Transliteration.

• Initial Stop Consonants. The Tamil stop consonants p , t , 8 , c , k are usually represented in initial position as [p, t, t, s] and [k]. Where words borrowed from Indo-Aryan, English, Portuguese, or some other language preserve voiced stops in initial position, we use the Roman letters [b, d, d, j] (but sometimes also [s], and [g]. Actually retroflex consonants never occur in initial position in Tamil words, so they will never occur in Literary Tamil, and in ST only in borrowed words.

• Medial Voiceless Geminate Stops. Medially, stops are voiceless in Literary Tamil only if geminate, i.e. doubled, and we follow this convention as well, writing two consecutive consonants [pp, tt, tt, cc, and kk] for what Literary Tamil writes as pp , tt , 88 , cc , and kk . Geminate [cc] is to be understood as being phonetically [cc]. Geminate Literary Tamil 66 never occurs in ST as such, but is always realized as [tt], e.g. pa66i `about' is [patti] in ST. After short vowels, the phonetic value of geminate consonants is to be understood as longer in duration than a single voiceless consonant; after long vowels, geminate consonants are actually not as long as after a short vowel, but we represent them as double and long, anyway, in order to avoid the confusion that results when a single stop occurs intervocalically. That is, Literary Tamil poo8u `put, place, serve' is [poodu] in ST, but poo88u `having put, placed, served' is [poottu] in our transcription, rather than [*pootu].

• Post-nasal Stops. The stop consonants p , t , 6 , 8 , c , k following nasals are always voiced in both Literary and Spoken Tamil. Except for - 6 -, which does not occur in such clusters in ST, they are represented in this transcription by the roman letters [b, d, ( ) d, j], and [g.] Thus tampi `younger brother' is [tambi], angkee `there' is [ange], ke5cu `supplicate' is [kenju], etc.

• Medial Stop Consonants. Single stop consonants (i.e. not double or geminated) in medial position, (i.e. between two vowels) in Tamil are typically laxed and fricativized. Thus the stops consonants p , t , 6 , 8 , c , k in medial position are actually phonetically [b] (or sometimes [v] or even []), [], [r], [D] [s], and [h], respectively. That is, some of them are laxed and voiced, some are flapped (e.g. the retroflex stop), but some (the palatal and velar) are only laxed but not usually voiced (although the velar may in some speech be voiced as well, i.e. pronounced [].) In order to stick to a Roman transcription that does not require elaborate phonetic symbols that complicate our typography, we will use the voiced Roman letters [b] (or [v]), [d, r, d] and the voiceless fricatives [s] and [h] for these lax and sometimes fricativized intervocalic Tamil stops. As noted above, Tamil, unlike some Indian languages, does not have a single standard transliteration system. Our choice was one that could be used by lay persons as well as scholars, hence the use of [h] for intervocalic k rather than a morphophonemic //k// or a phonemic /g/. This does not follow a purely linguistic (e.g. phonemic) convention, or those used in most other dictionaries, but we have found through years of teaching Tamil that most dictionary representations of Tamil are not phonetic enough to permit non-Tamils to approximate usual spoken pronunciations.

  In all of our phonetic representations we give preference to those that preserve morphological clarity, so that even though it may be common for many speakers to convert Literary Tamil intervocalic 8 to a flapped rhotic alveolar [r], we still represent these as [d]. Since the completive marker (Literary Tamil vi8u ) may be realized in the speech of many people as non-retroflex, i.e. vantuvi8uki6ee2 `I am definitely coming' may come out in ST as [vandirr], we prefer the transcription [vandidr], as this shows more clearly that there is a completive aspect marker vi8u present, even if it is realized phonologically only as [d]. Otherwise it is hard to explain why the past of it is always, in all dialects [vanditt] Literary Tamil vantuvi88ee2 . This will help avoid confusion with `perfect' forms also marked with iru , contrasting with completive - vi8u .

• Laterals and Rhotics. The Tamil sonorants l, 9, r , and 7 are represented in our transliteration as [l, l, r,] and [], respectively. Tamil 6 is not treated as a sonorant in its underlying form, but as a stop. Neither in Literary Tamil nor in ST can it occur in initial position; intervocalically it does occur in ST, where in most dialects it is indistinguishable from r , so in this position it is simply transliterated by [r]. In ST, Literary Tamil clusters such as 26 are simplified in non-Brahman dialect to [nn] after short vowels (e.g. Literary Tamil e26u `having said' becomes simply [-nnu]); after long vowels in ST 26 becomes simply [n], i.e. too26u `seem, appear' is [toonu].

