There has been much debate in the philosophy of mind about the nature of representation, its functional and causal role, how it is related to meaning, the importance of the syntax-semantics distinction, and so on. Much less has been written about the question of what is represented; indeed it is often taken for granted that this question is unproblematic. This essay will show just how problematic the question actually is.
It is often presupposed, in both cognitive science and the philosophy of mind generally, that cognition is the recovery by an animal or human of objective features of reality. "Reality" is understood to be an "in-itself," a realm of objects, properties, events and laws which are independent of any observer. (I call this position Realism.) The task of the cognitive system is to form true, or at least pragmatically useful, representations of the salient features of reality. (Representationalism)
I think this presupposition is wrong. Objects can only be individuated and classified within a system of categories. The categories which are relevant to understanding cognition must be those that the animal itself uses to constitute its world. Hence, "what is represented" is not reality, but the unities into which the experienced world is segmented by the organism. From this perspective the central issue is no longer representation, but how objects in the world come to be individuated and categorized in the first place.
Indeed, the issue is the nature of a "world."(1) My aim in this article is to use the notion of supervenience to establish a concept of world which is thoroughly naturalistic. I present the concept as an alternative, on the one hand, to Realism and Representationalism and their dilemmas, and to the transcendentalism of some phenomenologists on the other.
First, a note on terms. I will use "object" in a general way to include features and events as well as things. "Subject" will be used for an organism in its cognizing capacity, that is, for animal, human or other cognitive system. I will keep the term "world" for the structure within which subjects experience or perceive objects in opposition to "reality," which I will reserve for how things are "in-themselves." "Physical world" or "physical nature" will refer to the set of objects and relations hypothesized by current theories of physics and chemistry, particularly micro-physics. The words "constitute" and "transform" will also be used in special senses which will become clear as I proceed.
I will start by illustrating my thesis with an example which focuses
on colour vision, an area recently discussed by Thompson, Palacios and
Varela, and also by Hardin, Dennett and others. In Part III I will generalize
these results into a overall view of the relationship between a subject
and the objects in its world. Then I will explain the concept of supervenience
and use it to illuminate the notion of world. This will lead me to examine
the various dependencies, some causal, some non-causal, between subject,
object, world and micro-physical nature. In Part IV, I will discuss the
ontological status of objects in a world.
Let me start by analyzing a simple scenario:
But this cannot be right. Our physical sciences tell us that what is real is a causal network of atoms, sub-nuclear particles, electro-magnetic waves, mass-energy clusters or whatever. None of these things are represented by Alysha as she moves to watch the sunset, nor does the syntax of her representations mirror the laws of physics. The objects Alysha is responding to are objects in her world, the objects towards which her visual system and bodily responses are oriented.
Consider, for instance, the blue sky Alysha perceives above her and, in the west, the red sunset and green forest. Alysha does not move her chair to see the astronomical event of the sun crossing the horizon, an event which occurs every day and can be seen every clear day. She didn't sit out last night because as we say, "there was no sunset to see." We don't mean, of course, that the sun didn't set, we mean there was no red glow in the west. Her seeing the red glow against the blue sky is what has motivated her action. It is the red colour which differentiates the sunset in the west from the rest of the sky, although the boundaries may be vague.
Yet the redness and blueness are not elements of reality or of the physical world; they are categories of Alysha's visual system, and probably the visual systems of some other species. "Sunsets are red" means that they have something, namely their colour, in common with other red things, e.g., berries, rainbows, and hot objects. There is, however, no physical property common to these things which "redness" signals: berries are red due to their surface reflectance; the red in rainbows is caused by dispersive refraction; red-hot objects are red because of incandescence. The physical processes which cause the set of light frequencies to be radiated are entirely different in each of these cases. But even if the causes are different, is not the wavelength of the light emitted the same for all red things? No, even that is not true, since any number of combinations of frequencies may lead to the experience of redness. The only feature red things have in common is the capacity to provoke a positive response in the red-green opponent system of our visual systems. While light with a wavelength of 650nm will produce such a response, so will a vast number of combinations of other wavelengths ("metamerism"). The capacity to see things as varied as blood, rubies and lasers as all having the "same" colour is to be found in primates and a number of other species, but not in many others. Excluded, for instance, are most other mammals, who are monochromats, and pigeons, who can see many more, and totally different colours, than we do.(2)
What then could it mean to say that Alysha has a representation of a red sunset? It cannot mean that some process in Alysha's brain (or mind) corresponds to an objective property of reality, for redness does not exist "objectively," that is, without reference to particular kinds of visual systems. Thompson, Palacios and Varela criticize this objectivist approach to colour vision:
The typical emphasis in computational vision [is] on optimally 'recovering' prespecified features of the environment (i.e., distal properties whose specification is thought to be independent of the sensory-motor capacities of the animal.)(3)
By comparing the world as it must be perceived by different visual systems, Thompson, Palacios and Varela present excellent and detailed arguments to show that objectivism with respect to colour must be rejected. What is the alternative? Since there are no extradermal coloured objects, the most obvious alternative is subjectivism. C.L. Hardin, for instance, in Color for Philosophers, identifies colours with chromatic properties of neural states. Thompson, Palacios and Varela, however, are no happier with subjectivism than with objectivism and reject this forced dichotomy which Hardin seems to assume without question. They propose instead an "enactive" analysis based on the evolutionary interaction between each species and its environment. This ecological approach claims that there are indeed colours in the world, or at least "context-dependent and interest-relative structural analogues of color,"(4) but that these structures have evolved through a process of co-determination between perceiving animals and their environments. Animals select properties of the world relevant to their capacities while the environment selects for certain corresponding capacities in the animal.
