[Chapter 1 of The Primacy of the Subjective, by Nicholas Georgalis]


            Two crucial aspects of the human mind are phenomenal experience and intentionality. Under the influence of behaviorism, functionalism, and early identity theories of the twentieth century, accounts of these two aspects have been advanced that are independent of each other; certainly the two have received separate treatments by these theories.[i] Moreover, consciousness, insofar as it has been considered at all, has been associated with phenomenal experience. The separation of the two and the alignment of consciousness with just phenomenal experience are fundamental errors resulting in distorted accounts of both aspects, especially that of intentionality. I argue that the major source of these errors is the reliance on a strictly third-person methodology. Both phenomenal experience and intentionality are conscious phenomena, and a proper and unified understanding of them can only be obtained by adding a first-person methodology. While others also emphasize the importance of first-person methodologies, they typically use it to talk of just phenomenal states. A few rely on the first-person perspective in an attempt to expose phenomenal or phenomenological features of intentional states.[ii] To the extent that such efforts are designed to bring intentionality back under the umbrella of consciousness I am sympathetic to them but dubious of the means used to achieve this end. My theory is importantly different from all such efforts, as I deploy a first-person methodology to uncover a unique kind of non-phenomenal narrow content.

            Some Preliminaries.

There are at least two striking features that any theory of intentionality must confront:

(1)   There is an asymmetry between oneself and others with regard to access to the contents of our thoughts.[iii]

(2)   An individual may neither know nor be in a better position than someone else to ascertain what his own thought is about.

The resolution of this apparent conflict turns on the recognition that a correct analysis of intentional states involves not only two kinds of content, but two kinds of methodology. The sense of ‘content’ that preserves the first feature requires a first-person analysis, while the sense that saves the second requires a third-person analysis. An exclusive use of either type of analysis cannot provide an adequate theory of intentionality.

Advocates of a strictly third-person analysis of content abound but, to be successful, they must either provide for the first feature or show that it is a mere appearance. Including the first-person perspective in the study of the philosophy of mind runs counter to the views held by a formidable array of contemporary philosophers. For example, consider D.C. Dennett’s unequivocal rejection of the first-person perspective, “I declare my starting point to be the objective, materialistic, third-person world of the physical sciences.”[iv] This starting point he himself characterizes as “­the orth­odox choice today in the English-speaking philosophical world”[v] And later he says, “I propose to see­ just what the mind looks like from the third-person, materialistic perspective of contemporary science. I bet we can see more and better if we start here, now, than if we try some other tack.”[vi] In between these quoted passages he disparagingly cites a notable exception to the orthodoxy, Thomas Nagel’s insistence on the importance of the first-person perspective, or the subjective, for a proper understanding of a number of things about the world and ourselves. I think Nagel is right. However, in this chapter my employment of the first-person perspective will not consider the qualitative or phenomenal aspects of mental states often associated with Nagel’s work, I will turn to that in chapter 3. Instead, I will argue that the first-person perspective plays an indispensable role in uncovering a non-qualitative kind of content, one that plays a crucial role in studies of intentionality. (Further support for this is advanced in chapter 4.) Keeping the faith with the orthodoxy, rather than enabling us to “see more and better,” runs a significant risk of derailing the whole endeavor. There is no need to choose between the exclusive use of one or the other perspective in studies of the mind; forcing a choice is a mistake. Chapter 2 details various failures of strictly third-person methodologies.

In what follows, I ask my readers to realize that at times they must project themselves into the perspectives of the individual thinkers that I consider, to think the situation described as though they themselves were in the situation. Where I am not explicit, I let the context of my discussion serve as clues as to whether the first- or third-person perspective is the appropriate one to adopt. In what follows I will introduce a kind of narrow content that has not been considered in the literature. It is a non-phenomenal narrow content that is invisible from the third-person perspective. It is worth stressing that if I am right about this, then my interlocutor must assume the first-person perspective, if she is to fairly criticize what I say. For my arguments for it cannot even be comprehended unless one, at least temporarily, assumes that perspective on the cases I develop. This is a harmless request since temporary adoption of the first-person perspective should not in itself beg any questions against opposing views.

Some possible misunderstandings of what follows may be avoided if I unequivocally state at the outset that I am not concerned with the ontology of mental states, whether they are intentional or phenomenal mental states. My interest lies with the analysis of statements concerning such states. The kind of analysis that I will employ attempts to make sense of various simple statements that we find difficult to reject upon considering our experiences, such as the two features of intentional states described at the beginning of this section. I intend the analysis to be neutral with regard to ontological commitments; others, if they wish, may attempt to draw ontological conclusions from what I say, but they are not my conclusions, nor are they forced by what I say. I will return to why ontological claims should be avoided from time to time in this and subsequent chapters, and I will offer a more systematic discussion concerning the general conclusion about the meaninglessness of ontological claims in chapter 8.

Privileged Access and Minimal Content.

The asymmetrical access we have to our own thoughts is sometimes referred to as privileged access. Many different conceptions of the latter have been advanced, but two general reasons why privileged access is held in disrepute by some is that exaggerated claims have been made on its behalf, and it has kept company with dualism. However, privileged access is not necessarily connected to dualism and certain alleged features (such as complete transparency and incorrigibility) may be dropped, while still preserving an important point to the special access we have to some of our states. The special access at issue amounts to no more than one’s ability to non-inferentially know the content, in some sense, of at least some of one’s thoughts. It is in some such limited guise that privileged access remains a compelling doctrine.[vii] Although the asymmetrical access amounts to no more than this, we shall see the consequences are great and have not been adequately recognized.[viii]

            Let no one worry that by allowing this privileged access all the work is (mysteriously) done. Frankly, the access that the first-person perspective provides does not explain anything, still, it and the content to which we have special access are central features that must be explained or shown to be a mere appearance. They cannot be ignored.

            To illustrate the kind of access in question, consider an example.[ix] Suppose you ask me to form an image of my grandmother. On informing you that I have done so, you inquire how I know it is an image of her, and not someone else. For such an inquiry to make sense, it must be possible that I could be mistaken when I think the image I formed is an image of my grandmother. But as I formed it expressly to be of my grandmother, such a possibility must be ruled out: It is constitutive of the forming of the image that it is of her.

There is room for error on the agent’s part of a sort, one that is harmless to the point here. For example, if the woman who I have come to think of as my grandmother has really been an impostor. In that case, however, I would not be making a mistake about whom my image is of; rather, it is a mistake about my blood relationship to her. Another kind of case that may be put aside is one where some image randomly comes to my mind. Here, although the image is mine, I would be in no privileged position to ascertain of whom it is; indeed, it would appear to be no more of anything than are the “stars” I experience on receiving a blow to the head. If one were to maintain that it was of someone in particular, the criteria for deciding would be at best unclear; I certainly would not be in any privileged position to know this in such a case.

            So, while I can be wrong about images of mine in some ways, I cannot err in identifying whom my image is of when I deliberately form it to be of some particular individual. I cannot err in the latter simply because the possibility of error in these circumstances does not make any sense, not because I have some special mental powers or that I am cognizant of a special kind of entity. That it is of the particular individual in question is a constitutive element of the very act of forming the image. It could not be that act, if it were not of that individual. Given this, plus the fact that the content at issue is only part of the content of a thought, such “infallibility” is not to be confused with the Cartesian kind.

            Someone might think there is a possibility of error here because they hold some “resemblance” criterion for what an image is an image of. Thus, suppose the image I formed is in fact a rather poor resemblance of my grandmother. The image itself may even be an excellent resemblance of someone other than my grandmother, say, your neighbor; nevertheless, the degree of visual faithfulness to her is irrelevant. It is, after all, the image I deliberately formed to be of her. The criterion determining who, or what, it is an image of cannot be based on what it is the most (visually) similar to. It is not as if I conjured up an image and then began to wonder who it depicts; the image was conjured precisely to be an image of her. On the other hand, were I to find a photograph of someone, I might well wonder who is depicted in the photo and use resemblance as one criterion for deciding. In this case resemblance would be appropriate because of the causal relations involved in producing the photo. The causal relations are radically different in the case of a deliberately formed image; so, while there may be a resemblance, this is totally irrelevant to the determination of the individual represented by the image. The formed image is a direct result of my act to produce, not simply an image, but an image of a certain individual. That being a constitutive element of the act, I cannot perform the act without the result being of that particular individual; otherwise, it would be a different act.

            The same point could be made with a sketch, which has the advantage of being publicly observable. If you ask me to sketch my grandmother, the result may indeed look more like your neighbor than my grandmother, and we may even agree on this. But a poorly drawn sketch of my grandmother is still a sketch of her. That is why it is said to be poorly drawn; it is not said to be a well-drawn sketch of whomever it most closely resembles. In parallel to the imagery, and in contrast to the photo case, I neither have to infer nor is there any possibility of error on my part as to whom my deliberately drawn sketch is of. There is a possibility of error in your judging whom my sketch is of, and your judgment will be based on inference and may rely on resemblance, but my judgment does neither. That difference is just a manifestation of the asymmetry of access.

