Chapter Three

REPRESENTATIONALISM VERSUS DIRECT REALISM

* * *

3.1 What representationalism is

3.2 Arguments for representationalism

3.3 That contemporary foundationalists are representationalists

3.4 That representationalism fails

3.5 Defeating the skeptic quickly

3.6 Illusions, relativity, and direct realism

* * *

3.1 What representationalism is

Contemporary foundationalists share a cluster of fundamental beliefs about the nature of perception, about the initial legitimacy of the challenge of skepticism, about the connection between doing epistemology and doing metaphysics, and, accordingly, about how a defense of foundationalism should proceed. They differ over many of the details, of course, yet a broad agreement on these several philosophically basic issues characterizes the approach of the major 20th century foundationalists. That broad area of agreement is representationalism.

Representationalism's distinctive epistemological thesis is that knowledge or justified belief about external objects is indirectly arrived at. Such knowledge is based on prior knowledge about phenomenal states. The prior knowledge about one's phenomenal states is held to be justified non-inferentially, i.e., it doesn't depend upon other beliefs. External reality sets the standards for truth — i.e., truth is still defined in terms of correspondence — but the justification of belief proceeds entirely from the "inside," i.e., subjectively and without reference to external reality.

Idealists attack the notion that one can derive knowledge about external objects from knowledge about phenomenal states. Let's call this the "You can't get there from here" position. If this is right, and if perception at best yields phenomenal knowledge, then one is stuck "inside"; one's knowledge, including one's standards for justification and truth, become entirely dependent upon the subject. Typically, versions of the coherence model are then advocated for both justification and truth.

Direct realists agree with the idealist "you can't get there from here" charge against the representationalists. But they deny the representationalist claim that one's knowledge of external objects is based upon prior knowledge of phenomenal states. In fact, they typically reverse the order of dependence: knowledge of phenomenal states, being self-reflective, presupposes knowledge of external fact.[1] One begins directly in contact with reality and builds from there. Reality provides one's standards for both justification and truth.

The claims of this chapter are that all of the major contemporary versions of foundationalism are representationalist in principle, that representationalism fails, that the major arguments against foundationalism attack only (with one exception) representationalist foundationalism, and that the problems that typically motivate representationalism can be solved without resort to representationalism. If this analysis is correct, then it follows that a foundationalism based upon direct realism is still an option, whether or not the arguments often advanced against representationalist foundationalism are sound.

3.2 Arguments for representationalism

Historically, representationalism arises in the context of wondering how best to deal with two epistemological problems — skepticism, and perceptual relativity and illusions.

Skepticism challenges claims to knowledge or justified belief in a number of forms. The most powerful (and hence most widely used) skeptical challenge comes in the form of a question directed toward any claim to justifed belief or knowledge: "How do you know you're not X right now?" — where any of the following can be substituted for the X: dreaming, subject to a malicious demon's pranks, a brain in a vat. Not being able to answer the question leads one to hedge one's claims and to search for a source of certainty immune to the skeptic's question. Representationalism has been seen as providing such a haven, as we shall investigate below.

Representationalist accounts of perception and accompanying representationalist accounts of justification have also been the normal response to illusions and relativity. One's account of perception is obviously central to an account of foundationalism, since perception is, if anything is, the locus of our most fundamental commerce with reality. Therefore, facts about perception that are problematic, e.g., illusions and relativity, are special concerns for foundationalism. To see how such accounts directly affect the question of foundationalism, let us review the progression.

Take a standard case of perceptual relativity, say, a coin that appears elliptical in varying degrees as one views it while moving around it. We know that the coin does not vary, yet something about our experience of it does. We have two facts, then, to analyze and integrate:

(1) The real object (the coin) does not vary.

(2) Something about the awareness does vary.

We have two options at this point. We could say that the object of awareness changes, in which case it follows that the object of awareness is not the real object. Or we could say that the awareness changes. The most common answer, until recently, has been the former, which is the standard representationalist line. Representationalists will introduce intermediary objects of perception and hold that the perceiver is aware of the intermediary objects only — at least that is all the perceiver is aware of directly; the perceiver can be said to be aware of the real, external object indirectly, by means of an inference. Let us use the term "sense data" to stand for all of the candidates representationalists offer as intermediary objects between the perceiver and the object.

(The argument from illusion generates the same conclusion as the argument from relativity via the same form of argument: the real railroad tracks are straight; my awareness does not show what I am aware of to be straight; therefore, what I am aware of cannot be the real railroad tracks.[2])

But before the representationalist conclusion follows, the other alternative — that what changes is the awareness — must be eliminated as an option. Perceptual representationalists accomplish this, often implicitly, by denying that awareness could change. Awareness, on this view, is pure: it has no features of its own. It is simply pure inspection, it is a "mirror of nature," it is "diaphanous."[3] So there is no way it could vary, which means it must be the same in any instance. Then the needed conclusion follows: any variation can only be in the object of awareness.

The reason behind this diaphanous/mirror of nature view seems to be this: if one's awareness had its own identity, which would be partly responsible for the variations in perceptions, then its own identity would distort the resulting awareness, and then it could not know its object. The point is clear in Aristotle: nous cannot have any form of its own, for this would interfere with or distort its reception of the forms of its objects.[4]

The claim can be separated from Aristotle's form-impression account, leaving it in the following form: if awareness is to be a cognitive phenomenon, it must both take an object and not contribute anything to the awareness — which means it must be pure inspection. And if it is pure inspection, the changing qualities of perception cannot be ascribed to it.[5]

The point is also clear in Aquinas when he contrasts the perfect knowledge of intellectual substances with ours: the intellectual substances, "knowing something external to themselves, in a certain sense they go outside of themselves."[6] Aquinas is here indicating the implicit standard: to have perfect awareness of something external, there must be no intermediary organ to distort the resulting awareness; perfect awareness is a pure merging of subject and object. We humans of course cannot effect this pure merging; we have no way to jump outside our heads, so to speak, in order to attain the perfect grasp. The next best thing, then, is for our medium of awareness to be diaphanous, i.e., to have no distorting identity of its own.

