In this paper we propose a frequency analysis of French liaison that focuses on the liaison environments attested in the PFC database. The results of the analysis show the existence of a significant relationship (statistically interpreted as a power-law distribution) according to which a very restricted set of liaison environments has very high frequency of occurrence in the corpus and is substantially untouched by phonological and sociolinguistic variation, while a large "periphery" of infrequent uses appears to show significant aspects of style- and speaker-dependent variation. The study therefore demonstrates the importance of basing any variationist analysis on very large data sample, such as those provided by contemporary, well-reasoned linguistic corpora.