Elastic s+C and Left-moving Yod in the Evolution from Latin to French

Abstract Elastic s+C is the idea that s+C clusters are heterosyllabic by default in all languages, and that some repair will occur in case, pending on language-specific circumstances, a heterosyllabic parse is illegal (preceding long vowel, preceding coda, beginning of the word). The repair at hand is the branching of the s on the following empty nucleus. This generalization is derived from the behaviour of left-moving yod in the diachronic evolution from Latin to French. The floating yod (here coming from palatalization k+i,e > j+ʦ) anchors as a coda if the preceding syllable is open (placēre > plaisir), but is lost in case it is closed (cancellāre > chanceler), except when the syllable-final C is s (cresc(e)re > croistre (mod. croître)). We know independently that intervocalic s+C clusters are regular coda clusters: they block diphthongization (testa > teste (mod. tête)). Hence s is elastic: s+C is a regular coda cluster unless there is a demand for s to vacate its coda position. It is shown that among all syllabic identities for s+C that are entertained in the literature only one is compatible with this pattern: in CsC clusters, i.e. in absence of a preceding vowel, s branches on the following empty nucleus, i.e. the one that separates it from the following C. This is confirmed by an independent pattern: the middle consonant of CCC clusters is lost unless it is s (CsC), but is regularly dropped in sCC clusters. Here as well s+C is a regular coda-onset cluster when preceded by a vowel (sCC), but s elastically becomes a non-coda when preceded by a consonant (CsC). This empirical generalization appears to be an unprecedented finding: s in s+C is a coda when preceded by a vowel, but a (branching) non-coda when not preceded by a vowel. It is shown that it may solve a good deal of the notoriously mysterious behaviour of s+C clusters as such, i.e. in other languages and in synchronic analysis. Word-initially s+C is not followed by a vowel and therefore a non-coda, thus accounting for the typical cross-linguistic pattern whereby s+C is exceptional word-initially, but not word-internally (where it is followed by a vowel). Also, the branching analysis solves the mysterious fact that s only shows exceptional behaviour when it is followed by a consonant: there is no empty nucleus it could branch on when followed by a vowel.