This diaphanous standard is also a dominant theme in Eastern philosophy: one must seek to merge oneself with reality; one should become one with the universe in the sense that one should erase all distinctions between subject and object, because the subjective equals the illusory and unreal, and if the subject retains any individual identity then true reality is unattainable.

What is at work here is a fundamental unease over the distinction between object and subject. In its purest form, as in much of Eastern philosophy, this unease manifests itself in a desire to erase the subject. The underlying premise seems to be that if there is any difference between the knowing subject and the object, then the one cannot be said to know the other. In other words, subject and object must literally be identical. Let's call this the Identity Thesis.

While Aristotle and Aquinas do not accept the extreme version of the Identity Thesis, they do accept the view that awareness must be a diaphanous phenomenon; this is to enable a partial identity of subject and object while at the same time preserving a distinction between the two.

Representationalists since Locke and Descartes have delimited the scope of the diaphanous portion of our faculties. It is clear that our sense organs themselves are not mere diaphanous directions upon an object. They are physical organs; they have definite identities of their own. But if the diaphanous standard is correct, then it must follow that our sense organs do not put us directly in contact with reality as it really is. The pure inspection must come at a later stage, after the work of the senses is done, which means one's diaphanous awareness inspects what one's non-diaphanous senses present it.

We have now the fundamental ingredients of the representationalist model of perception. The only way to combine the three premises — that the real object does not change, that one's diaphanous awareness does not change, and that the senses are not diaphanous modes of awareness — is to hold that the senses present one with intermediary objects of awareness. This is the representationalist conclusion.

We start, then, with awareness of intermediary sense data. But there is an added benefit to this conclusion: since one's diaphanous awareness is simply pure inspection of sense data, propositions about one's sense data would seem to be incorrigible. There is at this point no distoring medium between awareness and the object of awareness. Hence one also has in propositions about sense data a source of certainty immune to skeptical objections. Thus the introduction of sense data also satisfies the second of our desiderata. This is certainly attractive to many foundationalists. The next stage is to infer the existence of external objects: it follows, therefore, that foundationalists who accept this account of perception thus follow the standard pattern of attempting to show how one can derive justified beliefs about external objects indirectly via incorrigible beliefs about sense data. Hence representationalist theories of perception generate representationalist theories of justification. Only such representationalist accounts, it is felt, can meet the demands of answering the skeptic and accounting for perceptual illusions and relativity.

Yet the posited intermediary sense data have been found to be extremely problematic objects. Are they mental or physical? Do they have backsides? The documented difficulties in answering these and related questions have led many to reject the sense data analysis, including Chisholm and Moser.[7] But in rejecting sense data we are thrust back into the problem of integrating the two facts about perception:

(1) The real object (the coin) does not vary.

(2) Something about the awareness does vary.

If we reject intermediary sense data, then we must opt for the only alternative: what varies is the awareness. Awareness is thus not an unchanging direction upon an object. The solution must be to understand the sensuous qualities of perception as modifying the awareness and not the object of awareness, if indeed one exists. Awareness is an action, and since adverbs are what modify descriptions of actions, acts of sensing should be described adverbially.

So far the adverbialist option is not necessarily representationalist. One still could hold that acts of sensing are intentional and take real things as their objects, or, more generally, that acts of awareness are necessarily intentional in the strong sense: that the intentional relation is a real relation. One could argue that the object of perception doesn't change in the case of the elliptical coin but that only the manner or form in which it is perceived changes. The adverbial description of the awareness would then be integrated with an account of its relation to its object. This would amount to a direct realist position — which is the position I think is right and which will be explored further below.

Some adverbialists (e.g., Chisholm and Moser), however, drop the real intentional relation and hold that the act of awareness has no necessary direction upon an object. At least two additional considerations motivate this. One is a desire not to make ontological commitments where unnecessary in doing one's epistemology.[8] A second is hallucinations: hallucinations seem to be an experiential phenomenon in which no real object exists corresponding to the experience. In the case of hallucinations, it cannot be that a real object is being distorted or relativised, as is possible to claim in the case of illusions and relativity; the hallucinatory experience seems to be created ex nihilo. Add to this the standard skeptical conjectures about brains-in-a-vat, and the conclusion seems to be that awareness could, in certain cases, occur entirely without an object.[9] We have no automatic method of telling whether a given awareness is a perception or a hallucination, so in the interests of having a unified theory of experience, of which perception and hallucination are two species, the non-intentional nature of hallucinations is decisive. What is crucial is the manner and character of the experience; what is contingent is the external relation. These two considerations dictate that, given the failure of the sense data theories, acts of sensing are to be described non-relationally as self-contained phenomena; and if an act of sensing is in fact related to a real object, then this is a contingent fact that is external to the act of sensing, and one that must be analyzed separately.

The important point for foundationalism is that by dropping the necessary relation between acts of awareness and real objects one makes indirect and thus more difficult the process by which one arrives at justified propositions about external reality. The nonrelational adverbial account is still in principle compatible with a direct realist account of perception, but it does require a representationalist account of justified belief. A further price to be paid is that by dropping the real intentional character of experience, it is hard to see how experience can be a cognitive phenomenon. In addition, it becomes that much harder to see how a hierarchy of propositions (which are cognitive) can be justified by something that is not cognitive, i.e., experience.[10]

Let us conclude this section with a summary statement of the progression of this representationalist argument.

