Kevin Lee
5/5/1997

The question of what concepts are and what it is to have them is the central issue in Georges Rey's "Concepts and stereotypes," a critical response to Smith and Medin's (S&M) work Categories and Concepts. Rey argues that their research in how people access concepts is irrelevant to the nature of concepts themselves. S&M acknowledge that they were only interested in questions of access, not a theory of identity. They also see a need for a metaphysical conception of what a concept is, and seem to agree with Rey that their epistemological characterization of concepts is insufficient for any question of what concepts are really are. Departing from their views, I will first discuss how the question of trying to identify what a concept is is biased from the outset in that it assumes a metaphysical conception is even possible. I will try to show that there is no firm justification for this bias, as it is based on a purely psychological belief in essentialism (defining conditions, or metaphysical reality). Ultimately, concepts are what they are only through their use in a linguistic community. Certain sciences may serve one particular interest of categorization very well, but their authority is derived only relative to this particular interest. The overlap in extension and use of a common symbolic token by different contexts of categorization are neither a confusion of different concepts nor an assertion of conceptual identity, they are merely the effects of dynamic usage in a given community.

The question that Rey is addressing throughout his discussion is "precisely what is a concept?" (237) He wants to find a theory of concepts, something that specifies the identity conditions of them. This issue is hopelessly biased from the outset. He assumes that it is possible to clearly individuate concepts, that defining conditions of the concept 'concept' exist and that his task is merely to discover them. What this means is that he has presupposed a metaphysical conception of concepts before he has even begun. This does not show that Rey is wrong; on the contrary, it asserts that he is being consistent in that his conclusion for a metaphysical account follows from (indeed is assumed in) his premises. Any conclusion significantly different from the one he arrives at must either start from an entirely different notion of 'identity' or deny the sort of metaphysical reality he wishes to assert.

It is fairly uncontroversial to say that some world must exist outside our knowledge of it, but to say that the world divides itself in some specific way and that finding the categories of division is purely a matter of discovery is more problematic. One of the things that concepts allow us to do is mentally separate one type of entity from other types. The question is how this cutting-up of the world happens. It might be a purely pragmatic way of dealing with a massive amount of information, or it may map onto natural divisions in the world. If the world is somehow pre-divided, then it may be possible to say that the correct definition of a concept exists independently of whether we know it or not.

Rey cites the work of Putnam and Kripke to support this view. Consider natural kinds such as 'gold' (discussion on pp. 252-255). It is possible that the features that we use to identify 'gold', such as 'the precious, yellow metal mined in the Sierras' or 'the metal used in wedding rings' turn out to be wrong. That is, 'gold' might never have been discovered or used or ever known, and yet it would still be 'gold'. It has a definition independent of anything we know about it: 'basic chemical element with atomic number 79'. Even if we do not know this fact, or never had atomic theory, it would still be metaphysically necessary that 'gold' has the atomic number 79. This seems to show that concepts can have definitions that competent users may not know. Rey takes this argument a step further, stating that before atomic theory, no one could define gold, "perhaps no one actually can today either (atomic theory might be wrong), or ever... the issue of definitions [is divorced] from the issue of anyone's ability to provide them" (255)

To see why this assertion of metaphysical truth is mistaken, let us take his example of 'gold' again. It seems intuitive to agree that some of the features we use to identify 'gold' may be wrong. People in the past may have mistaken pyrite for gold, and even now we tend to think of gold as yellowish in color when pure gold is white. However, there is a certain relation between what we think is gold and what we defer to experts to tell us really is gold. In general, we acknowledge that the sciences provide a relatively cohesive way of sorting things in the world in useful and meaningful ways. Even though we talk about 'gold' in lay terms, we know that there is the word 'gold' in chemistry, and the extension of the 'gold' in chemistry overlaps greatly with the extension of 'gold' that we are using. This is not to say that either the overlap in extension or use of the same word are arbitrary designations, but that the lay conception of 'gold' contributed, at least in part, to labeling that particular chemical substance 'gold', and continues to be related to it. Moreover, when we talk about certain aspects of 'gold' (in lay terms), we may be referring to the chemical properties of 'gold' in the sense that our interest in talking about 'gold' is the same interest that chemists have when they talk about it. So to answer Rey's argument, it is possible that atomic theory is wrong, but nothing in a different system of categorizing matter is going to be called 'gold' unless it somehow corresponds to what we already know is 'gold' (in both lay terms and our old atomic theory).

