Does the poorer performance of older people in laboratory tests of episodic memory result from metamemory problems? Memory of 20 young and 20 old adults for 60 "unrelated" paired associates was measured by an associative matching task. Two metamemory measures also were taken: Each pair was rated at the time of study for the likelihood that the person would recall it (prediction measure) and on the matching task each response pairing was judged as to its correctness (response evaluation measure). Young adults were correct on 50% of the associative matches, old only 30%. The two groups did not differ in the predictive metamemory measure: Both were able to predict relative memorability. Old persons, however, underestimated task difficulty (i.e., they overestimated the number of correct associative matches they would make, whereas the young did not). Both groups were quite accurate in the response evaluation measure.