EXTERNALISM: THOUGHTS AND THEIR CONTENT

KNOWING WHAT YOU THINK vs KNOWING

THAT YOU THINK IT^*

Fred Dretske

I am an externalist about meaning. I am an externalist about all forms of representation. That includes mental representation. What gives something intentional content, what makes it represent, mean, or say something about other affairs are not its intrinsic properties, but, rather, something about its purpose or function in an informational system. That is why alcohol in a glass tube--an ordinary household thermometer--is able to mean or say, truly or falsely as the case may be, that the temperature is 70^o F. It has the function (a function we give it) of telling us, providing us with information, about temperature. Remove this informational purpose, this indicator function, by (say) bottling the alcohol for medicinal purposes, and the glass encased liquid becomes representationally lifeless. It still expands and contracts as the temperature varies, but it no longer says anything about temperature that (like a statement or a belief) could be false.[1] The same is true of the splotches of ink in books and newspapers and the sounds we produce when talking to one another. It is not their shape, color, size, volume, or wave length (intrinsic properties), but rather something, broadly speaking, about their role in a system of communication that gives them their meaning. Except for the source of the functions (natural vs. conventional), the same is true of the events in our brains. They become representations, they acquire intentional content, by developing via some appropriate history an informational function.

______________________________________

^*Thanks to the members of El Instituto de Investigaciones Filosóficas, Universidad Nacional Autónoma de México, Mexico City, for the tough questions they asked when I read an early version of this paper. I'm especially grateful to Maiate Ezcurdia for her questions about the scope of knowledge attributions. Thanks, also, to Andreas Kemmerling, Heidelberg, for his constructive skepticism about an even earlier version of this material.

I have just described my own brand of metaphysical externalism about the mind. Like other forms of externalism, it denies that thought supervenes on the neurobiology of thinkers. The facts that determine what you think, are (some of them anyway) facts about the relations you, or your thoughts, bear (or bore) to external, affairs. The thoughts are in your head (just as words are in books), but what gives them their content (just as what gives words their meaning) isn't there.

This, we are told, creates a problem. It appears to be incompatible with the kind of knowledge we have of what is going on in our own mind. How can we know, in the special, authoritative, way that we do, that we are thinking about peanuts if thinking about peanuts consists, in part, of relations that exist between the thinker (or thought) and other parts of the world? If, as some externalists hold, you cannot have peanut thoughts--cannot, that is, think that, or wonder whether, something is a peanut--without standing (or having stood) in causal relations to peanuts, then it seems to follow that to know, in that special authoritative way we call introspection, that you are thinking about peanuts is to know, in that same authoritative way, that there are (or were) peanuts. But one cannot know, not at least in that way (by, so to speak, gazing inwardly) that there actually are (or were) peanuts. So thinking about peanuts cannot consist of relations (causal, functional, informational, or whatever) to peanuts. It cannot depend on there being, or having been, peanuts. It cannot consist of any relations of the sort externalists propose since this would imply that we could know, in the same way we know that we are thinking about peanuts, that we stand (or stood) in these relations to the independently existing conditions that (according to externalism) make such thoughts possible.[2]

This argument has by now spawned a substantial literature. I do not propose to review this literature (many of the seminal articles are collected in Ludlow & Martin 1998) because, as I see it, the debate rests on a false assumption, an assumption that (with few exceptions) both parties to the debate, both externalists and their critics, make. They assume, mistakenly, that what we know by introspection is not only (in the case of thought) what we think, the content of thought, but also that we think it, the attitudinal aspect of thought. If this is false, if it is possible to know by introspection that it is peanuts one is thinking about (that one's thought has this content) without knowing, at least not in the same way, that one is thinking something with this content, then there is no epistemological threat to externalism. What the externalist says is constituted by external relations to peanuts--thinking about peanuts--is not what introspection reveals about thought--that it is peanuts we are thinking about.

To my dismay, however, I have discovered that many people (even a few of the non-philosophers I've tried it on) find this thesis unintelligible. They tell me it just doesn't make sense. We cannot know what we are thinking about without knowing, in the same way, that we are thinking about it. If we know, by introspection, that it is peanuts we are thinking about, then we must know, by introspection, that we are thinking about peanuts.

