Eight male subjects were exposed to either bright light or dim light between 0600 and 0900 h for 3 consecutive days each. Relative to the dim light condition, the bright light treatment advanced the evening rise in plasma melatonin and the time of sleep termination (sleep onset was held constant) for an average approximately 1 h. The magnitude of the advance of the plasma melatonin rise was dependent on its phase in dim light. The reduction in sleep duration was at the expense of rapid-eye-movement (REM) sleep. Spectral analysis of the sleep electroencephalogram (EEG) revealed that the advance of the circadian pacemaker did not affect EEG power densities between 0.25 and 15.0 Hz during either non-REM or REM sleep. The data show that shifting the human circadian pacemaker by 1 h does not affect non-REM sleep homeostasis. These findings are in accordance with the predictions of the two-process model of sleep regulation.