Ask LLOG: Covenant?

« previous post | next post »

From L.J.:

Could you explain the real pronunciation of "covenant"? In American English, the first vowel is not like that in British English as shown in the Oxford English Dictionary. Please see the attachment. However, theoretically, a word cannot have three schwas because the stress cannot be assigned. Could the first vowel be actually a longer one?

I'll answer the question, while indulging in a short rant about the faults of the International Phonetic Alphabet (and the worse faults of its use, in dictionaries and elsewhere).

To start with, here's the OED's audio for its two top-row pronunciations

"British English" /ˈkʌvənənt/ and "U.S. English" /ˈkəvənənt/

Curiously, the two non-IPA spelled pronunciations are identical, both

KUV-uh-nuhnt

…which is actually fine, since I perceive no difference in perceptual vowel quality between the first syllables of the two (audio) versions.

(If there's a standard term for non-IPA spelled pronunciations of that general kind, I don't know what it is — so I'll call them  "lah-lah-lah".)

Even more curiously, the alternative "British English" pronunciation (given in IPA as /ˈkʌvn̩ənt/ instead of /ˈkʌvənənt/, and in lah-lah-lah as "KUV-uhn-uhnt") is actually bit-wise identical in audio waveform samples to the first British pronunciation — that is, the audio is the exact same file, decorated with a (therefore empty) claim about a difference in syllabification.

OK, now we get to the question, about how stress can be assigned to one of the three schwas in /ˈkəvənənt/.

The background problem is a fundamental equivocation about what "schwa" means.

Many (not all) unstressed vowels in English are "reduced", which means that they have a unrounded central quality if pronounced in a hyper-articulated syllable-by-syllable style, plausibly represented by IPA [ə] ("schwa"). However, in actual speech these vowels can be deleted (= reduced to non-existence), raised or lowered, fronted or backed, rounded or unrounded, nasalized or not, etc., depending on the sounds around them, and the phonetic realization principles of the speaker's idiolect.

IPA offers no symbol to represent this situation, so most people (phoneticians, dictionary editors, dialect coaches, whoever) use [ə] ("schwa") to fill the gap. This then creates an ambiguity. Does [ə] mean an unstressed-and-reduced vowel, whose actual quality might be [ɪ] or  [ɛ] or [ʌ] or etc., depending on circumstances? Or does it mean an unreduced (and perhaps stressed) vowel , in the phonological category that J.C. Wells assigns to the STRUT lexical class, and whose phonetic quality in a particular case might be perceived to correspond best to IPA [ə]?

For most people — though apparently not (some of) the OED's editors — the solution is to reserve the use of [ə] for the meaning "unstressed reduced vowel", and to used [ʌ] to represent the pronunciation of stressed STRUT vowels. This is obviously not what the OED editors have done, though their choice is undermined by their own lah-lah-lah transcriptions and by the phonetic quality of the audio clips they provide.

For comparison, see the entries for covenant from Merriam-Webster, Wiktionary, Cambridge, and so on.

As etymological background, see the Wikipedia entry for Hebrew "shva".

And in place of a longer IPA rant, you can take a look at

"Why I don't love the International Phonetic Alphabet", 1/2/2008
"How truck starts", 8/9/2011
"On beyond the (International Phonetic) Alphabet", 5/19/2018
"Farther on beyond the IPA", 1/18/2020
"First novels", 2/13/2022
"Pronunciation evolution", 3/15/2022
"More post-IPA astronauts", 3/16/2022
"What 'IPA' means now…", 9/28/2022
"Steele v. Monboddo", 7/12/2025
"Comftrable", 9/21/2025

Update — There's obviously more to said about the OED's different editorial standards for IPA vs. lah-lah-lah pronunciation representations. As another example, here's tactical:

 



38 Comments »

  1. Daphne Preston-Kendal said,

    November 1, 2026 @ 9:50 am

    All of the deviant choices of phonetic symbols made by the OED for the third edition are at least somewhat regrettable. If they were going to choose a different symbol than ʌ for the GenAm STRUT vowel, ɐ have been preferable to merging with unstressed ə.

