Wednesday, 23 November 2011


One or two of the people commenting on nt-reduction (blog, 18 Nov.) also mentioned the possibility of twenty having the vowel ʌ.

Kensuke Nanjo said
According to my daily observation of American English, I think this variant is worth including in pronouncing dictionaries. Quite a few Americans use it and as you may know, this is the second variant for "twenty" in the Merriam-Webster Collegiate Dictionary.
There are indeed plenty (“plunny”?) of Americans who seem to pronounce twenty with a seriously backed and lowered quality as compared with their default DRESS vowel.

However, in deciding whether this is a sporadic irregularity found just in this word (and perhaps in plenty too), we must first establish what is their default DRESS vowel. We need to discount the possible effects of what, following Labov, has come to be known as the Northern Cities Vowel Shift.
The “northern cities” (of America) in which this sound change flourishes are clustered around the Great Lakes: places such as Buffalo, Cleveland, Detroit, and Chicago. The geographical extent of the shift varies depending on which vowel is involved and in which phonetic environment(s); and in any case it is also socially and stylistically variable. But what it can do is to make DRESS words sound as if they have the STRUT vowel — perhaps all of them, perhaps particularly those in which the vowel is followed by a nasal. Note that the STRUT vowel shifts too, so that we do not normally get loss of the distinctions exemplified in get – gut, bed – bud, wren – run etc.

So someone who says ˈtwɛ̈ni, with a thoroughly retracted vowel, is not necessarily saying ˈtwʌni (“twunny”), to rhyme with funny.

Others, though, are. They include rirelan, who mentioned
twenty: /ˈtwʌni/ (along with "plenty" /ˈplʌni/. "plentiful" is still /ˈplɛntəfəl/ though.)

Furthermore, Americans from other, mainly southern or western, parts of the country may merge pen and pin as pɪn (i.e. merge DRESS with KIT before a nasal). For them, twenty may rhyme, if not with funny, then with skinny as well as with many.

Kensuke reckons that a reasonably exhaustive pronunciation dictionary ought to give AmE twenty as ˈtwenti, ˈtwʌnti, ˈtweni, ˈtwʌni. Seems reasonable, though perhaps we ought to add ˈtwɪnti, ˈtwɪni, too.


  1. I don't think /ˈtwʌnti/ is due to the NCVS, I'd say it's just an informal or reduced variant of twenty, similarly to /ðət/ of that. They are phonemically different from /ˈtwenti/ and /ðæt/, but accents with the NCVS have the same underlying phonemic structure as those without: only the phonetic (surface) realizations of the phonemes are different. So cad is /ˈkæd/ everywhere, even if a speaker with NCVS may pronounce it something like [ˈkeəd]. The same applies to the pin-pen merger: it's phonetic only, too.
    Dictionaries giving phonemic transcriptions should not show phonetic-only forms like [ˈtwɪnti] for twenty or [pɪn] for pen, or a raised [ˈpeən] of GenAm for pan, or any shifted or nt-reduced pronunciations, as they can be deduced from their phonemic transcriptions, possibly with the help of a reliable book on dialects.
    Giving too many variants can be more confusing than helpful, especially for an EFL student.

  2. perhaps particularly those in which the vowel is followed by a nasal
    ...or /l/, I think.

    I also suspect that the underlying fortis consonant might be playing a part, in which case there would be people distinguishing penny from twenty as [pEni] and [tw3ni]; has anyone else noticed this, or am I hallucinating?

    Seems reasonable, though perhaps we ought to add ˈtwɪnti, ˈtwɪni, too.
    If that only happens with people merging DRESS with KIT in *all* words where it precedes a nasal, that seems pointless to me, as giving a LOT alternative to all THOUGHT words.

  3. I agree with you: a dictionary is not supposed to show cot-caught merged pronunciations, since it is not a phonemic merger, either.

