EpAsis; - 7rt~j, 7r6Xts,3aaoXEb03, etc.), 7rXov s; ye'vos, IYEVI, K ai&vSw, s, oo-Obdo 7 monosyllabic contracted substantives of all three genders (e.g. ral~, Xas, Kps, a7p, Yppp, etc.,), case-forms which are always Op , ols, oi~, ~cp&s, contracted even in otherwise non-contracting paradigms (rCo&v< ,rqa'wi, -w < -ao in the corresponding masculine type, &tpnrov < tw6p roo), contracted derivatives in -ovs, -aLos,-ido, -o1io, -cn < -fos, -aios, -4ios, -bi'os,-ecw',contracted pronominal forms (like A.oi, ooi, oi,
2 The Greek dialects are not being considered in this paper, the data concerning their systems of intonation being too fragmentary and uncertain. 3 Several cases of autonomous circumflex have been disregarded because of the uncertainty of the tradition. Adverbs in-i seem to be in reality adverbs in -5 (datives). Grammarians disagree about the intonation of LxOvs, pvs, bopvs, and itvs; and the intonation of Homeric Z71vandBov is scarcely sure enough to be treated

matic verb have been disregarded because of the uncertainty of their etymology.

on a level with cases like iaX5, rLjLjS, 0o6s, etc.

The endings -EL,-E L of the the-

204

JERZY KURYLOWICZ

t , etc.), contracted presents in -a4, -w , -b and typiZv, W-Esl, tC, in -4w, verbal forms which are contracted even in otherwise futures non-contracting paradigms (rTOLdiat,reOdihat; CPpov, Epfpn < 'VPipEo, (pfpeat; (p~pELt< (ppeev), contracted subjunctive and optative forms (of the -yt verbs, of the aorists in -n77, -tvP, of the medio-passive perfects like etc. We can admit that in view of the important etc., safely 0#E3'XOuaL), r'le played by contraction in Greek grammar, groups (2), (3) of our classification contain only a very small part of all cases of autonomous circumflex intonation. The groups (1) and (2) are generally explained as properly Greek phenomena. The difference between the two groups consists in that the circumflex of (1) is phonetically conditioned, whereas in group (2) it is morphological. But in our opinion a deeper interpretation of its morphological character throws a light also on the circumflex of group (3), and seems to suggest, that there is no fundamental difference between group (2) and (3). But if it is so, if the circumflex of group (3) is also morphologically conditioned, then there is no reason whatever to suppose that the Greek intonations have been inherited from Indo-European. Let us therefore examine, item by item, group (2): All monosyllabic neuters of the third declension offer the circumflex, if the radical vowel is long or a diphthong, e.g. Kip, etc. This OwKp, 7rp, rule appears as if a corollary of the rule that all polysyllabic neuters of the third declension shift the accent as far as possible towards the initial syllable of the word (there is no exception to this rule), cf. 6voya,&XEupap, oVELbos, r'Xayos,aivart, etc. The two moraeof the monosyllableact as two syllables. The simplest explanation of this phenomenon is the following: in all contracted dissyllabic neuters the circumflex is phonetic because they must have been paroxytone before the contraction (as > ope, ppbap > ppip, iosr> ois etc.). That is, if we disregard the pahos small group of inherited monosyllables like Kip, the neuters of the third declension presented either the recessive accent or the circumflex. After the contraction had been effected, the inherited monosyllables followed the general pattern. The proportion Zeb : Zei = avp : vep (as concerns accentuation) implies that the diphthong of the monosyllable is treated as if resulting from contraction (which in fact is not the case). But if it resulted from and contraction then of course the difference of intonation between ZEbS the must admit As above we would be ZeM phonetically justified. influence of really contracted forms, that is the working of the following formula: the vocative presents either the recessive accent or the circum-

