The conventional regression method for partitioning heat production (HP) in growing animals between HP associated with either maintenance or growth assumes maintenance HP to be independent of feeding level (FL). However, there are indications that this assumption is not correct and an alternative method is proposed in this study from a reanalysis of 3 trials. In trial 1, 73-, 152-, and 237-kg calves received one milk replacer at 77, 84, 92, and 100% of their ad libitum metabolizable energy (ME) intake. In trial 2, 70-kg barrows received one diet at 60, 80, and 100% of their ad libitum ME intake {2600 kJ ME/[kg body weight (BW)(0.60) ⋅ d]}. In trial 3, 60-kg barrows received a basal diet [1700 kJ ME/(kg BW(0.60) ⋅ d)] or 4 diets consisting of the basal diet plus 850 kJ ME/(kg BW(0.60)⋅d) of starch alone or starch with corn gluten, casein, or vegetable oil. In the 3 trials (n = 48, 18, and 28, respectively), HP and activity-related HP were measured on individuals pigs and calves in respiration chambers for 6 d (fed state) and fasting HP (FHP; at zero activity) was calculated as the asymptotic value of HP kinetics on d 7 (feed-deprived state). The FHP changed by 0.22 kJ in calves and 0.14 kJ in pigs/kJ ME intake change during the previous days. The efficiency of using ME for maintenance and growth [k(mg); 1- (HP - FHP)/ME] was not affected by FL (calves: 84%, pigs in trial 2: 74%). In trial 3, k(mg) varied between diets in connection with variations in efficiencies between nutrients (from 55% for corn gluten to 85% for lipid). This new method of representing partitioning of ME intake considers FHP as variable with FL, does not require estimates of maintenance ME requirements, includes efficiencies that depend on diet characteristics, and is not biased by metabolic adaptations of the animal to FL.