The purpose of the study was to examine the effect of instruction using Dienes' perceptual variability principles on primitive generalization and mathematical generalization. The following was studied: the effect of achievement-to-criterion on one, two, or three non-symbolic embodiments of an objective using a selected class of variables on the achievement-to-criterion of a symbolic embodiment of a similar objective using an extended class and an isomorphic class of variables. Three instructional treatments in programmed format were devised for three sets of four objectives on the concept of equivalent fractions: T1, using an array of blue and red disks; T2, using the disks and an array of shaded and unshaded circles; and T3, using the disks, the circles, and partially-shaded rectangles. Seventy-eight fifth- and sixth-grade students were selected; 49 were randomly assigned to the three treatments and received instruction on all three sets of objectives. Twenty-nine students were randomly assigned to the three treatments but received instruction on only one set of four objectives. The proportion of T1 students reaching criterion on a primitive-generalization test and a mathematical generalization test was not less than the proportion of T2 students reaching criterion on the tests; T2 was not less than T3. It was concluded that instruction using one, two, or three concrete embodiments of a concept had essentially the same effect on the ability to generalize the concept. (Author/MS)