            My deployment of these cases not only illustrates a relevant asymmetry of access, a kind of privileged access, but they also implicitly introduce a restricted kind of content, since what each of us has non-inferential knowledge of in these examples is basically the “subject of one’s thought.” Just what is meant by this locution is unclear at this stage. It will in fact also turn out to be ambiguous between what I will call the minimal content and the objective content. I will argue that both are required in the analysis of wha­t a thought is about; each signals what the thought is about in a different way. Neglect of either is a neglect of a critical element required for a complete analysis of the thought.

What was non-inferentially known in each of the above examples was what the minimal content represented. (Representation is discussed in chapter 4.) It is worth pointing out that one’s special access to what is represented in such cases is known straight-out. In many cases, one may non-inferentially know the subject of one’s thought and do so without having to have a further thought about the initial thought. Far too often awareness of content is mistakenly characterized as requiring second-order thought, but these cases serve as counterexamples to that claim. I will discuss this point more fully below and in chapter 2.

I examine another example[x] to  illustrate both the concept of minimal content and  that awareness of it does not require a second-order thought. Suppose I make a diagram while lecturing on the battle of Borodino. I make Xs to mark the location of Napoleon’s troops, and Os to mark Kutuzov’s. Though there are countless errors I may make in my lecture as well as in the accuracy of my diagram, it makes no sense to ask me how I know the ‘X’s represent Napoleon’s rather than Kutuzov’s troops. In this kind of case and since the diagram is mine, the ‘X’s cannot fail to represent what I intend them to represent. Suppose, on looking at my diagram, I have the thought that Napoleon had too heavy a concentration of troops in the northeast. On having this thought, I non-inferentially know that an ‘X’ represents (a certain number of) Napoleon’s troops. I know this straight-out, without recourse to a thought about this thought. It would make no sense for me to look for evidence of this, to puzzle whether an ‘X’ really represented (a certain number of) Kutuzov’s troops. Though it does make sense that you might be puzzled about such a matter, you may have to infer on the basis of evidence, which could be the simple fact that I told you, what my ‘X’s represent. While I may be wrong in thinking there were too many troops in the northeast (as a matter of military strategy), I cannot be wrong in thinking that one of my ‘X’s represents (a certain number of) Napoleon’s troops.

            Consider a new twist on this case. I may wonder what Kutuzov would have done had he been in Napoleon’s position, my speculations being based on my knowledge of his psychology. But now the Xs (and my corresponding minimal contents) represent (a certain number of) Kutuzov’s troops. When one considers this situation from the first-person perspective it is evident that here too, I know the minimal contents of my thoughts straight-out, having this non-inferential knowledge clearly does not require a second-order thought, as is commonly supposed. This is in marked contrast to another who would need to consider a second-order thought in order to acertain what my thought is about. (Further support of this last contention is offered below.)

            Intentional States and Minimal Content.

A peculiar feature of an intentional state is that it is directed at an object, the intentional object. The explanation of this directedness and what the intentional object consists in is difficult; different authors provide considerably different accounts. These differences have serious consequences. As an aid to isolating the notion of minimal content, I will explain some aspects of Searle’s characterization of intentional states[xi], as my view of intentionality is a development of his. For Searle, the intentional object just is the actual object or state of affairs to which the intentional state is directed. For example, if someone loves Sally and also believes that it is currently sleeting outside, then the intentional objects for these two states are the flesh and blood woman, Sally, and the coming down of freezing rain. But, also, these states have representative contents which represent the women and the relevant state of affairs, respectively. (Searle sometimes calls the representative content ‘intentional content.’) On the other hand, there are intentional states which have no intentional object, though they must still have a representative content. For example, Ponce de Leon, in searching for the fountain of youth, was in an intentional state which had a representative content (consisting in part of a representation of what he was seeking), but his intentional state did not have an intentional object. So, on Searle’s view, every intentional state has a representative content, and it is in virtue of this content that the intentional state is directed at an intentional object but not all intentional states have intentional objects.

            It is of paramount importance to keep the intentional object and the representative content distinct. In particular, in those cases that Searle would describe as failing to have an intentional object, one must resist the temptation of identifying the representational content with the intentional object. The temptation to do so is abetted by the consideration that the thought must be about something—it is not about nothing. Since there is no relevant actual object or state of affairs for it to be about, the reasoning continues, it must be about the representative content.This move, a shift of types, has parallels with Frege’s identification of the referent of an expression in oblique contexts with its customary sense. On my view this shift to a new type of referent is not required. [xii] The idea behind such views is that in order to avoid talking or thinking about “nothing”, there always must be an intentional object of some sort.

Pursuing Searle’s view of intentionality, sometimes the representative content is propositional in form, as when one believes that a certain state of affairs obtains; sometimes it is not, as when one desires a certain object. In either case, however, an object is “signaled” by the representative content. (The difference in the cases turns on whether just an object is signaled, or a state of affairs that involves an object.) I accept everything in Searle’s view of intentionality presented to this point, but central to my theory is a specific addition. It is my contention that in the analysis of any intentional state there are two distinct but correct ways of characterizing the object signaled, what the intentional state is about. The minimal content represents the subject of the intentional state as the agent conceives it; whereas the objective content indicates the subject an objective observer of the agent would ascribe as the subject of the agent’s intentional state. Both minimal and objective content are restricted to the subject of a thought; neither includes what is thought of the subject, what is attributed to it. Just as the minimal and objective content must be kept distinct from one another, neither is to be identified with the intentional object. The latter is an actual object or state of affairs, whereas minimal content and objective content are concepts employed in the analysis of mental states. They correspond to first- and third-person descriptions, respectively, of what is the subject of the agent’s intentional states. Nor do I may any claims as to instances of these concepts being actual ingredients of mental states, for I refrain from ontological claims.

Minimal and objective content must not be identified for several reasons: (1) they play different roles in the analysis of intentional states; (2) they may signal different objects; (3) the agent’s access to these two contents is importantly different. Because of this difference in access, I spoke above, and will continue to do so, of minimal content representing and objective content indicating what the thought is about. Though the­se te­rms mark a differenc­e, I do not claim to explain the difference here. When a generic term is required, I will speak, as I did earlier, of signaling what the thought is about.

The situation may be depicted as follows: The schema for an intentional state is Y(R) where Y is some psychological mode (believing, desiring, etc.) and R is the representative (or intentional) content and is distinct from the intentional object, as defined by Searle.[xiii] If I am right, any such schema requires a decomposition into (i) Y(F(m)) and (ii) Y(F(o)), where m is the minimal content and o is the objective content. F marks what is attributed of “the subject of one’s thought.”[xiv] Such an analysis provides for the resolution of any conflict between (1) and (2), while preserving what is correct in each.

A trouble with exclusively third-person analyses of intentional states is that minimal content is invisible on any such analysis. Since the orthodox methodologies in contemporary analytic philosophy of mind employ a strictly third-person perspective, it is not surprising that minimal content has gone unnoticed. However, once a limited first-person perspective is permitted along side that of the third-person, minimal content is seen as a legitimate component in the analysis of intentional states,[xv] a component which gives (1) its due without contradicting (2). In contrast, the employment of a strictly third-person perspective encourages, if it does not imply, the extremely counter-intuitive rejection of (1). I stress that in maintaining the importance of a subjective first-person perspective in the analysis of intentionality, I am not denying the possibility of an objective account of it. Indeed, my introduction of the notion of minimal content and the analysis of any intentional state into two characterizations, Y(F(m)) and Y(F(o)), are steps in the direction of offering just such an objective account of the subjective.

            Support for the Distinction From an Unlikely Source.

It is interesting, if not surprising, to garner some support for something like my distinction between minimal and objective content from a paper by W. V. Quine (undoubtedly against his will). His paper also brings out an additional advantage of treating intentional states as I do, viz., a basis is provided for accommodating the treatment of sentences expressing intentional states, decomposed as (i) and (ii), within the syntax of predicate logic. Quine is not alone when he asserts “that no theory is fully clear to me unless I can see how this syntax [that of predicate logic] would accommodate it.” (1994, 144) My proposal may be viewed as a suggestion that also enhances the clarity of our understanding of intentional states in this regard, one that is different from Quine’s quotation based solution that he offers in the mentioned paper.

Quine attaches great importance to the subject’s point of view in the treatment of propositional attitudes in the paper cited. For example, he notes that the failure of substitutivity in belief contexts stems from the fact that “The subordinate clause of the construction is uttered from the subject’s point of view…and…the subject…didn’t know the things were identical.” (145, emphasis added) My formulation differs from Quine’s, but it is the same kind of distinction at play. On my theory, the subordinate clause expresses the thinker’s minimal content and requires an instantiation of Y(F(m)) rather than of Y(F(o)), and the thinker does not know that m and o signal the same thing. Additionally, Quine notes that the “quantification $x (Ralph believes that x is a spy) raises an ontological problem of the value of ‘x’…” (146) I agree, as I do with his diagnosis that “The difficulty, again, is just discrepancy between the real world, to which the outlying ‘$x’ relates, and the empathized world—Ralph’s—in which the recurrence of ‘x’ is trapped.” (146) I suggest a different resolution of this problem raised by Quine: Employ expressions for the objective content when the real world is at issue, and employ expressions for Ralph’s minimal content when it is a question pertaining to the empathized world, for in the “empathized world”, the subject of the thought as conceived by the agent is at issue, not the subject that an objective observer would ascribe.

Illustrative Applications of the Concepts.

