CHAPTER V
CONCEPTS
Concepts are the active structures that do the representing. All knowledge rests in the possession of concepts. While all knowing is done through representing, the ability to enter into knowing relationships with the world, all knowledge in posse, as potentiality, rests in concepts. This chapter considers concepts, what they are, how they work, and why they are not representations.
5.1: Why do Concepts Present such a Problem?
Some of the peculiarities of the operation of concepts account for much of the plausibility of the attempt to make representations come alive and to take representations as mental pictures. In particular, concepts account for memory, imagination, and abstract thought free of dependence on actual perception. The recognition that we have these abilities results in some very important conclusions for the theory of knowledge.
The fact that we have memory of actual experiences that seems to be identical in content, if not in feel, to actual experiences of the world seems to imply that the phenomenological content of our experiences is not dependent on our actual interaction with the object. This, of course, led to the view that perception impresses ideas or pictures into our mind that we can call up and look at with our mind's eye when we have a memory. (Hume appears to be the clearest example of this.) This also implies that what we are looking at with our mind's eye when we are actually perceiving is not the object itself, but a vivid impression of it in our mind, a mental picture or idea. Combine this view with a mind-body split and it becomes very difficult to see how our perception can accurately represent or picture reality; it becomes very difficult to see how anything can be like an idea but an idea. Memory shows us that the phenomenological character of our perceptions, the properties we perceive, are not attributes of the object as it is by itself, but are impressions upon our mind which retain indelibly the contributions made by this medium. The fact that we can have the same perceptions with or without interaction with the object shows that the phenomenological content of our perceptions is subjective, due to our constitution and not the object.
The abilities of imagination[1] cause even more problems for the attempt to see representations as pictures. For in imagination we see that the phenomenal properties we get in perception can be called forth and combined at will. It becomes clear that the mind is not simply a wax tablet for the impressions of sense. Mind is active, and in imagination it seems as if our concepts are active. With this it becomes much more difficult to regard concepts or ideas as pictures. General or abstract ideas must be abilities to make and manipulate many pictures.[2] Having such abilities allows us to think of things that are not present and to think of abstract properties such as redness or triangularity. Imagination frees our representative powers from their ties to sense perception. It is even less clear, however, how abstract ideas can mirror particular reality. They no longer even seem to be pictures; as Berkeley and Hume showed, there can be no abstract images.
Once it is seen that in our imaginings the images produced are the results of the activity of our concepts, it is no great leap to the realization that the images in perception are also a result of the activity of concepts. This, of course, was the leap that Kant made. With this move we see that the images we get in perception are not even impressions upon our mind; they are our own constructions. They are representations of reality constructed according to other representations, concepts.[3] The veil of ideas is doubled[4] in thickness. When concepts are seen as active and as representations, it becomes necessary to regard them as representations of a world that we construct. This, of course, is Kant's Copernican Revolution. It is the source of the internal model of objectivity.
I will attempt, in the following sections, to give an account of concepts that allows us to account for memory, imagination, and abstract thought without regarding concepts or ideas as pictures in the mind and without leading to an internal model of objectivity. This will require taking concepts as active abilities, as Kant did, but not as representations, as Kant did.
Concepts, as we have seen before, are more or less discrete networks of dispositions to represent domains of interest in certain ways. They are dispositions to connect modes of interaction in acts of representing. Many such dispositions pertaining to a single domain of interest can be grouped together into one concept by connection with the set of basic abilities that direct our activity to that domain of interest and tend toward direct interaction with it.
The discreteness of these clusters is only relative, since there is no reason why a single disposition to connect two modes of interaction cannot be connected with more than one set of basic referring abilities. Thus my disposition to represent a person perceived by my senses as having blonde hair may be part of my concept of my mother, my father, and the numerous other people who I know to have blonde hair. It would have been highly wasteful for the mind to have duplicated this disposition anew in each concept in which it appears. Thus, it makes sense to assume that it is a single disposition with connections to many different sets of basic referring abilities, and, hence, a member of numerous overlapping clusters of dispositions that make up various concepts.[5] Thus we would expect the boundaries between concepts to be fuzzy; we would expect concepts to affect each other, and we would expect the particular operation of particular dispositions to vary with the conceptual context they are operating in.
