Published online by Cambridge University Press: 22 April 2026
This is a phonemic analysis of the Sanskritized literary Tamil spoken in Madras by those whose vernacular is Tamil and whose fluent knowledge of Sanskrit is due to their heritage and upbringing. This variety of the language is distinct, on the one hand, from the several geographically determined dialects and, on the other, from Taṉittamiḻ, ‘pure Tamil’, which, free (at least in intention) of all Indo-Aryan influences, has records possibly two thousand years old and is in use today by certain people in the Tamil area. Sanskritized Tamil is a cultivated, literary language subject to strict canons of grammar and style, but it is a living speech and not an artificial medium used only by scholars.
1 The pronunciation here recorded is that of Sri E. K. Natesa Sarma, formerly of Loyola College, Madras.
2 Thus Naṉṉūl 147. Although these rules are for the transliteration of the Devanāgarī script (i.e. for the spelling of Sanskrit words in Tamil), their ultimate effect is to dictate the pronunciation. The rules are explicit and detailed, resulting in such conversions as of Skt. Ṙṣabha ‘bull’ to Tam. iḍapam and of Skt. lakṣmi ‘fortune’ to Tam, ilakkumi.
3 The following symbols are used in this article. The Devanāgarī script is transliterated in the accepted manner. The transliteration of the Tamil alphabet follows the usage of the Tamil lexicon of the University of Madras, and the symbols chosen for phonemic writing likewise are those of that romanization. The phonetic symbols are those of the IPA (rev. 1951) with the following exceptions and additions: a subscript dot marks retroflexion; underscoring marks alveolopalatal articulation; the letter r with superscript dot and breve denotes alveolopalatal flaps, tense and lax respectively; [ñ] is a palatalized nasal; [ḫ] is a lax glottal spirant. — A semicolon separating two English glosses of one Tamil word indicates that, so far as can be determined, the meanings are of homonyms, not of the same word.
4 See §5 and fn. 22.
5 Intonation and stress have been omitted from this study. The following observations have been made of stress. In two-syllable words stress is equally distributed. In polysyllables, a three-syllable law exists for short open syllables: the accent is recessive, but only as far as the antepenult; if there is a second accent, that also falls no farther back than three syllables before the antepenult (e.g. ‘kuṟṟiya'lukaram ‘the short u sound'). Long vowels in pretonic position are shortened (e.g. āṟu ‘six’ + pattu ‘ten’ yields aṟupatu ‘sixty‘).
6 Verbs glossed in this form are cited in the singular imperative, which is the normal way of naming a verb in Tamil.
7 Eight of these contrast in an identical environment: pāy ‘mat’, tāy ‘mother’ (sic), kāy ‘unripe fruit’, māy ‘conceal’, nāy ‘dog’, vāy ‘mouth’, cāy ‘lean’, yāy ‘mother’ (sic.)
8 /b/ occurs after word juncture only in words borrowed from Sanskrit.
9 /d/ occurs after word juncture only in words borrowed from Sanskrit?.
10 The name of the city of Madras, [mataṟās], a word of non-Dravidian origin, in rapid speech appears to have a lax dental stop before the lax flap; but in careful pronunciation the medial syllable is retained, and the very nearly interdental articulation of the [d] even in rapid speech indicates that the following vowel never completely vanishes.
11 There are three alternants, the distribution of which need not here be mentioned. It should be noted, however, that the phonemic problem of [g] comes to the fore again in the analysis of these morpheme variants, since if /g/ were not established as a phoneme /gal/ could be removed from the list of plural alternants. In the plural ‘trees’, then, [g] as the lax allophone of /k/ would presuppose an immediately preceding lax phoneme, while the velarity of the nasal would presuppose a following velar (§2.25).
12 /g/ occurs after word juncture only in words borrowed from Sanskrit.
13 The nasal appearing before [dʒ] is the alveolopalatal allophone of the dental phoneme /n/(§2.25, §2.29).
14 The [ŋaṉAm] of the grammarians (e.g. Naṉṉūl 106) cannot exist after word juncture, and all derivatives have the requisite /g/.
15 The word is of Dravidian origin.
16 A less transparent example is the word for ‘chair’, cited under (1) just above. This = nāṟ = nālu ‘four’ + kāl ‘foot’ + the nominal suffix i.
17 If a generalization were to be made from this one instance, it would have to be that the intensive /ṟ/ is obliterating the extensive /r/. The change from intensive to extensive is contrary to what has generally been observed in morphology and syntax.
18 The lack of lateral positioning should be emphasized. This /ḻ/ is the final phoneme in the name Tamil, of which the English spelling, the romanization tamiḻ, and the phonemic writing /tamiḻ/ here adopted, if mistakenly read as phonetic, all give a false impression.
19 Cf.fn.7.
20 The name (of Sanskrit origin) also contains the non-Tamil cluster [str].
21 No examples have been found of double /ḻ/ and double /w/.
22 R. Caldwell, Comparative grammar of the Dravidian or South Indian family of languages 138. The law is, in its author's phrasing, completely untenable, since it embodies a confusion of sound and symbol and of historical and synchronic analysis. It is possible that the graphic representation of Tamil sounds at some time in the past showed a complementary distribution of surds and sonants at some preceding period—and that, in modern terms, is what Caldwell is saying. (‘A sonant cannot commence a word, neither is a surd admissible in the middle, except when doubled.‘) But that the law operated diachronically, holding the language in a perpetual complementary distribution of these sounds, is not credible. In modern Sanskritized Tamil, surd and sonant (i.e. tense and lax) stops are in contrast (§§2.21-24).
23 The example is Caldwell's. The word is not in common use in modern Tamil, being almost completely restricted to reference, in elegant language, to the tusk of an elephant.