^* Research for this paper and other publications to appear in the near future was carried out with support from the American Council of Learned Societies and the University of Texas.

¹ The letter B before a number refers to Tocharian B texts published from the Berlin collections; an A to A texts from the same source; H identifies texts of the Hoernle collection; F and L are the labels for Paris texts edited by Filliozat and Lévi respectively. Local provenance is indicated in parentheses or, in text within parentheses, in brackets.

² The formulation as offered here is a restatement of what is found in KT 76–7; references to alleged events in the mind of a speaker of Tocharian have of course been omitted.

³ An asterisk preceding a Tocharian form indicates that this form has not been found attested in a published text.

⁴ The last two dual forms are cited from KT 77.

⁵ Further Indic loanwords to be added here are *ahār, du. (a)hārāñc (B 177 [M] b 2–3), *bhavāṅk, du. bhavāṅkäñcä (B 149 [MQ] b 3), and compounds of bhāw in B 175 (M) b 2.

⁶ In KT 119, piltā seems to have been taken as a plural form; this eliminates the need for emendation in the modifying adjective, but leads (implicitly, to be sure) to the postulation of a declensional pattern without a parallel (sg. -a : pl. -ā). The new approach as found in KT can therefore not be considered a significant improvement.

⁷ The accusative of eṅkwe is eṅkweṃ, (FM 1 b 6) ; the accusative marker of personal masculines is not part of the stem.

⁸ The forms given here (from B 550) show characteristics of the Eastern dialect of B (cf. JAOS 75.222–4 [1955]) ; their ‘standard’ (Central) equivalents would be kauṃñãkte, kauṃñäkti, kauṃñäktene.

⁹ Cf. KT 77.

¹⁰ I am assuming here the use of a euphemism. This phenomenon can be illustrated for B by at least two examples : kāswo denotes a skin disease in FS b 5 (cf. B 282 [MQ] a 4) ; it seems inadvisable to separate this word from A kāsu ‘good thing’. A second case is closer to asāñcne: ‘penis’ is rendered in B 334 (MQ) a 4, etc., by kwipe ike ‘shame place’; the more direct term seems to be tso in the Stein fragment Ch. 00316a². (Cf. W. Couvreur's discussion of this text in KZ 72.222–4 [1955], which fails to recognize the real nature of the passage.)

¹¹ The plural eśaiṃ (in a bahuvrīhi, trai-eśaiṃś) is perhaps to be explained as a recent formation after the analogy of what would be regular *klautsaiṃ ‘ears’.

¹² Cf. KT 77.

¹³ Uncertainty exists concerning the dual of piśpik ‘nipple’ (Sieg) or ‘breast’ (Filliozat). Krause lists (KN 3), after Sieg, a ‘paral’ form piśpikne which would lead to an alignment with [a]; but FY 1 a 1–2 has according to the photograph merely klaiñe teki piśpi—[n.]sa tsqṅkau tākaṃ ‘wenn eine Frauenkrankheit an den Brustwarzen entstanden ist’ (Sieg, KZ 72.64 [1955]). That is, both piśpikanesa and piśpiśanesa would be possible conjectures, of which the second seems to have a higher degree of probability.

¹⁴ The connection of yneś with ek is made by S. Lévi, Fragments de textes koutchéens 160 (Paris, 1933), though -eś is not analyzed any further. The same applies to A. J. Van Windekens, Lexique étymologique des dialectes tokhariens 170 (Louvain, 1941). Sieg, too, in his Glossar does not specify the formal relationship of -eś and ek, although H. Pedersen, Tocharisch vom Gesichtspunkt der indoeuropäischen Sprachvergleichung 74 (København, 1941), had already made the identification of -eś as a dual form. More recently, the same explanation of -eś has been offered by F. Bernhard in his mimeographed dissertation, Die Nominalkomposition im Tocharischen 26 (Göttingen, 1958), and, interestingly enough, also in KT 55, without, however, introducing any adjustments in the basic presentation of the paral problem on pp. 76 ff.

¹⁵ Bernhard (27) takes -ne to be the locative marker; this is a genuine possibility.

¹⁶ Note B 224 (MQR) b 1–2 eśän(e tä)ṅwäññane.

¹⁷ A further example of this construction is B 384 (S) a 6, where I would prefer to read (bo)dhisātwentso sa(rr)i(w)enta ek(ša)lye, not er(ša)lye.

