The doubly labeled water (DLW) method is an isotope-based technique for the estimation of the CO(2) production, and hence energy expenditure, of free-living animals and humans. Several methods are available for the calculation of CO(2) production from the isotope fluxes, depending on different assumptions about the behavior of isotopes during the elimination process. We used the DLW method to estimate the daily energy expenditures (DEE) of 55 field voles (Microtus agrestis) held in a captive facility at 8 degrees C. We calculated DEE using both plateau and intercept approaches for estimating the sizes of the isotope dilution spaces, three different assumptions about fractionation processes, and two ways of treating the different dilution spaces of the oxygen and hydrogen isotopes. We compared the resultant DEE estimates with metabolizable energy intake (MEI) measured during a 3-d feeding trial immediately before the DLW measurements, during which the animals were in energy balance. By making different assumptions about the apparent energy absorption efficiency, we generated a range of direct estimates of MEI. When we compared DEE and MEI, we found that the two-pool model formulations consistently underestimated energy demands by up to 29.8%, depending on the assumptions made in the reference calculation. However, while our data suggest that some correction for fractionation is necessary, with the present data we were unable to separate the two most common treatments of fractionation. These data strongly support the previous suggestion that for small mammals single-pool models provide more accurate estimates of energy demands than two-pool formulation of the DLW method.