Copyright (c) Amy Kind 2003
PSYCHE, 9(14), December 2003
Previously held: http://psyche.cs.monash.edu.au/v9/psyche-9-14-kind.html
KEYWORDS: concepts, empiricism, imagism, intentionality, categorization.
REVIEW OF: Prinz, J. (2002). *Furnishing the Mind: Concepts and Their Perceptual Basis*. Cambridge, MA.: The MIT Press. 358 pp. $40 hbk. ISBN 0262162075.
ABSTRACT: This essay reviews Prinz's *Furnishing the Mind*, a book in which he develops a Lockean-inspired version of concept empiricism that he calls *proxytype theory*. Prinz compares this view to rival theories of concepts, arguing that his view should be preferred because it is better able to meet the desiderata that a theory of concepts should meet. This review discusses these desiderata, Prinz's criticisms of the rival theories, and his development of his own view. I suggest that some of the rival theories rejected by Prinz might be in a better position to meet the desiderata than he allows.
1. THE ARGUMENTATIVE STRATEGY
In a key passage from *An Essay Concerning Human Understanding*, Locke questioned how the mind comes to be furnished. Prinz aims in this book to vindicate a Lockean answer to this question, namely, that all concepts derive from experience. Once Locke's concept empiricism has been properly modernized, Prinz argues, we can see that it is compatible with the latest research in cognitive science; in fact, such research can even be seen to support an empiricist view about concepts. Moreover, Prinz claims that his Lockean view--a view that he calls *proxytype theory*--offers a more satisfactory theory of concepts than any of the leading alternatives.
This last claim is defended by measuring how well such theories satisfy a list of seven desiderata laid out in Chapter One. According to Prinz, an adequate theory of concepts must
- have sufficient *scope*. Given the many different kinds of concepts (sensory, formal, theoretical, abstract, etc.) that humans are capable of having, we cannot focus on one kind of concept at the expense of the others.
- explain the *intentionality* of concepts. We need to explain how a concept comes to represent what it does.
- account for the *cognitive content* of concepts. Conceptual content often comes apart from referential content. Concepts that are coreferential can differ--for example, the concept of the morning star has different cognitive content from the concept of the evening star, though both of these concepts pick out the planet Venus. Moreover, concepts that are not coreferential can shar cognitive content. Compare my concept of water with the concept of water held by my doppelganger on a Putnam-style Twin Earth. My concept refers to H20, while her concept refers to XYZ, yet there is nonetheless 'some intuitive sense in which our two extensionally distinct concepts are alike.' (p. 8) A theory must thus balance the respects in which concepts are fine-grained and the respects in which they are coarse-grained.
- explain the *acquisition* of concepts. If there are any innate concepts, how were they introduced into the human genome? How did humans evolve so as to be able to acquire the learned concepts, i.e., how did the human conceptual faculty evolve? And for any given learned concept, how does it come to be acquired? Are the many different kinds of concepts all acquired in the same way? A theory should shed light on all of these questions.
- accommodate *categorization* abilities associated with concepts. When a person has a concept, she can identify instances of that concept. She can also identify the attributes of things that fall under that concept. A theory should be compatible with the various empirical findings about these abilities. (This psychological research is nicely summarized by Prinz in the first chapter.)
- account for the *compositionality* of concepts. A theory must be able to explain our ability to build compound concepts by combining other concepts.
- accommodate the *publicity* of concepts. No theory will be adequate unless it enables concepts to be shared across individuals.
Having given this list, Prinz addresses various complaints that can be raised about it. One problem arises from the inclusion of an allegedly unwelcome desideratum. For example, some might object to the categorization desideratum. Why presume that concepts, the constituents of thoughts, must be intrinsically connected with object-classification mechanisms? Prinz notes that psychologists generally 'regard the categorization desideratum as the main motivation for postulating concepts.' (p. 99) But that fact in itself does not give us a philosophical motivation for accepting this desideratum. Another problem arises from the exclusion of some other allegedly important desideratum. Most notable, perhaps, is the absence of any desideratum connected to linguistic meaning. Even without endorsing the strong view that concepts just are the meanings of words, we might naturally want our theory of concepts to contribute to our theory of linguistic meaning.
