Schwa

« previous post | next post »

xkcd (3/15/24)

(image URL) (explanation; transcript; discussion)

Selected reading

[h.t. Edward M. McClure]

 



78 Comments

  1. Jarek Weckwerth said,

    March 25, 2024 @ 3:31 am

    This requires an analysis where the STRUT vowel is stressed schwa. Not everyone buys it ;)

    But the other camp also has some fun graphics on offer:

    https://www.google.com/search?client=firefox-b-d&q=schwa+never+stressed

  2. AntC said,

    March 25, 2024 @ 3:45 am

    /ˈənj(ə)nz./ in the transcript:

    The parens /(ə)/ presumably means the schwə is omissable or reduced. But then doesn't the /j/ count as (at least) a semivowel? Is Randall bending the rules here?

    Neither 'Doug' nor 'tunnel' first syllable is a /ə/ for me (Br.E Southerner, with N.Yorkshire leanings). Since 'Doug' is a personal name, and topic/subject in several of the sentences, how would the vowel be unstressed/reduced?

  3. AntC said,

    March 25, 2024 @ 3:49 am

    @Jarek the STRUT vowel is stressed schwa

    As is standard in East Yorkshire. (But not North nor West; neither the other side of the Humber.)

  4. J.W. Brewer said,

    March 25, 2024 @ 6:23 am

    My instinctive negative reaction (manifested upon seeing this a day or two before it was posted here) seems, upon reflection, to be based on the perhaps unexamined axiom that "If it's not unstressed we don't *call* it 'schwa.'" I must admit, again upon reflection, that the non-stress differences in articulation between a stressed STRUT vowel and an unstressed schwa are, in my idiolect, so subtle and nuanced that I will need to defer to those with more expertise in phonetics/phonology than I possess to determine if they are, in fact, differences.

  5. Cervantes said,

    March 25, 2024 @ 6:53 am

    The cartoonist seems to have conflated the schwa with /ʌ/. I'm a beta solver for cryptic crosswords, which frequently use homophones, and they basically assume that most unstressed vowels can be a schwa, which is close to being true. But stressed vowels normally cannot be, and that goes for "u".

  6. Jarek Weckwerth said,

    March 25, 2024 @ 6:59 am

    @ J.W. Brewer: Well you've bumped into the main philosophy-of-phonology question: What is a difference.

    From what I've seen, even those approaches that follow the route of STRUT=stressed-schwa (or the reverse) will usually claim that there's much more variability in the unstressed position, as should be expected. So that's one thing.

    But the level of notation in the xkcd strip, and most of the time elsewhere that you see so-called IPA, is phonemic. It deliberately only shows phonemic distinctions, with two consequences that are difficult to digest for non-phonologists: (1) what you show depends 100% on your theory, like here; (2) if you want to interpret the transcription as strict IPA, you run into very many very muddy puddles.

    I started my trip into phonetics from JC Wells's Accents of English, and that is a different approach: Make your phonemic system diaphonemic, i.e. as overly distinctive as you need it, and then account for any "mergers" with "production" rules.

    @ AntC: I don't know where Randall took that transcription from, but the bracketed notation means "either a schwa, or optionally a syllabic nasal/lateral". For onion, I don't think it works, because at least on the phonemic level you don't get syllabic n after j in English. Even if the nj is taken to stand for ɲ (the palatal nasal).

    One place that this kind of thing tends to originate from is the CMU Pronouncing Dictionary. Shows how very careful you need to be if you open-source something. It's the mother of all pronouncing dictionaries in speech tech (for cheap people at least), and it does have STRUT=stressed-schwa; it even does unstressed schwa as /ʌ/ which really rubs me the wrong way and always leads to trouble. And its All Over the Speech Tech World. But it doesn't do optional schwa.

  7. Andy said,

    March 25, 2024 @ 8:45 am

    And no-one here has picked up that the first word ('What') doesn't contain a schwa in many dialects too.

  8. Chris Button said,

    March 25, 2024 @ 9:35 am

    @ Jarek Weckwerth

    This requires an analysis where the STRUT vowel is stressed schwa. Not everyone buys it ;)

    Good point. Although I'd suggest that it's not really a question of "buying it" so much as a question of a tangible difference for many (possibly most?) speakers of English outside of North American English.

  9. David Marjanović said,

    March 25, 2024 @ 10:13 am

    This requires an analysis where the STRUT vowel is stressed schwa. Not everyone buys it ;)

    Rather, this analysis is obviously correct for some Englishes and obviously incorrect for others.

    The cartoonist seems to have conflated the schwa with /ʌ/.

    The cartoonist, like most of the US (NYC seems to be the great exception), has merged the schwa with /ʌ/: [ʌ] and [ə] are the stressed and the unstressed allophone of a single phoneme in such accents.

    Accents like RP allow unstressed [ʌ] and contrast it with unstressed [ə], so they are separate phonemes /ʌ/ and /ə/ there, and this /ə/ really is never stressed like the meme has it.

    From what I've seen, even those approaches that follow the route of STRUT=stressed-schwa (or the reverse) will usually claim that there's much more variability in the unstressed position, as should be expected.

    Oh yes. You can even get vowel harmony phenomena there: the second vowel of data seems to be /ə/ for everybody, but I've heard different Americans pronounce the word as [ˈdeɪ̯ɾə], [ˈdæɾɐ] or [ˈdɑɾʌ].

  10. Rodger C said,

    March 25, 2024 @ 10:46 am

    I don't pronounce a STRUT vowel of any kind in "what's," "gonna," or "'cause." West Virginia, born 1948, learned "standard" American as a 2d dialect. Plus, my STRUT vowel in my basilect is mid-back.

    My friend the novelist Silas House (from Eastern Kentucky) tells about how he learned that prejudice against Appalachians existed when he was at a writers' dinner in the Bluegrass and pronounced "butter" with a mid-back vowel, and the table full of so-called sophisticates mocked him.

