In our 24-h society, nocturnal sleep-related accidents are common. Because all individuals are not equal in their responses to sleep loss, it is very important to identify predictors of vulnerability to sleep deprivation in normal subjects. We investigated the performance of a cognitive test of sustained attention, electroencephalogram theta/alpha power, subjective sleepiness, and two circadian markers (core temperature and melatonin) in 18 healthy men (nine morning types and nine evening types, 21.4 +/- 1.9 years) during a 36-h sleep deprivation in a constant routine protocol. Sleep need (self-reported) and baseline sleep structure were also investigated. Nighttime performance impairment was defined as the difference between the mean nocturnal number of lapses (00:00-07:30 [corrected] hours) and the mean diurnal number of lapses (07:30-20:30 hours) expressed as a percentage. Feeling fully alert in the morning just after awakening and/or sleepy in early afternoon were the only two factors (Multiple R > 0.80, > 60% of explained variance) which better predicted the decrease in performances of nocturnal operational tasks requiring sustained attention.