Guide to Chapter Seven of Gareth Evans'
The Varieties of Reference

Rick Grush (Department of Philosophy, UC San Diego)
Copyright  © Rick Grush 2003
Copyright holder grants permission to distribute, quote, and reproduce for any non-commercial purposes provided authorship and copyright are clearly indicated.

Chapter Seven: Self-Identification

Table of Contents:

- 7.1 Introductory
- 7.2 Immunity to error through misidentification
- 7.3 Bodily self-ascription
- 7.4 Mental self-ascription
- 7.5 Memory
- 7.6 The possibility of reference failure
- 7.7 Conclusions
- 7.A Appendix

7.1 Introductory

This initial section covers four issues. First, Evans discusses similarities and differences between self-identification and other forms of identification he has discussed earlier. Second, he argues that 'I' is a referring expression, contrary to a number of prominent views. Third, Evans argues that self-identification is not a matter of some elusive pure subjectivity nor is it even exclusively a matter of the mental. Fourth, Evans issues some cautions about the persistence of the subject's identity through time.

From the top of page 205 to the bottom of page 207, Evans discusses some initial similarities between 'I'-thoughts and other kinds of thoughts discussed already, such as 'here'-thoughts, in order to motivate the treatment of self-identification (he also notes some differences). The main kind of similarity is in the form of some apparent functional isomorphisms between 'I'-thoughts and the thoughts discussed in Chapter Six. In the case of 'here', for example, the subject's 'here'-thinking involves an Idea that is constitutively tied to certain ways of gaining information, and certain ways of acting on the subject's part. For example, getting information along an appropriate channel to the effect that ' ... is hot' is instantiated serves for the subject to have information to the effect that 'here it is hot'; and similarly for the action component: being too hot here must, ceteris paribus, be such as for me to act to leave the vicinity. Neither of these components will be involved in our Ideas of other locations.

Similarly, the kind of self-thinking we are interested in is thinking that has analogous functional connections to perception and action. My thought about 'the largest member of the UCSD philosophy department' need not have any special connections to my perceptual input or action; but my 'I' Idea must have such connections. There are information channels such that getting information to the effect that ' ... is facing a tree' is instantiated just is for me to have information to the effect that I am facing a tree. Similarly, if my knowledge that a bear is about to attack me, will have certain kinds of direct consequences for my actions — consequences by no means guaranteed by my being informed that the largest member of the UCSD philosophy department is about to be attacked by a bear.

Though there do appear to be similarities of these functional sorts between 'I'-Ideas and other kinds of Ideas, Evans points out that 'I'-Ideas have features that are novel. The first element concerns the role that our memories play in our 'I'-Ideas, and this will be the topic of 7.5. The second involves the fact that our 'I'-Ideas essentially involve a kind of self-reference. (This will be discussed to an extent in Part One of the Appendix.)

From the top of page 208 to the bottom of page 209 Evans discusses and reject the view that our 'I' notion is not genuinely contentful, the supposition that it has content being an illusion of grammar. The idea is that the verbal expression 'I' has a role in the language, and this is mistakenly taken to imply that it has a referent, or that it is associated with some kind of content. This view was maintained (in slightly different forms), by Geach, Strawson, Wittgenstein, and Anscombe. Evans argues that this view is mistaken, that there is substantive content to our 'I'-Ideas. Geach and Strawson both point out that, in effect, that in my judgment that I am in an awful muddle or that I am in pain, there does not appear to be anything corresponding to an identification of something that is in a muddle, or is in pain. The judgments might as well be expressed as 'What an awful muddle', or 'There is pain'.

Evans points out that the premise in both cases seems to be that there is nothing in such judgments that corresponds to getting myself to think of the right object (as there is in communicative situations that might require the use of 'I'). But the conclusion, that one is in fact not thinking of something that is oneself, that nothing corresponding to a substantive 'I'-Idea is involved in such judgments, does not follow. Evans points out that his own account has a place for such 'identification-free' thoughts, but that the Generality Constraint forces us to recognize that there must be more to our 'I'-Ideas than this. Our 'I'-Ideas must not only be such as to accommodate the kinds of identification-free thoughts that we evidently can have about ourselves, but it must also be capable of supporting thoughts about ourselves that are not tied to such circumstances. Evans' account will have the advantage that it can deal with both kinds of situation. (The specifics of how it handles both kinds of situation will unfold throughout the chapter and appendix.)

Beginning at the very end of page 209 (with the sentence 'It is vital to remember...') to the middle of page 213, Evans discusses the view that 'I'-thoughts are concerned exclusively or primarily with some kind of mysterious subjectivity or with mental states exclusively. This discussion opens with Evans pointing out that on his own view, our 'I'-Ideas are such that they are adequate only when we not only have an Idea with the correct functional structure that accommodates identification-free knowledge, but we must be able to identify ourselves with an element of the objective order. (It is this second aspect that accounts for the satisfaction of the Generality Constraint by out 'I'-thinking.)

Evans points out that Nagel, for example, has claimed that such an identification, of ourselves with an element of the objective order, is not possible, because of an essential divide between the subjective and objective. Evans' recap of Nagel's argument is something like this: When we conceive of an objective state of affairs, such as T.N. standing before a burning house, then the identification of myself with T.N. (or with any other person in the objectively construed world) makes no difference to how the world is conceived. All the objective facts are the same. Therefore, there isn't really any content to the supposition that I am identical with an objectively construed person. Since such content would presumably be that content a grasp of which would constitute knowledge to the effect that I am identical with an objectively construed person, there would seem to be no such thing as knowing what such an identity consists in.

Evans' replay is that while it is true that no difference will be manifest in how things are objectively conceived, there will be great differences in how 'the subject's immediate environment' is conceived. If I am identified with T.N., for example, then my environment will include a burning house; if I am identified with R.G., then it will not. Evans admits that such knowledge can only be expressed via indexical expressions, but says that this is no objection to treating it as knowledge. (At the very least, it puts the burden of proof back on Nagel's shoulders.)

