ACT-R as a Unified Architecture of Cognition: A Symposium in Honor of

This talk presents a theory and ACT-R-based model of human sentence comprehension that embodies the following claim: Sentence comprehension consists of a series of cuebased retrievals from short-term (and long-term) memory, subject to similarity-based interference and activation decay. ACT-R does not merely serve as an implementation language for the theory; rather it serves as the vehicle for bringing sentence processing into detailed contact with general and independently established principles of memory and skilled behavior. These principles, together with some representational assumptions from linguistic theory, provide explanatory accounts of many parsing phenomena (such as difficulty on embeddings and locality and recency effects), and generate novel predictions which can be empirically tested. Experimental work on several languages is summarized. The work includes experiments that examine the effects of interference and decay on distinct components of sentence processing, experiments that distinguish activation decay and distance-based accounts, and experiments that begin to answer one of the basic questions motivated by the theoretical framework: exactly what kinds of similarity (syntactic, semantic, phonological, positional) matters in sentence processing? This is joint work with Shravan Vasishth, Julie Van Dyke, and JJ Nakayama. No linguistic or psycholinguistic background will be assumed.