788 PHILOLOGY [ARYAN Sanskrit mldha, Zend: mlzhda, Greek (i.ia06$, Slavonic mlzda, Gothic mizdo. 1 Up to a very recent date the Aryan vowel-system was considered not to have contained more than the three " primitive " vowels a, i, u, and the diphthongs at and a (regardless of quantity). Tlie sounds of c and o, which are frequent in the European languages (and also in Armenian, as has been pointed out before), but do not occur in Sanskrit,- were regarded as later developments from the original a. We know now that these views were erroneous. Aryan not only had the five common vocalic sounds a, c, o, i, u, both long and short, but also often used the liquids and nasals r, 1, in, n, i), as vowels, that is, with syllabic value (as, for instance, in English battle, bottom, mutton, pronounced bat-tl, bot-tm, mut-tn), also both short and long. Besides these simple vocalic sounds, there were twelve diphthongs proper, ai, ci, oi, au, cu, ou, and di, ei, oi, du, eu, on, setting aside the similar combinations of , c, o, &c., with liquids and nasals. It will be observed at a glance that the Greek vowels and diphthongs V uv Primitive a, c, o. are exactly those of the Aryan system. The only case, indeed, where Greek has changed the Aryan sounds is that of the syllabic liquids and nasals, as will be shown hereafter. The first proofs for the priority of the European a, c, o in com parison v th the uniform Indo-Iraniaii a were discovered independ ently by Amelung and Brugmann. 8 Since then the number of proofs has been considerably increased. The most striking of all is perhaps the observation, made independently by Yerner and Collitz, 4 that the original back gutturals of Aryan are changed into palatals in Indo-Iranian when followed by i, y, or an a corre sponding to a European c, but are preserved without alteration when followed by other sounds, especially an a corresponding to a Euro pean a or o. We thus find not only forms like Sanskrit cid corre sponding to Greek TL, Latin quid, but also Sanskrit ca, panca,janas, &c. , corresponding to Greek re, TrevT*, yfros, Latin quc, quinque, genus, while the old guttural is kept in words like Sanskrit katara, (jarbha = Greek worepos (Ionian /corepos), Slavonic kotoryj, Gothic hwathnr, and German kalb. A special instance of this Indo-Iranian law of palatalization is exhibited in the formation of the redupli cative perfect, where initial gutturals are changed into palatals before the vowel of the reduplicative syllable, which is c in Greek arid elsewhere ; compare Sanskrit perfects like cakara, jagrdbfta with Greek rerporpa., toura., &c. If, then, the Indo-Iranian a( = European c) once had the same influence on preceding gutturals as the palatal vowel and semi-vowel i and y, it must necessarily itself have had a similar palatal, that is e-like, pronunciation distinguishing it from the other a s that go along with the non-palatal European a and o. The proofs for the coexistence of a and o in primitive Aryan are no less convincing than those for the existence of the palatal " a-vowel, " that is e, but they are too complicated to be discussed here. 5 The Aryan syllabic liquids and nasals were also discovered by Brugmann. In Sanskrit the short syllabic liquids are preserved in the so-called r-vowel and Z-vowel, as in krtd, klptd ; the long ones have passed over into Ir or ur, as in stlnid, purnA, and gurti. These Sanskrit vocalic r and I are the only direct remnants of the whole class. In all other cases the original system has been more or less destroyed. Thus, to give only a few instances, the syllabic nasals appear as a in Sanskrit and Greek, as in Sanskrit tatd, Greek raros for Into- (past part, of Jten, in Sanskrit tanomi, Greek reLvu for *Tfi>jw), Sanskrit c.atd, Greek t-Karfo, "hundred" (fork } nto-m); or as an before vowels, as in Sanskrit tunil, "thin," Greek ravus, for dissyllabic tn-fr. In Latin and Celtic an c has been developed before 1 See Osthotf, in Zeitschr. /. vergl. Spmcltf., xxiii. p. 87, and Kluge, ibid., xxv. p. 313. 2 It must be borne in mind that the Sanskrit sounds generally transcribed by e and o were originally diphthongs, = ai and au. 3 See A. Amelung, Die liildung der Tempusstamme durch Vocalsteigerung,~Ber m, 1871, also in Zeitschr.f. vergl. Sprachf., xxii. p. 300, and Zcitach.f. deiitsche* Alter- thum, xviii. p. 161 s</.; and Brugmann, in Curtius s Studien, ix. pp. 287, 363. In his earlier publications Brugmann wrote a, a-,, a 3 for e, o, a respectively ; A was then substituted for a s by De Saussure ; others, again, introduced u e and for Brugmaim s a l and a- 2 , and simple a for his a-j. The spelling e, o, a, now gener ally adopted, was first proposed by Collitz.