For instance, the evolution of ultra-violet perception in the visual system of the bee has occurred in tandem with the evolution of flowers having ultraviolet pigmentation, since the flowers need the bees for pollination as much as the bees need the flowers for honey. Most colours in the natural world are organic colours and have evolved in correlation with the visual capacities of organisms. The enactive theory claims, then, that colours are extradermal (against Hardin's subjectivism) but nonetheless are not independent of the existence of perceivers (against objectivism).
I think this position adopted by Thompson, Palacios and Varela is a
major advance for our understanding of the problem, although I have some
reservations to offer later. I want to go further, however, and claim that
colour is but one illustration of a more general structural interrelationship
between subjects and their objects in general. This generalization will
lead to the concept of world.
1. Generalization from colour to all perceived objects
I have examined in some detail the nature of colour because I think it illustrates very clearly the "prejudice of reality," as we might call it, that is, the presupposition that there is a reality-in-itself, whose structure is independent of the subject and which a subject represents more or less successfully, i.e., truly or falsely. The prejudice is clear and explicit in objectivism. Subjectivists like Hardin however also presuppose objective reality; they just claim that colours are not among its features, hence they must exist only in the subject.
It is not only in the investigation of colour that we find this presupposition. David Marr, in his very influential theory of vision, holds that the function of the visual system is to recover from retinal stimulations information about how things are objectively. One essential step in the perception of objects is the use of computation to recover "edges." The visual system detects rapid spatial variations in the matrix of luminescence values reported by the retina. The detections are performed on a number of scales, that is, at a number of spatial frequencies, and the results are superimposed. Where the results are similar at all scales, the system reports an edge which can be assumed to exist in the real world beyond the retina. These edges, together with various constraints, allow the visual system to reconstruct real surfaces and objects.(5)
In this analysis, Marr assumes that unlike retinal variations, edges are real properties definable by purely physical parameters. It seems to me that Marr here neglects an essential reference to the subject. He claims that the function of the vision system is to reveal to us the objects and features of the visible world. "Visible world," however, cannot mean physical world in the sense of the world of physics. When he speaks of edges, he is clearly thinking about the edges of tables or chairs. But what about molecules? Do they too not have edges? And the galaxies? Why are their edges not discussed? Marr is presuming that an "edge" is defined only on a certain scale. This scale is relevant to, indeed defined by, the size of the animal, its capabilities and its interests. Edges are defined by the organism; they are properties of the world it experiences and lives in, not of some reality-in-itself.
The same can be said for "surfaces." The many surfaces that make up a human face, and their variations in reflectance, are crucially important to human beings for guiding our social interaction. It is highly unlikely the visual system of pigeons would segment the world in a similar way. On the other hand what we see as the more or less uniform grey surface of a pigeon's wing may be broken into intricate patterns of differently coloured surfaces to other pigeons' pentachromatic visual systems, patterns that pigeons probably use to identify individuals. Surfaces are aspects of visual worlds, not of objective reality. As Thompson, Palacios and Varela put it, "although the reflectance at any point in the scene can be specified in physical terms, what counts as a surface may in fact involve tacit reference to a type of perceiver."(6)
I have examined colours, edges and surfaces and argued that they are dependent for their definition on perceivers, or subjects. But visual objects are defined by colours, edges and surfaces, so we can now conclude that perceived objects are not prespecified parts of reality, but unities whose very definition depends on the perceptual, and hence behavioural, capacities of perceivers. Not only the sunset, but the chair, the house and the movement are undefined without reference to Alysha or organisms like her.
I can now offer a simple answer to my initial question about what the
brain or the mind represents: if it represents anything at all, it represents
objects in the world. These objects, however, are not the unities of physical
theory, atoms, gravity, electro-magnetic radiation or whatever, but unities
defined by the capacities of subjects.
How does this appeal to objects in the world enable us to understand behaviour? Let us look at a simpler and more artificial example than the case of Alysha and the sunset.
Consider how a psychologist might explain my action, during a game of chess, of moving a knight to protect my king from your queen. My move takes place in a chess world; the objects in it are defined and individuated by an interlocking set of chess categories. As Wittgenstein puts it, to label something as the king, we must first understand the basics of the game.(7) (Indeed, worlds are very similar to Wittgenstein's language games.) Once she understands what chess is, and only then, the psychologist can see why the position of your queen is a threat to me. Then, given my beliefs about what will protect my king, she can explain my moving the knight.
