A Broader Naturalism: Metaphysical Intentionalism

A Broader Naturalism: Metaphysical Intentionalism

For Naturalism Seminar Presentation by John Dilworth, 10/20/03

http://homepages.wmich.edu/~dilworth/A Broader Naturalism_MI.htm

I am currently developing concepts of representation, content, and intentionality that are 'naturalized' in a broader sense of that term. The general idea is that, metaphysically speaking, there are not only physical objects and properties, but also consequently information about those objects and properties. This information is ontologically dependent on the relevant physical items, but nevertheless it has an objective status, independent of any human or non-human processing of that information.

The close connection of naturalism with information arises as follows. Naturalist approaches in philosophy rely on the methods and results of the sciences. But these methods and results are themselves centrally concerned with the acquisition and testing of objective information about the world. Science is an epistemic enterprise that is intimately bound up with concepts such as those of justified evidence and reliable information. So even a purely physicalist science has as a main goal the obtaining and organizing of accurate information about the world, rather than its simply coming up with a laundry list of the physical contents of the universe.

To repeat, such information is ontologically dependent on the relevant worldly physical items, but nevertheless it has an objective status, independent of any human or non-human processing of that information. Thus it is also mind-independent information, hence removing any fundamental barriers to naturalizing the philosophy of mind in a broad scientific framework.

One of my goals is to explain apparently mentalistic concepts such as those of intentionality and content in terms of this metaphysical concept of information. The resulting view, which I call 'metaphysical intentionalism' (see Selection A below) attempts to explain concepts such as those of meaning, content, representation and intentionality in terms of structures of objective information that are about the physical world. (However, this metaphysical concept of information must be sharply distinguished from the usual nomic causal covariation concept of information employed by Dretske, Fodor et al.)

A central concept in the current enterprise is that of representation. Many naturalists, and their opponents, hold that strictly speaking, nothing in the physical world represents, or is about, anything else (see Selection B below from Dallas Willard, who is an opponent of naturalism). I agree with this view, at least insofar as it is applied to purely physical objects and properties. Thus if A is a physical intentional, e.g. perceptual, state of some person, then we need some further explanation as to how state A could represent some physical item B.

Willard and others go on to say that interpretation, or taking as, is required to explain how external representation of B by A works, something supposedly not explicable in naturalistic terms. But my strategy is to explain this 'taking as' in terms of a further concept of representation--'internal' representation, as opposed to the usual 'external' representation--which internally represents relevant information about B. Thus the loose relation of external representation between A and B is explained in terms of a tighter relation of internal representation between A and information about B.

In more conventional terms, I am trying to explain the ordinary concept of the intentional content of a thought or perception about object B in terms of a new relation of internal representation of information about object B. Thus on this view, the content of a thought is not something mental, or 'inside the head', and hence the view is entirely compatible with recent externalist views of content (Putnam, Burge etc.). A related ordinary concept is that of 'having information' about something: strictly, on my view one never 'has' the information itself, but instead one only internally represents it. Selection C, 'Content and Representational Duality', discusses some of these points in more detail.

But how is this explanation still a broadly naturalistic one? Answer: because the criteria for when some person or organism counts as having internally represented information about object B are naturalistic, broadly functional criteria. Roughly speaking, we may legitimately claim that an organism A 'has' information about object B to the extent that A can appropriately respond to the presence or absence of B. It is the effective, practical functioning of perceptual or cognitive states that provides our best evidence for their intentional content--i.e., for the information that the relevant organism is able to thus access.

As for the explanatory primacy of internal over external representation, one good way to demonstrate this is to analyze cases of ordinary concrete representational objects such as pictures, whose representational status is relatively unproblematic (unlike that of cognitive or perceptual states of persons). In Selection D, 'External and Internal Representation' I show that we must rely on our judgments concerning what a picture internally represents in order to justify a claim concerning its external representational referent. That selection also shows how there is a fundamental internal versus external ambiguity in the concept of pictorial representation, that parallels the ambiguity found in the discussion of cognitive and perceptual representational states.

Selection A: from 'A Defense of Metaphysical Intentionalism'

John Dilworth

Metaphysical intentionalism (MI) is a novel philosophical position that seeks to 'intentionalize naturality' rather than naturalize intentionality, by showing that central parts of natural science cannot avoid postulating or presupposing intentional concepts, which concepts hence are themselves central to a broadened and more adequate conception of scientific naturalism.

