Does correctly answering a test question about a multiterm fact enhance memory for the entire fact? We explored that issue in 4 experiments. Subjects first studied Advanced Placement History or Biology facts. Half of those facts were then restudied, whereas the remainder were tested using "5 W" (i.e., "who, what, when, where", or "why") or analogous questions. Each question assessed a specific critical term of the fact. In the first 3 experiments, 1 test question was posed per tested fact; in the fourth experiment, up to 3 different test questions were posed per tested fact. After a delay of at least 24 hr, a final test involved questions that assessed the same terms that were tested during training, as well as questions that assessed a different term from that previously tested. Results showed that testing produced piecewise fact learning: Tested terms benefited relative to restudy, but untested terms did not. That pattern held when either fill-in-the-blank or multiple-choice questions were used during training, when 1 or 2 test trials were used during training, for both history and biology facts, and when more than 1 term from each fact was tested during training. Thus, across a range of circumstances, taking tests on complex facts results in a selective memory benefit for tested terms. In analogous applied settings, testing on multiple response terms should promote more comprehensive retention.