After nerve injury, a sympathetically dependent allodynia may be observed. Spinal alpha-2 agonists inhibit preganglionic neurons. The nature of the effect of alpha-2 agonists on allodynia induced by L5 and L6 nerve ligation (Chung model) was thus examined. Rats were implanted with spinal intrathecal catheters directed at the upper lumbar (L1-L2) or the lower cervical (C5-C6) spinal levels. After nerve injury, rats displayed a tactile allodynia (mean withdrawal thresholds, < 2.3 g). Lumbar intrathecal injection of alpha-2, but not alpha-1 or an opiate agonist, resulted in a dose-dependent reversal of the allodynia, with the ordering of activity (ED50 in micrograms in parentheses) being dexmedetomidine (0.9) > oxymetazoline (14) = guanfacine (17) = 5-bromo-6-(2-imidazolin-2-ylamino)quinoxaline (19) = 2-(2,6-diethylphenylamino)-2-imidazoline (21) = clonidine (22) > morphine (> 30) = methoxamine (> 30 micrograms). The effects of clonidine and 2-(2,6-diethylphenylamino)-2-imidazoline were reversed by the intrathecal injection of yohimbine. At equivalent doses, clonidine delivered systemically, i.c.v. by chronic ventricular guides, or at the level of the cervical spinal cord, produced substantially less antiallodynic action. These results jointly suggest that the sites of the antiallodynic action of spinal alpha-2 agonists are located at the level of the spinal preganglionic neurons and correspond to their ability to diminish preganglionic sympathetic outflow. The failure of morphine to exert an antiallodynic action reflects the fact that 1) opiates act presynaptically on small primary afferents and the allodynia is mediated by large afferent input and 2) opiates, unlike alpha-2 agonists do not have an effect on autonomic outflow.