Oxford University and NYU
What is involved in the consciousness of a conscious, "occurrent" propositional attitude, such as a thought, a sudden conjecture or a conscious decision? And what is the relation of such consciousness to attention? I hope the intrinsic interest of these questions provides sufficient motivation to allow me to start by addressing them. We will not have a full understanding either of consciousness in general, nor of attention in general, until we have answers to these questions. I think there are constitutive features of these states which can be identified by broadly philosophical investigation, and in the early part of this paper I will try to do some of that identification.
Beyond the intrinsic interest of the topic, the nature of such conscious attitudes is highly pertinent to a philosophical account of psychological self-knowledge. So I will also say something about the significance of the constitutive features of these conscious attitudes for a philosophical account of how it can be that a thinker has a distinctive kind of knowledge of some of his mental states. The general challenge in this area is to find anything intermediate between the unexceptionable but uninformative, on the one hand, and the absolutely unbelievable on the other.
Perceptual experiences and sensations, on the one hand, and so-called "occurrent" conscious propositional attitudes, on the other, differ in many respects. But there is one property they share. They both contribute to what, subjectively, it is like for the person who enjoys them. A person may try to recall who was Prime Minister of Czechoslovakia when the Soviet Union invaded. It then occurs to this person that Dubcek was the Prime Minister. Its so occurring to him contributes to the specification of what it's like for the person then. It would be subjectively different for the person if it occurred to him (falsely) that it was Husak; and subjectively different again if nothing comes to mind about who was Prime Minister. This example is a case of memory, but the point is in no way restricted to memory. The same is true of ordinary thinking in general, as when it suddenly strikes you that you have left the kitchen tap (faucet) running. In that case, your thought is not a conclusion inferred by you from other premisses; but reasoned conclusions may be conscious in this familiar sense too. When, on an appointing committee, you conclude in thought "On balance, Smith would be the best person", your so concluding can be a partial specification of what it's like for you then. On the general point that occurrent conscious propositional attitudes are often subjective states, I am in agreement with such writers as Flanagan (1992, p.214) and Goldman (1993, section 8). It is important to note, though, that acceptance of this point does not require any internalist theory of conceptual content. I would want to defend the view that the intentional contents which, in each of these examples, contribute to the specification of what it's like for the thinker are composed of concepts which are in part externally individuated.2 I will return to this issue.
When a thought occurs to you, or you make a conscious judgement, your attention is engaged. Your attention will often be shifted from whatever external events which may have been the object of your attention at the time. The engagement of your attention in conscious thought is by no means confined to those moments at which thoughts occur to you or you make a judgement. Any one of the following can occupy your attention:
Ryle, whose late writings on thinking do not contain the slightest hint of behaviourism, once considered an example involving a blindfolded chess-player. Ryle noted that "...when, after struggling to remember the positions of the pieces, the chess-player does remember, then his seeing them in his mind's eye, if he does do this, is not something by means of which he gets himself to remember. It is the goal, not a vehicle, of his struggle to remember" (1971, p.398). Ryle is emphasizing that the occurrence of a memory image cannot be identified with, nor taken as the vehicle of, the thinking that led to the image. My present point is that this thinking, what Ryle calls "the struggle to remember", is itself occupying the player's attention prior to any success he may have in that project.
If we are to describe correctly the relation between conscious thought and attention, we must respect the distinction between the object of attention and what is occupying attention. In a normal case of perceptual attention to some physical object, feature or event, there is something to which the subject is attending. The object of attention is perceived, it causally affects the subject. No doubt we may want to say that there is, or can be, some sort of object of attention in a pure case of perceptual hallucination. But those cases are plausibly understood as parasitic on the central case of genuine perception.In the cases of pure hallucination, it is for the subject as if there were a genuine object of attention.
In conscious thought, by contrast, there is no object of attention (nor is it as if there is). The notion of an object of attention which is inapplicable in conscious thought is that of an experienced object, event or state of affairs. In mental states other than those of conscious thought, a genuine object of one's attention might be a material object; or a continuing event; or the continuing or changing features of an object or event; or an object's changing relations to other objects or events. Having a sensation is also an experience. A pain, for instance, can equally be an object of attention. But thinking is not experiencing. There are objects of thought, but an object of thought is not thereby an experienced object, and is not an object of attention in the sense in question.
All the same, in conscious thought, your attention is still occupied - as it is also occupied in the perceptual cases, and in cases of imagination. It would be a crude nonsequitur to move from the true point that there is no object of attention in conscious thought to the false conclusion that conscious thought does not involve attention. I will attempt some further analysis of the notion of the occupation of attention a few paragraphs hence.
It has to be said that those who have recognized the involvement of attention in conscious thought have not always been helpful to its best elaboration. It is a great virtue of William James's justly famous discussion of thought and attention that he recognises a general category of what he calls "intellectual attention". But the effect is somewhat spoiled by his distinguishing the intellectual variety of attention by its alleged distinctive objects, which he says are "Ideal or represented objects".3 It emerges from his other discussions that by "Ideal object" he means a certain kind of concept. But genuine objects of attention are, in the central cases, experienced objects. Correlatively, in central cases genuine objects of attention are also such that their continuing and changing properties at the time of the state involving attention causally contribute to the way that object is given in the conscious state. Neither of the objects James proposes as objects of intellectual attention, concepts of a certain kind or the objects thought about (the "represented objects"), need stand in that sort of relation to conscious thought. I suspect that it is not an accident that those, like James, who have supposed that there are objects of intellectual attention have also been drawn to perceptual models of knowledge of one's own mental states. It was James who wrote, in making the transition from the volume of the Principles of Psychology containing his famous chapters on The Stream of Thought, Attention and Memory, to the volume dealing with perception, "After inner perception, outer perception!" (1983, p.651 - the first sentence of volume II).
To believe in a single general kind of attention which is occupied in both in cases of perceptual attention and conscious thought is not to be committed to a perceptual model of thought, nor to a perceptual model of our knowledge of it. In addition to the applicability to both perceptual attention and conscious thought of a pre-theoretical ordinary notion of there being something it's like for their subject, there are also some explanatory consequences of the hypothesis that there is a single general kind of attention of which perceptual and sensational attention, and conscious thought and imagination are all subspecies. It is a familiar truth about attention that any one of these kinds of attention can interrupt any one of the others. Perceptual attention can be interrupted by conscious thought; conscious thought can be interrupted by external events which capture the thinker's attention; either of these two subspecies of the occupation of attention can be interrupted by imagination; and so on. What we have here is not merely some family resemblance between varieties of conscious states, but apparently some form of competition for the exclusive use of a limited single faculty of attention. The familiar facts about attention are explained if there is a single, suitably high-level, resource, drawn upon either by perception, conscious thought or imagination, a resource with access to some of its own recent states and to memory representations generated by its own previous states. I do not say that it is absolutely impossible to explain any one of these facts in some other way. I conjecture, though, that other explanations will be ad hoc, and unable to explain the full range of familiar facts. For instance, one might try to explain the facts about the interruption of the occupation of attention in thought by perceived events as follows. One might say that there are two radically different kinds of attention, drawing on quite different resources, but that there is some separate explanation altogether of why a normal thinker is subject to severe limitations in his ability to be in both kinds of state in parallel. There is, though, a real danger that any such theory will need to postulate some additional system which which favours now one of the two alleged attentional systems, now the other, but cannot favour both. This additional system would appear to have just the features of the limited high-level resource to which this account was proposed as an alternative.
