Experience and expectation predict fine details of perception and production
It is well known that phonemes with otherwise similar phonological behavior may be produced differently (e.g. VOT, (Cho & Ladefoged, 1999)) or perceived differently (e.g. consonant confusions (Harnsberger, 2001)) depending on the language. Discussion of these facts has often remained at the observational level. In this talk I will present evidence that fine details of production and perception owe to the fact that speakers and listeners have statistical expectations about segments and words in their languages. I will draw evidence for this conclusion from a series of production and perception studies, drawing on durational difference between singleton and geminate consonants in Cypriot Greek and Italian, the perceptual similarity of consonants and word duration in Kaqchikel Mayan, and lexical mistrieval in English. These phenomena are all conditioned by high-level factors such as functional load, contextual predictability and prior phonetic experience. Some implications of the interplay between these high-level factors and low-level phonetic details are discussed, touching on the areas of phonological reduction, contrast enhancement and neutralisation, and sound change.