1. The periodicity of activity and hypothermia in Eliomys quercinus was studied in 12 animals exposed to daily lightdark cycles for 18 months continuously, and in 12 animals exposed to continuous total darkness for 5 months in winter. 2. The occurrence of a systematic yearly fluctuation in the mean duration of bouts of torpor, both in conditions of seasonally varying photoperiod or varying ambient temperature and in constant photoperiod and temperature indicates that an endogenous circannual rhythm is involved in the regulation of duration of torpor. 3. Under the experimental conditions, at 12° C, bouts of torpor occur all year round and tend to be short in summer and longer in winter. No sharp onset or end of hibernation can be distinguished in the Garden dormouse. 4. Onset of torpor usually occurs in the middle of the (subjective) night, arousal from torpor in the late (subjective) day. With seasonally varying photoperiod both onset and arousal correlate better with lights out than with lights on. 5. Long bouts of torpor tend to start earlier and to be terminated earlier in the day than short bouts. Synchronization in the experimental conditions was probably restricted to the normothermic phases between bouts of torpor. 6. In the absence of a daily zeitgeber the rhythm of onset of and arousal from hypothermia freeruns with a period differing from 24 hrs. The persistence of the circadian rhythm during hypothermia is proved by peaks about 24-hrs apart in the frequency distribution of lengths of bouts of torpor. 7. Estimates of the circadian period in intervals between successive arousals differed from estimates derived from activity onsets in normothermy. Both significantly negative and significantly positive differences occurred, indicating that torpor may have both slight accelerating and decelerating effects on the circadian clock. 8. The hypothesis is advanced that low body temperature has a phase-delaying effect on the circadian rhythm of heterotherms during the subjective day, and a phase-advancing effect during the subjective night. The implications of such a mechanism for temperature compensation and entrainability by temperature cycles are discussed.