Friday, 22 January 2010

aw, shucks

Phonetic symbols are not magic. And there is no super-phonemic system somewhere in the sky, consisting of universal sounds in one-to-one relationship with IPA symbols. Rather, the IPA (or any other symbol system) offers us a battery of rather vaguely defined symbols on which we can draw in order to refer to human speech sounds in the real world, or to abstractions from them.
Faced with an unknown language, or in the phonetics classroom with general-phonetic nonsense words, we produce an ‘impressionistic’ transcription that makes no prior assumptions about phonemic categories. We can, if we choose, use all manner of unusual letters and diacritics to record subtleties that we may have observed. On the other hand we may start out pretty unsubtle and then subsequently need to refine our transcription (make it “narrower”) if we find that we are overlooking subtleties that turn out to be important in the language concerned.
But when we want to use IPA symbols in a dictionary or language textbook, that sort of thing won’t do. There, we need a simple straightforward system that is not burdened down by unnecessary complications. In transcription systems there is always a trade-off: we can either show subtle details in the “text” (the dictionary entries, the transcribed passages), which makes the text complicated and difficult to follow — particularly for language learners who do not wish to be phonetics specialists — or we can reserve these details to the “conventions”, the rules for interpreting particular symbols in context in the language in question.
That is why pedagogical transcriptions are normally phonemic rather than allophonic.
Jerry Friedman wrote
I hope you can answer a question that I've asked unsuccessfully on
linguistics forums. Is there an IPA symbol for the General American realization of /ɔ/ that distinguishes it from RP /ɔ/, not to mention French and other versions? (My pronunciation of THOUGHT-class words and some others is pretty close to GenAm.)
while Bao Zhi-kun, a Chinese teacher of English, wrote
We know that the vowel in the words like “talk”, “call”, “law”, “bought” is pronounced differently in BrE and AmE: /ɔː/ in BrE but /ɒ/ in AmE. Why is this difference not shown in your LPD?

So the answer to Jerry Friedman is that there are several possible symbol choices that would bring out this difference. (Let’s overlook the problem of defining what exactly might be meant by “the” General American realization of the THOUGHT vowel. And let’s ignore the fact that about half of all Americans use their LOT vowel in THOUGHT words.) We could attach a diacritic to the RP symbol to show that the vowel is closer or more rounded than the GenAm: we could write ɔ̝ː or ɔ̹ː. Or we could even decide to write it . Or we could attach a diacritic to the GenAm symbol to show it as opener or less rounded: ɔ̞ or ɔ̜. (It is not typographically practicable to place two inferior diacritics toɡether on this symbol.) Or we could decide to write it ɒ (though that would create confusion for those accustomed to the use of this symbol for the intrinsically short vowel of RP LOT). If we were concerned only with AmE, we would use no length marks, since the length distinctions of RP are irrelevant in GenAm.
In the first edition of LPD I actually represented the AmE THOUGHT vowel as ɒː, differently from BrE ɔː, which would have pleased Bao Zhi-kun. But in the second edition I decided to use the symbol ɔː for both varieties. Apart from anything else, this makes for simpler entries, since the same transcription of words like thought θɔːt and law lɔː will do for both BrE and AmE.
So the same symbol e will do for the very different qualities of the DRESS vowel that we find in various kinds of BrE and AmE. The same symbol æ will do for very divergent qualities of TRAP. And the same symbol ɔː will do for different kinds of THOUGHT.

Here’s a rather different kind of example: we use the symbol l for the lateral approximant of many different languages. But in reading a transcription of German you have to remember to make it very clear; in French clear, but not quite so clear as in German; for most kinds of English it is generally darker, and may vary in colouring according to position; in Korean it has a noticeably retracted place of articulation; in some languages it is dental rather than alveolar. We don’t want to be bothered with such detail at every relevant dictionary entry: it’s better to just learn and remember that such-and-such a language or variety has such-and-such a quality of sound.
OK, Jerry, the French open-mid back vowel is rather centralized. But in a pronunciation dictionary of French it still makes sense to write it ɔ, rather than fussing about with ɔ̈ or the like.

If you want to speak RP-ish BrE you need to learn to use the actual sounds that are appropriate. It’s not sufficient to draw on some general-phonetic battery of universal sounds. Likewise if you want to speak network-standard GenAm. You have to establish an entire setting for the sounds of one variety or the other, and then operate within that setting — not faff around with transcriptional complications.


