In clinical practice efforts are made to apply a fixation plate on the side opposite the strongest muscle pull. This achieves an optimal distribution of compression between the fragment ends (principle of tension band plating). This is however frequently impossible for anatomical or surgical reasons. In an 'in vivo' study lasting 8 weeks a standardized oblique osteotomy was performed on the tibia of 16 sheep in four different models of tension band plating (a contoured and an overbent plate with or without an interfragmentary lag screw) were assessed. Tension on the plate surface was recorded by strain gauges for different gait speeds on the treadmill. These measurements were performed throughout the experiment. Radiographs were taken at regular intervals in order to assess stability and polychrome sequential labelling and microradiographs served to investigate the healing process. Possible relationships and/or interactions between plate tension and bone healing were investigated. Implant loading under bending strain was reduced the most for the combination of plate overbending with a lag screw. The insertion of a lag screw reduces the surface strain on the plate whether it is contoured or overbent. The bending and torsional forces are greatest if a straight plate is used alone and the principle of tension band plating is not applied. Direct bone healing was only observed in the group with contoured plate and lag screw. Overbending combined with a lag screw provided only a relatively unstable fixation. A residual gap immediately beneath the plate permits "dynamic compression" since the screws slide towards the osteotomy when loaded producing bone resorption under the plate and signs of screw loosening. The models with contoured and overbent plates without a lag screw were histologically assessed as very unstable with signs of secondary fragment displacement, obvious callus formation, resorption at the fragment ends and under the plate, delayed and diminished Haversian remodelling and corrosion sites at the screw heads and at the adjacent site on the plate hole. In all groups, stripping of the periosteum under the plate was associated with porosis of the corresponding cortex as a sign of temporarily impaired blood supply. A relationship between implant loading and/or unloading (stress shielding) could not be demonstrated. Callus formation, measured quantitatively on the radiographs, is directly related to the strain on the plate. Direct bone healing is rapid and is seen histologically three weeks postoperatively, particularly for fixations with contoured plate and lag screw. The early appearance of fixation callus in the presence of an intact blood supply indicates a primary instability of the osteosynthesis. Later, it may be an indication of secondary instability. The time at which osteons appear, their number and location provides information on the stability of the osteosynthesis. At a time when indirect fracture reduction and stabilization using minimally invasive techniques and implants is being propagated, additional ways and means must be sought to assess clinically the load on the implants and the risk of implant failure.