Page 22 - Fall 2011

ed. In another study, I found that the more attractive female participants rated a male model talker, the more they imitat- ed his vowels (Babel, 2012). Social preferences and liking thus modulate the process of spontaneous phonetic imita- tion. This suggests that the relationship between speech per- ception and production may be somewhat labile in nature. In a study of gender bias in imitation, Namy and colleagues found that women imitated more than men, but that women’s imitative behaviors were focused on a particular male voice used in the experiment (Namy et al., 2002). This indicates there was a particular aspect of this male model’s voice which encouraged females to imitate it more than the other voices used in the task. It is clear that social factors play a role in phonetic imita- tion. Language-specific internal factors are also involved in the process of imitating speech, just as lexical frequency played a role in Ms. Winfrey’s variable pronunciation of /ɑ͡􏰀/ or /a/. Using a modified version of an auditory naming task, Nielsen (2011) presented listeners with a block of model pro- ductions that had been digitally modified such that the aspi- ratory puff of air that accompanies the /ph/ sound in pint or pound was longer than it typically is in natural speech pro- duction. (Note that such a puff does not occur in /b/ initial words like beer or baseball.) Nielsen found exposure to these modified words caused participants to not only increase the duration of the aspiration of /ph/, but also to generalize this increase in aspiration to /kh/ initial words like canoe and kite. This indicates that imitation can be abstracted and general- ized across one’s linguistic system. Measuring imitation How do researchers determine whether imitation took place? The complexity of the speech signal makes the method of measuring imitation an important topic. There are two primary ways to gauge or measure phonetic imitation— acoustic or perceptual—each with its positive points and drawbacks. The choice between them is guided by the goals of the study. Let’s say a researcher would simply like to demonstrate that the speech signal was imitated in some way. The most common way to accomplish this is to have naive listeners rate perceptual similarity using an AXB task, in which listeners are presented with three tokens in a trial and are asked to judge whether the A or B token is more similar to the X token. The X token is a speech sample from Talker 1, the model talker who was presumably imitated. The A and B tokens would be speech samples from Talker 2; one token would be a “baseline” sample recorded before exposure to or interaction with Talker 1, and the other token would be a sample recorded during or after exposure to or interaction with Talker 1. When listeners consistently choose the post- exposure token as more similar-sounding to the X produc- tion of the same word, there is evidence for phonetic imita- tion. There are two primary virtues to using an AXB similar- ity task to assess phonetic imitation: (1) it is a holistic meas- ure that allows for imitation of any part of the acoustic signal to contribute to listener judgments of perceived similarity, and (2) if imitative behaviors serve as a seed to sound change, as has been argued (Garrett and Johnson, in press), then these behaviors must be perceptible to listeners. Oddly enough, this issue of sound change is critical in choosing to measure imitation acoustically as well. It has been argued that phonetic imitation is the means by which sound changes spread through communities. Sound change tends to affect particular aspects of the speech signal. Let’s return to the earlier example of the pronunciation of /u/. It was noted above that the pronunciation of this vowel in words like dude varies as a function of region. In addition to this regional variation, there is a sound change in progress with this vowel across most varieties of North American English. The sound change is not random: the second reso- nant frequency of this vowel is becoming higher due to talk- ers adopting a more fronted tongue position. Acquiring spe- cific acoustic evidence about what was imitated has the potential to address the issue of whether phonetic imitation is the seed by which sound change is spread. In imitating words with /u/, is this increase in the second resonant fre- quency one of the acoustic features that listeners imitate? If imitation studies demonstrate that what is imitated corre- sponds to the acoustic phonetic details involved in sound change, then researchers have promising evidence for how sound changes might spread through speech communities. While specific acoustic measures of imitation offer valuable insight into what is imitated, perceptual measures of imita- tion attenuate a potential experimenter bias: in acoustic measures of imitation, the researcher must predict which TUNE INTO ZERO’s SOUND SOLUTIONS ZERO is a world-wide leader in high-performance acoustical control for doors, windows and walls. Nobody does sound control better — we use advanced technology and testing to master the challenges of creating an effective barrier and preventing gaps in that barrier for the life of the assembly. Our systems are rated for use in sound studios and recording facilities, music halls, etc — up to 55 STC. Let us help you close the door on noise — contact us for a copy of our 20 page Sound Control brochure, and our 72 page Product Catalog, or visit our website. 1-800-635-5335 / 718-585-3230 FAX 718-292-2243 zero@zerointernational.com www.zerointernational.com Imitation in Speech 21