Benjamin Munson - munso005@umn.edu
University of Minnesota
164 Pillsbury Dr., Minneapolis, MN 55455
Kathleen Currie Hall, E. Allyn Smith
Ohio State University
Columbus, OH 43210
Popular version of paper 2aSC48
Presented Tuesday morning, May 19, 2009
157th ASA Meeting, Portland, OR
We use language to ask and answer questions, share thoughts and ideas, and relay information. But this is not the only function of language: just as teenagers use fashion to shape their identities, so too do we use language to reveal personal characteristics of ourselves and our perceptions of ourselves. There are many ways we can use language to project identity, including the use of particular words (such as "wicked" as in it was wicked cool), particular grammatical forms (such as the double negation in I haven't seen nobody), or particular pronunciations of a given sound (such as using a [w] sound instead of an [r] sound in the phrase Rascally Rabbit). This presentation focuses on this last type. We looked at the acoustics of people's productions and the perceptions by others of those acoustic differences to see what kind of social identity traits they might indicate.
Listen to the following two sound files: LISTEN A, LISTEN B.
There are multiple kinds of meaning at play in the sentence you hear. One is what we might call the asserted or proffered meaning; for the sentence "if the drywall begins to crack, call me," the asserted meaning is that you (the hearer) should call the speaker if your drywall begins to crack. The second type of meaning is what we'll call the social meaning, which, like the examples given above, indicates something about the identity of the speaker. In the case of sentence (a) v. (b), the way the // in crack is pronounced, which is the only difference between the two, potentially tells you something about the speaker who produced the message. Rather than thinking that the variation in the pronunciation of // could inhibit a listener from understanding that the word being spoken was crack in both cases, as researchers we find that hearers often use those pronunciation differences as cues to different characteristics of the talkers who produced them.
We can distinguish these different pronunciations of // from a scientific point of view. One is called 'tense' //. This pronunciation is characteristic of some North American dialects of English, including the 'Northern Cities' dialects, spoken in cities around the Great Lakes like Milwaukee, Chicago, Detroit, Cleveland, and Buffalo, New York (LISTEN). It sounds sort of like an "ee" vowel, followed by an // vowel, followed by an "uh" vowel. The other variant is produced with a more open, more-retracted jaw and more-steady vowel quality, which we call 'retracted' // (LISTEN). It sounds almost halfway between the // vowel (like in crack) and the /a/ vowel (like in crock). It is characteristic of some other North American dialects of English, including that spoken in Northern California. Acoustic records (called spectrograms) of these two types of vowels in the word bad are shown below in Figures 1 (retracted) and Figure 2 (tense). The use of tense versus retracted // does not depend solely on regional dialect, however; there are substantial differences within dialects, as well. Our primary question at this point is which different groups of speakers within a dialect produce different variants of //, and whether listeners in these dialects can use these acoustic characteristics to infer identity characteristics about speakers. Previous studies have shown that listeners infer characteristics about talkers from short stretches of speech, even in single vowels. For example, two other researchers, Bartlomiej Plichta and Dennis Preston, found that listeners infer a talker's regional American dialect from whether they say a 'drawled' vowel in words like 'pie' (as in its pronunciation in the following sentence). (LISTEN)
It has been shown that listeners associate variation in // with talkers' sexual orientation. In a previous study, the first author of the current paper (Munson) and his colleagues compared the speech of a group of self-identified gay men to a group of age-, dialect-, and education-matched self-identified straight men in Minnesota. The original focus of this research was how listener's beliefs about a talker's sexual orientation influenced how they perceived their speech. As part of this, Munson et al. collected baseline data on how talkers convey sexual orientation in the dialect region they examined, and how Minnesotan listeners perceive sexual orientation through phonetic variation. They found that gay men were more likely to produce an /æ/ vowel whose acoustic characteristics were qualitatively similar to retracted /æ/ variant than were heterosexual men. There was some overlap between the groups, showing that the // variant a Minnesotan man produces isn't always an indicator of his sexual orientation. In a perception experiment, nave listeners identified the retracted // variants as more gay-sounding than the tense // variants, which is exactly what is expected: in both production and perception, there was an association between the retracted variant and gayness. This wasn't the only difference that they found between gay and straight men. Differences were also found in the vowel /u/ (as in the word shoe) and /o/ (as in the word show), and in the /s/ sound. As a result of this work, one could hypothesize that // pronunciation encodes a particular social meaning in Minnesota and perhaps elsewhere, but the test was only conducted in one geographic region: retracted // 'means' gay and tense // 'means' straight.
The current study is a follow-up on this finding. After establishing the different pronunciations of // as having social meaning, we wanted to compare this social meaning to other kinds of meanings in natural language. For example, in the sentence "I don't want to mow the gosh darn lawn," the primary, asserted meaning is that the speaker does not want to mow the lawn, but there is a second kind of meaning indicated by the use of the epithet gosh darn that tells us that the speaker is irritated, in a bad mood, or something similar. This is not necessarily a (semi-)permanent characteristic of the speaker's identity, but it nevertheless tells us something about the feelings of the speaker, and so we wondered whether this kind of expressive meaning was similar to social meanings like //. To this end, we formed the current research team with a unique combination of skills and areas of expertise in semantics (meaning), sociolinguistics, acoustics, and the representation of sound structures in mind, so that we are in a position to look at the fine details of this kind of meaning.
The study that we are presenting at this ASA meeting is just one small piece of this project. In this presentation, we are trying to resolve an apparent contradiction in our findings. The results of a new study that we did in 2008 appeared to contradict the original finding that we published. It showed the opposite relationship between tense and retracted // and judgments of sexual orientation; that is, tense // was associated with gay speech and retracted // with straight speech. The new experiment differed from the earlier one in that it used trained speakers producing prototypical retracted and tense // tokens. The acoustic analysis that we did for the current study shows that the apparent paradox really isn't a paradox at all: the acoustic characteristics of the intentionally retracted and tense // tokens were quantitatively very different from the natural variation in the // tokens that were produced by untrained gay and straight talkers in the earlier experiments. When we do statistical analyses relating the measures of perceived sexual orientation to acoustic measures, we see that there is really much more consistency between the two experiments than we originally thought. The only difference is that the fine phonetic detail in the trained speakers' productions differs from that in the natural tokens that were stimuli in the earlier studies.
This finding provides an important lesson: impressionistic labels like "tense //" and "retracted //" do not predict listeners' behaviors as well as detailed acoustic analyses. More importantly, our work is an illustration of how much more an interdisciplinary team can learn about a problem than can individual scholars working in isolation.
 |
| Figure 2. Spectrogram of a tense vowel in the word bad. |
