Stanford Encyclopedia of Philosophy
This is a file in the archives of the Stanford Encyclopedia of Philosophy.

Concepts

First published Mon Nov 7, 2005; substantive revision Wed Feb 22, 2006

Concepts, pretheoretically, are the constituents of thoughts. But the pre-theoretic notion only goes so far as an entry point into philosophical theories of concepts. This is partly because concept has become a term of art among philosophers and partly because of the diversity of projects and concerns that tend to get lumped under this one heading. Adding to the confusion is the fact that disputes about concepts often reflect deeply opposing approaches to the study of the mind, language, and even to philosophy itself. Bearing all of this in mind, this entry is organized around five significant issues that are focal points for many theories of concepts. Not every theory of concepts takes a stand on each of the five, but viewed collectively these issues show why the theory of concepts has been such a rich and lively topic in recent years. The five issues are: (1) the ontology of concepts, (2) the structure of concepts, (3) empiricism and nativism about concepts, (4) concepts and natural language, and (5) concepts and conceptual analysis.

1. The ontology of concepts

We begin with the issue of the ontological status of a concept. The three main options are to identify concepts with mental representations, with abilities, and with Fregean senses.

1.1. Concepts as mental representations

The first of these views maintains that concepts are psychological entities, taking as its starting point the representational theory of the mind (RTM). According to RTM, thinking occurs in an internal system of representation. Beliefs and desires and other propositional attitudes enter into mental processes as internal symbols. For example, Sue might believe that Dave is taller than Cathy, and also believe that Cathy is taller than Ben, and together these may cause Sue to believe that Dave is taller than Ben. Her beliefs would be constituted by mental representations that are about Dave, Cathy and Ben and their relative heights. What makes these beliefs, as opposed to desires or other psychological states, is that the symbols have the characteristic causal-functional role of beliefs. (RTM is usually presented as taking beliefs and other propositional attitudes to be relations between an agent and a mental representation (e.g., Fodor 1987). But given that the relation in question is a matter of having a representation with a particular type of functional role tokened in one's mind, it is simpler to say that occurrent beliefs just are mental representations with a characteristic type of functional role.)

Many advocates of RTM take the mental representations involved in beliefs and other propositional attitudes to have internal structure. Accordingly, the representations that figure in Sue's beliefs would be composed of more basic representations. For theorists who adopt the mental representation view of concepts, concepts are identified with these more basic representations.

Early advocates of RTM (e.g., Locke (1690/1975) and Hume (1739/1978)) called these more basic representations ideas, and took them to be mental images. But modern versions of RTM assume that much thought is not grounded in mental images. The classic contemporary treatment maintains, instead, that the internal system of representation has a language-like syntax and a compositional semantics. According to this view, much of thought is grounded in word-like mental representations. This view is often referred to as the language of thought hypothesis (Fodor 1975). However, the analogy with language isn't perfect; obviously, the internal symbol system must lack many of the properties associated with a natural language. Nonetheless, like a natural language, the internal system's formulae are taken to have subject/predicate form and include logical devices, such as quantifiers and variables. In addition, the content of a complex symbol is supposed to be a function of its syntactic structure and the contents of its constituents. Returning to Sue's beliefs, the supposition is that they are composed of such symbols as DAVE, CATHY and TALLER and that her beliefs represent what they do in virtue of the contents of these symbols and how they are arranged .

The mental representation view of concepts is the default position in cognitive science (Pinker 1994) and enjoys widespread support in the philosophy of mind, particularly among philosophers who view their work as being aligned with research in cognitive science (e.g., Carruthers, 2000, Fodor 2003, Harman 1987, Laurence & Margolis 1999). Supporters of this view argue for it on explanatory grounds. They maintain that concepts and structured mental representations play a crucial role in accounting for the productivity of thought (i.e., the fact that human beings can entertain an unbounded number of thoughts) and in accounting for how mental processes can be both rational and implemented in the brain (Fodor 1987; see also the entry language of thought hypothesis).

Critics of this view argue that it is possible to have propositional attitudes without having the relevant mental representations tokened in one's head. Daniel Dennett (1977), for example, argues that most people believe zebras don't wear overcoats in the wild—and a million other similar facts—even though they have never stopped to consider such matters. Dennett also notes that computing systems can lack representations corresponding to the explanations we cite in characterizing and predicting their behavior. For example, it may make perfect sense to say of a chess-playing computer that it thinks that it is good to get one's queen out early, even though we know from how the computer is programmed that it has no representation with that very content (see Dennett 1978, 1987 for these and related criticisms and Fodor 1987 for a response).

Other critics claim that RTM is too closely associated with commonsense psychology, which they argue should be abandoned as a stagnant and degenerate research program (Churchland 1981; see Horgan & Woodward 1985 for a reply), or that developments in computational modeling offer alternatives, particularly to the language of thought version of RTM (e.g., see Port & Gelder 1995 on dynamic systems, and Elman et al. 1996 on connectionism; see Fodor & Pylyshyn 1988, Marcus 2001 for critical discussion of theories that hope to avoid combinatorial structure.)

1.2 Concepts as abilities

According to the abilities view, it's wrong to maintain that concepts are mental particulars—concepts are neither mental images nor word-like entities in a language of thought. Rather, concepts are abilities that are peculiar to cognitive agents (e.g., Brandom 1994, Dummett 1993, Millikan 2000). The concept CAT, for example, might amount to the ability to discriminate cats from non-cats and to draw certain inferences about cats.

