Since the inception of Labov's variationist approach to the study of linguistic change, there have been two approaches to phonology: one theoretically motivated, addressing relatively invariant idealizations of language; and another empirically motivated, in which language is taken to be fundamentally variable. Since the time of the controversy over the incorporation of probability into grammar in the Variable Rule (VR) model, these two approaches have largely diverged. Critics have expressed skepticism about probabilistic rules, challenged their psychological reality, and rejected the variationist's methods of aggregating speakers (cf. papers collected by Singh 1996). Variationists have responded with improved methods, more thorough demonstrations of variability, and articulations of the relevance of variation to linguistic change (e.g. Labov 1994). The controversy continues today; and as readers of the present volume will see, it has entered new territory.