“Who”: Perception of talker properties

Abstract

Spoken language simultaneously provides listeners with linguistic content and information about the person speaking. Listeners can identify specific individuals from auditory-only speech signals. Furthermore, listeners can identify a range of physical and socio-cultural talker characteristics (indexical characteristics) with above chance accuracy. Listeners are sensitive to relatively stable aspects of a talker's identity, such as their age, height, gender, race/ethnicity, sexual orientation, region of origin, native, and language status. Talkers also transmit more transitory information (e.g., current health status or emotional state). The mental representations that listeners build for individual talkers and speaker groups interact with the processing of linguistic information, such that the same physical acoustic signal can be interpreted in different ways depending on the talker's perceived indexical characteristics. In this talk, I will focus on how these mental models can allow for more efficient and accurate processing of speech signals. For example, listeners are more accurate at identifying words from familiar than unfamiliar talkers. Likewise, listeners more accurately understand speech from non-native speakers after a period of exposure. Research on the “who” of speech perception demonstrates the need for considering the interaction between talker characteristics and listener knowledge when making decisions about speech privacy.