ON THE DEVELOPMENT OF GREEK INTONATION

205

flex (if the word is a contracted monosyllable). Hence the analogical transfer of the circumflex to monosyllables which never have been subject to contraction. The monosyllabic forms of the personal verb (with long vowel or diphthong) have always the circumflex, e.g. /i < *gadt (athematic aorist). As in the preceding cases the old uncontracted forms (03ietc.) adopt the pattern of contracted forms (ara- < asre etc.). The circumflex of the latter is phonetic, because all personal forms of the Greek verb have the recessive accentuation. All these examples show, for certain grammatical categories, an intimate relation between the recessive accent in dissyllabic and polysyllabic words, and the circumflex intonation of monosyllables. For the two last cases this relation has been pointed out by Vendryes (Traite d'accentuation grecque 47). Our task will be now to show not only that this formula applies to a certain number of other cases, but that, conveniently enlargened, it covers practically all cases of group (3). The monosyllabic stems of fos~, aPs and ypaDs (genitive pobs, vabs, ypabs) must be compared with dissyllabic u-stems. In fact the existence of case-forms like po-(F)6s, j0-(F) , po-(F)Cov, etc. as opposed to the nominative P3o?,etc. suffices to make the speaking subject conceive the latter as fb-v-s (va-v-s, ypad-v-).4 But it is known that all substantive u-stems have an inherent recessive accent in Greek (only the oblique cases of u-stems are oxytone, if dissyllabic, e.g. yov6~F, 5opF's beside Iyb6v,56pv). Therefore poos, vabs, ypavs (with circumflex intonation) are nothing else but the result of an interpretation *pb-v~-s, *d-v-s, *ypa-v-s (and not *po-b-s, *va-'-s *,pa-b-s). The same is true for the monosyllabic stems of 6s(a-s), j1isand pDts (genitive br, orbs, ivbs, apvbs), conceived as *i-v-s, *0b-v-S, * b-v-s,*5pb-v-s. There is a very instructive difference between these four stems and the derivatives in -5- and -rS-, like txo-'v, qr7v's etc. Whereas in a prehistoric period the u-stems had obtained a recessive accent, all u-stems had remained oxytone. There was no pattern to change the intonation of '77rv's (gen. 5Tirbos) to a circumflex, because the accented suffix -(t)$- had always been felt as a unit and as different from the unaccented suffix -u- and did not, therefore, admit an analysis -(t)u + u-. The analysis -u + u- was possible only with root-monosyllables like bs, where the oblique cases were oxytone, because the oblique cases of u-stems are either oxytone (like yovFUr, aopFbs)or else have recessive accent (like pAOvos, rfeXKvos); l-stems, on the other hand, accent always the suffix, never the ending
SCf. the secondary length in Pv7vs, ypqvs.

206

JERZY KURYLOWICZ

or the root. The oblique cases of -(t)ii-stems, where is always -uytherefore have been could not associated with the accented, corresponding cases of u-stems, where u is never accented. Our explanation is confirmed by the paradigm of Zebswhere the lack of case-forms beginning in ZEF-(as corresponding to Vedic dy6hh) *ZEF6 for the lack of intonation in the nom. accounts circumflex sufficiently
Zebs.

All substantive stems in - t- have the recessive accentuation. Therefore the monosyllables KI~ 'curculio' and Xis 'lion'5 (gen. KLbS, XL6t), have the recessive accent, hence their conceived as *KL-L-s,*XLt--S, s: FF = KLS,XL6 *K-LS,*XL-L. But circumflex intonation; cf. bFL and not *W~s has no circumflex because its genitive is "force" "1s and therefore its nominative cannot be conceived as a 'v6s contraction from The indefinite adverbs wrj,Rwc,w&s, rol are enclitic and oxytone, the corresponding interrogatives, dependent interrogatives, and relatives are all perispomena. If we compare dissyllabic adverbs which have different accentuation according to their function, we remark that they are oxytone when indefinite (as 6ri, wor7, Yrodv) and paroxytone when interetc. We (i-re,R6re,wrt6ev). rogative, may assume that the influence of the forms with recessive accent on the monosyllables was made possible by monosyllabic forms resulting from contraction (as roD, ob). confinSome of the Greek grammarians distinguished also viv from vvY, of former the of to that the to strict sense the latter sequence or ing time, inference (= 5' or o3v). The difference between an oxytone monosyllable with indefinite meaning and corresponding perispomenon with definite meaning exists also, in our opinion, in the case of JEs 'one', which is oxytone when used in the sense of 'anyone' in the compounds Aq-eds,obi-els. The relation of els to ,LA?7els is the same as that of i to 9reei. The circumflex of eTsis of course secondary (like that of ri5, r 6, w~7s, ro). It must be relatively recent, being posterior to the passage of *bs > Eis. The interesting detail is that the functional value of the circumflex intonation creates new circumflexes which have nothing to do with any kind of contraction. Like els the nom. sing. ris is of a relatively recent origin (<*rdhvs).6
5 All grammarians agree about the circumflex intonation in the accusatives. As to the nominative, Herodian and Aischrion seem to admit KSS, X~s against Aristarchos (Chandler, A practical introduction to Greek accentuation2 162). 6 In the Septuaginta we also find a late acc. sing. irav (after the nom rais).

