In 1979, 304 healthy elderly individuals in New Mexico were recruited for a longitudinal study of nutrition and aging. Repeat measurements on a yearly basis of commonly requested clinical chemistry analytes allowed the calculation of reference intervals, between and within-subject variance components, and percentiles for change in concentration between two yearly measurements. The latter was further divided into analytical and biological variance components. The upper 95th percentile for the difference between two yearly measurements, expressed as a percent of the population mean, ranged from 4% for Na+ to approximately 20% for total cholesterol and to greater than 90% for ferritin. Year-to-year differences attributable to the biological component ranged from a low of 2% of the population mean for Na+ to 70% for gamma-glutamyltransferase.