If I may make a personal remark, one sign of old age is that people ask you to write commentaries on new(er) work. In the present case the invitation for me to write something may be linked to the publication of the Medin and Schaffer (1978) context theory of categorization model more than 30 years ago and/or the Smith and Medin (1981), Categories and Concepts book, almost as old. This ought to provide enough distance to view cumulative progress in this area of research and theory. Of course there was more than a little earlier work by Posner and Keele (1968), Reed (1973), and Smith, Shoben and Rips (1974) relevant to models and by Rosch, Mervis and others (e.g. Rosch, 1973, 1975; Rosch & Mervis, 1975; Rosch et al., 1976) laying out basic levels and goodness of example or typicality effects that reverberated through the cognitive sciences. The basic levels work was so important that it now has the status of being presupposed in developmental studies on the interaction of language and conceptual development (e.g., Waxman, 1989, 2002; Waxman & Lidz, 2006).

One way of assessing progress in an area is to evaluate how it is doing with respect to narrowness and insularity versus breadth. Cutting edge research seems like something that is inherently good, but it may be useful to examine what is being cut and how that edge is related to broader configurations.

In 1931, A. R. Luria traveled to rural Uzbekistan with a question: Is thinking influenced by the social and cultural environment? The fieldwork he did there (Luria 1976) was likely the first attempt to answer this question using methods of experimental psychology. Rural folk in Uzbekistan were at that time in the midst of a socioeconomic transition that involved, among other things, the collectivization of agriculture and the growth of schools and literacy. Luria saw in this a natural experiment. He found groups of participants with different levels of involvement in this transition – including different levels of schooling and literacy – and gave them a variety of cognitive tasks. Some of these were designed to elicit deductive inferences from premises like “precious metals don't rust” and “gold is a precious metal.” Luria found that subjects with no formal schooling often balked at such problems, rejecting the premises and saying, for example, that “one can speak only of what one has seen.” In contrast, subjects who had been to school were more likely to use the hypothetical premises to draw conclusions (e.g., “gold doesn't rust”) with no obvious basis in personal experience. Luria also exercised some experimental control, manipulating whether the content of an argument was familiar or unfamiliar to participants. Participants who had not been to school treated the two kinds of content differently, drawing conclusions from premises more often when the content was familiar than when it was unfamiliar.

INTRODUCTION

Like adults, children use categories as a basis for inductive inference. Having learned that some property is true of some individual (e.g., “My dog, Magic, likes marshmallows”), a child might assume both that other members of the same category (dogs) share this property and that members of other, similar categories (e.g., cats) might also share this property. In short, inductive inference may be guided by and reflect categorical relationships, hence the term “category-based induction” or CBI. Cognitive and developmental researchers have used a category-based induction paradigm not only to study the use of categories in reasoning, but also to draw inferences from patterns of reasoning about the nature of conceptual structures themselves (Atran et al., 2001; Carey, 1985; Gelman & Markman, 1986; Gutheil, Bloom, Valderrama, & Freedman, 2004; Gutheil, Vera, & Keil, 1998; Hatano & Inagaki, 1994; Inagaki, 2002; Johnson & Carey, 1998; Medin & Smith, 1981; Ross, Medin, Coley, & Atran, 2003; Waxman, Lynch, Casey, & Baer, 1997).

One of the most influential examples of the use of inductive projections to draw inferences about conceptual organization comes from research in the domain of folk biology involving categories of living things, including humans, nonhuman animals, and plants. Developmental evidence has revealed, in particular, certain systematic asymmetries in inductive strength among these categories. For example, researchers have found that children are more willing to project properties from humans to dogs than from dogs to humans.

SIMILARITY AND INFERENCE

The Property Problem

Ever since Goodman's (1972) warnings about the “false friend” similarity, psychologists have been cautious about explaining inductive inference in terms of resemblance. Innumerable shared properties unite any pair of objects, inviting their dubious certification as similar along with the inference that some further property is also shared. Genuine resemblance appears to be a three-place predicate, relating two objects only in the context of a set S of properties (often implicit) that determine the respects in which objects may be compared.

When it is plausible that the same set S is brought to mind in judging similarity and inductive strength, it can be revealing to explain the latter in terms of the former. Thus, famous Linda resembles the average feminist bank teller more than the average bank teller; and whatever properties support this judgment are likely also in place when people judge her chance of being one or the other. Indeed, the probabilities people assign in Linda-like problems are accurately predicted from the relevant similarities rated independently (Shafir, Smith, & Osherson, 1990).

Wittgenstein's admonition “don't think, but look” has had the important effect of stimulating psychologists to reconsider their common practice of equating concept formation with the learning of simple definitional rules. In the early 1970s, psychologists like Eleanor Rosch (e.g., Rosch, 1973), responding to the difficulty of identifying necessary and sufficient conditions for membership of all kinds of categories, proposed alternative models of category representation based on clusters of correlated features related to the categories only probabilistically. Without denying the importance and impact of this changed view of concepts (reviewed, e.g., by Smith & Medin, 1981), we think that in certain respects the “don't think, but look” advice may have been taken too literally. There are problems with equating concepts with undifferentiated clusters of properties and with abandoning the idea that category membership may depend on intrinsically important, even if relatively inaccessible, features. For example, on the basis of readily accessible properties that can be seen, people presumably will not judge whales to be very similar to other mammals. However, if they think about the fact that whales are mammals not fish, they will probably acknowledge that with respect to some important, although less accessible property or properties whales are similar to other mammals.