Thirty patients with failed back surgery syndrome were studied to evaluate the effectiveness of magnetic resonance (MR) imaging with gadolinium-diethylenetriaminepentaacetic acid/dimeglumine (Gd-DTPA) in differentiating postoperative epidural fibrosis (scar) from recurrent disk herniation. Pre- and postcontrast MR images were interpreted without access to other diagnostic, surgical, or pathologic findings. Seventeen patients had surgical and pathologic correlation of the MR findings at 19 disk levels. The precontrast studies had a sensitivity, specificity, and accuracy of 100%, 71%, and 89%, respectively. The enhanced MR studies correctly depicted the character of abnormal epidural soft tissue in 17 patients at all 19 levels. Scar showed heterogeneous enhancement on the early T1-weighted spin-echo images obtained within 10 minutes after contrast material administration. Herniated disk did not show significant enhancement on the early studies but showed variable degrees of enhancement on delayed images in nine of 12 cases. Other criteria were found to be less useful than the pattern of enhancement. Results indicate that precontrast and early postcontrast T1-weighted spin-echo studies are highly accurate in separating epidural fibrosis from herniated disk.