It has been hypothesized that the histological pattern of fracture healing is controlled at least in part by the local mechanical strains in the interfragmentary region. To test this “interfragmentary strain hypothesis,” we applied cyclic bending deformations to tibial osteotomies in 11 sheep. An instrumented flexible plate spanning a 1-mm osteotomy gap was deformed to create a gradient of tissue elongation from 10% under the plate to 100% at the opposite cortex. The cyclic deformations were applied three times per minute, 24 h per day, for 1–5 weeks. However, as a result of tissue differentiation, the bone-plate complex increased in stiffness with healing time, resulting in a marked reduction of the gap deformation at approximately 4 weeks. Fracture healing was evaluated using vascular injection of India ink and conventional histology. A nonlinear three-dimensional finite element model of the interfragmentary tissue at the initial stage of healing was used to predict the complex tissue strains. The ingrowth of vascularized soft tissue into the interfragmentary gap, as well as the subsequent differentiation of this tissue, occurred earlier and to a greater degree in regions of lower strain. In contrast, the proliferation of callus tissue was greatest at the periosteal and endosteal surfaces of the cortex opposite the plate. Direct comparison of the finite element predictions with the histology demonstrated that the spatial distribution of bone resorption at the fracture fragment ends directly corresponded to the locations of elevated tissue strain and stress. However, there was no consistent numerical relationship between the magnitude of these local peak strains and the corresponding volume of cortical bone resorption over the bone cross section.