  11. mg said,

    March 25, 2024 @ 10:50 am

    Point of information: Randall Monroe is American, living in the Boston area (but doesn't have a Boston accent – I'm not sure where he's from originally). So when you read this, hear it in American English.

  12. Coby said,

    March 25, 2024 @ 11:12 am

    The French have an old tradition (going back to the 19th century, maybe earlier) of pronouncing the STRUT vowel in English loanwords as /œ/. I wonder if this represents an older English pronunciation similar to that of a stressed schwa.

  13. David L said,

    March 25, 2024 @ 11:54 am

    This has puzzled me for a long time. M-W equates the schwa with what I learned to call the 'short u.' But that doesn't seem right to me. The word 'abut' illustrates the difference – when I say it, I hear a difference between the initial schwa and the stressed vowel, and my speechifying organs are in a different configuration when I make the two sounds.

    (I'm from the UK although have spent most of my life in the US).

  14. Philip Taylor said,

    March 25, 2024 @ 11:56 am

    For me (native speaker of southern <Br.E>, ≈ "low-RP") —

    Megan: /hwɒts ʌp ? wɒz dʌɡ ˈɡoʊɪŋ tʊ kʌm? dʌɡ lʌvz brʌntʃ./
    Ponytail: /noʊ, dʌɡz stʌk bi ˈkɒz əv ə ˈtʌnəl əbˈstrʌkʃən. ə trʌk dʌmpt ə tʌn əv ˈʌn jənz./
    Megan: /ʊx./

  15. David Marjanović said,

    March 25, 2024 @ 12:59 pm

    The French have an old tradition (going back to the 19th century, maybe earlier) of pronouncing the STRUT vowel in English loanwords as /œ/. I wonder if this represents an older English pronunciation similar to that of a stressed schwa.

    I think it dates back to a time when the STRUT vowel (in some RP-oid accent) was actually [ɵ] or something like that.

    [ɵ] seems to be the most common value for "schwa" in Germany.

  16. J.W. Brewer said,

    March 25, 2024 @ 1:12 pm

    Now I'm wondering if the first syllable of Munroe's surname is stressed in pronunciation or not, which (on some theories but maybe not all!) could affect whether it has a "schwa." According to the internet, Munro was born in the Lehigh Valley region of Eastern Pennsylvania and attended high school in the suburbs of Richmond, Virginia. I don't know at what age he moved from the first location to the second (or whether there were any intermediate stops). In any event, my impression is that the Richmond suburbs have in recent generations had a high enough percentage of transient Yankee residents that you can grow up there without acquiring any trace of an actual "Southern" accent, esp. if your parents/grandparents don't have such an accent.

  17. Philip Taylor said,

    March 25, 2024 @ 1:47 pm

    Having searched the thread in vain for any mention of "Munroe"/"Munro", I now assume that he to whom you refer is the cartoonist, in which case I have no idea which syllable might take primary stress. However, in the context of the Scottish Munros (so named after Sir Hugh Munro, who first published his table of mountains over 3,000 feet in 1891), the stress is definitely on the second syllable (/mʌn ˈroʊ/).

  18. Jarek Weckwerth said,

    March 25, 2024 @ 2:22 pm

    @ David Marjanović: Rather, this analysis is obviously correct for some Englishes and obviously incorrect for others.

    That's a brave statement ;)

    I would say it's one of the classical grey area examples in phonemic theory. Classically, to decide that two articulations belong to two different phonemes, you need minimal pairs. For distributional reasons (occurrence in stressed vs. unstressed syllables), finding those for STRUT vs. commA is rather difficult.

    But there are other factors where the analysis is primary, and whether the two things sound the same is of secondary importance. For example, STRUT is a checked vowel (limited to closed syllables). It's part of the strong vowel system. Native speakers tend to think of it as "short u" (see e.g. David L above). It alternates with MOUTH as in pronounce > pronunciation. Etc. etc.

    That's what I wanted to imply when I said some people don't buy it ;)

    And wrt US accents that don't merge the two — isn't California / general West another one? In my mind, I tend to associated the "stressed schwa" quality of STRUT with stereotypical country music. Lurve babe!

  19. Jerry Packard said,

    March 25, 2024 @ 2:28 pm

    I get a clear 3-way contrast with (IPA)

    poot = [p’Ut]
    put = [p’ʊt]
    putt = [p’ʌt]

    And I guess [ə] might be seen as the unstressed [ʌ] (or vice-versa) though I really don’t seem to distinguish them in my output (western Mass, eastern NY dialect area).

  20. Philip Taylor said,

    March 25, 2024 @ 2:31 pm

    I concur with two of your three, Jerry, but have [p’uːt] for "poot".

  21. Jonathan Smith said,

    March 25, 2024 @ 2:52 pm

    Re: French, English /ə/~/ʌ/ represented as [ø] or [œ] seems (to me) to be typical of French accented English, that is, the lips are definitely involved… I had thought of this as a lesson that even doing literally nothing with the vocal apparatus is by no means easy when one is unaccustomed…

  22. J.W. Brewer said,

    March 25, 2024 @ 3:06 pm

    The cartoonist's name is in fact Munroe (my Munro variant was a glitch as to him, although that's also extant). He was misspelled as Monroe upthread but that's perfectly understandable because "Monroe" is an order of magnitude more common in the U.S. than "Munroe" is, and perhaps even more well-known than other surnames of comparable frequency of occurrence since it was the surname of our fourth president.

  23. Philip Taylor said,

    March 25, 2024 @ 3:11 pm

    You don’t think that a certain Marilyn Monroe might also be partially responsible for the occasional inadvertent substitution of "Monroe" for "Munro[e]" in American English, JWB ?

  24. Terry K. said,

    March 25, 2024 @ 3:55 pm

    Note that the comic is not saying that all English speakers only use one vowel sound in those words. Just that it's not necessary to learn more than one vowel sound to speak English if you stick to a limited selection of words.

