Three studies were conducted to investigate reasons for the difficulties that children under 5 years of age experience with class inclusion tasks. The studies tested the claim that such tasks have a structural complexity that is beyond the children's processing ability. In experiment 1, class inclusion tasks were compared with other classification tasks. A total of 27 children between 4 and 8 years of age participated. It was found that the inclusion relation per se was a source of difficulty independently of question asymmetry and the use of descriptors for the superordinate class. Experiment 2 used a new task that was isomorphic to inclusion, and that avoided the features that had led to misinterpretation of the inclusion task. Subjects were 18 children 3 or 4 years old and 18 children 5 or 6 years old. It was found that children under 5 continued to have difficulty because they were unable to integrate relations between subclasses and between the subclasses and the superordinate class. Experiment 3, which involved 90 children of 3-7 years, used the easy-to-hard paradigm for studying capacity limitations. It was found that children's class inclusion reasoning was capacity-limited. Results are interpreted as showing that structure-mapping theory can account for processing load effects in classification tasks. (RH)