^{page 939 note 1} Cf. Brugmann, Grundriss, i, § 973; Grammont, Traité de Phonétique, p. 251, both of which give this example pride of place.

^{page 939 note 2} Bloch, L'Indo-Aryen, p. 56; cf. Formation de la Langue Marathe, p. 166.

^{page 939 note 3} Or, from the transcriptional point of view, the time-track or linear dimension.

^{page 939 note 4} Cf. Nida, Morphology, p. 26.

^{page 939 note 5} Cf. Bloomfield, Language, p. 213; Firth, J. B., in English Studies, xvii, February, 1935Google Scholar; Wells, R. S., “Automatic Alternation,” in Language, xxv, 1949, pp. 99 ff.Google Scholar Similar objections may be made against the Prague concept of” neutralization of oppositions”, applied to the present case by Trubetzkoy, Principes de Phonologie, p. 249.

^{page 939 note 6} Palmer, Introduction to Modern Linguistics, p. 31.

^{page 940 note 1} The “dot-and-dash” system of transliteration (ṅ for ṇ, etc.), though highly undesirable in phonetic discussion, is here retained as being in general use by Sanskritists.

^{page 940 note 2} RV., v. 42 and 56; A V., iii. 75;Tait., xiii. 6; Vᾱj., iii. 83 and 94.

^{page 940 note 3} VIII. iv. 1 ff.

^{page 940 note 4} Macdonell, Sanskrit Grammar, § 65. On the apparent RV. exceptions rāṣṭrānām and uṣṭrānām see Wackernagel, Altind. Gramm., § 145b; Macdonell, Ved. Gr., § 42a.a.

^{page 940 note 5} Sanskrit Grammar, § 189.

^{page 940 note 6} , being retroflex sounds at the time and place of the crystallization of the rule, as attested by the Pānīnya Śikṣā and by the Pāṅinean scheme in general (cf. Varma, Critical Studies in the Phonetic Observations of Indian Grammarians, pp. 6 ff.).

^{page 940 note 7} RV. Prāt., v. 56 (“avyavetaṃ… vighnak dbhih”).

^{page 940 note 8} The use of this symbol may have practical advantages when it comes to considering such changes from linear to prosodic function as are attested by Middle Indian developments of the type Skt. āria- > Pkt. aṭṭa- (cf. Pischel, § 289).

^{page 941 note 1} As may be seen in part from Professor Emeneau's, M. B. “The Nasal Phonemes of Sanskrit”, Language, xxii, 1946, pp. 86 ff.,Google Scholar with the findings of which I am in agreement; cf. also Fry, A. H., “A Phonemic Interpretation of Visarga,” Language, xvii, 1941, pp. 194 ff.Google Scholar

^{page 941 note 2} Cf. J. Vachek, Ada Linguistica v, p. 92 f.

^{page 941 note 3} Skt. ṣ is in many cases the reflex of IE ^*s by a process or processes marked at an early period not only by , r, but also by i, u, k, and it might be objected that in examples such as niṣaṅṅa- the marker is thus not ṣ but i. There are, however, good reasons for rejecting this argument: the following facts are relevant:— (a) From the physiological point of view it is difficult to conceive why a process of retroflexion should be associated with the diverse units i, u, k. (b) Parallel processes, with identical markers, are attested in Iranian and Slavonic. (c) The reflex of (b) is not, as in Skt., ṣ (ş), but Av. š (∫), O.C.S. ch (x). From (b) it follows that the beginnings of the process(es) are to be put back at least to the Indo-Iranian period. A possible interpretation of (a) + (c) is that the reflex of IE ^*s at this period varied according to the various markers (? iç, UX^W, rş, kx)—cf. also Morgenstierne, “The Language of the Ashkun Kafirs,” Norsk Tidsskr. f. Sprogvidenskap, ii, 1929, p. 199 f., Entwhistle, TPS., 1944, p. 33 f. The reasons for the ultimate “phonemic divorce” of these variants from s, and their convergence to Skt. ṣ, Av. š, O.C.S. ch, (as also the development ^*t > Skt. ṣṭ, Av. št) are probably to be sought in the total phonological systems of the respective languages. Thus, by the time of our attested Skt., the units i, u, k, could no longer be treated as markers of a process: in sequences of the type iṣ, etc., ṣ, not i, was now the marker (viz. of the prosody of retroflexion), and we may therefore, in the present problem, treat it as parallel to r. Where a process ^*i + s >iṣ, etc., appears to be active in the attested language (e.g. in new composition) it is to be explained as the result of “structural pressure” from inherited forms, intensified by the canonization of an observed phenomenon to the status of a Sandhi Law; its lack of physiological justification is at least hinted at by anomalous forms such as avaṣṭambh-, pratistambh-, and the sandhi agnis te, beside Vedic agniṣ ṭe, etc.

^{page 941 note 4} As shown by its development in Middle Indian (cf. Pischel, § 224).

^{page 942 note 1} Cf. the pronunciation of Hindi pah, where distinct retroflexion is heard in the vowel.

^{page 942 note 2} On word-palatograms and their uses see J. R. Firth, “Word-palatograms and Articulation,” BSOAS., xii, 1948, pp. 857 ff.

^{page 942 note 3} Cf. also Daniel Jones’ distinction between assimilation and “similitude” (Outline, § 835 ff.).

