Memory retrieval often enhances still later memory as evidenced by the testing effect. Divided attention (DA) is known to produce different effects on encoding and retrieval, substantially disrupting the former and often producing little effect on the latter. The present experiments examine whether the mnemonic consequences of retrieval are similarly resilient in the face of distraction or if they have a similar sensitivity to DA as study-based encoding. In 2 experiments, participants initially studied a set of word pairs (Phase 1) then engaged in restudy of some pairs and retrieval practice of others (Phase 2), followed by either an immediate or (24-hr) delayed final cued-recall test (Phase 3). Phase 2 restudy and retrieval practice occurred under full attention (FA) or DA. Phase 2 performance replicated earlier research in finding little effect of DA on retrieval success, and greater disruption to the secondary task in the retrieval than restudy condition. More importantly, the testing effect on the final test (the difference between the retrieval and restudy conditions) was greater in the DA than FA condition. Final recall was substantially reduced by DA in the restudy condition but not in the retrieval condition. This pattern was found for related and unrelated word pairs, with feedback during retrieval practice and without, for immediate as well as delayed final tests, and for high as well as low performance on the secondary task. The encoding effects of retrieval appear to be more resilient in the face of distraction than the encoding effects of restudy.