a) An Analysis of the Argument

The argument from the relativity of sensing is one of the most convincing arguments in the history of philosophy. From Locke to Meinong and Husserl, from Berkeley to Russell and Moore, no philosopher has been able to resist its spell. Reid, as we have seen, tried to resist, but even he succumbed in the end. What precisely is the argument, and how is it treated in a contemporary setting? I have chosen two examples, one from Russell, the other from Husserl.

Russell begins the argument in this way:

Although I believe that the table is ‘really’ of the same color all over, the parts that reflect the light look much brighter than the other parts, and some parts look white because of reflected light. I know that, if I move, the parts that reflect the light will be different, so that the apparent distribution of colours on the table will change. It follows that if several people are looking at the table at the same moment, no two of them will see exactly the same distribution of colours, because no two can see it from exactly the same point of view, and any change in the point of view makes some change in the way the light is reflected.

(B. Russell, The Problems of Philosophy, pp. 8-9)

The most important part of this description of a familiar situation is Russell’s claim that of several people, no two will see exactly the same distribution of colors. This claim, I think, is false. All of the people who are looking at the table (at the same time, from different points of view) will see that the table is, let us say, light brown. Russell’s mistake is not apparent because he also talks about what the table looks like; for example, that parts of it look white. And he also talks about a change, not of the color of the table, but of the apparent distribution of colors on the table. While we agree that parts of the table may look white and that the apparent distribution of colors may change, we also insist that what the persons see is a light brown table.

Next, Russell draws the first ominous conclusion from his description: “It is evident from what we have found that there is no colour which preeminently appears to be the colour of the table, or even of any one particular part of the table—it appears to be of different colours from different points of view, and there is no reason for regarding some of these as more really its colour than others” (Russell, 1977, p. 9). This conclusion is false. The table is seen to be light brown. Every observer agrees on that. What color is this table? One look from any point of view suffices: it is light brown. Repeat the test with another observer and you get the same result.

Russell now goes on to reason that since the colors which we “see” vary with our points of view, they cannot be colors of the table: “This colour is not something which is inherent in the table, but something depending upon the table and the spectator and the way the light falls on the table” (Russell, 1977, p. 9). We must deny, Russell maintains, that the table in itself has any one particular color. What holds for color, holds of course for texture, shape, etc. “Thus it becomes evident,” Russell concludes, “that the real table, if there is one, is not the same as what we immediately experience by sight or touch or hearing. The real table, if there is one, is not immediately known to us at all, but must be an inference from what is immediately known” (Russell, 1977, p. 11; the first italics are mine). With these last words, Russell has opened the door to skepticism. The skeptic can now argue, soundly in my opinion, that there can be no such inference which is grounded on experience. Hence we cannot know whether or not there is a table and, if there is one, what properties it has.

I called the argument from the relativity of sensing convincing. From the description of an ordinary situation one arrives in two swift steps at the conclusion that we do not “immediately experience by sight” the table and its color. But notice that in the process the perfectly good and normal ‘see’ has been replaced by the strange and awkward ‘immediately experience by sight’. Imagine that Russell had instead concluded that we do not see the table and its color. (Notice also the ‘real’ in front of ‘table’; such are the tricks of the philosophical trade.) Even a student of philosophy might then not have been impressed by the argument. “Lord Russell,” he may have said, “Let us go over the premises again; since the conclusion is obviously false, at least one of the premises must be false (or else, the argument must be invalid).” Russell and many other sophisticated philosophers, on the contrary, accepted the conclusion and then tried, ever more desperately, to discover the kind of inference that would lead from the immediately given, the sense-data, to the perceptual object, the table.

Russell speaks of sense-data and means by this expression “the things that are immediately known in sensation: such things as colours, sounds, smells, hardness, roughness, and so on. We shall give the name ‘sensation’ to the experience of being immediately aware of these things. Thus whenever we see a colour, we have a sensation of the colour, but the colour itself is a sense-datum, not a sensation” (Russell, 1977, p. 12; the first italics are mine). By a sensation Russell thus means the mental act of sensing. What is sensed is the sense-datum. A color is a sense-datum; for example, the color shade light brown is said to be a sense-datum. We find here the same mistake as in earlier philosophers: The sense-datum (sensation), which is an individual thing, is confused with one of its properties. The sense-datum, we repeat, has a color, a shape, etc., but it is not a color, shape, etc. It is clear that Russell uses “sensation” quite differently from us. We use it for the sense-datum; he uses it for the mental act of sensing the sense-datum. But this difference need not confuse us. My criticism of Russell’s version of the argument aims at the way in which he replaces talk about seeing a color by talk about sensing a color. He supplants the mental act of seeing by the quite different act of sensing. No wonder that the table and its properties disappear from the world of “experience,” for experience is confined to sensing. What is sensed, I agree with Russell, is a sense-datum and its properties, and not the table and its properties. Thus if all there were to “immediate experience by sight” is sensing, then we could indeed not know the table and its properties by “immediate experience by sight”. And how but by inference could we then know the table and its properties? But there is more to “immediate experience by sight.” There is seeing! And what is seen is not the sense-datum and its properties, but a table and its properties. When you look at the table, you see a light brown table, and you sense (experience) a certain sense-impression (sensation). It is not a question at all of inferring the color of the table from the color of your sense-datum. You see the color of the table.

