Drunkenness at the LSA

« previous post | next post »

One of the papers that caught my eye at the just-complete LSA meeting in Baltimore was Abby Kaplan, "Articulatory reduction in intoxicated speech". Here's the abstract:

Voiceless stops are commonly voiced post-nasally and intervocalically. Such alternations are often attributed to articulatory ‘effort reduction’: a hypothesis that voiced stops are ‘easier’ in these environments. My experiment tests this hypothesis by comparing productions of intoxicated subjects with those of sober subjects, assuming that intoxicated subjects produce more ‘easy’ articulations. Intoxicated subjects did not uniformly increase voicing of post-nasal or intervocalic stops; rather, the range of voicing durations contracted for both types of stops. I conclude that considerations of effort do not straightforwardly predict post-nasal and intervocalic voicing: the traditional effort-based account of these processes must be refined.

Kaplan's LSA presentation is not on line, as far as I can tell, but the same experiment is described in a handout for the Stanford Phonetics and Phonology Workshop, 11/30/2009,  "Lenition as Effort Reduction? Evidence from Intoxicated Speech".

There have been dozens of research publications on intoxicated speech, but most of them have a forensic motivation.  I thought that Kaplan's idea of using intoxication as a proxy for reduced effort was an interesting one,

In fact, there were two presentations on intoxicated speech at this LSA meeting, which is more than I recall having seen at one of these gatherings before. The other one was Thomas C. Purnell, "Motor control of vowel space: Effect on dialect features under intoxication":

Intoxicated speech affects motor control of speaking leading to coarticulatory, precision and timing differences from sober speech. Past research has not accounted for dialect or contemporary vowel changes, nor secondary vowel features such as nasalization and pharyngealization. Because vowels can be style shifted, i.e., they are under some conscious motor planning, this paper investigates how speakers produce vowel qualities while intoxicated and how changes in vowel space interact with what is know about the speaker’s dialect patterns. Four hypotheses were tested by comparing Rainbow passage vowels as spoken by six females recorded prior to intoxication with samples from post-peak intoxication.

Checking this topic on Google Scholar, I stumbled across a discussion from more than a century ago — in Edward Wheeler Scripture, The Elements of Experimental Phonetics, 1902 (which you can download in its entirety from Google Books):

The various hypotheses that have been put forth as explanations of phonetic changes might be directly tested by reproducing the conditions and recording the speech results. Thus the hypothesis that the changes known as Grimm's Law are the results of increasing rapidity of speech might be tested by recording language spoken at different speeds; the finer details that cannot be detected by the ear could be measured in tracings made as described in part I. […]

The experiments may be extended to unusual, abnormal or pathological conditions of the individual. Thus the progressive defects in coordination of muscular action whereby thickness of speech is caused may be carefully observed in the progress of alcoholic or etheric intoxication. The effects of rapidity can be observed at each degree of speed in the slowing increasing rapidity of speech in some cases of mania.

It is quite possible that experimental data may definitely confirm any one of the hypothetical causes assigned, but the well-known facts of mental life make it quite as possible that all these causes and others also are involved. I may add here that arrangements have already been made for making gramophone plates of speech under various conditions of excitement, emotion and fatigue so that the curves can be traced off the machine described in Chapter IV. Plans are also being developed for the systematic preparation and preservation of speech records in phonetic libraries.

It has taken a bit longer than Scripture planned for "the systematic preparation and preservation of speech records in phonetic libraries" to become a reality. But in fact, if you're interested in doing your own study of the phonetics of drunken speech, you can now start with a published "phonetic library", documented in Florian Schiel et al., "ALC — Alcohol Language Corpus", LREC2008.

Here's a bit more on Edward Wheeler Scripture, who was not previously known to me.  He founded the experimental psychology laboratory at Yale University in 1892.  After leaving Yale, according to a 1965 note by Edwin G. Boring in the American Journal of Psychology, he was

Associate in Psychiatry and Director of the Research Laboratory in Neurology at Columbia University, Investigator in the Carnegie Institution of Washington, Lecturer on Experimental Phonetics at the University of Marburg, Nonorary Lecturer on Phonetics at King's College, London, and finally Professor of Experimental Phonetics at the University of Vienna.

