Latent semantic analysis: A theory of the psychology of language and mind

In a recent article in Discourse Processes (Vol. 25, Nos. 2 & 3, 1998), Landauer, Foltz, and Laham described a computational model called Latent Semantic Analysis (LSA) and summarized its successful simulations of a variety of human performance phenomena that depend on word and passage meaning. Subsequent articles in the same special issue of Discourse Processes reported details of several of these studies. Charles Perfetti (1998), in a commentary, agreed that LSA is a useful research tool but argued that it should not be regarded as a plausible theory of mind because it is based on learning from co-occurrence data. In this response, I show why this objection lacks merit, and I clarify what LSA has to offer. LSA does not handle all aspects of language processing, but it offers a biologically and psychologically plausible mechanistic explanation of the acquisition, induction, and representation of verbal meaning.