Fall 2002

Theory of Knowledge Perception and Sense-Data

Dicker begins his discussion with the question: "Suppose you perceive x as F. What more is required for you to know that x is F?"

(Notice we're taking up questions of knowledge again. For a while there, we had just been focusing on questions about justification. Some discussions of perception focus on knowledge and others focus on justification.)

The first proposal Dicker considers is:

Dicker argues that this proposal won't work. Suppose you perceive everything as red. Now you look at a wall, that in fact happens to be red. You wouldn't be able to know that it's red, just because it looks red to you. Everything looks red to you. Yet Proposal 1 says you would be able to know that the wall is red. So Proposal 1 is wrong. (See Dicker p. 14.)

Next Dicker considers a more sophisticated proposal:

This proposal doesn't require you to know that the perceptual conditions are normal. It's enough if they are normal. (Compare the intermediate stance we discussed earlier: it just has to be true that your perceptual beliefs lack certain defects. You don't also need to have evidence that they lack those defects.)

Dicker considers the following objection to Proposal 2.

Suppose your system for detecting temperature by touch is like a thermometer that works only intermittently: sometimes it tells the right temperature, other times it's way off. Then you won't be able to rely on your temperature-detecting system, even when it happens to be working properly, unless you have some independent way of knowing that it's working properly at that time. So you won't be able to know by touch what temperature the water is, even if your temperature-detecting system is working properly right now, and so, even if at the moment, the perceptual conditions are normal.

If this objection is right, then Proposal 2 has to be rejected. The perceptual conditions are in fact normal, and you perceive the water as hot, but you don't thereby acquire knowledge that the water is hot. (See Dicker pp. 18-20.)

Dicker discusses replies to this objection; then he presents a more difficult objection, which he takes to be fatal to Proposal 2. Dicker's objection is this:

Suppose you don't have any independent reason for believing that the perceptual conditions are normal. Then they might be normal, or they might not. You have no way of knowing. If they really are normal, then it would be somewhat "lucky" or "accidental" that this is so. (Recall BonJour's Accident Argument.) Therefore, Dicker argues, it would also be somewhat "lucky" or "accidental" that this case of perceiving x as F is one where x really is F. And you don't count as knowing that x is F if it's just an accident or matter of luck that you're right about x's being F. (Recall the appeal to the notion of "accidents" in our earlier discussion of the Gettier Problem.) So in cases like these, Dicker thinks, you would not count as knowing that x is F. In order to know that x is F, it's not enough that the perceptual conditions happen to be normal. You also need to have evidence that they're normal. (See Dicker pp. 21-22.)

You can see here that Dicker is arguing from a conservative framework, like we saw BonJour arguing earlier. Dicker thinks that it's not enough for you to lack evidence that the perceptual conditions are unfavorable. It's not enough for the perceptual conditions to in fact really be favorable. If you're going to know that x is F on the basis of perceiving it as F, you also need to have positive evidence telling you that the perceptual conditions are normal.

Nothing less will do. This is an example of the conservative stance we were discussing earlier.

Well, at this point a problem naturally arises. How do you find out whether your perceptual conditions are currently normal? (Obviously, you can't just say "By perceiving them to be normal." If you said that, the question would just re-arise: How do you know that the perceptual conditions were normal when you perceived your current perceptual conditions?)

The threat of skepticism looms very large here. If we accept Dicker's Proposal 3, then we can't know anything on the basis of perception unless we have independent reasons for believing our perceptual conditions are normal, but how are we supposed to know that, if not by perception?

Sense-Data

At this point Dicker introduces the notion of sense-data. These are supposed to be aspects of your experience that you can know about no matter how unfavorable the conditions for external perception are. For instance, when you partially submerge a straight stick in water, it will look bent. In such a situation, you won't be able to know by perception that the stick is bent. (It isn't bent.) But the Sense-Data Theorist says that there is some shape you are aware of which is bent. This is your visual sense-datum of the stick. You can know that the sense-datum or visual appearance of the stick is bent, even if you're uncertain whether or not the stick out there in the world is bent. Similarly, if the wall looks red to you, you may not be sure whether the wall really is red, or whether it's just illuminated by red lights. In such a situation, though, you can know that your sense-datum or visual appearance of the wall is red.

So if we believe in these sense-data, they will be things we can know about in every case of perception, regardless of how favorable the external conditions of perception are, or are known by us to be.

