3

It seems to me that both /dʒ/ and /ʒ/ become voiceless (or almost) when they occur in word final position. Is this true?

Examples:

age, wage, courage, judge

garage, sabotage, collage, mirage

Does this happen due to the process of devoicing? If so, and as a result of this process, do the example words above lose all of their vocal fold vibration, or do they only lose some of it?

If the words above are only partially devoiced (don't lose all of their vocal fold vibration), how should they be pronounced in consequence? Can they be compared to a musical note that has to be played piano or pianissimo?

10
  • Please specify whether you have a particular brand of English in mind. The Brits pronounce garage "GAR-ridge." Is the second of the three consonants in the word "exit" voiced or unvoiced? Is it played pianissimo or forte? I'm thoroughly confused now. I don't remember ever hearing anyone saying "couraitch" or "mirash." Please clarify.
    – Ricky
    Nov 21, 2015 at 11:22
  • 2
    It's certainly true that English speaker distinguish between the words etch and edge, and batch and badge. Nov 21, 2015 at 12:12
  • 1
    That would depend on one's financial situation, the quality of drinks at one's pub, and one's willingness to consume those.
    – Ricky
    Nov 21, 2015 at 12:19
  • 1
    @Ricky My answer might help clear that up. It'll tend to be voiced or unvoiced according to the following sound. It's very difficult for us to tell just by listening to the sound at the end of the word whether it's voiced or not. Our brains tell us that if the preceding vowel is long then it's voiced (we pay no attention to the actual consonant itself). :) Nov 21, 2015 at 15:34
  • 2
    @Mitch I think that would be the clipping of the vowel that's tricking your brain. It will certainly feel like it's a voiced sound. So these words will all still be clearly distinguishable. That's the theory/science anyhow :) Nov 21, 2015 at 22:10

2 Answers 2

4

Voiced obstruents - the plosives /b, d, g/; the affricate /dʒ/; the fricatives /v, ð, z/ and /ʒ/ - all undergo devoicing when adjacent to unvoiced sounds, including when next to silence.

They may sometimes be fully devoiced all the way through. However, they tend to be devoiced at the beginning of the consonant if preceded by an unvoiced sound or at the end if followed by an unvoiced sound. In the parametric diagram below the line for vocal fold vibration (voicing) is flat when there's no voicing and wavy when there is.

enter image description here

For the word bond said in isolation, we can see that the beginning of the /b/ and the end of the /d/ are both devoiced. In a more detailed parametric diagram, we might put the little circle currently under the b and the d as a superscript to the right or left of the consonant to indicate which side is devoiced.

If the consonant is at the end of a word but followed by a voiced sound there will be no devoicing. This type of devoicing is an assimilatory effect, meaning that the consonant is taking on qualities from the sounds next to it. If there is no voiceless sound next to the consonant, it won't normally become devoiced.

Have a look at the waveform below for the words news items taken from a BBC radio news broadcast:

enter image description here

Here, the dark black fuzz that we see at 3 and 4 (the red numbers on the diagram) shows voiceless sound. The kind of wavy (periodic) line that we can see from the beginning of the word up till the beginning of 2, is voiced sound. That kind of regular repeating pattern there is what gives us the impression that a sound has musical pitch. This, essentially, is what voicing is. Now the transcription for news item is like this: /njuz aɪtəmz/. Notice that there are two /z/s here/. The first is in between two vowel sounds. The second, at the end of the second word, precedes silence. Now, if you look at 1 on the the diagram you will see that the first z is voiced all the way through. That wavy pattern carries on going all the way through the z section. However, if you look at 4, which represents the z at the end of the word next to silence, you will see that it is a messy smudge of black. This is because it is nearly entirely devoiced.

In short then, a /ʒ/ at the end of a word will become devoiced when not followed by another voiced sound. So will a voiced obstruent at the beginning of a word too, if not preceded by another voiced sound. (Only one word in English begins with the consonant /ʒ/ - the word genre).

Note to learners

It is not necessary to try and artificially recreate these types of devoicing effects. In fact, it will be harmful for your English. English speakers don't listen to (and cannot hear, unless they are specially trained) whether a consonant has voicing or not. We understand whether a sound is a /z/ or an /s/ because of other sounds in the word, not because of the /z/ or /s/ itself. An /s/ at the end of a syllable will cause the vowel to be shortened (this will happen in almost any language). This does not happen with /z/ - whether or not the /z/ has become devoiced. If you try and substitute in an unvoiced /z/, you will end up putting in an /s/ instead, which will affect the vowel and may lead to misunderstandings - and a very strange accent. Just use a normal /z/ and the rest will take care of itself.


