Over the past decades, a major challenge to a widely accepted view of the human mind has developed across several disciplines. According to a long predominant view, human beings are endowed with a general set of reasoning abilities that they bring to bear on any cognitive task, whatever its specific content. Thus, many have argued, a common set of processes apply to all thought, whether it involves solving mathematical problems, learning natural languages, calculating the meaning of kinship terms, or categorizing disease concepts. In contrast to this view, a growing number of researchers have concluded that many cognitive abilities are specialized to handle specific types of information. In short, much of human cognition is domain-specific.

The notion of domain specificity is not new. Indeed, intriguing (although brief) hints of domain specificity emerge in the epistemologies of Descartes and Kant and in the psychologies of Thorndike, Vygotsky, and de Groot. For example, in Mind in Society, Vygotsky argues that

Still, in recent years, increased and detailed attention has turned toward the question of domain specificity.

Ten years ago, Jerry Fodor published The Modularity of Mind, a book that received much well-deserved attention. His target was the then-dominant view according to which there are no important discontinuities between perceptual processes and conceptual processes. Information flows freely, “up” and “down,” between these two kinds of processes, and beliefs inform perception as much as they are informed by it. Against this view, Fodor argued that perceptual processes (and also linguistic decoding) are carried out by specialized, rather rigid mechanisms. These “modules” each have their own proprietary data base, and do not draw on information produced by conceptual processes.

Although this was probably not intended and has not been much noticed, “modularity of mind” was a paradoxical title, for, according to Fodor, modularity is to be found only at the periphery of the mind, in its input systems. In its center and bulk, Fodor's mind is decidedly nonmodular. Conceptual processes – that is, thought proper – are presented as a big holistic lump lacking joints at which to carve. Controversies have focused on the thesis that perceptual and linguistic decoding processes are modular, much more than on the alleged nonmodularity of thought.

In this chapter, I have two aims. The first is to defend the view that thought processes might be modular too (what Fodor [1987: 27] calls “modularity theory gone mad” – oh well!). Let me however echo Fodor and say that, “when I speak of a cognitive system as modular, I shall… always mean ‘to some interesting extent’” (Fodor, 1983: 37).

The point of a cognitive approach to cultural representations is to put forward a series of causal hypotheses in order to account for certain features of cultural phenomena. Central to such an inquiry is the notion of cognitive constraints; given the general properties of human minds, certain types of representations are more likely than others to be acquired and transmitted, thereby constituting those stable sets of representations that anthropologists call “cultures.” To many anthropologists, cultural phenomena seem to lie outside the scope of cognitive constraints, due to three types of reasons: their ontological status, their variability, and their transmission. Cultural anthropology generally focuses on abstract systems of “symbols,” “codes,” or “meanings,” the properties of which are supposed to be independent of the way they are represented in human minds. Second, cultural representations are considered as intrinsically variable; as a consequence, it seems difficult to appeal to universal properties of human minds in their description or explanation. Finally, the content and organization of cultural representations, in competent members of a culture, seem to be entirely constrained by what subjects were taught through social interaction. Against these assumptions, I will take as a starting point the following principles:

1. Cultural systems can and must be studied as sets of mental representations acquired and stored by human minds, because acquisition and memorization processes impose strong constraints on the contents and organization of cultural representations.

2. Their undeniable variability should not lead us to ignore important recurrent features, which deserve an explanation.

3. […]

Laypersons, teachers, and psychologists believe that learning takes place in people's minds. They all believe that learning takes place as a consequence of instruction, if we interpret instruction in its widest sense (Atran & Sperber, 1991).

Western laypersons have folk psychology notions of the mind and its functioning. But their notions are not all that naive because Western adults have considerable practice at learning and, even more important, thinking about learning. Entry into school, the culturally designated place for learning to take place, generally begins at age 6. Thus students who have finished their high school education have been in formal learning situations for 12 years. Over a considerable period of time, then, they have practiced what they and others think it takes to learn. In the sense that they have not studied formally about the mind and learning in, say, psychology courses, they can be considered laypersons. But it is a big stretch to call them laypersons and holders of a naive psychology after they have been in places where learning is the main goal and where they have been reflective about what that learning is.

Teachers are professionals who have been educated to teach in such a way as to cause change in others’ minds. That change is learning. The cause of that change is thought to be teaching. The differentiation between teaching and learning is often unclear, even in the educational literature. Some of our best friends refer to it as the teaching/learning process.

