Fred Dretske


That perceptual beliefs are representational is not much disputed these days. More controversial is the idea that perceptual experiences are too. Even more controversial is the claim that perceptual experiences are not only representational, but that their phenomenal character--the qualities that determine what it is like to have the experience--are completely given by the properties the experience represents things to have. That is the thesis I mean to examine (and defend) here.

In speaking of perceptual experiences as representational one might only mean that these experiences are (normally) of things. They possess intentionality in the philosophers technical sense of that word. We see books, hear bells, and smell garlic. When veridical, then, visual, auditory, and olfactory experiences are of (or about) books, bells, and garlic. If this is all one means by a representational theory of experience, it is hard to see how to avoid a representational theory. I mean more than this. I mean that experienced qualities, the way things phenomenally seem to be (when, for instance, one sees or hallucinates an orange pumpkin), are--all of them--properties the experience represents things as having. Since the qualities objects are represented as having are qualities they sometimes--in fact (given a modicum of realism) qualities they usually--possess, the features that define what it is like to have an experience are properties that the objects we experience (not our experience of them) have. If qualia are understood (as I understand them) to be qualities that, in having an experience, one is consciously aware of--those qualities (therefore) that, from a first person point of view, distinguish one type of experience from another, then qualia are a subset of objective physical properties.

That is a mouthful so I'll break it up into more digestible chunks. First a few words about representation.


1. What are Representations?

In speaking of representations I am guided by familiar instances of (what I call) conventional representations--instruments, gauges, stories and pictures. Thermometers represent temperature, speedometers speed. Pictures and stories represent the objects and events they are pictures of and stories about. Some representations are pictorial, others are not. I call such representations conventional because their power to represent is, in one way or another, underwritten by our collective purposes and intentions. Change the way we use or regard these artifacts and you change (perhaps even eliminate) their meaning, what they say or represent about the rest of the world . If experiences are representational, they are presumably not conventional in this way. They are natural or original (Haugeland 1981) representations, representations whose power to say how things stand in some other part of the world is not derived from agents who (by having intentions and purposes) already possess this power. Aside from this difference, though, a representational theory holds that experiences resemble ordinary representations in important respects. If they did not, I would see little point in calling them representations. What follows is a brief catalog of these important respects.

There are representational vehicles--the object, event, or condition that represents-- and representational contents--the condition or situation the vehicle represents as being so. In speaking about representations, then, we must be clear whether we are talking about content or vehicle, about what is represented or the representation itself. It makes a big difference. In the case of mental representations, the vehicle (a belief or an experience) is in the head. Content--what is believed and experienced--is (typically) not. The same is true with ordinary, familiar representations. Is the story in the book? The story-vehicle is, but not the content. That is why we cannot infer that because there are dragons in the story and the story is in the book, that dragons are in the book. When I speak of representations I shall always mean representational vehicle. If I mean content, I will say so.

Representations need not be about a particular object (spatio-temporal particular) in order to have a content. Following fairly standard causal thinking, I take the object(s) of a representation to be the object(s) that stand in the right causal relation to it. If there is no object that stands in the right causal relation to R (the representational vehicle), then R is not about an object. R nonetheless still has a content--what it would be saying about an object if it had an object. Think of radar misrepresenting an aircraft approaching from the east. There is a slowly moving blip that--when things are working right--is the kind of blip an aircraft produces. In this case, though, no such aircraft exists. The radar still "says" there is. What it says about this non-existent aircraft if that it is approaching from the east. The pattern on the screen is about (or, if you don't like this language, the pattern says) something in the same way fictional stories are about (or say) things. It is about an approaching aircraft in the same way Shakespeare's play, Hamlet, is about an indecisive Danish prince.[1] Representations lacking an object have a content fixed by the ways they (mis)represent the world to be.[2] On a representational view of experience, the phenomenal character of our experience is determined not by the objects we experience--these can change or be non-existent while the experience remains, subjectively, exactly the same kind of experience--but by the way the experience represents things to be, the properties it represents objects (if there are any) to have In speaking of content, then, we must always be understood to mean attributive content, what the representation says or represents about a possibly non-existing object. As Davies (1996) puts it, the content of a perceptual representation is not object-involving.

