Tibiae from 60 male Wistar rats, aged 13 +/- 1 weeks, were divided into six groups for mechanical and histological testing. Bones were loaded repetitively in torsion at 90 deg.s-1. Group 1 was subjected to 5,000 loading cycles at a twist angle of 3.6 degrees, groups 2-5 to 10,000 cycles at 3.6, 5.4, 7.2, and 9.0 degrees, respectively, and group 6 was tested to failure. Six transverse sections from the middiaphysis were then cut, bulk-stained in basic fuchsin, and hand ground to 30-50 microns to examine the presence of microcracks. Cracks were classified as running parallel to lamellae, crossing lamellae, crossing the full thickness of the cortex, or invading vascular canals. Results for fatigue testing showed that the tibiae exhibited a gradual decrease in torque (P less than 0.05), average stress (P less than 0.01), stiffness (P less than 0.01) and energy absorbed (P less than 0.01) from the initial loading cycle. Analysis of microdamage showed an increase in the variety of cracks from groups 1-5. Analysis of deviance demonstrated a strong dependence of crack probability on the level of loading for all crack types (P less than 0.05) except those crossing lamellae. This study reinforces the evidence that yielding of bone observed during repetitive loading is caused by diffuse structural damage such as microcracking or debonding.