Bybee is concerned with providing an understanding of a vast amount of evidence from language use that remains unaccounted for in the usual models of generative phonology, including its currently popular descendant, Optimality Theory.

It may seem surprising, but it is true nonetheless, that much evidence of language use is missing in the generative tradition, which has typically—some will say necessarily—been satisfied with presumptive facts intuitively available to generative phonologists—for example, the replacement of syllable-final [s] by [h] in Cuban Spanish. Just part of Bybee's evidence concerning this example (p. 140) is that [s] occurs more often than not before a pause, has been typically replaced by [h] only within morphemes, and 48% of the time is replaced by [h] even before vowels. Bybee would replace the generativist analysis with a connectionist account of the facts, which she shows is better able to deal with the pervasive evidence of variation and the crucial role of frequency in determining this.

The trend of laboratory phonology begins to weaken the generative phonological reliance on intuitions, and perhaps the time is ripe for appreciation of Bybee's thorough review of the variationist facts and for her formal theory in which, basically, all simple and complex words known to memory appear in allophonic detail in the lexicon. This is true for many phrasal units as well, and there are no phonological rules, only schemas, or frequency-weighted generalizations over memory traces in which form and meaning co-occur.

Generative phonologists may say that such lexical complexity is unreasonable, that linguistic theory is not ready to deal with variation, that dealing with it must await better understanding of the grammar of an ideal speaker-hearer, and that schemas remain a vague notion. However, Bybee has made clear the extent of the problem and how ridiculously far short of the facts the generativist account falls. She has made apparent the fundamental role of frequency in determining variation, and how variation can at least begin to have influence—if not formal expression—in phonological theory.