I suggest several ways that externalism could influence psychological theorizing. Externalism could just capture our everyday intuitions about concepts and meanings. More profoundly, it could enter into psychology through evolutionary theory, guide our hypotheses about conceptual abilities, and, most significantly, it could influence our accounts of learning and conceptual change.

Millikan's account of substance concepts fails to do away with features. Her approach simply moves the suite of relevant features into an encapsulated module. The crux of the problem for scientists studying human infants and nonhuman animals is to determine how individuals reidentify objects and events in the world.

Thinking of concepts as explicit lists of features used to pick out referents neatly is indeed mistaken; but there are other alternatives than making concepts mere pointers. These alternatives are suggested by the difference between meaning X and having the concept X, problems of conceptual change, implicit conceptual schemata, the conceptual requirements of the division of cognitive labor, and how concepts figure in perception versus language.

There are three points of difference between psychologists' assumptions and those that Millikan suggests: (1) concepts as representations versus concepts as reflecting a capacity; (2) concepts having a role in categorization and inference versus a role in reidentification; and (3) the “basic level” as an aspect of the “vertical” dimension of categories versus being a kind of category, on a par with natural kinds.

A realist story of concepts like Millikan's can and should accommodate facts about how the context of items available for comparison during concept formation affects just what concept is formed or reidentified. Similarly, the contribution of the goals and purposes of the conceptualizer are relevant to how concepts are acquired and deployed, but can be understood as entirely consistent with a view of concepts as objectively evaluable.

I discuss neuroscientific and phenomenological arguments in support of Millikan's thesis. I then consider invariance as a unifying theme in perceptual and conceptual tracking, and how invariants may be extracted from the environment. Finally, some wider implications of Millikan's nondescriptionist approach to language are presented, with specific application to color terms.

Even in infancy, concept formation has to do with creating meaning, not with tracking substances. Preverbal infants can identify a substance such as a dog, but their first concept of this substance is not dog but animal. It is difficult to account for such global concepts by the perceptual processes involved in object identification, yet these concepts are the foundation on which later concepts are built.

Millikan's account of substance concepts is based on a notion of “substance” expanded from realist notions of individuals and natural kinds. Her metaphysical notion, based on “inductive potential,” is shown to be too puristic and needs to incorporate cognizing subjects. This could preserve the realist/nondescriptionist insight that the extension of substances is determined by the world.

The problems that Millikan addresses in theories of concepts arise from an extensional view of concepts and word meaning. If instead one assumes that concepts are psychological entities intended to explain human behavior and thought, many of these problems dissolve.

Millikan's theory of substance concepts has advantages for psychological theories, including those in cognitive development. However, the disadvantage is that it cannot be generalized even to some of the most common concepts that children acquire in the early years of life. For a general theory we must get beyond substances.

Millikan's externalist account of concept acquisition cannot completely avoid the distinction between central (defining) and peripheral (characteristic) features, because some knowledge is required to achieve reference and to decide what kind of information to record about the identified substances. However, the emphasis on external reference may provide the requisite principled way to make this distinction.

In presenting her attractive theory of concepts, Millikan makes an unwarranted assumption about the role of language in concept acquisition. The phoneme string, rather than the “word” as a semantic entity, may suffice to play the crucial role in the acquisition of substance concepts. Hence Millikan may underestimate the degree of similarity between language and other media of perception.

User intentions invoked to account for the distinctive way in which public-language natural-kind terms gather their extensions are inapplicable in the case of Millikan's substance concepts. I suggest that theoretical justification is preferable and available and raise exploratory questions about the applicability of the notion of substance concepts to the genesis of self concepts.

Millikan's proposal of a common structure of substance concepts does not explain certain conspicuous findings in the psychological literature such as typicality effects, the context sensitivity of these effects, and slips of the tongue. Moreover, it is unclear how antidescriptionism could be relevant to psychological theorizing. Finally, it does not seem to be true that concepts of individuals, stuff, and real kinds have a common structure in older children and in adults.

Semantic realism fits Millikan's account of kind terms in its focus on information-theoretic abilities and strategic ways of gathering information in human communication. Instead of the traditional logical necessity, we should interpret rigid designation in a dynamic semantics as a legislative act to constrain possible ways in which our belief may change.

Evidence from language acquisition suggests that words are powerful mechanisms in the acquisition of substance concepts. Infants initially approach language with the general expectation that words refer to real kinds, regardless of grammatical cues to the contrary.

I have no problem with Millikan's saying that Mama, milk, and mouse are substances, but I do not see why this list cannot be extended with white, red cows, things, vovetas, lhenxa, GRUE, and so on. In the right circumstances, given the right training, the characteristics of substances that Millikan provides work equally well for each of them.

We applaud Millikan's psychologically plausible version of the causal theory of reference. Her proposal offers a significant clarification of the much-debated relation between concepts and beliefs, and suggests positive directions for future empirical studies of conceptual development. However, Millikan's revision of the causal theory may leave us with no generally satisfying account of concept individuation in the mind.

A number of clarifications of the target article and some corrections are made. I clarify which concepts the thesis was intended to be about, what “descriptionism” means, the difference between “concepts” and “conceptions,” and why extensions are not determined by conceptions. I clarify the meaning of “substances,” how one knows what inductions to project over them, the connection with “basic level categories,” how it is determined what substance a given substance concept is of, how equivocation in concepts occurs, and the role of language in the conception of substances. Finally, I clarify exactly why I said that concepts of individuals, real kinds, and stuffs have “a common structure,”

Since the BBS article in which Premack and Woodruff (1978) asked “Does the chimpanzee have a theory of mind?,” it has been repeatedly claimed that there is observational and experimental evidence that apes have mental state concepts, such as “want” and “know.” Unlike research on the development of theory of mind in childhood, however, no substantial progress has been made through this work with nonhuman primates. A survey of empirical studies of imitation, self-recognition, social relationships, deception, role-taking, and perspective-taking suggests that in every case where nonhuman primate behavior has been interpreted as a sign of theory of mind, it could instead have occurred by chance or as a product of nonmentalistic processes such as associative learning or inferences based on nonmental categories. Arguments to the effect that, in spite of this, the theory of mind hypothesis should be accepted because it is more parsimonious than alternatives or because it is supported by convergent evidence are not compelling. Such arguments are based on unsupportable assumptions about the role of parsimony in science and either ignore the requirement that convergent evidence proceed from independent assumptions, or fail to show that it supports the theory of mind hypothesis over nonmentalist alternatives. Progress in research on theory of mind requires experimental procedures that can distinguish the theory of mind hypothesis from nonmentalist alternatives. A procedure that may have this potential is proposed. It uses conditional discrimination training and transfer tests to determine whether chimpanzees have the concept “see.” Commentators are invited to identify flaws in the procedure and to suggest alternatives.