Self-Identification and Self-Reference *

Ingar Brinck

1. Introduction

[1] To know who one is, and also know whether one's experiences really belong to oneself, do not normally present any problem. It nevertheless happens that people do not recognise themselves as they walk by a mirror or do not understand that they fit some particular description. But there are situations in which it really seems impossible to be wrong about oneself. Of that, Ludwig Wittgenstein once wrote:

It is possible that, say in an accident, I should feel pain in my arm, see a broken arm at my side, and think it is mine, when really it is my neighbour's. And I could, looking into a mirror, mistake a bump on his forehead for one on mine. On the other hand there is no question of recognising a person when I say I have toothache.... it is as impossible that in making the statement "I have toothache" I should have mistaken another person for myself, as it is to moan with pain by mistake, having mistaken someone else for me. (1958: 67)
In the passage in which this remark is found, Wittgenstein distinguishes between two kinds of use of "I". The first use, as object, as in "I have broken my arm" or "The wind is blowing in my hair", he holds, involves the recognition of a particular person, and there is the possibility of error as concerns the identity of the person. In the other use, as subject, as in "I think it will rain" or "I am trying to lift my arm", no person is recognised. No mistake can be made about who the subject is.

[2] By this distinction, Wittgenstein drew attention to a phenomenon that later has been dubbed immunity to error through misidentification (henceforth IEM) (see Shoemaker 1968). It occurs in cases in which it would be absurd or nonsensical to describe one's predication by saying: "Someone is F-ing (e.g. yawning, seeing, walking, etc.), but is it I?" In such cases, it appears impossible to be wrong about who the subject is.

[3] Of course, IEM does not pertain to cases in which one has knowledge about oneself by observation, as when one sees the reflection of a person walking across the street in the windows along the pavement, and infers that the person must be oneself. On the contrary, in cases that typically exhibit IEM, the subject does not need to infer that she instantiates property F.

[4] Wittgenstein explained the fact that it sometimes is impossible to misidentify oneself in saying "I" by denying that "I" in such uses, i. e. in its use as subject, refers. Statements like "I have pain" are not, says Wittgenstein, about a particular person. "I" in its use as subject does not function like a name, and it does not rely on a descriptive recognition of the speaker (1953: sections 404, 410; 1958: 67). Elisabeth Anscombe, for one, followed Wittgenstein in denying a reference to "I" (1975).

[5] Other writers, notably Sidney Shoemaker, have, when trying to explain IEM, focused on statements in which "I" occurs together with mental predicates. Shoemaker maintains that a statement is IEM relative to "I" if it contains a psychological predicate such that it is known by the speaker to be instantiated in a special way.

[6] This way is such that if one is aware that the predicate is instantiated, one thereby (non-inferentially and non-observationally) knows that it is instantiated in oneself. In saying "I feel pain", the speaker has not identified somebody that he knows to be doing so as himself (Shoemaker 1984: 9ff). Thus, a speaker does not have to first identify himself before he can use "I" correctly in a sentence that contains a predicate of the right kind. Self-identification and self-reference depend fundamentally on the self-ascription of those predicates. Shoemaker writes:

[T]hat a statement "a is f" 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 f, but makes the mistake of asserting "a is f" because, and only because, he mistakenly thinks that the thing he knows to be f is what "a" refers to. (1984: 7ff)
This error occurs, for example, if a person sees a figure dressed in baggy clothes in a mirror and in an upset voice exclaims "I look awful!", not at first realising that he is looking at somebody else. Statements are not subject to error through misidentification if the speaker cannot mistake another person for himself in the way described in the quotation.

[7] Contrary to Shoemaker, Gareth Evans held that not only psychological predicates, but also predicates referring to bodily states occur in statements that are IEM (Evans 1982: 216ff). This means that the way of gaining information about oneself that lies behind IEM does not only concern mental states.

2. Evans on immunity to error through misidentification

[8] Gareth Evans treated IEM at length in Varieties of Reference. At the bottom of our capacity never to, in a fundamental sense, misidentify ourselves, lies, according to Evans, a general capacity of gaining identification-free knowledge. Such knowledge is based on a certain way of receiving information, and is characterised by not being dependent on its source being conceptually identified by the receiver of information.1 The source is identified by its spatio-temporal location.

[9] Evans introduced the notion of immunity to error through misidentification (IEM) in relation to demonstrative identification. Demonstrative thoughts are disposed to be controlled by information that the subject has gained in a way that relies on a continuous information-link between subject and object. On the basis of that information, the subject can locate the object.

[10] Evans held that the information-link, in order to guarantee IEM, must be supplemented by a so-called fundamental identification. Such an identification consists in conceiving of an object as an element of the objective order, which in turn depends on knowing what it is for the object to be located at a position in space. This knowledge is excersised as a practical ability, but is based on patterns of reasoning, that constitute a theory of inferences about the subject's own location and movements through space (Evans 1982: 223). It is the theory that makes us realise that we are part of an objective order and that our thoughts are general.

[11] By "objective order", Evans meant a conception of space in which the egocentric and public perspectives have become integrated. This conception arises when perceptions of public space has an impact on individual action and the subject directs her action at the public space. When the subject thus can impose a conception of egocentric space on the public one, she has an adequate Idea of a point in public space (Evans 1982: 168).2 The objective, spatial world is "a world of objects and phenomena which can be perceived but which are not dependent on being perceived for their existence" (Evans 1982: 222).

