Traversing the Fitness Landscape

The Henry and Theresa Biggs Lecture in Linguistics: Paul Kiparsky, Stanford University


Two mysteries of linguistic change are DRIFT, the incremental spread of innovations over long periods of time, and CONVERGENCE, independent change in different languages along similar unidirectional pathways. I propose an explanation for them based on the idea that learners at each stage of acquisition prefer the most probable language consistent with what they have already learned. A well-defined framework for grammar comparison with a simple formal measure of the probability of a language is provided by Optimality Theory. This measure is RANKING VOLUME, defined as the number of constraint rankings that generate the language, divided by k!, where k is the number of constraints. Ranking volume is already known to be a a good predictor of typological distribution (Bane & Riggle 2008) and of orderly variation patterns (Anttila 1997, 2010). I apply the model to the drift from head-final to head-initial syntax observed in the world’s languages, showing that the minimal constraint system that characterizes the possible word orders also predicts the attested diachronic pathways and stable states in the typological landscape.