The top paragraph of page 212 contains some additional considerations to the effect that the subjective and objective cannot be separated by an unbridgeable gulf. The argument here is enigmatic:

This might not carry complete conviction. Think of it this way. If it were really the case that our conception of the world were purely 'objective', meaning that it had no essential connection to anything that was immediately subjective, then it would be an odd sort of conception — something like an impersonal description that one might find in a very complete history book. But if there are no connections to anything that is immediately subjective, then the difference between an impersonal description of the objective world and an impersonal description of a fictional world vanishes. If Nagel is right then the supposition that I am R.G. is much on a par with the supposition that I am Sherlock Holmes. But of course no two suppositions could be more distinct. Part of what makes my objective conception of the world an objective conception of the real world is the fact that it is the world that I am in — not that it is merely a world such that I can entertain the supposition that I am identical to one of its inhabitants, but is such that I have no choice but to accept such an identity.

From the middle of page 212 to the middle of page 213, Evans discusses the view that our notion of ourselves is exclusively that of a mental subject. Evans here claims that it is true that it is essential that our self notion be a notion of a mental subject — a subject of thought and judgment. But this does not impugn the notion that bodily and other physical self-ascriptions are also necessary. There is no real argument here. And in fact, there isn't a distinct developed argument for this anywhere in the chapter. But materials for such an argument are scattered throughout the chapter and the appendix.

Finally, from the middle of page 213 to the end of the section of the middle of page 215, the topic is personal identity through time. The issue is that certain kinds of illusions about the identity of a subject over time exist because ways of describing situations covertly build substantive assumptions into our conception of some states of affairs in such a way that remarkable conclusions seem to follow.

The example Evans gives is Anscombe's conclusion that self-identification has the mysterious property that misidentifications are not possible: in various of my 'I'-thoughts over time, there is no possibility that there is an inadvertent switch of referent. Anscombe takes this to be reason to be suspicious of the idea that in self-identification one is identifying anything at all. But Evans remarks that the remarkable guarantee in this case is a consequence of our description of the case, and not anything genuinely mysterious. As Evans points out, if it is correct for us to describe the situation as one in which a single subject makes a number of self-identifications and identifies himself as the same subject in all cases, then of course the subject cannot be mistaken. Of course a subject might mis-identify himself, but in those cases the description that was used to describe the situation will not apply. Cases of this sort of thing will come up in the course of the chapter.

7.2 Immunity to error through misidentification

This section has two parts. First, Evans explains why 'here' rather than 'this' is the closer parallel to 'I' as far as the kinds of mechanisms involved go. Essentially, it is because 'I' thoughts, like 'here' thoughts but unlike 'this' thoughts involve information links that can be latent, so long as the subject has the appropriate dispositions to act on any information that arises. Second, Evans discusses the phenomenon of immunity to error through misidentification. His primary concern is to argue that the phenomenon is not tied exclusively to self-ascription of mental predicates, but rather is a function of ways of gaining information. Hence, the phenomenon is explained by mechanisms he has independently motivated in the previous chapters, and is not some odd exclusively mental phenomenon.

The first part of this section, running from the beginning on page 215 through the first two-thirds of page 216, Evans compares 'I' thoughts with 'here' and 'this' thoughts. The reason is that although the Ideas in all cases constitutively involve information links, in the case of 'I'- and 'here'-thoughts, these links can be latent, provided the subject has the appropriate dispositions to act on any such information in the appropriate way should any arise. (The limits of the sufficiency of such latent dispositions is discussed in 7.6.)

The role of such information-links in 'I'-thoughts implies that such thoughts will exhibit the phenomenon of identification-freedom, in the sense Evans discussed in Chapter 6. Evans explores this in the second and larger part of this section, from the bottom third of page 216 to the end of the section at the middle of page 220. Evans discusses other treatments of the phenomenon of immunity to error through misidentification (henceforth IEM) and points out that while it is often taken to be a phenomenon concerned with self-ascription of mental predicates, it is actually tied to ways of gaining information, and hence applies to some cases of bodily self-ascription, and does not apply to some cases of mental self-ascription. For a mental and physical case in which the ascription exhibits IEM, consider 'I am seeing red' thought/said as one is looking at a red wall in good light; and 'My legs are crossed', thought/said on the basis of normal proprioiceptive information. In these cases, neither 'Someone is seeing red, but is it I who am seeing red?' nor 'Someone's legs are crossed, but is it I whose legs are crossed?' seem to make sense. For a case of each that does not exhibit IEM: a neuroscientist is looking at readouts of a number fMRI machines that are reading the neural activity of several subjects, one of whom is himself (though he does not know which image corresponds to which subject) and looking at one of the scans that shows activity in the appropriate region thinks 'Someone is seeing red, but is it I who am seeing red? (the example is imperfect in some respects); an anaesthetized subject in a room or mirrors who sees crossed legs in a mirror and says/thinks 'Someone's legs are crossed, but is it I whose legs are crossed?'. In all such cases, what makes or breaks the exhibition of IEM is whether or not the information to the effect that the predicate is instantiated does or does not require supplementation of an identity claim in order to make the self-ascription.

Evans will argue that one of the uses that the phenomenon of IEM is sometimes put to is doubly fallacious. That use is to discredit the notion that we manage to identify a mental subject at all in our self-thoughts (Wittgenstein and Anscombe have made such arguments). After quoting the Wittgenstein passage from The Blue and Brown Books, he points out an error that can be made on its basis. The phenomenon is that in certain cases of self-ascription, the information that is the basis of the ascription carries no identification information. Evans agrees that this is true. From this, it is concluded that in such ascriptions, nothing is identified. Evans argues that this is false.

One way to see Evans point here is as follows. A judgment based on information, including self-ascriptive judgments, identify an object as the object that instantiates the property in question. There are two ways this can happen. First, the deliverances of the information links (or set of cooperating information links) may provide information about the identity of the object along with information about the property being instantiated. Second, the information link might be constitutively dedicated to a certain Idea in such a way that any information incoming along this channel to the effect that some property is instantiated is information to the effect that the object corresponding to that Idea instantiates the property. If one does not recognize the second sort of possibility, then cases in which the information channel does not carry any identification information (that is, all cases providing for identification-free ascriptions) will seems to be extremely mysterious, for it will seem to be the case that they are ascribing a property to nothing.

