Performance on the psychomotor vigilance task (PVT) sensitively reflects a circadian modulation of neurobehavioral functions, as well as the effect of sleep pressure developing with duration of time awake, without being confounded by a learning curve. Sixteen healthy volunteers underwent two 40-h constant posture protocols in a balanced crossover design. During these protocols, either low sleep pressure conditions were attained by an alternating cycle of 150 min of wakefulness and 75 min of sleep (NAP) protocol, or high sleep pressure conditions were achieved by total sleep deprivation (SD) protocol. During scheduled wakefulness in both protocols, the PVT was carried out every 225 min. Quantitative analysis of the lapses, slowest (90th percentile) and fastest (10th percentile) reaction times (RTs) during the protocols, indicated that the lapses and slowest RTs were sensitive to changes in homeostatic sleep pressure. Our data indicate that the difference between the fastest and slowest RTs (interpercentile range 10th-90th percentile) was particular sensitive to detect very early effects of growing sleep pressure. On the other hand, decrements in PVT performance which were related to circadian phase did not depend significantly on any categorization (such as percentiles of the RTs).