The long-term precision of chromium-51 ethylenediamine tetraacetic acid (51Cr-EDTA) measurements of glomerular filtration rate (GFR) was evaluated in a retrospective study of data obtained over a 12 year period. Each GFR measurement was derived from plasma samples taken at 2, 3 and 4 h following injection of 3 MBq 51Cr-EDTA. The records of 7507 patients were reviewed, from which 55 subjects were identified as having had studies on 10 or more occasions. The mean number of studies per patient was 12.9 (range 10-23) over a mean period of 9.4 years (range 4.3-11.8 years). Plots of GFR, clearance half-life, (T1/2) and volume of distribution (VD) were drawn for each patient and used to identify subjects showing linear changes with time that could be fitted using linear regression. Each residual was expressed as a percentage of the expected value calculated from the regression line and all the residuals combined to give histograms for GFR, T1/2 and VD. Each histogram was fitted with a normal distribution between the -3 S.D. and +3 S.D. limits using weighted least squares. Final results for the coefficient of variation were: GFR 9.8%, T1/2 6.7%, VD 9.4%. The precision errors were used to calculate the smallest statistically significant change measurable by the 51Cr-EDTA technique. With 10% significance and 80% power, the smallest measureable change was 30% for GFR and 20% for T1/2. Unless there are clinical grounds for thinking that a patient's volume of distribution has changed, T1/2 is the optimal parameter for identifying real changes in renal function.