OF GREEK INTONATION ON THE DEVELOPMENT

207

Its circumflex intonation corresponds to the recessive accent found in certain cases of the paradigm. Whereas the genitive ravr6s and the dative ravnr are regularly oxytone, the forms r&ivrotv (gen.-dat. dual), irawrov (gen. plural) and riro- (dat. plural) have the recessive accent. As in all the preceding categories it is not our task here to investigate what the reason of this recessive accentuation is. All we allege is that the primary phenomena of accent-shifts in dissyllabic and polysyllabic forms are reflected, as in a mirror, by intonation-changes in closely related monosyllables. The long vowels of eIsand irasdo not result from contraction but from compensatory lengthening after the disappearance of v (or after the loss of nasalization). In general such long vowels have the acute intonation, if there is no particular reason for the circumflex intonation (as in the two preceding cases). This is very clear in the case of the acc. plur. where we have -obs and -a's in the two first declensions, but -oos, if this explanation of Wackernagel's -eEl(< -b6v, -f's : eirevei < *e&Cyieps is correct) in the third declension, because the intonation of the corre(< -bes,-4s) has been introduced into sponding nominatives in -ovs, -CEs the accusative. In the two first declensions such influence has been impossible, because the endings of the nom. plural are -o0, -al. Compare also the acute intonations of the nom. sing. of the participles in -obv,-eLs,-&'s, etc. Goiob, 6'vros, a6vTL, a6vra, etc., (the accentuation is not recessive, but columnal7 cf. &rosobs,&dro66ro3 and not *&rb7ovs, * rborvros). The formula concerning the relation between recessive accentuation and circumflex intonation may be generalized as follows: Whenever, in a grammatical category, forms accented on the final syllable coexist with forms accented on the penult, the former obtain the circumflex intonation, if the final syllable contains a long vowel or a diphthong and if it may be conceived as a contracted syllable. It is clear that our first formula becomes only a special case of the general formula, because a monosyllabic word may be considered as a word accented on its last syllable. Examples: The ending of the dative sing. of the first decl. (-? or - ) and of the dative sing. of the second decl. (-Cp). In the 3-d decl. the oxytone stems have either the accented ending -i (only with monosyllabic stems) or a complex consisting of accented suffix + unaccented t (-f&, -pL, -15t, -6at, etc.), or, finally, a long syllable resulting from contraction -dr7r -4-? in e&-yevet, -o! < -6-L in al=oL). As the ending of the dative < (-ed
7 For the meaning of this term see below.

208

JERZY KURYLOWICZ

sing. of the first and second decl. contains the characteristic element i it is conceived which appears also in the contracting group (type circumflex. In other as resulting from a contraction and obtains the e'reve), has a phonetic circumflex, because -eZwas once dissyllabic words: elyeveZ has a morphological and accented on the first syllable (-4-t), and KaXw3 circumflex, because it is modelled on the pattern of the type eyeve. The ultimate reason why the monosyllabic complex -c has the circumflex intonation on its last syllable is that the dissyllabic complexes -C', -lat, etc. have the recessive accent within their two syllables. But this pattern of the dissyllabic complexes has worked only through the medium of the contracted types (-e6etc.). It is not because they result from an Indo-European contraction, but because they are modelled after forms proceeding from Greek contracc have the circumflex intonation. tion, that the endings -a and The accentuation of oxytone nominal stems may be either 'columnal' or 'marginal' (we use the expressions introduced first by de Saussure in his famous article on Lithuanian accentuation, cf. Recueil des publications scientifiques 532): an oxytone paradigm has columnal accentuation if the accent is always on the same syllable counting from the beginning of the word (e.g. 7art-p, rarpbs, WraTepa, rarp&o-,s where the accent is always on the second syllable), and if the shortest form of the Another example: ~XTls, eXTrios, XrtLa, paradigm is oxytone (arwT"p). has 'marginal' accentuation, if all its forms etc. An oxytone paradigm are accented on the last syllable, e.g., KaXbs, -oi, -4, etc. or rut, -1s, -1, etc. The following rule applies to Greek oxytone paradigms: forms with marginal accentuation have the circumflex, if there exist corre7 and sponding forms with columnal accentuation. For instance, 7rt/ KaXwhave the circumflex intonation, because there are corresponding forms with -t ending and columnal accentuation (as irotpit, XTrirt, etc.). etc.). The same is true of rtEtis (s-ending as in rotzivos, c'XrLa0s, This correspondence between columnal and marginal paradigms has of course been established through the medium of contracted paradigms. For instance the direct model of -is (< -ps) is the circumflex of cases like -ois < -6bos, -4(o)os, -6(t)os, the indirect model is the columnal etc. accentuation of cases like -Evos,-ipos, -5os, -bos, -Amros, The same is true of the genitive pl. ending -wv which must have the circumflex every time it is accentuated (KaXMv, robov), because there are corresponding forms with columnal accentuation (rotphvwv, X'Xriwv, etc.). In adverbial formations the ending escapes from this influence
s We disregard of course the peculiar accentuation of the vocative.

