A nonequivalent control group design study, with a math unit on ratio and proportion as the treatment, was conducted to examine the validity of some written tasks as a measure of formal reasoning. Subjects in the study were 68 fifth grade students who enrolled in two classes in a small urban school. The experimental group learned a 12-lesson math unit, which dealt with techniques to solve written problems of ratio and proportion. During this period the control group continued to study the regular mathematics curriculum, which at that time was decimal fractions. Pre- and posttests included ratio and proportion tasks from different group tests of formal reasoning. Findings showed that school learning did not improve students' ability to solve x/y=c type tasks, but did not improve their ability to solve tasks of multiplicative compensation. Since no improvement was observed in the ability of students in the control group to solve any of the above tasks, it was concluded that performance of some of the tasks that are used in group tests of formal reasoning are highly dependent on specific previous learning, and as such they should not be used as a sole indicator of formal reasoning. (Author/TW)