Fifty nerve root infiltration studies were evaluated prospectively in a consecutive series of 50 patients referred to the Sheffield Problem Back Clinic with complicated back problems. All were over 35 years of age (mean, 51 years). Ten (20%) demonstrated abnormal segmentation of the lumbar spine. Sixteen (32%) had undergone previous surgery. Before nerve root infiltration all patients were investigated by radiculography and computed tomography. Nerve root infiltration identified two types of response. In 20 patients, infiltration reproduced the symptomatic pain, which then was abolished by local anesthetic--the positive result. These patients were considered suitable for surgery. One patient in this group had spontaneous resolution of pain and thus did not undergo operative treatment. In 30 patients, infiltration did not reproduce the symptomatic pain regardless of the level studied, or only partially reproduced the pain at two or more levels--the negative result. Radiculography and computed tomography in these patients also was inconsistent. This group were considered unsuitable for surgery. In those patients undergoing surgical decompression, nerve root infiltration correctly identified the symptomatic level in 18 of the 19. Computed tomography and radiculography identified the level in 14 and 12 patients, respectively.