Tuesday 27 September 2011


At the running club these days I can, alas, do no more than jog a mile or two if neither my arrhythmic heart nor my arthritic hip are playing up. But it’s still a great place to socialize with friends old and new.

One of our coaches is something of a linguistic paradox: a highly educated man, with a degree from one of our oldest universities, established in his chosen profession. But he retains a strong working-class south London accent. As he was giving out the announcements recently, he told us in connection with some forthcoming event that further ˈdɪiʔɛ̈oz were available on the club website.

This pronunciation of details exemplifies, inter alia, intervocalic t-glottalling and l-vocalization, stigmatized features that one would not usually hear from someone with his educational background. Naturally he also has the usual British word-initial stress for this lexical item. (Compare AmE: in Yuko Shitara’s poll, 75% of Americans voted for final stress in detail.)

This set me musing about similarities and differences between the two major sound changes now vying for supremacy in the world of English non-initial t, namely voicing and glottalling. I am thinking above all of cases such as butter and a lot of, where AmE normally has a voiced tap ɾ and BrE may have any of t, ɾ, ʔ.

It is clear that an American-style voiced tap is by no means uncommon in Britain, particularly in high-frequency items such as a lot of. Conversely, a British-style glottal stop seems to be not unknown in north America.

But the two rival developments affecting t do not operate in identical environments. Yes, their environments overlap, as in the cases quoted. But t-voicing is blocked by a non-vowel right-hand environment (as in the two plosives in that’s right!), an environment in which glottalling is clearly more frequent than prevocalically. ˈðæʔs ˈraɪʔ

Furthermore, t-voicing is blocked (somewhat mysteriously, from my point of view) if the following vowel is unreduced and there is no word boundary. You get it in later ˈleɪɾə(r) and my late uncle maɪ ˈleɪɾ ˈʌŋkl̩, but not in latex ˈleɪteks.

For Americans, then, there is no chance of voicing the t in detail. For those who stress the second syllable, the t is like a word-initial t, always voiceless. For the 25% of Americans who prefer initial stress the same constraint as in latex comes into play, and again there is no chance of voicing the t. But for Brits who revel in glottalling — like our running coach — this is just one more candidate item for a glorious glottal stop.


  1. I find North American t-glottalling particularly striking between two nasals, as in "sentence" or, say, "Clinton".

    When I hear this from speakers who have clearly had speech training, such as newsreaders, it's hard for my BrE-trained ears to avoid a feeling of stylistic incogruity :)

  2. There seems to be some variation in glottalling between England and Scotland. During the last election, Gordon Brown frequently said "constituency" with the second T glottallised. I don't think that many English people would glottalise that T.

    I think that T-glottalisation is the modern equvialent to H-dropping. H-dropping seems to be on the way out now, whereas T-glottalling can be heard almost everywhere in Britain.

  3. @ Ed

    H-dropping on its way out? I really can't believe that, I'm afraid.
    Maybe I should poshen up my social circle...

  4. Northern Ireland also has intervocalic T-voicing, in almost exactly the same environments as American English, but with one or two exceptions:

    Before vocalic L, as in bottle NI preserves the voicelessness and uses a lateral release: 'bɑtˡl̩. Compare AmE 'bɑːɾl̩~'bɑːɾəl.

    Before vocalic R, as in better, some NI speakers use a voiceless dental plosive: 'bɛt̪ɹ̩ (cf AmE 'bɛɾɹ̩). But the dental allophones of the alveolar plosives are stigmatised in Northern Ireland and many speakers use a voiced/tapped T in this position: 'bɛɾɹ̩ (as in AmE).

  5. I think t-voicing in American English may have come from Northern Ireland. I also think other things, like the "cot-caught merger" may have come from there.

  6. Jack Windsor Lewis says:
    John, when you speak of "two major sound changes now vying for supremacy in the world of English non-initial t, namely voicing and glottalling.... above all .. cases such as butter and a lot of, where ... BrE may have any of t, ɾ, ʔ" I feel inclined to mention another possible contender, at least as a far as the inter-word situation is concerned, namely the postalveolar approximant ɹ by which I was surrounded in my Cardiff childhood and I hear from time to time from various speakers among the least likely being the Scot Lord Lamont. In Cardiff we cd have it, as well in butter also, as has been pointed out by my fellow Cardiff originary Bev Collins, even by many in starting and hotter.

  7. There's also the non-sibilant alveolar fricative possibility, particularly associated with Scouse and Irish English, but I have it quite frequently and I'm from neither Liverpool nor Ireland.

  8. So those words would be homophonous with borough, starring and horror?

  9. Nobody here is distinguishing between detail (n) and detail (v). For me (an American), detail is one of those words where this makes a difference, the stress falling on the first syllable of the noun and the second of the verb.