Once we are freed from the illusion that representations must connect properties that are pictures or images of certain aspects of reality, we can see that concepts can connect varying types of interactions including perceived properties, actions, feelings, emotional responses, and other concepts. Thus my concept of my mother is a complex set of dispositions to perceive her, act toward her, emotionally respond to her, and think about her in various different ways. It seems to me that this view of concepts is closer to our normal pre-reflective view of concepts than are the philosophical views of concepts as list of properties or abstract pictures.
Concepts on this view must be physically and neurologically embodied structures. There is nowhere else for them to be. It is simplest to see the dispositions of which they are composed as neurological connections between neural structures that are usually activated by bodily and sensory interaction with the world.[6] These neural structures are what I have been calling modes of interaction. I call them this because they are defined by their connections to sensory and bodily receptors and motor neurons, hence they provide for and control all our cognitive interaction with the world.
Concepts also have another characteristic without which it is (at present) impossible to explain their operation. The operation of concepts gives rise to consciousness. It is impossible, on this view, to give an account of how concepts work without considering how they affect and are affected by the contents of consciousness.[7] Although it is rare in contemporary discussions of knowledge to consider the role of consciousness in knowledge, it is not surprising, at least to me, that knowledge and representing require consciousness. I for one would be a much less efficient knower if I were not conscious (or at least if I were less conscious than I usually am).
Therefore, saying that concepts involve dispositions cannot simply be a shorthand way of saying that they are structures that do things. (Although, since we do not normally have access to the actual neural structures involved, we must describe concepts in terms of what they do.) The dispositions must be viewed as what I call teleological tendencies. They are teleological not only in the sense that they have an object or are directed at some goal, in this case the connection of the two modes of interaction. These tendencies actively pursue their object, and, most importantly, they have a particular subjective character that guides and drives the tendency towards its completion or expression. That is, there is a way that the disposition feels when active and the consciousness of this feel is instrumental in guiding the activity of the organism to the expression of that tendency.
Consider a simple example: I have an itch. This can be seen as a disposition to connect my visual and tactile location of the itch with my specific motor abilities known as 'scratching'. The fact that I have such a disposition does not explain why I in fact scratch the itch or how if I am unsure of the exact location of the itch how this disposition guides my activity to find the right spot to scratch. It is the subjective character of the disposition or teleological tendency that drives the disposition towards its expression and which guides it along the way. Scratches that miss the spot don't feel as good as those that hit the spot (literally and figuratively). Thus, in the process of its interaction with the world the teleological tendency drives itself towards its own expression. As we shall see in the next chapter, agency and the possibility of knowledge, or the awareness that we are engaged in an act of objective representing, require that our concepts have these characteristics. Before we move on to that, we need to see more about how concepts operate in memory,imagination, and abstraction and why they are not representations.
The sketches in this section will be radically incomplete. I will not pretend to know how memory, imagination, or abstraction actually work. I will however begin to sketch how an account of them might be given according to the model of concepts outlined in this chapter.
First we need to get a bit clearer about what a mode of interaction is, how it is activated, and how they are combined with or projected onto one another. As noted earlier, what I am calling modes of interaction are neural structures defined in terms of their connections to efferent and afferent peripheral neurons. Their activation involves the manifestation of properties, feelings, memories, associations, etc., in the case of afferent modes of interaction and various types of bodily actions and movements in the case of efferent modes of interaction.
Representing involves the connection of these structures or the projection of one onto the other. The point of calling it a projection rather than simply a connection is to emphasize the fact that the activation of one of the modes of interaction is primary in time and leads to the activation of the other in virtue of their connection by a concept. In the technical terminology of Chapter One, the primary mode of interaction is the destination, the secondary is the source. For example, my eyes activate a visual mode of interaction. This is connected to tactile modes of interaction by concepts, so these tactile structures are activated apart from actual input from tactile peripheral neurons. This causes the projection of the tactile modes of interaction onto the visual mode of interaction. The experienced result is my visual perception of the tactile properties in my experience of the object or scene. The tactile properties are re-presented in terms of the visual ones. For example, when I see visual images on the television, I project the other types of properties I associate with these images onto my visual sensation and see people doing things instead of flickering images of light, which is what an animal without our conceots, such as a dog, might see.
