I know the rule of pronouncing s at the end of words(plural nouns and singular verbs) that if s follows a voice sound(d, l, etc.), it will be pronounced as /z/ sound. For example,

"words" is pronounced as word/z/.

However, I believe that I have heard many native speakers pronounce it as /s/ sound.

Did I hear it wrong? or Americans simply pronounce s at the end of "words" as /s/ not /z/

  • There are lots of different American accents, and I'm sure they include differences in this pronunciation. There's also not that much difference between the two sounds, so there's a spectrum of pronuciations between them.
    – Barmar
    Commented Jan 22, 2015 at 20:31
  • Stereotypically, in Chicago urban dialect Bears (the name of the Chicago football team) is pronounced /bers/. But only by fans, of course. Commented Jan 22, 2015 at 20:59
  • There are no words in English which are differentiated by the difference between a final /-ds/ and /-dz/. So one can pronounce words either way, and be understood perfectly well. I wouldn't be surprised if there are some Americans who pronounce it /wɚds/. Native English speakers probably won't even notice the difference (as long as you don't pronounce it wurts; this is a phonemic difference in English, e.g., herds vs. hurts). Commented Jan 22, 2015 at 21:01
  • 1
    Native Americans may have reservations about this.
    – tchrist
    Commented Jan 22, 2015 at 22:37
  • @PeterShor Not quite true. If you deliberately use an /s/ there it will cause the voiced part of the syllable to be shorter meaning that it will indeed be understood as hurts instead of herds! Commented Jan 23, 2015 at 16:06

2 Answers 2


Many phoneticians and phonologists use the term lenis and fortis to describe different types of consonant. Those phonemes which we typically think of as being voiced are lenis and those which we characteristically think of as voiceless are termed fortis.

There are two reasons for using these terms. The first is that lenis, or so-called "voiced", consonants in English have other properties which they share apart from the fact that we characteristically consider them voiced .

The second reason is that in actual speech, most voiced consonants undergo devoicing when they are next to voiceless sounds. Voiceless sounds here includes silence. This means that, for example the /b/ in bed, which we normally consider to be voiced, will be partially and sometimes fully devoiced at the beginning of a sentence when preceded by silence. Similarly, the /d/ in bed, again normally considered voiced will be partially or fully devoiced when occurring at the end of a sentence next to silence. This, as described, will also be true of the /b/ in hotbed where it occurs next to voiceless /t/.

The same is true to a lesser extent for fortis consonants. In certain environments consonants which we normally consider voiceless may become voiced if they occur between two vowels. So for example in the word better in expressions such as we'd better go, the /t/ will often be realised by a voiced alveolar tap. Similarly the /h/ in ahead is often fully voiced instead of being voiceless.

There is obviously a problem with talking about devoiced voiced sounds and so on. But there is a deeper problem, which is that we still need terms to talk about consonants that are usually voiced, even if they happen to occur in a position in which the actual phonetic sound will be entirely voiceless. This is because the most important properties of these sounds remain the same. It won't have escaped the quick-witted reader here that I have just said above that the /b/ and /d/ in bed may both be voiceless if the word is said in isolation. This is absolutely true. If you record a speaker speaking naturally and cut out just the /b/ or/d/ segments they will sound exactly like a /p/ and a /t/ most of the time.

I can already hear screams of gibberish, claptrap and the like from the audience. Quite understandable. It doesn't take one second to realise that any native speaker is able to freely distinguish between the words bed and pet when produced by another native speaker. However, the reason that this is true, is that fortis and lenis consonants keep their fortis and lenis characteristics whether they are voiced or not.

Most importantly, fortis and lenis consonants affect the other sounds around them. It's been proven that English listeners are practically deaf when it comes to distinguishing the voicing of word initial sounds by listening to the sounds themselves. The reason that we know that the sound at the beginning of bet is a /b/ and not a /p/ as in pet is that the vowel after /b/ is immediately voiced, whereas there is a delay after /p/ before the voicing for the vowel kicks in. During this voiceless period of the vowel there is an audible hissing as the air from the consonant is released from the vocal tract. This is what is known as aspiration. Fortis consonants in English are aspirated when in word initial position, and it is the aspiration after /p/ and not the sound of the consonant itself which helps us to distinguish /p/ from /b/.

Fortis and lenis consonants have a completely different effect when occurring at the end of a word. Fortis consonants in all languages have the effect of shortening the preceding vowel when they occur at the end of a syllable. This is known as prefortis clipping. If you say the words bead and beat, and listen carefully, you might be able to notice that the FLEECE vowel, /i:/, is considerably shorter in the word beat. This is because the fortis consonant, /t/, causes the vowel to be shortened. Even though the /d/ in bead will usually be devoiced, and therefore is phonetically voiceless, it remains lenis and causes no shortening of the vowel. It is the length of the preceding vowels or other voiced segments which tells a listener whether a syllable final consonant is fortis or lenis. It has nothing to do with whether the actual realisation of the consonant actually contains any voicing, any vibration from the vocal folds.