The psychologist's explanation only makes sense if she accepts that there are chess pieces and moves in the world and that these are capable of acting as causes of my behaviour. This explanation makes no appeal to the physical processes which make up the game. Since it is irrelevant whether the chess pieces are made of wood, marble, ink on paper or pixels on a computer screen, the actual physical components which compose the pieces have no place in the psychological explanation. On the other hand, the pieces are not in my head; they are on the chess-board.
In the psychologist's explanation, we need to distinguish two phases. First she needs to understand the constitution of my world of chess, that is, how I have come to be living in this world as opposed to another, why I have adopted these categories. She presupposes this prior constitution when she identifies the queen as the cause of my action. A representationalist might want to say that the cause of my behaviour is not the queen as such, but my internal representation of it. I have reservations about this way of putting it, but note that even if we accept it for the sake of argument, what is alleged to be represented is a "queen," not a piece of wood or the pile of molecules that makes it up. I am suggesting rather that the cause of the behaviour is an external object, the queen (or the event of moving it.) External here does not mean physical: queens do not exist in the physical world, but only in the world of chess. Of course, such a world is dependent on human beings: only in so far as I have a neural system capable of responding to objects in the chess world, a capacity that dogs, for instance, lack, is it possible for queens to motivate my action.
A representationalist, such as Fodor, might say that the only way I can have a neural system with this power is if there are processes whose causal-syntactic structure models the semantic content "queen." He might be right; I suspect this is an empirical question. Given the new story about parallel distributed processing, this theory seems less and less likely. The psychologist's explanation, however, doesn't depend on the outcome of this issue, but only the presence of some neural structures which enable me to constitute and live in a chess world. There is a queen for me in this external world if there are some circuits, any circuits (maybe different from those that other people have) which enable me to respond to the unity "queen," and which would respond in the same way whether the queen were made of marble, wood or whatever.
Understanding the chess world and the way I live in it, grasping how
I define the pieces and moves, and seeing how this world constitutes its
objects and my ability to perceive them, are necessary prerequisites for
the psychologist to explain my behaviour.
3. Worlds depend on supervenience
With this simplified example behind us, I am now ready to propose a general approach to the relationship between worlds, subjects and objects based on the notion of supervenience. My central thesis is that a world is a brain-object structure which supervenes on micro-physical processes and within which supervenient causal regularities are to be found. A world is defined by reference to a species of subject or organism which is part of that world and whose perceptual and behavioural responses define the supervenient unities or objects which are to be found in that world. Each world has evolved from a previous world by an internal transformation.
In unpacking this position, I will start by briefly explaining the notion
of supervenience and then showing how it can be applied to illuminate the
nature of worlds. Then I will look at the intricate interrelationships
between the causal, non-causal and supervenient dependencies which the
analysis requires.
Supervenience is a materialist notion developed by Horgan, Kim and others which presupposes as fundamental the existence of micro-physical laws.(8) Everything that exists is determined by these laws and so, in principle, if we understood these laws and the initial conditions to which they were applied, we could understand and predict anything that happens. In practice, of course, this is seldom if ever possible on the macro-level. Predicting where a leaf will fall is beyond the competence of anything but a divine physicist. It is not that the laws are too complex; the problem lies in the complexity of the initial conditions needed to apply each of the relevant laws.
We can, however, discover higher level regularities and causal laws which depend on these fundamental laws and their constraining initial conditions, even if we seldom know exactly how the dependency works in detail. We can understand that a thermostat regularly causes the furnace to come on even if we are incapable of calculating the trajectories of all the micro-physical particles in its mechanism. The Law of Thermostats, "Low temperature causes the furnace to start," is an example of a causal law which we know supervenes on micro-physical laws, even if we don't know exactly how in each token case. A further complication is that there are different designs of thermostats: some thermostats use bimetal strips, others use air-chambers, and some may use solid-state devices. Indeed there is an indefinitely large number of ways to construct thermostats, all of which would obey the same causal Law of Thermostats, and all of which are physically deterministic. Hence the Law of Thermostats is not, in its general form, reducible to the laws of micro-physics, although it certainly supervenes on them. It is the initial or boundary conditions, within which the laws operate which result in the Law of Thermostats and these constraints are not themselves explicable by the micro-physical laws alone.
The Principle of Supervenience is most commonly used to describe the relationship between physical processes and the mental properties which are supposed to supervene on them. The notion is intended to preserve the principle of physical determinism while allowing that mental properties may have multiple physical realizations and are therefore in some sense autonomous. In its weakest form, the principle claims that there can be no differences on the supervenient, mental level of the world without some difference on the micro-physical level.
There is, however, no reason to restrict the notion to mental properties.
