Prosodic structure and prominence constraints on epenthesis: evidence from hiatus resolution across Portuguese varieties

Languages tend to avoid hiatus and different means may be used to achieve this end, including the deletion or semivocalization of one of the vowels (Casali 1997, Frota 2000, Cabré & Prieto 2005). These processes usually apply within particular prosodic domains, and are often blocked under stress clash configurations (Nespor & Vogel 1989, Frota 2000, Cabré & Prieto 2005). Another way of breaking hiatus is through epenthesis, a common process usually motivated by phonotactic requirements or metrical structure (e.g. word minimal size, rhythmic alternation of stressed and unstressed vowels – Hall 2011, 2013). In this talk we focus on a particular case of epenthesis that is active in non-Standard, Northern and Central varieties of European Portuguese, which consists of the insertion of [j] to break a hiatus formed by central mid or low vowels (a-a), when V2 bears word stress (e.g. a aula [ɐjˈawlɐ]). Besides playing a role in the definition of the context of epenthesis, we show that prosodic and prominence conditions further constrain the phenomenon in non-trivial ways. Although [j]-insertion has long been reported in the literature (Vasconcelos 1895, Cintra 1971, Segura 2013), important details on the phenomenon are largely ignored, which we address here: (i) its prosodic domain of occurrence; (ii) possible effects of higher levels of prominence and of the prosodic status of the unit V1 belongs to (prosodic word – PW, or clitic – CL); (iii) its geographic area; (iv) external factors that may affect its occurrence (e.g. type of speech, age). Speech data were recorded from three districts in the North (Oporto, Braga, Viana do Castelo) and one in the Centre of Portugal (Portalegre). The preliminary data reported below, from 20 subjects, include 3 urban and 3 rural data points. Collected speech comes from three sources: (i) a sub-corpus of a reading task (27 sentences, produced twice by each participant), and (ii) semi-spontaneous tasks (a Map Task and an Interview). In the read speech corpus, prosodic constituency and prominence are controlled for: the hiatus sequence occurs inside or across prosodic words (PW), prosodic word groups (PWG), phonological phrases (PhP) and intonation phrases (IP); and V2 appears as the head/non-head of each domain inside IP. Speakers are evenly distributed in two age groups (20-45 years-old and above 59 years-old). Results from the three tasks show that all of the regions exhibit some amount of insertion, but the frequency and contexts of glide occurrence vary. Globally, there is more insertion when V1 belongs to a CL than to a PW (28% vs 12%, respectively), but some areas almost never show insertion when V1 belongs to a PW (i.e., Portalegre). In the latter cases, the prosodic phenomenon seems to be restricted to maximally local prosodic contexts. Results from the read speech corpus show that, when V1 is part of a PW: (i) only in one region (Viana do Castelo) glide insertion optionally applies across PhP; in most areas, glide insertion optionally applies inside and across PWG; IP boundaries, by contrast, always block insertion; (ii) to the exception of the rural point in Oporto region, higher levels of prominence matter across regions, since there is preference for insertion when V2 is the head of PhP or of PWG (21%), in contrast with non-head contexts (5%). Although older subjects tend to insert more, across age groups there is also a clear effect of type of speech, as semi-spontaneous data favours glide insertion (53% vs 17% in the reading task). This suggests that rather than simply a pattern of change, we may be in face of dialectal struggle within bidialectal communities, especially evidenced in more formal speech: prosodic and prominence constraints favour epenthesis, while external constraints (i.e. Standard prestige) press towards inhibition of glide insertion. Overall, this research contributes to determining language internal and external factors (and their respective weights) in glide epenthesis (along the lines of Coetzee 2009, 2012), and, more in general, to deepen our understanding of the mechanisms involved in phonological variation and change