The goal of the present study was to replicate Burt et al. (2009) and Hou et al. (2013) findings by determining the contribution of peers' deviance to changes in participants' (monozygotic [MZ] twins') self-reported delinquency from mid- to late adolescence while controlling for possible gene-environment correlations (rGE) through the use of the cotwin method in the context of a cross-lagged design. Two separate samples were used to this purpose. The first sample included 289 MZ-twin pairs (48.42% female; 65.5% of European descent, 21.7% of African descent, 12.8% other; average age at first assessment = 16.20 years) and relied on perceived measures of peer deviance, as in the two original studies. The second sample comprised 181 MZ-twin pairs (50.67% female; 84% of European descent, 3% of African descent, 2% of Asian descent, 2% Native North Americans, and 9% unspecified; average age at first assessment = 13.05 years) and used direct measures to assess peers' deviance. Participants in the first sample were part of a large representative sample of American adolescents, whereas the second sample included families with twins born in a large Canadian city. Results showed that within-pair differences in self-reported delinquency were stable from midadolescence to late adolescence in both samples. As in Burt et al. (2009) and Hou et al. (2013) within-pair differences in peer deviance did not predict changes in within-pair differences in self-reported delinquency in both samples (i.e., no socialization effect). In contrast to Burt et al. (2009) and Hou et al. (2013), however, we found no predictive link from twins' self-reported delinquency to peers' deviance (i.e., no selection effect). Finally, as in Burt et al. (2009) we found that peers' deviance was stable from middle to late adolescence and that peers' deviance and participants' delinquency were cross-sectionally correlated in the first sample, but not in the second sample. These results are interpreted in the context of the ongoing debate in regard to peers' socialization vs selection effects.