The differences in age-related fracture risks among men and women must reflect gender differences in the relevant variables. We are concerned here with gender differences in structural variables that relate to the size and shape of bones. As children grow, their bones grow in diameter through periosteal modeling. Studies show that radial growth is driven by mechanical forces and is not just "genetically programmed". Moving bone mass farther from the center of the diaphysis makes it more effective in resisting bending and twisting forces, and disproportionately so in comparison to changes in bone mass. Gender differences in long bone structure appear to arise because the bone cells of males and females function in different hormonal environments which affect their responses to mechanical loading. In girls, bone formation on the metacarpal periosteal surface essentially stops at puberty, and is replaced by formation on the endosteal surface, reducing endosteal diameter until about age 20. Bone strength is 60% greater in male metacarpals than in those of females because bone is added periosteally in boys and endosteally in girls. At menopause endosteal resorption resumes, accompanied by slow periosteal apposition, weakening cortical structure. Similar phenomena occur in such critical regions as the femoral neck. Another fundamental gender difference in skeletal development is that whole body bone mineral content increases in linear proportion to lean body mass throughout skeletal maturation in boys, but in girls there is a distinct increase in the slope of this relationship at puberty, when estrogen rises. Frost's hypothesis is that this reflects an effect of estrogen on bone's mechanostat set point, and this is increasingly supported by data showing that estrogen and mechanical strain act through a common pathway in osteoblast-like cells. If Frost's hypothesis is correct, the mechanostat is set for maximal effect of mechanical loading on bone gain during the 2-3 years preceding menarche. During the childbearing years, the set point is at an intermediate level, and at menopause, it shifts again to place the skeleton into the metabolic equivalent of a disuse state. The most direct approach to resolving this problem would be to simulate the putative effect of estrogen on the set point itself.