    John Wells’s critique of the choices made for the BrE transcription (including another misuse of ə) is also still relevant: https://web.archive.org/web/20250217184047/http://www.phon.ucl.ac.uk/home/wells/ipa-english-uni.htm#:~:text=7%2E%20Upton%27s%20scheme,-This

  2. HTI said,

    November 1, 2026 @ 9:54 am

    The pictured transcriptions are phonemic (hence the slashes) and not phonetic, so I’m not sure why we’re worried about reserving the use of [ə] (note the square brackets) for anything.

    /ʌ/ and /ə/ are distinct phonemes in many dialects of 21st-century English, either because they’re (I.) very conservative — no FOOT–STRUT split and therefore */ʌ/ is simply /ʊ/ and not /ə/ — or (II.) very advanced — FOOT–STRUT split, but STRUT has subsequently merged with some other phoneme.

    Note that in the (II.) case the “other phoneme” is not always /ə/ as in most (all?) Canadian or U.S. English. I believe the outcome is sometimes FATHER /ɑː/, although other options seem plausible too.

    The (I.) case is true in, for instance, the north of England. Some other parts of the U.K. can have /ʌ/ ≠ /ə/ with the FOOT–STRUT split, although others agree with U.S. speakers on this one.

    Cases where /ʌ/ and /ə/ merge treat /ə/ like any other vowel which can occur in stressed and unstressed syllables alike. The OED is correct to write /ˈkə…/ in such cases (e.g. the U.S.) and /ˈkʌ…/ in cases which might have the U.S. pattern but might have /ʌ/ ≠ /ə/ instead (e.g. the U.K.).

  3. HTI said,

    November 1, 2026 @ 9:58 am

    Preston-Kendal says “ɐ have been preferable to merging with unstressed ə” but this remark only makes any sense if one is trying to represent phonics, which the OED are not (and definitely should not be). Phonemically there are no minimal pairs to distinguish STRUT from COMMA in such dialects and, moreover, native speakers with no linguistic training intuitively feel that STRUT and COMMA are the same exact thing.

  4. ktschwarz said,

    November 1, 2026 @ 12:13 pm

    "Pronunciation respelling" is Wikipedia's name for things like "KUV-uh-nuhnt"; I don't know if it's used more widely.

  5. ktschwarz said,

    November 1, 2026 @ 1:43 pm

    The reader who was anxious that "a word cannot have three schwas because the stress cannot be assigned" must have been taught the myth that schwa can't be stressed. Geoff Lindsey has a blog and video debunking that myth: STRUT ʌ, schwa ə and American English.

    The OED's IPA transcriptions for "covenant" are fine, as explained by HTI and Lindsey. There's some potential for confusion with the lah-lah-lah transcription, as the post says, because it uses different respellings for the stressed and unstressed syllables. This is because they're trying to cover both British and American pronunciations in the same respelling. The key to their "Simple Text Respell" system is here:

    https://www.oed.com/information/understanding-entries/pronunciation/simple-text-respell/

    The keyword for "u" is "strut", and for "uh" is "letter" (a thoughtless and British-centric choice, it should have been "commA" to cover both). That's okay for readers who understand that they're both the same in GenAm, but some people don't.

  6. NW said,

    November 1, 2026 @ 4:54 pm

    There might not be straightforward minimal pairs, because STRUT is almost always stressed and schwa is unstressed, but we can get close. For the word 'income', I can pronounce it two relevant ways; I think they're in free variation in my speech. It can have a weak schwa or it can have the STRUT vowel of 'come'.

    This is when I use the homorganic velar nasal. I have a third possibility, where the /n/ does not assimilate, and then the 'come' part can only have the STRUT vowel, which I usually notate as [a]. Phonetically it is [a-], about midway.

  7. HTI said,

    November 1, 2026 @ 6:27 pm

    Assuming you have /ə/ = */ʌ/ that would be /ˈɪNkəm/ and /ˈɪNˈkəm/ respectively. (You can write the latter with secondary stress but I have my doubts about the utility of going beyond the binary stressed–unstressed distinction.)

  8. ktschwarz said,

    November 1, 2026 @ 6:27 pm

    Lindsey mentions (in the blog that I linked above) an example of unstressed STRUT in the first syllable of "undone". But, from another of his blogs: "STRUT and commA … have practically no contrastive value with respect to each other: they’re defined in a mutually exclusive way, and alleged ‘minimal’ pairs such as an equal and unequal are not structurally identical."

