Is self-knowledge compatible with externalism?

 

 

 

Pierre Jacob

Institut Jean Nicod, CNRS,

1 bis avenue de Lowendal,

75007 Paris, France.

Email: jacob@ehess.fr

 

 

 

Abstract

 

Externalism is the view that the contents of many of a person's propositional attitudes and perhaps sensory experiences are extrinsic properties of the person's brain: they involve relations between the person's brain and properties instantiated in his or her present or past environment. Privileged self-knowledge is the view that every human being is able to know directly or non-inferentially, in a way unavailable to anybody else, what he or she thinks or experiences. Now, if what I think (or experience) is not in my brain, then it seems indeed as if I cannot have any privileged authoritative first-personal access to the content of what I think. Hence, externalism seems inconsistent with privileged self-knowledge. The purpose of this paper is to provide a road towards a conciliation between self-knowledge and externalism.

 

 

Keywords externalism, informationally based teleosemantics, self-knowledge, introspection, compatibilism.

 

 

For quite obvious reasons, externalist theories of mental content are widely thought to be incompatible with privileged self-knowledge. Broadly speaking, externalism is the view that what a person thinks is not entirely determined by his or her inner cognitive resources but depends on items and properties present in his or her environment. Privileged self-knowledge is the view that each person has a direct privileged access to the contents of his or her own thoughts. This tension is the topic of the present paper.

 

1. The Problem

Externalism is the view that the contents of many of an individual's propositional attitudes and perhaps also - although this is more controversial - the contents of an individual's sensory experiences are not intrinsic properties of an individual's brain: they supervene on relations between an individual's brain and properties instantiated in his or her environment. In other words, content is an extrinsic property of the individual's brain: it may involve relations between the person's brain and properties instantiated in his or her present or past environment. Just as the meaning of a token of a word of a natural language is not an intrinsic property of the token, the content of a mental representation is not an intrinsic property of the mental symbol (if any) tokened in the individual's brain. As Putnam (1975) famously put it, “meanings ain't in the head”.

 

1.1. Varieties of Externalism

I think it is useful to distinguish a social and a non-social version of externalism. According to the non-social version of externalism, content may supervene on relations between an individual's brain and physical, chemical, biological properties instantiated in the individual's environment. According to the social version of externalism, what an individual thinks may even supervene on what other members of the individual's community think. Social externalism may be illustrated by thoughts involving so-called “deferential” concepts. A deferential concept is a concept expressed by a word whose meaning (e.g., “elm” or “arthritis”) an individual layperson does not quite master or masters incompletely. He or she will therefore not unreasonably be inclined to defer to experts from his or her community to sharpen his or her understanding of the concept. At the root of social externalism is the view that incomplete or partial mastery of a concept does not prevent a person from having genuine thoughts involving that concept.[1]

Notice that although the content of an individual's thought involving a deferential concept may depend upon what other members of his or her community happen to think, presumably the non-conceptual content of an individual's sensory experience does not so depend either upon what other members of his or her community think or upon what they experience. The non-conceptual content of a sensory representation does not involve concepts - let alone deferential ones[2]. I take it that the contents of my visual experiences do not depend upon what other members of my community experience. This, however, is perhaps not so clear for the non-conceptual contents of an individual's emotions: the non-conceptual content of an individual's emotion may perhaps depend upon the non-conceptual contents of the emotions of members of his or her community. Depending on one's view of the relation between concepts and percepts (or thoughts and sensory experiences), the content of an individual's sensory experience may perhaps depend upon what other members of his or her community think if it depends upon one of his or her thoughts involving a deferential concept and if the latter in turn supervenes upon what members of his or her community think. Assuming that there is a level of visual perception which is modular by Fodor's (1983) criteria, i.e., informationally encapsulated or non-epistemic in Dretske's (1969, 1990) sense, then the content of an individual's non-epistemic visual experience will not depend upon what other members of his or her community happen to think. Unlike non-social externalism then, social externalism clearly applies to the contents of conceptual thought, not so clearly to the non-conceptual contents of an individual's sensory representations. In the sequel, only externalism of the non-social variety will be relevant.

 

1.2. Self-Knowledge

Privileged self-knowledge is the view that each of us is able to know directly or non-inferentially, in a way unavailable to anybody else, what he or she thinks or experiences.[3] If what I think (or experience) is not in my brain - if what I think is metaphysically constituted by relations between my brain and properties instantiated in my brain's environment -, then it may seem as if I cannot have any privileged authoritative first-personal access to the content of what I think. To put my cards on the table, I think there are two separable pieces of knowledge involved in full self-knowledge: one is the knowledge of what one thinks (or the knowledge of the content of one's thought); the other is the knowledge of the fact that one thinks (the knowledge that one experiences so-and-so or the knowledge that one believes such-and-such). Full self-knowledge involves both kinds of knowledge: knowledge of what one thinks and knowledge that one thinks it.[4]  My view is not that in introspection, one may as a matter of fact come to know the content of one’s own thought without knowing that one thinks it. My view is that, for the analytic purpose of understanding how introspective self-knowledge arises, full self-knowledge can be usefully decomposed into two components.[5]

