When presented with some information multiple times, people learn it more effectively when engaging in retrieval practice in a variable rather than constant manner. However, they believe constant learning to be more beneficial. We thus aimed at helping participants appreciate the benefits of varied learning. Using foreign word translations as study materials, we had participants complete two study-test blocks with half of the translations presented in a varied manner, and half in a constant manner. We asked for global predictions of performance before each test, and for global postdictions after each test. Study and test experience did not alleviate the preference for constant learning, and neither did instructions explaining the benefits of varied learning inserted in between the blocks. Item-by-item performance feedback on the first test, however, managed to help participants appreciate the benefits of varied retrieval practice. This underscores the effectiveness of item-by-item strategy comparisons for mending metacognitive errors.