Human plasma melatonin concentrations can be measured accurately and sensitively by gas chromatography-negative chemical ionization mass spectrometry. With this assay, we have shown that: in rats and in humans, plasma melatonin is exclusively derived from the pineal gland; propranolol and clonidine reduce melatonin levels in human; some blind people appear to have free-running melatonin secretory circadian rhythms; bright light can acutely suppress human melatonin production according to a linear fluence-response relationship; manic-depressive patients appear to be supersensitive to light, even when they are well; melatonin levels are greater in manic patients than in depressed patients; in experiments to test the clock-gate model and the hypothesized phase-response curve, two different effects of light appear to present in humans: an acute suppressant effect (mainly in the evening during long photoperiods) and an entrainment effect (particularly during the morning but also in the evening). When blood is sampled for measuring melatonin levels as a marker for circadian phase position, bright light should be avoided after 5 p.m. (the dim light melatonin onset). Bright-light exposure in the morning appears to advance circadian rhythms, whereas bright-light exposure in the evening appears to delay them. Once a patient has been 'phase typed' (phase-advanced vs. phase-delayed), predictions can be made about whether morning or evening light would be more effective in treating the sleep or mood disorder.