Understanding dependencies in real timea crosslinguistic investigation of antecedent complexity and dependency length

Abstract

Words in a sentence are dependent on each other for syntactic and semanticinterpretation. The existence of dependencies between non-adjacent words impliesour memory system needs to integrate information across intervening linguisticmaterial by either actively maintaining information in memory and/or retrievingpreviously encoded representations. In this dissertation, I investigate the memorymechanisms involved in the processing of dependencies between non-adjacentwords, exploring both local and unbounded dependencies in typologically diverselanguages ¿English, Spanish and Basque. Specifically, I investigate two factors thataffect the activation level of memory targets and, hence, their subsequent retrievalfrom memory in sentence comprehension: (a) the representational complexity ofantecedents (Chapter 2), and (b) activation decay as a function of time (Chapters 3 & 4).I show that increasing the syntactic complexity of antecedents increases theiractivation level and facilitates their subsequent retrieval from memory. I also showthat activation decay is a major determinant of comprehension difficulty in both localand unbounded dependencies in Spanish, suggesting locality effects are a generalphenomenon in VO languages. Lastly, I find evidence that expectation-drivenfacilitation overrides activation decay in Basque, supporting the idea that all OVlanguages show antilocality effects, regardless of the degree of word order flexibility.In sum, this dissertation broadens the available evidence to further understand themechanisms underlying the comprehension of linguistic dependencies in real time.