1. The pharmacological profile of GR94839, a kappa-opioid agonist with limited access to the central nervous system, has been investigated. Its antinociceptive activity has been compared with that of GR103545, a centrally-penetrating kappa-agonist and ICI204448, the previously described peripherally-selective kappa-agonist. 2. GR94839 was a potent agonist in the rabbit vas deferens in vitro assay for kappa-opioid receptors (IC50: 1.4 +/- 0.3 nM; n = 6), but had limited activity at mu- or delta-opioid receptors. 3. In the mouse abdominal constriction test, GR94839 was 238 fold more potent when given i.c.v. (ED50: 0.008 (0.004-0.029) mg kg-1; n = 18) than when s.c. (ED50: 1.9 (0.7-3.1) mg kg-1; n = 30). In comparison, GR103545 was equipotent when given i.c.v. or s.c. 4. After intravenous administration, the maximum plasma to brain concentration-ratio attained by GR94839 was 18 compared with 2 for GR85571, a structurally-related kappa-agonist that is centrally-penetrating. 5. GR94839 inhibited the 2nd phase of the rat formalin response at doses 7 fold lower than those required to inhibit the 1st phase (ED50 vs 1st phase: 10.2 (6.7-17.1) mg kg-1, s.c.; ED50 vs 2nd phase: 1.4 (1.0-1.8) mg kg-1, s.c.; n = 18). GR103545 was equipotent against the two phases. 6. Intraplantar administration of the opioid antagonists, norbinaltorphimine (100 micrograms) or naltrexone (1 microgram), reversed the antinociceptive effect of systemic GR94839 (3 mg kg-1, s.c.) against the 2nd phase of the formalin response and intraplantar injection of GR94839 (30-100 micrograms) selectively inhibited the 2nd phase.7. GR94839 and IC1204448 reversed the hyperalgesia in the zymosan-inflamed rat paw at doses (ED50 GR94839: 2.0 (1.1-3.2) mg kg-', s.c.; ED50 IC1204448: 1.2 (0.8-1.7) mg kg-', s.c.), lower than those required to raise the noxious pressure threshold in the non-inflamed paw (EDSO GR94839: 16.4 (8.6-46.7) mg kg', s.c.; ED50 IC1204448: 68.0 (22.1-32000) mg kg', s.c.). GR103545 raised the noxious presure threshold in the inflamed and non-inflamed paws at the same doses.8. GR94839 was sedative in the rat rotarod test (ED50: 35 (12-245) mg kg-', s.c.) at doses higher than those required to inhibit the 2nd phase of the formalin response or reverse hyperalgesia in the zymosan-inflamed rat paw. The doses were comparable to those that inhibited the 1st phase of the formalin response and raised the noxious pressure threshold in the non-inflamed paw.9. The results suggest that GR94839 is a selective kappa-agonist which has antinociceptive activity against inflammatory pain at doses that produce limited central effects. These antinociceptive effects are probably mediated at peripheral opioid receptors.