Arai, M. 1 , Nakamura, C. 2, 3 & Hirose, Y. 3
1 Seijo University
2 Massachusetts Institute of Technology
3 The University of Tokyo
Numerous studies demonstrated that comprehenders predict upcoming argument information (e.g., Altmann & Kamide, 1999). Much less known is whether language comprehenders predict a specific verb based on input received earlier in a sentence. Such a finding would provide support for the theory of incremental language processing that estimates predictability of a specific word based on the change in probability distributions, uniformly applicable to the verb, arguments, and adjuncts (Hale, 2001; Levy, 2008). We conducted two experiments using a novel priming method. Participants first read either a head-final Japanese sentence that encouraged prediction of a specific verb or one that did not. The prediction-encouraging sentence ended with a less expected, though plausible, alternative verb (i.e., the predicted verb never appeared). In a separate following trial, participants saw either an unrelated verb (control) or the previously predicted verb (Experiment 1), or a synonymous verb (Experiment 2) and performed lexical decision. The results showed faster response times to the predicted verb compared to control following prediction-encouraging sentences but no difference following non-encouraging sentences. No such interaction was observed with synonymous verbs, demonstrating that the effect is lexically-specific in nature. Our results provide evidence for the prediction of a specific verb during sentence comprehension.