Use of computational models with kinematic boundary conditions to study the knee joint contact behavior for normal and pathologic knee joints depends on an understanding of the impacts of kinematic uncertainty. We studied the sensitivities of tibio-menisco-femoral joint contact behavior to variations in knee kinematics using a finite element model (FEM) with geometry and kinematic boundary conditions derived from sequences of magnetic resonance (MR) images. The MR images were taken before and after axial compression was applied to the knee joint of a healthy subject. A design of experiments approach was used to study the impact of the variation in knee kinematics on the contact outputs. We also explored the feasibility of using supplementary hip images to improve the accuracy of knee kinematics. Variations in knee kinematics (0.25mm in medial-lateral, 0.1mm in anterior-posterior and superior-inferior translations, and 0.1 degrees in flexion-extension and varus-valgus, 0.25 degrees in external-internal rotations) caused large variations in joint contact behavior. When kinematic boundary conditions resulted in close approximations of the model-predicted joint contact force to the applied force, variations in predictions of contact parameters were also reduced. The combination of inferior-superior and medial-lateral translations accounted for over 70% of variations for all the contact parameters examined. The inclusion of hip images in kinematic calculations improved knee kinematics by matching the femoral head position. Our findings demonstrate the importance of improving the accuracy and precision of knee kinematic measurements, especially when utilized as an input for finite element models.