1. If perception is a cognitive phenomenon, then it takes an object.

2. But in the case of perceptual illusions and relativity, what one is aware of can't be the real object.

3. But perception is cognitive, so it must take an object.

4. Hence, there must be sense data.

5. Sense data are unacceptable.

6. Therefore, we must drop the taking-an-object requirement for perception.

7. Hence, it cannot be that perception necessarily takes an object.

8. Hence, we must formulate perceptual experiences adverbially and nonrelationally.

9. Thus, from 1 and 7, perception is not a cognitive phenomenon.

10. It also follows from 7 that establishing percept/external object relations require an inference.

3.3 That contemporary foundationalists are representationalists

This representationalist pattern is clearly at work in the writings of Lewis and Chisholm. Lewis is explicitly committed to phenomenalism. While Chisholm is not a phenomenalist, he follows Lewis in making incorrigibility a requirement for foundational propositions. Both therefore feel constrained to limit their subject matter to psychological experiential data; since propositions about such data assert no necessary connection to anything beyond the data, they can be validated in a manner that "is absolutely decisive and theoretically complete."[11]

The same procedure is followed in Moser's early writings. There, nonpropositional events of immediate apprehension are said to justify subjective propositions solely about phenomenal content; these subjective beliefs in turn justify physical-object beliefs.[12] The best-explanation model is then pressed into service in answer to questions about how one gets from subjective to physical-object propositions.

A modified version of this representationalist procedure is adopted in Moser's more recent writings. The immediate data that give rise to propositions are taken nonrelationally, as subjective, psychological contents: "[s]uch contents do not entail the existence of anything independent of the perceiver's experience of them."[13] As skeptics have pointed out, any given experience could be, from the subject's standpoint, any number of things other than an awareness of an external object.[14] So realism with regard to physical objects is taken initially as a hypothesis to be evaluated relative to others. And so one needs to proceed indirectly in arriving at a justified external-world proposition.[15]

It is also clear that representationalism is taken as characterizing contemporary foundationalism by its opponents. Rorty has argued that foundationalism is a delusion peculiar to post-Cartesian philosophy. Note the epistemological direction that Rorty is accusing Descartes of having sent modern philosophy upon: representationalism.[16] Michael Williams, another opponent of foundationalism, has made the claim that in rejecting foundationalism one is in effect defending direct realism.[17] Clearly Williams is assuming that foundationalism is committed to phenomenalism, which is a key representationalist thesis. In attacking foundationalism, Sellars also proceeds by noting the representationalist model that Descartes, Locke, and Kant introduced. According to this model, our cognitive access to the world is indirect: we don't have direct access to the facts themselves, but we do have direct, diaphanous access to the representational states themselves.[18] As for foundationalism, Sellars sees whether some form of this representationalist model is workable as "a central theme" regarding the tenability of foundationalism.[19] Once again, foundationalism and a representationalist methodology are seen as necessarily linked.

There is certainly some historical justification for seeing the two as necessarily linked. And so there is also some justification for inferring that if representationalism fails, foundationalism also fails. This is exactly the pattern of argument that antifoundationalists follow. But if foundationalism can be separated from representationalism, then from representationalism's failure antifoundationalist conclusions will not follow. In the next section I argue that representationalism does fail; here I am largely in agreement with the antifoundationalists and the idealist arguments that they have traditionally brought to bear against foundationalism.

3.4 That representationalism fails

Representationalism is adopted to allow a foundationalist methodology to proceed without begging questions against skeptical worries, whether those worries be about perceptual illusions or evil demons. Representationalism is claimed to allow the foundationalist to proceed "neutrally," that is, without making any ontological commitments about the status of perception or about the general relationship between consciousness and reality.[20] Yet it is held that representationalism can be adopted without falling into idealism; that is, representationalists hold that it is possible to generate justified external-world propositions from nonrelational, subjective starting points. The purpose of this section is to show that representationalism fails on both counts: it does not proceed neutrally (i.e., without making ontological commitments) in response to skepticism, and justified external-world propositions in principle cannot be generated either from subjective propositions or experiential states taken as nonrelational subjective states.

Let us first examine the claim that representationalism, at least in the form in which it appears in the writings of Chisholm and Moser, is ontologically neutral. Older versions of representationalism cannot make a claim of neutrality, for at a minimum they are committed to the existence of sense data, which are posited as a new ontological category. But, as we have seen, Chisholm and Moser have made it clear that their use of adverbial appearance formulations is not to be taken as involving ontological claims. The skeptic is right, they believe: the mere having of an experience does not imply the existence of an object corresponding to it; so the experience should be taken nonrelationally.

But there is a problem here. In avoiding one type of ontological commitment, namely a commitment to external objects corresponding to experiential states, representationalists have made another type of commitment, namely to a certain model of mind. Skeptical considerations all depend upon holding that given experiential states could occur just as they do in the absence of an external object. Generalizing this claim, the skeptical premise is that the entire contents of one's consciousness could be generated just as they are experienced even if there were no external reality.[21] This is the skeptical position Descartes arrives at in the First Meditation. But this premise presupposes a commitment to the view that consciousness is not in any fundamental way dependent upon reality for its contents. And that is to say that consciousness generally is a nonrelational phenomenon. This is the fundamental source of the view that the relationship between consciousness and reality has to be argued for. But this is hardly a neutral position ontologically. This is to deny what realists take as an axiom — that to be conscious is to be conscious of something, that consciousness cannot create its own contents, that consciousness is a relational phenomenon. The denial of realism is not a position neutral with respect to some third alternative. Consciousness is either essentially relational or it isn't; either position involves a basic ontological commitment. And if one argues that relations between conscious states and external realities are not essential to the conscious state, that instead they are contingent relations to be established separately, then one has made one's commitment. This is the position Chisholm and Moser accept.

If representationalism is not an ontologically neutral response to skepticism, then what of its other thesis, namely that it is possible to bridge the epistemic gap between nonrelational subjective states and propositions about external objects?

