Symmetrical, asymmetrical, and custom asymmetrical stems without collars were manufactured and inserted without cement, in that order, into each of eight femora. Displacement transducers were mounted at different locations to measure relative motion between the stem and the bone. A load was applied to simulate the forces acting on a hip joint in a single-limb stance. A force of 1000 newtons was applied three times in order to seat the prosthesis, and then a cyclical force of 1000 newtons was applied to the head of the prosthesis; readings of relative motion were taken every 500 cycles, until 2500 cycles had been reached. The relative motion diminished as the number of cycles increased, with most of the reduction occurring within the first 500 cycles. The symmetrical stem had the least relative axial motion, with an average of six micrometers, but the other stems also had low values: eleven micrometers for the custom stem, and nineteen micrometers for the asymmetrical stem. In terms of axial rotation, however, the symmetrical stem had several times more motion than did the other two stems. Motion of the stem tip toward or away from the canal in the frontal plane and in the sagittal plane also was higher for the symmetrical stem. Axial migration (permanent sinkage after 2500 cycles) was very similar between all three types of stem. In a torsional loading test, the symmetrical stem showed about ten times more relative rotational motion than did the other two stems. Over-all, the custom stem and the asymmetrical stem had similar results.(ABSTRACT TRUNCATED AT 250 WORDS)