Résumé : A previous study suggested that the kinematics of a loose-hinge total elbow arthroplasty (TEA) are those of a truly semiconstrained joint. This study addresses the effects of malposition of the implant. The three- dimensional elbow kinematics during simulated active motion were studied in six cadaver specimens using an electromagnetic tracking device. In addition to simulated active elbow flexion, flexion arcs were obtained under an elbow varus or valgus moment, to calculate the structural varus-valgus laxities. The results after four different Coonrad-Morrey TEA positions of implantation were compared with those of the intact elbow. The flexion-extension amplitudes were not significantly decreased after TEA implantation, except with external rotation of the ulnar component, which resulted in a loss of extension. In the intact elbow and after TEA implantation in any position, the mean varus-valgus deviations throughout elbow flexion were in a narrower range than the structural limits imposed by the ligaments (intact elbow) or the TEA hinge design. With internal malrotation of the humeral component over 10°, however, the valgus structural limit was reached and, conversely, the varus limit with external rotation over 10°. The clinical improvement observed with the semiconstrained TEA is derived from the benefits of the less constrained articulation. The proximodistal changes of TEA implantation have no consequence on the kinematic pattern. Relational malpositioning of either humeral of ulnar component should be avoided, the first because it changes the kinematic pattern toward the structural limits of the implant and, therefore, may lead to excessive stresses at the bone-cement-implant interfaces and to early loosening, and the latter because it causes loss of extension.