    There was a previous long and detailed discussion of STRUT vs. schwa for various accents in Language Log comments a couple of years ago, with several links to Geoff Lindsey:
    https://languagelog.ldc.upenn.edu/nll/?p=63169

  9. JPL said,

    November 1, 2026 @ 6:46 pm

    I perceive a difference between the BE pronunciation clip and the AE version: the AE does seem more "closed", as the Brits would say, and the BE is more "open" (i.e., in terms of "tongue height"). But I don't recognize that as the standard AE pronunciation, which is I would say virtually identical to the BE. So which came first here, the performance or the description? BTW, in the AE clip, the first syllable is stressed, but with a more closed vowel quality than the first syllable in the BE version. So does the term 'schwa' refer to the purely phonetic vowel quality, or to the reduced status of unstressed vowels in English, which phonetically is typically, but not always, the more "closed" vowel height, rather than the more "open" one? There appears to be a difference between the phonetic quality described by the "inverted "e" symbol and the phenomenon described by the phonological term 'schwa', such that they might pick out the same sound, but for different reasons. So the inverted "e" symbol belongs to phonetic analysis, and the term 'schwa' to phonological analysis, and these should not be mixed up. The distinction between the inverted "e" and the inverted "v" describes a phonetic distinction, which is still necessary to distinguish. The term 'schwa' belongs to phonology, and the precise quality of the vowel so named is given by a more "narrow" transcription. (There are a lot of cases where in order to disambiguate two homophonous pronunciations of different word forms, the full version of the two vowels is pronounced, often with contrastive stress.)

  10. Chas Belov said,

    November 1, 2026 @ 11:07 pm

    I couldn't hear any difference between those two British pronunciations, nor between the way they said "covenant" and the way I say it. I'm not so concerned with the first vowel as with the last. It sounds to me not like a schwa but rather like a short I.

    The closest IPA I can come up with is kəv'ənɪnt.

  11. Mark Liberman said,

    November 1, 2026 @ 11:59 pm

    @Chas Belov: "I couldn't hear any difference between those two British pronunciations"

    That's a good thing, since the second "British" audio link points to the exact same file as the first one!

    "I'm not so concerned with the first vowel as with the last. It sounds to me not like a schwa but rather like a short I."

    Agreed — that's normal for reduced vowels between coronal consonants.

  12. Pedro said,

    November 2, 2026 @ 5:04 am

    ktschwarz said,
    "Pronunciation respelling" is Wikipedia's name for things like "KUV-uh-nuhnt"

    I've heard that "pronunciation respelling" before, but I think it's too grand a term for what it really refers to – basically sticking a H after every other vowel while adding very little new information about the pronunciation. "Lah-lah-lah" is a much better term.

    What's going on here is that the standard British & American pronunciations are identical for this word: the first vowel is pronounced as a short U (rather than a short O as you might expect from the spelling), the accent is on that first syllable and the other two syllables are pronounced with weak vowels (schwas). But for some reason the OED uses a slightly different set of symbols for the two standards: they transcribe the STRUT vowel as ʌ in British English but ə in U.S. English, so it looks like the two standards use different pronunciations.

    That's confusing because there is a family of words beginning with cov- where the British & American standards genuinely differ. For example Coventry is /ˈkɔvəntri/ in British English but /ˈkʌvəntri/ in U.S. English – that is, the first vowel is a short O (as written) in the UK but a short U (just like covenant) in the U.S.

    It's also a bit misleading to transcribe the British & American vowels with different symbols just because it's rendered differently in R.P. and General American, because there's far more variation between British accents (e.g. [ʌ], [ɤ], [ʊ] or indeed [ə]) than there is between these two standard forms.

  13. Pedro said,

    November 2, 2026 @ 7:48 am

    Why are they saying the pronunciation with the syllabic nasal [ˈkʌvn̩ənt] is only available in British English? Surely any native English speaker in any country can choose between [ˈkʌvənənt] and [ˈkʌvn̩ənt] ?

  14. Jerry Packard said,

    November 2, 2026 @ 8:57 am

    I pronounce the last vowel in ‘covenant’ as [ɛ] (upper NewYork State dialect area).

