06-Luis-Chap05.qxd 22:02:05 9:08 AM Page 164 5 Evans and the Sense of “I” JOSÉ LUIS BERMÚDEZ ! This paper focuses on two enduring features of Gareth Evans’s work. The first is his rethinking of standard ways of understanding the Fregean notion of sense, and the second his sustained attempt to undercut the standard opposition between Russellian and Fregean approaches to understanding thought and language. Evans’s The Varieties of Reference provides the only worked-out development of how a Fregean should understand singular thoughts. Singular thoughts are ways of thinking about objects that require the existence of the object being thought about. Paradigm singular thoughts are perceptual beliefs of the sort that might be communicated through demonstrative expressions such as “this” or “that” and beliefs about oneself or the present moment that might be expressed through token-reflexive expressions such as “I” or “now”. Suppose we characterize Russellian approaches to understanding language and the thoughts it expresses in very broad terms as holding that all that is required for a correct and adequate specification of a thought (the meaning of a sentence or the content of a propositional attitude) is a specification of the object(s) being thought about and the properties attributed to it/them. It looks very much as if singular thoughts will lend themselves to a Russellian approach, and there have been a range of influential arguments to this effect (Kaplan 1989; Perry 1977). In The Varieties of Reference Evans offers a counterweight to this line of argument. Starting out with general reflections on Frege’s notion of sense, Evans argues in Part I, on the basis of a distinctive way of My thanks to Mark Sainsbury, Chris Gauker, Fiona Macpherson, Jane Heal, and Stephen Butterfill for helpful comments on earlier drafts of this paper. 06-Luis-Chap05.qxd 22:02:05 9:08 AM Page 165 Evans and the Sense of “I” 165 understanding Frege’s notion of a mode of presentation, that Fregean senses are in general object-dependent in precisely the way that singular thoughts are object-dependent. He then goes on in Part II (and in the related paper “Understanding demonstratives” (1981b)) to work out in considerable detail how this understanding of sense can be applied to thoughts that would most naturally be expressed using demonstrative and token-reflexive expressions. The discussion goes to the heart of many of the most central issues in the philosophy of language and of thought. An important set of issues come into focus when we ask whether indexical expressions have a Fregean sense.1 John Perry’s two influential papers “Frege on demonstratives” (Perry 1977) and “The problem of the essential indexical” (Perry 1979) argued powerfully that indexicals cannot be accommodated within a Fregean theory of sense, because there are requirements upon the Fregean notion of sense that the context dependence of indexical expressions precludes them from satisfying. Instead, we need to separate out different strands of what we might intuitively think of as the “meaning” of sentences expressed using demonstratives. In particular, he suggests in “Frege on demonstratives” that we need to distinguish the sense of a sentence employing indexical expressions from the thought expressed by that sentence. Evans responded to Perry in ‘Understanding demonstratives’. In that paper, and in Part II of The Varieties of Reference, he offers a positive account of how we might understand the senses of sentences involving a range of indexical expressions. His account develops the idea that the Fregean senses of referring expressions are object-dependent and involve ways of identifying and keeping track of the object or objects being thought about (taking “object” in a loose enough sense for the present moment to count as an object). My primary concern here is with just one part of the general debate opened up by Evans’s work: namely, the status of the indexical expression “I”. I shall be exploring the peculiar difficulties that “I” poses for a Fregean theory and show how Evans’s account of the sense of the first-person pronoun can be modified to meet those difficulties. I begin, in section I, with some general comments on 1 For the purposes of this paper I shall talk about demonstrative and tokenreflexive expressions as distinguishable categories within the general category of indexical expressions. Both demonstrative and token-reflexive expressions are context-dependent, but whereas demonstrative expressions “latch onto” objects through some sort of act of ostension (such as pointing), using token-reflexive expressions to pick out an object requires no such gesture. 06-Luis-Chap05.qxd 166 22:02:05 9:08 AM Page 166 José Luis Bermúdez the Fregean approach to thought and language, identifying three different explanatory tasks that the notion of a thought is called upon to perform. Section II explores the distinctive difficulties that indexical expressions in general, and “I” in particular, pose for Fregean approaches to thought and language. We will consider in section III Perry’s proposal to deal with these difficulties by fractionating the notion of a thought so that there is no single thing that performs the three functions identified in section I. Rejecting this proposal takes us, in section IV, to Evans’s own account of the sense of ‘I’. Evans’s account can be seen as a working-out of Frege’s well-known dictum that “everyone is presented to himself in a special and primitive way in which he is presented to no one else” (Frege 1918: 25). Evans’s account has several dimensions, one of which stresses the importance in self-conscious thought of judgments based upon forms of information derived from sources that can only provide information about oneself. Evans is quite explicit that his account of the sense of “I” has the consequence that “I”-thoughts are not shareable. We see in section V that, despite Evans’s best efforts, this cannot be reconciled with the objectivity of “I”-thoughts. It is, I will argue, a constraint upon an account of the sense of “I” that it respect the symmetry requirement, so that what I say in a given context using “I” should be the same as what you say in a given context using “you”. In section VI I propose a development of Evans’s account that meets the symmetry requirement. I FREGE’S CONCEPTION OF THOUGHT: A SKETCH For Frege thoughts are simultaneously (1) the bearers of truth and falsity (the things2 that are true or false); (2) the objects of propositional attitudes (the things that are believed, hoped, etc.); 2 The description of thoughts as things should not be taken too literally. It need not be taken to suggest more than, first, that we can refer to thoughts (by talking, for example, of the thought that p); second, that we have some grasp of the identity conditions of thoughts (of what makes thoughts the same or different); and third that thoughts are not in any sense reducible either to individual ideas or acts of thinking or to the sentences which express them. It is true that Frege suggested in various places that thoughts exist in a “third realm”, but we can set these comments aside. They are not essential to the overall picture. 06-Luis-Chap05.qxd 22:02:05 9:08 AM Page 167 Evans and the Sense of “I” 167 (3) the senses of assertoric sentences (the things expressed by sentences that purport to describe the way things are). The interconnection of these three different dimensions of the notion of a thought sets the background for the problems posed by indexical expressions.3 Frege’s distinctive idea is that the essence of a thought is given by the fact that it is a candidate for truth or falsity. Fundamental to the analysis of a thought is its truth condition: namely, what would have to be the case in the world for that thought to be true. Starting from the truth condition, we can analyze the parts of a thought in terms of how they contribute to determining that truth condition. For Frege, judgment is the fundamental propositional attitude. To form a judgment is to commit oneself in thought to a certain state of affairs holding in the world. The judgment is true just if that state of affairs does hold, false otherwise. Judgment itself is a psychological state—a relation to the thought, rather than a part of the thought. The thought itself is the postulate (for want of a better word) that the relevant state of affairs holds in the world. To judge the thought to be true is to endorse the postulate—to take it to be true. And what we endorse when we form a judgment is precisely what we understand when we understand an assertoric sentence (where an assertoric sentence is one which characterizes the world as being a certain way). Frege’s notion of sense is correlative with the notion of understanding, so that the sense of a linguistic expression, whether that expression is a proper name, a logical constant, a predicate, or a complete sentence, is what a competent language-user understands when he understands it. The connection with the other two aspects of the notion of thought comes because Frege takes the understanding of a sentence to be basic, and the notion of truth to be fundamental to understanding. We understand a sentence insofar as we understand its truth condition, and we understand the parts of a sentence insofar as we understand how they contribute to determining the truth-value of a sentence.4 3 It should be clear that my account of Frege owes much to Michael Dummett, and in particular to Dummett 1973. 4 There are various complicating factors here, not least of which is that something needs to be said about fictional statements that are not directly truth-apt. Evans himself has some very interesting things to say about this in ch. 10 of The Varieties of Reference. Evans’s views are illuminatingly discussed in Sainsbury 1999. 