French Liaison in the Light of Corpus Phonology: From Lexical Information to Patterns of Usage Variation

Based on the PFC corpus (Phonologie du Français Contemporain) and a dataset of 27,788 liaison sites, we show that the phenomenology of French liaison, far from being captured by a phonological process stricto sensu, is regulated by the lexical information carried by the liaison consonant. The principles according to which a small amount of high-frequency and syntactically bound left words produce obligatory liaison do not account for the large periphery of variable liaison, which in turn is subject to lexical constraints and can be the locus of diachronic and socio-stylistic variation. Many of the liaison words occurring with the highest token frequency in the corpus can also be produced with non-realized liaison, thus pertaining to the large periphery of variable liaisons. It is then argued that liaison in itself is stored as a variable process.