Let us take another look at Russell’s argument. We would describe the situation in this way. When you look at the table, you see its color; it is light brown. There occurs an act of seeing which is neither a sensing of a sense-datum nor an inference from a sense-datum. You also experience a certain sense-impression which has a certain shape and a “distribution of colours.” If you pay attention to this sense-datum, you will notice, for example, that parts of it are almost white while others are light brown. You will also notice that this distribution of colors changes when you move around the table. Of course, the color of the table does not change when you move around it. Nor does your perception of the color change: you see that the table is light brown, no matter where you stand. What changes as you move are the colors of your sense-impressions, not the color which you see. Russell therefore describes, as he depicts the situation, the changing colors of our sense-impressions, and not the color which we see. One could say that he confuses the color seen with the colors sensed. Or one might say that he simply does not recognize that there is a color seen; all he acknowledges are the colors sensed.

As a result of this mistake, he thinks that no color appears to be the color of the table. This is quite true as long as we only consider the sensed colors. However, we must also keep in mind that from a certain point of view the color seen coincides with the color sensed. Perhaps the example of shape is clearer. If you were to look at the table top while suspended from the ceiling, straight down, you would see that it is rectangular. Furthermore, your sense-impression would also be rectangular. Under this condition, the shape which you see and the shape which you sense coincide. We may therefore call it “the normal situation.”

Since the colors which we sense vary with lighting conditions, position, and even with the state of the nervous system, these colors are not the colors of the table. The table, we know, does not change its color when we walk around it. But this does not mean that only the sense-impressions are colored, that color is not “something inherent” in the table: the table does have a particular color. And we know this color, because we see it.

Let us turn to Husserl:

The perceived thing in general, and all its parts, aspects, and phases, whether the quality be primary or secondary, are necessarily transcendent to the perception, and on the same grounds everywhere. The color of the thing seen is not in principle a real phase of the consciousness of color; it appears, but even while it is appearing the appearance can and must be continually changing, as experience shows. The same color appears ‘in’ continuously varying patterns of perspective color variations. Similarly for every sensory quality and likewise for every spatial shape.

(E. Husserl, Ideas, par. 41)

The table, we agree with Husserl, transcends our perception of it. It is not a part of our perceiving, but its object. And the same holds for the color of the table. But it is not true, as Husserl then goes on to claim, that we know the color of the table through or by means of perspective color variations, that is, through the colors of our sense-impressions. The table’s color is not transcendent in the sense that it transcends what is “directly before the mind” when we look at it. The color of the table is “directly before the mind” because it is seen.

As I said before, most modern philosophers have accepted the argument from the relativity of sensing. But this argument, as I also pointed out, leads directly to skepticism. The skeptic correctly argues that if we can know the color of the table only by inference from the colors which we sense, then we cannot know it at all. For, in order to know the table’s color, you must know something to the effect that the experience of a yellow sense-impression indicates under certain conditions that the table is light brown. But how could we possibly discover a regularity of this sort? There is only one way: We must compare the color of the sense-impression with the color of the table. We must “take a look,” first at the color of the sensation, noting that it is yellow, and then at the color of the table, noting that it is light brown. But if Russell and Husserl are correct, then we cannot “take a look” at the color of the table. We can only look at the color of our sense-datum. Therefore, we cannot compare the color of the sense-impression with the color of the table. Hence we cannot discover the required regularity. Therefore we cannot know the color of the table.

Let me formulate the argument more generally in order to show how powerful it is.

(1) We assume that we know the colors of our sensations under certain conditions.

(2) We also assume that we know the color of the table, not by acquaintance, but only by means of the following description: The color of the table is the color which causes us under certain conditions to experience sense-impressions that are, say, white, light brown, etc. More generally, we know the color of the table only as the color which stands in a certain relation, R, to the colors of our sensations. I take this assumption to be the conclusion of the argument from the relativity of sensing.

(3) But to know the color of the table in this manner is not to know what color the table has. We do not know, for example, whether it is light brown or dark brown, or whether it is green or navy blue.

The skeptic’s argument thus rests on the principle that to know something merely as the A which stands in R to B, is not to know what properties A has. Assume that you want to know who murdered the Dean of Women (assuming that just one person committed the foul deed). If you are told that it was the brother of the president of the university, then you still do not know what particular person killed the Dean, unless you also know what particular person is the president’s brother. Assume that the president’s brother has been introduced to you at a cocktail party as Mr. Henry Miller, and that you had a pleasant conversation with him. Since you are acquainted with Mr. Miller, you know the person who killed the dean, and you could pick the person out of a line-up. But obviously you could not do so if you were not acquainted with the person who killed the dean.

We know something about the color of the table if we know that it is the color which causes us, under certain circumstances, to experience a white sensation. We know, to be specific, that it stands in a certain relation to the color of our sensation. However, we do not know which of several colors it is. But this is not what ordinarily happens. Ordinarily, we claim to know what particular color the table has, what shape it has, etc. The table is light brown, the top is rectangularly shaped, etc. Thus we know that the skeptic’s conclusion is false. Since we accept its first premise, and since we accept the principle about knowledge by relational description mentioned above, we must reject the second premise. It is not true that we can know the color of the table only by relational description. We know it by acquaintance. We know it by sight.

b) The Appearance of Appearances

In his response to Hume’s version of the argument from the relativity of sensing, Reid speaks of the appearance of a perceptual object: “In a word, the appearance of a visible object is infinitely diversified, according to its distance and position” (T. Reid, Essays on the Intellectual Powers of Man, p. 227). Appearances have played a large role in more recent discussions of the argument and we shall now take a look at them.

