Three experiments explored whether group membership affects the acquisition of richer information about social groups. Employing a minimal-groups paradigm, 6- to 8-year-olds were randomly assigned to 1 of 2 novel social groups. Experiment 1 demonstrated that immediately following random assignment to a novel group, children were more likely to generalize negative behaviors to outgroup members and positive behaviors to ingroup members and to report a preference for ingroup members. Experiments 2 and 3 showed that this initial ingroup-favoring bias interacts with subsequent learning, thereby attenuating the effect of negative information about the ingroup and enhancing the effect of negative information about the outgroup. These effects were more powerful with respect to preferences than induction: After hearing that some ingroup members behaved badly, children predicted that ingroup members would behave more negatively than outgroup members, but they did not express preferences for the outgroup over the ingroup. Together these data shed light on the construction of social category knowledge as well as the processes underlying the absence of own-group positivity among children from lower-status social groups.