Distinguishing theories of syntactic expectation cost in sentence comprehension: evidence from Japanese

Abstract Previous research in the sentence comprehension literature has established that people expend resources keeping track of partially processed phrase structures during the process of comprehending sentences. An open question in this literature has been what units of syntactic expectation cost the human parser utilizes. Two viable options from the literature are (1) incomplete syntactic dependencies; and (2) predicted syntactic heads. This article provides a self-paced reading experiment from Japanese — a head-final language — that tests the incomplete dependency hypothesis. The materials in the current experiment manipulate the number of dependents of an upcoming verb, by manipulating (1) the presence/absence of a locative postpositional phrase modifier of the verb and (2) the presence/absence of a dative argument of the verb. The results failed to show any support for the incomplete dependency hypothesis, but were completely consistent with the predictions of the predicted head hypothesis. Taken with the results from the literature, these results offer support for the predicted head hypothesis.