The purpose of this study was to develop and evaluate a biomechanical model of lumbar back extension over a wide range of positions for the lumbar spine, incorporating the latest information on muscle geometry and intra-abdominal pressure (IAP). Analysis of the Visible Human data was utilised in order to obtain anatomical information unavailable from the literature and magnetic resonance imaging was used to generate subject-specific anatomical descriptions. The model was evaluated by comparisons with measured maximal voluntary static back-extension torques. Predicted maximal specific muscle tensions agreed well with in vitro measurements from the literature. When modelling the maximal static back-extension torque production, it was possible to come fairly close to simultaneous equilibrium about all the lumbar discs simply by a uniform muscle activation of all back-extensor muscles (the caudal part showed, however, less agreement). This indicates that equilibrium in the lumbar spine is mainly regulated by passive mechanical properties, e.g. muscle length changes due to postural changes, rather than due to complex muscle coordination, as earlier proposed. The model showed that IAP (measured during torque exertions) contributes about 10% of the total maximal voluntary back-extensor torque and that it can unload the spine from compression. The spinal unloading effect from the IAP was greatest with the spine held in a flexed position. This is in opposition to the effects of changed muscle lever arm lengths, which for a given load would give the largest spinal unloading in the extended position. These findings have implications for the evaluation of optimal lifting techniques.