A rabbit ulnar non-union model was used to evaluate the effect of recombinant human osteogenic protein-1 on the healing of a large segmental osteoperiosteal defect. A 1.5-centimeter segmental defect was created in the mid-part of the ulnar shaft of adult rabbits. The defect was filled with an implant containing either recombinant human osteogenic protein-1 or naturally occurring bovine osteogenic protein. The recombinant human osteogenic protein-1 implants consisted of a carrier of 125 milligrams of demineralized, guanidine-extracted, insoluble rabbit bone matrix (the collagen carrier), reconstituted with 3.13, 6.25, 12.5, twenty-five, fifty, 100, 200, 300, or 400 micrograms of recombinant human osteogenic protein-1. Animals that received recombinant human osteogenic protein-1 were compared with animals that received an implant of 250 micrograms of a preparation of naturally occurring bovine osteogenic protein mixed with the collagen carrier. Limbs that served as controls received either the collagen carrier alone or no implant at all. The treated and the untreated defects were examined radiographically and histologically at eight or twelve weeks after implantation. Mechanical testing was performed on six animals. All implants of recombinant human osteogenic protein-1, except for those containing 3.13 micrograms of the substance, induced complete radiographic osseous union within eight weeks. The defects that were treated with an implant of bovine osteogenic protein also healed within this time-period. The bone induced by both types of implants had new cortices with advanced remodeling and marrow elements. Histological evaluation of this new bone at eight weeks postoperatively revealed primarily lamellar bone, with the formation of new cortices and normal-appearing marrow elements. The average torsional strength and energy-absorption capacity of the union induced by recombinant human osteogenic protein-1 was comparable with that of intact bone. The control defects that had been implanted with collagen carrier alone and those with no implant showed no bridging of the defect.