  1. This is of course all very true. Additionally, I would think a vast majority of EFL learners will never encounter the IPA outside EFL teaching materals, and they will not use it for other languages. So all this detail is indeed unnecessary at this level. (That is, if you make the learner understand that the same transcription can actually stand for two noticeably different pronunciations. This is not so self-evident, especially for the uninitiated.)

    But by extension, the same can be said of any transcription system in this setting. So, at least to me, the main argument against e.g. Upton's system would have to be tradition and its minority status.

    Two additional advocatus-diaboli sub-comments: (1) Of course tradition will also be on the side of the various (typically American) "transliteration" systems. (2) If the cross-Atlantic differences in e.g. TRAP or THOUGHT are not worth showing, why have GOAT represented in two different ways?

  2. Very useful post and comments. These will help me in my work with private clients (both actors and civilians) and cast members.

    Using e for DRESS instead of ɛ continues to be a burr under my saddle. I do understand the historical reasons, but I think the time has come to let them go. The sound at the beginning of FACE is demonstrably not the sound in DRESS, in the "standard" (loaded word, I know) pronunciations of BrE and AmE.

    I run into a similar problem in LPD with listing ɝ for GenAm* for NURSE (pro) but not using ɚ as the second part of the r-colored diphthongs in START, NORTH/FORCE, NEAR, SQUARE, CURE (con). R-coloration is the norm in the US, and not differentiating ɹ (consonant) from ɚ or ˞ (vowel) causes difficulties. ʌv kɔɚs I can get around this, but it makes me cranky!

    I have resisted and resisted making these comments whenever the opportunity came up to make them, but I just can't fight it any longer. Thanks for letting me get them off my chest. I do realize that LPD has to serve more purposes than mine, but as I said, I'm cranky. Please forgive me!

    * There is, of course, no such thing. Incidentally, your final graf on page xxi confuses the heck out of me. Clearly I will have to study it more carefully.

  3. I left out one more difficulty with not using ɚ: father, standard are different from comma, the in rhotic American.

  4. I find it funny how dictionaries use different symbols for the GOAT vowel between American and British English yet the same symbol for the TRAP vowel. To my ears, the GOAT vowel is much more similar between the two than the TRAP vowel. Also, there are many people in Britain who do use /oʊ/ in GOAT words.

  5. I think this is a fair point about GOAT. Blame Gimson. On the other hand, just think of the people who would write to me complaining if I were to use the symbol for something that is increasingly neither rounded nor back.

  6. > the same transcription of words like thought θɔːt and law lɔː will do for both BrE and AmE.

    I agree that a phonemic transcription is simpler and removes unneeded redundancy that can be learned through various rules. However, I disagree with the above generalization. AmE has both /ɔː/ and /ɒː/ as separate phonemes. Lets assume that you transcribe AmE "law" [lɒː] and "low" [lɔː] as / lɔː/. There is no way to distinguish between the two. One could learn that /ɔː/ → [ɒː], but then not be able to properly pronounce AmE "low". Or they could learn that /ɔː/ → [ɔ ː], but then not be able to properly pronounce AmE "law". Outside of context, a native AmE speaker would have a difficult time understanding "law" [lɔː].

  7. @Anon at 16:09: law /lɔː/ is [lɒː] for some people, [lɑː] for others. Low is /loʊ/, generally [loʊ], with varying degrees of diphthongisation (e.g. [loː]) and centralisation of the first element. No danger of confusion.

  8. Amy: I share your irritations exactly.

    Anondude: Lets assume that you transcribe AmE "law" [lɒː] and "low" [lɔː] as / lɔː/

    Well, you could, but I wouldn't, because I don't say low anything like that. Whatever else it may be, it's certainly not a monophthong.

    John: Instead of dealing with the subtleties of English vowels, you could be struggling with Hawai'ian or Samoan instead of English, and deciding whether to write /t/ or /k/ in phonemic transcriptions!

  9. I suppose that ultimately it's a question of understanding your audience. I suspect that the majority of people who purchase a pronunciation dictionary like LPD are EFL/ESL learners or their teachers. These people are generally choosing a model to follow (either the BrE or AmE one) and sticking with it, though valuing the ability to compare the 2 transcriptions.

    Part of the challenge of what you call reserving the details for the "conventions" is that so many (lazy) readers don't bother to READ the darn things. So the fact that /ɔ/ can mean different things in different versions of English gets missed. So we assume that they mean the same thing when they do not, because the symbols look the same.

