¹ De Saussure, Mémoire sur le système primitif des voyelles indo-européennes (1879). The Hittite evidence was pointed out by Kurylowicz, Symbolae Grammaticae in Honorem Ioannis Rozwadowski 95-104 (1927). For a fuller statement of his views, see his Études indoeuropéennes 1.27-76, 254 f. (1935), which he has supplemented in Prace filologiczne 17.90-6 (1937).

² See especially Lang. 14.269-74, 15.181 fn. 2. I take this opportunity to correct three errors in the latter passage, for which I am responsible rather than the author, since I read the proof. In line 2, for *e-, read: *'e-. In line 10, for *γọ, read: *. In line 14, for ḷ, ṛ, read: ḷ, .

I shall elsewhere undertake to prove what in this paper is assumed, namely, that the four laryngeals were present in Indo-Hittite, but were lost (leaving various traces) in the Pre-Indo-European period.

³ Couvreur, De Hettitische Ḫ, een bijdrage tot de studie van het Indo-Europeesche vocalisme 257, 261-4; Kurylowicz, Prace filologiczne 17.95 f.; Sturtevant, Lang. 14.109 fn. 20a, 15.148 fn. 11.

⁴ This line of reasoning was first presented by Couvreur; but his statement of the case can be improved.

⁵ The vocalism of the first syllable, and thereby the initial laryngeal, are determined by the Greek cognate; see Walde-Pokorny 1.16.

⁶ On seḫur and meḫur, see Lang. 12.182-7, 14.109 fn. 20a.

⁷ See Lang. 15.153.

⁸ JAOS 52.1-12, HG 74-84.

⁹ Leipziger Semitistische Studien 7.13-25.

¹⁰ Cf. Ungnad, Materialen zur altakkadischen Sprache = Mitteilungen der vorderasiatischen Gesellschaft 20.2 (1916); Thureau-Dangin, Le syllabaire accadien pp. IV f. (1927), Tablettes cappadociennes (deuxième série) 4-6 (1928). I am under obligations to my colleague Goetze for instruction in this matter.

¹¹ This explains the facts noted in HG 84, that zz is more than twice as frequent as z between vowels, and that scarcely any words always show z in that position; z stands for ts and is properly written double as a voiceless sound. Whether the enclitic -za, the only common word that always shows a single z between vowels was voiced [dz] is not clear; there are other peculiarities also in the writing of this word, and its etymology is unknown.

Similar reasoning suggests that we assume a voiced s in such words as kis-(ki-i-ša, ki-ša, ki-ša-ri, ki-ša-a-ri, ki-i-ša-ri, pret, ki-ša-at, etc.) ‘become’ and was- (wa-a-ši) ‘buy’. Similarly voiceless l, r, m, and n are indicated by ma-al-la-i 'he grinds', kat-te-ir-ra-az ‘below’, am-mu-uk ‘me’, an-na-aš ‘mother’, etc. This whole matter demands further investigation. Cf. below p. 84.

¹² Muršilis Sprachlähmung 28-32.

¹³ JAOS 58.200 f.

¹⁴ Études indoeuropéenes 1.54 f.

¹⁵ He writes .

¹⁶ Lang. 14.104-11.

¹⁷ Sturtevant, Lang. 14.77.

¹⁸ Cf. Ernout-Meillet s. v. ācer.

¹⁹ Lang. 14.269-71. There is a fuller treatment in an unpublished Princeton dissertation by W. M. Austin.

²⁰ IH ; is required if we connect Skt. ava ‘down’, Lat. au- ‘away’ and Hitt. awan. Otherwise we may as well assume' or x.

²¹ HG 150.

²² Cf. Sturtevant, AJP 59.96.

²³ ἄγνυμι, has often been connected with ῥήγνυμι on the basis of an Indo-European alternation of wr- with w- (cf. Lat. frango: Skt. bhanákti ‘break’). This unconvincing etymology is now further weakened by the short augment in . I am inclined to suggest a connection with Hitt. ḫwek-, ḫuk- (ḫu-e-ik-zi, ḫu-ik-zi, pl. ḫu-u-kán-zi, ḫu-kán-zi, pret. 1 pl. ḫu-u-ga-u-en) ‘set free from witchcraft'. That this rather than 'bewitch, charm’ is the primary meaning of the verb is indicated by its constant use of magic favorable to the person upon or for whom the rite is performed, and still more strikingly by its occasional opposition to ḫamenk- ‘bind’ in the same text (see instances in Friedrich, ZA NF 4.187 fn. 3). The most important objection to the new etymology is that the durative is regularly written with -kk- (e.g. ḫu-uk-ki-iš-ki-iz-zi). To be sure there are other cases of double mute in the durative beside single mute in the primary verb; but at least in the case of ak-ku-uš-ki-iz-zi: e-ku-zi ‘drinks’ the double writing is the one that is phonologically regular. At any rate my former connection (HG 96) of Hitt. ḫwek- with IE wek^w- must be given up on account of (1) the meaning of the Hittite verb, (2) the lack of u after k in Hittite, and (3) the short augment in Sanskrit and Greek.

²⁴ Origines de la formation des noms en indo-européen 1.8, 26.

²⁵ Quaestiones Epicae 165.

²⁶ Lang. 15.153 fn. 29.

²⁷ F. Bechtel, Hauptprobleme 252; Wackernagel, Altindische Grammatik 1.5.

²⁸ Cf. H. Pedersen, KZ 36.76 f.