Three imaging modalities were compared for diagnostic efficacy in the detection of simulated osteoarthritic lesions created by a series of dentin chips placed at four locations on the head of a temporomandibular joint condyle in a dry human skull. The modalities included panoramic-based biplanar temporomandibular joint views, parasagittal linear zonography, and sagittal-view, computer-controlled, multidirectional tomography. Equal numbers of images with and without chips were viewed by six examiners. Observers recorded the presense or absence of lesions on a 100-point confidence scale. Responses were used to construct ROC curves. Pooled data sorte by modiality produced areas under the curves of 0.7421(Zonography), 0.6627 (panoramic imaging), and 0.7478(tomography). No statistially significant difference in lesion detection was found among the three modalities, between observers, or within observers.