    As a side note, there's actually no such thing as 'phonemic merger' in a synchronic sense, unless it means the 'identical phonetic realization of two or more underlying phonemes'. Diachronically speaking we could say that two phonemes have merged into one in all dialects of a language over time, but we obviously compare today's dialects (and dictionaries give pronunciations) synchronically, not diachronically. Thought should be given simply as /θɔːt/, which incorporates all the possible realizations of [θɔːt, θoːt, θɔɑt, θɑt, θɑːt, θɒːt] and many more. Why waste paper and ink, bytes and bits?

    And no, you're not hallucinating. You can hear twenty even as [ˈtwəni] (yes: stressed schwa), which is of course different from [ˈtwʌni]. I could imagine that this sort of retraction is because of the preceding [w], which has a velar component. Speakers that have a velarized [ɫ] even before vowels might end up pronouncing common words like plenty as [ˈpɫəni] or [ˈpɫʌni], too, but less frequent words like plentiful could retain their normal DRESS sounds. That's my theory. Can anyone confirm this? Does anybody who pronounces a clear [l] before vowels rhyme plenty with 'twunny'?

  4. I agree that /ˈtwɪn(t)i/ is pointless; the DRESS-KIT neutralization is sub-phonemic and automatic.

  5. teardrop: Blogspot thought it was spam, and put it in the spam folder. As you have seen, I have now rescued it and restored it (above).
    Meanwhile, I have also deleted six spam comments (on previous postings) that Blogspot had not spotted.

  6. I grew up in Seattle (b. 1961), far from the pertinent group of "northern cities," and I can assure you that for me and my peers and family members, "twenty" always rhymed with "runny," and "plenty" certainly did not. I share Teardrop's hunch that one cause of this pronunciation is the preceding [w], but I would add another, namely the fact that the word "twenty" often occurs in a relatively unstressed position, namely in compounds like "twenty-one," in which position it is all the more likely to be reduced to a schwa, which is then carried over into the stressed position. Compare the American pronunciation of "because," even in stressed position, as "becuz" (notably enshrined in the famous rhyme from The Wizard of Oz: "Because, because, because, because, because--because of the wonderful things he does!").

  7. Teardrop:

    As a side note, there's actually no such thing as 'phonemic merger' in a synchronic sense

    Well, "split" and "merger" are only meaningful in describing an accent relative to another accent. This may be a contemporaneous one: some accents have the thin-fin merger, eliminating the /θ/ phoneme, and others do not. Alternatively, it may be a historical one: all current accents have the vain-vein merger, though the spelling tells us that earlier accents did not. Most often it is both.

    unless it means the 'identical phonetic realization of two or more underlying phonemes'

    Better to call that a neutralization: thus some accents have a pin-pen neutralization, but none (I think) have a full merger of DRESS and KIT vowels. Likewise, some have a /t~d/ neutralization intervocalically, but none have a full merger of /t/ and /d/.

    Thought should be given simply as /θɔːt/, which incorporates all the possible realizations of [θɔːt, θoːt, θɔɑt, θɑt, θɑːt, θɒːt] and many more.

    Well, that works for thought, but nobody notates bath or cloth with a separate symbol; the convention is to give two pronunciations, representing the split and unsplit accents (or in the case of cloth, the remerged accents as well). Similarly, there's no getting away from writing two pronunciations of lieutenant; it's too idiosyncratic to be handled with a special symbol that can be read as /(j)u/ or /ef/.

  8. John Wells: thanks, I wonder which part of my comment made Blogspot consider it was spam...

    John Cowan:
    I don't think the thin-fin (TFM) and the pin-pen (PPM) mergers are that different. If a TFM speaker hears [fɪn] out of context, they can't tell if it's thin or fin. If a PPM speaker hears [pʰɪn] out of context, they can't tell if it's pin or pen, either. So TFM and PPM are alike on the hearer's side. As a speaker, the only difference between the two mergers is that a TFM pronounces all occurrencies of a phoneme differently, while a PPM only some of them. But I don't think that's an essential difference, because neither of them is done on purpose. That's why I consider both TFM and PPM neutralizations.

    The reason I don't like the idea of eliminating a phoneme from a dialect or defining own phoneme sets for each dialect is that the definition of phoneme concerns a language, not a dialect, which makes perfect sense to me since a phoneme is releated to meaning, and meaning is shared among all dialects of a language.

