Lexical diffusion is not lexical analogy

Abstract

This paper refutes Kiparsky's claim that lexical diffusion “behave[s] like lexical analogy in every respect” (1995:643). Evidence from American English glide deletion, t/d deletion, and vowel reduction/deletion before /r/ is elicited to demonstrate the gradient nature of lexical diffusion, as opposed to the quantal nature Kiparsky attributes to lexical analogy. Kiparsky's claim that “the mechanism [for lexical diffusion] is analogical in just the sense in which, for example, the regularization of kine to cows is analogical” is refuted by reference to sound changes which affect the most frequent words first, not the least frequent, as happens in truly analogical changes. In fact, two stress shifts in English, both of which would affect Stratum 1 phonology, exhibit differing patterns of diffusion. It is shown that the use of the term lexical analogy is ambiguous, that Kiparsky uses it to mean both the motivation behind the actuation of the change and the means by which the change is implemented in the lexicon. Lexical diffusion, in contrast, is always a method of implementation, and to the extent to which the implementation of a change requires something different of speakers than does its actuation, the fit between the actuation of a shift and its implementation will always be imperfect.