To further clarify these distinctions and to illustrate their usefulness, consider the following example. A chemist in the mid-eighteenth century might have thought that phlogiston was abundant in charcoal. The objective content of such a thought could not indicate phlogiston, since there is no such thing; nevertheless, the minimal content of the chemist’s thought, the subject of her thought as she conceived it, did represent phlogiston. While the chemist had a special access to her minimal content, she certainly had no special access to the objective content or what it indicates. This absence of special access to the objective content and what it indicates is not simply because, as is the case in the present example, that there is no such thing, for even when the indicated object exists and is coextensive with what her minimal content represents, there is still no privileged access to it, the existing intentional object. This last point is simply the familiar one that whatever privileged access there is, there is none that we have to actual objects in the world.[xvi]

            Though the objective content and one’s minimal content may signal different objects, they need not. Suppose I am correct in thinking that a certain woman is my grandmother, but I have many false beliefs about her, and there are a number of important things true of her of which I am ignorant. Suppose I entertain the thought that she was a gracious woman. The correctness of that judgment is a function of what she actually did or would do in certain circumstances. My judgment will be a function of what I believe, rightly or wrongly, about her. I may harbor so many false beliefs pertaining to her that another, someone who has a more accurate view of her, would say of me, in a colloquial vein, “He does not know her at all.” Despite my lamentable epistemological status, any of my intentional states concerning her will have a minimal content that represents her (the right woman), and I can know this non-inferentially. So both the minimal and objective content can signal the same object, in spite of all the misinformation I have regarding the intentional object.

            Note my grandmother is the sort of thing that is relatively easily individuated, even if one labors under a vast amount of misinformation about her. Lest it be thought that that is why both the minimal and objective content can signal the same subject, consider the following. Suppose I entertain the thought that quarks are difficult to experimentally detect, and the physicists who theorize about such things are right in thinking there are quarks. Although I have barely the foggiest idea of what quarks are, it is still plausible that my minimal content and the objective content of my thought signal the same object. For, at the very least, had the physicists not developed their theory of quarks, I would not have my thought about quarks. Because I have so little knowledge of quarks, the subject of my thought, as I conceive it, is intended to be whatever the physicists are talking about. I do not add to it in any way that could make what is signaled deviate from the objective content, from what those in the know take it to be.

Here, perhaps, we have operative a germ of truth in direct theories of reference. This is not to say that to be meaningful my thoughts require the objects signaled by them exist (as with Russell’s logically proper names); the phlogiston example makes this clear. Quite the contrary, it is the claim that minimal content represents something to me—whether what is represented exists or not, whether or not my representation of it is accurate, whether I have or do not have correct information—and it does so on pain of nonsense. For, at the very least, I must have some conception of what I am representing; otherwise, the very meaningfulness of the claim that my thought is about anything—as opposed to some sentence I might utter—is brought into question. In this respect, the content is minimal. 

            When my minimal content does match the objective content, but I labor under a number of false beliefs regarding it, I clearly have limited understanding of the object represented, and my ability to explicate my thoughts pertaining to it is thereby limited. I explain below the contrast between being aware and understanding what is represented by one’s minimal content. For now, I simply point out that to have a thought about something, e.g., quarks, I need have no great understanding of them; indeed, they need not even exist. When my understanding is slight, I will, no doubt, be unable to adequately explain my thought. Still, to have such a thought at all, I must be able to represent quarks in thought and, therefore, to individuate them to some degree.

Further clarification of what it is to know one’s own minimal content and why such knowledge is privileged may be obtained by considering what has already been said in the image case. Earlier I argued that it makes no sense to question whether a deliberately formed image of, say, my grandmother was of her because its being of her was constitutive of the forming of the image. The case is similar vis a vis the minimal contents of my thoughts. Constitutive of my entertaining some particular thought is that its minimal content be what I take it to be. It would not be the thought that it is, if it did not have the subject that I take it to have. Of course, as we have seen, this is compatible with the objective content of the very same thought actually signaling something different than does the minimal content or nothing at all, but one can fail to know this. There is never privileged access to the objective content of one’s thought, nor certainly, to what it indicates.

            Partial Recapitulation.

The analysis of a single intentional state requires both a minimal content and an objective content, because what the thinker takes her thought to be about may not in fact exist, or be seriously misconceived by her, or be entirely different from what it is about objectively. The occurrence of any or all of these mishaps in no way vitiates the thinker’s awareness of the subject of her thought as she conceives it, though they may mislead another who is not adopting her point of view. The minimal content represents what the agent conceives her thought to be about. The objective content indicates, from an objective, third-person perspective, what the thought is about; more precisely, what there would be inter-subjective agreement on what the thought is about. A complete analysis of the subject of the agent’s thought requires both of these. Ideally, what the two signal will be one and the same, but this cannot be guaranteed. When the two contents signal different objects, as in the phlogiston case, it is not noticed from the first-person perspective, for as soon as the agent learns of such a divergence, she will alter her minimal content to bring it in line with the objective content. To allow for either the possibility of a match or a divergence in what is signaled by these two contents, a correct analysis of an intentional state requires both; but one has privileged access only to one’s own minimal contents.

Awareness Contrasted with Understanding—Minimal Content.

I now introduce a series of cases in order to clarify a number of central ideas:

1.      Minimal content

2.      What it is for an individual to be aware of minimal content

3.      The character of an individual’s privileged access to her minimal contents

4.      The difference between being aware and understanding one’s own minimal contents

5.      The contrast between an individual’s individuating versus understanding or explicating the subject of her thought.

The terms ‘awareness’ and ‘understanding’ are widely used but vague. It is certainly questionable whether they have univocal uses in their wide applications. In this section I will contrast the two terms and show when each is preferred. (I do not claim that the contrast developed is or should be universally used, but it does mark, I think, an important difference.) Awareness, when taken as non-inferential knowledge, is preferred in discussions pertaining to the individuation of the subject of a thought, construed as minimal content. Understanding, on the other hand, comes to bear when the explication of a thought or the determination of its subject, construed as the objective content, is at issue. I will show this by way of reminding you of some familiar truths about formal theories.

            Consider a formal theory that has as one of its models the natural numbers. It is well known that such a formal theory has other models, e.g., a model which has sets, and not numbers, in its domain. So the theorems may be viewed as truths about numbers or as truths about sets. Now consider two individuals, A and B: A learned the theory as axioms about natural numbers, and only them; she doesn’t even know it could have another model. In contrast, B learned the same formal theory as axioms exclusively about sets.

            What seems quite uncontroversial is that each of these individuals can increase her respective understanding of the theory as she explores ever further the consequences of the axioms. The increase in understanding may be measured by the extent to which they can prove theorems and explain their proofs. Suppose, somewhat artificially, that A and B are identical in their symbol manipulating abilities. The first can go on in fascinating detail with truths about numbers, the second can go on in equally fascinating detail about sets. The point, of course, is that their capacity to respond differentially and appropriately to the symbols is, by hypothesis, identical. There is absolutely nothing that would differentiate their responses vis a vis the manipulation and concatenation of the symbols of the theory. Yet A and B take themselves to be proving truths about different things; their respective minimal contents are different and represent different objects.

            We, too, may acknowledge that what A and B take their respective contents to be are quite distinct, for the one it is numbers, for the other it is sets,[xvii] and we may do so without our making any ontological commitment to the nature of either of these. Indeed, we may acknowledge that A and B take themselves to be making statements about different objects even while we hold that they are ultimately not different (the numbers just are certain sets), or that neither numbers nor sets are actual objects at all (say, along nominalistic lines). In short whether our ontological commitments coincide with A’s or B’s or neither, it will have no bearing on our discussion of what their minimal contents are. Thus, while we acknowledge what is clearly the respective minimal contents of our theorists’ thoughts, it might well be that the objective content of one or the other’s thoughts indicate something other than what is represented by her minimal content.

            The above is intended to bring out, in a preliminary way, the importance of keeping separate discussion of issues surrounding symbolic manipulation, understanding, and ontology, from those surrounding minimal content. This follows from the fact that A and B are identical with respect to their symbol manipulation and understanding, yet they have different minimal contents, and that all of this is consistent with several different ontologies. Furthermore, though A’s minimal contents represent numbers, as A becomes more sophisticated, she may conclude that there are no numbers only, say, sets. Her minimal content could on occasion still represent numbers, though she does not believe that there are any. Just as we may entertain thoughts of mermaids without believing that any exist.[xviii] So symbol manipulation, understanding, minimal content, and ontology are independent in at least the ways here indicated. For the next few pages I will focus on how awareness of minimal content is largely independent of both understanding and symbol manipulation. I will also touch on the relation between the latter two. I will not address any ontological questions as such.

            Recall that understanding involves the ability to determine a thought’s objective content and to be able to explicate the thought in such a way that there is inter-subjective agreement. It is undoubtedly the case that a significant factor in the measure of one’s degree of understanding is the extent to which there is appropriate and differential responses to a relevant set of symbols. Such differential responses probably are extremely relevant to accounts of understanding. At the very least, they would be essential as a measure of one kind of understanding. However, whatever bearing differential symbol manipulating capacities have, or do not have, vis a vis the question of understanding, they do not have much to do with the concept of minimal content or with the fact that an individual is sometimes aware of her minimal content.