[ Consult Russell here? Armstrong, Feldman MS on question whether we're infallible about our own mental states? ]

There is a notorious argument concerning sense-data, which goes by several names. Sometimes it is called The Argument from Illusion. Dicker calls it The Argument from Perceptual Relativity. That argument goes like this.

Consider one of the situations where the external world isn't exactly the way it appears to you to be. Some of these cases count as illusions: there really is something there that you're perceiving, you're just misperceiving some of its properties. For example, when the stick is partially submerged in water, it looks bent but in fact it's still straight. If you look at a circular plate tilted on its side, the plate may look elliptical to you. It looks elliptical but in fact it's round. If there's tricky lighting, or you're wearing special contact lenses, the wall looks red to you but in fact it's white. And so on. Other cases count as hallucinations: there isn't anything out there in the external world that you're perceiving, it's just all made up. For instance, Macbeth hallucinates a dagger floating in front of him. You might want to count dreams and being in the Matrix as a kind of massive hallucination.

In cases where the world is the way it appears to you to be, we say that your experiences of the world are "truthful" or veridical.

Some cases of perception are hard to classify. For example, the situation where you're focusing on a distant object and it looks to you like you're holding up two fingers in the foreground, but in fact you're just holding up one finger. Here your experiences are clearly not veridical. There look to be two finger-shaped objects there, but in fact there is only one. But it's not clear whether we should count this as an illusion or a hallucination. (To say it's an illusion is to say that you're perceiving your finger, but just misperceiving it as having some properties that in fact it lacks. What might those properties be? "Being in two places at once"? Perhaps it's better to say that one of the finger-shaped objects is a hallucination. But if so, which one? Are both of them hallucinations? That doesn't seem right; after all, we don't want to deny that you're perceiving your finger. The problem is that you're perceiving it twice. Maybe we need a special category for cases like this.)

In any case, consider some case where your experiences aren't veridical. Some case of illusion, or hallucination, or seeing double. For the sake of example, I'll talk about the stick that appears bent when in fact it's straight.

The Argument from Illusion has two stages. Stage I goes as follows.

1. The stick appears bent.

2. So there is some bent shape in your visual field, which you're aware of.

3. But the real stick you're looking at, out there in the physical world, isn't bent. It's straight. (And if you were hallucinating a bent stick, there wouldn't be any real object you're seeing.)

4. So the bent shape you're aware of can't be identical to the real stick. The bent shape is not a physical object. It's something mental. We can call it your "sense-datum" of the stick.

A Sense-Datum Theorist is someone who believes in these mental objects or sense-data. On his view, these sense-data are what we're really, or most directly, aware of in cases of illusion or hallucination. To figure out what the stick's real physical properties are, we have to extrapolate from the appearances or sense-data we have of it. (For instance, we can look at the stick from a number of different angles, feel it with our hand, and so on, to figure out what its real shape is.)

In Stage I, the Argument from Illusion is talking about cases of illusion and so on. It's trying to argue that in those cases, there are these things, sense-data, that really have the properties that external objects seem to us to have, and that are what we're really or most directly aware of. Stage II of the argument tries to extend this result to cases of veridical perception, as well.

From Stage I, we have:

5. In cases of illusion and so on, what we're directly aware of aren't real physical objects, but only the sense-data or appearances those objects produce in us.

The argument continues:

6. From the inside, cases of veridical perception seem to be just like cases of illusion or hallucination.

7. So what we're directly aware of in cases of veridical perception must also be sense-data. If what we were directly aware of in the one case were things in our own mind, and in the other case were physical objects in the external world, you'd think we'd notice some difference.

The Sense-Datum Theory was a very popular way of thinking about perception until the 1960s or so. It always had some detractors, but it gives us a very neat, powerful model of how perception works. You look at a stick. The stick causes your brain to be affected in certain ways, and this produces a sense-datum or image of the stick in your mind. What you're aware of is just this sense-datum. It may or may not correspond to the way the stick really is. For instance, the sense-datum may be bent while the stick is really straight. Many philosophers thought that although our sense-data have colors, science shows us that nothing in the external world is really colored. That's just a feature of our own sense-data. And so on.

But the Sense-Datum Theory always had opponents, too; and nowadays, it's out of fashion. There were a number of worries about sense-data that prompted this.

One worry is about where these sense-data are supposed to be located. They seem to have some kinds of spatial properties. For instance, one of my current sense-data is to the right of another one of my sense-data. But sense-data don't seem to be located in physical space. So do I have some special "mental space" where all my sense-data reside?

