Non-steroidal anti-inflammatory drugs (NSAIDs) are frequently used as analgesics. Although the results of clinical studies indicate considerable disparity in the analgesic efficacy of NSAIDs, the pre-clinical models generally used for the study of nociception do not allow a clear distinction to be made between the analgesic properties of agents belonging to this family. As clinical pain is characterized by hyperalgesia, we evaluated the effects of NSAIDs with similar chemical structures but different selectivities for cyclo-oxygenase (COX)-1 and COX-2 in a new behavioural model of central hyperalgesia in rats. We assessed the effects of lornoxicam, piroxicam, and meloxicam on the reduction of hindpaw nociceptive thresholds to thermal stimulation produced by a 10% formaldehyde (formalin) injection into rat tail. Each drug was administered intraperitoneally (i.p.) at its ED(50)for the anti-inflammatory effect (namely the inhibition of carrageenan-induced hindpaw oedema). At this dose (1.3 mg kg(-1), 1.0 mg kg(-1), and 5.8 mg kg(-1), respectively), lornoxicam, piroxicam, and meloxicam produced the same anti-inflammatory effect, did not modify thermal nociceptive thresholds, and significantly reduced the hyperalgesia. However, only lornoxicam was fully effective for prevention of hyperalgesia. Our results indicate a dissociation between the anti-inflammatory and the anti-hyperalgesic activity of NSAIDs, where the latter seems to be more evident after the block of both COX-1 and COX-2. Finally, they suggest that our experimental model of thermal hindpaw hyperalgesia can be effectively utilized to assess the ability of different drugs to reduce central sensitization, and thus hyperalgesia.