Thirty-eight patients undergoing lobectomy or pneumonectomy for carcinoma had preoperative computed tomography (CT) of the chest. Twenty-seven had both 1.5 mm and 10 mm collimation scans, and eleven had 10 mm collimation images only. These images were analyzed for the extent and severity of emphysema, and the analysis compared to the pathologic findings in the corresponding transverse slice of lung. The latter was graded by a modification of a panel of standards and by a grid system numerically expressing extent and severity. The grid system is theoretically superior to the panel of standards because it allows better quantitation of early emphysema and, contrary to the set of standards, is designed to analyze transverse CT images and corresponding pathologic slices. There was good correlation between the CT score and the pathologic score using the panel of standards (r = 0.81, p less than 0.001) but a lower correlation with the grid system (r = 0.70, p less than 0.001). The correlation improved slightly with 1.5 as compared to 10 mm collimation scans. Close comparison between the CT and grid scores showed that CT was sensitive in demonstrating early distal acinar and irregular emphysema. However, CT consistently underestimated the extent of centriacinar and panacinar emphysema because most lesions less than 0.5 cm in diameter were missed. We conclude that CT is insensitive in detecting the earliest lesions of emphysema.