The assembly strains produced in cadaver femurs during uncemented femoral arthroplasty were measured using strain gages and photoelastic coatings. Resecting the femoral neck, reaming the canal with power drills, and rasping with an optimal size rasp, as determined by preoperative radiographic templating, produced small assembly strains, up to 300 microstrain. Insertion of an optimal-size prosthesis after preparing the femoral canal with instruments the same size as the prosthesis produced moderate assembly strains, up to 1,000 microstrain. Half a millimeter press-fit of optimal prostheses produced larger assembly strains, up to 2,000 microstrain. Half a millimeter press-fit of a prosthesis that was also one size (1.0 mm) larger than that determined to be optimum produced even larger assembly strains (2,000-6,000 microstrain) and longitudinal linear fractures in the femoral cortex. Insertion of prostheses that were smaller than the rasps produced minimal strains in the femoral cortex. The magnitude of peak strains produced by press-fitting the femoral components and the small amounts of disparity between the size of the recess and the prosthesis necessary to produce these strains show the narrow range of tolerances available to the surgeon. Cementless femoral arthroplasty requires great care in preparing the femoral canal to the appropriate size as determined from preoperative templating, using accurate and precisely toleranced instrumentation and prosthetic components in order to avoid femoral fractures, yet obtain a stable fit.