Onset insertion across words in present-day dialectal European Portuguese. A case for competing grammars?

The paper investigates two cases of incremental sandhi in European Portuguese (EP) dialects: the insertion of a nasal onset to vowel-initial postverbal definite articles (V-DA sequences; e.g. {'ʃɐmɐ̃w̃nu=Chamam o} homenzinho para dentro) and preverbal 3 person accusative pronouns (PT-AP; ‘PT’ stands for ‘proclisis trigger’, e.g. Depois {nɐ̃ȷñu=nem o} penteavam). Besides tracing how voweland consonant-initial variants are geographically patterned, intraspeaker variation is analyzed. This variationist dimension requires the following points to be addressed: a) does [n] insertion occur in V-DA sequences in the same areas as in PT-AP?; b) do [n]-initial definite articles in the speech of an informant entail [n]-initial accusative pronouns (or the other way round)? The data have been drawn from the ‘verbatim transcription’ section of CORDIAL-SIN (Martins, coord., 2010), a dialectal corpus containing speech samples collected in 42 localities. The method consists in calculating the proportional frequency of onset insertion. It corresponds to the percentage of occurrences of [n]-initial DAs or preverbal APs, where the total number of both consonantand vowel-initial occurrences in PT-AP and V-DA contexts is 100% (Hoekstra and Versloot 2019). Minimal contrasts between these realizations are captured in terms of diatopic variation (distribution of different mappings over different areas; Seiler 2004: 394) and free variation (random co-occurrence of mappings, defined on an individual basis). Zones populated by speakers tending to apply concurrently the two variants are paid special attention. Intra-speaker variation documented in CORDIAL-SIN is approached in terms of competing grammars, thus relying on mutually exclusive methods of processing input data. The analysis starts from the assumption that each time preverbal APs and postverbal DAs are treated differently with respect to onset insertion, speakers must be aware that they belong to distinct morphosyntactic categories. Cues prompting speakers to come up with divergent solutions are identified, accordingly. The analysis reveals as many as six different non-standard grammars (mappings between syntactic and prosodic domains) at work. Five of them are category-sensitive. Tellingly, no purely phonological processing of V-DA and PT-AP sequences is found, i.e. no interviewee employs consonant-initial DAs only on a par with consonant-initial APs only (although there are speakers who exclusively apply [n]-initial forms in one of the configurations). In addition, 8 informants in 6 six localities display a hit-and-miss grammar with both consonantand vowelinitial articles and both consonantand vowel-initial preverbal clitic pronouns. Consonant-initial preverbal APs are found in 15 interviews. The archipelago of the Azores and the North of the mainland Portugal (with a rich historical antecedence; Maia, 1986) stand out as the areas with most prominent variation. In each locality, though not in the speech of every informant (11 out of 26 speakers apply this variant uniquely), [n]-initial forms appear alongside the standard vowel-initial ones.