Kim, a proponent of supervenient causation, has applied the principle to
the understanding of phenomenal objects like tables.(9)
Let me show how this would work by applying the notion to my favourite
example, colour. The colour of a particular flower to a bee can certainly
be explained in principle by the micro-physical laws which determine the
reflectance of the pigment in that flower. If the bee were to see something
different, then something about these laws would have had to change. In
the case of another species of flower, however, the same colour may be
explained by reference to entirely other laws governing the constituents
of a different pigment. In other words, the causal regularities to be found
in the interaction between bees and ultra-violet flowers may be realized
by multiple micro-physical causal processes and laws. Further, we can say,
counterfactually, that if some new pigment were developed which stimulated
the bee's visual system in the same way, it too would give rise to the
same causal regularities, regardless of which micro-physical process were
involved. Every token interactive event can be explained by micro-physical
laws, but it doesn't follow that the general causal laws governing the
colour-bee regularities, and which would explain enactive co-evolution,
can be reduced to micro-physical laws.
5. Supervenience applied to worlds
Let me now apply the notion of supervenience to illuminate the concept of world. There are a very large number of causal systems which could supervene on micro-physical laws. The evolutionary emergence of a world may be seen as the selection of one of these and its development as a brain-object supervenient causal system. On the side of the species, perceptual and motor capacities must emerge which can respond to extradermal supervenient causal unities. On the other side, these unities must be valid supervenient causes which are governed by laws. If unities were selected which were not regular in the requisite way, then this evolutionary step would fail; the species would not survive. The organism's perceptual system must latch onto causal unities which are in fact supported by micro-physical laws. When a new species does survive then a new supervenient system, a new pattern of causal regularities has emerged. It is this new pattern that I am calling a world.
To call something an object in an organism's world is to classify it by means of categories particular to this world. The categories serve to individuate the unities in space and in time and to assign them to a general kind.
I'll use Alysha's sunset again as an illustration. The sunset causes Alysha's visual system to recognize a sunset. This causation supervenes on the micro-physical processes of electron vibration, emissions of photons of various energies, rearrangements of visual purple molecules, and so on. Her visual system constrains the physical processes in a number of ways. For instance, only certain frequencies of light are "permitted" to have effects on her neural system; ultra-violet is screened out, for example. Only those combinations of frequencies which positively fire her red-green opponency circuits are permitted to effect her sunset recognizing system. Hence one area of the sky is granted the power to act as a single coordinated cause which results in her recognition of the sunset as its effect. This area is thereby spatially individuated, differentiated from its background, since the rest of the sky is not able to cause this same effect. Generalization occurs in a similar way: since this power is not momentary, but is built into her brain, these clusters of micro-physical events would produce the "same" effect another day. Further, any cluster of micro-physical events which fire her opponency circuit in this way will cause her to perceive a sunset. We have then a system of unified causes and effects within the brain-sky structure, the subject-object structure, which supervenes on micro-physical laws. This is the structure of Alysha's world, or at least part of it.
None of this explanation requires us to think of there being some pre-existing physical object in the sky which some organisms secondarily categorize as a sunset. From the viewpoint of micro-physics, electromagnetic radiation of various frequencies and intensities is coming from all areas of the sky all the time. There is no physical criterion for selecting one part of this radiation or one period of the day and segmenting it out as a not-yet-labelled-sunset. Nor is there any physical reason to associate what is happening in the sky tonight with what happened last Wednesday, when there was a sunset, as opposed to last Friday, when there was not. The only criteria for individuation in space or in time, or for assignment to a kind, are perceptual ones: Alysha and the rest of the human species has the capacity to respond in a standard manner to a hotchpotch of events which they define as a sunset. The boundaries of this new entity may be "arbitrarily (perhaps infinitely) complex and bizarre," as Haugeland puts it in a similar case.(10) Haugeland, writing of decisions rather than sunsets, but the point is the same, says:
The point is not that the "boundaries" of decisions and actions are
micro-physically fuzzy, indescribable, or even negotiable; rather there
are no such boundaries at all: decisions are not distinguishable (even
in principle) as separate individuals at the level of micro-physics.
(11)
If this scheme is correct, then an organism's percepts are caused by objects in the world. However, it is also true that
objects depend on subjects while both make sense only in a world which depends superveniently on micro-physics. There are at least five kinds of dependencies here which must be clearly distinguished from each other. (I take the opportunity to introduce some technical meanings of terms.)
2. A red sunset depends for its definition on the visual system of human subjects, or organisms with similar visual systems. Let me call this subjective dependence. Note that the fact that there is such a dependence is completely objective.
3. Both the perceptual system of an organism and the objects which are perceivable by it depend on a supervenient causal structure. It is this structure that I am calling a world. Sunsets cause visual and behavoural effects in humans because both humans and sunsets are constituted by the human world on which they depend.
4. The existence of a world depends on an historical process by which earlier worlds were transformed into this world in accordance with biological, evolutionary principles. Alysha's seeing of a sunset depends on a transformation of a monochromatic world defined by some ancestor of primates which led to a world in which both red-seeing and red sunsets could occur.
5. Finally, all of the above depend on the micro-physical processes upon which they supervene. This is supervenient dependence.
6. Interrelationships between the five dependencies
The most likely confusion is between subjective or constitutional dependence (#2 or #3) and causation (#1). If one were to hold that cognition involved setting up representatives of reality in the head, then the sunset, like other perceptual objects, would be an event in the head and it would be reasonable to propose the brain as its cause; the brain would construct the sunset within itself. My position, on the other hand, is that the sunset is a distal event occurring in the world, and while it is dependent on the subject and on the world, these are not causal dependencies.