  25. Philip Taylor said,

    March 25, 2024 @ 4:21 pm

    I would like to hear a recording of someone who uses schwa universally throughout the monologue to see if, for someone who would normally use several different vowel sounds within the passage, the meaning is immediately clear. Just because the speaker is "speaking English" does not imply that he or she would necessarily be understood by other English speakers …

  26. Philip Taylor said,

    March 25, 2024 @ 4:28 pm

    Sorry, it is, of course, a dialogue. Mea culpa.

  27. Jerry Packard said,

    March 25, 2024 @ 5:38 pm

    @Philip Taylor
    Yes, my high back rounded vowel is also lengthened.

  28. JoshR said,

    March 25, 2024 @ 7:14 pm

    Phillip Taylor asked,

    "You don’t think that a certain Marilyn Monroe might also be partially responsible for the occasional inadvertent substitution of "Monroe" for "Munro[e]" in American English, JWB ?"

    I don't see how Marilyn would be more responsible than the 5th president and originator of the Monroe Doctrine, which Americans all learn about in school, and the fact that the Monroe spelling is just far more common. Especially for anybody born after 1960, for whom Marilyn's fame and appeal are more informed, rather than experienced.

  29. Chris Button said,

    March 25, 2024 @ 7:53 pm

    @ Jarek Weckwerth

    I would say it's one of the classical grey area examples in phonemic theory.

    I think it's just another case of Ladefoged's "phonemic conspiracy". In this case, we should just be talking about schwa as the default feature of syllabification. Instead, we search in vain for minimal pairs that don't exist.

  30. Chris Button said,

    March 25, 2024 @ 8:02 pm

    So, for example, where's the schwa phoneme in a syllabic sonorant? Does writing it with an accompanying superscript schwa make it any more "vocalic" than writing it with an accompanying vertical line below (or above) it?

  31. JPL said,

    March 26, 2024 @ 12:18 am

    "Let's hear it for the mighty sound of THEE famous Sonic Boom of the South!" Ya gotta have that distinction.

  32. Jon said,

    March 26, 2024 @ 12:36 am

    Philip Taylor –
    There was a Two Ronnies sketch in which Ronnie Barker, apparently posh, spoke every vowel as a schwa if I remember rightly.

  33. Philip Taylor said,

    March 26, 2024 @ 5:41 am

    Not familiar with the sketch, Jon, but Steven Dodd once informed me that in high RP all four of "tyre", "tier", "tear (vb)" and [a fourth which I now forget] could take the same vowel sound (/ɪə/) and therefore be aurally indistinguishable. I therefore wonder whether RB was actually using /ɪə/ throughout rather than a simple /ə/.

  34. David Marjanović said,

    March 26, 2024 @ 10:14 am

    Having searched the thread in vain for any mention of "Munroe"/"Munro", I now assume that he to whom you refer is the cartoonist

    Yes; his name is Randall Munroe.

    (And thank you – I would not have guessed the last-syllable stress.)

    I would say it's one of the classical grey area examples in phonemic theory. Classically, to decide that two articulations belong to two different phonemes, you need minimal pairs. For distributional reasons (occurrence in stressed vs. unstressed syllables), finding those for STRUT vs. commA is rather difficult.

    Yes, but the trick is that the accents that have two separate phonemes here do allow unstressed /ʌ/. For example, both vowels in undone are /ʌ/ in generic southern British English including RP.

    I recommend these two videos by an actual phonetician (I just play one on the i[ɾ̃]rtoobz) – whose native accent has the merger-by-redistribution, but the accent he uses in his videos keeps /ʌ/ and /ə/ separate.

    Missing in these videos is another vaguely transatlantic difference: words like culture. In accents with the merger, the (stressed) vowel is /ə/. The resulting sequence /əl/ is treated like any other /əl/ that lies entirely within a syllable; this means the /ə/ can phonetically drop out, leaving a syllabic [ɫ̩], or it can be backed & rounded by the /l/, resulting in [ʊɫ] and suchlike. In accents without the merger, the stressed vowel is /ʌ/, which does not interact with the following /l/ – you simply get [ʌl] at the phonetic level.

    And wrt US accents that don't merge the two — isn't California / general West another one? In my mind, I tend to associated the "stressed schwa" quality of STRUT with stereotypical country music. Lurve babe!

    There are very few accents of English that have a stressed [ə]. Whether /ʌ/ and /ə/ are separate phonemes is orthogonal to that: the claim is that most of the accents with the phonemic merger treat [ʌ], which is only occurs stressed in these accents, and [ə], which only ever occurs unstressed in these accents, as the stressed and the unstressed allophone of a single phoneme.

    (…with some further phonetic variation on the unstressed side as mentioned.)

    Note that the comic is not saying that all English speakers only use one vowel sound in those words. Just that it's not necessary to learn more than one vowel sound to speak English if you stick to a limited selection of words.

    If you actually do that on the phonetic side, you'll sound rather… Welsh, as far as I understand.

    where's the schwa phoneme in a syllabic sonorant?

    A syllabic sonorant is a phonetic phenomenon. Whether it should be analyzed as a consonant or as a sequence of vowel phoneme + consonant phoneme (or the other way around) depends on the language/dialect/accent.

    In the kinds of English I've heard, syllabic sonorants can be, and most likely should be, analyzed as sequences of /ə/ + consonant, because:
    1) there is an /ə/ in the system that also occurs in other environments and often surfaces as [ə];
    2) even individual speakers are remarkably inconsistent (again: in my limited experience) in whether they put a vowel in or not.

    In my kinds of German (both Austrian Standard German and a Central Bavarian dialect), syllabic sonorants should not be analyzed as sequences of vowel + consonant, and cannot be without great theoretical convolutions, because:
    1) there is no /ə/ in the system (and no [ə] either) – unstressed e is either unreduced [ɛ] or absent, there's nothing in between;
    2) with full consistency, no vowel ever surfaces, unless people want to be extra-clear (as in shouting over the telephone) and use spelling-pronunciations with [ɛ].