06-Luis-Chap05.qxd 168 22:02:05 9:08 AM Page 168 José Luis Bermúdez Our understanding of the constituent words of a sentence is parasitic on our understanding of the sentences in which they can feature. We understand a word insofar as we understand the contribution that word makes to determining the truth condition of such sentences. For a proper name, that understanding consists in knowing what it would be to single out an object as its semantic value. For a predicate, it consists in a means of determining whether or not an object falls within the predicate’s extension. We understand logical operators (or at least the truth-functional logical operators) insofar as we understand how they contribute to determining the truth conditions of complex sentences on the basis of the truth-values of their constituent sentences. The three aspects of Frege’s conception of a thought dovetail together. For Frege, thoughts are the bearers of truth and falsity, the objects of the propositional attitudes, and the meanings of assertoric sentences. Indexical expressions are problematic, because they seem to prise these aspects apart. Section III illustrates how this fractionation of the notion of a thought threatens not only to reduce the Fregean notion of sense to incoherence, but also to render problematic the very notion of psychological explanation. First, though, we will turn in the next section to considering how Evans proposed to put flesh on the bones of this general schematic account. II EVANS’S INTERPRETATION OF FREGE The notion of sense is ineliminably bound up with the notion of truth. The sense of a sentence is its truth condition, which is a function of the senses of its subsentential components. This is often put by saying that sense determines reference. But how should “determination” be understood here? According to popular interpretation, the relation of determination (as applied to referring expressions) is the relation of satisfaction. The sense of a referring expression is a definite description or cluster of definite descriptions. The referring expression refers to a given object because that object is the unique satisfier of the associated description or cluster of descriptions. Proper names, therefore, are synonymous with definite descriptions, and in order to understand a proper name, one needs to understand the sense of the synonymous definite description. 06-Luis-Chap05.qxd 22:02:05 9:08 AM Page 169 Evans and the Sense of “I” 169 This way of understanding how sense determines reference has been subjected to a battery of criticism, most of which has to do with what are taken to be its counter-intuitive modal implications. Kripke has been the prime instigator of this line of attack (Kripke 1980). According to Kripke, referring expressions such as proper names are rigid designators, which is to say that they refer to the same object in all possible worlds (in which that object exists), whereas definite descriptions are non-rigid designators that can shift reference across possible worlds.5 The repudiation of the descriptive view is central to Evans’s development of the notion of sense. Evans’s conception of sense actually requires treating proper names and non-descriptive referring expressions as rigid designators.6 Evans takes his lead from what he calls Russell’s principle, according to which a subject cannot make a judgment about something unless he knows which object his judgment is about. Evans interprets “knowing which” as requiring that the thinker possess discriminating knowledge of the object being thought about, where that in turn requires being able to distinguish that object from all other things (1982: 89). We have, he maintains, an intuitive grasp of three different forms that such discriminating knowledge might take. A thinker might possess discriminating knowledge of an object by perceiving it. Or by being able to recognize it.7 Or by knowing distinguishing facts about it. Each way of possessing discriminating knowledge corresponds to a distinct type of linguistic referring expression. 5 This is not the place to discuss Kripke’s arguments. It is worth pointing out, though, that there is a far simpler argument from the behavior of proper names in complex sentences to show that proper names cannot be synonymous with definite descriptions—an argument that does not appeal to problematic intuitions about modality. Proper names cannot be synonymous with definite descriptions because they do not display scope ambiguities with respect to negation, whereas definite descriptions do (when they are construed in Russell’s manner). There are two ways of understanding the negation of the sentence “The teacher of Alexander was the greatest living Stagyrite”, but only one way of understanding the negation of the sentence “Aristotle was the greatest living Stagyrite”. But, of course, the issue is not really whether proper names are synonymous with definite descriptions, but rather whether they function as non-rigid designators. 6 Evans allows a category of so-called descriptive names that function as rigid designators although their reference is fixed descriptively. See Evans 1982: §2.3 and the introduction to this volume. 7 It is not simply perceptual recognition that is at stake here. Evans is operating with a broad sense of recognition that extends, e.g., to recognizing the sort of information that would be relevant to determining the truth-value of sentences about the object. 06-Luis-Chap05.qxd 170 22:02:05 9:08 AM Page 170 José Luis Bermúdez The discriminating knowledge that derives from standing in a perceptual relation to an object will most obviously be expressed linguistically by means of demonstrative expressions. The understanding of most proper names will deploy recognitional discriminating knowledge, while knowledge of discriminating facts about an object will most naturally be expressed linguistically through definite descriptions. It is Frege’s basic idea that we should think about the sense of referring expressions in terms of knowing what it would be to identify objects as their semantic values. We can now see how Evans develops this basic idea. Different types of referring expression will be correlated with different ways of identifying objects: demonstratives with perceptual identification, (most) proper names with recognitional identification, and definite descriptions (and some proper names) with what I characterized earlier as identification through satisfaction. Evans devotes relatively little attention to identification through satisfaction (apart from the category of descriptive names), and we will follow him in this. Most of The Varieties of Reference is devoted to recognitional and perceptual identification, both of which rest upon the thinker’s standing in certain information links to the object being thought about. These information links yield the thinker’s discriminating knowledge (or what Evans calls the subject’s Idea, where an Idea of an object is the conception of the object that allows the thinker to differentiate that object from all other objects). In the case of demonstrative referring expressions, these information links are obviously perceptual, but Evans envisages a far wider range of information links in the case of recognition-based thoughts, including information links based upon memory and testimony. In each case the sense of the relevant expression is understood in terms of the ability to exploit the relevant information links. Descriptive identification, in terms of which many authors have proposed that the sense of referring expressions can be completely understood, plays at best a peripheral role for Evans, applicable almost exclusively to those expressions that have the surface structure of definite descriptions. It is this basic distinction between descriptive identification, on the one hand, and recognition-based and demonstrative identification, on the other, that allows him to argue that a great many singular referring expressions are Russellian. The following passage is instructive. 06-Luis-Chap05.qxd 22:02:05 9:08 AM Page 171 Evans and the Sense of “I” 171 In the case of description-based identification, a subject may have a fully coherent Idea-of-an-object, a, even if there is nothing that would be satisfied by that Idea. . . . But this is not so in the case of either of the other modes of identification. In those cases, if there is no object that would be identified by the purported mode of identification, then there is no coherent Idea-of-anobject, even in this sense. In these cases it is therefore true, in the strictest sense, that where there is no object, there is no thought. (1982: 136) It remains unclear, however, how Evans’s general approach to the sense of referring expressions is to apply to the first-person pronoun, which does not seem obviously to involve either recognition-based identification or demonstrative identification. I sketch out the basic contours of Evans’s account of the sense of the first-person pronoun in section IV. First, though, we need to consider why indexical expressions in general, and the first-person pronoun in particular, pose prima facie difficulties for a Fregean notion of sense. That will be the task of section III. III INDEXICALS AND FREGEAN THOUGHTS The object picked out by an indexical expression is a function of the context in which the expression is uttered. In the case of demonstratives such as “this” or “that”, the object referred to is most frequently determined by the ostensive gesture accompanying the utterance (or by whatever else makes a particular object peculiarly salient in that context). Simple rules determine a particular person as the referent of token-reflexive expressions such as “I” and “you”—roughly, “I” refers to the utterer of the sentence and “you” to the person (or persons) addressed. These simple rules are the basis from which we can understand more complex cases involving, for example, deferred reference (as when you write a letter on my behalf using the first-person pronoun). Given that indexical expressions have varying reference, any account