O processamento de orações relativas e pseudorelativas: a concordância de número e o fenómeno da atração em português europeu

In this thesis we examine the availability of "Pseudo-relatives" structures (PR) in European Portuguese (EP) and the attraction phenomenon of involving number agreement in both offline and online tasks. Grillo & Costa (2012, 2013) argue that previous work on the attachment of relative clauses (RC) to complex noun phrases in object position (Cuetos & Mitchell, 1988) did not control the availability of PR across languages and structures. PRs are string-identical to RCs, but the two types of clauses have different structure and interpretation. The authors also argue that the unexplained variation in preferences to link the OR (eg., prosody, referentiality), comes from the fact that the PR is grammatically forced to an apparent non-local attachment. Regarding the processing of number agreement, Lourenco-Gomes (2008) and Lourenco-Gomes, Costa and Maia (2011), comparing the RC attachment with the sentences disambiguated by gender and number agreement in EP, noted that number agreement generated higher error rates than gender agreement. Lourenço-Gomes & Lindemann (2012) further demonstrated an effect of number agreement and segmentation on error rates. We propose that these results are best explained as attraction phenomena (Bock & Miller 1991), rather than attachment. The phenomenon of attraction occurs when a complex NP with two or more nouns with different morphological marking leads to agreement mistakes in the verb (AcuñaFariña, 2009). To investigate these questions we conducted three experiments with offline measures to compare preferences of attachment between OR and PR and one self-paced reading experiment, with online and offline measures, in order to compare these preferences with the number agreement. After methodological improvements made in the experiments 1 and 2, which showed unclear or conflicting results, the experiments 3 and 4 showed generally a stronger preference for high attachment (HA) in PR-context than in context of genuine RC; faster response times for PR than RC; higher error rates for low attachment (LA) than HA in PR-context and faster response times for HA than LA in PR context. These findings support Grillo & Costa's proposal (2012, 2013). With regard to the attraction phenomenon, we observed higher response times in the condition in which N2 singular intervened between N1 singular and the RC, which allows us to interpret the results as supporting an attraction-based account.