One is almost tempted to argue that such inferences attempt to derive an "is" from an "ought".[22] Once we accept the idea of a strict epistemological order, we cannot explain how a warranted inference to the existence of the physical world is possible.[23] Now it is a familiar but forceful idealist objection to the correspondence theory of truth that if the theory were correct we could never know whether any of our beliefs were true, since we have no perspective outside our system of beliefs from which to see that they do not correspond.[24]

The problem here is not a new one. On representationalist grounds, one has access only to one's immediate experiential states. Propositions about such experiential states include no terms relating the experiential state to an external reality. So how can a proposition that makes reference to an external reality acquire justification?

Deduction from the subjective propositions will not work. In a valid deduction one cannot have in the conclusion terms that are not in the premises.[25] Subjective propositions do not contain terms referring to an external world. The only way to preserve a deductive structure would be to attempt the rationalist move of positing innate or a priori principles that do contain an external-world term in them. But there are two well-known and insurmountable difficulties inherent in any such project. First, one would have to justify the claim that such principles could be innate or a priori. Second, supposing that one had overcome the first difficulty, one would have the further enormous difficulty of showing that the innate or a priori principles have anything to do with the reality they are allegedly about.

An enumerative induction will not work, either. All that follows from such a pattern of reasoning is a proposition of the form, "All cases of being appeared to X-ly are cases of being appeared to Y-ly."

For these reasons representationalists adopt the best-explanation or hypothetical induction model, popularized by Russell during his representationalist phase. As we have seen, Chisholm and Moser agree that this is the most promising approach. Chisholm argues that we have to beg the question against the skeptic anyway, and since the external world propositions square best with our pre-analytic intuitions, they are usually the best explanation. Moser argues that good explanations do not introduce objects not given in the experience, and thus concludes that the usual external world propositions beat out skeptical hypotheses on grounds of simplicity.

Moser's is the more interesting position on this point, since he holds that best explanation models allow one to defeat the skeptic and not merely lessen the degree to which one begs the question, as is the case with Chisholm. Yet on Moser's explanatory criteria, it is not clear that any sort of realism emerges as the winner. The usual external-world proposition posits entities that exist independently of consciousness, i.e., that do not depend upon minds for their existence. This does introduce objects that are different in kind from experiences, and thus opens Moser's accounts to Berkeleyan objections. Berkeley can and would argue that idealism can better satisfy Moser's explanatory criteria than can realism's external-world, mind-independent-object propositions, since idealism can account for the same phenomena (experiences) by having an ontology consisting of two sorts of things (minds and their experiences) rather that Moser's three (minds, their experiences, and external mind-independent objects).

If this is right, then Moser is back in the same position as is Chisholm, who appeals to realist pre-analytic intuitions, or as is Russell, who solves the problem by claiming that belief in mind-independent matter is instinctive and that we should stick with our instinctive beliefs unless they conflict with other instinctive beliefs.[26] Yet neither intuitions nor appeals to instincts are convincing sources of justification. And so we have not been able to justify the representationalist leap. Chisholm, for example, notes in passing that on his grounds he cannot distinguish perception from hunches in terms of the one's providing superior grounding for external world propositions; but he doesn't pursue the point.[27] Only animal faith can bridge the gap.[28]

Since neither deduction nor any form of induction can bridge the gap, it seems that the only way for an external-world proposition to acquire justification is for one to be in a position to assert a connection between the external reality and either the experiential state or the subjective proposition. Being in the required position is not a problem for the relation between the experiential state and the subjective proposition: one has immediate access to each. But in the case of the external-world proposition, some sort of direct access to external reality would be required, and this is something that representationalism denies that one has. Hence, the project seems impossible, as idealists since Kant have argued forcefully. Any sort of justification would presuppose that the subject be able to take a third-person perspective encompassing both the external reality and the subject's own first-person perspective. Putting the point more casually, the subject would have to be able to take a God's-eye view of the situation or be able to "jump outside his head."[29] This is clearly impossible.

But even if it weren't, the same representationalist problems would arise: even from the God's-eye perspective there would have to be some means by which the subject becomes aware of the reality, and to whatever means is posited the following questions can be addressed: Does it give the subject the reality as it really is? Or is the resulting awareness a function also of the means of awareness, as in the case of the other sensory organs? Could an evil demon be deceiving the subject? On representationalist grounds, one would be pressed to conclude that even the God's-eye perspective is only a higher-order subjective and nonrelational state, and that any connection between it and an external reality is a contingent matter that needs to be established independently. Positing a yet higher-order God's eye view is clearly futile, once the pattern is grasped. Thus, we have not yet generated any justification for the external world proposition, and it is a mystery how any such justification could be generated on representationalist premises.

This is the argument with which idealists have refuted representationalists since Kant. In some ways Kant was more sophisticated than many of the idealists who followed him. In place of an external material world, which they took Kant as refuting, many later idealists felt safe in filling the void with various replacements: God, the Will, the Will to Power, projections of one's noumenal self, the Collective Self, the Absolute, and so on. But the argument is quite general: from premises solely about the phenomenal nothing follows about the noumenal, not even that there is such a thing. If the subject has access only to a nonrelational first-person perspective, then there is no way to get the perspective needed to justify, however minimally, the belief in one noumenal candidate over another.[30]

Twentieth century antirealists have, I think, correctly grasped the full force of Kant's argument against representationalism and used it well to argue against 20th century foundationalists. It is the power of the Kantian argument that explains why the representationalist versions of foundationalism that have been offered this century have not been greeted with enthusiasm. Representationalism, if consistently followed, leads to the view that justification is solely a matter of internal networking, and that truth is either an empty concept[31] or one that needs to be reworked along entirely conventionalist lines. Both of these conclusions are antithetical to foundationalism.

So in order to defend foundationalism we need to go back and address the concerns that initially motivate representationalism. Since the representationalist claim is that both skepticism and problems of perception necessitate a nonrelational view of consciousness from which to initiate justification, it follows that we have two projects: answering the skeptic, and dealing with perceptual illusions and relativity on realist grounds. In the following two sections I will argue that from represent- ationalism's failure antifoundationalism does not follow, by arguing that the considerations that have led to the adoption of representationalism in the first place — skeptical challenges and perceptual issues — can be dealt with without resorting to representationalism.

