We explore how context change and item variation during natural category learning influence memory and generalization to new examples. Participants studied either images of the same bird or varied birds from each of several categories. These images could be presented in a constant background color or different background colors. During test, birds were presented in only one of the studied background colors. Performance at test depended on the context overlap between study and test, with better performance when there was minimal context change during study. Also, contrary to previous findings, we found that learners generalized better when items were repeated during study and remembered old items better when items were varied during study. When there is a moderate degree of context change, there is no benefit of repetition or variation for either novel or old items. These results indicate that context change and item variation have complementary effects on learning.