I can live with implausibility but unintelligibility is something else. So I take this criticism seriously. I mean to address it in this essay. What I hope to show that it is possible to know, by introspection (or by whatever method it is one knows it), that it is p one thinks without knowing, by that same method, that p is something one thinks.

Although it won't quite get me to where I want to go, let me begin by talking about young children--people who think but do not know, do not even understand, what it means to think. Psychologists tell us that a typical three-year-old does not have a developed concept of belief. These children have beliefs, of course, and it is easy enough to find out, by asking them the right questions, what it is they believe, but that they believe or think these things is beyond their comprehension. They lack the concept of thought, but they are, nonetheless, authorities about what they think. They enjoy a special kind of access to the content of their thought. If you want to know whether three-year old Suzy thinks Daddy is home or the dog is loose, just ask her. Is Daddy home? Is the dog loose? Her answers will tell you, quite unerringly, what it is she believes . You cannot, to be sure, ask Suzy directly, and in just these words, what it is she believes because (we are assuming) Suzy doesn't yet understand, at least she haven't fully mastered, what it means to believe something. But there are indirect ways of finding out. If Suzy understands what it means for the dog to be loose, you can find out whether she thinks the dog is loose by asking her whether the dog is loose. She is an authority, mind you, not on whether Daddy is home or the dog is loose (about these topics you may know better than she), but on whether she thinks Daddy is home or the dog is loose. Her authority on these topics is in no way diminished by her ignorance of what it means to think such things.

Children who think (and say) that the dog is loose must, of course, understand what it means for the dog to be loose but they needn't understand what it means to think it. They would not--indeed, they could not--say (or even think) that they know what they think since saying (or thinking) this requires them to refer to what they think (that the dog is loose) as something they think and, lacking an understanding of what it means to think this, they are unable to do this. Nonetheless, we can certainly describe what they know in this way. The child knows what it thinks--viz., that the dog is loose--despite not knowing (not even thinking) that it thinks this.

I will be accused of playing fast and loose with the scope of knowledge attributions. It may be intelligible--even true--to say that Suzy knows something about the proposition--that the dog is loose--she believes, but that is not the same as saying that Suzy knows what she believes. That would be like saying that Suzy knows the answer to a difficult mathematical problem--that it is, say, 24--simply because she knows what number is written on the board (viz., 24) when 24 happens to be the answer to the problem. Suzy may know that the number 24 is on the board, but unless she knows this number under the description, "the answer to the problem" she doesn't know what the answer to the problem is. Likewise, Suzy may think that the dog is loose, and she may know this proposition--that the dog is loose--under some description or other (perhaps as "what I told Mommy"), but unless she knows it under the description, "what I think" she doesn't know what she thinks. All Suzy really knows is what she told Mommy. What she told Mommy is what she thinks, of course, but she doesn't know this.

This is a fair objection. At least it is an objection that any philosopher who has come this far is likely to make. So I concede the point. It is why I said at the beginning that consideration of children, people who lack the concept of thought, would not get me quite where I wanted to go. It only gets me to the point of having established that there is some sense in which children can know what it is they think without knowing that they think it. They can know what they think without knowing they think it in the same sense you can know what my brother is doing without knowing it is my brother doing it. But this, as my critic has been quick to point out and as I am now willing to concede, only shows that Suzy knows of what she thinks that it is that the dog is loose (the phrase "of what she thinks" kept carefully outside the scope of the attributed knowledge). It does not show that Suzy knows that what she thinks is that the dog is loose (the phrase "what she thinks" here occurring inside the scope of the knowledge attribution). So it does not, not in any relevant sense, show that Suzy knows what she thinks--much less that she knows what she thinks without knowing she thinks it.

So, to make the next step in my argument, let me shift to a person who, unlike a child, possesses the relevant concepts and the relevant beliefs. I will describe an analogous situation--knowing what someone said--and suggest that it provides a plausible model for knowing what one thinks.