In their first few years of life, children are making sense of the world at two levels at once: at the fine-grained level of everyday object categories (deciding which things are trees and which are dogs and which are cookies), and at a broader level that some have called commonsense “theories.” Both are remarkable achievements. First, consider categorization. If children's vocabulary is any indication, by the age of 6 they have carved up the world into thousands of distinct categories (Carey, 1978). Many children undergo a vocabulary “explosion” at roughly 18 months of age (Halliday, 1975; McShane, 1980; Nelson, 1973), when the rate of acquisition suddenly rises exponentially. One child studied in detail by Dromi (1987) produced as many as 44 new words in one week, and roughly 340 new words in her first 7 months of speech. No other species acquires symbolic communication at this rate. Even studies that successfully teach apes to acquire sizeable vocabularies in sign language are incomparable, with no noticeable vocabulary explosion (e.g., after more than 4 years of exposure to sign language, Washoe acquired only about 132 signs; Gardner & Gardner, 1989).

At around the same time that children learn to classify individual entities and undergo rapid vocabulary growth, they are developing broad systems of belief about the world. Not only do children learn to identify certain objects as “dogs,” but they also learn that dogs belong to the class of animals, and that animals engage in characteristic biological processes such as growth, inheritance, and self-generated movement. Children are learning about physical laws such as gravity, mental states such as dreams, and social relationships within units such as families.

This is a two-part chapter, the first part of which attributes a moral domain to the infant and presents a model of the domain; and the second part of which shows how the primitives of the domain provide an invariant form in which to express the diverse moral beliefs of different cultures.

Part I: Infant's model

The subject matter of morality is social behavior, the relation among individuals, ultimately how one individual treats another. Human social competence is highly developed, and the ability to make moral judgments about social behavior is part of the competence. Some concepts of moral judgment are not unique, but are shared. For instance, the attribution of intention, which is central to morality, is a fundamental component of theory of mind (Leslie, 1987; Premack & Woodruff, 1978; Wellman, 1990; Wimmer & Perner, 1983), while aesthetics, which enters into moral judgment, is a hidden yet significant component of pedagogy (Premack, 1984, 1991). However, morality is not simply constructed from pieces of other social competences. Judgments concerning the “rightness” or “wrongness” of acts, an individual's “rights” and “responsibilities,” the concept of “ought,” are sui generis, and cannot be derived from concepts belonging to other parts of social competence.

What is the source of these distinctly moral concepts, of “right,” “wrong,” “ought,” “responsibility,” and the like? Are they irreducible primitives, or can we find their origin in other sources?

The human sciences can be characterized as a working out of two sets of tensions in interpreting human experience: tensions between the biological and the social and between the particular and the general. In this chapter, I examine the relations between two lines of thinking, each commanding increasing attention among psychologists and social scientists, that appear to be contradictory. The first, a position I term conceptual rationalism, seeks biological foundations for specific concepts that are central and, perhaps, universal in human development. The second, a position that has come to be known as situated cognition, argues that knowledge is acquired in and attuned to specific social and historical situations and that conceptual development can be understood only in terms of the situational contexts of action. I argue here that the rationalist and situationist views, far from being contradictory, share important epistemological assumptions and can – perhaps must – be combined to provide a theory of cognitive development and functioning. I develop a view of learning and development that I call situated rationalism, illustrate it with some examples from mathematics and science learning, and consider its implications for education.

The conceptual rationalist argument: Biological preparedness

In recent years, there has been a reassertion of interest in the biological basis of human learning and thinking (e.g., Gelman & Carey, 1991; many chapters in the present volume). This new line of thinking grows out of recent research on language development, concept development in infancy and early childhood, and animal cognition and learning. The core proposal of those pursuing this line of thinking is that there exists a set of biological constraints on learning and cognitive development.

By establishing that domain-specific machinery is necessary to explain human cognitive performance, psychologists who advocate modular or domain-specific approaches have found themselves in an unanticipated situation. Metaphorically speaking, it is as if they had laboriously built a road up one side of a nearly impassable mountain range into unexplored terrain, only to find themselves met at the top by a foreign road construction crew – evolutionary functionalist researchers – who had been building a road upward to the same destination from the far side of the mountains. Quite unexpectedly, cognitive psychologists find their field intimately connected to a whole new intellectual landscape that had previously seemed remote, unfamiliar, and all but irrelevant. Yet the proliferating connections tying together the cognitive and evolutionary communities promise to transform both fields, with each supplying necessary principles, methods, and a species of rigor that the other lacks. Although the sudden conjunction of these two communities has led to the customary level of mutual misunderstanding, the long-run significance of these developments is unmistakable. From this emerging integrated perspective, the domain-specific mechanisms or modules cognitive psychologists have been studying can be readily recognized for what they are – evolved adaptations, produced by the evolutionary process acting on our hunter-gatherer ancestors (Cosmides & Tooby, 1987).

Natural selection and ancestral environments

Viewed from a more encompassing scientific framework, the confluence of these two research communities seems inevitable (Tooby & Cosmides, 1992). The human brain did not fall out of the sky, an inscrutable artifact of unknown origin, and there is no longer any sensible reason for studying it in ignorance of the causal processes that constructed it.