It should be noted that what we identify as properties are not the individual contributions of each mode of interaction to the finished experience. What we call the property redness is the result of a complex process of abstraction from already finished experiences. Since we only become conscious or have experiences after modes of interaction are connected, we have no access to what they actually contribute to the experience apart from this abstraction.
Memory and imagination show us that our perception involves the activation and connection of modes of interaction. They also show us that these structures can be activated independent of actual interaction with the world. On the model presented here, one way of viewing concepts are as autonomous activators of groups of modes of interaction. Concepts tend toward the activation of the modes of interaction independent of activation of these structures from interaction with the world. (Of course, we must assume that concepts have a mutually inhibitory effect on each other to prevent too many of them from activating too many structures at once.)
Memory can be explained, although superficially, without much trouble on this model: We have an experience. It involved an act of representing or connecting the activation of modes of interaction. This was caused in an act of direct referring, where both modes of interaction are directly in contact with the world and were activated by that interaction. This resulted in a concept connecting these two modes of interaction and tending toward their joint activation independent of stimulation from the external world. Since this concept is a member of a cluster of concepts relating to that domain in virtue of connection with certain basic referring abilities, various different types of activations within that diffuse cluster can bring about the activation of this concept and result in the representing of the memory. It should be noted that nothing in this model requires the storage of the original experience; what is stored is simply the ability to engage in that act of representing again.
Imagination can be explained (in only its roughest outlines) in a similar manner. Concepts tend towards their expression in acts of representing, and these often result in images, among other things. What imagination shows us is that the activation of these concepts is under our control to some degree.[8] An important aspect of this ability is the way in which we are able to take connections or concepts in one cluster, relating to one domain of interest, and apply them to another domain by connecting that concept with the referring abilities that refer the new cluster to the new domain.[9] It should be noted that this metaphorical projection of concepts is itself an act of representing which takes as its source the concept to be projected and its destination the new cluster of concepts and, in particular, the set of referring abilities that unites the cluster as relevant to a specific domain.
Abstraction is also a type of representing that takes a concept as one of its modes of interaction. It seems to be the exact converse of normal perception, where a direct sensory mode of interaction is the destination and a concept by which the sensory interaction is connected with other properties and knowledge is taken as the source which is projected onto the sensory interaction. The other properties and knowledge are represented in terms of the sensory experience. In abstraction a sensory experience is represented in terms of a single property. Take for example our seeing an object as red. Here a concept that attempts to connect a mode of interaction whose activation we experience as redness with another mode of interaction is the destination. We attempt to project interactions onto this concept. That is, we try to see things we are interacting with as red.
This is a very important ability. It allows me to find the red ball on the desk, among other important things. I interact visually with objects on my desk, all the while activating the red mode of interaction. When I experience non-red things this activates other modes of interaction, and the mutually inhibitory effect of the modes of interaction inhibits the activation of the red mode of interaction. When I light upon a red object the activation from the interaction reinforces the activation from my concept. Luckily, I am aware of all this activation and inhibition; I feel it as a satisfaction[10] or frustration of my teleological tendencies. Therefore, I know that I have found the red object I was looking for.[11]
This ability to actively attempt to see the world in terms of a concept and then have the connections aimed at by the concept either reinforced or inhibited by interaction with the world forms the basis of our ability to gain objective knowledge of the world. Before moving on to a more detailed discussion of this in Chapter Six, we need to see why concepts are not representations.
5.4: Why Concepts aren't Representations
There are three main types of considerations that make it impossible for concepts to be considered representations. Of course, the most simple argument is that if the model of representing presented here is correct, then representing is essentially an act. Therefore, concepts cannot be representations because they are the active structures that do the representing, not the acts themselves. But rather than derive the conclusion that concepts are not representations from this model of representing, I would like to see it as an independently accessible conclusion that supports this model. Let us, then, review the three main considerations that rule out concepts as representations.
