Adaptive changes in bone modeling in response to noninvasive, cyclic axial loading of the rat ulna were compared with those using 4-point bending of the tibia. Twenty cycles daily of 4-point bending for 10 days were applied to rat tibiae through loading points 23 and 11 mm apart. Control bones received nonbending loads through loading points 11 mm apart. As woven bone was produced in both situations, any strain-related response was confounded by the response to direct periosteal pressure. Four-point bending is not, therefore, an ideal mode of loading for the investigation of strain-related adaptive modeling. The ulna's adaptive response to daily axial loading over 9 days was investigated in 30 rats. Groups 1-3 were loaded for 1200 cycles: Group 1 at 10 Hz and 20 N, Group 2 at 10 Hz and 15 N, and Group 3 at 20 Hz and 15 N. Groups 4 and 5 received 12,000 cycles of 20 N and 15 N at 10 Hz. Groups 1 and 4 showed a similar amount of new bone formation. Group 5 showed the same pattern of response but in reduced amount. The responses in Groups 2 and 3 were either small or absent. Strains were measured with single-element, miniature strain gauges bonded around the circumference of dissected bones. The 20 N loading induced peak strains of 3500-4500 mustrain. The width of the periosteal new bone response was proportional to the longitudinal strain at each point around the bone's circumference. It appears that when a bone is loaded in a normal strain distribution, an osteogenic response occurs when peak physiological strains are exceeded. In this situation the amount of new bone formed at each location is proportional to the local surface strain. Cycle numbers between 1200 and 12,000, and cycle frequencies between 10 and 20 Hz have no effect on the bone's adaptive response.