Calendar Detail

Speaker:
Meghan Sumner

This event is part of an event series „In_equality Colloquium“.

This talk builds upon four facts: Spoken language is highly variable. Our system for understanding spoken language is complex, redundant, and dynamic. As listeners we understand spoken language quickly and adeptly, somehow encoding our linguistic experiences to memory. And speech inherently conveys a linguistic (what we say) and a social message (how we sound when we say it). This results in a massive amount of variation in speech. Each talker has their own natural range of variation within and across social styles. Productions from other similar talkers expands this range. And productions from more diverse talkers, crossing macro- and micro-social categories, expands the range again. In this talk, I suggest that only with a socially-modulated approach can we fully understand spoken language comprehension. From such an approach, a path toward a fuller understanding of implicit biases that are pervasive in society becomes clear.

Once we understand talker variation in spoken language, I provide an overview of some listener behaviors to this variation to highlight the interrelated nature of three components of this process: Recognition, Cognitive Resource Allocation, and Memory. First, I highlight strikingly similar listener behaviors for a range of talker-related variation (e.g., informal and formal speech styles, accents, emotions) across some standard paradigms in spoken word recognition. I show that independent of variation type, listeners are highly sensitive to within- and across-talker variation and use this information fast and early when recognizing spoken words. Second, I show that these behavioral patterns can be understood if we think about recognition, resource allocation, and memory encoding as social, interrelated, dynamic and context dependent. I illustrate this point with some data from the visual world paradigm and I then work through the logical connections to data on asymmetrical memory encoding. Finally, I show the downstream effects these patterns have on the representational structure of voice/group-associated stereotypes for different talker populations.

Meghan Sumner is an Associate Professor of Phonetics at Stanford. Her work considers sound patterns that exist in languages and their associated usage patterns (who says what, how and when), and investigates the social meaning humans associate with these patterns (and how they come to make these associations). She addresses how, cognitively, this social information affects attention, perception, recognition, memory, and comprehension. With that understanding, she approaches areas in which language and society interact addressing how stereotype and bias are reinforced through spoken language.

Link for online participation