Forty-six male train drivers (mean age=46.5, SD=5.1) were recruited to participate in a diary study for 14 consecutive days with questions about their sleep and working hours. A polynomial mixed-effect regression model showed a curvilinear relation (p < .001) between shift-start time and sleep duration for shifts starting at 03:00-12:00 h and with a near linear increase for ones starting between 04:30 and 09:00 h of approximately 0.7 h for every 1 h the shift was delayed. The longest sleeps were estimated at ∼8 h before shifts that started at ∼10:00 h. The shortest sleeps were found for shifts that started before 04:30 h and were estimated at ∼5 h. Individual differences were estimated with a random-effect standard deviation of 0.51 h, independent of shift-start time (p =.005). One-half of the between-subject variance was explained by subjective health. A one-step decrease in health was associated with a 26 min increase in sleep length. The results have practical implications for constructing shift schedules. Early morning shifts reduced sleep length substantially and should be mixed with later start hours to avoid the accumulation of sleep dept. Delaying the shift-start past 10:00 h had little effect on sleep opportunity; however, delaying shift-start to between 04:30 and 9:00 h had a strong impact on sleep length, with 70% of the extra time used for sleep, suggesting large positive effects for this range of shift-start times.