I have a general philosophical position to announce and defend, to be called metaphysical intentionalism (MI), which seems to be novel--at least in its broad outlines and some of its basic details. A useful initial characterization of MI is in terms of what it opposes: whereas any number of philosophers have tried to naturalize intentionality, by trying to explain, reduce or even eliminate it in terms of acceptable basic scientific concepts such as those of causal covariance,[1] proper biological function,[2] neural networks[3] and so on, MI instead is a position that attempts to intentionalize naturality, by showing that even central parts of natural science, and its practices of scientific investigation, cannot avoid postulating or presupposing intentional concepts--such as those of meaning or semantic content--which concepts hence are themselves central to a broadened and more adequate conception of scientific naturalism.

Another useful initial placement for metaphysical intentionalism (MI) is that it is a continuation or natural extension of the broadly empiricist, epistemological tradition that originated with Descartes, Locke and Hume: both scientific and metaphysical arguments concerning the nature of reality cannot succeed in transcending the limits of the available epistemic methods of human and scientific investigation, which result in knowledge or information about the world that is itself thoroughly permeated with intentional factors.

To be sure, there are various other metaphysical positions, such as idealism, or sundry forms of ontological or property dualism, that would affirm the reality of intentional phenomena against the encroachments of natural science. But what differentiates MI from such positions is that it seeks to intentionalize science itself--and more broadly, metaphysics and ontology as well.

Next, how is the truth of MI to be established? ... First, I shall briefly demonstrate the metaphysical centrality of the intentional concept of informational aboutness.

1. The Ontology and Metaphysics of Intentionality

Is the ontology of the world limited to those items currently recognized by science, such as certain objects, properties and facts, or should intentional phenomena also be recognized as having a legitimate scientific status? According to the metaphysical intentionalism (MI) position, the latter view is correct, in that intentional or semantic items may be shown to be equally legitimate basic ontological items. Here are some arguments for this view.

First, arguments about the place of intentionality usually take, as their point of departure, that prima facie there seem to be some intentional properties, such as that of a thought being about some particular object. The issue then becomes one of whether that property of aboutness may be eliminated, or reduced to some group of physical properties, or left as irreducible.

However, this way of discussing the issue already concedes too much to traditional metaphysical views, for even if the intentionalist is successful in staving off elimination or reduction for her properties, the resulting irreducible intentional properties are still just some small subset of properties in general; they have no possibility of affecting other, non-intentional properties, and they are ripe for being sidelined or trivialized as merely 'non-instantiated' properties, as part of some scientific program of reform.[4]

Instead, intentional items should be regarded as a fundamental ontological category in their own right, alongside objects, properties, facts, or whatever other items are claimed to have some ontological status. But how could such a proposal be justified?

At this point it is appropriate to introduce a specifically intentionalistic concept of information. Central to the idea of intentionality is the concept of one item being about, or being directed towards, another item. The desired additional ontological category is that of items of information, each of which is about some more standard or basic non-informational item, such as an object or property.[5] Also, it could be argued that such items of intentional information (hereafter simply: items of information) have a relative ontological necessity, for if there are any basic ontological items at all, then there must also be appropriate informational items that are about those basic ontological items.

In support of this view, consider the following two generally accepted positions. The first is that scientific knowledge or information has some objective status, independent of the individual scientists who produced it; but this view is already tantamount to a claim that such objective information is a genuine ontological category in its own right, yet one whose members are about more basic non-informational items--since of course science gives us knowledge about the world.

The second and related fairly standard position is that both knowledge and belief are propositional in form, where propositions are viewed as mind-independent, objective entities, that have some metaphysical status whether or not anyone ever thinks about or linguistically expresses them. Here again, if this view is correct then propositions are both a legitimate ontological category in their own right, and also one whose members are about basic ontological items--whether truly or falsely.

Indeed, given the general acceptance of these two standard positions, one might legitimately wonder why metaphysical intentionalism is not already a commonly defended and well-entrenched philosophical position. My provisional diagnosis as to why this is so would focus on three factors: 1) that the relevant intentional concept of informational aboutness is not a well-understood or adequately analyzed concept; 2) that it has generally been assumed that any item of discussion whatsoever must fit into the standard ontological strait-jacket of being an object, a property, or a fact etc.; 3) plus the fact that many discussions of intentionality are still mired in more traditional subjectivist or mentalistic views about cognitive activities, hence making them seem fitting targets for scientific pruning rather than intentionalistic elaboration. I shall make a start on providing the needed revisions in succeeding Sections.