[12] To have a conception of this world, the subject must be able to "think of his perception of the world as being simultaneously due to his position in the world, and to the condition of the world at that position". The idea of there being an objective world is simultaneous with the idea of the subject being in the world, his location determined by what he perceives. So the idea of an objective world and of the subject's being at a particular location cannot be separated. The objective self comes into existence when egocentric action and general thought are woven together.

[13] That information controls demonstrative thought means that there is no gap between the perceptual information that the subject receives and the concept formed from that information. According to Evans,

a bit of information (with the content Fx) is in the controlling conception of a thought involving a subject's Idea of a particular object if and only if the subject's disposition to appreciate and evaluate thoughts involving this Idea as being about an F thing is a causal consequence of the subject's acquisition and retention of this information. (1982: 122)
The subject of a demonstrative thought does not infer that the perceptual information and the conceptual content concern the same object.

[14] Evans did not restrict IEM only to thoughts about oneself, as is customary. He defined it in terms of indentification-free knowledge:

When knowledge of the truth of a singular proposition, "a is F" can be seen as the result of knowledge of the truth of a pair of propositions, "b is F" (for some distinct Idea, b) and "a=b", I shall say that the knowledge is identification-dependent: it depends (in part) on the second basis proposition, which I shall call the identification component. We might say that knowledge of the truth of a singular proposition is identification-free if it is not identification-dependent. (1982: 180) 3
He subsequently introduced a narrow sense of identification-freedom in order to exclude singular propositions that are not "information-based". Such singular propositions identify the referent purely descriptively as opposed to by some direct information-link between object and subject and do not stand in a causal relation to the referent. Evans wrote:
[K]nowledge of the truth of a singular proposition is identification-free in the narrow sense if (i) it is not identification-dependent and (ii) it is based on a way of gaining information from objects. (1982: 181)
Note that in identification-free knowledge, predication and identification are simultaneous. There is no room for mistakes concerning the identity of the referent.

[15] Similarly to demonstrative thought, self-conscious thought about oneself rests upon a disposition to have one's thoughts controlled by information. This information concerns both our mental and bodily states. And as in the case of demonstrative thought, the object of thought -- in this case, oneself -- must, to be subject to a fundamental identification, that is, be conceived of as an element of the objective order. The identification involves, except for locating oneself spatiotemporally, conceiving of oneself as a person, that is, as a being of a certain (general) kind.

[16] The ability to conceive of oneself as an instance of a category (the category of persons) is closely related to another ability, captured in what Evans called the Generality Constraint. According to it, conceptual thought is essentially structured. Evans wrote that

if a subject can be credited with the thought that a is F, then he must have the conceptual resources for entertaining the thought that a is G, for every property of being G of which he has a conception. (1982: 104)
Likewise, the subject can think of other objects than a, of which he has a conception, that they are F. Structured thoughts can be generalised and combined, and the combinations can be transformed.4 The two applications of the Generality Constraint together make it possible for the subject to think of himself as part of the objective world. To grasp the generality of a concept involves understanding that it applies in the same way to other people as to oneself, and that the evidence that underlies ascription of beliefs to others is of the same kind as is used when one ascribes beliefs to oneself.

[17] "I"-ideas develop from the subject's capacity to act on incoming information and thereby place herself in the objective order. Her interaction with the surroundings will give rise to an egocentric self-concept, that reflects what I call an indexical self-awareness. As long as we only consider situations in which the subject acts on information gained from the actual context, this self-concept will account for the knowledge that the subject has of herself. But to grasp propositions about oneself whose value cannot be decided on the basis of information accessible in the actual context, another kind of self-concept is needed. This concept requires that the subject has a detached self-awareness, that enables the subject to think about herself as detached from any particular context.5

[18] This means that the information-link and the action-link together do not exhaust what being a subject amounts to (Evans 1982: 208). Evans maintained that one's Idea of oneself also must comprise

a knowledge of what it would be for an identity of the form [I=dt] to be true, where dt is [...] an identification of a person which... is of a kind which could be available to someone else. (1982: 209)
This identification would be available to someone else and would conform to the Generality Constraint. The content of such an identification is what it is for a particular subject as a person to be located at a point in spacetime in the objective world.

[19] To entertain thoughts about oneself that are IEM, it is, apart from conceiving of onself as a person, necessary that one is disposed to receive information about oneself, and that the information gives rise to identification-free knowledge. There are two ways to gain such knowledge about one's bodily states (Evans 1982: 220ff). The first one is by perception of one's body, through, for instance, proprioception, sensitivity to heat, pressure, and so on. No gap will open up between knowing that F is instantiated and knowing that it is so in oneself, because "to have or to appear to have the information that the property is instantiated just is for it to appear to him that he is F" (1982: 221).

[20] The second one is indirect, and consists in knowing one's position, orientation, and relation to other objects by perceiving the external world. Evans made a connection between having this kind of knowledge and having a concept of the objective world. As mentioned above, a concept of the objective world is necessary for locating objects and thereby for demonstrative thought in general. The perceptual experiences of the world and of oneself are interdependent, and so are the objective and subjective spheres. In Evans account, there is, as it were, no telling which comes first.

[21] We gain knowledge about ourselves through our mental states by self-ascribing states like beliefs or other attitudes. The self-ascription does not rely on directing oneself inwards (Evans 1982; e.g., 225, 230). In ascribing beliefs to myself, for instance, I do not scan my inner self, but direct my attention outwards, to the states-of-affairs about which I have beliefs.

[22] Self-ascription of perceptual experiences is different (1982: 226ff). Perceptual experiences consist, according to Evans, in informational states with a non-conceptual content that can be true or false. Judgements are reliably caused by these states. In going from an informational state to a judgement, the information carried by the state becomes conceptualised. The experience then becomes conscious to the subject.

