On the Persistence of Structural Priming: Mechanisms of Decay and Influence of Word-Forms

When people speak, they have to choose a syntactic structure for their utterances. Research shows that this choice is not made independently for each utterance, and that it is influenced by properties of recent utterances (Bock, 1986). But the learning mechanisms responsible for this influence, and its duration are not completely clear. Pickering and Branigan (1998) suggest that structural repetition is due to trailing-activation in the language production system. However, this account lacks a formal implementation. Chang, Dell, and Bock (2006) have developed a computational model that tries to explain structural priming based on error-based learning. In this paper we report a formal model that relies on an unsupervised learning mechanism and use this to replicate behavioral data for both structural priming and lexical influence on this priming. We use this model to investigate the factors that lead to these priming effects and its rate of decay.