Although Scripture died in 1945, this notice did not appear until 1965, because

a failing older man, he dropped out of professional attention and his death received no professional notice, and […] the English Who's Who continued his sketch until the 1959 volume.

Boring ends his note with a comment on the ending of Scripture's 1932 Autobiography, which "is almost entirely concerned with intellectual reminiscences about people, ideas, and research":

"I notice a paucity of personal details in my account. I have forgotten most of them and am not interested in the rest; I do not think the reader would be interested either. In order to be dated and placed I have to state that I was born in 1864 in a village in New Hampshire, U.S.A."

Dated and placed! Almost, he failed to be. Perhaps that was his wish.

Edward Wheeler Scripture's other published works include The New Psychology (1897) and Stuttering and Lisping (1912), both of which are also available for full download from Google Books (classified as "Family and Relationships" and "History" respectively, but still). There are also ten volumes of Studies from the Yale Psychological Laboratory that appeared under his name — the tenth and last is here.

Scripture left the U.S. in 1913, twelve years before the LSA was founded, and as far as I know, he never attended an LSA meeting. The quaint pdf-format index to Language has no entry for him.


  1. sh said,

    January 11, 2010 @ 12:35 pm

    Judy Duchan, Getting Here: A short history of Speech Pathology in America

    [(myl) Thanks! This is an excellent short biography, which fills in many of the gaps in Boring's 1965 note (and corrects it in some ways as well). It was especially interesting to learn about Scripture's conflict with George Trumbull Ladd.

    Scripture was fired from Yale in 1903, in a dispute with the chair of the department, George Trumbull Ladd (see Seashore's description of the difficulties between these men in Murchison's Biography-History of Psychology in Autobiography, Vol 3, 1930). The disagreements originated from the philosophical differences between Ladd, who subscribed to what Scripture called "armchair psychology" and Scripture who led the "new science" movement in which experiments were the only definition of the science. Ladd also was let go (given early retirement) by Yale's president because of the conflict with Scripture.

    I'm ashamed to have missed this link, which is Google's top hit for Edward Wheeler Scripture. I guess that I started with Google Scholar and never really moved on.]

  2. Jarek Weckwerth said,

    January 11, 2010 @ 12:47 pm

    I have only had a brief look at Kaplan's handout, and it surely looks impressive. But my general first impression was that trying to investigate effort from acoustic data only will always be difficult. (Which is also true for some other research into lenition or "effort reduction" that I've seen.) She does ask the question of "what exactly does effortful mean" right on the first page, and that is what really needs an answer. Direct gestural approaches might be more successful in trying to provide one (e.g. I would imagine the first step in "effort reduction" might be gestural mistiming, with all kinds of acoustic results) but of course that requires actually measuring gestures using ultrasound or articulometry or a similar technique… And of course there's aerodynamics…

  3. tom purnell said,

    January 11, 2010 @ 3:05 pm

    The connection to Scripture's observation is nice, but perhaps somewhat problematic. I have to admit that prior to looking at the vowel space of the speakers I would have agreed with Scripture primarily because I anticipated vowel reduction as being roughly equivalent to vowel centering per synchronic and diachronic processes (e.g., final English /e/ being reduced to schwa and then deleting). In fact, previous literature said as much.

    However, at the conferenceI was struck by the similarity in conclusions that Abby and I reached–and this is my flight home interpretation–that alcohol reduction, while very telling about certain facts of language, may not be precisely coextensive with linguistic reduction. What we both found (Abby is free to comment on this herself) was that vowels (in my case) and consonants (in Abby's case) aligned to an acoustic continuum (regression line) and not directionally towards the expected reduced acoustic space. For example, in my analysis speakers' vowels appeared to rotate the front vowel continuum so that for the speaker doing the most rotation, BOT moved forward and BEET moved back. Likewise, the head of the BITE vowel moved forward in the mouth towards the regression line and BAN moved from a peripheral to less peripheral (but not centralized) position. BAT moved up the continuum, not centrally. In other examples, BOT and BOUGHT seemed to keep their distance even when rotated. (The slides can be found on my website.)

