The prevalence and variability of ventilation-perfusion (VA/Q) inequality was examined in 26 stable, symptomatic, asthmatic subjects (mean FEV1/FVC, 79% predicted; mean FEF75, 43% predicted) studied once a week for 9 consecutive weeks. We used a recent modification of the multiple inert gas elimination technique allowing frequent serial studies without the need for sampling arterial blood. The VA/Q inequality was expressed as log SD (the second moment) of the distributions of blood flow (Q) and ventilation (V) on a log scale. Log SDQ averaged 0.74, and in every patient log SDQ exceeded the 95% upper limit of normal (0.60) in 2 wk or more. In only 5 patients was mean log SDQ less than 0.6. The ventilation distribution was less abnormal, with mean log SDV exceeding the 95% normal upper limit in only 4 patients. Bimodal blood-flow distributions containing low VA/Q units were observed at some point in 24 of 26 subjects, but occurrence was variable, and in only one third of all measurements was bimodality found. Analysis of variance showed that 70 to 75% of the total variance of log SD was due to intersubject differences, about 20% was due to random changes over time, and the remaining 7 to 9% was not explained by either and was due mostly to experimental error. Arterial PO2 measured 3 times in each subject was inversely related to log SDQ (r = 0.76), but only 60% of the variance in PaO2 was explained by VA/Q mismatch, the rest being due presumably to variation in mixed venous PO2 and similar extrapulmonary factors.(ABSTRACT TRUNCATED AT 250 WORDS)