Turkish phonology

Consonants

/m, p, b/ are bilabial, whereas /f, v/ vary between bilabial and labiodental.

Some speakers realize /f/ as bilabial [ɸ] when it occurs before the rounded vowels /y, u, ø, o/ as well as (although to a lesser extent) word-finally after those rounded vowels. In other environments, it is labiodental [f].

The main allophone of /v/ is a voiced labiodental fricative [v]. Between two vowels (with at least one of them, usually the following one, being rounded), it is realized as a voiced bilabial approximant [β̞], whereas before or after a rounded vowel (but not between vowels), it is realized as a voiced bilabial fricative [β]. Some speakers have only one bilabial allophone.

/n, t, d, s, z/ are dental [n̪, t̪, d̪, s̪, z̪], /ɫ/ is velarized dental [ɫ̪], /ɾ/ is alveolar [ɾ], whereas /l/ is palatalized post-alveolar [l̠ʲ].

/n/ is realized as velar [ŋ] before /k, ɡ/, and palatal [ɲ] before /c, ɟ/.

/ɾ/ is frequently devoiced word-finally and before a voiceless consonant. According to one source, it is only realized as a modal tap [ɾ] intervocalically. Word-initially, a location /r/ is restricted from occurring in native words, the constriction at the alveolar ridge narrows sufficiently to create frication but without making full contact, [ɾ̞]; the same happens in word-final position: [ɾ̞̊].

/ɫ/ and /l/ are often also voiceless in the same environments (word-final and before voiceless consonants).

Syllable-initial /p, t, c, k/ are usually aspirated.

/t͡ʃ, d͡ʒ/ are affricates, not plosives. They have nevertheless been placed in the table in that manner to save space.

Final /h/ may be fronted to a voiceless velar fricative [x]. It may be fronted even further after front vowels, then tending towards a voiceless palatal fricative [ç].

In native Turkic words, the velar consonants /k, ɡ/ are palatalized to [c, ɟ] (similar to Russian) when adjacent to the front vowels /e, i, ø, y/. Similarly, the consonant /l/ is realized as a clear or light [l] next to front vowels (including word finally), and as a velarized [ɫ] next to the central and back vowels /a, ɯ, o, u/. These alternations are not indicated orthographically: the same letters ⟨k⟩, ⟨g⟩, and ⟨l⟩ are used for both pronunciations. In foreign borrowings and proper nouns, however, these distinct realizations of /k, ɡ, l/ are contrastive. In particular, [c, ɟ] and clear [l] are sometimes found in conjunction with the vowels [a] and [u]. This pronunciation can be indicated by adding a circumflex accent over the vowel: e.g. gâvur ('infidel'), mahkûm ('condemned'), lâzım ('necessary'), although this diacritic's usage has been increasingly archaic.

/b, d, d͡ʒ, ɡ, ɟ/ are devoiced to [p, t, t͡ʃ, k, c] word- and morpheme-finally, as well as before a consonant: /edˈmeɟ/ ('to do, to make') is pronounced [etˈmec]. (This is reflected in the orthography, so that it is spelled ⟨etmek⟩). When a vowel is added to nouns ending with postvocalic /ɡ/, it is lenited to ⟨ğ⟩ (see below); this is also reflected in the orthography.

In addition, there is a debatable phoneme, called yumuşak g ('soft g') and written ⟨ğ⟩, which only occurs after a vowel. It is sometimes transcribed /ɰ/ or /ɣ/. Between back vowels, it may be silent or sound like a bilabial glide. Between front vowels, it is either silent or has a [j] sound (e.g. düğün 'marriage', where the [j] is even mandatory in fast speech to distinguish it from dün 'yesterday'), depending on the preceding and following vowels. When not between vowels (that is, word finally and before a consonant), it is generally realized as vowel length, lengthening the preceding vowel, or as a slight [j] if preceded by a front vowel.

According to Zimmer & Orgun (1999), who transcribe this sound as /ɣ/:

Word-finally and preconsonantally, it lengthens the preceding vowel.

Between front vowels it is an approximant, either front-velar [ɰ̟] or palatal [j].