I now attempt some further analysis of the occupation of attention by conscious thought. When you have a thought, it does not normally come neat, unconnected with other thoughts and contents. Rather, in having a particular thought, you often appreciate certain of its relations to other thoughts and contents. You have a thought, and you may be aware that its content is a consequence, perhaps gratifying, perhaps alarming, of another conclusion you have just reached; or you may be aware that its content is evidence for some hypothesis that you have formulated; or indeed that it is a counterexample to the hypothesis. Now when you think a particular thought, there is of course no intention in advance to think that particular thought. But there can be an intention to think a thought which stands in a certain relation to other thoughts or contents. It is thought carried out in accordance with such an intention that is directed, as opposed to idle, thought.
The relation that one intends one's subsequent thoughts to bear to one's earlier thoughts may be that of logical consequence; or that of being evidence for the earlier thought; or that of being an amusing observation about some event, or the like, if one is engaged in writing an after-dinner speech. As always, there is a distinction between the intended and the actual relation a later thought bears to earlier thoughts. On a particular occasion, there may in fact be no intended relation, but an actual thought appreciated as bearing certain relations to others may nonetheless occur to the thinker. Or there may be an intended relation, but one may be distracted from one's goal in thought by the occurrence of a thought bearing a relation different from the intended relation; or, as in the case of the uninspired drafter of the after-dinner speech, nothing bearing the intended relation may come to mind.
It is worth reflecting further on the striking fact about attention and consciousness that your attention can be occupied by your trying to do something in thought. Your state is subjectively different in the case in which you are trying, in thought, to achieve a particular kind of result from that in which you are casually drifting in thought. This can be so even if the same sequence of thoughts occur to you in the each of the two cases. Though striking, the point seems to be a special case of a more general phenomenon. In general, a subject's trying to do something (and what it is he is trying to do) contributes to what it is like for the subject. It does so in a way which goes beyond any occupation of attention by external events, sensations or thoughts. The phenomenon can be illustrated, beyond the realm of conscious thought, in the first instance by some perceptual cases.
Consider an example in which doing something is occupying your attention. Your attention might be occupied by driving down a narrow street without scraping the cars parked on each side; or by getting the cursor from the top left hand corner to the bottom right hand corner of your computer's screen; or by getting someone with whom your are in conversation to decide to take a certain course of action without pressurizing them. Your attention's being occupied in such actions cannot be identified with your attending to the events in the external world which they involve. In the example of driving down the narrow street, you could attend to exactly the same external movements and objects without being the driver at all. Similarly, in the example of moving the cursor, your pattern of attention to motions and symbols on the screen could be exactly the same as when someone else is operating the mouse which controls the cursor. Nor can the action's occupying your attention be identified with attention to some further external events or perceptual states. In the driving example, it does not consist in your attending to the movements of the steering wheel, or to sensations of pressure on the wheels and pedals. The experienced driver will not be attending to such things when his attention is occupied with the action of driving down the narrow street. Nor, again, is the object of your attention is any event of trying (whatever that might mean).
So it seems that the occupation of your attention by your doing something always goes beyond mere perceptual attention to particular events or objects. It follows that there could not be events or objects your mere perceptual attention to which constitutes the occupation of your attention by your doing something. There could not be so, because trying, in thought or action, to achieve a certain goal is a subspecies of consciousness in its own right, and, when present, is additional to perceptual attention and the occupation of attention by occurrent thoughts. To make this point is not to deny that attention is a perceptual phenomenon in at least one important respect. Attention is a perceptual phenomenon at least in the sense that a full specification of what it is like for a perceiver must include a statement of which of the perceived objects, events, properties or relations he is attending to in having that experience. The present point is just that a specification of whether the subject's attention is occupied in trying to do something must also be included in an account of what it is like for him.
Corresponding to this relation between attention and action are certain divisions between imaginative possibilities. You can imagine seeing your hands and arms making certain movements in front of you, from your standpoint in the imagined world as the owner of the hands and arms. That is one thing, but it is another visually to imagine moving your hands and arms to make those motions. In characterizing imaginings, we can distinguish between what is suppositionally imagined to hold in the imagined world and what it is imagined to be like for the subject in the imagined world. When you imagine seeing a suitcase with a cat wholly obscured behind it, you suppositionally imagine (S-imagine) that there is a cat behind the suitcase. Suppositionally imagining that there is a cat behind the suitcase is to be distinguished from imagining what it would be like for someone in the imagined world, i.e. imagining from the inside the subjective state of the person seeing the suitcase in the imagined world.4 Now the distinctive contribution made to imagination by imagining doing something seems to fall on the side of what it is like subjectively, for the person in the imagined world, rather than merely on the side of what is suppositionally imagined. This is just what one would expect in advance, given the general principle that to non-suppositionally imagine something is to imagine being in a subjective state of a certain kind, and given that doing something contributes to the specification of a subjective state. The point that imagining doing something falls on the non-suppositional side also receives some confirmation when one reflects on the different requirements for suppositionally-imagining that one is doing something, as opposed to imagining (not merely suppositionally) doing it. To imagine, non-suppositionally, doing something requires that one know how do it. For this kind of imagining, only someone who knows how to play the Appassionata sonata can imagine, from the inside, playing it. This last species of imagining is distinct from imagining moving one's hands on the keyboard in any old fashion and hearing the sounds of the Appassionata come out. Merely to suppositionally imagine that one is playing that sonata does not require any knowledge of how to play it.
There is disagreement within psychology about the subpersonal mechanisms underlying perceptual attention at the personal level. There would, though, be less disagreement on the proposition that perceptual attention serves a function of selection. It selects particular objects, events, or particular properties and relations of objects and events, in such a way as to improve the perceiver's informational state concerning the selected items. The details of the nature of the improvement are a matter for empirical investigation. The improvement might be a matter of more detailed, and new, kinds of informational content; or it might be a matter of the speed with which states of given informational content are attained. Whether this capacity for improved informational states for selected items is used effectively or wisely is another matter. Attention is a resource which may be drawn upon whatever the subject's purposes.
If what I have said about the occupation of attention by conscious thought is along the right lines, then the occupation of attention at least in directed thought also performs a function of selection. One can expect that the parallel cannot be precise, just because of the difference noted between the presence of objects of attention in the perceptual and sensational cases, and their absence in the case of conscious thought. Nonetheless, when a thinker is engaged in directed thinking, he is in effect selecting a certain kind of path through the space of possible thoughts - thought contents - available to him. There is not selection for particular thoughts, of course: that would involve the rejected view that there are intentions to think certain particular thoughts. But there is selection of a certain kind of thought, given by the content of the thinker's aim in thought. Without such selection, human thought would be chaotic, at the mercy of associational connections not necessarily at all pertinent to the thinker's current goals. And as in the perceptual case, this capacity may be used wisely and effectively, or not.
