So, lately I've been really interested in devoicing voiced consonants to their counterparts. I've been studying painstakingly, so I would like to share it with you, and you can tell me if you agree with me or not.

So, I'll start:

1. I have two bookshelves. In this case the word 'have' is pronounced with /v/ sound. However, due to the fact that the next word starts with /t/ sound, it becomes /f/ sound.

'Bookshelves' is pronounced with /v/ sound and ends with /z/ sound, however, due to the fact that it is at the end of a sentence, the /v/ and the /z/ sounds will transform into a very but very ultra week /f/ and /s/ sound.

So the sentence /aɪ hæv tu ˈbʊkˌʃɛlvz/ actually becomes /aɪ hæf tu ˈˈbʊkˌʃɛlfs/ (It's how Google translator pronounces it)

2. Vegetable -> /ˈvɛʤtəbəl/ pronounced with /ʤ/, however, if you try to pronounce 'vegetable' with /ʤ/ sound, it is possible but it's hard, and not only that, but it sounds unnatural.

So, due to the fact that the next sound is /t/ sound, it means that the word 'vegetable' will actually become /vɛʧtəbəl/. The /ʤ/ sound will convert into its voiceless counterpart /ʧ/. Google translate

3. The phrase you've changed a lot. The /ʧ/ sound is actually a mix between the /t/ sound and the /ʧ/ sound. So it means that the preceding sound (/v/) will be devoiced and will become a weak /f/ sound rather than a hard /v/ sound.

Not only that, but it seems like the /ʤ/ sound in 'changed' will become a very weak /ʧ/. So you've changed a lot (/juv ʧeɪnʤd ə lɑt/) becomes /juf ʧeɪnʧt ə lɑt/. Not only that, but it seems like the /d/ in 'changed' will become more of a /t/ sound. Keep in mind, they're becoming very weak counterparts. Here is the audio.

4. He's trying. The same will happen here. He's is normally pronounced with /z/ sound, however, due to the fact that the next word starts with /t/ sound, it will actually become more of a /s/ sound. And it's more common to pronounce the /tr/ cluster as /ʧ/ sound, so /hiz ˈtraɪɪŋ/ will become more like /hisˈtraɪɪŋ/. Audio

5. I moved, same thing happens here. 'Moved' is at the end of a sentence, so it means that the /v/ sound in 'moved' will become a very ultra light /f/ sound.

The /t/ and /ʧ/ sounds seem to devoixe many sounds.

'You move to me', 'I have to go', 'I've been engaged twice' (the /ʤ/ sound will become more of a /ʧ/ sound)

It also happens at the end of words. Never in my wildest dreams -> The /z/ in dreams will actually be a very ultra light /s/ sound, so light that it will sound like a little whispery /s/ sound.

Same happens with 'I have a page' -> the /ʤ/ sound will become very ultra /ʧ/ sound, since it's at the end.

However, if you say 'In my dreams I see you', then the /z/ sound will be strong.

If you say 'I'm gonna change my name', then the /ʤ/ sound in 'change' will be strong.

'I'm gonna move again', in this sentence the /v/ sound will remain and it won't change to /f/ sound.

One last more: It won't change to /f/ sound. In change to, the /ʤ/ sound will actually change to /ʧ/ since the /t/ sound devoices it.

Same with, I have a garage, when it's at the end it seems it becomes a very ultra light /ʧ/ sound instead of the /ʒ/ sound. But if you say 'my garage is big', then /ʒ/ will not change to /ʧ/ and will remain strong.

