Children’s Acceptance and Use of Unexpected Category Labels to Draw Non-Obvious Inferences

Abstract

Children’s Acceptance and Use of Unexpected Category Labels to Draw Non-Obvious Inferences Vikram K. Jaswal (jaswal@psych.stanford.edu) Department of Psychology, Jordan Hall, Bldg. 420, Stanford University Stanford, CA 94305-2130 USA Ellen M. Markman (markman@psych.stanford.edu) Department of Psychology, Jordan Hall, Bldg. 420, Stanford University Stanford, CA 94305-2130 USA Abstract Language provides an efficient, uniquely human way of transmitting non-obvious category information between individuals and across generations. To explore whether it can serve this purpose even for very young children, we conducted two experiments: one with 24-month-olds, and the other with preschoolers. Children made non-obvious inferences about perceptually misleading animals. Those who heard the animals called by counter-intuitive category labels made inferences different from those who did not hear the labels, demonstrating an important influence of language on thought, even at 24 months of age. However, preschoolers appear to have been less influenced than toddlers, suggesting that there are limits on children’s willingness to accept anomalous category labels. Introduction Categorization is fundamental to human cognition, enabling communication and serving as the basis for the representation of objects and for predicting and explaining their behavior (Anglin, 1977; Markman, 1989). Children as young as 3.5 months (Eimas & Quinn, 1994), adults (Rosch et al., 1976), and non-human animals (e.g., Freedman et al., 2001) alike readily use perceptual similarity to determine the category to which something belongs. But when reasoning about an object, or explaining or predicting its behavior, perceptual appearance is not always criterial of category membership. For example, even though eels look like snakes, in order to more accurately characterize their ancestry, behavior and physiological processes, experts categorize them as fish. Given sufficient experience, children as young as 30 months can form non-obvious categories by noting causal (Gopnik & Sobel, 2000) or functional (Kemler Nelson et al., 2000) regularities between objects. Under some circumstances, even non-human animals can learn to categorize objects in perceptually non-obvious ways (Herrnstein & DeVilliers, 1980). However, recognizing non-obvious similarities can be a slow and laborious process, often requiring experience that is difficult to obtain. Moreover, it requires every individual in every generation to have the experience for him or herself (Tomasello, 1999). Another, arguably more reliable and efficient way to obtain non-obvious category information is through language (Gelman et al., 2000): When a trusted source uses an unexpected category label for an object, it reflects a particular perspective that others have found useful when thinking and reasoning about that object in the past, and it can cause us to revise a classification immediately. For language to have this effect, however, listeners may have to give up a compelling, perceptually based classification in favor of a classification they do not immediately understand. As adults, we can accept linguistically provided non- obvious classifications (e.g., a whale is a mammal, not a fish) because we implicitly assume that something deeper than surface similarity binds category members together (Medin & Ortony, 1989). Such an essentialist assumption would facilitate the rapid uptake of non-obvious category terms. However, it is not clear whether very young children also expect categories and category terms to encode more than surface similarities. For example, it has been argued that children begin with or quickly develop an expectation that category labels encode similarly shaped objects (e.g., Imai, Gentner, & Uchida, 1994; Smith, 1999). Furthermore, whereas children readily learn words for basic-level categories (e.g., “table”), they have difficulty learning words for superordinate categories (e.g., “furniture”), which have fewer perceptual features in common than basic-level ones (e.g., Horton & Markman, 1980; Mervis & Rosch, 1981; Rosch et al., 1976). Finally, young children sometimes object to parental attempts to correct perceptually based categorization errors (Mervis, 1987). In two studies, we investigated children’s willingness to accept perceptually counter-intuitive classifications on the basis of linguistically provided information alone. In our first study, we asked this question of 24-month-olds. Gelman and her colleagues have found that children as young as 32 months can make inferences on the basis of linguistic information rather than perceptual appearance (Gelman & Coley, 1990; Gelman & Markman, 1986), but their procedure involves memory demands, verbal responses, and linguistic comprehension abilities beyond those of younger children. Our procedure uses an imitation paradigm that minimizes these task demands (see also Baldwin, Markman, & Melartin, 1993; Mandler & McDonough, 1996; Welder & Graham, 2001).