THE LIMITS OF COOCCURRENCE - TOOLS AND THEORIES IN LANGUAGE RESEARCH

The development of new tools such as those described in the articles o f this special issue marks an important advance in discourse research. These tools include ways to test competing models and ways to track v ery large language databases. The former aid the field's ability to de monstrate theoretical progress, and the latter can perform computation s beyond what a human expert could do in a lifetime. There is a differ ence between tools and theory, however, and in this article, I point o ut some of the ways in which large dimensional space methods such as L atent Semantic Analysis and Hyperspace Analog to Language fall short o f being plausible theories about psychological reality. I examine in-p rinciple failures and wrong-kind failures that arise in the systems an d point out the limitations of systems based exclusively on co-occurre nce.