But before finishing this Section, here is a simple argument that shows why no item of information can be identical with any ordinary object or property, etc. It is that every informational item has the relational property of being about some other item, whereas no non-informational item has that property. Or in other words, the intentional nature of items of information guarantees their ontological distinctness from any more usual items in other ontological categories.

Selection B: from 'Knowledge and Naturalism' Dallas Willard

http://www.dwillard.org/articles/artview.asp?artid=64

Why Truth Can't Be "Naturalized"

Suppose, then, that truth in the sense of the matching up of representation with subject matter lies at the heart of knowledge. Can it be captured within the categories of the narrower Naturalism? I believe it clearly cannot. The argument against it is an old and simple one that has been reworked in many forms in recent decades.

Suppose that we have an acceptable list of physical properties and relations. We might take them from physical theory, as the properties and relations corresponding to the concepts of current physics: location, mass, momentum and so forth. (Who knows what the future or ultimate physics will look like?) Or, moved by the above doubts about what philosophy can soundly derive from the sciences, we could turn to the primary qualities of Modern philosophy, and, for that matter, add on the secondary ones as well: color, odor, etc. I don't think we need, for present purposes, to be very scrupulous about the list. Let us agree that whatever goes on such a list will count as physical properties, and that narrow Naturalism is the proposal to confine our inquiries and conclusions to whatever shows up on the list and combinations thereof.

The argument, then, is simply that no such property or combination of properties constitutes a representation of anything, or qualifies their bearer as being of or about anything. The properties of those properties and combinations thereof are not the same as the properties of representations (ideas, thoughts, propositions, beliefs, statements). If this is correct, and if the narrower Naturalism admits only these properties, then there are no representations in the world of the narrower Naturalism. Truth then disappears from that world, because in it no subject matter is represented; and hence it can never happen that something "is as it is represented or thought to be." With truth, knowledge also disappears. The ontological structure of knowledge cannot be present in the world of narrower Naturalism.

Note that my claim is that such physical properties never constitute a representation. I say nothing here about representation (mental qualities) not emerging from the physical properties of, say, the human brain. This is not because I think they may so emerge, although some form of interaction between them and the brain surely does happen. Rather, it is because I can only regard talk of the emergence of irreducibly mental properties from the brain or the central nervous system as mere property dualism cum apologies.

Significantly, Hilary Putnam and Daniel Dennett, in defending their own views of representation, belief and the intentional, emphatically support the view that the physical is devoid of the representational. Putnam asks us to imagine an ant crawling around in a patch of sand in such a way as to trace out "a recognizable caricature of Winston Churchill." "Do the lines thus produced depict or represent Churchill?" Putnam asks. He thinks most people would say it does not. The ant has simply traced some lines "that we can 'see as' a picture of Churchill." Putnam's view is that nothing (in the brain/mind or out) in itself is a representation of anything, but is of or about (depicts or denotes), something other than itself, if it does so, only because we "take it" as depicting or denoting that other thing.

Dennett holds a similar view, and specifically with reference to the states of human beings, such as beliefs, desires, etc. In the typically "naturalistic" mode he declares his "starting point to be the objective, materialistic, third person world of the physical sciences," and holds "that philosophy is allied with, and indeed continuous with, the physical sciences." Like Putnam (who of course is not a Naturalist), depicting, denoting, etc. is for Dennett only a matter of how we treat something. "The intentional stance" or "intentional strategy," as he calls it, "consists of treating the object whose behavior you want to predict as a rational agent with beliefs and desires and other mental stages exhibiting what Brentano and others call intentionally." The existence of belief, etc. with its intentionality is to be confirmed only by the success of the intentional stance as a strategy for predicting behavior. And when a better predictive strategy comes along, all the mental clutter of "folk psychology" will go the way of phlogiston and witches. Of course if that happens they aren't "really there" now.

For Dennett as for Putnam there is nothing in the brain or out that by its nature represents something else. There are no "natural signs." There are only human events of "taking as." And these events of "taking as" also, it would seem, must themselves lack any natural capacity for representation (of what is taken as). Rather they, in their turn, can only be taken as representing what is taken as intentional states or systems of the human organism, or what is taken as pictures or symbols. They do not inherently represent them.