  In many spoken dialects of Tamil the sonorant 7 is merged with 9 , and never occurs in ordinary speech. Because this sound is sociolinguistically highly preferred, however, and foreigners who are able to pronounce it are praised for their efforts, we give [r] as the usual transliteration of 7 , even though many speakers, even educated ones, may be heard to use 9 . That is, we give [mae] for Literary Tamil ma7ai `rain', even if many speakers say [male] or even [male].

  There are certain stem-final consonants in this group that tend to be not realized at all in ST, while others occur but with an epenthetic [u] vowelFor its pronunciation, see below. added. Thus it is typical for the final 9 in Literary Tamil pronouns ava9, avarka9, niingka9 (`she, they, you') to be deleted in final position in ST: [ava, avanga, niinga]. The same is true of 9 that occurs as the final segment in certain person-number-gender endings, as in irukki6iirka9 `you are located' which is realized phonetically as [irukkriinga] if nothing follows it. If anything follows as a suffix, however, 9 is not deleted in ST: niingka9aa ? `you?' is realized in ST as [niingalaa], essentially phonetically the same as the spelling pronunciation of Literary Tamil. Final 9 of other words may be treated in different ways by different ST dialects. Some dialects delete 9 in naa9 `day' to give [naa], but others add an epenthetic [u]: [naalu]. We give preference the dialects that do the latter, i.e. preserve morphological clarity.In this sense we take bits and pieces of different dialects as `standard', since this is sounder pedagogically, even though there may be no speaker who actually replicates each and every pronunciation we prefer.

  Monosyllabic words with short vowels ending in laterals (there are none that end in rhotics) such as pal `tooth', ka9 `toddy', col `say', etc. are realized in ST with doubled laterals and an added epenthetic [u] vowel: [kallu], [kallu], [sollu], etc.

• Nasals. Literary Tamil has graphemes for a labial nasal m , a dental nasal n , an alveolar nasal 2 , a retroflex nasal 3 , a palatal nasal 5 , and a velar nasal ng . Spoken Tamil does not need this many phonemic or phonetic distinctions, requiring only [m, n,] and [n]. We therefore transliterate m as [m], both n and 2 as [n], With the exception that clusters 26 usually become [nn], as already noted. 3 as [n], while 5 and ng which usually occur only before a homorganic nasal (i.e. the palatal and velar nasal, respectively) can be transcribed with [nj] and [nk] with the assumption that English speakers, at least, will pronounce these with palatal and velar articulations automatically. 5 occasionally occurs in prevocalic position in ST, in which case we do transliterate it as [ n], e.g 5aa2i `sage' [ naani].

  Monosyllabic words with short vowels ending in alveolar nasals (those that end in retroflex nasal follow the pattern of the laterals mentioned above; labial nasal does not occur) such as e2 `my', o2 `your' convert the nasal segment into nasalization of the vowel: [], [ o],

  Monosyllabic words with short vowels ending in retroflex 3 , such as pe3 `girl', ka3 , `eye' etc. follow the pattern of doubled laterals with an added epenthetic u: [ponnu],This form also undergoes vowel rounding; for an explanation see below. [kannu] in ST.

• Glides. The Tamil glides v and y are usually transliterated as [v] and [y], respectively. In certain cases v is closer phonetically to [w] or even to the bilabial [], similar to Spanish `v' in `vaca', but we ignore this degree of phonetic detail. Often v and y in Literary Tamil forms will not have any phonetic representation at all in ST, since intervocalic v in particular is deleted in weak positions, resulting in forms like ST [kondaa] from Literary Tamil ko38uvaa . In such cases no representation of v will appear in ST. Similarly, the presence of y in Literary Tamil often conditions palatalization of adjacent consonants in ST, with subsequent assimilation or deletion of the y in ST. Thus, Literary Tamil aayntu `five' is [anju] in ST---the y causes palatalization of the dental nt , after which y disappears, leaving only ST [nj]. In final position also, many Literary Tamil y 's are deleted, e.g. the adverbial ending - aay is realized simply as [aa] in ST.

