Obtaining anatomic knee kinematics after a total knee is likely to improve outcomes. We used a crouching machine to compare the kinematics of standard condylar designs with guided motion designs. The standard condylars included femoral sagittal radii with constant radius, J‐curve and G‐curve; the tibial surfaces were of low and high constraint. The guided motion designs were a medial pivot and a design with asymmetric condylar shapes and guiding surfaces. The machine had a flexion range from 0° to 125°, applied quadriceps and hamstring loading, and simulated the collateral soft tissues. The kinematics of all standard condylar knees were similar, showing only small anterior–posterior displacements and internal–external rotations. The two asymmetric designs showed posterior displacements during flexion, but less axial rotations than anatomic knees. The quadriceps forces throughout flexion were very similar between all designs, reflecting similar lever arms. It was concluded that standard condylar designs, even with variations in sagittal radii, are unlikely to reproduce anatomic kinematics. On the other hand, designs with asymmetric constraint between medial and lateral sides, and other guiding features, are likely to be the way forward. The mechanical testing method could be further improved by superimposing shear forces and torques during the flexion–extension motion, to include more stressful in vivo functional conditions.