Our ordinary, everyday notion of a conscious attitude does not apply only to occurrent attitudes. Each one of us has myriad conscious beliefs, intentions, desires, hopes, fears, suspicions, and the rest, and obviously these cannot all be contributing to what it's like for each one of us at any given time. Often when we talk of a conscious attitude we are concerned with an underlying state, capable of producing manifestations in conscious occurrences which do contribute to a specification of what it's like for the subject. In the case of a conscious belief that p, some of these manifestations in consciousness - judgements and propositional impressions that p - may be prompted by an explicit question of whether or not it's the case that p. Much else may also trigger such a conscious state. It may be triggered by reflection on related subject-matters, on other things learned in the same circumstances, or indeed on virtually anything that associative memory may link to the belief that p. In the case of belief, we will also want to distinguish what is already believed, from beliefs formed when the issue arises. Length of time to conscious retrieval, and difficulty of retrieval or formation, which will itself be relative to the thinker's current intellectual and perceptual context, make these partially dispositional notions of a conscious but non-occurrent propositional attitude essentially a matter of degree along many dimensions.
Conscious thoughts and occurrent attitudes can, like other conscious mental events, give the thinker reasons for action and judgement. They do so also in the special case in which they give the thinker a reason for self-ascribing an attitude to the content which occurs to the thinker, provided our thinker is conceptually-equipped to make the self-ascription. On the position I am developing, we can, for instance, take at face value the statement that someone's reason for self-ascribing the belief that Dubcek was Prime Minister of Czechoslovakia when the Soviet Union invaded is his just then judging that Dubcek was Prime Minister at the time of the invasion. To spell it out in more detail, we can distinguish three stages a thinker may pass through when asked "Whom do you believe was Prime Minister there when the Soviet Union invaded?". First, after reflection, he may have
(1) an apparent propositional memory that Dubcek was Prime Minister then.
Since he is, we may suppose, taking memory in these circumstances, and for this sort of subject-matter, at face value, he moves to endorse the content of the apparent memory, and makes
(2) a judgement that Dubcek was Prime Minister then.
This judgement makes it rational for him to make
(3) a self-ascription of the belief that Dubcek was Prime Minister then.
To say that (2) is the thinker's reason for making the judgement in (3) is not to say that he infers the self-ascription from a premiss that he has made such a first-order judgement. A mental event can be a thinker's reason for doing something (including the special case in which what is done is making a judgement), without the case being one of inference. An experience of pain can be a thinker's reason for judging that he is in pain. To try to construe this as an case of judgement reached by inference would make it impossible to give an epistemology of the self-ascription of sensations. (Am I supposed to rationally reach the conclusion that I am in pain from the premiss that I am in pain?) The pain case shows too that the model need not be that of perception, either. The conscious pain itself, and not some alleged perception of it, is reason-giving.
Let us call an ascription of an attitude with a certain content, by a subject to himself, made for the reason that he has an occurrent conscious attitude of a certain kind, with that same content, a consciously-based self-ascription. In the example as imagined, the self-ascription of the belief that Dubcek was Prime Minister when the Soviet Union invaded is a consciously-based self-ascription.5 This characterization of a consciously-based self-ascription also includes examples in which the self-ascription is made on the basis of the conscious occurrence of a mental event of the very kind ascribed, as when you judge "It has just occurred to me that p" because indeed it has just occurred to you that p. Reaching a self-ascription of a belief by basing it upon a conscious state is of course only one of several means, each of them special to self-ascription, by which a thinker may knowledgeably come to self-ascribe a belief. It is a very important point that some knowledgeable self-ascriptions are not based on any intermediate conscious state at all.6 I return to those cases and their significance in §5 below. My task in this section and the next is just to try to understand the consciously-based self-ascriptions better.
The description of a self-ascription made on a particular occasion as consciously-based should not be regarded as in competition with the description of it as reached by Evans's procedure (1982). In employing Evans's procedure, "I get myself in a position to answer the question whether I believe that p by putting into operation whatever procedure I have for answering the question whether p" (1982, p.225). Searching your memory to see if you have any information about who was Prime Minister when the Soviet Union invaded is precisely one of the methods you have for answering the first-order question of who was Prime Minister then. Coming to self-ascribe a belief on the basis of the deliverances of stored information is a special case of use of Evans's procedure, rather than any kind of rival to it.
The idea of consciously-based self-ascriptions is sometimes regarded with great suspicion. In fact, in respect of the rational sensitivities required for consciously-based self-ascription to proceed properly, these ascriptions are importantly similar to other, very different cases. Consider for a moment beliefs which are reached by inference. When a belief is reached by inference from certain premisses, the contents of some of the thinker's states are taken by the thinker to support the inferred conclusion, and they do so in the case of valid inference. Now the thinker who successfully reaches new beliefs by inference has to be sensitive not only to the contents of his initial beliefs. He has also to be sensitive to the fact that his initial states are beliefs. He will not be forming beliefs by inference from the contents of his desires, hopes or daydreams.
Another pertinent case is that of beliefs reached by endorsing the content of one of the thinker's perceptual experiences. Here too the thinker makes a transition - and this time not an inferential transition - from one state with a certain content, to a belief with an overlapping, or an appropriately related, content (depending on your views about the nature of perceptual content). Again, the sensitivity does not involve merely some grasp of relations of content between the two states involved in the transition. The thinker is also sensitive to which kind of initial state it is that has the content. He will not be prepared to take the content of imaginings, for instance, at face value in the same way.
In cases of consciously-based self-ascription of attitudes and experiences, a thinker similarly makes a transition not only from the content of some initial state, but also makes it because the initial state is of a certain kind. There is, though, also a difference from the cases of inference and perception mentioned in the last paragraph. In the case of consciously-based self-ascription, the distinction between those events which are occurrent attitudes of the right kind to sustain the resulting judgement and those which are not is a distinction which is (partially at least) conceptualized by the thinker. The self-ascriber thinks of his state as belief, or as experience of a certain kind, or whatever it may be. He also thinks of himself as the state's subject. Possession of these important conceptual capacities does, of course, go far beyond the ability to make judgements rationally in response to one's own conscious states.
Taking shortcuts in reaching knowledgeable judgements, without making explicit in conscious thought all the intermediate steps, a ubiquitous phenomenon in human thought. Judgements in which a thinker self-ascribes a particular attitude with a particular content are no exception. For instance, many a thinker, when asked our earlier question about the Prime Minister of Czechoslovakia, will move straight from the apparent memory (1), in a context in which the deliverances of memory are taken at face value, to the self-ascription in (3). Such shortcuts are permissible, and can still result in knowledgeable judgements, provided that they are taken only in circumstances in which the thinker could take the longer route, with each transition in the longer route made for the right sort of reasons. This is not to make a concession to a purely reliabilist theory of knowledge. A pure reliabilism would not require the possibility of taking the longer route with full rationalization at each step.
It is worth noting a certain phenomenon when a shortcut is taken. We can continue with the example of the shortcut in the Dubcek example. For the thinker who takes this shortcut, the conscious apparent memory (1) is certainly causally and rationally influential in producing his self-ascription of the belief that Dubcek was Prime Minister at the time. However, even in a context in which memory is taken at face value, an apparent propositional memory that Dubcek was Prime Minister then is not a reason for self-ascribing the belief the Dubcek was Prime Minister then. This may be puzzling: what is going on here?