  15. J.W. Brewer said,

    November 2, 2026 @ 10:38 am

    I assume what's going on in Jerry Packard's idiolect is that secondary stress on the final syllable is preventing vowel reduction, and I suspect he's not the only Anglophone for whom that is the case although I don't know if they'd all have the same unreduced vowel there. I'm in that bad place where self-consciousness of the issue is making me doubt that I can accurately assess my own usage, but I also wonder if an unreduced vowel in the final syllable is maybe more likely when covenant is used as a (rarer) verb than when used as a noun? Using the derived noun "Covenanter," which is important in discussions of Scottish history, will tend to force an unreduced vowel in what becomes the penultimate syllable, but of course dictionaries are often lazy about separately specifying pronunciations for such derived words.

  16. Jerry Packard said,

    November 2, 2026 @ 11:33 am

    I think J.W. is correct, as I pronounce ’covenant’ with primary stress on the first syllable and secondary stress on the third syllable. So for me, syllable one is [ʌ], syllable two is [ə] and syllable three is [ɛ].

  17. Jerry Packard said,

    November 2, 2026 @ 1:14 pm

    “…I also wonder if an unreduced vowel in the final syllable is maybe more likely when covenant is used as a (rarer) verb than when used as a noun?”

    On analogy with syllable-final stress in {CON-vert con-VERT}-type pairs, the final syllable is more likely to be unreduced if it is the noun member of the pair.

  18. Jerry Friedman said,

    November 2, 2026 @ 2:00 pm

    @HTI: native speakers with no linguistic training intuitively feel that STRUT and COMMA are the same exact thing.

    Some do, but I've seen native speakers use some representation of the KIT vowel for schwas, and I've seen one say that schwa is the FOOT vowel (not always very rounded among Americans). On that subject, note "rookie" from "recruit", presumably pronounced with a schwa in the first syllable. The /r/s might have something to do with that.

    For a minimal pair, the closest I can get at the moment is "annoy" versus "unoiled", though after "an", "unoiled" might be accented on the first syllable—which might be part of a case that negative "un-" has secondary stress. Another one is "enable" versus "unable". Merriam-Webster says the first vowel of "enable" is KIT and American Heritage says it's DRESS, but I think it's a schwa for me and many Americans.

  19. JPL said,

    November 2, 2026 @ 5:28 pm

    For a word like "lobotomy", I would predict, in one of the typical pronunciations, that the phonetic qualities of the first unstressed vowel are slightly different from the phonetic qualities of the second unstressed vowel, by hypothesis due to the tongue in the first case moving toward the back position in anticipation of the stressed back rounded vowel, and in the second case the tongue moving forward towards the high front position for the final vowel. If that's the case, then I would suggest that the term 'schwa' should not be so closely tied to one particular vowel quality, but should refer to a "reduced" vowel in unstressed position, regardless of the specific phonetic quality of the particular case. And then you can give general rules for what tends to happen to unstressed vowels in particular phonetic contexts. (As Mark does above for "between coronal consonants".)

  20. Coby said,

    November 2, 2026 @ 5:46 pm

    It seems that the French in the 19th century heard what is now written as [ʌ] as though it were [ə] or [œ], whence their pronunciation of such words as club or lunch. Littré (1872-77) explicitly writes that kleub is a peu près la prononciation anglaise.

  21. Chas Belov said,

    November 2, 2026 @ 6:27 pm

    @Pedro

    "Lah-lah-lah" is a much better term.

    I was actually puzzled by it. "Pronunciation spelling" may be grand, but I instantly understand it from seeing an example.

    Maybe if it was "LAH-lah-lah" or "lah-lah-LAH."

  22. Chas Belov said,

    November 2, 2026 @ 6:29 pm

    *pronunciation respelling

  23. HTI said,

    November 2, 2026 @ 7:00 pm

    @Jerry Friedman

    KIT vowel for schwas

    The weak vowel merger is a distinct phenomenon. But yes, it does result in the KIT–COMMA distinction being neutralised in unstressed syllables. The distinction is maintained in stressed syllables e.g. kit (∈ KIT) vs. cut (∈ COMMA=STRUT).

    I've seen one say that schwa is the FOOT vowel (not always very rounded among Americans).

    Also the weak vowel merger. Once again only applies in unstressed syllables: put vs. putt.

