Proprioception was measured in two groups of patients following successful total knee arthroplasty (TKA). In one group, the posterior cruciate ligament was retained and an unconstrained cruciate-retaining total knee component was used; in the other group, the posterior cruciate ligament was excised and a cruciate-substituting design was implanted. Threshold to detection of passive motion was quantified as a measure of proprioception. The degree of preoperative arthritis was objectively classified according to Resnick and Niwoyama. There was no difference in threshold to detection of passive motion in cruciate-retaining versus cruciate-substituting TKA. In patients with a moderate grade of arthritis before surgery, the postoperative scores were virtually identical. When the grade of preoperative arthritis was severe, patients with cruciate-substituting TKAs performed significantly better than those with cruciate-retaining TKAs.