3. Some problems with Evans' account

[23] The problems of Evans' account are related to the fact that the information-links by themselves neither can determine contents of thought or guarantee the IEM of judgements based on purely informational states. As Evans himself pointed out, they are necessary, but not sufficient to do this (1982: 88, and section 6.2, especially at 148). The reason why information-links cannot determine contents of thought is that they do not help the subject single out which object he is thinking of. An object must be identified in some particular way to be an object of thought. The information-link cannot on its own provide the subject with an object.

[24] Furthermore, to be disposed to treat an object as relevant to the truth-value of a proposition is not sufficient for conceptual thoughts, unless the subject's thought of the object obeys the Generality Constraint (1982: 147). This means thinking of it as independent of any particular context, but as located in the objective order. An informational or causal link of Evans' kind will not do for this, since generality implies that the content can be detached from any context and thus cut loose from any such link. The subject should be able to think about the object also when the situation excludes demonstrative thought.

[25] The identification required for generality cannot consist in a definite description, if it is supposed to be involved in thoughts that exhibit IEM. Descriptions are not IEM, since they are not guaranteed to pick out the same (numerically identical) object in all contexts. They are satisfied by whatever object happens to fit them in the actual context.

[26] The view that causality is not sufficient to determine the object of thought can be developed in various directions. Causal links lack a conceptual connection with intentionality and thereby with thought. This means that an intentional relation prima facie cannot be reduced to a causal one. If the aim is to naturalise intentionality, one has to find some way to restrain the causal link in order that it univocally picks out the right object. This is hard to do without using intentional concepts, as indicated by the massive critique that has been raised against such attempts.

[27] If a causal link nevertheless would adequately pick out the object, it does not follow that the object has been selected in a way that serves its cognitive purpose in the context of thought. Indeed, it would not be guaranteed that the object was displayed in a way that would at all be intelligible. It may even be that the representation lacked content and only occured as a vehicle.6

[28] These possibilities emerge because of the difference between the causal and the conceptual realms. Causality does not imply conceptualisation, but intentional thought demands concepts. Reflection depends on having concepts. The cause and the content of a representation cannot be equated by fiat.

[29] In my view, the role of the causal link as concerns I-thought is to tie the conceptual realm to the natural world. Reflection is confined to that realm. Therefore, we need both the causal relation that ensures that thoughts have, as it were, real objects, and concepts that ensure that our thoughts are subject to rational conditions.

[30] To the information-links, Evans adds the fundamental identification, that is supposed to single out the source of the information without the possibility of error. The fundamental identification involves identifying the source as part of the objective order. As concerns I-thoughts, this means to identify oneself as a person.

[31] As far as I can see, it really is impossible to conceive of oneself only from a third-person perspective, since one cannot tie general beliefs to oneself without a first-person conception of oneself. Lewis' example of the two Gods, who know all the facts there is to know about the world, but cannot tell who they are -- who is the God on the tallest mountain and who is the God on the coldest one -- because they lack contextual knowledge, is an excellent illustration of this (see Lewis 1979).

[32] On the other hand, a subject who can only have indexical thoughts about herself will obviously not have a full understanding of "I". She would not be able to entertain context-independent thoughts about herself or think about herself as an instance of a category of beings similar to herself. These points are emphasised by Evans when he asserts that thoughts about oneself must conform to the Generality Constraint.

[33] No doubt, generality is necessary for the ability to entertain a full range of thoughts about oneself, including future-directed, conditional, and counterfactual ones. But is it necessary to explicitly think about oneself as instantiating the Generality Constraint, or is it sufficient to apply it blindly? And which is the connection between generality and IEM?

[34] Clearly, if I-thoughts necessarily involved the performance of an explicit identification of the sort [I=dt], they would always be open to error through misidentification, since there is no guarantee that one knows that one is a particular person P, and thereby instantiates a certain description. Actually, it seems that I-thoughts that rely on the proposition [I=dt] are identification-dependent, according to Evans own definition. This means that such I-thoughts could, by definition, not be IEM.

[35] The condition that I-thoughts have to be general is, as far as I can see, offering a background to particular thoughts. The identification of oneself as a person is of a preparatory kind, providing a condition that in principle should be fulfilled, but that may not hold in practise. It is presupposed by the use of "I". A subject may lose her understanding of the generality of thought, or not take it into consideration, or she may temporarily not be conscious of what it means to be a person, but all the same be able to entertain thoughts about herself that are IEM. Generality constitutes a formal property of thought in all its guises, while IEM concerns a limited class of thoughts. An account of IEM should reflect the particular basis of the latter kind of thought.

[36] But if we take the fundamental identification of oneself as a person to provide a background condition for I-thoughts that does not have to be performed on every occasion of thought, this will provoke a problem for Evans' solution to how to constrain the information-links, which are specific to a certain context. Remember that the information-links are necessary, but not sufficient to determine particular contents of thought or guarantee IEM. They cannot by themselves single out which object the subject is thinking of in a particular context. This means that, as things stand now, either the subject does not for some reason have to perform an identification of herself when thinking about herself in particular contexts of thought, or we need another suggestion as to what constrains the links.

[37] There is yet another problem concerning the identity of the subject. The model of thought that Evans takes as his starting-point brings with it a distinction between the source and the receiver of information conveyed by the link. As regards demonstrative thought, the model appears quite natural. But as regards I-thoughts, it seems inappropriate to distinguish between source and receiver. Rather, the identity of the subject should be assumed from the start, in line with the principle that all entities are identical to themselves, and only be annulled in extraordinary cases.