While the abilities view is maintained by a diverse group of philosophers, the most prominent reason for adopting the view is a deep skepticism about the existence and utility of mental representations, skepticism that traces back Ludwig Wittgenstein (1953/1958). One of the most influential arguments along these lines claims that mental representations are explanatorily idle because they reintroduce the very sorts of problems they are supposed to explain. For example, Michael Dummett cautions against trying to explain knowledge of a first language on the model of knowledge of a second language. In the case of a second language, it is reasonable to suppose that understanding the language involves translating its words and sentences into words and sentences of one's first language. But according to Dummett, one can't go on to translate words and sentences of one's first language into a prior mental language. “[T]here is really no sense to speaking of a concept's coming into someone's mind. All we can think of is some image coming to mind which we take as in some way representing the concept, and this gets us no further forward, since we still have to ask in what his associating that concept with that image consists” (Dummett 1993, p. 98). In other words, the mental representation itself is just another item whose significance bears explaining. Either we are involved in a vicious regress, having to invoke yet another layer of representation (and so on indefinitely) or we might as well stop with the external language and explain its significance directly. (For critical discussion of this type of regress argument, see Fodor 1975, Crane 1995, Laurence & Margolis 1997).

Not surprisingly, critics of the abilities view argue in the other direction. They note difficulties that the abilities view inherits by its rejection of mental representations. One is that the view is ill-equipped to explain the productivity of thought; another is that it can say little about mental processes. And if proponents of the abilities view remain neutral about the existence of mental representations, they open themselves to the criticism that explication of these abilities is best given in terms of underlying mental representations and processes (see Fodor 1968 and Chomsky 1980 for general discussion of the anti-intellectualist tradition in the philosophy of mind).

1.3 Concepts as Fregean senses

The view that concepts are Fregean senses identifies concepts with abstract objects, as opposed to mental objects and mental states (e.g., Peacocke 1992, Zalta 2001). Concepts are said to be the constituents of propositions. For proponents of this view, concepts mediate between thought and language, on the one hand, and referents, on the other. An expression without a referent (“Pegasus”) needn't lack a meaning, since it still has a sense. Similarly, the same referent can be associated with different expressions (e.g., “Eric Blair” and “George Orwell”) because they convey different senses. Senses are more discriminating than referents. Each sense has a unique perspective on its referent—a unique mode of presentation. Differences in cognitive content trace back to differences in modes of presentation. It's for this reason that the thought that George Orwell is Eric Blair lacks the triviality of the thought that George Orwell is George Orwell. Philosophers who take concepts to be senses particularly emphasize this feature of senses. Christopher Peacocke, for example, locates the subject matter of a theory of concepts as follows: “Concepts C and D are distinct if and only if there are two complete propositional contents that differ at most in that one contains C substituted in one or more places for D, and one of which is potentially informative while the other is not” (Peacocke 1992, p. 2). In other words, C and D embody differing modes of presentation. (See the entry Frege for discussion of the sense/reference distinction and for more on the explanatory functions associated with senses. To avoid terminological confusion, we should note that Frege himself did not use the term 'concept' for senses, but rather for the referents of predicates. Similarly, it is worth noting that Frege uses the term 'thought' to stand for propositions, so for Frege 'thoughts' are not psychological states at all.)

The view that concepts are Fregean senses, like the abilities view, is generally held by philosophers who are opposed to identifying concepts with mental representations. Peacocke himself doesn't go so far as to argue that mental representations are explanatorily idle, but he does think that mental representations are too fine-grained for philosophical purposes. “It is possible for one and the same concept to receive different mental representations in different individuals” (Peacocke 1992, p. 3). He is also concerned that identifying concepts with mental representations rules out the possibility of there being concepts that human beings have never entertained, or couldn't ever entertain.

If we accept that a thinker's possession of a concept must be realized by some subpersonal state involving a mental representation, why not say simply that the concept is the mental representation? Just this proposal is made by Margolis and Laurence (1999, 77). Mental representations that are concepts could even be typed by the corresponding possession condition of the sort I favour. This seems to me an entirely legitimate notion of a kind of mental representation; but it is not quite the notion of a concept. It can, for instance, be true that there are concepts human beings may never acquire, because of their intellectual limitations, or because the sun will expand to eradicate human life before humans reach a stage at which they can acquire these concepts. ‘There are concepts that will never be acquired’ cannot mean or imply ‘There are mental representations which are not mental representations in anyone's mind’. If concepts are individuated by their possession conditions, on the other hand, there is no problem about the existence of concepts that will never be acquired. They are simply concepts whose possession conditions will never be satisfied by any thinkers. (Peacocke, 2005, p. 169).

Advocates of the mental representation view would respond to these arguments by invoking the type/token distinction with respect to mental representations. According to the mental representation view, concepts that haven't been acquired are just representations of a type that have never been tokened (for related discussion, see Laurence & Margolis 1999).

Critics of the sense-based view have questioned the utility of appealing to such abstract objects (Quine 1960). One difficulty stems from the fact that senses, as abstract entities, stand outside of the causal realm. The question then is how we can access these objects. Advocates of the Fregean sense view describe our access to senses by means of the metaphor of “grasping”—we are said to grasp the sense of an expression. But grasping here is just a metaphor for a cognitive relation that needs to be explicated. Moreover, though senses are hypothesized as providing different modes of presentation for referents, it is not clear why senses themselves do not generate the mode of presentation problem (Fodor 1998). Since they are external to our minds, just as referents typically are, it isn't clear why we can't stand in different epistemic relations towards them just as we can to referents. In the same way that we can have different modes of presentation for a number (the only even prime, the sum of one and one, Tim's favorite number, etc.), we ought to be able to have different modes of presentation for a given sense.

1.4 Is the issue terminological?

Stepping back from the details of these three views, there is no reason, in principle, why the different views of concepts couldn't be combined in various ways. For instance, one might maintain that concepts are mental representations that are typed in terms of the Fregean senses they express. For this reason alone, it's fair to wonder whether the dispute about ontology is a substantive dispute. Perhaps there is only a terminological issue about which things ought to be granted the label “concepts”. If so, why not just call mental representations “concepts1”, the relevant abilities “concepts2”, senses “concepts3”, and leave it at that?

