Human subjects vary in the extent to which their body's protein and fat compartments are mobilized for fuel during starvation. Although an inverse association between the initial adiposity and the contribution of protein as fuel during starvation has been known for nearly a century, interest in the quantitative importance and functional significance of the initial percentage fat as a determinant of biological variation in energy-partitioning between protein and fat (and hence in determining the partitioning characteristic of the individual) is relatively recent. The present paper addresses these issues by revisiting the classic Minnesota experiment of semi-starvation and refeeding from a standpoint of system physiology. In a quantitative analysis of the relationship between the initial body composition (ration FAT0: fat-free mass (FFM)0) and the composition of weight loss (ratio delta FAT: delta FFM) in the thirty-two men in the Minnesota study, the arguments are put forward that the fraction of FFM lost when the fat stores reach total depletion is independent of the initial percentage fat, and that this fraction represents the 'dispensable' component of the protein compartment that is compatible with life (i.e. the protein energy-reserve, rp). The concepts are developed that (1) the initial percentage body fat (which reflects the initial ratio FAT0:FFM0) provides a 'memory of partitioning' which dictates the control of partitioning between protein and fat in such a way that both the protein energy-reserve (rp) and the fat energy-reserve (rf) each complete depletion simultaneously, a strategy that would ensure maximum length of survival during long-term food scarcity, and that (2) variability in the relative sizes of these two energy reserves (i.e. in rf:rp) could, in addition to the initial percentage fat, also contribute to human variability in energy-partitioning. The basic assumptions underlying this re-analysis of the Minnesota data, and the concepts that are derived from it, have been integrated in the simple mathematical model for predicting the partitioning characteristic of the individual. This model is used to explain how variability in the fraction of the protein compartment that could function as an energy reserve (rp) can be as important as the initial percentage fat in determining inter-individual variability in protein-sparing during the early phase of starvation, in fuel partitioning during prolonged starvation, or in the maximum percentage weight loss during starvation. The elucidation of factors underlying variability in the size of the protein energy-reserve may have important implications for our understanding of the pathophysiology of starvation and age-associated susceptibility to muscle wasting, and in the clinical management of cachexia and obesity.