Considering the use to which Prinz wants to put this list, these complaints are non-trivial. He first wields this list against rival theories of concepts, arguing that each such theory fails to satisfy one or another of the desiderata. He then argues for his own theory by attempting to show that it can satisfy all the desiderata on the list and accordingly should be preferred to its rivals. Thus, one wishing to argue against Prinz's version of concept empiricism might naturally try to take issue with his desiderata. Alternatively, of course, one might argue either that one or more of the other theories fares better with respect to the desiderata than Prinz suggests or that Prinz's own theory does not fare as well with respect to the desiderata as he claims. My own inclination is towards the first of these alternatives. If Prinz is right that his own theory can satisfy these desiderata, then I expect that some of the other theories are not as inadequate as he claims. Just as Prinz borrows strategies for accommodating some of the desiderata from rival theories, the rival theorists might be able to borrow Prinz's strategies for accommodating some of the other desiderata. I discuss this in further detail below.
2. ARGUMENTS AGAINST RIVAL THEORIES
The six rival theories that Prinz considers in Chapters Two through Four are imagism, definitionism, prototype theory, exemplar theory, theory theory, and informational atomism. Although Prinz's discussion of these theories focuses on their failures with respect to various of the desiderata, what seems most important, and thus what I will focus on here, are (1) his discussion of imagism, given that his own view is a modernized version of this view; (2) his arguments that the first five theories all fail to satisfy the intentionality desideratum; and (3) his argument that the sixth theory, which offers a promising account of intentionality, nonetheless fails to satisfy the categorization desideratum.
2.1. IMAGISM
Imagism has considerable philosophical tradition behind it. Deriving from the work of the British empiricists, the imagist claims that our concepts can be identified with picture-like images in the mind. Though the view Prinz himself ultimately offers is a modernized version of imagism, he thinks that traditional imagism fares quite poorly with respect to the above seven desiderata, failing to accommodate scope, intentionality, categorization, compositionality, and publicity. Imagism seems restrictive in scope, for example, since there seem to be many concepts that we have for which there could not be corresponding images--consider abstract concepts like justice or democracy. Likewise, the specificity of images proves problematic with respect to the publicity desideratum: "If my concepts are constituted by my images of the objects I have experienced, they differ from your concepts for the ssimple reason that we have experienced different objects. ... To satisfy the publicity requirement, it seems that concepts must abstract away from the differences between particular category members. Images, as imagists traditionally conceive them, do not do that." (p. 30)
Though imagism does not have many supporters in contemporary discussions of concepts, one might nonetheless here be concerned that Prinz targetsan overly naive version of imagism. This concern will come out more clearly once we turn to his own view below. But for now, let us consider Prinz's criticisms of the imagist account of the intentionality of concepts. Prinz takes imagism to rely upon a resemblance-based account of intentionality. As he then notes, resemblance is insufficient for intentionality. Because images resemble many different things, they have ambiguous reference. Any image I have that resembles a car will also resemble a truck, and thus refer to both cars and trucks. Because I can have a concept that refers unambiguously to cars, imagism runs into trouble. Further problems arise from the fact that swhile resemblance is symmetric, reference is not. Not only does my image of a car resemble my car, but also my car resembles my image of a car. It thus seems that just as my image of a car refers to my car, my car should refer to my image of a car.
For these reasons, the resemblance-based account of imagism cannot account for the intentionality desideratum. At this point, given these criticisms, a question naturally arises: Must imagism itself presuppose a resemblance-based account of intentionality? This question is left unanswered--a fact that is especially surprising given that Prinz's own view (inspired by imagism) adopts a different account. Moreover, an argument can be made that the answer to this question is "no." Conside non-mental images, like photographs. The representational story about photographs is not a resemblance-based account, or at least, not solely so. A photograph of a person will still be a representation of that person even if it does not resemble her. Likewise, a photograph of one identical twin is not of the other identical twin even though it resembles them both. When it comes to photographs, at least, it seems we need to invoke a causal connection to explain their representational content. However this representational story is spelled out, perhaps we can give a similar story when it comes to the intentionality of mental images.
2.2. DEFINITIONISM, SIMILARITY-BASED THEORIES, AND THEORY THEORY
Like imagism, definitionism has considerable philosophical tradition behind it. Owing to rationalists such as Plato and Frege, this view claims that the concept of X can be identified with a list of features, each of which is a necessary condition for being X and which together are sufficient for being X. The definitionists attempt to account for the intentionality of concepts by claiming that a concept refers to those things that satisfy this list of conditions. But according to Prinz, the definitionists merely "postpone the problem of intentionality rather than solving it"; ultimately, the account "tells us absolutely nothing, because "satisfaction" is a synonym for "reference"." (p. 40)
The next three theories that Prinz considers grow primarily out of the psychological literature. In the 1970s, research in cognitive and developmental psychology showed that people typically group objects together on the basis of judgments of similarity. These findings led to two similarity-based theories of concepts: prototype theory and exemplar theory. On a prototype theory, concepts are identified with mental representations (called *prototypes*) of what we take to be the best instances of a given category. On an exemplar theory, concepts are identified with collections of mental representations (called *exemplars*) of previously encountered instances of a given category. Though Prinz believes that we likely use both of these types of representations when we engage in categorization, he argues that neither type of representation should be identified with concepts. Importantly, these accounts fail to satisfy the intentionality desideratum for the same reason as imagism. They rely on similarity, or resemblance, to provide an account of intentionality, and as we have already seen, this is insufficient.
