This thesis investigates how educated speakers of Brazilian Portuguese (BP) process subject-verb agreement with unaccusative verbs and postverbal subject DPs during sentence comprehension. It draws on the theoretical frameworks of Generative Grammar, Psycholinguistics, and Variationist Sociolinguistics. Through a series of four experiments, it examines factors such as the linear distance between the subject and the verb, subject animacy, DP subject number, and verbal salience. The results, analyzed within the framework of real-time linguistic computation, reveal nuances in the processing of agreement in unaccusative constructions. These variations depend on the task and the variables investigated. In the first experiment, an acceptability judgment task using a 5-point Likert scale, the factors of subject-verb distance and DP subject number were manipulated (the verb was always singular). The main effects of both variables and interaction were observed. Higher ratings were assigned to shortdistance conditions and singular DP conditions, with the short-singular-DP condition receiving the highest ratings. In contrast, the short-plural-DP and long-plural-DP conditions received lower ratings, with no significant difference between them. Since the sentence remained accessible for 8 seconds, incongruence between the verb s number (singular) and the DP subject s number (plural) may have caused dissonance despite the distance. The second experiment involved a self-paced reading task using the same independent variables as in Experiment 1. A main effect of DP subject number was observed, with a preference for singular DPs. Pairwise comparisons highlighted differences between the long-singular-DP and long-plural-DP conditions. Higher reading times were associated with the latter, indicating a preference for singular forms. The third experiment consisted of a sentence repetition task that examined verb tense (present vs. past) and DP subject animacy as independent variables. The prediction that animate DPs would induce more congruent repetitions was not confirmed. Verb tense as a main factor was also not significant. However, a difference emerged for inanimate DPs, with more correct repetitions observed for inanimate past tense forms. This finding suggests a possible phonological salience effect. The final experiment employed a self-monitoring listening task to evaluate the effects of verb number and DP subject animacy. No main effects or interaction effects were observed. When considered alongside the other tasks, these results suggest that oral contexts favor the acceptability of both redundant and non-redundant agreement, potentially reflecting an instance of speakers’ internal bilingualism (Roeper, 1999) in BP. It can be concluded, therefore, that the investigated variables exert complex effects on the computation of agreement relations in unaccusative constructions and that the cognitive operations and demands elicited by the experimental tasks, as well as the modalities of stimulus presentation (written or oral), influence the processing of these relations, pointing to a context of diglossia.