Carbon dioxide production (rCO2) was estimated in four sheep over a period of 10 days using doubly labeled water (2H and 18O) and was compared with simultaneous respiration chamber measurements of CO2. The excess 2H and 18O measurements were corrected for the empirically determined effects of isotope rebreathing within the confines of the chambers. A weighted monoexponential curve was then fitted to the data from which isotope flux rates and ultimately rCO2 and water turnover (rH2O) estimates were made. The curve fits were weighted assuming a Poisson model. Selection of this weighting policy did not bias the results, and curvature in the data also appeared to have little effect on the rCO2 estimates. Fractionated evaporative water loss expressed as a fraction of rH2O (X) was estimated from water balance and breath water production estimates; the mean X was 0.145 and ranged from 0.108 to 0.183. Corrections for 2H loss in fecal solids reduced the mean rH2O (4,746 g/day) by 35.5 g/day and increased the mean rCO2 (332.3 l/day) by 21.2 l/day. Further corrections to account for 2H loss in methane (mean production rate 27.2 l/day) reduced rH2O by 33.8 g/day and increased rCO2 by 20.3 l/day. The final isotopic estimates of rH2O were 14.6 +/- 6.59% (n = 4) lower than direct measurements and the mean rCO2 was 3.5 +/- 14.48% (n = 4) lower than the chamber measured rCO2. However, in one of the animals studied the rCO2 deviated markedly from the chamber-derived value, and this discrepancy has yet to be explained. When this animal was excluded from the comparisons, the standard deviation was greatly reduced (+/- 3.6, n = 3) and the mean overall error on rCO2 was +3.6%.