Although I argue in favor of a conciliation between externalism and self-knowledge, I am thus less of a compatibilist than Burge (1988) is. On his view, having privileged self-knowledge about a thought consists merely in having the ability to think that thought - whether or not having this ability requires the thinker to stand in some appropriate relation to items in his or her social and non-social environment. To know that I am thinking that water is wet, it is enough that I be able to think that water is wet - whatever is required for me to think that water is wet allows me to know that I think that water is wet. Unlike Burge (1988), I claim that my grounds for knowing what I think are not ipso facto grounds for knowing that I think it: in order to know that I believe that water is wet, it is not enough that I be able to think that water is wet, I must also appropriately form a higher-order belief about my belief that water is wet and hence apply my concept of belief to myself.[6] In order to know that I know that water is wet, I must possess the concept of knowledge (which itself involves the concept of belief). I may be able to think that water is wet and be devoid of the concepts of belief and knowledge.

 

1.3. Informationally based teleosemantics

In this paper, I want to argue that a particular kind of externalism is compatible with partial privileged self-knowledge - privileged self-knowledge of the first sort, i.e., knowledge of what one thinks. The kind of externalism I have in mind is a particular brand of a teleosemantic approach to mental content which combines teleological ideas with informational semantics. According to this approach, which I will call the indicator-functional version of teleosemantics and which is roughly Dretske's (1988, 1995), to have content is to have an indicator (or information-carrying) function: to mean that x is F is to have the function to indicate F. The relevant notion of function is the teleological notion: a device's teleofunction (or proper function) is what the device is supposed to do.[7] Hence, it can capture the normativity of content. Like most teleosemanticists, I think that the best account of a device's teleofunction is Wright's (1973) etiological account, according to which Z is the proper function of trait X in system S iff Xs were selected for doing Z because X's doing Z contributed to the proliferation of S's ancestors. In other words, a device's function is what it has been designed to do. Device S (or some state of S) will carry information about property F if it covaries (or is correlated) with instantiations of F. S will have the teleological function to carry information about F if it has been designed by some selection process or other to carry information about F.[8]

Clearly, such a teleosemantic theory is externalist since, on this view, what I think is a matter of two broad extrinsic factors: first, states of my brain will not carry information about some property instantiated in my environment unless they are nomically correlated with instantiations of the property in question. Secondly, they will not have indicator functions unless my brain has been suitably shaped by some phylogenetic and/or ontogenetic historical processes. Prima facie, the claim that an informationally based teleosemantic theory of mental content is compatible with some amount of privileged self-knowledge should seem surprising. It should seem surprising precisely because on this approach, the content of what I think and experience depends (or supervenes) on nomic correlations between properties of states of my brain and properties instantiated in my environment; and it depends (or supervenes) on both the phylogenetic and the ontogenetic history of my brain. And I certainly do not have authoritative first-personal privileged access to nomic correlations between properties of my brain and properties instantiated in my environment. Nor do I have authoritative first-personal privileged access to the ontogenetic history of my brain, let alone to its phylogenetic history. So if correct, the claim that the indicator functional theory of mental content can accommodate some aspect of partial privileged self-knowledge is interesting.

           

2. The Scope of the Principle of the Reflexivity of Content

2.1. Introspection as Displaced Perception

Now, Dretske (1995) has an intriguing account of introspection, i.e., the process whereby a person comes to know about his or her own mind. Dretske calls his account “displaced perception”. Displaced perception, which applies first and foremost to knowledge of the external world, is a process whereby a person comes to know through perception facts involving things which he or she does not perceive. Dretske gives a number of telling examples of this process: one comes to know how much one weighs by looking at the hand of a bathroom scale. What is perceived is the scale. What is learned is a physical fact about oneself. One may come to know that someone else is at the door by hearing the bell ring. What one perceives is a sound. What one learns is the presence of a human being at the entrance of one's home. One may come to learn about distant events in the world by reading a newspaper or by listening to the radio, and so on. What goes on in all these cases is that one derives a conceptual representation of some fact from the perceptual representation of a distinct fact. So the idea is that one derives a conceptual representation of the fact that e.g., k is F from perceiving the G-ness of h. According to Dretske, this is typically what happens in the case of introspective knowledge of one's perceptual experiences.

In this paper, I will not say much about displaced perception. However, the difference between perceptual knowledge of the external world and implications of perceptual knowledge of the external world will be relevant later. My present goal is to argue for the (partial) compatibility between externalism and first-personal authoritative access to the contents of one's own mind. First, I want to spell out a principle which I will call the principle of the Reflexivity of Content, and which, I claim, falls out of the indicator-functional version of teleosemantics. Furthermore, I will argue that, if correct, this principle entails a limited amount of privileged self-knowledge or first-personal authoritative access to the contents of an individual's thoughts and experiences. Thirdly, I will examine the scope and limits of the kind of privileged self-knowledge which follows from the Reflexivity of teleosemantic content. Finally, I will briefly examine three objections to the limited kind of self-knowledge which falls out of the indicator-functional view of mental content.

