The contrast diversity effect: Increasing the diversity of contrast examples increases generalization from a single item.

In five experiments, we established and explored the contrast diversity effect-the effect of diversity of negative evidence on inductive inferences drawn from a single observation of a target exemplar. In Experiments 1 through 3, we show that increasing the diversity of negative evidence in a contrasting category led people to infer that a target exemplar corresponded to a higher level category and led to greater generalization of a novel property associated with the target. Further, we demonstrated two boundary conditions in which the effect only occurred when the negative evidence was consistent with a higher level category that both united the contrast exemplars and distinguished them from the target (Experiment 4) and when the negative evidence and the target shared an obvious parent category (Experiment 5). Taken together, these findings demonstrate that increasing the diversity of negative evidence alone increases generalization from a target so long as the negative evidence is drawn from a single contrast category that excludes, but shares a common parent with, the target. Implications for general theories of induction are discussed. (PsycINFO Database Record (c) 2020 APA, all rights reserved).

Do Americans Have a Preference for Rule-Based Classification?

Six experiments investigated variables predicted to influence subjects' tendency to classify items by a single property (rule-based responding) instead of overall similarity, following the paradigm of Norenzayan et al. (, Cognitive Science), who found that European Americans tended to give more "logical" rule-based responses. However, in five experiments with Mechanical Turk subjects and undergraduates at an American university, we found a consistent preference for similarity-based responding. A sixth experiment with Korean undergraduates revealed an effect of instructions, also reported by Norenzayan et al., in which classification instructions led to majority rule-based responding but similarity instructions led to overall similarity grouping. Our American subjects showed no such difference and used similarity more overall. We conclude that Americans do not have a preference for rule responding in classification and discuss the differences between tasks that reliably show strong rule or unidimensional preferences (category construction and category learning) in contrast to this classification paradigm.