The second error is one Evans already mentioned and revisits at page 218: people take IEM to be tied exclusively to mental self-ascriptions. When these two errors are put together, the erroneous conclusion to the effect that in our mental self-ascriptions we are not identifying anything at all is reached. Thus we are told that if 'I' refers, it refers to the subject of mental predicates, and since in such cases nothing is identified, there is no such subject, and hence 'I' does not refer. (This is essentially Anscombe's argument in 'The First Person'.)

The section closes with remarks that are essentially exegetical in character. Evans counters the suggestion that he is misunderstanding Wittgenstein's purpose in the passage,  and the use to which he puts the observations there described.

7.3 Bodily self-ascription

This section has two parts, both aimed at self-ascription of physical predicates. The first part has to do with bodily self-ascription proper — our self-ascription of predicates that concern our own bodies and their condition. The second has to do with our self-ascriptions of location and our immediate vicinity, such as 'I am in my bedroom', or 'I am facing a tree'.

The first part of this section, running from the beginning at the middle of page 220 to the top of page 222, discusses our self-ascription of bodily predicates. This would include things like 'My legs are bent', 'My hair is blowing around', 'I am being pushed'. Evans claims that in such cases it certainly appears to be the case that we have dedicated sensory channels whose deliverances we take to be of immediate, identification free relevance to ourselves.

To the suggestion that there might be deviant causal chains such that in fact the information coming in is not from the subject's own body, and thus in some sense a misidentification is made, Evans makes two counters. The first is that this fact does not entail the conclusion to the effect that in normal cases the identification is based on an identity claim. This argument parallels the argument to the same effect in section 6.6.

The second is that if the proposal that in normal cases an identity claim were involved, then it would not be clear what to make of the entities on either side of the identity operator. The proposal would be that the self-ascription that I am F would be the product of two judgments: b is F; and I am b. Evans claims that the only Idea that could be in play for the entity b is a descriptive one along the lines of 'the body from which this information derives', and Evans claims that our normal self-ascriptive thoughts have nothing like this complexity. And as far as the Idea of 'I' involved in the identity, Evans claims that if this Idea were such as to always require the kind of identity-dependent ascriptions envisioned by this proposal, then it is not clear how one could identify oneself with a physical thing at all. It is not immediately clear what this means, however. It is a reference to arguments, some of which Evans has outlined elsewhere — others of which he was familiar with due to, e.g. Strawson — to the effect that the kind of identification we have with out bodies is stronger than that provided via any sort of inference based on evidence. If this were not so, my identification of myself with my body would be the result of an inference based on a conclusion to the effect that only body x is so placed such that its causal encounters with its surroundings make sense of the pattern of sensory input I have (or on the action side: only body x is the one whose movements conform to the intentions I formulate). This sort of identification with one's body certainly seems to be quite unlike the way in which we actually identify ourselves with our bodies.

The second part runs from the top of page 222 to the middle of page 224, and investigates another kind of bodily self-ascription — the self-ascription of predicates involving one's position, orientation and comportment with respect to one's environment. For example, I self-ascribe to myself the location of being in my bedroom when I wake up and see the familiar objects. Evans notes again that this appears to provide for identification-free knowledge.

Evans claims that such an ability for identification-free ascriptions of one's location are a necessary part of the ability to conceive of an objective world at all. The argument here is a bit thin, but clearly Evans has in mind the sort of thing that he has argued for in his 'Things without the mind' as well as his teacher Strawson's work in this respect in Individuals and The Bounds of Sense.

More concretely, given that on Evans account we must have an adequate Idea of ourselves in order to entertain genuine 'I'-thoughts, and that an adequate Idea of ourselves is made available in part via a capacity on our part to identify ourselves with an entity (not in the sense of making a true identity judgment) in the objective order, the theory described at the top of page 222 is the theory a mastery of which underwrites this ability. Is it, then, what makes our Ideas of ourselves adequate.

Evans briefly runs through arguments to the effect that such knowledge cannot be identification-dependent, and these considerations exactly parallel the same arguments concerning bodily self-ascription.

In the final paragraph, Evans remarks that because the 'I'-Ideas involved in our self ascriptions have these physical and bodily aspects, the supposition that our notions of ourselves is primarily of a non-physical mental thing cannot be maintained.

7.4 Mental self-ascription

This section has five topics. First is the self-ascription of belief states. The account is roughly that one ascribes belief states to oneself not via inner perception, but by reusing the same processes of belief formation in operation in coming to beliefs about the world. The second topic is the self-ascription of experiential states. Again, the account lacks anything corresponding to inner perception. Third, Evans discusses certain illusions that arise concerning self-knowledge, and how his account can help to avoid them. Fourth, it is pointed out that the sorts of ascriptions he discusses give rise to self-knowledge that is identification-free. Finally Evans makes a few remarks about why solipsism, though untenable, gets it appeal.

The first topic, from the top of page 225 to the middle of page 226, is the self-ascription of belief states. There are two parts to this account: the resources that supply the content of the belief and the means by which one can discover that one holds the belief; and the conceptual mechanisms that make self-ascription of beliefs in general possible. The first part here is rather straight-forward: in ascribing a belief to myself, say in answering the question 'Do you believe p?' what I do is not to look inwardly at a mental substance, but rather to bring to bear those cognitive resources that would be used to answer the question 'Is p true?'. If the yield of that procedure is a 'yes, p is true', then the answer to the first question is 'yes, I believe that p'.

But forming beliefs cannot be sufficient for the self-ascription of beliefs, of course. What else is needed is not an inward glance at the belief state that has been formed, but rather the concept of belief itself. The subject must have something like an appearance-reality distinction at hand (such a distinction of course being central to any concept of belief), and with this the subject will perhaps have the wherewithal to understand that her own assessment to the effect that p is true is an instance of a belief state. The concept of belief must be the sort of thing that is in principle applicable to subjects other than herself. Evans closes this discussion with an allusion to Kant's famous pronouncement that the 'I think' must be able to accompany all my representations, but it is merely formal. By this Evans means to be giving voice to the idea that in absence of the concept of belief, a subject might form what are in fact beliefs, but would not herself be in a position to recognize that what she is doing is forming beliefs. From the subject's point of view, so to speak, it is just manifest that the world is thus and so. The notion that she is holding beliefs that might be in error is not within her cognitive horizons. In such a case, the 'I think' has no grip. We as theorists looking at the subject can say, truthfully, 'She thinks that p', but her self-ascription of that state is not possible. It is doubtful that Kant had anything like this in mind in his discussion of the purely formal 'I think', however.