[38] In connection with bodily self-ascription of such properties that give rise to judgements that are IEM, Evans notes how absurd it would be to question the identity of a body that one has gained information about by, for instance, proprioception or through interaction with one's surroundings (Evans 1982: 221ff). Evans nevertheless allows for the possibility of deviant information-links, and thereby for error through misidentification, in exceptional circumstances. He claims that such possible situations show that error might occur, but that ordinary judgements based on information-links are not subject to IEM. Information-links only guarantee IEM under "normal" conditions (1982: 249).

[39] It is not clear how we determine what counts as normal conditions, or on what grounds we can exclude other cases. For reasons of circularity, we cannot say that the conditions under which the information-links guarantee IEM are normal. To me it seems unsatisfactory to allow for the possibility that source and receiver diverge as in the case of I-thoughts. It signals that something is wrong with Evans' model of thought about oneself.

[40] In Evans' model there is actually two identifications involved for I-thoughts, one of the source a as such, on the basis of the information-link, a is F, and one between source a and receiver b of information, a is b, where b is I, that is, the subject of the thought. The identity between b and I should hold as a result of the identification of the subject as an element of the objective order (which we just saw is identification-dependent as well, since the subject must be identified as a person). This means that I-thoughts are inherently identification-dependent, something that, by Evans' definition, is excluded for judgements that are IEM. Surprisingly, Evans apparently introduces a possibility of error through misidentification in his overall model for I-thoughts.

[41] This example of identification-dependence, pertaining to the source and the receiver of information respectively, is similar to the one brought up above, in relation to identities of the form [I=dt], where dt is a general identification of a person. It constitutes in fact an extension of it, since not only thoughts about oneself that contain descriptive elements, but any I-thoughts that rely on an identification brought on by the split between source and receiver will be identification-dependent.

[42] There is no guarantee that the source and the receiver are identical, and the receiver cannot just assume that she gets information from the right source. In giving the self two roles, both as source and receiver, or as subject and object, a gap opens up that may be difficult to patch up. How is the identity between the two guaranteed? -- Not by an identification built on the information-link, since the problem arises with the link. Constraints on the information-link cannot be taken from the link itself.

[43] The causal relation between subject and object stands in need of a foundation. If the subject should be able to identify the source as hers, it appears that we need a self-concept prior to gaining information about ourselves. Obviously, this runs counter to Evans' theory, since he holds that self-identification is simultaneous with the gaining of information.

4. The nature of IEM

[44] The idea that in order that one's thought latches onto an object, it is necessary to identify the object, sounds peculiar when the object is oneself. It also opens the possibility of error. In a context that leads to judgements that are IEM, moreover, the question of whether the source reflected in the received information is oneself should never be allowed to come up. Still, it is hard to deny that whenever we think about something, we must have a discriminating conception of it. In the case of I-thoughts that occur in judgements that are IEM, this conception must be of a special kind to exhibit identification-independence.

[45] As I pointed out at the outset of the last section, the problems with Evans' theory that I focus on all have to do with the idea that the information-links cannot determine the object of thought or guarantee IEM on their own. It seems that there is something wrong with the fundamentals of Evans' conception of IEM. To find the source of the problems, we need to re-examine IEM. In this section, I will therefor bring up a few different interpretations of it, in order to get a better understanding of its nature.

[46] In the first section, I quoted Wittgenstein on what Shoemaker later dubbed immunity to error through misidentification. It appears that the passage from which the quotation is taken has given rise to a number of interpretations, all directed at illuminating IEM.

[47] Wittgenstein seems to be saying that in some specific cases when the subject believes that a certain property is instantiated, one belonging to a particular kind, no question concerning in whom it is instantiated will arise. The intuition is that in such cases, the subject cannot be wrong about who is the subject of the predicate that designates the property -- if there even is a subject. These cases are expressed by statements containing "I" in its use as subject.

[48] According to Shoemaker, error through misidentification occurs when a subject knows a certain predicate to be instantiated, but is wrong about in whom. There are predicates such that if the subject knows that one of those is instantiated, he cannot be wrong about in whom or about how to identify that person. These predicates are assured to give rise to IEM. Shoemaker spells out Wittgenstein's intuitions about the nature of the predicate and the impossibility of error. He does not agree with Wittgenstein about the reference of "I", though, since he maintains that it is a referential expression. The latter point is, however, of minor importance for the moment.

[49] Anscombe puts forward two conditions for IEM: first, that there is a guarantee against radical reference-failure, that is, the referent exists, and second, that there is a guarantee against mistaken reference, that is, what the subject takes to be the referent is the referent (1975: 56ff). Her definition does not explicitly focus on the kind of property that gives rise to judgements that are IEM, but on the fact that the subject does not make any mistakes concerning the identity of the object. In this respect, what she writes about IEM differs somewhat from Wittgenstein's and Shoemaker's comments about it. They apparently see a connection between the kind of property that is instantiated and the impossibility of mistaken identification.

[50] Evans maintains that knowledge of the truth of a proposition is identification-free, and thus IEM, when the subject knows that property F is instantiated, and her knowledge of the truth of the proposition "a is F" does not depend on knowledge of the truth of a proposition expressing identity, "a is b". Then he adds, as a way of spelling out what lies behind IEM as described by the first two conditions, that the knowledge of the truth of the proposition "a is F" is based on a certain way of gaining information about the object.