What we have here can be called "a failure of pseudo-transitivity". The apparent memory gives a reason for judging that Dubcek was Prime Minister then, the occurrence of such a judgement gives a reason for self-ascribing the belief that he was; but the apparent memory does not give a reason for the self-ascription. More generally: from
(4) A conscious event of f- ing gives a thinker reason to J that p
taken together with
(5) A conscious event of J-ing that p gives a thinker a reason to H that p
it does not follow that
(6) A conscious event of f- ing gives a thinker a reason to H that p.
The Dubcek example is an illustration of (6) not following from (4) and (5).
I call this only a failure of "pseudo-transitivity", because what is rationalized in (4) is not literally the same as what does the rationalizing in (5). What is rationalized in (4) is the action-type of J-ing that p. What does the rationalizing in (5) is the occurrence of an actual event of J-ing that p. This indeed is the key to why pseuo-transitivity should fail. There must actually be an event of J-ing that p for the thinker to be given a reason to H that p. (4) does not itself ensure that there is such an event, even if there is a conscious event of f -ing. For the thinker may not act on the reason which is given by f -ing. We should, indeed, positively expect pseudo-transitivity to fail when what gives the reason is the occurrence of a conscious event of a certain kind.
By contrast, when the rationalizing is done just by a relation between contents, transitivity must hold when it is conclusive reasons which are in question. From the premisses that
(7) p gives a conclusive reason for accepting q
(8) q gives a conclusive reason for accepting r
it does follow that
(9) p gives a conclusive reason for accepting r.
The conclusiveness of the reason stated in (8) does not require anyone to be in any particular state of consciously accepting that q. The relation of conclusive support holds between the contents themselves, independently of any psychological relations any particular thinker bears to them. What then would be a general characterization of the cases in which pseudo-transitivity fails? It fails in instances where the content of the rationalized judgement (or other attitude) mentioned in the second premiss is fixed at least in part by the nature of the psychological state which does the rationalizing, and not just by the content of that state. In a case of pure logical inference, it would be fixed just by the content of that state, and transitivity would not fail.
Am I cheating by trading on the fact that I have chosen a relation of conclusive support to illustrate cases in which transitivity does hold? No cheating is occurring here. This is shown by the fact that even in cases where the reasons are conclusive, as long as the rationalizing of the content of the final judgement (or other attitude) is done by the nature of the rationalizing psychological state, and not just its content, pseuo-transitivity still fails. Thus we have both
(10) An experience of pain gives a thinker a reason to judge that he is in pain
(11) A judgement that he is in pain gives a thinker reason to self-ascribe the belief that he is in pain.
It still does not follow, however, that
(12) An experience of pain gives a thinker reason to self-ascribe the belief that he is in pain.
The position on consciously-based self-ascription which I have so far sketched occupies an intermediate stance between two more extreme positions, elements of which can be found in some recent discussions. One of these two more extreme positions can be called the "no-reasons" account of self-ascription of attitudes. At the other extreme is a form of internalist introspection. Those how have leant towards these positions may not have intended them to apply to consciously-based self-ascriptions. I think it will provide us with a better understanding of such self-ascriptions if we consider how neither extreme can be applicable to them.
The spirit of the "no-reasons" account can be introduced by quoting a paragraph from Shoemaker. Shoemaker is discussing the self-ascription of belief, and he writes (1993, pp.78-9):
Compare another sort of case in which mental states "automatically" give rise to other mental states - that in which a set of beliefs, B, give rise to a further belief, C, whose content is an obvious consequence (deductive or inductive) of their contents. No doubt there is a microstory (as yet unknown) about how this takes place. But one sort of microstory seems out of the question. It would be wrong-headed to suppose that having identified the underlying mechanisms or structures in which the possession of the various beliefs (and the various concepts they involve) is implemented, one must postulate additional mechanisms, completely independent of these, to explain how it is that B gives rise to C. Given a neural or other subpersonal mechanism, nothing could justify regarding that mechanism as an implementation of a given belief if the nature of the mechanism is not such that the microstory of its existence and operation involves an implementation of the inferential role of that belief (and so involves relations to a larger system). It is equally wrong, I think, to suppose that having identified the underlying mechanisms in which beliefs, thoughts, sensations, and so on, are implemented, and those in which the possession of concepts of these is implemented, one must postulate yet other mechanisms, independent of these, that explain how it is that these states give rise to introspective beliefs about themselves.
Actually, it should be uncontroversial that it is an error to postulate mechanisms "completely independent of" the mechanisms linking the subpersonal realizing states. Such complete independence would lead to familiar problems about overdetermination. But complete independence is not the issue I want to focus on. I want to consider the position of someone who says there never is a personal-level, causal and reason-giving explanation of why a thinker has the belief that he has a certain belief, in normal cases. There is, according to this "no-reasons" position, a genuine explanation at the subpersonal level, but that is not at the personal, reason-giving level. It is, according to this position, written into the functional role of the concept of belief that normally when someone has a first order belief, he is willing to self-ascribe that first-order belief (if he has the concepts of himself and of belief, and if he considers the question). This, according to the "no-reasons" position I am considering, is a definitional remark. It is like the remark that valves allow only a one-way flow. It should not be confused with a causal explanation.
The epistemology which naturally accompanies the no-reasons theory is that of reliabilism. The reliabilist epistemology in this area is summarized by Shoemaker: "Our minds are so constituted, or our brains are so wired, that, for a wide range of mental states, one's being in a certain mental state produces in one, under certain conditions, the belief that one is in that mental state. This is what our own introspective access to our own mental states consists in. [...] The beliefs thus produced will count as knowledge, not because of the quantity or quality of evidence on which they are based (for they are based on no evidence), but because of the reliability of the mechanism by which they are produced" (1994, p.268). Reliabilism in this area may be elaborated in various different ways. Shoemaker's own view is that "believing that one believes that P can be just believing that P plus having a certain level of rationality, intelligence and so on..." (1994, p.289). On that version of the approach, it would, as Shoemaker immediately notes, be wrong to regard the second-order belief as caused by the first-order belief. But equally, someone who holds that the first-order belief causes the second-order self-ascription, in ordinary cases, could endorse the reliabilist epistemology. What will be common to all variants of the no-reasons theory is the claim that there are no reasons in the offing of the sort which would be required for the second-order beliefs to be knowledge on any more reason-based approach to epistemology.
The no-reasons theory is, then, one of the two extreme positions I want to identify.7 The other extreme position is occupied by the internalist introspectionist, exemplified by Alvin Goldman in some of his recent writings (1993). The internalist introspectionist holds not only that conscious attitudes are subjective states, are such that there is something it is like to have them. The internalist introspectionist believes further in an internalist theory of content. The internalist introspectionist holds that there is some level of intentional content of which we have knowledge, and which is not individuated by anything outside the thinker's head, neither his perceptual nor his social environment. For this internalist introspectionist, any externally individuated features of a state - any such feature of its content or of its role - are not available in consciousness itself.
It seems to me that each of these two extreme positions is correct in the criticisms available to it of the position at the other extreme; and that each of them is wrong in its own positive account. Suppose it consciously occurs to someone that that liquid (presented in perception) is water. The content here has many externalist features, even for someone who does not endorse Evans' (1982) and McDowell's (1984) "object-dependent" view of senses. The content of the perceptual demonstrative involves the liquid being presented as occupying a certain region of space relative to the thinker, and such contents can only be elucidated in externalist ways (Peacocke, 1993). The concept water is itself the topic of some of the most famous externalist discussions (Putnam, 1975). When it occurs to you that that liquid is water, or you hear someone else asserting that it is so, it seems that these very externally-individuated conceptual contents enter the content of your consciousness. We do not have a full description of your subjective state if we omit what you are thinking, or what you hear the other person as saying.
