Circadian rhythms of core body temperature and melatonin are commonly used as phase markers of the circadian clock. Melatonin is a more stable marker of circadian phase when measured under constant routine conditions. However, little is known about the variability of these phase markers under less controlled conditions. Moreover, there is little consensus about the preferred method of analysis. The objective of this study was to assess various methods of calculating melatonin and temperature phase in subjects with regular sleep schedules living in their natural environment. Baseline data were analyzed from 42 healthy young subjects who were studied on at least two occasions. Each hospital admission was separated by at least 3 weeks. Subjects were instructed to maintain a regular sleep schedule, which was monitored for 1 week before admission by sleep logs and actigraphy. Subjects spent one habituation night under controlled conditions prior to collecting baseline temperature and melatonin measurements. The phase of the melatonin rhythm was assessed by 9 different methods. The temperature nadir (Tmin) was estimated using both Cleveland and Cosine curve fitting procedures, with and without demasking. Variability between admissions was assessed by correlation analysis and by the mean absolute difference in timing of the phase estimates. The relationship to sleep times was assessed by correlation of sleep onset or sleep offset with the various phase markers. Melatonin phase markers were more stable and more highly correlated with the timing of sleep than estimates of Tmin. Of the methods for estimating Tmin, simple cosine analysis was the least variable. In addition, sleep offset was more strongly correlated with the various phase markers than sleep onset. The relative measures of melatonin offset had the highest correlation coefficients, the lowest study-to-study variability, and were more strongly associated with sleep timing than melatonin onsets. Concordance of the methods of analysis suggests a tendency for the declining phase of the melatonin profile to be more stable and reliable than either markers of melatonin onset or measures of the termination of melatonin synthesis.