This investigation is a further study of the long-term effects of a weekly rotating day-night shift schedule on circadian rhythms. It was performed 3 a after the introduction of an increase in nighttime work. Six male typesetters were observed in connection with the beginning and end of the nightshift week and in connection with their free days in the "day week." Measurements were taken of body temperature during both the awake and sleep periods and of self-rated activation. For the same subjects a comparison was also made of variations in awake temperatures and self-related activation after one month, one year and three years on the new shift schedule. Results showed that after 3 a of experience the awake temperature curves had a flattened shape in connection with both the beginning and end of the nightshift week and also during the free days. During sleep there was however a short-term adjustment of the temperature curves within the nightshift week. Comparisons over years revealed significant changes in the course of variation in body temperature during both night and day work, while no significant changes were found for self-related activation. The possibility that experience has a differential influence upon the phase adjustment of temperature maxima compared to minima was discussed.