To evaluate intraindividual and between-machine variability in DLCO, one technician performed triplicate measurement of DLCO in eight healthy subjects on three consecutive days using three different systems (Collins [C], Gould [G], SRL [S]) on five separate occasions over the course of 1 yr. For each subject and machine, we calculated the coefficient of variation (CV) of the DLCO and its associated alveolar volume (VA) and inspiratory vital capacity (IVC) over the five different measurement periods using the mean value for each period. CV (in %) for DLCO in each of the eight subjects for each machine varied from 2.4 to 7.1% for machine C, 4.1 to 10.5% for machine G, and 2.5 to 8.7% for machine S. Two-way ANOVA showed significant differences in CV for DLCO and IVC but not for VA across subjects (p less than 0.01) and for DLCO, VA, and IVC across machines (p less than 0.03): CV for DLCO was higher for G than for C or S; CV for VA and IVC was lower for C than for G or S. Multiple regression did not show a significant relationship between within-individual CV for DLCO and that for VA or IVC. We conclude that variability in DLCO measurements is affected by both subject and machine, but, at least in healthy cooperative subjects, it does not appear to be affected by variability in VA or IVC. For the average subject and the two less variable machines we tested, DLCO measurements could vary within the same subject as much as +/- 9.6% (two-tailed 95% confidence interval) with repeated measurements over 1 yr.