In order to make informed instructional decisions, teachers need psychological knowledge about relational categories. We conducted two 2 x 2 experiments to examine effective designs for learning relational categories in the context of teacher education. In both experiments, a blocked compared to an interleaved example format was more beneficial for learning relational categories when generating new examples. Experiment 1 (N = 176) additionally showed that student teachers generated new examples more successfully when they had to classify rather than read examples. Moreover, Experiment 2 (N = 95) revealed that student teachers who learned with examples taken from their individual subjects rather than examples from random subjects generated new examples more successfully. A mediation analysis showed that the subject-specific task value mediated the effect on example generation. The detrimental effects of an interleaved format could be partly compensated for by learning with subject-specific examples.