¹ I am again indebted to Professor Franklin Edgerton for valuable remarks.

² Cf. Tedesco, Dialektologie der westiranischen Turfantexte, Monde Oriental 15.184–258 (1921, appeared 1924), henceforth quoted as Dial. The reasons for the designation of the North Dialect as Median, not Parthian, I hope to set forth sometime in the future.

³ Bartholomae, Altiranisches Wörterbuch (henceforth quoted Wb.) 425 top.

⁴ The difference between OPers. kaščiy, čiščiy (with šč), from *kasčit, *čisčit, and pasā (with s), from *pasčā, as shown Dial. 210, is due to dialect mixture: s is the genuine Persian, and šč a Northwestern development.

⁵ Cf. Pali etc. imassa below, §14 end.

⁶ For the Armenian forms cf. Meillet, Esquisse¹ 61, ² 87.—Bonfante, JAOS 64.183 fn. 71, sees in Armen. o- not *k^u̯o-, but *i̯o-; erroneously, for *i̯o- nowhere becomes interrogative or indefinite.

⁷ Cf. Bartholomae, Grundriss §415; Wb. 422–7; Wackernagel-Debrunner 3.§258, especially γ, note.

⁸ Wb. 426 middle.

⁹ Wb. 424 middle.—Thrice čiš occurs as a neuter: V. 5.50, 18.36, and N. 72 (čiš zaotarš kairim aŋhaṯ 'what is the zautar's business?'), cf. Wb. 423 and 427, no. 2. This phenomenon, too frequent to be dismissed as a mere corruption, is the more strange as all of Middle Iranian (Persian, Median, Sogdian, and Saka) distinguishes 'who' and ‘what’. But considering the texts where it occurs, this neuter čiš cannot be something archaic, as Vaillant, BSL 37.104, believed. Rather, Middle Iranian ‘poets’, whose language had č° only in the neuter, used indiscriminately all forms with č°, čiṯ, čiš, and (see below) čīm, for ‘what’. The Late Avestan increase of či-forms for kaṯ is Middle Iranian. But čiṯ as opposed to kā in the old passage Y. 9.3 (see above) is probably archaic: the old *čit, anterior to kaṯ (§7), seems to have subsisted longer as adjective than as substantive.—An interrogative acc. masc. čīm occurs in Y. 19.20 and 20.4 (so), two worthless passages; as acc. neuter, Y. 62.8.

¹⁰ Wb. 423 bottom.

¹¹ Wb. 426 middle.

¹² Wb. 426 bottom, 427.

¹³ Göttinger Nachrichten, Philol.-hist. KL, Fachgruppe 3, Neue Folge 1.131 (1934).

¹⁴ Once interrogative kíḥ 10.52.3, Wackernagel 3.559.

¹⁵ Macdonell 303 fn. 7.

¹⁶ Grassmann 307 (1873): “kím (das ältere kád allmählich verdrängend)”.—Lanman, Noun-Inflection (JAOS 11) 377 (1880): “kím, which is gradually supplanting the more antique kád even in the Veda [is perhaps a phonetic variation of *ká-m]”.—Arnold, Vedic Metre 30 (1905), quotes kát as indicating early date; but kím, correctly, is not named among the forms which indicate late date (§86).

¹⁷ According to Wackernagel 3.562.

¹⁸ He compares the Rigvedic enclitic particle kīm, with long ī: mā́-kīm ‘ne’ 6.54.7 (mā́kir neśan mā́kīṁ riṣan mā́kīṁ sáṁ śāri kevaṭe 'none [of the cows] may be lost, none maybe harmed, none be crushed in a ravine'), 8.45.23, ná-kīm 'non' 8.78.4,5, a-kīm ‘from’ 1.14.9, to which corresponds Av. -čim (YAv. mā-čim, Gāth. naē-čīm, YAv. naēδa.čim). But I do not think this word is identical with kím ‘what’. Rather, if we can neglect the long ī (as does the comparison with kím 'what'), the close connection with mã-kih and ná-kih and the fact that, like these, -kīm does not go beyond the Rigveda, point to a stereotype acc. masc. *kim (= Av. čim, Gk. τιv-ά).—The other Av. čim, 'why', only twice in the late Aogamadaëčā, is perhaps the Middle Iranian form of the ablative *čahmāt, just like (according to Bartholomae, WZKM 30.34) MPers. čim ‘reason’.

