Facet surfaces of the cricoarytenoid joints from two cadaver larynges were digitized. The data were used to compute the optimal axis of rotation for each of the joints in the sense that the computed axis minimized the variance of the joint gap over the full range of joint motion. The optimal axis corresponded to a rocking motion of the arytenoid on the corresponding cricoid. This motion was consistent with experimental data from digitized recordings of vocal fold movement. Using the rigid laryngoscopic view, a similarity in vocal process movement, over the range in motion, between the rocking axis and the vertical axis described in the literature was found, resolving the controversy between two conflicting views of motion of the vocal processes.