This thesis deals with the processing of gender and number features in Brazilian Portuguese (BP) when establishing agreement in predicative structures containing a complex DP in the subject position. It investigates structures such as O telhado das casas estava quebrado na lateral (The roof of the houses was broken on the side) and A bilheteria dos teatros ficou inundada no temporal (The box office of the theater was flooded in the storm), where There is number agreement between the subject and the auxiliary verb, and gender and number agreement between the subject and the participle. To determine whether gender and number features are processed at the same time, whether the semantic phenomenon of distributivity can interfere in the processing of these features, two experiments were conducted: a sentence production task designed to elicit speech errors and a self-paced reading experiment, both with university students who are native speakers of Brazilian Portuguese. The results indicate that: (i) the university student participants, who are users of a formal variant of BP and are therefore aware of the rules of agreement, in fact produce number and gender agreement errors; (ii) the participants are sensitive to sentences with agreement errors in gender and number in reading comprehension; (iii) distributivity is an interfering factor in agreement processing, particularly in production and (iv) the morphophonological gender mark, in the head of the subject, affects agreement processing both in speech production and reading comprehension. The influence of distributivity is explained using Rodrigues (2006) model of a monitored parser in production, which is compatible with the idea of an autonomous syntactic processor that is not subject to semantic interference. According to Rodrigues model, the influence of semantic factors can be accounted for by how complex subject DPs are represented and maintained in working memory and analyzed by the parser during language production. Proposals on gender and number representation (Picallo, 1991/2008; Ritter, 1993; Di Domenico, 1995) are considered in the discussion on dissociation of gender and number features in agreement processing of the predicative structures investigated in this thesis.The results of the production and the comprehension experiments conducted with speakers of BP are compared to the results of studies conducted with Spanish speakers involving an induced error experiment (Antón-Mendez et al., 2002), in the case of production and, in the case of comprehension, ERP (Barber and Carreiras, 2005) and eye-tracking (Acuna et al., 2014) experiments. Comparison and discussion of the experimental results in the two languages covers not only the influence of distributivity on agreement, but also the question of how the number features on the auxiliary verb and the participle are computed (whether it is a single or dissociated process) and the influence of gender marking, particularly participants tendency to produce participles in the masculine form, which in BP is the default, unmarked form.