3.5 Defeating the skeptic quickly

The standard skeptical arguments based on hallucinations, the Cartesian dream argument, and worries about evil demons or being a brain in a vat share a common concern. For each variant the claim is that everything one experiences could be unreal, i.e., that one's entire experiential reality could be completely out of relation to any external world. The question then asked is, How does one know it isn't? If one is not able to establish that it isn't, one's failure is taken as tantamount to establishing the nonrelational view of consciousness. And from the nonrelational view follows either a general skepticism or an attempt to construct a representationalist theory of justification.

I believe the skeptical argument is invalid. Thus I believe that one of the traditional motivations for representationalist theories of justification is misguided. Let us take Descartes's evil demon as our working example in analyzing the root skeptical argument.

1. Imagine an extremely powerful and evil demon who systematically deceives one about every conclusion one reaches.

2. If this is in fact going on, one cannot trust any of one's conclusions.

3. So, to trust one's conclusions, one has to show that there is no such demon.

4. This cannot be done.

5. So, one cannot trust any of one's conclusions.

Line 4 follows because any attempt to show there is no demon would require some premise from which the conclusion follows that there is no demon; but any such premise would be subject to the same conditions: showing that that premise can be accepted would first require showing that there's no demon. The project thus mushrooms exponentially and impossibly.[32]

Skepticism can be put on the defensive by a number of self-refutation charges; but however valid such charges are, they do not tell us what about the skeptical strategy is wrong. For that I think we need to note that the tricky inference in the above evil demon argument is from lines 2 to 3. By investigating this inference I think that we get to the heart of the problem with the skeptic's strategy.

What actually follows directly from line 2 is:

6. To trust one's conclusions, there has to be no demon.

The question then is whether line 3 follows from line 6. Does trusting one's conclusions require showing that there is no demon, or, more generally, showing that nothing in reality undermines the conclusion? Skeptics generally do make this a requirement. Barry Stroud, for example, formulates the skeptic's principle as follows: For any possibility incompatible with our knowing something, we must know that the possibility does not obtain if we are to know the thing in question.[33] But to show that nothing in reality undermines our knowledge-candidate, we would already have to know everything in reality. The skeptical argument thus requires omniscience as a precondition for acquiring knowledge, and this is obviously an impossible condition to satisfy.

The point of the skeptic's principle is that one has to rule out possibilities that are incompatible with accepting the conclusion. But — and this is the important point — in the case of the evil demon, we would have to have some reason to think the demon is a possibility that needs to be ruled out. This is an issue that skeptics rarely address explicitly.[34] Why should one accept the claim that an evil demon is a possibility? If the demon is a possibility, then one bears the burden of ruling it out. But until the demon is established as a possibility, there is no need to undertake to rule it out. The skeptic does bear an initial burden.[35]

To establish the demon as a possibility, one line of argument often resorted to by skeptics is as follows: If it hasn't been shown that a demon is impossible, then it is possible that a demon exists. And since it hasn't been so shown, it follows that it is possible that a demon exists. This argument is widely used. In discussion and in print the skeptic's procedure is often as follows: Imagine an evil demon; How do you know there isn't one?; You don't; Therefore, skepticism. With the raising of the question, the assumption is that the failure to show that there is no demon establishes it as a possibility that undermines all of one's knowledge-candidates. But however widely the argument is used, it fails, for it is an argumentum ad ignorantiam. In textbook form, ad ignorantiam appears as "p has not been shown not to be so, thus we can conclude p is so." In this form, p is granted the highest epistemic status merely from the lack of evidence for its denial; this is clearly fallacious. Yet the same fallacy can be committed when lesser degrees of epistemic status are at issue. The principle behind making ad ignorantiam a fallacy is that no degree of positive epistemic status is achieved merely from pointing out the lack of positive epistemic status of its denial. A teacher cannot show that a student may have cheated on a test by noting that there is no evidence to show that the student didn't cheat. A prosecutor cannot show that the defendant may be guilty from the fact that nobody has presented any reason to think her innocent. Now, to grant a proposition the status of being a possibility is to confer positive epistemic status upon it. So it is invalid to claim that a demon is a possibility from the fact that a demon has not been shown not to be possible.

The skeptic, accordingly, needs to establish the demon as a possibility by finding some actual positive grounds for it. The only grounds for this ever offered is based upon human imaginative capacities: one can imagine a powerful evil demon bent upon systematic deception. The principle at work, then, is: If something is imaginable, then it is possible.[36]

To claim that something is imaginable means that the proposed scenario contains no known internal contradictions. A unicorn is imaginable; a round square is not. But from the fact that nothing to one's knowledge contradicts an imaginary construction, it does not follow that the imaginary construction is possible in the sense the skeptic needs. The skeptic's claim is not merely that we can noncontradictorily construct an evil demon scenario in our heads — the claim is that out there there may be a demon doing demonish things.[37] This is a stronger sense of epistemic possibility. So when the skeptic asks, How do you know that there is no evil demon?, we can properly first demand an answer to the question, What grounds do you the skeptic have for claiming that out there a demon may exist? And the skeptic cannot simply rely on the fact that a demon is imaginable, for the mere fact that our imaginative powers enable us to construct a scenario gives us no reason to think that reality may instantiate that scenario. For something to be a possibility in the needed sense we have to have some actual grounds for thinking that reality could be certain way. This would require some input from reality. But this is not something the skeptic can provide, on principle, for the skeptic is committed to denying that any knowledge of reality is possible.

