Assessment of the significance of an observed set of serum chemical values for determining a person's state of health requires comparison with a set of defined reference values. We tested the assumption that a reference group of individuals, categorized by age and sex, gives a narrower range of variation than does a larger mixed population. If this were true, the demographic set would be a more sensitive reference than is the customary "normal range" for interpretation of values occurring in the individual. The ratio, R, of intra-personal to inter-personal (group) standard deviations was similar for defined age/sex classes and the overall group for 16 serum constituents. When the "raw" intra-individual variation (biological plus analytical variation) was adjusted to remove the average analytical component, the resulting R was less than 80 for all constituents except creatine kinase, which indicates that these are all particularly strong "discriminators" of individuality. These results imply the need for individual rather than population-based reference ranges, even if the latter are from persons of similar age and the same sex.