  35. J.W. Brewer said,

    March 26, 2024 @ 10:57 am

    I'm sure the more recent Monroe doesn't hurt the prominence of "Monroe" via "Munroe" but it's probably overdetermined. I didn't know or couldn't recall the backstory of how she came to get her stage name, but "Monroe" turns out to have just been the maiden name of her mother, the much-married Gladys Pearl Monroe Baker Mortensen Eley. (MM's biological father was a fourth fellow with none of those surnames.)

  36. Chris Button said,

    March 26, 2024 @ 1:59 pm

    @ David Marjanović

    For example, both vowels in undone are /ʌ/ in generic southern British English including RP.

    A better test might be be "until" since the first syllable may be reduced to schwa or not as the case may be.

    A syllabic sonorant is a phonetic phenomenon.

    Im not sure what you mean. A syllabic sonorant is what it is. A phonetic analysis could discuss how some speakers might pronounce it with a schwa and a non-syllabic sonorant as opposed to as a syllabic sonorant. But on an underlying level, this is straying into phonological questions around what is a syllable and how a sonorant can bear the schwa (i.e. the syllable) while an obstruent cannot–hence the schwa is no longer optional.

    There are very few accents of English that have a stressed [ə].

    Square brackets are used for broad or narrow phonetic transcriptions of how things may surface when uttered depending on how precise you need to be. Do you mean a stressed /ə/ with phonemic slashes here?

  37. Terry K. said,

    March 26, 2024 @ 4:11 pm

    If you actually do that on the phonetic side, you'll sound rather… Welsh, as far as I understand.

    Depends on the "you". I'd sound like an American, speaking normally. There's two sounds in there, plus one in the "title text" (the mouse-over text), that I might sometimes pronounce differently. The first vowel in "obstruction" might get pronounced to rhyme with Bob (a case of over-articulation, I think). And onion and cousin could possibly get an /ɪ/, but schwa and/or syllabic n is also very normal. In other words, the vowel sound there might very in a way that doesn't matter there, but does for distinguishing, for example, Rosa's, Rosie's, and Roses.

    A perhaps important point is that length is not phonemic in American English. The stressed vowels being longer does not make them a different vowel.

  38. Benjamin Ernest Orsatti said,

    March 26, 2024 @ 5:21 pm

    Terry,

    That's the thing that perplexes me so much about certain varieties of British English — take the innocent-enough-sounding word "home". Here in Western Pennsylvania, I believe the IPA for this word, as pronounced, would be "hōm", as in, "D'jinz gō hōm æftər ðə Stiɬɬər gæm?" But in certain varieties of (what I guess to be posh) British English, they've seen fit to stuff ALL of the vowels into that poor word, so that it comes out sounding like "haeioum".

    Why do we need all those vowels in there?

  39. Jonathan Smith said,

    March 26, 2024 @ 7:40 pm

    @Benjamin Ernest Orsatti
    hm your representations remind me of "phonetic triphthong" descriptions of U.S. "southern drawl," for which Wikipedia on the subject offers e.g. "sat" [sæi̯ət], "set" [sɛi̯ət], and "sit" [sɪi̯ət].

  40. Chris Button said,

    March 26, 2024 @ 8:26 pm

    "Welc.ome" [ˈwɛlk.əm] vs "be.come" [bɪˈkʰʌm] is a nice one. The stress placement conditions the aspiration and hence the syllable break too.

    It's probably worth noting that the strut vowel is most commonly ɐ rather than ʌ, so we should really be writing [bɪˈkʰɐm] rather than [bɪˈkʰʌm].

    The John Wells take from his Longman Pronunciation Dictionary: "LPD distinguishes between the vowels ʌ and ə, although in AmE they can generally be regarded as allophones of the same phoneme, and for some speakers are more or less identical phonetically too."

  41. John Swindle said,

    March 27, 2024 @ 12:34 am

    @Benjamin Ernest Orsatti: No u-offglide in Western Pennsylvania "home"?

  42. Philip Taylor said,

    March 27, 2024 @ 2:34 am

    Benjamin — "Why do we need all those vowels in there?" — /ɛʊ, sɛʊ ðɛt wi kɛnɒt 'pɒːsɪblɪ bɪ mɪstaɪkən fʊə hɔɪ pɒlɔɪ, ɒv kɒːs …/.

  43. Heddwen Newton said,

    March 27, 2024 @ 4:27 am

    Dr Geoff Lindsey did a great video on this last year, highly recommended for this subject, and funny as well :-) https://youtu.be/wt66Je3o0Qg?si=TL0apYg1i_wqOjwV

  44. Benjamin E. Orsatti said,

    March 27, 2024 @ 7:43 am

    John,

    Dispensing with the u offglide and reducing most vowels to "æ" is one of the distinctive features of W.Pa.Eng.! E.g., "dauntaun" –> dæntæn. Curiously, there does exist an example of vowel proliferation, and that's in the word "no", which becomes, "'næʊ, 'hɔʔɔ".

    Going to school in Philly, there were u offglides everywhere, and I kept getting them on me with the result being that I had to shower much more frequently than at home.

    Philip,

    I struggled with your transcription until I read it in Sir David Attenborough's voice; then it clicked immediately! Spot on, mate!

  45. Benjamin E. Orsatti said,

    March 27, 2024 @ 8:06 am

    Heddwen, Ah,, the “punchline” is at 9:15 in the video. I get it now! Incidentally, the Northern British accent almost seems like a non-rhotic version of general Am.Eng.

  46. David Marjanović said,

    March 27, 2024 @ 8:54 am

    this is straying into phonological questions

    Exactly. I'm trying to keep the phonetic and the phonological questions apart.

    a sonorant can bear […] the syllable[…] while an obstruent cannot

    Obstruents can be syllabic nuclei; however, there are very few languages where that occurs on the phonetic level, and even fewer where it's inbuilt in the phonology rather than a matter of surface realization of other things.

