The biomechanical events which accompany functional loading of the human mandible are not fully understood. The techniques normally used to record them are highly invasive. Computer modelling offers a promising alternative approach in this regard, with the additional ability to predict regional stresses and strains in inaccessible locations. In this study, we built two three-dimensional finite element (FE) models of a human mandible reconstructed from tomographs of a dry dentate jaw. The first model was used for a complete mechanical characterization of physical events. It also provided comparative data for the second model, which had an increased vertical corpus depth. In both cases, boundary conditions included rigid restraints at the first right molar and endosteal cortical surfaces of the articular eminences of temporal bones. Groups of parallel multiple vectors simulated individual masticatory muscle loads. The models were solved for displacements, stresses, strains, and forces. The simulated muscle loads in the first model deformed the mandible helically upward and toward its right (working) side. The highest principal stresses occurred at the bite point, anterior aspects of the coronoid processes, symphyseal region, and right and left sides of the mandibular corpus. In general, the observed principal stresses and strains were highest on the periosteal cortical surface and alveolar bone. At the symphyseal region, maximum principal stresses and strains were highest on the lower lingual mandibular aspect, whereas minimum principal stresses and strains were highest on its upper labial side. Subcondylar principal strains and condylar forces were higher on the left (balancing or nonbiting) side than on the right mandibular side, with condylar forces more concentrated on the anteromedial aspect of the working-side condyle and on the central and lateral aspects of the left. When compared with in vivo strain data from macaques during comparable biting events, the predictive strain values from the first model were qualitatively similar. In the second model, the reduced tensile stress on the working-side, and decreased shear stress bilaterally, confirmed that lower stresses occurred on the lower mandibular border with increased jaw depth. Our results suggested that although the mandible behaved in a beam-like manner, its corpus acted more like a combination of open and closed cross sections due to the presence of tooth sockets, at least for the task modelled. Although restricted by an oversimplified representation of mandibular structure, limited data which described the behaviour of its various material components, and approximation of the behaviour within and between groups of jaw muscles, our FE-based predictions were consistent with current notions of mandibular mechanics derived from experimental recordings in man and nonhuman primates. © 1992 Wiley-Liss, Inc.