A very complicated issue in understanding morpho-phonological alternations concerns those phenomena that are pervasive, frequent, and phonotactically-motivated, and yet exceptionful and lexically-sensitive. To what extent are such processes, that apply idiosyncratically to different morphemes, words and even phrases, represented in a way that generalizes to novel forms?

This talk examines this issue via the “well-plowed ground”* of French liaison, specifically those liaison consonants that come and go at the boundary between nouns and the words preceding them (e.g. les [z] amis, ‘the friends’). My main goal is to assess the predictions that different analyses make about (i) how exceptional nouns resist liaison, and (ii) how productive liaison should be with novel nouns. I report the results of an adult production + acceptability study, primarily with L1 speakers of Laurentian French in Quebec, which find liaison to be somewhat intermediate in its application as compared with regular vs. exceptional nouns. I also discuss computational simulations and child corpus studies that try to understand how learners eventually reach the knowledge state of our adult participants.  The overall picture is one in which learners and speakers are making really interesting generalizations about their lexicons and their grammars, including some more abstract than almost any model might predict?… but which also require substantial individual variation.