This article proposes an extension of the connectionist analysis of group biases by Van Rooy et al. (2003) to provide a more parsimonious account for ingroup favoritism as caused by self-anchoring (i.e., projecting features of the self in defining the ingroup) and to test the novel hypothesis that out-group derogation is caused by self-competition (i.e., using opposite features of the self for defining the outgroup). Moreover, our connectionist account predicts that these two effects are more pronounced under conditions of self-priming that render characteristics of the self more accessible. Two experiments using minimal groups demonstrate (a) that ingroup favoritism is due to self-anchoring or the projection of self features in the ingroup, (b) that outgroup derogation is a direct consequence of self-competition (i.e., contrasting away from the self) rather than of contrasting away from the ingroup, (c) that self-priming (by making self-evaluations first) leads to more pronounced evaluation of the self, and to more contrastive evaluation of the ingroup and outgroup. A connectionist simulation in which these predictions are simulated shows a large overlap with the experimental data, supporting our connectionist perspective. The discussion centers on other predictions based on a connectionist approach of group categorization.