 

2.2. Stating the Principle of the Reflexivity of Content

Suppose system S - a thermometer (or a watch) - has the function to carry information about the temperature (or to keep track of time). Call F the indicated property. S - the thermometer (or the watch) - has the function to indicate the determinable property F. When S represents the temperature as being 20°C (or the time as being 5:15 PM) - a determinate property -, S is ipso facto in a state which carries information about its being 20°C that S is representing the temperature to be (or about its being 5:15 PM that S is representing the time to be). Whether the temperature is actually 20°C or not, if S represents it to be 20°C, then S must be in a state which carries information about the way it represents the temperature. If broken (or malfunctioning), S might represent the temperature inaccurately: it might misrepresent the temperature. If it does, then it does not carry information about temperature: it does not indicate the temperature. But even if S does not carry information about the temperature - even if it misrepresents it - still S must occupy a state carrying information about how it is representing or misrepresenting the temperature.

This principle, which I call the principle of the Reflexivity of Content, can be stated thus:

 

Reflexivity of Content:

A system cannot represent (whether correctly or incorrectly) things to be F unless the system is in a state carrying information about F-ness, i.e., about how it represents things to be. In other words, in the process whereby system S represents something to be F, S has automatically available information about the content of its own representation.

           

Note that the principle holds of representational systems with indicator functions. I am not assuming that such systems are ipso facto minds. A representational system derives its indicator functions either from an intentional or from a non-intentional process of selection. If the latter, then it is a mental representational system: it can be a mind. If the former (i.e., if it is an artefact), then it is a non-mental representational system. Although a non-mental representational system S may occupy states with representational content, it cannot, I assume, have experiences, beliefs and desires. Only systems whose indicator functions derive from a non-intentional selection process may have experiences, beliefs and desires. If S derives its informational functions from an intentional process, it can nonetheless carry information. In other words, information may be available to it. From the fact that the information that p is available to S, it does not follow that S threreby knows that p. S will not know that p unless it can form the belief that p. If S is a non-mental representational system, it will not be able to form such a belief and hence will not be able to know that p. Thus, the principle of the Reflexivity of Content merely says that by virtue of representing the instantiation of a property, a representational system - be it mental or non-mental - has available information about the property being represented. It does not say that the device knows which property is being represented.[9]

Now let us assume that beliefs are representations. On the indicator-functional teleosemantic theory of belief contents, when I form the belief that p, I have privileged information about the fact that what I believe is that p or that p is the content of my belief. Whether I believe correctly or incorrectly that p, my believing that p provides me with information about the way I represent the world - about the fact that I represent the world in such a way that ‘p’ is true. Let us assume that perceptual experiences are representations. Then on the indicator-functional teleosemantic theory of the non-conceptual content of perceptual experience, when I have a visual experience as of a yellow lemon, my visual experience carries information about how I represent the world too. Whether my visual experience is veridical or hallucinatory, it provides me with privileged information about how I am representing the world, namely as containing a yellow lemon in front of me.

 

2.3. A Teleosemantic Argument for the Principle of the Reflexivity of Content

Now, I will briefly sketch the teleosemantic argument in favor of the principle of the Reflexivity of Content. According to the theory I accept, S will not represent things to be F unless S has the function to indicate (or carry information about) F. If S works according to its design - if it performs its indicator function -, then S represents F correctly and it does carry information about F. So if S correctly represents something as F, then it does carry information about its being F that it is representing. In virtue of correctly representing something as F, S is in a state carrying the information that F is being instantiated. This information is thus available to S. So we get the principle of the Reflexivity of Content for veridical representations. What about non-veridical representations? If S misfunctions, then it will misrepresent F when F is not being instantiated. Suppose that S does not perform its indicator function according to design: suppose that S incorrectly represents or misrepresents F as being instantiated when F is not instantiated. If S represents x to be F when x is in fact G, not F, then S does not carry information about the property instantiated by x, namely, G. However, in misrepresenting x as being F, S turns out to be correlated with what would be the case if, contrary to fact, S were performing its function. Even if S misrepresents the fact that F is being instantiated when it is not, still S carries information about the property which would be instantiated if it were doing its job according to design. If F is the property which, by design, S has the function to indicate, then F is the property which would be instantiated if S were representing the fact that F is being instantiated and S were performing its function according to its design. Hence, if S has the function to indicate F, when S represents something as being F, then S carries information about the fact that it is representing F, whether it is representing F correctly or incorrectly. This is the principle of the Reflexivity of Content.

 

3. The Limits of the Principle of the Reflexivity of Content

Now I want to examine the scope and limits of the teleosemantic principle of the Reflexivity of Content and the kind of self-knowledge it secures. According to this principle, a device cannot represent something as F without having available and reliable information about the content of its representation. Does it follow that it also has information about the fact that what it is doing is representing? If a device has information about how it is representing what it represents, does it ipso facto have the information that it is representing? A device cannot represent x as F without occupying a state carrying the information that the way it represents x is as being F, not G. If I have a visual chromatic experience representing a lemon in front of me as being yellow, then I occupy a state carrying the information that yellow is how I represent the lemon to be. The question which the principle of the Reflexivity of Content raises is the following: Does it not entail that any device with representational powers must have the information that it is a representational device? From the fact that one cannot represent x as F unless one is in a state carrying information about how one is representing x, it seems to follow that one must ipso facto have the information that what one is doing is representing - that one is engaged in representation. If a creature has a visual experience of x as F, it would seem to follow from the principle of the Reflexivity of Content that the creature has the information that it has experiences. If a creature believes that x is F, it would seem to follow from the principle of the Reflexivity of Content that it automatically has information about the fact that it has beliefs. I want to argue that such consequences do not follow. In effect, I want to argue that information is not closed under known logical implication.

