Toe-ly gumby a sound change

« previous post | next post »

On Sunday 9/10/2017, Steve Bannon was interviewed on 60 Minutes. Looking at the interview from the perspective of a phonetician, I was struck by pervasive evidence of a little-studied sound change in progress. Word-internal intervocalic coronal consonants — /t/, /d/, /n/ — in weak positions (i.e. not followed by a stressed vowel) are deleted, and the surrounding vowels are merged. This process is increasingly common in American English, and is frequently exemplified in Steve Bannon's speech, at least in this sample.

Let's look at a few examples. Early in the interview, Charlie Rose says "You are attacking on many fronts people who you need to help you, to get things done", and Bannon responds:

They- they're not gonna help you
unless they're put on notice they're gonna be held accountable

if they do not support the president of the United States.
Right now there's no accountability.

They have totally
they do not support the president's program, it's an open secret on capitol hill,
everybody in the city knows it.

I've put in bold nine words where this process might apply. Let's take them one at a time.

(1) In the first example, "gonna" (which I'll assume is a lexicalized reduction of "going to") is pronounced as [ˈɡʌ.nə] — the intervocalic /n/ is a short (25 msec.) ballistic tap, but it's clearly there, and the syllable coalescence doesn't happen:

they're not gonna help you

gonna help

(2) The second example goes the other way — coalescence applies, and the sequence "gonna be" becomes just two phonetic syllables, [ˈgʌ̃.bi]. The /n/ leaves a nasalized vowel as its residue, and the second syllable of "gonna" is gone, with maybe some additional shading of the vowel following the [g]:

put on notice they're gonna be held accountable

gonna be

(3-4) The third and fourth examples are more striking, because some additional reductions some into play, and the whole sequence "president of the United States" comes out as something like [ˌprɛz.juˌnɑʲˈsteʲts]. To get this, we need not only to merge the second and third syllables of "president" and of "united" and reduce to nothing their final consonants, but also to elide "of the". Presumably for Mr. Bannon, "president of the United States" has become a low-entropy fixed expression, subject to extreme reductions like those that turn "Worcester" into [ˈwʊ.stɚ]:

if they do not support the president of the United States

president of the United States

(5) In the fifth example, coalescence applies, and "totally" become [ˈtoʷ.li]:

They have totally


(6) In the sixth example, "president's" becomes something like [ˈprɛ.zɪ̃z]. This exemplifies not only coalescence of the second and third syllables, but also simplification of the final consonant cluster, which is partly due to loss of the nasal murmur and partly to the loss of the /t/ closure. Both of these last phenomena are also widespread and almost obligatory — for example, Americans rarely pronounce the /t/ in final /sts/ clusters (as in "artists").

they do not support the president's program


I'll leave you to check the remaining three cases — "capitol", "everybody", and "city" — but my evaluation is that none of them show coalescence, so that the final score is five out of nine.

There's obviously more to be said. Are coalescences of this kind the end point of gradient phonetic reductions, or instances of an essentially quantal or symbolic "phonological rule", or new pronunciations entered in the mental lexicon as a result of either phonetic or phonological changes? Or maybe all three? For more discussion (than you probably want to read), see my paper "Towards Progress in Theories of Language Sound Structure", in a forthcoming festschrift for John Goldsmith.



  1. Jarek Weckwerth said,

    September 12, 2017 @ 6:53 am

    Well, I've been listening to quite a lot of political material from the US lately, and I've found it really striking how the phrase President of the United States is highly reduced for most of the talking heads. So I think the "entropy thing" is a major factor (as it always is, of course).

  2. David Gil said,

    September 12, 2017 @ 7:19 am

    I was with an American friend at a restaurant in Sri Lanka. He was thirsty, and asked the waiter for some [r], just a syllabic [r] with some schwa colouring perhaps, and maybe, or was I imagining it, just a hint of rounding at the onset of the syllable. The waiter had no idea what he was asking for, and my friend was getting more and more frustrated, and kept on repeating his [r], longer and louder. All to no avail, until I finally took pity on them both, and in my best British accent, said [wɔ:tə].

