Monday, 18 January 2010


The terrible events in Haiti mean that both the name of the country and the name of its capital, Port-au-Prince, are repeatedly in the news.
Dictionaries agree in recommending an anglicized pronunciation of Haiti itself, namely ˈheɪti. This has not stopped occasional reporters offering variants such as ˈhaɪti and even hɑːˈiːti (and LPD, CPD etc duly report these). Many educated English people will be aware of the French pronunciation aiti, and possibly even of the French spelling Haïti. Although French is one of the two official languages of Haiti, the other one, and the language of most of the inhabitants, is Haitian Creole (a dialect of French Creole). The Creole spelling of the name of the country is Ayiti.
I think the Oxford BBC Guide to Pronunciation is on solid ground when it states firmly that in English the country is called hay-ti /ˈheɪti/ and labels this the “established anglicization”. The adjective Haitian is correspondingly given as ˈheɪʃən. (The French haïtien is aisjɛ̃.)

Similarly with the name of the city that now lies in ruins. OBGP, LPD and CPD all record the anglicization of Port-au-Prince as ˌpɔːt əʊ ˈprɪn(t)s (or the corresponding AmE ˌpɔːrt̬oʊ-). OBGP, however, while recognizing that as the established anglicization, “also common in Haiti”, comments further that
A French pronunciation, por-oh-pra(ng)ss, and a French Creole pronunciation, port-oh-pra(ng)ss, are also both in current use.
This is very true: plenty of TV reporters say ˌpɔːt əʊ ˈpræn(t)s.

They are probably not aware that in standard French the name is actually pronounced pɔʀopʀɛ̃ːs. There is normally no liaison in the case of the word Port: compare Port Arthur pɔʀaʀtyʀ, Port-Royal pɔʀʀwajal, and Port-sur-Saône pɔʀsyʀsoːn.

In Haitian Creole, however, the name of the city is Pòtoprens, presumably pɔtopxɛ̃s.

* * *

I’ve put another video up on YouTube, about the pronunciation preference polls in LPD.


  1. This comment has been removed by the author.

  2. The tilde is supposed to be above the vowel in pɔtopx̃ɛs, isn't it?

  3. Wikipedia is an unreliable source, but for whatever it may be worth, it reports:

    Port-au-Prince (pronounced /ˌpɔrtoʊˈprɪns/; French pronunciation: [pɔʁopʁɛ̃s]; Haitian Creole: Pòtoprens)

    I have no personal attachment to any of these transcriptions, nor to the pronunciations they are meant to represent. Until it was brought to my attention (in Linguism, Jan. 16) that in French, liaison from Port was incorrect, I would have pronounced that pesky t in what I mistakenly believed was good French.

    My last French lesson was much longer ago than I care to recall; but I've had nobody to speak French with since. All that remains - somewhat like the Cheshire cat's grin - is the accent.

  4. Well the t wouldn’t be liaised in Port-Royal pɔʀʀwajal, and Port-sur-Saône pɔʀsyʀsoːn in the first place, would it?

    Thanks for the bracketed [t]s in ˌpɔːt əʊ ˈprɪn(t)s and ˌpɔːt əʊ ˈpræn(t)s. I was trying to argue on an earlier thread that it's impossible to be prescriptive about t and d in such contexts, and if their phonetic presence or absence is in free variance or can be predicted from the context in terms of epenthesis, they have no phonemic status in that context.

  5. My first point was intended to mean that Port-Royal and Port-sur-Saône have presumably had no t since French has been French, whereas the Creole Pòtoprens shows that the t must in fact have been liaised at some point there. I don't know about Port Arthur.

  6. Kraut: the tilde is above the vowel (more or less) on my screen. This is a matter for your browser's rendering engine and the font: the coding is correct.

    mallamb: you're right of course about liaison, but I couldn't think how to make that point succinctly.

