One of the chief functions of the mind, both in our day-today living and over the long evolutionary haul, is to relate us to the rest of the world, especially by way of perception and action. To put the point in the simplest possible terms, by perception we take in information about the world, we then coordinate this information both consciously and unconsciously, and make decisions or otherwise form intentions, which result in actions by way of which we cope with the world. In this chapter we will consider the relations between perception and the world that exists apart from our perceptions, what philosophers like to call, misleadingly, the “external world.”
Why is there supposed to be a problem? If I extend my arm forward, I see my hand in front of my face. What could be easier than that? There is a tripartite distinction between me, the hand, and the actual conscious experience of perceiving by way of which I perceive the hand. There is, of course, a complex neurobiological story to be told about how the reflection of light off of the hand attacks the visual system and sets up a series of neuronal processes that eventually result in the conscious experience of seeing the hand. Furthermore, there are some philosophical niceties, as we saw in our discussion of intentionality, about the form of the causal self-referentiality involved in the conditions of satisfaction of the visual experience. But so far it does not seem very difficult. However, I have to tell you that there are few problems in the history of philosophy that have given more trouble than the problem of perception.
The view of perception that I have just adumbrated is a form of perceptual realism and is sometimes called “direct realism” and sometimes even “naïve realism.” Most of the great philosophers in the history of the subject are convinced that it is false. They believe (and by “they” I mean such great philosophers as Descartes, Locke, Berkeley, Hume, and Kant) that we do not see the real world. We do not see independently existing objects and states of affairs in the world. All that we ever actually perceive directly—that is, perceive without the mediation of any inferential processes—are our own inner experiences. In the past century philosophers usually put this point by saying “We do not perceive material objects, we perceive only sense data.” Some of the earlier terminology used for sense data were “ideas” (Locke), “impressions” (Hume), and “representations” (Kant). But if asked, “What is the direct object of a perceptual verb, taken literally, strictly, and philosophically?” the tradition has almost always been to say that the direct objects of perceptual verbs are not expressions naming independently existing material objects but expressions naming our own inner experiences, our sense data.
What are the arguments for this apparently counterintuitive view? There are two famous families of arguments, the argument from science and the argument from illusion. I will consider each in turn.
The scientific account of perception shows how the peripheral nerve endings are stimulated by objects in the world, and how the stimulation of the nerve endings eventually sends signals into the central nervous system and finally to the brain, and how in the brain the whole set of neurobiological processes causes a perceptual experience. But the only actual object of our awareness is that experience in the brain. There is no way we could ever have direct access to the external world. All we can ever have direct access to is the effect that the external world has on our nervous system.
This argument seems to presuppose that we were talking about the actual perception of the real world when we described how objects in the world cause the stimulation of our nerve endings; but in fact the argument concludes that such a perception is impossible. Bertrand Russell once ironically stated this apparent paradox by saying: “Naïve realism leads to physics, and physics, if true, shows that naïve realism is false. Therefore naïve realism, if true, is false; therefore it is false.”1
The point that I take it Russell is making is that naïve realism seems somehow self-defeating. If you try to take seriously the idea that we are in direct perceptual contact with the external world, and do science on that basis, science will give you the result that we cannot be in direct perceptual contact with the external world.
I think the argument most likely to convince most people in the history of this subject is the argument from science. But in the history of philosophy the argument that has been more influential among philosophers is called the argument from illusion.
If we try to take naïve realism seriously it seems to lead to some sort of inconsistency and self-contradiction. Here is how. Suppose I now hold a knife in my hand and I see the knife. But Macbeth, in a much more dramatic situation, also had the experience of seeing a knife, specifically a dagger. However, Macbeth was having a hallucination. He did not see a real dagger, but only a hallucinatory dagger. So in Macbeth’s case we cannot say that he saw a material object. But he definitely saw something. We might say he saw the “appearance of a dagger” or a “hallucinatory dagger.” But now, and this is a crucial step, if we are going to say in Macbeth’s case that he only saw the appearance of a dagger, then we should say it in every case, because there is no qualitative difference between the character of the experience in the veridical cases and in the hallucinatory cases. That is why Macbeth was deceived: there was no difference between the experience he had and the experience of actually seeing a dagger. But if we say that in every case we only see an appearance and not the object itself, we should surely get a name for these appearances. Let us call them “sense data.” Conclusion: we never see material objects, but only sense data. And now the question arises, What is the relationship between the sense data we do see and the material objects that apparently we do not see?
