|Philosophy of Mind
The Self-Knowledge That Externalists Leave Out
Lisa L. Hall
ABSTRACT: This paper addresses the relationship between self-knowledge, practical reason and Externalist theories of mind. Specifically, I argue that the kind of self-knowledge defended by Externalists is insufficient for intentional action. I claim that we know how to act only if we have access to beliefs about how our circumstances are related to our intended actions. I then go on to argue that the kind of mental content we need to characterize these beliefs is incompatible with the Externalists assumptions.
I once wandered for hours in a Tokyo neighborhood searching for a traditional takoyaa shop that sold handmade kites. I knew a lot about the place I was looking for. I knew the name of the proprietor and the shop. I knew the address, though this proved useless since I found no street signs. I even had a photograph showing the shop's facade. All this was more than enough to fix the identity of my destination. Yet, still my destination eluded me. What I never knew, at any point in my search, was where the shop was in relation to me. Lacking this, everything else I knew did me no good at all.
I think this simple story illustrates something important, and often overlooked, about the kind of self-knowledge required for intentional action. Most philosophers would presumably agree that, to act intentionally, an agent needs to know something about the contents of her beliefs and desires. Otherwise, she has no way of knowing what might be reasonable for her to do. But what specifically must an agent know about her mental contents in order to act? Not all philosophers give this further question the attention it deservesor so I'll attempt to show.
The philosophers whose views I'll discuss defend Externalist theories of mind. Externalists hold that an agent's environment partly constitutes her thoughts and that, because of this, her thoughts don't supervene on her internal states. Critics have sometimes charged that this view conflicts with the apparent directness and authority with which agents know their own minds.(1) Although some Externalists deny that self-knowledge has these features, (2) most defend what McGeer calls "Compatibilism": They maintain that self-knowledge and Externalism can be reconciled once we abandon our misconceptions about the mind. (3)
In what follows, I'll not assess the adequacy of existing Compatibilist theories. Instead, I'll simply grant for argument's sake that some version of Compatibilism succeeds: I'll assume that agents generally do have direct, authoritative access to their mental contents, where such contents are understood in the Externalist's sense. But I will maintain that self-knowledge so-conceived is insufficient for intentional action. I'll argue that we know how to act only if we have access to beliefs about how our circumstances are related to our intended actions. I'll then go on to argue that the kind of content we need to characterize these beliefs contradicts Externalist assumptions.
Externalism and Mental Content
To see why self-knowledge might seem problematic for Externalists, we need to consider their assumptions about mental content. No brief exposition can capture the full range of Externalist views on this subject. However, two general theses seem to command consensus. The first, which I call Extensional Differentiation (ED), is a thesis concerning how mental contents are individuated. Burge formulates ED as follows:
This requires unpacking. By "that-clause," Burge means the kind of subordinate sentential expression that occurs in predicates used to report on mental attitudes. Following Burge, I'll say that these expressions "provide the contents" of the attitudes they report. (5) So, in the sentence Don believes that Elvis is alive, the expression that Elvis is alive provides the content of Don's belief. What Burge calls "counterpart expressions" are terms that play identical grammatical roles in containing expressions. Thus, the name Elvis in that Elvis is alive is the counterpart of Lennon in that Lennon is alive.
ED says that whenever such counterparts differ in extension, application or reference, that-clauses containing them provide different contents. This has seemed plausible to philosophers for at least two reasons. First, extensionally-distinct counterparts make different semantic contributions to containing expressions. This explains why Elvis is alive has different truth conditions than Lennon is alive and, likewise, why Don believes that Elvis is alive has different truth conditions than Don believes that Lennon is alive. (6) Assuming that-clauses function semantically to provide mental contents, we then have an argument for ED. Secondly, differences in the extensions of counterparts correspond intuitively to differences in what that-clauses and attitudes are "about." The expressions that Elvis is alive and that Lennon is alive are intuitively "about" different individuals just in case their contained occurrences of Elvis and Lennon differ in extension. Corresponding attitudes follow the same rule. Assuming differences in "aboutness" entail differences in content, we then have a further argument for ED.
Taken together, these considerations support the following three theses concerning the role of mental contents in semantics and psychology:
Call anything that meets these conditions "identifying content," or "contentI." To understand the role of contentI in arguments for Externalism, we need to distinguish weak and strong readings of ED. Weak ED says that contentI corresponds to at least one important sense of the term "mental content." It thus allows that differences in extension and aboutness may be consistent with identity in other intentional properties. Externalists typically endorse one of two stronger readings of ED. (7) They either hold that contentI corresponds to the only correct or coherent sense of "mental content," (8) or that it corresponds to the only sense of "mental content" relevant to some broad category of thoughts. (9) On either of these stronger readings, ED entails that, for many ordinary thoughts, differences in extension and aboutness entail differences in all intentional properties.