Actually, infidelity to the phenomenology would be the least of it. It would be an epistemological disaster to suppose that, in having a conscious belief, or understanding someone else's utterance, we are aware only of something weaker than an externally-individuated intentional content. For it is a datum that we do know the full, ordinary, externally-individuated intentional content of our own thoughts, and of other people's utterances, without reliance on inferences from, or presuppositions about, something weaker which is all, in some alleged stricter sense, we would be aware of on the internalist introspectionist's view. How this ordinary noninferential awareness and knowledge of one's own thoughts, and of the meaning of others' utterances, is possible at all would remain a mystery on the internalist introspectionist's view.
The problem would not be confined to knowledge of the content of one's own thoughts, and of the meaning of others' utterances. Let us take it as granted that the content of non-occurrent beliefs, desires and intentions is externally individuated, and likewise for the representational content of perceptual states. To deny then that conscious occurrent attitudes have an externally individuated content would lead to trouble on the following three fronts.
(a) It is a truism that conscious thinking can lead rationally to the formation of beliefs which constitute knowledge. How could such belief-formation have the status of knowledge if the content of the formed belief is externally individuated, but that of the rational thinking leading to it is not? The belief formed would require that the thinker stand in certain environmental relations which, on the internalist introspectionist's account, are not ensured by the contents of the conscious thoughts which rationally produce the belief.8
(b) Conscious thought can provide the rational explanation of an action under one of its environmental descriptions, such as reaching in one direction rather than another. Yet it seems that an action, under the given environmental, relational description, cannot be rationalized by conscious thoughts which are not relationally individuated. There would be a gap between the content of the thoughts and the pertinent relational property of the action which is, to all appearances, the property explained by the thought.
(c) If the content of perceptual experiences is, at least in part, externally individuated, then conscious judgements made rational by perceptual experience would involve a massive loss of information and specificity if the content of the conscious judgement were not externally individuated. I see no reason in principle why this information must be thrown away in rational thought. As a fully rational intermediary between perception on the one side and belief, desire, intention and action on the other, it seems that conscious thought must retain the externalist character of what rationalizes it, and what in turn it rationalizes.
It would be possible to continue in this vein, but the general problems which are emerging are enough to motivate a question. Why should anyone feel forced into a position in which they deny that conscious, occurrent states with externally-individuated contents can give reasons for thought and action, including - amongst others - knowledgeable self-ascriptions?
There seem to be three reasons influencing those who make the denial. The first is the idea that the nature of knowledgeable self-ascription of intentional states is inconsistent with externalism about content. This was certainly influencing Goldman, who supports his view by observing that "Cognizers seem able to discern their mental contents - what they believe, plan or desire to do - without consulting their environment" (1993, p.25). Like many others, I hold that cognizers are indeed able to do this, but that this truth is consistent with externalism about content. Burge (1988) and Davidson (1987) were pioneers in addressing the question of how the reconciliation is to be effected. I discuss the correct way of effecting the general reconciliation elsewhere (1996), and to keep this paper within tolerable bounds, must simply refer the reader to those several discussions.
A second factor influencing those who reject the conception I have been outlining is reflection on particular examples. Even an objector who believes in the consistency of externalism with a range of authoritative intentional self-knowledge may still have a more specific concern. He may worry that it is subjectively exactly the same for someone on earth who has the occurrent thought that water is wet as it is for his twin on twin-earth who has the occurrent thought that twater is wet. A natural way of elaborating this second source of concern runs as follows. When John thinks that water is wet, and twin-John thinks that twater is wet, they are thinking of different liquids in exactly the same way - only their contexts differ.9 According to this second objection, the subjective character of the occurrent thought that water is wet is fully captured by this way - `W' we can call it -in which water is thought of, together with the way - <is wet>, let us say - in which the property of being wet is thought of. These ways, the objection continues, are common to the thought that water is wet and the thought that twater is wet. So according to this objector,
John thinks that water is wet
will receive some analysis with the initial structure
Thinks( John, water, W^<is wet>)
Twin-John thinks that twater is wet
receives some analysis with the initial structure
Thinks( Twin-John, twater, W^<is wet>).
Our objector's point can then be formulated concisely: it is that only the third term of this relation contributes to the subjective character of the occurrent thought.
The objector who presses this point need not be disagreeing with the main thrust of the argument I have been presenting. The objector can agree that the fact that it occurs to a thinker that water is wet can contribute to fixing what it is like for him then. He is just offering a particular view about the way in which the first-order content of this occurrent thought is to be analyzed. Equally, this objector can be hospitable to the idea that a self-ascription of an attitude can be rationally based on a subjective state or event with a certain content.
Suppose we were to grant that there is some component of intentional content which cannot vary between twin-earthly counterparts, and that only that component contributes constitutively to the specification of subjective phenomenology. (I am not sure it is right to grant this, but let us do so for the sake of argument.) Granting that by no means implies that all subjective similarities involving intentional content are internally individuated. Suppose John has the perceptual-demonstrative thought, of a liquid he sees in a glass, that that liquid is drinkable; and that twin-John equally has the perceptual-demonstrative thought, of a liquid he sees in a glass, that that liquid is drinkable. We can fill out the example so that the liquids are thought of in the same perceptual-demonstrative way, in the objector's use of this term. This way must involve the liquid's being presented as in a certain direction and distance from the subject. It is overwhelmingly plausible that this way can be individuated only in external terms. One who has a perception, or thought, whose content includes it stands in certain complex, potentially explanatory relations to a place a certain distance and direction from him (Peacocke 1993). But this perceptual-demonstrative way certainly contributes to what it is like for one who has an occurrent thought `That liquid is drinkable'. I would also say the same of the perception itself which makes available the perceptual-demonstrative way of thinking of that liquid in the subject's vicinity. The perception has an externally-individuated content, and contributes to the nature of the thinker's subjective state. Both the occurrent thought and the perception can give reasons for the thinker to make knowledgeable self-ascriptions of attitudes with externally-individuated intentional contents. The upshot is that insisting that there are no subjective differences between an earthly person and his twin-earth replica is by no means to eliminate the phenomenon of conscious states with externally-individuated attitudes giving reasons for knowledgeable self-ascription of those very states.
There is also a third influence, of a rather different kind, upon those who think that conscious events with externally-individuated contents cannot give reasons for thought and action. The factor is well-identified - though not, I hasten to add, endorsed - in Paul Boghossian's discussion of a paradox. He writes:
"We sometimes know our thoughts directly, without the benefit of inference from other beliefs. [...] This implies that we know our thoughts either on the basis of some form of inner observation, or on the basis of nothing." (1989, p.5).
Later he continues:
"Ordinarily, to know some contingent proposition you need either to make some observation, or to perform some inference based on observation. In this sense, we may say that ordinary empirical knowledge is always a cognitive achievement and its epistemology always substantial." (p.17)
Similarly, Crispin Wright suggests that if we are to have "a substantial epistemology of intentional states",
"then it seems that the only relevant possibilities - since one does not know a priori of one's own beliefs, desires, etc. - are observation and inference" (1989, p.631).
