Large-deformation bending and buckling have long been proposed as failure mechanisms by which the strength of trabecular bone can be affected disproportionately to changes in bone density, and thus may represent an important aspect of bone quality. We sought here to quantify the contribution of large-deformation failure mechanisms on strength, to determine the dependence of these effects on bone volume fraction and architecture, and to confirm that the inclusion of large-deformation effects in high-resolution finite element models improves predictions of strength versus experiment. Micro-CT-based finite element models having uniform hard tissue material properties were created from 54 cores of human trabecular bone taken from four anatomic sites (age = 70+/-11; 24 male, 27 female donors), which were subsequently biomechanically tested to failure. Strength predictions were made from the models first including, then excluding, large-deformation failure mechanisms, both for compressive and tensile load cases. As expected, strength predictions versus experimental data for the large-deformation finite element models were significantly improved (p < 0.001) relative to the small deformation models in both tension and compression. Below a volume fraction of about 0.20, large-deformation failure mechanisms decreased trabecular strength from 5-80% for compressive loading, while effects were negligible above this volume fraction. Step-wise nonlinear multiple regression revealed that structure model index (SMI) and volume fraction (BV/TV) were significant predictors of these reductions in strength (R2 = 0.83, p < 0.03). Even so, some low-density specimens having nearly identical volume fraction and SMI exhibited up to fivefold differences in strength reduction. We conclude that within very low-density bone, the potentially important biomechanical effect of large-deformation failure mechanisms on trabecular bone strength is highly heterogeneous and is not well explained by standard architectural metrics.