Double weak theory as a framework for discovery

Abstract

A weak theory is not very attractive, a double-weak one even less so. The double-weak theory of speech perception serves mainly as a cover story for postponing rumination on the deeper principles that ultimately underlie human speech communication. Instead, it adopts minimal assumptions about the relation among symbols, sounds, and gestures that provides a rationale for immediately pursuing some hunches about empirical generalizations, or primitive “laws.” It is hoped that some of these may be akin to Kepler's laws of planetary motion or eighteenth century “gas laws” that may serve as grist for later, more deeply explanatory theories. The active search for such empirical laws in speech perception starts from existing hypotheses about the general structure of surface phonological representations (e.g., features, phonemes, syllables) and proceeds by constructing detailed statistical models that accord with the fine structure of human responses in parametrically controlled speech perception experiments. Some examples and possible extensions of existing models are presented.