The observed fit of bone mass to a healthy animal's typical mechanical usage indicates some mechanism or mechanisms monitor that usage and control the three longitudinal growth, bone modeling, and BMU-based remodeling activities that directly determine bone mass. That mechanism could be named a mechanostat. Accumulated evidence suggests it includes the bone itself, plus mechanisms that transform its mechanical usage into appropriate signals, plus other mechanisms that detect those signals and then direct the above three biologic activities. In vivo studies have shown that bone strains in or above the 1500-3000 microstrain range cause bone modelling to increase cortical bone mass, while strains below the 100-300 microstrain range release BMU-based remodeling which then removes existing cortical-endosteal and trabecular bone. That arrangement provides a dual system in which bone modeling would adapt bone mass to gross overloading, while BMU-based remodeling would adapt bone mass to gross underloading, and the above strain ranges would be the approximate "setpoints" of those responses. The anatomical distribution of those mechanical usage effects are well known. If circulating agents or disease changed the effective setpoints of those responses their bone mass effects should copy the anatomical distribution of the mechanical usage effects. That seems to be the case for many agents and diseases, and several examples are discussed, including postmenopausal osteoporosis, fluoride effects, bone loss in orbit, and osteogenesis imperfecta. The mechanostat proposal is a seminal idea which fits diverse evidence but it requires critique and experimental study.