The second topic is self-ascriptions of experiential states, and it runs from the middle of page 226 to the top of page 231. This account is similar to the account of self-ascription of belief states in a certain respect, in that it places at its core the reuse of capacities for representing the world. The structure of these capacities is as follows: the gaining of perceptual beliefs is, according to Evans, a two-stage process; first the perceptual systems create in the subject an informational state that carries information about the perceived environment. These informational states are seemings (this was discussed in Chapter Five), and are non-conceptual. In forming a perceptual belief, one is applying concepts, to the environment, on the basis of this non-conceptual information state. It is crucial to note that what one is not doing in this case is applying concepts to the information state itself.

Nevertheless, one can learn about this information state indirectly by doing exactly what one does when forming perceptual beliefs about the environment, and then exploiting the content of these beliefs to get a handle on the content of the associated information state. For example when I see a stick partially submerged in water a certain information state is formed. On the basis of this information state, concepts such as 'bent' have their application induced, so to speak, and I am inclined to form the perceptual judgment 'the stick is bent'. However, I might also make a judgment about my experiential state such as 'it seems as though there is a bent stick'. But again it should be noted that in neither case is the information state that to which the concept 'bent' is applied. In both cases it is applied to an (apparent) aspect of the subject's environment.

An analogy might help. Someone who wears contact lenses will make perceptual judgments on the basis of information states that are the way they are in part because of the operation of the lenses. The perceptual judgment 'That line is straight' concerns not the lenses, but some feature of the environment. Now the subject cannot perceive the lenses themselves (unless she removes them, of course, but assume she does not — if you don't like that example rephrase it in terms of the biological lenses in a subject with normal vision). But she can indirectly gain information about the lenses by simply using them. So if she wants to know if the lenses are smudged, she looks at some visual scene that she has reason to believe has only crisp contours, and if anything looks fuzzy or distorted, she can judge 'my lenses are smudged'. Similarly, my judgments to the effect that 'It seems to me as though ...' are arrived at not by applying any perceptual tools at my own mental states or at the informational state, but rather again at the world. But I use the deliverances to tell me about the informational state, exploiting the dependence between the specifics of the informational state and the perceptual judgments to which it inclines.

Evans turns now to a certain kind of infallibility that appears to be present in cases of self-ascription of experiential states. While some kinds of self-ascription are fallible, such as 'I seem to see eleven points of light' when in fact the subject miscounted, others appear not to be, such as 'I seem to see red'. This fact might seem to constitute an objection to Evans' account, and Evans responds to this objection from the middle of page 229 to the top of page 230. What Evans fails to do is clearly spell out the objection to which he is responding. The objection is this. On Evans' account, experiential judgments involve two components: an informational state and a judgment based upon (though not about) that state. Another account would have it that the information state is not distinct from the exercise of the concepts used in the judgment, but rather just is the disposition to apply the concept. If this is the case, then Evans' account, which posits a content-carrying information state independent of any judgments made on its basis, is incorrect. And the fact that there are infallible judgments such as 'I seem to see red' might appear to nail the case for the opposition, for infallibility is secured when the disposition to make a judgment employing a certain concept is identified with the information state itself. And it would also seem that on Evans' account, mistakes should be possible, exactly because the disposition to make the judgment and the information state on which the judgment is based are not identical.

Evans' response is two-pronged. First, he points out that the opposition position is implausible. Two reasons are given. First, he points out that perceptual judgments simply do not appear to be cases in which we have a disposition to make a certain kind of judgment. As Evans puts it '... it is not the case that we simply find ourselves with a yen to apply some concept — a conviction that it has application in the immediate vicinity.' The second is that we appear to be able to have experiential states whose content far outstrips our conceptual repertoire, and on the opposition account (which simply identifies the content of the experiential state with the disposition to apply a concept) this should be impossible.

Evans then points out that there is another, more plausible, explanation for this infallibility. It is that a subject can only be credited with these sorts of observational concepts to the extent that they apply those concepts to perceptual information states of that sort. A putative error would be better evidence for non-grasp of the concept; and since we are assuming that the subject in question has mastery of the concept, error is ruled out.

Evans briefly mentions experience that is not characterized by a content, such as 'I feel an itch in my foot'. His suggestion is that in such cases the informational state in question might have additional features over and above its information-carrying features. In the case of itching, it might have functional connections to actions, such as scratching. (Though Evans does not expand on this, it seems that there are similarities to this account and his account of spatial perception, in which experiences are said to be imbued with spatial content if they have the appropriate connections to behavior.)

Thirdly, Evans turns to what he calls 'certain illusions about the self'. This discussion runs from the top of page 231 to the top of page 233. The feature Evans is alluding to is the fact that one's perceptual and judgmental attention is always directed outward, even when making self-ascriptions. The illusion that this gives rise to is the belief that the self does not figure in the content of anything that one is aware of. Evans reply is that the content of self-ascriptive judgments is not limited to the deliverances of the information that is the basis of the self-ascription. Any judgment, qua thought, involves Ideas of the entities and properties judged/thought about. In a self-ascriptive judgment, one's 'I' idea is in play. This Idea is constitutively linked to the appropriate information channels in such a way that one has a disposition to employ it in self-ascriptive judgments when information comes in on these channels. Thus the content of the judgment outstrips the informational deliverances of the information link.

Fourthly, Evans points out (first full paragraph on page 233) that the kinds of self-knowledge he has discussed give rise to knowledge that is identification-free and hence immune to error through misidentification. This is closely related to the previous topic about illusions of the self. It is because the information gives rise to judgments that are identification-free that they appear to have no grip on any object — any such grip would amount to an identification of that object, and would leave open the possibility that the object so identified is not the self.