Strong ED entails Externalism given a second thesis which derives from twin arguments. This thesis, which I'll call "Indexicality," (10) says that environmental factors that may be unknown to agents figure essentially in determining the extensions of many ordinary terms that occur in that-clauses correctly ascribed to them. Assuming even Weak ED, this entails that environmental factors partly constitute the contentsI of many ordinary thoughts. Moreover, since a possessor of these contentsI needn't be aware of their environmental constitutents, it entails that contentsI don't, in general, supervene on internal states. We get Externalism by combining Indexicality with Strong ED: We may then infer that there is no intentional property of thoughts that generally supervenes on internal states, since all depend on factors that may be unknown to individual thinkers.
We're now in a position to see why self-knowledge might seem problematic for Externalists. Anyone who accepts both Strong ED and Indexicality seems compelled to give up the presumption that agents generally know their mental contents. There just isn't any sense of the term "content" which makes this true. Moreover, given Indexicality, the process of acquiring self-knowledge seems implausibly indirect. Apparently, an agent must investigate her environment to know her thoughts. And finally, by the same reasoning, an agent must concede authority about her thoughts to others insofar as their knowledge of her environment exceeds her own.
Self-Knowledge and Practical Reasoning
A number of Externalists have by now addressed the topic of self-knowledge. Most have defended Compatibilism: They've argued that there are pragmatic and/or theoretical grounds for assuming that agents normally do have direct, authoritative access to their contentsI. These arguments are controversial, but I'll concede the Compatibilist's point here. I'm happy to give the Compatibilist the self-knowledge she wants, because I think philosophers should worry more about the self-knowledge she leaves out.
To explain what's missing from the Compatibilist's story, I need to introduce a pair of assumptions. First, I take it to be a premise of intentional psychology that agents at least sometimes engage in practical reasoning. When we appeal to an agent's beliefs and desires to explain her intentional actions, we're trying to explain, among other things, why she believed those actions were reasonable, in the sense of being appropriate or expedient means to her ends. Second, I assume that a certain level of self-knowledge is required to engage in practical reasoning. Any agent who can deliberate about appropriate means to her ends must have at least some knowledge of both what she desires and values and what she believes. More specifically, I assume that such an agent must know, or at least have access to, those mental contents that provide essential premises in her practical inferences. My thesis is that some of these contents are not contentsI; hence, self-knowledge of contentsI is insufficient for intentional action.
My defense of this depends on two subordinate theses. The first I call Circumstantial Agency (CA):
Some definitions are in order. An action, as I use the term, is a particularsomething performed by a particular agent under a particular set of circumstances. A way of acting, by contrast, is a typesomething different agents can do under different circumstances. Constituting the circumstances under which an agent acts are such things as her space-time location and any individuals on which she acts. Thus, if I kiss B at one place and time and C at another, I perform two different actions, but I act in the same way in each instance. As this case illustrates, the particular circumstances under which an agent engages in a given way of acting may be rather important in determining whether her action is reasonable. Presumably, kissing in itself is neither reasonable nor unreasonable, but kissing a certain individual at a certain place and time may well be one or the other. As Aristotle observed, the really hard questions for practical reason arise here, where abstract ways of acting give way to concrete actions. Ideally, one wants to kiss all and only the right people at the right places and times. Virtuosity in action requires knowing which these are.
CA tells us that practical reason aims at virtuosity in this sense: It aims at rationality in concrete actions, not just in ways of acting. This isn't a controversial thesis on its face, but I think its full implications are rarely recognized. It particularly goes unnoticed that believing that it's reasonable for one to perform an action requires believing that the circumstances under which one acts are appropriately related to those under which one believes one should act. Thus, I may believe that it's reasonable for me to drink a certain liquidsay, H2Oat time t and location l, yet fail to believe that it's reasonable for me drink when presented with some potable H2O at t and l. My problem is not a failure of logic, but rather lack of a critical belief: At t, I don't believe that my circumstances are identical to those under which I believe I should drink. (Either I don't believe, at t: "Here and now is the place and time for me to drink;" or I don't believe of the liquid in the glass: "That's H2O.")
The importance of this kind of belief is made explicit in another thesis, which I call Relational Rationality (RR):
I'll use the term "relational beliefs" to refer to the kind of beliefs that RR identifies as essential to practical reasoning. (11) RR implies that an agent who lacks appropriate relational beliefs may fail to see how she should act under a given set of circumstances even if she possesses other appropriate beliefs. My Tokyo experience provides one illustration. I knew where I wanted to go in non-relative terms, but I lacked appropriate beliefs about how a critical constituent of my circumstancesnamely, my own locationwas related to my intended destination. Consequently, I didn't know which direction to walk in order to get there. Consider also Jim S. Jim believes he should leave the party wearing his own coat and not another guest's. But Jim can't decide which of two "twin" coats hanging on the coat rack he should wear. Jim knows which coat to wear in non-relative terms, because he has lots of identifying information about himself. What Jim lacks is a relational belief: He doesn't believe of any coat on the rack: "That one is mine."
