The study aimed at finding the way to facilitate students’ ability to solve proportion word problems. Previous research suggested two different ideas on abstract framework: abstract framework was less effective than examples; abstract representations would be effective in complex problems. Therefore, we focused on the abstractness of framework and the complexity of the problems. We compared problem solving performance of students in three classes with different abstractness in teaching how to solve proportion word problems. In “no formula” class (N=20), the teacher only showed correct expressions and answers; in “abstract formula” class (N=24), the teacher showed abstract formula to solve proportion problems; and in “situation formula” class (N=23), the teacher showed more grounded formula. The results indicated that the students in “no formula” did better than those in “abstract formula” condition. The effects of abstract frameworks and examples, and the complexity and transformation of the problems were discussed.