Many experiments provide evidence that practicing retrieval benefits retention relative to conditions of no retrieval practice. Nearly all prior research has employed retrieval practice requiring overt responses, but a few experiments have shown that covert retrieval also produces retention advantages relative to control conditions. However, direct comparisons between overt and covert retrieval are scarce: Does covert retrieval--thinking of but not producing responses--on a first test produce the same benefit as overt retrieval on a criterial test given later? We report 4 experiments that address this issue by comparing retention on a second test following overt or covert retrieval on a first test. In Experiment 1 we used a procedure designed to ensure that subjects would retrieve on covert as well as overt test trials and found equivalent testing effects in the 2 cases. In Experiment 2 we replicated these effects using a procedure that more closely mirrored natural retrieval processes. In Experiment 3 we showed that overt and covert retrieval produced equivalent testing effects after a 2-day delay. Finally, in Experiment 4 we showed that covert retrieval benefits retention more than restudying. We conclude that covert retrieval practice is as effective as overt retrieval practice, a conclusion that contravenes hypotheses in the literature proposing that overt responding is better. This outcome has an important educational implication: Students can learn as much from covert self-testing as they would from overt responding.