            When I hold that one’s minimal content is independent of symbol manipulation, I am not holding that one can have and be aware of minimal content without being able to engage in some relevant symbol manipulation. The claim of independence results from my argument that the symbol manipulation, no matter how sophisticated, may remain constant, while the minimal content, sets or numbers, varies. Qua minimal contents, A’s and B’s understanding is different, despite their identical symbol manipulating abilities; since A’s minimal content represents numbers and B’s represents sets, and this is so regardless of what the true ontology includes (numbers but no sets, sets but no numbers, both, or neither). This is not to say that either would have any understanding if they were unable to perform some relevant symbol manipulations, only that symbol manipulation is not sufficient to determine understanding since, at least in this case, it is not sufficient to determine either the thinker’s minimal content or the objective content of her thought.

            Furthermore, as the consideration below of the C theorist will show, one may be aware of the same minimal content as A, even though one’s symbol manipulating ability is considerably inferior to that of A’s. While such inferior symbol manipulating ability may well be a sign of one’s poorer understanding of numbers, it in no way diminishes the fact that the agent’s thoughts are about numbers. So just how much symbol manipulation is required for awareness of minimal content is not clear. What is clear is that different minimal contents are consistent with high levels of identical symbol manipulation (e.g., the A and B theorists) and the same minimal content is consistent with radically different levels of symbol manipulation abilities (A and C theorists). So, rapid and smooth symbol manipulation is not necessary and very little is sometimes enough. (Compare my discussion of Van Gulick below.) To this extent, and in this sense, awareness of minimal content is independent of symbol manipulation. Let me explain further.

            Consider a third individual, C, who like A, learned the theory strictly as a theory about numbers, but is not nearly as adept at proving theorems as is A. On the basis of the disparity in their symbol manipulating skills, we would say that A has a better understanding of the theory of natural numbers than does C. Yet, C, though bumbling in her symbolic manipulations, still knows without inference the minimal content that those symbols or her corresponding thoughts have for her. Her difficulty consists in providing proofs of particular truths about numbers. In spite of this, whether it is a statement she can easily prove or one she cannot prove at all, she is equally aware that it is a statement about numbers (whatever their ultimate ontological status). The present important point about minimal content is that C is directly aware of the same content in both cases and, in either case, it is correct to say that C has the same minimal content as does A, in spite of A’s greater facility at theorem proving, thus, greater understanding. In contrast, B’s minimal content is different from that of both A’s and C’s, despite the fact that B’s symbol manipulation ability is identical to A’s.

            These considerations of our three theorists depend on projecting ourselves into their respective circumstances, adopting their first-person perspective. When we do so it is evident that while each can be aware of the minimal content of her thoughts as she manipulates symbols, the appropriate differential responses themselves have very little, if anything, to do with either what that minimal content is or the agent’s awareness of it. What they take the objects to be is the same or different independently of those responses, independent in the sense explained above.

            C’s limitations in her theorem proving may well indicate limitations in her understanding of number (or, perhaps more accurately, limitations in her understanding of number-theory), though this limitation neither affects what her minimal content represents, nor her awareness of it.[xix] Any proper account of understanding, undoubtedly, will be one that admits of degrees of understanding. Indeed, the varying degrees of understanding may well be reflected, if not partly constituted by, the varying degrees of successful symbol manipulation. In important contrast, awareness of minimal content does not admit of degrees. Either our theorist is aware she is proving (or attempting to prove) theorems about numbers or she isn’t. (Recall that both A and C have the same minimal content despite their radically different ability to prove theorems about numbers.) The relative ease with which she is able to construct such proofs appears irrelevant to her awareness of minimal content. She may prove, say, “2 is a prime number,” or “There is no greatest even number,” with ease, but despite her best efforts is unable to prove, say, “Every number has a prime factorization.” In any of these cases, however, C is equally aware that her minimal content represents number, be it a specific one, as in the first sentence, or number in general, as in the ­latter two.

            Thus, it is not the theorist’s symbol manipulating ability that determines what her minimal content represents nor her awareness of it. This is so even if it is maintained that her differing abilities to prove the two statements is a consequence, or an indication, or even constitutive, of her different degrees of understanding of the two statements. Clearly then, no matter what the full account of understanding is, it is importantly different from one’s awareness of what the minimal content of one’s thought represents. Our understanding admits of degrees, our awareness of minimal content does not. While we may well have more or less understanding of, say, numbers (as evidenced by how adequately and thoroughly we can explain or prove that numbers have certain properties), it is nonsense to say that one’s awareness of numbers has increased or decreased (or is more or less than someone else’s awareness of numbers). C, our weak number theorist, knows her thoughts are about numbers. Period. It does not matter whether it is a thought that she can easily prove or explain, or not. Nor does it matter that generally she is not very good at explaining the properties of numbers, nor at proving theorems about them. These abilities bear on the strength of her understanding and gives rise to talk of degrees of understanding. Such abilities and the resulting degrees are not transferable to awareness of minimal content.

            One of the distinct advantages of the notion of minimal content is precisely that it permits our avoidance of the troublesome issues surrounding that of understanding, while still allowing an advance in the analysis of intentionality. Of course, understanding does contribute to knowledge of the objective content and, so, to ‘content’ thus understood. But though understanding ultimately plays some essential role in our knowledge of the objective content, it is not similarly involved in our knowledge of what our minimal contents represent. That it is distinguishable is important for current discussions of intentionality, since so little is understood about understanding. The scope of the term ‘understanding’ is clearly broader, if not also vaguer, than is one’s non-inferential knowledge, awareness, of one’s own minimal content. One can know the latter with very little understanding of the thought of which it is the minimal content, as the discussion of theorist C brings out (or the example of quarks discussed earlier).

Understanding, Explicating, and Individuating.

Towards the end of his (1988), Burge draws a distinction, which is related to the above, between thinking a thought and explicating it. It is correct to do so. I can think a thought that mercury is an atomic element, his example, without being able to give a proper explication of that thought. When I am unable to explicate my thought, presumably it is because I do not adequately understand it. I may know very little about the periodic table and how mercury fits into it, its atomic structure, etc. But Burge mistakenly conflates explicating with individuating when he goes on to say: “One clearly does not have first-person authority about whether one of one’s thoughts is to be explicated or individuated in such and such a way.” (Burge 1988, p. 662) While one does not have first-person authority over the explication of one’s thought, say, that mercury is an element, to have the thought at all one must be able to somehow individuate mercury as what is represented by the minimal content of that thought, no matter how ignorant one might be, or how many misconceptions one might have of mercury.  Explicating is strongly related to understanding, but individuating a subject of a thought, in the sense of being aware of the one’s minimal content, is rather independent of explicating and understanding. I will criticize some of Burge’s relevant views in detail in chapter 6.

            There are two problems here of concern. First, the one just mentioned of conflating individuating and explicating. Secondly, there is an ambiguity in the reference to a thought. Take the latter first. At the outset I indicated a schematization of a thought as Y (R) and stated that this schematization is ambiguous between the two schemes: Y(F(m)) and Y(F(o)) (where Y is some psychological mode, F is some attribute, m is the minimal content, and o is the objective content of the thought). Now insofar as the quotation from Burge pertains to a thought in the sense of Y(F(o)), I have no disagreement with him. There is no first-person authority over one’s thought in this sense of thought. However, when the thought is construed as Y(F(m)), things are quite different. Here, the first problem mentioned, the conflation of individuating and explicating comes to bear.

            To have a thought at all, in the sense of entertaining it, the thinker must be able to individuate the thought, even when she is unable to explicate it. An inability to explicate manifests her lack of understanding of the thought but, even so, the thinker would have first-person authority with regard to how the thought and its minimal content are individuated. Individuating a thought or a minimal content is different from explicating either of them, as I now will explain.

            To individuate a thought Y(F(m)) one must (1) be able to differentiate it from other thoughts—either because the minimal content is different or something different is attributed to the same minimal content. Obviously, one must have already individuated the minimal content to do this. And (2) one is able, when the occasion arises, to reidentify it as the same thought as one had before. To explicate a thought, on the other hand, one must explain how that thought relates to others. One must state a good number of properties the represented object is to have and, where appropriate, show that it possesses those properties. In the case of numbers, this amounts to providing proofs of, and commentary on, theorems. Again, our C theorist is unable to do this to any great extent. So, our C theorist cannot explicate many of her number thoughts. Her shortcomings, however, do not interfere with her ability to individuate her thoughts and minimal contents, but her inability to explicate her thoughts does provide reason for saying she does not understand them or, at least, that she does not understand them well.

            One can be aware of one’s thought Y(F(m)) and minimal content (m) without being able to do much explicating; nevertheless, one cannot even begin to explicate a thought unless one has already individuated it. This is true no matter how inadequate one’s explications may be. Understanding entails the ability to explicate a thought, and doing this presupposes one has individuated the thought. But part of individuating a thought is individuating its minimal content. Therefore, understanding presupposes minimal content.