    Accent-for-actors coaches have until recently stuck to impressionistic transcriptions doggedly; it wasn't until the start of IDEA that we began, as an industry, to use the Lexical Set words to discuss vowels rather than phonemic transcriptions followed by phonetic ones (e.g. /ɔː/ → [ɒː] or some such--we now tend to say "THOUGHT = [ɒː]"). Thank you Doug Honorof! But many of us still use transcription of text as a teaching method (a quite effective one I might add) where we transcribe quite narrowly what's going on. If one has to switch between many different accents, this kind of narrow transcription is very handy, rather than trying to remember the "conventions" you're attempting to learn. By the time you're ready to go on stage, you better not be thinking about phonetic symbols! But when you're beginning, they can be very helpful.

  10. My own preference is to use length marks when transcribing all the major dialects (erring a bit on the broader side of things), so for the common North American realization of /ɔ:/, I use [ɒ:].

    I'm also a fan of using [ɑɚ], [ɛɚ], [ɪɚ], [ɔɚ], [ʊɚ]. As a Massachusetts native who makes the Mary-merry-marry distinction, I was always getting irritated with American dictionaries for using transcriptions like [fɛr] for "fair". This practice makes no allowance for the full range of dialectal variation, and it leads merged speakers to believe that their "merry" is identical to mine, with the difference lying in the pronunciation of "Mary", whereas I am strongly convinced that this is wrong.

    I've seen phonetic treatments of AmEng (by merged speakers) suggesting that unmerged northeasterners have a special phoneme [ɛə] (and likewise [ɪə], etc) that we use in "Mary" (with the absurd implication that we pronounce "mare" differently), and based on all my observation and intuition, I want to tell them that they have the phonemics completely wrong. I suggest that the merged and unmerged pronunciations of "Mary" are phonemically identical, and that the difference lies in "merry", with merged speakers pronouncing it as unmerged speakers pronounce "Mary".

    It all makes sense if you frame the phonemic difference thus: unmerged dialects allow sequences of checked vowels + /r/ (or more properly, "non-r-combining vowels" + /r/), whereas merged dialects don't. I think this is the only reasonable way to understand the package of "Mary-merry-marry", "hurry-furry" and "serious-Sirius" mergers. It means that we must consider /ɑɚ/, /ɛɚ/, /ɪɚ/, /ɔɚ/, /ʊɚ/ as free r-colored vowel phonemes, shared by all rhotic dialects, whether merged or unmerged. It is the sequences of [ærV], [ɛrV], [ɪrV], [ʌrV] and [ɒ:rV] which are unique to the unmerged dialects. (The last one is a bit anomalous as it doesn't involve a checked vowel per se; but it's the distinctive sequence that us Massachusetts folk use in words like "sorry" and "horror".) My apologies for the long post, but I think dictionaries do a disservice when they suggest that GA "bear" and "bet" use the same phoneme. It's a workable, if unsatisfying, analysis if we pretend that fully merged GA is the only dialect of English, but as soon as we move on to other dialects, I think it leads us to look at the phonemics in a completely wrongheaded way.

  11. As usual, Eric Armstrong delivers an admirably thoughtful contribution to the discussion. I hope I can be like Eric when I grow up.

    John said: "Likewise if you want to speak network-standard GenAm. You have to establish an entire setting for the sounds of one variety or the other, and then operate within that setting — not faff around with transcriptional complications."

    Though it would, for the finicky (who, me?), be less accurate than using ɹ and ɚ, I think using r for both would work well, but only if ɝ were abandoned. After all, you don't use both l and ɫ in LPD; you just use l, and we get the idea of "some kind of l-ness.") So you could have, under GenAm:

    NURSE = ɜr (or, perhaps, əːr - but now I'm fomenting outright rebellion)
    LETTER = ər
    START = ɑːr
    NORTH/FORCE = ɔːr
    NEAR = ɪər
    SQUARE = ɛər
    CURE = ʊər (whether CURE functions as a unified set any more seems questionable to me, but that is by the bye.)

    I'd still have to explain to nearly all my clients the idea of r as both consonant and vowel, and why I prefer to use ɹ and ɝ, ɚ, but I think having only one symbol for "some kind of r-ness" would work well for your larger readership - and I do think it would actually be less confusing to EFL learners than having to contend with ɝ but ər. I know it would be less confusing to me!

  12. @Lazar: Rock on!

    Although my speech is fully rhotic, as a New Yorker, I treat r between vowels as syllable-initial ɹ, not syllable-final ɚ. In other words, most Northeasterners still treat r between vowels the way most speakers of BrE do. Syllable-final ɚ allows for a variety of two-way and three-way mergers among the various American regional varieties. And the variety is considerable.