It is important to distinguish the property an object is represented as having (call this M) and the property of representing an object to be M, call this Mr. If a perceptual experience represents an object to be moving, M, then, if the representation is veridical, it is the object that is M while the experience is Mr. If qualia are properties one is aware of in having perceptual experiences--the properties objects (if there are any) phenomenally seem to have--then it is M, movement, a property (if the experience is veridical) of the object experienced, not Mr, a property of the experience, that is the quale.[3] According to a representational view of perceptual experience, therefore, qualia are properties that physical objects, the ones we experience, normally have. They are not properties that experiences have. We are aware of movement, M, not Mr. I have no idea what the property of representing something to be moving, Mr, looks (smells, feels, sounds) like. I suspect it doesn't look, sound, smell, or feel like anything and for roughly the same reason that means dog (a property of the word "dog") does not look, sound, or feel like anything. Certainly not like a dog. Or the word "dog."

One last point before getting down to business. Some people think of all mental representation as essentially conceptual. To mentally represent x as M is to subsume x under the concept M, to believe, judge, or think that x is M. If a person does not know or understand what M is, does not have the concept M, then neither S nor anything in S (e.g., a perceptual experience) can represent something as M. You can't experience movement if you don't understand what movement is.

I have no quarrel with conceptual representations--I think of perceptual beliefs that way myself--but it seems clear that if experience is to be understood in representational terms, the representations cannot be conceptual. Seeing a clock hand move is different from believing that it is moving. You don't have to see it moving to believe it is moving, and you can see it moving without believing it is moving. Even birds, bees, and human infants experience movement, and they don't need the concept movement to do it. One no more needs the concept movement to experience movement than one needs numerical concepts to experience multiplicity. An experience of five objects is different from one of four (six, eight, etc.) objects, and it is different even for animals (including children) who do not know how to count and are, therefore, incapable of believing or judging that there are five objects before them. If one lacks the concept five, one will be unable to describe the five fingers one sees as looking like five fingers, but we--those who have the requisite concepts--can surely describe things in this way for those unable to do it for themselves. Perhaps they cannot describe themselves as hungry either, but we can. A representational theory of experience says that a visual experience can represent the fingers on one's hand as five (not as four or six) even when the experience occurs in a person (or animal) lacking the ability to say or think that this is how things are being experienced.[4]

I'll not try to say what the difference is between an experiential and a conceptual representation. I've tried to do that elsewhere. In Dretske (1981) I made the distinction in terms of an analog and a digital encoding of information. In Dretske (1995) the same idea was cast in terms of systemic (phylogenetic origin) and acquired (ontogenetic origin) forms of representation. But this isn't the place to develop or review these efforts. It is enough if we agree that a representational theory of experience must distinguish, in representational terms, between an experience of an object's properties--in the case of vision, its movement, color, orientation, shape, size, texture, and so on--and a judgment (belief, knowledge) that some object has those properties. If these are both representations, they are different kinds of representation, and a representational theory of the mind should, in the final analysis, be able to account for this difference. That, though, is a project for another time. For now we are interested in the question of just how plausible it is to think of experience exclusively in representational terms whatever the final analysis of representation turns out to be.

2. Experiences as Representations.

There is two--maybe (depending on how one counts) even three--compelling reasons for thinking of experiences in representational terms.

There is, first, the fact that the properties that individuate experiences, the ones that distinguish one type of experience from another--qualia, in short--are not (at least they need not be) qualities of anything in the head of the experiencer and, therefore, given that experiences occur in the head, not qualities of the experience. They need not, in fact, be qualities of anything. Nothing needs to have the properties we experience (at least not at the time we experience them). There need be nothing orange and pumpkin shaped --certainly not in the head at the time the experience is occurring--for us to have an experience as of an orange pumpkin. How is this possible? The same way it is possible to have stories (Cinderella, for instance) about (coaches turning back into) pumpkins without there being anything in the book (vehicle) resembling magical pumpkins. The properties and situations one is aware of in having an experience are just like the properties and situations described in stories: intentional. The world needn't contain them in order to be represented as containing them.