The first premise of the argument from the relativity of sensing states that the table looks different from different angles and distances. It has been said that it appears to have different colors from different angles and distances. Both expressions hide the implausibility of the first premise. When you look at the table from one point of view, you see that it is light brown; and when you look at it from another point of view, you see again that it is light brown. Imagine that the first premise had been formulated in this way: When you look at the table from one point of view, you see that it is dark brown; from another point of view, you see that it is white; from still another point of view, you see that it is light brown; and so on. We would immediately object that these perceptual mistakes do not ordinarily occur. But if you say that the table looks dark brown from one angle and white from another, or that it appears to be dark brown from one point of view and white from another, then you may get away with it.

To say that the table looks dark brown from one point of view and white from another, or that it appears dark brown from one point of view and white from another, is simply false. When we say that the table looks dark brown, we mean that we see it to be dark brown, but have some doubt that it really is dark brown. As Vesey puts it: “What an object looks like to somebody is what, on looking at it, that person would take it to be, if he had no reason to think otherwise” (G. N. A. Vesey, “Seeing and Seeing as,” p. 69). In our example of veridical perception, the person would take the table to be light brown from every angle and distance.

In a classic discussion of the argument, C. D. Broad talks about a penny, lying on a table, and viewed from different positions: “We know, e.g., that when we lay a penny down on a table and view it from different positions it generally looks more or less elliptical in shape” (C. D. Broad, “The Theory of Sensa,” p. 86). It is a certain fact, he claims, that the penny does look different as we move about. I claim that this is neither certain nor even a fact. When we view the penny it does not look more or less elliptical: it looks round. And as we move about, the penny does not look different (in shape): it looks the same. Broad realizes that something is wrong with his description of the perceptual situations. He knows that a person looking at the penny would never say: “This is elliptical” or “I see that this penny is elliptical.” “When I judge that a penny looks elliptical,” Broad explains, “I am not mistakenly ascribing elliptical shape to what is in fact round. Sensible appearances may lead me to make a mistaken judgment about physical objects, but they need not, and, so far as we know, commonly do not. My certainty that the penny looks elliptical exists comfortably alongside of my conviction that it is round” (Broad, 1965, p. 87). Broad here uses “looks like” and “appears” not in the ordinary way. In his sense, you may say that the penny looks oval even when you see it to be round. What is this sense?

When you look at the penny, you see that it is round. But you also have, according to our view, an elliptical sense-impression. This suggests that Broad is talking about your sense-impression when he says that the penny looks elliptical to you or that it has an elliptical appearance. You may be reporting, not the shape of the penny, not the shape you see, but the shape of your sense-impression. On many ordinary occasions, there may be no reason for you to notice the properties of your sensations. But occasionally the shape or color of one of your sense-impressions may become important. H. H. Price gives the example of someone looking at the sun as it approaches the horizon on a winter evening. He knows of course that the sun is round, and he also sees that it is round. But he may be struck by the fact that under those conditions he experiences an oval sense-impression. And he may express his astonishment by saying: “Look at that sun, it looks oval, almost like an egg.” In short, the sensible appearances of Broad’s example are sensations and their properties: “Thus when I look at a penny from the side, what happens, on the present theory, is at least this: I have a sensation whose object is an elliptical, brown sensum; and this sensum is related in some special intimate way to a certain round physical object, viz., the penny” (Broad, 1965, p. 89). In our words, I sense (experience) an elliptical, brown sensation, and this sensation is caused by a round penny.

Broad uses the appearance terminology in order to “introduce” sense-data. This has become a favorite practice in response to a spreading rejection of sense-data. One seems to reason like this. Nobody can object to the appearance terminology; we all use it in ordinary discourse. If we can therefore introduce the sense-data terminology as a mere variant of the appearance terminology, we have secured the existence of sense-data. Or, at least, nobody can then object to the use of a sense-data terminology. But this kind of reasoning seems to me to have things backward. That there are sensations, visual sensations, auditory sensations, etc. is a fact that requires no philosophical argument. To deny that there are sensations is as silly as to deny that there are perceptual objects. Sense-data, therefore, need no philosophical introduction. They need not be made palatable to philosophers. If a philosopher cannot swallow them, the worse for the philosopher. When we look at the penny, we have an elliptical sense-impression, as everyone can prove by drawing the penny in perspective. And perhaps, just perhaps, we can also express this fact by using expressions like “looks like” or “presents an elliptical appearance” in an unfamiliar way.

But I must add a word of caution. To hold that there are sensations is not the same as to hold that there are sense-data in all of the many meanings which philosophers have given to this term. Sensations are “ordinary, everyday types of things.” Sense-data, on the other hand, are often philosophical props. How else can one explain, for example, Moore’s quite serious question of whether or not sense-data are parts of the surfaces of perceptual objects? Or how else can one explain the ridiculous claim that there are unsensed sense-data? While there can be no doubt that there are sensations, I am sure that there are no “sense-data” of that sort.

