This thesis investigates how children growing up in Rio de Janeiro city (Rio) deal with variable input regarding the morphophonological expression of number agreement in Brazilian Portuguese (BP). Plural in BP may be expressed only in the determiner (non-redundant, non-standard variety) or in all agreeing elements (redundant, standard variety). The level of variation is influenced by social factors such as level of education and socioeconomic status (SES): the more educated the speaker is, the more redundancy is morphologically expressed (NARO, 1981; NARO; SCHERRE, 2015; SCHERRE; NARO, 1998). In Rio, these varieties co- exist, given that people from different SES interact on a daily basis. It is proposed that exposure to variable input yields underspecification (ADGER, 2006; ADGER; SMITH, 2010). In the case of number agreement in BP, it is suggested that variation results in underspecification (see ROORYCK, 1994) of morphophonological features in the lexicon pertaining to plural agreement redundancy. We hypothesize that there is a level of bilingualism regulated by SES in number agreement in BP: gradual specification of morphophonological information is dependent on social factors, resulting in a sort of bilingualism. An elicited production task was carried out with preschoolers, from both private (Priv) and public (Pub) schools in Rio’s suburban area, in order to verify whether preschoolers exhibit preference for any of the morphophonological expressions of number agreement; type of school is taken as a social variable (ALVES; SOARES; XAVIER, 2014). Results show that preschoolers exhibit a considerable level of variation. However, they differ in terms of production of non-standard varieties, being Pub more likely to produce non- standard forms than Priv. It is argued that during morphophonological encoding in language production, the access to morphemes stored in a Pool of Variants (ADGER, 2007) becomes subject to frequency (LEVELT, 1999) and influence of social factors. Obligatory plural marking in the determiner is explained both from the perspective of language knowledge and its representation and from an online computational model for language production (CORRÊA; AUGUSTO, 2007, 2011), assuming a two-phase DP (PICALLO, 2017) which is phase-based transferred (CHESI, 2007) to morphophonological encoding. The effect of schooling/ literacy in number agreement variation is also verified with Priv and Pub 6th graders. Results show an effect of type of school/ SES in which Priv produces more standard responses than Pub. Additionally, an effect of overall academic performance was obtained among Pub 6th graders from the same school: Pub A (above-average academic performance) are less subject to variation than Pub B (below-average academic performance). Overall, 6th graders results show that academic performance and SES interact in a gradient continuum of number agreement redundancy production: Priv > Pub A > Pub B. Furthermore, a test of assessment of linguistic abilities regarding number agreement, MABILIN II (CORRÊA, 2000), verified whether Pub B 6th graders are able to process grammatical information pertaining to number, given that their performance was similar to Pub preschoolers, despite their age difference. Preschoolers and 6th graders results are discussed in terms of bi/multilingual-like contexts: a) BP speakers acquire both varieties simultaneously, as in Bilingual First Language Acquisition (MEISEL, 1994); b) Good academic performance may enhance metalinguistic awareness allowing for code-switching depending on the level of proficiency in one of the varieties (CRAIG; WASHINGTON, 2004); c) good academic performance and higher SES may result in a sort of passive bilingualism or passive bidialectalism regarding the non-standard forms (see CORNIPS, 2014). Finally, preschoolers and 6th graders results are discussed in terms of a multilingual continuum, characterized as underspecification as variation within one grammar in one extreme to access to different, independent specified grammars in another extreme. More broadly, this thesis is inserted in an interdisciplinary field, combining sociolinguistics, psycholinguistics and theoretical formal linguistics.