Having a new scientific conception of 'gold' may eventually change what we include in our category of 'gold', and this change might even continuously progress so that at some point none of what we once called 'gold' is now still considered 'gold'; when this occurs, it becomes difficult to say that we were somehow wrong before and correct now, we are simply talking about different things. In other words, there is never anything defining some static concept 'gold'. Even if we eventually do converge upon some scientific conception that greatly overlaps with our common conception, much of our common usage will not require or imply a scientific conception. Our lay theories about deciding what counts as 'gold' may share some of the same interests as scientific categorization of what counts as a simple substance, and so in some cases we do mean to appeal to science to provide a more cohesive account. This appeal may eventually have implications in our common usage of the word, and it is through the relationship between that usage and a scientific one (and perhaps other ones) that a full concept of 'gold' emerges. Conceptual change takes place within this background of shared interest, and it is the large overlap in and common name for identifiable entities throughout a linguistic community that determine what the concept associated with that name is.

If concepts are purely products of a linguistic community, then how is it conceivable that concepts may have definitions that are not known at all? Let us say, as Rey does, that psychological essentialism is true - that generally, people believe that concepts do have essential conditions and definitions. Because we often run into cases where defining conditions seem to exist even though we do not know them, we might assume that they exist for all concepts. This can work at a community level, so that even pooling all the knowledge of experts, we continue to believe in defining conditions for concepts that no one can define. This view is not necessarily entirely wrong, but it is based on the idea of having better or worse categorization schemes.

Presumably, the 'best' way of categorization will provide us with defining conditions. We might assume that in our current community we do not have adequate methods of categorization (or theories upon which we base these methods), and with better ones we may come up with the proper definitions. This issue leads us to understand why science seems to have more authority in providing definitions than other ways of categorizing. What it appears to do is provide an interest-free way of simply discovering what natural divisions there are in the world, and this should lead to great conceptual coherency. To take a contrasting example, I may classify animals as 'wumps' if they are dangerous to me, and as 'sumps' if they are not. The boundary between 'wumps' and 'sumps' is a fuzzy one, and some animals are one or the other at different times. In fact, what may be a wump to me may be a sump to somebody bigger and stronger. The reason is because these categories are interest-relative; my divisions are entirely based upon some way that makes a difference to me. Better or worse ways of categorizing 'wumps' and 'sumps' make sense only relative to the interest of wanting to distinguish between dangerous and non-dangerous animals, and so the interest itself serves as something like a defining condition. Science is often construed as a means of eliminating this personal element, of classifying things in some objective way that is equally valid at all times and regardless of personal interest. The approach that science attempts to take is one of finding out what is already out there. If this approach is right, then better or worse categorization is based on how well we match our categories to the world, not to some arbitrary interest. The unknown essential definitions would simply be categorizations of the world as it really is, and science might provide a way (to Rey, the best and perhaps only way) of knowing them.

Although in this account I have granted the possibility that concepts may have defining conditions specified by the interest involved in their categorization, it is doubtful that Rey would acknowledge that there is a metaphysical condition that determines being a 'wump' versus a 'sump'. First of all, it is a purely stipulative definition, and does not seem to have anything to do with how the world really is. There is great variance and lack of coherency in that category, and moreover the defining condition (of being 'dangerous') is itself ambiguous. Rey's criterion of "if p then necessarily p" (254) may not ever capture defining conditions of such a volatile category. So if science is qualitatively different than these interest-relative methods, then it may have a stronger claim to authority in supplying defining conditions. This question is not about whether the current scientific world is successful or not in providing defining conditions, but whether it is in principle any different than any other method of categorization. It is not within the scope of this paper to present an entire critique of science, nor is the intention to show that science is useless. The precise issue at hand is that science is useful, and that we do science because it allows us to predict how phenomena will act, why the world is as it is, etc. In short, science captures the world in meaningful and interesting ways to us. It is possible to use scientific methods and have entirely different categories; this depends on whether those categories make a difference to us. But this is precisely how any method of categorization works. The fact that certain sciences present very cohesive accounts and theories does not necessarily mean that the world divides itself up in that way (and not other ways), but that dividing the world in that way is significant to us. There is a problem here with specifying a difference between how the world divides itself and how we divide the world, because there is a sense in which our world is how we conceptualize it. Issues such as theory-laden perception and paradigm entrenchment show certain difficulties in separating reality from the perception of it. The distinction between epistemology and metaphysics blurs here; it is not that one becomes eliminable in service of the other, but rather that we must have a different view about the nature of reality itself.

Let us return to the 'gold' example. The lay concept of 'gold' is pretty well established in modern culture. It is a precious metal that has a yellowish tint and is fashioned into various forms of jewelry with varying amounts of purity. It also has a long history as a standard of currency, was involved in various get-rich campaigns in California and Alaska, etc. At the same time, this substance is a simple element abbreviated Ag with an atomic number of 79. But are all the characteristics of 'gold' other than atomic structure irrelevant to its identity? Rey's characterization of Putnam's argument is not entirely plausible. It seems that in a real sense, 'gold' is yellowish in color. What is not yellowish is the substance of which 'gold' is mostly composed. When we are talking about 'gold' as lay people, we might really mean to indicate that color is a very important part of the substance. Some of us may even consider it to be an essential defining condition of gold. Psychologically, we may have placed the gold color of gold in part of the essence slot. If this is true, then it seems that chemists have a different concept of 'gold' than what lay people have. Rey might say that the lay people are wrong to the extent that chemists have a more 'optimal' theory, and so even though lay people think they know defining conditions, they are mistaken.