            In saying this I am not holding that either the identity of concepts or the meaning of terms are ultimately determined by the individual. So when Burge elsewhere maintains that “…the meanings of many terms…and the identity of many concepts…are what they are even though what the individual knows about the meaning or concept may be insufficient to determine it uniquely,” (Burge 1992, p. 46) there is no disagreement, in one sense. Our individual ignorance is varied and extensive. Such ignorance, however, pertains to the objective content and what may be correctly attributed to it. Still, in another sense, I do disagree with Burge: there is a first-person authority over what one’s thought is about in the sense of minimal content, the subject of the thought as conceived by the thinker, regardless of the extent of the thinker’s misconceptions pertaining to it. (These matters will play an important role when I discuss, chapter 6, Burge’s famous thought experiment.)The greater the misconceptions the less able one is to correctly explicate one’s thought, the less one understands it. One neither has first-person privilege over the objective content nor of what counts as understanding, but this is different from individuating or knowing the minimal content of one’s thought; here, one does have first-person privilege.

            I repeat for stress that the expression, ‘the subject of one’s thought’, is ambiguous between the minimal content and the objective content of the thought. Minimal content is invisible when a strictly third-person methodology is employed. When there is some dim awareness that something like it is operative in discussion of certain issues, e.g., when one unwittingly considers one’s own case, the singular importance of the case is not generally recognized; typically, it is ignored or consigned to the periphery as an annoying detail. If one reads Burge’s comments as pertaining to the objective content, I am in complete agreement with him. Specifically, there is no privileged access to the objective content of one’s own thought. But, I have argued that it is essential that any analysis of intentional states countenance minimal content in addition to the objective content, and the situation is quite otherwise with respect to minimal content, as I have tried to show.

            It is precisely because minimal content is a distinct element of analysis from that of the objective content that we are able to avoid talk of understanding and still have something significant to say about intentionality or representation. Once we distinguish awareness of minimal content from knowledge of the objective content or understanding the thought objectively, we can go further than we have in showing that the capacity for responding differentially to symbols is not sufficient for determining awareness of minimal content. To see this we can collapse our two individuals A and B into one, dropping the restriction that the individual doesn’t know about alternative models. Call our new theorist ‘D’. The syntactic strings that she manipulates are identical, whether the theorem she is proving is “about” numbers or sets. Yet, on some occasions D’s minimal contents represent numbers and on others sets; she may alter what the symbols represent to her at will. (This is similar to the “new twist” on the Kutozov example discussed above.) Surely, it is wrong to say that her differential symbol manipulations account for, or in any way determine, the different contents that she can be aware of. Nor, for the same reason, does the converse relation hold. So, appropriate differential responses to symbols on the one hand, and awareness of one’s minimal content on the other, are once again seen to be utterly independent (in the sense indicated above).

            These last considerations also bring out the asymmetry of access to minimal contents from the first- and third-person point of view. Our ascertaining D’s minimal content is necessarily based on inferences from her behavior and it does make sense from that perspective to question whether we have made the right inference. In contrast, she makes no such inference; she knows her minimal content straight-out, and it makes no sense to question whether she is right about this.

Higher-Order Thoughts are Not Necessary.

It is also important to notice that in considering what D’s thought is about we must move to a second-order thought about her thought, given our third-person perspective of her. It is easy to see, therefore, that if we were to restrict ourselves to an exclusively third-person methodology in the analysis of such cases, we would endorse the view that D must have a second-order thought to be aware of what her thought is about, for in considering another’s thought, we naturally adopt a third-person perspective. But it is clear that from D’s first-person perspective, this is not required—she does not need to have a second-order thought. This asymmetry between us and D can be overcome by entertaining the other beliefs we think D holds and the circumstances in which she thinks she finds herself and consider what the subject of our thought would be if we were similarly situated. Of course, we can be wrong in what we conclude, but such is the way for all empirical matters that we bear no special relation to. (Compare chapter 3 and last section of chapter 4.)

The various versions of the view that a second-order thought is required for this kind of knowledge are known as Higher Order Theories of intentionality or HOTs.[xx] Insofar as HOTs require a second order thought in their analysis of intentionality, they must be rejected, if I am right that D’s awareness of her own minimal content is first-order. The earlier examples—the A, B, C theorists, the grandmother image, and the battle diagram, with and without the “twist”—serve as additional counterexamples to the widely held claim that a second-order thought is required for such knowledge. But, again, the truth of my claims can only be recognized when one adopts the first-person perspective, the considerations that I appeal to are invisible from a third-person perspective. That this is so is a direct manifestation of the important asymmetry between the first- and third-person perspective that I have argued for. If this is correct, the above also makes clear how an exclusively third-person methodology may lead us astray; thus, the cases provide reason to reject an exclusive use of that methodology. The very idea that a second-order thought is required for knowledge of this sort is an artifact of a strictly third-person methodology. In the next chapter, I will show that a particular attempt to account for privileged access within the confines of a third-person methodology, one that relies on a higher order thought, fails.

The New Problems of Absent and Inverted (nonphenomenal) Content.

I have argued in my discussion of the A-D theorists that no matter how correct and elaborate the symbol manipulations are, this is not sufficient to individuate one minimal content from another; still, the content is easily and directly individuated and differentiated by the agent, and this is achieved from her first-person perspective. The various theorists (A, C, and D) differentiate numbers (or particular numbers) and are able to distinguish them from sets (or particular sets). This ability neither requires that they make an inference, nor that they entertain a second-order thought, though both are required from a third-person perspective.

I have also argued that a difference in minimal content for an individual does not require a difference in symbolic manipulation (compare theorist D). Thus, the symbol manipulation is neither necessary nor sufficient for determining different minimal contents. Since the symbol manipulation is not sufficient to determine either A’s or B’s minimal content, there is here a sort of indeterminacy of minimal content from the third-person perspective, though each has a definite minimal content that is different from the other’s and is unproblematically accessible to the one having it. (This kind of indeterminacy problem is discussed in detail in chapter 7.) As a result of this, we have an analogue to the problem of the inverted spectrum. Given the behavior of A and B as observed from a third-person perspective, we can attribute thoughts about numbers or sets to either, but we cannot determine on that basis which is which, an inverted minimal content problem.

 We can now go further. Such manipulations, which are accessible from a third-person perspective are not even sufficient to say that the symbols even have any content for the processor. Programs for proving theorems can and have been developed. Some are able to prove more difficult theorems than others, as well as doing it more efficiently.[xxi] Still, we have no reason to think that the symbols have any content for the program or for the machine running the program.[xxii] We could even build into the program the “disposition” to display ‘number’, or ‘2’, when asked appropriate questions. The incorporation of this little programming task yields no basis for holding that the computer running the program has special access to its minimal content, not even that it has minimal content. None of this would provide reason to hold that the symbols it manipulates have content for it. [xxiii]

            Indeed, the same result that applied to a computer running a program could be achieved with a person. Let E be yet another theorist who learns the same formal theory as did A and B, but learns no semantics for it beyond that for the logical constants. E may still become quite adept at manipulating the symbols in accordance with the formal rules and uninterpreted axioms of the theory, as adept as A and B. But with E, we go beyond this indeterminacy of minimal content relative to the third-person perspective, just discussed, and move to absent content. All the relevant symbol manipulations may be realized in E’s activities, but E has no minimal content relative to these symbols or their manipulations.

            The choice of the expression ‘absent content’ is obviously chosen because my point parallels one that Ned Block raised against the functionalist long ago (1978 in Block, 1980), what he dubbed ‘The Absent Qualia Argument.’ The general idea that he argued for, in opposition to functionalism, was that while all the functional roles that a functionalist might require could be in place in some system, yet it would still be plausible (in some such cases) to deny that the system has any qualia. My arguments establish that all the relevant functional roles might be in place, yet the minimal content is not thereby determined (A or B theorists, inverted spectrum analog) or is in fact totally absent (E theorist, absent qualia analog, which is also the intentionality analogue to the so-called zombie case).[xxiv] However, my support for absent and inverted minimal content is stronger than is that for the inverted spectrum, absent qualia, and zombie problems, since my arguments do not depend are mere conjectures of possibilities—they are verifiable from the first-person perspective.

Grasping Content.

Having exploited the familiar idea of multiple models for a single formal theory in my discussion of the A-D theorists, additional light may be cast by considering a converse relation: cases where the same object is characterized by a variety of formal theories. G. Rota, et. al., (1989) argue that such situations are quite common in the practice of mathematics. Among other examples, they discuss the real number line and groups as cases where the practicing mathematician has a pre-axiomatic grasp, an understanding of these objects that is free from any particular axiomatization.[xxv] This grasping of the object is critical, they claim, to identifying different axiomatizations as being axiomatizations of the same object.[xxvi]

            They recognize that a student learning a theory will sometimes be unavoidably dependent on some particular axiomatization of that theory. This is a dependency that is especially strong in a case such as group theory, less so in other cases, e.g., the real number line. But, in any case, it is a dependency that is overcome once she becomes familiar with the theory. To cast this in my terms, becoming familiar with the theory is becoming aware of groups, becoming (as I would put it) aware of a minimal content that represents groups, in a way that transcends the particular axiomatization by which the student was introduced to groups. They argue that “To the mathematician, an axiom system is a new window through which the object, be it a group, a topological space or the real line, can be viewed from a new and different angle that will reveal heretofore unsuspected possibilities.” (382, 1989)

            In holding the above, the authors are neither arguing nor claiming that these “grasped” objects have some special ontological status: In particular, they stress that they are not arguing for Platonism; they insist that they “...are [just] acknowledging the actual practice of mathematics...” (382, 1989) Nor are they arguing that the mathematician is exercising any special faculty, mental or otherwise, when grasping these objects. They take a neutral stand with respect to both the ontological status of these objects as well as to the character of the faculty by which one is aware of them. Their analysis is developed from the first-person perspective of a practicing mathematician, not from a third-person perspective. So, their approach is methodologically similar to my own, and it is appropriate to characterize the mathematician’s grasping of the objects of his own thoughts as another manifestation of one’s being aware of the minimal content of one’s thought, a minimal content that may well match the objective content.

