Twelve non-lactating Boer goat does with live masses ranging between 20 and 60 kg and of different body fatness were used to develop equations predicting body moisture, protein, ash and fat from tritiated water space. Calculated coefficients of variation, with the standard deviation obtained from deviations from regression showed that body moisture, protein, ash and fat could be predicted with an accuracy of 1,8%,7,5%, 7,3% and 7,4% respectively. Combining body mass and tritium space in a multiple linear regression increased the accuracy of predicting body fat substantially. Applying prediction equations derived for sheep could underestimate body fat in Boer goat does.