Final Devoicing in Citation Forms

Once you have seen kitap → kitabı a few times, an obvious question arises: if the stem is "really" kitab-, why does the dictionary list it as kitap with a p? The answer is the deepest of the three consonant rules, and the one that explains the other two: Turkish does not allow a voiced obstruent at the very end of a word, so any underlyingly voiced final stop is devoiced in the bare citation form. Kitap is not the true stem — it is the disguised, word-final version of kitab-. Understanding this flips your whole mental model: the inflected form is the honest one, and the dictionary form is the one that has been altered. (For the softening this exposes, see p → b, ç → c, t → d; for the sounds themselves, the consonant overview.)

The rule: no voiced stop at the end of a word

Turkish phonology bans voiced obstruents — chiefly the stops b, c, d, g and their soft partners — from the final position of a word standing alone. When a stem that "wants" to end in one of these is forced to the word edge with no suffix, it surfaces with the matching voiceless partner instead:

The mechanism is the same in every row: at the word edge the stop devoices (b → p, d → t, c → ç, g → k); add a vowel suffix and the consonant is no longer final, the ban lifts, and the true voiced consonant comes back. This is why the "softening" of the previous pages is better described as revealing — the voiced consonant was there all along, merely hidden.

In each sentence the same noun appears twice: once bare and devoiced (kitap, kâğıt), once with a vowel suffix and voiced (kitabı, kâğıdın). Seeing the pair in one breath is the fastest way to feel that they are one word with two faces.

Why this matters: the citation form hides whether a word softens

Here is the practical payoff, and it is significant. Because devoicing hits the citation form, the dictionary entry alone cannot tell you whether a word has a voiced stem or a genuinely voiceless one. Two words can be spelled with the same final letter and behave oppositely:

At is "honestly" voiceless — its true stem is at-, so the accusative is atı with the t intact. Kanat is "dishonestly" voiceless — its true stem is kanad-, devoiced at the edge, so the accusative is kanadı. From the bare forms at and kanat you cannot predict this. The only reliable witness is a vowel-suffixed form, which strips off the devoicing and shows the stem's true colours.

The lesson is concrete: learn nouns in pairs — citation form plus accusative or possessive — from the very beginning.

Once kitap / kitabı and at / atı live in your memory as pairs, you never have to guess whether a stop softens — the pair has already told you. (The lexical question of which words have voiced stems is catalogued on the non-softeners discussion.)

The same thing happens inside verbs

Final devoicing is not a noun-only quirk; it shapes verb stems too, and ignoring it produces audibly wrong conjugations. The verb git- ("go") has the underlyingly voiced stem gid-. Before a consonant suffix the stem is at the edge and devoices to git-; before a vowel suffix the d returns.

Git-mek (infinitive, t before the consonant suffix -mek) but gid-iyor ("is going") and gid-emiyor ("can't go") once a vowel follows. The same alternation hits et- → ed-iyor ("doing"), tat- → tad-ı ("its taste"). These verb-stem shifts have their own treatment on the verb stem changes page, but the cause is identical to the noun case: a voiced stem disguised by final devoicing.

A contrasting case: a genuinely voiced final letter

Not every word ending in a voiceless-looking letter is hiding a voiced stem, and — importantly — some words really do end in a voiced stop, which is allowed for a few letters. The word ad ("name") keeps its d because here the d is the true, stable final consonant in normal usage:

Ad → adı, adını — the d is steady. Note the contrast with at → atı: adı (from ad, "its name") and atı (from at, "the horse") differ by exactly the voicing of that consonant, and getting it wrong swaps the meaning. This is the clearest possible argument for learning the inflected pair: only the suffixed form distinguishes a true voiced stem from a devoiced one.

Spelling: the devoiced letter is what you write

A final practical reassurance: Turkish spelling keeps the devoiced final letter. You write kitap, kâğıt, git, ağaç — the surface, devoiced forms — not the abstract underlying kitab, kâğıd, gid, ağac. The underlying voiced stem is a mental model for predicting suffixed forms, not a spelling you ever produce on its own. So the rule to hold is: write the citation form as devoiced, but expect the voiced stem to reappear the moment a vowel suffix lands.

Common mistakes

In isolation, the final stop devoices and that is what you write: kitap, not kitab.

Kanat hides a voiced stem kanad-; before a vowel the true d returns: kanadı.

The verb stem is really gid-; before the vowel of -iyor the d comes back: gidiyorum.

"Name" is ad → adı (true voiced d); atı would mean "the horse." Only the inflected form keeps them apart.

Key takeaways

• Turkish devoices voiced final stops in isolation, so the citation form hides the true stem: kitap disguises kitab-, git disguises gid-.
• A vowel suffix removes the word-final position and lets the true voiced consonant resurface (kitabı, gidiyor).
• The dictionary form alone cannot tell you whether a word softens — at and kanat look alike but behave oppositely.
• Learn every p/ç/t/k noun in a pair (citation + accusative) so the true stem is never a guess.
• Spelling keeps the devoiced letter: you write kitap, git, ağaç, and let the voiced stem return only under suffixation.