Clyde gets a telephone call from his good friend Harold. Harold tells him that he, Harold, is going on vacation for two weeks. Clyde hears him say this and, let us suppose, hears him say it under ideal telephonic conditions (no static, clear articulation, etc.), the kind of conditions that would ordinarily lead us to say that Clyde knows what Harold said. So far, I hope, there is nothing suspicious. Now the twist. There are several people, all practical jokers, who, unknown to Clyde, enjoy telephoning Clyde and imitating Harold. They are very good at it. As far as Clyde can tell, the call he received from Harold could have been from any one of these people. It sometimes is one of these other people, Unaware of the past deceptions, and, therefore, the very real present possibilities they create, Clyde not only believes (correctly as it turns out), without doubt or hesitation, that it is Harold he is talking to, but (incorrectly as it turns out) that he knows it is Harold.[3]

I have Gettierized[4] Clyde's belief that it was Harold on the phone while leaving intact his evidence for what it was that Harold said to him. The question I'm interested in is this: does the fact that Clyde does not know it was Harold who said he was going on vacation mean that he doesn't know what Harold said to him? If asked ("What did Harold say?"), Clyde will tell you, confidently and truthfully, exactly what Harold said. And if asked whether he knows--and, if so, how he knows--that this is what Harold said, Clyde will tell you, once again confidently and (I submit) truthfully, that he knows this because he heard Harold say it. If anyone ever knows what another person says on the phone, Clyde, given the circumstances, surely, knows what Harold said. Yet, Clyde doesn't know it was Harold who said it. Clyde thinks he knows. This, indeed, is why he so confidently reports what he knows by referring to the caller as Harold. But the truth of the matter is that Clyde is ignorant about who called him.

Unlike the earlier case of the child, we now have an example in which the agent, Clyde, does understand the phrase ("what Harold said") being used to pick out the proposition that he heard expressed on the phone. He not only understands it, he confidently (and truly!) believes it refers to what he heard. That is why he describes what he heard the caller say as "what Harold said." Unlike the case of the child who does not believe, does not even understand, that "what I think" (when said or thought by her) is a correct description of the mental state whose content (viz., that the dog is loose) she has special access to, Clyde does understand--indeed, he truly and confidently believes, that "what I heard Harold say" is a correct description of the content he has special access to, the proposition he heard expressed on the telephone. Why isn't this enough to know not (once again) that it was Harold who said he was going on vacation, but that what Harold said was that he was going on vacation? If it is enough, then, it seems, we have an attractive externalist model of introspection. Just as Clyde can know (by hearing) what it is Harold said without knowing, at least not by hearing, that it was Harold who said it, why can't a person know what it is he thinks (by, say, introspection) without knowing (not by introspection) that he thinks it?

I will be told (I have been told) by impatient skeptics that I am still ignoring subtle, but nonetheless quite valid, distinctions of scope. Since Clyde doesn't know it was Harold who said he was going on vacation, it would be wrong to put the phrase "what Harold said" inside the that- clause that expresses what Clyde knows. Clyde doesn't know that what Harold said is that he is going on vacation. All he really knows is something of or about what Harold said--that it is that he is going on vacation. We can say that Clyde knows what the caller said, and the caller was Harold, but since Clyde doesn't know the caller was Harold, he doesn't know what Harold said. Clyde believes the caller was Harold, and he is (we may suppose) fully justified in this belief, but (given the special circumstances) he doesn't know it. So it would be wrong to describe him as knowing that what Harold said is that he was going on vacation. Clyde doesn't--not really, not strictly--know what Harold said.

I don't believe in this real, this strict, form of knowledge[5], but I'm willing, once again, to concede the point to those who believe that, strictly speaking, nothing belongs inside the scope of a knowledge attribution that isn't known to belong there by the agent to whom the knowledge is ascribed. I am, after all, interested in convincing even a stubborn internalist that there is a workable--indeed, a plausible--externalist model of self-knowledge. So I'll work with what I'm given. My third (and final) example--a slight variation on the second--is meant to comply with these strictures on scope.