[51] It seems that both Evans and Shoemaker think that IEM is connected to the fact that the predicate is true about the subject -- even that it depends on it. But if we take a look at sentences that express judgements that are IEM relative to "I", it turns out that they remain IEM also in case the predicate is misascribed. Compare with the following examples.

(1) I hear bird-song

(2) I am overweight
Even if it turns out that I did not hear anything but hallucinated, or that I am not at all overweight, but on the contrary skinny (as in a bad case of aneroxia), this would not automatically make me revise my opinion about who instantiated these properties. I would still be referring to myself, although I was attributing the wrong predicates. In (1) and (2), even if we have a misascription of predicates, we do not have a misidentification of the referent.

[52] This means that in these and similar examples the predicate does not override the subject-term. Let us compare (1) and (2) with a few examples in which the subject never can be wrong about whether the predicate is instantiated.

(3) I have pain

(4) I think it is going to rain
Some of Evans' examples have a different character, although belonging to the same kind as (3) and (4), since he allows predicates that express bodily states in judgements that are IEM, as in
(5) I am hot
and possibly
(6) I am being pushed
In (3)-(6), predication and identification are interdependent. The basic idea is that there are properties such that if one knows that they are instantiated, and this knowledge depends on a certain way of gaining information, -- the one that Evans tries to specify, the properties cannot be instantiated in somebody other than oneself.

[53] This position calls for comments. For one thing, if IEM depends on a certain way of gaining information, it might be that the property does not in fact have to be instantiated in the subject, as long as the belief that it is so depends on the proper way (to IEM) of gaining information. This means that any belief that a certain kind of predicate is instantiated (the kind that occurs in judgements that are IEM) is based on information that could only be gained in such a way as to be gained from oneself. Then the information that the belief is based on is either true and then about oneself, or distorted and not true about oneself, but not true about somebody else either. The subject of the judgement is in either case oneself. The information could not be about somebody else, since the way it was gained is such that it could not be about someone else than oneself.

[54] Evans surprisingly appears to use a strong form of IEM that I will call immunity to error through misascription (IEA), that only occurs if the judgement as a whole is true. IEA depends on the truth of the predicate. No mistake can be made about the identity of the subject due to misascription of the predicate, in case the predicate is such that if the subject makes the judgement that the property designated by the predicate is instatiated, it is instantiated. The property that gives rise to IEA is self-presenting and infallible. That a property F is self-presenting means that it is such that if I believe that I am F (and I actually am F), then I am certain that I am F. Some states are such that I could not believe that I am in them, without actually being in them, as pain or happiness, or the state of thinking.

[55] But it does not seem that the predicate must be true about the subject for IEM to occur. Even if the predicate is wrong, the subject will in some cases not be misidentified, notably in those that are based on information gained in the way described in the last paragraph. The position that a judgement can be IEM regardless of whether the predicate is true about the subject or not is, it seems, allowed for by a definition put forward by Andy Hamilton. He defines IEM in the following way:

An assertion of "I am F", "I fed", "I shall f", etc., is IEM if and only if the justification subjects would offer for it is such that if they subsequently come reasonably to doubt the assertion, no matter what the cause, it will be senseless for them to cite the original justification as a reason for claiming that none the less someone is F (or fed, etc.). (1995: 335)
Hamilton goes on to say that the original justification is that which the subject would offer if asked to justify the first-person utterance. According to Hamilton, IEM thus concerns what the subject actually believes, and not how she has received the information. He believes that an account of IEM should be internalist. But it is not farfetched to assume that the reason why the original justification cannot be used to claim that somebody else than oneself is F depends on how the information was gained in the first place.

[56] The case which in accordance with Hamilton's definition would give rise to error through misidentification is such that the predicate will take precedence over the subject-term. If the predicate does not fit the subject, we are on some occasions licensed to draw the conclusion that somebody other than whom we initially thought instantiates the property in question.

[57] But what is it that guarantees that the subject is not misidentified in judgements that are IEM, if the information on which the judgement is based can be incorrect? We have reached the position that IEM does not depend on having true information about oneself in the context of judgement, but so far, we do not have a hint of an explanation of what else would guarantee IEM. There are at least two possibilities.

[58] It might be that as long as the information is not gained indirectly, the subject-term will pick out the subject who has gained it, even if it is false. The way of gaining information guarantees IEM. But this suggestion does not avoid the difficulties we had with Evans' account of IEM. We still cannot be certain that the receiver of the information is the same as its source, or that the subject is the receiver. The information-link remains insufficient to explain IEM. And in some cases, there does not seem to be any information of the relevant sort at all connected to the judgement that is IEM, for instance, in the case of hallucinations and imagined experiences. The subject may have faulty beliefs about herself that do not rely on gaining information about oneself directly or immediately, but that could be conjured up from sheer fancies.

[59] A second possibility is that there exists an identification-free default way of referring to oneself, that in some particular circumstances can be overriden by the kind of predicate that relies on such information as renders the predicate an identificatory role. This way of referring does not directly depend on information-gaining, but has been determined beforehand. I will explore this line of thought in the next section.

5. Self-identification and self-reference

[60] We have reached the position that whatever it is that guarantees IEM relative to "I", it cannot only be information gained in a direct or immediate way. This means that IEM must be guaranteed at an earlier stage. We also know that IEM cannot be based on a description of the subject, since descriptions are inherently identification-dependent. To get a grip on what we are looking for to account for the foundation of IEM, let us return to what Evans had to say.

