An automated complete blood count with white blood cell differential was performed yearly on successive groups of healthy second-year medical students from 1979 through 1987. For three classes (1984-1987), the counts were repeated on the same people nine months later. These data demonstrated that the mean value of all hematologic parameters was quite stable over nine years. This allowed for an estimation of the upper limit for the combined effects of drift in accuracy, precision, and biologic stability. The stability was achievable despite an evolution in technology and quality assurance methods over that period. A comparison of intraindividual versus interindividual variation demonstrated that a normal range based on population statistics may be less sensitive than a normal range established for a person during routine health maintenance, especially for the platelet count.