ON THE DEVELOPMENT OF GREEK INTONATION

209

of barytone forms and offers the acute intonation: cIrro0WV, 7rpo-roCbv, fba -rw'V. qz'obw'V, The adverbs in -ws are accented as the genitive plural of the corresponding adjective, i.e., they retain the accent of the adjective. If the latter is oxytone, we obtain the three types rvx60rwS, (> ebyEvYls) EbyEvYiL and KaXWs with circumflex intonation after eibvrGjs (the type with marginal accentuation resulting from columnal accentuation by means of contraction). Originally the -os suffix must have been restricted to the o-stems, as it is probably related to the Indo-European ablativeending, but the spreading of -ws over all consonantal stems must antedate the circumflex intonation of the type KaXs. One may make the same remark with regard to the adverbs in -ot (originally only with o-stems). The attaching of -ot to other stems (cf. also i~v8o, 'ot) and the subsequent contractions, e.g. lHaLavoZi (< Illepatoi Hllaava), (< livx6s; (< IIvtoi (< IIv06) caused the -ot of forms like ,vyxoZ IlEpateis), to be 'in the interpreted as resulting from a contraction of a interior') with the first syllable carrying the accent. dissyllabic complex The optative suffix -ot- of the thematic verbs is treated as if it had the circumflexed intonation (raL~ebots, raicebot and not *wraLEvoLs, *ralaevoL). This is easily accounted for by the fact that the optative element -1was still productive after the period of the limitation of accent: rL0eEepv, and not *rLWELjEv, < *~iowev. The existt56oAe-V 6lab-j-eUl r-ti---~~,y, forms was sufficient to make the speaking subject decompose ence of such -oL- into -o + t-. In summing up we may say that it is unmethodical to draw a distinction between group (2) and group (3). In (2) we have the cases where the morphological circumflex is quite clear and generally accepted. But if we try to apply the corresponding morphological principle to group (3) we find that there is not one case left for which it would be necessary to recur to Indo-European. The morphological principal is the following: monosyllabic forms become circumflex, if related dissyllabic and polysyllabic forms offer the recessive accent. Last syllables of dissyllabic and polysyllabic words become circumflex if these syllables correspond to dissyllabic paroxytone complexes in related forms. In this manner recessive intonation in the syllable corresponds to recessive accent in the word. This equivalence between circumflex and recessive accent (which, as one knows, holds good for really contracted forms), could have been established only after the first contractions, i.e., after the disappearance of intervocalic a and t. It cannot be essentially older than Homer, because the Homeric text knows both contracted

210

JERZY KURYLOWICZ

and uncontracted forms; therefore the main bulk of the text was composed when the first vowel contractions had already begun. A counterproof is furnished by the innumerable cases where a final long vowel or a final diphthong has the acute intonation, because it cannot be conceived as resulting from the contraction of a form with recessive accent. The dual endings -&, -a' (in the first two declensions) do not show any trace of the ending -e and are therefore not conceived as resulting from contraction (contrast -4, -is, where the elements -L, -s characteristic have been recognized). The same is true of of the dative and genitive, -.3, the endings -ol, -at of the nom. plural (no final -s). From the Greek point of view, to take another example, the verbal endings -lia, -aat, con-rat, -vraL cannot be analyzed and interpreted as products of tractions and are therefore treated as having the acute intonation, i.e. they admit the proparoxytonesis of the corresponding verbal forms. This applies also to infinitives in -vat, -(a)al, and to adverbs like xaiac where the t- element has ceased to be felt as connected with the dativelocative ending t because of the special function of these forms. It seems that besides the forms proceeding from contraction there is only one morphological group owing its circumflex to a phonetic and not to a morphological reason, the vocatives fo6, -ypai, IA, 3atLXEv, iXTV, 7xoZ,Zei, etc. As a rule all such vocatives, unless influenced by formally identical nominatives (as in the case of tfb7P,alcv, 7rotAfV, bX'p's, the peculiar bor p etc.) present the circumflex intonation. Of course falling intonation of the vocative became phonemic and morphological only after the opposition between acute and circumflex had arisen. A very instructive parallel is found in Serbo-Croatian, where every vocative has the falling intonation on the first syllable, regardless of the accentuation and the intonation of all the other forms of the paradigm (compare also the accentuation of the Indo-European vocative, which very often does not agree with the vowel-degrees of the suffix and of the root). Such details prove that the 'natural' intonation of the vocative is so rooted in actual pronunciation as to resist the analogical action of seems to me related forms. This explanation of the circumflex of ZED preferable to that given by Vendryes (47d). The Greek intonations seem thus to be a linguistic phenomenon properly Greek, not inherited, independent of apparently similar phenomena in Balto-Slavic. Their origin was purely phonetic, but, like the vowel-degrees of Indo-European ablaut, they were soon charged with a grammatical function in certain morphological categories.

Vous aimerez peut-être aussi