[61] It seems that Evans maintained that in normal circumstances one does not need to perform an identification of oneself every time that one has an I-thought. He asserted that an egocentric self-concept develops as a result of the subject's constant and long-term interaction with the surroundings. It gives, according to Evans, rise to an objective conception both of oneself and the world. This means that the subject is located spatiotemporally through this interaction, because she is at the centre of a network of information- and action-links. In this case, no further contextual identification is called for.

[62] The egocentric self-concept consequently provides the subject with a location in the objective order. She does not need to keep track of herself through different contexts, since the concept evolves in a self-perpetuating process, as the subject moves around in a changing environment. Evans points out that no special skill is necessary to produce suitable judgements about oneself that span past, present, and future. It simply is a feature of the cognitive dynamics of "I" that such judgements can be made (1982: 237ff). As soon as the subject has access to the conceptual sphere, she will also understand that she can be categorised as a person, and thereby have a general self-concept that conforms to the Generality Constraint. She will then be able to think about herself in an infinite number of ways.

[63] But if we take the causal model for I-thoughts, that is evoked by Evans to explain IEM, for granted, it is hard to see how we can avoid a theory in which the subject is held to re-identify herself on each occasion she tokens an I-thought, and in which case an identification of the source at time t with the evolving, continuous subject would be required. How might we retain the view that the content of I-thoughts relies on discriminating knowledge obtained by an information-link, while avoiding the need for an identification in the context of thought?

[64] It appears that one source of the problems with Evans' account is that he tried to answer two questions at once. The first question concerns IEM, the second the idea or concept of oneself that is necessary for entertaining an I-thought or using the word "I". If we look for the raw material needed for answering both questions in Evans' texts, the answer to the first question should most probably be found in the interaction that lies behind the unfolding of the egocentric self-concept, while the answer to the second one would have to do with the way the subject is presented with herself in the actual context. Let us, for the moment, direct our attention to the first matter. I would describe it as follows.

[65] The interaction with the surroundings that gives rise to an egocentric self-concept, as descibed by Evans, will also involve an immediate awareness of oneself as an information-gainer and experiencing subject. This indexical self-awareness has developed from the gaining of information that represents the context from the subject's point of view, making the subject aware of her position in relation to other objects in the context and of herself as enduring through time. It is inherently subjective and accompanies every self-representation, indexing them, as it were, with the mark of subjectivity.

[66] The indexical self-awareness is a direct consequence of being the nexus between incoming information and action. It is independent of having or receiving information either about oneself or the surroundings once it has developed. Consequently, it constitutes a disposition to have thoughts indexed with the point of view of the subject. If the information-links are temporarily disrupted, it remains intact, since it does not depend on a continuous flow of information. But it will only matter to cogntion and agency if tied to contextual information about the subject and her spatiotemporal location.

[67] All information that is gained immediately or directly by way of a causal link will automatically be indexed as subjective. It is not an act of will or intention on behalf of the subject. So will every tokening of "I" be. But unfortunately indexical self-awareness cannot guarantee that uses of "I" are IEM. It will at most give rise to a Kantian I-think, that guarantees that every thought one has actually is one's own. What is needed for IEM is an awareness that this (my) thought is about me.

[68] It sounds plausible that unless the subject has evidence for the opposite, she will take information about herself that has been gained in the way described by Evans to be about herself. Information that has not been gained in this way will not be indexical or centred on the subject. In that case, nothing at all indicates that it is about the subject. If in a particular context it becomes evident that a certain predicate is false about the subject, and the belief that it is true has not been caused in the relevant way, the subject is licensed to draw the conclusion that it may instead be true about somebody else.

[69] IEM would consequently depend on how the subject had gained information about herself, but the information or the information-links as such could not alone guarantee that she did not misidentify the subject of the predicate. In the absence of evidence to the contrary, however, it would be correct of the subject to assume that the predicate was about herself, and she would intuitively do so.

[70] Nevertheless, another type of argument could be given for the IEM of I-thoughts, from evolutionary design. States that carry information about oneself, gained either immediately or directly, play a crucial role in agency. Several conditions on content having to do with the context in which it occurs must be fulfilled for agency to take place. First of all, purely descriptive thought content would not move the agent to action. It has to be connected to the context, that is, the spatiotemporal location of the subject. Further, the content has to be provided with the point of view of the subject: it must concern the subject.

[71] Third, there should not be any room for hesitation as regards the identity of the subject. The information must concern the subject in an immediate way, or many actions will fail to reach their goal. Timing is an essential ingredient in agency. Some of the information that the subject receives will function as a cue for action, and to do so, it must lock into the informational state that the subject already is in, and move her at the precise moment.

[72] Informational states take care of the most basic processes and actions, those that underlie conscious reflection. They provide the impulse that sets off basic actions like running from the enemy, pulling away the hand when burned, drinking when in need of liquid, and so on. It would be an extreme waste of cognitive resources if there were a gap between this part of the cognitive world and the conscious part that springs from it -- if the subject always could doubt whether it really was she who had those experiences. Much energy would be put into finding out to whom those experiences belong, and the hesitation would cause failures, both because of that and because of problems with timing.

[73] I submit that if the perceptual and cognitive faculties of a subject function properly (that is, as designed), the first-person judgements of a subject that are caused by informational states about herself will display IEM. This suggestion cannot provide a constraint on the information-links or a logical guarantee of IEM. What it can do is to tell us in what situations it simply becomes senseless to refer to IEM at all, that is, in cases of mental or physical illness, when the perceptual apparatus is malfunctioning, or in case the perceptual apparatus has been manipulated with. It does not seem that we can get any further in this matter.