While Broad uses the appearance terminology in order to introduce sense-data, other philosophers have used it in order to introduce a special sense of “being appeared to.” They say that the penny appears elliptical to you, but deny that there is an elliptical appearance (see, for example, Roderick M. Chisholm, Perceiving, pp. 115-25). It is a fallacy, they claim, to conclude that there is an elliptical appearance, if the penny appears elliptical to you. I shall be brief and go to the heart of the matter, since this view is by now well-known. The heart of the matter is that these philosophers acknowledge acts of sensing but no sense-impressions. But to reject sensations, is as silly as to reject perceptual objects. What reasons are there in this case for denying the obvious?

It is usually maintained that there are a number of puzzling questions about sensations which do not arise for the corresponding acts of sensing (see, for example, James W. Cornman, Perception, Common Sense, and Science, pp. 78-89). For example, one claims that the question of whether one can be mistaken about the properties of one’s sensations is particularly troublesome. Cornman asks: “Does anyone fail to experience a property that one of his sensa has?” I think that the answer is straightforward and not difficult at all. As the question is phrased, the answer is negative: Nobody can fail to experience a property of a sensation which he experiences. This is a mere truism. It amounts to no more than that one cannot fail to experience what one experiences. Sensations cannot have properties which are not experienced, for a sensation cannot be other than it is. But perhaps Cornman means to raise a different and more important question, namely: Can a sensation have a property which one does not notice? But to this question the answer is clear-cut as well: Of course, it can. There is no difference between sensations and perceptual objects in this regard. To take Cornman’s example, assume that a tiger appears at the edge of the jungle a few yards away from you. He stands there for a second or two and then vanishes back into the jungle. Assume also that the tiger has exactly twenty stripes on the side which was turned toward you. Did you notice that the tiger has twenty stripes? Of course, not! You were much too excited to count his stripes. You saw a striped tiger, but you did not see that the tiger had twenty stripes. What about your visual sense-impression of the tiger? Cornman thinks that this question presents us with a dilemma. Presumably, neither the view that the sense-impression has twenty stripes nor the view that it does not is satisfactory. But there is no dilemma. Your sense-impression had twenty stripes, just like the tiger. But you did not pay any attention to your sense-datum and did not count its stripes. Thus you did not notice that it had twenty stripes. If you are really in doubt about the number of stripes of your sense-impression, you merely have to lure the tiger back out of the jungle and then pay attention to your sense-impression.

Why are we supposed to be impaled by this answer? Cornman states that “it is quite unreasonable to say that the sensum someone briefly experiences because of a quick glimpse of a tiger, has a definite number of stripes, although he does not experience it to have some definite number of stripes” (Cornman, p. 81). Notice the ‘experience’ here. We must again correct the expression: It is not unreasonable at all to hold that a sensum which is only briefly experienced should have properties which are not noticed. By now it is clear that the alleged difficulty revolves around the problem of whether or not our knowledge of sense-impressions is infallible. And it should be obvious that we come down squarely on the side of those philosophers who have held that we can and do make mistakes about our sensations. Not only do sensations have properties which we do not notice when we experience them, we can even make mistakes about their properties when we pay attention to them. We shall return to this topic in the next main section and then explain in what sense all knowledge of mental things is fallible.

Another reason why sensa are taken to be suspect is contained in the question of where sensa are spatially located (Cornman, pp. 84-85). Again, I fail to see the difficulty. It is clear that sensa are not located in perceptual space, since they are not perceptual objects. But this says nothing more, we must quickly add, than that they are not spatially related to perceptual objects. The oval sensation caused by the tilted penny is neither to the left of the penny nor to its right. Nor does it stand in any other spatial relation to the penny. But it does stand in spatial relations to other sense-impressions. For example, it may be to the left of a yellow sensum which is caused by a pencil. Visual sensations have spatial properties (size and shape) and stand in spatial relations, just as perceptual objects do. They share these spatial features with perceptual objects, just as they share colors with them.

So much for two of the alleged difficulties with sensations. It remains to emphasize how peculiar this kind of attack on sense-impressions is. One denies the existence of sensations on the sole ground that they raise certain problems. But so do perceptual objects. And so do physical objects. Should we then also deny the existence of apples and electrons, and switch to talking about “seeing applely” and “measuring electronly”?

But we have strayed from our main topic: The nature of appearances. Appearance is often contrasted with reality, and thereon hangs another interesting tale. If you mistakenly think that the penny looks oval to you just because you experience an oval sensum, then you may also make a second mistake. Calling the sensations “appearances of the penny,” you may think of the properties which the penny appears to have as properties which your sensations really have. And then it follows, as Broad puts it, “that sensa cannot appear to have properties which they do not really have” (Broad, 1965, p. 93). For, if sensa could appear to have properties which they do not have, then we would have to conclude that there are “super-sensa” which really do have the properties that the ordinary sensa merely appear to have. And this reasoning would lead to an infinite series of sensa. But we must protest: It is not the function of sensations to solve the appearance-reality problem, which consists in the fact that a penny, though round, may appear to be oval. It is false that the sensation caused by the penny always has the shape which the penny appears to have. The penny may appear to be round, even if its sensation is oval. The appearing terminology, we insist, is not ordinarily a sense-data terminology. When you look at the tilted penny, it does not appear to be oval at all. You see that it is round; you are certain that it is round. And if you were to use the word “appear,” you would say: “It appears round to me.” On the other hand, if, for philosophical reasons, you use the appearance terminology as a replacement for the sense-data terminology, then you can no longer play with the appearance-reality distinction. Then to say that the penny presents an oval appearance is to say nothing more nor less than that it causes an oval sense-impression. And then it follows that sensations cannot have appearances, but it does not follow that they cannot appear to be other than they are, in the ordinary sense of this phrase. They cannot have appearances, because sensations do not cause you to have sensations of sensations in the way in which perceptual objects cause you to have sensations. But this fact does not imply that it is impossible to make mistakes about them.