    I certainly have a hurry-furry split, say orange and Florida with ɑ.ɹV, not ɔɚ, and have a Mary-merry-marry split - although, oddly enough, not on the actual words Mary and marry, which are both ˈmæɹi in my idiolect. I can say ˈmɛəɹi with ease (and also ˈmɛɚi, if called for) , but I only do so in character as someone else. Fairy would be a better example of ɛəɹi in my speech.

  13. @Amy: Thanks! Like you, I make all the distinctions despite being fully rhotic. (In short, I would describe my dialect as GA phonetics on top of Eastern New England phonemics.) And I'm familiar with unmerged speakers using /æ/ in places where [ɛə] is found in RP. For example, I knew someone from Rhode Island who pronounced "vary" as ['væri], and myself I use ['pærənt] for "parent".

  14. @Amy:

    For you (or any other GenAm speakers with the 3-way split), which vowel is the first syllable in "Maryland"?

    I (near-RP California exile) always hear the TRAP vowel from GenAm speakers.

    Like most ignorant Brits, I had always imagined that "Maryland" was pronounced "Mary" + "land" -i.e. with successive SQUARE, happY and TRAP vowels. I was completely mystified when my first US roommate said he was from (what sounded like) "Marilyn" and had no idea what he was talking about for a while :)

  15. I find it very annoying how many Brits and Aussies seem to get offended when we (Americans) can't pronounce their placenames "correctly", but of course they can't pronounce our placenames right either.

  16. And you don't see us complaining about it near as much as they do (that's what I meant to add at the end of that last post).

  17. @vp: The standard unmerged pronunciation is ['mɛrələnd], with the DRESS vowel. To your point, I can confirm that I have heard ['mɛərilænd] several times from Brits. One example that comes to my mind is in the movie "Gettysburg": the English/Irish actor William Morgan Sheppard uses this pronunciation in the opening narration.

  18. Thank you for the very complete answer to my question.

    I felt I could use "General American" the way I did because of a sentence on page 145 of _The Accents of English, Vol. 1_ that compares GenAm /O/ and RP /O:/. (Sorry, I can't paste into this window, so I don't know how to get IPA and I'm using "Kirshenbaum" ASCII IPA.) italic

  19. In fact, I'm having more trouble with the interface. That was supposed to be a preview testing HTML tags, not a finished post!

    Anyway, a related point is that most Americans' (and my) vowel in NORTH and FORCE words sounds very different to me from that in THOUGHT words. (Are the main exceptions New York and eastern New England?) I would imagine this confuses EFL students, but I don't have any experience with that.

    So can I use [ɔ̝] (ah, that's better now that I have a profile) if I want to explain that I pronounce the last syllable of "centaur" differently from "tore" and the first syllable of "Taurus", and differently from "tar"? The other two words where I might have [ɔ̝r] are "dinosaur" and "Bryn Mawr", but I might say either with [ɑr]. (It's my mother's fault—she's from Pittsburgh.)

  20. @Jerry Friedman: In Eastern New England (excluding Rhode Island), the THOUGHT and LOT classes both use an open [ɒ:], while the FORCE set uses [ɔə] and the NORTH set uses either [ɔə] or [ɒ:] (the latter being more common in older speakers). In New York, THOUGHT, NORTH and FORCE all use either [ɔ:] or [ɔə]. Rhode Island is rather intermediate between the two dialects.

  21. @Lazar: Thanks for the correction. Are there other parts of the country that use the same vowel in the THOUGHT class as in the NORTH and/or FORCE classes?

  22. @John W

    > just think of the people who would write to me complaining if I were to use the symbol oʊ for something that is increasingly neither rounded nor back.

    Why choose oʊ? Increasingly it won't do for BrE, as you say, but it looks a bit as if əʊ increasingly will do for AmE. Not least because an increasing number of AmE speakers actually use something like that, I suppose because it has evolved independently along the lines of BrE əʊ, but perhaps also to some extent because it has been influenced by it. I feel sure some AmE speakers affect some of the BrE values for it.

  23. I don't think it was influenced by BrE. That's kind of silly.

  24. @Jerry: There are some pockets in the South where the NORTH-FORCE distinction is maintained, although I can't say whether this causes NORTH and THOUGHT to coalesce as in New England - I'm not as knowledgeable on southern dialects as on northeastern ones.

