p . 1 Probability in Language Change

1 Kie Zuraw 0. Introduction Why do languages change? If children are able to infer surrounding adults' grammatical systems from their utterances, and if adults adjust their lexicons and perhaps grammars to achieve better communication with their interlocutors, any linguistic innovations that might somehow arise should be quickly stamped out. This is a combination of Weinreich et al.'s (1968) " actuation problem " (how and why does a particular change occur at a particular time) and what we might call the " continuation problem " : what sustains the momentum of a change, causing an innovation to increase in frequency, to spread from word to word, or to spread from speaker to speaker, rather than stalling or receding? Any answer to the continuation problem that has been proposed must rely on a probabilistic model of the language faculty. If the rise in frequency of an innovation results from snowballing mislearnings, we require a model of how the learner responds to her or his variable environment. If the rise in frequency results from individuals' adopting the speech patterns of some social group, we require a probabilistic model of the speech community, in which individuals probabilistically and incrementally update their grammars and lexicons in response to interlocutors' behavior. Moreover, when the rise in frequency of an innovation involves variation within individuals, as we can often see that it does in written records, we require a probabilistic model of language representation and/or use. Otherwise, we have no way of representing difference between a generation whose members use a new variant 20% of the time and a generation whose members use a new variant 40% of the time. The fact that language change happens seems to demand a probabilistic view of the language faculty, in phonology, morphology, syntax, semantics, processing, acquisition, and the social use of language. The probabilistically oriented study of language change therefore relies on probabilistic models of all the areas of linguistics discussed in this volume. This chapter surveys the role of probability in the study of language change. The first section describes the use of probabilistic tools in establishing language relatedness through vocabulary comparison, an important task when historical and textual records are lacking, and inferences about language change must be drawn from the ways in which related languages differ. The second section examines how the frequencies of linguistic traits change over time in the historical record, and how the …