    Square brackets are used for broad or narrow phonetic transcriptions of how things may surface when uttered depending on how precise you need to be. Do you mean a stressed /ə/ with phonemic slashes here?

    No, I really mean a stressed mid-central vowel as a matter of measurable phonetic fact. That occurs in very few accents of English (but, for example, half the accents of Welsh). A stressed /ə/ occurs in a lot more English accents (and usually surfaces as [ʌ], but not always).

    Depends on the "you". I'd sound like an American, speaking normally. There's two sounds in there, plus one in the "title text" (the mouse-over text), that I might sometimes pronounce differently. The first vowel in "obstruction" might get pronounced to rhyme with Bob (a case of over-articulation, I think). And onion and cousin could possibly get an /ɪ/, but schwa and/or syllabic n is also very normal. In other words, the vowel sound there might very in a way that doesn't matter there, but does for distinguishing, for example, Rosa's, Rosie's, and Roses.

    I'm talking about the stressed vowel of obstruction, the u.

  47. Chris Button said,

    March 27, 2024 @ 10:09 am

    No, I really mean a stressed mid-central vowel as a matter of measurable phonetic fact. That occurs in very few accents of English (but, for example, half the accents of Welsh). A stressed /ə/ occurs in a lot more English accents (and usually surfaces as [ʌ], but not always).

    I don't agree. To quote John Wells again:: "LPD distinguishes between the vowels ʌ and ə, although in AmE they can generally be regarded as allophones of the same phoneme, and for some speakers are more or less identical phonetically too." He then gives the example of "above" in AmE and transcribes it with two schwa.

    Exactly. I'm trying to keep the phonetic and the phonological questions apart.

    Respectfully, I still think you're blurring them. The occurrence of a nasal sonorant as a syllable in its own right can be part of a perfectly valid phonological analysis.

    Obstruents can be syllabic nuclei

    Technically yes. But you are straying into a very different discussion by bringing it up, and it is really besides the point here.

  48. Jarek Weckwerth said,

    March 27, 2024 @ 12:07 pm

    @ David Marjanović: these two videos by an actual phonetician I think it would be important to point out that it's the one and only Geoff Lindsey, by far, by a really long mile, the best phonetic presence on YouTube. Thank you @ Heddwen for providing the "citation" ;)

    very few accents of English that have a stressed [ə] "Very few" may be a bit strong. Lindsey gives his own accent as an example. The usual narrow-ish transcription of KIT in New Zealand and South Africa is [ə]. And Standard Southern English has stressed [ə] in gonna, 'cause.

    WRT unstressed STRUT: Any full vowel can occur in unstressed syllables*. That is, I think, slightly beside the point here. The phonemic question is, does it minimally contrast with the schwa? The online resource that I usually use for minimal pairs in so-called RP gives four very very stretched examples. So the evidence is weak.

    (*) There's also the reverse analysis, reportedly more prevalent in the US, where if you find an unreduced vowel, you assume it must bear some level of stress.

    @ BE Orsatti: the IPA for this word, as pronounced, would be "hōm" Well, to me this is American-style dictionary respelling, not IPA. If you want real IPA for this, it would be [hom]. I don't know if that is the case in Western Pennsylvania, but Philadelphia is very certainly one of the handbook examples of [əʊ] for GOAT in the US, so I'm puzzled by your reaction to English English realizations, unless you mean pretty advanced "Estuary". I hear central-onset GOAT from American speakers all the time, in particular in the South and California.

    @ Terry K: onion with syllabic n? I really don't think so, unless it's THREE syllables long. The main sequence would be (1) tongue tip on the alveolar ridge for the first [n]; (2) Tongue tip removed from the alveolar ridge for the "j"; (3) Tongue tip back on the alveolar ridge for the second [n]. So the actual phonetic sequence would be [ˈənin], because [j] is non-syllabic [i] followed by a vowel, and there's no following vowel. Therefore you can only have a syllabic [n] at the end if it forms an additional separate syllable of its own, thus [ˈə.ni.n̩], three syllables. If you want a non-syllabic [j], you must have a schwa after it, which makes the second [n] non-syllabic, [ˈən.jən].

  49. Benjamin E. Orsatti said,

    March 27, 2024 @ 12:26 pm

    Jarek said:

    "I don't know if that is the case in Western Pennsylvania, but Philadelphia is very certainly one of the handbook examples of [əʊ] for GOAT in the US, so I'm puzzled by your reaction to English English realizations, unless you mean pretty advanced "Estuary". I hear central-onset GOAT from American speakers all the time, in particular in the South and California."

    Philly / Pittsburgh // Kent / Liverpool.

    "Pittsburghese" prevails over Western Pennsylvania, Eastern Ohio, and the northern panhandle of West Virginia. Somewhere around Morgantown, it "blends" into "standard Appalachian", and by the time you get to Cleveland, you're more or less fully "Midwestern". The "Philly" accent is within the same Sprachbund as New Jersey and New York, and probably starts around Lancaster or so (with Central PA having its own Pennsylvania Dutch-influenced dialect).

    I looked up "estuary" — that's probably the culprit behind "haeioum" (see Wikipedia entry on "GOAT" and "MOUTH" vowels in Estuary English). In Pittsburgh, the only thing that distinguishes the word "math" from "mouth" is vowel length.

  50. Terry K. said,

    March 27, 2024 @ 12:28 pm

    Jarek Weckwerth, notice I said "schwa and/or syllabic n". I was not claiming onion can be pronounced with a syllabic n. I was simply not making the distinction between pronunciation with a schwa or with no vowel at all, a distinction which is not important for the discussion of if a vowel besides schwa appears in these words, and a distinction which I'm not going to attempt to analyze.

  51. Jarek Weckwerth said,

    March 27, 2024 @ 12:38 pm

    @ Terry K: OK, point taken. Sorry for the confusion!

  52. Benjamin E. Orsatti said,

    March 27, 2024 @ 12:39 pm

    Myb t wld jst b bttr f w ddn't wrry bt vwls t ll xcpt n css f "strng vrbs", lk "swim/swam/swum", tc., whr thy srv grmmtcl fnctn.

