Abstract

ABSTRACT Our past studies have found that for young children, the act of comparing perceptually similar entities can lead to insight into conceptual relations shared by the two entities. Thus, paradoxically, comparing perceptually similar entities can promote abstract responding in children. In this paper we ask whether there are limits to the benefits of perceptual similarity: that is, how close is too close. We examined four-year-olds' categorization patterns after comparing either two similar (but discriminable) or two nearly identical category exemplars. As expected, those who viewed the similar exemplars classified objects based on conceptual over perceptual commonalities. However, those who saw nearly identical exemplars classified objects based on perceptual similarity. These findings suggest that introducing category exemplars that are nearly identical inhibits abstraction and generalization. KEYWORDS: comparison, categorization, similarity, structural alignment. The process of comparison is a major force in children's learning and development. For example, if children compare sharks with tigers, they can discover that these two species have many deep commonalities despite their many surface differences. When the two animals are compared, their commonalities-that both make their living by hunting other animals, and are therefore fast and strong, and armed with razor-sharp teeth - stand out against their surface differences. Although comparing different examples is a good way for learners to see important abstractions, much research has shown a relational shift in children's comparison processing (Gentner, 1988; Ratterman & Gentner, 1998). For example, a 5-year-old asked to say how a camera is like a tape recorder says "Both are black;" and asked to say why a cloud is like a sponge, says "Both are round and fluffy" (Gentner, 1988). That is, young children focus on surface similarities, rather than on the deeper commonalities of greater utility and importance. In contrast, a 9-year-old focuses on relational commonalities, such as common function ("Both are used to record things you want to remember") or causation ("Both hold water and later give it back"). Thus, the very people most in need of learning-young children-may not be able to take advantage of analogical comparison. Fortunately, our research has shown that young children can gain the benefits of comparison, if the examples are very similar (Gentner & Namy, 1999; Namy & Gentner, 2002). We found that comparing instances of a category enables preschool children to move beyond compelling perceptual commonalties to extract deeper conceptual commonalities. We began with a standard forced-choice wordextension task. We taught 4-year-old children a new name (e.g., riffle) for an object (e.g., an apple) and then asked them to indicate which of two alternatives would also be called a "riffle". We found, as in previous studies, that children's choice was governed largely by perceptual, rather than conceptual, similarity: that is, they tended to choose a perceptually similar object from a different category (e.g., a balloon) rather than a perceptually distinct match from the same category (e.g., a banana). The same result occurred if the initial object was a pear instead of an apple: children chose the perceptually but not conceptually matching alternative. But when a third group of children was given both the apple and the pear and told "These are both riffles. Can you see why both are riffles?" their choice shifted to the conceptual match. These results are baffling within a traditional stimulus-generalization account: in this account, inviting a perceptual comparison should lead to greater reliance on the shared perceptual commonalities, and more perceptual-match responses relative to either standard object presented alone. However, these effects of comparison are consistent with the structure-mapping theory of similarity and analogy (Gentner, 1983; Gentner & Markman, 1997; Gentner & Medina, 1998). …