Initial learning can interfere with subsequent learning (proactive interference [PI]), but recent work indicates initial testing can reduce PI. Here, we tested 2 alternative hypotheses of this effect: Does testing reduce PI by constraining retrieval to the target list, or by facilitating a postretrieval monitoring process? Participants first studied 4 lists of unrelated words. The study-only group performed a distractor task following each list, whereas the tested group recalled each list. After these initial lists, both groups studied and were tested on a final list. Replicating prior work, the tested group recalled more of the final list items and had fewer prior-list intrusions than the study-only group (i.e., initial testing reduced subsequent PI). To test the 2-alternative hypotheses, Experiment 1 used a modified recall test for the final list, whereby participants were asked to recall the final list of words and also report any items from prior lists that inadvertently came to mind. Contrary to the constrained retrieval hypothesis, initial testing did not reduce the number of prior list items that came to mind, but consistent with the postretrieval monitoring hypothesis, testing increased the likelihood that the intrusions would be correctly attributed to prior lists. Experiments 2 and 3 further tested the postretrieval monitoring hypothesis by testing the final list twice. According to the hypothesis, testing all of the lists should render prior testing nondiagnostic of list membership, thereby impairing retrieval monitoring in the test group and minimizing its ability to reduce PI. This prediction was confirmed.