We examined the process of learning to see 2-dimensional forms, 3-dimensional solids and short distance in eight congenitally or early blind patients who acquired varying degree of sight after surgery. When we first asked them to recognize geometric solids such as cube, cylinder, cone and tetrahedron by using vison, it was found that this task was too difficult for them, even though they had the visual ability to identify 2-dimensional forms. The learning processes involved in attaining successful discrimination and identification are described here. The front distance of them became estimated mostly by the size of object at last. This study reveals the difficulty in transitioning from 2-dimensional discrimination to a 3-dismensional knowledge base.