  53. J.W. Brewer said,

    March 27, 2024 @ 12:58 pm

    Benjamin Orsatti's proposal of a Philly/NYC Sprachbund seems a bit eccentric, although maybe it depends on the feature under discussion. But the notably fronted GOAT vowel characteristic of the Delaware Valley that was under discussion is not at all present in NYC-area accents, although I don't quite know where the isogloss (if that's right word here?) runs through New Jersey.

    GOAT-fronting is also characteristic of the Baltimore area. It's about a 230-mile drive from Baltimore to Pittsburgh if you stay on the back roads where you can more easily stop and talk to people, but I have no very good sense of where along the way the distinctive Baltimore features disappear from local accents and where the distinctive Pittsburgh features appear and what may lie in between. At one point in the late 1990's I attended a wedding in Hagerstown but I didn't take advantage of the occasion to do any dialect fieldwork.

  54. Benjamin E. Orsatti said,

    March 27, 2024 @ 2:34 pm

    Other Keystoners, please correct me, but I'd say that, if you were carving up PA into various dialect districts,

    https://wiki.millersville.edu/download/attachments/36668375/pennsylvania_county_map.gif

    you'd draw one quadrilateral connecting Greene, Mercer, Jefferson, and Fayette ("Pittsburghese"); and another connecting Venango, Erie, Wayne, and Pike ("Northern Lake People Talk", except for Erie, where they talk like people from Buffalo). Everything within the triangle comprising Somerset, Centre, and Lancaster is a weird mélange of Pennsylvania Dutch, Appalächian, and other odd infusions (In Altoona, they call a single-family residential structure a "hæls"!).

    I dunno about what goes on further East, but at some point, they just start going crazy with diphthongs, like they'd just been discovered. Maybe "Sprachbund" is too strong a word, because even between Brooklyn and Long Island, there are significant differences; but in terms of vowel proliferation (e.g., MOUTH, GOAT), you might as well call all of PHL/NYC/NJ "Greater Diphthongia".

  55. Philip Taylor said,

    March 27, 2024 @ 5:10 pm

    « I looked up "estuary" — that's probably the culprit behind "haeioum" ». I think that that is rather unlikely, Benjamin — to my mind (and in my experience) Estuary English is the very antithesis of high-RP, and it is high-RP in which I believe that your "haeioum" is most likely to be encountered.

  56. Jon Lennox said,

    March 27, 2024 @ 8:54 pm

    @Andy: In my dialect the word "what" normally has the FATHER vowel /ɑ/, but the phrase "What's up" the word has a stressed schwa. I just noticed this.

  57. J.W. Brewer said,

    March 28, 2024 @ 1:21 am

    The GOAT vowel is a diphthong in normal/default GenAm pronunciation, typically represented as /oʊ/. A non-diphthongized alternative (exhibited not merely in Pittsburghese but e.g. in the Minnesota/"Fargo" accent*) is a noticeable regionalism. Sometimes the fronted Philly-to-Baltimore version is shown as /eʊ/, although maybe there are alternative notations where it's a more complex triphthong or something. The MOUTH vowel is likewise a dipthong in GenAm. Which isn't to say that there's no regional diphthongization going on, just that those aren't the right examples.

    *My own "Pittsburgh" grandmother was born and raised in St. Paul, Minnesota and did not relocate to Pgh. until she was in her early thirties, so that probably to some extent distorted my early sense of how Pittsburgh folks talk.

  58. Benjamin E. Orsatti said,

    March 28, 2024 @ 6:53 am

    It's the Vikings' fault! I'd venture to guess that the Nordic influence that monophthongified /oʊ/ into the Midwestern /å/ is the same one that "flattens" it into the Pittsburgh /o/ (i.e. the Danish influence on the language of the ancestors of the Scots-Irish folk who settled in W.Pa., who are (believed to be) responsible for our "quaint" manner of speaking.

    Oh, one more question (maybe) before the Fates cut this thread — do Britishers get confused by the similarity in pronunciation between Am.Eng. "can" /'kæn/ and "can't" /'kænʔ/ (i.e., whereas in Br.Eng. it's more like "can" /'kăn/ and "can't" /'kɒnt/?

  59. Jarek Weckwerth said,

    March 28, 2024 @ 11:09 am

    @ Benjamin: Are you familiar with the story of the Great Vowel Shift? If not, I recommend the very reasonable Wikipedia article for a start. Again, my knowledge of Western Pennsylvania is close to none, but since monophthongal GOAT is a conservative feature of English, the null hypothesis does not need other languages and it should be that it was carried into the area by speakers of conservative accents. (On the assumption that when those places were settled, the innovative diphthongal realization was standard on the coast which I'm absolutely not sure it was.) And since there are still plenty of such accents in the British Isles today (notably in much of the North of England, Scotland and parts of Ireland), then your logical expectation should be that there were even more in the 18th and 19th century. No Vikings or Dutch people needed; or perhaps only as a very secondary reinforcing influence. Only if you disprove that hypothesis can you consider the hypothesis of innovative monophthongization (which would be a reversal of an otherwise pan-English trend).

    In terms of isoglosses etc., I recommend this monster map by Rick Aschmann which seems to be mostly based on maps from Labov et al.'s Atlas of North American English (I think). The latter is of course a tour de force of American accent studies. If your local library has it, get your hands on it and be amazed.

  60. Philip Taylor said,

    March 28, 2024 @ 11:15 am

    Benjamin — « do Britishers get confused by the similarity in pronunciation between Am.Eng. "can" /'kæn/ and "can't" /'kænʔ/ (i.e., whereas in Br.Eng. it's more like "can" /'kăn/ and "can't" /'kɒnt/? ». "Britons", Sir, not "Britishers" ! I'm afraid that as a Briton I cannot answer your question from first-hand experience, but given your phonetic transcription it does seem extremely likely. For me the two are "can" (/kæn/) and "can’t" (/kɑːnt/).

