Several devices are now commercially available to measure, monitor or predict alertness state. While leading technologies rely on signals from the driver (i.e., eye closure, brain activity) or vehicle (i.e., steering deviation), ocular parameters remain a primary candidate for accurately monitoring driver state. A novel ocular ‘biomarker’ known as ‘pupillary unrest’ reflects alterations in sympathetic nervous system activity via instability of pupillary response. The extent to which the pupillary unrest index (PUI) accurately predicts subsequent performance impairment is unknown. Seven healthy young adults (4 men, 21.9 ± 0.6y) underwent a three night stay in the laboratory, comprising one night of baseline sleep, a 40-hour extended wake period, and a night of recovery. Two hours post habitual wake-time, and thereafter bi-hourly, participants completed a subjective rating of sleepiness (KSS), an 11-min test of PUI, and a 10-min visual PVT. Data were averaged across the first 16 hours to create a baseline period, and each time point compared for time awake effects. To examine the predictive capacity of PUI, compared to subjective ratings alone, each PVT was classified as “impaired’ if lapses exceeded a 25% (mild), 50% (moderate) or 75% (severe) threshold increase above individual baseline levels. ROC analyses were then performed. PUI, PVT lapses and KSS showed a similar time course across the 40 hours of sleep deprivation. All metrics increased as a function of time awake (p < 0.04), peaking 26-30hours post wake. While the predictive capacity of both PUI and KSS were moderate-high (PUI: 0.66–0.79; KSS: 0.86–0.89), this was dependent on the level of impairment: PUI was a better predictor at lower levels of impairment [AUC 0.79 (25/50%) vs. AUC 0.66 75%), while KSS remained high for all (AUC > 0.86). While self-reported sleepiness accurately predicts performance impairment, few drivers engage in self-monitoring and/or adaptive behaviours. Objective markers are therefore essential in combating drowsy driving. Our data provides evidence for the utility of this novel ‘biomarker’ to be developed as a predictor of performance impairment, particularly in the early stages of alertness failure. This research was supported by VicRoads.