Two procedures were investigated in an attempt to decrease the variability of overlearning response latencies in a study-test paradigm, paired-associate task matching CVC's with response keys: (1) self-pacing the task by presenting test trial stimuli whenever the subject pressed a "home" key; and (2) instructing and shaping subjects to keep home key depressed until they selected a response key and measuring the period of home key depression as the latency of response onset. Self-pacing was found to decrease the variability of S-R latency, but only during the early stages of overlearning drill. There was no apparent utility in timing response onset as opposed to the complete S-R response. (Author)