Previously, we reported that gabapentin, a nonselective gamma-aminobutyric acid agonist, reduced 'positive' subjective effects of cocaine without reducing cocaine self-administration. We speculated that the self-administration procedure used in that study was not sensitive to subtle shifts in the reinforcing effects of cocaine. Thus, this study examined the effects of gabapentin maintenance on cocaine self-administration using a purchase-cocaine choice procedure. During this 48-day inpatient/outpatient study, nontreatment-seeking cocaine abusers (n = 12) were maintained on gabapentin (0, 600, 1200 mg/day); four doses of cocaine (0, 12, 25, 50 mg) were each tested twice under each gabapentin condition. All cocaine testing was conducted while participants were inpatients. Before the start of each session, participants were provided with 25 dollars (five 5 dollar bills, one for each choice opportunity) and smoked the 'sample' cocaine dose once. Subsequently, participants were given five opportunities to purchase the sampled dose of cocaine (at 5 dollars per dose) or to keep 5 dollars for that choice trial. Choice to self-administer cocaine increased significantly with escalating cocaine doses; gabapentin maintenance did not alter choice to self-administer cocaine. These results concur with findings from our previous investigations of gabapentin and with those from a clinical trial examining the effects of larger gabapentin doses on cocaine use by treatment-seeking cocaine-dependent individuals. Together, the data indicate that gabapentin does not show promise as a treatment medication for cocaine dependence.