The variability of overnight urinary albumin excretion rate (AER) and albumin to creatinine ratio was assessed in eight normal subjects and two groups of insulin-dependent diabetic patients divided on the basis of an initial overnight urinary albumin excretion rate below (n = 15) or above (n = 12) 30 micrograms/min. The latter group is known to be at risk of developing clinical diabetic nephropathy. An albumin to creatinine ratio of 2.6 and above identified all patients with an initial albumin excretion rate greater than 30 micrograms/min. The mean of the coefficients of variation, calculated from five successive overnight urine collections, for all subjects was 38% for albumin excretion rate and 37% for albumin to creatinine ratio. There was no significant difference in the variation of albumin excretion rate and albumin to creatinine ratio within or between the groups. Subsequent AERs from diabetics with an initial rate greater than 30 micrograms/min changed category more often (chi 2 = 11.9, p less than 0.001) than those from diabetics with lower initial rates and normal subjects. This was due to four subjects with initial values close to the cut-off level, whose subsequent values varied around it. Albumin excretion rates in normal subjects never exceeded 11 micrograms/min. Whether a patient's risk status is influenced by the degree of variation of albumin excretion rate around a risk level, or whether the classification of risk is improved by multiple collections, awaits testing in prospective subjects.