Rapid loss of cancellous bone after menopause occurs by a mechanism that removes some structural elements completely, leaving those that remain more widely separated and less well connected. Slow loss of cancellous bone continues by a mechanism that reduces the thickness of the structural elements that survive the initial phase of rapid loss. Both processes have advanced further in patients with vertebral compression fracture due to osteoporosis, than in healthy subjects of similar age: whether this is because they began sooner, proceeded more rapidly or continued for longer is unknown. This overall concept, first developed about ten years ago, has now been confirmed by a variety of different methods including node-strut analysis, star volume and the change in curvature with profile dilation. Because part of the architectural contribution to compressive strength is captured by non-invasive densitometric methods, and the contribution of cortical bone to compressive strength is significant, it has been difficult to demonstrate an independent architectural component of bone fragility by in vitro comparison of structural measurements with biomechanical testing. Nevertheless, three independent clinical studies, comparing subjects with and without vertebral fracture, have each strongly suggested an independent architectural contribution to fracture risk. Complete removal of structural elements is initiated by focal perforation of trabecular plates, but the mechanism of perforation remains controversial. Generalised or focal reductions in trabecular thickness, a stochastic relationship between the frequency distributions of resorption depth and trabecular thickness, and preferential occurrence of remodelling activation on thinner trabeculae may all contribute, but based on indirect reasoning, it is impossible to account for the observed facts by these mechanisms alone, implying the necessity during the first few years after menopause for a substantial increase in resorption depth. This process has rarely been observed in human subjects, but has been independently demonstrated in three different animal species — monkeys, rats and minipigs, in the latter two as a consequence of oestrogen deficiency. Architecture has important implications for the prevention as well as for the pathogenesis of vertebral fracture, since increasing the thickness of existing trabeculae has only minimal effect on connectivity. Because lost trabeculae cannot be replaced, once destruction of the skeleton has been allowed to proceed, it is likely that no pharmacologie intervention will have much effect on the strength of the bone that remains. Consequently, more emphasis must be given to prevention, either by oestrogen, or by some substitute for oestrogen that will also reduce resorption depth as well as the rate of remodelling activation.