  7. From the endings -rs, -rd and -rt normally no liaison is made. Rather, the r elides.
    Tu pars en vacances. [pa ʀɑ̃]
    Il le mord à la jambe. [mɔ ʀa]
    Cela ne sert à rien. [sɛ ʀa]

    Two common exceptions to this rule are:
    1) -rd and -rt in inversions of verb endings: Sort-on? [sɔʀ tõ]
    2) -rs when the "s" indicates plurality: chers instants [ʃɛʀ zɛ̃stɑ̃]

    As for "port" itself, Fouché rights in his venerable "Traité de prononciation française" Klincksieck, 1959 :
    " Port-au-Prince, capitale de la république d'Haïti, se prononce soit [pɔʀtopʀɛ̃ːs] (prononciation locale), soit [pɔʀopʀɛ̃ːs]. Mais on ne lie pas dans les autres composés de Port : Port-à-Binson, Port-à-l'Anglais, Port-au-foin, Port-aux-Perches, Port-en-Bessin. "

    (By the way, I too am experiencing problems viewing the nasal diacritic above the vowel in question. This is a new problem for me on this page.)

  8. If you care to look at Graham Pointon's blog, you'll see that I encountered the same problem with the tilde not being above the vowel (not even more or less). It was not until I entered the decimal code for the tilde first that the result proved satisfactory in Firefox.
    When I look at JW's tilde position and compare it mine in the transcription of the Haitian Creole pronunciation, I must say (not without pride ;) ) that my version looks better (humour alert).

  9. I work on a Mac, so the use of "decimal codes" is quite foreign to me. Both in Safari and Firefox the tildes are displayed incorrectly. Again, this did not used to happen. I wonder what the _real_ problem is. It seems to be a new one…

  10. That video was quite condescending. Americans have been pronouncing "which" and "witch" the same for years. It's not like that's an innovation that just took place. I don't who you interview to get your data.

  11. Nedecky,
    I take it you mean the t elides, but your first three examples are all non-inverted verbs. With the inverted ones French can even make liaisons that aren't even there to make, except etymologically: a-t-il, va-t-on, even porte-t-il, so I suppose the wonder is that it makes liaisons where there is much less close linkage, as in most of the other examples, including the Port-au-Prince of the topic. I would say it's because chers instants is so fossilized that you get that liaison, whereas chers amis is freer. Similarly with vas-y and s'en va-t-en guerre, but va à ta chambre and s'en va en ville, mis en bouteille, but prends-en. All right, it should be la mise en bouteille, but if the French can't get it right… What the heck, someone else have a go.

    Because what you really need is some mention of exceptions to your rule with nouns and adjectives ending in these sequences.

    The diacritic problem is not caused by our browsers. I have come across sites in the past where this happens, and found that you have to put the diacritic first. I don't think that has always been the case on this one, but it does work on Linguism, so here goes:

    pɔr o pr̃ɛs

  12. Dear mallamb,

    my "rule" of not making liaison from -rd, -rt, -rs endings extends far beyond non-inverted verbs (although my examples were all verbs, as you point out.) Bear in mind one of the main points of the original post: the noun "port", treated above as a proper noun. Excepting some expressions, common nouns with these endings should never take liaison : Ce port en France. [pɔ ʀ̃ɑ]

    There are many -rd and -rt nouns which behave in the same way (that is "enchaînement" of the "r", not a liaison as [t] ), including:
    nord (du nord au sud)
    lourd (un lourd objet)
    bord (bord à bord)

    vert (le vert et le jaune)
    mort (la mort est…)
    tort (faire du tort à quelqu'un)
    part (de part en part)

    -rs not in verbs and plural nouns is rarer. The prepositions "vers" and "hors" spring to mind — they both behave the same way: no liaison of [z], as in vers eux [vɛ ʀø]

    Another tricky exception is when words with these endings occur as adjectives and adverbs before words they directly modify. Compare "Toujours en retard" [-ʒu ʀ̃ɑ] with "toujours heureux" [-ʒuʀ zø-] , and also "fort excès de" [fɔ ʀɛ-] with "fort aimable" [fɔʀ tɛ-] .

    The interpolated "-t-" you mention is a curious case which prevents hiatus.

  13. Thank you for agreeing we needed n and adj examples. I hope you didn't think I hadn't thought of any of yours myself, but I was really looking for some key examples such as you have now come up with: "Toujours en retard" [-ʒu ʀ̃ɑ] versus "toujours heureux" [-ʒuʀ zø-] , and also "fort excès de" [fɔ ʀɛ-] versus "fort aimable" [fɔʀ tɛ-] .