            Partial Summary.

Earlier I utilized the familiar fact that we can have different models for the same formal theory and, as far as one’s own thoughts are concerned, one can recognize these models as having different objects apart from any formal manipulations. More recently, I have incorporated what others have argued, viz., given that the same object can be treated by different formal theories, the recognition that they are different formalizations of the same object requires the representation of this object, in some sense, apart from the formalization. (These points are important for my interpretation of Quine’s Indeterminacy of Translation and the distinction between representers and information bearers, discussed in chapter 7 and 4, respectively.) As a result, we can make sense of these results by employing the concept of minimal content; minimal content is a representation of an object apart from formalizations. Secondly, that we each have asymmetrical access to our own minimal content. Thirdly, that our non-inferential knowledge of our minimal contents is independent of overt symbol manipulation (though the latter probably plays a role in an account of the more complicated intentional notion of understanding). Fourthly, there are analogue problems of absent and inverted qualia pertaining to nonphenomenal content. All of this has been achieved without making any presuppositions as to the ontological status of either minimal content nor to what is thereby represented.

            More on Grasping—Frege Comparison.

The minimal content of a thought is the subject of a thought as conceived by the thinker. Some may understand minimal content as a Fregean sense, understood as a “mode of presentation”, for it is natural to take the expression ‘as conceived by’ as though it were a certain mode of presentation, or the subject of the thought under a certain “guise”. But this would be seriously misleading. My view of minimal content departs significantly from Frege’s concept of ‘sense.’ First a minor difference: both minimal content and objective content can, by definition, only signal an object, whereas Fregean senses may also signal states of affairs. A much more critical difference is that I hold that a single thought is properly analyzed as having both a minimal content and an objective content; in contrast, for Frege a thought has only one sense. Finally, and most importantly, minimal contents, depending as they do on how the individual conceives things, are subjective in a way in which Frege insisted senses are not.[xxvii]

This last point is of great importance when making a comparative evaluation of Frege’s claim that we grasp senses with my claim that we grasp minimal contents. The idea of grasping plays a central role in both of our views. But since minimal contents are unabashedly subjective and are in part constituted by the thinker, the idea that we grasp our own minimal contents seems clearly right. Just how a thinker is to grasp an objective Fregean sense, a sense that she herself does not even in part constitute, appears a bit mysterious and is not addressed by Frege. The concept of minimal content is clearly different than, and has a clear advantage over the concept of Fregean sense, at least in this respect. From the first-person perspective, it is constitutive of my entertaining some particular thought that it have the minimal content that it does. It simply would be a different thought if it did not have the subject that I take it to have. (Cf. the discussion above on the individuation of a thought and minimal content.) The grasping of minimal content by the thinker is unproblematic because the act of thinking that particular thought is constitutive of its minimal content.

            Given the thinker’s role in constituting her minimal contents, a role she does not play regarding the objective content, we should expect that the ways minimal content and the objective content determine their respective references are distinct. For the agent is in no better position to know what is indicated by the objective content of her thought, in virtue of its being her thought, than is anyone else—she in no way constitutes this, which is in marked contrast to the agent’s ability to know what is represented by her minimal content.

            In making my case for the necessity of both minimal content and the objective content in any adequate analysis of intentional states, there have been no presuppositions, certainly no claims, as to the ontological status of either kind of content. I have emphasized how access to them is importantly different, and it is only with respect to minimal content that the first-person perspective is crucial. While I am able to thus reconcile the two apparently conflicting features of intentional states mentioned at the outset, I also go beyond it inasmuch as I hold that minimal contents are a function of how we conceive the subjects of our thoughts, and since one cannot take one’s thought to be about anything which one does not have the capacity to (rightly or wrongly) conceive, our having the capacity to be aware of our minimal content is a necessary condition for our having intentional states.[xxviii]

Explaining the Appearance.

Given that it is at least initially plausible that we sometimes have privileged access to the content, in some sense, of our intentional states, a special restriction is placed on any proposed account of intentionality: It must either include an account of this feature or explain why it appears to be the case but is not.

            The force of this restriction might be made clearer by an analogy. George Berkeley argued that material substance did not exist (or that the notion itself was incoherent). But ordinary thinking, ordinary experience, seems to present material things as an uncontroversial datum. In sections 34-81 of A Treatise Concerning Principles of Human Knowledge, Berkeley particularly addresses himself to objections based on this type of consideration, that his view denies various obvious truths, or obliterates various obvious distinctions, such as that between real things and chimera.[xxix] In point of fact, while rejecting material substances, he took great care to show exactly why one cannot, say, simply walk through real walls while one can walk through imagined walls, even though both are collections of ideas. (Whether he was ultimately successful is entirely immaterial to the little moral I wish to draw from my discussion of him.) In doing so, he thereby offered an explanation of how certain obvious data, which seemed to refute his view, were not only consistent with but explained by his account.

            That Berkeley did explain why such obvious data is in accord with his theory is precisely why Samuel Johnson’s famous “refutation” of Berkeley,[xxx] though cute in its vivid and immediate appeal to what seems to be a conflicting datum, can never be taken seriously. Had Berkeley failed to provide these explanations, Johnson would have had a formidable point. Unlike Berkeley, many modern day philosophers prefer to reduce or eliminate the mental rather than the material; nevertheless, if they are to avoid a Johnsonian type refutation (in reverse), they, too, must explain certain appearances, though their task is quite different from that which Berkeley confronted.

It is the kind of privileged access to one's content brought out by the various examples discussed earlier that constitutes the sort of datum, which produces such severe obstacles for third-person methodologies, e.g., externalism with regard to mental content. And it does so as a datum. It does not require some elaborate theoretical analysis (internalist or otherwise) to shore it up so that it may play this role. Not only do we have mental states but, also, we (at least sometimes) directly know (some of the content of those states). When I speak here of ‘being aware of the content’ or ‘knowing what the content is’, I repeat that I do not refer to its ontological status, but only to how it is described in virtue of the subject's non‑inferential knowledge of it. Whatever its ultimate ontological status, it is data of the utmost importance.

In an article by Heil (to be discussed in the next chapter) he cites with approval Davidson's claim that we are victims of a certain misleading picture of the mind, to wit, “... the content of one's mental states are taken to be based on inward glimpses of those states or on the grasping of particular entities' (P. 247). Davidson recommends aban­doning such a picture, and Heil claims that once we do, “. . . we remove at least one of the reasons for supposing that externalism undermines privileged access” (P. 247). I applaud this abandonment. However, this picture is not the only obstacle for an externalist. Abandoning it does not, thereby, clear the way for an externalist account of privileged access, not even the limited sort of privileged access that I and even some externalists are concerned with preserving. In arguing that externalism cannot account for any privileged access to contents, I have not relied on any analysis of this privileged access that presupposes such “inward glimpses of those states or on the grasping of particular entities”; for I am neither committed to this model of introspection, nor to the grasping of entities. I explicitly reject the idea that minimal content is an entity. It is a tool of analysis that enables us to correctly say what the subject of the agent’s thought is—as she would conceive it—it is one that requires a first-person methodology. Distinguishing it from objective content, which is probably what the externalist is getting at, enables us to reconcile (1) and (2). Ignoring it leaves us with paradoxes and quandaries.

The metaphor used to characterize or single out the datum is inconsequential relative to the datum, itself. How we are aware of content, whether by inward glimpses, or whatever, is immaterial to what is at least the apparent fact that we are sometimes aware of (at least part of) the content of our states. It is only the latter that I require. Additionally, whether those contents are genuine entities, of whatever sort, or not entities at all is, also, immaterial to this, at least, apparent fact. I do not think that content is an entity, nor that we have inward glimpses of it, nor did my arguments depend on any such assumptions. Still we are sometimes aware of content, though we are not aware of it in the sense of having special or detailed knowledge of its nature or ontological status. To hold the latter, would be to invoke the thoroughly discredited view that our privileged access to our own mental states gives us infallible and incorrigible knowledge of the very nature of those states. I certainly do not hold that view.

This epistemically privileged, first-person access to part of our content (of some of our occurrent states) lays claim to being an important datum. In contrast, asserting the existence of some unspecified external condition and some unspecified determination of content relation, as the externalist does, is not only unacceptably vague, its alleged relevance to an account of privileged access is a highly theoretical claim. Moreover, I have argued that even allowing appeals to external conditions determining content and coupling such appeals to second-order thoughts fail to account for the datum of privileged access. [xxxi] As such, and in so far as it does not square with the datum, it must be repudiated.