Now if the categories "by observation; or by inference; or by nothing" were exhaustive, then one can see how pressure against the position I have been advocating increases. As Boghossian says, observation of a coin cannot tell us about its relational properties, such as its place of minting. Yet it is apparently relational properties of a mental state that we know in knowing its content. Nor are conscious attitudes known only by inference. If those three categories were exhaustive, we would already be moving on to the freeway which leads to the conclusion that psychological self-ascriptions are judgement-dependent. But the three categories are not exhaustive. Consider your knowledge of some feature of the content of your current visual perception. You do not observe your perceptual states; nor do you know about them only by inference; and there is certainly a good case to be made that many, if not all, of the contents featuring in perceptual experience are in part externally individuated. There is no evident obstacle to holding that a perceptual experience's having a certain content makes reasonable, for one conceptually equipped to think it, the first-person judgement that he is having an experience with a certain content. The reasonableness of such a judgement does not at all rely upon some level of internalist content which the experience has, and for which the transition to a judgement about the experience is unproblematic. On the contrary, for perceptual experience we seem to have externalism about representational content all the way down. I would make corresponding points for self-ascriptions of thoughts with particular contents, when they are consciously-based.
On the other hand, the internalist introspectionist is right at least to the extent that he emphasizes the subjective aspect of conscious attitudes, their ability to contribute to a specification of what it is like for the thinker. Such conscious states can give reasons, and there is equally no evident reason to deny that, for a conceptually-equipped thinker, they give reasons for a self-ascription of the attitudes that they are, with the contents they have. If one state gives a thinker's reason for a second state, they must be distinct states, and the reason-giving character of this explanation places it at the personal, not the subpersonal level. These points rule out the no-reasons theory.
If each of the extreme views' criticisms of the other is sound, as I suggest, then indeed neither of the extreme views can be correct. We need to recognize conscious propositional attitudes as a non-perceptual, non-sensational category of subjective, conscious states in their own right, with their externally individuated contents contributing ineliminably to the particular conscious states they are.
We have discussed the reasons which appear to, but do not in fact, support internalist introspectionism; but what of the considerations which lead to the no-reasons view? Of course, the spurious trilemma "by observation; by inference; or by nothing" could be influential here too. But Shoemaker also offered some more specific analogies with other kinds of case.
The parallel with logical inference mentioned in the quoted paragraph from Shoemaker actually seems to me to count against, rather than in favour of, a no-reasons theory. Consider a belief with a content containing a logical constant, where the belief is reached by inference from another conscious belief. Suppose the logical transition in question is one the willingness to make which, without further inference, is partially constitutive of grasp of the logical concept in question. Acceptance of the premiss of this transition does give the thinker a reason for acceptance of the conclusion, and, in standard cases, the premiss will be the thinker's reason for accepting the conclusion. Like any other statement about reasons, this is a statement about the personal, not the subpersonal, level. In fact, the proper way of individuating the logical concept in the content would itself make reference to, or at least entail, that acceptance of certain contents containing it is, in certain very central cases, the thinker's reason for accepting the corresponding conclusions. The crucial point is this: the fact that certain transitions are involved in the very identity of a concept is entirely consistent with the fact that entering the second state of the transition is done for the reason that the thinker is in the first state in the transition. The consistency of this combination is a commitment of anyone who holds the plausible view that concepts are individuated in part by what are good reasons for making judgements involving them, or by which contents involving them give good reasons for judging other contents.
An ordinary thinker's mastery of the concept of belief can then equally include as one component his willingness to self-ascribe the belief that p when he makes a conscious judgement that p (for the reason that he'd made the judgement, and in circumstances in which the question arises). Such a component of ordinary mastery adverts to an explanation of the self-ascription which is neither wholly subpersonal, nor definitional in a way which excludes reason-involving explanation of one state by another. Nor, as we emphasized, does recognition of such explanations involve a reversion to perceptual models of self-knowledge. Once such explanations are acknowledged, we would also expect the short-cut mechanisms mentioned in §2 to come into play: states which would lead to a judgement that p can come to produce a self-ascription of the belief that p without proceeding through the middle stage of the first-order judgement.
The intermediate position I have been advocating for the case of consciously-based self-ascriptions would actually be in agreement with Shoemaker when he writes that second-order belief about one's own beliefs "supervenes on the first-order state plus human intelligence, rationality, and conceptual capacity" (1993, p.79). The truth of this supervenience claim cannot be used to decide the issue between a no-reasons theory and the intermediate position. Suppose it is indeed so that having the concept of belief involves taking certain conscious judgements as reasons for self-ascribing the corresponding belief, and that first-order beliefs produce such reason-giving conscious judgements.Then failure of the supervenience claim Shoemaker makes would be as impossible on the intermediate view as it is on the no-reasons view. The intermediate view will hold that the first-order beliefs produced judgements which, in someone with minimal rationality and suitable conceptual capacity, will rationalize the self-ascription of the first-order state. Correlatively, the holding of the supervenience claim cannot be used as evidence for the view that second-order beliefs about one's own first-order beliefs are not distinct existences from the first-order beliefs they are about.
Neat and simple as this exposition of the intermediate position for consciously-based self-ascriptions may sound, I think it would be oversimple unless we draw a further important distinction. Within the class of conscious states, we can distinguish a proper subclass of states each member of which has the following property: it is either individuated in terms of what are good (non-instrumental) reasons for being in the state, or its individuation has consequences for what are good (non-instrumental) reasons for being in the state. A paradigm example of a state in this proper subclass is that of having a belief with a given content. This is a state which is either individuated by what are good reasons for being in it, or its individuation has consequences for what are good reasons for being in the state. Paradigm examples of states outside the proper subclass are those of having an experience with a given representational content, and the state of having a certain kind of sensation. These are states which are outside the immediate control of reason.
This classification is a classification of states themselves, rather than anything to do with the etiology of a thinker's coming to be in one of the states on a particular occasion. There may be no good reasons at all producing the paranoid's belief that others are out to get him. But the individuation of the belief does have consequences for what would be good (non-instrumental) reasons for being in that belief state. States in our distinguished class are not necessarily responsive to reason, but they are responsible to reason. I call states in the distinguished class reason-led states. (An acknowledged leader may be disobeyed, but it is disobedience which is in question, rather than a weaker relation of not doing what fulfilment of the content of the command would require). There is some rough overlap between class of states which are not reason-led and those which are exercises of a faculty of receptivity, rather than of spontaneity, in the sense used by Kant and revived by McDowell (1994). Some overlap is to be expected, since states in which one receives information about the world by perceptual contact are not ones the thinker is in because he has certain reasons.10
To be in a reason-led state is to be committed to something: to a certain content's being the case, or to one's doing something, or to something's being good in a certain respect (in the case of reason-led desires). Because of this element of commitment, there is always the possibility of a thinker's raising a practical question of whether he should be in a given reason-led state - whether he should believe that p, or form a certain intention.11 To enter one of these states is an action, which may be undertaken rationally or irrationally, and the general apparatus of action-explanation is applicable to it.
