More than seventy years of probabilistic phonology

Abstract

AbstractThe idea that phonology is probabilistic goes back to Pāṇini, and the classic distinction between accidental and systematic gaps in the lexicon is implicitly probabilistic. This idea took on new vigour just over seventy years ago with Shannon’s work on information theory. It is now supported by a wide variety of studies concerning all levels of representation in the theory of sound structure. Models that use probabilities have had striking successes in capturing how people learn the sound patterns of their language from experience, and deploy their knowledge productively and adaptively. However, people’s patterns in production often deviate systematically from the statistical patterns in the input they receive. Cognitive biases, propensities for regularity and structure, and social influences on attention and memory, all play important roles.