We compared the magnitudes of motion between the prosthesis and bone during axial and torsional loading in seven matched pairs of fresh-frozen femora of cadavera in which an uncemented, collarless, isthmus-filling, straight-stem (Harris-Galante) prosthesis had been placed in one femur and an uncemented, collarless, proximal-filling, curved-stem (anatomic) prosthesis had been placed in the other femur. The comparison was performed in order to determine the effect of the geometry of the stem on the magnitude of motion. Single-limb-stance loads and combined axial and torsional loads were applied to the implanted femoral prostheses with the use of a jig that simulated acetabular and trochanteric loading. Extensometers were used to measure motion at the prosthesis-bone interface. The prostheses were then removed and were reinserted, with cement applied to the proximal porous coating to simulate ingrowth of bone. The single-limb-stance and combined axial and torsional loads were reapplied and the magnitude of motion was recorded again. No significant differences in the magnitudes of the motion were found between the femora in which the straight stem had been implanted and the femora in which the curved stem had been implanted, during either simulated single-limb-stance or low-intensity torsional loading. When large torsional moments (twenty-two newton-meters) were applied, significantly less motion occurred at the bone-prosthesis interface, both proximally (p = 0.019) and distally (p = 0.0013), in the femora with the curved-stem implant than in the femora with the straight-stem implant. When cement had been applied proximally, proximal and distal motion between the prosthesis and the femur was decreased during simulated single-limb-stance and during torsional loading in the femora with the straight stem and the femora with the curved stem.