This intentional aspect of representation is evident in even such familiar measuring instruments as speedometers. Nothing need be going 100 mph for my (malfunctioning) speedometer to represent me as going that fast. Even when the representation is veridical, it (the representation) need not have the properties it says the vehicle has. Ordinarily, of course, a speedometer (located in the car whose speed it represents) has the same speed it represents the car as having, but the police have stationary devices that can represent a car as going 100 mph. According to a representational theory, experiences are like that. The representational vehicle, the thing in your head, doesn't (or needn't) have the properties it represents the world as having. The thing in your head that represents the object out there as moving needn't itself be moving. That is why looking in a person's (or bat's) head won't reveal the qualities being experienced by the person (bat) in whose head one looks. What I experience (see) when I look in another person's head are the representational vehicles--electrical-chemical events in gray, soggy, brain stuff; what the person experiences (sees), on the other hand, is representational content--a bright orange pumpkin.

I don't know of any other theory about the nature of sense experience that tells this satisfying a story about first vs. third person aspects of experience. If we agree that experiences of orange pumpkins exist in the brain of the person seeing the pumpkins, why can't other people tell what these experiences are like by looking in the brain, at the experiences, of the person having them? For the same reason I can't tell what story is being told by looking inside a book written in Chinese. It may, for all I can tell, be about coaches turning into pumpkins. I can see the representations clearly enough, but, with no understanding of the code (in this case, the language), I fail to understand their meaning (content). You can't identify representational content by looking at the objects (vehicles) that have it.

Since, as far as I can tell, a representational theory is the only materialistic theory that accounts for these otherwise puzzling facts about experience, only a representational theory successfully bridges the explanatory gap. A representational theory of experience doesn't, I admit, solve the "hard" problem of consciousness. It bridges this explanatory gap only be opening up an equally puzzling gap somewhere else: how do electrical and chemical events in gray soggy brain stuff manage to represent bright orange pumpkins? We all understand how marks on paper--viz., the words "bright orange pumpkin"--manage to do this. We give the symbols this power by making them mean bright orange pumpkin. We could, collectively, make these words mean something else. But no one, presumably, gave the events occurring in our brains their meaning. If they are representations, they are not, like words, conventional representations of the objects we see and hear. So where do brains get their power--this original or natural power--to represent (non-conceptually) the things we experience?

I have a view about this (see Dretske 1995), but this isn't the place to drag it out. The problem I am addressing here is not what gives brains their power to represent the qualities that constitute experience, but the question of whether a representational story gives a satisfying account of phenomenal experience. If it does, then we can turn to the next question: what, then, is a satisfying account of representation? But one thing at a time. Let's first be clear about whether there are aspects of experience, ways we see and feel the world, that cannot be interpreted as ways the world is represented? If there is, then even if experience is representational, it is not only representational. So even if we have an adequate theory of representation, one capable of marking the difference between conceptual (belief) and non-conceptual (experience) forms of representation, we will, in the end, need something else, something more, to obtain a complete theory of experience.

Before turning to the question of whether a representational theory provides a complete account of phenomenal experience, let me finish giving reasons for thinking it is, at least, part of the story. The second reason for favoring a representational analysis of experience may be nothing more than a restatement of the first reason, but I think it gives a sufficiently different slant on things to justify mentioning it. As we have already seen, qualia, on a representational view, are a subset of ordinary physical properties, the ones that objects are represented (in experience) as having. When the representations are veridical, when things actually are the way they look, objects actually have the properties they are represented as having. Square and moving, for instance, are two of the qualia of experiences (perceptions) of moving squares. These are properties of the objects we see--the moving square. They are not (or need not) be properties of our experience of moving squares. Your experiences of moving squares aren't moving squares.

This means that the properties we use to individuate experiences are the objective properties of the objects we experience, not the properties of the experiences themselves. We distinguish experiences not in terms of their properties, but in terms of the properties that their objects (if there are any) have. This sounds remarkably like a representational mode of classification, remarkably like the way we classify, say, stories and pictures. A is a biography of Oscar Wilde, B a history of the Spanish Civil War. We put these books on different library shelves not because they--the books--have such different properties, but because the things they describe are so very different. It is the same with pictures. C is a picture of a Turkish business tycoon, D a picture of the mayor of Chicago. These are different pictures, doubtless to end up in different photo albums or in different museums, not because they are so different (they could be indistinguishable; the mayor might be the tycoon's twin brother), but because what they represent, what they are pictures of, is so different.

Once we realize that the properties we use to classify subjective experiences are not the properties they have but the publicly accessible properties that external objects (the ones these experiences are normally of) have, it becomes--or so I think--irresistible to explain this unusual classificatory procedure in representational terms. Experiences differ because the objects they are (normally) of differ in these experientially detectable ways.