Nor should these last comments be construed as my harboring some view such as that there exits “pure data”, that there is a “given”, or that common sense beliefs are somehow epistemically privileged. I hold no such views. Such views are, to my mind, also, thoroughly discredited. (See my discussion of some of these and related issues in chapter 5.) Though there are no “pure data”, it is far from evident that the indicated datum—that sometimes we have privileged access to (at least part of) our own content—is such that it is essentially dependent on this particular (wrong‑headed) picture of the mind. Rejecting the picture does not, ipso facto, eliminate the datum in question. Even once that picture is rejected, there remain puzzles, puzzles, I have argued, that cannot be solved using a strictly third-person methodology. As I see it, any externalist account must fail in accommodat­ing privileged access to content. If that is correct, the only option for one employing a strictly third-person methodology is to dissolve the datum by explaining it away as mere appearance, that is, by showing privileged access to our own contents is just an illusion.

Eliminativists or reductionists are not the only ones that must explain these appearances. Anyone attempting to give an account of intentional states in terms that are prima facie of a radically different type must, in some way, make plausible how the elements of this different type could give rise to the salient features of experienced intentional states. We must be able to make plausible how, say, neurons firing and various chemicals transported over synapses, or how the functional organization of the brain, or whatever the proposed theory is, can give rise to one’s (sometimes) having special access to the minimal content of one’s own intentional states. I offer somewhat more than a framework for the solution of this problem in chapter 3.

A problem for all, regardless of one’s philosophical persuasion, is either to provide an explanation of the appearance that we sometimes do have special access to the minimal contents of our thoughts or show that this is mere appearance. Being everyone’s problem, failing to resolve it does not of itself count against one’s theory. However, if one has failed to resolve this problem and a consequence of one’s theory is an outright denial of the appearance, the theory must be rejected. In the next chapter, I will show that a wide class of orthodox theories in the philosophy of mind fail to account for this appearance. The conclusion that any analysis that relies on an exclusively third-person perspective cannot accommodate the relevant kind of privileged access is fortified in chapters 4 and 6.

Some Additional Comparisons: Descartes and Searle.

Does our awareness of our minimal content amount to a privileged access of the Cartesian sort? The latter is typically portrayed so that the contents of one’s mind are completely and infallibly transparent to oneself. I am not convinced that this is a fair portrayal of Descartes, but I will not attempt here a scholarly defense of a different reading of Descartes. This common view of Descartes, whether it is his view or not, has been widely, and rightly, criticized. Certainly, the objective content is not completely and infallibly transparent to the agent. I do not challenge this. So long as ‘content’ is construed as the objective content there is no privileged access to it. But this is no reason for all privileged access to any kind of content of one’s thoughts to fall into disrepute. Minimal content is a distinct content from either the objective content or the representative content; it is only to minimal content that we sometimes have privileged access. Any infallibility with regard to what the minimal content represents results from the senselessness of an attribution of error rather than our having some special ability or faculty. Thus, a non-Cartesian form of our privileged access is preserved.[xxxii]

            Minimal Content and the Chinese Room. The problem that I am occupied with concerns a subjective, nonqualitative content. In chapter 3,  I will discuss how discussions of subjectivity and consciousness are distorted by concentrating on the phenomenal, but first I turn to how my view differs from others whose discussions do focus on nonqualitative content. A notable example is John R. Searle’s. My view is most sympathetic to Searle’s, and it shares a number of features. Importantly, the capacity for awareness of minimal content, being a necessary condition for an agent’s having intentional states is a commitment to intrinsic intentionality.

Searle’s idea of intrinsic intentionality is famously, or depending on your view, infamously, presented in his Chinese Room thought experiment; one might wonder whether I have presented a remodeled Chinese Room.[xxxiii] In this thought experiment, the central concept is that of understanding; the question whether the room system has subjective phenomenal states is not at issue (pace Van Gulick, see below). Though Searle does not speak explicitly in terms of content in that work, it is clearly involved: The question, “Does the system understand?” is the question whether the strings of marks have any content for the agent (or system) manipulating them. Or, as Dretske sometimes put it, “Is there any content for the system itself, as opposed to merely being such that content can be assigned to the system?” (I do not mean to suggest that Dretske and Searle are as one here. There are obvious and serious differences.) My earlier discussion of minimal content and an interpreted formal theory shares certain features with Searle’s Chinese Room. Casting the issues as I have, however, has several advantages over talk of understanding and natural languages.

            First, rather than trying to contrast the Chinese Room with a genuine speaker of Chinese who understands (with all the attendant murkiness of this notion), the contrast is instead drawn in relatively simple and clear terms: straight out, one is aware that one’s minimal content is, say, sets or numbers. So we take several steps back from understanding to minimal content. Crucially, and unlike understanding, minimal content does not come in degrees (see above). As a result, we do not have to rely on, or attempt to resolve, conflicting intuitions as to whether or not a system such as the Chinese Room understands. In the cases I considered, each individual clearly is aware of the their respective minimal contents—despite radically different levels of understanding, whether those contents are the same (e.g., A and C theorist compared) or different (A and B theorist compared).

            Another advantage is that a common move functionalists, and others, make against the Chinese Room is diffused. They often wish to bring in causal interaction of the system with its environment. Since such interactions are relevant in our own case for language understanding, they argue that such causal interaction must also be extended to the room system, prior to its getting any serious attention. I think this move is fundamentally mistaken (see chapters 2 and 3), still, the importance of such interaction may initially seem plausible when it is a question of language understanding in general. In part, it may seem plausible because of the pervasiveness of terms referring to things that do occur in our environment. Whatever the reasons, it is clear that such causal interaction with the environment (other than the trivial ones for input and output) is not even initially plausible when we speak of abstract contents such as numbers or sets.

Closing Thoughts.

The central point may now be noted. By preserving the distinction between minimal and objective content, as well as the different access to them, one is able to jointly maintain the two seemingly contradictory, but crucial features of intentional states: Firstly, the individual who’s state it is, is often in a privileged position as to ascertaining what it is about (minimal content). Secondly, an individual may neither know, nor be in a better position than someone else to ascertain, what his own state is about (objective content). The resolution of these two claims requires the recognition that potentially two different subjects are at issue, and this requires the admission of the legitimacy of the first-person perspective. On the other hand, to restrict the investigation of intentionality to an objective third-person perspective results in concentrating on some equivalent of the objective content. Such a methodology of necessity ignores the subjective aspect of how the thinker conceives things, minimal content. Since that feature appears to be among the distinctive features of intentional states, unless strictly third-person accounts can either accommodate this feature or show it to be a mere appearance, such accounts must be rejected.[xxxiv]

The methodological dominance of a strictly third-person perspective in studies of phenomenal states and intentionality artificially restricts attention to features available to that perspective, features such as causal relations, symbol manipulations and, as I will argue more fully in the next chapter, the appearance that second-order thoughts are required. But if I am right, these features are not common to all subjective or intentional states and, so, they introduce distorting factors. On the other hand, the concepts of minimal content and our awareness of it are more fundamental than that of understanding, for the latter presupposes the former (see above). Furthermore, if I am right, minimal content and awareness of it are elements that are common to the analysis of all conscious states (see chapter 3). Thus, being simpler, more fundamental, and common to all such states, they provide a better basis for the analysis of mind, subjectivity, and intentionality than do the concepts of phenomenal state, understanding, second-order states, or symbol manipulation. Since the important central role that minimal content plays can only be seen when we project ourselves into the positions of the various theorists considered, when we adopt the first-person perspective, it is not surprising that it has gone largely unnoticed.




[i] I will explain the influence of these theories on the separation of these aspects in chapter 3.

[ii] For an example of the latter see Charles P. Siewert, The Significance of Consciousness, Princeton University Press (1998).

[iii] This feature only depends on the fact that we sometimes know without inference part of what our thoughts are about in some sense. It is not that we know all of it in this way. It is not that we always non-inferentially know part of it. It is not even that we need explicitly entertain what the content is. To make my case, it is enough if the agent has been, or could become non-inferentially aware of part of the content of his thought in some sense; I do not need that he is always so aware.


[iv] Daniel Dennett, (1987, p. 5).


[v] Dennett, (1987, p. 5).


[vi] Dennett, (1987, p. 7).


[vii] David Armstrong in his “Is Introspective Knowledge Incorrigible?” (The Philosophical Review LXXII, 4 (October 1963): 417-32, reprinted in  The Nature of Mind, ed., David M. Rosenthal, 126-133. Oxford University Press 1991) argues against the claim that all introspective reports are incorrigible. Even assuming his argument for this is sound it does not refute my claim that some introspective knowledge is incorrigible. Moreover; Armstrong’s arguments assume that incorrigibility claims must be confined to a “perceptual instant”, but none of my arguments for infallible knowledge of minimal content are contingent upon this assumption. Similarly, R. Nisbett, T.D. Wilson, and A. Gopnick—See McGeer Journal of Philosophy 1996, p. 496—argue that psychology rejects the presumption that first-person reports should be accepted, barring good reason to override; they appeal to various evidence in support of this. Here too the kinds of error they find are irrelevant to my concept of minimal content, defined below, and our infallible knowledge of it.