 

3.1. Perceptual knowledge is not deductively closed

Let us turn our attention temporarily from self-knowledge to knowledge of the external world. I want to reflect on a similar failure of closure: perceptual knowledge is not closed under known logical implication in at least two respects.

Consider a typical case of perceptual knowledge. Suppose I see a yellow lemon in front of me. From an informational point of view, if my visual system is in good working condition, if the yellow lemon is in good viewing conditions and if the illumination is appropriate, when I see a yellow lemon in front of me, my visual chromatic experience provides me with the information that there is a yellow lemon in front of me. If my visual system is reliable, then the information which it provides is reliable. If so, then the visual information that the lemon in front of me is yellow is reliable. I am thereby in a position to know that there is a yellow lemon in front of me or that the lemon in front of me is yellow. And this is a piece of knowledge I gained via visual perception.

Arguably though, having the visual experience of a yellow lemon will not suffice for me to know that there is a yellow lemon in front of me. Having the visual experience is one way to know. Another way to know would be to be told by someone trustworthy that there is a yellow lemon in front of me. But having the visual experience is not enough in and of itself to provide knowledge. Presumably, a creature may have a visual experience of a yellow lemon without believing that there is a yellow lemon in front of her. Roughly, the visual experience is what Dretske (1969) has called non-epistemic vision. However, in order to know that there is a yellow lemon in front of one, one must believe that there is. And arguably one cannot believe that there is a yellow lemon in front of one unless one has the concept of a lemon. So, provided that I have the required conceptual apparatus and provided that my visual system is reliable, then having the visual information that there is a yellow lemon in front of me can provide me with the knowledge that there is.

Now, I am tempted to suppose that the proposition that the lemon in front of me is yellow (or the proposition that there is a yellow lemon in front of me) could not be true unless there were physical objects or unless there were an external world. In other words, the truth of the proposition that there is a yellow lemon in front of me presupposes the existence of lemons, the existence of physical objects and the existence of an external world. However, as argued by Dretske (1969, 1981), I want to claim that from the fact that I see that there is a yellow lemon in front of me and from the fact that nothing can be a yellow lemon unless it is a physical object, it does not follow that I see (or can see) that there are physical objects (let alone that there is an external world).

I can see the yellow lemon in front of me. I can also see that there is a yellow lemon in front of me. If my visual experience is reliable and provided that I have the concept of a lemon, I thereby come to know that there is a yellow lemon in front of me. It can be true that there would not be a yellow lemon in front of me for me to see unless it were a physical object and therefore unless there were physical objects. Nonetheless, the fact that I see the yellow lemon in front of me does not entail that I see that there is a physical object, let alone that there is an external world. Seeing a yellow lemon in front of one has a sensory phenomenology. Seeing that there is a yellow lemon in front of one too has a sensory phenomenology. There is something it is like for one to see a yellow lemon. However, it is by no means obvious that there is a phenomenology - a sensory phnomenology - to seeing (or thinking) that there are physical objects. Or rather, if there is a phenomenology, it is not obvious that this phenomenology is sensory. If a lemon is yellow, then it is colored. Although there is a sensory phenomenology to seeing a yellow lemon, it is by no means obvious that there is a sensory phenomenology to seeing (or thinking) of a lemon that it is colored.

The view I want to hold then is that I learn that there is a yellow lemon in front of me by visual perception. There are true propositions which are implied by what I learn by visual perception. But these true propositions are not such that I can learn about them by visual perception. Thus, perceptual (or visual) knowledge of the external world is not closed under known logical implication. Let us say that there is a broad and a narrow sense of “perceptual (e.g., visual) knowledge”. In the broad sense perhaps, anything which I come to know by exploiting visual perception counts as perceptual knowledge. But in the narrow sense, not every consequence of what I am perceptually aware of counts as a piece of perceptual knowledge.

In fact, within an informational framework, there is a distinction to be made between the source of information and the information channel.[10] The former is what the information transmitted is about. The latter is the means by which the information is communicated to a receiver. From an informational point of view, I can thus get knowledge by perception if the perceptual channel which delivers the information is reliable - if, in other words, the information delivered is reliable. However, a signal cannot provide a reliable information about a source and at the same time supply the information that the information provided is reliable. It cannot, while delivering a reliable piece of information about a source, provide the additional information that it is reliable or that the channel through which the piece of information has been delivered is a reliable channel. A watch cannot simultaneously tell you what time it is if you did not know it in advance and tell you that it is reliable. You can tell whether a watch is reliable by looking at the position of its hands if you antecedently know what time it is. If you do not know antecedently the time, you will learn what time it is by looking at the watch's hands if the watch is reliable. But you will not jointly learn what time it is and whether the watch is reliable just from looking at its hands.

I learn that the phone is ringing by hearing the phone ring. I learn that my wife is wearing perfume by smelling her perfume. I learn that lemons are acid by tasting lemons. All these things I learn by perception (be it visual, auditory, olfactory or gustatory). And, if my perceptual systems are reliable, all these things are things I come to know by perception. All these facts are facts of which I have perceptual knowledge in the narrow sense. But I do not thereby come to know by perception that my visual system, my auditory system, my olfactory system and my gustatory system are reliable. I do not have perceptual knowledge in the narrow sense of the fact that my perceptual systems are reliable.

