Taking Talkers for Granted: Individual and Group Effects in the Recognition of and Memory for Spoken Words

The endeavor to understand how humans understand spoken language is faced with the biggest confound of all: the speech signal. Inherently, speech conveys information about sounds and words and information about talkers. Research for decades focused on the perception and recognition of spoken words with no hint of the role of social meaning in these processes.  Even when research investigated accented speech, for example, it was under the heading of "adverse listening conditions", where the questions focused more on how listeners grapple with variation, than on the mapping of phonetic variation that couple linguistic and social meaning. Over the past decade or so, research aimed at understanding how listeners perceive, recognize, and remember spoken words has embraced the notion of a speech as a multi-faceted information source.  Yet, we are still far removed from understanding the complex patterns exhibited across studies, and understanding what information is used when, and how it is used by listeners.   To illustrate this point, I draw on work looking at talker-specificity effects and accent effects.  On the one hand, we have a theory that suggests that subtle talker differences impact lexical access, spoken word recognition, and memory (Nygaard et al., 1994; Johnson, 2006, Sumner, 2013) and on the other, we have evidence of group-effects that occur independent of any particular talker (Sumner & Samuel, 2009; Sumner & Kataoka, 2013).  Over the past few years, in our lab, I have noticed again and again, that these effects are highly malleable, and depend a great deal on context.  Behavior in perception, recognition, and memory studies fluctuates for any given talker depending on other voices that are present in the study.  This observation leads me to question whether information that cues talker vs. group effects is used differentially by listeners, depending on the context (micro-social differences between talkers that share macro-social category membership vs. macro-social differences between talkers that differ in macro-social category membership).  In this talk, I pursue this line of thinking.  First, I provide an overview of number of studies investigating the recognition and memory of spoken words across accents, and show that regardless of talker (swapping one voice out for another), group effects arise.  I then review some research illuminating talker-specific effects.  Finally, I present some work designed to investigate talker vs. group effects, and suggest that the relevance of a particular social contrast between voices best accounts for the behavioral differences we find across the literature.