¹ For my views about the laryngeals, cf. Lang. 16.81–7, 179–82, 273–84, and references. In addition to the symbols given above we shall have to use H asa cover symbol for ', '., or x, and Ạ for.' or x.

² Most recently by Kurylowicz, Études indo-européennes 46–72.

³ A. Meillet, MSL 10.276–8; cf. Hjalmar Frisk, Göteborgs Högskolas Årsskrift 42.2.42 f.

⁴ See Sommer, Hdb. d. lat. Laut- und Formenlehre² 173 f.; Leumann, Lat. Gramm. 138, and references.

⁵ Symbolae Grammaticae in Honorem Ioannis Rozwadowski 1.95–104, Cracoviae (1927).

⁶ That the analogical vocalism and accent are as early as Primitive Indo-European is indicated by the agreement of Sanskrit and Greek.

⁷ Probably Indo-Hittite had no feminine gender, but it had forms in -a'. and -'. that functioned as collectives and perhaps as abstracts, and some of these forms became Indo-European feminines.

⁸ The place name Πλάταια for expected *Παλταια must be due to πλατύs, πλατ∈ȋα, etc. Skt. pthivî ‘earth’ shows for expected ur and th for expected t by the analogy of pthús, pthv, etc.

⁹ But cf. Pollux 6.74: vθa δ‘ πλάττοντο oi ρτοι, πλάθανον ὸμοίΩs δι‘ οὐ πλάττοντο. The reconstruction with IH ple- rather than pḷ- is based upon the Greek accentuation and assumes later assimilation to the vocalism of πλατ ύ s, etc.

¹⁰ One may, of course, assume either phonologic a or analogic e in the penult.

¹¹ Gk. σθνοs ‘strength’ has ∈ for expected α by the analogy of other neuter s-stems, such as βλοs, yvos, δοs, μλοs, μνοs, νμοs, πλκοs, os, στγοs. Sommer, Griechische Lautstudien 65–8, supports the etymology in spite of the 0 beside the τ of στημι; his explanation of the aspirate is no longer needed. Boiling, AJP 21.315 f., connected the word with χΩ and Skt. sáhate ‘prevails’ on the basis of a variant root with a labio-velar, for which he found evidence in Vedic ásaghnos, sághat, etc. To one who thinks that the IE palatals were derived from the pure velars this evidence is inadequate.

¹² E. G. Kluge, KZ 26.88–92; Berneker, IF 9.363 f.; Pedersen, KZ 40.178; Persson, Beiträge 420 fn. 1; Brugmann, Grundr. 1².632 f., KVG 150–5, 158–61, 164–6, 168 f.

¹³ In Celtic the inherited voiceless aspirates were not distinguished from the voiceless non-aspirates and in Germanic only in certain positions. I do not include the labiovelars and the palatals; for the former there is no certain evidence; for the latter the evidence is scanty, and in the centum-languages their development is, of course, the same as that of the velars.

¹⁴ For Sanskrit and Avestan see most recently Kurylowicz, Études indo-européennes 1.46–55; Hjalmar Frisk, Göteborgs Högskolas Årsskrift 42.2.3–50. For Armenian see especially Meillet, BSL 36.109–20, Esquisse d'une grammaire comparée de l'armenien classique², 34–38. For Slavic see Meillet, Le slave commun² (revised by Vaillant) 22–6.

¹⁵ It is generally assumed that the verb originally meant ‘thrust away with the foot’, but this meaning does not occur either in Latin or in Sanskrit. It is citable for Lith. spiriù.

¹⁶ There is no sound evidence for Lat. b from dh after u (see H. Pedersen, Hittitisch und die anderen indo-europäischen Sprachen 50 f.), and neither is there for Lat. b from ih in that position.

¹⁷ So Sommer, Kritische Erläuterungen 64. Walde-Hofmann, 1.633, suggest that ā may have been shortened under the influence of OHG hamo ‘fish-net’.

¹⁸ For the Indo-Hittite predecessors of the Indo-European ā-stems, see Sturtevant, Lang. 14.239–47; Hittite evidence shows that the IH suffix -x- had factitive meaning and the suffix -;- collective meaning. It is certain that the suffix -x- had other values (Lang. 16.283) and -;- may have formed abstracts as well as collectives. For the present it seems that when other evidence is lacking ū-stem nouns in Indo-European should be assumed to have suffix -x- if they have factitive meaning and suffix -'- if they are collectives. I am also, in this paper, assuming that abstract ū-stems represent IH suffix -.'-. We may then assume either IH rotax 'a causing to run' (for the meaning of the root ret- [rather than reth-], see Walde-Pokorny 2.236) or IH rota', ‘a running’.