Speaking in favor of a representational account of perceptual experience there is, finally, the fact that it solves an old and a very troubling philosophical problem. Although the appeal of this consideration is limited to realists of a certain stripe, I mention it anyway. Most people are perceptual realists. Besides, since solutions to philosophical problems do not grow on trees, this consideration should carry some weight.

By perceptual realism I mean the view that in ordinary perception we are directly aware of physical objects and events--things that exist independently of our perception of them. Seeing a tree is not to be understood as awareness of some mental intermediary (an image, a sense-datum) having the properties the tree appears to have. Almost everyone, in their unguarded moments, and outside the philosophy classroom, believes this. Perceptual realism of this sort, however, has always had trouble with hallucinations. What is it one sees (experiences) when hallucinating a pink rat? Certainly not a pink rat--everyone agrees about that--but is it, nonetheless, something that, like a pink rat, is pink and rat-shaped? Is it something that exemplifies these properties? There certainly seems to be something, but, if there is, where is this pink, rat-shaped, thing? It isn't out there. Just ask your friends. For materialists it isn't in the head either. There is nothing pink and (probably) nothing rat-shaped in the brain of a person hallucinating pink rats

This difficulty in answering this question has inspired some pretty desperate moves in philosophy--e.g., adverbial theories of perceptual experience. A representational theory provides a happy rescue. In hallucinating pink rats we are aware of something--the properties, pink and rat-shaped that something is represented as having--but we are not aware of any object that has these properties--a pink, rat-shaped, object. We are aware of pure universals, uninstantiated properties. A representation, remember, doesn't need an object that has the properties the representation represents something to have. The radar doesn't need an object, external or internal, much less an airplane, approaching from the east to represent something at 35,000 feet approaching from the east. Representations are first and foremost representations of properties--altitude, direction, temperature, pressure, shape, color, size, hardness, distance, velocity, position and so on. If nothing stands in the right causal relationship to the representation to qualify as the object being represented, then the representation, unconnected to a proper object, says that something has these--as it turns out--uninstantiated properties when nothing, in fact, has them. There is no need to invent internal objects to have the properties an unconnected representation mistakenly attributes to an object. Just as radar can "hallucinate" an airplane approaching from the east at 35,000 feet, a visual experience can represent a pink rat in the corner without there being anything--certainly nothing pink and rat-shaped--it represents to be that way. There does not have to be anything X represents to have the property Y for X to represent there to be something having the property Y. Problem solved.


3. Problems: some examples.

That is the good news. Now the bad news. What speaks against this account of perceptual experience?

One thing that might (if it were true) tell against a representational theory is if there were no plausible theory of original or natural representation that could put some flesh on these bones. If we can't imagine how electrical-chemical activity in the nervous system could represent there to be a pink rat eight feet in front of a person, then, despite the philosophical benefits of thinking of visual (auditory, etc.) experience in this way, the theory never really gets out of the gate. It would be nice if it were true, yes, but we can't imagine how it could be true.

There are, of course, philosophical theories of representation (including my own: see Dretske 1995) that purport to give a naturalistic account of representation, an account that describes how biological systems actually perform this wondrous feat. But there isn't much agreement (at least not much positive agreement) about the plausibility of these theories. One of the stumbling blocks is that any plausible theory of representation (just like any plausible theory of meaning) is externalistic. It locates a system's powers of representation in the network of relations (causal, informational, etiological) the system bears (or bore) to external affairs. This has the consequence that representational content (hence, on a representational theory, the quality of experience) does not supervene on the biology of a system. Biologically identical organisms (with different histories, say) can be representing their environments in completely different ways. On a representational theory of perceptual experience, then, they can be having much different experiences of the things they see. But how can biologically identical organisms be having different experiences?[5] Although many philosophers have been willing to accept this result for belief (one person believes the liquid he is seeing is H2O, his biological twin believes it is XYZ), fewer are willing to accept it for a person's perceptual experience of the liquid. Whatever difference there may be in their beliefs about the liquid, it must look the same to both of them.

