The first study of the reproducibility of a widely accepted grading system for femoral cementation in total hip arthroplasty is reported. Sixty postoperative radiographs were graded on two occasions at least 2 weeks apart by two experienced observers. The test-retest reproducibility rate was 60% for observer 1 and 70% for observer 2. The interobserver reproducibility varied between 60 and 63% for the different assessments. Between 5 and 6% of cases differed by more than one grade between tests. Kappa statistical analysis confirmed that reproducibility was poor in all cases. This cement grading system, even in the hands of experienced observers, has too high a rate of intraobserver and interobserver error to recommend its use.