In two studies, we examined how 5-year-olds weigh similarity against other factors in deciding from whom to learn. Specifically, we examined the factors of history of and reasons for inaccuracy in Experiment 1 (n = 64) and of competence and authority in Experiment 2 (n = 32). In the 1st phase of Experiments 1 and 2, children’s social biases were tested: 5-year-olds met both a similar informant (SI) and a dissimilar informant (DI). These informants were puppets (Experiment 1) or human teachers (Experiment 2). Children could select either informant as a source of object names. Across experiments children systematically preferred learning from the SI over DI. In the 2nd phase of Experiment 1, both informants first provided accurate information and then 1 of them became inaccurate during an event that clearly explained (being blindfolded) or did not explain (wearing a scarf) the inaccuracy. For half the children, the SI was accurate and the DI was inaccurate. Only after inaccuracy events that were causally unclear did children showcase similarity biases in their subsequent learning preferences. Experiment 2 showed that identifying a DI as a teacher (a profession associated with positive attributes) failed to counter children’s similarity bias. These findings provide important insights on contextual factors that contribute to children’s favoring of socially meaningful others.