Poster A69, Saturday, March 25, 5:00 – 7:00 pm, Pacific Concourse
Speaker-specific predictions about category membership during language comprehension
Rachel Ryskin1,2, Shukhan Ng3, Katie Mimnaugh3, Sarah Brown-Schmidt4, Kara D. Federmeier3,5; 1Massachusetts Institute of Technology, 2Boston University, 3University of Illinois at Urbana-Champaign, 4Vanderbilt University, 5Beckman Institute for Advanced Science and Technology
During language comprehension, listeners predict features of upcoming words. Previous work (Federmeier et al., 2010) examined predictive processes when readers judge the fit of a word with a category (“A kind of tree”). Atypical exemplars (ash) elicit a larger frontal positivity compared to typical exemplars (oak) and anomalous words (tin). This frontal positivity may index processes associated with a prediction being disconfirmed (Federmeier et al., 2007). An open question is whether such predictions are context-dependent. Here, we first successfully replicated Federmeier et al. (2010) in the auditory modality. In a second experiment, we extended the design to test the context-specificity of predictions using two-speaker contexts (Bob and Susan). Speakers alternated providing category cues (e.g., Susan: “Bob, name a kind of tree.”) and answering (e.g., Bob: “oak”). Critically, participants were provided with advance information about one of the speakers (e.g., Bob; counterbalanced across subjects) through short interviews that revealed that Bob had a strong preference for, e.g., ash trees. We hypothesized participants might make speaker-specific predictions, expecting “oak” if Susan was being asked about trees but “ash” if the person being asked was Bob. Indeed, we observed a significantly increased frontal positivity in response to a typical word when it was said by Bob compared to Susan. Interestingly, no speaker-differences were observed in the N400 window. These results suggest that participants made speaker-specific predictions about upcoming words and that these predictions had consequences for the integration of new information into the existing representation, but not for the initial semantic processing.
Topic Area: LANGUAGE: Semantic