of the sense of those expressions confronts a stark choice. Does a given indexical expression have a single sense holding across different contexts of utterance, or does it have a different sense for each different context of utterance? For the first-person pronoun, is there a single sense of “I”, or is each person’s use of “I” to refer to himself associated with a different sense of “I”? It is clear that there are pressures in each direction. On the one hand, it 06-Luis-Chap05.qxd 172 22:02:05 9:08 AM Page 172 José Luis Bermúdez seems plausible to demand that what I understand when I hear you saying “I am hungry” is the same as what you understand when you hear me say “I am hungry”. There is a sense in which we are saying the same thing. But this means abandoning the principle that the sense–reference relation cannot be one–many. If sense determines reference and reference is context-dependent, then sense must be context-dependent. In the face of these conflicting pressures John Perry has suggested fractionating Frege’s unitary notion of sense into two distinct notions: the notion of sense, on the one hand, and the notion of a thought, on the other (Perry 1977).8 The sense of a sentence containing an indexical expression is the context-insensitive component of what we might intuitively think of as the “meaning” of that sentence. So, the sense of a simple sentence of the form “I am F” is composed of the incomplete sense of the predicate “F” conjoined with an element corresponding to the linguistic rule that a token of “I” refers to the utterer of that token. In a simple sentence of the form “It is F now”, the incomplete sense of “F” is complemented with an element corresponding to the linguistic rule that a token of “now” refers to the time of utterance. And so on for the other indexical expressions. The thought expressed by the utterance of an indexical expression, on the other hand, is somewhat closer to a singular thought in the sense discussed earlier. It is composed of the actual object picked out by the relevant indexical expression, together with the sense of the relevant predicate expression. Let us return to the three dimensions of Frege’s notion of a thought identified in section I. Thoughts are: (1) the bearers of truth and falsity (the things that are true or false); (2) the objects of propositional attitudes (the things that are believed, hoped, etc.); 8 Perry uses somewhat different terminology in “The essential indexical”, where “sense” is replaced by “belief state” and “thought” by “belief”. Here is how he characterizes the distinction there: “The shoppers, for example, are all in a certain belief state, a state which, given normal desires and other belief states they can be expected to be in, will lead each of them to examine his cart. But, although they are all in the same belief state (not the same total belief state, of course), they do not all have the same belief (believe the same thing, have the relation of belief to the same object). . . . Belief states individuated in this way enter into our common sense theory about human behavior and more sophisticated theories emerging from it. We expect all good-hearted people in that state which leads them to say ‘I am making a mess’ to examine their grocery carts, no matter what belief they have in virtue of being in that state” (Perry 1979: 181). 06-Luis-Chap05.qxd 22:02:05 9:08 AM Page 173 Evans and the Sense of “I” 173 (3) the senses of assertoric sentences (the things expressed by sentences that purport to describe the way things are). In the case of indexicals, Perry suggests, the third dimension comes apart from the second. The thoughts that are the objects of the propositional attitudes expressible by sentences involving demonstratives are context-sensitive and object-dependent, whereas the senses of those sentences are context-insensitive and do not include the objects referred to by the indexical expressions they contain. It is senses that carry the burden of psychological explanation, while it is thoughts that are picked out in propositional-attitude ascriptions. Perry thinks that this fractionation of Frege’s notion of sense is independently motivated. He has two arguments, each starting from the distinctive explanatory role of the states of mind expressible in sentences involving demonstratives. The first line of argument comes across clearly in the following passage. We use senses to individuate psychological states, in explaining and predicting action. It is the sense entertained, not the thought apprehended, that is tied to human action. When you and I entertain the sense of “a bear is about to attack me”, we behave similarly. We both roll up in a ball and try to be as still as possible. Different thoughts apprehended, same sense entertained, same behavior. When you and I both apprehend the thought that I am about to be attacked by a bear, we behave differently. I roll up in a ball, you run to get help. Same thought apprehended, different sense entertained, different behavior. Nevertheless, it is surely the thought apprehended that is the indirect reference of a sentence containing a demonstrative reference in the scope of “believes”. (Perry 1977: 494–5) For sentences involving indexical expressions, Perry claims, content and functional role, come a part. The very person picked out by the first-person pronoun is part of the content of a belief expressed by a sentence involving “I”—and the beliefs expressed by two different persons each uttering equiform tokens of a given “I”-sentence are correspondingly different. But two people will respond in the same way to the sentence “A bear is about to attack me”, even though that sentence gives rise in each of them to a different belief. So, we need a further notion to capture what it is that leads these two people to behave in similar ways. The second argument also stresses the demands of psychological explanation. Perry starts from the assumption (discussed in section I) that Frege is committed to the thesis that all referring expressions 06-Luis-Chap05.qxd 174 22:02:05 9:08 AM Page 174 José Luis Bermúdez are synonymous with definite descriptions. This leads him naturally to the thought that, if a given indexical expression has a Fregean sense, it must be possible to substitute some coreferring definite description not involving any indexical expressions for that expression in any sentence in which it might feature.9 Perry shows with considerable plausibility that any candidate definite description will lack the immediate implications for action possessed by the putatively synonymous demonstrative expression. Hence, on the plausible assumption that synonymous expressions cannot have diverging implications for action in a single context, it follows that indexical expressions cannot be synonymous with definite descriptions. Clearly, this second line of argument will have a limited impact on those Fregean theorists, such as Evans, who deny that all referring senses have to be synonymous with definite descriptions. But Perry’s points do make very clear an important constraint upon any non-descriptive notion of sense applicable to demonstratives: namely, that it should capture the cognitive dynamics of such expressions. Leaving aside the claims about definite descriptions, let us turn to Perry’s suggestion that we need to distinguish between the sense of a sentence involving a demonstrative and the thought expressed by that sentence in order to accommodate the requirements of psychological explanation. Yet it is counter-intuitive to sacrifice the basic tenet of propositionalattitude psychology that people act the way they do because of what they believe. According to Perry, what people believe (the content of their belief) is not directly relevant to explaining how they behave. My belief that I am about to be attacked by a bear is a belief about me (and correspondingly different from your belief that you are about to be attacked by a bear), and this fact can only be captured at the level of the thought expressed by the indexical sentence “I am about to be attacked by a bear”. But the fact that the sentence “I am about to be attacked by a bear” is about me, as opposed to you, is not on Perry’s view a part of the explanation of why a certain type of behavior is correlated with uttering that sentence. Suppose we leave aside the analysis of sentences involving demonstratives and instead ask about indexical beliefs simpliciter. 9 The qualification rules out, e.g, taking “I” to be synonymous with “the utterer of this token sentence”. 