Memory retrieval affects subsequent memory in ways both positive (e.g., the testing effect) and negative (e.g., retrieval-induced forgetting, RIF). The changes to memory that retrieval produces can be thought of as the encoding consequences of retrieval, examined here with respect to attention. In three experiments, participants first studied category--example word pairs, and then practiced retrieval for half the pairs from one-third of the categories (the R+ items) and restudied half the pairs from a different third of the categories (the S+ items), while the final third of the categories were in the nonpracticed control condition (the Np items). This was followed in turn by a final test over all categories and examples, including the unpracticed examples from the retrieval-practice and restudied categories (the R- and S- items, respectively). The middle phase (of retrieval practice and restudy) was conducted under full attention (FA) or under divided attention (DA) in which participants also performed a distracting secondary task. DA had little effect on final recall in the retrieval practice (R+) condition but significantly reduced final recall of the restudied (S+) items, producing a net increase in the testing effect relative to the FA condition. RIF (measured as the difference between the R- and Np items) was substantial in the FA condition but was eliminated by DA. This occurred because the final recall of R- items significantly increased in the DA compared to FA condition, a highly unusual result in which distraction actually improved an aspect of memory performance. In sum, DA during retrieval practice dissociated the positive and negative effects of retrieval on subsequent memory, increasing the positive effect, embodied by the testing effect, but decreasing the negative effect, embodied by RIF. The theoretical implications are discussed.