First there is a problem about the ontological status of concepts. If they are particular images, they cannot be representations, because there can be no similarity between images and objects as they are in themselves. The other alternative seems to be that they are abstract entities. Besides the metaphysical extravagance involved in this thesis, it does not make it any easier to see how concepts represent. Abstract objects do not interpret themselves any more than particular ones. When concepts are finally seen as active abilities, they are no longer accessible to our view. Putting them behind the scenes as dispositions or abilities makes it impossible for them to ever re-present anything to us since they are never present to us. It seems a fairly safe principle that what we are never conscious of cannot be a representation, and we are never conscious of concepts. Even if when, through neurology, we can begin to get an idea of what they are, it becomes no easier to see them as representations. It is slightly ludicrous when one is asked for a representation of a horse to present the asker with a set of slides of the neural structures that carry out our visual and mnemonic representing of horses. Even if they were to find a tiny image of a horse on one of the slides, there is no reason to take this as a representation of the horse apart from an act of interpretation than there is to take a footprint as a representation of a foot apart from some act of interpretation.
Once again, it should be noted that none of this means that we cannot analyze our concepts, make new connections between them, or play around with them in our heads. It means simply that any access we have to concepts is through what they do, how they cause us to represent. Since we are also aware of the operation of concepts, although not of their content, through the felt character of the teleological tendencies that make them up, we can also effectively guide their operation. Yet, even though we are aware of how they are guiding us towards their satisfaction and of how close they have come to this satisfaction, we are only aware of a particular expression of the general potential or concept itself.
Second, it seems that concepts as opposed to images are irreconcilably general and indeterminate in their application to particular things. This presents a number of problems in taking them as representations. On one hand, if they are taken to represent in virtue of similarity, then it seems that they can only represent general objects. My concept of redness is not similar to the red ball, but only to its redness. It represents an abstract object in the realm of forms, not a particular object. If, on the other hand, they are seen to represent in virtue of having been active in the construction of the objects they represent, it becomes difficult to explain how these concepts can be both pictures with similarity to their objects[12] and active agents that construct images. As we saw, if the application of concepts depends on their correctness or similarity, then it will be impossible for concepts to apply to anything to which they are not completely similar. The application of concepts requires an Intentionality based on the inclusion of referring abilities. Since these must be independent of representational content, a concept that can apply itself cannot be a representation.
Kant's solution of providing an intermediary faculty of Imagination to schematize the concepts will not do. Apart from the dubiousness of the ad hoc assumption that there can be entities that are half way between abstract concepts and particular sensuous images, the project seems to land one in a representational counterpart of Zeno's paradox. For how is the application of the concept to the schema determinately accomplished? It should require another intermediary; after all the schema is only half similar to the concept. Why shouldn't it require another schema half way between the concept and the original schema? But the real problem lies in the move from schema to particular, for there will always be determinate particular aspects of the image that are not in the schema. Plato's gradation of the forms from the Good down to the most particular form did not help him with the problem of participation. So the attempt to provide mediation between general concepts and particular images will not solve the problem of application for Kant.[13] How it is the schemata perform their function remains an "art hidden in the depth of the human soul, the true secrets of which we shall hardly ever be able to guess and reveal." (Kant 1966, A141, B180-81, p. 123)
The obvious alternative to Kant's view is to see concepts as sets of abilities that include abilities to refer to a particular. Such concepts can apply themselves because they are not representations that can only apply in virtue of their similarity to particular objects. They are formed in interaction with certain domains and include within themselves dispositions to refer back to those domains. Thus, they can apply in particular circumstances by producing acts of representing without having another representation outside of them that determines their application.
The third problem is a fairly obvious one. If concepts are seen as both representations and active in the construction of what they represent, it is clear that the external model of objectivity must go. If the representation constructs the object, there is no hope of its content being caused by the object. Objectivity becomes dependent on concepts rather than the other way around. I am assuming for this argument that no one would accept an internal model of objectivity unless forced to by lack of alternatives. One can avoid an internal model by taking concepts as the active structures responsible for representations, rather than representations themselves. The next chapter explains how this is possible.
[1] By imagination I mean simply the ability to creatively construct and manipulate images. The Kantian notion of imagination is related to the attempt to solve problems peculiar to Kant's system. The normal uses of the word make it an ability we have in virtue of our concepts, not a mediator between concepts and sensation.