Introduction

How is knowledge stored and organized in the human mind? By knowledge I do not mean school knowledge or book knowledge, but the basic everyday knowledge that provides us with a frame of orientation in daily life, in interaction with other people, in our natural modus operandi.

In my view, there can hardly be a better way of approaching this question than by analyzing language (and languages). “Languages are the best mirror of human mind … and precise analysis of the meanings of words would allow us – better than anything else – to know the operations of the mind” (Leibniz, 1949: 368).

In language, grammar provides a basic framework for the interpretation of the world and of human existence in the world, with its fundamental semantic categories such as person, number, gender, tense, aspect, mood, “evidentiality,” degree, and so on (cf. Jespersen, 1924), whereas the lexicon divides and organizes the “contents of the world” into more or less coherent and selfcontained domains. By studying the structure of the lexicon, we can discover what these domains are and how they are organized, and thus reveal fundamental aspects of human interpretation of the world and the organization of knowledge and experience.

Our understanding of Agency is, in part, the result of domain-specific learning. The nature of this domain-specific learning needs to be understood in relation to the organization of information processing in the infant. As a result of adaptive evolution, the infant is a specialized processor of information with an architecture that (in part) reflects properties of the world. On this assumption, it should be possible to establish links between properties of the world, processing subsystems specialized for tracking those properties, and domains of knowledge. It is argued in the case of Agency that three main classes of world properties are reflected in three corresponding processing subsystems producing three distinct levels of knowledge. These three related triples are, respectively, mechanical Agency, actional Agency, and attitudinal Agency. Each of these three linked property classes, processing subsystems, and knowledge levels are discussed in turn but the focus will be mainly on mechanical Agency. In developing these ideas this discussion deals more generally with the nature of early mechanical understanding and its relation to conceptual development. A number of the ideas put forward in Leslie (1988) are revised and extended.

One lesson of cognitive science is that different types of knowledge often have different locations within the global organization of human information processing. In development, different types of commonsense knowledge may originate from different locations in core cognitive architecture. Early mechanical understanding and the notion of Agency can be studied within such a framework.

This is a book about domain-specific cognition – the proposal that at least some human conceptual abilities are specialized for some types of contents and not for others. In this chapter we address the development of a single domain: everyday understanding of the mind. We suggest that this development is best understood as the formulation of a succession of naive theories. Moreover, this “theory theory” can help to characterize cognitive domains more generally and to explain domain-specific development. Our chapter, therefore, joins company with a number of recent discussions drawing parallels between theory change in science and cognitive development (Carey, 1985, 1988; Gopnik, 1984, 1988; Karmiloff-Smith & Inhelder, 1974; Keil, 1987; Wellman, 1985, 1990).

Theory of mind

In the past five years there has been an explosion of interest in children's early understanding of the mind (Butterworth, Harris, Leslie & Wellman, 1991; Astington, Harris, & Olson, 1988; Frye & Moore, 1991; Whiten, 1991). The research tackling this question has come to be called “children's theory of mind.” This title reflects a leading explanatory position among investigators in this area. According to this position, our everyday conception of the mind is an implicit naive theory; children's early conceptions of the mind are also implicit theories, and changes in those conceptions are theory changes. We refer to this explanatory position as the “theory theory.”

If the theory theory has any content, it should be falsifiable. It should lead to empirical predictions about the course of development, and these predictions should be different from those of alternative accounts.

There is a growing body of evidence that infants attend selectively to some fundamental aspects of number and music. Such findings suggest that attention to and learning about number and music are perhaps due to the presence of innate, skeletal principles in each domain. In our chapter, we develop this position while showing how it is consistent with the different ways that cultures support learning and development in specific knowledge areas. Pairing considerations of number and music enables us to show that domain specificity and cultural variation need not be treated as antithetical.

Whereas it is still common for scholars in some fields to assume that “primitive” peoples lack a concept of number (see the following discussion), there is wide acceptance of the idea that all peoples develop musical competence, mainly because the latter can happen without benefit of formal instruction, use of special symbol systems, or the need to represent abstract, relevant dimensions like pitch, key, harmony, rhythm, and so forth. Indeed, accounts of the evolutionary function of the human music capacity often include the idea that, in preliterate societies, music serves to efficiently organize information that cannot be written down. For example, Gardner (1983) describes a possible role for music in organizing religious rites and work groups in the Stone Age, and Sloboda (1985) hypothesizes that music provides a mnemonic framework within which the structure of cultural knowledge and societal relations is stored and communicated. Whatever the role of music in preliterate groups, Donald (1991) points out that these societies always have complex rituals based on some form of music.