    @mallamb: For what it's worth, I pronounce that phoneme as [ɤʊ], with a back but unrounded onset. My view of things may well be distorted, but I perceive [ɤʊ] as the mainstream realization, and [oʊ] as slightly marked (evoking either Midwestern or Canadian speech, or the old-fashioned General American of a 1950s movie). But like Anonymous, I quite doubt that British influence had anything to do with it.

  25. I assuming by "Midwestern" you mean Upper Midwestern. If that's the case, then I agree. I think Sarah Palin pronounces it the second way you mention and that pronunciation is salient (if that's the right word) to a lot of Americans, including me (and even her fellow Alaskans for that matter). I think of my GOAT as pretty "mainstream" and I hear my onset as being more of an open-mid back unrounded vowel, but it may be more like the way you describe it, as you very well may have better ears than I do.

  26. I meant to say "I'm" at the beginning.

  27. @Anonymous: Yes, Upper Midwestern.

  28. I think a number of American dictionaries used to treat the New York usage as the default. I remember struggling to figure out dictionary pronunciation codes that used ô for both THOUGHT and FORCE words. Completely senseless for a Californian who pronounces all THOUGHT and LOT words with ɑ.

  29. @mallamb:

    Labov et al. give statistics for the relative numbers of speakers in the US with what they define as a fronted value for the GOAT vowel (in checked position). They find it as characteristic of the South, South Midland and Philadelphia regions: I would estimate that this represents around 1/3 of the USA's population.

    Of course in Britain there are plenty who don't have a fronted vowel in northern England and the Celtic countries.

  30. @Lazar:
    Thank you. I had been thinking [oʊ] was slightly marked. But I wasn’t making any great claims about British influence (WHAT British influence?) I was quite circumspect actually:
    it looks a bit as if əʊ increasingly _will do_ for AmE. Not least because an increasing number of AmE speakers actually use _something like that_, I suppose because it has evolved independently along the lines of BrE əʊ, but _perhaps also to some extent because it has been influenced by it._

    And when you consider what it _will do_ for in BrE, even within the relatively restricted range of RP and near-RP, doesn’t it seem increasingly uncontentious for AmE? Schwa alone covers a huge area, and with ə being used for both BrE ə and ʌ, and the preference for ʌ for both which has been expressed on here, the only strong argument for AmE oʊ is from tradition, and that didn't stop the Brits taking up əʊ while Americans were still using y and w for diphthong codas!

    Don’t take my observation about some AmE speakers affecting some of the BrE values for it too seriously. I have spent very little time in the US, and this is not something you can study surveys for! But I still get the impression that some sort of BrE values of əʊ is the most likey feature for an American in Britain to pick up.

    I always think GOAT was not the best choice for identifying this in the first place, being as you say in checked position. Have you ever observed what a strikingly clipped realization that can give it in BrE?

    I hope John considers whether your 1/3 might be grist to my mill!

  31. A lot of Americans think of a fronted GOAT as sounding Southern and, as you probably know, a lot of Americans hate Southern accents and would try to do everything they can to avoid sounding Southern. So this isn't exactly a sound that they hold in high esteem. By values do you mean phonetic values or do you mean how they regard the sound?

  32. @vp: I say ['mɛrələnd].

    "Labov et al. give statistics for the relative numbers of speakers in the US with what they define as a fronted value for the GOAT vowel (in checked position)." How recent are these stats? I'd have expected to see a species of fronted GOAT diphthong showing up in SoCal, too.

    So far as I can tell, oʊ̝ is a pretty fair notation of what I use in GOAT.

    @mallamb: "It looks a bit as if əʊ increasingly will do for AmE." I guess I need to get out more. I don't hear it except in speakers from a limited number of regions. I'll probably notice it more now that it has been brought to my attention.

    But as for British influence on American GOAT, that dog won't hunt. You only get that sort of thing in the neither-fish-nor-fowl "junior year abroad" accent as demonstrated by Madonna. Normal people grow out of it, or get it laughed out of them, once they get back home to the States. If əʊ is becoming the norm here, it has surely evolved entirely independently of British Influence, Austin Powers imitations notwithstanding.

  33. @mallamb:

    I'm not sure I agree with your advocacy of /əʊ/ as a notation for both "BrE" and "AmE".

    Assuming "BrE" means "British English", is a fronted starting point for GOAT found anywhere except in the south and midlands of England (and in RP)? That represents less than half of the combined population of the British Isles.

