The testing effect refers to the finding that retrieving previously encoded material typically improves subsequent recall performance more on a later test than does restudying that material. Storm et al. (2014) demonstrated, however, that when feedback is provided on such a later test the testing advantage then turns to a restudying advantage on subsequent tests. The goal of the present research was to examine whether there is a similar consequence of feedback when the difficulty of initial retrieval practice is modulated. Replicating prior research, we found that on an initial delayed test, recall of to-be-learned items was better following difficult than easy practice. Critically, however, providing immediate feedback on an initial delayed test reversed this pattern. Our findings are consistent with a distribution-based interpretation of how feedback at test modifies recall performance.