Structual models of geometry achievements were compared across age levels of subjects. College and junior-highschool students read a geometry text with figural illustrations on the computer screen. Then multiple choice tests were administered on the same screen. Data for this study was drawn from tests for imagery rotation, symmetry ideas, analogy and tests for ability to induce the formulae for obtaining the area of a triangle and that of a trapezoid by producing a parallelogram from two symmetrical triangles or trapezoids. To see how students monitored their own progress toward solution of geometry problems, a rating scale for self-efficacy was inserted after each of those geometry test questions. Correlational findings showed that for older students, correct estimation for their own analogy performance was related to correct ratings for their own success in inducing a formula for obtaining the area of a trapezoid, while for younger ones, correct rating for their own ability to mainipulate imagery predicted correct rating for their understanding of the formula for obtaining the area of a trapezoid. Results of path-analyses showed that for older subjects, their correct estimation for understanding of the point symmetry ideas has predicted correct ratings for their own success in inducing the formula for obtaining areas of treapezoids and triangles. The role of analogical transfer in geometry learning has been discussed.