Anticonvulsants are widely used in the treatment of neuropathic pain, and are assumed to act preferentially on lancinating, shooting pain. In the present study, the effects of gabapentin, a novel anticonvulsant, were evaluated systematically on both spontaneous and evoked pain in 18 patients with peripheral nerve injuries or central lesions. Gabapentin was administered orally in gradually increasing doses up to a maximum of 2,400 mg/day. Evaluations of spontaneous ongoing and paroxysmal pain, allodynia and hyperalgesia were performed at the beginning of the study ('baseline') and 6 weeks after the steady-state dose had been reached. Quantitative sensory tests were used to measure detection and pain thresholds to mechanical and thermal stimuli and the responses to suprathreshold stimuli. Gabapentin induced a moderate and statistically significant relief of ongoing spontaneous pain and was particularly effective in reducing paroxysmal pain. A striking finding was the significant effect on brush-induced and cold allodynia. In contrast, no effects were observed on detection and pain thresholds to static mechanical and hot stimuli. Side effects were generally minor and did not interfere with everyday activities. The present study suggests that gabapentin has preferential antihyperalgesic and/or antiallodynic effects, and is equally effective in pain due to peripheral nerve injuries and central lesions.