Introduction. The results of the original mobile bearing Oxford unicompartmental knee replacement (UKR) in the lateral compartment have been disappointing because of high dislocation rates (11%). This original implant used a flat bearing articulation on the tibial tray. To address the issue of dislocation a new implant (domed tibia with biconcave bearing to increase entrapment) was introduced with a modified surgical technique. The aim of this study was to compare the risk of dislocation between a domed and flat lateral UKR. Methods. Separate geometric computer models of an Oxford mobile bearing lateral UKR were generated for the two types of articulation between the tibial component and the meniscal bearing: Flat-on-flat (flat) and Concave-on-convex (domed). Each type of mobile bearing was used to investigate three distinct dislocation modes observed clinically: lateral to medial dislocation, with the bearing resting on the tray wall (L-M-Wall); medial to lateral dislocation, out of the joint space (M-L); anterior to posterior dislocation, out of the joint space (A-P). A size C tray and a medium femoral component and bearing were used in all models. The femoral component, tibial tray and bearing were first aligned in a neutral position. For each dislocation the tibial tray was restrained in all degrees of freedom. The femoral component was restrained from moving in the anterior-posterior directions and in the medial-lateral directions. The femoral component was also restrained from rotating about the anterior-posterior, medial-lateral and superior-inferior directions. This meant that the femoral component was only able to move in the superior-inferior direction. Different bearing sizes were inserted into the model and the effect that moving the femoral component medially and laterally had on the amount of distraction required to cause bearing dislocation was investigated. Results. The average femur distraction to allow bearing dislocation in the A-P, M-L and L-M-wall directions was 1.62 mm (27%), 0.51 mm (26%) and 1.2 mm (24%) greater respectively for the domed bearing. There was a 3% increase in femoral distraction required to cause L-M-Wall dislocation, per increment of bearing thickness for both the domed and lateral bearings. There was on average a 7% increase in femoral distraction required to cause L-M-Wall dislocation per mm increment of medial femoral component movement. Discussion. Dislocation over the tray wall is a particular clinical problem and using a domed bearing can lead to an increased required femoral distraction of between 25% and 37%. This may be significant during everyday activities and demonstrates that the new domed design should reduce the incidence of bearing dislocation by increasing the amount of entrapment. Increasing the thickness of the bearing has a small effect on the distraction required to allow bearing dislocation. Lateral placement of the femoral component markedly reduced the femoral distraction required for bearing dislocation over the tray wall. Medial placement of the femoral component is advisable so long as impingement with the tray wall is avoided