Otherwise, intervocalic /ɣ/ is phonetically zero (deleted).

Before the loss of this sound, Turkish did not allow vowel sequences in native words, and today the letter ⟨ğ⟩ serves largely to indicate vowel length and vowel sequences where /ɰ/ once occurred.

Phonotactics

Turkish phonotactics are very simple, and it can be described as CV(C)². Although Turkish words can take two final consonants, the possibilities are limited.

Turkish only allows complex onsets in a few recent English, French and Italian loanwords; such as Fransa, plan, program, propaganda, strateji, stres, steril and tren. Even in these words, the complex onsets are only pronounced as such in very careful speech.

Some loanwords add a vowel in front of them to break the complex onset; for example the French station was borrowed as istasyon to Turkish.

All syllables have a nucleus

No diphthongs (/j/ is always treated as a consonant)

No onset /ɰ/

No complex onsets (except for the exceptions above)

In a complex coda:

The first consonant is either a voiceless fricative , /ɾ/ or /l/

The second consonant must be a voiceless plosive

Two adjacent plosives and fricatives must share voicing, even when not in the same syllable

Consonant assimilation

Because of assimilation, an initial voiced consonant of a suffix is devoiced when the word it is attached to ends in a voiceless consonant. For example,

the locative of şev (slope) is şevde (on the slope), but şef (chef) has locative şefte;

the diminutive of ad (name) is adcık (little name), but at (horse) has diminutive atçık (little horse).

Vowels

The vowels of the Turkish language are, in their alphabetical order, ⟨a⟩, ⟨e⟩, ⟨ı⟩, ⟨i⟩, ⟨o⟩, ⟨ö⟩, ⟨u⟩, ⟨ü⟩. There are no diphthongs in Turkish and when two vowels come together, which only occurs in some loanwords, each vowel retains its individual sound. (e.g. aile [a.i.le], laik [la.ic])

/e, o, ø/ are phonetically mid [e̞, o̞, ø̞]. For simplicity, this article omits the lowering diacritic even in phonetic transcription.

Most speakers lower /e/ to [æ] before the coda /m, n, l, r/, so that perende 'somersault' is pronounced [perænˈdɛ]. There are a limited number of words, such as kendi 'self' and hem 'both', which are pronounced with [æ] by some people and with [e] by some others.

Some scholars transcribe /ø/ as /œ/. This article uses the former symbol for simplicity, and because /ø/ has the same height as the other mid vowels /e/ and /o/.

/a/ has been variously described as central [ä] and back [ɑ]. For simplicity, this article uses the diacriticless symbol ⟨a⟩, even in phonetic transcription. /a/ is phonologically a back vowel, because it patterns with other back vowels in harmonic processes and the alternation of adjacent consonants (see above). The vowel /e/ plays the role as the "front" analog of /a/.

/i, y, ɯ, u, e, ø/ (but not /o, a/) are lowered to [ɪ, ʏ, ɯ̞, ʊ, ɛ, œ] in environments variously described as "final open syllable of a phrase" and "word-final".

Vowel harmony

With some exceptions, native Turkish words incorporate either exclusively back vowels (/a, ɯ, o, u/) or exclusively front vowels (/e, i, ø, y/), as, for example, in the words karanlıktaydılar ('they were in the dark') and düşünceliliklerinden ('due to their thoughtfulness'). /o ø/ only occur in the initial syllable.

The Turkish vowel system can be considered as being three-dimensional, where vowels are characterised by three features: front/back, rounded/unrounded, and high/low, resulting in eight possible combinations, each corresponding to one Turkish vowel, as shown in the table.

Vowel harmony of grammatical suffixes is realized through "a chameleon-like quality", meaning that the vowels of suffixes change to harmonize with the vowel of the preceding syllable. According to the changeable vowel, there are two patterns:

twofold (/e/~/a/): Frontality is preserved, that is, /e/ appears following a front vowel and /a/ appears following a back vowel. For example, the locative suffix is -de after front vowels and -da after back vowels. The notation -de² is shorthand for this pattern.

fourfold (/i/~/y/~/ɯ/~/u/): Both frontality and rounding are preserved. For example, the genitive suffix is -in after unrounded front vowels, -ün after rounded front vowels, -ın after unrounded back vowels, and -un after rounded back vowels. The notation -in⁴ can be this pattern's shorthand.

