20th World Congress of Philosophy Logo

Philosophy of Mind

In Defense of Direct Perception

Robert Hudson
Concordia University

bluered.gif (1041 bytes)

ABSTRACT: My goal in this paper is to defend the claim that one can directly perceive an object without possessing any descriptive beliefs about this object. My strategy in defending this claim is to rebut three arguments that attack my view of direct perception. According to these arguments, the notion of direct perception as I construe it is objectionable since: (1) it is epistemically worthless since it leaves perceived objects uninterpreted; (2) it cannot explain how perceived objects are identified; and (3) it is ill-prepared to assign objective content to perceptual states.

bluered.gif (1041 bytes)

What is involved in the claim that one directly perceives an object? The notion of direct perception that I propose to defend in this paper is this: that one 'directly' perceives an object if one's perception of this object is not mediated by beliefs. Put another way, a 'direct' perceiver does not believe anything about an object in (directly) perceiving it. On this construal of the notion of direct perception, it follows that if one directly perceives an object, one does not describe this object; for any description of an object is expressed as a belief, and direct perceptions do not involve beliefs. The direct perceiver, I claim, does not (and indeed may be completely unable to) give a description of the perceived object, without this lack (or inability) detracting from the fact that the object is directly perceived.

In defending this view of direct perception, we need to become clearer on how it is possible for a belief to mediate one's perception of an object. There are (at least) two ways in which this can occur. Here's the first.

A belief can be said to mediate one's perception of an object if a belief serves as a premise from which one deduces that an object is perceived. That is, in a perceptual situation, a belief mediates one's perception of an object if one infers that one is perceiving this object on the basis of some belief or beliefs that one has. Clearly, this is a form of what many people would call indirect perception. It includes such cases as perceiving something on the basis of, or by means of, perceiving something else. And, I submit, if this is what it means for a belief to mediate one's perception of an object, it follows that there are, in fact, such things as direct perceptions, and that the object of a direct perception is an undescribed object. For, to begin with, any chain of inferences has to start somewhere; and, eventually, there has to be a belief, state of mind, or state of the world, of which one is aware, without one's having been aware of some other belief, state of mind, or state of the world beforehand. Such a belief, state of mind, or state of the world is directly, that is, noninferentially, perceived. Furthermore, suppose that the object of a direct (noninferential) perception is a described object. Since, once again, any description of an object is expressed as a belief, and since we are assuming here that the way for a belief to mediate one's perception of an object is by means of an inference, this means that the belief expressing the description is playing an inferential role in perceiving the object—which contradicts our assumption that the object of perception is uninferred. And so, by reductio, the object of a direct (noninferential) perception is an undescribed object. Thus, if all one means by the directness of perception is its uninferredness, then there's little doubt that there are direct perceptions, and that if one directly perceives an object, one does not describe this object.

However, a belief can mediate one's perception of an object in another sense. On this other sense, although no belief is needed as a premise from which one infers that one is perceiving an object, a belief is still needed in an auxiliary capacity in the context of a perception in so far as one needs to have a belief that to some extent describes the perceived object. Clearly, if mediation in this new sense is unavoidable in a perceptual situation, then the notion of direct perception I favor is untenable. Thus, I propose to defend my notion of direct perception in the following way: I shall examine three reasons why one might claim that perception is mediated in the way suggested; and, by rebutting each of these reasons in turn, I hope to show that my view of direct perception, according to which directly perceived objects are not described objects, is on philosophically sound footing.

Here, then, are three reasons supporting the claim that perception needs to be mediated by descriptive beliefs.

Reason (1): for every perception, there is an attendant, descriptive belief because a perception devoid of such a belief lacks epistemological value—in short, it cannot generate any information about its object. We need, at this point, to consider two senses of the phrase 'epistemological value'. On the first sense, the stated worry is exactly right; and if this were the only sense of 'epistemological value', we would have a strong, epistemological argument against the claim that perceptions are unmediated by descriptive beliefs, and so a strong, epistemological argument against the claim that the objects of perception can ever languish undescribed. The first sense of 'epistemological value' is this: a perception has epistemological value if it is capable of generating true beliefs about its object. And, to be sure, in a situation where one perceives an object without the presence of any attendant, descriptive beliefs about this object, one perceives this object without any attendant, true, descriptive beliefs about this object. So, if we are committed to the claim that perception has epistemological value, and if we believe that it has epistemological value in the sense described above, then perception must be mediated and indirect, and all perceived objects must be objects one is capable of describing.

