To examine the usefulness of exercise testing in impairment evaluation, we reviewed the evaluation of 348 asbestos-exposed shipyard workers. We compared work capacity predicted from history, physical examination, chest roentgenogram, resting electrocardiogram, and resting pulmonary function tests with measured work capacity during an incremental cycle exercise test. The predicted work capacity was often incorrect when compared with measured maximal oxygen uptake (VO2). One third (22 of 66) of those predicted to have reduced work capacity had normal measured work capacity, and 46 of 148 workers (31%) predicted to have normal work capacity were found to have low maximal VO2 during exercise. Of 134 men for whom predicted work capacity was uncertain, maximal VO2 during exercise was low in 49 (37%), normal in 81 (60%), and remained indeterminate in 4 (3%). Thus, of the 138 workers who had low measured VO2, 43 were correctly predicted to have normal work capacity, 46 were incorrectly predicted, and the prediction was uncertain in 49. Only a few were limited by respiratory disease, and cardiovascular disorders limited 69% of those with a low maximal VO2 during exercise. Accordingly, resting VC, FEV1, and DLCO had a poor correlation with exercise performance. Finally, we found that resting DLCO was a poor predictor of abnormal exercise AaPO2, dead-space/tidal volume ratio, or arterial end-tidal PCO2 difference. We conclude that exercise testing is needed for accurate work capacity assessment in impairment evaluation. Exercise testing also facilitates the identification of the major limiting system in those with low work capacity.