This happens in many languages. For example, in many languages the letter d transliterates to IPA's t if it is the last letter of the word (instead of d). Consider, in English the "d" in "demon" and in "baked". We sort of attach a default phoneme to the letter, and then there are exceptions where the letter is pronounced differently.

Is there a word for this phenomenon?

  • Are you asking about phonology or phonetics?
    – fdb
    Oct 25, 2017 at 12:53
  • 1
    For the specific phenomenon you describe, there is the term Auslautverhärtung (aka final devoicing). Oct 26, 2017 at 12:13

4 Answers 4


You seem to be talking about "neutralization". For example in German and Russian, word-final d and t are both pronounced [t]. We have a version of that in English, where the past tense d is pronounced [t] after certain sounds (as in "baked, based, laughed"). To make any sense of the English pattern, especially, you have to not think in terms of "letters", and instead think in terms of the sounds that the letter-sequences represent (somewhat obscurely in English). Given some two similar but distinct sounds of a language (like t,d) the language may have some rule which eliminates (neutralizes) the phonetic difference between those two sounds so that they are pronounced the same way. It could be that d becomes pronounced like t, or t becomes pronounced like d, or both are pronounced the same way like [ɾ], for example English write, ride having the contrast between t and d but writer, rider being pronounced the same way, in most American dialects, with [ɾ] in place of t,d.


The pronunciation of "baked" is actually a special case in English.

English phonotactics do not allow /t/ and /d/ to contrast at the end of a word when there is a preceding voiceless obstruent (a sound like /k/, /p/, /f/, /s/), so the suffix spelled "-ed" is realized as /t/ in this context (and /d/ in other contexts). This is a type of neutralization, as described by user6726, and the realization of the neutralized segment as a voiceless sound can be explained as the result of "assimilation" in voicelessness to the preceding voiceless phoneme.

But /t/ and /d/ do contrast at the end of a word in other contexts in English: there is a clear difference in pronunciation, easily heard for a native English speaker, between words like "meld" and "melt", or "lend" and "Lent".

On the other hand, it is well known that in some European languages, such as standard German and Dutch, there is a "neutralization" between voiced and voiceless obstruents word-finally no matter what the preceding sound is. A typical example is standard German "Rad" and "Rat": they are considered to be homophones.

This is often treated as the result of a phonological rule that causes consonants to be recategorized as "devoiced" or "fortis" or something like that at the end of a word, although in some languages, such as Dutch or Russian, the situation is complicated by the fact that the neutralized sound may in fact be voiced, not voiceless, if it occurs in the middle of a phrase before a voiced obstruent (in the onset of the first syllable of the next word). I found a paper that talks about this phenomenon in Dutch: "Final Devoicing and Voicing Assimilation in Dutch Derivation and Cliticization," Janet Grijzenhout & Martin Krämer

Another complicating factor is that apparently, in these languages, word-final neutralization of voicing contrasts is often "incomplete" in the sense that careful measurements may reveal some kind of differences in production that are just too small to be perceived by human listeners. See e.g. "The Nature of Incomplete Neutralization in German: Implications for Laboratory Phonology," Bodo Winter and Timo Röttger. This seemed very surprising to me when I first read about it, but I believe I have read about similar cases of a mismatch between production and perception of word-initial or word-final geminate/long consonants in some languages. It's not obvious to me if this is relevant to a theoretical analysis of the phenomenon of devoicing.

  • 2
    I don't know that the /d/ > /t/ change after a voiceless obstruent has to do with word boundary -- it may instead happen at the end of a syllable.
    – Greg Lee
    Oct 25, 2017 at 22:07

It is one special case of what the Prague School called the demarcative function of language: "The term demarcative is applied to those elements or features that in particular languages serve to indicate the occurrence of the boundaries of words and phrases and, presumably, make it easier to identify such grammatical units in the stream of speech."

(Neutralization, referred to by @user6726, is also a Prague School notion.)


Final obstruent devoicing—where a lenis consonant is systematically and cumpulsorily replaced by its fortis counterpart word finally and before fortis consonants—is a feature of many languages, for example Catalan, German or Turkish. This is a case of one phoneme being replaced by another.

Notice that this is a phonological rule and is very different from the phonetic devoicing undergone by lenis consonants, in English for example, when they occur next to unvoiced sounds or next to silence. These consonants remain lenis, even when devoiced and the devoicing can in no way be thought of as being 'compulsory'. When said in isolation the [d] in bed is likely to be devoiced, but will nonetheless remain a [d].

  • I guess your "phonetic devoicing" is what I would call an allophonic phonological rule and your "phonological rule" is what I would call a morphophonemic phonological rule. The devoicing of d in baked is morphophonemic -- that is, /d/ becomes (fortis) /t/.
    – Greg Lee
    Oct 26, 2017 at 22:38
  • @GregLee Well,tthe phonetic devoicing I was talking about is what devoices a [d] in the word bid when said in isolation. It isn't a rule. But I understand what you are talking about above (it isn't the same thing though). The devoiced [d] in bed is not a /t/. Oct 26, 2017 at 22:44
  • The only things wrong with your phonetic devoicing example are (1) it invites confusion, since the original question asked about /t/ in baked, (2) your wording suggests it is not "phonological".
    – Greg Lee
    Oct 27, 2017 at 2:10

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.