¹ Meillet connects the peripheral position of Italic, Celtic, Hittite, and Tocharian with the probability that they broke away from the parent language before the others. It is obvious that changes in the latter would encounter an unsurmountable barrier after the separation of the former.

² We do not have to decide at this time the question whether the labial element was originally a distinct semi-vowel . See Sturtevant Lang. 6. 224. Whatever it was, Tocharian (see below) reveals evidence of its past existence which places it in this respect also with Hittite and the centum-languages.

³ That a similar innovation should have been performed independently by two languages in the extreme West and two others in the far East is in this instance not improbable because the roots of the innovation were present in the entire IE territory. After a third singular like Hit. es-ari was once used as an indicative beside its original volitive use, spread of the r-form to other indicative forms was easily accomplished.

⁴ In the comparison of Hittite forms it must be remembered that double consonants have no double etymological value, and that the signs for voiced and voiceless stops designate the same sound, apparently the voiceless.

⁵ I use the terms primary and secondary for convenience, although not accurate historically, for the IE perfect contributed its share to the latter in Hittite and Tocharian.

⁶ With the dropping of the unstable vowel ä in -mät- of sika-t-är cf. e.g. the numeral okät ‘eight’ beside okta-, spät ‘seven’ beside äpta-.

⁷ Quoted henceforth with the abbreviation SSS.

⁸ This formulation necessarily abstracts from later secondary changes which took place in each language separately; e.g., the change of Hitt. *-ti to -zi as in es-zi = Gr. στí, IE *es-ti, or the Tocharian palatilization of t as in mācar : Lat. māter, IE *mātēr.

⁹ So Hrozný, Spr. d. Heth. 70 f.

¹⁰ Sturtevant, JAOS 52.5.

¹¹ Friedrich, ZDMG 76. 159.

¹² SSS 6.

¹³ Not only the root, but also the suffix of praksam corresponds to Skt. pcchá-ti, for IE -so-, Skt. -(c)cha- became Toch. -sa-, which is -ska- in the B dialect; see SSS 358 f.

¹⁴ Hrozný, op. cit. 4.

¹⁵ Sturtevant Lang. 4. 3 f.

¹⁶ Toch. is pronounced n, and often is n etymologically.

¹⁷ SSS 5.

¹⁸ Sturtevant, Lang. 3. 109 ff., seemed at the time to have made out a good case for his earlier theory that initial IE bh- became Hitt. h-, but later (JAOS 50. 125 ff.) retracted in view of a number of convincing etymologies in which it appears as Hitt. p-, which identification has the greater compelling power because this development is exactly what would be expected since it is in line with that of medial bh and of every other series of explosives. Unless therefore, as is highly improbable, influence upon Hittite of some other remoter IE dialect, in which IE bh- became h-, could be established, it is necessary to reject such comparisons as handami ‘I set in order, prepare’ = Skt. bandhāmi ‘I fasten together, construct’, hanna- ‘litigate, decide a suit’: Skt. bhána-ti ‘speaks, tells’, and others not mentioned by Sturtevant, as haddulatar ‘health’ : Skt. bhadrá-h ‘fortunate’, Goth. batiza ‘better’. We must even reject the comparison of the root-syllable of humant- ‘all’ with Toch. pu-k ‘all’ and Skt. bhū'-ri- 'much'.

¹⁹ Sturtevant, Lang. 6. 216.

²⁰ Sturtevant, loc. cit.

²¹ Sturtevant, JAOS 51. 126.

²² Fraenkel, IF 50. 227.

²³ E. g. P. Poucha, Arch. Orient. 2. 322.

²⁴ Boisacq, Dict, etymol. s.v. πήχυs; Fraenkel IF 50. 7. The o of poke is not yet explained, for Toch. o normally represents an IE u-diphthong. However, with IE ō cf. Toch. ñom ‘name’ : Lat. nōmen Skt. nā'ma, and, with original short o, orkäm 'darkness' : Gr. òρφvós ‘dark’, IE root *org(h)-, cf. Hirt IF 12. 226; Scheftelowitz BB 28. 293, 29. 17.

²⁵ Friedrich, loc. cit.

²⁶ Sturtevant, Lang. 8. 130.

²⁷ Fraenkel, IF 50. 7.

²⁸ See Sturtevant, Hitt. Glossary 70.

²⁹ Sturtevant, Lang. 5. 10 f.

³⁰ Cf. SSS 169.

³¹ Mudge, Lang. 7. 253.

³² Sturtevant, Lang. 3. 166 f.

³³ Sturtevant, Lang. 6. 27 f.

³⁴ Change of IE d- to Hitt d/t is doubted by Sturtevant AJPh. 48. 249, who suggests Hitt. ta- from IE *do-.

³⁵ So, e.g., Sturtevant, Lang. 6. 214.

³⁶ Hrozný, op. cit. 61.

³⁷ Probably Hitt. a for e in the first syllable is due to Hittite vowel assimilation rather than inherited ablaut differences.

