Four older and 4 younger men were given extended exposure to a continuous-recognition memory procedure. Experimental variables included the type of stimulus (alphanumeric strings, words, or sentences), the intervals separating repeated items, gains and losses for correct and incorrect recognitions, and the extent of practice with the memory task. Signal detection analyses indicated that the older men generally were less accurate (sensitivity), particularly when the stimuli were strings, but that age differences decreased with practice. Under conditions in which the payoff matrix was neutral, the older and younger men showed equivalent rates of hits and false alarms (bias). Alteration of the matrix to require more liberal or more conservative patterns of recognition responding led to corresponding changes for men of both ages. Adjustments by the older men, however, were not as close to the bias values called for by the new matrices.