  • 1
    Just briefly because I have to run now, some of this is accurate but some of it seems not quite right. You’re venturing into the realm not of phonemes but rather of phonetic allophones under various assimilatory effects, something that almost by definition native speakers listening to utterances normally ignore in their mental model of which word was said based on its phonemes. If you are a native Spanish speaker, how much attention do you give the voicing of the unvoiced phoneme /s/ to [z] in words like mismo and desde under regressive assimilation of the voiced consonant following /s/?
    – tchrist
    Jan 25, 2019 at 21:26
  • 1
    Note that in a case like bookshelves, even if /vz/ does end up getting devoiced to [fs], the main phonetic hint to a native English speaker that it is a phonemically voiced cluster is the length of the preceding vowel (or vowel + sonorant), which does not change. So there is still a clear, audible difference between shelf’s [ʃɛ̆ɫfs] and shelves [ʃɛɫˑvz ~ ʃɛɫˑfs]. Jan 27, 2019 at 12:21

1 Answer 1


This is more or less correct, but in most of your examples a native English speaker doesn’t hear these as their voiceless counterparts because the preceding vowel is longer than it would be if the consonant was its "voiceless" counterpart.

“I have two” = [a͡ɪː hæːf tu]; it’s not pronounced the same as “I haff to”, where the middle vowel would be short (or “clipped”).

English "voiced" plosives /b d g/ (and the affricate /d͡ʒ/) tend to be most strongly voiced when they are between two other voiced sounds—vowels, semivowels, or sonorant consonants such as /l r m n/—and least strongly voiced when they are next to voiceless obstruents such as /p t k tʃ s ʃ f θ/. (I think it is impossible for "voiced" plosives to be surrounded on both sides by voiceless obstruents.) However, the exact degree of voicing given to these sounds varies between speakers. The key things that differentiate /b d g/ from /p t k/ are that /p t k/ are aspirated at the start of a stressed syllable, and that vowels are "clipped"--which means that they are phonetically shorter--before a "voiceless" phoneme.

Transcription notes

Some people don’t like to transcribe the consonant found in contexts like "have two" with the symbol “f” because even though it is voiceless (or mostly voiceless), it is also “lenis”, a vague word that I think in the context of voiceless fricatives is mostly related to the length of friction ("lenis" fricatives are shorter than "fortis" ones). Rather, they transcribe it with the "v" symbol along with a "voiceless" ring diacritic: [v̊]. So "I have two" would be something like "[a͡ɪː hæːv̊ tu].

The difference in pronunciation between "vegetable" and a hypothetical "vetchtable" isn't as obvious to me, maybe because this is a polysyllabic word, or maybe because the consonant involved is /d͡ʒ/. Nonetheless, I would recommend pronouncing vegetable as [ˈvɛːd̥͡ʒ̊təbl̩] rather than as [ˈvɛt͡ʃtəbl̩].

Special circumstances where a voiced fricative is truly replaced with its voiceless counterpart

There are a few special phrases where a historically voiced fricative is truly turned into its "voiceless" counterpart phoneme. One of these is "have to", in the sense "must". This is pronounced the same as "haff to" for many speakers, with /f/ instead of /v/. We also see this with /s/ instead of /z/ in "has to", "used to" and "supposed to". And I've heard that this can occur for some speakers inside some historically compound words, such as "newspaper": Merriam-Webster lists a pronunciation with /s/ as well as a pronunciation with /z/. But these are exceptions, not the rule.

  • Here m.youtube.com/watch?v=WPeDfu8cgs4 at minutes four of the video she says I have to go. The V becomes an f sound and she explains why. I'm confused...yes Jan 25, 2019 at 22:18
  • I see. I understand. The length thing. Save vs safe. Save will be a little bit longer since it ends in the voices v sound while safe will be shorter. Same with badge vs batch. Badge the ae will be longer since it ends in a voiced sound and the ending is weaker. While batch the ae vowel is shorter and the ending is a little.bit longer since it ends in an unvoiced sound. Jan 25, 2019 at 22:24
  • But still. When you say. I have to go. Do you make that V an F sound? Jan 25, 2019 at 22:24
  • Would you pronounce "vegetable" with the 'CH' sound, instead of the 'J' sound? Jan 26, 2019 at 1:53
  • 1
    @tchrist: I'm not sure that the schwa here would be epenthetic; I think that the three-syllable pronunciation is historically the result of syncope
    – herisson
    Jan 27, 2019 at 3:35

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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