Toxic bird pits?

« previous post | next post »

David Leonhardt, "Did He Say 'bird'?", NYT 2/10/2023:

As President Biden was reciting a list of bipartisan accomplishments during his State of the Union address this week, he seemed to use a phrase that I had never heard before: toxic bird pits.

Was it some major news story that I had missed while on leave over the past few months? Or was it the latest Biden malapropism, destined to dominate post-speech commentary? I tried to figure out the answer by typing the words into Google and Twitter, but they offered no clarity. Google had nothing for me. A Twitter search yielded dozens of people tweeting a version of “toxic bird pits???” and not much else.

Like a large percentage of the mass- and social-media pieces over the past few weeks, Leonhardt's article is actually about ChatGPT — subtitled "We talk with Sam Altman, the C.E.O. of OpenAI, about the artificial-intelligence craze."

There's plenty to say about Large Language Models — I'm fated to talk about them this afternoon, in a panel discussion organized by Penn's Critical Writing Center — but this post is actually about phonology and phonetics.

Executive summary: Biden didn't mis-speak; Leonhardt mis-heard. Or to be fair to both: Biden's pronunciation of "burn pits" was within the norms of current American speech patterns, though maybe a little under-articulated given the low contextual likelihood of that phrase for his listeners; and Leonhardt's mishearing was also reasonable, given the signal and his mental language model.

And the details are exactly on point for the current topic in a course I'm currently teaching — Phonetics II: Data Science — where we're discussing the nature of "allophonic variation", i.e. the different ways that speech sounds of a given category can be realized in different contexts.

Specifically, syllable-final nasals in English (like the /n/ in "burn") cause the previous vowel to be nasalized, but depending on the context, the canonical "nasal murmur", which the dictionary says should follow, may be extremely short or even entirely absent. Such lenition or deletion is especially likely when the next syllable starts with a voiceless consonant (like /p/). And syllable-final coronal stops, like /t/ or /d/, are also subject to lenition, to the point of apparent deletion when a consonant follows, even in "standard" formal speech — consider the (often imperceptible) difference between tends and tens, or tents and tense, or tend to and ten to.

And of course the vowels of burn and bird have been phonologically the same in (most varieties of) English for several centuries. So the difference between "burn pits" and "bird pits" will come down to how nasalized the first word's vowel is — often hard to hear — and the difference between a (perhaps short and soft) nasal murmur and the residue of a /d/ closure, which may not be well distinguished (if at all) from the following /p/ closure.

I hope it's clear that my point is not to adjudicate between Biden and Leonhardt, or to defend either one, but rather to illustrate the interesting complexities of the real-world relationship between phonological symbols and phonetic signals.

Here's the relevant passage of the SOTU (larger context here):

You know, we're often told
that Democrats and Republicans can't work together
but over the past two years, we've proved the cynics and naysayers wrong.
we disagreed plenty.
And yes, there were times when Democrats went alone,
but time and again,
Democrats and Republicans came together.
Came together to defend a stronger and safer Europe.
We came together to pass one in a generation-
((once in a)) generation infrastructure law,
building bridges connecting our nation and our people.
We came together to pass most significant law ever,
helping victims exposed to toxic burn pits.

And the final "burn pits" phrase:

And here's the waveform and spectrogram of the relevant two words:

The red square outlines the open and voiced part of the /ɚ/ vowel (in "bird" or "burn"); the blue square outlines the /n/ murmur, about 50 msec. long; and the green square outline the /p/ closure, which is about 55 msec. long. As I said at the start of the post, this pronunciation is well within the bounds of "standard" American English variation, though it's arguably under-articulated given how the low contextual likelihood of the phrase for most of his listeners.

In a Postscript to his article, Leonhardt explains why the contextual likelihood of that phrase was greater for Biden:

Postscript: Biden was actually referring to — or meant to refer to — toxic burn pits, the name for bonfires in which the U.S. military incinerated trash while fighting overseas. These fires caused health problems for many troops, and Biden believes they contributed to the cancer that killed his son Beau.

(For a deeper dive into the theoretical issues around allophonic variation, see my 2018 paper "Towards progress in theories of language sound structure".)



  1. Cervantes said,

    February 10, 2023 @ 10:07 am

    Which brings up another issue that interests me. R is actually a vowel, but it isn't allowed to be orthographically in English. (It is in slavic languages, Viz. Kent Hrbek, whose name we have no trouble pronouncing.) The preceding U, I or E are all silent and do not affect the pronunciation. We could spell it brd or brn, but that isn't allowed. Preceding A is pronounced as in father, and o as aw, but what you really have in those cases is a diphthong. R is not allowed to be a vowel orthographically in romance languages either. I wonder what the historical reason is.

  2. Benjamin E. Orsatti said,

    February 10, 2023 @ 12:53 pm

    Cervantes said,
    R is actually a vowel, but it isn't allowed to be orthographically in English. […] I wonder what the historical reason is.

