January 13, 5:00pm

Anthony Kroch

Statistical Independence in Syntactic Change

University of Pennsylvania

As long ago as 1979, David Sankoff and William Labov, in their defense of the "variable rule," noted that external (i.e., social) and linguistic factors affecting the probability of a binomial linguistic choice tended to be statistically independent in logistic regression (= variable rule analysis). While they did not pursue this point, it has turned out to be of importance in subsequent work, especially in quantitative investigations of diachronic syntax. Indeed, statistical independence among factors influencing linguistic variation has proved more general in the diachronic realm than might have been expected. In the first place, linguistic factors affecting binomial morphosyntactic choices in historical corpora have turned out generally, if not universally, to be statistically independent of text date, an external factor. This is the so-called "Constant Rate Effect." Secondly, and equally significantly, choices between syntactic options undergoing change have repeatedly turned out to be independent of other grammatical options and constraints. In this talk, I will present some of these independence results from the history of English and other languages and discuss their implications for the relationship between grammar and usage.