^{page 44 note 1} Some abbreviations: IA = Indo-Aryan; O = Old; M = Middle; N = New; IE = Indo-European; II = Indo-Iranian; Skt = Sanskrit; Pkt = Prakrit; AMg = Ardha-Māgadhī; Pischel = R. Pischel, Grammatik der Prākrit-sprachen (cited by paragraphs); Geiger = W. Geiger, Pāli Literatur und Sprache (cited by paragraphs); Edgerton = P. Edgerton, Buddhist Hybrid Skt Grammar (cited by paragraphs); Burrow, Skt Lang. = T. Burrow, Skt Language; Burrow, Khar. Lang. = T. Burrow, Language of the Kharoṣṭhī Documents (cited by paragraphs); Whitney, Gram. = W. D. Whitney, Skt Grammar (cited by paragraphs); Whitney, Roots = W. D. Whitney, Roots and Verb-forms of the Skt Language; Berger = H. Berger, Zwei Probleme der mittelindischen Lautlehre.

^{page 44 note 2} “ig yaṇaḥ samprasāraṇam”: “ik (= i, u, r, ḷ) which replaces yaṇ (= ya, va, ra, la) is called samprasāraṇa”. The Kāśikā explains that the name is given by some to both the sound and the change.

^{page 44 note 3} Burrow, , Skt Lang., p. 108Google Scholar.

^{page 44 note 4} Pischel, 151.

^{page 45 note 1} Burrow, , Skt. Lang., p. 46Google Scholar.

^{page 45 note 2} See Alsdorf, L., “A specimen of Archaic Jaina Māhārāṣṭrī,” BSOS., viii, 325Google Scholar.

^{page 45 note 3} Aśokan -isiti may represent -isiti, -īsiti, -issiti.

^{page 45 note 4} Geiger, 10.

^{page 45 note 5} To postulate *dvatiya- from which dutiya- is derived by samprasāraṇa (Berger, p. 61) seems unnecessary. Such a form need never have occurred and to invent it seems to be taking a very mechanical view of samprasāraṇa.

^{page 45 note 6} Three forms exist in Skt: dva, du, and dv-i with -i- suffix. The use of the weak grade du- instead of dvi- is parallel to the use of tṛ- in tṛtīya- instead of the tri- seen in Greek τρίτος. There seems to be no example of normal grade tra- in IA, unless AMg tacca- is < *tratya- not *tṛtya-. In the latter -ṛ- might be expected to develop > -i- in the proximity of the palatal -cc-.

^{page 45 note 7} Whitney, , Gram., 233aGoogle Scholar.

^{page 45 note 8} For such variants in BHSkt see Edgerton 34.14. For variations between IA and Iranian see Bailey, H. W., TPS., 1953, p. 24Google Scholar.

^{page 46 note 1} Patañjali on Pāṇini, vi, 1, 108.

^{page 46 note 2} Geiger 25, though some of his examples are due to contraction, not apophony (see below).

^{page 46 note 3} Burrow, , Skt Lang., p. 85Google Scholar.

^{page 46 note 4} Berger (p. 78) states that thīna- is not a samprasāraṇa form, but represents a weak-grade formation. Since samprasāraṇa has been shown to be merely the Skt grammatical term for weak grade, his distinction is difficult to understand.

^{page 46 note 5} Whitney, , Gram., 993cGoogle Scholar.

^{page 46 note 6} So Pischel, 152, though in 584 he seems to derive -tūna < -tuāna < *-tuvāna < -tuāna. He is followed in this by Schwarzschild, L. A. (JAOS., 76, p. 113)Google Scholar who calls this phonetic development “samprasāraṇa” because of a wrong notion of what samprasāraṇa really is (see below).

^{page 46 note 7} The weak grade and absence of final -ṃ in śrutu show that this must be an absolutive, not the infinitive form *śrotu.

^{page 46 note 8} For the change -īk, -ūk > -iṃ, -uṃ, cf. Pāli manaṃ < manāk.

^{page 46 note 9} When the weak grade sadhrīc- already exists in Skt it seems unnecessary to postulate a MIA innovation (showing samprasāraṇa) *sadhrik < sadhryak. See Lüders, , Philologica Indica, p. 494, n. 1Google Scholar.

^{page 46 note 10} See Bailey, H. W., TPS., 1953, p. 24Google Scholar, and Edgerton, 34.13.

^{page 46 note 11} Showing the same grade as in brahma-. In Pāli normally bṛ- > b(r)u-, cf. abbuyha < ābṛhya.

^{page 47 note 1} For the alternation his-/hes- see Burrow, T., JRAS., 1956, p. 200Google Scholar. The attempt made in Geiger 10 to derive hes- < hiṃs- by phonetic changes must be abandoned.

