The doubly labelled water technique (DLW) allows the estimation of the carbon dioxide production (r<SUB>co2</SUB>) of a free-living animal from which, with knowledge of the respiratory quotient (RQ), energy expenditure may be predicted. The aim of this project was to assess the performance of this technique in ruminant animals using sheep as a model. In a series of preliminary experiments physiological processes which may affect this technique were examined. It was found the administered isotopes equilibrated throughout the body water after 6 hours following either an oral or intravenous dosing. <SUP>2</SUP>H lost in the form of methane was estimated to lead to errors in r<SUB>co2</SUB> of between -3.27 and -6.54% and may be corrected for by use of a methane to urine enrichment ratio of 0.6536. Losses of <SUP>2</SUP>H in exchangeable sites in faeces were estimated to lead to errors in r<SUB>co2</SUB> of between -5.35 to -6.54%, again corrections factors were proposed. Sequestration of <SUP>2</SUP>H into carcase fat was also quantified, a maximum error of 0.67% was estimated on r<SUB>co2</SUB>, although more <SUP>2</SUP>H was detected in the fat free carcase this was assumed to be exchangeable <SUP>2</SUP>H. A maximum error of -2.39% on water turnover (r<SUB>H20</SUB>) was estimated to arise from the recycling of <SUP>2</SUP>H due to the catabolism of labelled body solids, although this was probably an overestimate. No convincing evidence was found to suggest the fractionation factors applied to human subjects are not equally applicable to ruminant animals. The baseline body water enrichment was altered by a change in diet from synthetic concentrates to fresh chopped grass. Isotopic estimates of r<SUB>co2</SUB> were compared with respiration chamber measurements of r<SUB>co2</SUB>. A number of curve fitting procedures were considered together with three multi-point, three ratio plot and three two-point estimations of r<SUB>co2</SUB>. Using a multi-point calculation and after making corrections for fractional water loss, loss of <SUP>2</SUP>H in methane and faecal solids, r<SUB>H20</SUB> was underestimated by 16.24%. The isotopic r<SUB>co2</SUB> estimates were on average 1.52% (SD 15.31) lower than the chamber measurements, although one animal was markedly different from the others, omitting this animal led to an average error of + 5.95% (SD 4.34). In applying this technique to animals in the field the influence of a free-living existence on the performance of the method was assessed. A number of r<SUB>co2</SUB> calculation procedures were again considered. Preliminary chamber measurements were used to estimate methane production and faecal dry matter output in the field. Estimates of r<SUB>H2O</SUB> in the field (mean 7.218.1. day<SUP>-1</SUP>) were lower than those measured in the chambers (mean 9.812 1.day<SUP>-1</SUP>). From predictions of dry matter intake in the field RQ was estimated (mean 1.067), using the isotopic r<SUB>co2</SUB> estimates (mean 590 1.day<SUP>-1</SUP>) energy expenditure was then predicted (mean 11586 kJ.day<SUP>-1</SUP>). In the light of this work it was concluded the DLW technique may be usefully applied to ruminant animals and a number of practical suggestions were made.