[2] Hume actually arrived at a theory of abstract ideas similar to this. Abstract ideas were seen as using concrete particular ideas, but having their content or meaning consist in a habit or custom of thought that could produce other particular images at will. See the Treatise, Book I, Part I, Section 7; especially pp. 20-21. (Hume 1928)
[4] It is actually tripled; we must count the sensible manifold as a level of representation as well.
[5] It should be noted that this claim, like most of the description of the working of concepts in this chapter and throughout this work, is an empirically testable hypothesis about how neurologically embodied connections between modes of interaction operate as coordinated wholes.
[6] Although I became aware of this work too late for it to be incorporated into the content of this chapter, there has been recent work on neurological and mathematical models of cognitive activity that seems to correspond closely to the models presented here. Paul and Patricia Churchland have presented some recent work in neurology that takes neural sheets and the connections between them as the basic units of cognition. These neural sheets can be identified with what I have called modes of interaction and the connections can be identified with the dispositions that make up concepts. (See Paul Churchland 1986, 1987 and Patricia Churchland 1986, Chapter Ten). A group of researchers in cognitive science called the Parallel Distributed Processing Group (PDP) have provided mathematical models in which states of neural activation represented mathematically as vectors are transformed or connected to other states through weighted connections mathematically represented as matrices or tensor functions. The vectors correspond to the modes of interaction, the matrices to the dispositions or concepts. These models also agree with the view presented here in seeing the content of cognitive activities as resting in the connections made between the states of neural activation. (See Rumelhart 1986 and Patricia Churchland 1986, Chapter Ten).
[7] It may be possible in the future to give a physical account of what we now describe as the contents of consciousness and its operations. This would require giving an account of the necessary connection between the mental properties and the physical properties that allows one to see how they spring from a common causal nexus. At present I am convinced by Thomas Nagel's and Donald Davidson's arguments that it is not possible to give such an account in purely physical terms, at least as the physical is now understood. (See Nagel 1974, 1986 Chapter Three, and Davidson 1970.)
[8] This can mean simply that the dominant teleological tendency in our makeup at a particular time determines the activation of certain of our concepts, which then go on to determine the next stage in the process. How agency is possible without assuming a homunculus or transcendent governing agency will be discussed in Chapter Six.
[9] For convincing accounts of the prevalence and importance of this type of representing see Lakoff and Johnson 1980, Johnson 1987, and Lakoff 1987. Two things should be noted about this type of metaphorical elaboration of concepts: (1) I do not think it involves making new concepts, but simply the extension of existing ones into new domains by connecting them with new clusters and the referring abilities that define them. The concepts then act differently or metaphorically when exercised in their new context upon their new domain. (2) Metaphorical projection of concepts implies and requires reference that is based on basic referring abilities that are independent of the content of the concepts involved. Without the establishment of domains independent of conceptual content, metaphorical projection would be a wholly conceptual matter, just another way of connecting concepts. What makes it metaphorical is the connection of old concepts to new aspects of reality, not connection to new concepts.
[10] I use the word satisfaction to distinguish it from particular feelings. Satisfaction is not, as the utilitarians thought, a particular feeling that we can call pleasure. Pleasure or satisfaction is a way of feeling the particular phenomenological character of our interactions as an expression or culmination of a teleological tendency.
[11] This explains the necessity of consciousness for knowledge. For it is not just the fact that the desired connections are reinforced when I scratch the right spot. I must be conscious of this, that is, I must connect it with the tendencies to continue the reinforced activity. If I were not conscious of it, the right connections might be being reinforced like mad, while I go on happily trying to put my elbow in my ear. The directedness of the activity towards a goal and the guidance of that activity towards the goal by the subjective character of the process are necessary for knowledge to be achieved; this is what adding consciousness to the analysis adds.
[12] See Kant 1966 (A137, B176, p. 121) for his statement of the principle that a concept must be homogeneous or similar (gleichhartig) to any object it represents.
[13] Kant sometimes speaks of schemata as procedures for the construction of images instead of representations half-way between images and concepts. It is no more clear, however, how there can be a procedure for determinately applying concepts to particular images. It is also unclear how this characterization of schemata is coherent with Kant's general strategy in the Schematism. See the passage referred to in the above footnote for a fairly unambiguous indication that schemata are to serve their function by being homogeneous or similar to both concepts and images.