In an attempt to clarify the correlation between the spread of a local anesthetic and a radiographic contrast medium in the epidural space, a lumbar epidural catheter was inserted in 21 cancer patients. In group I, 10 patients, a single dose of bupivacaine 0.5%, 8 ml, was injected. The epidural block was evaluated by assessing loss of pin-prick sensation. The following day a contrast medium, iohexol 180 mg I/ml, 1.2 ml, was injected and the position of the catheter determined. A second dose of iohexol 180 mg I/ml, 8 ml, was injected and epidurograms were taken. The vertebral spread was measured in cephalad and caudad directions from the catheter tip. Group II, 11 patients, underwent the same procedures as group I with the exception of the volumes injected. On the first day they were given bupivacaine 0.5%, 16 ml, and on the second day iohexol 180 mg I/ml, 1.2 ml by 16 ml. In both groups the position of the patients was supine and the speed of injection standardized. In group I there was no statistically significant correlation between the spread of contrast medium and the extension of the epidural block either in the cephalad or in the caudad direction. In group II there was a statistically significant correlation in both the cephalad and caudad direction. It is concluded that epidurography can provide an estimate of the epidural block, but cannot predict the exact segmental distribution of the block.