    "enable" versus "unable"

    Yes, enable typically starts with unstressed KIT (and therefore simultaneously COMMA, STRUT, and FOOT as well) in these kinds of dialects. I suspect that in connected speech unable could well be a homophone but in careful speech the morpheme un- is stressed. In the latter case we do have a minimal pair but the distinction is unstressed COMMA vs. stressed COMMA, i.e. the same exact phoneme (modulo neutralisations).

  24. Michael said,

    November 3, 2026 @ 4:03 am

    Phonemically there are no minimal pairs to distinguish STRUT from COMMA in such dialects and, moreover, native speakers with no linguistic training intuitively feel that STRUT and COMMA are the same exact thing.

    I agree with this — to me, STRUT and COMMA feel like they are the same vowel.

    But that vowel isn't the vowel that I think of as the paradigm for schwa, which would be closer to FOOT. (What do I mean by "the paradigm for schwa"? Sometimes I may jocularly pronounce a string of characters that is clearly not pronounceable. When two consonants abut each other, it will often be necessary to put some kind of vowel in between them, and since the 'spelling' calls for nothing, that vowel will represent my mental model of a schwa, reduced as far as possible.)

    In particular, for "covenant", the second syllable is ambiguous in my mind and the third syllable cannot use a STRUT-like vowel and must instead use a FOOT-like vowel.

    But I don't recognize that [OED audio] as the standard AE pronunciation, which is I would say virtually identical to the BE. So which came first here, the performance or the description?

    I felt similarly. Your second sentence reminded me of being frantically contacted on behalf of a Chinese undergraduate who was finishing up a thesis and urgently needed some audio files of words pronounced by a native speaker (hopefully to be me!). I was fairly bemused at the idea of writing the paper first and collecting the audio samples second.

    Another one is "enable" versus "unable". Merriam-Webster says the first vowel of "enable" is KIT and American Heritage says it's DRESS, but I think it's a schwa for me and many Americans.

    This resonates with me. I've always been uncomfortable putting "any" into IPA; my feeling is that according to the conventions I learned for IPA transcription of English, "any" has no appropriate transcription. (Because of the vowel in the first syllable.)

    For what it's worth, I can't reliably distinguish unstressed KIT from schwa. They're the same to me, or perhaps more accurately my mental phonological model has no space for unstressed KIT.

    I do think that whatever the first syllable of "enable" might be, there's no way it can be DRESS.

  25. Michael Watts said,

    November 3, 2026 @ 4:14 am

    (I am the "Michael" of the preceding post)

    From Lindsey's post:

    But, sooner or later, this strategy is bound to go wrong. In the video, I show a clip of an American YouTuber insisting that the word undone is /əndʌn/. Of course it isn’t: for those with a STRUT-commA contrast, it’s /ʌndʌn/. The YouTuber is aware of the stress difference between the two syllables and has been led to believe that stress differences require him to write different symbols. So he makes this mistake precisely because he stresses schwa every day of his life!

    This feels a little off to me. Certainly I have also 'learned', in some sense, that schwa is necessarily unstressed in English. But if you asked me about "undone", I would tell you that the vowel in both syllables was wedge. Neither is schwa. I would have gone on to conclude that "undone" has spondaic rather than iambic stress. (Which might not be true!) So, despite having the same education that Lindsey says causes this problem, I am somehow… probably not immune, but resistant, to it. More description of the problem is needed. It's very clear to me that the first syllable of "undone" can't be schwa. As in my comment above about "covenant", a schwa there would have to sound more like FOOT. And the syllable would have to be weaker than in fact it is. (OK, I guess I do have an opinion about the spondaic stress.)

  26. Liwei Jiao said,

    November 3, 2026 @ 9:59 pm

    Hi, I am posting this on behalf of Prof. XU Ximing of Ningbo University in China. He published a paper on this topic a few days ago. Here is the English abstract. The article is written in Chinese. He will join the discussion if his paper is found interesting.
    DOI: 10.20101/j.cnki.1001-5124.202601090