Now there are certainly causal processes in vision. Light from the sky in the west causes retinal neurons to cause certain firing patterns in the optical lobes of the brain. But there is no causal process that originates in the brain and which has as its effect the redness of the sunset in the world. The subject is neither like a painter colouring an internal, mental canvas, nor like the arsonist whose smoke colours the sky. Subjects are not like movie projectors which cause the sunset to appear on the screen of the heavens. Nothing is projected by the subject into the world. So when I hold that the feature "red" in its world depends on the subject, I am not claiming any direct causal dependence. I am simply saying that redness can only be defined by reference to a certain kind of organism. Definitional dependence is not causal.
Similarly, to say that the sunset, and its perception, depend on the
human or primate world, does not imply that the world is an object which
causes the sunset as its effect. In "the cat is on the mat" the agreement
in number of the verb "is" with the noun "cat" depends on the structure
of English, but this does not mean that the English language causes
this agreement. A language is not an object like a cat, mat or speaker
which could act as a cause of a sentence. Similarly, a world is a supervenient
structure within which causal events can occur; it is not itself an object
susceptible of acting as a cause on its own objects or subjects. Constitutional
dependence is not causal. (This was Aristotle's argument against the causal
power of Forms in Plato.)
7. The enactive approach to transformation
Even if we grant that neither subjective dependence nor constitution are causal relations, is there not a causal story to tell nonetheless? Is not the transformation of worlds (i.e., #4 above) a causal process that can be explained by physical science? I interpret Thompson, Palacios and Varela as doing exactly this with their "enactive" theory, to which I referred above. In their attempt to account for colours, they too reject any purely neural-causal explanation. For them, colours are properties of the extradermal world because they have evolved in tandem with the development of organisms whose perceptual systems are responsive to these colours. This is an evolutionary-causal process in which "perceiving animals as both assemblies of sensory-motor networks and as organismic unities ... shape the extradermal world into an environment in their interactions."(12) They mean, of course, causal interactions. They point out that most colours in the world are to be found not in the inorganic world, but on organisms which have evolved these colours precisely because they can be perceived by animals.(13) The enactive theory might claim (this is my example, not theirs) that the red-green opponency system in the vision of certain organisms gave them an evolutionary advantage because it enabled them to easily spot red berries against a green background. The other half of the story is that berries evolved their redness because it promoted their reproduction, given the development of red vision. Note one thing that the enactive theory does not say: although the theory claims that there may be causes for why our vision sees some things as red, and these causes are related to the causes of the things being red, it does not claim that our visual system causes them to be red. With this I am in complete agreement, as my argument above indicates.
In fact, I think that enactionism in general is true as far as it goes and is an important advance in our understanding of constitution, indeed, in our understanding of dependencies #1, 2 and 3. But I don't think it goes far enough; as it stands it has a number of inadequacies. For one thing, it oversimplifies the relationship between perceivers and perceived. The colour of flowers, for instance, may have co-evolved with the vision system of insects, but primates too have come to see flowers as coloured, probably much later. It is unlikely that anything in primate evolution has co-determined the pigmentation of flowers and yet they are coloured objects in primate worlds. And even the most exaggerated version of the Gaia hypothesis could hardly maintain that red sunsets have co-evolved with human visual systems.
Since it is my contention that not only colours, but all objects and features of any world are constituted, the enactive, interpretation of transformation cannot be sufficient. The edges of tree-trunks or of mountains, the reflecting surfaces of leaves or of pigeon feathers, and the unity of dogs or sunsets, all of them features of the human world, cannot be accounted for by any co-determination by evolving human visual systems. I think they can be accounted for by appealing to the kind of supervenient categorization I am proposing.
Further, I think that enactionism, even where it is successful, still presupposes supervenient worlds. Thompson, Palacios and Varela make it clear that the features of a perceived world cannot be given a purely physical-level specification.(14) Hence the kind of causation involved in the evolutionary enactive process they describe cannot be purely physical causation either. That is, the objects or events that play the roles of either causes or of effects in the co-evolutionary process are not defined by physical theory, but only by the perceived worlds within which they are categorized. For instance, the colour of flowers which they claim has co-evolved with insects by circular causation cannot be defined by the chemistry of the pigments. Presumably different flowers have achieved their colour by means of different pigments. What have all these pigments in common? Only that the visual system of certain insects responds to them in the same way. They cannot be defined as in-themselves, but only as for-insects. There are no causal laws in physics or chemistry which would take these kinds of unities as causes or effects.
That is, the enactive explanation already presupposes the categorization of the world into supervenient objects before the co-evolution explanation can work.
This is not to say that we should reject a causal explanation of the
development of or transformation of worlds, that is, of dependency #4.
I am only claiming that the causation involved is supervenient, not micro-physical.
We can explain the transformation of one world into another in terms of
supervenient causation alone, without appealing to physical laws.
