Carey marshals significant evidence supporting the idea that children have a series of core domains of knowledge that give them a rudimentary understanding of the world (Carey Reference Carey2009). Over time, that knowledge is expanded to provide more elaborate representations that ultimately support adult-like competence. For example, even infants are sensitive to basic causal events in which one moving object contacts a second and causes it to move. Early on, however, infants do not seem to perceive events that create a change of state in an object as causal.

What is impressive about this work is that a community of researchers has carefully mapped domains of knowledge to identify the limits of infant knowledge and to determine the progression of changes in that knowledge over time. From this pattern of knowledge change, it is possible to make detailed proposals about psychological mechanisms that may play a role in the development of knowledge, such as the role that language may play in the development of numerical competence.

This approach contrasts sharply with the standard mode of research on cognitive processing in adults. Most research on adults focuses on the structure of adult knowledge, but not its content. Whereas this strategy has enabled impressive gains in our understanding of cognitive processing, in many areas of research we have reached the limits of what we will learn about cognitive processing without at least some focus on the content of adult's knowledge.

Research on analogy provides an instructive example. In analogy, most researchers in the field subscribe to a basic set of principles that are encapsulated in Gentner's (Reference Gentner1983) structure mapping theory (Gentner & Markman Reference Gentner and Markman1997; Holyoak & Thagard Reference Holyoak and Thagard1989; Hummel & Holyoak Reference Hummel and Holyoak1997). This theory proposes that people are able to draw analogies between distant domains (such as the atom and the solar system) by representing relations using structured representations that explicitly encode the connection between a relation and its arguments. Analogies involve finding parallel relational structures that are found by matching identical relations and ensuring that the arguments of those relations are also placed in correspondence. Analogical inferences allow one domain to be extended by virtue of its correspondence to another by copying knowledge from the base domain to the target domain when that knowledge is connected to the correspondence between domains (Holyoak et al. Reference Holyoak, Novick, Melz, Holyoak and Barnden1994). This theory makes proposals only about the structure of people's knowledge, but not its content.

This framework has allowed the field to learn a lot about the way children and adults form analogies and use that knowledge to learn new information. For example, studies have demonstrated that children and adults are better able to process analogies when they involve higher-order relations that bind together a lot of representational structure than when they involve only low-order relations (Clement & Gentner Reference Clement and Gentner1991; Gentner & Toupin Reference Gentner and Toupin1986) . Other studies have demonstrated that these relational connections are crucial for licensing which information will be carried from base to target as an inference (Markman Reference Markman1997). Finally, this work supported the discovery that there is a psychological distinction between alignable differences in which each item has a corresponding property that differs, and nonalignable differences in which one object has a property that has no correspondence in the other (Markman & Gentner Reference Markman and Gentner1993).

Basic research on analogical reasoning has stagnated, however. There is some interest in factors that promote retrieval of analogies from memory, but the area is not a thriving source of new publications. A key reason for this decline in productivity is that it has become difficult to generate new predictions about analogical reasoning without knowing anything about the content of what people are reasoning about. Most of the studies in the 1980s and 1990s focused on novel pictures, analogies, stories, and insight problems that had little or no connection to the knowledge people had when they entered the laboratory (Gentner et al. Reference Gentner, Rattermann and Forbus1993; Gick & Holyoak Reference Gick and Holyoak1980).

The developmental research that Carey reviews suggests that it would be fruitful for researchers to focus more systematically on the content of people's knowledge to develop insights about analogy. There has been some research on the history of science that has examined the content of the knowledge used in analogies (Gentner et al. Reference Gentner, Brem, Ferguson, Markman, Levidow, Wolff and Forbus1997; Nersessian Reference Nersessian1987). There has also been work with expert designers that has looked at the influence of domain knowledge on the use of analogy in generating ideas for new products (Christensen & Schunn Reference Christensen and Schunn2007; Dunbar Reference Dunbar, Ward, Smith and Vaid1997; Linsey et al. Reference Linsey, Wood and Markman2008). However, there hasn't been any systematic study of particular domains of knowledge by communities of researchers. Consequently, there is little continuity in research from one laboratory to the next.

The impressive gains in our understanding of the development of children's knowledge in core domains has emerged from the commitment of a community of researchers to explore a common topic, despite theoretical disagreements about the underlying developmental processes. This success suggests that a similar strategy would be worthwhile in the study of cognition in adults.

Research on decision making provides an instructive case study here. As Goldstein and Weber (Reference Goldstein, Weber, Busemeyer, Hastie and Medin1995) point out, for many years research on decision making was dominated by studies of how people made choices from a set of risky gambles. Gambles were used because you did not need to know much about people's preference structures to know that they are likely to prefer more money to less money. In the modern study of decision making, studies of gambles have been supplanted by work on consumer behavior that explores what people know about brands of products and how they use that knowledge to make choices. In addition, research on decision making has begun to map out a broad set of people's motivational structures as a way of understanding how the waxing and waning of people's goals influences their preferences. The progress in this area suggests that a similar revolution ought to take hold in other areas of cognitive research, such as analogy, categorization, and problem solving.