Thus, the situation with respect to knowledge of the external world is the following. I come to know in the narrow sense various things by perception. What I come to know by perception in the narrow sense implies a number of propositions such as that there is an external world. Although such propositions are logically implied by what I have learned by perception, I do not come to know by perception all the consequences of what I know perceptually in the narrow sense. I do not know by perception - if I know it - that there is an external world. If the information which is provided by my perceptual system is reliable, then what I learn by perception is perceptual knowledge in the narrow sense. From the fact that my perceptual system is reliable and from the fact that what I learn by perception is knowledge, it does not follow that I can know by perception that my perceptual system is reliable. In other words, from the fact that my perceptual system is reliable and from the fact that what I learn by perception is knowledge, it does not follow that I can know by perception that I am not hallucinating a yellow lemon when I perceive one or that my senses do not deceive me.

 

3.2. Self-knowledge is not deductively closed

I now want to argue that the situation with respect to self-knowledge is parallel to the situation with respect to knowledge of the external world. If the principle of the Reflexivity of Content is correct, then a representing creature has information about how it is representing whatever it represents in one of the senses of “how it is representing what it represents”. A creature has privileged first-personal authoritative access to the contents of his or her thoughts and experiences. But it does not follow that she thereby has privileged first-personal authoritative access to the fact that she is thinking, believing or experiencing. A creature can have beliefs and experiences without having either the concept of belief or the concept of experience. According to the indicator-functional theory of mental content, a representational system has the function to provide information about properties instantiated in its environment. It does not thereby have the function to provide information about properties instantiated by itself. In virtue of exercising its representational capacities, a creature gets information about how it represents things in its environment. It does not thereby learn that it is a representational system. In order to learn that it is a representational system, it must have the concept of representation. In order to entertain the thought that it is thinking, a creature must have the concept of thought. In order to entertain the thought that it has beliefs or experiences, a creature must employ either the concept of belief or the concept of experience.

Let me sum up what I have argued. According to the principle of the Reflexivity of Content, in representing x as F, a system has available information about the fact that it represents x as F and not otherwise. Does it not follow then that such a system is ipso facto able to know that it is representing something as F? No, it does not. In order to know the latter, it must represent the fact that it is representing. And unless it possesses the concept of representation, it will not be able to do so. Externalism of a certain sort is therefore compatible with partial self-knowledge to the following extent: a representational system has direct authoritative access to how it is representing what it is representing. It does not have direct authoritative access to the fact that what it is doing is representing. Another way to put the point would be to say that I have first-personal authoritative access to what I think, believe or experience (to the content of my thoughts, beliefs and experiences), not to the fact that I am thinking, believing or experiencing. I have first-personal authoritative access to the contents of my thoughts, not to the attitudes under which I entertain these thoughts.

 

4. Three Objections

4.1. First Objection

At this point, I can well imagine the following objection. When I introspect and come to know that I believe such-and-such, it certainly does not seem to me as if I first become immediately aware of the content of what I think, and then only through a second stage do I learn that my belief is a belief and not a desire, a hope or a fear. It rather seems as if it all comes as a package: introspection makes me aware of my beliefs qua beliefs, of my desires qua desires, just as it makes me aware of their contents and attitudes. Awareness of the difference between belief and desire seems as immediate as the difference between any two propositional contents.

I agree that introspection cannot disentangle awareness of content from awareness of the attitude under which the content is being entertained. The reason is, I think, that we do have higher-order concepts such as the concepts of representation, thought, belief or desire. And we cannot refrain from making use of these concepts when we reflect on our own minds. Consider the distinction between what Dretske calls epistemic vision and simple or non-epistemic vision (or visual representation) of things in the external world - a distinction often made in terms of the distinction between the conceptual and the non-conceptual content of a representation. As I said above, unlike the exercise of epistemic vision, the exercise of simple or non-epistemic vision can proceed, it is claimed, independently from any conceptual resources and/or belief-forming mechanism. A philosopher who argues for this distinction is not claiming that unless a person can refrain from exercising his or her conceptual abilities and/or from forming beliefs, he or she will not be able to achieve simple or non-epistemic vision. Rather, what is being claimed is that simple or non-epistemic vision of an object does not require the formation of a belief - non-epistemic vision is compatible with the fact that the person fails to exploit his or her conceptual resources and to form any belief about the object. It is, therefore, consistent with the distinction between epistemic and non-epistemic vision that we normally form beliefs and exercise our conceptual abilities upon seeing things.

Similarly, I want to say, we have available information about the content of what we think. But arguably, in making introspective use of this information, we exploit our conceptual resources about our own minds, i.e., we exploit our concepts of belief, desire and experience. Indeed, we can hardly resist exploiting these concepts. And this is why introspectively it does not seem as if we are aware of what we think independently of our awareness of the way we entertain the content of our thought - independently of our awareness of the fact that our thought is e.g., a belief or a desire. Importantly, I claim that we directly have available information about the content of our thoughts (and experiences), not about the attitude under which the content is being entertained, hence, not about the fact that we think or experience. However, I do not claim that we know the content of our thought without knowing that we are thinking. Full introspective knowledge is knowledge both of what we think and that we think it. And furthermore, in turning available information about the content of what we think into knowledge of what we think, we must apply some attitude concept or other. A creature cannot come to know the content of its representation without deploying the concept of representation. We cannot come to know the content of our thought without applying the concept of thought. Application of such concepts is the price to pay for transforming available information about the content of what we think into knowledge of our mind.