Finally (from the middle of page 233 to the top of page 235) Evans turns to the issue of solipsism. He states that the phenomenon of identification-free knowledge needs to be treated with care for it can be a pressure toward solipsism. This way of introducing the issue seems unclear, for nothing in the subsequent discussion seems in any way to be, or be an antidote to, pressure toward solipsism. Rather, it would seem that the discussion here might have been better framed in the following way: The solipsist has been challenged, e.g. by Strawson, to give an account of what he means by the 'I' that is the putative center of his world (Strawson, in Individuals, argues that a true solipsists has no self-concept). The solipsist might respond in the following way: I am the entity such that, when it is in pain, certain experiences are to be felt; I am the entity such that when it sees red and experience of the following sort is to be expected; and so forth. Evans' point here is simply that this won't work, because the solipsist needs to be able to covertly invoke the notion of the self in order to correctly delimit those experiences that the definition makes use of. The full version of the definition would be; I am the entity such that when it is in pain, certain experiences are to be expected by me. By illicitly phrasing it in impersonal terms, the solipsists makes it seem as though a workable, non-circular definition of the self is being offered.

7.5 Memory

This long section consists of three main parts. First, Evans discusses the cognitive dynamics of belief retention, aka memory. This is a process by which a belief formed at one time is retained over time. Second, Evans discusses another way that memory can operate, where what is retained is not a belief state but an informational state, and at a later time a belief state is formed on the basis of the retained information. Finally, Evans discusses at length the issue of whether the deliverances of memory are immune to error through misidentification.

The first issue is the cognitive dynamics of belief retention, and it runs from the beginning of the section at the top of page 235 to the top of page 239. The cognitive dynamics issue is similar in many respects to the topic of the first section of the Appendix to Chapter Six, where Evans discussed Dynamic Fregean thoughts. There, recall, the point what that there appears to be a temporally extended cognitive dynamics such that the same thought might persist over time in such a way that manifests a sensitivity to the subject's changing spatiotemporal position in an objective order. Given this, the beginning of this section is puzzling:

This is puzzling because if we are to believe the discussion of the first section of the Appendix to Chapter Six, and Evans' article 'Understanding demonstratives' from which that material is taken, then there is precisely such a simple rule for the case of 'now'. Evans' example there was the thought 'today is fine' and the continuation of the subject's grasp of just this thought as its verbal expression changes to 'yesterday was fine'. Thus, it would seem that if I remember being in a position, at time t, to assert 'today (now) is fine', I will be in a position, on the day after t, to assert 'yesterday (then) was fine'. The difference is perhaps better captured by noting that the cognitive dynamics of memory in relation to 'I' thoughts has the unique feature that a single persisting entity is tracked over time in such a way that the entity a that has a property F ascribed to it at an earlier time 'a is F at t₁' is the same as the entity that is having a property ascribed to it at the later time 'a is G at t₂'. The case of today and yesterday is not parallel, for in that case, though the time of the ascribing is different, the time at/during which the entity is being asserted to have the property is the same. (The expression 'property F is ascribed to a at time t' has an ambiguity that might make this point hard to grasp: it can mean either that the ascription itself took place at t (though a may have had F at some time other than t), or at can mean that the time at which a has F is t (though the event of ascribing this takes place at some time perhaps other than t).)

That distinguishes 'I' from 'now'. The difference between 'I' and 'this' is that, while a single object is tracked over time, it requires the active maintenance of the information link and demonstrative Idea of the object in order to do this. In the case of 'I', no such activity is required on the subject's part.

Evans points out that the sort of identification-freedom involved in judgments involving memory is not the same as that involved in present tense self-ascriptions. While the basic account of identification-free knowledge is the same, of course, the particular mechanisms by which it comes about are not the same. This is clearly seen by noting that a present tense self-ascription that is identification-dependent (I see a body in a mirror that I take to be mine and I say 'I am sunburnt') will yield a past-tense memory-based self-ascription that is identification-free in the sense that it is not the product of 'someone was sunburnt' and 'I am that someone'. More generally, there are two places, so to speak, where an identification might be made in a past-tense self-ascription. It might be made at the time in the past t when the remembered present tense ascription took place. This would involve an identity between ‘I’ at t and the identified body at t. Or it might involve an identity between ‘I’ at the present time of the recollection t’ and a subject at the earlier time t; or both. (The second sort is in play for a subject who suspects he is q-remembering — see below.)

Evans states that this cognitive dynamics can be regarded as another part of the functional characterization of 'I'-Ideas: they not only have certain kinds of direct connections to various information channels, but they are also characterized by a kind of cognitive dynamics that, e.g., exploits information retained in memory in an identification-free way.

The second topic runs from the bottom of page 238 to the middle of page 240, where Evans discusses another way in which memory can operate, where what is retained is information, rather than a persisting belief state. The difference here is parallel to the difference between self-ascription of belief states and self-ascription of experiential states, discussed in 7.4. There, Evans claims that during perception there was an information state induced in the subject through sensation that was subsequently conceptualized to yield a belief state, with self-ascription of experiences dependent upon the informational state, and self-ascription of belief states dependent upon the belief state. In the current case of memory, the distinction concerns two possible ways that  a memory can be achieved. Either a), at t₁ the perceptual systems produce an informational state that is, also at t₁, conceptualized to yield a belief state, and the content of this belief state is retained in memory from t₁ to t₂ at which point it is remembered; or b) at t₁ the perceptual systems produce an informational state that is retained in memory from t₁ to t₂, and at t₂ this informational state is conceptualized so as to yield a memory-based belief state. The (b) type of case is what is currently under discussion. And Evans points out that such cases are subject to the same sort of cognitive dynamics as the retained belief states, and that the retained information is not information about the experience, but information about the world.

The final topic, from the bottom third pf page 240 to the end of the section at the bottom of page 248, concerns whether or not memory of the (b) sort gives rise to knowledge that is identification-free. This discussion has a number of components. First, at the bottom of p. 240 and very top of p. 241, Evans claims that such cases appear to exhibit IEM. Next, he points out that Shoemaker has argued against this conclusion from considerations based on the possibility of quasi-memories.

Shoemaker's argument is described from the top of p. 241 to the middle of p. 242. It is this: it is possible to have memories that seem to be memories of one's own but are not (perhaps they are someone else's memories), call these q-memories; it is also possible that one might either suspect or know that one's own apparent memories are q-memories; therefore, it can make sense for one to apparently think, on the basis of a remembered informational state 'Someone watched a tree burning, but was it I?' Thus, Shoemaker concludes, this sort of memory appears not to exhibit IEM.