Meillet's equation of Armen. inč ‘something’ with Skt. kíṁ cit (cf. Brugmann, Grundriss² 2.2.358, Wackernagel-Debrunner 3.562) is still attractive. It would presuppose that the phonetic development of k^u̯ before i was č (hence č in the particle), and that in the pronoun Pre-Armenian replaced *či- by *ki- after the analogy of *kos ‘who’, just as Sanskrit did. Z-i ‘what’ would then be from older *in. But even if this derivation were correct, it would not yet establish an ‘IE’ neuter *k^u̯im. Rather Pre-Armenian *kim would be a secondary transformation of *kid just like Skt. kím (see next paragraph).

¹⁹ Lithuanian-Latvian has kàs for ‘who’ and ‘what’.

²⁰ Some more recent views on the relation of the stems *k^u̯o- and *k^u̯i- are Leumann's, Lat. Gram. 288 (1928): “Idg. *q^u̯o- ist wohl eine Umgestaltung von *q^u̯i- nach *to-” (but contrast Gk. Horn. τo: τoο, Slav, cneso: togo, Goth, has: sa, ha: pata); Benveniste's, Studi baltici 3.128 (1933): *k^u̯o- originally accented and interrogative, *k^u̯i- enclitic and indefinite (this was Caland's theory, cf. Wackernagel 3.560 below), see here fn. 24; and (a strange idea) Vaillant's, BSL 37.103–4 (1936): *k^u̯i- from an oblique case.

²¹ Cf. Wackernagel 3.561 (who does not separate the genders).—Hittite, as against kwi-in the nom. and acc. (kwis, kwin, neut. kwit), has kwe- in the gen., dat., and abl. sg. (kwel, kwetani, kwez; Sturtevant §256 and §275), probably for animate as well as inanimate.

²² W. Geiger §111; Andersen, Pāli Reader Gloss. 75b; Diet. PTS sub ka and kin (read kiŋ).

²³ Pischel §428.—In Hindī, kis is obl. masc., while the neuter is kāhe.

²⁴ This Avestan (and Rigvedic) restriction of čiš to the indefinite, not mentioned by Benveniste, Studi baltici 3.128, would fit his idea that *k^u̯o- was originally interrogative and *k^u̯i- indefinite. But otherwise this theory seems not to be supported by the facts.

²⁵ W. Geiger §85, Pischel §377.—True, -i may possibly not only be the old form, but to some extent represent a secondary transfer to the masculines.

²⁶ A Rigvedic Prākritism in verb inflection was pointed out by me Lang. 20.215: the active inflection of the aorist subjunctives marāti, marāma, maranti RV 1.191.

²⁷ A ‘popular’ hymn according to the theory of Arnold, Vedic Metre 22.

²⁸ The substantive, except for the ī-stems, ū-stems, and panth-, is drawn up by Debrunner (cf. p. iii).

²⁹ Macdonell §342.

³⁰ Or, possibly, only *cakkhuḥ became cakkhu.

³¹ Except for the AV voc. fem. sahasra-cakṣo, which, as shown above, is of different kind.

³² Grassmann; Macdonell 294 bottom.—For the Atharvaveda, Macdonell l.c. quotes several cases where the meter requires *-u instead of the textual -uḥ; these are probably Prākritisms.

³³ Arnold 272, but not on p. 22.—Note isolated late parṣā for normal párṣi.

³⁴ But Sheth, Pāiasaddamahaṇṇavo, gives khavā- f. ‘rātri’ without reference, and khavājala- n. ‘avaśyāya; rime’ from Ṭhāṇaṅgasutta (AMāg.).—Bartholomae, Wb. 549, sets the Av. nom. sg. xšapa up as an ā-stem xšapā-. But I do not understand why this form may not be the nom. of the frequent an-stem xšapan- f. (cf Wb. sub xšapā- note 1).

³⁵ To RV imásya Wackernagel l.c. adds Ait.Ār. imásmai, MBh. imaiḥ, etc., and says: “Diese formen sind wohl meistens mittelindischen Ursprungs” (capitals mine). Apparently at least the RV form is not yet regarded by him as Middle Indic; and I should agree with this.

^36a I do not of course mean the language of the Rigveda hymns with kím, but the spoken language of the kím-poets, which no doubt had also other Middle Indic features.

³⁶ (*) denotes hymns parts of which Arnold designates as popular, but not the verses containing kím; the kím-passages are: 3.53.14; 6.47.10; 7.55.3. (On p. 280, however, Arnold notes the last verse as popular.)

³⁷ Geldner, Translation.

³⁸ Whitney-Lanman, Atharva-Veda Translation, for AV 18.1.4, 7, and 12.