Fifty patients who had patellar subluxation and thirty control subjects were examined using axial roentgenograms of the patellofemoral joint that were made with the knee in 30 and 45 degrees of flexion, as well as computed tomography scans that were made with the knee in full extension. The amount of lateral patellar tilt was quantitatively assessed using the lateral patellofemoral angle, as described by Laurin et al., and the congruence angle, as described by Merchant et al. In both the control subjects and the patients, the angle of Laurin et al. changed significantly when the knee was flexed from full extension to 30 degrees. The difference between the groups was statistically significant at each angle of flexion of the knee, and the difference between the groups was most prominent on the computed tomography scans that were made with the knee in full extension (p less than 0.001). In the patients, the average congruence angle (as described by Merchant et al.) was 5 degrees and in the control subjects, -10 degrees. This indicated that, in our patients, the extent of the patellar subluxation was less than that in previously reported series, and, as a result, the sensitivity of the congruence angle in diagnosing patellar subluxation was only 0.30. In contrast, the sensitivity and specificity of the computed tomography scans for diagnosing patellar subluxation were 0.96 and 0.90, respectively--that is, they were higher than the values that were obtained using any axial roentgenograms. Thus, our results indicated that patellar subluxation can be detected more accurately by using computed tomography with the knee in full extension than by using conventional axial roentgenograms.