    [(myl) You mean [wɒ˞]? Totally obvious, like [o] for "aqua".]

  3. J said,

    September 12, 2017 @ 7:37 am

    For 'president of the United States', I don't hear [ˌprɛz.juˌniˈsteʲts] but [ˌprɛz.juˌnɑʲˈsteʲts]. I think it's just a typo, but … I'm not a world-famous phonetician! :-)

    [(myl) More like a braino — blame the Great Vowel Shift. Thanks for the proofreading, it's fixed now…]

  4. BZ said,

    September 12, 2017 @ 8:21 am

    Is any of this really new? All of these reductions are unremarkable to me in fast speech; I might speak like this myself (minus the accent). It reminds me of the supposed "Fluffia" regionalism, but try saying "Philadelphia" quickly and tell me it doesn't sound like "F'luffia".

    [(myl) No, it's certainly not new. But the range of reductions/deletions/changes involved hasn't been systematically studied.]

  5. Jerry Friedman said,

    September 12, 2017 @ 8:30 am

    You mention the loss of the /t/ closure in "president's", but I've been told Americans pronounce "prince" with a /t/, and I think I do, though not much of one. Do most of us have a prince-prints merger? Do some of us say it with a /t/ and some without?

    [(myl) It's complicated. But let's start by establishing the fact that most Americans do not actually pronounce a phonetic [t] in the final cluster of words ending in unstressed /Vnts/. For example, in the 2015 Fresh Air interview with Dr. Vincent DeVita about cancer treatments, there are 11 instances of the word "patients" spoken either by Dr. DeVita or by Terry Gross, and not a single one of them has a phonetic [t]:


  6. Mark Meckes said,

    September 12, 2017 @ 9:45 am

    I thought I remembered it coming up here before that "isn't" is sometimes pronounced with a short flap or tap or stop in place of the s.

  7. Gregory Kusnick said,

    September 12, 2017 @ 12:07 pm

    During Bob Dole's 1996 presidential run, I recall hearing a piece on NPR about Dole's truncated pronunciation of "presidency" (something like "prezny" if I remember right). But a quick Google search fails to turn it up.

  8. Mark Meckes said,

    September 12, 2017 @ 1:01 pm

    Ah, here is the post I was thinking of. Not "isn't", but "didn't".

    [(myl) Thanks for the cross-reference — the coalescence of two syllables across /d/ in "didn't" is indeed exactly the same process as noted above in "gonna", "president(s)", and "totally". And the loss of the final /t/ is related to process involved in the final cluster of "presidents" (and "patients"), though the environment is a little different.]

  9. Chris C. said,

    September 12, 2017 @ 4:49 pm

    There's a female personality on my local radio station who does this often, even when speaking slowly and carefully. It's particularly noticeable in a spot she often reads for a charity raffle that's run a couple of times a year. It sounds much more odd in slow speech than when spoken quickly, and I can never escape the impression that she sounds drunk.

  10. Rubrick said,

    September 12, 2017 @ 5:44 pm

    Given that, as you said, there's been limited study on this phenomenon, I'm curious what evidence leads to the conclusion that it is in fact on the rise. (Curiosity here is genuine; I've been reading you much too long to suppose you pulled that assertion out of your hindquarters!)

    [(myl) Well, this is clearly a phonologically-conditioned phonetic change, with perhaps some lexicalization as well, so it counts as a sound change. And it's variable, as Mr. Bannon's two pronunciations of "totally" illustrate, so it's a change in progress.

    Could it be a case of stable variation? For a change involving phonetically natural but extreme reductions, which are not stigmatized or associated with a salient internal group identity, that seems unlikely. (Though this particular change seems to be further advanced in AAVE, and also maybe in some regional variants — Steve Bannon grew up in Virginia — it's not something that most Americans trigger on, as far as I can tell.)