    I clearly didn't think hard enough, as they are obvious examples of the principle I was trying to formulate: they have different degrees of conventionality in their structural relationships. And as I said before, this doesn't necessarily mean the conventionality is fixed one way or the other. Thus here again "Toujours en retard" [-ʒu ʀ̃ɑ] cannot have liaison but "toujours heureux" [-ʒuʀ zø-] is merely less likely not to have it, which is a distinction of less than fully fixed conventionality, whereas "fort excès de" [fɔ ʀɛ-] cannot have it and "fort aimable" [fɔʀ tɛ-] must have it, which are fully fixed conventions. Am I right?

    Another case of the fossilization I mentioned, but what has become fossilized here is I think the adjectival behaviour in "fort excès de", parallel to un lourd objet, and an adverbial behaviour parallel to toujours heureux, but in a bound expression which no longer permits of an alternative variant.

  14. ə, ɘ or a narrower ə between consonants are in some international phonetic alphabet English dictionaries

  15. On Port-au-Prince, are the people who don't say [prɪn(t)s] saying it like English prance (northern pronunciation) or are they actually using a pseudo-French nasalised vowel? I think I'd use the latter.

  16. @ mallamb,

    yes, I agree with all your comments above.

    We classical singers pay a lot of attention to what is sometimes referred to as "lyric diction". In the case of French, the so-called "style soutenu" is what we use on stage and in concert in Mélodie (French Art Song) and opera. The pronunciation dictionary by Warnant is unparalleled in its attention to detail at both the everyday level of speech (parler courant) and this formal level (style soutenu), with variations for both levels of speech indicated. Such details include vocalic harmonization, the "e caduc", and optional liaisons; the latter is what we're on about with the example "toujours heureux". As you indicate, it is possible not to hear the liaison, even though it is permitted in this adverbial usage. In formal song, we'd sing it with liaison. Most of the optional liaisons are observed in opera and Mélodie in order to preserve the consonant-to-vowel "flow" so inherent to the language, and indeed, sought after in fine singing.

    It is a shame the last edition of the Warnant (1987: Duculot) is out of print — very luckily, I got my copy from Warnant's widow, who was contacted on my behalf by a former colleague and collaborator of the author.

  17. nedecky, mallamb: on my screen your tildes appear neatly above the ʀ, not above the vowel. (I'm using Firefox on a PC running Vista.)
    It's nothing to do with Mac vs PC, nedecky: both platforms are Unicode-compliant, and in HTML/XML you can enter Unicode numbers either in decimal or in hex (with a leading x in the case of the latter). Or in my case use a phonetic keyboard and type the IPA characters direct.

  18. Dear Prof. Wells,

    I entered the tildes that appear on your screen over the R _before_ the nasal vowels, trying out Kraut's recommendation here. On my screen, they are the ones that appear correctly, whereas the ones I did in my first post today (entered after the vowel) appear incorrectly! Very odd…

  19. I have just learned more about the proper use of liaison (under some circumstances at least) than I ever learned in French class. My thanks to everyone who has increased my store of knowledge. I still can't speak French, but from now on, I shall not-speak it better!

    John, I meant to say something earlier about the video. One of the things I value most in LPD is the survey information. So while I don't think I learned anything new about the surveys, I greatly enjoyed hearing you talk about them.

    As for which/witch, there are many, many Americans who still use ʍ (or hw, if you prefer) in the former. So many that I'm often surprised at the number of w-saying Americans who haven't noticed the ʍ-saying Americans. I'm w-saying myself, but I switch to ʍ easily - an outgrowth of instruction from my best friend (we were both 10) that only ʍ was "correct." I learned it so I could be correct, too - even though nobody I knew (my best friend's family excepted - her parents were from Massachusetts) used ʍ. Such is the power of peer pressure.

    I advise my clients to learn whichever pronunciation is unfamiliar to them, so that they too can switch with ease when a role requires it. Other than that, I can't say I think it matters much. The use or non-use of ʍ has never led to misunderstandings in my experience, although the use of ʍ can kill any number of jokes that depend on the use of w.