Evans points out that the situation here is parallel to the case of 'here' discussed in Chapter 6. From the fact that in some situations a question can be raised such that it can make sense to think 'somewhere it is noisy, but is it here that is noisy?' in such a way as to require an inference in order to establish that in fact it is here that is noisy, it is illicitly assumed that in the normal case such an inferential articulation is present. In the present case, from the fact that a subject can be aware that it is a possibility that she is q-remembering in such a way that an inference would be required to establish herself as the real subject of the memories, it is illicitly assumed that such an inferential articulation — one that would render memory identification-dependent — is in play in the normal case as well.

And just as in the case of the discussion of 'here'-thoughts, Evans not only points out the bad move, but provides independent reasons for thinking that memory cannot, in general, be identification-dependent. These supplemental reasons are articulated from the top of p. 243  to the middle of p. 246. First, the proponent of the identification-dependent memory theory must assume that the subject has a workable Idea of herself that is of use in the identity judgment 'that person is me', and Evans claims that in order to have an adequate Idea of 'me' to be of use here, one must be able to, inter alia, locate oneself in space, re-identify places, and these sorts of capacities require that memory be of use in an identification-independent way. The argument is shorter than it ought to be, but the general gist is clear: Evans is challenging this theorist in a certain way. Evans has gone into a good deal of detail as to what is involved in one being able to think 'I' thoughts, and on this account, it is not clear that the identification-dependent memory theorist has a coherent theory. At the least, that theorist should make good on the 'I'-Idea that is showing up in the identity judgment. Evans is surely right that when we are exploring these depths of capacities such as memory, perception, etc. we can no longer simply take for granted things like a capacity to think 'I' thoughts is isolation from other fundamental cognitive capacities.

He goes on to argue that the 'that person' side of the identity is also dicey. Not only would we not have an adequate Idea of any such person (Evans here appeals to his dictum that information-links are not sufficient), the judgment in question — 'someone was F' or 'that person was F' — would not constitute knowledge, for if one's memories are subject to such distortion, the inference from their deliverance of someone's having been F to someone actually having been F would be a lucky shot in the dark, if true; and simply false if false; and in neither case sufficient to support a knowledge claim.

Evans' analysis, at the top of page 246, of Parfit's example is, given what has been said, clear enough — mostly repetition of the same points.

From the middle of p. 246 to the end of the section, Evans discusses why many resist his conclusions. The text is structured in such a way as to give the impression that there are two different points here — one about Williams' remarks on imagination, and another about a fallacy concerning the nature of q-memory — but I think it is more revealing to treat them together. The fallacy is to move from the conceptual possibility of q-memory, and even of a subject's coming to believe that her apparent memories are q-memories, to the conceptual possibility of a kind of memory that presents contents to the subject that leave the identity of the subject open. Call this q*-memory. The slide is easy enough to describe: a subject who realizes that she is q-remembering has apparent memories, and also information to the effect that those memories are not her own, and thus she arrives, via inference, to conclusions such as 'someone, perhaps other then myself, was F' on the basis of those memories. From this, it is concluded that the faculty of q-memory is actually q*-memory, providing informational states whose content leaves the identity of the subject unspecified — their content being simply 'someone was F'. Once this illicit slide from q-memory to q*-memory is in place, it is all to easy to make a further slide to the effect that in the normal case, one's memories provide one with information whose content is such as to require an inference in order establish that oneself is the subject of that experience. It is here that Williams' observation is relevant. One way to read the import of Williams' point that to imagine being in the West Indies is not the same as to imagine oneself being in the West Indies is as suggesting that the difference is to be made up by some sort of surplus premise or inference to the effect that the one that one is imagining being in the West Indies is oneself. Whatever this surplus is, it is then taken up by the q*-memory theorist as independent evidence for the sort of inference required for her account of memory. Evans' reply is to the effect that Williams' observation is correct about imagination, but that we cannot draw conclusions about memory on its basis.

And Evans is surely correct to point out that this gets things exactly backwards — the normal operation of memory provides states whose content is to the effect that oneself was the subject of the experience, and it is only in unusual circumstances that one will engage in an inference that overrides that element of the content. In short, there is no such thing as q*-memory. The no-inference null hypothesis of the deliverances of memory are that the subject of the memory is also the subject of the remembered experience, and if there is an inference involved, it is one that is required to move from the prima facie self-ascription to a guarded content that leaves the identity of the subject of the experience open.

7.6 The possibility of reference failure

In this section, Evans discusses a possibility that his views on reference and 'I'-thoughts provide: that a subject's 'I' thoughts may lack a referent. Two kinds of case are discussed. The first are cases where the controlling conception of oneself is being fed information from more than one object. The second sort of case is the brain in a vat.

In a book full of novel, fascinating and counter-mainstream positions, this section is perhaps the most extreme. Having rejected the 'no ownership/'I' does not refer view, Evans now rejects its main rival, the Cartesian view to the effect that 'I' thoughts have a guaranteed referent. The first sort of case in which a subject's 'I' thought might lack a referent involves elements such as quasi-memories and information links that are configured in deviant ways. This discussion runs from the beginning of the section at the top of page 249 to the top third of page 250.

The point here is that one's 'I'-Idea is one that has as its target a single object from which the various information channels and memories derive. And just as in cases where the unknown substitution of one object for another in what is taken to be a temporally extended demonstrative thought, or unknown movement during an attempted series of 'here' thoughts, can have the result that there is no object and hence no thought, Evans here argues that a parallel lack of a single object from which the information in one's controlling conception of oneself derives (or via which one's intentions to act are manifested — the other side of the functional characterization of 'I'-Ideas) can render one's 'I'-thoughts ill-grounded, and without an object.

The second kind of case, whose discussion runs from the middle of page 250 to the end of the section at the middle of page 255, involves the brain in a vat. In the first sort of case, Evans pointed out that while in some cases one's attempted  'I'-thoughts might be ill-grounded and lack an object, one might always retreat to a more limited sort of 'I'-thought that would have an object. But in the case of the brain grown from conception in a vat, there would be no such fall-back position, and no possibility of an 'I'-thought. This can initially seem to be obviously false, for in the standard brain-in-a-vat case, which we are to imagine, the brain is connected to inputs and outputs in such a way as to exactly mimic the inputs and outputs a normal brain would have, and also such as to yield a subject of thought that is fooled into thinking that it has a body, is moving around in a vast world, etc., when in fact none of this is true. And surely it will be claimed that in such a case the subject is capable of thinking 'I' thoughts, and had in fact thought quite a number of them, assuming that the brain is around for a while and continues to a state in which the subject is in a state comparable to that of a normal adult (as was the case for most humans in The Matrix, for example).