We have to remember that we needn't suppose that Clyde can know what Harold said without knowing it was Harold who said it. All I need to show is that the way Clyde knows it was Harold who said it can be different from the way he knows what Harold said. If this is possible, then the way is clear to concede (to the yet unconverted) that although you cannot know what you think without knowing that you think it, the way you know what you think may be entirely different from the way you know that you think it. You can know what you think by introspection, but introspection may not the way you know you think it. To know that you are thinking about peanuts may require--or so externalists are free to maintain--a more indirect method, a method (empirical investigation?) compatible with an externalist theory of what it takes to think about peanuts. All that introspection tells you is that it is peanuts you are thinking about..

To illustrate this possibility we need imagine only a small modification of our last example. Clyde (overly suspicious from too much philosophy) finds out who called him by tracing the call. He discovers that the call originated from a phone to which only Harold had access. So he knows it was Harold who called him. He knows both what Harold said (viz., that he was going on vacation) and that it was Harold who said it, but his way of knowing the one is different from the way he knows the other. He heard what Harold said, but he did not--indeed (given the sound-alike impersonators) could not--have heard (by listening to his voice on the phone) that it was Harold who was saying it. Given the conditions, there is nothing distinctive about Harold's voice to enable one to know, by hearing him talk on the phone, that it is Harold. Clyde knows it was Harold (who said that he was going on vacation) as a result of an empirical investigation. But no investigation was required to find out what Harold said. Clyde heard him say he was going on vacation.

Using this slightly modified example as a model for introspection, then, the proposal is that the way we find out what it is we think (desire, wonder, fear, expect, etc.) is different from the way we find out that we think (desire, wonder, fear, and expect). The first method we call introspection. Whatever, exactly, introspection comes down to, it does not involve empirical investigation of external circumstances. That is what makes it introspection. But this is quite consistent with an empirical investigation being required to find out that the content revealed by introspection is the content of a mental state, a state whose possession of content is constituted, in part, by a network of external relations. Given our model, this should be as sensible as saying that Clyde heard what Harold said but needed an investigation (tracing the telephone call) to find out it was Harold who said it. Or, to give an example that might appeal to baseball fans, it is like needing a program to find out that it was Lou at bat, but not needing a program to know that Lou hit a home run. You saw him hit a home run. That you know by direct perception. But that it was Lou who hit it you know indirectly, by looking at your program.

We have, then, the following picture of self-knowledge: when thinking about peanuts, we can know, with a special kind of first person authority, what it is we are thinking about--peanuts. Nothing illicit is smuggled into the scope of the knowledge attribution since we know both that it is peanuts we are thinking about and that we are thinking about them. So we can, in both word and thought, and with full knowledge, pick out and refer to what we are thinking about--peanuts--as something we are thinking about. Nonetheless, our way of knowing that it is peanuts we are thinking about is, or it may be, quite different from the way we know that we are thinking about them. Although we enjoy first person authority about the first fact, we may enjoy no privileged access to the second fact. It may be, as externalists have it, that to know you are thinking about peanuts requires information not obtained by looking exclusively inward. Introspection doesn't tell you that you think, only what you think.

If this is right, we have a viable theory of self knowledge. Special authority about, and privileged access to, one's own thoughts and experiences is compatible with an externalist theory of thought and experience. The only remaining question is whether this theory gives, besides a coherent account of self-knowledge, a plausible account of self-knowledge. Is this really all that, introspection yields? Do we, in fact, use a different method to find out that we think from the one that tells us what we think?

My purpose was only to argue for the intelligibility of a hypothesis, not to argue for its truth. Since I've already done this, I should quit now. Nonetheless, I cannot resist adding a few remarks--three will do--about the plausibility of this picture of the mind's knowledge of itself.