  20. My village in France was called Megrit. I diligently omitted the for a long time until I heard the locals pronouncing it.

    What an excellent video in that it keeps pace with current trends and removes the set-in-stone approach of former dictionaries!

    The ʍ or hw sound certainly disappeared from the OED a while ago, but I used to hear it regularly in the UK in my childhood in the 60's, and Dame Judi Dench as 'M' in the Bond films still employs the technique. When ð in 'mother' becomes in some speech (muvver), and ɵ in 'think' becomes (fink), we feel this is uneducated. Were it not for the work of educators, the LPD in 2050 might have to include these, and our precious digraphs be consigned to history. In the days when we used to say the in 'eight' and 'daughter' as /x/ we didn't have the benefit of a pronunciation dictionary.

    I have always said /ˈlɪkərɪʃ/ but I didn't know there was another word 'lickerish' meaning lecherous, greedy, longing, and fond of fine foods!
    Just as I cannot write a good looking 'Iota' above, I cannot see any tildes over vowels in the post in Windows IE 7.
    Maybe we ought to resort to screen-shots.

  21. Interesting! I did not know that you had a YouTube channel. I'm really a fan of your dictionary. Thanks!
    Kid regards from Argentina,

  22. @All:
    I tried IE v8 having selected UTF-8 for character encoding – the result is not very satisfactory: The tilde is exactly above the next letter, phonetic symbols vary in size.
    I also tried Flock v1.2.7 with UTF-8 – the characters have identical heights but the tilde is right above the next letter. I normally use Firefox (my current v3.5.5 – UTF-8): the tilde is above the next symbol. With all three browsers the tilde is above the epsilon in my second post (18 January 2010 14:23).

    Your conclusion that you ”still can't speak French, but from now on, [you] shall not-speak it better” made me laugh. Thanks for that.

  23. The placement of diacritics typically depends on the *font* rather than anything else. I know Arial Unicode MS puts tie bars in the "wrong" places (they are supposed to be code between the two characters they're supposed to tie), so I wouldn't be surprised if it also put combining tildes in the wrong places. Try DejaVu Sans/DejaVu Serif.

  24. (And if someone comes to pronounce "prince" indistinguishably from "prints", then the /t/ *is* phonemic for them.)

  25. John W, having finally got round to following up the video, I was intrigued by your pronunciation zʊˈɒlədʒɪ at 3.42 or so.

    I feel sure it has been around all my life, but have said zəʊˈɒlədʒɪ myself as long as I can remember, and so have my immediate family and beyond, all people not given to spelling-pronunciations or etymologizations. And "as long as I can remember" is almost to my infancy: I remember "correcting" my "defective" r to a retroflex one (I always had a tap between vowels, for which I was mocked, but the "defective" allophone was more girly than plenty that got by, and I didn't stand up to the mocking!) and I remember another kid correcting his lisp, which was quite impressive – he was entitled to one, as he could touch the tip of his nose and the point of his chin with the tip of his tongue!

    But I don't remember being "corrected" to zəʊˈɒlədʒɪ, or adopting it from Greek, or analogizing from some word such as protozoon, and I think I would remember. And there are a great many words like that, from zoon itself to zoophyte and beyond for which I also have zəʊ. Does anyone pronounce any of them with zu?

    I now find that OED says, zəʊˈɒlədʒɪ, popzuːˈɒlədʒɪ in the 1989 update for that entry, I think for the first time, but an astonishing number of dictionaries both online and off don’t recognize it at all. Not even Wiktionary.

    What some of them do recognize for AmE alongside zoʊˈɑlədʒi is zəˈwɑlədʒi, obviously in various variously inadequate but decipherable transliterations. Amy &Co, my guess is that that is for zuˈɑlədʒi, with the strong w glide that u has in AmE and the broken ʊw reduced to əw. But conceivably it could also be the ow of zoˈwɑlədʒi reduced to əw. Any thoughts?

  26. Despite the reduced outcome, I was mentally aiming at zəʊˈɒlədʒi !

  27. army, consider the ways in which they may pronounce them indistinguishably:

    If with [ns] there is nothing t to BE phonemic and if with [nts] then the t is still not phonemic as it is not functional in that environment, and can be predicted as a purely phonetic phenomenon due to the transition between the n and the s, not the realization of a phoneme. So again there is no phoneme there.

