The phonology of Tamil is characterised by the presence of "true-subapical" retroflex consonants and multiple rhotic consonants. Its script does not distinguish between voiced and unvoiced consonants; phonetically, voice is assigned depending on a consonant's position in a word, voiced intervocalically and after nasals except when geminated. at p. 371 Tamil phonology permits few consonant clusters, which can never be word initial.
Vowels
The vowels are called ('life letter'). The vowels are classified into short and long (five of each type) and two diphthongs.
The long (netil) vowels are about twice as long as the short (kuṟil) vowels. The diphthongs are usually pronounced about 1.5 times as long as the short vowels, though most grammatical texts place them with the long vowels.
+ Monophthongs Front Central Back short long short long short long Close ( ) Mid Open
Tamil has two diphthongs: and , the latter of which is restricted to a few lexical items like vau and vauvāl. Some like Krishnamurti consider the diphthongs as clusters of /a/ + /j, ʋ/ as they pattern with other VC clusters. There are clusters of /ej, oj, uj, eːj, aːj, oːj, uːj/ as in cey, koy, uy, tēy, kāy, kōy, tūymai but they are never taken as diphthongs. The way some words are written also varies e.g. avvai as (avvai), (auvai) or (avvay) (first one most common). Word final is pronounced as , it is called a (kuṟṟiyalukaram) "short u" (as it has only half a sound unit, compared to 1, 1.5 or 2 of other vowels) in tolkāppiyam and it is unrounded even in literary Tamil; in spoken Tamil it can occur medially as well in some words after the first syllable. Word final [u] occurs in some names, chiefly nicknames like rājēndraṉ as rāju but also some other terms of the form (C)VCu as in karu, puḻu. There is a minimal pair naṭu "center", naṭŭ "to plant". In declensions the full u is not deleted like the half u like naṭuvillai - naṭillai.
Old grammars mention that there are 1 (regulars), 2 (long vowels, geminates), 3 (extra long vowels), 1.5 (diphthongs), 0.5 (kuṟṟiyalukaram, kuṟṟiyalikaram) and 0.25 (makarakkuṟukkam) units of sounds: kal, kāl, orūu, vau, atu, pōṉm. 3 units are for the extra long vowels as in kurīi, ēe, kiḻāaṉ, makaṭūu.
Colloquially, an initial or may have a onglide; likewise, an initial or may have a onglide, e.g. and . In Karnatakan dialects short versions of them may further become [a, ʋa], eg. eṉṟu, utai /ɐnnʉ, ʋɐd̪iki/. This is very light or does not occur in Sri Lankan dialects.
Indian Colloquial Tamil also has nasalized vowels formed from word final vowel + nasal cluster (except for /Vɳ/ where an epenthetic u is added after it). Long vowel + nasal just nasalizes the vowel, short vowel + nasal may also change the quality, for example, /an/ gets fronted to [ɛ̃] becomes [aʋɛ̃] ([aʋæ̃] for some speakers), /am/ gets rounded to [õ] becomes , becomes , becomes , the remaining vowels only get nasalized. There is word final nasals in loanwords like ṭaim, ōm, sāmān. Karnataka's dialects have [ʊ̃] for /an/ and -m is just deleted, eg. maram [mɐrʊ].
In spoken Tamil sometimes an epenthetic vowel u is added to words ending in consonants, e.g. nil > nillu, āḷ > āḷu, nāḷ > nāḷu (nā in some dialects), vayal > vayalu etc. If another word is joined at the end, it is deleted.
Colloquially, the high short vowels , are lowered to and when next to a short consonant and . For example, becomes ; and becomes . This is an instance of lowering umlaut. It doesn't happen in pronouns and some other words e.g. ivaṉ and evaṉ are different words. also monophthongises to an but it causes the lowering of before it, e.g. ilai > ele/ela. Additionally, the front long vowels and are subject to retraction when present in the first syllable of a bisyllabic word and followed by a retroflex consonant. As such, "house" becomes , but its inflected form remains . Likewise, "search!" becomes , but "(he) searched" remains . The presence and degree of retraction for each vowel may be different; it varies between dialects and even individual speakers. Almost all words end with vowels in spoken Tamil.