  In contrast, Tamil words beginning with the mid-vowels [e] and its long counterpart [ee] have an automatic [] onset in Tamil (as in many South Dravidian languages). Thus e22a `what' is phonetically [enna]. Since this is automatic and predictible, we do not supply this [] glide in our transcription, and transcribe e22a always as [enna]. Similarly, the rounded vowels [o], [oo], [u], and [uu] are preceded in initial position with an automatic [w]-glide in Tamil. uur `town, city' is phonetically [ur()], but again since these w-glides are predictible (in fact hardly even salient to a Tamil speaker), we do not provide them. In connected speech in non-sentence-initial position, these automatic glides are usually deleted: avar peer e22a ? `What is his name?' is in ST [avar peer-enna] rather than [avar peer-psupyenna].

• Oral Vowels. The Tamil vowel system consisting of five cardinal vowels a , i , u , e , o and their long counterparts aa , ii , uu , ee , oo are represented in our transliteration as [a, i, u, e, o] and [aa, ii, uu, ee, oo], respectively. The diphthong ai is usually simplified to [e] in ST; thus the accusative avar - ai `him' is [avare] in ST. This is actually phonetically [], but we represent it as [e] for simplicity. In monosyllables ai is not monophthongized, but the [i] element is lengthened, or followed by a glide [y]. Thus the verb vai `put, keep' becomes [vayyi] in isolation (e.g. as a verb stem or imperative), but in more complex morphological constructions, e.g. followed by tense-markers, will change to [e]: vaittee2 `I put, kept' will be [vacceen] or [vecceen]. Here the [i] element triggers palatalization of the tt to produce [cc].

  A special note must be made of the phonetic qualities of the short u vowel when it occurs after the first syllable of a word, and in particular in final position. Its pronunciation in initial syllables is [U], but after the first syllable its phonetic quality is unrounded and somewhat fronted, i.e. more like IPA [] or []. This is similar in quality to the short `oo' vowel in `book' as pronounced in southern American English, to the Russian [] (jery) or to the final [u] in Japanese. Since it is again totally predictable when a Tamil u will be pronounced in this way, we do not represent it as different from phonetically rounded [u]. This pronounciation is not different from the spelling pronunciation of Literary Tamil [u], so anyone with a knowledge of the pronunciation of Literary Tamil will have no trouble predicting this.

  This situation is complicated by the fact that in ST, many short [i] vowels (phonetically [I] also merge with []. For example, the vowel of the past tense-marker of Class 3 verbs spelled i2 as in vaangki2ee2 `I bought, acquired, fetched' is pronounced like [] in ST: [vaangneen]. Some linguists who have worked on ST have regularly substituted [] in these positions, but since this pronunciation is predictible, and differs from the spelling of Literary Tamil, we do not give either [u] or [] here, but transcribe it as [i], leaving it to the knowledge of the speaker to provide the correct phonetic realization. The Literary Tamil diphthong au is rare even in Literary Tamil, and does not occur in our data except in loan words, e.g. English `pound', which we would represent as [paundu] or [pavndu].

• Long Vowels. In final position in ST, Literary Tamil long vowels are often shortened, so that what may be written with a long vowel may always occur short in ST. Thus angkee `there' is always [ange] in ST, unless followed by another vowel, as in angkeeyee `there (emph.)' [angeeye]. Here the non-final ee remains long, but the final one is shortened. Sometimes to keep morphological processes clear, however, we represent long vowels in final position as long (in transcription), even though they are phonetically short. In rapid speech, moreover, long vowels anywhere in a word will be shorter than when the word is in isolation, and short vowels may be completely deleted.

• Nasal Vowels. Spoken Tamil possesses a set of nasal vowels [ a], [ o], [], and [ u], some of which also have long counterparts [ a a], [ o o], and []. These nasal vowels are not found in Literary Tamil, but arise from the nasalization of vowels followed by 2 or m in final position. Thus, [ a] arises from the sequence - a2 in Literary Tamil, e.g. ava2 `he' becomes [av a]; in some dialects ava2 becomes [av] instead, which accounts for some instances of []. [ o] arises from the Literary Tamil sequence - am , so that Literary Tamil maram `tree' becomes [mar o]; [ u] arises from the Literary Tamil sequence - um , e.g. niingka9um `you, too' becomes [niingal u] in ST.