Surely there is something wrong here. If we are in a world where nothing is naturally representative of something else, and we see the lines traced by the ant as a picture of Winston Churchill, then our seeing also is not naturally of the lines, and of the lines as depicting Churchill. Either there is going to be at some point a "taking as" which does not itself represent anything (even what is "taken")--which certainly sounds like a self-contradiction and is at best unlike the instances of "taking" featured in Dennet's explanations--or there is going to be an infinite regress of "takings." This inclines one to say that unless there are some natural signs--things that refer or represent simply because of what they are--there will be no signs at all. But natural signs are, precisely, impossible in the world of strict physicalism--and, for his own reasons, in Putnam's more generous world as well.

Selection C: from 'Content and Representational Duality' John Dilworth

1. The Relation of Content and Intentional State

An Intentional state is about something, and it also has a content--that much is generally agreed on. But what exactly is content, or a content, and how is it related to the intentional state whose content it is? I shall offer unusual answers to both of these questions, with the aid of some novel considerations, as well as clarifying what it is to be an intentional state.

The debate concerning whether some kinds of content are externally rather than internally determined--or whether they are wide or narrow--is relevant to these questions in the following way. If some content item C is external or wide, then it cannot be a monadic property of its relevant intentional state S. For any monadic property of S is supervenient on S alone, independently of its environment, whereas any external content item C is, by definition, at least partly determined by environmental factors that are independent of S.[6] Similar reasoning would show that C cannot be identical with a physical part or state of the cognitive system that has intentional state S. Thus insofar as such a wide content item C has any status at all, it must stand in some non-monadic, non-exemplification relation to S.[7]

Also, it is theoretically desirable, if possible, to have a uniform theory of content, in which any kind of content, whether broad or narrow, is connected in the same way with its intentional state. This possibility is still open, even if some content is narrow, because narrow content is so defined in terms of its supervenience on S alone, rather than in terms of its being a monadic property. (Though any monadic property must be supervenient on S alone, it does not follow that anything supervenient on S alone is a monadic property.) My proposal will implement such a uniform theory.

At this stage in the wide versus narrow content debate it seems safe to assume that at least some content is wide, so I shall do so. And with my promissory note of a uniform theory, we may provisionally conclude that content items C in general stand in some non-monadic relation R to their intentional states S.

2. Two Kinds of Representation

An important clue to the relation R between an intentional state S and its content C may be derived from the following consideration. It is a truism that organisms are able to represent items in their environment because they have acquired information about such items with the aid of their senses, which information or informational content is in some way stored or encoded by the organism in its internal physical structures or states.

Thus information is generally regarded by naturalists as being available to organisms, not directly as pure information or content (in the manner of some substance or property dualist view), but instead via that informational content being physically encoded in some way in the organism. It is this encoding relation that we will identify with relation R.

Here is an example that invokes the concept of encoding. In a discussion of mental causation, Frank Jackson says:

"Any view of content that insists that content properties are not intrinsic, “syntactical”, shape or neurological properties, or some such, is inconsistent with content being causally efficacious. The usual response on behalf of such theories is to grant that content is not causally efficacious while insisting that a property intimately connected with content is. Something of the shape, syntactical or neurological category, encodes the relevant semantic property, and the encoding property does the causing—that is how content gets to be causally relevant without actually causing.”[8]

Jackson here explicitly invokes a relation of coding (or encoding) between some physical property and the relevant semantic content property.

The encoding relation is also familiar from computational contexts. For example, there are various ways in which any given real number may be digitally encoded in computer memory. In such cases, a given physical state of a memory register will be the physical basis for the encoding of the number by the register. Or an encryption--a kind of linguistic coding scheme--might be used to encode ordinary words as meaningless strings of symbols, which may be decoded by appropriate inverse operations.

But what exactly is the relation of encoding? It seems unavoidable that it is some species of representation. For example, various encrypted forms of a given word are different representations of the same word; an appropriate string of binary digits is a standard binary representation of a given number, and so on. When a physical structure is used to encode some information, it is being used to represent the relevant information.[9]

Let us now re-examine the initial truism, that organisms are able to represent items in their environment because they have acquired sensory information about such items, which informational content is in some way stored or encoded by the organism in its internal physical structures or states. If information is itself represented by the organism, as a way of storing the information needed for it to represent some environmental item, then we must distinguish two varieties of representation, to be called internal versus external representation.