  10. I have never 100% understood the occurrence of t-voicing in American English.

    I do tell my teenage pupils that inside one word, between a stressed syllable leftwards
    and an unstressed syllable rightwards, t is regularly voiced in North America.

    However, t-voicing also seems to happen across word-boundaries in contexts that
    I fail to manage to summarize in a rule, as in "but I", or "it all".

    But the trickiest of all is with the initial "t" of "to", as can be heard in
    Chicago-based "Plain White T's" 's song "Delilah" :


    from 0mn50s to 1mn11s
    "Oh, it what you do to me"

    and at 2mn33s
    "and you're to blame".

    Of course, this is consistent with the long-attested existence of the contracted form
    "wanna" for "want to".
    Still, cross-word-boundary t-voicing remains a bit tricky to predict...

    Jérôme Poirrier
    Grenoble, France

    1. As an American from the Northeast, it would not be uncommon in casual speech to ask:

      "What do you want to do tomorrow?"
      wa:d'y'wannadu:d'm)ro (where ' represents schwa)

      The word initial "t" in tomorrow in this environment would sound "too proper"; not incorrect, but not congruent to the audience. The production of voiceless t would also cause a "break in the flow" of the sentence, being that all of the words before it were linked and only the word tomorrow would be separated, unnecessarily, from the rest of the sentence.

  11. Detail (n.) undergoes rhythmic stress shift for me: "That's a deTAIL" vs. "Show me the DEtails" vs. "DEtails, DEtails!" I haven't had a chance to look at Shitara's work yet, but I will.

    I very much hope that the day will return in the anglophone lands when, as in Elizabeth I's time, an accent represents only a point of origin and nothing more, and people are accepted (or not) for the content of their character and not the deTAILs of their phonology. There are still of course stigmatized accents in North America, like the New York City accent and the Inland Southern accent, but I think the old prejudices are beginning to fade. (It's hard to sort out anti-AAVE-accent prejudice, anti-AAVE prejudice, and anti-AA prejudice, so it's hard to speak of that: certainly there is much less stigma against a mere AAVE accent than against speaking AAVE itself.)

  12. @Jérôme: I think the word to (in its weak form) is an exception: in American English it's pronounced but no other initial Ts are treated this way (I think).

    Northern Ireland has the same exception: the weak form of to is pronounced (Scots tae + initial T voicing) and the strong form is tʉː (or basilectally diː).

  13. I suggest the underlying factor for recent British English pronuciation is the modified attitude to RP (as Daniel Jones defined it), with consequent growing acceptance of hitherto low prestige home counties variants. The second part of DJ's RP definition concerned acquired RP. It's easy to forget today just how strong the pressure used to be to adapt to RP in many situations in order to gain or keep employment and promotion.

    We see today that this pressure has relaxed over the past half century, and very likely at least a generation earlier for professions like science and engineering where skill and creativity are usually more important than vowel timbre. In my old home town on the Thames estuary, they still tell a story about the distinguished nuclear physicist Dr William Penney (possibly apocryphal, I've never been able to find a printed source). The prime minister was to be briefed about the nuclear program at the lab but the director Professor Penney was advised to leave the talking to more comprehensible colleagues.

    Swap the name and location and we have the story of John's coach. And the stories of all the others who are no longer pressured to acquire RP to hold down a career. Glottal stops and vocalised laterals have occured locally for over 100 years, today there's just not the need to hide them so much.

  14. As VP observes, the realization of /t/ as [ʔ] is very common in American accents following a stressed syllable ending in /n/, such as sentence or Clinton. But in the northeastern US, or at least in eastern Massachusetts where I now live, it occurs very often in contexts in which, to someone who grew up in another part of the US, such as I, it sounds decidedly foreign: e.g., curtain, button, and even across word boundaries, though I can't confidently recall any examples. I have not yet figured out what the pattern is.

  15. My own pronunciations of "detail" are the same as Peter Shor's. Parts of speech make a difference.

  16. MKR: As someone who first picked up English in Northeast New Jersey, I can confirm that a glottal stop in curtain and button is natural for me. Ditto in high-frequency words like certain and written.

    If /t/ follows a vowel or a sonorant, I think a following syllabic /n/ is enough to trigger a glottal stop quite regularly in this style of speech. A syllabic /m/ or /l/, however, doesn't usually trigger a glottal stop (I have a voiced tap for bottom and bottle). On the other hand, a non-syllabic /m/ or /l/ practically requires a glottal stop, as in atlas or apartment.

  17. Pete, I don't think I would go as far as saying that the weak form of to is in American English, which doesn't explain what happens in examples like have to or need to, where the /t/ is never voiced.

