1. Nociceptive thresholds to noxious mechanical (paw pressure) and thermal (tail flick) stimuli were recorded in conscious rats. The effects of three selective kappa-opioid receptor agonists on the responses to these stimuli were determined following intravenous, intracerebroventricular or intrathecal administration. Results were compared with those obtained with morphine. 2. Following intravenous administration PD117302, U50488, U69593 and morphine produced steep parallel dose-response curves indicating antinociceptive activity when evaluated in the paw pressure test. When U50488 and U69593 were tested at a single dose of 3.3 mg kg-1 no effect was seen in the tail flick test. 3. When given by the intrathecal route only morphine was effective at increasing the nociceptive threshold. PD117302, U50488 and U69593 were without effect in either the paw pressure or tail flick tests when tested at doses up to 100 micrograms per rat. PD117302 caused flaccid paralysis of the hindlimbs following intrathecal administration at the top dose tested. This effect was not reversible by naloxone. 4. All three kappa-opioid receptor agonists produced naloxone-reversible antinociception in the paw pressure test, and to a lesser extent in the tail flick test, when injected directly into the third cerebral ventricle with the maximum effect occurring between 5 and 10 min after administration and declining back to control levels by 60 min. Morphine had a much slower onset of action with the peak effect being observed 30 min after dosing. 5. It is concluded that, under our experimental conditions in the rat, the antinociceptive effects of kappa-agonists are likely to be operated via an action at a supraspinal rather than a spinal site.