In line with variationist sociolinguistics interests (LABOV, 2006(LABOV, [19662008[1972), this Master's thesis focuses on variable (L) in syllable coda (alto vs. arto 'tall') and in complex onset (cliente vs. criente 'client') in São Miguel Arcanjo (SMA), a town in the countryside of São Paulo state.The collected sample consists of 24 sociolinguistic interviews in SMA, stratified by sex/gender (masculine; feminine), age group (18-25; 30-40; 50+ years old), level of education (elementary; high school) and region of residence (rural or urban). In the case of the last category, SMA is clearly divided in these two regions, which differ from one another socioeconomically and, hypothetically, also sociolinguistically. One of the main goals of this research is to investigate whether local linguistic usage correlates with the rural-urban differentiation.Both qualitative and quantitative analyses have been carried out, with the former showing that rotic (L) in syllable coda is strongly avoided by SMA speakers, with rotic tokens not even reaching 1% of the data sample. Therefore, rotic (L) in syllable coda can be categorized as a local stereotype, as per Labov (2008 [1972]) and Bortoni-Ricardo (2011). In complex onset, results of statistical tests suggest that the rotic variant tends to be avoided by younger speakers, those with a higher level of education and those who live in the urban area. This confirms the hypothesis that rotic (L) would tend to occur more in rural speakers' speech.Quantitative analyses further suggest a change in progress in SMA, where the norm would have been to pronounce complex-onset (L) as a rotic; but with a pattern of change towards the non-rotic pronunciation of (L) in this context, with [l] being used as standard or in accordance with the "normative norm" (FARACO & ZILLES 2017). This change could move towards the rural areas, where we currently observe stable variation and resistence to linguistic standardization.Finally, this thesis shows that, although stigma can be associated with rotic (L) in complex onset, since women and speakers with higher levels of education tend to avoid it, half of our informants do not even notice rotic (L) when they are asked "What do you think of this sentence: O Dougras imprica com tudo 'Douglas questions everything'" (with rotic L). Thus, in spite of the obtained results of statistical tests, it is not possible to firmly state that rotic (L) is actually stigmatized SMA.