[74] Let us return to the question what is required to token "I". I argued above that a causal link cannot alone determine the object of thought, or reference. A conceptual element must be added. But how can it be, if we are talking about judgements or statements that are IEM relative to "I"? The challenge is to show how reference can be conceptual but still remain IEM.

[75] Evans held that demonstratives and indexicals are referent-dependent, that is, that their meaning depends on that the referent occurs in the context of utterance (Evans 1985: 294ff). The function of sense, according to him, is not to determine the reference of the term, but to provide a way of thinking of the referent, a mode of presentation.7 He writes that "[T]o give an account of how a thought concerns an object is to explain how the subject knows which object is in question". To know which object is in question involves being able to distinguish it from all other things. That requires discriminating knowledge. Evans brings up the following sufficient conditions for that: when one can perceive the object at the present time; when one can recognise it if presented with it; and when one knows distinguishing facts about it (1982: 89).

[76] As regards demonstratives and indexicals, knowing which object means being able to identify the object as it appears in the actual context and not by some antecedently given condition (Evans 1985: 303). The identification depends on observation and on the information-links between subject and object. But we do not observe the self, and it seems that the referent of "I"-thoughts, according to Evans, is presented in a different way.

[77] Evans' conditions for discriminating knowledge mentioned above make use of general concepts as well as of information gained in the actual context. Above, I pointed out that, as concerns "I", the general concept is not sufficient to constrain the information-links. Nor is it sufficient to account for the cognitive role of "I". It leaves the latter without an explanation.

[78] Part of the specific cognitive role of "I" is contributed by the subjective perspective of thought content as given in indexical self-awareness. If the information that a subject gains about herself was not centred on the subject, it would not move her to action or influence her behaviour. All kinds of information could be registered, but would not be localised, or causally connected to any particular sources. The information would come from a purely descriptive place in the universe, impossible to pick out in relation to the subject herself, since she would not be able to place herself indexically in the universe. An inert and causally impotent subject that exists isolated from the causal realm would not last long.

[79] The other part of the cognitive role is given by the way the subject is presented with herself in the actual context. This presentation will influence her actions, lines of thought, and decisions. Subjects do not normally function in the same way cognitively even if placed in the same kind of situation, not only because they happen to be at different places in time and space, but also because they happen to be in different individual states at that location.8

[80] Evans is quite right that the information-links are necessary to identify the subject by its spatiotemporal location. But if "I" is to have the adequate cognitive role, not only in agency, but also in reasoning, one has to conceive of onself as something more than a point in a system of coordinates -- which is what the spatiotemporal identification provides. Any subject, or person, could be at a particular point p. A conceptual, individual, and contextual mode of presentation is necessary to individuate the subject.

[81] Such a mode of presentation is given by so-called de re senses.9 Evans and John McDowell have both worked on modifications of Frege's theory that aim at making sense context-sensitive. 10 McDowell describes de re senses as conceptual, but depending essentially for their occurrence on the existence of the referent (McDowell 1984: 283ff). De re senses constitute the content of token expressions, a content which could not be determined without the presence of the referent in the context of utterance (287ff). They present the referent to the speaker in a certain way, which makes the de re sense specific to its res, and provides the cognitive significance of token expressions.

[82] De re senses function as a conceptualisation of the referent. As I see it, they permit the speaker to focus her attention on the referent and give her the means to discriminate it from its surroundings by its non-relational properties.11 As regards "I", the referent is identical with the subject that has those experiences which constitute the foundation for the de re senses. The de re sense is caused by an informational state carrying non-conceptual content about oneself. Saying "I" constitutes a direct expression of the particular point of view of the subject.

[83] Thus, self-reference depends on knowing the linguistic meaning of "I" as well as on being presented with oneself by a de re sense. Since the de re sense is caused by an informational state about the subject, and it constitutes a conceptualisation of the information carried by the state, "I" will always be used from a certain perspective or point of view and be indexed as subjective.

[84] But there are nevertheless uses of "I" in which the referent can be misidentified. As described in section 4, that happens when the predicate takes precedence over the subject-term. "I" can then, although referring to the speaker on the basis of a de re sense, be exchanged for another subject-term that refers to the person who actually instantiates the predicate in question.

[85] Two kinds of judgement (containing "I") that do not exhibit IEM relative to "I" can be discerned. First we have those that focus on properties of oneself that cannot be known immediately or directly, but the knowledge of which is gained from the context, like "weighing 120 lbs" and "facing south". Then we have such that attribute properties the information about which cannot be gained from the context of utterance (although it might be gained in the context of utterance). They are such as "I am Liz Taylor", "I won all my fights", and "I am the editor of Soul".

[86] Judgements that exhibit error through misidentification relative to "I" are ambiguous. According to a first interpretation, the speaker talks about herself as instantiating a certain property, that she actually does not instantiate. On a second interpretation, the speaker talks about a certain person who has a particular property and who she thinks is identical to herself, but is not so.12 But despite the fact that the speaker misidentifies herself, there is no doubt that in saying "I" she intends to refer to herself.

[87] Nevertheless, it seems that the intended referent is the person who instantiates the relevant property. This means that the identification of the referent relies on indirectly gained information. The referent is identified descriptively. On the other hand, "I" will, all the same, refer to the speaker in the context. This is taken care of by the lingustic meaning of the type expression "I", which is to refer to the user (presumably a person) in the actual context.

[88] The ambiguity shows that there is not any point in using the teminology of intended or actual referent relative to uses of "I" that exhibit error through misidentification. The actual referent may equally well be the speaker as whoever instatiates the predicate that did not fit the speaker -- if anybody does so. How we interpret the situation will be a pragmatic matter, depending on cues in the context of use.

