Subjective alertness may provide some insight into reduced performance capacity under conditions suboptimal to neurobehavioural functioning, yet the accuracy of this insight remains unclear. We therefore investigated whether subjective alertness reflects the full extent of neurobehavioural impairment during the biological night when sleep is restricted. Twenty-seven young healthy males were assigned to a standard forced desynchrony (FD) protocol (n = 13; 9.33 h in bed/28 h day) or a sleep-restricted FD protocol (n = 14; 4.67 h in bed/28 h day). For both protocols, subjective alertness and neurobehavioural performance were measured using a visual analogue scale (VAS) and the psychomotor vigilance task (PVT), respectively; both measures were given at various combinations of prior wake and circadian phase (biological night versus biological day). Scores on both measures were standardized within individuals against their respective baseline average and standard deviation. We found that PVT performance and VAS rating deviated from their respective baseline to a similar extent during the standard protocol, yet a greater deviation was observed for PVT performance than VAS rating during the sleep-restricted protocol. The discrepancy between the two measures during the sleep-restricted protocol was particularly prominent during the biological night compared with the biological day. Thus, subjective alertness did not reflect the full extent of performance impairment when sleep was restricted, particularly during the biological night. Given that subjective alertness is often the only available information upon which performance capacity is assessed, our results suggest that sleep-restricted individuals are likely to under-estimate neurobehavioural impairment, particularly during the biological night.