On the geometric interpretation of seminormality