    Could it at least be a very slow change, taking place on a time scale of centuries rather than decades? That's certainly possible.

    I have the impression that this kind of coalescence is more common than it was when I was a kid — you can form your own opinions about what part of my anatomy this impression comes from — but I don't have any concrete evidence. And it might simply be that American English in formal settings is less formal than it used to be, just as dress codes have changed over the decades; or that I'm now living in Philadelphia rather than Boston or northern New Jersey.]

  11. Michael Watts said,

    September 12, 2017 @ 7:06 pm

    I recently deleted the /b/ in "maybe"; how sure are we that this is specific to coronal consonants?

    [(myl) Word-internal intervocalic position following a stressed syllable and preceding an unstressed syllable is a weak place for ALL consonants — in some sense, we all want "stress feet" to turn into single syllables. Coronal stops and nasals are more weakenable (if that's a word) than others, because the tongue tip is low in mass and relatively flexible, but other consonants have shorter closures, smaller areas of contact, etc., in the same weak positions. The "flapping" of /t/,/d/,/n/ presumably starts as a combination of those general factors, but clearly more is going on, because most American speakers now (for example) produce a voiced flap (or an approximant, or nothing) in "city" even in relatively slow, careful, and formal speech.]

  12. Lazar said,

    September 12, 2017 @ 7:38 pm

    I knew a couple kids in high school who had deletion of weak /t/ and /d/, for example saying "better" as [ˈbɛɚ].

  13. Anne Cutler said,

    September 13, 2017 @ 12:29 am

    When someone gets around to doing this research, I hope that attention is given to the duration of the sounds that still occur. A vague recollection has stirred concerning the "disappearing 'the'" in Yorkshire English; if the recollection is based on fact rather than wishful thinking, then the closure in "at mill" for "at the mill" was longer than the closure in a matched phrase such "at Myf's". So would the [o] in the bisyllabic "totally" be longer than the [o] in, say, "holy"?

    [(myl) A good question! My guess, in advance of the facts, is that this varies — as it must, over historical time, as the lexical forms change from (e.g.) aqua to eau.

    As for the Yorkshire "the", could it be like the Luxembourgish d' or Dutch 't, which seem sometimes (in my inexpert impression) to turn up as lengthening of an adjacent consonant?]

  14. Joyce Melton said,

    September 13, 2017 @ 2:36 am

    My experience with this phenomenon is based on visits to New Jersey. Everyone in the south half of the state seems to talk like this and the north half seems headed in that direction.

    Have you ever been to 'lãi:si:'i (Atlantic City) in the off season when almost everyone you meet is a local? I sat in a coffee shop for two hours just listening to cops come and go and speak in this bizarre abbreviated patois. Bizarre for an Arkansawyer who grew up in Inland California, at least.

  15. ardj said,

    September 13, 2017 @ 4:26 am

    re: @Jerry Friedman & MYL response
    Neither phonetician nor American, I can't help wondering if the difference between the t in "prints" and in "patients" is (partly ?) because the latter t comes at the end of an unstressed syllable.

    [(myl) That's clearly a factor. My impression is that the fate of the /n/ is an important part of the picture — when there's a clear vowel + nasal murmur + fricative sequence, an intrusive stop segment is likely. When the vowel + nasal murmur part turns into just a nasalized vowel, or just a syllabic nasal, I think that the stop — whether intrusive or original — becomes much less likely. This should be checked, obviously.]

  16. Jerry Friedman said,

    September 13, 2017 @ 9:45 am

    MYL: Thanks for the comments to me and ardj and for the examples. I think I hear a tiny little /t/ in some of the samples, though less than half. This may only prove that I'm not a linguist.

    I can imagine a firs' step of comparing "patience" and "patients", "residents" and "residence", etc.