However, the participants in the dispute don't generally view it as a terminological one. Perhaps this is because they associate their own theories of concepts with large-scale commitments about the way that philosophers should approach the study of mind and language. Undoubtedly, from Dummett's perspective, philosophers who embrace the mental representation view also embrace RTM, and RTM, as he sees it, is fundamentally misguided. Likewise, from Fodor's perspective, RTM is critical to the study of the mind, so an approach like Dummett's, which disallows RTM, places inappropriate a priori constraints on the study of the mind.

These differences in perspective remain present once a more fine-grained terminology is adopted. For example, it would still be a matter of dispute whether there are mental representations and whether they can do the sorts of explanatory work that proponents of RTM require of them or whether these explanatory roles provide the most important or coherent cluster of roles associated with the term “concept”. Previously, these issues would have found expression by posing the question of whether concepts are mental representations. However, if we adopt the proposed new terminology, much the same set of issues would arise concerning the nature and existence of the various more fine-grained categories—concepts1, concepts2, and concepts3.

2. The structure of concepts

Just as thoughts are composed of more basic, word-sized concepts, so these word-sized concepts—known as lexical concepts—are generally thought to be composed of even more basic concepts. In this section, we look at different proposals about the structure of lexical concepts (see Margolis & Laurence 1999 for different approaches to the issue of conceptual structure).

2.1 The classical theory

In one way or another, all theories regarding the structure of concepts are developments of, or reactions to, the classical theory of concepts. According to the classical theory, a lexical concept C has definitional structure in that it is composed of simpler concepts that express necessary and sufficient conditions for falling under C. The stock example is the concept BACHELOR, which is traditionally said to have the constituents UNMARRIED and MAN. If the example is taken at face value, the idea is that something falls under BACHELOR if it is an unmarried man and only if it is an unmarried man. According to the classical theory, lexical concepts generally will exhibit this same sort of definitional structure. This includes such philosophically interesting concepts as TRUTH, GOODNESS, FREEDOM, and JUSTICE.

Before turning to other theories of conceptual structure, it's worth pausing to see what's so appealing about classical or definitional structure. Much of its appeal comes from the way it offers unified treatments of concept acquisition, categorization, and reference determination. In each case, the crucial work is being done by the very same components. Concept acquisition can be understood as a process in which new complex concepts are created by assembling their definitional constituents. Categorization can be understood as a psychological process in which a complex concept is matched to a target item by checking to see if each and every one of its definitional constituents applies to the target. And reference determination, we've already seen, is a matter of whether the definitional constituents do apply to the target.

These considerations alone would be enough to show why the classical theory has been held in such high regard. But the classical theory receives further motivation through its connection with a philosophical method that goes back to antiquity and that continues to exert its influence over contemporary thought. This is the method of conceptual analysis. Paradigmatic conceptual analyses offer definitions of concepts that are to be tested against potential counterexamples that are identified via thought experiments. Conceptual analysis is supposed to be a distinctively a priori activity that many take to be the essence of philosophy. To the extent that paradigmatic conceptual analyses are available and successful, this will convey support for the classical theory. Conversely, if the definitions aren't there to be discovered, this would seem to put in jeopardy a venerable view of what philosophy is and how philosophical investigations ought to proceed (see section 5 below).

The classical theory has come under considerable pressure in the last thirty years or so, not just in philosophy but in psychology and other fields as well. For psychologists, the main problem has been that the classical theory has difficulty explaining a robust set of empirical findings. At the center of this work is the discovery that certain categories are taken to be more representative or typical and that typicality scores correlate with a wide variety of psychological data (for reviews, see Smith & Medin 1981, Murphy 2002). For instance, apples are judged to be more typical than plums with respect to the category of fruit, and correspondingly apples are judged to have more features in common with fruit. There are many other findings of this kind. One other is that more typical items are categorized more efficiently. For example, subjects are quicker to judge that apples are a kind of fruit than to judge that plums are. The problem isn't that the classical theory is inconsistent with results like these but that it does nothing to explain them.

In philosophy, the classical theory has been subjected to a number of criticisms but perhaps the most fundamental is that attempts to specify definitions for concepts have a poor track record. Quite simply, there are too few examples of successful definitional analyses, and certainly none that are uncontroversial (Wittgenstein 1953/1958, Fodor 1981). The huge literature on the analysis of knowledge is representative of the state of things. Since Edmund Gettier (1963) first challenged the traditional definition of KNOWLEDGE (as JUSTIFIED TRUE BELIEF), there has been widespread agreement among philosophers that the traditional definition is incorrect or at least incomplete (e.g., Dancy 1985). But no one can seem to agree on what the correct definition is. Despite the enormous amount of effort that has gone into the matter, and the dozens of papers written on the issue, we are still lacking a satisfactory and complete definition. It could be that the problem is that definitions are hard to come by. But another possibility—one that many philosophers are now taking seriously—is that our concepts lack definitional structure.

2.2 The prototype theory

What other type of structure could they have? A non-classical alternative that emerged in the 1970s is the prototype theory. According to this theory, a lexical concept C doesn't have definitional structure but has probabilistic structure in that something falls under C just in case it satisfies a sufficient number of properties encoded by C's constituents. The prototype theory has its philosophical roots in Wittgenstein's (1953/1958) famous remark that the things covered by a term often share a family resemblance, and it has its psychological roots in Eleanor Rosch's experimental treatment of much the same idea (Rosch & Mervis 1975, Rosch 1978). The prototype theory is especially at home in dealing with the typicality effects that were left unexplained by the classical theory. One standard strategy is to maintain that, on the prototype theory, categorization is to be understood as a similarity comparison process, where similarity is computed as a function of the number of constituents that two concepts hold in common. On this model, the reason apples are judged to be more typical than plums is that the concept APPLE shares more of its constituents with FRUIT. Likewise, this is why apples are judged to be a kind of fruit faster than plums are.