1. As the above examples show, we often know that x is y by some direct method (hearing, seeing, introspection) without being able to know, by that same direct method, that it is x that is y. If we know that it is x that is y, our way of knowing this may be, and often is, quite different from our way of knowing that x is y. I don't have to see, even be able to see, that it is water that is boiling to see that the water is boiling. There is, after all, nothing about water to distinguish it from gin, vodka, and a variety of other liquids. I needn't be able to see that it is water in order to see that the water is boiling. So if it is water, and if I reasonably and truly believe it is water, there is nothing to prevent me from saying I can see that the water is boiling. That, I submit, is how I know that the water is boiling. If I know it at all, though, it isn't how I know it is water that is boiling. If I actually know it is water, I probably know that in some way other than the way I know it is boiling. The fact that I came to know (or believe) that it is water by chemical analysis doesn't mean I can't see that the water is boiling. Why shouldn't the same be true of introspection. The fact that I found out I think by having someone (parents? teachers? friends? Descartes?) tell me doesn't mean I can't now discover what I think by simple introspection.

2. The analogy with ordinary perception can, I think, be pushed a little farther with, I hope, illuminating results. Perception of ordinary dry goods tells us what is in the physical world, not that there is a physical world. I see that there are cookies in the jar, people in the room, and (by the newspapers) continued violence in the Middle East. That is how I know there are cookies, people, and violence in these places. Cookies, people, and violence are physical things, things that exist independently of my perception of them. Do I, therefore, know, by perception, by seeing, that there are things that exist independently of my perception of them? Can I see that there is a material world and that, therefore, solipsism is false? I don't think so. It seems more reasonable to say that assuming there is a physical world, or assuming we know (in some other way) that there is such a world, perception tells us what sorts of things are in it--cookies, people, and violence. Visual perception has the job of telling me what (physical objects) I see, not that I see physical objects. If my perceptual faculties had the latter job, the job of telling me that I was (in effect) not hallucinating, not aware of some figment of my own imagination, then they would be incapable of discharging their responsibilities. For, as we all know, hallucinatory cookie jars can, and sometimes do, look much the same as real cookie jars. You can't see the difference. If it’s a real object you see, perception can tell you whether it’s an orange or a banana (a difference that is plainly visible), but perception cannot tell you whether it's a real orange or just a figment of your imagination. That difference isn't visible.

Memory has a similar structure. Memory tells us what happened in the past--the specifics, as it were, of personal history. It does not tell us there is a past. I can remember (hence, know) what I had for breakfast this morning. No trick at all. I distinctly remember that it was granola. Nonetheless, despite this implying that the past is real (if it isn't real, I didn't have breakfast this morning; hence, do not remember having granola for breakfast this morning) this doesn't mean I can remember that the past is real. If I know the past is real, I don't know this by remembering that it is real. That isn't a way to answer Russell's skeptical question about the past.[6] If I know the past is real, I know it in some way other than by memory. Memory is a faculty that tells me some of what occurred in the past given that there was a past just as perception informs me what is in the physical world given that there is a physical world. Maybe I have to know the past is real in order to remember what I had for breakfast this morning (I doubt it, but let it pass), and maybe I have to know there is a physical world to see whether there are cookies in the jar (let that pass too), but the point is that I do not have to know these things by memory or by vision in order for memory and vision to tell me (give me knowledge of) what I had for breakfast this morning and what is in the cookie jar.[7]

Introspection is like that. Introspection tells me what is in my mind, what it is I am thinking, wanting, hoping and expecting. It doesn't tell me I really have a mind, mental states with such content. If I know that at all, I know it in some way other than by introspection, the faculty that, given that I have thoughts and feelings, tells me what I'm thinking and feeling.

3. This account of introspection also squares neatly with a representational theory of mind. This won't persuade anyone who isn't antecedently sympathetic with a representational scheme, but it may tip the balance for those who are.

Think about a simple representational device, a measuring instrument, say, whose job it is to provide information about some magnitude--temperature, say. A pointer registers 70° on a scale calibrated in "degrees Fahrenheit" and, thereby, represents the temperature to be 70° F. In very crude terms, and setting aside for the moment the difference between a conventional (in which the indicator functions derive from us) and a natural (where they derive from evolution or learning) forms of representation, this state of the thermometer (the pointer pointing at "70") is the thermometer's "belief" that the temperature is 70° F.[8] The instrument occupies a state that represents, says, or means (in a way that might be false) that the temperature is 70° F. Even if we refuse (as we should) to credit instruments with knowledge (since knowledge of the sort now in question implies belief, and instruments do not have beliefs), we can ask what kind of information they have about their own representational states. Does the thermometer have information about how it is representing the temperature? Does it have the information that it is representing temperature? If thermometers knew things, could it know it was representing the temperature to be 70^o F?

