On each of five study-test trials, young and old adults attempted to memorize the same list of 60 words (e.g., bed, rest, awake), which were blocked according to their convergence on four corresponding associates. Half of the participants in each age group were given an explicit warning about the DRM paradigm prior to encoding and were asked to attempt to avoid recalling any associated but nonpresented words (e.g., sleep). Lists were presented auditorily at either a fast (1,250 msec/word) or a slow (2,500 msec/word) rate. Without a warning, the probability of veridical recall across trials increased for both age groups; however, the probability of false recall across trials decreased only for young adults. When a warning about false recall was provided, young adults virtually eliminated false recall by the second trial. Even though old adults also used warnings to reduce false recall on Trial 1, they were still unable to decrease false memories across the remaining four study-test trials. Old adults also reduced false recall more with slow than with fast presentation rates. Taken together, these findings suggest that old adults have a breakdown in spontaneous, self-initiated source monitoring as reflected by little change in false recall across study-test trials but a preserved ability to use experimenter-provided warnings or slow presentation rates to reduce false memories.