¹⁹ Perhaps IH rotẠos was, to start with, the genitive of rotaẠ ‘of running’ = ‘running easily’, and this, being reinterpreted as the nominative of an o-stem, gave rise to a complete o-stem declension. In an article to appear in TAPA 71, I have called attention to a number of apparent parallels in Hittite and the Indo-European languages. The historically attested meanings of Lat. rota and Skt. ráthas would suggest that the IH gen. rotẠos meant ‘of wheels, wheeled’; but the parallelism of οή ‘a flowing’ : óos ‘stream’ and a considerable number of similar pairs makes one incline rather to the abstract meaning given above. Lat. radius makes it probable that there was originally ablaut of the radical syllable in the noun IH rotaẠ (nom. *rótẠ : *rtáẠ?). Possibly the adjective in -iyos was not formed until a concrete meaning had been developed, and it may have arisen independently from IH rotaẠ ‘wheel’ and IH rotẠos ‘chariot’.

²⁰ Whitney, Sanskrit Grammar §1136 and c.

²¹ Göteborgs Högskolas Årsskrift 42.2.3–6 (1936).

²² As will appear presently this suffix is to be connected with the abstract-forming suffix -tā, and I assume that the latter is from IH -ta; see fn. 18. Hence I write -t;os rather than -txos or -t'os.

²³ With analogic restoration of the suffix -t;os > -thos. Horn, μαστὁs shows the IE t that would be regular in the related suffix IH -ta; > IE -tā.

²⁴ Études indo-européenes 1.48, but cf. Frisk, op.cit. 41.

^24a Kerns and Schwartz have suggested this connection on a different basis in a paper that has not yet been published.

²⁵ Walde s.v.; Walde-Pokorny 1.240 f.; Feist, Vergleichendes Wörterbuch der gotischen Sprache³ s.v. weipan, and references.

²⁶ For denominative verbs of this type, cf. Sturtevant, Lang. 14.239–44.

²⁷ Of course such a root as wyeth- may, without more ado, be assumed to have a by-form weith-. Note, however, that IH witHrós obeys Sievers' law!

²⁸ So Walde-Hofmann s.v. folium; Walde-Pokorny 2.176 f.

²⁹ Études indo-européennes 1.46.

³⁰ Kurylowicz, op.cit. 46 f., Prace filologiczne 17.90 f. IH ' rather than; or x is required by Gk. πατΩ, πάτησα, etc. < IH pté'-; it is less probable that we should connect άπάτη 'deceit' and άπατάΩ ‘deceive’ (cf. Schrader, KZ 30.466; Prellwitz² 44) and assume an α-colored laryngeal.

³¹ Grundr. der iranischen Philologie 1.1.8.

³² Göteborgs Högskolas Årsskrift 42.2.41.

³³ Cf. above 000 and fn. 19. The declension without ablaut assumed in Lang. 14.241 to account for the Indo-European case forms may well have been secondary.

³⁴ He writes -te₂ t₂.

³⁵ Cf. fn. 19 and fn. 33.

³⁶ BSL 36.110.

³⁷ Sommer, Hdb. d. lat. Laut- und Formenlehre² 122.

³⁸ Walde-Hofmann are inclined to accept the unsatisfactory connection with caudex, codex ‘stick, block of wood’.

³⁹ Hitt. spandh (ši-pa-an-ta-a-ḥi, ši-pa-an-da-, išpandahe beside 3d sg. ši-pa-an-ti) may have restored by analogy with da ‘I take’, te ‘I place’, etc. It is also possible to read spanda and assume IH spondxa.

⁴⁰ See Walde-Pokorny 2.541 f.

⁴¹ The latter does not occur before the second century b.c.; it may be an Indic loan word.

⁴² One must add that in this word there is some independent evidence for the laryngeal in the long vowel of Skt. kh 'well, khātás 'dug', etc. The last named form may represent IH κtós.

⁴³ Holger Pedersen, La cinquième déclinaison latine 48 fn. 1.

⁴⁴ Otherwise Kerns and Schwartz, JAOS 60.182–5.

Sievers' law (see Edgerton, Lang. 10.235–65) requires alternation of -la- and -la-(Edgerton writes -ḷla- and -la-) after long syllables and short syllables respectively. It is here assumed that the alternation is as old as Indo-Hittite, but that it continued to function actively until Primitive Indo-European at least, so that, if the quantity of the preceding syllable changed, -la- was substituted for -la- or vice versa. It is, however, assumed that IH -A-, like the resultant IE -Ә-, had the prosodic value of a short vowel.

There is ample proof for similar alternation of -l- and -ḷ- between non-syllabics (Gk. άμαλδὐνΩ ‘soften’ : Lat. mollis ‘soft’ < IE mldwis, Gk. καρδία : κραδίη ‘heart’ < IE brd- : d-, etc.); it is probable that here too was normal after a long syllable, as is assumed in the reconstructions in the next paragraph.