    As long as the majority of speakers make a difference between thin and fin or pin and pen, they should be treated as containing distinct phonemes of the English language, i.e. in all contemporary dialects, even if in some of them the distinction gets neutralized conditionally (PPM) or unconditionally (TFM). In current English, /hw/ of white is losing ground to /w/, and in a couple of decades, we may find this neutralization widespread enough in all the dialects to eliminate /hw/ as a phoneme, and then we could relabel this neutralization of today to be a historical merger, just like the vein-vain merger.

    Splits like bath are systematic, and such splits can be represented by using so called diaphonemes: placeholders that map to phonemes depending on the dialect. For example, lost could be represented as /lɔst/ (here I choose to use italic to indicate a diaphoneme, but any other way would do) that translates to BrE /lɒst/ and AmE /lɔːst/, which in turn are realized as BrE [lɒst, [lɔst], AmE [lɔːst, lɔɑst, lɑst, lɑːst].

    Unsystematic differences, such as the variants of lieutenant, can't be represented with diaphonemes: they are phonemically different, just like the two variants of either.

  9. Diaphonemes are only useful (or, in fact, necessary) when specifically trying to describe multiple "dialects" at a time. But there's a whole body of research that understands mergers (and splits) as characteristic of individual idiolects, not dialetcs, let alone languages.

  10. Teardrop: if the currently spoken accents of a language are part of the language, why are accents no longer in use not part of it? The distinction you are making seems artificial to me (as my distinction seems artificial to you).

  11. wjarek:
    Since phonemes —by definition— are units of a language, not of dialects, a phonemic or diaphonemic transcription can essentially do nothing but transcribe an entire language, i.e. multiple (all) dialects at a time. Current practice, however, gives phonemic transcriptions of individual dialects (or even idiolects), but this clearly violates the definition of the phoneme. Contemporary mergers (i.e. neutralizations) do not occur at the phonemic level of the language, so no phonemic transcription is supposed to reflect them.

    John Cowan:
    Theoretically accents no longer in use could be included, but it would be quite impractical. Phonemic sets change by time. A dictionary is expected to give current pronunciations based on the current phonemic set, and accents no longer in use may not meet this criterion.

  12. @teardrop:

    According to whose definition of "phoneme"?

    Many people where I live have an extra distinction, not present in most English dialects, between two subsets of the FACE lexical set: "wait" and "weight" form a mimimal pair. Should this distinction, which many speakers of English never encounter, be recognised in phonemic transcriptions of English?

    You're also running into the question of what exactly makes a dialect into a separate language anyway.

    I actually largely agree with you about dictionary entries -- no-one shows all STRUT words with two vowels to include dialects without the FOOT/STRUT merger (and I don't think that's just because of dialect snobbery) and it would seem a waste of space to do so -- but I think that's a different question from phonemic transcriptions.

  13. @JHJ: Thanks, exactly what I wanted to say.

    In my definition of "phoneme", for example, the essential part is that phonemes are lexically distinctive. The language-specific part is an afterthought that is only useful (?) in comparative analyses; a rather unfortunate shorthand for "system-specific".

    Think of how you would determine the phonemic system in a "new" language. You would test the speakers on what lexical distinctions they make, and possibly on what phonetic distinctions they (fail to) perceive as lexical. So if a speaker has /fIn/ for both fin and thin, then there's no phonemic contrast in their system. End of story. What other speakers do doesn't matter. It just so happens that speakers in close contact will end up with very similar systems.

    Otherwise, you could take the diaphoneme story wherever you wanted. /nXYt/ for night and Nacht, /bZt/ for boat and boot, whatever.

    Delimiting dialects and languages is an arbitrary thing with the exception of clear Abstand situations.

  14. JHJ:
    I have seen dozens of definitions containing the word language, but I have yet to come across one with words like dialect or accent. Have a look at this collection:

    Dialect vs. language: I don't think there's a clear-cut boundary between them. Still, nobody thinks a different language is spoken in Chicago than in LA, but everybody agrees they're different dialects/accents of English. Even without knowing exactly what distinguishes a dialect from a language, we all think of language as a group of dialects — and such groups of dialects are mentioned in the definitions above.

