In French, some words ending in a vowel use a consonant-final variant before vowel-initial words (e.g. grand [ɡʁɑ̃] ∼ [ɡʁɑ̃t] ‘little-MASC ’). The consonants occurring at the end of consonant-final variants are called liaison consonants. Liaison consonants are challenging for phonological theory because of evidence that they pattern ambiguously between stable word-final consonants and word-initial consonants. Some researchers have proposed specific phonological representations to account for this ambiguous behavior, including floating consonants and gradient underlying representations.

In this talk, I will propose an alternative account where the ambiguous patterning of liaison consonants is analyzed as a paradigm uniformity effect: in a word 1 – word 2 sequence, the liaison consonant ends up being ambiguous between a stable word-final consonant and a word-initial consonant because of a pressure to make contextual variants of word 1 and word 2 similar to their citation forms (i.e. words as pronounced in isolation).

I will use two case studies to support this analysis: (i) a study of liaison enchaînée in Swiss French using acceptability judgments, and (ii) a phonetic study of liaison consonants in affrication contexts (t#i) in Quebec French. I will show that the data of Study 1 and Study 2 can be modeled using a probabilistic grammar including independently motivated paradigm-uniformity constraints, without any need for special phonological representations.