    This is not to say that this reshuffling of vowels isn't reduction and should suggest intoxication shouldn't be used to test linguistic generalizations. But, we need to avoid thinking that different speech states are merely the result of turning up or down the volume on a radio. To begin with, in both Abby's study and my study speakers weren't sloppy drunk. Intoxication in my study entailed the descending arm of intoxication in which, as described to me by John Curtin in the UW-Madison psych department, human behavior is different in the ascending and descending arm of intoxication (i.e., with respect to a peak of 0.08).

    What I'm trying to get at here is that in order to properly evaluate Scripture's claim, we need to be more aware of factors related to equivalency between atypical and typical speech states as well as the appropriateness of measurement techniques describing those states.

  4. Sili said,

    January 11, 2010 @ 4:23 pm

    I assume that this is one case where it wasn't hard to get student volunteers for the experimental.

  5. Rubrick said,

    January 11, 2010 @ 7:20 pm

    I'm sorry you didn't make the part below the line a separate post, just so you could title it "Boring Writing About Scripture".

    [(myl) We clearly need to hire you as a headline writer…]

  6. IrrationalPoint said,

    January 11, 2010 @ 7:54 pm

    Maybe I'm just naive (or is it cynical? Sometimes it's hard to tell the difference.), but I'm slightly surprised that people consent to have their drunken talk recorded, and then continue to consent the morning after.

    [(myl) Perhaps you're just making a joke, but (as the linked handout explains) the procedure involved ingesting measured doses of alcohol in a laboratory setting, making the recordings, and then staying in the lab until blood alcohol levels returned below a safe threshold. The consent was of course obtained before the alcohol was ingested.

    It's tempting to imagine that such studies should be like this one, but the reality is hardly scandalous at all, I'm afraid.]

  7. John said,

    January 11, 2010 @ 9:19 pm

    assuming that intoxicated subjects produce more ‘easy’ articulations"

    Is it not perhaps that this assumption is mistaken and not the theory about voiceless stops?

  8. Ben said,

    January 11, 2010 @ 11:41 pm

    Or was the assumption that intoxicated subjects would display an increase in easiness across the board?

  9. Chris said,

    January 11, 2010 @ 11:50 pm

    In an effort to bring some academic gravitas to this thread, I hereby present to you the world's greatest drunken linguist: Foster Brooks.

  10. Abby Kaplan said,

    January 12, 2010 @ 12:11 am

    That was the initial idea, at least in my study – of course, John's objection could very well be right. As Tom Purnell mentioned in his comment, both of our results show that it's not at all simple. What we can confidently say is that drunk speech doesn't match up neatly with the typology of sound change; whether that's because intoxication doesn't induce "easy" productions, or because articulatory ease isn't what's driving these changes, is another matter.

    At any rate, I think it's precisely the difficulty involved in getting reliable data on articulatory effort that means we need to think very carefully about how we incorporate considerations of effort into phonological theory…

  11. IrrationalPoint said,

    January 12, 2010 @ 4:23 am

    MYL: I understood how the consent was obtained, it just surprised me that people would consent to have whatever inane or ridiculous things they say while pissed recorded (it is of course possible that what I think of as the threshold for "pissed" in students is higher than would ever be used in a lab setting). Since it is usually a clause in every consent form that you can withdraw your participation at any time, including after the recording has been completed, I'm slightly surprised it isn't a problem that people go "omg, did I say *that*? I can't have *that* on tape!"


    [(myl) The things that are recorded — at least the things that are discussed in the cited papers — are performances of specified words and phrases, not free conversation. So the only source of embarrassment would be slurred speech. (I don't mean to suggest that investigations of the effects of alcohol or other intoxitants on conversational speech would be unethical or impossible to clear with a human-subjects review board, just that issues would be involved in that case that don't seem to come up here.)]

  12. Adam said,

    January 12, 2010 @ 5:29 am

    I'm surprised (well, pleasantly surprised) that in the USA these days you can get away with an experiment that involves giving people alcohol.

  13. uberVU - social comments said,

    January 12, 2010 @ 4:00 pm

    Social comments and analytics for this post…

    This post was mentioned on Twitter by linglounge: RT @nmashton: Shout-outs to #linguistsgonewild at #lsa2010. RT @languagelog: Drunkenness at the LSA http://bit.ly/6AOVFe

  14. I said,

    January 13, 2010 @ 11:55 am

    I was hoping this would be a video of drunk linguists. Damn.

RSS feed for comments on this post