06-Luis-Chap05.qxd 22:02:05 9:08 AM Page 175 Evans and the Sense of “I” 175 The content of an indexical belief, according to Perry, is an objectdependent proposition. But what work does this object-dependent proposition do within the cognitive economy, given that the things that explain behavior are not object-dependent and are more general? The object-dependent proposition is what is either true or false. The bearers of truth and falsity are also what feature in inferences. This is most obvious when the inferences are truthfunctional. If the validity of an inference is a function of the truth-values of its constituents, then those constituents must be truth-apt. But the constituents of truth-functional inferences must also be capable of featuring in non-truth-functional inferences. With this in mind Perry’s position becomes even more puzzling. The sorts of things to which we appeal in giving psychological explanations of behavior are not the sort of things over which inferences are defined. How, then, can we make sense of the idea of a person reasoning their way towards acting in a certain way. It looks very much as if the conclusion of a practical inference must be an object-dependent proposition. Let us suppose that this practical inference issues in action. It is natural to think, then, that an adequate explanation of the relevant action will correctly identify the conclusion of the practical inference. And yet this is precisely what Perry is committed to denying.10 10 It is important to clear up a potential confusion at this point. Perry draws a close parallel between his notion of sense and Kaplan’s notion of character (Kaplan 1979, 1989). Kaplan, like Perry, fractionates the notion of thought into two components: character and content. The content of an utterance, roughly speaking, is the proposition it expresses in a given context. This proposition can be understood as a function that assigns a truth-value to each index, where an index specifies a world, a time, an agent, a location in that world, etc. We can understand the contents of subsentential linguistic expressions accordingly. The character of a linguistic expression, in contrast, is the component that corresponds to linguistic meaning. An expression’s character is a function that assigns a given content to each context. So, for example, the character of “I” is the function that assigns to each context the constant function from possible worlds to the agent of the context. In other words, the character of “I” is effectively that the denotation of any token of “I” is always the utterer of that token. One of Kaplan’s principal contributions to the study of indexicals is his development of a logic of demonstratives in which the notion of character plays an integral part. However, Kaplan’s logic of demonstratives is perfectly compatible with my claims about the split between inferences and psychological explanation. Although Kaplan’s character (Perry’s sense) does feature in his logic of demonstratives, it does so in a very subordinate role. Kaplan holds that the character of a linguistic expression is given by linguistic rules such as those deployed in the recursive definitions that serve to define truth and denotation in a context. There are no inferences at the level of character, although we need character to identify the 06-Luis-Chap05.qxd 176 22:02:05 9:08 AM Page 176 José Luis Bermúdez Given the counter-intuitive position in which Perry ends up, it is worth looking again at his arguments that object-dependent propositions are too specific for psychological explanation. Perry’s argument rests on two implicit principles. The first is that all tokens of a given behavior-type require the same explanation. The second is that giving the same explanation for two different behaviors that are tokens of the same type involves attributing the same psychological states to the authors of the two behaviors. It is easy to see how those two principles rule out giving explanations of different tokens of a given behavior-type in terms of object-dependent propositions—provided, of course, that we really do have different tokens of a different behavior-type in the cases that Perry discusses. Perry is assuming that if our bodies move in comparable ways, then we are behaving in the same way, so that there is just one behavior that is instantiated both when you roll yourself up in a ball and when I roll myself up in a ball. It is true that there is a way of thinking about behavior in which this is indeed the case. This is the sense of “behavior” in which a behavior is just a series of bodily movements. But one might think that psychological explanations explain actions rather than bodily movements, where actions are at least in part individuated by their goal. Looking at the aims of psychological explanation in this way, it is no longer so obvious that my rolling up in a ball is the same action as your rolling up in a ball. After all, my rolling myself up in a ball is intended to save me from the bear, while your rolling yourself up in a ball is intended to save you from the bear. One might be inclined to say that in this case we have two different actions that are effected by similar sets of bodily movements.11 If one looks at the matter like this, then Perry’s two principles do not yield the same result. What I am doing is saving myself from the bear. This action-type can have other tokens, such as, for example, my shooting the bear. These tokens would, in accordance with Perry’s two principles, have the same explanation: namely, my belief that I am about to be attacked by a bear. Your rolling yourself up into a ball is a token of a completely different action-type: namely, the action-type of your saving objects over which inferences are defined and evaluated. Truth and validity arise only with reference to specific contexts, whereas the whole point of the notion of context is that it abstracts away from the specifics of context. (I am grateful to Chris Gauker for encouraging me to address this issue). 11 A similar point is made by Evans (1982: 203–4). 06-Luis-Chap05.qxd 22:02:05 9:08 AM Page 177 Evans and the Sense of “I” 177 yourself from the bear, and, like the other tokens of that type, it is to be explained in terms of your belief that you are about to be attacked by a bear. Of course, our actions are related. But this does not mean that there is a single thing (the sense of the sentence “I am about to be attacked by a bear”) that explains both our behaviors. We can do justice to the similarity by noting that there is a significant similarity between the explanations that might be offered of my action and yours respectively. Consider the proposition expressed by the sentence “JLB is about to be attacked by a bear” (leaving open for the moment whether that proposition is to be understood in Perry’s way or in some other way). We see that proposition as made up of an element corresponding to the proper name “JLB” and a propositional function corresponding to the incomplete expression “—— is about to be attacked by a bear”. Both my thought that I am about to be attacked by a bear and your thought that you are about to be attacked by a bear involve that very same propositional function being completed by constituents that in some sense correspond to the first-person pronoun we might use to express our beliefs. Surely the fact that both thoughts can be interpreted in terms of a single propositional function captures the respect in which they are similar. This brings us back to a new formulation of the problem with which this section began. Suppose we accept, pace Perry, that there must (in the case of indexicals in general and “I” in particular) be a single thing that does the explanatory work that Perry parcels out between thought and sense. Is this to be understood in terms of Frege’s notion of sense? The following section considers Evans’s attempt to answer the challenge. IV EVANS ON THE SENSE OF “I” Evans discusses “I” in two different places. In “Understanding demonstratives” he offers an account of the sense of “I” in the context of a general discussion of demonstratives (Evans 1981 b). In The Varieties of Reference Evans devotes an entire chapter to selfidentification. There he investigates what he calls each person’s “I”-idea—the discriminating knowledge of oneself that allows one to think “I”-thoughts. I shall assume in this paper that the two 06-Luis-Chap05.qxd 178 22:02:05 9:08 AM Page 178 José Luis Bermúdez discussions are continuous, so that what Evans says about “I”-ideas can be taken to represent his views about the sense of the first-person pronoun. An “I”-thought, according to Evans, is a thought that might typically be expressed through a sentence containing the first-person pronoun. Since Evans adopts an orthodox Fregeanism, according to which the sense of a sentence can be equated with the thought it expresses (as opposed, e.g., to positions such as that of Perry), it seems to follow that the sense of the first-person pronoun should be equated with the first-person component of the “I”-thought. Frege famously wrote that “everyone is presented to himself in a special and primitive way in which he is presented to no one else” (Frege 1918: 25). One strand in Evans’s account of the sense of “I” interprets this idea in a way that harmonizes with the general thesis (discussed in section II) that the sense of a referring expression is to be understood in terms of the utterer’s discriminating knowledge of the referent of that expression. Everyone who uses the first-person pronoun with understanding does so in virtue of ways they have of thinking about themselves that are both primitive and not available to anyone else. Using the first-person pronoun to refer to oneself requires being in touch with a particular object (namely, oneself), but not in a way that involves picking oneself out as the referent of the pronoun. Rather, it involves being in receipt of certain types of information— information that is distinctive precisely in virtue of being the sort of information that does not require identification of a particular object as the source of that information. The types of information underwriting what Evans somewhat confusingly calls selfidentification are such that there can be no question but that one is oneself the source of the relevant information. Evans characterizes these types of information by using Shoemaker’s notion of immunity to error through misidentification. This notion is defined in terms of the impossibility of a certain type of error: to say that a statement “a is !” is subject to error through misidentification relative to the term ‘a’ means that the following is possible: the speaker knows some particular thing to be !, but makes the mistake of asserting “a is !” because, and only because, he mistakenly thinks that the thing he knows to be ! is what ‘a’ refers to. (Shoemaker 1968: 557) Judgments are immune to error through misidentification when this type of error is impossible. Immunity to error is an epistemological 06-Luis-Chap05.qxd 22:02:05 9:08 AM Page 179 Evans and the Sense of “I” 179 notion relativized to the grounds upon which a judgment or statement is made, so that, e.g., the same sentence can be employed to make two different statements, only one of which is immune to error, if the grounds on which the statement is made on the two occasions of