The St Georg modular knee replacement has been studied in 59 cases with an observation period ranging from 28 to 73 months. In 47% of the knees both compartments were replaced; five of the six poor results were in this group. These were revised to a hinge arthroplasty or, in one case, to an arthrodesis. Other complications were few and insignificant. There were no infections. We concluded that unicompartmental knee arthroplasty can be recommended when joint involvement is localised to one compartment only. With more generalised joint disease we prefer a semiconstrained total condylar prosthesis.