The prototype theory does well in accounting for a variety of psychological phenomena and it helps to explain why definitions may be so hard to produce. But the prototype theory has its own problems and limitations. One is that its treatment of categorization works best for quick and unreflective judgments. Yet when it comes to more reflective judgments, people go beyond the outcome of a similarity comparison. If asked whether a dog that is surgically altered to look like a raccoon is a dog or a raccoon, the answer for most of us, and even for children, is that it is remains a dog (see Keil 1989, Gelman 2003 for discussion). Another criticism that has been raised against taking concepts to have prototype structure concerns compositionality. When a patently complex concept has a prototype structure, it often has emergent properties, ones that don't derive from the prototypes of its constituents (e.g., PET FISH encodes properties such as brightly colored, which have no basis in the prototype structure for either PET or FISH). Further, many patently complex concepts don't even have a prototype structure (e.g., CHAIRS THAT WERE PURCHASED ON A WEDNESDAY) (Fodor & Lepore 1994, Fodor 1998).

One solution to these problems is to hold that a prototype constitutes just part of the structure of a concept. In addition, concepts have conceptual cores, which specify the information relevant to more considered judgments and which underwrite compositional processes. Of course, this just raises the question of what sort of structure conceptual cores have. One common suggestion is that conceptual cores have classical structure (Osherson & Smith 1981, Landau1982). This won't do, however, since it just raises once again most of the problems associated with the classical theory (Laurence & Margolis 1999).

2.3 The theory theory

Another and currently more popular suggestion is that cores are best understood in terms of the theory theory of concepts. This is the view that concepts stand in relation to one another in the same way as the terms of a scientific theory and that categorization is a process that strongly resembles scientific theorizing (see, e.g., Carey 1985, Gopnik & Meltzoff 1997, Keil 1989). It's generally assumed, as well, that the terms of a scientific theory are interdefined so that a theoretical term's content is determined by its unique role in the theory in which it occurs.

The theory theory is especially well-suited to explaining the sorts of reflective categorization judgments that proved to be difficult for the prototype theory. For example, theory theorists maintain that children override perceptual similarity in assessing the situation where the dog is made to look like a raccoon, claiming that even children are in possession of a rudimentary biological theory. This theory, an early form of folk biology, tells them that being a dog isn't just a matter of looking like a dog. More important is having the appropriate hidden properties of dogs—the dog essence (see Medin & Atran 1999 on folkbiology). Another advantage of the theory theory is that is supposed to help to explain important aspects of conceptual development. Conceptual change in childhood is said to follow the same pattern as theory change in science.

One problem that has been raised against the theory theory is that it has difficulty in allowing for different people to possess the same concepts (or even for the same person to have the same concept over time). The reason is that the theory theory is holistic. A concept's content is determined by its role in a theory, not by its being composed of just a handful of constituents. Since beliefs that enter people's mental theories are likely to be different from one another (and are likely to change), there may be no principled basis for comparison (Fodor & Lepore 1992). Another problem with the theory theory concerns the analogy to theory change in science. The analogy suggests that children undergo radical conceptual reorganization in development, but many of the central case studies have proved to be controversial on empirical grounds, with evidence that the relevant concepts are implicated in core knowledge systems that are enriched in development but not fundamentally altered (see Spelke 1994 on core knowledge).

2.4 Conceptual atomism

A radical alternative to all of the theories we've mentioned so far is conceptual atomism, the view that lexical concepts have no semantic structure (Fodor 1998, Millikan 2000). According to conceptual atomism, the content of a concept isn't determined by its relation to other concepts but by its relation to the world.

Conceptual atomism follows in the anti-descriptivist tradition that traces back to Saul Kripke, Hilary Putnam, and others working in the philosophy of language (see Kripke 1972/80, Putnam 1975, Devitt 1981). Kripke, for example, argues that proper names function like mere tags in that they have no descriptive content (Kripke 1972/80). On a description theory one might suppose that “Gödel” means something like the discoverer of the incompleteness of arithmetic. But Kripke points out we could discover that Schmitt really discovered the incompleteness of arithmetic and that Gödel could have killed Schmitt and passed the work off as his own. The point is that if the description theory were correct, we would be referring to Schmitt when we say “Gödel”. But intuitively, that's not the case at all. In the imagined scenario, the sentence “Gödel discovered the incompleteness of arithmetic” is saying something false about Gödel, not something trivially true about Schmitt. Kripke's alternative account of names is that they achieve their reference by standing in a causal relation to their referents. Conceptual atomism employs a similar strategy while extending the model to all sorts of concepts, not just ones for proper names.

2.5 Pluralism about structure

At present, the nature of conceptual structure remains unsettled. Perhaps part of the problem is that more attention needs to be given to the question of what explanatory work conceptual structure is supposed to do and the possibility that there are different types of structure associated with different explanatory functions. We've seen that conceptual structure is invoked to explain, among other things, typicality effects, reflective categorization, cognitive development, reference determination, and compositionality. But there is no reason to assume that a single type of structure can explain all of these things. As a result, there is no reason why philosophers shouldn't maintain that concepts have different types of structure. For example, notice that atomism is largely motivated by anti-descriptivism. In effect, the atomist maintains that considerable psychological variability is consistent with concepts entering into the same mind-world causal relations, and that it's the latter that determines a concept's reference. But just because the mechanisms of reference determination permit considerable psychological variability doesn't mean that there aren't, in fact, significant patterns for psychologists to uncover. On the contrary, the evidence for typicality effects is impressive by any measure. For this reason, it isn't unreasonable to claim that concepts do have prototype structure even if that structure has nothing to do with the determination of a concept's referent. Similar considerations suggest that concepts may have theory-structure and perhaps other types of structure as well (see Laurence & Margolis 1999 on different types of conceptual structure). Moreover, different types of concepts may well have correspondingly different types of structure—the human conceptual system is, after all, incredibly diverse.

3. Empiricism and nativism about concepts

One of the oldest questions about concepts concerns whether there are any innate concepts and, if so, how much of the conceptual system is innate. Empiricists maintain that there are few if any innate concepts and that most cognitive capacities are acquired on the basis of a few relatively simple general-purpose cognitive mechanisms. Nativists, on the other hand, maintain that there may be many innate concepts and that the mind has a great deal of innate differentiation into complex domain-specific subsystems.