utterance are appropriately different. Not all judgments that would be expressed by sentences involving the firstperson pronoun are immune to error through misidentification (relative to the first-person pronoun). The relevant judgments are just those that are grounded in and based upon the appropriate sources of information—sources of information that can provide information only about oneself. Autobiographical memory is one such source. When one remembers something “from the inside”, there can be no question but that the person who had the remembered experience is oneself (leaving aside the possibility of what Shoemaker has called quasimemories—see Shoemaker 1970 and Evans 1982: §7.5). Introspection is a further source. If I come to believe that I am having a certain thought, I might well be mistaken, but I cannot come to believe through introspection that someone is having a certain thought and be mistaken about who that person is. Ordinary visual perception is a source of self-specifying information about one’s own spatial location and trajectory. One’s field of view is structured in a way that offers information about one’s spatial relations to perceived objects and about the route one is taking through the perceived environment (Bermúdez 2002, 2003). This information can be inaccurate, but cannot be information about anyone else’s spatial location or trajectory. Somatic proprioception also provides selfspecifying information that is immune to error through misidentification. If, for example, I form the judgment that my arms are folded on the basis of feedback from joint position sense, then there is no sense (as the world currently is) in which I can be mistaken about whose arms it is that are folded.12 We can now see how Evans proposes to develop Frege’s suggestion that “everyone is presented to himself in a special and primitive way in which he is presented to no one else”. Understanding Frege’s “way of being presented to oneself” as involving ways of thinking of oneself (in line with his general way of understanding 12 I am prescinding here from debates posed by the possibility of being hooked up to someone else’s body. For further discussion see Ayers 1991 and Cassam 1995. 06-Luis-Chap05.qxd 180 22:02:05 9:08 AM Page 180 José Luis Bermúdez what Frege says about modes of presentation), Evans elucidates what is “special and primitive” in terms of the capacity to think thoughts, and to make statements, that are immune to error through misidentification relative to the first-person pronoun. The sense of the first-person pronoun is the ability to keep in touch with the person that one in fact is, the person referred to by one’s use of the first-person pronoun. This ability is a function of one’s possessing sources of information about oneself that are immune to error through misidentification. There are two further dimensions to Evans’s account of the first person. One has to do with the “output” side of first-person judgments. Being appropriately in touch with the person one in fact is is a function of one’s ability to act on judgments derived from the appropriate information sources. Evans’s account speaks directly to the issues about psychological explanation and cognitive dynamics raised by Perry. If I see a bear coming towards me, I acquire information that is suitably immune to error through misidentification to the effect that there is a bear coming towards me. So I act in the appropriate manner, immediately and without needing to identify myself as the person towards whom the bear is coming. There is no room for the thought that might be expressed as “The bear is coming towards someone, but is that me?” and no delay in moving straight into action. It might turn out that the bear has designs on the person behind me, rather than on me, but this would not be an error of misidentification in the required sense. It would be a mistake about the ramifications of the fact that the bear is coming towards me. With respect to both the input and the output sides of the sense of the first-person pronoun, Evans can be seen both as fleshing out some of Frege’s sketchy comments on “I” and as developing an account of the functional role of “I”. The final strand in his account of “I”-ideas departs both from Frege and from considerations of functional role, as standardly understood. Indispensable though these familiar ingredients (an information component and an action component) are in any account of the Ideas we have of ourselves . . . they cannot constitute an exhaustive account of our “I”-ideas. So long as we focus on judgements which a person might make about himself on the basis of the relevant ways of gaining knowledge, the inadequacy may not strike us. A subject’s knowledge of what it is for the thought “I am in pain” to be true may appear to be exhausted by his 06-Luis-Chap05.qxd 22:02:05 9:08 AM Page 181 Evans and the Sense of “I” 181 capacity to decide, simply upon the basis of how he feels, whether or not it is true—and similarly in the case of all the other ways of gaining knowledge about ourselves. However, our view of ourselves is not Idealistic: we are perfectly capable of grasping propositions about ourselves which we are quite incapable of deciding, or even offering grounds for. I can grasp the thought that I was breast-fed, for example, or that I was unhappy on my first birthday, or that I tossed and turned in my sleep last night, or that I shall be dragged unconscious through the streets of Chicago, or that I shall die. (Evans 1982: 208– 9) We can, Evans emphasizes, effectively think about ourselves from a third-person point of view, entertaining possibilities to which we have no informational links (as in the vivid examples he gives). The truth conditions of these thoughts are far more complex than the truth conditions of thoughts about oneself that are based upon introspection, proprioception, and other sources of information that is immune to error through misidentification. Consider a sentence of the form “I am F”, where “F” is a predicate whose applicability to oneself can be determined through one of the informational links we have been discussing—hence where the sentence as a whole is immune to error through misidentification relative to the first-person pronoun. The truth condition for “I am F” is intimately tied to the subject’s ability to use the relevant information link to determine whether the predicate applies. Because “I am F” is immune to error through misidentification, the subject does not have to identify himself in order to grasp the truth condition of the thought that the sentence expresses—the thought will be true just if the subject can detect F-relevant information in the favored way. The issue, from the subject’s point of view, is simply whether or not there is F-ness. But when it comes to “I am F” sentences that are not immune to error through misidentification, the subject’s grasp of the truth condition must involve, not simply sensitivity to the presence or absence of F-ness, but an ability to determine what it would be for it to be he who is F. And this in turn means that the subject must be able to think of himself in a much richer way than is required in order to grasp the truth condition of sentences that are immune to error through misidentification. Evans returns at this point to Russell’s principle and to the idea that thinking about an object involves having discriminating knowledge of that object. For Evans this requirement holds no less when the object in question is oneself than when it is an object in 06-Luis-Chap05.qxd 182 22:02:05 9:08 AM Page 182 José Luis Bermúdez the distal environment. So, the final strand in Evans’s account of “I”-ideas (and hence of the sense of the first-person pronoun) is that the thinker possess discriminating knowledge of himself, where this knowledge is understood in terms of the subject’s ability to locate himself within the objective spatio-temporal world.13 Here is how Evans puts it. It seems to me clear that as we conceive of persons, they are distinguished from one another by fundamental grounds of difference of the same kind as those which distinguish other physical things, and that a fundamental identification of a person involves a consideration of him as the person occupying such-and-such a spatio-temporal location. Consequently, to know what it is for [" ! I] to be true, for arbitrary ", is to know what is involved in locating oneself in a spatio-temporal map of the world. (Evans 1982: 211) Being able to locate herself within a spatio-temporal map of the world allows a thinker to determine the truth condition of any thought (whether that thought involves predicates that are susceptible to error through misidentification, or immune to such error). With this third dimension in his account of “I”-ideas, Evans moves away from Frege and closer to a broadly Kantian conception of self-conscious thought, on which the ability to think about oneself is intimately tied to the ability to think about the objective order of things.14 V PRIVACY VERSUS OBJECTIVITY Evans explicitly endorses Frege’s suggestion that the sense of “I” is private. Against Perry, who maintains that the idea of a private sense is incompatible with Frege’s conception of the objectivity of thought, Evans argues forcefully that there is no incompatibility between privacy and objectivity. What is absolutely fundamental to Frege’s philosophy of language is that thoughts should be objective—that the existence of a thought should be independent of its being by anyone, and hence that thoughts are to be distinguished from ideas or the contents of a particular consciousness. 13 See Evans 1982: §7.1 and the appendix to ch. 7. This aspect of Evans’s view has been taken up in recent books by Campbell (1994) and Cassam (1997). See also Bermúdez 1998. 