    The phenomenon of miscellaneous noises which may happen to sound like the realizations of phonemes is a commonplace of phonology. It is called epenthesis, or anaptyxis if the interpolated noise is vocalic.

  28. BTW Amy, in some Gaelic-influenced dialects of BrE wh- is regularly [xw]. Almost Anglo-Saxon!

    And if it is not, one finds an etymological reason for that, e.g. whelk never has xw, and OED says "the unetymological spelling with wh begins in the 15th cent.", but a quick scan of the citations reveals no example of that spelling earlier than the 17th, and it doesn't seem to have taken hold until the late 19th.

    To me it seems that the unetymological pronunciation with h is a greater sin than the spelling, and James M, it hasn't disappeared from the OED. The spelling-pronunciation with hw is the only one they give for whelk! In the 1989 updated entry!

  29. nedecky, mallamb: on my screen your tildes appear neatly above the ʀ, not above the vowel. (I'm using Firefox on a PC running Vista.)

    Same here, with the tildes in the blogpost itself fine (Opera, Vista).

    Video: You probably did that simply to make your point more clearly, but you sound - rightly, I think - as if you were saying foreigners tended to pronounce your as [juːə] rather than like the NS variant [jʊə].

    Had the poll asked about Asia with [jə] or [ɪə] as well?

  30. Mallamb
    Sorry I should have said it has disappeared from MY OED the Ninth Edition printed in 1995. There is no reference that I can find to wh sounding as hw or ʍ. Not even for whelk!
    I hadn't realised it was unetymological and I value your opinion.

  31. Thank you James for that truly relevant update. Most enlightening and reassuring. But there are no tsarinas around here to mug for the funds for new print editions. The OED online sub is ruinous enough. They occasionally try to mug me for the latest CD, but it would probably be even more ruinous to keep getting updates of that.

    They are also pushing a new periodical revoltingly called Illuminea. How do they intend one to pronounce THAT inanity? At least when the idiots at the Post Office squandered vast sums on the idiotic rebranding "Consignia" and more vast sums to change it back again, the opening fanfare included the pronunciation.

    OUP cannot I think realize that there are any number of other things also given this name, probably by the same superstar admen, but the others are not at the same risk of making themselves the laughing-stock of the literate world.

  32. @mallamb:

    If I understand your discussion of "prince" correctly, you are saying, in the terminology of Accents of English, that the distinction between /nts/ and /ns/ is neutralized for some speakers.

  33. I have noticed American reporters pronouncing "Haiti" with a voiced "d" instead of "t," although the NBC live broadcasts from the island seem to favor the "t." The Oxford American Dictionary that came with my Mac tells me we Americans are expected to use "d."

  34. Well, dear fellow readers, it seems to me that at least on Windows XP there is indeed something strange going on with the fonts. I've actually taken a couple of screenshots from Firefox 3.5.7 and Opera 10.01. You can see them here. Same font (DejaVu Sans), same system.

    On my machine, Opera does it right but FF seems to have the problem people have been describing. It looks like they differ in how they mix the "general" font with the font required for IPA. Look at the shapes of "j" and "y". Opera is smart enough to use DejaVu for the whole transcription, thus rendering the diacritics in the correct places. FF mixes the two fonts (which is also apparent from the somewhat uneven size after zooming in so much -- I did zoom in a lot for the screenshots), and the interaction causes the problem.

  35. OK, a quick update: On my (old) XP machine, it's Verdana that's doing this. If you want to know if your system is susceptible to this, check here. Annoyingly, Word 2007 is doing exactly the same.

  36. This comment has been removed by the author.

  37. The bl**dy tilde positiong problem is a matter of your browser option settings and the fonts that you have on your harddisk:
    pɔtopx͂ɛs 1. tilde, 2. epsilon
    pɔtopxɛ͂s 1. epsilon 2. tilde

  38. @Kraut: Nope, the problem is Verdana, as described here. I can attest that the fix they describe works 100% correct, and that's why it mainly affects people on FF/XP. As they say, and as I wrote las night, Opera does have an internal fix but underlyingly it's a font problem.

