The hypothesis that exercise limitation of respiratory origin can be predicted accurately from the lung function impairment has been tested using maximal oxygen uptake (VO2max) as the dependent variable in a multiple regression analysis. The subjects were 157 men who met objective criteria for exercise being limited by respiratory impairment. VO2max (mean value, 1.38 L min-1) was described by FEV1 and single-breath lung transfer factor (diffusing capacity) for carbon monoxide (TL') singly or in combination, but the accuracy was poor (at best, standard error of the estimate, 0.36 L min-1; r2, 29.1%). FEV1 could be replaced by FVC and FEV1/FVC. Description of VO2max was improved by also including in the equation the variables age, fat-free mass, and submaximal exercise ventilation (VE). Transfer factor did not then contribute significantly. VO2max as percent of predicted (mean value of 60%) was described by %FVC or %FEV1, but the accuracy was poor (SEE, 16.0%; r2, 14%). Prediction was improved somewhat by the alternative use of inspiratory vital capacity and FEV1/FVC. Transfer factor did not contribute additional information; however, inclusion of VE materially improved the accuracy (SEE, 12.9%; r2, 44%). Among a subgroup of 35 men whose lung disease was due to asbestos, %TL' or transfer factor measured using a multibreath estimate of residual volume (%TLCO) made a small contribution to the explained variance, e.g.: %VO2max = 0.44% FEV1 -0.78 VE + 0.16% TLCO + 52.3 SEE 7.27%. This equation also described the %VO2max of all subjects (SEE, 13%).(ABSTRACT TRUNCATED AT 250 WORDS)