Background context: Lumbar nerve root blocks and epidural steroid injections are frequently employed in the management of degenerative conditions of the lumbar spine, but relatively few papers have been published that address the complications associated with these interventions. Serious complications include epidural abscess, arachnoiditis, epidural hematoma, cerebrospinal fluid fistula and hypersensitivity reaction to injectate. Although transient paraparesis has been described after inadvertent intrathecal injection, an immediate and lasting deficit has not been previously described as sequelae of a nerve root block. Purpose: We present three cases in which either persisting paraplegia or paraparesis occurred immediately after administration of a lumbar nerve root block and propose a mechanism for this devastating but previously unreported complication. Study design/setting: Case reports of three patients. Patient sample: Three patients, two women and one man ranging in age from 42 to 64 years, underwent three procedures performed at three different facilities, in the hands of two different injectionists. In each instance, penetration of the dura was not thought to have occurred. In two procedures the needles were placed transforamenally, one at L3–4 on the left and one at L3–4 on the right, and in the third the needle tip was placed immediately lateral to the S1 nerve root. Outcome measures: Patient follow-up data from medical office records. Methods: In each case, needle placement was verified with injection of a contrast media in conjunction with either computerized tomography or biplanar fluoroscopy. No backbleeding or cerebrospinal fluid was encountered upon aspiration in any of the procedures. Magnetic resonance imaging (MRI) was performed within 48 hours of injury in all patients. Results: In each patient, paraplegia suddenly ensued after instillation of the steroid solution and, in each instance, postprocedure MRI revealed increased signal in the low thoracic spinal cord on T2-weighted imaging consistent with edema. The sudden onset of neurological deficit and the imaging changes noted in the spinal cord point to a vascular explanation for these injuries. We postulate that in these patients the spinal needle either penetrated or caused injury to an abnormally low dominant radiculomedullary artery, a recognized anatomical variant. This vessel, also known as the artery of Adamkiewicz, in 85% of individuals arises between T9 and L2, usually from the left, but in a minority of people may arise from the lower lumbar spine and rarely even from as low as S1. The artery of Adamkiewicz travels with the nerve root through the neural foramen and irrigates the anterior spinal artery. Injury of it or injection of particulate matter into it, as what may happen with the commonly used epidural steroid injectates, may result in infarction of the lower thoracic spinal cord, producing the clinical and imaging findings seen in these three patients. Conclusions: We present the cases of three patients who had lasting paraplegia or paraparesis after the performance of a nerve root block. We propose that the mechanism for this rare but devastating complication is the concurrence of two uncommon circumstances, the presence of an unusually low origin of the artery of Adamkiewicz and an undetected intraarterial penetration of the procedure needle.