    I'll bet it's the same as the reason for a lot of our present conventions — "'cause Latin don't have it!".

  3. Terry K. said,

    February 10, 2023 @ 1:20 pm

    Cervantes, In which Romance languages would you say R is (sound-wise) a vowel? It's not in Spanish.

    And in English, it would depend on where in word, and the variety of English. At the beginning of a syllable (alone for following a consonant) it's more in the category of w and y. And some varieties of English only have R in those contexts. And then there are variant pronunciations of the R phoneme.

  4. Cervantes said,

    February 10, 2023 @ 3:11 pm

    I wrote: "R is not allowed to be a vowel orthographically in romance languages either. " You overlooked the "not." But in Spanish, it's part of a diphthong, i.e. "ir" is pronounced like English "eer" and "er" is pronounced like English "ayr," "ur" as "oor," though the "oo" can be schwaed out, giving you pretty much "r" as a stand-alone vowel. But it's still essentially a vowel sound in any case. I don't know how Virgil was pronounced in ancient Rome, but we pronounce his name as Vrgil.

  5. Terry K. said,

    February 10, 2023 @ 3:22 pm

    I did not overlook that "not". That's precisely what I was replying to. In the context, saying that it's not orthographically allowed to be a vowel implies that it is or can be a vowel in Romance languages, but isn't allowed to be spelled that way. Otherwise, that point would be irrelevant to what you said previously.

    And I'm not a native Spanish speaker, but your description of Spanish does not fit with any description of it I've ever seen, either in learning Spanish or in linguistic descriptions.

  6. J.W. Brewer said,

    February 10, 2023 @ 3:44 pm

    But no one thought Biden was saying "toxic burr pits"? Which I suppose would be like landfills for burrs* which are toxic. You wouldn't want to fall into one.

    *Or "burs," which is apparently a preferred-by-some alternative spelling for the relevant sense of what I would spell burr, glossed as e.g. "a seed pod with sharp features that stick in fur or clothing."

    [(myl) Good point — but if you look at the spectrogram, you can see a rising F2 towards the end of the vowel, which cues coronal place, and is apparently enough to move the perception towards a syllable-final /d/ or /n/. Also, maybe the extension of voicing 50 msec. into the closure argues perceptually for a (weak) voiced (stop or nasal) consonant.]

  7. Taylor, Philip said,

    February 10, 2023 @ 7:16 pm

    "consider the (often imperceptible) difference between tends and tens, or tents and tense, or tend to and ten to". The first two pairs I completely understand — in my idiolect, "tends" and "tens" are both realised as /tenz
    /. and "tents" and "tense" as /tens/ (the LPD informs me that there is a vestigial /t/ before the /s/, but I neither can hear it nor can I feel my mouth making it). But for me, "tend to" and "ten to" are very different — a clear /dt/ in the first, and a simple /t/ in the second.

  8. amy said,

    February 11, 2023 @ 1:15 am

    "Toxic bird pit" seems like the perfect phrase to describe Twitter.

  9. Brett said,

    February 11, 2023 @ 3:49 pm

    I have to say, Biden's pronunciation sounded a lot more like "toxic bird pits" to me—even though I am quite familiar with the burn pits issue. (My brother attributes many of his unfortunate current medical problems to the burn pit adjacent to his barracks in Iraq.) When I first heard him say it, I wasn't being primed to expect "bird" over "burn"; if anything it would have been the opposite. Yet "bird" was still what my aural parser initially came up with, although I immediately recognized what he was talking about, corrected the word to "burn" in my mind, and forgot all about it until I saw in mentioned here.

  10. John Swindle said,

    February 13, 2023 @ 3:33 am

    I recently bought a manual coffee grinder. Closed captions for the online instructional video describe the business end of the grinder as "the conical bird."

  11. John Swindle said,

    February 13, 2023 @ 4:07 am

    Or rather.”the ceramic bird,” sorry.

  12. Keith said,

    February 16, 2023 @ 4:06 am

    Cervantes, your comment is interesting. I'd never thought of the letter R as representing a vowel. I'm familiar with dltng vwels n wrttn lngg to save space, but never thought of brd and brn as being a possible alternative to bird and burn, with just the letter R to represent the vowel sound.

    For me, in both non-rhotic English and German, I see the R after a vowel as modifying the preceding vowel sound, generally lengthening it and shifting its place in the mouth.

    To go back to why R "isn't allowed to be orthographically [a vowel] in English", I think that it must be because of relics or fossils of older orthography and pronunciation.

    The vowels represented by IR and UR would have been pronounced differently in times past, the older form "bridd" /bridd/ becoming "bird" /bird/ then "bird" /bɜːd/, while "bern" /ˈbɛrn/ became "burn"/bɜːn/; the middle vowel merged, but the spelling retained a different written vowel.

  13. Graeme said,

    February 17, 2023 @ 10:07 am

    Late to this. But Mark, if there really was a comma between ‘most significant law ever’ and the clause about burn pit victims …

RSS feed for comments on this post