It is unnecessary to impose such rigid terms. What lay people mean when they speak of gold is not identical to what chemists mean. But that does not imply that the token 'gold' is being used to indicate two different concepts, or that the layperson's concept is wrong. And just because a layperson may defer to chemists when talking about what 'gold' really is, it also does not imply that they are talking about the same exact concept. As mental representations, concepts allow us to pick out certain groupings in the world. Communication is possible because the same token is often used to denote the same or a large number of the same entities whenever it is used. That is, there is a large overlap between things with atomic number of 79 and things we think are made of that shiny yellow metal. The different possible contexts in which categorization can take place (with different goals in the categorization process) become problematic when the extensions (or even possible extensions) of each grouping do not overlap. The use of a single token for multiple possible groupings may lead to disputes, because there is the psychologically real belief that the single token denotes some single individuated set of entities, i.e. that there are defining conditions. In these cases, other issues such as conceptual coherency or theoretical stability may determine which groupings are to be considered more central. The interrelated nature of these many possible groupings indicate that the groupings themselves are not independent, but are intertwined with each other and other concepts. In this way, concept individuation, or concepts themselves, are merely useful ways of simultaneously organizing and constituting the world. They do not have defining conditions because knowing them is a holistic and dynamic process. It is misleading to say that they are holistic entities, though, because it implies that people might have different concepts based on small differences in their overall mental states. What I want to propose is that they are not 'entities' at all, that individuating them is possible only in rough and general ways, and that the multiple and changing ways of using a token are constantly forming what is relevant to that concept. In this sense, Smith & Medin (and other psychologists') work in showing typicality effects and the problem of locating defining features is relevant to how to characterize concepts. They do not directly claim that the metaphysical conception of concepts is wrong, but point to a way of reconceptualizing what it is to be a concept.

It might be argued that some concepts such as nominal kinds or certain technical terms have real defining conditions because they have been explicitly stipulated as such. Even in these cases, however, the process of conceptual change occurs quickly. For example, kin terms no longer have their purely stipulated meaning - what it is to be a 'mother' is hotly debated. This does not mean that the stipulated definitions are not still defining in some sense, but accepting this definition alone is not an acceptable basis for claims of correctness. Someone claiming that biological motherhood is the 'correct' motherhood because it is stipulated as such is not saying anything interesting about motherhood. Similarly, an unmarried male who has taken a vow of celibacy is not only a bad example of a 'bachelor', he does not seem to be a bachelor at all. To claim that he is one simply because he fits the stipulated criterion of bachelorhood as 'unmarried man' is not acknowledging the development of the concept beyond its stipulated origin. If two people hold different stipulated definitions for a term and they understand each other (I claim motherhood is biological, and you claim it is purely raising a child, and we are both aware of each other's definitions), then it seems there is no ground for argument. We are just defining the token 'mother' in different terms. At worst, it would be confusing if we all used the same token in completely different ways, but nonetheless there is nothing to argue about with pure stipulations. The reason that there does seem to be argument is that we wish to associate our (perhaps controversial) ideas about a concept with an accepted notion of that concept and the rich ways in which it is used. This association plays a legitimizing role by appealing to the apparent authority of the accepted notion. In this manner, a pure stipulation is useless without an intricate connection to the ways in which a concept is used. Similarly, even starting from a stipulated definition, the actual use of a concept by a community can change the range of that concept's applicability. Retaining the original stipulation may have some practical import for technical precision or reference purposes, but any more complex use of a term cannot escape the psychological opportunities for dynamic change.

In this paper, I have argued against the idea that concepts have defining conditions. To do this, it was necessary to have a completely different view of what a concept is, that is, that it is not anything in particular. Through usage, tokens that denote a certain grouping of things gain meaning and import. The usage of the token can change through time, and various interests in categorization may bring about changes in what entities the token picks out. Concepts, then, are not things that can be individuated and distinguished clearly, they are part of a holistic conception of all our knowledge. Moreover, they seem to be dynamic groupings that have a historical development and so carry implications in areas that we may not be aware of (as in scientific use). Everything we associate with a concept may contribute to what that concept is for us (including conditions that are outside of us), and issues of communality and cohesiveness may in general determine centrality of features. It is not proposed that concepts are defined by these terms, but that the traditional view of a metaphysical/epistemological distinction collapses with concepts because it does not make sense to speak of their identity outside of how and with what interest they are used by a community.