8. The transformation of worlds
How, for instance, did a world with sunsets in it come about? A just-so story of how Alysha's world came into being would take us back into a hypothetical evolutionary history of the primates. Some ancestor living in a monochromatic world underwent a neural mutation which permitted some or all of the things we now call red to stand out and be recognized, that is, which gave red things the power to cause some specific neural events. This had some adaptive advantage, maybe facilitating berry-picking, and so became selected for subsequent generations. Hence the monochromatic world transformed itself into a coloured world.
This historical account makes no call on micro-physical principles to
explain what happens, although of course it must supervene on these principles.
We understand how a new world comes into existence by seeing how one world
transforms itself into another, sometimes enactively.
9. A world is a condensed history
So Alysha's world is the world of her species; she has inherited it from her ancestors. The genetic history of her visual system and probably the conceptual history of her culture are responsible for sunsets being part of Alysha's world. It makes little difference for my analysis whether this inheritance is genetic or cultural. Probably the capacity to see red is genetic in normal human beings whereas the categorizing of these reds and the associated phenomena into a "sunset," an object of note and admiration, is cultural, but it would be of little consequence for my position if it turned out that colour classifications are cultural or that "beautiful sunset" is an innate concept. The crucial point is that an object is what it is for a subject because of the past. The history of the species or culture is the original source of the constitution; the individual, for the most part, just borrows a ready-made world to live in.
In a sense, a world is a condensed history, and this is as true for the objects in it as for the subjects. I think it would be a mistake to emphasize exclusively the evolution of cognitive capacities in the subject, as I think representationalism does, and ignore the evolution of objects.
Representationalism emphasizes a priori knowledge because it conceives of the units of evolution as isolated individuals and places the burden of history solely on the internal genetic or memory processes of each phenotype. It tends to say that primates know the difference between red and green, they know that red berries are good to eat, they have an innate concept of sunset and so on. It thinks of each individual as having innate knowledge, knowledge that was figured out millions of years ago by some primitive visual system and passed on in genetic code to the current generation.
I think this is wrong on two counts. First it interprets the influence of the past as invariably resulting in knowledge where I think it results in capacities whose origin is usually unknown to the organism. An organism is the result of eons of adaptive development but this does not mean that the organism knows this history, or that it knows why it sees the world as it does. Knowledge here is not the point (though it is not excluded). The present way that the world is constituted is the evolutionary result of facts in the past. An organism does not see red because it knows or racially remembers that red berries contain vitamin C, nor because it has an innate concept of red. Rather, the capacity to see red is a causal result of the fact that its ancestors adapted well by absorbing vitamin C from berries. The visual system doesn't know anything about the past; it is not an accumulator of facts or of any knowledge. It doesn't remember facts; it is a product of these facts. It is part of an evolved world in which the visual capacities are the result of historical success and failure.
Secondly, I think the role of the subject is overemphasized here. I would prefer to say that not only the subjects but also the objects which are in the world have been historically constituted and so these objects encapsulate both certain survival values for the species and a set of supervenient physical regularities. The objects which are not found in a world may be as important as those which are: their absence is a result of past attempts by organisms to define objects in ways which were either superveniently impossible or which were antagonistic to survival. This historical wisdom is stored as much in the objects as in the subject's cognitive capacities. As Dreyfus puts it:
When we are at home in the world, the meaningful objects embedded in
their context of references among which we live are not a model of the
world stored in our mind or brain; they are the world itself. ...
My memories are stored in the familiar look of a chair or the threatening
air of a street corner where I was once hurt. My plans and fears are already
built into my experience of some objects as attractive and others as to
be avoided. ... The whole I/O model makes no sense here. There is no reason
to suppose that the human world can be analyzed into independent elements,
and even if it could, one would not know whether to consider these elements
the input or the output of the human mind.(15)
10. A shared world defined by individuals
Individuals who share a history thereby also share a world. I don't just mean that they can see similar objects, generically identical; I mean that an individual can perceive and interact with an object which is the self-same, numerically identical object another individual can perceive. If individuals were motivated only by their own internal representations, each person would respond to her own private model, a different token for each individual. An object in a world, however, is external to the perceiver, it is a causal unit, and so is accessible to anyone with perceptual capacities which match its structure. Hence it is possible for John, if his visual and conceptual systems are the same as Alysha's, to respond to the self-same sunset. They live in the numerically same (i.e., token-identical) world.
Public as it may be, the world still depends on and is maintained by its subjects. That is, while both objects and subjects are constituted by their world, the relationship between them is not symmetrical. Objects do not define subjects. If all the supervenient objects in the human world were to vanish tomorrow, perhaps by an Evil Demon changing the laws of micro-physics, human subjects could in principle still continue to exist for a while, though in a very confused, and soon, hungry state. Their eventual extinction would be causally explained. If human subjects vanished tomorrow, however, sunsets, chess pieces and other objects exclusive to human worlds would no longer even be possible objects. They would not have been causally destroyed. Objects depend on subjects, or the capacities of subjects, in a way that subjects do not depend on objects. This is the dependence I have been calling "definitional." (#2 above)
It is the capacities that each individual has which allow it to live
in its world. A world could not be superimposed on an individual without
these capacities: there are some things everyone must do for themselves;
having a world is one of them. Nor should we think of worlds as theoretical
entities invented by philosophers or scientists in the third-person mode
to give a coherent account of the behaviour of those they are studying.
