UC San Diego

This dissertation explores the processing of syntactic (1) and anaphoric gap-filler dependencies (2) in Korean relative and adjunct clauses ('because' clauses), using event-related brain potentials (ERPs), self-paced reading times, and eye-tracking. 1) [RC _i senator-ACC attacked- REL] reporter-NOM error-ACC admitted. 'The reporter who attacked the senator admitted the error. (2) [BECAUSE _ i senator-ACC attacked-BECAUSE] reporter-NOM error-ACC admitted. 'Because hei attacked the senator, the reporteri admitted the error.' Following a discussion of different syntactic analyses of Korean relative clauses (especially wh-movement and unconditional binding of null arguments) in Chapter 2 and an overview of previous experimental studies (Chapter 3), Chapter 4 investigates the processing of syntactic dependencies using Korean subject (SRs) and object relative clauses (ORs). The results show a processing advantage of SRs over ORs, supporting processing models based on the accessibility hierarchy (Keenan & Comrie, 1977) and the phrase-structural distance hypothesis (O'Grady, 1997). Furthermore, the ERP results show that at the head noun, where the gap and filler are associated, ORs elicit an anterior negativity compared to SRs, similar to the pattern found in English relatives. This effect is attributed to higher working memory load for gap-filler association in ORs. Chapter 5 investigates the processing of backward anaphoric and syntactic long- distance dependencies; the goal is to investigate possible similarities and differences in underlying cognitive/ neural processes. The results show that the processing of anaphoric dependencies is also sensitive to grammatical constraints defined in terms of the accessibility hierarchy or hierarchical structural distance. In the ERP experiment, both syntactic and anaphoric object dependencies elicit a LAN effect compared to control sentences at the matrix subject position. This is taken to suggest that in both types of dependencies, gap-filler association is immediate. However, compared to anaphoric dependencies, syntactic dependencies elicited a larger LAN effect. At the next word, relative to syntactic dependencies, anaphoric dependencies elicited a sustained LAN effect that continued through the end of the clause. The different time-course of these (left) anterior negativities was taken to suggest different gap-filler association requirements for syntactic versus anaphoric dependencies. These results are discussed in light of the syntactic analyses in Chapter 2