This form of argument has been run on a wide variety of different sorts of examples. Here is another one. When I hold a finger up in front of my face and focus my eyes on the wall at the far side of the room, a phenomenon occurs known as double vision. I see my finger double. But now, when I see my finger double, I do not see two fingers. There is only one finger there. But I obviously do see two of something. What do I see two of? Well let us call these somethings that I see, the appearances of a finger—and I do indeed see two appearances of a finger. But now—and again this is a crucial step—there is no qualitative difference between seeing the appearances of a finger and seeing the real finger. I can prove this to myself by refocusing my eyes so that the two appearances coalesce into a single appearance. Where I was previously seeing two appearances I am now seeing only one appearance. So if we are going to say in the double vision case that I only see appearances and not material objects, we should say it in every case. Let us get a name for these appearances; let us call them “sense data.”
Here is a third argument. If I put a straight stick in a glass of water, because of the refractive properties of light, the stick looks bent. But now, the stick is not really bent; it just looks bent. Still, when I see the stick I am directly seeing something bent. What is it? I am directly seeing the appearance of a stick and the stick does indeed present a bent appearance. But the stick is not itself bent; the appearance is bent. But then what I directly see is bent, so what I am seeing is the appearance and not the stick. And by now you will recognize what the next step is going to be: if I am going to say in this case that I do not see the stick but only the appearance, I should say it in every case because there is no qualitative difference between the cases. We need a term to describe these appearances. Guess what? We will call them “sense data.” Conclusion: I never see material objects but only sense data.
I could keep going all day with these examples, but just a couple of more to give you the full flavor of the style of argument. Suppose I get up from my chair and walk around the table, while keeping my eyes on the table. As I walk around, something is changing; furthermore, something I directly perceive is changing. The table is not changing. The table remains absolutely unchanged throughout my walk. But what does change? Obviously, it is the appearance of the table. The table presents to me a different appearance from different points of view. But now, since what I see is changing and the table is not changing, and what I see is the appearance, it seems that I am seeing only appearances and not the table. Furthermore, since there is no qualitative distinction between this experience and any other, I seem forced to the conclusion that I never see anything but appearances. We need a technical term to name these appearances. We will call them “sense data.”
Here is another example, also famous. I take from my pocket a coin and hold it up. As I look at it straight on it looks round. But if I turn it slightly at an angle it no longer looks round; it looks elliptical. But now we know one thing for certain: the coin itself is not elliptical. It has not changed its shape as I turned it on an angle. But we also know that I am directly perceiving something elliptical. There is no question that right here now in my visual field there is something elliptical; I directly see it. But it seems then that what I am seeing is not the coin, for the coin is round. What I am directly seeing, what I am seeing without any inferential process at all, is the elliptical appearance of the coin. And if I am going to say in this case I only see appearances, I should say it in every case for there is no qualitative change when I turn the coin directly upright so that it now presents a round rather than an elliptical appearance. The conclusion is obvious: we should say in every case that I see appearances, not material objects, and these appearances can be called “sense data.”
Nearly all of the famous philosophers of the past 350 years, and most of the respectable philosophers until about the middle of the twentieth-century, accepted some sort of sense-datum theory. Hume, indeed, thought that naïve realism was so obviously false that he hardly bothered to refute it. At one point he says that if you are tempted to naïve realism you can refute it by just pushing one eyeball. When you push one eyeball you see everything double and, according to Hume, the naïve realist would have to conclude that the universe simply doubled in the number of objects that it contains. But since it obviously did not double, Hume thinks it follows that we are not seeing material objects.3
The argument from illusion has a logical structure that is common to all of these examples. Here is how it goes:
1. Naïve realists assume that, in the typical case at least, we see material objects and that we see them as they really are.
2. But there are lots of cases, as even the naïve realist would admit, where we do not see material objects (for example, in the hallucination cases), or do not see them as they really are (as for example, in the bent stick case and the elliptical coin case).
3. But even in theses cases we do see something and we do see it as it really is. In the cases where there is no material object there at all, as for example in the Macbeth dagger case, Macbeth did see something. There was something directly present in his visual field. And in the cases where there is a material object there but we do not see it as it really is, as in the examples of the elliptical coin and the bent stick, we do see something elliptical and we do see something bent. Both the elliptical entity and the bent entity are directly present to us in our visual field.
