Rabbitt (1979, 1981) has argued that the slowing of choice RT performance with old age reflects less sensitive control over speed of responding. Based on the finding that old subjects' errors were as fast as young subjects', whereas their correct responses were much slower, Rabbitt suggested that old subjects often overshoot when trying to increase or reduce response speed. Data collected from a different serial, four-choice RT task in this experiment were not, however, consistent with Rabbitt's account. Error as well as correct RTs were longer in old than in young adults, with both showing similar increases with age. A post hoc examination of the aged subjects' data also indicated important individual differences within aged samples. Some subjects' RTs and error rates were no different from those of the young; other subjects' longer RTs and lower error rates were consistent with an increased emphasis on accuracy.