Both gamma-aminobutyric acid (GABA)A and GABAB receptor subtypes have been implicated in spinally mediated antinociception in acute pain models. In the current study, the formalin test was used as a model of protracted nociception to examine the effect of intrathecally (i.t.) administered baclofen (GABAB agonist), muscimol (GABAA agonist) or midazolam (a benzodiazepine) on antinociception. At doses that did not affect motor function, baclofen (0.3 and 1.0 micrograms, i.t.) decreased the flinch response in a dose-dependent manner during Phase 1 and Phase 2. This effect was reversible by the GABAB-specific antagonist, CGP35348 ([P-(3-aminopropyl)-P-diethoxymethyl-phosphinic acid]). Muscimol (0.3 and 1.0 microgram i.t.) evoked a dose-dependent, bicuculline-reversible decrease in flinching during Phase 1 and Phase 2, but midazolam had no effect on either phase. No attenuation of the quiescent period between Phase 1 and Phase 2 was seen upon administration of baclofen, muscimol or midazolam. Additionally, no increase in nocifensive behavior was observed upon administration of either GABAA or GABAB antagonists alone. Therefore, our conclusions are that both GABAA and GABAB agonists are antinociceptive at the spinal cord level and that endogenous spinal GABA levels are insufficient for a GABA potentiator to act alone in an antinociceptive manner.