Page 21 - Fall 2011

(1996) took speech samples from Mr. King and twenty-five of his guests, measuring the long-term average spectra (LTAS) in a low frequency band-pass filter from each speech sample. Using these LTAS measures, correlation coefficients of the actual conversations were compared to those of pseudo-con- versation pairings. Actual conversations had significantly higher correlation coefficients, suggesting that interacting talkers accommodated their spectral patterns. The researchers used a measure of LTAS variability and estab- lished that Mr. King’s interviewees fell into a dominating, low deference group and a high deference group. Mr. King’s LTAS measures indicated he took a deferent stance toward the dominating group and a more dominating position in inter- views with those in the high deference group. Influential politicians of the time, like former President George H. W. Bush and former President Bill Clinton, fell into the domi- nating group, triggering a more deferring response from Mr. King. On the other hand, former Vice President Dan Quayle was a member of the high deference group who accommo- dated more to the speech patterns of Mr. King. (These inter- views were taken from broadcasts from April 1992 through July 1993; it may help to keep the social and political context of that era in mind.) Undergraduate students completed sur- veys to rate the social status of the interviewees and to evalu- ate how perceived social status affected Mr. King’s accom- modative behavior toward the interviewees. These subjective student-elicited measures echoed Mr. King’s patterns of pho- netic accommodation: for example, former President Clinton was at the top of the social status ranking, while former Vice President Quayle was at the bottom. Simply, the researchers found that Mr. King accommodated to the speech patterns of his guests who were of higher social status, while the lower status guests accommodated low frequency spectral charac- teristics of their speech toward those of Mr. King. Phonetic imitation in the speech laboratory Talk show hosts are, of course, not the only ones who imitate and accommodate to interlocutors during spoken language interaction. In recent years, phonetic imitation has been a hot topic in laboratory-based studies of speech. These studies often take one of two forms: a task guised as a sort of guided spontaneous conversation or an auditory naming task. Let’s discuss these in turn, starting with the guided spontaneous conversations. These are often map tasks or spot-the-difference tasks, typically involving two partici- pants. They are guided in the sense that they are centered around cooperative activities dictated by the experimenter, but are spontaneous in that the detail of the conversation is freely determined by the natural interaction. In one recent and influential study on phonetic convergence, Pardo (2006) examined phonetic convergence in same-gender dyads involved in jointly completing a map task, where one mem- ber of the dyad was the giver of map directions and one was the receiver whose task it was to navigate the described path. Dyads were found to have converged on 62% of the experi- ment trials. Female dyads were found to converge toward the speaker who was receiving instructions, whereas male dyads patterned oppositely; they converged toward the speech of the talker giving instructions. Pardo concluded that particu- lar social factors dependent on the situational context of a conversation—factors such as gender and the power dynam- ic of a giving-receiving interaction—determine the direction of phonetic accommodation. Another recent study examined convergence between dyads completing a spot-the-difference task (Kim et al., 2011). The conversational dyads in this study were pairs of native English speakers and native Korean speakers who either did or did not speak the same dialect, and dyads of native and non-native speakers of English. More accommodation was found in the same-dialect dyads than in the different-dialect or the cross-language pairs. Kim and colleagues concluded the process of convergence is facilitated when members of a dyad share a language background, indi- cating that convergence is easier when the target of the con- vergent behavior is within an individual’s pre-existing pho- netic repertoire. The second design frequently employed in the literature is an auditory naming task. An auditory naming task consists of a listener hearing a model talker produce a word over headphones or loudspeakers, and the listener’s task is to iden- tify the word by saying it out loud: that is, to name the audi- tory object. While this method eliminates the natural social context for imitative behaviors to emerge, it offers a more controlled environment for speech researchers to query par- ticular aspects of what might facilitate or inhibit the imitation process. Using this methodology, Goldinger (1998) estab- lished that less common words are imitated more than words that are used more frequently. This finding suggests that pho- netic imitation may play a role in how we learn about our native languages. For example, the number of times you have heard the word potato uttered around you is likely many times more than the number of times you have heard kohlra- bi. Exposure to variation in how potato is pronounced is unlikely to sway your production of the word: in a sense, you confidently know how to say potato. Hearing a slightly differ- ent pronunciation of the word kohlrabi, on the other hand, may cause you to cast your own pronunciation of the word into doubt. On some level, this is an inaccurate way to describe the role of lexical frequency in phonetic imitation; an auditory naming task is too fast-paced to allow for per- sonal reflection on pronunciation insecurities. However, if you have amassed fewer experiences with a particular word, you have fewer memories about how to pronounce it, making a single new exposure to kohlrabi all the more prominent a perceptual experience. Other research using an auditory naming paradigm has explored the interaction of social and phonetic factors in speech imitation. In my own work, I have examined how social biases and preferences moderate phonetic imitation. In an auditory naming task where New Zealanders were pre- sented with an Australian model talker, implicit social biases in New Zealanders’ positive or negative views about Australia were measured using an Implicit Association Task, a stan- dard social psychology tool (Babel, 2010). A strong relation- ship was found: while overall New Zealanders imitated the Australian model, the more positive the New Zealanders’ implicit social biases toward Australia, the more they imitat- 20 Acoustics Today, October 2011