Context and topic marking affect distinct processes during discourse comprehension in Japanese

In languages like English or German, definite and indefinite markers serve to distinguish given/old from new information in the discourse model. Japanese, in contrast, lacks definiteness markers, but has a topic marker. The present paper examines how the information status of a noun phrase (NP) in Japanese is represented and integrated into the discourse model. An ERP experiment investigated the processing of topic-marked and non-topic-marked NPs following three different context sentences (making available a given, inferred, or new reading). The results revealed an increase in the N400 as a function of contextual cueing, i.e. the less accessible a referential expression is in the discourse model (i.e. given, inferred, new referent), the more pronounced is the amplitude of the N400. In addition, a late positivity was observed for topic shift, i.e. when topic-marked NPs occurred after contexts in which they were not already the established topic. Crucially, topic shift was facilitated by a particular semantic relation. Since acceptability ratings cannot account for the observed pattern, the positivity is considered to reflect costs from the establishment of a new discourse topic. From a cross-linguistic perspective, the data reveal that contextual cues universally influence discourse integration (N400), while processes subserving the updating of discourse structure (e.g., topic shift) differ across languages (late positivity): Discourse updating is guided by the given-new distinction in German and by topicality in Japanese (at least in canonical word order).