So, although full introspective knowledge is knowledge of what one thinks and knowledge that one thinks it, I think that the distinction is justified. Its justification, it seems to me, stands or falls with the distinction between propositional content and propositional attitudes which in turn holds for unreflective thoughts. On my view,[11] although non-mental representational devices (artefacts such as gauges, thermometers and the like) do not have attitudes such as beliefs and desires, still their states may have contents, i.e., they may represent the instantiation of such and such a property in the environment. Such a device may even represent the content of another device. So, for example, the state of an alarm system may represent the state of a thermometer which in turn represents the temperature.[12]

The reason I stress the fact that the distinction between content and attitude holds for unreflective thoughts is that, in the case of reflective thoughts, the distinction does not always hold. Suppose, for example, that (in Cogito-like thoughts) what I think is that I am presently thinking or that what I presently believe is that I have beliefs. In such cases, the distinction between content and attitude may collapse. And so does the distinction between knowledge of what one thinks and knowledge that one thinks it. The latter collapses in so far as (or because) the former collapses. But in so far as one and the same content can be entertained under different attitudes, the distinction between knowing what one thinks and knowing that one thinks it does, I think, obtain. Notice, however, that such reflective thoughts are different from introspective self-knowledge. Although they are higher-order thoughts, such reflective thoughts do not in and of themselves amount to knowledge about the content of one’s own mind. My thought that I am presently thinking and my belief that I have beliefs are typical reflective thoughts. But they differ from my introspective claim to know what I believe when e.g., I am aware that I believe that water is wet. Unlike Burge (1988), I do not think it is sufficient to be able to believe that water is wet in order to know that one believes that water is wet, since one may have this belief and lack the concept of belief. In order to know that one believes that water is wet, one must hold the belief that water is wet and form the higher-order belief that one believes it. The reflective thought that one is thinking is just that: a reflective thought. It is not a claim about self-knowledge. Nor is the belief that one has beliefs. Both are reflective thoughts that may arise as logical consequences of any thought or any belief.

 

4.2. Second Objection

A second and powerful objection is the following. Consider Putnam's (1975) celebrated thought-experiment. Oscar on Earth and Twoscar on Twin-Earth express different natural kind concepts when each of them utters a token of the word 'water': the former refers to water, the latter to twater. Any externalist account of the contents of concepts - and, as I pointed out, the informationally based teleosemantic account clearly is an externalist account - is bound to claim that, before the discovery of the molecular structure of water, Oscar on Earth and Twoscar on Twin Earth must have had introspective access to different concepts with different contents. But how could they? Prior to the discovery of the molecular structure of water, water on Earth and twater on Twin Earth could not be distinguished by any method and by anyone. Only after the discovery of the molecular structure of water on Earth could chemists distinguish the molecular structure of water from that of twater. So how could Oscar's introspectively formed belief about his water concept differ from Twoscar's introspectively formed belief about his twater concept? This challenge raises a problem for any externalist account of natural kind concepts - not just an informationally based teleosemantic account.

As I see it, the best response available to an externalist - any externalist - is to bite the bullet and accept the view that Oscar and Twoscar do have introspective access to two different concepts and that they did so even before the discovery of the molecular structure of water. They do so even if, were Oscar to move from Earth to Twin-Earth or were Twoscar to move from Twin-Earth to Earth, neither could recognize any difference. The introspective situation of Oscar and Twoscar with respect to their distinct concepts (of water and twater) is rather like the situation I might be in were I acquainted with two identical twins: Sara and Mara. My seeing Sara differs from my seeing Mara in virtue of the fact that Sara is not Mara. My belief that Sara has blond hair differs from my belief that Mara has blond hair for the same reason. For all that, I might be completely unable to distinguish my beliefs and my visual experiences just because I might be unable to distinguish Sara from Mara.[13]

Does not the strategy I am recommending here amount to a recognition that externalism is not after all compatible with priviledged self-knowledge of the content of one’s own concepts? I think that the strategy I am recommending is consistent with self-knowledge properly construed. What, I think, it is inconsistent with is not self-knowledge but the view that self-knowledge is in effect self-ominiscience.