Perhaps there are (or will be) plausible theories of representation that (unlike my own) avoid this consequence by making representational content supervene on the current, internal, state of the agent. I don't see how this can be done, but maybe my imagination is too feeble. Or perhaps (this is my own view) this (to some) intolerable consequence can be neutralized by a deeper understanding of exactly what it is (i.e., phenomenal experience) we are giving a representational theory of. Since I have already said all that I can usefully say on this topic (Dretske 1995: Chapter 5; 1996), I leave the matter here without further comment. About these matters--matters of relative plausibility--people must judge for themselves.

I turn, instead, to an aspect of this theory that has probably attracted the most critical attention. This is the idea that the way things seem (phenomenally) is completely given by the representational character of the experience. Once you have said how the experience represents things to be, you have said everything about the experience that is subjectively accessible to the person having the experience. As might be expected from its uncompromising generality, that claim has attracted a torrent of counterexamples. The argument is not that there are phenomenal qualities that, as a matter of fact, the human nervous systems cannot represent. That would be an argument about the representational capacity of biological systems, and I do not hear that empirical argument being mounted by philosophers. The argument, rather, is that there are certain qualities we experience that cannot be understood as qualities anything is being represented as having and this for reasons having to do either with the nature of representation itself or the specific quality in question. The argument goes like this: in having experience of type E, things seem F to us; but nothing in us can represent something to be F; therefore experiences of type E are not (contrary to representationalism) completely determined by their representational properties.

Once the dialectical position is expressed in this way, and once we appreciate the enormous variety of representational systems (e.g., stories, instruments, pictures), it is clear that convincing arguments of this form will not be easy to find. How does one find a property that nothing in us can represent something as having? Isn't the very act of specifying the property a way of representing it? Wouldn't a property that nothing could be represented as having be one that was absolutely undetectable? If it was detectable, then we could build an instrument to detect it. This instrument could then represent (possibly misrepresent) objects as having the property. So the property, if it is detectable, is a property we can imagine instruments (and if instruments, why not nervous systems?) representing objects to have. It seems, then, that counterexamples to a representational theory of experience will be forced to appeal to undetectable properties as the ones that the theory cannot give an account of.

These are, I confess, first reactions--intuitions, if you will--of someone committed to a representational theory of experience. It is my way of saying that I do not see how a representational theory of experience can be refuted by arm-chair philosophy. Once we are clear about exactly what biological representations are, perhaps it can be refuted by the scientific facts. But not a priori reasoning. Still, it is worth looking at the arguments. Maybe I'm missing something.

I won't be looking at all the arguments. There are too many. I'll pick and choose. My choices are self-serving, of course, but I think something can be learned by looking at two examples in particular. They exhibit, each in their own way, a common tendency to misinterpret the representational story. Their failure, therefore, is instructive. Whether all counterexamples can be handled as easily as these is an important question that I won't even try to answer.

In criticizing my (Dretske 1995) representational account of sense experience Kent Bach (1997: 467) cites the following kind of case:


The most obvious objection to phenomenal externalism is that there are some phenomenal properties that really are attributable to experiences themselves. . . . For example, visual experiences can become blurry, as when one removes one's glasses, without their objects appearing to have become fuzzy. Their objects look different, of course, but do not look to have changed.


The objection is that although (in removing one's glasses) one experiences blurriness, one's experience does not represent the experienced object as blurry. Hence, there are some phenomenal properties that are not properties things are represented as having. These properties, Bach suggests, are properties of the experience itself.

One is tempted to ask whether there is supposed to be something in the head (that, we are assuming, is where the experience occurs) that gets blurry when one removes one's glasses. Is it this, the representational vehicle, some part of the visual cortex, then, whose blurriness one becomes aware of when things look blurry?

Surely not. Putting on rose-colored glasses (so that objects look rose colored) does not change the color of things in the head. The visual cortex doesn't turn pink. Why, then, should taking off one's glasses so that the objects look blurry make one aware of something (an experience in the head) that is blurry. Once we abandon a sense-data theory of perception, we realize that nothing--neither the object seen nor the experience of it--need be blurry for objects to look blurry. And once we appreciate that fact, we are back to a representational theory of experience. Blurry is the way experience represents objects, and you don't need a blurry representation to represent things as blurry. You can do it, for example, with sharply printed words.

