Asymmetries and generalizations in early French minimally-rising syllable-contact clusters: revisiting Proto-French stop epenthesis

This paper presents a highly-granular syllable-contact analysis of Proto-French (2nd-9th centuries) stop epenthesis in minimally-rising sibilant or sonorant + sonorant clusters (cisera > cisdre ‘cider’, cumulu > cumble ‘peak’, camera > chambre ‘room’, molere > moldre ‘grind-INF’) within a parallel optimality-theoretic framework. Capitalizing on Split Margin Approach syllable-contact constraints, place agreement and faithfulness, and phonotactic constraints against [*tl, dl] sequences, the account offers a cohesive analysis explaining both the asymmetric gap in stop epenthesis for word-medial /sl/ clusters (i(n)sula > isle, *istle/*isdle/*iscle ‘island’) as well as for deviant clusters whose epenthetic velar stop defies the homorganicity requirement with the input coronal coda, such as medial /nl/ > [ŋ.gl] (spinula > espingle ‘pin’) and word-initial /#sl/ > [(e)skl] (slaitan > (e)sclater ‘burst-INF’). The analysis elucidates these non-canonical reflexes, which have proven elusive in the literature, by drawing on other attested phenomena affecting early French minimally-rising clusters such as manner change (/tN, dN/ > [z.n], e.g. retina > resne ‘rein’), regressive nasal assimilation (computare > conter ‘count-INF’), repairs resulting from the phonotactic aversion to /tl, dl/ (Late Latin place change to [kl, gl], e.g. vetulus > ve[kl]us; Proto-French deletion, e.g. /tl, dl/ > [ll] spatula > espalle ‘shoulder,’ ustulare > usler ‘burn-INF’) and the onset of phrase-level prothesis of word-initial /#sC/ → [#(e)sC] as of Late Latin, proposing a modified dating of /#sl/ > [#skl] and explaining how and why /sl/ and /nl/ flout the canonical stop epenthesis pattern.