They can serve this secondary, explanatory purpose only in so far as they
are available to the subject herself in the first place. It is because
Alysha sees the red sunset that she sets up the chair to watch it. If the
sunset were only a theoretical concept available to those adopting a particular
psychological theory of behaviour, then it could affect Alysha's behaviour
only if she were herself such a psychologist. If Alysha herself doesn't
recognize sunsets, if sunsets are not objects of her world, then this kind
of explanation of her behaviour is void. The objects in a world are for
an individual organism. In one sense, then, the responsibility for there
being a sunset in her world is Alysha's.
I maintain that the five kinds of dependencies mentioned above can be
accounted for by analyzing worlds in terms of supervenience. All evolved
structures supervene on the micro-physical. However, with the possible
exception of the historically first world, each world structure is best
understood by seeing how a previous world transformed itself into this
one on the basis of its own internal principles, rather than by appeal
to micro-physical laws. Each world is a supervenience causal structure,
that is, it constitutes objects as unified causes and it constitutes the
correlated cognitive capacities of the subject. When a particular worldly
object causes an effect in the brain of a subject we have a token case
of perception.
If the scheme I have outlined above is generally correct, we now need
to ask about the ontological status of objects in the world. Are worldly
objects subjective duplicates of reality? Are they illusions of our brains?
If not, are they "real," and what could that mean?
1. Are perceived objects duplicates of physical objects?
A recurrent theme in the philosophy of mind for three and a half centuries is that the mind is in contact only with its own products, its own mental objects or structures. Ideas, impressions, sense-data, representations have often been presented as models of the extradermal world, surrogates inside the head which act as representatives or delegates of physical objects. Fodor's Representational Theory of Mind is a contemporary example in which the mind is understood to compute over mental representations.
To believe that such and such is to have a mental symbol that means such and such tokened in you head in a certain way; it's to have a token 'in your belief box.' ... Mental processes are causal sequences of tokenings of mental representations.(16)
There is a functional isomorphism between psychological structures in the brain and the extradermal features which they model. In the Representational Theory of Mind objects get duplicated inside the organism.
But what objects does Fodor believe are mentally represented? Fodor refers to horses and chess-boards. Yet a horse is not a physical object in my technical sense of an object of physics; a horse cannot be defined in micro-physical terms. Nor is a horse a piece of reality-in-itself: the segmentation of the world into horses involves tacit reference to perceptual systems like those of humans (and, undoubtedly, of horses).
Anyone who, like myself, proposes that an organism lives in its "own" world is bound to be interpreted by many as holding that what a subject actually experiences are only internal duplicates or representatives of objects. But in my case this would be a misinterpretation. The objects in an organism's world are not mental models which correspond, or fail to correspond, to equivalent objects in the physical world. They are not representatives; they are distal unities in the world, but unities which are definable only by reference to a subject. The sunset is not in Alysha's head; it is in the sky, over to the west. A sunset is public, not private.
But this does not mean that it is an object of physics, for its criteria
for individuation and generalization are not those of any physical theory.
A sunset is only such for a particular kind of perceiver, one capable of
three-dimensional vision, of colour vision, and whatever else is needed.
That this pattern is relative to an observer does not mean that it only
occurs within that observer. It's place is in the world not the head. The
matter it is made up of is physical. It differs from the objects of physical
theory only in being categorized differently. To say that the world constitutes
objects is not to say that it creates duplicate models or mental representatives
of objects, but to say that it defines these objects in the first place.
2. Are objects in the world real or illusory?
Are these objects in the world then illusory? After all, I am claiming that they are dependent on perceivers, and what else can "real" mean but "independent of the observer"? It should give us pause in this line of thinking to remember that those very epitomes of "reality," the objects of quantum theory or the attributes of events in Einsteinian space-time, are not independent of the observer.
Let us approach the problem from the other end. What is not real is illusory; it only "seems" to be the case. Dennett has made much of this way of putting things, contrasting how things "seem" to consciousness in its "heterophenomenological world" with how they are in reality.(17) But, as Husserl long since pointed out, expressions such as "to seem" or "to appear" have at least two meanings. In one sense, sense A, the terms are used to distinguish something which is true from that which is false and only "seems" to be the case, as in "He appeared honest, but was really a crook." It is in another sense, sense B, that we use them to refer to how things appear to us in consciousness, or, by analogy, how things are given to any organism. "The sunset appeared to Alysha as a huge redness covering the western sky" does not imply that what Alysha is seeing is false, just that her perspective cannot be ignored.
Dennett seems to play (or really plays?) on these two meanings of the term. He uses "real" to refer exclusively to processes in so far as they are categorized by physics, and contrasts them, as he should, with the way events are categorized by the experiencer, how they "seem" to consciousness. This is a legitimate use of sense B. However, he slides illegitimately into sense A when he then declares many of these appearances to consciousness to be illusory.