Suppose I cannot tell the difference between my seeing Sara and my seeing Mara. Does it follow that I am not really having two genuine and genuinely distinct visual experiences? As an externalist, I do not think so. My belief about my visual experience of Sara might be indistinguishable from my belief about my visual experience of Mara. Similarly, my belief about my visual hallucinatory experience might be indistinguishable from my belief about my visual experience caused by a real object. Still, the two experiences will differ in that only one of them is an experience of an object. When I compare my belief that Sara has blond hair with my belief that Mara has blond hair, I form a higher-order belief about my two distinct first-order beliefs. My higher-order belief forming mechanism might fail to identify any recognizable difference between my two first-order beliefs. On the view I argue for here, I have special access to the content of my own thoughts (beliefs and experiences) in the sense that my access to my own thoughts differs from either your access to my thoughts or my access to your thoughts. But this special first-personal introspective access is not omniscience: it does not guarantee that for any difference between any two of my thoughts (sensory experiences, concepts and/or beliefs), I can find out about it. I am authoritative about the contents of my own thoughts in the sense that I do not need to observe my own behavior to determine what I think (the way I need to observe your behavior to determine what you think). But differences between two of my thoughts may escape the notice of my introspective resources. To use a term of Boghossian (1989: 481), self-knowledge might not “exhaust” the content of one’s mind. It might not exhaust the contents of sub-personal states. It might not exhaustively recognize each and every difference between the contents of personal states either.

 

4.3. Third Objection

A third objection is the following. On the present account of introspective self-knowledge based on the Principle of the Reflexivity of Content, a person's knowledge of her own mind is two-tiered. On the one hand, from the very fact that she entertains a thought, a person has direct privileged access to the contents of her thoughts. But she does not have the same access to the fact that e.g., she thinks or believes whatever it is she thinks or believes. She does not have access to the fact that what she is doing is thinking or believing in the same direct way she has access to the content of what she is thinking. In order to know the fact that what she is doing is thinking or believing, she must form a higher-order belief about this fact. And this she cannot do unless she uses her higher-order concepts of thought and/or belief. A person can have experiences, entertain thoughts and form beliefs without possessing the concepts of experience, thought and belief. If so, then she will not be able to form the higher-order thought that she is having experiences, entertaining thoughts and forming beliefs.

Now, I want to claim, a person cannot represent the fact that she is representing the world as being thus-and-so unless she can misrepresent what she is representing. An internal state cannot be a representation unless it may be a misrepresentation. The same is true for believing (and/or knowing) that one believes that p. In other words, a person cannot know that she believes that p unless she believes that she does. And she cannot believe that she believes that p unless her higher-order belief could be a false belief. It follows that on the present account, a person's full self-knowledge of her own mind - contrary to the Cartesian tradition - turns out not to be infallible; it is fallible.

Now the question arises: How could a person be mistaken about the fact that she believes what she does? A person, I claim, has direct privileged access to the fact that what she believes is that p (or about the content of her experiences). How then could she be wrong about the fact that she entertains this content as a belief? Of course, a creature could not think that she is thinking that p unless she had the concept of thought. But assuming that she does and assuming that she has several distinct concepts for distinct propositional attitudes, how could she misapply them? To see how she could, consider the difference between knowing and believing that p. I assume that knowing that p involves believing that p, but it involves more. Believing that p is necessary though not sufficient for knowing that p. Now a person may well think that she knows that p where in effect she merely believes that p. She might erroneously take herself to know that p because she wrongly takes her belief that p to be reliable when in fact it is not. In this case, she would misjudge the reliability of her belief that p and therefore be wrong in taking herself to know that p. If so, then although perhaps a person might not be mistaken in thinking that what she is thinking is that p, still she might be mistaken in thinking that she knows it. Perhaps another case of erroneous self-ascription of an attitude might be the kind of self-deception involved in wishful thinking for example. Suppose I come to believe that a certain proposition q is true under the influence of wishful thinking. Do I really believe that proposition q is true? Am I right in self-ascribing to myself a belief in the truth of q? Suppose someone (e.g., my psychoanalyst) points out that the evidence in favor of the truth of q is very thin indeed and she convinces me that I very much want q to be the case. I might then revise my higher-order belief that I believe that q is true and come instead to the view that I wish that q were true: I so much wished q to be true that I managed to convince myself that I believed q to be true whereas the evidence overwhelmingly suggests otherwise.

In examining the previous second objection, I argued that self-knowledge should not be confused with omniscience about the contents of one’s own thoughts. I now want to point out that, on the view I am recommending, there is room for error in the self-ascription of concepts of propositional attitudes. This follows from the assumption that in order to know how one entertains a thought e.g., as a belief, one must form a higher-order belief which involves the concept of belief. There is no guarantee that the application of the concept of belief will be correct. Nor can it be ruled out that the higher-order belief is false. As Boghossian (1989: 481-2) puts it, it is one thing to recognize “a profound asymmetry between the way in which I know my own thoughts and the way in which I may know the thoughts of others”. It is quite another to claim that “self-knowledge is both infallible and exhaustive”. In response to the previous objection, I argued that it is not exhaustive. In response to the present objection, I claim that it is not infallible.

           

5. Conclusion

In this paper, I have assumed an externalist view of the contents of propositional thoughts, i.e., an informationally based teleosemantic account. First, I have argued that this account entails a principle which I have dubbed the principle of the Reflexivity of Content which in turn entails that a person has direct privileged access to the content of her thoughts: by virtue of entertaining a thought, a person has available information about what she thinks. Secondly, I have argued that from the fact that a person has first-personal direct authoritative access to the content of her thoughts it does not follow that she is thereby similarly authoritative about the fact that she thinks or has thoughts. So on the picture of self-knowledge which I have sketched, full self-knowledge turns out to have two components: knowledge of what one thinks and knowledge that one thinks it. Finally, I have examined three objections to my two-tiered account of self-knowledge and I have tried to dispel them.

