Metabolic rate has been linked to several components of fitness and is both heritable and repeatable to a certain extent. However, its repeatability can differ among studies, even after controlling for the time interval between measurements. Some of this variation in repeatability might be due to the relative stability of the environmental conditions in which the animals are living between measurements. We compared published repeatability estimates for basal, resting and maximum metabolic rate from studies of endotherms living in the laboratory with those living in the wild during the interval between measurements. We found that repeatability declines over time, as demonstrated previously, but show for the first time that estimates from free-living animals are also considerably lower than those from animals living under more stable laboratory conditions.