With all this in mind, let us go back to the case of a person who self-ascribes the belief that Dubcek was the Czech Prime Minister at the time of the Soviet Union's invasion, and does so because he has a conscious propositional memory which represents Dubcek as having that office then. The self-ascription of the belief on this basis goes beyond a mere report of oneself as having an apparent memory with that content. One makes the self-ascription in part because one is endorsing the content of the apparent memory. The self-ascription of the belief would after all not be correct if one were suspending judgement on the content of the apparent memory. So when a self-ascription of a reason-led state is made on the basis of the occurrence of a conscious state in this way, it is not a mere report of that state. It involves the same kind of endorsement and commitment as would be made in entering the first-order, reason-led state itself. This is a major difference from consciously-based self-ascriptions of experiences, sensations, or mental images. The distinction can give one some limited sympathy with those who say that self-ascriptions of belief and certain other states are not just descriptions of mental states. To say that the self-ascriptions are not consciously-based at all, however, would overstep the limits of legitimate sympathy.
For any concept, it is always illuminating to try to answer this question: why does it have to be as it is if proper applications of the concept are to track instances of the property picked out by the concept?12 In the case of an observational concept, we have at least the rudiments of an answer to this question. The possession condition for the concept would mention the role of a particular kind of perceptual experience in possessing the concept. We would begin to answer the question, as applied to a perceptual concept, by explaining how ordinary perceptual experience involves a sensitivity of the perceiver to instantiation of the property picked out by the concept. It is clear, in outline at least, how other theories of the concept of belief would aim to answer the same question as posed for the concept of belief. A no-reasons theorist would speak simply of a reliable sensitivity of a thinker's self-ascriptions of beliefs to instances of the property of believing the relevant contents, a sensitivity built into the account he would offer of possession of the concept of belief. Equally, a theorist such as Wright (1989), who treats the correctness of belief-ascriptions as constitutively partially dependent upon a subject's willingness to self-ascribe them, will also have his own answer to the question. For such a judgement-dependent theorist, the correctness of an ordinary self-ascription of a belief will not be independent of the thinker's willingness to make the self-ascription. This is a very different explanation of how the offered account of mastery connects with the truth of the self-ascriptions of belief. It may not be plausible; but at least it addresses the question. I, however, have rejected the pure no-reasons view, and I have not endorsed any judgement-dependent theory. So how do I answer the question of why the concept of belief has to be as it is if its proper applications are to track instances of the property it picks out?
There are at least two subquestions to be distinguished here.
Subquestion (a): What account, if any, is available in the present framework of the nature of the first-order property of believing a given intentional content, the property whose instances in the thinker are to be tracked by properly-made self ascriptions? Subject to some elaborations further below, I would answer this first subquestion by drawing on the material in A Study of Concepts (1992). The nature of the property of believing that p, for a given content p, is fixed by the possession conditions for the various concepts which comprise the content p, together with their mode of composition. Each such possession condition for a concept contributes, at a point determined by the mode of composition, a requirement on a state's being a belief with a content containing that concept. The totality of such requirements fixes the condition for a state to be a belief with the given content p. This condition fixes a property whose instances may or may not be tracked.
The condition is fixed in a quite specific way, but there is no plausibility that this way - which involves a detailed statement of the possession-conditions for the concepts involved - is a route employed at the personal level when a subject knowledgeably self-ascribes a belief. Ordinary thinkers need not have any personal-level conscious knowledge of the possession conditions of the concepts they employ. So this response to the first subquestion takes us only as far as a fixing of the property about which the opening question of this section can be raised. It is not an answer to that question, for it does not explain how any tracking takes place. It merely specifies the property which might be tracked.
Subquestion (b): How do the points in the preceding sections of this paper, about the nature of conscious, occurrent propositional attitudes, and their availability as reasons in their own right, contribute to answering the question of how properly-made self-applications of the concept of belief track instances of the property of believing a given content? If they are irrelevant to answering the question, should they not be omitted from an account of ordinary mastery of the concept of belief? If they are relevant, how do they contribute to tracking the property?
To address the cluster of issues in subquestion (b), I suggest that we need to step back a little and consider the nature of belief. If we were asked about the nature of belief, we would very plausibly include the following principles:
(I) To make a judgement is the fundamental way to form a belief (or to endorse it when it is being reassessed). Judgement is a conscious rational activity, done for reasons, where these reasons are answerable to a fundamental goal of judgement, that it aims at truth.13
(II) Beliefs store the contents of judgements previously made as correct contents, and these stored contents can be accessed so as to result in a conscious, subjective state of the thinker which represents the stored content as true. (The contents will often need to be adjusted as the spatio-temporal location of the thinker varies. This is why a cognitive dynamics, in Kaplan's sense (1989), is essential in the description of the mental life of any creature with beliefs.)
(III) Beliefs equally aim at the truth, rather than being a mere record of what was once judged. Hence their contents are always potentially open to revision by later judgements, perceptions, memories and reconsiderations of reasoning and evidence.
There is vast amount more to be said about the nature of belief and judgement, but it seems that features (I) - (III) of belief are core characteristics which will be retained in any more extensive elaboration.
Now we can bring this to bear on our second set of questions (b) above. I claim that our distinctive ways of coming to knowledgeable self-ascribe beliefs are correct methods in part because of the nature of belief given in characteristics (I) - (III). It is important for this claim that the characteristics (I) - (III) do not themselves mention knowledgeable self-ascription. It also matters for the claim that these are a priori characteristics of belief. They are not a posteriori in the way in which, as a true principle about its nature, the statement "Water is H2O" is a posteriori.
Consider the procedure of making a self-ascription of a belief that p for the reason that one has just consciously judged that p. By characteristic (I) of belief, the first-order judgement will, when all is working properly, be an initiation (or continuation) of a belief that p. So the self-ascription will be correct.
Consider also the case in which someone has a propositional memory representation of its being the case that p, is taking these representations at face value, and rationally moves, directly or indirectly, from this particular memory to the self-ascription of the belief that p. Again, when all is working properly, the memory representation will be a manifestation of an underlying stored belief that p, the kind of representation which, according to characteristic (II), stored beliefs will generate. So the self-ascription will be correct. Equally, when there is new evidence which means that the memory must be re-evaluated, characteristic (III) of beliefs explains why a rational self-ascription cannot just consist in re-endorsement of the stored information (or misinformation).
In summary, when all is working properly, knowledgeable self-ascriptions track the property of belief for this reason: the very means by which they are reached are ones whose availability involves the thinker's having the relevant belief. When all is working properly, these means would not be available were he not to have the relevant first-order belief.
On this approach, representations in conscious thought of contents as true can be produced by underlying beliefs; can contribute to what it's like subjectively for the thinker; and can be involved in reasonable self-ascription, of a kind involved in ordinary mastery of the concept of belief. All of this is consistent with knowledgeable self-ascriptions tracking the instantiation of the property of belief in the appropriate content. The position I have been developing seems to me to endorse all those propositions simultaneously.
It is worth noting the special link - whose existence one would well expect in advance - between the means for coming to make knowledgeable self-ascriptions and the first-person. A conscious memory representation, and equally a conscious judgement, can give a reason for making a self-ascription (or indeed for doing anything else) only to the person who has that memory, or who is making that judgement. So these rational means are restricted to self-ascriptions of beliefs and other mental states and events. Certainly, a premiss that someone else has a certain memory, or is making a judgement, can give a thinker reason to ascribe beliefs to someone else. But I have emphasized that consciously-based self-ascription is not a case of inference, and a fortiori not a case of moving inferentially from premisses about one's own mental states to a conclusion. The mental event itself, rather than some premiss about the event, is the thinker's reason for making the judgement.
