Abstract

Both ambiguity and `locality’, the distance between syntactically-related constituents, have been shown to contribute to on-line sentence processing complexity. We’ll describe a parsing model for Combinatory Categorial Grammar which integrates them into a single-cost metric. The model makes more fine-grained predictions about both the step-by-step, incremental complexity of sentence processing and also leads to a more predictive account of the role of extrasyntactic information—such as lexical frequency, semantic and pragmatic plausibility, and prosody—in resolving ambiguity incrementally.

We’ll review the experimental psycholinguistic evidence in favour of the model, but mostly focus, in the spirit of Hawkins (1994), on the predictions the model makes concerning the organisation and usage of language, and on the evolutionary trajectories of languages. We’ll argue that: 1) it is supported by corpus-based evidence about usage and the distribution of prosodic boundaries; 2) the model of ambiguity resolution is optimal from the perspective of an evolutionary model of language development and change.