The appearing terminology, to sum up, is not a natural way of talking about sensations. Quite to the contrary. Sensations are not appearances. To believe that they are may lead one to the mistaken conclusion that we cannot make mistakes about them. Nor need we be defensive about talking about sensations. That there are sense-data is a matter of common sense. But this fact has been hidden behind the shifting clouds of philosophical confusion. If one studies recent discussions about the existence and nature of sense-data, then it becomes clear why some philosophers are tempted to escape from all that heat by jumping into the fire of the adverbial view.

c) The Causal Theory

What is “immediately given by sight,” according to the argument from the relativity of sensing, is not the real table. If the real table is known at all, then it is known only mediately, indirectly. Thus there is a gap between what we know directly, sense-impressions, and what we know only indirectly, namely, perceptual objects. I do not think that this gap can be closed and, therefore, I believe that the skeptic will triumph. But there have been attempts to close it and thus to close the door to skepticism. We shall now take a look at the most famous of these attempts, the so-called “causal theory of perception” (see, for example, H. H. Price, Perception, pp. 66-102).

Return with me to Russell in order to take a measure of his philosophical plight. Why should we believe that there is a “real” table “out there” that causes our sense-impressions? We are told that the assumption that there are perceptual objects (and that they have the properties which we think they have) is the most simple hypothesis which explains the behavior of our sense-impressions: “If the cat appears at one moment in one part of the room, and at another in another part, it is natural to suppose that it has moved from the one to the other, passing over a series of intermediate positions. But if it is merely a set of sense-data, it cannot have been in any place where I did not see it; thus we shall have to suppose that it did not exist at all while I was not looking, but suddenly sprang into being in a new place” (Russell, 1977, p. 23). I have felt since my undergraduate days that this is one of the most flimsy of all of Russell’s arguments. I “see” a bundle of sense-impressions at one place; a little later I “see” a similar bundle at a different place. Am I supposed to explain how they got from one place to the other? Of course, if I assume that there is a cat that causes my sense-impressions, then nothing is more natural than to assume that the cat that caused the first bundle of sense-impressions has moved across the room and now causes the second bundle of sense-impressions. Nor is it clear to me what Russell’s argument has to do with simplicity. A thoughtful skeptic, I think, would respond to Russell with the question: “What cat?”

But the fatal and obvious objection to Russell’s line of reasoning is that we do not know of cats by inference to the simplest hypothesis. We see the cat, first in one corner, then in another. Russell does not feel too comfortable with his explanation and adds: “Of course it is not by argument that we originally come to our belief in an independent external world. We find this belief ready in ourselves as soon as we begin to reflect: it is what may be called an instinctive belief” (Russell, 1977, p. 24). Thus it turns out that we do not, after all, infer the existence of an external world, but are born with an instinctive belief in it. But Russell, to his credit, does not feel happy with the instinct theory either. So he confuses the situation by pointing out, quite correctly, that a web of beliefs forms a hierarchy that must rest on some basic beliefs. He tailors this fact to his particular purpose: “All knowledge, we find, must be built up upon our instinctive beliefs, and if these are rejected, nothing is left” (Russell, 1977, p. 24). But our knowledge of the external world is not based on instinctive beliefs. Instinct has nothing to do with it. It is built on perception. We know that there are perceptual objects and what they are because we can see them, hear them, smell them, etc.

All right, let us assume that we instinctively believe that there are perceptual objects. But how do we know what they are like? Is there also an instinctive belief that apples are red and that elephants like peanuts? Of course not. But, then, there really are no perceptual objects in Russell’s so-called external world. “Out there,” Russell believes with the majority of philosophers and physicists, are elementary particles and the like. And these physical objects have no colors and do not eat peanuts. How does this “scientific world view” jibe with the instinct theory just mentioned? Do we instinctively believe that there are cats that move from place to place or do we instinctively believe that there are elementary particles? If we believe the former, is not our instinctive belief false? And what then happens to the system of knowledge built upon this and other such instinctive beliefs? On the other hand, how plausible is the view that we instinctively believe in electrons?

According to Russell, there are sense-data with which we are directly acquainted and there are the physical objects science talks about. The gap, we see, is between knowledge of sense-impressions and knowledge of physical objects. How is this gap to be bridged? Well, there is presumably the instinctive belief in “external causes” for our sense-impressions. But this merely assures us of the existence of a Kantian Ding an sich. Surely science claims to know more than that there are unknown causes of our sensations. At this point Russell invokes two new devices. Firstly, there is knowledge by description. Secondly, there is knowledge by isomorphism. Knowledge by description, as we have all learned from Russell, explains how we can know things with which we are not acquainted. We may know a perceptual object, for example, as the object which causes us to experience a certain sense-impression under certain circumstances. There is nothing wrong with Russell’s famous distinction between knowledge by acquaintance and knowledge by description. Unfortunately, knowledge by description is not able to close the skeptical gap, as I pointed out. Even if we grant what a skeptic need not grant, namely, that we are somehow justified in making a causal inference from our sense-impressions to “external objects,” there remains the fatal objection that we do not know the external world by causal inference. Russell, as we just saw, concedes this point.