  39. For Firefox users (tested with v3.5 under WIN):
    1. Have a unicode-compliant font installed;
    2. Under Tools -> Options -> Content push the button "Advanced..." next to "Fonts & Colors".
    3. Disallow pages to choose their own fonts.
    4. Select a unicode font for both serifed and non-serifed fonts.
    5. Press OK.
    Hope that helps!

  40. BTW, the elision in last night above was totally accidental. But how appropriate for this venue!

  41. @Kraut: Yes, you can do that, too. But then you'll be using the same Unicode font everywhere in FF, and the behaviour won't be fixed in e.g. Word. Updating the font is a system-wide solution.

  42. This comment has been removed by the author.

  43. @wiarek: Thanks for drawing my attention to the cause(s) of the problem and the solution. The Microsoft update can be found here:

  44. 1. nasal x
    2. fine

    Opera, no extra fonts installed or settings changed. Here's the source code:
    <Variable name="bodyfont" description="Text Font"
    default="normal normal 100% Verdana, sans-serif">

  45. Sorry, hadn't refreshed, so I hadn't seen the last six comments.

  46. @Kraut: All part of the service!

    @mallamb: "Amy &Co, my guess is that that is for zuˈɑlədʒi, with the strong w glide that u has in AmE and the broken ʊw reduced to əw. But conceivably it could also be the ow of zoˈwɑlədʒi reduced to əw. Any thoughts?"

    My first thought is that personally, I say zoʊˈɑlɪdʒi, though perhaps zoʊˈɑlədʒi when speaking especially rapidly.

    My second thought is that there is a (fairly) popular book here in Leftpondia called There Is No Zoo in Zoology: And Other Beastly Mispronunciations, by Charles Harrington Elster. I have never read it, and I suspect from the subtitle that I'd find it full of regrettable overstatements if not outright errors, but the title is catchy — I've never forgotten it.

    My third thought (three thoughts in one day!) is that probably most Americans pronounce zoology more or less as zuˈɑlədʒi or zʊˈɑlədʒi, but that a strong w glide is pretty much limited, at least in my experience, to broad Noo Yawk accents. It is, of course, entirely possible that you mean something different by "strong" than what I read into it.
    Let me remind the reader(s) that:

    (a) My own experience of American accents is necessarily limited;
    (b) I'm fundamentally self-taught, hence the greater number of large gaps than filled-in places in my knowledge and understanding of phonetics;
    (c) I'm not an American, I'm a New Yorker!

  47. @mallamb: "in some Gaelic-influenced dialects of BrE wh- is regularly [xw]. Almost Anglo-Saxon! […]

    Oy vey. My head is spinning. I'll read that more carefully and try to take it all in another time!

    @alarob: "I have noticed American reporters pronouncing "Haiti" with a voiced 'd' instead of 't,' […] The Oxford American Dictionary that came with my Mac tells me we Americans are expected to use 'd.'"

    I have not been listening to American "reporters" because what passes for "news" in the US bears the same relationship to news that my ass does to my elbow. But if the OAD really says that Americans use a d in Haiti, than the OAD, like the law, is a ass. What most Americans use for t between vowels - provided the syllable before the t is the stressed one - is a voiced t [t̬] (in case it doesn't read correctly, the diacritic is meant to be a subscript wedge. I think of it as a tiny little v (as in voiced) beneath an otherwise voiceless consonant). Some people use [ɾ] to indicate this sound, which I suppose is a reasonable choice for anyone who makes their r sounds as alveolars. My ɹ, ɾ, and r are post-alveolar, so that would not be a good choice for me.

    For those of my clients who can't or don't want to read IPA, I advise them to notate the voiced t as "t/d" — "half t, half d." It may not be pretty, but it works!

    Not that I'm opinionated or anything ... but Americans do NOT say d in place of t. We just don't. Stupid dictionaries ...

  48. That notion was made popular by British phoneticians as a revenge for your papers claiming we say veddy vor very.

  49. @Amy:
    You say: "What most Americans use for t between vowels - provided the syllable before the t is the stressed one - is a voiced t [t̬]": I'm afraid I can't fully agree with you on this statement. What about "at all". Don't you flap the /t/ here? It's a [ɾ], isn't it? And what about words such as 'latex, syntax' or 'quite obvious, get over here, start off' and a few others?