But there is another sense of epistemic value we need to consider. It is the epistemological value found in any process (be it cognitive or not) that tracks unerringly the presence of a particular object. A process with such value, in the best case, always signals when a particular object, x, is present, and never signals when x is absent. Such a process is clearly an excellent tool with which one can learn about this object. And there is no reason why a perceptual process cannot have this value even if x reposes undescribed for the perceiver. Many inanimate objects, such a thermometers, do an excellent job of tracking particular objects (in this case, a temperature level) without the aid of descriptions. This is not to say that descriptions exercised by perceivers hinder the production of this sort of epistemic value. The point is that descriptions are not needed for such value to accrue.

Thus, our conjecture—that for every perception there is an attendant, descriptive belief because without such a belief, without, that is, a description for the object of perception, perception cannot provide us with information about its object—has proved faulty. Perception freed from descriptive beliefs can provide us with the information that a particular object or kind of object is present, albeit not elaborating to the perceiver what the nature of this object, or what the nature of this kind of object, is.

But, one might respond here, what epistemic value could a process have if, despite the fact that it unerringly tracks a particular object or kind of object, the perceiver cannot say what this object is? Consider, as a case in point, a situation where one wants to test one's system of beliefs by exploring its consistency with the results of perception. If perception is to have any significant epistemic value, surely it must be suited to accommodate this task. Yet direct, uninterpreted perception, as I have construed it, cannot serve in this role since there is no way that perceptions, disaffiliated as they are with beliefs, can be compared with or impact on the constituents of one's system of beliefs (except, perhaps, in the vaguest of ways, as in Quine's web of belief model).

This objection is easily disarmed. For the object of a direct perception is, I claim, an undescribed object. But that is not to say it is an indescribable object. In all likelihood, the object of a direct perception can be described, and so can participate, in this way, in the commerce of confirmatory discourse. The only point in saying that it's an object of direct perception is to say that it's perceived without the benefit of descriptive beliefs. Whatever has been so perceived can, subsequently, I don't doubt, be the object of a descriptive belief, and so can subsequently bear truth-valuably on other beliefs.

Here is a second reason why one might claim that, for every perception, there is an attendant, descriptive belief: (2) the perceptual identification of an object is inexplicable without the aid of a belief describing this object. For instance, if I say that I see my watch, and, when asked what my watch looks like, I am either unable to give any description of it or I give a patently false description of it, one might well conclude that I don't see my watch, after all. The problem here, I take it, is that if a perceiver is unable to describe a perceived object to some degree, there is no explanation for how the perceiver succeeds at identifying the object when the object passes through its perceptual field, whereas if the perceiver can describe a perceived object, then the question of how one can explain its ability to identify the object is straightforward.

Now, when one describes a perceived object, one lists a number of properties that the object has, which properties must themselves be perceived. One wonders: does the perception of these properties also involve the mediation of descriptive beliefs? It is clear that at some point there must be a way to perceive properties—or, more exactly, to perceive objects possessing properties—without the mediation of descriptive beliefs, on pain of there being a regress. For example, if the perception of object O involves a description containing properties P1, P2 and P3—or, more exactly, a description involving objects O1, O2 and O3 with these properties—and if O1, O2 and O3 are themselves perceived in ways that involve descriptions, and couldn't be perceived otherwise, then, once again, O1, for example, will involve further properties P4, P5 and P6 and further objects O4, O5 and O6. Clearly this process will continue until some objects are perceived without the benefit of descriptions. Now, why is the prospect of such a regress a problem? It is a problem if one wants to explain how one identifies a perceived object. For there is no complete explanation here if explaining such a perceptual identification involves the use of a description, which itself involves further perceptual identifications and further descriptions, and so on. On the other hand, there is a complete explanation (of sorts) if at some point the recourse to additional descriptions is averted, which is the case if direct perceptions are possible. So, if one is serious about the problem of how to explain the identification of perceived objects, then one is eventually going to have to face the prospect of perceptions unmediated by descriptive beliefs. Nevertheless, suppose one insists that the perceptual identification of objects must involve the use of descriptions. The required description, presumably, involves noting that the perceived object possesses a property or properties not shared by other objects in one's perceptual field. For instance, I can perceive a lamp because I can recognize it as possessing properties distinct from the properties possessed by a desk, a chair, table, and so on. But let's simplify the case to where a perceiver P's perceptual field contains only two objects, in particular, two identical lamps, which the perceiver labels A and B. Side by side, perceiver P can describe a difference between the two—say, A is to the left of B. Thus, on the account we are criticizing, P can perceive A because P can recognize it as possessing a property not shared by B—to wit, "being to the left of a lamp". Now, let the lamps be shuffled without the perceiver knowing, after shuffling, which is which. And, let the lamps return to their original positions. Does the perceiver now see lamp A to the left of B? In this post-shuffling situation, P no longer recognizes the lamps P had originally labelled. In particular, P doesn't recognize A as possessing the property, "being to the left of a lamp". Yet surely it is correct to say that P sees A to the left of B. For the perceptual situation here, following the shuffling, is both phenomenologically and actually identical to the situation before the shuffling; and before the shuffling we found it acceptable to say that P saw A to the left of B. So why not say this now? P certainly doesn't see B to the left of A. Nor does P see "neither A nor B" on the left, for the lamp on the left is either A or B. Consequently, if we agree that, in the post-shuffling situation, P perceives A to the left of B, it follows that P's ability to perceive A to the left of B is inexplicable if the explanation for this ability involves the use of identifying descriptions.