Below is a speech pressure waveform and spectrogram for the words beat and bead respectively. The wavy readout at the top is the waveform, the diagram underneath is the spectrogram. You will see that the vowel on the left, represented by the first big burst of activity in the waveform readout is considerably shorter than the one on the right. You can also see the vowel in the spectrogram underneath. The vowel is the the first vertical grey band. It's about half the width in the first word as in the second - a result of pre-fortis clipping.

_____ / bi:t / __________________ / bi:d / ________

enter image description here

The Original Poster's question

If the Original Poster has a very sharp ear, they may indeed have heard that the /z/ at the end of words is actually voiceless when said by native speakers. In other words, the phonetic quality of the sound itself will be very close to our canonical idea of /s/. This is because it will be devoiced here at the end of the word. However, native speakers' ears and brains will tell them that this is the /z/ phoneme, and not /s/. The reason is that the /d/ segment is clearly /d/ and not /t/, because the vowel in /wɜ:ds/ is fully long here. The following plural marker will be understood as /z/ simply because it follows the lenis consonant /d/.

However, we shouldn't think that this means that we can try to say /s/ at the ends of English words instead of /z/. This would be a terrible mistake. Using /s/ where we need /z/ at the end of a word will cause the vowel to be reduced. If you try to say the word peas with an /s/ instead of a /z/, you will clearly just be saying the word peace because the /s/ will cause the vowel to be shortened. If you want some peas it is probably not a good idea to say:

  • Please give me some peace!

Hope this is helpful!

  • Comments are not for extended discussion; this conversation has been moved to chat.
    – tchrist
    Commented Feb 3, 2018 at 19:28

I'd guess that you heard it wrong, though it's very possible that you heard /z/ pronounced as [s]. The slashes enclose phonemes (articulatory targets), while the brackets enclose phones (i.e., actual sounds). The distinction in English between phonemically voiced /b, d, z, .../ and phonemically voiceless /p, t, s, .../ is phonetically complicated. The voiceless phonemes are fortis (strongly articulated), aspirated (in the case of stops p, t, ...), and voiceless, typically, while the voiced phonemes are lenis, unaspirated, and voiced, typically. But /z/, for instance, can perfectly well be pronounced as voiceless, and it can still be told apart from /s/, provided it is lenis (weakly articulated).

It is also possible to distinguish /z/ from /s/ not by voicing, but by length of the preceding vowel, since vowels are longer before /z/ than they are before /s/. So if I say [rais] and [rai:s] for you and ask which was "rice" and which was "rise", you'll be able to tell easily by the length of the diphthongs [ai] versus [ai:]. (Or, in Canadian English, you can listen for the difference between [ʌi] and [ai].)

  • 1
    Do you have some kind of source for a fortis–lenis distinction besides a voicing distinction for /s z/? I cannot in any way produce a lenis but voiceless /z/ in English that is distinct from /s/. I’m not even sure what the theoretical distinction would be—something like [z̰̊] or [z̤̊]? Commented Jan 22, 2015 at 22:52
  • 1
    It’s just the phrasing that “/z/, for instance, can perfectly well be pronounced as voiceless, and it can still be told apart from /s/, provided it is lenis” that puzzles me. If /z/ is pronounced voiceless in English, it is automatically fortis to me: it’s voiceless/fortis vs. voiced/lenis. So if words is pronounced with an unvoiced /z/, it is just [wɚˑd̥s]. (With the plosives, I agree there’s a three-way distinction, but not with the sibilants.) Commented Jan 22, 2015 at 23:27
  • 1
    That’s not relevant to a fortis–lenis distinction at all. That deals only with articulatory voicing which, phonetically, is a completely separate process from forte. Of course /z/ can be pronounced as either voiced or voiceless—I’m not disputing that. What I’m saying is that to me, there is no three-way distinction in English between a fortis voiceless [z̤̊], a lenis voiceless [z̥], and a (lenis) voiced [z], and therefore a voiceless /z/ cannot be told apart from a /s/. For English sibilants, voice and forte are not orthogonal, though they are, marginally, for plosives. Commented Jan 22, 2015 at 23:53
  • 2
    Well, I can only say that I have no such distinction. The /z/ in witches and zip are equally (un)voiced and fortis/lenis to me—that is, there is just as much variation in both voicing and forte in one as there is in the other. This contrasts, as I've said, with plosives, where I can easily distinguish three phones matching only two phonemes. (I am not mixing up phonemic and phonetic distinction; that would be a rather grave mistake for a phonetician to make.) Commented Jan 23, 2015 at 0:27
  • 2
    Actually, @Araucaria, reading your post made me realise that Greg here is (I presume) talking about forte as a phonemic characteristic, rather than as a phonetic articulation, in which case I don't disagree at all. I fully agree that /z/ is marked for lenisness whether it's unvoiced or not—my beef was that Greg’s answer seemed to be saying that unvoiced /z/ could also be phonetically articulated as lenis, which I would say is almost a physical impossibility (at least to me). Commented Jan 24, 2015 at 0:26

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.