This paper analyzes sandhi phenomena in European Portuguese in which coda consonants in word-final position are resyllabified to become onsets in two ways: by epenthesis when they occur at utterance boundaries (excluding /ʃ/), and by associating with following onsetless words (within the same utterance). I present an Optimality Theoretic account for why this resyllabification occurs, which includes a constraint against assigning moras to consonants (*Cμ), and a constraint against having codas (NO-CODA). These constraints work together to produce the facts we see in the European Portuguese data: /ɾ/ and /l/, which I argue are moraic codas, are resyllabified in both environments mentioned above, but /ʃ/, which I argue is a non-moraic coda, is only resyllabified utterance-medially before onsetless words. I then discuss the ramifications that resyllabification across word boundaries has for the relationship between syntactic and phonological words, with reference to Selkirk’s (2011a; 2011b) Match Theory; although there is correspondence between words on these two levels, those corresponding items need not consist of exactly the same number of segments.
Burkinshaw, K.D. (2016). Mismatches between European Portuguese lexical and phonological words. Calgary Working Papers in Linguistics, 29(Fall), 4-18.