- H. Collitz, in Bezzenberger s Jieitrdge, iii. p. 177 s i. ; Vemer s discovery
was communicated by Osthoft , in Morpholovitche Untersitchungen, i. p. 116, and by Hubschmann, in Xeit-ichr. f. venjl. SfTachf., xxiv. p. 409. See also the full discussion of this problem by Joh. Schmidt, ibvl., xxv. p. 1 si/. 5 Besides the references given above, compare for this and the following especi ally F. Klugr-, Bdtr. zur Geschichte der germun. Conjugation, Strasburg, 1879 ; F. Masing, Hat Verhaltniss der griech. Voculabstiifitng zur SantkrltitCMn, St Petersburg, 1879 ; F. de Sau.ssure, Meinoire sur le sj/sieme jtrimitif des voi/elles duns hf languea indo-eurnpeennes, Leipsic, 1879 ; G. Mahlow, Die lange.n Vocale il, e, u in den europ. Sprachtn, Berlin, 1879 ; OsthofT and Brugmann, Morpho- loglsche Untersuchungen aufdem Gebiete der indng. Sprachen, 4 vols., Leipsic, 1878, f/. ; G. Meyer, Griechische Grammatik, Leipsic, 1880 ; and a long series of articles by K. Verner, Brngmann, Meyer, Osthoff, Joh. Schmidt, in Zeitschr. f. vergl. Xprachf., vol. xxiii. sr/., and by BezzenberKer, Collitz, and Fick, in Bezzen- berger a lifitriigf., vol. ii. sr]. ; also Fick in Gcittinger gekhrte Anzeigen, 1880, i. p. 417, and 1881. ii. p. 1418; Paul, in Paul and Brauiie, Beit-rage zur Geschidtte der dcutschen Soroche und Littratiir, vi. p. 103 ; H. Moller, ibid , vii. p. 482. the nasal, Latin centum, tcnu-is, Irish cet (for *cent in Germanic a u, Gothic Jiund, Old High German dunni. Original syllabic r and I are in the same way represented by Greek pa (ap) and Xa (aX), as in (5pa.Kov, ppaSvs (for *npa6fa), TrXari^y ( = Sanskrit ddrytm, inrdii, prthu), and in Germanic by ur, ul (more seldom ru, In), as in Gothic thaurttu-s, "dry" (foi*thursus), u-ulfn, "wolf" ( Sanskrit trshii, vrku], and so forth. The most brilliant result, however, of these recent researches was not the more exact fixing of the phonetic values of the single Aryan vowels, and of the rules of correspondence between these and the vowels of the individual languages, but the discovery that the system of etymological vowel-change which pervades the whole of Aryan word-formaaon and inflexion, and which had until then generally borne the name of vowel -gradation, was chiefly deve loped under the influence of stress and pitch. It is well known how the theory by which the old Sanskrit grammarians tried to explain vowel-differences in words or forms derived from the same " root " considered the shortest form of a root-syllable discernible Shortest in all its derivations as the most primitive shape of the root, and form let the fuller forms be developed from it through a process of of root. increase, which Sanskrit grammar is accustomed to call guim and vrdd/ii. Taking, for instance, the inflexions of perfects like rt da, vettha, veda (originally pronounced vaida, &c. ), plur. vidmd, ridd, vidus, or cakara, cakdrt/ta, cakara, plur. caknnd, cakrd, cakrtis, past part, krtd, they would start from vid and kr as "roots," and say that red- (raid-) and kar- in veda (vaida] and cakara, &e., were derived from these through guna, that is, through the insertion of an a before the original root-vowels i and r. This doctrine has been adopted by Bopp, and thus become one of the fundamental theories of comparative philology, although the objections that can be raised against it are both numerous and obvious. Even if we pass over the difficulty of giving a satisfactory phonetic explana tion of the assumed process of insertion, how are we to account for the fact that in cases like ydjdmi, past part, ishthd, or perfects like jagrdbha, plur. jagrbltmd, the "inserted" a stands after the " root-vowel " instead of before it ? Or, if we look at forms like paptimd, perf. plur. of put ami, "I fly, " or studs, sthd, srititi, plur. of dsmi, "I am," must we not tsiku pt and s as the original roots, and is it possible to imagine that such roots could ever have existed ? All such difficulties disappear by assuming the new theory, that the fuller forms are more original. As the above instances .show, the fuller forms appear wherever the "root-syllable " is accentuated, Functio that is, stressed ; the shorter ones are confined to stressless syllables, of stresi What, then, more natural than to assume that the a of the fuller forms was the original "root-vowel," and that it was dropped in the shorter forms on account of their being unaccentuated ? Loss of stressless vowels is one of the most frequent phonetic phenomena in all languages, and we have only to look to modern English pro nunciation to find the most striking analogies to the processes assumed above. Every -day pronunciations like p tdio, S pttmbcr for the written potato, September are exact parallels to the Sanskrit 2ia-p timd, and the common ml(ldrd), ml(hidy) against the usual full my to the Sanskrit vidmd against vdida ; even the r-vowel is quite well known in rapid speech in forms like 1 prpfac, or hintry, iiatshral for the written propose, history, natural. -v So far the new theory of vowel-gradation may be summed up as follows. Every root-syllable originally contained one of the three primitive vowels a, e, o, either short or long ; i, u, the liquids and nasals, only occurred as semi -vowels or consonants, that is, form ing monosyllabic (diphthongic) combinations with these vowels, which may either precede or follow the consonants. Thus, taking the combinations with the short vowels as an instance, we get the following table ai ci oi and ya ye yo au eu ou .. ica we wo &c. In originally stressless syllables long vowels were shortened and short vowels dropped. If the original short vowel were sur rounded by mutes, the mutes would come into contact through the loss of the vowel, as in Sanskrit pap-timd from *papatimd, or Greek (ir-T6fjLT)i> from irdro/jLai, or Zff-xw from ?x u (f r * ff *X w )- If. however, the root- vowel were combined with a semi-vowel (/, u, or?/, ?/), liquid or nasal, the latter would, on account of their vowel-like character, become syllabic (that is, vocalic) if followed by another consonant, but remain consonants if followed by a vowel ; compare the follow ing instances taken from Sanskrit (for the sake of distinctness we write the original ai, au for the common f, o). (vdida vidus f tutduda tutudus (daddrm dadrct -I ydjdmi ish hd vdktum uktd I jagrdbha jagrbhils jigaya jigyiis cakara cakrus ( tatdna tatd (for tntd, see above)
,, tatnirt
In the same way we find in Greek olda I5af, (pfvyu <pvy(.lv, depKo^ai HSpaKov (for *?5pKov = Sanskrit &drc,am ; see above), rptiru trpairov (for *trpirov}, &c., and correspondingly in the other languages.
