Fatigue risk associated with work schedules of hospital doctors is coming under increasing scrutiny, with much of the research and regulatory focus on trainees. However, provision of 24 h services involves both trainees and specialists, who have different but interdependent work patterns. This study examined work patterns, sleep (actigraphy, diaries) and performance (psychomotor vigilance task pre- and post-duty) of 28 anaesthesia trainees and 20 specialists across a two-week work cycle in two urban public hospitals. Trainees at one hospital worked back-to-back 12 h shifts, while the others usually worked 9 h day shifts but periodically worked a 14 h day (08:00-22:00 h) to maintain cover until arrival of the night shift (10 h). On 11% of day shifts and 23% of night shifts, trainees were working with> or =2 h of acute sleep loss. However, average sleep loss was not greater on night shifts, possibly because workload at night in one hospital often permitted some sleep. Post-night shift performance was worse than post-day shift performance for the median (t((131))=3.57, p<0.001) and slowest 10% of reaction times (t((134))=2.91, p<0.01). At the end of night shifts, poorer performance was associated with longer shift length, longer time since waking, greater acute sleep loss, and more total work in the past 24 h. Specialists at both hospitals had scheduled clinical duties during the day and were periodically scheduled on call to cover after-hours services. On 8% of day shifts and 14% of day+call schedules, specialists were working with> or =2 h of acute sleep loss. They averaged 0.6 h less sleep when working day shifts (t((23.5))=2.66, p=0.014) and 0.8 h less sleep when working day shifts+call schedules (t((26.3))=2.65, p=0.013) than on days off. Post-duty reaction times slowed linearly across consecutive duty days (median reaction time, t((131))=-3.38, p<0.001; slowest 10%, t((160))=-3.33, p<0.01; fastest 10%, t((138))=-2.67, p<0.01). Poorer post-duty performance was associated with greater acute sleep loss and longer time since waking, but better performance was associated with longer day shifts, consistent with circadian improvement in psychomotor performance across the waking day. This appears to be the first study to document sleep loss among specialist anaesthetists. Consistent with observations from experimental studies, the sleep loss of specialists across 12 consecutive working days was associated with a progressive decline in post-duty PVT performance. However, this decline occurred with much less sleep restriction (< 1 h per day) than in laboratory studies, suggesting an exacerbating effect of extended wakefulness and/or cumulative fatigue associated with work demands. For both trainees and specialists, robust circadian variation in PVT performance was evident in this complex work setting, despite the potential confounds of variable shift durations and workloads. The relationship between PVT performance of an individual and the safe administration of anaesthesia in the operating theater is unknown. Nevertheless, the findings reinforce that any schedule changes to reduce work-related fatigue need to consider circadian performance variation and the potential transfer of workload and fatigue risk between trainees and specialists.