  17. Andrew Usher said,

    September 13, 2017 @ 5:44 pm

    I think that would be the wrong question. All native speakers, so far as I know, merge pairs like patients/patience in normal speech. The question of whether any [t] is audible is a matter of idiosyncratic or random variation, rather than spelling – just like linking/intrusive 'r' at word boundaries of non-rhotic speakers.

    There are other clusters in which such an 'epenthetic' stop can easily appear/disappear, but none that are so uniformly merged. But I can guess the people that regularly have [t] in those words are also likely to insert a [t] in 'else' and say the 'p' in 'empty' and 'Thompson' (I don't). In fact if I were asked for a word with silent 'p', 'empty' is the first that would come to mind, but I wouldn't say it because I think I might be laughed at.

    k_over_hbarc at

  18. Kenny Easwaran said,

    September 13, 2017 @ 9:17 pm

    Is this the same as the phenomenon of the word "alright" being pronounced (and even spelled) "a'ight"? I actually noticed myself saying it just before I remember noticing seeing the spelling elsewhere.

  19. Andrew Usher said,

    September 13, 2017 @ 10:18 pm

    No, surely that isn't a sound change because R before a stressed vowel doesn't delete anywhere else. I think the particular king of laziness that produces 'a-ight' is peculiar to interjection words. I don't know if I've ever said it that way, but certainly I do say the more common L-deleting pronunciation 'awright'; that may indeed be a sound change, though the only common words it applies to are alright, already (ophthalmologist being the only other I can think of right now – the L must be followed by m, n, or r, then a stressed vowel) – by the way, this is another illustration of how 'alright' is not 'all right' (L can't be deleted in the latter, unless you habitually vocalise L) and pedants that prescribe only the latter spelling are ignorant.

    Addressing the subject, I can't take seriously any transcription that puts STRUT in the first syllable of Bannon's "gonna" – it's very clearly schwa, as it normally is for most Americans including me. It is inexplicable how one could really think it sounds like 'gun a'. In fact, the closest other vowel to that schwa is FOOT, and before I learned about phonemes that's what I thought I used in 'gonna' and adverb 'just' (the latter is spelled 'jist' in eye dialect, showing the other vowel closer to schwa than STRUT is).

    But it is true that he entirely elides the second syllable of 'totally' – evidence of sound change? No, not itself, not he, nor anyone else, would say 'totally' that way in isolation. Without much stronger evidence it just seems like more reduction of 'low entropy words', which everyone exhibits to an extent, and the reductions are not always predictable or recorded in dictionaries. Even if the word became a perfect rhyme for 'holy', thus, it could not be mistaken in that context.

  20. J said,

    September 14, 2017 @ 8:29 am

    " the fact that most Americans do not actually pronounce a phonetic [t] in the final cluster of words ending in unstressed /Vnts/. For example, in the 2015 Fresh Air interview with Dr. Vincent DeVita about cancer treatments, there are 11 instances of the word "patients" spoken either by Dr. DeVita or by Terry Gross, and not a single one of them has a phonetic [t]"

    This is why I read Language Log! Finally, confirmation that I'm not MAD! :-)

  21. Anthony said,

    September 14, 2017 @ 8:59 am

    I have no 't' in apartment; it doesn't disappear but assimilates to 'p'.

  22. Andrew Usher said,

    September 14, 2017 @ 7:25 pm

    Aparpment? I don't think so. It has been shown such assimilations are not complete. While the release of the stop may seem labial because of the continuous transition into the following consonant, the closure is assuredly (unless you speak in a way unfamiliar to me) alveolar (and/or glottal).

    Possibly some non-native speakers do produce such un-English assimilations; I haven't noticed.

  23. Daniel said,

    September 15, 2017 @ 12:27 pm

    This reminds me of this older post about the country song "Somethin' 'Bout a Truck".

  24. Chas Belov said,

    September 16, 2017 @ 8:27 pm

    @Jerry Friedman: I do add a t to prince (prints), as well as a p to Amsterdam (Ampsterdam). But I too have caught myself t-dropping.

RSS feed for comments on this post