³⁸ With palatalization of initial t- to ś before ä in the numeral śäk ‘ten’ = Skt. dáça, Lat. decern, IE *de; so, e.g., Poucha 325. It is not clear, however, why the result of the palatalization of t is here ś instead of the regular c; cf. SSS 350.

³⁹ Friedrich, loc. cit.

⁴⁰ So, e.g., Poucha 325 f.

⁴¹ SSS 15.

⁴² Friedrich 167.

⁴³ The root knā- is of course the IE *ĝnō- of Skt. jñā-, Gr. γvω-; cf. Poucha 324.

⁴⁴ So, e.g., Zimmern, OLZ 25. 200 f.

⁴⁵ Cf., e.g., Sturtevant, Lang. 6. 225.

⁴⁶ Otherwise Poucha 322 f.

⁴⁷ SSS 7.

⁴⁸ Cf. SSS 2.

⁴⁹ SSS loc. cit.

⁵⁰ Otherwise (: Gr. ∈λγχω) Sturtevant, Lang. 6. 218.

⁵¹ Götze, IF 42. 327 f., who, however, needlessly assumes borrowing from the Skt. yugá-m, cf. Sturtevant, Lang. 6. 218.

⁵² Poucha 323.

⁵³ Poucha, loc. cit.

⁵⁴ Hrozný, op. cit. 23.

⁵⁵ Sturtevant, Lang. 6. 218.

⁵⁶ Mudge, loc. cit.

⁵⁷ Hrozný, op. cit. 164.

⁵⁸ This connection seems to me so obvious that I do not hesitate to draw the conclusion that various words with radical l, e.g., Lat. celāre ‘hide’, Gr. καλιά ‘hut’, which are cited by Uhlenbeck, Et. Wörterb. d. ai. Spr. 304, as cognate w^Tith Skt. çaraá-, must be separated from the latter in favor of the above derivation.

⁵⁹ Friedrich, IF 41. 369 f.

⁶⁰ See note 33.

⁶¹ See note 38.

⁶² Friedrich, IF 41. 372 ff.

⁶³ Poucha 324.

⁶⁴ Otherwise Sturtevant, Lang. 6. 226, who connects with Goth, qistjan ‘destroy’ and assumes a labio-velar; but the loss of the labialization in the Hitt. word as well as the Lith. gestù gèsti ‘be quenched’, which he also compares, is insufficiently motivated. In view of the latter, however, it is possible that the Hitt. and Toch. words had a velar instead of palatal.

⁶⁵ SSS 53.

⁶⁶ See note 43.

⁶⁷ Kurylowicz, Symbolae Grammaticae in honorem Joannis Rozwadowski 101; Poucha 324.

⁶⁸ So Friedrich ZDMG 76. 159, Fraenkel IF 50. 11. Although less probable, IE ĝh is also possible. See Sturtevant Lang. 6. 216, and cf. Skt. máhi ‘great’.

⁶⁹ Sturtevant, Lang. 6. 216.

⁷⁰ See note 20.

⁷¹ See note 24.

⁷² So already Hrozný, op. cit. 62.

⁷³ Sturtevant, Lang. 6. 221 f.

⁷⁴ Friedrich, loc. cit.

⁷⁵ Friedrich, loc. cit., also found IE g in Hitt. huis- ‘live’, with which he compared Lat. vīvus, Gr. βíos, etc. It is obvious that this etymology cannot stand, in view of the Hitt. representation of the labio-velars in the certain instances.

⁷⁶ Another instance of IE g is Toch. ko ‘cow’ = Skt. gāú-, Gr. βo, IE *gōu-s. Evidently therefore ko has a u-diphthong and is another instance in which Tocharian shows no trace of the labialization.

⁷⁷ See also note 24.

⁷⁸ With the change to u of the labial element of the labio-velar plus the following palatal vowel is to be compared the change of IE *te ‘thee’ to Toch. cu, classified as ‘oblique.‘ Here too the c for t is evidence of the former existence of a palatal vowel.

⁷⁹ The Tocharian shift of consonants agrees with the Hittite in effacing or tending to efface the difference between the orders, but is altogether different from the Armenian, which, like the Germanic, shifts all orders, so that the distinction is maintained. It is therefore wrong for e.g. Pokorny, Die Stellung des Tocharischen in Kreise der idg. Sprachen, or Poucha p. 326 to cite the Tocharian shift as evidence of close relation to the Armenian.

⁸⁰ For the treatment of Sanskrit loan words in Toeharian cf. SSS 55 ff., P. Toucha Arch. Oriental. 2. 300 ff.

⁸¹ SSS 96.

⁸² Fraenkel, IF 50. 105.

⁸³ Since the vowel of the first syllable was originally u, the first c for t was really due to assimilation to the second, cf. Poucha 325 f. Another instance of c from dh is ca-sä-s pret, of tā- ‘place’ : IE *dhē-.