    FOOT-STRUT is actually a historical split, unlike the wait-weight merger. There will always be dialects with archaic features when compared to the majority of the dialects. Whether or not to include them in the phonemic system could depend on the number of speakers. It's up to decision.

    Speakers don't have /fɪn/, they mostly have no clue of what a phoneme is, they just pronounce e.g. [fɪn]. It's only the linguists that categorize their speech into abstract segments to be able to describe how these entities are realized in the various dialects of a language, and to have a unified way to describe its phonology and phonotactics regardless of how the phonemes are actually realized. If phonemes were limited to a single dialect, you couldn't even speak about e.g. English phonology, only about thousands of independent dialectal phonologies that would be phonemically incomparable because a phoneme makes sense only within its own system. This would be a rather limited use of phonemes, but they are capable of a lot more: they provide an overview and the link between dialects.

  15. @teardrop:
    They may use the word "language" rather than "dialect", but that doesn't necessarily mean that they mean "language" in the group of dialects sense. Look at how linguists actually use the word "phoneme". Here's Larry Trask, in Language: The Basics. "Incidentally, perhaps you are wondering just how many phonemes there are in English all together. The answer: forty-odd. Why such a vague answer? Because not all English speakers use exactly the same set of speech sounds. For example, do you pronounce the words buck and book differently or identically?" and so on. A few lines later "However, very few English speakers have fewer than about forty phonemes, or more than about forty-five".

    As for your last paragraph, no it doesn't mean that we can't talk about English phonology. The phonologies of different dialects of English may be different, but they have a lot in common. We can describe features which are common to all dialects, or most dialects, or to a group of dialects, and we can describe ways in which they differ. We can make a detailed study of a particular dialect; if I read a study of RP there are a few things which don't apply to how I speak, but there's still an awful lot in common.

  16. @teardrop: That's a nice list of definitions. (BTW, thanks for pointing us to Onelook; seems like a useful resource.)

    To keep the discussion appropriately lighthearted, let me just counter this with these:

    (1) From our noble host

    (2) From Daniel J himself

    (3) In a well-reputed dictionary of phonetics, search for the phrase "phonemes of a particular speech variety", and go to p. 265.

    (4) And from another biggie, search for "the number of phonemes may differ" and go to p. 45.

    Also, yes, properly speaking you can't speak of "a/the phonology of English". Anything that is presented as such is only the phonology of one specific accent. If it tries to do more, it has to say so. And yes, speakers do have /fIn/, even if they don't have a clue of what a phoneme is. This is just a shorthand notation for the fact that, on the sound side, THIN = FIN, i.e. they're homophones, and that the speaker has the three things as distinctive elsewhere in their system. And no, phonemes do not provide links between dialects. Shared governments, armies and histories do. Sometimes, mutual comprehensibility. But don't get me started on the subtitled versions of Trainspotting in the US.

  17. Awww, JHJ, you didn't waste time looking for those links ;)

  18. For me, twenty and plenty do not rhyme. Twenty has ʌ, plenty does not.

    Twenty does rhyme with funny in one sense, but not another. Because of the t phoneme in twenty (which doesn't usually gets pronounced), the two words don't feel like they rhyme, for me, but, if I listen to how I saw them, yes, they rhyme.

  19. JHJ:
    If language can be used in different senses, so can phoneme, since the latter inherits this ambiguity in its definition from the former.

  20. @teardrop: but, like I said, look at how linguists actually use the word. As well as the example I gave, look at wjarek's links. Can you find any examples of linguists using "phoneme" in your way?

  21. @teardrop:
    If language can be used in different senses, so can phoneme, since the latter inherits this ambiguity in its definition from the former.

    According to whom?

    "Phoneme" may be used in various different senses. By far the most interesting one, to me anyway, is the psychological sense of "phoneme": where phonemes are the means by which words are encoded in an individual's mental dictionary