Whether or not the pineal gland has a significant physiological role in humans is not known. There has nevertheless been speculation about the potential therapeutic use of melatonin (in view of its hypnotic and possible zeitgeber properties) in conditions such as insomnia and jet lag, and in shift-workers. Our work concerns the effects of melatonin administration in humans and the interactions between melatonin and other circadian variables. Chronic (one month), timed (1700 h), low-dose (2 mg daily) melatonin administration to normal subjects without environmental control consistently increased evening fatigue and slightly modified the 24 h prolactin rhythm without effect on cortisol, growth hormone, luteinizing hormone, thyroxine, testosterone or self-rated mood. In five out of 11 subjects the endogenous melatonin rhythm was advanced by one to three hours. During fractional desynchronization of circadian rhythms by increasing imposed 'day' length (26-29 h, 24 days, 500 lux), 5 mg melatonin per os at lights-out in two subjects resulted in better entrainment of the fatigue rhythm to the zeitgeber than in five out of six control subjects, without major consistent effects on other measured circadian variables. Using a new radioimmunoassay for 6-hydroxymelatonin sulphate (aMT6s), the major melatonin metabolite, we have shown that the urinary aMT6s rhythm is closely correlated to that of melatonin in plasma and is completely suppressed by an acute dose of atenolol (100 mg per os), a peripheral beta-adrenergic antagonist. During fractional desynchronization by increasing imposed 'day' length in one subject and decreasing imposed 'day' length in two subjects, the urinary aMT6s rhythm behaved similarly to that of core temperature. The results suggest that fatigue (or alertness) may be entrained by melatonin, but whether critical performance rhythms can be suitably manipulated remains to be clarified. It is likely that melatonin production is linked to the so-called 'strong' circadian oscillator.