The vowel /ø/ does not occur in grammatical suffixes. In the isolated case of /o/ in the verbal progressive suffix -i⁴yor it is immutable, breaking the vowel harmony such as in yürüyor ('[he/she/it] is walking').

Some examples illustrating the use of vowel harmony in Turkish with the copula -dir⁴ ('[he/she/it] is'):

Türkiye'dir ('it is Turkey') – with an apostrophe because Türkiye is a proper noun.

gündür ('it is the day')

kapıdır ('it is the door')

paltodur ('it is the coat').

Compound words do not undergo vowel harmony in their constituent words as in bugün ('today'; from bu, 'this', and gün, 'day') and başkent ('capital'; from baş, 'prime', and kent, 'city').

Vowel harmony does not usually apply to loanwords and some invariant and irregular suffixes, such as -ki ('belonging to ...') and -ken ('while ...-ing'). In the suffix -e²bil ('may' or 'can'), only the first vowel undergoes vowel harmony. There are a few native Turkish words that do not have vowel harmony such as anne ('mother'). In such words, suffixes harmonize with the final vowel as in annedir ('she is a mother').

Suffixes added to foreign borrowings and proper nouns usually harmonize their vowel with the syllable immediately preceding the suffix: Amsterdam'da ('in Amsterdam'), Paris'te ('in Paris').

Consonantal effects

In most words, consonants are neutral or transparent and have no effect on vowel harmony. In borrowed vocabulary, however, back vowel harmony can be interrupted by the presence of a "front" (i.e. coronal or labial) consonant, and in rarer cases, front vowel harmony can be reversed by the presence of a "back" consonant.

For example, Arabic and French loanwords containing back vowels may nevertheless end in a clear [l] instead of a velarized [ɫ]. Harmonizing suffixes added to such words contain front vowels. The table on the right gives some examples.

Arabic loanwords ending in ⟨k⟩ usually take front-vowel suffixes if the origin is kāf, but back-vowel suffixes if the origin is qāf: e.g. idrak-i ('perception' acc. from إدراك idrāk) vs. fevk-ı ('top' acc. from ← فوق fawq). Loanwords ending in ⟨at⟩ derived from Arabic tāʾ marbūṭa take front-vowel suffixes: e.g. saat-e ('hour' dat. from ساعة sāʿa^t), seyahat-e ('trip' dat. from سياحة siyāḥa^t). Words ending in ⟨at⟩ derived from the Arabic feminine plural ending -āt or from devoicing of Arabic dāl take the expected back-vowel suffixes: e.g. edebiyat-ı ('literature' acc. from أدبيّات adabiyyāt), maksat, maksadı ('purpose', nom. and acc. from مقصد maqṣad).

Front-vowel suffixes are also used with many Arabic monosyllables containing ⟨a⟩ followed by two consonants, the second of which is a front consonant: e.g. harfi ('letter' acc.), harp/harbi ('war', nom. and acc.). Some combinations of consonants give rise to vowel insertion, and in these cases the epenthetic vowel may also be front vowel: e.g. vakit ('time') and vakti ('time' acc.) from وقت waqt; fikir ('idea') and fikri (acc.) from فِكْر fikr.

There is a tendency to eliminate these exceptional consonantal effects and to apply vowel harmony more regularly, especially for frequent words and those whose foreign origin is not apparent. For example, the words rahat ('comfort') and sanat ('art') take back-vowel suffixes, even though they derive from Arabic tāʾ marbūṭa.

Stress

Main stress occurs regularly on the last syllable of a word, except for forms including suffixes with inherent stress, adverbs, proper names, and some loanwords (particularly from Italian and Greek) such as masa /ˈmasa/ ('table, desk'), lokanta /loˈkanta/ ('restaurant'), and iskele /isˈcele/ ('pier'). The lexical exceptions in Turkish stress have been important to linguistic theories of how phonological exceptions should be represented grammatically.