With respect to temperature, this thermometer is like the child, Suzy (described above), with respect to the dog's whereabouts. The instrument "thinks" the temperature is 70°, but it has no concept of representation. It, therefore, has no way to represent itself as thinking about (i.e., representing) temperature. It has been given the power of thinking about temperature (this is what it reports on) but it hasn't been given the power of thinking about itself as thinking about temperature. It doesn't report on itself, on its own representational efforts. Nonetheless, it does carry (in a somewhat degenerate sense) information about the way it represents the temperature. The way we find out how the instrument is representing the temperature, after all, is by looking at the instrument. Is the pointer pointing at the numeral "70" on a scale suitably calibrated in "degrees Fahrenheit?" Yes? Then it is representing the temperature to be 70^o F. The actual temperature may not be 70^o F, of course, but if the pointer is pointing at "70" then the device is "saying", perhaps falsely, that the temperature is 70^oF. So if that is where we look to see how the instrument is representing the temperature, to see what the instrument "says" or "thinks" about temperature, then that is where the instrument itself should look if it wants to find out what it "thinks" about the temperature. In that sense, if the instrument wants to know what it thinks, it should look inward, at itself. It should introspect. That's where the information is about what it thinks, about the way it is representing the world.

But though the instrument, given that it is representing temperature, necessarily has information (in its own intrinsic state--i.e., pointer position) about how it is representing the temperature, about what it is representing the temperature to be, it is not so clear that it has--in fact I think it clearly does not have--the information that it is representing temperature or, indeed, representing anything at all. If somebody wanted to be sure that this device was a thermometer (something that actually represents temperature) and not a badly designed wind gauge, a door stop, or an object d'art (things that do not represent temperature) she clearly shouldn't look at it. Looking at it wouldn't provide the information because, typically, there is nothing about a device that tells one what its purpose or function is. To find out what the purpose or function of an object is, or whether it even has a purpose or function, one has to know (in the case of artifacts) the intentions and purposes of its designers and builders or, perhaps, the way the object is normally used. These facts though, are extrinsic, relational, facts, that are not, or need not, be evident in the object itself. There is nothing about a gauge itself, its intrinsic properties, that tells you whether it is a pressure gauge, an altimeter, a badly designed (absolutely dysfunctional) speedometer, or a fancy paperweight.[9]

If this is correct (I argue it at greater length in Dretske, forthcoming (b)), then every representational device necessarily has, in its own intrinsic state, information about how it represents the world, but little--possibly no information--that it represents the world. It can look inward to see what it represents the temperature as--about this subject, the content of its representational states, it has privileged and authoritative access-- but it has to look outward to see whether, in fact, it represents anything at all. That, I submit, is an accurate picture of self knowledge, and it is, I remind you, a picture that externalism recommends.

REFERENCES

1. Dretske, F. forthcoming (a). "How Do You Know You Are Not A Zombie?" in Privileged Access and First Person Authority, Brie Gertler, ed. Ashgate Press.

2. Dretske, F. forthcoming (b). "Externalism and Self -Knowledge," Externalism, Skepticism and Self-Knowledge, edited by Susana Nuccetelli.

3. Dretske, F. 1969. Seeing and Knowing. Chicago; University of Chicago Press.

4. Dretske, F. 1970. "Epistemic Operators," Journal of Philosophy, LXVIII, 24; 10097-1023.

5. Dretske, F. 1971. "Conclusive Reasons," Australasian Journal of Philosophy, 49.1: 1-22.

6. Dretske, F. 1972. "Contrastive Statements," Philosophical Review, LXXXI, 4: 411-437

7. Ludlow, P. and N. Martin, eds. 1998. Externalism and Self-Knowledge. Stanford, CA; CSLI Publications.

NOTES