Let me turn to a third reason why one might claim that, for every perception, there is an attendant, descriptive belief: (3) given that perceptions have objects, what possibly can be the object of a perception where this perception is not associated with any beliefs? That is, how could a direct perception be a perception of something (leaving aside the question of how the perceiver might identify it), if possessing ofness is a cognitive achievement? The requisite ofness here is rudimentary, given the presence of descriptive beliefs enhancing one's perception. For such a perception is of whatever satisfies the attendant description. On the other hand, because direct perceptions are devoid of descriptions, the question of their objective content remains. Of course, one might interpret direct perceptions and so grant them objective content. But surely, if this were the only way direct perceptions could have objective content, the cited objection would be affirmed.

In response, I'm not entirely clear why anyone would consider this to be a serious problem for direct perceptions, for their ofness is easily found relative to the ofness of indirect perceptions. Why? Well, to begin with, any perceptual state will involve a perceiver and a perceived object. And with direct perception, the object of perception is whatever this perceived object is. On the other hand, with indirect perception—in particular, in a case where one perceives an object through the use of various descriptive beliefs --, the object of perception is ambiguous: it could be the directly perceived object or it could be the object that satisfies one's descriptive beliefs. For instance, a perceiver may see an object through the description, "the green lamp", where this object is neither a lamp nor green, though it appears as such. What, in this case, does the perceiver see—a green lamp, or something that's not a green lamp? It can't be both. But, then, it can't be neither. If the descriptive beliefs are true regarding the directly perceived object, then, fortunately, this ambiguity vanishes with veridicality—in our example, the perceiver would be said to see a green lamp. In other cases, however, cases where the attendant description is false, there is a serious problem in saying what the object of perception is. Thus, I'm a bit puzzled by any concern one might have regarding the ofness of a direct perception. There is no particular problem here, so far as I can see, leaving aside any special problem one might have in saying what perceptual states are to begin with; and, at any rate, whatever problem one might have here would attend to the account of indirect perception as well since it, too, presupposes the existence of perceptual states. On the other hand, a serious problem arises with indirect perceptions since it can be hard to say what in fact is being indirectly perceived, due to the possibility of error in a perceiver's set of mediating, descriptive beliefs.

I said above that the object of indirect perception is ambiguous: it could be the directly perceived object or it could be the object that satisfies one's descriptive beliefs. That is, the notion of indirect perception itself involves the notion of direct perception. Need this be so? Indirect perception, as we have defined it, utilizes an interpretation of the perceived object—the object indirectly perceived is interpreted. But indirect perception must utilize more than an interpretation, for there must be an object that is interpreted. What is this object? It could, itself, be an interpretation—but then we've simply displaced the problem one step, for the question will arise again with this interpretation, to wit, what is its object. Nor could the object of an interpretation simply be what satisfies the interpretation (or, put another way, whatever satisfies the descriptive beliefs associated with a perception) for an object could satisfy this interpretation, without being the de facto object of perception. Indeed, an object of indirect perception might not, in actuality, even satisfy one's associated descriptive beliefs—and still it will be indirectly perceived. Thus, the object of an indirect perception must be what is provided by a direct perception—what other object could be a candidate? That is, the notion of an indirect perception relies on a prior notion of a direct perception, and is indeed inconceivable without it.

bluered.gif (1041 bytes)


Back to the Top

20th World Congress of Philosophy Logo

Paideia logo design by Janet L. Olson.
All Rights Reserved


Back to the WCP Homepage