Twenty patients with failed back surgery syndrome were analyzed prospectively with MR imaging. In addition, 10 of these patients were analyzed with high-dose contrast-enhanced CT or gadopentetate dimeglumine-enhanced MR imaging. Imaging results were compared with surgical and pathologic findings in all cases. In the 10-patient subset, abnormal epidural soft-tissue specimens were also assessed with light and electron microscopy for vascular density, size of the extracellular space, and collagen orientation and thickness. The average vascular density of epidural fibrosis on light microscopy was found to be 1.19%; the average size of the extracellular space on electron microscopy was 4.29%. Scar 4 months of age or less had a larger extracellular space than did older scar; high- (grade 4 or 5) intensity scar had a larger extracellular space than did less intense scar on long TR/short TE images. Scar 1 year old or less enhanced more intensely on CT than did older scar. The MR signal intensity and CT enhancement characteristics of epidural scar were also found to differ according to epidural location. The percentage of scar that was hyperintense on long TR/TE images was as follows: anterior, 82%; lateral recess, 70%; lateral, 47%; and posterior, 20%. However, no relationship was found between the degree of CT enhancement of scar and vascular density. Gap junction status and extracellular space size, therefore, are more important than vascular density in predicting the degree of enhancement. The accuracy of contrast-enhanced CT and unenhanced MR in separating scar from herniated nucleus pulposus is 80%. This accuracy is related to the partial overlap in imaging characteristics of scar and recurrent herniated nucleus pulposus.