Previous research suggests a direct relationship between criminal friends and criminal thinking style; however, social identity theory proposes that identity mediates the impact of social group members on development of thinking styles. This research project is the first attempt to empirically test the mediating role of criminal social identity in the development of criminal thinking styles within a recidivistic prison sample (N = 312). The structural equation model of criminal thinking style presented and tested in this study supports the central predictions of social identity theory, with findings demonstrating an indirect effect of antisocial friend associations on criminal thinking through in-group affect and in-group ties with criminal in-group members that reflect two of three dimensions of the measure of criminal social identity applied in the current study. Further implications in relation to theory and previous studies are discussed. © 2013 Copyright Taylor and Francis Group, LLC.