It is easy to confuse: (1) properties of a representation with (2) properties the representation represents the objects (being represented) as having (i.e., the intentional properties). This is especially so with pictorial representations. Imagine two pictures, one a blurry picture (e.g., a photograph taken with an out-of-focus camera) of a sharply defined object (a block of ice), the other is a sharply focused picture of a fuzzy-edged object (e.g., a wispy cloud). The sharp picture of the fuzzy cloud might resemble the blurry picture of sharp ice. So, if the picture of the ice is blurry, if blurriness is a property of this representation, then the sharp picture of the fuzzy cloud, looking much the same, should also be blurry. But the picture of the cloud is not at all blurry. Maybe the cloud is blurry (I don't think so), but the picture certainly isn't. It is perfectly in focus. Blurriness, when applied to pictures, refers to the way an object is represented. A picture is blurry if it represents a sharp object as having fuzzy edges--when, that is, the property an object is represented as having does not correspond to the property of the object being represented. That is why the picture of the cloud--though it looks exactly like a blurry picture--is not blurry. There is no misrepresentation. The borders of the object being represented are as fuzzy as they are represented as being.

When Bach describes an experience as being blurry he is confusing an intentional property of a representation--how the experience represents things to be--with a property of the representation itself. This is easy to do with blurry pictorial representations since pictorial images of a sharp object actually have the property (fuzzy edges) they represent the sharp object to have. So it is easy to mistake the intentional property for a property of the representation. No one confuses these properties in the case of verbal representations. That is because the words describing something as blurry (e.g., "blurry") need not (like a picture) have fuzzy edges to do it.

Some of Peacocke's (1983) examples involving constancy phenomena might well involve a similar confusion although they are harder to analyze since, in this earlier work, Peacocke is not distinguishing, as I am doing here (and as he does later--Peacocke 1992), conceptual from non-conceptual representations. If we set aside this difference as best we can, though, his examples are worth discussing because they raise deep and puzzling questions about the nature of perceptual experience and the phenomenal qualities that define it. I will discuss only one of his examples.

You see two trees, one of them at one hundred yards (call this one Close Tree), the other (Far Tree) at two hundred yards. The trees are the same size, and (constancy mechanisms being what they are) they look to you to be the same size: that is, "taking your experience at face value you would judge that the trees are roughly the same physical size." (1983, p. 12) Peacocke concedes that this property (viz., same physical size) is a representational property of the experience. It is the way your experience represents the objects. Nonetheless, Close Tree takes up more of your visual field, it (as psychologists like to put it) subtends a greater angle than does Far Tree. This feature is as much a feature of your experience of the trees as is the fact that they look to be the same size. In this sense (relative amount of visual field occupied), then, Close Tree looks bigger than Far Tree.

The question is whether one's experience of the trees not only represents Close Tree as the same size as Far Tree (as a result of the operation of size constancy mechanisms) but also, in some different (but phenomenologically accessible) sense, represents Close Tree as larger than Far Tree.[6] I see nothing wrong with saying that the experience represents both things. This, of course, is no contradiction. It is merely the difference between an object-centered (allocentric) description of the trees (as being the same size) versus a perceiver-centered (egocentric) description of one as being larger than (i.e., occupying more of the perceiver's visual field) the other. The trees are also being represented as being at different distances from the perceiver, and the way trees of the same size are represented as being at different distances from the perceiver is by representing them as occupying different areas of the visual field.

All this seems straightforward and not only consistent with, but supportive of, a representational account of perceptual experience. I mention the example, nonetheless, because of my suspicions that some people will find the example convincing against a representational theory for the wrong reasons--exactly the kind of reasons that led Bach to confuse blurry awareness of objects with awareness of blurry objects. The thinking goes like this: in seeing trees as the same size we are consciously aware not only of the property the trees are represented as having (same size) but also the non-representational properties in virtue of which they are seen that way--viz., the comparative amounts of the visual field the trees occupy. Close Tree is "larger" than Far Tree in the same way two pigmented areas on a perspective drawing of the same-sized trees (at different distances) are of different size. The drawing represents the two trees as being the same size, and the way it does this is by splotches of pigment of much different size. In the drawing, Close Tree is represented by a 2" splotch of green, Far Tree by a 1" splotch of green. In looking at the drawing, we are aware of both the property the trees are represented as having (same size) and the property of the drawing (representation) by means of which they are represented this way (2" splotch in lower left corner, 1" splotch in the upper left). It is easy to confusedly think that our ordinary, unmediated, perception of trees is like this. We are aware of both the way (same size) our experience represents the trees and (if we attend to it) the properties of the representation (relative size of the representations) by means of which they are represented that way.

