Gardner, B. , Sullivan, A. , Trine, T. , Jaeger, . T. F. & Kurumada, C.
Department of Brain and Cognitive Sciences, University of Rochester
Recent studies show that listeners cope with variation in pragmatic language use by adapting their expectations; when the speaker is introduced to be abnormal and incapable of informative language use, interpretation of a scalar adjective as indicating contrastive inference (e.g., Pass me the tall glass) is suppressed (Grodner & Sedivy, 2011). Yet, this mechanism is of little use if listeners cannot adapt from a limited amount of input; Grodner and Sedivy (2011) used more than 150 instructions, making it unclear how quickly listeners modulate their comprehension. We present a new visual-world experiment to test if listeners adapt their online inferences rapidly.
[Design] 40 participants were randomly assigned to Reliable-Speaker or Unreliable-Speaker condition. The Unreliable-Speaker 1) was introduced as a patient of linguistic and social impairment; 2) mislabels objects; and 3) frequently over-modifies, whereas the Reliable-Speaker modifies appropriately. Participants first watched a video where the speaker produced 3 example sentences, and then completed 16 target and 32 filler trials. [Results] Anticipatory eye-movements based on the scalar adjectives dissipated in the Unreliable-Speaker condition. A pilot study (N=40) without a video instruction yielded null results, suggesting that the embodied speaker presence is a key to the rapid speaker-based adaptation of online pragmatic comprehension.