The theory theory, which developed in the psychological literature in the mid-1980s, views concepts as mini-theories of the categories that they represent. These mini-theories contain considerable information, including beliefs about hidden essences, casual and explanatory relations, and broad ontological categorizations. But while the theory theory does a nice job of satisfying the scope, cognitive content, and categorization desiderata, it too fails to provide an adequate account of the intentionality of concepts. Most likely, the theory theorist will suggest that we construe concepts as definite descriptions that refer by picking out unique features of their referents. Prinz offers several objections to this suggestion. First, theories do not typically specify necessary and sufficient conditions for category membership. Second, even in cases in which theories do specify essential features of what they refer to, those essential features are often specified circularly. Finally, because theories often contain false information, it looks as if many theories will fail to pick out any unique referent if they are construed on the model of definite descriptions. Thus, as it stands, the theory theory cannot satisfactorily accommodate the intentionality desideratum.
2.3. INFORMATIONAL ATOMISM
While the theory theory goes to one extreme by packing a maximal amount of information about the category represented into a given concept, the last theory that Prinz considers--informational atomism--goes to the other extreme, denying that concepts have any complexity or structure. According to informational atomism, which is primarily associated with the work of Jerry Fodor, concepts are unstructured symbols that carry information about things by way of a nomological connection to them. As such, informational atomism provides a particularly promising account of the intentionality of concepts--a desideratum that none of the other theories was able to satisfy. Unfortunately, however, the atomistic nature of the view means that it is strikingly unable to account for categorization. If concepts are atomistic, and hence have no constituent features, then we cannot explain our categorization abilities in terms of information contained in the concept. The atomist must claim that identifying the category that a given object belongs to or producing information about the standard features of an object in a given category has nothing to do with one's concept of that category.
3. DEFENSE OF PROXYTYPE THEORY
Having surveyed these rival theories, Prinz turns in the rest of the book to developing and defending his own version of concept empiricism. Because he attempts to ground conception in perception, he sees his view as being in the spirit of imagism. To accommodate contemporary empirical research, Prinz updates the traditional imagist view by abandoning the supposition that concepts are conscious picture-like entities. He also borrows heavily from some of the other theories that he has already rejected; by doing so, he believes that he can avoid the objections that proved fatal to traditional imagism. Most notably, (1) his proxytypes are hybrids of the sorts of representations invoked by traditional imagism, similarity-based theories, and theory theory; (2) to account for the intentionality of these proxytypes, he adapts the nomological connection account utilized by informational atomism.
Let us take these points in reverse order. On Prinz's view, concepts get their intentional content by standing in informational relations with what they represent. But to avoid the problems associated with informational atomism mentioned above, Prinz needs concepts to have structure. He thus sees proxytypes as *detectors* rather than as *indicators*: "An indicator is an unstructured entity that falls under the nomological control of some property (e.g., a light that is flashed when a particular letter is presented to a letter-detecting device). A detector is a mechanism that mediates the relation between an indicator and the property it indicates (e.g., a template that causes a red light to be flashed when As are presented)." (p. 124) Detectors are typically structured entities, and they enter into nomological relations with properties at least partly in virtue of their structure.
But what kind of representations are proxytypes? Prinz sees them as perceptually derived representations of categories that are or can be activated in working memory. (A given representation stands in, or serves as a *proxy*, for the category that it represents.) These perceptual representations are meant to be quite different from the picture-like representations postulated by the traditional imagists because, for example:
a) There is a wide range of kinds of representations, corresponding
to the multiple levels of perceptual processing.
b) Not all of the representations need be conscious.
c) The representations are highly structured.
d) The representations are schematic and can abstract away from
specific features.