[89] At the outset of this article I said that knowing who one is does not normally present any problems. At least it does not if the locution "knowing who" only means to be able to place oneself in space and time relative to other objects and to categorise oneself as a person. Problems may arise as soon as we go beyond these rather modest demands. Still, in the absence of psychological illnesses and neurophysiological injuries, or perhaps science fiction come true, knowing who one is does not usually cause any trouble. The existence of persons is as described above inherently subjective. We do not really run the risk of conflating ourselves with somebody else.

Ingar Brinck
Lund University, Sweden


Anscombe, E. A. (1975) "The First Person". In Mind and Language, S. Guttenplan (ed.). Oxford: Oxford University Press.

Brinck, I. (1997) The Indexical "I". Dordrecht: Kluwer Academic Publ.

Donnellan, K. (1966) "Reference and Definite Descriptions". Philosophical Review 75: 281-304.

Evans, G. (1982) The Varieties of Reference. J. McDowell (ed.). Oxford: Oxford University Press.

----- (1985) "Understanding Demonstratives". In Collected Papers Oxford: Oxford University Press.

Hamilton, A. (1995) "A New Look at Personal Identity". The Philosophical Quarterly 45, No.180: 332-349.

Lewis, D. (1979) "Attitudes de dicto and de se". The Philosophical Review Vol. LXXXVIII, No. 4, October: 513-543.

McDowell, J. (1984) "De Re Senses". The Philosophical Quarterly Vol. 34, No. 136: 283-294.

----- (1986) "Singular Thought and Inner Space". In Subject, Thought, and Context, P. Pettit and J. McDowell (eds.). Oxford: Oxford University Press.

Shoemaker, S. (1968) "Self-Reference and Self-Awareness". Journal of Philosophy Vol. LXV, No. 19: 555-567.

----- (1984) Identity, Cause and Mind. Philosophical Essays. Cambridge: Cambridge University Press.

Wittgenstein, L. (1953) Philosophical Investigations. New York: Macmillan Publishing Company.

----- (1958) The Blue and Brown Books Oxford: Blackwell.


* I would like to thank Johannes Persson for valuable comments. (back)

1 By "information" I will throughout this article intend content that is directly causally derived from a state or an object and that is not conceptual. Properties that distinguish conceptual from non-conceptual content are, e.g., having a determinate truth-value, having a subject-predicate structure or a constituent structure, and obeying the principle of compositionality. Non-conceptual content is more fine-grained than conceptual content. Perceptual experience is often taken as a paradigm-case of states that carry non-conceptual content. Non-conceptual content moreover appears to be essentially first-personal, in that it always is from a point of view. In saying that non-conceptual content nevertheless is content, I mean that it does not influence the subject purely causally, but by its representational properties, that is, by being about something and by presenting that thing in a certain way to the subject. For a discussion of non-conceptual content, see chapter 5 of (Brinck 1997). (back)

2 The subject has a number of dispositional connections with places and these form a network of links that constitute the subject's unified egocentric space. (back)

3 That a and b are distinct Ideas of a particular object roughly means that they constitute two separate ways of thinking or conceiving of that object. See (Evans 1982: 104). (back)

4 Full-blown conceptual thought shares these features with language, but that does not mean that thought and language must have a similar representational form. (back)

5 For the concepts of indexical and detached self-awareness, see (Brinck 1997: section 5.5 and chapter 6). (back)

6 A syntactic representation may, of course, have a causal effect on behaviour. It cannot, however, be an element in reasoning or reflection, since it is not transparent to the subject. Neither can it influence the subject by its representational properties, as described in footnote 8 above. (back)

7 Evans denies that sense is something that preceeds the encounter with the object of thought, in the way sense usually is thought to do in order to determine reference. He rejects the interpretation of sense as a way of determining reference and instead underlines its role as providing thought content. See, e.g., page 294 and also footnote 6. (back)

8 The character of the individual state depends, in turn, on the history of the subject, the subject's physiological and chemical make-up, and other similar factors. (back)

9 In (Brinck 1997), I argue that tokens of "I" refer by way of a de re sense. See especially chapters 4 and 5. (back)

10 For example see (Evans 1985); (Evans 1982) chapters 1, 6, 7; (McDowell 1984) and (McDowell 1986). Evans and McDowell claim that their interpretation of Frege is no less correct than the one that tends to understand sense in terms of Russellian descriptions (see McDowell 1986: 143 ff). (back)

11 Attention-focusing is an active state of the subject, that involves having a certain attitude to the object of attention, as when inspecting a mental image or having a propositional attitude. It also requires that the subject categorises and identifies the object. (back)

12 We could name the first reading "referential" and the second "attributive", after the distinction introduced by Keith Donnellan between attributive and referential uses of descriptions. Donnellan points to the difference betweeen denotation and reference. In the former case, the description is used to state something about (attribute properties to) a certain subject, while in the latter case the description primarily is a tool for calling attention to the referent, a job that with an equally good result could be done by some other device. See (Donnellan 1966). (back)

1998 Ingar Brinck

EJAP is a non-profit entity. EJAP may not be printed, forwarded, or otherwise distributed for any reasons other than personal use. Do not post EJAP on public bulletin boards of any kind; use pointers and links instead. EJAP retains the right to continuously print all the articles it accepts, but all other rights return to the author upon publication.

EJAP is hosted at The University of Louisiana at Lafayette. Please send comments and questions to