    As we have already seen, the majority of Americans do not front GOAT, so it doesn't seem appropriate for "AmE" either.

  34. GOAT-fronting accents include London (with some exceptions), the Home Counties, RP and some Scouse (Liverpool), but I'm not sure about East Anglia, the West Country, the West Midlands or the East Midlands. You also might get GOAT fronting in the Republic of Ireland these days (especially in the Dublin area), if you're including that. I'm not sure how they pronounce it in Northern Ireland. It's probably like how they pronounce it in Scotland in many cases, which is definitely not fronted or even diphthongal for that matter.

  35. Some people in northern England have a fronted monophthong for GOAT. This is particularly associated with Hull, where it is so fronted that it can sound like RP NURSE; various websites (including a BBC "Guide to Hull dialect") say that in Hull "err nerr" is "an expression of dismay". But a trend in the same direction can be found in other parts of Yorkshire, e.g. my GOAT is not a fully back vowel.

    All that said, I think that if I were looking for a symbol to cover GOAT across all forms of English, I'd go for a nice simple /o/...

  36. You might also get that fronted monophthong in broad Geordie, where it is apparently only used by males. Thanks for reminding me about that.

    Here's an example of someone using that variant:

  37. No no no no. By 'will do for' I mean 'is indeterminate enough for'. In the case of both BrE and AmE. Because ə is conveniently ill-defined. We start from the position that əʊ won't really cover all the 'phonetic values' (obviously - if people think I'm making value judgments they aren't trying very hard) even for the British Isles, so why on Earth would we expect it to do so in N America?

  38. Should I say I'm sorry for bringing up the GOAT conundrum? ;)

    Anyway, I think the whole discussion nicely highlights the need for JW's post in the first place. I think his point was that, for many of the points of difference between BrE and AmE, it makes better sense to pass over the minutiae and minimise the set of symbols within a determinedly phonemic application such as a dictionary. The symbol set is supposed to be as phonemic as it gets, so to speak. That was the gist of pointing out that if we agree to have the same symbol for both THOUGHTs, then why not for both GOATs.

    But, on the other hand, most of the comments -- a majority of them from native speakers I believe -- show that due to the origin and nature of the IPA, there will always be some point of contention as to whether a given symbol is appropriate. The IPA may not be "magic", but there still is a default feeling that, well, [ə] stands for something unrounded... (The case of [ʌ] for STRUT notwithstanding -- people do forget after some time.)

  39. By the way:

    Eric Armstrong: readers don't bother to READ the darn things [i.e. the introduction]. So the fact that /ɔ/ can mean different things in different versions of English gets missed. So we assume that they mean the same thing when they do not, because the symbols look the same.

    My thoughts exactly. American-style respelling systems à la Webster are much more dialect-independent (even though we do have the comment from Julie above about ô). But they are very difficult to read for non-natives; a standard exercise I use with my students is "transliterating" Webster-style respellings into IPA -- a constant socurce of bewilderment... IPA-style systems have one great advantage for non-natives: in a majority of languages using the Latin alphabet, ā, ē, ŭ etc. (or similar renderings) don't trigger a reading of FACE, FLEECE, STRUT. /iː/, /eɪ/ etc. do.

  40. (The last comment, I promise. It's been a long thread.)

    I've always associated the centralised (NOT fronted in the strict sense) GOAT in the US with Philadeplphia. And in general it's a Midland thing, as evident from the second map linked to by mallamb above. Notice that "F2 of GOAT > 1350 Hz" really puts the nucleus in schwa territory. Well, almost. But then, a friend who used to work with Labov in Philly and then moved to England once said to me, "that centralisation they talk about is nothing in comparison with what goes on here".

    And finally, JW: just think of the people who would write to me complaining if I were to use the symbol oʊ for something that is increasingly neither rounded nor back. This is a double-edged argument, I think... Do people complain about having e.g. /eə/ for something that is increasingly not a diphthong? /æ/ for something that is increasingly open? [Tongue in cheek, of course.]

  41. On a phonemic level, there are other issues, especially if one or both parts of the diphthong occur as phonemes on their own. (That would be the case, for instance, for schwa, but not for /o/ in many accents.)

    By the way, I find it interesting that the social prestige in RP seems to depend only on the closedness, not on how far to the front or the back the diphthong is or even how weak the second part is.

  42. @Lipman: the social prestige in RP seems to depend only on the closedness, not on how far to the front...