14 06-Luis-Chap05.qxd 22:02:05 9:08 AM Page 183 Evans and the Sense of “I” 183 When Frege stresses that thoughts can be grasped by several people, it is usually to emphasize that it is not like an idea. A true thought was true before it was grasped by anyone. A thought does not have to be owned by anyone. The same thought can be grasped by several people. (PW [Frege 1969] 251) His most extended treatment of the nature of thoughts—‘The Thought’— makes it clear that it is the inference from shareability to objectivity which is of paramount importance to Frege, rather than shareability itself. Since an unshareable thought can be perfectly objective—can exist and have a truth value independently of anyone’s entertaining it—there is no clash between what Frege says about ‘I’-thoughts and this, undeniably central, aspect of his philosophy. (Evans 1982: 313) Evans proposes two respects in which thoughts involving the sense of “I” (as he is proposing to understand it) will count as objective. First, they are not reducible to the contents of an individual consciousness. Second, they exist and have a truth-value independently of anyone entertaining them. Why might thoughts of a given class not be reducible to the contents of an individual consciousness, even though they were not shareable? There would be such irreducibility if there were thoughts of that class that have never been thought by anyone. This certainly makes perfectly good sense for, e.g., mathematical thoughts (on all non-constructivist views of mathematics). It cannot, however, be applied to non-shareable “I”-thoughts (or indeed to any nonshareable token-reflexive thoughts), simply because the identity of any given token-reflexive thought is a function of the episode of thinking. Recall that the reference of “I” in a sentence involving the first-person pronoun is fixed by the token-reflexive rule that (in standard cases) picks out the speaker of the relevant token sentence as the person to whom the pronoun refers. Since an “I”-thought is a thought expressed by a sentence involving the first-person pronoun, one would expect the reference of the first-person component of the thought to be fixed in an analogously token-reflexive way. The whole point of token-reflexivity, and indexicality in general, is that the identity of the thought is determined by the context in which it is thought—in which case, there is no thought without an episode of thinking. If this episode has to be an episode featuring the subject of the thought (as it must be if the thought is non-shareable), then I see no sense in which the thought can exist and have a truth-value independently of being thought by the particular person who is the subject of the thought. 06-Luis-Chap05.qxd 184 22:02:05 9:08 AM Page 184 José Luis Bermúdez Senses are not just the objects of propositional attitudes and the bearers of truth and falsity. They are also the units of communication— what is understood when someone utters a sentence with understanding and when someone else hears that sentence with understanding. A thought that is non-shareable is a thought that is non-communicable. Frege himself seems to be in two minds about how communication is secured in sentences involving the first-person pronoun. On the one hand, the famous discussion in “The thought” that we have already considered appears to entail the non-shareability of the sense of “I”. On the other hand, however, in the posthumously published “Logic”, Frege suggests that the thought that I entertain is the same as the thought that I communicate in a sentence involving the firstperson pronoun. It is not necessary that the person who feels cold should himself give utterance to the thought that he feels cold. Another person can do this by using a name to designate the one who feels cold. (Frege 1914: 1969: 235) The emphasis here is not only on the communicability of “I”-thoughts, but even on the possibility of expressing them with sentences not featuring the first-person pronoun. There is an ambiguity in the passage from “Logic”. What is the name that someone else might use to designate me when I feel cold? Is it a proper name or a pronoun? This is particularly important in the present context. If the objectivity of “I”-thoughts requires shareability, then it must be possible for someone else to think a thought that I would express using “I”. So we can ask: how would that person express this thought? What words would they use to express the same thought that I would express using “I”? We are looking for a token utterance that would be synonymous with my token utterance. Clearly, this would not be an utterance employing the first-person pronoun. No two utterances involving “I” can be synonymous if they are uttered by different people. As we saw in section III, there are good reasons for thinking that utterances involving “I” will generally not be synonymous with utterances involving proper names and definite descriptions, even when those names/descriptions pick out the same person as the use of “I”. Of course, there will be some cases in which proper names and/or definite descriptions will work just fine. And there are no doubt cases where someone might use the third-person pronoun successfully to express the same thought that I would express using “I”. But there 06-Luis-Chap05.qxd 22:02:05 9:08 AM Page 185 Evans and the Sense of “I” 185 seems no prospect of finding a rule to the effect that, whenever there is a thought expressed by a sentence involving “I”, that thought will equally be expressible by a sentence involving a proper name, or a definite description, or the third-person pronoun. The most plausible candidate for such a rule would be a token sentence involving the second-person pronoun, accompanied by a suitable way of fixing reference (either anaphorically or ostensively). As Mark Sainsbury has emphasized, a token of “I” can be synonymous with a token of “you”, even though the type-expression “I” is not synomymous with the type-expression “you” (Sainsbury 1998). The thought that I express with a token of “I” in a given context just is the thought that you express with a given token of “you” in that same context (or a suitably related context), so that when you understand my utterance “I am F” you do so by grasping a thought that you would yourself express by saying “You are F”. Let us say, then, that the objectivity and shareability of “I”-thoughts imposes a constraint. This is the constraint of preserving the token-equivalence of “I” and “you” within a given context. I will call this the symmetry constraint. Evans’s account of the sense of the first-person pronoun does not meet the constraint. Since he holds that the sense of the first-person pronoun is private, my utterance of “I” cannot be synonymous with your utterance of “you”. If my understanding of the firstperson pronoun is tied to my sensitivity to, for example, proprioceptive information from my own body, then there is no sense in which my utterance of “I” can be synonymous with your appropriately related utterance of “you” (for the understanding of which proprioceptive information from my body is not in any sense an issue). The point holds more generally. What Evans calls the information component of an “I”-idea is a set of information links that can hold only between a person and himself. A fortiori, therefore, thoughts containing such an “I”-idea are not shareable—and so (on our assumed equivalence of each person’s “I”-idea and his grasp of the sense of the first-person pronoun) we have the consequence that the sense of the first-person pronoun is not shareable. Evans’s argument that objectivity does not require shareability is an unsuccessful attempt to show that this result is not as unpalatable as it might immediately appear. But, even if it had been successful, objectivity is not the only issue here—and an objective thought that cannot be communicated or denied hardly fits the 06-Luis-Chap05.qxd 186 22:02:05 9:08 AM Page 186 José Luis Bermúdez Fregean picture of thought and linguistic understanding. Evans is well aware of this, and is explicit about his rejection of the Fregean model of communication in The Varieties of Reference. We do not have Frege’s full model of the role of sense in communication; for we do not have the thesis that communication between speaker and hearer requires them to think of the referent in the same way (in any plausible or natural sense of that phrase). The nearest we come to the full Fregean model is with expressions like “here” and “now”; the furthest we move away from the full model is with expressions like “I” and “you”. (Evans 1982: 316) There is a tension between this and some of the other things that Evans says about communication. A few pages before the quoted passage Evans comments that “it is a fundamental, though insufficiently recognized, point that communication is essentially a mode of the transmission of knowledge” (Evans 1982: 310). It is true that Evans is careful to elucidate this transmitted knowledge in a de re manner, so that what is transmitted in a sentence of the form “x is F” is knowledge of x to the effect that it is F. But many would agree with Frege that the transmission of knowledge in communication is best seen in terms of a fully determinate de dicto proposition. If I know of x that it is F, then there must be some way in which I am thinking of x. Suppose that it is different from the way in which you think of x. Then it is perfectly possible for the proposition that I apprehend when you tell me that x is F to be a proposition that you would deny (and would be perfectly rational in denying). It is hard to see how this can count as an instance of the transmission of knowledge. Applying this back to the case of the first person yields further support for the symmetry constraint. VI RETHINKING EVANS’S ACCOUNT OF THE SENSE OF “I” Evans’s account has three different strands: what he calls an information component, an action component, and a location component. The information component has to do with the thinker’s sensitivity to sources of information about himself that are immune to error through