Regular final stress

As stated above, word-final stress is the regular pattern in Turkish:

The metrical weight of a syllable in terms of moras has no effect on the placement of stress in the regular pattern. Light (L) syllables in Turkish are open syllables (V or CV) which consist of a single mora while heavy (H) syllables have consonantal codas (VC or CVC) and consist of two moras.

The regular pattern persists in derived words as well. (See: stress and suffixation section.)

Sezer stress

Proper names (of both places and foreign people) follow a different stress pattern, known in the linguistics literature as Sezer stress (after the discoverer of the pattern, Engin Sezer). In this lexical domain, stress occurs on the antepenult if the penult is light and the antepenult is heavy, and otherwise on the penult. The weight of the final syllable is irrelevant.

Penultimate stress:

Antepenultimate in …HLσ words:

The Sezer stressed form /aˈda.na/ would be expected to have the unattested form */a.daˈna/ under the regular stress pattern. Thus, it can be seen that the regular and the Sezer pattern are contrastive.

The Sezer stress pattern is productive in spite of it being observed on a smaller set of lexical items. Suffixed words that have the regular pattern can shift to the lexical class of placenames (via zero-derivation). When these words are used as placenames, the regular stress pattern shifts to the Sezer pattern. For instance, the word /toɾ.baˈɫɯ/ torbalı ('with a/the bag') has regular stress in its normal use, but when a placename it has Sezer stress /ˈtoɾ.ba.ɫɯ/.

The Sezer stress pattern is completely regular, including for loanwords whose source language version has a different stress pattern. That is, source stress is not preserved in Turkish. For example, the English word Arkansas has antepenultimate stress (i.e. /ˈar.kən.sɔː/), but the loanword in Turkish has penultimate stress.

One approach to the metrical analysis of the Sezer pattern posits a general disyllabic iambic rhythm that is aligned with the right word edge with a restriction against having a nonfinal foot (or alternately requiring an extrametrical final syllable) and a requirement that heavy syllables carry stress (weight-to-stress). Thus:

The words with antepenultimate stress have a rhythmic reversal to a trochee to prevent a heavy antepenultimate syllable from not being stressed, that is an illicit *(H'L)σ form:

Stress and suffixation

Turkish is usually considered a syllable-timed language. Stressed and unstressed syllables do not differ greatly. Pitch and stress are very important in Turkish. The regular stress pattern occurs on words with a stem combined with suffixes. Here the stress is consistently word-final and appears to shift rightward away from the stem as suffixes are concatenated.

The above is not the case in stems with Sezer stress. Stems with Sezer stress retain the main stress of the underived form:

(Turkish orthography requires an apostrophe between proper nouns and attached suffixes.)

Adverbs do not generally take final stress:

néreye? nérede? ('whereto?', 'where?')

Words ending with a personal predicative suffix are generally stressed on the preceding syllable. This stress pattern can be useful in disambiguating homographic words containing possessive suffixes or the plural suffix:

Other suffixes which do not take stress are the interrogative and negative suffixes mi and ma, and the adverbial and adjectival suffixes le and ce:

Geldi̇́ mi? ('Did he/she/it came?') Yápma! ('Don't do/make [it]!') bu surétle ('in this way') Tǘrkçe ('Turkish') yálnız ('just/only'; cf. yalníz, 'alone')

On the other hand, the verbal tense/aspect/mood morpheme is usually stressed:

geliyórum ('I am coming') gidérsin ('you[sing.] go')

In negated verbs, the stress typically falls on the syllable preceding the negation morpheme:

gélmedim ('I didn't came') istémiyorum ('I am not wanting [to/it]')

In compounds, the first compound element retains its stress (prior to compounding) while the second element loses its stress.

There are some exceptions to these stress rules, including diminutives, which have word-initial trochee (= initial stress). In addition, -en/-an adverbs differ in that they have nonfinal right-aligned trochee weight-to-stress (i.e. stress H penult, else: stress antepenult):