Fluoroscopic evaluation of total knee arthroplasty (TKA) has reported sudden anterior translation of the femur relative to the tibia (paradoxical anterior motion) for some cruciate-retaining designs. This motion may be tied to abrupt changes in the femoral sagittal radius of curvature characteristic of traditional TKA designs, as the geometry transitions from a large load-bearing distal radius to a smaller posterior radius which can accommodate femoral rollback. It was hypothesized that a gradually reducing radius may attenuate sudden changes in anterior–posterior motion that occur in mid-flexion with traditional discrete-radius designs. A combined experimental and computational approach was employed to test this hypothesis. A previously developed finite element (FE) model of the Kansas knee simulator (KKS), virtually implanted with multiple implant designs, was used to predict the amount of paradoxical anterior femoral slide during a simulated deep knee bend. The model predicted kinematics demonstrated that incorporating a gradually reducing radius in mid-flexion reduced the magnitude of paradoxical anterior translation between 21% and 68%, depending on the conformity of the tibial insert. Subsequently, both a dual-radius design and a modified design incorporating gradually reducing radii were tested in vitro in the KKS for verification. The model-predicted and experimentally observed kinematics exhibited good agreement, while the average experimental kinematics demonstrated an 81% reduction in anterior translation with the modified design. The FE model demonstrated sufficient sensitivity to appropriately differentiate kinematic changes due to subtle changes in implant design, and served as a useful pre-clinical design-phase tool to improve implant kinematics.