Secondly, there is knowledge by way of isomorphism. This is a most spellbinding bridge over the skeptical gap. The basic idea is that even though we cannot infer the properties of the external world from the properties of our sensations, we can know the structure of the external world, because that structure is isomorphic to the structure of the realm of sense-data. For example, Russell says that “we may assume that there is a physical space in which physical objects have spatial relations corresponding to those which the corresponding sense-data have in our private spaces” (Russell, 1977, p. 31; my italics). And similarly for time: “Thus, in so far as time is constituted by duration, there is the same necessity for distinguishing a public and a private time as there was in the case of space. But in so far as time consists in an order of before and after, there is no need to make such a distinction;” (Russell, 1977, p. 32). And Russell concludes with these words: “Thus we find that, although the relations of physical objects have all sorts of knowable properties, derived from their correspondence with the relations of sense-data, the physical objects themselves remain unknown in their intrinsic nature, so far at least as can be discovered by means of the senses” (Russell, 1977, p. 34). Brentano, in the same vein, describes the task of physics in this fashion: “Physical science is the science which tries to explain the succession of physical phenomena of normal and pure . . . sensations by assuming that it is the effect on our sense organs of a world which is space-like extended in three dimensions and time-like running in one direction” (F. Brentano, Psychology from an Empirical Standpoint, vol. I, p. 138).

The same appeal to knowledge by way of isomorphism is used by Meinong to escape from skepticism and idealism (see A. Meinong, Ueber die Erfahrungsgrundlagen unseres Wissens). According to Meinong, there exists a noumenal world beyond the phenomenal world of our mental phenomena. This noumenal world is similar to the phenomenal world in that it contains things and their properties. In addition, however, these noumenal entities stand in the same relations (“relations of comparison”) of difference, similarity, and equality in which their phenomenal counterparts stand. One can therefore surmise, for example, that if two sensations have similar properties, then their noumenal counterparts have similar properties. Furthermore, number is a projectible feature: If we hear two tones, then there correspond to these sensations an equal number of noumenal objects in the noumenal world.

What principle distinguishes between projectible and nonprojectible features? Russell does not give a straightforward answer, but Meinong does:

It is the wide, even unlimited, field of application of those two kinds of ideal object [relations of comparison and numbers]. One cannot think of a pair of objects which are not different in the extended sense just used; as little [can one think] of objects which do not form a complex with a determinate number of parts. Under such conditions, an application to what is not phenomenally given, that is, to noumenal objects, is naturally completely unobjectionable. The situation is quite different for complexes like melodies which are only applicable to tones. . . .

(Meinong, p. 103)

Any two objects of the phenomenal world are different from each other, and any kind of object of this world can be numbered. Therefore, Meinong argues, these features must also occur in the noumenal world. What is projectible are the features which every possible phenomenal world must have. The noumenal world may have quite a different “content” from the phenomenal world, but it must share its categorial structure.

But what reason is there for believing that the two worlds must share the same categorial structure? From the fact that something is a categorial feature of the phenomenal world it simply does not follow that it must also be a categorial feature of the noumenal world. Nor can we argue that we can only conceive or imagine the noumenal world in terms of the categorial features of the phenomenal one and that it, therefore, must have these features. It may well be the case that the noumenal world, if there is one, is beyond our conception and imagination. I think that we are left with Russell’s “instinctive belief”: The only reason we can have for believing that the noumenal world must be isomorphic to the phenomenal world is blind instinct, and that of course is no reason at all.

Notice, finally and most importantly, how badly this talk about a noumenal world behind the mental world jibes with the professed belief in the world of physics. Compared to the world of physics, Russell’s and Meinong’s “external world” is as bloodless as a skeleton. Open a textbook on physics and you will be impressed and perhaps even enchanted by the colorful properties elementary particles are said to have. No uncertainty here. No ignorance. We know, within the bounds of scientific certainty, what the features of the physical world are. The “external world” of Russell and Meinong, it turns out, is not the world of physics at all. If there were such a world, then it would have to transcend the world of physical objects. We have just witnessed a most amazing turn of events: The causal argument and the thesis of isomorphism do not bridge the gap between the mental and the physical realms, but lead, at best, from the mental world to a strange world of noumenal objects.

The real world, we are told, is not the world of perceptual objects, not the world of apples and colors, but the world of physics. In the name of science, the perceptual world must be exorcised. We protested that such a drastic measure is unnecessary, that apples and elementary particles dwell harmoniously in the non-mental world. But in vain. And then something incredible happened. Philosophy raised its wise old head and introduced the argument from the relativity of sensing. Suddenly, those who insisted that the “external world” is the world of physics found themselves at a total loss: knowledge of the physical world becomes impossible. The very best they can come up with is an instinctive belief in a noumenal world. The best they can do is to postulate an isomorphic world of noumena. Thus the physical world is lost to them as well. “Out there,” in the end, are neither apples nor elementary particles. “Out there” are unknown things with unknown features.