    I think Jérôme's examples are better analysed as t-voicing applied to the initial /t/ of the weak form . Notice that to follows a vowel or a sonorant in all of his examples (do to me, you're to blame, want to), consistent with the usual conditions for t-voicing. It's almost as if to is a postclitic; notice also that if it follows /t/, one of the /t/s is often deleted, as in got to, used to, and supposed to.

  18. @ Paul Carley: It seems that way to me, but it might be pot luck of young people I come across.

    More generally: I have always liked this Janis Ian (American) song from 1967, and have noticed that some of the glottalisation is typically British. Listen to "not our" at 0:55 and "let us" at 1:46. [?] for /t/ does exist in America as well.

  19. Jongseong: "...which doesn't explain what happens in examples like have to or need to, where the /t/ is never voiced."

    I agree with you about have to. But for me I think the /t/ in need to can be voiced, especially in very informal speaking styles.

  20. @ yuriive: I've considered this again, and I need to revise my position. I agree that the /t/ in need to can indeed be voiced in a more relaxed speaking style. The same goes for had to. Maybe the final /d/ interacts with the /t/ and causes the latter to elide, just like one of the /t/s in got to, used to, etc. are elided.

    My argument that the weak form of to is also in American English still stands.

  21. John W,
    « t-voicing is blocked (somewhat mysteriously, from my point of view) if the following vowel is unreduced and there is no word boundary.»

    I think it might be less mysterious a phenomenon, if like Saussure you would accept that the t of your example 'latex' may function both as the coda of the first syllable and the onset of the second. Your LPD3 entry's syllabification for 'latex' explains the clipping the vowel of the first (?unnecessarily, especially for AmE) but not the blocking of the t-voicing of the second which I think you are right to predict even for your coach:
    latex ˈleɪt eks latex|es ɪz əz

    But your LPD3 syllabification ˈdiː teɪ ə l for 'detail' explains the American non-t-voicing but not the coach's glottal stop (no doubt with clipping):
    detail noun, verb ˈdiː teɪ ə l ; di ˈteɪ ə l, də- detail|ed d detail|ing ɪŋ detail|s z — Preference poll, American English: ˈ•• 75%, •ˈ• 25%.

    For the coach you would need to recognize the syllabification ˈdiːt.eɪᵊl.

    I do realize this happens across apparent word boundaries too: ˈsəɪʔə for 'see to'.

  22. It occurred to me that to describe all this, one should resort to the notion of “unaspiration” first of all. I’ve been testing the following set of rules:
    « UNASPIRATED T has got two variants, i.e. ʔ and ɾ
    - Certain British speakers (from working class) will change unaspirated t into ʔ as long as the t follows a vowel.
    - Some other British speakers will change unaspirated t into ʔ rather if the t follows a vowel AND precedes a consonant, or if the t follows a vowel AND is word-final.
    - Americans (and the Beatles) will change unaspirated t into ɾ (voiced t) in a… …"VERY VOICED" environment, i.e. if the t is in immediate contact with vowels or sonorants on the right and on the left simultaneously.
    - However, Americans may change ɾ into ʔ if the t is in contact with sonorants both on the right and the left ».
    Cross-wordboundary t-tapping, as in « But at, at all, it all, my late uncle » is justified by the absence of aspiration in the t. In “latex”, t-tapping seems blocked by the “e” being a strong vowel and thus causing the t to be aspirated.

    Jérôme Poirrier
    Grenoble, France

    PS : however, for a Frenchman like me, aspiration is something tricky. For example, I have no idea on how to realize the t of eighteen…

  23. Erratum :
    I wrote "I have no idea on how to realize the t of eighteen"; well, actually, I do. But I meant that in case of stress shift ("18 months"), I am not able to HEAR whether the t is aspirated, although my belief is that since the "ee" is supposed to be a strong vowel, then the t is supposed to be aspirated.

  24. @Mister_P:

    I think you may find considerable variation in the degree of aspiration of the "t"'of "eighteen". Personally I always stress that word on the first syllable, and I don't aspirate /p t k/ much unless initial or stressed: my VOT would not be zero but would be far less that that in, say, "team". However other native speakers will probably aspirate far more.

  25. Is T in words like Clinton, button, mountain:
    1. an unreleased T
    2. a glottal stop

    in General American?

  26. This comment has been removed by the author.

  27. @Pianoman:

    Is T in words like Clinton, button, mountain:
    1. an unreleased T
    2. a glottal stop

    in General American?

    I would say a released glottal stop.

    I suppose it could also be an unreleased T immediately followed by a released glottal stop:

    [t̚ʔn̥], which I would have trouble distinguishing auditorily from a plain [ʔ]

    Disclaimer: I don't speak General American, although I do hear a lot of it.


Note: only a member of this blog may post a comment.