Four groups of subjects performed a 48min, computer-controlled, visual watchkeeping task. Two groups were run under fixed, non-adaptive conditions, ,one with immediate knowledge of results' ( K R ) and the other without (NKR). The KR group showed the usual superiority in detection rate over the NKR group, and made fewer cornmissive errors (false alarms). Two othergroups, also K R and NKR, ran under adaptive conditions, wherein the size of the signals they watched for was adjusted during the vigil according t o past performance, so as to maintain a preset detection rate. The resulting curves for the adaptive variable closely resembled the traditional performance measure, detection rate. Various adaptive strategies are discussed