Testing, as a form of retrieval, can enhance learning but it can also induce forgetting of related memories, a phenomenon known as retrieval-induced forgetting (RIF). In four experiments we explored whether selective retrieval and selective restudy of target memories induce forgetting of related memories with or without initial retrieval of the entire learning set. In Experiment 1, subjects studied category-exemplar associations, some of which were then either restudied or retrieved. RIF occurred on a delayed final test only when memories were retrieved and not when they were restudied. In Experiment 2, following the study phase of category-exemplar associations, subjects attempted to recall all category-exemplar associations, then they selectively retrieved or restudied some of the exemplars. We found that, despite the huge impact on practiced items, selective retrieval/restudy caused no decrease in final recall of related items. In Experiment 3, we replicated the main result of Experiment 2 by manipulating initial retrieval as a within-subject variable. In Experiment 4 we replicated the main results of the previous experiments with non-practiced (Nrp) baseline items. These findings suggest that initial retrieval of the learning set shields against the forgetting effect of later selective retrieval. Together, our results support the context shift theory of RIF.