Another worry is whether we mean the same thing when we say that a sense-datum is "straight" or "round" or "big" or "red" as we do when we say that a stick is straight, or that an apple is round and red, and so on. When I look at an elephant, it appears to be very big to me. When I try to push it, it feels very heavy. Does that mean that my sense-datum really is that big and heavy?

A third kind of worries has to do with the relation between sense-data and our brains. Are the sense-data just part of our brains? (This question is connected to the previous question: how can my sense-datum be as big as an elephant, if it's part of my brain and my brain isn't that big.) Are they produced by our brains? Does believing in sense-data commit you to some kind of mind/brain dualism?

A fourth kind of worry has to do with whether sense-data can have properties other than the ones they appear to us to have. (At first glance, you'd think not. The whole idea of sense-data is that they're supposed to have exactly the properties they appear to have, and no more. But there are problem cases here which make this issue hard to decide.)

Most importantly for our present purposes is the way that philosophers nowadays view the Argument from Illusion. The general consensus these days is that, if you construe this argument as an attempt to prove the existence of sense-data, the argument is question-begging. Recall the first two steps of the argument:

1. The stick appears bent.

2. So there is some bent shape in your visual field, which you're aware of.

Here it is simply assumed, without argument, that you are aware of some thing which really does have the property that the stick appears to you to have. This assumption can be rejected.

Opponents of the Sense-Datum Theory would say, instead, that you merely seem to be aware of a bent shape. They deny that there is anything which is really bent, anywhere, whether in physical space or in some private "mental space," for you to be aware of. Macbeth merely seems to see a dagger; there is no dagger, not even a mental dagger, which he really does see. In the seeing double case, there aren't really two finger-shaped things anywhere, which you're successfully and correctly perceiving. You merely seem to be aware of two finger-shaped things. In fact, there is only finger-shaped thing. And it's an ordinary physical object. Not some special mental object or sense-datum.

That's what the opponents of the Sense-Datum Theory say. It would only be if you already bought into the existence of sense-data, mental objects which really do possess the properties that external objects seem to us to possess, that you would find the step from (1) to (2) in the Argument from Illusion plausible. It's only people who already believe in sense-data who will accept this move. If you're dubious about sense-data, then you will think this move is illegitimate. If you're dubious about sense-data, you won't think it follows from the fact that something appears F to you that there is any object which really is F.

So the Argument from Illusion relies crucially on a move that would only be accepted by people who already believe in sense-data. That is why, if this argument is supposed to prove the existence of sense-data, it begs the question.

Dicker says at the end of his Chapter 1 that we shouldn't think of the Argument from Illusion (or as he calls it, the Argument from Perceptual Relativity) as trying to prove that there are sense-data. He writes:

Perhaps this is a good way to understand what's driving the fans of the Sense-Data Theory. Dicker thinks that if we don't introduce sense-data, then we won't be able to know anything by perception, since (he thinks) knowing things about the external world by perception requires having independent reason to believe that your perceptual conditions are "normal" or favorable: that is, that your senses are reliable, there's no tricky lighting, and so on. It's totally unclear how we could get that kind of knowledge about our perceptual conditions. At least if we have sense-data, perception will give us knowledge about them. Then we'll have our foot in the door. Maybe we'll be able to build up to knowledge of the external world, using our knowledge of sense-data as a starting point.

[ Here discuss Armstrong, Feldman MS on whether we're infallible about our own mental states? ]

Notice that this whole story turns on accepting Dicker's conservative view about what's required to have perceptual knowledge of the external world. BonJour and Dicker both think that your perceptual belief that P can't count as knowledge unless you also have independent reason to believe that this belief was formed reliably, in favorable perceptual conditions, and so on.

I want to spend the last few classes exploring whether we might give up that conservative view. We've already looked at the intermediate, externalist view. Let's start thinking more seriously about the liberal view. This view can agree with the conservative view that whether you're justified depends only on "internal" factors. But it says that all you need, to be justified in believing P on the basis of perception, is that evidence of defects in your belief be lacking. It's enough if you lack evidence that your visual system is unreliable, if you lack evidence of tricky lighting, and so on. You don't also need positive evidence that your visual system is reliable, that the lighting is plain white light, and so on.

[Theory of Knowledge] [Syllabus] [Notes and Handouts] [James Pryor] [Philosophy Links] [Philosophy Dept.]