For some speakers in spoken Tamil the front vowels /i(:), e(:)/ get rounded to their corresponding rounded back vowels when they are after a labial consonant /m, p, ʋ/ and before a retroflex consonant, some words with it are quite acceptable like பெண் /peɳ/ > பொண்/பொண்ணு [poɳ~poɳ:ɯ] but others like வீடு /ʋi:ʈu/ > வூடு [ʋu:ɖɯ] are less accepted and may even be considered vulgar.
Another change in spoken Tamil is vowel harmony, where vowels change their height to be more similar to nearby vowels: e.g. literary Tamil /koʈu/ > spoken Tamil [kuɖɯ].
Consonants The consonants are known as ('body letters'). The consonants are classified into three categories with six in each category: valliṉam ('hard'), melliṉam ('soft' or nasal), and iṭayiṉam ('medium'). Tamil has very restricted consonant clusters (for example, there are no word-initial clusters). There are well defined rules for voicing stops in the written form of Tamil, Centamiḻ (the period of Tamil history before Sanskrit words were borrowed). Stops are voiceless when at the start of a word, in a consonant cluster with another stop and when geminated. They are voiced otherwise.
Tamil is characterized by its use of more than one type of coronal consonants: like many of the other languages of India, it contains a series of retroflex consonants. Notably, the Tamil retroflex series includes the retroflex approximant () (example Tamiḻ; often transcribed 'zh'). Among the other Dravidian languages, the retroflex approximant also occurs in Malayalam, old Badaga, old Telugu and old Kannada. In most dialects of colloquial Tamil, this consonant is seen as shifting to the retroflex lateral approximant in the south and palatal approximant /j/ in the north.
The proto-Dravidian alveolar stop *ṯ developed into an alveolar trill /r/ in the Southern and South Central Dravidian languages while *ṯṯ and *ṉṯ remained (modern ṟṟ, ṉṟ).
[n] and [n̪] are mostly in complementary distribution and are predictable; namely, [n̪] occurs word initially and before /t̪/, while [n] occurs everywhere else. [n̪] occurs in other places from compounding of words starting with it and geminates from sandhi, eg. annēram (a-nēram), ainūṟu (ai-nūṟu) but speakers might use the alveolar nasal there instead.
/ɲ/ is rare word initially and is mostly only found before /t͡ɕ/ word medially; it occurs in geminated form as in āññai or maññai, mostly from Sanskritic loans with jñ or sandhi where the 2nd word starts with ñ aññāṉam (a-ñāṉam), in singular form in some rare words pūñai, añal and in compounds like aṟiñaṉ. Only around 5 words have doubled intervocalic [ŋ], all are different forms of the word aṅṅaṉam "that manner", apart from that [ŋ] only occurs before /k/.
A chart of the Tamil consonant phonemes in the International Phonetic Alphabet follows:
+Standard Tamil consonants Labial Dental Alveolar Retroflex (Alveolo-) palatal Velar Glottal Nasal Plosive/Affricate ~ Fricative Tap ɾ Trill Approximant Lateral approximant
+ Common Spoken Tamil consonants Labial Dental Alveolar Retroflex (Alveolo-) palatal Velar Glottal Nasal Plosive/Affricate ⠀ ⠀ ⠀ ⠀ ⠀ Fricative () ⠀() () () () () Tap Trill () Approximant () Lateral approximant
, , and are only found in loanwords and may be considered marginal phonemes, though they are traditionally not seen as fully phonemic. Intervocalic is pronounced as [ɣ~h] by Indian Tamils and [x] in Sri Lanka. For most speakers in spoken Tamil the distinction between the tap and trill is lost except in the southern Kanyakumari dialect. In most spoken Indian dialects, /rr/ [tːr] merges with /t̪t̪/ while some others keep them as [tːr] . /nr/ merges with /nn~ɳɳ/ if the preceding vowel is short or /n~ɳ/ if it is long like in literary Tamil. In speech, /pɐrri, pɐt̪ɐrrɐm, kɐɻɐnrɐt̪u, onru, muːnru, panri/ are realized as [pɐt̪t̪i, pɐðɐʈ(ː)ɔ̃, kɐɭ̆ɐɳɖɐðɯ, (w)oɳɳɯ~(w)onnɯ, muːɳɯ, pɐnni]. /t͡ʃ/ in spoken Tamil varies significantly. Some speakers pronounce it as [s] intervocalically and as an affricate initially, while others have [s] both initially and intervocalically. A final group of speakers has [t͡ʃ] before certain vowels and [s] before others, e.g. சின்ன [t͡ʃin:a] "small" but சாவி [sa:ʋi] "key". However, there are words where the pronunciation is fixed—for example, ceṉṉai and sēlam cannot be pronounced as [senːaɪ̯] and [t͡ʃeːlam]. [ʃ] is also attested as an allophone in certain dialects, eg. cari [ʃɐɾi]. /t͡ʃː ɲt͡ʃ/ are always [tːʃ n̠ʲd̠ʒ]. In spoken Tamil /j/ might cause palatalization to the adjacent consonants and then get assimilated or deleted, e.g. literary Tamil aintu, spoken Tamil añju. In spoken Tamil intervocalic /k, ʋ/ may be deleted sometimes as in /poːkiraːj/ as [poːrɛ] and sometime /ɻ/ with compensatory lengthening of the vowel as in /poɻut̪u/ as [poːd̪ɯ]. Word finally glides, mainly /j/ are generally deleted unless if the word is monosyllabic where its doubled e.g. cey > seyyi, rūpāy > rūbā. Word final l and ḷ in polysyllabic words are deleted especially in pronouns but reappear when a suffix is added e.g. nīṅkaḷ > nīṅga, similar to French liaison. l and ḷ gets assimilated to ṟ, ṭ before plosives and ṉ, ṇ before nasals in literary Tamil, e.g. vil, kēḷ, nal, veḷ > viṟka, kēṭka, naṉmai, veṇmai, before c, ṇ too becomes ṭ, eg. kāṇ-ci > kāṭci; in spoken Tamil it is deleted and the next consonant is doubled. Before coronal stops, the stops assimilate to the lateral's POA and the lateral is deleted like in the past tense forms of verbs ending with l and ḷ, e.g. kol-ntu, koḷ-ntu, vil-ttu > koṉṟu, koṇṭu, viṟṟu. As in Proto Dravidian, literary Tamil words can't begin with an alveolar or retroflex consonant but in spoken Tamil some words begin with r and l because of deletion of the initial vowel, e.g. literary Tamil iraṇṭu, ulāttu, spoken Tamil raṇḍu~reṇḍu, lāttu. In loanwords, a short i, u or a is added before them, e.g. Skt. loka, Tamil ulakam. In loanwords, the voiced and aspirated plosives are all loaned as the plain plosive. Of the fricatives, h is loaned as k, the sibilants as c, ts as cc and kṣ as ṭc; sometimes s and ṣ are loaned as t and ṭ as in mātam, varuṭam, ilaṭcam (Skt. māsa, varṣa, lakṣa) and kṣ as k(k) or c(c) as in kēttiram~cēttiram, piccai from Sanskrit kṣetra, bhikṣā. Words can only end with /m, n, ɳ, l, ɭ, ɾ, ɻ, j/ in literary Tamil. Natively the only allowed clusters are C:, ṭP, NP, RP, RP:, RNP (where P = plosive, N = nasal, R = liquid, : = gemination, C = any consonant) most common are C: and NP; any heterorganic cluster indicates a morpheme boundary. Others clusters occurring in loanwords are split by vowels or simplified, e.g. varṣa > varuṭam.Zvelebil - Dravidian Linguistics - An Introduction (1990) pg. 12 In Sri Lankan Tamil dialects, word initial voiceless plosives may be aspirated dialectally, intervocalic /k/ may be [x~ɣ~h] and /t̪/ maybe [d̪~ð] or rarely [θ~t̪]; word final nasals are always preserved; <c> tends to be [t͡ʃ~ʃ~s] or rarely [t͡s] initially depending on the dialect and speaker, a rare [z] was noted for a speaker intervocalically, eg kaṭutāci [kɐɖuðaːzi]; <ṟṟ, nṟ> may be [tt~t̪t̪, nd~ɳɖ]. Some colloquial dialects merge ṇ, ḷ with n, l.Variability of Retroflex Perception and Production in Heritage Tamil Speakers, Kamala Muthukumarasamy, Graduate Program in Linguistics York University Toronto, Ontario August 2023 Kongu Tamil has word final /ŋ/ as word final /ŋkV/ becomes /ŋ/, e.g. literary Tamil vāṅka, Kongu Tamil vāṅ.
The voiceless consonants are voiced in different positions.
+ Tamil stop allophones Place Initial Geminate Medial Post-nasal Velar ~~~ Palatal , ~ Retroflex — ~ Alveolar — tːr r (d)r Dental ~ Labial ~