^{page 47 note 2} Kuhn, E., Beiträge zur Pali-grammatik, p. 54Google Scholar.

^{page 47 note 3} Pischel 152.

^{page 47 note 4} For the change -a- > -u- after a labial, cf. kammuṇā (Pischel 104).

^{page 47 note 5} Geiger 19, n. 2.

^{page 47 note 6} Berger, p. 32.

^{page 47 note 7} This may not be < śvapāka- at all, but < *śavapāka- “corpse-cooker”. Pāli has sapāka- also.

^{page 47 note 8} It may on the other hand be an early example of the secondary development found later in NIA (see Grierson, On the modern IA vernaculars, paras. 239, 254) whereby -e-, -o- are frequently interchangeable with -ya-, -va-. Other examples are found in the Niya Pkt (Burrow, , Khar. Lang., 7Google Scholar). Such an alternation may have an historic origin, and may be originally a bye-form of the normal grade: , parallel to . Alternations of the latter type exist in Skt, e.g. brahman-, barhiṣṭa-; they were more frequent in the NW Pkt, e.g. karma-, krama-. The alternation vyadh-/vedh- occurs in Skt only in Epic (Whitney, Roots s.v.) and may represent a MIA innovation, cf. Pkt vehai.

^{page 48 note 1} This phenomenon is found also in Skt bhos < bhavas, and in NIA, e.g. Pashai kok < kawak (see Morgenstierne, , Pashai Language 2. Texts, p. xv)Google Scholar.

^{page 48 note 2} Pischel 153.

^{page 48 note 3} Geiger 26.

_{page 48 note 4} Pischel 594.

^{page 48 note 5} Edgerton 35.49–51.

^{page 48 note 6} Edgerton 1.102, 3.115. In fact from both Pischel and Edgerton it can be seen that some of the examples they quote do not show an alternation -ya/-i, but rather -ya > -yi, similar to that discussed above, e.g. koppi < -kupya, paśyi < -paśya.

^{page 48 note 7} Burrow, , Khar. Lang., 9Google Scholar.

^{page 49 note 1} For the weakening of final -a see Bailey, H. W., BSOAS., xi, 789Google Scholar, and cf. Apabhraṃśa inst. sg. -eṃ < -ena (Pischel 363).

^{page 49 note 2} This is Pischel's explanation (594), rejected by Edgerton (35.49) and Gray, (BSOS., viii, 575)Google Scholar. The latter's observation that it is “quite improbable” is hard to understand in the light of the similar change in Niya.

^{page 49 note 3} Grierson, loc. cit., para. 182. It is doubtful if this change can be called samprasāraṇa, even in its broadest sense, since the -i which remains is probably not due to the vocalization of -y-, but arises from the epenthetic -i- evolved before it when the group -ry- was resolved > -riy- instead of being assimilated.

^{page 49 note 4} Pischel 594.

^{page 49 note 5} Pischel 593.

^{page 49 note 6} See Bailey, H. W., JRAS., 1955, p. 24Google Scholar.

^{page 49 note 7} See Schwartzsehild, L. A., JRAS., 1956, p. 183Google Scholar.

^{page 49 note 8} L. A. Schwartzsehild (loc. cit., p. 184) examines the reasons for the change -āya > -āe. Since it is unlikely that this change can be separated from the corresponding change in the absolutive ending of verbs in -ā, it seems likely that it does not arise from the dative ending -āyai, but is a phonetic change -ya > -ye after a long vowel (see Alsdorf, , BSOS., viii, 329Google Scholar), similar to the change of internal -ya- > -yi- discussed above. For the corresponding change of final -va > -vo see Burrow, , Khar. Lang., 53Google Scholar.

^{page 49 note 9} See Geiger, 30–1

^{page 49 note 10} See S. Sen, Comparative Grammar of MIA, paras. 22, 26.

^{page 49 note 11} Found only in the grammarians. See Pischel 584.

^{page 50 note 1} Schwartzschild, L. A., JAOS. 76, p. 113Google Scholar.

^{page 50 note 2} Burrow, , Khar. Lang., 79Google Scholar.

^{page 50 note 3} Burrow, , Skt Lang., pp. 167, 178Google Scholar.

^{page 50 note 4} The fact that absolutives in -i are not so far attested in OIA is not conclusive since many unquestionably old features in MIA have no apparent antecedents. It may, however, be significant that there seems to be no absolutive ending in -i with the -nam suffix. All other absolutive endings appear with and without -nam, e.g. -yā, -yānam; -ccā, -ccānam; -tvī, -tvīnam; tvā, -tvānam; -tu, -tūnam.