    A Continuum of Phonetic Changes in English Viewed from “Relative Prominence”: from Relative Reduction to Absolute Reduction
    XU Ximing (Faculty of Foreign Languages, Ningbo University, Ningbo 315211, China)
    Abstract: Based on Liberman & Prince’s metrical theory of “Relative Prominence”, this paper takes English words “gymnast” and “modest” as cases to explore the fact that -nast in the former can be uttered as either [-nst] or [-nst](Am. variant), which is identical to the reduced vowel in [-dst](-dest) of the latter. The heavy syllable gym-[dm-] is seen as relative prominence (metrical stress) of the strong position at foot level, while the (super) heavy syllable [-nst] and (super) light syllable [-nst] are respectively treated as relative reduction and absolute reduction of the weak position in line with the metrical appeal for strong-weak opposition to avoid stress clash. It is argued here that although attraction of stress by heavy syllables is matched in harmony, they are possibly reduced as [] syllables or even omitted if they lose stress assignment, thus shaping a rhythmically distinct pattern in alternation of stressed and unstressed syllables. The relationship between relative reduction and absolute reduction is a continuum of phonetic changes rather than a metrical structure in stress-unstress opposition. This research can provide a theoretical basis and reference for the study and teaching of English phonology and phonetics, and for Chinese natives’ study of stress acquisition in particular.

  27. Liwei Jiao said,

    November 3, 2026 @ 11:06 pm

    Because the IPA does not display properly, I post the paper on Google Drive, with permission from the author.

    https://drive.google.com/file/d/1QUHRZ_KidRCFyJtxYbkr6gupiTWbKlJa/view?usp=drive_link

  28. Jarek Weckwerth said,

    November 4, 2026 @ 12:30 pm

    I tend to think the situation in the OED is the result of having two separate teams (led by Upton and Kretzschmar) for the two major dialects in the Oxford Dictionary of Pronunciation, without enforcing any alignment between them. Wells and Roach/Setter tried to minimize the differences in their Longman and Cambridge dictionaries. Oxford did not make the effort, leading to theory-based decisions producing puzzling unnecessary differences in the representations.

    The recording situation is, however, universal. Publishers just use recording banks that they happen to have made for other purposed (usually learner dictionaries) and they don't seem to be verified at any stage beyond the recording studio. The CD versions of Longman and Cambridge have hundreds, possibly thousands of cases where the recording is at odds with the transcription.

  29. Michael Watts said,

    November 4, 2026 @ 5:40 pm

    Because the IPA does not display properly

    Something weird is going on. The published comment shows "tofu" blocks indicating unicode points U+F07B, U+F07B (again, but apparently meant to contrast with the first one), U+F040, U+F05A, U+F049, U+F07B, U+F040, and U+F040.

    These code points are not valid for the public internet; they belong to Unicode's Private Use Area (U+E000 – U+F8FF).

    Context suggests to me that U+F07B was meant to depict ash, æ, U+00E6, and that U+F040 was meant to be schwa, ə, U+0259. Those characters have defined Unicode points and they display appropriately in a browser.

    Some software that Liwei Jiao is using seems to represent IPA glyphs in some way that doesn't match Language Log's specified UTF-8 encoding. I'm not really sure what exactly happened, though.

  30. Michael Watts said,

    November 4, 2026 @ 5:57 pm

    Ah, I've learned what happened: Liwei Jiao's comment is presumably pasting text that was originally written in this font: https://www.azfonts.net/fonts/ipa-samm-uclphon1-silmanuscript/regular-186161 , which defines U+F040 as schwa and U+F07B as lowercase ash.

    So the affected part of the abstract should read like this:

    Based on Liberman & Prince’s metrical theory of “Relative Prominence”, this paper takes English words “gymnast” and “modest” as cases to explore the fact that -nast in the former can be uttered as either [-næst] or [-næst](Am. variant), which is identical to the reduced vowel in [-dəst](-dest) of the latter. The heavy syllable gym-[dʒɪm-] is seen as relative prominence (metrical stress) of the strong position at foot level, while the (super) heavy syllable [-næst] and (super) light syllable [-nəst] are respectively treated as relative reduction and absolute reduction of the weak position in line with the metrical appeal for strong-weak opposition to avoid stress clash. It is argued here that although attraction of stress by heavy syllables is matched in harmony, they are possibly reduced as [ə] syllables or even omitted if they lose stress assignment, thus shaping a rhythmically distinct pattern in alternation of stressed and unstressed syllables.

    I don't know what was intended for the contrast between "[-næst]" and "[-næst] (Am. variant)"

  31. JPL said,

    November 4, 2026 @ 7:28 pm

    A possible hypothesis, consistent with my comments above, might be the following.