[viii]  Donald Davidson (1987, 1984) is an important exception. He argues for the importance of first-person authority and points out that some ignore or dismiss it, while others fail to see the seriousness of the problem it poses for externalist views of mental content. He notes (1988) that while Burge (1988) acknowledges first-person authority, he does not adequately show how such authority can be reconciled with his view. See my discussion of Burge in chapter 6, where I show this failure undermines his famous thought experiment.  Though the problem is serious, Davidson argues that first-person authority can be reconciled with an externalist view of content. The asymmetry of access that I argue for is like that upon which Davidson’s bases first-person authority. This access turns on the agent’s non-inferential knowledge of the content of one’s thought, but the content that I argue for which is governed by first-person authority is much narrower than Davidson’s. For him, there is special access to our occurrent beliefs, desires, and other propositional attitudes as these are ordinarily understood (1987, p. 447), what is often referred to as wide content. The content that I argue we each have privileged access to is far more restricted than this; still, as will be evident, it is not to be identified with what is normally referred to as narrow content. Putnam (1975) first introduced the terms ‘wide’ and ‘narrow’ in connection with these issues.

                Some other attempts to provide for a limited privileged access that is neither infallible nor transparent and available to the externalist may be found in Heil (1988) and Shoemaker (1990). But see below where I argue that Heil’s attempt fails and chapter 2 where various problems are raised for externalist attempts to provide for even this limited privileged access.

[ix] Introduced by John Heil (1988).

[x] Heil 1988, p. 248. Although the example is Heil’s, I exploit it in ways that he would not endorse.

[xi] Though Searle’s most extended account is to be found in his (1983), see also his (1984) for a particularly clear discussion of the differences between intrinsic intentionality and derivative or metaphorical intentionality.

                It is worth noting that acceptance of Searle’s framework for an analysis of intentionality, as he presents it in the early chapters of Intentionality, does not commit one to the acceptance of the causal account of intentionality he gives in the final chapter.

[xii] More on the relations between Frege’s view and mine later.

[xiii] This way of putting it is borrowed from Searle (1983, 6). The decomposition of this schema discussed immediately below goes beyond Searle.

[xiv] One might think that the same rational that leads to distinguishing m and s would force a similar distinction with respect to F. This is not so. The point in making the distinction between minimal content and the objective content is that there are strong grounds for analyzing the ‘subject of the agent’s thought’ in two different ways. For purposes of analysis, there are no similar grounds for holding that F should be similarly analyzed. Reconciliation of any similar apparent conflict between reports from the first- and third-person perspective pertaining to F is adequately resolved by appealing to m and o, together with the recognition that F applies truely or falsely to the subject of the thought, whether one is considering m or o. Of course, when m and o do not signal the same thing, it may happen that F(m) and F(o) differ in truth-value.

[xv] Holding this does not entail that whenever I am in a state that has minimal content that I must also be explicitly aware of that minimal content. (Cf. My warning in note 8.) The awareness spoken of here is simply the non-inferential knowledge of what is represented by the minimal content, not the noticing of some special feature of the state or its minimal content. And that’s a good thing, for it would seriously count against my view. As to why, see, e.g., Saul Kripke (1982), especially the latter part of part 2.

[xvi] Talk of “actual objects in the world” is to be understood here, and throughout, as simply those objects upon which there is inter-subjective agreement within a given community. No deep ontological claim is being made. The reasons for this and some of the consequences will be apparent in chapters 7 and 8.

[xvii] Someone may object at this point and insist it is not clear that the content is numbers in one case and sets in the other, since it is not even clear what numbers and sets are apart from the role they play in a theory. Compare Paul Benacerraf (1965) or Michael D. Resnik (1981). If the question were what does the objective content indicate, I would be inclined to agree, but that is not the issue. There is a clear sense in which what A and B take their respective objects to be are distinct. No matter how that difference eventually gets cashed out, and no matter what the ultimate nature of numbers and sets—as minimal contents they are distinct. Taking the position that numbers and sets are nothing apart from the relations that they enter into (and, thus, insofar as they enter into the same relations, they are the same contents) is to give a theoretical account of the ontological status of numbers and sets, one that might even be right. Similarly, if one holds that numbers are ultimately certain kinds of sets, that there are no numbers. Thus, such considerations bear on our understanding of numbers and sets; they do nothing to explain the appearance of awareness of minimal contents, and different ones at that.

[xviii]  I suggest that minimal content can be useful in an analysis of fictional objects. Normally, an agent attempts to have her minimal contents signal the same subjects as do the objective contents of her thoughts; whenever she learns they do not match, she aligns her minimal content so that it does. Things are otherwise with fictional objects. A fictional object is simply the subject of a thought as conceived by the thinker, her minimal content, that represents an object that she knows does not exist, she knows the objective content of her thought is empty, but efforts toward realignment are deliberately excluded.

[xix] To challenge this one would have to give an extremely strong holistic account of meaning. One that would have as a consequence that the content of ‘number’ is different for our theorists A and C because of their differing ability to manipulate the symbols. If we were to accept this, we might wonder whether two individuals ever could have the same content, inasmuch as it is doubtful that two individuals would ever actually have identical symbol manipulating abilities. (The unrealistic assumption that A and C have identical symbol manipulating abilities was made strictly for the sake of argument.) The results here further support the importance of distinguishing explicating and individuating. (Compare below pp. 10-12.) See also my discussion of Burge in chapter 6, where I discuss the conditions for saying two individuals have the same concept even when there is not an exact match.

[xx] Armstrong ’68, ’80, Lycan ’87, ’92, and Rosenthal ’86, ’90, ’91 are examples of advocates of higher order theories of intentionality.

[xxi] To the extent that appropriate symbol manipulation is a component of understanding, to that extent it would be correct to speak of such programs as instantiating understanding of number theory, and some instantiating a better understanding than others. Whether the sense in which such programs “understand” is anything more than metaphorical is another question, because whether appropriate symbol manipulation is merely a measure of understanding or is actually a constituent or sufficient for it is yet to be determined. Here we are touching on problems raised by Searle in his famous Chinese Room, discussed below.

[xxii] Cf. Fred Dretske (1985). He states: “unless the symbols being manipulated mean something to the system manipulating them, their meaning, whatever it is, is irrelevant to evaluating what the system is doing when it manipulates them.” (Pp. 27-28) I agree with Dretske that the important and relevant case is that in which the symbols mean something for the system processing them, though I disagree with him on what this amounts to or whether his approach can adequately deal with it. (Compare my discussion of Twin-Earth in chapter 2.) In addition, I prefer to put the point in terms of the system’s privileged access to its minimal content. For Dretske’s more recent development see his (1997).

[xxiii] The capacity to have access to one’s minimal content is closely related to a symbol’s having content for the processor. The idea of a symbol having content for a thinker is developed further in chapters 4, 6, and 7; in chapter 2, I argue that strictly third-person methodologies fail to provide for this.

[xxiv] I will turn to a more explicit treatment of functionalism in the next chapter and to the importance of the ideas of absent and inverted content in chapter 3.

[xxv] The ‘pre’ in ‘pre-axiomatic’ should not be understood temporally. Perhaps a more accurate prefix would be ‘extra.’ It may well be questionable whether the objects are grasped apart from any theory or background language whatsoever (cf. chapters 7 and 8), though I do not think that Rota, et. al. are making such a strong, unrestricted claim.

[xxvi] They state:

Since there is no one group, but groups come in incredible variety, such a pre-axiomatic grasp of the notion of group cannot be attributed to familiarity with a single group. One could argue that such an understanding is derived from familiarity with several groups and their “common” properties, but this would amount to begging the question, since familiarity with more than one group presupposes an unstated understanding of the concept of group that permits one to recognize several instances as being instances of the same general mathematical structure. (Rota, et. al., 1989, pp. 381-382)

[xxvii] In contrast to the minimal content, the objective content is objective and, in this respect, it is more like a Fregean sense, or a truncated Fregean sense, for unlike the latter, it only indicates objects, not states of affairs. Of all the concepts that I have employed, however, it is probably that of representative content, understood with an objective content rather than with minimal content, which is closest to a Fregean sense.

[xxviii] I repeat, holding this does not entail that whenever I am in a state that has minimal content that I must also be aware of that minimal content. (Cf. my warning in note 1 and note *** [Kripke].)


[xxix] Compare my discussion of Galileo in chapter 4, who also attempts to explain the appearances that seem to conflict with his view..

[xxx] It is reported that Johnson uttered, “Thus, I refute Berkeley”, while kicking a stone.

[xxxi] I will argue in the next chapter that the externalist leaves both the determination relation and the external condition required for his view in a highly unspecified way. See section entitled, “Indeterminacy in Determining Content and More on Second-Order Thoughts”.

[xxxii]  One might argue that my distinction between minimal content and objective content aligns with Descartes’s distinction between objective and formal reality, respectively. See Meditation 3.

[xxxiii]  This is developed in Searle (1980). The case somewhat simplified is essentially as follows: An English speaking person is locked in a room and provided with a large ordered store of sequences of Chinese characters together with a set of elaborate syntactic transformation rules. The latter are formulated in English, but operate on strings of Chinese characters. The individual in the room does not understand Chinese. However, given new strings of Chinese characters from someone outside the room, he is able to apply the transformation rules to the received strings so as to produce new strings that he then passes out of the room. A Chinese speaking person on the outside is passing meaningful Chinese sentences to the man in the room, and the latter, unknown to himself, is returning meaningful and appropriate Chinese responses. 


[xxxiv] For example, those of Davidson (1987) and Heil (1988). They argue that externalism can accommodate the kind of privileged access herein discussed, though what they take the access to be to differs from what I do. In chapter 2, I argue that Heil in particular is unsuccessful and that any externalist account must fail, including those like Davidson’s, which appeal to the natural history of the agent.