On the current account, ordinary or external representation of an object X by a state S may be explained by the ability of an organism to encode or internally represent the informational content C, in virtue of which internal representation of content it satisfies various functional criteria for it to externally represent X, such as by its being able to use the information to recognize and appropriately react to object X.

As a somewhat broader approach to encoding and internal versus external representation, consider the familiar view that mental thoughts have propositional content. As McGinn puts it, "It seems constitutive of the mental realm that it be populated by states with specific contents, i.e. targeted onto an array of propositions. The mental (qua mental) is not thinkable apart from such targeting, with the concomitant targeting onto the world"[10] (italics mine). McGinn here clearly specifies two different kinds of targeting or aboutness or representation, which I would analyze as internal representation of propositions plus a concomitant external representation of the world.[11]

3. Ways that Objects are Represented as Being

More evidence of the need for a concept of internal representation is provided by our natural use of certain common locutions in connection with intentional states. We say, for example, that a perceptual state represents some object as having certain properties, or as being some way, whether or not it actually has those properties, or is that way. In a related vein, it is a truism for representationalists that intentional states do not necessarily show an object as it actually is, but rather only show how it is represented as being, or, again, the properties it is represented as having.[12]

Such locutions could be explained by invoking the concept of internal representation, as follows. An intentional state internally represents its contents, which consists, for example, in those properties that the actual object is represented as having, while this 'represented as having' locution is itself explained in terms of the same concept of internal representation.

Clearly some such account is needed, because the above 'representation as' locutions could not be explained simply in terms of the standard concept of external representation, since the relevant externally represented object may not have some of the characteristics that it is represented as having, or it may even not exist at all in more extreme cases of misrepresentation.

Nevertheless, it is still the externally represented actual object which is internally represented as having certain properties, so that both concepts of representation are integrally related, and play complementary roles in an account of intentionality. Arguably a cognitive system can successfully externally represent an actual or putative object only by representing it as having certain features (or representing that it has those features), but which representation as, or that, involves internal representation of relevant content items. While at the same time, such cases of internal representation of content are normally--in cases such as veridical and non-fictional ones--only a subsidiary part of cognitive activities that are primarily focused on achieving successful external representation.

4. More Connections

This brief Section will investigate further the connections between content versus non-content items, and internal versus external representation. My procedure will be to adopt some premises on the basis of some preliminary arguments, and then to justify their retention in the rest of the paper via the overall plausibility of the resulting position.

To begin, here is a simple preliminary argument to show that no content item is identical with any non-content item. In Section 1, content items were truistically described as providing, or as consisting of, information about objects and properties. But if any content item is about something, this means that it has semantic properties, including properties of correctness or incorrectness. But no ordinary object, or no ordinary property (whether monadic or relational) has semantic properties. Hence, as mentioned, no content item can be identical with any non-content item.

A second argument to the same effect is based on the explanatory relevance of content and internal representation. Consider a fuller version of the truism from Section 1, namely that organisms are able to represent items in their environment because they have acquired information about such items with the aid of their senses. Now if such informational content is to genuinely explain the ability of an organism O to externally represent items, it cannot simply consist of the relevant ordinary objects or properties as internally represented by O, because such a view would amount to a mere reduplication of the relation, and relata, of external representation. (One cannot explain why A externally represents X by saying simply that it is because A internally represents X).

A similar point would apply to broader resemblance views of representation, that would hold that A externally represents X in virtue of the resemblance of its content to X. If the content consisted of ordinary objects and properties Y, then 'A internally represents Y' would be in as much need of explanation as 'A externally represents X', and hence it could not explain external representation. Hence again, content items cannot consist of non-semantic, ordinary objects or properties, on pain of explanatory irrelevance.

Selection D: from 'External and Internal Representation' John Dilworth

1. The Ambiguity of the Concept of Representation

To begin, I shall show that the concept of representation has two distinct aspects by demonstrating an apparent ambiguity in the concept. One standard test for ambiguity of a concept C is to find a case in which C both is, and is not, applicable to some object or objects; then one may proceed to repair the damage by distinguishing different senses, aspects, or varieties of the concept 'C'. Here is a pictorial test case of this kind for the concept of representation.

First, it is surely true that a picture P might represent a man, even if there is no actual man about whom it is true that he is the man thus represented by picture P. For the person who painted A might have intended merely to represent 'a man', without having any particular actual man in mind.