^{page 942 note 4} In view of the still vexed problem as to the date of the earliest writing of Sanskrit, and hence its relationship to the more ancient grammatical and phonetic treatises, it may be advisable to interpret “orthography” in a wider sense than is usual, i.e. as extending, if such a thing is possible, to a system of phonemic “oral spelling” antedating the use of any script.

^{page 943 note 1} In the case of an ancient language such an examination must, of course, be largely hypothetical, and in part based on the very orthographic data which we are rejecting; however, there is no reason why, having stood on these phonemic stepping-stones whilst building our prosodic bridge, we should any longer tolerate their obstruction of the stream.

^{page 943 note 2} The fact that y appears as a nom-interfering articulation is not unexpected. Intervocalic y in Skt. is in many cases to be treated simply as a prosodic marker of syllable-junction—e.g. in dhiyā, etc., and probably in many instances of the type dāyalca- <*dā-aka- (cf. Wackernagel § 187; note also the alternation of y with v, another typical prosodic marker, at all stages of Indo-Aryan); as such, it is likely to have involved less constriction than a phonematic y (cf. the “laghuprayatnatara” y of Jaina Pkt., or of Hindustani aya, etc. As for post-consonantal position, we may remember the frequent alternations of y and ^* under the conditions of “Sievers’ Law”, and the wider extension of the vocalic realization in Vedic Skt., as attested by metrical requirements (-bhyas to be realized as -bhi-as, etc.). The weak articulation of y is moreover supported for non-initial position in general by the ancient treatises (cf. Varma, op. dt., pp. 126 ff.).

^{page 943 note 3} The operation of “Verner's Law” is a case in point. Cf. also Twaddell, On Lefining the Phoneme, p. 49.

^{page 944 note 1} Cf. Grammont, op. cit., pp. 269 ff.; R. G. Kent, “Assimilation and Dissimilation,” Language, xii, 1936, pp. 245 ff.

^{page 944 note 2} The addition of “potential” is necessary to cover reduplicated forms, such as babhῡva, where although the bh, in fact, belongs to the second syllable, babh- is potentially a radical syllable (like labh-, dabh-, etc.)—cf. the attested secondary root dadh- from da-dhā-ti. The non-application of the rule in other cases may be illustrated by such a form as abhi-dhāa-bhiḥ.

^{page 944 note 3} E.g. budh-: bhot-syati. The aspiration may vanish from the syllable altogether if the dominant junction-prosody is such as to involve the transfer of the syllable-prosody to another syllable, as in the case of “Bartholomae's Law” (^*budh-ta- > buddha-, etc.). Variations such as dhug-dhvam beside dug-dhi present a difficulty of interpretation.

^{page 944 note 4} Indicated as bwht in Harley's Colloquial Hindustani; cf. also Hoenigswald, “Declension and Nasalization in Hindustani” (JAOS., 68), p. 143 f., n. 15.

^{page 945 note 1} E.g. J. R. Firth, “The Structure of the Chinese Monosyllable in a Hunanese Dialect (Changsha),” BSOAS., viii, pp. 1055 ff.; N. C. Scott, “The Monosyllable in Szechuanese,” BSOAS., xii, pp. 197 ff.; E. J. A. Henderson, “Prosodies in Siamese: A Study in Synthesis,” Asia Major, I, ii; J. Carnochan, “A Study in the Phonology of an Igbo Speaker,” BSOAS., xii, pp. 417 ff.

^{page 945 note 2} Language, xx, pp. 181 ff. Cf. also C. F. Hockett, “Peiping Phonology,” JAOS., 67, 1947, pp. 253 ff., “Componential Analysis of Sierra Popoluca,” UAL., 13, 1947, pp. 258 ff.

^{page 945 note 3} E.g. Skt. snāna- > Pkt. ṅhāṅa-, etc., IE ^*ismeros > Gk. ἴμερος;. Note also the Alexandrian TT prosodic writing of ῥ (i.e. ^H_r-) for the Greek sound which was the reflex of inter alia IE ^*sr- (e.g. ῥέω): the Gk. ῥ is transcribed in Latin as rh- (e.g. rhētor), in Armenian as hr- (hretor); Tsaconian šinda < Lac. ῥίδδά, suggests a realization as breathed r- (cf. Sturtevant, Pron. of Gk. and Lat., § 67b). The phonematic s in such cases has been replaced by prosodic breathing, just as the pre-IE phonematic “laryngeal” units are replaced by various prosodies in the attested languages: e.g. length in Gk.τίθημι < ^*-dhe-, aspiration in Skt. tiṣṭhuti < ^*-ste-, voice in Skt. pibati < ^*pip eti, hiatus (a prosody of syllabic junction) in Gk. θ ος; < *dhuos (cf. θῡμός, Hitt. tu wis).

^{page 946 note 1} The same peculiarity is shown, as we might expect, by forms such as pranaṣṭa- beside praṅaśati, or pariniṣṭhā- for ^*pariṅiṣṭhā-, the ” dissimilating influence ” here extending also to the form pranaṣkasyati for *praṇankṣyati.

^{page 946 note 2} With a partial parallel in Av. āyzrāμayeiti for expected ^*āyž-.

^{page 946 note 3} The relevant features of the parallelism may be diagrammaticaUy illustrated as follows:—

^{page 946 note 4} Ed. Goetz and Schoell, Frag. 76. Cf. 200.12 (“H quod adspiratio sit non littera”).