misidentification. The action component has to do with the thinker’s disposition to take a certain category of thoughts 06-Luis-Chap05.qxd 22:02:05 9:08 AM Page 187 Evans and the Sense of “I” 187 about himself to have immediate implications for action—those thoughts whose typical linguistic expression would be the firstperson pronoun. The location component has to do with the thinker’s ability to locate himself within the spatio-temporal order. Of these three components it is the first that most directly generates the difficulties identified in the previous section. My sensitivity to information about myself that is immune to error through misidentification is not something that you can share. Hence, building such sensitivity into the sense of the first-person pronoun will have the result that “I”-thoughts are incommunicable. There are two motivations for including an information component in an account of the sense of the first-person pronoun. The first is an analogy between the first-person pronoun and other referring expressions. Evans’s general picture of the sense of referring expressions involves the holding of an information link between thinker and object. In the case of “I”, the most obvious information links are precisely the ones that Evans discusses— introspection, somatic proprioception, autobiographical memory, and so on. But these information links do not perform the same function as the information links invoked to explain other referring expressions. On Evans’s general picture, information links allow the subject to identify a particular object as the referent of the referring expression. But the whole point of the information links that he discusses with respect to the first-person pronoun is that they do not make possible any sort of identification of a given person as oneself. They are immune to error through misidentification precisely because the judgments to which they give rise do not involve any such identification. So, the work that the information component is doing in explaining the sense of the first-person pronoun cannot have to do with making possible identification of the thinker as the referent of the first-person pronoun. Evans’s second motivation for introducing an information component into an account of the sense of the first-person pronoun is the idea that we can understand the truth conditions of certain thoughts about ourselves simply by understanding how we would determine their truth-value on the basis of the deliverances of sources of information that are immune to error through misidentification. So, for example, there need be nothing more to my grasp of the truth condition of the thought that I am in pain than my ability to judge that I am in pain when I feel pain—and there need 06-Luis-Chap05.qxd 188 22:02:05 9:08 AM Page 188 José Luis Bermúdez be nothing more to my grasp of the truth condition of the thought that I am thinking about my car than my ability to judge, on the basis of introspection, that I am thinking about my car. But why should this have anything to do with the sense of the first-person pronoun? Surely what is at stake here is not a way of thinking about myself, but a way of thinking about what it is to be in pain—or to be thinking about a certain object. Consider the corresponding linguistic expressions “——is in pain” and “——is thinking about——”. Evans holds that we understand these linguistic expressions by understanding what it would be for them to be “completed” by an arbitrary name of an object to yield a full sentence expressing a thought. Understanding this thought is, of course, a matter of understanding what it would be for the thought to be true. In the case of the predicate “——is in pain”, therefore, one would expect this understanding to include the difference between its first- and third-person application criteria—the difference between applying the predicate to ourselves on the basis of how we feel and applying it to others on the basis of what they say and how they behave. Similarly for the dyadic predicate “——is thinking about——”. Here too we would expect an account of what it is to understand this predicate to distinguish between the introspective grounds on which one might apply it to oneself and the rather different grounds on which one might apply it to others. It looks, therefore, as if the sensitivity to peculiarly self-specifying information that Evans builds into his account of the sense of the first-person pronoun is already accommodated in any account of what it is to understand the relevant predicates. It is true that a thinker can only understand thoughts that might be expressed with sentences of the form “I am F”, where “F” is a predicate that can be applied to oneself on the basis of information that is immune to error through misidentification, if they are appropriately sensitive to the relevant information and know that they are. But this sensitivity is part of what it is to understand the predicate “——is F”, rather than forming part of the sense of the first-person pronoun. Neither motivation for introducing the information component into an account of the sense of the first-person pronoun is compelling. There is a significant disanalogy between the firstperson pronoun and other referring expressions in this respect. Evans is surely right (given his assumptions about Russell’s principle and the 06-Luis-Chap05.qxd 22:02:05 9:08 AM Page 189 Evans and the Sense of “I” 189 dependence of successful reference upon discriminating knowledge) to propose a general model for referring expressions in which the ability to identify the object referred to depends upon the existence of some sort of information link between thinker and object. And he is right also to stress that we do have dedicated sources of information about ourselves. But he is wrong to think that these information sources serve as an information link underpinning mastery of the first person pronoun of the sort that we find underpinning mastery of other referring expressions. Suppose, then, that we drop Evans’s information component from an account of the first-person pronoun. One immediate consequence is that the sense of the first-person pronoun ceases to be so obviously private and incommunicable. But can we move from this to an account that explains both what it is to utter the first-person pronoun with understanding and what it is to understand someone else’s utterance of the first-person pronoun? And, in particular, can we move to an account of the sense of the firstperson pronoun that meets the symmetry constraints discussed in the previous section? Let us start with what is involved in understanding someone else’s utterance of the first-person pronoun. One needs to be able to identify the utterer of the token sentence—one needs to know who uttered the sentence in question. But in what does this knowledge consist? It is not necessary to be able to track down the physical source of the token, since it is frequently enough to know what one would have to do to track down the physical source of the token. If I overhear someone whose voice I do not recognize say “I” in the next room, then I know what I need to do to track down the physical source. I need simply go next door. Yet the ability at stake is clearly the ability to locate a given individual—to know what one would have to do to arrive at a position where one can apply the token-reflexive rule. Equally clearly, the location that needs to be determined is egocentric. What matters is that I should be able to locate the utterer of “I” relative to my own position. In fact, the canonical way of understanding a token of “I” involves being able to put oneself in a position relative to the speaker such that one can think about him in second-person terms (it is irrelevant that one might not be the addressee of the sentence in question) or in those third-person terms in the expression of which “he” functions 06-Luis-Chap05.qxd 190 22:02:05 9:08 AM Page 190 José Luis Bermúdez primarily as a demonstrative pronoun (more or less synonymous with “that man”).15 This account of what it is to understand someone else’s utterance of “I” clearly meets the symmetry constraint. It also preserves one element in Evans’s account of the sense of the first-person pronoun. Evans thinks that an integral element in one’s “I”-idea is the ability to think of oneself as located on a spatio-temporal map of the world—to be able to identify oneself with an element in the objective order. My understanding of another person’s utterance of “I” is constituted by my ability to identify a particular person at a particular spatio-temporal location as the utterer of the token. If I am not able to locate (or at least be in a position to go about locating) the utterer of the token within the objective order, then I cannot properly be understood as having understood the relevant utterance. Everything said so far, therefore, is fully compatible with Evans’s account of the sense of the first-person pronoun (now that the information component has been dropped). But this, after all, is the easy case. Does it give us any clues regarding an account of what it is to use “I” with understanding? Is there a comparable requirement on one’s own self-referring use of “I” that one be able to locate oneself within the objective order? It is tempting to think not, since I can truthfully utter the sentence “I have no idea where I am”. More generally, an objection to any attempt to place substantive requirements upon the ability self-consciously to refer to oneself in thought or speech is that it seems possible for a thinker who does not meet the requirement to wonder what is happening. Surely, it is argued (most famously in Anscombe 1975), an anaesthetized amnesiac temporarily deprived of the use of his senses is perfectly able to reflect upon what is happening to him. It is clear why we do not want to deny that the anaesthetized amnesiac would be able to think about himself self-consciously.16 15 It