But we know better. We can explain how we know of elementary particles and their properties. We do not lose the world of physics for the shadowy world of noumenal objects, because for us the physical world is firmly tied to the perceptual world. Is it not ironic that those who believe that they have to choose between the perceptual world and the physical world end up with neither? What shall it benefit a philosopher to dump the world of perceptual objects into the mind, if he will lose the fascinating world of physics?

d) “Parts of the Surfaces of Inkstands”

Russell’s version of the argument from the relativity of sensing concerns the color of a table. What holds for color, holds for shape and other properties as well. The true properties of the table are presumably hidden behind a veil of appearances. But this is not the only curtain that shields the table from the mind. Even if we could see its color and its shape, it is argued, we could not see the “whole table,” but only a spatial part of it. The table, according to some philosophers, is hidden from mental view in a complicated twofold manner. We cannot perceive it, but only parts of it; and these parts are not really perceived either, but are hidden behind appearances (sensations, sense-data). This, in my opinion, is the gist of Husserl’s famous theory of aspects. He describes the perspectible feature of perception in these words:

It is not an accidental caprice of the Thing nor an accident of “our human constitution” that “our” perceptions can reach the things themselves only and merely through their perspective modifications. On the contrary, it is evident, and it follows from the essential nature of spatial thinghood (and in the widest sense inclusive of “visual illusion”) that Being of this species can, in principle, be given in perception only by way of perspective manifestation.

. . . it is an essential necessity to be apprehended as such with apodeictic insight that spatial Being in general can be perceived by an Ego actual or possible only when presented in the way described.

(Husserl, 1962, par. 42)

I shall not discuss Husserl’s theory of aspects, but consider Moore’s version of this theory. Moore’s discussion is interesting for two reasons. Firstly, he connects the perspectible feature of perception with the problem of how to close the skeptical gap. He tries to convince himself that sense-impressions are parts of the surfaces of perceptual objects, namely, of those parts that are turned toward the perceiver. He may have asked himself: Since a spatial aspect of a perceptual object is obviously not identical with the object, why not identify this aspect with a sense-datum? Secondly, it is instructive and humbling to observe how the great champion of common sense, once he has strayed from the path, takes one wrong philosophical turn after another. Here is a paradigm of how a most cautious and astute philosopher can arrive at an absurd conclusion by starting from what he considers to be obvious and certain premises. Whenever I am inclined to believe that a line of reasoning in this book runs smoothly and in the right direction, I remind myself of Moore’s paper, “Some Judgments of Perception.”

Moore considers the perceptual judgement, This is an inkstand, and claims that it seems to him “so very certain” that:

in all cases in which I make a judgment of this sort, I have no difficulty whatever in picking out a thing, which is, quite plainly, in a sense in which nothing else is, the thing about which I am making my judgment; and that yet, though this thing is the thing about which I am judging, I am, quite certainly, not, in general, judging with regard to it, that it is a thing of that kind for which the term, which seems to express the predicate of my judgment, is a name.

(Moore, 1970b, p. 229)

In plain words, when Moore asserts that this is an inkstand, he is not asserting that what the ‘this’ represents is an inkstand. He is not making this judgment, because the this of his judgment is not an inkstand. Far from being “so very certain,” this claim is plainly false. What Moore asserts is either true or else it is false. Now, he obviously means for his judgment to be true. But if it is true, then the this of his judgment must be an inkstand. How, otherwise, could the assertion that this is an inkstand be true?

Why is Moore so certain that the this of his judgment is not an inkstand? This is where perspective enters into the considerations: “Nobody will suppose, for a moment,” Moore says, “that when he judges such things as This is a sofa,’ or ‘This is a tree,’ he is judging, with regard to the presented object, about which his judgment plainly is, that it is a whole sofa or a whole tree; he can, at most, suppose that he is judging it to be a part of the surface of a tree” (Moore, 1970b, p. 230; my italics). Once again, in plain English, since Moore sees only part of the sofa, the this of his judgment cannot be the whole sofa, but must be the part that he sees. What sets Moore’s tortuous train of thought in motion is the assumption that he cannot possibly see a “whole” perceptual object, but can see only a part of it.

Next, Moore identifies the part of the surface of the inkstand which he sees with a sense-impression: “The object of which I have spoken as the object, about which, in each particular case, such a judgment as this always is a judgment, is, of course, always an object of the kind which some philosophers could call a sensation, and others would call a sense-datum” (Moore, 1970b, p. 231). Notice the “of course”! Hardly any philosopher, and certainly not I, would call a part of the surface of a perceptual object a sensation. Any temptation to call it a sensation would be immediately quenched by the realization that this would turn surfaces of perceptual objects into mental things.

Moore, then, is quite certain (1) that the this of his judgment is not the inkstand before him, (2) that it is part of the surface of the inkstand, and (3) that this part of the surface is a sensation. I think that all three of these propositions are false.