My conclusion is that the skeptic cannot establish the demon as a possibility. The same analysis would apply to brains-in-a-vat and other scenarios of the same form. Possibilities are more than imaginabilities; they require actual evidence of some sort of connection between the imaginary construct and the real world they are claimed to be about. In the absence of any such evidence, the offered skeptical propositions are arbitrary posits. (The arbitrary nature of such posits is also suggested from the fact of how easy it is to generate them.) Completely arbitrary posits have no positive epistemic standing. And if a posit has no positive epistemic standing, then I think that one is under no obligation to attempt to refute it. From this it follows that since these considerations apply to the general skeptical scenarios, one is not under any obligation to attempt to refute skepticism. One can show that its method of attack is fundamentally misguided and set it aside from the outset.

It also follows that one can even accept the skeptic's principle — that for any possibility incompatible with one's knowledge-candidate, one must rule out that possibility — without falling into the problem of omniscience. This is because we now have stronger requirements for establishing possibilities: mere imaginability is not enough. One needs some actual evidence, however minimal, that undermines the proposition in question. Doubts are thus always particular: they arise in the context of investigating a particular knowledge-candidate, and they are based upon particular concrete evidence that leads one to consider as a possibility an incompatible proposition. General doubt of the sort skepticism requires does not and cannot arise. Hence no general nonrelationalism follows; and this in turns undercuts one of the sources of the perceived need for representationalist accounts of justification.

3.6 Illusions, relativity, and direct realism

Perceptual relativity and illusions have traditionally been the other impetus for representationalist theories of justification. The apparent problem is that such phenomena reveal that consciousness is not a mirror of nature: a coin can appear differently when viewed from different angles, straight railroad tracks can appear to converge, and so on. This leads to the representationalist conclusion that what appears cannot be the real object.

If the fact that consciousness is not a mirror of nature is taken as a problem, then when representationalists come to investigate the relationship between consciousness and reality they must be operating upon the premise that consciousness could be directly aware of reality only if it is a mirror of nature or some sort of diaphanous medium. Following Kelley I will call the view of consciousness this principle embodies the diaphanous model.

I think the diaphanous model of consciousness must be rejected. This is primarily not on the grounds that the diaphanous model leads to representationalism's insurmountable problems and that therefore, modus tollens, whatever leads to representationalism must be false — though I think that that is a powerful argument. I think the model can and should be rejected on its own grounds. If this can be done, then we will be in a position to approach perceptual illusions and relativity without being tempted into representationalism.

While the diaphanous model plays a fundamental role in determining one's approach to epistemology, it gains impetus from two other philosophical sources. Noting these sources is important to understanding its appeal.

The first is coming to epistemology with a prior commitment to ontological mind/body dualism. Philosophers reach ontological dualism for a variety of reasons, but for whatever reasons dualism encourages perceptual representationalism, as follows. If there is a metaphysical gap between the mental and the physical, then, since the sense organs are on the physical side, it seems natural on dualist grounds to interpret the workings of the senses as being completely non-conscious and merely as providing an internal display. The nonphysical mind then grasps this internal display somehow, but since the mind is considered to be nonphysical, there can be no intermediary mechanism needed at this point, and so it must be via diaphanous direct confrontation that the nonphysical mind grasps the internal display. The nonphysical mind next has to infer a connection between the internal display and an external reality. That is the familiar representationalist problem.[38] And since dualism has been the historically dominant philosophical position on the mind/body relation, it should not be surprising that epistemological representationalism has also been dominant.[39]

Even so, representationalism is not necessarily tied to ontological dualism. One can have a non-ontological compartmentalization of the mind and the senses that also encourages representationalism. If one identifies the conscious self solely with the mind — where concepts, propositions, and self-consciousness exist — and sees the senses as mere conduits of information that are attached to the conscious self, then epistemological representationalism follows. The senses then are not part of one's consciousness: they merely register and transmit information to one's consciousness, which is waiting at the end of the tunnel, so to speak. In this case, as with ontological dualism, there are then two gaps to bridge, one between the conscious self and the end result of sensory processing, and one between the conscious self and external reality. The former is said to be bridged by direct confrontation — i.e., diaphanously — and the latter by means of an inference of some sort. Again, these are the classic ingredients of representationalism.

Kelley has hypothesized that there is an epistemological source for the diaphanous model, arising from the fact that perception can be investigated from two perspectives, the first-person and the third-person.[40] The former yields a phenomenological account of perception from the standpoint of the subject, the latter a scientific account of the processes and mechanisms of perception. Each perspective yields information not accessible to the other. From the first-person perspective, the subject is not aware of, for example, the array of rods and cones being stimulated in his or her eyes during an experience of color. From the third-person perspective, the scientist investigating rods and cones is not aware of the feel of the subject's experiencing. An ideal account of perception should harmoniously integrate the two — but this is precisely where the problems arise. Kelley suggests that the problems arise because people first approach perception from the first-person perspective, and only in adulthood, if ever, take account of the third-person evidence. But from the first-person perspective, perception seems to be a diaphanous revealing of things as they are. The subject is aware of no processing taking place nor of any effect of his or her sensory organs upon the experience. One merely opens one's eyes and confronts reality — that's common sense. If the first-person experience of diaphanous revelation is generalized and taken, implicitly or explicitly, as the standard for valid perception, then the facts of perceptual relativity and illusions will be taken as being puzzling and threatening to the validity and directness of perception.[41]

Naive realists hope to preserve diaphaneity between subject and object, usually by writing off illusions and other problem cases as aberrations. This seems to preserve common sense, but, for good reasons, it has not been popular among philosophers.[42] Representationalists abandon the naive realist's version of diaphaneity, but preserve it for the relation between conscious subject and sense data by adopting an internal mirror of nature model; then, in order to get justified common sense propositions about the external world, they argue that the reflections in the mirror are pretty good guides to what is really out there. Idealists and antirealists point out that consciousness is hardly diaphanous, thus disqualifying naive realism, and argue further that there is no way to determine whether the reflections in the mirror have any worth, thus disqualifying representationalism.

