OBJECTIVE To report the frequency of radiocarpal joint (RCJ) perforation during two proximolateral carpal sheath (CS) approaches and to investigate whether the presence or absence of a natural communication between the RCJ and CS. METHODS Experimental, randomized, controlled study. METHODS Twelve adult horses and 12 cadaver front limbs. METHODS Proximolateral CS tenoscopy was performed on both carpi of 12 horses under general anaesthesia. Limbs were randomly assigned into two groups. In group 1, the instrument portal was centered at the level of the distal radial physeal remnant (PR). In group 2, the instrument portal was centered 2 cm proximal to the PR. Immediately after tenoscopy, dye-coloured fluid was instilled arthroscopically into the dorsolateral RCJ, and the open tenoscopic portals were evaluated for leakage. Twelve cadaver front limbs were examined for naturally occuring RCJ and CS communication by using positive-contrast computed tomography (CT). Frequency of perforations were compared with McNemar's test for two matched proportions. RESULTS In group 1, 10 of 12 RCJ were perforated during tenoscopy, and, in group 2, four of 12 RCJ were perforated (P = 0.03). No diffusion of contrast from the RCJ to the CS was detected by CT. CONCLUSIONS Perforation of the RCJ was common when CS instrument portals were placed at the level of the PR. No natural communication was detected between the RCJ and CS. CONCLUSIONS Carpal sheath tenoscopic instrument portals may need to be placed >2 cm proximal to the PR to prevent inadvertent RCJ perforation, especially in horses undergoing tenoscopic treatment of a septic CS.