The results of using segmental cortical autogenous bone grafts to reconstruct defects created by resection of tumors were analyzed in forty patients. Thirty-three patients had dual grafts while seven had a single fibular graft. Dual grafts were used for major bones (humerus, femur, and tibia without fibula) while single grafts were used for the radius and for the tibia when the ipsilateral fibula was intact. Thirty patients had good or excellent results; seven, fair; and three, poor results. In twenty-five patients primary union was achieved within tweleve months and in two, in twenty months, while twelve patients required a second, supplementary cancellous graft at the site of non-union to obtain stability. One patient required removal of an infected graft and had a poor result. Stress fractures of the grafts occurred in eighteen of the forty patients after union had occurred. The stress fractures healed in fifteen of these patients: in six with no treatment (the fracture was identified retrospectively), in seven with external immobilization, and in two after bone-grafting of the ununited fracture. There were three persistent non-unions of stress fractures despite bone-grafting, internal fixation, and electrical stimulation, and these account for two of the three poor results. The length of the defect did not affect the incidence of non-union but it did affect the number of fatigue fractures. The shorter grafts (7.5 to twelve centimeters) were associated with a 33 per cent incidence of non-union (four non-unions of twelve grafts) while the longer grafts (twelve to twenty-five centimeters) had a 32 per cent rate of non-union (nine non-unions of twenty-eight grafts). The incidence of fatigue fractures in the longer grafts (58 per cent) was much greater than that in the shorter grafts (17 per cent). The grafts decreased in density during the first six months but gradually regained their mass and were generally comparable to normal cortical bone at two years. As the patients became functiona, most (55 per cent) of the the grafts became more dense than normal, some (34 per cent) remained the same, and a few (11 per cent) became less dense. Similarly, some (32 per cent) hypertrophied, most (58 per cent) remained the same size, and a few (9 per cent) atrophied. There was little morbidity (three of forty patients) associated with graft procurement. In twelve patients an additional graft was implanted experimentally, labeled with tetracycline, and subsequently removed at the time of a secondary procedure. These grafts were analyzed to determine if human grafts were repaired in the same fashion as grafts in experimental animals. The studies showed that human grafts are repaired in the same fashion, but that the sequence takes approximately twice as long as it does in the dog.