    In English, only the vowels with stress (primary and secondary stress) are phonemic (make differences), while unstressed vowels are variable and not associated with any particular vowel quality. The particular unstressed vowel quality in each case is determined by the phonetic context of a word, but would be roughly in the area of mid (closed) central. In cases of ambiguity a phonemic difference could be found in the historical or etymologically-based spelling of the written citation forms. The term 'reduced vowel' raises the question "reduced from what?". That presumes that at some previous historical point a definite vowel was there independently of the written historical or etymological written forms; borrowings would be assimilated to the English practices (e.g., the French vs AE pronunciations of 'croissant'). In syllable-timed varieties of English that lack a central vowel, some available vowel quality has to be found in the pronunciation of the unstressed vowels, and typically these are found in the written forms of the words, which are learned originally in school and then become conventional.

    I don't know if this hypothesis would be borne out, and I don't know if anyone already understands it this way. (I thought of it because of the use of the little empty boxes as symbols representing a variable content for particular vowel qualities in Liwei Jiao's comment. That could be a symbol for "schwa" separate from and independent of the IPA vowel symbols.(In the IPA system the inverted e and all the other mid central symbols (barred i, etc.) don't have other "names", and :schwa" is not used in phonetic description.))

  32. Jarek Weckwerth said,

    November 5, 2026 @ 9:10 am

    @JPL In some varieties, there's plenty of minimal pairs with the traditionally construed schwa. Standard Southern British, for example, distinguishes -ed from -ered as -ɪd vs. -əd, with Wells's standard example being chatted vs. chattered. Same for -es vs. -ers, e.g. boxes vs. boxers. Would you consider that ɪ as a vowel that is not reduced? It's certainly not stressed in any usual sense; and e.g. in singing -ed may be rendered as -ɛd, which I would think is the "unreduced" version. What about -man vs. -men?

  33. JPL said,

    November 5, 2026 @ 6:39 pm

    @Jarek:

    Thanks for bringing the data, and I'm sure there's a lot more.

    In the case of "chatted" vs "chattered" and "boxes" vs "boxers", the phonetic quality of the unstressed vowel is what it is because of differences in the structure of the two word forms, but this involves not just the distinction between the two word forms, but also distinction in the two different lexemes of which they are instances. The disambiguation involves distinguishing between two different lexemes: 'chat' vs 'chatter' and 'box' vs 'boxer', where the second member includes an additional morpheme, marked by an "er" suffix (I think these are two different "er" suffixes.) Speakers know that the "er" is there, because the "r" is pronounced in some contexts. (The vowel for the "agentive" suffix seems to be variable, at least when disambiguation is necessary.) So the difference in phonetic vowel quality in the word forms reflects this underlying morphological distinction, manifested in vowel length as well as the ghost of "r" in the one case, vs the influence of the coronal consonants in the other, determining the difference. These are two variants of the unstressed vowel, not instances of two different phonemes (so I would not say that the fronted variant (barred "i") has stress), although I suppose you could say that the variants of the unstressed vowel are allophones of something. My suggestion was that the notion of "unstressed vowel" should not be tied to one particular phonetic quality, but the specific quality is determined by context, and the same would go for the term 'schwa', so on a list of English phonemes, "schwa" would not be one of them. What exactly the vowels are "reduced from" is a mystery, at least to me.

    I'm certainly not ready to declare the conjecture confirmed, so any data would be appreciated. I think there are cases where two word forms are distinguished by two different variants of what I've been calling "the unstressed vowel". But as far as I can remember, these two variants are determined by conventional spelling differences.

    For "singing", "ed" and "man" vs "men" something seems to be missing. Maybe you can clarify. (The "agentive" suffix includes a vowel, the inflectional suffixes ("-d", "-s") do not.)

  34. JPL said,

    November 5, 2026 @ 7:02 pm

    Another way of stating the conjecture: No distinctions between two different lexemes are made in the unstressed vowel position. This seems unremarkable, I don't know if it's a fact, but, if true, what does it mean for the phenomenon of unstressed vowels?

