There were 15 healthy female subjects, differing in their position on the "morningness-eveningness" scale, studied for 7 consecutive days, first while living a sedentary lifestyle and sleeping between midnight and 08:00 and then while undergoing a "constant routine." Rectal temperature was measured at regular intervals throughout this time, and the results were subjected to cosinor analysis both before and after "purification" for the effects of physical activity. Results showed that there was a phase difference in the circadian rhythm of core temperature that was associated with the morningness score, with calculations that "morning types" would be phased earlier than "evening types" by up to about 3 h. This difference in phase (which was also statistically significant when the group was divided by a median split into a "morning group" and an "evening group") could not be attributed to effects of waking activity and existed in spite of the subjects keeping the same sleep-wake schedule. Moreover, it persisted when the subjects' data had been purified and when the data were obtained from the constant routine. That is, there was an endogenous component to this difference in phase of the core temperature. The morning group also showed a greater fall of core temperature during sleep; this was assessed in two ways, the main one being a comparison of constant routine and nychthemeral data sets after correction for any effects of activity. Even though the morning group was sleeping at a later phase of their circadian temperature rhythm than was the evening group, neither group showed a fall of temperature due to sleep that varied with time elapsed since the temperature acrophase. It is concluded that another factor that differs between morning and evening types is responsible for this difference.