Loosening after total ankle arthroplasty may result from absorption of tensile and shear forces by relatively small bone-cement interfaces. A surface replacement arthroplasty similar in principle to the Oxford knee arthroplasty offers theoretical advantages over other designs. The ability of such a prosthesis to provide a stable articulation was tested by examination of ankle laxity before and after replacement of the tibiotalar articular surfaces. The motion response characteristics of four specimens were measured for internal-external rotation moment (+/- 3 nm), inversion-eversion moment (+/- 3 nm), and anteroposterior displacement (+/- 50 N). The prosthesis restored normal internal-external and inversion-eversion rotatory stability, but the mean anteroposterior laxity was increased from 5.5 to 9.7 mm (P less than 0.05). Anteroposterior stability could not be restored by increasing the thickness of the bearing interposed between the prosthetic surfaces. The findings can be explained by consideration of the anatomy of the ligaments in relation to the contour of the normal and prosthetic articular surfaces. The normal tibial articular surface, which is concave in the sagittal plane, provides restraint against anteroposterior motion which is lost when the surface is replaced by the flat prosthetic surface. The findings question the suitability of this type of prosthesis for the ankle.