What might have moved Moore to assert the near-contradiction that in the true judgment This is an inkstand, the this is not an inkstand? The little word ‘whole’ in ‘whole inkstand’ is the clue. Moore seems to reason that the this cannot be the whole inkstand because he does not see the whole inkstand, but does see the this. Why does Moore believe that he does not see the whole inkstand? A whole, of course, has parts. The whole inkstand has many spatial parts. Moore, I surmise, adopted the principle that one cannot be said to have perceived a whole perceptual object unless one has perceived all of its parts. Listen to Kant’s version of this principle: “When we see a house in the distance, we must necessarily have a representation of the different parts of the house . . . For if we did not see the parts, we would not see the house either. But we are not conscious of this presentation of the manifold of its parts” (I. Kant, Logik, p. 34).

It cannot be denied that Moore cannot see the back of the inkstand as long as he looks at it from the front. It is therefore true that he does not see all of the spatial parts of the inkstand when he looks at it from the front side. Thus if the just announced principle were true, it would follow that we cannot see (whole) inkstands (in one glance, from one point of view). It would be true, as Husserl puts it, that “our perceptions can reach the things themselves only and merely through their perspective modifications.” But the principle is false: perceptual objects are presented to us “directly” and not through their parts or perspective modifications. The fault lies with the phrase “seeing a whole inkstand.” I suppose that one may only be said to have seen the whole inkstand if one has walked around it, looked at it from all sides, and looked inside as well. In this sense, Moore has not seen the whole inkstand as long as he looks only at its front. But it does not follow from this fact that what he sees is part of the surface of an inkstand. The expression could also mean something else, namely, something like “not just seeing part of an inkstand.” Assume that an apple has been cut into halves and that you see one half of it before you on the table. In this case, what you see is not the whole apple, but merely a part of it.

To sum up, I have argued that just as we see the properties of perceptual objects “directly,” so do we see the objects themselves “directly,” and not through or by means of their spatial parts. My argument rests on a rejection of the principle that one can only see a perceptual object if one sees all of its parts.

My analysis applies not only to spatial parts, but also to temporal parts of perceptual objects. Broad, after he has concluded that “The perceptual situation contains as a constituent something which is in fact part of the surface of the bell,” goes on to claim in the same vein that “at most we can say that it contains as a constituent a short event which is in fact a slice of a longer strand of history, and that this longer strand is the history of a certain bell” (C. D. Broad, The Mind and its Place in Nature, p. 149). But just as we do not see part of the surface of the inkstand when we see the inkstand, so do we not see a temporal slice of the bell when we see the bell. And this for the same reason. The bell has temporal parts; it has a temporal history. But to see the bell is not the same as to see these temporal parts. In the spatial as well as in the temporal case, a certain type of argument is at work:

(1) An inkstand has (consists of) many spatial and temporal parts.

(2) Therefore, to see an inkstand is to see all of these parts.

(3) But we cannot see all of these parts (from a point of view, at a moment).

(4) Therefore, we cannot see the inkstand.

I have said that premise (2) of this argument is false. We can now see clearly why it is false. Assume that the inkstand has a scratch at the bottom, and let us express this fact awkwardly by saying that the inkstand has the property of having a scratch. Now, it is clear that from the fact that the inkstand has the property, it does not follow that you see that it has this property when you see the inkstand. For example, when you see that the inkstand is half full of ink, you do not see that it has a scratch at its bottom. The description “the half full inkstand in front of Moore’ describes the very same thing as the description ‘the inkstand in front of Moore with a scratch at its bottom’. But it does not follow from this fact that the two expressions can be substituted for each other salva veritate in all contexts. The inkstand before Moore is identical with the inkstand consisting of such and such spatial parts. But it obviously does not follow that when you see that the inkstand before Moore is half full, you see that the inkstand with such and such spatial parts is half full. In short, the principle adopted by Moore for spatial parts and by Broad for temporal parts is false, because the so-called principle of extensionality is false (see my The Categorial Structure of the World, pp. 375-380). Formulated linguistically and tailored to our case, the principle states that description expressions which describe the same thing can be substituted for each other salva veritate in all contexts. The principle of extensionality is so obviously false that one can only wonder why it has ever been adopted. Surely, when I believe that the earth is round I do not ipso facto believe that two plus two equals four, even though the sentences ‘the earth is round’ and ‘two plus two equals four’ are both true. Similarly, for the case illustrated above, when two expressions describe the same thing. It must be emphasized that the principle of extensionality is not the same as the Leibnizian law of substitutivity, according to which expressions which represent the same thing can be substituted for each other salva veritate. Perhaps the principle of extensionality has sometimes been accepted because it was confused with the law of substitutivity.

If Broad were correct in that the “most that we can grant is that a small spatio-temporal fragment of the ontological object is literally a constituent of the situation,” then we can construct another skeptical argument, similar to the argument from the relativity of sensing (Broad, 1960, p. 150). In a nutshell and in Broad’s terminology, how do we come by the conviction that this fragment is “not isolated and self-subsistent . . .; but that it is spatio-temporally a part of a larger whole of a certain characteristic kind . . .”? (Broad, 1960, p. 151). Broad sees clearly that we do not come by this conviction through inference, and he also admits that it cannot be justified by inference (Broad, 1960, p. 151-152). What is left, then, is the skeptical conclusion that this conviction is nothing more than “instinctual belief.” But we know how to avoid this conclusion. We have defended the view that, contrary to Moore and Broad, we see perceptual objects and not just spatio-temporal parts of them.