We examined the hypothesis that interpolated testing in a multiple list paradigm protects against proactive interference by sustaining test expectancy during encoding. In both experiments, recall on the last of 5 word lists was compared between 4 conditions: a tested group who had taken tests on all previous lists, an untested group who had not taken any tests on previous lists, and 2 other groups (one tested and the other untested) who were warned about the upcoming test prior to study of the fifth list. In both experiments, the untested/warned group performed significantly better than the untested/unwarned group on both correct recall and prior list intrusions but did not achieve the same recall accuracy as tested groups. In Experiment 2, an instruction manipulation check further narrowed the gap between the untested/warned group and the tested groups. In addition, we verified that a reduction in test expectancy indeed occurred in the untested group compared with the tested group by asking participants to indicate how likely they believed they were to receive a test on each studied list. These findings suggest that testing protects against proactive interference largely via attentional processes and/or more effective encoding.