In this chapter, I distinguish among three types of cognitive activity: the cognitive processes of children; the cognitive processes of individual scientists; and the collective, public enterprise that constitutes the history of science. I propose that when we focus on individual scientists rather than the organized discipline within which they work, there are some neglected similarities between the way that they think and the way that young children think. On the other hand, I argue against the proposal that there are important and interesting analogies between the cognitive processes of children and the collective history of science. In particular, I argue against the claim that individual children show patterns of cognitive development in early childhood that resemble the shifts in theoretical stance that are found in scientific disciplines viewed across several decades or centuries.

On general grounds, this analogy is implausible. First, science as a progressive enterprise is a highly specialized activity involving few individuals even in this century and virtually none throughout most of human history. By contrast, many facets of cognitive development in children are universal, and probably have been throughout most of our history. Second, theory change in science is a collective activity that depends on competitive, interindividual and intergroup communication. Yet, cognitive development in individual children is not in any obvious sense a collective or competitive enterprise.

The standard basis for the alleged analogy is that science exhibits four important cognitive traits: theoretical economy, corrigibility, domain specificity, and an absence of magical thinking. First, a large set of observations is explained – economically – in terms of a small set of theoretical postulates or principles.

The revival of interest in domains of cognition, especially in the contexts of cross-cultural and developmental studies, is a welcome new awareness of how different sorts of concepts and belief systems might become tailored to particular kinds of lawful regularities in our physical and social worlds. To make much progress, however, this new emphasis requires more precise distinctions between types of domains and better descriptions of the ways in which each type might vary across development and cultures. We can talk about domains as rarefied as a cardiologist's knowledge of arrhythmia to those as commonplace as everyday folk psychology. Domains can vary from the highly concrete causally rich relations in a naive mechanics of physical objects to the highly abstract noncausal relations of mathematics or natural language syntax. Lumping together all of these different sorts of domains so as to have similar effects on cognitive development is likely to be misleading and uninformative. In this chapter, I consider some distinctions and their implications for questions concerning the origins of concepts.

The focus of this chapter is on the emergence of biological thought. Concepts of living things may offer an especially clear illustration of how domains might be involved in the origins of more specific concepts, and conversely of how specific concepts become intertwined within larger belief systems. In addition, views of concept structure and use have increasingly invoked the special importance of those belief systems known as theories.

This volume presents research and theoretical discussion on domain specificity in human thought. “Domain specificity” is the idea that all concepts are not equal, and that the structure of knowledge is different in important ways across distinct content areas. The notion of domain specificity has received much attention in recent years, but surprisingly it has not yet been given a unified treatment. A sense of how widely this concept has been discussed can be seen by viewing the range of disciplines represented in this volume: philosophy, psycholinguistics, linguistics, cultural anthropology, biological anthropology, developmental psychology, cognitive neuroscience, and education. We hope that the volume will thus be of interest to scholars in a broad range of disciplines.

The present volume is based on a conference, “Cultural Knowledge and Domain Specificity,” held in Ann Arbor, Michigan, October 13–16,1990. The conference was an attempt to discover if the notion of domain specificity could be discussed profitably (even intelligibly!) across disciplinary lines. Most important, we had the strong hope and belief that knowing more about other traditions could be valuable in informing our own local interests. In preparation for the conference, participants distributed their papers well in advance. Accordingly, we requested attendees to devote little time to formal presentations so that we could focus on discussions, both formal among the entire group and informal among participants over coffee breaks and on walks around campus. We were delighted to discover that we had in fact a lot to say to one another. We believe that this is evident in the chapters that follow.

The understanding of some conceptual domains is clearly helped by a domainspecific competence, as many of the chapters in this volume establish. This is the case for middle-sized objects physics (Carey & Spelke), living kind classification (Atran, Keil), theory of mind (Gopnik & Wellman, Leslie), and numerosity (R. Gelman & Brenneman). Other conceptual domains lack such an underlying domain-specific competence, including, for instance, astronomy, particle physics, computer technology, or (as the chapters by Boyer and Vosniadou suggest) religious representations and cosmology. In yet other domains the question is moot, as for instance in the case of chemistry or artifacts. The issue is quite undecided too in the domain of social categories (Turiel, 1983).

How does knowledge develop in domains for which there is no ad hoc innately specified competence? The mechanism invoked most often is analogy and transfer from better grounded domains. In particular, it has been suggested that acquisition of social categories is based on a transfer from the biological domain rather than on a domain-specific competence (Atran, 1990; Boyer, 1990; Rothbart & Taylor, 1990). In this chapter, I present evidence and arguments suggesting that the acquisition of social category does not depend on such a transfer. This could be so either because there is a domainspecific innately specified competence just for the social domain, or because social categories fall from the start within the extension of wider competences. I discuss these and other possibilities in the conclusion.

There is no well-described psychological mechanism of analogical transfer but there are lots of plausible examples. Many rich examples are found in the social sciences, in particular the history of science and anthropology.