Notes:

The technical term for consonants that are usually voiced is lenis. The technical terms for sounds that are usually unvoiced is fortis. I haven't used these terms here, because there was enough new jargon already for unfamiliar readers.


References

The parametric diagram is from the Speech Internet Dictionary.

The waveform is from Swphonectics.com. Courtesy of both Sydney Wood and SWPhonetics.

Both were accessed on 21 November 2015.

5
  • 3
    The vowel length certainly plays a large part in decoding the phoneme. But I would think the voiced/unvoiced quality must also play some role in our decoding the phoneme. Otherwise, Americans wouldn't have any more trouble that Brits distinguishing bidder and bitter (and we do). Nov 21, 2015 at 21:19
  • 1
    Dude, you're scaring me. Actually, it's a very good point: we're conditioned to detect key sounds only, and simply assume the rest. On a larger scale, our brains are also conditioned to look for key words only in a sentence, and form an opinion about the rest without listening to it. Which leads me to believe that we expect others to speak exclusively in stock phrases, and those others expect us to do the same. The eerie part here is that there's no room for actual thinking in this picture: it's all just reflexes. Now that's reassuring.
    – Ricky
    Nov 21, 2015 at 21:26
  • @PeterShor I believe that the reason that the US one is more likely to cause problems there is that there's a lack of aspiration which Brits would get following the /t/. Because aspiration is basically a quality of the vowel following the /t/ before the voicing kicks in there's no aspiration after a voiced /t/. So again, weirdly, we're seeing an effect on neighbouring EDIT: vowels. What I do know is that in BE at least, you can change people's perception of a voiced sound by chopping the vowel before it ... Nov 21, 2015 at 21:28
  • @PeterShor However, I'm not an expert on Gen Am. It would be interesting to see what the actual data shows ... (for those speakers who use a voiced allophone of /t/ intervocalically). Nov 21, 2015 at 21:29
  • @Ricky Yeah, it's kind of weird but wonderful, ain't it! Nov 22, 2015 at 2:52
2

... how should they be pronounced... Can they be compared to a musical note that has to be played piano or pianissimo?

The final consonants in those words are all voiced. The voicing tends to be accompanied by a prolonged preceding vowel and unvoiced consonants by a shortened preceding vowel. Concentrate on that vowel, not on the consonants.

I heard Angela Merkel the other day use the English word "fans" (which I suppose has become a loan-word in German). She pronounced it "fence". It's a good example of how the vowel is affected in English:

faaanz

fence

P.S. So, to take up your analogy, imagine the choir director telling the singers to hold the note which corresponds to the vowel and then in your mind's eye see the director bringing his two fingers and thumb together gently, to reflect how the "voiced obstruent" closes off the sound unabruptly.

5
  • @That's a bit of a wide-spread misconception. The voiced consonant doesn't lengthen the preceding vowel, exactly; it's that unvoiced (i.e. fortis) consonants cause the vowel to be clipped (i.e. reduced). So the /i:/ in bead is ever so slightly shorter than the /i:/ in bee. The /i:/ in beat however is about half the length of the /i:/ in bead. :) Nov 21, 2015 at 13:58
  • The vowel in my bead is longer than the vowel in my bee. But I agree about beat. I've reworded to remove the implication of causality. The vowel in my bees is longer than the vowel in my bee.
    – TRomano
    Nov 21, 2015 at 14:08
  • Do you have a slightly palatal offglide on bee but not on be?
    – TRomano
    Nov 21, 2015 at 14:15
  • This fella knows how to compromise: "Because of the influence of voicing on vowels —they have shorter duration before a voiceless obstruent— English has a cline of four durational patterns illustrated in bead, beat, bid, bit." ling.ohio-state.edu/~odden/Vowel%20Length.pdf
    – TRomano
    Nov 21, 2015 at 14:58
  • Sorry for the delay, have been doing an answer for this question. I think bee and stressed be are the same for me, but its difficult to tell if that's how I actually normally talk. Yes that sounds like a good description from OS. Nov 21, 2015 at 15:21

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.