In an effort to cross-validate a Japanese study, a 16 item test of logical connectives was administered to 223 boys and girls at each grade level (third through ninth grades) and two groups of college sophomore girls. The purpose of the test was to assess their understanding of class inclusion and exclusion, class intersection, and class union. Half of the groups received a test in which set elements were pictures; the other half had words as set elements. Although there were significant differences betwee n grades for all three types of questions, (a) inclusion and exclusion are understood by a majority of even the youngest children, (b) intersect is understood by a majority of all bu t the youngest children, and (c) union is not understood by the majority of subjects except at the college level. Those taking the test, in which set elements were pictures, performed better than those taking the test in which set elements were words. These results, in general, support the findings of the Japanese study, although Japanese children as a group scored higher than American children. The most complete explanation of the present data seems to be an analysis of performance in terms of component operations for processing and storing information. (MH)