  Long nasal vowels [ a a], [ o o], and [] may have several sources in Literary Tamil. [ a a] may result from the nasalization of both Literary Tamil - aa2 or - aam , i.e. vantaa2 `he came' becomes [vant a a], but irukkalaam `it may be, let it be' also has final [ a a], i.e. [irukkal a a]. Because of the shortening of long vowels in final position, these long nasal vowels also are shortened finally; but to preserve morphological clarity we usually represent them as long in our transcription. The long vowel [ o o] arises from the nasalization of the sequence oom , found typically as the marker of second person plural, as in vantoom `we came' [vand o o]. Because of the shortening rule, however, it may be realized phonetically as [vand o], but we usually avoid this representation, again for morphological clarity. The long vowel [] usually arises from nasalization of the Literary Tamil sequence - ee2 , found most typically in the person-number-gender ending for first person singular, as in vantee2 `I came', i.e. [vand]. Again, by the shortening rule this usually becomes [vand], but for clarity we avoid this representation. It does not become short in monosyllabic environments, so Literary Tamil ee2 `why?' remains long: [y], contrasting with e2 `my', which is [y].

  Monosyllabic words with long vowels ending in Literary Tamil 2 usually do not nasalize, but instead an epenthetic [u] [] is added, e.g maa2 `stag', becomes [maanu]. Literary Tamil words ending in 3 also do not produce nasalized vowels in ST, but if position final, simply add [u], e.g. tuu3 `pillar' becomes ST [tuunu].

• Vowel Shifting. A number of other differences between vowels in Literary Tamil and their realization in ST have to do with certain phonological changes in the Tamil vowel system since Tamil orthography was fixed.
  • Lowering. Literary Tamil words with short high vowels i [i] and u [u] in an initial syllable followed by one consonant and the vowel a [a] or ai [ai] lower these vowels to e [e] and o [o] respectively in ST. Thus Literary Tamil forms like ilai `leaf' and ku7antai `child' become [ele] and [korande], respectively, whereas forms like illai `no, not', where the vowel is followed by a double consonant, do not lower. This change is totally regular, so that even some borrowed words, such as English `glass', borrowed usually as ki9aas may, in some dialects, become [kelaas] or even [kelas].

  • Rounding. Another process that is less regular, and may therefore still be in progress as a sound change, is the rounding of short and long front vowels i , ii , e and ee ( [i, ii, e], and [ee]) to their corresponding back vowels u , uu , o , and oo ([u, uu, o], and [oo].) This occurs usually when the initial consonant is a labial ([m, v, p]) and the following consonant is retroflex. Some forms that have undergone this change are quite acceptable, but others are considered to be somewhat substandard or casual (or even ``vulgar") so many speakers avoid this kind of rounding, or deny that they do it even when it is observed in their speech. Thus the rounding of the vowel in Literary Tamil pe3 to ST [ponnu] is quite normal, but the following are on a kind of sliding scale of acceptability: [porandadu] for Literary Tamil pi6antatu `it was born' (this form undergoes lowering first), [pudi] for Literary Tamil pi8i `like', [vuudu] for Literary Tamil vii8u `house', [voonum] for Literary Tamil vee38um `want, need, must', etc. Different speakers would rank these differently, but the general scale of acceptability is as given. We try to avoid what are considered the most egregious of these, but in an attempt to remain colloquial, some may be present in our examples. Sometimes the conditioning factor does not even include a retroflex consonant, as in the example of pi6a , where the following consonant is alveolar; in extreme cases no second conditioning consonant is present, and an initial labial alone is sufficient to cause rounding, as in Literary Tamil mitakkum `it will float' becoming, in some dialects, [modakkum]. This is obviously an example of a sound change in progress, and is therefore sociolinguistically marked.

  • Other Changes. There are a few other changes in vowel quality from Literary Tamil to ST that cannot be described under the previous rubrics. These are mostly idiosyncratic, but may have to do with what is often called vowel harmony, i.e., vowels changing in order to agree with another vowel in height or rounding. Thus, for example Literary Tamil ko8u `give' has a high vowel in its first syllable in ST, i.e. [kudu] rather than *[kodu], which is the reading pronunciation of LT. There is no good explanation for this change, except that the height of the two vowels agree; but there are many other counterexamples. Since this dictionary is neither an etymological nor a historical dictionary, its concern is not to explain such changes, but merely catalogue them.