Rhotics, final and otherwise.

In LT, there was a contrast between three r-like consonants:

<LI> ¯ r, a phonetically flapped or tapped r, more or less alveolar, phonetically [ '104 ]. This sound could not occur in initial position in older forms of LT, but in ST occurs initially, medially, and finally (but usually finally is followed by epenthetic [ '124 ] before pause.) This sound does not undergo gemination (doubling).

<LI> ± r was originally an alveolar stop in Proto-Dravidian and in older forms of Tamil. It did not occur initially or finally (just like retroflex stops) and when geminated (doubled) had the value of a voiceless alveolar stop: ±± was phonetically [tt] or [t] with r-like offset: [tr]. Intervocalically, ± was trilled: [], and in some dialects, mainly southern (Kanniyakumari etc.) a real phonetic contrast between this sound and the previous one is maintained. However, in the speech of most Tamils, a phonetic contrast is not maintained, even if speakers maintain the fiction that there is. This sound is orthographically maintained when writers depict spoken Tamil in plays and novels, since it is the marker, among other things, of the present tense. We therefore maintain it for spelling contrast in our transcriptions, e.g. the two `r's of ÂçØÅÔ£ varroom are not phonetically distinct, but we write them as ¯ and ± (in Tamil script) because Tamil linguistic culture prefers this. This sound is only found in native-Dravidian lexica.

<LI> Another sound that is sociolinguistically complex is the retroflex frictionless continuant ° = to0pt.25ex ##= by .25ex which, under ideal conditions, is phonetically [ '107 ].This was the symbol used by Firth (1934).

In contemporary Tamil, many speakers replace this sound totally with the retroflex lateral ª l and a plain lateral [l] has also been the symbol used by most Europeans for the name of the language: `Tamil' is »¾Õ° tami= to0pt.25ex ##= by .25ex ; the `Chola' Kingdom is ضÔà coo= to0pt.25ex ##= by .25ex a, but there is also the item `Coromandel' (from ضÔþÿ»Á£ coo= to0pt.25ex ##= by .25ex amandalam) for the coast south of Madras, with an /r/ instead of an /l/. Because of the `mystique' surrounding this sound (Tamils seem to believe it is `unique' in Tamil) it is learned only through literacy by many speakers, and even then, some never master it. Therefore, if foreigners can learn it, it gets them good karma, so we use ° = to0pt.25ex ##= by .25ex wherever it occurs in LT, even though it is rare in ST. Tamil second-language learners should note, however, that forms like ×´ÔÃÿ×» ko= to0pt.25ex ##= by .25ex ande `child', ÂÔ×à ½Ã£ vaa= to0pt.25ex ##= by .25ex e pa= to0pt.25ex ##= by .25ex am `banana, plantain' are more likely to be pronounced [kolande] and [va '161 le p '025 l] than with corresponding ° = to0pt.25ex ##= by .25ex I have heard, however, a hawker selling bananas on a railway station platform in Trichy, with clear [ '107 ] in the appropriate places.

The ° = to0pt.25ex ##= by .25ex sound is never geminated (another reason to consider it to be an r-sound) and does not undergo many of the morphophonemic rules that apply to ª l, e.g. it does not become replaced by nasals, or become a stop the way ª l does. Neither sound occurs in initial position, both because rhotics and liquids did not in LT, and because retroflex consonants never occur in initial position in native Dravidian words. In some cases, intervocalic ° = to0pt.25ex ##= by .25ex may be deleted, with compensatory lengthening of the vowel: LT ×½Ôêâ po= to0pt.25ex ##= by .25ex utu `time' --> ST ؽÔâ poodu.

</UL>

Our solution, therefore, is always to distinguish these three sounds, and the two l's, in all our transcriptions, even though many of them may be neutralized in many people's speech.