  35. ktschwarz said,

    November 6, 2026 @ 1:53 am

    Rosa's/roses and Lenin/Lennon are canonical minimal pairs for differences in unstressed reduced vowels; there are also near-minimal pairs Alice/Dallas and abbot/rabbit. There was an old Language Log post about whether these can really be distinct:
    https://languagelog.ldc.upenn.edu/nll/?p=5523

    Mark Liberman was highly skeptical there about the reliability of intuitive judgments, and about whether any differences in isolated words are still there in connected speech: "in normal usage, these differences evaporate, or are overwhelmed by other sources of variation, or both."

    Personally I feel like it's impossible to be sure whether I actually say Rosa's/roses differently, or I just think they're different because of spelling.

  36. Jarek Weckwerth said,

    November 6, 2026 @ 4:21 pm

    @JPL: I'm not sure I follow. In chattered vs. chatted, the surrounding context is 100% the same, so the vowel quality cannot be context-dependent, and therefore is not allophonic since the two words can be distinguished. They are distinguished by ɪ vs. ə; therefore the two are phonemes. That's your classical commutation procedure. The fact that the morphemes may be different is irrelevant. Banned vs. bend have different morphemic structure but they do work as a minimal pair for æ vs. ɛ, right?

  37. JPL said,

    November 6, 2026 @ 6:20 pm

    @Jarek:

    I'm not a phonologist, so I'm not familiar with the lore. But in the comment above I should have said, "… the differences in the phonetic quality of the unstressed vowel are what they are because of differences in the morphological structure of the two word forms …", rather than differences in the phonetic context, which in the case of these two word forms is the same. The "-er" (or "Vr"), (indicating iterativity of the act), distinguishes 'chatter' from 'chat' as lexemes, and speakers at least try to include that in utterances, resulting in lowering and lengthening of the unstressed vowel. (Compared to the "chatted" case, where the unstressed vowel is affected only by its position between two coronal consonants.) My point was that the unstressed vowel sounds in the the two word forms, in some text (not system), are not instances of anything that has an independent existence; that they are not instances of any phonemes. In this particular case there are detectable differences, but in other cases either of these contexts could have the other value, without any problem of ambiguity. Phonemes are the sounds a language uses to distinguish one Bloomfieldian form from another, and this notion applies to lexemes, not word forms in texts. My conjecture was that no two lexemes are distinguished by vowels in unstressed positions. So the vowel in an unstressed position is not an instance of a phoneme. Is it true? I don't know. Maybe there are counterexamples.

  38. JPL said,

    November 7, 2026 @ 6:52 pm

    In my comments above I mentioned that there are probably, maybe not a lot, but a good number, of cases of two different lexemes whose typical pronunciations in texts are indistinguishable, and which can be disambiguated by substituting a "full vowel" (not a variant of the "reduced vowel") for an unstressed vowel. We've been talking about unstressed vowels in medial positions, but maybe unstressed vowels also occur in word-initial or word-final positions. So if these are examples of the same phenomenon, an example of what I'm talking about might be 'effect' vs 'affect'. In normal casual speech the fact that the initial vowel is unstressed results in either potential ambiguity in the text, or the speaker/writer wondering which lexeme they are using; and the problem is solved by reaching for the conventional spelling distinction. But then the question arises of whether at some earlier time in the history of the language there was a full vowel in that position which over time became reduced, or the spelling distinction is an arbitrary post hoc choice. I think there are more examples of this phenomenon.

    In normal casual speech speakers distinguish one word from another by trying to maintain the normal phonemic distinctions, and this means at least trying to make the right articulations, treating them perhaps as a "target", even though the result may be affected by the phonetic contexts. But with unstressed vowels it seems that speakers are not "aiming" for anything, or trying for a specific target, but rather it's just a case of getting from one articulation to another vocally (or "vocalically") in the easiest and most natural way. And nobody tries to disambiguate by appealing to variants of the unstressed vowel.

    This question was bothering me these days because of the tennis player Frances Tiafoe. I learned that apparently his parents are from Sierra Leone, and so I wondered how they could give him a girl's name, because in Sierra Leone it is conventional that "Frances" is a girl's name, and "Francis" is for boys. In speech there is never any ambiguity because in Sierra Leonean English there are no "unstressed vowels", and people always use the full vowel associated with the spelling distinction.

RSS feed for comments on this post · TrackBack URI

Leave a Comment