Evans argues that in fact the subject cannot successfully think any 'I'-thoughts, for the subject has no adequate Idea of himself such as would be required fro such a thought. When appraised of her situation, the subject would have to realize that the entire universe of her experience has been a sham, that her body, including what she took to be her brain, are illusory. The one thing on which her 'I'-thoughts clung — her body qua spatiotemporally extended objective entity in the world — has been revealed to be non-existent, along with the rest of the world.

Evans discusses the possibility that the subject will attempt to think of herself as some brain somewhere. He says:

Evans is too concessive in this wording, for it makes it sound as though the subject's main challenge is not knowing where in the world this brain is, when in fact the position is far worse than this: the brain that is supposed to exist exists somewhere in a world (not the world) that the subject knows nothing of. If there is anything like an Earth, or if subjects really have anything like brains, is entirely dependent on the extent to which the brain in the vat's caretakers chose to model the virtually created world on the actual world. But if suddenly appraised of her situation, and if she thought her position through carefully, she would realize that all foundation for her thought about anything had been undercut.

The next possibility is that the subject might think of herself via description, as 'the thinker of this thought' (where the 'this thought' might either refer to this thought itself, or to some 'demonstratively identified' thought). To this Evans points out that in the case at hand, the subject has no adequate Idea of 'this thought' to use to anchor the attempted description. Normally, an adequate Idea of a thought would individuate it with respect to the subject entertaining it. (Note: we must assume that in this example thoughts are being individuated by something other than their contents, for if they are being so individuated, then of course a description such as 'the thinker of this thought' is not guaranteed to have a unique referent, as indefinitely many subjects may be entertaining exactly that content. The example is easier with cases such as 'the experiencer of this pain'. And here it is the subject's capacity to have any inkling of the numerical identity of the pain that Evans is questioning, since normally numerical identity of pains is established via reference to the subjects that experience them (unlike qualitative identity, of course), but in this case it is the subject's identity that is in doubt.)

Evans goes on to claim that in other kinds of case, such as paralyzed subjects, or subjects that were normal up to a point and then had their brains removed, residual dispositions might be sufficient to support 'I'-thoughts. Evans closes with some remarks on the contingency of the fact that in the normal case our control center (brain) is at the same location where we are, the center of our egocentric space, here, and ways in which this contingent fact influences intuitions on our 'I'- and 'here'-thoughts.

7.7 Conclusions

This section is an attempt to wrap up the points of this chapter in terms of four bullet points. The first, concerning aspects of our self-identification adds substantively to the material of the chapter. The other three points are basically quick restatements of some of the more important consequences of the view.

This concluding section is broken down into four numbered points. The first is by far the most substantive, and adds significantly to the prior discussion. The gist is that our 'I'-Ideas constitutively involve functional connections to sensory input and action output in such a way as to render any attempted reduction of 'I' to some description unworkable, because no description can guarantee the correct functional connections. I addition to these functional considerations, a subject's 'I'-Idea, in order to be adequate, must be such that the subject has the capacity to identify herself as an entity in an objectively conceived world.

The second point is a recap of the discussion in Section 7.1 concerning the view, which Evans opposes, to the effect that our self-identification involves a sort of mysterious subjectivity that is impossible to reconcile with the objective order. He points out how his account, because it accounts for the phenomena in an entirely naturalistic way, lets us move past such views.

The third point concerns the fact that, on Evans' account, our 'I'-Ideas span the physical and mental. They are based on ways of gaining knowledge that is identification free, and such that both bodily and mental ascriptions depend on these ways of gaining knowledge.

The fourth point concerns the dependence that our 'I'-Ideas have on certain empirical presuppositions (such as a continuing bodily referent). Because of this, the notion of the mind as a independent entity, cognitively self-sufficient in absence of a body, is an illusion.

7.A Appendix

Appendix Part 1

In this section Evans introduces and defends what he calls the 'self-reference principle'. This principle is meant to explain an important aspect of our self-conception and self-knowledge in such a way as to account for certain things that have led philosophers to privilege our thinking over our bodily self-conceptions, but in a way that avoids a number of implausible views, both Cartesian and 'no referent' views.

Evans begins by remarking on Anscombe's seminal paper 'The First Person'. As Evans reconstructs her argument, she claims that the import of the first person pronoun 'I' is not simply a device that the subject uses to refer to himself; it is this of course, but not all cases of referring to oneself are such that the first person pronoun is appropriate. On the other hand, trying to spell out what this special way of self-reference is without use of some pronoun that already has the import of the first person-pronoun is not possible. The conclusion (which involves considerations over and above these in Anscombe's paper) is that 'I' is not a referring expression — 'I' does not refer at all.

To this Evans introduces what he calls the 'self-reference principle' in order to capture what is special about self-referential thought in such a way that does not exploit something that already has the import of the first-person pronoun. Evans' idea is that the subject can intend to bring it about that he satisfy the concept-expression 'xi refers to xi'. Evans admits that it can seem as though formulating such an intention requires exactly the sort of pre-existing self-referential capacity that we are trying to explain, in that the subject cannot be intending to satisfy that concept expression is the way in which Oedipus satisfied it when he thought 'The slayer of Laius is F.' With such a thought, Oedipus was satisfying the concept expression 'xi refers to xi', but not in the way required. But Evans claims that the explanatory direction goes in the other direction: that it is the capacity to grasp and use the self-reference principle that (partially) explains the special character of our self-referential thought.

Evans then defends the principle in a way that makes it appear not only to be a device capable of critiquing Anscombe's argument, but also a necessary supplement to Evans' own position as articulated in the main part of the chapter. He points out that if the self-reference principle is not respected, then having the correct functional connections between one's putative 'I'-Idea and sensory input and behavioral output does not guarantee that one is self-conscious. I think that Evans' point here is something like this: the mere existence of the correct dispositions is not enough for self-conscious thought, because those connections might be there even for a subject who lacks the capacity for self-conscious thought (though of course such connections in at least dispositional form are necessary for a subject to have an adequate 'I'-Idea).