Given this updating of imagism, one surprising lacuna in Prinz's discussion is the lack of a treatment of any of the contemporary discussion of mental images. Though in the heyday of behaviorism there was a tendency to offer eliminativist accounts of mental images, there are now many philosophers who accept the psychological reality of images but who deny that they represent in a picture-like fashion. For example, some philosophers suggest that the introspective evidence suggesting that images represent in a picture-like fashion is best interpreted as evidence that imaging is like perceiving. Images are then construed as having the same kind of structure as percepts, a structure that need not be the naive "conscious pictures" of Locke, Berkeley, and Hume. Granted, images as traditionally understood will be far inferior to proxytypes in accounting for Prinz's desiderata. But what of images construed in this more contemporary way? A consideration of some of the literature in the imagery debate--both philosophical and psychological--might help to differentiate Prinz's proxytypes from various contemporary treatment of mental images.
Having offered this proxytype account, Prinz needs to show how it can accommodate the desiderata more successfully than the rival theories that he has discussed. Some of this falls out from his discussion of what proxytypes are. For example, (c) helps him to meet the compositionality desideratum. The compositionality of proxytypes is discussed in detail in Chapter Eleven.
Likewise, (d) helps him to meet the scope desideratum, or at least, to meet it more successfully than traditional imagism. Since proxytypes can abstract away from specific details, a proxytype theorist can explain how, for example, we get our general concept of triangularity. (It is often charged that this concept poses a problem for imagism, since all images of a triangle must be scalene, isosceles or equilateral.) But Prinz has more work to do to show that proxytype theory can satisfactorily answer the scope requirement, since even the abstractness of proxytypes does not explain how perceptual representations can be formed for concepts like causation or democracy. Though Prinz admits that he cannot provide a full explanation, in Chapter Seven he offers suggestions aimed to "provide grounds for optimism and directions for research." (p. 187) Many of the concepts that serve as traditional counterexamples to concept empiricism are problematic because they are concepts of things that not directly perceivable. However, Prinz argues that many of the referents of these concepts are nonetheless capable of being perceptually *tracked*. Take the example of an electron. Most of us form this concept by relying on word tracking, i.e., we defer to experts' uses of the word ¥electron". The experts track electrons perceptually by tracking the effects electrons produce on scientific instruments or by looking at photographs of electron traces. They build proxytypes which include these perceptual representations, which themselves are reliably caused by electrons. Ultimately, then, the concept is grounded in perception. Prinz extends this sort of analysis to other unobservables, noting that "Representations of words, emotions, and other perceptual states conspire with causal theories to track properties in the world. The range of properties that can be perceptually tracked far exceeds the range of properties that can be directly perceived. This gives proxytypes considerable expressive breadth." (pp. 187-88) Prinz's discussion, though not conclusive, suggests a plausible way that proxytype theory can protect itself from the charge that it cannot satisfy the scope requirement.
Less satisfying is Prinz's treatment of the publicity desideratum. As he notes, empirical research suggests that proxytypes may not be shared across individuals, thus threatening his ability to account for publicity. In response to this threat, he explicitly relaxes the desideratum:
This concern [raised by the empirical research] can be met by relaxing the publicity requirement. Rather than demanding strict identity between default proxytypes, we can settle for similarity. If you and I both agree about the most conspicuous walrus features, than we understand each other when we use the word "walrus," and we engage in similar walrus-directed behaviors. If the publicity desideratum is intended to explain such examples of coordination, a theory that predicts considerable conceptual similarity will suffice. (p. 158)
Though I am inclined to accept this relaxation of the publicity requirement, I worry that doing so calls into question his earlier treatment of the rival theories. Prinz criticizes the first five theories he considers for their inability to meet the publicity requirement. But these theories are all held to the strict publicity requirement, and it might well be the case that they would be able to meet the requirement once it has been relaxed. At the very least, Prinz owes us an explanation as to why that is not the case.
Similarly, one of the key threats to the first five rival theories is their inability to accommodate the intentionality desideratum. As we have seen, Prinz avoids this problem by borrowing the nomological account of informational atomism. But why isn't that account als available to any of the other theories? Again, at the very least Prinz owes us an explanation as to why they too cannot borrow that more promising account.
4. CONCLUSION
This book offers an interesting empiricist treatment of concepts, and Prinz does an especially good job of incorporating recent psychological research and motivating its philosophical import. The case that he mounts on behalf of his proxytype view is an impressive one, and I am convinced that it is an important view that deserves serious attention. Though I have suggested that proponents of rival theories of concepts have some room to maneuver in countering Prinz's objections, it is nonetheless clear that the objections cannot be ignored. I would not be surprised if much of the philosophical discussion about concepts in the near future were framed by the discussion in this book.
