There are few pharmacological studies of central neuronal measures in animal models of neuropathic pain. In the present study we have compared the effects of two anticonvulsants, carbamazepine and gabapentin, on spinal neuronal responses of nerve injured rats (selective ligation of spinal nerves L5 and L6, SNL) and sham-operated rats. The development and maintenance of cooling and mechanical allodynia of the lesioned hindlimb of SNL rats was followed with behavioural indices. The contralateral hindlimb of SNL rats and the ipsilateral hindlimb of sham-operated rats did not develop allodynia. Electrophysiological studies of SNL rats were then performed at two post-operative (PO) time points (PO days 7-10 and PO days 14-17). Spinal neurones of SNL rats, but not sham-operated rats, exhibited spontaneous activity at both PO days 7-10 and 14-17 (1 +/- 0.4 and 3 +/- 1 Hz, respectively). Paradoxically, the magnitude of electrical (C-fibre) and natural (mechanical and thermal) evoked neuronal responses of SNL rats at PO days 14-17 were smaller than the evoked neuronal responses of SNL rats at PO days 7-10 and sham-operated rats. The electrical evoked A-fibre responses of neurones were comparable for the three groups of rats. Both subcutaneous carbamazepine (0.5-22.5 mg/kg) and gabapentin (10-100 mg/kg) significantly reduced the spontaneous activity of spinal neurones of SNL rats at both PO time points. Carbamazepine had inhibitory effects on electrical C- and A-fibre and mechanical punctate (9 and 50 g) evoked neuronal responses of SNL rats which were significantly different to the lack of effect of carbamazepine on these measures in sham-operated rats. Gabapentin had comparable effects as carbamazepine on the electrical C-and A-fibre and mechanical punctate (9 and 50 g) evoked neuronal responses of SNL rats. In contrast to carbamazepine, gabapentin also reduced evoked neuronal responses of sham-operated rats and there was no difference between the effects of gabapentin in SNL and sham-operated rats. Robust behavioural changes in the SNL model of neuropathy are paralleled by a temporal increase in spontaneous activity and a paradoxical decrease in evoked spinal neuronal responses. The peripheral nerve dysfunction reveals an effect of carbamazepine which is maintained throughout the observation period, validating this experimental approach. Gabapentin, a novel treatment for neuropathic pain states, also reduced neuronal responses, but the actions of the drug were not dependent on nerve injury. Further studies at the spinal level may shed light on the physiology and pharmacology of the aberrant processes associated with neuropathic pain.