Seven measurements of albumin excretion in overnight, recumbent (OR) and daytime, ambulant (DA) urine samples were carried out at successive intervals of 3 months in 172 insulin-dependent diabetic patients; at entry into the study, all had a resting albumin excretion rate less than 300 micrograms/min. Urinary albumin excretion in both collections was expressed as a concentration (UA, mg/l), as a creatinine ratio (UA/UC, mg/mmol) and as an excretion rate (UAV, micrograms/min). The pooled within-subject standard deviation (log. (base e) transformed data) for each expression of the albumin excretion was: (1) OR sample--UA 0.6824 mg/l, UA/UC 0.5257 mg/mmol, UAV 0.5940 micrograms/min; (2) DA sample--UA 0.7830 mg/l, UA/UC 0.5780 mg/mmol, UAV 0.6334 micrograms/min. The results were used to calculate the 95% range for a difference between two measurements within an individual patient which was lowest with OR UA/UC (chi/divided by 4.42) and highest with the DA UA (chi/divided by 9.16). Variation in the OR sample was also studied in terms of the patterns of microalbuminuria (M; UAV greater than 15 micrograms/min) which were found to be closely associated with the initial level of albumin excretion: persistent non-M was most common in patients with a baseline UAV less than 15 micrograms/min; established M was most common in those with a baseline UAV greater than 70 micrograms/min, some of whom developed clinical albuminuria. The frequency with which patients should be re-screened for M may be determined by the initial value of albumin excretion and by the threshold used to define M.