That is the same mistake as in the Bach example. One misidentifies a represented property--comparative amount of the visual field the two trees occupy--with a property of the representation. In viewing a photograph of the two trees, I am aware of a property of the representation (the relative size of the two splotches of pigment), but in viewing trees I am not. Both the properties I am aware of when I see the trees--both their comparative size (allocentric property) and distance from me (egocentric property)--are (relational) properties the trees are represented as having. Unlike viewing a picture, I am not aware of (and do not represent there to be) two differently sized objects that (together with appropriate placement of these objects in the picture) represent trees as being the same size. The only objects I am aware of in seeing the trees are the trees themselves, and they are, and they are represented as being, the same size. If things are working right, these objects are also represented as being different distances from me and, therefore, as a matter of geometry, as necessarily subtending different angles and (what amounts to the same thing) as occupying different areas of my visual field. These, however, are all properties the trees are represented as having, not (as in pictures of trees) properties of those objects (splotches of pigment on the picture surface) that represent the trees.


4. Modes of Presentation.

I conclude this skimpy survey of counterexamples with a brief discussion of modes of presentation, an idea that enjoys wide appeal but that has no place in a representational theory of experience.

Ned Block (1995) invites one to compare a visual experience of a property with an auditory experience of the same property. His choice of property is overhead. Hearing a sound as coming from overhead is, he says, different from seeing something as overhead. The quality of the experience is different but the same property is being experienced (234). So, he concludes, there is a phenomenal difference in these experiences that is not traceable to representational differences. The property overhead has different modes of presentation--in this case, a visual and an auditory mode.

Michael Tye (1995: 157) argues that an auditory, but not a visual, representation of the property overhead represents loudness. So the phenomenal difference between seeing something overhead and hearing something overhead may not be a result of the different way the property overhead is being presented (a non-representational difference) but a difference in what else besides overhead is being represented. I suggested the same in (Dretske 1995). To show that a modal difference is non-representational it is not enough to show that there is a phenomenal difference in experiences of the same property. One has to show that the difference in the experiences are the result of that (and no other) property. Otherwise the phenomenal difference can be attributed to what else is being represented in the two modes.

And how might one do this? It will not be easy. It will not be easy because even if one could isolate a property that was experienced in splendid isolation in two distinct modalities (nothing else was represented in either mode) a representational theorist could always take refuge in the possibility that whatever phenomenal differences persist in the two ways of experiencing this property are to be accounted for in representational terms by a concurrent representation of modal differences. That is to say, in representing F-ness in mode V (vision) and T (touch), the phenomenal difference in our awareness of F-ness might be explained as the difference in representing F and (some aspect of) V in the first case and F and (some aspect of) T in the other. In representing a property, there is--or there may always be--a representation of the channel over which information about that property is received. If this were so, then even if there were phenomenal differences associated with different modes of access to objective properties, differences in our experience of objective properties would still be representational differences.

Is this implausible? I don't think so. Consider our perception of movement. We know that whether or not we sense movement in a perceptual object (whether it appears to be moving in the phenomenal sense of "appear") depends not simply on what happens on the retina. It also depends on information the brain receives and the commands it gives whether or not executed (Rock 1975: 187) concerning position and movement of the eyes. If you fixate a stationary object you sense no movement. If you track a moving object--thereby keeping the retinal image immobile (as immobile as when you fixate a stationary object)--you sense movement. So the experience of movement depends on information the visual system has about itself. On a representational view of experience this means that the quality of experience is given not just by information about what is happening on the retina (and points further out), but also by what is happening in the perceptual system itself (points further in). This being so, can't we suppose that modal differences (the alleged differences in the way properties are presented in two sensory modes) are really representational differences--differences in what else the perceptual systems represent about themselves?

I do not know how plausible this is. Maybe it is far-fetched. Still, the possibility is worth mentioning if only to indicate how hard it is to find effective counterexamples to a representational account of experience. The only thing that is impossible according to a representational theory are phenomenal differences with no representational differences. As long as there exist modal differences, though, a representational theorist can take refuge in the idea that, perhaps, different ways of gaining access to the world are themselves represented in our experience of the world.






