Evaluating case-based decision theory: Predicting empirical patterns of human classification learning