  61. Chris Button said,

    March 28, 2024 @ 11:38 am

    Usually the schwa reduction in "can" takes care of it in connected speech since you wouldn't reduce the negative form "can't" to schwa (and it would incidentally also sound like a very rude word!). Very occasionally, there can be confusion though in my experience.

  62. Rodger C said,

    March 28, 2024 @ 12:02 pm

    I do seem to notice that Britishers, and especially L2 English speakers trained in RP, have trouble distinguishing "American" can and can't. But the Southern pronunciations conventionally written kin and and cain't are not only easily distinguishable but, I think, isomorphic with the RP pronunciations.

  63. Rodger C said,

    March 28, 2024 @ 12:03 pm

    "Kin" and "cain't", of course.

  64. /df said,

    March 28, 2024 @ 12:51 pm

    Prompted by the discussion, I read out the cartoon dialogue with all the vowels schwa-ified and found that I'd turned into Bəb Dələn. Is nt wrryng bt vwls the way to a Nobel Literature prize?

  65. J.W. Brewer said,

    March 28, 2024 @ 1:24 pm

    The Aschmann map FWIW uses "Atlantic Midland" as its label for the Philadelphia-Baltimore regional accent, has Pittsburghese as a subset of "Allegheny Midland," and then has those two "Midland" accents physically separated by "East Midland," which is shown as covering a territory that looks suspiciously gerrymandered (i.e. a notably non-convex shape squiggling amidst other more convex regions). So that addresses my question above re what would you hear when driving from Baltimore to Pittsburgh on the backroads in between the invisible line where the Balt-specific features fade out and second invisible line where the Pgh-specific features fade in.

  66. David Marjanović said,

    March 28, 2024 @ 1:51 pm

    I completely forgot… accents that, like RP, are non-rhotic and lack the STRUT-COMMA merger could be analyzed as having a NURSE-COMMA merger: [ɜ] as the stressed and [ə] (which can drop out) as the unstressed allophone of a single phoneme. Seven is a homophone of the river Severn.

    I don't agree. To quote John Wells again:: "LPD distinguishes between the vowels ʌ and ə, although in AmE they can generally be regarded as allophones of the same phoneme, and for some speakers are more or less identical phonetically too." He then gives the example of "above" in AmE and transcribes it with two schwa.

    Wells is actually very annoying for not distinguishing phonetic and phonological transcriptions. He didn't even use brackets or slashes, just IPA in boldface.

    Now, if he said "and for some speakers are more or less identical phonetically too", I can't simply contradict that; but I have to say that "more or less" without any further detail is hard to evaluate. With my linguistic background I have much more difficulty distinguishing [ʌ] from [ɑ] than from [ə]; I'm sure that's not true of any native speaker of most kinds of English (maybe there are some in India or Nigeria).

    The occurrence of a nasal sonorant as a syllable in its own right can be part of a perfectly valid phonological analysis.

    As I tried to demonstrate, I think it is part of a perfectly valid phonological analysis for some languages – but not for others, and the kinds of English I'm halfway familiar with seem to be in the latter category.

    Technically yes. But you are straying into a very different discussion by bringing it up

    You brought it up by explicitly denying it. :-|

    very few accents of English that have a stressed [ə]

    "Very few" may be a bit strong. Lindsey gives his own accent as an example.

    His native accent, not the accent he uses in his videos; and on the phonological level, but we're never told about the phonetic one. Being John Wells's former student, Lindsey doesn't distinguish the two until he really has to (which is the case later in the video).

    The usual narrow-ish transcription of KIT in New Zealand and South Africa is [ə].

    This I agree with; I wasn't thinking around enough corners, so to speak! Scottish accents do something similar.

    WRT unstressed STRUT: Any full vowel can occur in unstressed syllables*. That is, I think, slightly beside the point here. The phonemic question is, does it minimally contrast with the schwa? The online resource that I usually use for minimal pairs in so-called RP gives four very very stretched examples. So the evidence is weak.

    Fair enough. The site doesn't list undone, though, and Lindsey talks about it in one of the two videos.

    In Altoona, they call a single-family residential structure a "hæls"!

    Is that the same phenomenon as the L in Bristol?

    similarity in pronunciation between Am.Eng. "can" /'kæn/ and "can't" /'kænʔ/

    Phonetically it's even worse: add some prenasal raising, and you get [k̟e͠ən] vs. [k̟e͠ənʔ̚] or [k̟e͠əɾ̃]…

    (I don't think a lengthened tilde over nasalized diphthongs is IPA standard, but it's available in Unicode and I like it…)

    Are you positing a phonemic glottal stop, BTW?

  67. Terry K. said,

    March 28, 2024 @ 2:44 pm

    In American English, as I've experienced and speak it, "can" without extra stress (emphasis), can has a schwa, "can't" doesn't. And to go with that, "can" is unstressed, whereas "can" has stress.

    When emphasized (extra stress), can has no T, and can't does have a a T. Not sure if it's always a fully released T, but certainly often enough in this emphasized form.

  68. Benjamin E. Orsatti said,

    March 28, 2024 @ 3:24 pm

    Jarek,

    Darnit, I forgot all about the Griet Vawol Sheft and saw zebras when I should have been looking for horses. Thanks! As for the Aschmann map — wow! There's even an embedded link to the "Pittsburgh Dad" videos! (http://www.youtube.com/watch?v=MJfNI9DppP4). As for the map itself, though; I'd nudge the "Pittsburgh" oval a little further to the southwest, so that it grabs Wheeling, but not State College.

  69. J.W. Brewer said,

    March 28, 2024 @ 3:54 pm

    Stressed "can" comes out with a variety of different vowels for different Americans. I think for me it's usually /ˈkɛn/, which wiktionary associates with Philadelphia AND New Zealand. For Americans more distant from Philadelphia it offers "/ˈkæn/, [ˈkʰæn], [ˈkʰɛən ~ ˈkʰeən] (see /æ/ raising)." I think my "can't" is boring old /kænt/, although maybe I can't exclude in certain context the possible utterance of the more interesting option "(æ-tensing) IPA(key): [kʰẽə̃nʔ ~ kʰɪ̃ə̃nʔ]."

