This preclinical study was designed to evaluate 20-degree lateral cephalometric and posteroanterior cephalometric techniques with digital subtraction enhancement for their ability to detect 1-, 2-, and 3-mm shifts in condylar position in two of three planes of space (horizontal or transverse and vertical). The model used was a dry human skull with a complete mandible mounted on a microscopic stage with positioning accuracy to 0.1 mm in x, y, and z planes. Analysis of variance showed no significant difference from expected values in the transverse plane (posteroanterior cephalogram only). There were significant differences in the sagittal plane (20-degree lateral cephalogram only) for the 2- and 3-mms shifts (P < .001). There were no significant differences among expected values in the vertical plane for individual increments or for increment comparisons. The magnitude of standard deviations and of the absolute value of differences indicates markedly incrased variability for both techniques as compared to tomograms produced with cephalostat repositioning. The authors conclude that even with rpecise repositioning and digital subtraction enhancement, cephalometric radiographs are inadequate for clinical monitoring of condylar position.