Bone marrow has been shown to contain a population of rare mesenchymal stem cells that are capable of forming bone, cartilage, and other connective tissues. We examined the effect of cultured autologous mesenchymal stem cells on the healing of critical-sized (twenty-one-millimeter-long) segmental defects in the femora of adult female dogs. Autologous mesenchymal stem cells were isolated from bone marrow, grown in culture, and loaded onto porous ceramic cylinders consisting of hydroxyapatite (65 per cent) and beta-tricalcium phosphate ceramic (35 per cent). The animals were randomly assigned to one of three groups. In Group A (six dogs), a porous ceramic cylinder that had been loaded with autologous mesenchymal stem cells was implanted in the defect. In Group B (six dogs), a ceramic cylinder that had not been loaded with cells was placed in the defect. In Group C (three dogs), the defect was left untreated (no ceramic cylinder was implanted). Radiographs were made immediately after the operation and at four-week intervals. At sixteen weeks, the animals were killed, the involved femora were removed, and undecalcified histological sections from the defects and adjacent bone were prepared. Histological and histomorphometric studies were carried out to examine the healing of the defects and the formation of bone in and around the ceramic implants. Atrophic non-union occurred in all of the femora that had untreated defects, and only a small amount of trabecular bone formed at the cut ends of the cortex of the host bone in this group. In contrast, radiographic union was established rapidly at the interface between the host bone and the implants that had been loaded with mesenchymal stem cells. Numerous fractures, which became more pronounced with time, developed in the implants that had not been loaded with cells. Histological and morphometric analyses demonstrated that both woven and lamellar bone had filled the pores of the implants that had been loaded with mesenchymal stem cells; the amount of bone was significantly greater (p < 0.05) than that found in the pores of the implants that had not been loaded with cells. In addition, a large collar of bone (mean maximum thickness, 3.14 millimeters) formed around the implants that had been loaded with cells; this collar became integrated and contiguous with callus that formed in the region of the periosteum of the host bone. The collar of bone remodeled during the sixteen-week period of study, resulting in a size and shape that were comparable with those of the segment of bone that had been resected. Callus did not develop around the cortex of the host bone or around the defect in any of the specimens in the other two groups.