Relational ContentCombined with my two initial assumptions, RR implies that agents possess the self-knowledge required for intentional action only if they have access to appropriate relational beliefs. I think this conclusion, by itself, contradicts the strongest reading of ED. (12) This is because the kind of content that characterizes relational beliefs isn't contentI. One way to see this is to consider what happens when an agent acquires a relational belief. Thus, suppose that Jim S. comes to believe of his coat: "That is my coat." At this point, Jim's psychological condition changes: Now he knows which coat to wear home. Yet, the extensions of counterpart terms in that-clauses ascribable to Jim don't change. When Jim says in these circumstances: "I believe that is my coat," his uses of the terms that and my coat pick out the same thing as would his use of the term Jim S.'s coat. But that Jim S's coat is Jim S's coat is something Jim presumably believed all along. Nor is there reason to think Jim comes to associate any new qualities or non-relative descriptions with his coat. He simply comes to believe that a particular object in his visual field is identical to one with which he previously associated certain qualities and non-relative descriptions. But then, by all relevant measures, Jim's contentsI don't change when he comes to believe "That is my coat."
Similarly, suppose that, while searching in Tokyo, I had come to believe truly: "The kite shop is 600 feet to my left." For the first time all day, I would have known which way to turn. But this change in my psychology wouldn't have been a change in contentsI. I wouldn't have associated any new qualities or non-relative descriptions with the shop. Moreover, had I reported my belief, the location picked out by my use of 600 feet to my left would have been identical to the address of the shopsomething I knew all along. The only change would have been in my understanding of how the shop's location related to mine.
By this reasoning, neither Jim's quandry about the coats nor my quandry in Tokyo can be attributed to ignorance of contentsI. What we both lack is knowledge of what I'll call "relational contents," or "contentsR." An agent knows the contentsR of her thoughts just in case she knows how she believes the individuals and properties her thoughts are about are related to herself and the circumstances under which she acts. For example, someone who knows that she believes the contentsR associated with "x is to my left" knows how she thinks the place picked out by her use of x is related to her spatial location at the time. And someone who knows that he believes the contentsR associated with "That's my coat," knows how he thinks the object picked out by his use of my coat is related to what he demonstrates at the time. To know such things is not to know contentsI because different individuals may be related in identical ways. Thus, depending on the direction I'm facing, my thought: "x is to my left," will be about different locations. However, my understanding of how that location relates to my body's orientation will be the same. Likewise, if Jim and the "twin" coat's owner each believes truly "That's my coat," each has a belief about a different coat. However, each thinks of that coat as identically related to his demonstratum.
Undoubtedly, some Externalists will dismiss these examples. Following Putnam, they'll argue that terms like that and my are exceptional. Such "absolutely indexical" terms have fixed meanings, though they vary in extension with context of use. (13) But this guarantees a purely relational content only for a few singular thoughts (e.g., the thought: "I am here.") Most thoughts are reported using common nouns, names and other terms whose extensionssays Putnamare context-invariant. Why believe we can assign a contentR to these?
I think this criticism underestimates the extent of context-variation in ordinary language. In fact, many expression types, though not ambiguous, denote different individuals under different circumstances of use. (14) More important, the criticism sidesteps the issue at hand. The question isn't whether our terms and concepts have fixed extensions, nor even whether their extensions are externally fixed. (15) It is: What must we believe about these extensions in order to act?
So, let's suppose that 'water' has a fixed extension, that I believe that water quenches thirst and that I know my belief's contentI. In that case, I must (in some sense) know that I believe that H2O quenches thirst. Will knowing this be of use to me when I'm thirsty? Not unless I also know what I believe about how I'm related to H2O. To decide how I should act under any concrete set of circumstances, I'll need know what I believe about things like how encounters with H2O are likely to affect my perceptual states, what I can do with H2O, and how I can communicate about H2O with others who speak my language. But knowing things like this about my mental contents can't be equivalent to knowing contentsI. After all, as Putnam showed, I might have believed the same things about XYZ had I been raised in a different environment. And so I think we must assign contentR even to natural kind terms like 'water.' Settling for self-knowledge in the Externalist's sense won't satisfy the imperatives of agency. (16)
(1) See, for example, Brueckner, A. L., 1986, "Brains in a Vat," The Journal of Philosophy, LXXXIII, 3:148-167; Boghossian, P., 1989, "Content and Self-Knowledge," Philosophical Topics, XVII, 1:5-26; McKinsey, M., 1987, "Apriorism in the Philosophy of Language," Philosophical Studies, 52:1-32, & McKinsey, M., 1991, "Anti-Individualism and Priveleged Access," Analysis, 51, 1:9-16.