For example, if I say, "It seems like the temperature is 20 degrees
but really it's 25" I am using the term in sense A. But if I say, "It seems
like the temperature is 20 degrees, but really it's just that the average
kinetic energy of molecules in my environment is so-and-so" I am using
"seems" in sense B. If I'm stoned, a mouse may seem to me, falsely (i.e.,
sense A), to be an elephant. If my neurons are operating normally, then
a certain pile of molecules may seem (sense B) to me to be a mouse. To
claim that in the second case a mouse is a fiction is to assimilate the
case of the mouse to the case of the elephant, to confuse sense B with
sense A.
3. Ontological status of worlds and objects
So what is real? The term has too many senses to be very useful. At the beginning of this article I offered a definition of reality as that which is independent of the observer. This definition, however, neglects the fact that many authors, such as Dennett, use the term in an honorific way to refer to whatever they consider most significant in their scheme and to use "unreal," "illusory," and similar terms pejoratively, as put-downs for that which they consider secondary. This hierarchical evaluation originally implied a Divine perspective: how things "really" are meant how they were for God. I suspect an implicit appeal to such a hierarchy is behind many contemporary uses of the term even if no appeal is made to God.
I think debates about what we should honour with the title "real" are largely sterile. Let us just distinguish the mode of being of objects categorized by physical theories from those objects of an organism's world which are defined by that organism, or, as I put it, which the world "constitutes." As long as we are clear that constituted objects are not thereby illusory, but are also not the objects of physics, how we attribute the honorific term "real" is irrelevant. (In "Real Patterns," Dennett drops the pejorative word "illusion" and adopts the term "semi-real" for patterns which are correlated with an observer as opposed to the "real" objects of physics. That makes him, he says, a "semi-realist" or a "mild realist."(18)) Positively what we need is a substantive discussion about the ontological status of objects. So we should not ask whether the objects of an organism's world are "real;" we should describe how they come about, how they are sustained in their being and what ways they are dependent, which is the approach that I've taken above.
Representationalists and Realists would restrict us to two ontological categories. Anything which is real is either an object definable by physical theory or it is a mental entity (which may also be physically definable). Anything else is illusion.
In my alternative position, objects in a world do not fall into either
of these restrictive categories of latter-day Cartesianism. Worldly objects
are not pieces of "reality-in-itself," i.e., independent of any organism,
subject, or world. They are constituted by a world and their definition
depends intrinsically on the capacities of an organism. But this does not
make them mental; they are extradermal causes which obey laws as objectively
as any other supervenient process, thermodynamics, for example. They are
neither illusions nor one of the two Cartesian-certified "realities."
I have made three points in this article. I have used supervenience
to offer a naturalized account of worlds and their objects. I have argued
that it is objects constituted in these worlds which act as causes on subjects
or organisms, not some reality-in-itself, or even the objects of micro-physics.
And I have hinted, though hardly demonstrated, that we may no longer need
the notion of representation once the companion concept of reality-in-itself
has been replaced by objects in a world.
1. I am making no claim to originality here. The concept of world has already been proposed by Husserl, Heidegger, Merleau-Ponty and many others. What I think is original is my attempt to naturalize the concept by appeal to supervenience.
2. All this is explained in C.L. Hardin, Color for Philosophers, Indianapolis: Hackett, 1988. See especially Part I, p 2 et seq.
3. Evan Thompson, Adrian Palacios and Francisco J. Varela, "Ways of coloring: Comparative color vision as a case study for cognitive science." Behavioral and Brain Sciences 15 (1992), p. 2.
4. Thompson, Palacios and Varela, p. 20.
5. David Marr, Vision, San Francisco: Freeman, 1982.
6. Thompson, Palacios and Varela, p. 17.
7. Ludwig Wittgenstein, Philosophical Investigations, Oxford: Blackwell, 1968. #31, p. 15e.
8. See, for instance, Jaegwon Kim, "Psychophysical Supervenience," Philosophical Studies 41 (1982) 51-70. The Southern Journal of Philosophy Supplemental Volume XXII 1984 is entirely devoted to the topic.
9. Jaegwon Kim, "Supervenience and Nomological Incommensurables," American Philosophical Quarterly 15 (1978) 149-156, p 155.
10. John Haugeland, "Ontological Supervenience," Southern Journal of Philosophy, Supplementary Volume XXII, 1984, 1-12, p. 2.
11. Haugeland, "Ontological Supervenience," p. 8.
12. Thompson, Palacios and Varela, p. 19.
13. Thompson, Palacios and Varela, p. 20.
14. Thompson, Palacios and Varela, p. 23.
15. Hubert Dreyfus, What Computers Can't Do, New York: Harper and Row, 1972, p. 265-266.
16. Jerry A. Fodor, Psychosemantics, Cambridge, MA: MIT Press (Bradford Book), 1987, p. 17.
17. Daniel C. Dennett, Consciousness Explained, Boston: Little, Brown and Company, 1991, p. 85.
18. Daniel C. Dennett, "Real Patterns," Journal of Philosophy, 88 (1991), 27-51, p. 31.