¹⁸ Note the vacillation in the endings of the modifying adjectives (-ana : -e).

¹⁹ Cf. Bernhard 21.

²⁰ Sieg's reading ai(w)e in B 49 (Š) a 8 is conditioned by his interpretation of ostuw' aiwenta as ‘houses [and] dwelling places’; it need no longer concern us (cf. also KN 6 fn. 1).

²¹ For these problems cf. A. von Gabain and W. Winter, Türkische Turfantexte IX 29–34 (Berlin, 1958) and Akten des Vierundzwanzigsten Internationalen Orientalisten-Kongresses München 520–1 (Wiesbaden, 1959).

^21a Bernhard (124) considers pauke-paine a compound; but note the caesura!

²² Instead of ‘he’ etc., ‘she’ etc. is of course possible.

²³ See as an example the picture of the cloth shown to Ajātaśatru as it appears on a wall painting found at Qïzïl (published most recently by H. Härtel in Indische und zentralasiatische Wandmalerei, plate 26 [Berlin, 1959]; cf. also p. 18).

²⁴ Umlaut by a lost *u seems to be limited to initial or secondarily initial position (cf. or ‘wood’ < *doru, ost ‘house’ < *wostu), as indicated by a form like mekwa ‘nails’ < *nogh-(?). It is therefore unlikely that a form *koṃ should have developed from *ǵonu.

²⁵ Examples of secondary -ñi- occur in B 107–116 (S), 324–325 (M), 326–332 (S), 369 (M), 433–489 (MQ IX), 493 (MQ IX), 591 (S), 592 (M).

²⁶ tanākko is literally ‘small grain’ (cf. FY 1 a 2, b 3). ‘Mark’ renders Sieg's ‘Mal’; as an alternative, ‘speck’ or even ‘pimple’ might be considered.

²⁷ Cf. Bernhard 135–6.

²⁸ From here on abbreviated as WTG.

²⁹ Cf. WTG 147 ff. The exact meaning of päl- or päl(t)s- cannot be safely extracted from the combination with ‘ears’; ‘prick’, ‘point’, or the like is just a guess.

³⁰ For a historical statement cf. §5.

³¹ The subject seems to be ñäkt(eṃts ñähte) ‘Indra’.

³² -a- in klausapilši is taken to represent stressed /ǝ/ because of the alternation of -a-and -ø- in klautsane/klautsne and because of the frequent occurrence of /ǝ́/ at the transition point in B compounds. For a critique of the different interpretation offered by Bernhard (46) and Krause (GGN 1961.88), see my forthcoming communication in JAOS.

³³ No distinction is made at this point between free forms and forms extracted from more complex formations.

³⁴ Cf. the references given in footnote 21.

³⁵ E stands as a symbol for neutral (‘e-coloring‘) laryngeal. For agreement in the reflexes of *y plus laryngeal in Tocharian and Greek cf. Evidence f or laryngeals 174.

³⁶ A palatalizing component is indicated by the sign '.

³⁷ In the case of keni, an adjective rather than the numeral ‘two’ must be considered the source of the spread of -i.

³⁸ Cf. E. Schwyzer, Griechische Grammatik 1.565 (München, 1939).

³⁹ For the problem of pai ‘feet’ cf. IF 66.320–3 (1961).

⁴⁰ Origin and nature of the -ai extension in the word for ‘arm’ are not clear.

⁴¹ While the same applies to B kuke(ne) ‘heels’, A kukäṃ belongs in one class with A esäṃ, etc.; cf. the remarks on B antsane/āṃtsne below.

⁴² šäptuk 'seventy' and nmuk 'ninety' are taken to show analogical spread of -uk.

⁴³ Like Latin, Tocharian has a contrast between nominative-singular stem and the stem of all other forms of a noun with an ablauting paradigm.

⁴⁴ It is doubtful that āmpe should be a masculine genitive; surely, if pissaṅkäṃtwe stands for *pissaṅkäntwis, it must be an alternant of the feminine form āmpine. One explanation of the unexpected form āmpine is that it may have been formed after the analogy of the very common adjectival relation nominative masculine -ši : genitive feminine -šine; another explanation is referred to in §9.2.

