The nutrient balancing hypothesis proposes that, when sufficient food is available, the primary goal of animal diet selection is to obtain a nutritionally balanced diet. This hypothesis can be tested using the Geometric Framework for nutrition (GF). The GF enables researchers to study patterns of nutrient intake (e.g. macronutrients; protein, carbohydrates, fat), interactions between the different nutrients, and how an animal resolves the potential conflict between over-eating one or more nutrients and under-eating others during periods of dietary imbalance. Using the moose (Alces alces L.), a model species in the development of herbivore foraging theory, we conducted a feeding experiment guided by the GF, combining continuous observations of six captive moose with analysis of the macronutritional composition of foods. We identified the moose's self-selected macronutrient target by allowing them to compose a diet by mixing two nutritionally complementary pellet types plus limited access to Salix browse. Such periods of free choice were intermixed with periods when they were restricted to one of the two pellet types plus Salix browse. Our observations of food intake by moose given free choice lend support to the nutrient balancing hypothesis, as the moose combined the foods in specific proportions that provided a particular ratio and amount of macronutrients. When restricted to either of two diets comprising a single pellet type, the moose i) maintained a relatively stable intake of non-protein energy while allowing protein intakes to vary with food composition, and ii) increased their intake of the food item that most closely resembled the self-selected macronutrient intake from the free choice periods, namely Salix browse. We place our results in the context of the nutritional strategy of the moose, ruminant physiology and the categorization of food quality.