    Well, Cruttenden firmly refers to realisations with a frontish starting point, along the lines of [ɛ̈ʊ], as Refined RP, and on p. 322 (2008 edition) speaks thus: "learners are advised to avoid the diphthong having a fronted onset [ɛ̈], this variant being regarded as excessively affected". Also Roach (2003: 26) on [eʊ]: "unfortunately, this gives impression of a 'posh' accent -- it sounds like someone trying to copy an upper-class pronunciation". And of course that is the general feeling among the population, isn't it? See e.g. this cartoon.

    So there is a very clear sociophonetic (and attitudinal) pattern here.

  43. Yes, I simplified, of course, and each of the (major group of) variants, including differences between the first and the second part, has its connotations, but inside RP.

    (Concerning the link, the pronunciations used are as authentic as the cartoon is funny.)

  44. @wjarek:

    Does Cruttenden actually use the transcription [ɛ̈ʊ]? I don't believe that's his Refined RP. And it would need to be closer to Roach's [eʊ] even to be "Refained RP".

    And I don't think for a moment you should be 'sorry for bringing up the GOAT conundrum', though it has been cracked up on this thread to be more of a conundrum than it is.

    > there still is a default feeling that, well, [ə] stands for something unrounded...

    Your suspension points do I hope imply that you would go on to stress that that is because it's in square brackets.

    Coz here we're talking about /əʊ/, and depending on the realizations we have to describe, we can predict that that schwa has a more or less rounded allophone in the context of the ʊ, or that the ʊ has a more or less fronted allophone in the context of the ə etc. In fact there is no reason not to say that the allophony of each is determined by the other in that combination.

    The point of a phonemic analysis is to distil the commutation classes and distribution out of all this mess. It's not impossible that even monophthongal [o:] could be consistent with an /əʊ/ analysis provided it was an allophonic combination of the two alongside the sequences of allophones.

    What is impossible if we are serious about doing any of that is to adopt the above suggestion that we should write /o/ and have done with it. Unless we want to do level-mixing with diachronics thrown in.


    I seem to have said something similar, but of course in spades as usual.

    Re the cartoon, wjarek did give it as an example of "the general feeling among the population", so I suppose that covers him for the pronunciations being complete rubbish. But I didn't think it was all that unfunny.

  45. Absolutely; I didn't mean to attack wjarek.

  46. @mallamb: Does Cruttenden actually use the transcription [ɛ̈ʊ]? I don't believe that's his Refined RP.

    Oh yes he does. It's not on Google Books, but you can check here on Amazon. Just seach for "excessively affected". It's the 1994 edition, so the wording is slightly different, and the page number is different. But the transcription has survived at least 16 years!

    Coz here we're talking about /əʊ/, and depending on the realizations we have to describe...

    Of course, that was the point of John's post. Agreed. My point was this: judging from some of the discussion here and elsewhere, at least some people some of the time take /əʊ/ to mean [əʊ] (etc., depending on what they find unsatisfactory about a particular symbol choice) in a very narrow sense. And I feel the obvious reason is that the same set of basic symbols can be used for both broad and narrow transcription, and they forget about this:

    The point of a phonemic analysis is to distil the commutation classes and distribution out of all this mess.

    Again, agreed. However, for printing, you need to decide which symbol to use (a typesetting decision in essence). Which may lead to the discussion above.

    Unless we want to do level-mixing with diachronics thrown in.

    I'm afraid you can't (in fact, shouldn't) avoid either in a pronunciation dictionary targeted at EFL learners. Doing phonology for the initiated is another matter altogether.

    cartoon [...] pronunciations being complete rubbish

    Eye dialect, sorry ;) Even worse, unsystematic, inconsistent eye dialect. But fun. BTW, I noticed only after posting that the Language Log post actually pointed to a post on <a href=">John's old blog</a>, which covered much of the same terrain as some of our comments here...

  47. O jarku, we can have commerce! Will you be my friend? Then will I be able to say "Jarek!" (I thought even before you explained it that the voc was in a parlous state, so perhaps my Polish was a plus quantity before it was nil.)

  48. BTW I am an ostensive definition of a not "excessively affected" counterexample of Cruttenden's unbelievable transcription for his Refined RP. And I am not Wells's U-RP either, though I do have most of the U indicators.

    Foulmouthed is how I would describe myself.

  49. I promise this will be my last post on the subject. Commenters be interested in Wikipedia's IPA for English page: as you can see a kind of archiphoneme approach is used, encompassing both RP and GenAm, and /oʊ/ is used for GOAT.

  50. @mallamb: we can have commerce! Will you be my friend?