  50. Of course you could analyse [nts] as underlying /ns/ for speakers who don't have [ns], but you'd have to complicate the rule for the formation of plural nouns/third person verbs so that they allow for /prInt/ to become /prIns/, and I can see little good reason to do that. I'd prefer to say that their phonotactics doesn't allow /ns/.

    And flapping does not depend directly on stress, but on the position in the syllable; you can flap the t in at all, for example, even if it's pre-tonic. I think the environment in which that happens typically are /Vt.V/, /Vrt.V/ and (in America) /nt.V/, where syllabic l and m count as V's. But Freddie Mercury sometimes also flapped them pre-pausally.

  51. @Amy:

    Many US speakers do lose the distinction between /t/ and /d/ intervocalically, regardless of what the _phonetic_ output is. For evidence, see "very suddle" as a typographical error for "very subtle".

  52. @Lipman: Yes, the limitations of eye-dialect cut both ways, but I do expect better from dictionaries.

    @Kraut: "I'm afraid I can't fully agree with you on this statement. What about "at all". Don't you flap the /t/ here? It's a [ɾ], isn't it?"

    [ɾ] is phonetically wrong for me (and for many other speakers of AmE), where t is between vowels in words such as Haiti, and a phrase such as what a dump. That is because my tapped r [ɾ] is post-alveolar (I don't use it in my every day speech, but I do in accents that call for it: RP ca. 1930 springs to mind.) My voiced t, on the other hand, is alveolar. I can feel the difference when I speak, even if others might not hear it.

    For many other speakers of AmE, [ɾ] would be a perfectly good symbol to describe the t sound in the middle of "Haiti" (as most Americans pronounce it), - but it's counter-intuitive for me to use that symbol for the sound, even if correct. I associate it too strongly with the letter r.

    "And what about words such as 'latex, syntax' or 'quite obvious, get over here, start off' and a few others?"

    For most Americans, these are not all the the same thing. In "latex, syntax" t is definitely [tʰ]. In the short phrases you mention, and also "at all,", the tendency of most Americans is to use voiced t, even thought you'd think these phrases would be analogous to "latex, syntax." I don't know why they aren't, but they aren't.

    There are American speakers who do "pop the t" in those phrases - mostly from parts of the American south, I think.

    @vp: Put that way, I'd have to say you're right, but I still wouldn't put d in the dictionary for those words, because there are more than enough Americans for whom it really isn't true. And isn't the OAD making phonemic distinctions, not phonetic ones? If the version I've just looked at on Amazon is anything to go by, OAD doesn't even use the IPA. So there's really no excuse for their putting a d in "Haiti."

    (I confess to being puzzled as to why someone who wanted a good American dictionary would buy one from Oxford. OAD has, just for example, entries for A4 and A5 (paper sizes), and these are not American English. Merriam-Webster would surely be the obvious choice. I suppose there must be a good reason, but I'm baffled.)

    In my own speech, the distinction between "latter" and "ladder" is pretty subtle (not suddle!) but I do feel a difference most of the time. The more rapidly I speak, the more the distinction disappears into a merger of t with d, but then the d is as much devoiced as the t is voiced.

    As for the spelling error: Yikes! I haven't come across that one before. I really don't think I could describe my own pronunciation of "subtle" as having a d in it. Just a t, sometimes fully unvoiced, sometimes voiced, but not ever an actual d.

    (I suppose I've made things less clear than I might have, by saying "voiced t" instead of "partly voiced t.")

    I'm a lousy typist, but my spelling is too good to wrap my mind around "suddle." You've provided the evidence, but I just can't fully take it in.

  53. In "latex, syntax" t is definitely [tʰ]."

    Really? In GenAm? I have never heard aspiration here (in a noninitial, nonstressed syllable), except in certain varieties of London speech.

  54. David Marjanović24 January 2010 at 15:49

    in some Gaelic-influenced dialects of BrE wh- is regularly [xw]. Almost Anglo-Saxon!

    There's an episode of Family Guy that revolves entirely about pronunciation of wh as [xʷ]. It's on YouTube somewhere.