3.1 Renewed interest in the empiricism/nativism dispute

In recent years, the debate over innate concepts has been reinvigorated as advances in cognitive science have provided philosophers with new tools for revisiting and refining the traditional dispute (see, e.g., Pinker 1994, Elman et al. 1996, Carruthers, Laurence, & Stich 2005). Philosophers have greatly benefited from empirical studies in such diverse fields as developmental psychology, evolutionary psychology, cognitive anthropology, neuroscience, linguistics, and ethology. Part of the philosophical interest of this work is that, while the scientists themselves take sides on the empiricist-nativist dispute, their theories and data are often open to interpretation.

As an example, one of the earliest lines of investigation that appeared to support traditional nativist conceptions of the mind was the study of language (Pinker 1994). Noam Chomsky and his followers argued that language acquisition succeeds even though children are only exposed to severely limited evidence about the structure of their language (Chomsky 1967, 1975, 1988; see also Laurence & Margolis 2001). Given the way that the final state (e.g., knowledge of English) outstrips the data that are available to children, we can only postulate that the human mind brings to language acquisition a complex set of language-specific dispositions. For Chomsky, these dispositions are grounded in a set of innate principles that constrain all possible human natural languages, viz., universal grammar (see Baker 2001 on universal grammar).

Not surprisingly, many philosophers have questioned Chomsky's position. The ensuing debate has helped to sharpen the crucial arguments and the extent to which nativist models should continue to command their central place in linguistic theory. (On the empiricist side, see Cowie (1999), Prinz (2002), and Devitt (in press); on the nativist side, see Laurence & Margolis (2001) and Crain & Pietroski (2001)); see also the entry innatism: linguistic). For instance, one of Fiona Cowie's criticisms of Chomsky's poverty of the stimulus argument is that any induction establishes a conclusion that outstrips the available data; hence, going beyond the data in the case of language acquisition doesn't argue for innate language-specific dispositions—or else there would have to be a specific innate disposition for every induction we make (for an earlier version of this argument, see Putnam 1967, Goodman 1969). Both Laurence & Margolis and Crain & Pietroski respond by teasing out the various ways in which the problem of language acquisition goes beyond general problems about induction.

3.2 Empiricism about concepts

Traditionally, empiricists have argued that all concepts derive from sensations. Concepts were understood to be formed from copies of sensory representations and assembled in accordance with a set of general-purpose learning rules, e.g., Hume's principles of association (Hume 1739/1978). On this view, the content of any concept must be analyzable in terms of its perceptual basis. Any purported concept that fails this test embodies a confusion. Thus David Hume ends his Enquiry with the famous remark:

When we run over libraries, persuaded of these principles, what havoc must we make? If we take in our hand any volume; of divinity or school metaphysics, for instance; let us ask, Does it contain any abstract reasoning concerning quantity or number? No. Does it contain any experimental reasoning concerning matter of fact and existence? No. Commit it then to the flames: For it can contain nothing but sophistry and illusion. (1748/1975, p. 165)

A similar doctrine was maintained by the logical positivists in the early Twentieth Century, though the positivists couched the view in linguistic terms (Ayer 1959). Their principle of verification required for a sentence or statement to be meaningful that it have empirical consequences, and, on some formulations of the principle, that the meaning of a sentence is the empirical procedure for confirming it (see the entry Vienna Circle). Sentences that have no empirical consequences were deemed to be meaningless. Since a good deal of philosophy purports to express propositions that transcend all possible experience, the positivists were happy to say that these philosophical doctrines are entirely devoid of content and are composed of sentences that aren't merely false but are literally gibberish.

Despite the current unpopularity of verificationism (though see Dummett 1993, Wright 1989, and Dennett 1991), a growing number of philosophers are attracted to modified forms of empiricism, forms that primarily emphasize psychological relations between the conceptual system and perceptual and motor states, not semantic relations. An example is Lawrence Shapiro's defense of the claim that the type of body that an organism has profoundly affects its cognitive operations as well as the way that the organism is likely to conceptualize the world (Shapiro 2004). Shapiro's claim is directed against philosophical theories that willfully ignore contingent facts about human bodies as if a human mind could inhere in wildly different body types. Drawing on a number of empirical research programs, Shapiro cites examples that appear to support what he calls the embodied mind thesis, viz., that “minds profoundly reflect the bodies in which they are contained” (Shapiro 2004, p. 167).

Jesse Prinz (2002) also defends a modified form of empiricism. Prinz claims that “all (human) concepts are copies or combinations of copies of perceptual representations” (Prinz 2002, p. 108). Though the reference to copies is a nod to Hume, Prinz certainly doesn't buy into Hume's verificationism. In fact, Prinz adopts a causal theory of content of the kind that is usually associated with atomistic theories of concepts (e.g., Fodor 1998); thus Prinz's theory of intentional content doesn't require a concept to inherit the specifically perceptual content of its constituents. Nonetheless, Prinz thinks that every concept derives from perceptual representations. Perhaps the best way to understand the claim is that the mental representations that are activated when someone thinks about something—no matter what the thought—are representations that originate in neural circuits with perceptual or motor functions and that the mental process is affected by that origin. Suppose, for example, that one is thinking about a hammer. Then she is either activating representations that inhere in visual circuits, or representations involved in circuits that control hand shape, etc., and her thought is affected in some way by the primary function of these circuits. Following Lawrence Barsalou (1999; see also Barsalou 2003), Prinz characterizes concept possession as a kind of simulation “tantamount to entering a perceptual state of the kind one would be in if one were to experience the thing it represents” (Prinz 2002, p. 150).

