The purpose of this study was to identify factors which predispose to aseptic loosening of the femoral component in cemented total hip replacement. Its design was based on rigid selection criteria, so that successful and loose replacements which employed the same surgical technique were compared. Measurements of patient anatomy and of the insertion of the femoral component were made, by an accurate computer technique, on initial post-operative radiographs. Loosening was associated with heavier patients with a wider medullary canal which was flared proximally. This difference in anatomy led to differing distributions of cement in the successful and loose replacements. Medial cement-bone demarcation, at the mid-stem level, was also associated with loosening. These findings indicate the importance of optimizing the size of the prosthesis with respect to the femoral morphology.