Objective. To verify whether magnetic resonance imaging (MRI)-proven joint injury is sensitive as compared with joint injury determined by physical examination. Methods. MRI of the wrist and finger joints of both hands was examined in 51 early-stage rheumatoid arthritis (RA) patients by both plain and gadolinium diethylenetriaminepentaacetic acid-enhanced MRI. Synovitis, bone edema, and bone erosion (the latter two included as bone lesions at the wrist joints); metacarpophalangeal joints; and proximal interphalangeal joints were considered as MRI-proven joint injury. Japan College of Rheumatology-certified rheumatologists had given a physical examination just before the MRI study. The presence of tender and/or swollen joints in the same fields as MRI was considered as joint injury on physical examination. The association of MRI-proven joint injury with physical examination-proven joint injury was examined. Results. A total of 1,110 sites were available to be examined. MRI-proven joint injury was found in 521 sites, whereas the other 589 sites were normal. Physical examination-proven joint injury was found in 305 sites, which was significantly low as compared with MRI-proven joint injury (P ؍ 1.1 ؋ 10 ؊12 versus MRI). Joint injury on physical examination was not found in 81.5% of the sites where MRI findings were normal. Furthermore, an association of the severity of MRI-proven joint injury with that of joint injury on physical examination was clearly demonstrated (P ؍ 1.6 ؋ 10 ؊15 , r s ؍ 0.469). Conclusion. Our present data suggest that MRI is not only sensitive but accurately reflects the joint injury in patients with early-stage RA.