4. In these cases we directly see appearances, etc. (sense data) and not material objects.
5. These cases are not qualitatively different from the standard case, hence if we are going to say in these cases that we see sense data and not material objects we should say the same thing in every case.
I I. CONSEQUENCES OF THE SENSE-DATUM THEORY
Direct realism is the view that we, at least typically, directly perceive objects and states of affairs in the world. Direct realism is denied when we say that we never perceive objects and states of affairs but only our own experiences, our own sense data. But once we make that move we have a very serious question: What is the relationship between the sense data that we do perceive and the objects that we apparently do not perceive? There are a number of answers to this question in the history of philosophy but I believe that basically they boil down to two families. One family, the most immediately appealing, is to say we do not perceive objects themselves, but we do perceive representations of objects. The sense datum that we do perceive is a kind of a picture of the object, and so we can find out about the object by inferring the presence and features of the object from the characteristics of the sense data. The actual object in the real world resembles the sense data at least in certain respects. Some philosophers, perhaps most importantly Locke, made a distinction between those features of the sense data that have corresponding resembling elements in the real world and those features that do not. The features of the real world that actually resemble sense data were called “primary qualities” and they consisted of shape, size, number, movement, and solidity. (Locke’s list is “solidity, extension, figure, motion or rest, and number.”)4 But there are other sense data for which there is no corresponding resembling feature in the real object. Locke misleadingly called such features of objects “secondary qualities.” This is misleading because strictly speaking there are no such qualities of objects. Rather, as Locke points out, the secondary qualities are just the powers that the primary qualities have to cause in us certain experiences. These secondary qualities are color, smell, taste, and sound. Our experiences of both primary and secondary qualities are caused by real features of the object; but the object itself does not have the features corresponding to our experiences of the secondary qualities.
This doctrine is called the representative theory of perception and it was worked out in some detail, especially by Locke. According to the representative theory of perception, we spend our conscious lives as if we were inside a movie theater. We can see pictures of the real world on the screen of the movie theater, but we can never go beyond the inside of the movie theater to see the real world itself, because the movie theater is entirely in our mind. All we ever see are more pictures, more representations. The representative theory was attacked, I think very effectively, by both Berkeley and Hume. There are a number of forms of the attack, but the basic argument, the one to which there does not appear to be an answer, is this: if we say that our sense data resemble objects and thus represent them in the way that a movie of a scene represents the actual scene, then the problem is that we have given no clear meaning to the notion of “resemblance,” and consequently no clear meaning to “representation.” How can we say that the sense data we do see resemble the object that we do not see if the object is by hypothesis totally invisible? It is as if I said I had two cars in my garage and they both looked exactly alike, but one was totally invisible. It makes no sense at all to say that there is a perceptual resemblance relation between something that has perceptual features and something that has no perceptual features.
When Berkeley saw this point he did not, as one might have hoped, go back to naïve realism and say he must have made a mistake when he moved from the naïve realist theory of perception to the sense-datum theory. Rather, Berkeley says the only things that exist are minds and ideas. The real world consists entirely of sense data. There are no such things as material objects in addition to our actual and possible experiences. Hume, though in a more complex fashion, adopted a similar conclusion. This view has various names but perhaps the most common name for it is phenomenalism. Material objects consist in collections of sense data; there are no material objects over and above, or in addition to mental phenomena.
Phenomenalism was intended as a logical thesis and thus can be most clearly stated as a logical thesis about language. Instead of saying objects consist of sense data, which makes it look as if we are disagreeing with the view that objects consist of molecules, what we should really say is that statements about objects, and indeed empirical statements in general, can be translated without loss of meaning into statements about sense data. The same verificationist impulse that led to behaviorism in the philosophy of mind led to phenomenalism in the philosophy of perception. Just as the only evidence we have for other minds is behavior, so it seems the only evidence we have for material objects is sense data. A truly scientific conception of minds must therefore be behavioristic, and analogously a truly scientific conception of the material world must be phenomenalistic.
I believe this whole way of thinking about perception is hopelessly misconceived. As I said earlier I believe it is the most disastrous theory in the history of philosophy over the past four centuries. Why? Because it makes it impossible to give a true account of how human beings and other animals relate to the real world. It leads almost inevitably from Descartes and Locke to Berkeley and Hume, and from there to Kant. And then things get really bad as the tradition leads to Hegel and absolute idealism. The whole thought of attacking it once again depresses me enormously, but I will not have done the job I promised you I would do in this book if I did not to attempt answer it point by point. So here goes.