One challenge to this view of cognition is its implication for abstract concepts. It's one thing to say that the concept HAMMER involves the activation of circuits related to hand shape; it's quite another to identify significant modal-specific representations underlying such concepts as TRUTH, DEMOCRACY, ENTROPY, and NINETEEN(Adams & Campbell 1999, Brewer 1999). Logical concepts are also a challenge. Prinz suggests as a perceptual basis for the concept of disjunction that it is based on feelings of hesitation. However, his more considered view seems to be that logical concepts are best understood as operations, not representations. The resulting theory is one in which thoughts lack logical form. The trouble is that this makes it difficult to see how to distinguish logically equivalent thoughts. A related problem is that, since composition for Prinz does not yield structurally complex representations, there seems to be nothing to distinguish the type of contents associated with judgements (propositional contents) from those associated lists or even single concepts (for related discussion Fodor & Pylyshyn 1988).

3.3 Nativism about concepts

Perhaps the most influential discussion of concepts in relation to the nativism/empiricism debate is Jerry Fodor's (1975, 1981) argument for the claim that virtually all lexical concepts are innate. Fodor (1975) argued that there are theoretical problems with all models of concept learning in that all such models treat concept learning as hypothesis testing. The problem is that the correct hypothesis invariably employs the very concept to be learned and hence the concept has to be available to a learner prior to the learning taking place. In his (1981) Fodor developed this argument by allowing that complex concepts (and only complex concepts) can be learned in that they can be assembled from their constituents during the learning process. He went on to argue that lexical concepts lack semantic structure and consequently that virtually all lexical concepts must be innate—a position known as radical concept nativism. This view has been criticized by numerous philosophers (Samet 1986, Sterelney 1989, Cowie 1999, Laurence & Margolis 2002) yet the logic underlying Fodor's argument continues to influence cognitive scientists who more or less universally reject Fodor's conclusion, but struggle to understand how the mind is able to develop a richer representational system on the basis of a more impoverished one (Niyogi & Snedeker forthcoming).

One further issue concerning innate concepts is whether the very idea of innateness makes sense. A common point among those who are skeptical of the notion is the observation that all traits are dependent upon interactions between genes and the environment and that there is no way of fully untangling the two (Elman et al. 1996, Griffiths 2002; see also Marcus 2004). Nonetheless, there are clear differences between models of the mind with empiricist leanings and models of the mind with nativist leanings, and the notion of innateness may be thought to earn its usefulness by marking these differences. (See Samuels 2002 for discussion of different proposals of what innateness is.)

4. Concepts and natural language

We turn now to the issue of how concepts and thoughts relate to language.

4.1 Can there be concepts without language?

Some philosophers maintain that possession of natural language is necessary for having any concepts (Brandom 1994, Davidson 1975, Dummett 1993) and that the tight connection between the two can be established on a priori grounds. In a well known passage, Donald Davidson summarizes his position as follows:

We have the idea of belief only from the role of belief in the interpretation of language, for as a private attitude it is not intelligible except as an adjustment to the public norm provided by language. It follows that a creature must be a member of a speech community if it is to have the concept of belief. And given the dependence of other attitudes on belief, we can say more generally that only a creature that can interpret speech can have the concept of a thought.
Can a creature have a belief it does not have the concept of belief? It seems to me it cannot, and for this reason. Someone cannot have a belief unless he understands the possibility of being mistaken, and this requires grasping the contrast between truth and error—true belief and false belief. But this contrast, I have argued, can emerge only in the context of interpretation, which alone forces us to the idea of an objective, public truth. (Davidson 1975, p. 170).

The argument links having beliefs and concepts with having the concept of belief. Since Davidson thinks that non-linguistic creatures can't have the concept of belief, they can't have other concepts as well. Why the concept of belief is needed to have other concepts is somewhat obscure in Davidson's writings (Carruthers 1992). And whether language is necessary for this particular concept is not obvious. In fact, there is an ongoing research program in cognitive science that addresses this very issue. A variety of non-linguistic tasks have been given to animals and infants to determine the extent to which they are able to attribute mental states to others (see Tomasello, Call, & Hare 2003 for work on chimpanzees and Onishi & Baillargeon 2005 for work on infants; see also Bloom & German 2000). These and related studies provide strong evidence that at least some aspects of theory of mind are nonlinguistic.

Davidson offers a pair of supplementary arguments that may elucidate why he is hesitant to turn the issue over to the cognitive scientists. He gives the example of a man engaging in a non-linguistic task where the man indicates his answer by making a choice, for example, selecting an apple over a pear. Davidson comments that until the man actually says what he has in mind, there will always be a question about the conceptualization guiding his choice. “Repeated tests may make some readings of his actions more plausible than others, but the problem will remain how to determine when he judges two objects of choice to be identical” (1975, p. 163). The second argument points to the difficulties of settling upon a specification of what a non-linguistic creature is thinking. “The dog, we say, knows that its master is home. But does it know that Mr. Smith (who is the master) is home? We have no real idea how to settle, or make sense of, these questions” (1975, p, 163). It's not clear how seriously Davidson himself takes these arguments. Many philosophers have been unconvinced. Notice that both arguments turn on an underdetermination claim—e.g., that the interpretation of the man's action is underdetermined by the non-linguistic evidence. But much the same thing is true even if we add what the man says (or to be more precise, if we add what the man utters). The linguistic evidence doesn't guarantee a correct interpretation any more than the non-linguistic evidence does.