Common to all three positions, however, is the diaphanous standard. Even those who reject the diaphanous model accept its underlying premise. Rorty, for example, argues correctly, to my mind, that the traditional mirror of nature view is false. But then he argues that only if consciousness were mirror of nature could human consciousness (or any form of consciousness, for that matter) be capable of coming to know facts objectively. Since it is not a mirror of nature, human consciousness is not valid. There is a package-deal offered here: Either the mirror of nature model is true and consciousness is valid, or the mirror of nature model is false and consciousness is invalid. But why should we accept the premise that the mirror of nature or diaphanous model is the only way for consciousness to grasp reality?

To focus the key problem the diaphanous model presents, let us state its argument in summary form.

1. If perception occurs by some means (e.g., a physical sense organ), then the means influences the outcome.

2. If the means influences the outcome, then perception is not diaphanous.

3. If perception is not diaphanous, then the subject cannot be directly aware of the object.

4. Therefore, if perception has to take place by a means, the subject cannot be directly aware of the object.

The contrapositive of 4 is the startling conclusion:

4a. If the subject can be directly aware of the object, then it must occur by no means.

Facts about illusions and relativity are all taken to support premise 1, by showing that environmental circumstances, the nature of the organ and its prior conditioning all have something to do with the nature of the resulting conscious state. Premise 2 is not controversial. Premise 3 is an implication of the diaphanous thesis; it does the major work in generating conclusions 4 and 4a.

Any model of consciousness that holds that direct confrontation must and can only occur diaphanously has to run into problems on metaphysical grounds. Advocates of the model, in attempting to maintain the cognitive link between subject and object, end up denying any identity to consciousness. Consciousness's grasping of its object takes place by no means; it is simply a direction upon an object; and this is made possible, recalling Aristotle's initial formulation,[43] by not having any identity of its own. Yet nothing can exist with no identity. To be is to be something; anything that exists exists as something specific. Consciousness, accordingly, must have an identity of its own — it cannot be a diaphanous nothing or an as-near-to-nothing as is possible.

The question then is, What is consciousness's identity? The question has two senses. The first is completely general: What is it for a thing to be conscious? What is consciousness as such? The second sense of the question is more specific: How does this particular type of consciousness (sensory, perceptual, conceptual) work? What are the details of its particular mechanisms and processes?

Consciousness, when spoken of generally, is spoken of as either a faculty or a state. To speak of the faculty is to speak of what it makes possible or accomplishes. And here it seems reasonable to speak of states of awareness. Awareness, though, is a relational phenomenon. To be aware is to be aware of something. Consciousness, accordingly, is to be defined fundamentally relationally, as awareness of an X. What can X be? Is it, as the necessary relatum, real or not necessarily so? Only the former can be possible, for if consciousness is simply an awareness of something, then there is only one thing for it to be aware of: reality. Before awareness can occur, there must be something to be aware of. Both logically and chronologically, therefore, reality comes first and consciousness second. This is realism's axiomatic general point about the metaphysical passivity of consciousness.[44] As a corollary, it follows that awareness of one's own existence as a conscious being cannot precede awareness of reality. As a necessarily relational phenomenon, prior to any contact with reality a consciousness exists only potentially, so it has no existence or source of data independent of reality to reflect upon. One can only grasp one's identity as a conscious being only after one has been conscious — which requires that one be aware of reality.[45]

The big issue is of course whether having an identity prevents consciousness from directly knowing its object, for it is the diaphanous model that has led representationalists and idealists to resist the above realist axioms. Here we integrate the general thesis that consciousness is metaphysically passive with the rejection of diaphaneity as the standard for direct awareness.

As a faculty of awareness, consciousness has to be aware of reality by a specific means and in a specific form. The means and form depend on the identity of the particular type of consciousness in question. This brings us to the second and more particular sense of our question about consciousness's identity: How does this particular type of consciousness work? — whether the particular type in question is sensory, perceptual, or conceptual. Since our foundational concerns here are with the senses and whether their workings necessitate representationalism, let us focus upon them, first taking the third-person perspective, to make the point.

The senses evolved as response mechanisms. Some organisms evolved with receptors able to make them aware of their surroundings. All receptors require energy transfer and differentiation, and different receptors evolved tuned to different types of energy and to different features of a given energy type. Realism's first point is that the organism's receptors' being appropriately stimulated by its environment is its awareness of its environment. If we ask what the receptor is responding to, the answer can only be, "Reality." That's the "what" of perception. But, obviously, organisms experience reality in different forms, since their receptors tune them to different features in the array of energy impinging upon them. This is to raise the question of the "how" of sensation, which depends upon the sensory receptor in question. Each type of receptor responds to its environment in its own way, causing the subject to be aware of reality in a particular form.

Suppose, for example, a cube suspended by a thread in a room. If a blindfolded person were to make contact with and manipulate the suspended cube in her hands, she would have certain experiences. If a sighted person walked into the room and around the cube, he too would have certain experiences. If a bat flew into the room and proceeded to fly around the room navigating by sonar, it too would have certain experiences. In each case, the object of the experiences is the same; each organism is responding to the same thing. But in each case, to switch to the subjects' perspective, the "raw feels" are very different. Does it follow that since the experiences are different, the object of the experience must be different? I don't think so. The nature of these raw feels is determined by the sensory receptor that is stimulated in each case, and it is this that determines how — not what — the organism experiences. In each case the object is the same — each organism is aware of the shape of the cube — but what varies is the form or manner in which the cube's shape is grasped. It would be an error to conclude that since the form is different in each case, none of the organisms is aware of the object's shape. To be aware an organism has to be aware by some means, which means it has to respond in some form. It can't respond in no particular form at all; magic is not an option here.

The wrong idea is to argue that since the form the awareness takes is not identical with the object, the awareness cannot be of the object. This is to imply that there can be no difference between the manner of the awareness and the way in which the object exists, which is a restatement of the impossible identity thesis. In a less extreme version it is to imply that there can only be one right way to be aware of the object, which is a restatement of the diaphanous model.

