Context-Sensitive Coding in Speech Recognition, Articulation and Development

This paper describes a theory of the coding of speech at the phonetic level and applies that theory to the problems of speech recognition, articulation, and development. The theory specifies the elementary structural components of words and has nothing to say about the higher-level syntactic and semantic aspects of language. The theory is not contradicted by any of the facts that I know, and it provides a very simple explanation of many facts concerning speech recognition and articulation. However, the theory is extremely speculative, and no one should be misled concerning the amount of direct evidence for the theory. My enthusiasm for the theory is based largely on the clarity and simplicity with which it handles many of the basic problems of speech recognition and articulation, rather than on the definitiveness of its empirical support.