Bach, K. 1997. Engineering the mind. Review of Naturalizing the Mind by Fred Dretske. Philosophy and Phenomenological Research, LVII.2 (June 1997), pp. 459-468.


Block, N. 1995. On a confusion about a function of consciousness. In Behavioral and Brain Sciences 18: 227-287.


Block, N. 1997. Author's Response. In Behavioral and Brain Sciences 20: 1. 159-166.


Davies, M. 1996. Externalism and Experience. Philosophy and Cognitive Science: Categories, Consciousness, and Reasoning. Eds. A. Clark, J. Ezquerro, and J.M. Larrazabai. Dordrecht: Kluwer Academic Publishers.


Dretske, F. 1981. Knowledge and the Flow of Information. Cambridge, MA; MIT Press.


Dretske, F. 1995. Naturalizing the Mind. Cambridge, MA; MIT Press.


Dretske, F. 1996. Phenomenal Externalism: If Meanings ain't in the Head, Where are Qualia? Philosophical Issues, 7 (1996), Enrique Villanueva, ed. Ridgeview Publishing Company; Atascadero, CA. Pp. 143-158.


Harman, G. 1989. Some philosophical issues in cognitive science: qualia, intentionality, and the mind-body problem. In Foundations of Cognitive Science, Michael Posner, ed. Cambridge, MA; MIT Press. 831-848.


Harman, G. 1990. The intrinsic quality of experience. In Philosophical Perspectives, 4: Action Theory and Philosophy of Mind, James Tomberlin, ed., Atascadero, CA; Ridgeview Publishing Co. Pp. 31-52

Haugeland, J. 1981. Mind Design. Cambridge, MA; MIT Press


Peacocke, C. 1983. Sense and Content: Experience, Thought, and their Relation. Oxford: Clarendon Press


Peacocke, C. 1992. A Study of Concepts. Cambridge, MA; MIT Press.


Rock, I. 1975. An Introduction to Perception. New York; Macmillan.

Tye, M. 1995a. Blindsight, orgasm, and representational overlap. In Behavioral and Brain Sciences 18, pp. 268-269.


Tye, M. 1995b. Ten Problems of Conscioiusness. Cambridge, MA; MIT Press.




[1] We must be careful to distinguish misrepresentation from non-representation. A measuring instrument turned on that registers 0 units to a condition of 5 units misrepresents this condition. Had it been turned off, it would not have represented (hence, misrepresented) anything despite registering the same value (0).


[2]In saying that Hamlet depicts the world as containing a Danish prince, I do not mean to suggest that in doing so it necessarily (though it might) misrepresent the world this way. Not all false representations are misrepresentations. Caricatures and (deliberate) fiction isn't.


[3] Harman (1989, 1990) makes essentially these same points. I nonetheless depart from Harman, a fellow representationalist, who takes experience of quality M to be essentially a belief that something is M. I take experience to be a non-conceptual representation of the properties being experienced. According to my lights, Block (1997, p. 164) is right in distinguishing experiences of M from recognizing (i.e., conceptually representing) something as M.


[4] In an earlier work, Chris Peacocke (1983, p. 7) limits the representational content of experience to properties the subject has the concept of. It is, he says (p. 19), a conceptual truth that one cannot have an experience with a given representational content unless one possesses the concepts from which that content is built up. Limiting experiences in this way he builds a strong case for a non-representational (sensational) aspect of experience. Later, however, Peacocke (1992) develops the notion of scenario content, a type of non-conceptual (but nonetheless representational) content possessed by experiences.

[5] I have found that some philosophers take this to be a denial of materialism. A little reflection, though, shows only that it is a denial of an extremely nave form of materialism. A materialists needn't suppose that just because two pieces of paper are physically indistinguishable, they must, if they are pictures of something, be pictures of the same thing. They might be photographs of twins or of different (but identical looking) paper clips. All that this (differences that do not supervene on the current material condition of objects) shows is that some perfectly respectable physical properties--and this includes being a picture of X--are relational properties.


[6] Peacocke argues that the comparative amount of the visual field occupied by the two trees is not a relation between them that they are represented as having because this relationship exists even for people who do not have the concepts of visual field, area, and so on (those concepts that, according to Peacocke, are needed for mental representation of any kind). This is not an argument that he would any longer accept and so I skip over it here.