 

 

 

References

 

 

 

Bernecker, S. (1996) Externalism and the Attitudinal Component of Self-Knowledge, in Bernecker, S. & Dretske, F. (eds.)(2000) Knowledge, Readings in Contemporary Epistemology (Oxford: Oxford University Press).

 

Boghossian, P. (1989) Content and Self-Knowledge, Philosophical Topics, XVII: 5-26.

 

Burge, T. (1979) Individualism and the Mental, in French, P., Uehling, T. Jr. & Wettstein, H.K. (eds.) Midwest Studies in Philosophy, vol. IV (Minneapolis: University of Minnesota Press).

 

Burge, T. (1988) Individualism and Self-Knowledge, in Bernecker, S. & Dretske, F. (eds.)(2000) Knowledge, Readings in Contemporary Epistemology (Oxford: Oxford University Press).

 

Cummins, R. (1975) Functional Analysis, in Sober (ed.) (1984) Conceptual Issues in Evolutionary Biology (Cambridge, Mass.: MIT Press).

 

Donnellan, K. (1993) There is a Word for that Kind of Things: an Investigation of Two Thought Experiments, Philosophical Perspectives, 7, 155-71.

 

Dretske, F. (1969) Seeing and Knowing (Chicago: Chicago University Press).

 

Dretske, F. (1981) Knowledge and the Flow of Information (Cambridge, Mass.: MIT Press).

 

Dretske, F. (1988) Explaining Behavior (Cambridge, Mass.: MIT Press).

 

Dretske, F. (1990) Seeing, Believing and Knowing, in Osherson, D., Kosslyn, S. & Hollerbach, J. (eds.) An Invitation to Cognitive Science, Vol. 2, Visual Cognition and Action (Cambridge, Mass.: MIT Press).

 

Dretske, F. (1995) Naturalizing the Mind (Cambridge, Mass.: MIT Press).

 

Fodor, J.A. (1983) The Modularity of Mind (Cambridge, Mass.: MIT Press).

 

Jacob, P. (1997) What Minds Can Do (Cambridge: Cambridge University Press).

 

McDowell, J. (1994) Mind and World (Cambridge, Mass.: Harvard University Press).

 

Millikan, R.G. (1984) Language and Other Biological Objects (Cambridge, Mass.: MIT Press).

 

Millikan, R.G. (1993) White Queen Psychology and Other Essays (Cambridge, Mass.: MIT Press).

 

Neander, K. (1995) Misrepresenting and Malfunctioning, Philosophical Studies, 79, 109-41. 

 

Putnam, H. (1975) The Meaning of ‘Meaning’, in Putnam, H. (1975) Philosophical Papers, Language, Mind and Reality, vol. II (Cambridge: Cambridge University Press).

 

Wright, L. (1973) Function, in Sober, E. (ed.) (1984) Conceptual Issues in Evolutionary Biology (Cambridge, Mass.: MIT Press).

 

 

Acknowledgments

An initial version of this paper was presented at the International Colloquium entitled “Categories and Beliefs”, organized by the Rosselli Foundation in Torino, from 13 to 14 November 1999. I am grateful to participants at this Colloquium, to participants at a Workshop on “Self-Consciousness”, organized by Albert Newen at the University of Bonn (Germany) from  8 to 9 July 1999, and to three referees for Mind and Society for helpful criticisms.

 



[1] See Burge (1979). Some philosophers - e.g., Donnellan (1993) - have offered a metalinguistic account of the phenomenon of incomplete or partial conceptual mastery on which Burge (1979) bases his argument for social externalism. I think, however, that the metalinguistic account faces a decisive objection. On the metalinguistic account, Mara's full belief (involving her partial mastery of the concept expressed by the word 'arthritis') that she has arthritis in her thigh is really the belief that she has in her thigh what experts in her community call 'arthritis'. If so, then two persons speaking different languages may not share the same belief involving concepts of which they have partial mastery. But I think this consequence is undesirable.  

[2]  Although it has been questioned by McDowell (1994), I assume here the distinction between conceptual and non-conceptual content.

[3] See Boghossian (1989).

[4]  The two-tiered view I am arguing for here is quite similar to the view defended by Bernecker (1996).

[5]  Much of the justification for the decomposition between the two kinds of knowledge derives from the claim (encapsulated in the Principle of Reflexivity) that by virtue of representing the instantiation of a property, a device has available information about the property being represented. Much of my defense of the two-tiered view relies on the differences between having available information and knowing.

[6]  See Bernecker (1996) for a discussion of Burge’s compatibilism.

[7] As Millikan (1984, 1993) and Neander (1995) have emphasized, unlike the non-teleological notion analyzed e.g., by Cummins (1975), the teleological notion of function has normative implications.

[8]  See Jacob (1997, ch. 4).

[9]  This point will be relevant again when I argue that perceptual knowledge and self-knowledge are not closed under logical implication.

[10] See Dretske (1981).

[11]  As on the views of Dretske (1995) and Bernecker (1996).

[12]  See Bernecker (1996, p. 504).

[13] Thanks to Diego Marconi, Mike Martin and Alberto Voltolini for discussing this very point at the Conference.