Evans, in the next paragraph (the bottom of p. 259) shows that what else is needed above the functional connections just discussed is a capacity for psychological self-ascription. And this can be seen as an additional layer of functional-dispositional aspects of one's 'I'-Idea. And the implication here is that with this layer of dispositions in the mix, the self-reference principle is satisfied, and the subject is capable of self-conscious thought. Evans adds that it is sufficient for respecting the self-reference principle that the appropriate functional links mentioned above, as well as those involving memory, etc., are in place. As he says:

The subject need not have the 'vertiginous' thought that 'I am the subject of this very thought', or anything of the sort.

Suffice it to say that a defender of Anscombe will not give in at this point, but exploring the thickets of this issue in the context of Anscombe's position is no trivial undertaking, and is beyond the scope of this Guide, so I will move on.

Evans closes with two remarks on the self-reference principle. The first is that the fact that satisfaction of the self-reference principle is necessary for self-conscious thought (thought about oneself) gives insight into why the notion that the self is fundamentally mental has been so pervasive. The explanation is that in one's inquiring into one's self, one is asking about the thing that is doing the inquiring, and since such inquiries are done via the medium of thought, the subject of the inquiring is, by the self-reference principle, a thinker. But Evans points out that, while it is true that self-conscious subjects are necessarily thinkers, the result that they are only thinkers is an accidental feature of the way such inquiry unfolds. Hence his example to the effect that if the inquiry took place in a purely verbal medium, where the self-reference at work was linguistic self-reference, we would get the result that the self is necessarily a speaker.

 

Appendix Part 2

In this section, Evans discusses aspects of the demonstrative Ideas he has discussed. In particular, Evans discusses the relation between the 'functional' aspect to such Ideas and a 'background' component. The upshot is that  a full account of these Ideas requires attention to both aspects, and the two aspect are not fully separable.

Evans first describes what he calls the functional component of the Idea types he has discussed — demonstrative, 'I'-Ideas, 'here'-Ideas, etc. The functional component has to do with the special, identification-independent links between these Idea types and relevant avenues of informational intake and behavioral expression. He next points out that this component by itself is not sufficient for thought about the implicated object or place, for the functional component by itself does not guarantee satisfaction of Russell's Principle: it does not allow the subject to know which object in the objective order the thought is meant to concern. For that, additional resources are needed. These points are not new. Evans has been making use of them since Chapter 6. There, recall, much effort went into describing how the information link with the demonstratively identified object did two things: it played a role in the functional aspect of the demonstrative idea type, providing for the identification-free status of at least some of the subject's thoughts of the object; and it also allowed the subject to locate the object in objective space, thus allowing the subject to know which object her thought concerned.

The second paragraph is pitched at an abstract level. Roughly, the idea is this: if we take the component of an Idea type that allows for satisfaction of Russell's Principle (what Evans calls the background component) to be glossed as 'conceptual', then the point is that we cannot analyze away the functional components of such Ideas in terms of the conceptual components involved. On this rough reconstrual, the point is one on which Evans is in agreement with Perry ('The essential indexical') and Lewis ('Attitudes de dicto and de se') as well as others. Less roughly, Evans' point is more general, in that it is not limited to what would normally be regarded as concepts, so that 'conceptual' would be roughly identical to 'can be expressed by a definite description'. Evans' point is meant to cover whatever is involved on the satisfaction of Russell's Principle. All of this might also reasonably be described as 'conceptual' to the extent that 'conceptual' and 'objective' are taken to be on a par.

The last paragraphs state that these two components are closely connected. Again, this is not new. The account of demonstrative Ideas in Chapter 6 was a case of exactly the sort Evans is abstractly describing here. The information link, whose primary application is in partly constituting the functional component of the Idea of the object, is also that which allows the subject to locate the object in egocentric space, and hence objective space, by which means the subject knows which object is in question.

 

Appendix Part 3

In this section, Evans raises a question for his account, but does not answer it. The question concerns what was called in Part 2 of the Appendix the 'background' aspect of our Ideas. The specific question is about what is involved in our Ideas being objective.

Evans begins by pointing out that while his particular apparatus of the fundamental level of thought may seem unnecessary, the job that it is intended to do is entirely necessary. The job it is intended to do is to explain what it is in virtue of which a subject's thought concerns an objective world. Evans states that since our thought clearly is objective, any adequate account of thought will have to have something playing this role, even if it is not exactly of the form of Evans' proposal. (The photograph model lacks any such apparatus, and throughout the book this failure has been what Evans has harped on, in one way or another.)

Evans then points out that how this is supposed to work, even on his own account, is not obvious. The problem is, roughly, that it is not at all clear how a subject can bootstrap herself up to fully objective conception of anything. The example Evans gives here is his account of objective space in 6.3,  according to which one's cognitive map is given by a representation of the world anchored to recognizable landmarks that holistically constitute the frame of reference. But these landmarks are themselves not independently objectively conceived.

This is a question that Evans does not have an answer to, but it is one that he has asked before. In 4.3 he says of the generality constraint that it is perhaps an ideal to which our actual mode of thought only approximates; in Part 2 of the Appendix just above, he intimates that the 'objective' aspect of an Idea is imperfectly and idiosyncratically built upon the 'functional' component.

 

Appendix Part 4

This very brief section, in which McDowell puts Evans in the third person, makes some remarks on some central ways in which Evans would likely have expanded his discussion of our 'I'-ideas had he been able. Specifically, there would have been more attention paid to the action component. In footnote 4 of this chapter Evans remarks that he is focusing on the informational input side of the functional characterization of 'I'-Ideas in part because it has received less attention. This section is too short and, in its own way, too straight-forward to need explanation. [Note: McDowell has an extremely helpful and fascinating discussion of the import of the action element of our self-conception, in a way that might very well have been at least roughly along the lines Evans would have pursued, in his 'Referring to oneself' (In Hahn, Ed. The Philosophy of P.F. Strawson).]

 

[End of Guide to Chapter Seven]

Copyright  © Rick Grush 2003

Copyright holder grants permission to distribute, quote, and reproduce for any non-commercial purposes provided authorship and copyright are clearly indicated.