The arguments for the sense datum-thesis are, without exception, fallacious. Let us consider them in order.
Science does not refute naïve realism. To say that because we can give a causal account of how it comes about that we see the real world, it follows that we do not see the real world, is to commit a famous fallacy. It is called the genetic fallacy. It is the fallacy of assuming that a causal account that explains the genesis of a belief, that explains how the belief was acquired, thereby shows the belief to be false.
The genetic fallacy is usually about beliefs, but the form of the fallacy can be generalized. The idea is that if you can show that the causes of a belief or other intentional content are insufficient to prove its truth, then you have somehow refuted the belief or other intentional state.
In my intellectual childhood, the most common forms of the genetic fallacy were in Freudianism and Marxism. You doubt the truth of Marxism? That only shows that you are misled by your bourgeois class background. You doubt the truth of Freud’s teachings? That only proves you are a victim of your own repressions. Nowadays, one does not hear the genetic fallacy much except from postmodernists. I used to wonder why the fallacy was so common in postmodernism until I read an account that explains why the postmodernists really have no other form of argument available to them.5
Anyway, the form of the genetic fallacy in the theory of perception goes as follows. We can show that when you apparently see your hand in front of your face, what is actually happening is that light reflected from the hand is causing you to have a visual experience, which you take to be a visual experience of your hand. Because we can explain why you think you are seeing a hand, we can show that you did not really see a hand in front of your face but only the visual experience, which was the effect of the neurobiological processes.
So stated, I hope it is obvious that this is a fallacy. The causal account of how I come to see my hand in front of my face does not show that I do not really see my hand in front of my face.
Replying to the argument from illusion is trickier. I will borrow both the ideas and the techniques of my teacher in philosophy, J. L. Austin in order to refute this argument.6
Notice that in every one of the arguments I gave, the linguistic strategy is to get a noun that will be the direct object of verbs of perception but that does not name a material object. So, in the case of Macbeth’s dagger, we were told that we did not see a real dagger but only a hallucinatory dagger. But the difficulty with this is that in the sense of “see,” I really see a knife in my hand; in the case of the hallucination, I do not see anything. Expressions like “hallucinatory dagger” cannot name a species of dagger. To put it in words of one syllable, when Macbeth had a hallucination, he did not see anything. At least not anything in the dagger line of business. No doubt he saw his hands. So from the fact that Macbeth had a hallucination that was phenomenologically indistinguishable from a real experience it does not follow that he saw a special kind of object or entity that is common to both veridical and illusory experiences.
Similar objections can be made to the double-vision cases. One should never accept the question uncritically. The question was, When I see my finger double what do I see two of? The answer to this is: when you see your finger double you do not see two of anything. You see one finger and you see it double.
In both the double-finger and the bent-stick examples, the notion of appearance is introduced to provide a direct object of the verbs of perception. The idea is that you do not see the object itself but only its appearance. But if you think about this, there is something self-contradictory about the idea that I might see the appearance of an object and not see the object. To see the appearance of an object is just to see the way it looks. And there is no way you can see the way something looks without seeing that something. Consideration of examples will make this completely clear. Suppose I ask you, “Did you see the way Sally looked at the party?” It makes no sense for you to say, “Yes I saw the way she looked but unfortunately I couldn’t see her. I could only see her appearance.”
Let us apply these considerations to the example of the table. I get up and walk around the table. The appearance of the table changes, because I see it from different points of view, but the table does not change; therefore it seems that I see the appearance and not the table. I hope it is obvious that this is a fallacy. Of course the table looks different from different points of view. But the changes in my visual experiences, which are themselves brought about by the fact that I am changing my position and therefore my point of view, do not show that I fail to see the table, but only something that, so to speak, gets between me and the table, its appearance. On the contrary, the whole discussion presupposes that I am actually seeing the table throughout, for there is no way that the table could continue to present to me different appearances from different points of view if I were not actually seeing the table.