It is by means of this latter that the facts of perceptual relativity are taken to necessitate representationalism. Ayer's analysis of the stick in water's appearing bent is a classic case. Comparing the appearance of the stick in the water and that of the stick in air, he asks which perspective is veridical, assuming that at most one could be. But what could the one veridical appearance be? Berkeley raises an exactly parallel issue in the first of his Three Dialogues about the so-called primary quality, size. Size, he points out, is as apparently variable as are the secondary qualities, since apparent size varies as one moves closer to or further from an object. We can ask Ayer's question: Which one perception of the object's size is veridical? It could only be the one in which the actual size appears. What perspective would that be? From what distance? It is clear that asking the question on this premise leads to the conclusion that no perspective from any distance would be correct — not even having one's nose pressed up against it — for each is just another perspective from some distance or other.

In pattern, then, advocates of the diaphanous model begin by noting that awareness of the same object can take different forms. Then they ask which form is correct or veridical, implying that only one form could be veridical. But then it seems that the only one that could be veridical is the one in which the awareness has all and only the features the object has in itself. But that is to say that the awareness is identical with the object. And if the awareness is identical with the object, then there really is only one thing — the object — which is to say that awareness of objects doesn't exist.

Against all of this, the right idea has to be that there is only one thing for receptors to respond to — reality. But the form the awareness takes depends on the sort of receptor involved. Many different receptors are possible, as biology teaches us. In some cases different receptors tune the organism to different features of the object — for example, touch to macroscopic texture and smell to microscopic chemical composition. In other cases, different receptors make organisms aware of the same feature of the object but by means of responding to different forms of energy — for example, cats' seeing shape and humans' feeling shape and bats' hearing shape. From the standpoint of philosophical investigations into conscious phenomena, no one of these means of awareness is more valid than any other; each simply is a way of becoming aware of some aspect of reality. Means of awareness are more or less informative than others, either by being more or less discriminatory or by being more or less wide-ranging. But the central epistemological point is the same for all of them: each makes possible an awareness of reality.

The idea we have rejected is that there is only one valid form of awareness — whichever one is a diaphanous revealing of the object as it really is in itself. Rejecting the diaphanous model allows us to recognize that the same thing can appear differently to the same receptor under different circumstances (e.g., viewing a coin from different angles) or to different receptors (e.g., touching and seeing the coin's shape).

Rejecting the diaphanous model also allows us to say that two very different things could appear the same. Only if we operate under the assumption that valid awareness is a diaphanous revealing of things as they are in themselves, will we expect that things with different identities will have to appear differently. For example, a puddle of water and a stretch of pavement are very different things. But it is a fact that a stretch of pavement in the distance on a hot summer's day can look like a puddle of water. But, according to the diaphanous model, different things should appear differently, and since in some cases they do not, we are said to be presented with an illusion and led to distrust our senses. Yet there is no a priori rule that says the way the two appear must be different. If it turns out that the arrays of energy from each that impinge upon one's eyes are similar, then they will appear the same. That is the way one's eyes respond to these real phenomena; to learn that the two things are different will simply require more information than is immediately available.

From this brief defense of direct realism, focusing on its rejection of the diaphanous model, it follows that representationalism is not needed in response to perceptual illusions and relativity. It also follows that the next important question for foundationalism is not how one justifies ordinary object propositions on the basis of subjective nonrelational perceptual states. "The critical problem," to agree with Maritain, "is not 'How does one pass from percipi to esse?'"[46] That way of approaching the issue is the representationalist model we get in most 20th century foundationalism. The question is where the abstract nature of our propositional awareness comes from and how we can say that it preserves the connection to reality that perception has. Showing how the connection is preserved is the same as the project of showing how such propositions are justified. Since that propositional form of awareness requires concepts, the basic question is how abstract concepts are formed on the basis of one's perceptual awareness of reality. The question is about the transformation of awareness of reality from a concrete perceptual form to an abstract conceptual form.

Most foundationalists nowadays do not address the abstractness issue. Operating on representationalist premises, they are largely concerned with getting to the external world and knowing that or being justified in believing that one is such. The direct realist approach to the issue sets aside such questions and moves directly to the question of abstractness.

That abstractness is derivable from particularity is also attacked by idealists and antirealists, and forms an integral part of their attack on the possibility of foundationalism. The attack comes as part of the attack on the "given" and consists in arguing that perception is inadequate as a source of justification of propositions. Sellars, for example, raises this issue explicitly:

Now the idea that epistemic facts can be analysed without remainder — even 'in principle' — into non-epistemic facts, whether phenomenal or behavioural, or public or private, with no matter how lavish a sprinkling of subjunctives and hypotheticals is, I believe, a radical mistake — a mistake of a piece with the so-called 'naturalistic fallacy' in ethics.[47]

This is the issue of the given and its justificatory connection, mythical or not, to propositions. This is the subject of our next chapter.

* * *

[3](see also Kemp Smith, for example [1918, pp. xxxix-xlv]).

[5] Kelley (1974, pp. 18-20) makes this point.

[7] See Michael Tye (1984) for a summary of the difficulties for the sense data theory and how these give impetus to

adverbial theories.

[8] For example, see Moser (1989a, p. 83, n. 25).

[10] It is experience so conceived that is the subject of many of the attacks on the "given," which we will be investigating

in Chapter 4.

[20] For example, see Lewis (1929, p. 65) and Moser (1988a, pp. 238-239).

[21] Skepticism's argument on this point is discussed and rejected below, in Section 3.5.

[32]"essentially sound" (1975, p. 9, p. 318).

[34]part as a corrective against complacency in one's opinions.

[39] Descartes and Locke are two obviously important cases — both were dualists, and both were epistemological representa tionalists.

[41]that one accepts a diaphanous account of perception.