IntroductionWinter presents a severe metabolic challenge to homeotherms. Just as food availability drops to its annual minimum, low ambient temperatures (T a ) increase energy requirements. Many mammals escape this energy bottleneck by hibernating, whereby they reduce body temperature (T b ) usually below 10°C and metabolic rate to a fraction of that of normothermic individuals. The suppression of metabolic rate through torpor has a clear adaptive value when feeding opportunities are sporadic or non-existent and fat reserves are limited. Hibernators, however, do not exploit the energy savings of torpor to the maximum; all hibernators studied to date arouse periodically (Lyman et al. 1982). Although they account for > 75% of the total energy requirement for hibernation (Kayser 1953;Wang 1978;Thomas, Dorais & Bergeron 1990), arousals remain among the least understood of hibernation phenomena.Although numerous factors have been proposed to account for the occurrence and frequency of arousals, these can be broadly grouped under three principal Summary 1. Using existing data on the rate of cutaneous and pulmonary evaporative water loss (EWL) for hibernating Little Brown Bats (Myotis lucifugus) and on the duration of torpor bouts, body temperature (T b ) and oxygen uptake (V O 2 ) of Golden-mantled Ground Squirrels (Spermophilus saturatus), the rate of EWL was modelled for ground squirrels hibernating at ambient temperatures (T a ) of -2, 2, 4 and 8°C. 2. Total EWL showed a curvilinear response to T a , being lowest at 2°C and increasing with both increasing and decreasing T a . EWL at -2°C was about equal to that at 4°C. The duration of torpor bouts showed the same curvilinear response to T a and torpor bout duration at -2°C was similar to that at 4°C (8·5 vs 8·3 days, respectively). 3. At T a ≥ 2°C, where T b of torpid S. saturatus is not metabolically defended, torpor bout duration is significantly related to T b , V O 2 and EWL, with the three variables having similar r 2 values. 4. Using the regression equations generated at T a ≥ 2°C to predict torpor bout durations at -2°C, where T b is metabolically defended, shows that the three variables do not have equivalent predictive abilities. T b and V O 2 predicted torpor bout durations of 15·2 and -40·4 days, respectively, compared with observed durations of 8·5 days at -2°C. EWL predicted torpor bout durations of 8·4 days or only 0·1 days less than that observed at -2°C. 5. The relation between torpor bout duration and total EWL was insensitive to major variations in cutaneous EWL. Over T a ranging from -2°C to 8°C, a stepwise multiple regression including T b , V O 2 and EWL as independent variables identified EWL as the only variable significantly correlated with torpor bout duration. 6. Our analyses suggest that torpor bout duration may be influenced by EWL, indicating that animals may need to obtain free water when they arouse. An analysis of the structural and temperature characteristics of ground squirrel hibernacula suggests that they may function as a biolog...