To examine in detail the events that lead to healing under different mechanical conditions, experimental fractures of the rabbit tibia were immobilized with either a metal compression plate or a plastic plate, to produce stable or unstable mechanical conditions respectively. To fracture the bone, a small saw cut is made in the medial cortex and three-point pressure is then applied to fracture the remaining cortical bone. A small amount of bone is removed by the saw cut. This is of no consequence when a plastic plate is used because the plate is designed so that a small gap remains between the fractured surfaces. When a compression plate is used, a gap approximately 100 pm wide remains under the plate. In addition, despite the use of compression, only part of the fractured surfaces are normally in contact, so other narrow fracture gaps may remain. The fractures were examined at experimental times from 4 days to 1 year by both light and electron microscopy. In stable mechanical conditions, a thin layer of periosteal callus starts to form within 4 days and healthy mesenchymal cells and capillaries are adjacent to the fracture. By 7 days the callus is fully developed and is covered by a new periosteum. Over the following weeks it becomes more compact as the cavities are filled by bone. The endosteal callus, if formed, is small. From 9 days onwards the fracture gaps of between 30 and 100 pm are invaded by mesenchymal cells; macrophages remove the debris and fibroblasts lay down a collagenous matrix. Capillaries are numerous. By 2 weeks, these gaps are lined by osteoblasts producing transverse layers of new bone on the fractured surfaces. The gap is rapidly filled by transversely orientated bone to achieve union by 3-4 weeks. Resorption is limited to very small, isolated regions of the fractured surfaces. The gaps between 10 and 30 μm wide appear to be too narrow for cells and capillaries to enter, and they are widened by osteoclastic action. Mesenchymal cells and capillaries then enter and bone is laid down on the surfaces of the newly excavated cavities. After union, the region is recognized as a wide band of irregularly organized bone. In compressed regions, areas of actual contact are very few, and narrow gaps can be seen. Union is achieved by direct Haversian remodelling across gaps up to 9 pm wide. Remodelling starts at about 3 weeks and it is part of the normal process that occurs after union of a fractured cortex. The events in the fractures healing under unstable mechanical conditions diverge immediately. The initial periosteal reaction is similar, but cancellous bone forms only away from the fracture gap. Mesenchymal cells in a fibrous matrix, but with very few, if any, associated capillaries, persist as a thick layer over the fracture gap. At 6 days islands of cartilage begin to develop, and by 12 days the whole central area is filled by a cartilaginous matrix containing thick and thin collagen fibrils and elastic fibres. The cartilage is replaced by bone by 3-4 weeks. At this point, healthy cells and capillaries are found near the fracture gap. A small endosteal callus may form in a similar way. The fracture gaps are now invaded. The debris is removed by macrophages, and fibroblasts produce a collagenous matrix. Osteoclasts cover the fractured surfaces, but the amount of resorption is minimal. Capillaries also enter the gap. Within 1-2 weeks the gap is filled by transverse lamellae of bone. Once the fractured cortex is united, the events are similar in both series of fractures. Resorption of any callus starts. In the periosteal callus, large cavities form particularly along the callus-cortex junction. Over the next 3 months the callus is resorbed and the cortex is extensively remodelled. By 18 weeks no callus remains. The cortical bone is compact, but large cavities filled with fat cells may occur. The overall dimensions of the bone may be larger than that of the contralateral normal tibia, but the layer of cortical bone is thinner. Thus the plate, whether of metal or plastic, affects the structure of the bone. This study serves to emphasize that the mechanical stability of a healing fracture determines the composition and size of the periosteal callus and the time at which union across a fracture gap is achieved. The fracture fragments must be mechanically stable, so that there is no movement at the fracture gap before cells can enter and lay down new bone. This may be achieved immediately, though artificially, by a metal compression plate, or slowly, though naturally, by the formation of a large callus of cancellous bone. The movements and differentiation of the cells involved in healing are discussed. This experimental fracture model will form a basis for future experimental studies of healing fractures.