    Of course! No, wait, I'm not on ebay or Facebook... There may be problems ;)

    the voc was in a parlous state

    Well, actually, on second thoughts, I think I overstated the case. It's slowly decaying but probably not yet on its last legs. "Increasingly stylistically marked" would be a publishable description.

    a not "excessively affected" counterexample of Cruttenden's unbelievable transcription for his Refined RP

    Well, he says "this variant [is] regarded as excessively affected". He doesn't call it affected.

    I'll spare you a certain turn of phrase from JW's Accents of English then.

  51. Oh no, don't spare me! I don't remember that, and I'm past even finding the book since I built a new study, never mind flicking through it.

    If I don't like it, I can always change my (only recently adopted ad-hoc) attempt to transcribe that allophone I think I have!

    Coz I'm not affected, honest I'm not! JW may have exchanged words with me recently enough to remember that (and he would distinguish between that and my being "touched", which I undoubtedly am). If he holds his peace, it doesn't mean he wouldn't back me up: I don't think he follows threads of this antiquity!

  52. Well, you asked for it. Look at the paragraph that straddles pages 105 and 106, especially the last sentence. (I couldn't link to the exact quote because if I did, it wouldn't show the bottom of p. 105, so I had to use a little trick to show p. 105).

    BTW, JW used a whole range of symbolisations, as seen e.g. here, along with another occurence of dispute-inducing phrasing: "now widely considered affected".

    And I think you can easily be foulmouthed in the U-est RP of all England. It's two different levels, isn't it? ;)

  53. Well that is quite a little trick you used there, wjarek. Thank you.

    But what Wells is talking about in the first passage you have given me is [eʊ], and I have never said, or said I said that, much less attempted to argue for it. So that certainly doesn’t apply to me. In fact I have no less distaste for that particular affectation than he appears to have! This may be deduced from the spelling of "refined" in my original remark about this: 'it would need to be closer to Roach's [eʊ] even to be "Refained RP"'!

    What I think is being guyed in the cartoon is [ëᵻ] and that was what I was guying myself in its monophthogized incarnation when I wrote re JW's blog entry " triphthongs, anyone?" (2009/12/10) “It’s sëː wɪːd that vɪəkl goes to vɪːkl but ˈvɪɪkl doesn’t, isn’t it?”

    The second passage you have given me is about variants of the [ɜʊ] type, and that is the representation I favour. (I think it was the one adopted by the first Collins to use IPA, and I totally approved of that, but I can't stagger round long enough to search out my copy of that either.) But the diacritics Wells has got under the ë and ɛ̈ there have defeated everything I have tried to make them legible, so some of it is lost on me. Can you see what they are? Anyway he goes on to say "this [ɛʊ] type enjoys some popularity among U-RP and even mainstream RP speakers born between the First and Second World Wars; but it is now widely considered 'affected' and has ceased to be fashionable among younger speakers." I can only think that [ɛʊ] was intended to have some sort of diacritic too.

    I would just about fit into the category I think he intends there, but I would certainly put myself closer to the [ɜʊ], with much slighter fronting of the first element.

    Just as well I thought I had nothing to fear from that "certain turn of phrase", and did ask for it! Can't have you thinking I'm affected, foul mouth and all!

  54. I think this shows how difficult (well, at least cumbersome) it is to represent such subtle variation in vowels using the IPA, even with diacritics. I think these are dots under the symbols, "old style" for raising. If I don't forget, I'll check in a printed copy in my office tomorrow.

    And, Heaven forbid, I would never think you're affected, mallamb ;) After all, already the earliest research into attitudes to sociophonetic variation showed that it's impossible to predict personality from pronunciation features. But it also showed that people are quite happy to make such judgements, and that there are stereotypes. Cruttenden etc. of course talk about those stereoptypes.

    BTW, [eʊ] lives on quite happily among many pronunciation teachers over here. As a result, some of our poor students, when in England, get mocked for talking like WW2 Royal Navy admirals.

  55. LOL I have observed that in products of pronunciation teachers of far more diverse provenance than Poland. Where BrE has not yet been supplanted by AmE, at least it shows they are trying! But the mockers' stereotypes are a bit out of whack as to RN rank if the poor _good_ students are reproducing it faithfully.

    I thought if anything the diacritics under the symbols looked like "old style" ogonek-type hooks for raising. Shall be interested to learn what you can see from your printed copy. Still can't locate mine.

  56. I meant ogonek-type hooks for lowering, of course.