Functional Linking Between Negative and Positive ERPs for Syntactic Processing in Japanese: Mutual Enhancement, Syntactic Prediction, and Working Memory Constraints

摘要

This study attempts to detect the differences in event-related potentials (ERPs) associated with two syntactic processes: the syntactic integration of discontinuous dependency and the detection of a violation of the syntactic island constraint. We recorded the electroencephalogram elicited by complex sentences in Japanese that included a dependency between a quantifier and its head noun, in which we changed the word order of the two words to manipulate the presence and absence of a syntactic integration and a syntactic island violation while keeping the lexical items and construction unchanged. We found significant negative and positive deflections for the syntactic integration only when a quantifier preceded its head noun. We also observed significant negative and positive deflections for the syntactic island violation, for which the negativity was more salient when a quantifier preceded its head noun. This study is the first to report a late positive ERP for a violation of the syntactic island constraints in Japanese, and the results showed that the ERP elicited by syntactic integration and that by syntactic island violation were different in terms of their latency, topography, and duration. More importantly, the ERPs elicited by the two syntactic processes were biphasic, and the amplitudes of the negative ERP and of the positive ERP were positively correlated. This positive correlation could be a characteristic of syntactic processing because it contrasted with the negative correlation reported for the ERP elicited by semantic anomalies in English. Furthermore, the amplitude of the ERP for syntactic integration was negatively correlated with the individual capacity of working memory (WM). That is, a reader with greater WM capacity showed smaller negativity and positivity for the syntactic integration, whereas the amplitude for the syntactic island violation showed no significant correlation with the individual capacity of WM. Our results suggested that linguistic ERPs functionally interacted with each other and that the ERP involving the retention and the retrieval of a distant word could be constrained by the individual differences in WM capacity. We discuss the possible reasons for the contrast between English and Japanese on the basis of the cross-linguistic differences in the two languages.