Objective. The purpose of this study was to estimate the inter- and intraobserver agreement for interpreting magnetic resonance (MR) images of the temporomandibular joint (TMJ). Methods. The study was based on MR images of 30 TMJs. The images were interpreted by seven observers for disk configuration, disk position, joint fluid, bone marrow changes, and diagnosis. The observers were not calibrated. Kappa statistics were used. Results. The kappa values were, for interobserver agreement of disk configuration, 0.10; for disk position in the sagittal plane with closed mouth, 0.35; for a combination of closed mouth and open mouth, 0.44; for disk position in the coronal plane, 0.17; for joint fluid 0.36; for bone marrow changes, 0.01; and for diagnosis, 0.39. Intraobserver agreement was generally higher than interobserver agreement. Conclusion. Agreement on disk position in the sagittal plane, on presence and amount of joint fluid, and on diagnosis was fair to moderate. Agreement on disk configuration, on disk position in the coronal plane, and on bone marrow changes was poor.