Twenty-five horses undergoing arthroscopic surgery were studied to develop a scheme for assessing pain in horses while investigating the effects of phenylbutazone (PBZ) analgesia. Fifteen of the 25 horses received PBZ 4 mg/kg intravenously (IV) before surgery and 2 mg/kg (IV) every 12 hours thereafter until 60 hours; the remaining 10 (placebo group) were given a corresponding volume of saline. In both groups, venous blood samples were collected for catecholamine, beta-endorphin, and cortisol assays before premedication and up to 72 hours after surgery. Postoperative pain was evaluated by measuring predefined behavioral and physiological variables. A total postoperative pain severity index (TPPSI) was calculated using all variables. There were no differences between PBZ and placebo groups in plasma beta-endorphin or catecholamine concentrations, but the TPPSI was higher in the placebo group than in the PBZ group, suggesting that perioperative treatment with PBZ has some analgesic benefit. This study shows the difficulties associated with pain assessment in horses.