This study evaluated the reliability of three rating systems for total knee arthroplasty. Twenty-nine patients were assessed by six observers with the American Knee Society Score, the British Orthopaedic Association Score, and the Oxford 12-item questionnaire. The inter- and intraobserver variations were calculated for the American Knee Society Score and the British Orthopaedic Association Score, as was the reproducibility of the Oxford 12-item questionnaire. Components of the scores were evaluated for agreement using Kappa statistics. The British Orthopaedic Association Score had smaller interobserver variation compared to the American Knee Society Score, and the greatest reproducibility of the three systems. This was attributed to the equal weighting of its component variables. The Oxford 12-item questionnaire, a self-administered questionnaire that eliminates interobserver error, emerged as the most reliable system. Observer experience affected the reliability of the American Knee Society Score and the British Orthopaedic Association Score. Subjective variables were more reliable and reproducible than the objective components.