Ten nurses have been studied for a period of about ten days during which they had rest days and between two and seven successive night shifts. Rectal temperature and wrist movement were monitored throughout, and the nurses kept an activity log. The process of adjustment to night work was assessed by comparing shifts in body temperature with those in mid-sleep, using the mean of the values during rest days as a control in both cases. When raw temperature data were used, the measured shifts were significantly greater than when the temperature data were first modified or 'purified' by a series of methods that took into account the direct effects upon temperature of sleep and physical activity. The greater the 'purification' of the data--whether by using an activity profile derived from a wrist accelerometer or the subject's own log--the less rapid did adjustment to night work appear to be. We conclude that conventional measurements of the adjustment of the temperature rhythm to night work that make use of raw data over-estimate it due to masking effects. We suggest also that the purification methods that we have developed can be used in the field and enable estimates to be made of the size of masking effects caused by sleep and activity, as well as the shift of the endogenous component of the circadian rhythm.