⁴⁵ As B kene is masculine in 514 (MQR) b 9, it is unlikely that A kaṃ should be feminine; the translation of W. Thomas, Die Gebrauchsweisen der Vergangenheitstempora im Tocharischen 56 (Wiesbaden, 1957), is therefore hardly correct, as is also borne out by the locative ending of the pronoun tāmaṃ. (The alternative suggested loc.cit. fn. 3 presents a genuine possibility.) The translation proposed here would help to narrow down the area of application of A kaṃ, B kene, as kaṃ in A 301 seems to be different from A waśeṃ (= B weśeñña) ‘voice’. Is kaṃ the tune played by an orchestra? Cf. JAOS 75.33 (1955).

⁴⁶ The B form could also be derived from *-oysi(n) ; cf. IF 66.311, 321 (1961).

⁴⁷ For a different analysis of the forms in question cf. Pedersen 66–7.

⁴⁸ A detailed discussion of the accent in B will be given elsewhere.

⁴⁹ Krause's explanation of ‘paralic’ wene, yene as against ‘dualic’ tai is rather curious. He writes about this contrast: 'Das ist psychologisch verständlich: Eine Zweizahl beim Personalpronomen der 1. und 2. Person bezieht sich eher und häufiger [N.B.] auf ein natürliches Paar als beim Demonstrativum.' This is a claim in terms of statistics, and one would want some data to substantiate it. But even so one wonders how the fact that almost all nominal 'parals' refer to nonpersons is to be reconciled with Krause's assumptions: a parallelism between noun and pronoun is, after all, to be expected.

^49a Bernhard (43) adduces from the Paris text PK NS 76 a 3 the passage wi aletsi ostašmeñcane ‘two strange householders’, which adds one more example each for [a] and [c].

⁵⁰ KN 9 lists a nominative tuciye instead; the correct form tute (FY 3 a 2) is given KT 152.

^50a An outright exception is ostašmeñcane, referred to in fn. 49a.

⁵¹ The relation kartsi : kreñc = old neuter : old nonneuter form would be the same as that found in lalaṃški : lalaṃškane.

⁵² This need not have occurred in the actual contexts preserved, so that Krause's first point (KN 10 fn. 2) becomes unimportant; his second point is eliminated by the different analysis of tiṃ proposed in §9.2.

⁵³ As opposed to conditioned partial elimination in plural forms of adjectives (cf. KN 10).

⁵⁴ In old consonant stems, there was of course no morphological contrast between masculine and feminine nouns, at least not in case forms surviving in Tocharian.

⁵⁵ For this formulation, Greek evidence was considered of paramount importance.

⁵⁶ It seems less likely that A we should reflect an old feminine plural form corresponding to Latin duae. For the development *dw- > Toch. w-, cf. IF 66.319 (1961).

⁵⁷ Note that in the forms for ‘both’ A -i corresponds to B -i. This seems to increase the likelihood that analogical interference is to be assumed in such forms as A kanweṃ : B keni, etc.; cf. §6, §7.1.

⁵⁸ A small residue of doubtful forms remains. Is it possible that a dual allaik to allek ‘other’ is preserved in the broken passages B 61 (Š) b 2, 166 (MQ) b 1, or even 532 (D) a 4? For the last form, note the dual in b 4. Is kontsaisa after āmapi (in B 325 (M) a 1) a -ne-less dual form, or is a dual missing between āmapi and kontsaisa because of the following wat 'or'? Note however that all these suggestions depend on the correctness of Sieg's assumption that āmapi is a scribal error for āntpi 'both'.

⁵⁹ The term ‘established pair’ (in lieu of the commonly used ‘natural pair‘) was introduced in the introductory section of this paper; it represents an attempt to give preference to the thései rather than phúsei nature of linguistic signs. For the importance of the notion of an established number cf. KZ 71.1–14 (1954).

⁶⁰ This is the reverse of Schwyzer-Debrunner's view.

⁶¹ The following discussion describes the phenomena as found in Tocharian B, since the loss of most final vowels in A led to a situation which could be analyzed only in a very much more involved manner.

⁶² I am using here the actually attested B form; for the time under discussion it would be better to posit *šǝrǝ.

⁶³ In this frame, other expressions implying the number ‘two’, above all ‘both’ or the dual form of the demonstrative, can be substituted for wi.