  70. Chris Button said,

    March 28, 2024 @ 4:19 pm

    @ David Marjanović

    Wells is actually very annoying for not distinguishing phonetic and phonological transcriptions. He didn't even use brackets or slashes, just IPA in boldface.

    This might help from p.609:

    "The phonetic notation in this dictionary is phonemic, with the following minor exceptions …"

    And one of those exceptions is:

    "For some speakers (not of RP), ʌ and ə are not in contrast."

    You brought it up by explicitly denying it.

    tsk, I stand by my statement that schwa is the default feature of syllabification.
    But ssshhh, let's not get into varieties of Berber.

  71. Sidney Wood said,

    March 30, 2024 @ 8:14 am

    Many comments above referred to STRUT pronunciations, both in England and America. The FOOT-STRUT split and the TRAP-BATH split were unique to SBE (both regionally and in RP), distinguishing SBE from NBE (spoken in the northern counties of England). The FOOT-STRUT split was completed early in the 17th century, and would have been taken to the American colonies by earlier settlers, to Massachusetts by migrants from East Anglia and Kent in 1629-1640, and to Virginia in 1642-1675 from the south and southwest (Fischer, "Albion's Seed", 1989). Fischer's remaining migration waves, from the Midlands and Northern England in 1675-1775, wouldn't have acquired either sound change anyway. The TRAP-BATH split wasn't complete in SBE until late 17th century, too late for any SBE migrants, so TRAP hasn't been raised towards [ɛ] in American speech. TRAP and STRUT have now changed again in SBE, regionally and in RP. Both are now fully open (F1 is higher than 600 Hz). TRAP is open front (no IPA transcription, why not 'open [æ]'?). Jones got Cardinal 4 [a] wrong by putting it at open front, but that's a different story. It's really open central, but Jones' hands were tied by his abolition of open central. STRUT has been lowered from [ʌ] to [a], regionally and in RP (the earliest speaker I've heard was born in the 1850s, and by Daniel Jones himself born in 1880).
    My own ancestors didn't migrate to Massachusetts 400 years ago, but stayed behind in Molash, acquiring Estuary English in Faversham and on Sheppey in the 19th and 20th centuries..

  72. Chris Button said,

    March 30, 2024 @ 9:01 am

    @ Sidney Wood

    Regarding [ʌ] being [a], presumably you mean inverted [ɐ] rather than [a] (to my earlier comment)?

  73. David Marjanović said,

    March 30, 2024 @ 10:07 am

    TRAP is open front (no IPA transcription, why not 'open [æ]'?). Jones got Cardinal 4 [a] wrong by putting it at open front, but that's a different story. It's really open central, but Jones' hands were tied by his abolition of open central.

    How is [a] defined in the IPA?

    The story I've patched together about the meaning of a in the IPA is that [a] vs. [ɑ] is supposed to represent the contrast found in conservative French varieties, so [a] is supposed to be open front (the bottom left corner of the vowel chart). I agree, unsurprisingly, that this [a] occurs as TRAP in some Englishes but not RP or GenAm. Even though open central vowels are much more common than open front or back ones worldwide, there's no IPA symbol for them because there doesn't seem to be a sound system that distinguishes more than two unrounded fully open vowels, so the "International Phonemic Alphabet" says you can write any front-mid or mid-back contrast as front-back [a]-[ɑ], or press [ɐ], the symbol for the near-open unrounded mid vowel, into service.

  74. Sidney Wood said,

    March 30, 2024 @ 10:14 am

    @Chris Button
    No I meant a sound change: former [ʌ]-like STRUT is now (since at least the 1850s) [a]-like. I believe the [ɐ]-like STRUT was Gimson's (among others) way of illustrating the earler /ʌ/ for STRUT:/ʌ/ is pronounced [ɐ] (without being a sound change). A shift from mid [ʌ] to low [a] (or half open to open) is large enough to be seen as a sound change. The articulation would require exchanging the upper pharyngeal constriction of [ʌ] for the lower pharyngeal constriction of [a], using different muscles.

  75. Sidney Wood said,

    March 30, 2024 @ 11:08 am

    @David Marjanović
    The IPA is reproducing and defending Daniel Jones' original decision, adopted from Passy around 1906-1907. Passy's version of the Bell vowel model (speculated tongue movement, height vs backness, from Visible Speech) required only front vs. back for French, overlooking that French /a/ is only relatively more front than French /ɑ/, not absolutely. The IPA's own illustration of French describes /a/ as central (by Fougeron & Smith). Since 1907 we've had Russell (1928), Chiba & Kajiyama (1941, translated 1959), Fant (1960), and the invention of the Spectrograph. So it's high time to update and revise Jones' conclusions.
    Wells (1982) revised RP TRAP to /a/, accepting Jones' cardinal 4 which is really open central and is needed for new open STRUT.

  76. Chris Button said,

    March 30, 2024 @ 9:47 pm

    @ Sidney Wood

    I think [ɐ] rather than [ʌ], and [a] rather than [æ], would be a more standard analysis.

  77. Chris Button said,

    March 30, 2024 @ 10:03 pm

    [a] by and large just as a more open [ɐ].

  78. David Marjanović said,

    April 3, 2024 @ 10:58 am

    I could and should have looked up much earlier where the problem lies: Wikipedia says that only [i], [ɑ] and [u] are defined as corners of the vowel system in terms of humanly possible articulations; I had assumed [a] is as well, but it's not. "The other vowels are 'auditorily equidistant' between these three 'corner vowels', at four degrees of aperture or 'height': close (high tongue position), close-mid, open-mid, and open (low tongue position)." It's not surprising that this has led to confusion!

    I also see there's this article that I should have read earlier…

RSS feed for comments on this post