is true that not all cases of apparently successful communication involving “I” will take this form. Telephone conversations with people one does not know proceed perfectly well when one is in complete ignorance of the location of the other person. But there are of course degrees of success in communication and degrees of understanding. A conversation can proceed perfectly well in circumstances in which one might hesitate to attribute full understanding to the parties concerned. The notion of understanding is an ideal to which we may approximate more frequently than we attain it. 16 Nor is Evans committed to denying that the anaesthetized amnesiac would be able to think about themselves. It is true that Evans does countenance the possibility 06-Luis-Chap05.qxd 22:02:05 9:08 AM Page 191 Evans and the Sense of “I” 191 Such a person could be asking questions to which there are perfectly determinate answers (“Where am I?”, “Are my arms above my head?”, “Why is this happening to me?”). And if he can think them, then he can obviously think about himself. Clearly, then, it cannot be a condition of his being able to think about himself that he know the answer to these questions, or any like them— and, a fortiori, nor can knowing the answer to any such question be part of the sense of the first-person pronoun. But does this rule out some form of location component in the sense of the firstperson pronoun? In one obvious sense it does. The anaesthetized amnesiac cannot identify himself as a particular person at a particular spatio-temporal location. But let us look again at what Evans says. It seems to me clear that as we conceive of persons, they are distinguished from one another by fundamental grounds of difference of the same kind as those which distinguish other physical things, and that a fundamental identification of a person involves a consideration of him as the person occupying such-and-such a spatio-temporal location. Consequently, to know what it is for [" ! I] to be true, for arbitrary ", is to know what is involved in locating oneself in a spatio-temporal map of the world. (Evans 1982: 211) The requirement here is not that one actually be able to identify oneself as the person occupying such-and-such a spatio-temporal location, but rather that one knows what is involved in locating oneself in public space. If one understands that one is such a person, then one can understand the truth condition for any sentence of the form “I ! "” where “"” is a name or description picking out a particular person. It seems clear that on this way of understanding the location component, it is easily satisfied by the anaesthetized amnesiac. It might seem that we have accommodated Anscombe’s problem case only at the cost of thinning down the idea of a location component in the sense of the first-person pronoun so far that it becomes impossible to meet the symmetry requirement. I understand that an “I”-thought might “fail to net any object at all” (1982: 253). But the examples he gives are ones in which either the information links to one’s own body are systematically distorted (as would happen if one were receiving kinaesthetic information from someone else’s body) or the thinker discovered himself to be a brainin-a-vat. Evans has the resources to deal with Anscombe’s case. I discuss one such resource below. 06-Luis-Chap05.qxd 192 22:02:05 9:08 AM Page 192 José Luis Bermúdez someone else’s uttterance of the first-person pronoun just if I can locate the utterer of “I” relative to my own position—and this, I suggested, is effectively the ability to put myself in such a position relative to the utterer of “I” that I can address him in the second person. But this seems very different from the bare knowledge of what is involved in locating oneself in public space. After all, that bare knowledge can take the form simply of knowing that one is here, without one’s being able to identify here as anywhere except the place where one is. The utterer of “I” does not seem to need any substantive capacity to locate himself. So how can there be the equivalence between what the utterer understands and what the hearer understands demanded by the symmetry thesis? At this point we can turn to one of the most suggestive aspects of Evans’s thinking about the first person: namely, his analysis of egocentric spatial thinking and the relation between “here”-thoughts and thoughts about space in general. Evans repeatedly emphasizes the interdependence of “I”-thoughts and “here”-thoughts. Consider the following passage. It is not the case that we first have a conception of which material object in the world we are (or what it would be to establish that), and then go on to form a conception of what it is for us to be located at a particular place. It is true that [p ! here] is the same thought as [I am at p]; but this does not mean that I identify p as where I am. This would raise the question ‘How do I identify myself, and make sense of my being located somewhere?’, but—if we had to keep the capacity to grasp ‘here’-thoughts out of the picture—would make it impossible to answer it. (Evans 1982: 153) Evans’s point is that the ability to think of oneself as being here is far richer than the trivial ability to think of oneself as being located at the place where one is. A subject cannot have “here”-thoughts in isolation. “Here”-thoughts are part of a complex network of egocentric spatial thoughts defining an egocentric space, in which the spatial relations of objects are coded in a frame of reference centered on the thinker.17 But nor, Evans argues, can a subject have a conception of egocentric space in isolation. This is where Evans makes one of his Kantian moves, arguing that the ability to think about space in egocentric terms cannot be separated from the subject’s capacity to think objectively about space (to possess what 17 As Evans points out (1982: 157), an egocentric space is not a special kind of space, but a distinctive way of representing space. 06-Luis-Chap05.qxd 22:02:05 9:08 AM Page 193 Evans and the Sense of “I” 193 psychologists and cognitive scientists call a cognitive map).18 What is un-Kantian about Evans’s understanding of spatial knowledge, however, is that he analyzes the thinker’s capacity to superimpose their representation of egocentric space on their representation of objective space (and vice versa) in essentially practical terms—in terms of the thinker’s ability to find his way about (Evans 1982: 165). On Evans’s account, therefore, the location component in the sense of the first-person pronoun involves an essentially practical ability to integrate egocentric and objective ways of thinking about space. This underwrites one’s sense of oneself as a physical object and makes it possible to grasp the truth conditions of thoughts about oneself. One obvious consequence of this is that, provided Anscombe’s anaesthetized amnesiac has retained this essentially practical ability to find his way around (even if only in a dispositional sense), then Evans’s account allows him to have genuine “I”-thoughts (on the assumption that we have dropped the information component from an account of the sense of the first-person pronoun). Less obviously, but more importantly, this points the way to seeing how the proposed modified version of Evans’s account meets the symmetry requirement. My use of the first-person pronoun with understanding rests upon an ability to grasp the truth conditions of the thoughts thereby expressed, where those truth conditions are states of affairs involving a certain person (namely, myself). My grasp of which person that is depends upon my being able to locate myself within the spatial environment. Consider now your understanding of my use of the first-person pronoun. You understand my use of the first-person pronoun just if you grasp that very same truth condition—that is, the state of affairs involving me. Your grasp of which person is at stake depends upon your ability to locate that person (namely, me) within the spatial environment. As we saw when discussing Perry in section III, it is often necessary to individuate actions in terms of their ends. The reasons that lead us to individuate actions in this way extend to abilities, so that the ability you exercise in understanding my utterance of “I” is in fact the same ability I exercise in uttering “I” with understanding. And so the symmetry constraint is met. 18 For further discussion of the relations between egocentric and objective ways of thinking about space, see Campbell 1993 and 1994. 06-Luis-Chap05.qxd 194 22:02:05 9:08 AM Page 194 José Luis Bermúdez VII CONCLUSION We see, therefore, how a modified version of Evans’s account of the sense of the first-person pronoun resolves the fundamental problems that indexical expressions and indexical thoughts pose for a Fregean theory of sense. The account that Evans offers resists the temptation to fractionate the notion of sense and provides us with a suitably unitary notion of the sense of “I”. It does so, however, at a price. Evans abandons the shareability of “I”-thoughts—and hence the possibility of understanding someone else’s “I”-thoughts (on the assumption that one cannot understand a thought without grasping it). The price is not worth paying. Fortunately, we need not pay it, because Evans is mistaken in thinking that what he calls an “information component” is required to explain the sense of the first-person pronoun. Once we drop that element of his account, and instead emphasize the points he makes about self-location, we see that the sense of the first-person pronoun is perfectly shareable. Building on Evans’s insights into first-person thought and the firstperson pronoun, we can develop an account of the sense of “I” that allows the thoughts expressed using “I” to count as fully Fregean thoughts, shareable and with determinate truth conditions. This is surely one of Evans’s most enduring contributions to the philosophy of thought and language.