Cognitive Limitations Impose Advantageous Constraints on Word Segmentation

In this paper, we examine the implicit learning constraints imposed by cognitive limitations on memory and processing in the context of word segmentation. To understand how incremental processing and memory restrictions improve performance, we analyze the properties of the incremental, constrained learner proposed by Venkataraman (2001) and a batch, ideal learner version of Venkataraman’ s learner that relies on the same probability model, with and without memory limitations. We show that both cognitive limitations can improve performance on the word segmentation task and present computational analyses of the pressures underlying these results.