The crucial false step in the argument structure I summarized was step 3: in every case you perceive something and perceive it as it really is. This is not true. In the hallucination cases you perceive nothing and in the other cases—the bent stick, elliptical coin, etc.—you do perceive the object but under conditions that may be more or less misleading. From the fact that the stick looks (sort of ) bent it does not follow that you are really seeing a bent entity, the look. No, you are really seeing a stick, an independently existing material object, which under those conditions looks bent.
It is an amazing fact about the history of philosophy that these arguments have had the influence they have had. I do not believe they will bear a moment’s scrutiny and I leave it to the reader, as a five-finger exercise, to see how we could apply these lessons to show the fallacy in the argument about the elliptical coin.
But, one might say, refuting the arguments against naïve realism is not sufficient to show that naïve realism is true. This is a correct objection. We need some argument to show that on at least some occasions we do actually perceive material objects and states of affairs in the world. What could such an argument possibly be?
The problem we are confronting here is a variation of traditional skepticism. The skeptic’s argument is always the same: you could have all the evidence you do have, indeed you could have all possible evidence, and still be mistaken. So prove, for example, that you really are seeing the table in front of you and not just having a hallucination, dreaming, being deceived by an evil demon, etc. There is no way that I can answer the skeptic directly about my present visual experience of the table. The whole point of the skeptic is that I could be having exactly this experience and still be mistaken. And if I could be mistaken in this case, why not in every case?
I do not believe it is philosophically astute to try to answer this argument directly. I do not believe that I can prove to the skeptic that I am now really seeing the table as opposed to having a hallucination, dreaming, etc. What I can do instead, is to show that a certain style of discourse, the one in which the skeptic is currently engaged, presupposes the truth of some version of direct realism. (I like to think of my version as “naïve” but it does not matter whether it is naïve or sophisticated.) The realism in question has to contain the view that we have at least on some occasions perceptual access to publicly observable phenomena. These are commonly thought of as “material objects,” but again that designation is not crucial. What is crucial is that different people can at least on some occasions perceive the same publicly observable phenomena—chairs, tables, trees, mountains, clouds, etc. The argument that I am about to present is a “transcendental” argument in one of Kant's many senses of that term. In a transcendental argument in this sense, we assume a certain proposition p to be true and then show that a condition of the possibility of the truth of p is that another proposition q should also be true. In this case we assume that there is an intelligible discourse shared publicly by different speakers / hearers. We assume that people actually communicate with each other in a public language about public objects and states of affairs in the world. We then show that a condition of the possibility of such communication is some form of direct realism. The key to the argument is to see that the sense-datum hypothesis has, without explicitly revealing it, reduced the publicly available world of material objects to a private world of sense data. Only I can experience my sense data. Only you can experience your sense data. But how, then, can we ever talk about the same object in a public language? How, in short, can we ever succeeded in communicating with each other about public objects? If material objects are reducible to sense data, and the only sense data I have access to are my own sense data, then I could never communicate with you about a public material object.
Here are the steps of the argument:
1. We assume that we successfully communicate with other human beings at least some of the time.
2. The form of the communication in question is of publicly available meanings in a public language. Specifically, when I say such things as, “This table is made of wood,” I assume you will understand the words in the same way that I do. Otherwise we are not succeeding in communicating.
3. But in order to succeed in communicating in a public language, we have to assume common, publicly available objects of reference. So, for example, when I use the expression “this table” I have to assume that you understand the expression in the same way that I intend it. I have to assume we are both referring to the same table, and when you understand me in my utterance of “ this table” you take it as referring to the same object you refer to in this context in your utterance of “this table.”
4. That implies that you and I share a perceptual access to one and the same object. And that is just another way of saying that I have to presuppose that you and I are both seeing or otherwise perceiving the same public object. A public language presupposes a public world. But that public availability of that public world is precisely the direct realism that I am here attempting to defend. The problem with the sense-datum hypothesis, as with phenomenalism in general, is that it ignores the privacy of the sense data. Once you claim that we do not see publicly available objects but only sense data, then it looks like solipsism is going to follow rather swiftly. If I can only talk meaningfully about objects that are in principle epistemically available to me, and the only epistemically available objects are private sense data, then there is no way that I can succeed in communicating in a public language, because there is no way that I can share the same object of reference with other speakers. That is what I meant when I said that a public language presupposes a public world. But the presupposition of that public world is precisely the naïve realism that I have been defending. We do not prove the truth of naïve realism; rather, we prove the unintelligibility of its denial in a public language.