However, if there is no actual man represented by picture P, then there is a clear sense in which P does not represent 'a man'--at least, not in the sense in which it would represent 'a man', if there were some actual man whom it represented.

At the same time, though, to repeat, clearly picture P might still represent 'a man' in the first sense above, in that, if the artist's painting is successful, then it will 'represent a man', in spite of the fact that it does not represent 'a man' in the sense of: an actual man.

Now to be sure, the ambiguity here might be argued to be located primarily in the whole phrase 'representation of a man', or in different senses of 'a man' in each case, rather than in the concept of representation itself. Nevertheless, the case at least shows the need to distinguish two fundamentally different aspects or categories of representation--each generically describable as: representation by a picture P of some thing X--that we confuse at our peril. Thus the label 'ambiguous' will remain descriptively convenient for such phrases as 'P is a representation of a man', and thus derivatively as applied to the concept of representation itself, even if it is only (very) different kinds of use, or different aspects, of that concept that are actually involved.

As for the evidence I have used to establish the ambiguity in question, it itself is arguably not controversial at all, since it is recognized already. For example, Richard Wollheim distinguishes representations of particular objects, or events, from representations of objects or events "…that are merely of some particular kind", and thus of a man etc., rather than of some particular man;[13] while Kendall Walton uses the fact that a picture can represent a man, even though there is no actual man thus represented, as part of his case that the concept of representation is primarily to be understood not in object-centered ways, but rather in terms of his own 'props used in imaginative games' approach.[14] And Nelson Goodman, in the face of such examples, distinguishes a picture of a man from a man-picture, in cases where no actual man is represented.[15]

However, by focusing on the fact that examples such as the above show that there is an apparent, or at least formal, ambiguity in the concept of representation, I hope to facilitate the raising of issues and approaches that have not yet been adequately investigated.

First, I shall distinguish the relevant distinct concepts or aspects of representation as external versus internal (or outer versus inner) representation. External or outer representation by a picture P is about cases where there is some actual object X, normally external to the representing object P,[16] which is represented by P. Thus, for example, picture P externally represents a man just in case there is some actual man, distinct from P, who is represented or portrayed by P.[17]

However, the formulation of a corresponding definition of internal or inner representation is inevitably more controversial. Here is an initial attempt: internal or inner representation is about whether there is an object X', that is internal to the object P, and which is represented by P.

That this definition is inevitably controversial can be seen by giving a substitution-instance of it: picture P internally represents a man just in case there is a man, who is internal to P, and who is represented by P.

This instance is likely to raise logical and metaphysical hackles on at least two grounds: first, that it clearly involves reference to, and reidentification of, 'the man' in question (since it seems to be said both that there is such a man, and who--that is, that same man--is both internal to P, and represented by P), and second, metaphysical worries about how 'a man' could possibly be 'internal' to a picture P.

Indeed, it seems likely that these, or related, logical and metaphysical obstacles have proved formidable enough to dissuade previous attempts at distinguishing an inner aspect, or kind, of representation from its more conventionally accepted outer aspect or kind.

Nevertheless, as already noted, at least the unanalyzed ambiguity, and the evidence on which it rests, is not controversial: that a picture can represent a man, without there actually being such a man, cannot be denied. Thus the initial issue is, of course, as to how linguistic contexts such as 'picture P represents a man' should be analyzed, in cases when there is not any actual corresponding man represented by P.

I take it that it would usually be assumed that this initial case just shows that there is a unitary concept of representation, which, among other things, is such that 'P represents X' does not entail 'there is an actual X represented by P'. However, that non-entailment can equally well be explained by postulating two distinct concepts of representation, on one of which, namely internal representation, the entailment does not hold, while on the other (external representation) the entailment does hold. For it is clear that standard examples of the non-entailment in question would, on my account, be cases of internal rather than external representation, and hence be adequately explained by that concept of internal representation.

Moreover, the issue of ambiguity is not so easy to evade, because, for instance, one could see or recognize that a picture P represented 'a man', whether or not there was, independently of one's act of perception, some actual man who was represented by picture P. Thus it is hard to deny that one could, on the one hand, have conclusive perceptual evidence that a picture P represents someone--which 'someone' one would describe or recognize as 'a man'--quite independently of whether one also has any evidence that something else, an actual man, is represented by picture P. Thus from an epistemic perspective the relevance or applicability of two distinct concepts of representation--or of two distinct kinds of 'object' of the concept--seems to be an inescapable part of the joint evidentiary situation in such a case.

