MRI studies of the knee were performed at intervals between full extension and 120° of flexion in six cadavers and also non-weight-bearing and weight-bearing in five volunteers. At each interval sagittal images were obtained through both compartments on which the position of the femoral condyle, identified by the centre of its posterior circular surface which is termed the flexion facet centre (FFC), and the point of closest approximation between the femoral and tibial subchondral plates, the contact point (CP), were identified relative to the posterior tibial cortex.

The movements of the CP and FFC were essentially the same in the three groups but in all three the medial differed from the lateral compartment and the movement of the FFC differed from that of the CP. Medially from 30° to 120° the FFC and CP coincided and did not move anteroposteriorly. From 30° to 0° the anteroposterior position of the FFC remained unchanged but the CP moved forwards by about 15 mm. Laterally, the FFC and the CP moved backwards together by about 15 mm from 20° to 120°. From 20° to full extension both the FFC and CP moved forwards, but the latter moved more than the former. The differences between the movements of the FFC and the CP could be explained by the sagittal shapes of the bones, especially anteriorly.

The term ‘roll-back’ can be applied to solid bodies, e.g. the condyles, but not to areas. The lateral femoral condyle does roll-back with flexion but the medial does not, i.e. the femur rotates externally around a medial centre. By contrast, both the medial and lateral contact points move back, roughly in parallel, from 0° to 120° but they cannot ‘roll’.

Femoral roll-back with flexion, usually imagined as backward rolling of both condyles, does not occur.

The posterior cruciate ligament (PCL) was imaged by MRI throughout flexion in neutral tibial rotation in six cadaver knees, which were also dissected, and in 20 unloaded and 13 loaded living (squatting) knees. The appearance of the ligament was the same in all three groups. In extension the ligament is curved concave-forwards. It is straight, fully out-to-length and approaching vertical from 60° to 120°, and curves convex-forwards over the roof of the intercondylar notch in full flexion. Throughout flexion the length of the ligament does not change, but the separations of its attachments do.

We conclude that the PCL is not loaded in the unloaded cadaver knee and therefore, since its appearance in all three groups is the same, that it is also unloaded in the living knee during flexion. The posterior fibres may be an exception in hyperextension, probably being loaded either because of posterior femoral lift-off or because of the forward curvature of the PCL. These conclusions relate only to everyday life: none may be drawn with regard to more strenuous activities such as sport or in trauma.

In 13 unloaded living knees we confirmed the findings previously obtained in the unloaded cadaver knee during flexion and external rotation/internal rotation using MRI. In seven loaded living knees with the subjects squatting, the relative tibiofemoral movements were similar to those in the unloaded knee except that the medial femoral condyle tended to move about 4 mm forwards with flexion. Four of the seven loaded knees were studied during flexion in external and internal rotation. As predicted, flexion (squatting) with the tibia in external rotation suppressed the internal rotation of the tibia which had been observed during unloaded flexion.