Evidence for a hypothalamic defect in septo-optic dysplasia.