Hence I would claim that, in spite of any initial difficulties or obstacles in defining internal representation, some defense should be attempted of that concept, that adequately captures or allows for such epistemic differences, as well as being acceptable on logical and metaphysical grounds.

2. Internal Representation

It might be thought that the concept of internal representation for a picture P of an object X at least is dispensable in cases where there is an external representation by picture P of object X. But I shall argue that in such cases internal representation still has an important epistemic role to play, in that it is in virtue of seeing what a picture P internally represents--its object X'--that one can justify a claim that P externally represents object X. I shall concentrate on this case in this current brief defense of internal representation.

As a preliminary, a rival candidate for what one thus sees—namely actual object X itself—should be ruled out. There is no magical way in which, merely by looking at a picture P of some actual object X--which picture is not identical with X--one can thereby ‘see through’ the picture to see X itself, no matter how one explains or analyzes what is involved in having perceptual access to picture P, and of what is involved in P's having representational qualities such that it represents X. For perceptual claims are veridical, no matter what interpretive juggling or re-arranging goes on with respect to 'what one sees' in such a case; so that if P is not X, then one cannot correctly be said to see X, solely in virtue of one's looking at P, whether or not P is a representation of X, and whatever is the correct analysis of 'P is a representation of X'.

Now to begin. My initial attempt at a definition of internal representation was this: internal or inner representation is about whether there is an object T', that is internal to the object P, and which is represented by P. For example, picture P internally represents a man just in case there is a man, who is internal to P, and who is represented by P.

As announced, I shall accept this definition or characterization, and justify it using epistemic considerations about perception of the picture. If a person with normal perceptual abilities, under normal lighting conditions, sees a picture of a man, he will be able to see a man when he looks at it, or equivalently, he will be able to recognize what he sees as a man--assuming that the picture is successful in representing a man, that is, that similar normal people under normal conditions can see or recognize a man when they look at it.[18]

Thus one's primary evidence as to whether a picture A is successful in representing a man (whether or not it externally represents a particular man) is derived from evidence as to general abilities of people to see or recognize a man when perceiving that picture A. For whatever the intentions of the person making the representation, she will only have succeeded in representing X if those general recognitional conditions are satisfied.

However, the concepts of seeing (as already noted) and recognition are veridical concepts, such that a claim to see or recognize X is true only if there is an X that one sees or recognizes. Thus if it is true that one sees a man X when perceiving picture A, then there must be a man X whom one sees--or one did not, after all, see or recognize such a man X when perceiving the picture.

Thus, short of a massive 'error theory' of ordinary reports of seeing or recognizing pictures, which would deny that we ever really do see or recognize items when seeing pictures, our ordinary epistemic claims to see or recognize X when perceiving A should be respected, as requiring us to accept that there is an internal X that is thereby seen or recognized--and hence to accept the current referential characterization of internal representation being argued for.

Now it might be thought that this conclusion is premature, on the ground that seeing or recognizing X when perceiving a picture might be a case of external rather than internal representation. However, as already pointed out, the ability to see a man while looking at a picture, or the recognition of what one sees as a man, is a skill independent of whether or not the picture also externally represents some particular actual man. In addition, it has already been argued that it cannot be the external actual entity X itself that one sees.

Next I shall briefly argue that it is in virtue of seeing what a picture P internally represents (its object X') that one can justify a claim that P externally represents object X.[19]

The basic reason as to why external representation claims are epistemically dependent on internal representation claims is because of the conjunction of the following two points that have already been presented separately, namely: a) claims of successful representation require perceptual or recognitional evidence, since mere evidence of artistic intentions to represent X does not demonstrate success in doing so, plus b) the only perceptual or recognitional evidence available is of internal representational kinds, since strictly speaking one cannot see or recognize a picture as externally representing X.

To be sure, one might believe that one sees that a picture externally represents Napoleon, but one cannot actually see it thus even if the picture is an external representation of Napoleon, for the reasons already given (because of 'no seeing through' plus the veridicality of the concept of seeing). Hence such claims of perception or recognition of external representation need to be reworked as objective, perception-independent claims about what a picture itself externally represents--but the only perceptual or evidential support available for such objective claims is from internal representational sources.