(2) See, for example, Ryle, G., 1949, The Concept of Mind, Hutchinson: London; pp. 155-185.
(3) McGeer, Victoria, 1996, "Is 'Self-Knowledge' an Empirical Problem? Renegotiating the Space of Philosophical Explanation," The Journal of Philosophy, XCIII, 10:483-515. Other arguments for Compatibilism can be found in Burge, T., 1988, "Individualism and Self-Knowledge" The Journal of Philosophy, LXXXV, 11:649-663; Bilgrami, A., 1992, "Can Externalism be Reconciled with Self-Knowledge," Philosophical Topics, XX, 1:233-267; Bilgrami, A. 1994, "Self-Knowledge and Resentment," Philosophical Topics; Davidson, D., 1984b, "First Person Authority," Dialectica, 38:101-11; Davidson, D., 1987, "Knowing One's Own Mind," Proceedings and Addresses of the American Philosophical Association 60:441-458, & Davidson, D., 1991, "Epistemology Externalized," Dialectica , 45:191-202.
(4) Burge, T., 1979, "Individualism and the Mental," p. 75. In Midwest Studies in Philosophy, vol. IV, ed. P. A. French, T. E. Uehling, and H. K. Wettstein, University of Minnesota Press: Minneapolis.
(5) Burge chooses this expression in an effort to remain neutral about the ontological status of mental contents. I will attempt to maintain the same neutrality here.
(6) This is another way of expressing Frege's idea that beliefs normally reported by these sentences differ in cognitive significance. Some philosophersmost notably Davidsonargue that knowing the contents of a subject's beliefs is equivalent to knowing the truth conditions of sentences she would sincerely assert. (See Essays #9 - #11 in Davidson, D., 1984a, Inquiries into Truth and Interpretation, Oxford: Clarendon.) ED follows from this assuming that the truth conditions of a subject's sincere assertions correspond to those of sentences embedded in accurate reports of her beliefs.
(7) With the possible exception of Putnam, Externalists seem determined to resist what Bilgrami calls the "bifurcation" of mental content and what McDowell (following McGinn) calls a "duplex" conception of mental content. (McDowell, J., 1992, "Putnam on Mind and Meaning," Philosophical Topics, 20, 1:35-48.)
(8) This seems to be Davidson's position as developed in a number of papers. See especially Essays #2, #9 and #10 in Davidson, 1984a. Bilgrami, who departs from Davidson on some other points, seem to retain this basic thesis. See Bilgrami, 1992.
(9) See, for example, McDowell, J., 1986, "Singular Thought and the Extent of Inner Space." In McDowell, J. & Petit, P., Subject, Thought and Content, Oxford: Clarendon.
(10) "Indexicality" is Putnam's term. (See Putnam, H., 1975 "'The Meaning of 'Meaning.'" In Mind, Language and Reality, Cambridge University Press: Cambridge.) I use the term here because of its familiarity and despite reservations about its appropriateness. The semantic property Putnam has in mind seems importantly different from that which characterizes the class of terms linguists typically call "indexicals." See more on this below.
(11) My use of this term differs importantly from that of some other philosophers writing in this area. Boghossian (1989) is one example. By "relational property," he seems to mean a property that depends essentially on a subject's relationship to some other individual, property or kind, where the identity of this relatum is essential to the identity of the relational property in question. By contrast, when I speak of a relational belief, I mean a belief about a relationship that may obtain among different relata.
(12) The version according to which all mental content is contentI.
(13) See Putnam 1975.
(14) Consider tensed expressions like He is sleeping, imperfect descriptions like the kitchen table and the book, and proper names like City Hall and Christmas Eve. A semantic theory sensitive to this kind of context-variation in ordinary language is developed in Barwise, J. and Perry, J., 1983, Situations and Attitudes, Cambridge, Mass: MIT Press/Bradford. See also the essays in Perry, J., 1993, The Problem of the Essential Indexical, Oxford University Press: Oxford.
(15) Even if some terms do have fixed extensions, this doesn't entail that their meanings determine or constitute their extensions. It's reasonable to suppose that rules governing the use of a kind term like water should stipulate that it picks out the same stuff on all occasions of use. (This is presumably part of what it means for water to function as the name of a natural kind.) But this is quite different from saying that the identity of the stuff picked out by water is determined by its meaning.
(16) I am grateful to both Joseph Buckley and Robert Feleppa for their comments on earlier versions of this paper. I am also deeply indebted to John Perry both for his inspiration and for many helpful discussions concerning the issues I've addressed.