Here, we present results from two foraging studies with the bumblebee Bombus terrestris L. We used the doubly labelled water technique which makes it possible to estimate energetic costs of free flight in a natural environment. One study was carried out in windy weather in the open, the second in a large greenhouse under controlled conditions. Individual mass-specific metabolic rates from the open-air study varied widely (159.5 to 750.2 W kg-1; n = 7) and differed significantly from expected values predicted from laboratory data. Results from the greenhouse study were closer to expectations and the range of individual metabolic rates was much narrower (405.0 to 485.5 W kg-1; n = 8). The range of field metabolic rates was three to four times larger than that of laboratory studies (Heinrich, 1975; Ellington et al., 1990; Cooper, 1993), which might reflect the fact that only a relatively small percentage (25 – 30 %) of bees are ‘co-operative’ in laboratory flight cost measurements (Hanauer-Thieser and Nachtigall, 1995; C.P. Ellington and T.J. Wolf, pers. observations). However, all bees flew in the field studies, even in strong winds. We conclude that estimates of energetic costs of free flight using laboratory data (time/activity/laboratory = TAL), an approach regularly used in tests of foraging models, might not be appropriate under many natural conditions. Whereas the TAL method provided a good estimate of mean metabolic rate it is a poor method for studies in which individual variations are of interest.