Now a full justification of a claim that picture P externally represents an actual man X would presumably require much more evidence than merely that one can internally see picture P as a man, or see a man when one looks at it --presumably one also must be able to see it as a man of a certain sort, or as a man having further, sufficient identifying characteristics. But the ability to internally see a man when one looks at the picture remains as an indispensable part of those evidential necessary conditions for external representation of the actual man X.

I take these points as showing that the epistemic foundations of internal representation are at least as secure as those of external representation, so that any potential attacks on my account of internal representation as somehow being 'unfounded' or 'subjectivist' have already been defused to a significant extent.

Next, I claim that the overt or intuitively natural logical form of statements of internal representation--that there is an object T', that is internal to the object P, and which is represented by P--can itself be given an epistemic defense, in that the evidential chain justifying claims of external representation itself requires that the internal epistemic evidence have such a logical form, that is, a logical form in which a particular 'internal' entity is referred to and reidentified as being represented by P.

To begin with, it has already been established that a claim to see a man when looking at picture P is, on the face of it, unavoidably referential--that is, implying a claim that there is a man whom one sees. An additional reason for accepting that point is because statements reidentifying that same man play an essential role in the epistemic justification of claims of external representation by a picture.

This is so because justification has a social dimension: in order to justify a claim that picture P externally represents an actual man X, there must be intersubjective agreement on the evidence for that claim. This requires that different persons must each be able to see picture P as internally representing the same man X', since otherwise the evidentiary basis for the external representation claim--the fact that a man X', that is, the same man X' in each case, can be seen by qualified observers when looking at picture P--will fail to be established.

For example, suppose that two people A and B look at the same picture P, and that each claims, on the basis of what he sees, that the picture externally represents Napoleon. If asked for the evidentiary basis of his claim, A replies that he can see, recognize or identify the man in the picture as Napoleon--that is, that the man internally represented by the picture is recognized by him as being Napoleon. And suppose that person B gives the same reply to the question.

Now my claim is that, in order for B to be corroborating A's claim, it must be assumed or presupposed that each is talking about the same man in the picture. For if A were to recognize one man X in a picture, and B were to recognize another man Y in the same picture, then there would be no single man Z such that both A and B recognize him as being Napoleon, and hence no secure ground for social agreement that the picture therefore externally represents Napoleon.

Indeed, so basic is this point that one can hardly even make sense of the claim that A and B might be recognizing different men in the picture--other than in the everyday sense that the picture multiply represents two distinct men, so that A and B might be talking at cross-purposes about those different men, that are visible in different areas of the picture.

In sum, I would claim that the overt or intuitive logical form of statements of internal representation cannot be overturned without at the same time overturning our whole present scheme of justification of claims that some pictures do indeed externally represent certain actual things. Thus the referential logical form of statements of internal representation is arguably at least as secure as is the referential logical form of statements of external representation, under our present scheme of justification.

3. Allaying Ontological Anxieties

I shall finish by seeking to allay any initial ontological anxieties about the status of objects of inner or internal representation.

In the previous Section I provided an epistemic justification of the referential logical form of the definition of internal representation as: there is an object T', that is internal to the object P, and which is represented by P, along with examples such as: picture P internally represents a man just in case there is a man, who is internal to P, and who is represented by P.

Now I shall address the status of the phrase 'there is' in the definition. I claim that this phrase is merely part of a claim that 'there is' an object of reference T' that is represented by a picture; or in other words, this 'there is' has no ontological implications as to the ontic status of any given such object of reference, but instead it functions merely as part of a claim that statements of inner representation are indeed referential--involving a reference to some object--independently of such other ontic issues.

A more specific ontological anxiety is a concern that an account of internal representation might be committed to the actual existence of such inner objects, since of course the idea of an actually existing man as somehow being internal to a picture of him seems confused or even self-contradictory. However, my reply is that a general account of internal representation must inevitably include cases of internal objects that could not be actually existent, since they are fictional in a broad sense--such as with pictures of unicorns, or Santa Claus, or of a character in some fictional novel or play. This again shows that the referential form of statements of inner representation does not in itself have any ontological implications as to the status of any particular objects that are thus referred to.

In conclusion, a start has been made on showing the legitimacy and centrality of a concept of internal representation. Also, I have argued elsewhere[20] that recognition of the concept enables a systematic and coherent account to be given of references to fictional entities in artworks, so that the concept is indeed an important one.

Notes