We tested the hypothesis that adenosine excites nociceptive primary afferents innervating the knee joint. Neuronal recordings were made from fine nerve filaments innervating the knee joint in rats anaesthetized with pentobarbitone. Drugs were injected close-arterially (i.a.) or into the articular space (i.art.). We studied normal and chronically inflamed arthritic joints, the latter 14-21 days after a single intra-articular injection of Freund's Complete Adjuvant, performed under halothane anaesthesia. Adenosine injected i.a. caused delayed (approximately 10 s) excitation of the majority of polymodal C-fibre afferents, and had similar effects when injected directly into the joint. Adenosine triphosphate (ATP) had biphasic effects on discharge, a fast (<1 s) excitation was followed by a delayed increase similar to that seen with adenosine. The adenosine A1 receptor agonists N6-cyclopentyladenosine (CPA) and N-[(1S,trans)-2-hydroxypentyl] adenosine (GR79236) also excited the C-fibre afferents. The A1 antagonist 8-cyclopentyl-1,3-dipropylxanthine (DPCPX) antagonized the responses evoked by adenosine, CPA, and the delayed increase seen after ATP, indicating that excitation of the nociceptive afferents was mediated via adenosine A1 receptors. Adenosine and ATP evoked delayed excitatory effects of similar magnitude, regardless of whether or not the knee joint was chronically inflamed. The increased basal discharge observed in arthritic joints was unaffected by DPCPX, which implies that the increase in spontaneous activity associated with arthritis is unlikely to involve tonically released adenosine. The results support the hypothesis that adenosine excites primary afferent nociceptive nerve terminals in the rat knee joint, an effect mediated by adenosine A1 receptors. ATP, adenosine, and A1 receptors may play a role in generating the peripheral nociceptive (pain) signal.