A number of researchers have suggested that the phase (timing) of circadian rhythms in depressed patients is abnormal. Longitudinal studies could help to elucidate the relationship between circadian phase and mood. Such studies would be facilitated by the development of a noninvasive method for measuring circadian phase. In normal volunteers, the concentration of salivary melatonin measurements has been shown to be significantly correlated with those obtained in plasma; however, it is unknown whether salivary melatonin measurements can reliably detect the unmasked time of onset of nocturnal melatonin secretion (a measure of circadian phase). In addition, the relationship between salivary and plasma melatonin measurements in medicated psychiatric patients is unknown. We measured plasma and salivary melatonin simultaneously in a sample of 12 medicated patients with rapid cycling bipolar disorder. The intraclass correlation coefficient between plasma and salivary measures of the dim light melatonin onset (DLMO) was 0.93. We therefore conclude that salivary melatonin can be used to determine the time of the DLMO in this population.