Davidson appears to be employing a very high standard for attributing concepts to animals. In effect, he is asking for proof that our attributions are correct. In contrast, most philosophers who are happy to attribute concepts to animals do so because of a wealth of data that are explained only by appeal to an internal system of representation (e.g., Bermudez 2003; for overviews within cognitive science, see Bekoff, Allen, & Burghardt 2002, Gallistel 1990, Hauser 2000). For example, many species of birds cache food for later retrieval. Their very survival depends upon their ability to successfully recover, in some cases, more than 10,000 different caches in a single season. Researchers studying one species of caching birds have shown that not only do the birds represent the location of the food, but they integrate this information with information about the quality of the food, its perishability, and whether their caching was observed by other birds. Evidence here comes from demonstrations of selective retrieval and recaching of food items under experimentally controlled conditions. Birds will retrieve more perishable items first. When highly valued food items become highly perishable, they shift strategies to retrieve a higher percentage of less perishable food items. And birds that have themselves stolen food from other birds will selectively recache stored food when they are observed caching it (see Clayton, Bussey, & Dickinson 2003, Emery, Dally, & Clayton 2004). Experimental data of this kind provide evidence for particular concepts in birds (of food types, locations, and so on) as well as surprisingly sophisticated cognitive operations defined in terms of them.

There is a great deal of controversy among philosophers about the implications of this type of research. Proponents of RTM are, of course, entirely happy with the idea that the scientific theories of what birds are doing can be taken at face value. Other philosophers maintain that if the scientific theories say that birds are computing an algorithm for determining a caching strategy, then this can only be read as a façon de parler. Still others will grant that animals have representations but go on to claim that these representation are of a lesser status, not to be confused with concepts (Brandom 1994, 2000, McDowell 1994).

This raises an interesting question about whether there is a motivated and principled difference between concepts in humans and mere representations in animals. Philosophers who maintain that there is such a difference often cite the role of concepts in reasoning. For example, Robert Brandom claims that representations in animals do little more than act as reliable mechanisms of discrimination. These representations are supposed to be like thermometers, responding to specific environmental features yet without entering into appropriate inferential processes. However, it's not clear what counts as an appropriate inferential process, and certainly there is room for differing opinions on this point. Moreover, whatever reasoning amounts to, comparative psychology is replete with examples that suggest that animals are capable of far more than reliable detection. Animals may not be as smart as humans, but that doesn't mean they are as dumb as thermometers (see Hurley & Nudds 2006 on reasoning in animals).

4.2 The priority between language and concepts

Even if it's agreed that it is possible to have concepts in the absence of language, there is a dispute about how the two are related. Some maintain that concepts are prior to and independent of natural language, and that natural language is just a means for conveying thought (Fodor 1975, Pinker 1994). Others maintain that at least some types of thinking (and hence some concepts) occur in the internal system of representation constituting our natural language competence (Carruthers 1996, 2002, Spelke 2003).

The arguments for deciding between these two positions involve a mixture of theoretical and empirical considerations. Proponents of the first view have claimed that language is ambiguous in ways that thought presumably is not. For example, the natural language sentence everyone loves someone could be interpreted to mean that everyone loves someone or other, or to mean that everyone loves one and the same person (Pinker 1994). Proponents of the first view have also argued that since language itself has to be learned, thought is prior to language (Fodor 1975; Pinker 1994). A third and similar consideration is that people seem to be able formulate novel concepts which are left to be named later; the concept comes first, the name second (Pinker 1994).

Proponents of the alternative view—that some thinking occurs in language—have pointed to the phenomenology of thought. It certainly seems as if we are thinking in language when we “hear” ourselves silently talking to ourselves (Carruthers 1996). There is also intriguing data that success on certain tasks (e.g., spatial reorientation that relies on combining landmark information with geometrical information) is selectively impaired when the linguistic system is engaged but not when comparable attention is given to non-linguistic distractors. The suggestion is that solving these tasks requires thinking in one's natural language and that some of the crucial concepts must be couched linguistically (Hermer-Vazquez, Spelke, & Katsnelson 2001; Shusterman & Spelke 2005; Carruthers 2002).

4.3 Linguistic determinism and linguistic relativity

Finally, one further issue that bears mentioning is the status of various claims regarding linguistic determinism and linguistic relativity. Linguistic determinism is the doctrine that the language a person speaks causes her to conceptualize the world in certain ways while delimiting the boundaries of her conceptual system; as a result, people who speak very different languages are likely to conceptualize the world in correspondingly different ways. Linguistic relativity is the weaker doctrine that the language one speaks influences how one thinks.

Linguistic determinism is historically associated with the writings of Benjamin Lee Whorf (Whorf 1956). Whorf was especially interested in the languages of the indigenous people of America. He famously argued that the Hopi both speak and think about time in ways that are incongruent with European languages and thought. Rather than viewing time as a continuum that flows evenly throughout the universe and that can be broken up into countable events occurring in the past, present, and future, the Hopi are supposed to focus on change as a process. Their conceptual system is also supposed to differ from ours in that it embodies a distinction between things that are or have been accessible to perception versus things that are not, where the latter category includes things in the future as well as mythical and mental constructs.

The claim that the Hopi lack our concept of time has not stood up to scrutiny. Whorf used clumsy translations of Hopi speech that concealed the extent to which they talk about time (references to yesterday, tomorrow, days of the week, lunar phases, etc.). More interestingly, Whorf provided no direct evidence of how the Hopi think. Instead, he used the circular reasoning that they don't think about time as we do because they don't talk about time as we do. In fact, the Hopi use numerous familiar devices for time keeping, such as calendar strings and sun dials, and their sensitivity to time is evident a wide variety of cultural practices (Malotki 1983).

Linguistic determinism isn't an especially promising doctrine and has few adherents these days, but linguistic relativity is the subject of a spirited debate (see Gumperz & Levinson 1996, Bowerman & Levinson 2001, and Gentner & Goldin-Meadow (2003)). Some recent examples of particular interest include whether language influences how we conceptualize space (non-linguistic spatial reasoning) (e.g., Munnich & Landau 2003; Levinson 2003), motion (non-linguistic reasoning about motion) (e.g., Papafragou, Massey, & Gleitman 2002), and sex (the impact of grammatical gender) (e.g., Boroditsky, Schmidt, & Phillips 2003)

5. Concepts and conceptual analysis

Some of the deepest divides in contemporary philosophy concern the limits of empirical inquiry, the status of conceptual analysis, and the nature of philosophy itself (see, e.g., Chalmers 1996, Jackson 1998, Block & Stalnaker 1999, DePaul & Ramsey 1998). And concepts are right at the center of these disputes. For many, philosophy is essentially the a priori analysis of concepts, which can and should be done without leaving the proverbial armchair. We've already seen that in the paradigm case, an analysis embodies a definition; it specifies a set of conditions that are individually necessary and jointly sufficient for the application of the concept. When all goes well, the intuitions are supposed to match the correct analysis perfectly (though generally speaking it's understood that there may be a tradeoff, where most intuitions have to match an analysis but where an otherwise successful analysis may lead to the discrediting of a few intuitions).

