How do social characteristics of a speaker influence how listeners process their speech? There is evidence that social characteristics, like a speaker’s age, gender, and so forth, can shift how listeners respond to their speech. For example, ambiguous sounding words are recognized quicker and more accurately when matched with pictures of speakers who are likely to say those words, and changing a visual cue about a speaker while keeping the audio constant can change listeners’ judgments of what they heard. An unanswered question is whether social information is directly affecting perception, or if it is only affecting later decision-making. Addressing this question will contribute to our understanding of the role of social information during speech perception and will further develop our models of human language processing.
Previous work has demonstrated that social cues can directly shift categorization behavior, but the paradigms used in the majority of this work do not allow conclusions to be drawn about the specific time-course of social and acoustic cue integration. The specific time-course of this process is critical to debates over the nature of linguistic representation (episodic vs. abstract) and processing (feed-forward vs. interactive models). This dissertation investigates how listeners use social cues during speech perception in real-time by measuring the cues’ influence on the earliest stages of speech perception. In what follows, I ask whether social cues can induce selective adaptation effects in perception and whether social cues can influence on-line perception as detected via eye-tracking during a compensation for coarticulation task.
For the selective adaptation experiments, I replicate previous work and show less “SH” categorizations when listeners are repeatedly exposed to clear exemplars of /S/. While we do observe evidence for potential subtle influences of speaker gender on the magnitude of the selective adaptive effect, we do not observe evidence for our critical prediction: we do not observe an effect of speaker gender guise on the direction of selective adaptation. That is, the selective behavior of an ambiguous [s]-[S] is not shifted by the perceived gender of a speaker. Additionally, we observe no evidence for the role of visual face gender cues or acoustic cues to speaker sexuality on the selective adaptation behavior of listeners.
Turning to the compensation for coarticulation eye-tracking experiment, I again replicate the classical effect. For both asta ashka and alda-arga stimuli, we observe perceptual compensation for the specific acoustics of C1 in listeners’ C2 categorizations. Additionally, this effect emerges gradually as the stimuli unfold, with observable gradient differences between listeners’ fixation data developing even before the stimulus has been completed. We do not, however, observe evidence for our core prediction of speaker gender influencing the compensation behavior of the ambiguous C1 step on later C2 categorization.
Taken as a whole, these results indicate that the role of sociophonetic cues in perception may be restricted to later decision stages, rather than exerting their influence during earlier perceptual stages. Though further investigations are required for a more robust conclusion, the experiments detailed in this dissertation present critical evidence for the precise role of these sociophonetic cues in our understanding of speech perception.
