Matrix tasks designed to assess the logical properties of multiple classification and multiple seriation were administered to 105 first, third, and fifth grade children. The tasks included cross-class and double series matrices, each of which had a reproduction and a transposition instructional set, and a revised set of cross-class and double series matrix tasks using cylinders and squares. Task presentation order effects and sex differences were not noted. On the traditional matrix tasks, grade level differences reached statistical significance for cross-class tasks only; reproduction of the double series matrix was significantly easier than transposition; no intermatrix reproduction task comparisons were significant, while cross-class transposition was significantly easier than double series transposition. Approximately 50% of the children who passed a cross-class task and failed double series responded with a cross-classification in the traditional blue cylinders format. When stimulus conditions, number of possible correct solutions, and components of visual memory were all held constant on the revised matrix tasks, cross-class was significantly more difficult at the third grade level in both green square and blue cylinder formats. The revised green square and cylinder formats were equally difficult. Accordingly, cross-classification appeared to be more difficult than double seriation for children of transitional status. (Author/RD)