5.1 Attractions of conceptual analysis

Conceptual analysis is attractive to philosophers for a number of reasons. One is that it makes sense of a good deal of philosophical practice—what George Bealer (1998) calls the standard justificatory procedure. Philosophers are always constructing thought experiments and eliciting intuitions. If this practice makes sense, then there has to be an understanding of what philosophy is that would vindicate its utility. Conceptual analysis is supposed to provide just what's needed here. Intuitions can be said to be of value to philosophy precisely because they help us to get clearer about our concepts, especially concepts of intrinsic philosophical interest (JUSTICE, KNOWLEDGE, etc.).

A related attraction is that conceptual analysis explains why philosophy can be an a priori discipline, as many suppose it is. If philosophy is primarily about concepts and concepts can be investigated from the armchair, then the a priori character of philosophy is secured (Jackson 1998).

A third attraction of conceptual analysis is that conceptual analysis has been argued to be a necessary precursor for answering questions about ontological reduction, that is, the sort of reduction that takes place when it's argued that genes are DNA segments, that sensations are brain states, and so on (Chalmers 1996, Jackson 1998). According to one way of filling this view out, one has to begin with an a priori analysis of the higher-level concept, particularly an analysis that makes explicit its causal relations. One can then appeal to empirical findings regarding the things that actually have those causal relations. For example, neuroscience may reveal that such-and-such brain state has the casual relations that analysis reveals to be constitutive of our concept of pain. In the course of doing this, neuroscience is supposed to be showing us what pain is (Lewis 1966, Armstrong 1968). But neuroscience is only in a position to do this against the background of the philosophical work that goes into articulating the concept. (For detailed treatments of this view of reduction, see Chalmers 1996 and Jackson 1998—though it should be noted that Chalmers argues that PAIN, and other concepts of conscious mental states, cannot be analyzed solely in terms of their causal relations and concludes from this that consciousness itself is irreducible.) This work has generated a great deal of debate (e.g., Block & Stalnaker 1999, Yablo 2000, Papineau 2002).

Finally, a fourth attraction is that conceptual analysis may offer normative guidance (Goldman 1986). For instance, epistemologists face the question of whether our inferential practices are justified and, if so, what justifies them. One standard answer is that they can be justified if they conform to our intuitions about what counts as a justified inference (Goldman 1986). In other words, an analysis of our concept of justification is supposed to be all that is needed in order to establish that a set of inference rules is justified. So if it ever turned out that different groups of people employed qualitatively different sets of inferential principles, we could establish the epistemically preferable one by showing that it does a better job of conforming to our concept of justification.

5.2 Objections to conceptual analysis

Many philosophers who are opposed to conceptual analysis identify their approach as being naturalistic (e.g., Papineau 1993, Devitt 1996, Kornblith 2002; see also the entry naturalism). A common theme of this work is that philosophy is supposed to be continuous with science and that philosophical theories are to be defended on largely explanatory grounds, not on the basis of a priori arguments that appeal to intuition. Accordingly, perceived difficulties with conceptual analysis provide arguments for naturalism.

One such argument centers around the failures of the classical theory of concepts. Earlier, in Section 2, we noted that paradigmatic conceptual analyses require concepts to have classical structure, an assumption that is increasingly difficult to maintain. For this reason, a number of philosophers have expressed skepticism about the viability of conceptual analysis as a philosophical method (e.g., Ramsey 1998, Stich 1992). Others, however, have called into question the connection between conceptual analysis and definitions (Chalmers & Jackson 2001).

Another objection to conceptual analysis is that the intuitions that philosophers routinely rely upon may not be shared. Anyone who teaches philosophy certainly knows that half the time students have the “wrong intuitions”. But who are we to say that they are wrong? And given that people disagree about their intuitions, these can hardly be treated as objective data (Cummins 1998).

Things become even more interesting if we branch out to other cultures. In a preliminary study of East Asian vs. Western intuitions, Jonathan Weinberg, Shaun Nichols, & Stephen Stich (2001) found that East Asians often have the “wrong intuitions” regarding variations on classic philosophical thought experiments, including Gettier-type thought experiments. At the very least, this work suggests that philosophers should be cautious about moving from their own intuitions to claims about the proper analysis of a concept.

What's more, the cultural diversity that the work in Weinberg et al. points to raises a troubling question for philosophers who want to establish normative claims on the basis of analyses of concepts, such as the concept of justification. Suppose, for example, that East Asian culture offers a different concept of justification than the one that is embedded in Western commonsense thought (assuming for sake of argument that there is a single concept of justification in each culture). In addition, suppose that East Asians employ different inferential practices than our own and that their practices do a fair job of conforming to their concept of justification and that ours do a fair job of conforming to our own. On what basis, then, are we to compare and evaluate these differing practices? Does it really make sense to say that ours are superior on the ground that they conform better to our concept of justification? Wouldn't this just be a form of epistemic prejudice? After all, the question arises whether, given the two concepts of justification, ours is the one that ought to be used for performing normative epistemic evaluations (Stich 1990).

Much is at stake in the debate between conceptual analysts and naturalists, and it is likely to be a central topic in the theory of concepts for the foreseeable future.

Bibliography

Other Internet Resources

Related Entries

belief | Frege, Gottlob | innateness: and language | language of thought hypothesis | mental content: nonconceptual | mental representation | naturalism | Vienna Circle