⁸⁴ The of Skt. k finds no explanation through Kretschmer's solution. On the other hand when Sturtevant separated Hitt. tegan from Gr. χθών etc., and connected with Lith. deñgti ‘cover’ (Lang. 6. 226 f.), he lost sight of Toch. tkam, which is so clearly the connecting link between tegan and χθών that we must assume connection between them all. Sturtevant's objection that Hittite shows a sibilant for the IE sound designated by þ in taks- ‘build’ beside Skt. tákā, Gr. τκτων ‘carpenter’, can be taken care of by assuming that the metathesis in IE *kþōm took place before the þ had become a sibilant as in taks-, and that in the order þk the first became a t (d). The question whether the e of Hitt. tegan is a secondarily developed vowel we must leave open for the present.

⁸⁵ At its best the entire complicated group of phonetically discordant forms raises difficult questions whatever is our judgment regarding the question as to which languages were the innovators in each detail. According to Sturtevant, e.g. Lang. 7.118f., Hittite represents the original state of affairs both in its retention of an IH h and the relative order of the vowels. Tocharian, however, which often corroborates Plittite in its oldest features, shows no sign of anything corresponding to such an h here or elsewhere.

⁸⁶ Also the IE neut. pl. nom. acc. is common in Hittite, cf. AJPh 51. 253.

⁸⁷ For the singular this conclusion is not absolutely certain, for Sturtevant, Lang. 8. 1 ff., has associated together a number of facts which seem to point to the existence of a pre-Indo-European ablative singular, and Hrozný had identified the Hitt. instrumental sing, in -t with the IE ablative in -t or -d. Cf. Donum natal. Schrijnen 367f.

⁸⁸ The use of Hitt. -as also in the dative pl. must be later than in the genitive.

⁸⁹ The more strangely looking plural suffixes of Tocharian, e.g. -α, -ant, -ntu, are probably stem-suffixes to be compared with the German -er and -en as plural endings. Since in some instances the dropping of final syllables with short vowels had deprived certain words of the final stem suffix in the nom. acc. sing., and only of the ending in the plural, singular and plural differed mainly by the preservation of the old stem-suffix in the latter.

⁹⁰ The reason for the palatalization of the n is not clear, but cf. Toch. ñom ‘name’ = Skt. nma Lat. nōmen.

⁹¹ Hitt. amug was IE *eme with added ĝ: (either the particle *ĝe or after IE *eĝ nom. sing.). The a of amug for e came from the plural anz-as, and the u from tu-g. Cf. Lang. 6. 169.

⁹² Short vowels of original final syllables (except in monosyllables) are dropped in Tocharian.

⁹³ Owing to the wiping out of IE quantitative distinctions of vowels in Hittite (AJPh 51.258) the IE verbs in the thematic vowel were confused with the ā-stems, so that Hitt. handa-mi may be taken as corresponding sound for sound with the Skt. word in its formative part.

⁹⁴ In daski-mi and the like the first i for a IE ō is probably due to vowel assimilation, or else is analogical to the 3. sing., where the i represents the e-grade of the thematic vowel.

⁹⁵ So Friedrich, ZDMG 76. 167.

⁹⁶ For the Hittite see now Sturtevant, Lang. 8. 119ff.

⁹⁷ I still consider this explanation more probable than the new one suggested by Sturtevant, 129 ff., sc. that the Hittite s-forms of the third sing. were perfects from aorist-stems, on the one hand because he has neglected the evidence of Tocharian, which does not vary in its stem-form before -t of the 2. sing. and -s of the 3 person, on the other because I cannot see anything significant anyway in the encroachment (analogical of course) of the e in Hitt. forms like uppe-sta from uppa-hi, on which he bases his argument that the two persons differed in the origin of the s.

⁹⁸ The difference in the sphere of usage does not weigh heavily—the Tocharian s-forms form a part of every preterite, while those of Hittite are confined to preterites of the hi-conjugation. The latter must be closer to the original state of affairs. The IE present perfect must still have existed in the period of Tocharian-Hittite unity, and the s-forms were then part of a complete s-aorist formed to express past time for these perfect-presents. However, as the s-aorist disappeared otherwise, and the historical perfect became the normal past tense, there resulted a composite past paradigm made up of the three s-forms and of perfects with past meaning. Originally these s-forms were confined to pasts from present perfects, whence Hittite used them only in the hi-conjugation, but in Tocharian this composite paradigm became the only preterite, assisted by the greater clearness of the s-forms e.g. in the second sing., where -t (IE -tha) had become the only present ending, and where --t therefore made possible the expression of past time.

⁹⁹ Otherwise Sturtevant, e.g. Lang. 6. 154.

¹⁰⁰ See AJPh 53. 207f.

¹⁰¹ So Sturtevant, Lang. 7. 251.

¹⁰² The primary Skt. ending -makē must be due to secondary ablaut (shifting) based on the ablaut IE i Skt. i : IE ai Skt. ē.

¹⁰³ The longer endings -wastat and -wastati are of course due to secondary extension of the shorter -wasta.

¹⁰⁴ Cf. Brugmann, Gr.² 2. 3. 642.

¹⁰⁵ At least we can say this of the B dialect, and the dialect A can scarcely be earlier. Cf. Hirt, Idg. Gram. 1. 43.