1. Nitrogen balance, weight loss and resting metabolic rate were measured in thirty-eight obese inpatients on 3.4 MJ (800 kcal)/d diets over 3 weeks. 2. All subjects were fed on 13% protein-energy in three meals/d for the first week. 3. In weeks 2 or 3, using a cross-over design, ten subjects were fed on 15 or 10% protein-energy as three meals/d; fourteen subjects were fed on five or one meal/d with 13% protein-energy; and fourteen subjects were fed on 15% protein-energy as five meals/d or 10% protein-energy as one meal/d. 4. N loss was least on the high-protein week and frequent-meal week: the largest difference was found when these effects were combined (P less than 0.001). 5. When protein-energy was held constant at 13% N loss decreased significantly (P less than 0.01) between week 2 and 3, but when the protein-energy was manipulated there was no significant N conservation in the third week. This suggests that the protein:energy value is more important than meal frequency in the preservation of lean tissue. 6. Weight loss was also least on the 'high-protein' week and 'frequent-meal' week, but this result reached significance only when the effects were combined (P less than 0.05). 7. Resting metabolic rate decreased with time but was not significantly altered by the dietary regimens. 8. Therefore, during the first 3 weeks at an intake of 3.4 MJ/d, a diet with a high-protein concentration, fed as frequent small meals, is associated with better preservation of lean tissue than an isoenergetic diet with lower-protein concentration fed as fewer meals. There was no evidence that meal frequency or protein concentration affect the rate of fat loss.