Middle school prealgebra students (n=157) learned about functions in a 20-day unit that emphasized: (1) representing problems in multiple formats, (2) anchoring learning in a meaningful thematic context, and (3) discussing problem-solving processes in cooperative groups. They produced smaller pretest-to-posttest gains on symbol manipulation tasks, such as solving equations, and larger gains in problem representation tasks, such as translating word problems into equations, tables, and graphs, than did a comparison group taught in the standard way. Although the groups did not differ in their pretest-to-posttest gains in calculating correct answers for word problems, the treatment group produced a larger gain in using mathematical representations while solving word problems than did the comparison group. The same pattern of results was obtained for lower-achieving students and language-minority students. Implications for cognitive theory and educational practice are discussed. Contains 72 references. (Author)