One event cannot be more probable than another that includes it. Judging P(A & B) to be higher than P(A) has been called the conjunction fallacy. This study examined a disjunction fallacy. Ss received brief case descriptions and ordered 7 categories according to 1 of 4 criteria: (a) probability of membership, (b) willingness to bet on membership, (c) inclination to predict membership, and (d) suitability for membership. The list included nested pairs of categories (e.g., Brazil–South America). Ranking a category more probable than its superordinate, or betting on it rather than its superordinate, is fallacious. Prediction, however, may be guided by maximizing informativeness, and suitability need conform to no formal rule. Hence, for these 2 criteria, such a ranking pattern is not fallacious. Yet ranking of categories higher than their superordinates was equally common on all 4 criteria. The results support representativeness against alternative interpretations.