I said that for some ways of coming to make knowledgeable self-ascriptions, the nature of belief and judgement is part of the explanation of their correctness. It is not the full explanation, and my exposition was peppered with occurrences of the qualifying phrase "when all is working properly". Someone can make a judgement, and for good reasons, but it not have the effects that judgement normally do - in particular, it may not result in a stored belief which has the proper influence on other judgements and on action. A combination of prejudice and self-deception, amongst many other possibilities, can produce this state of affairs. Someone may judge that undergraduate degrees from countries other than her own are of an equal standard to her own, and excellent reasons may be operative in her assertions to that effect. All the same, it may be quite clear, in decisions she makes on hiring, or in making recommendations, that she does not really have this belief at all. So the ways of coming to make self-ascriptions which I have been discussing are by no means infallible. Insofar as they are authoritative, in a wide range of ordinary cases, they are so only because those cases are not like the self-ascription of the belief about degree standards. It is an important question for further work exactly what further restrictions ensure an authoritative self-ascription. I do not think, though, that acknowledgement of this situation reinstates anything like a perceptual model of self-knowledge. Nothing here supports the conception of a distinction between the mental states and our perceptions, or experiences, of them. Our recently imagined subject is mistaken about whether she has the belief that degrees from other countries are of equal value with her own; but not every case of error is a case of perceptual error.
A second question meriting someone's further investigation is how far the model developed in this section is generalizable to knowledge of other mental states. What I have done in the case of belief is to argue that the explanation of the correctness of certain distinctive ways of coming to knowledgeable self-ascriptions is in part the essential nature of the mental state ascribed. Can we do this for other mental states too? Certainly it seems at first blush as if the approach can be developed pari passu for self-ascriptions of intention. This is partly because decision stands to intention and the self-ascription of intention somewhat as judgement stands to belief and the self-ascription of belief. If the approach can be developed for other mental states and their self-ascription too, we have the promise that the connection we noted in the case of belief between the nature of the mental state, reasons for knowledgeable self-ascription, and the first-person may generalize to other mental states. It may be another general connection between the mental and first-person thought.
I turn now to consider some consequences of this account of conscious occurrent attitudes and consciously-based self-ascription, and some questions and challenges they raise.
(a) It is worth reflecting further on what is involved in the consciousness of those conscious mental events which are capable of contributing to the rational explanation of self-ascriptions. If the conscious nature of the mental event is partially explanatory of the self-ascription, this consciousness cannot consist in the event's availability to belief or to verbal report. Similarly, if the conscious nature of the mental event is partially explanatory of some other non-verbal action or behaviour, this consciousness cannot consist in its disposition to produce such behaviour.
There are occasions on which a person expresses a first-order belief, or indeed makes a self-ascription of a belief, and in which these are not consciously-based in the way I have been discussing. Most of us, when it becomes conversationally appropriate to say "I know my name is NN" , or "I know my address is such-and-such", have no need to wait upon its surfacing in consciousness what our names and addresses are. We make these utterances intentionally and knowledgeably, but not because it has just occurred to us that our names and addresses are such-and-such. If we were to analyze the consciousness of an occurrent thought in terms of its availability to immediate first-person belief or report, we would be failing to distinguish the genuinely consciously-based self-ascriptions from these latter cases - from which they are distinct. We would also undermine the possibility of explaining certain examples of self-ascription in part by reference to a conscious event. On the theory we are currently rejecting, the consciousness of a mental event would consist in its disposition to produce self-ascriptions in thought or language. A disposition cannot be the cause of its manifestations.
How then do the conscious mental events which can support consciously-based self-ascriptions stand in relation to the well-known and, in my judgement, valuable distinction drawn by Block (1993, 1995a) between `phenomenal' consciousness and `access' consciousness? Phenomenal consciousness ("P-consciousness") has been characterized as experience in general, understood in Block's writings so widely as to include occurrent thoughts. Block characterizes a state as access conscious (A- conscious) "if, in virtue of one's having the state, a representation of its content is (1) inferentially promiscuous (Stich 1978), that poised for use as a premise in reasoning, (2) poised for rational control of action, and (3) poised for rational control of speech" (Block, 1995a p.231). When someone knows what he is thinking because he has just consciously thought it, his first-order thinking is access-conscious because it is phenomenally conscious, in Block's broad sense. For conscious mental events of a sort which support consciously-based self-ascriptions, their access-consciousness is not independent of their phenomenal consciousness.
In one of his papers, Block writes "I don't know whether there are any actual cases of A-consciousness without P-consciousness, but I hope I have illustrated their conceptual possibility" (1995a, p.233). He also notes that "If indeed there can be P without A, but not A without P, this would be a remarkable result that would need explanation" (1995b, p.272). Other writers have also placed some weight on the alleged nonexistence of cases of access consciousness without phenomenal consciousness.14 But consider the knowledgeable self-ascriptions like "I know that my name is Christopher Peacocke", where there is no intermediate conscious state rationally producing the utterance (nor the judgement). These seem to me to be straightforwardly cases in which the knowledge or belief is access conscious without being phenomenally conscious. In accordance with Block's definition, the representation of the content of the knowledge or belief is inferentially promiscuous, and can certainly be poised for rational control of action and speech. (It is actually controlling it when I say "I know my name is Christopher Peacocke").
On the views I am proposing, then, there is more than one sort of availability for verbal report, and we should distinguish between the case in which there is an intermediate conscious event playing a role in the explanation of the verbal report, and the case in which there is not. It should be noted that we do in fact still call a belief conscious if its content is expressed, or even reported, without the occurrence of any intermediate occurrent conscious state. This is one strand in our ordinary notion of a conscious state. Whether its presence is fundamental, or owed rather to its connection with another element which is fundamental, is a further question, to which I turn almost immediately.
(b) The cases of knowledgeable self-ascriptions not based on any intermediate conscious state raise many questions. One of the most pressing is the support they may seem to give the no-reasons view of §3 above. For the no-reasons theorist may counterattack against the arguments developed in §6 in two ways. First, the no-reasons theorist may say that a reliability account of a thinker's knowledge of his own attitudes is the only epistemology which can properly accommodate those knowledgeable cases where there is no intermediate conscious state. Second, the no-reasons theorist may even say that the correctness of his account is not confined to no-intermediate-conscious-state ("NICS") examples. It can be extended to the cases where there is an intermediate conscious state, since a conscious judgement that p is precisely something which is reliably correlated with a thinker's believing that p (certainly, he may say, sufficiently reliably to ground attributions of knowledge).
We need, then, to address two clusters of questions:
(i) Are reliability accounts the only epistemological accounts which can explain why the second-order beliefs in the NICS examples amount to knowledge?
(ii) Is the class of NICS examples philosophically more fundamental than the class of intermediate-conscious-state examples? Or vice versa? Or is neither more fundamental than the other?
There are general reasons for thinking that there must be some alternative to a reliabilist treatment of the NICS self-ascriptions. For the objections to pure reliabilism - its omission of any rationality or entitlement requirement - are well-known, and have not, so it seems to me, been overcome.15 It would be very puzzling if reliabilism were right about just one kind of psychological self-ascription, but wrong elsewhere. In fact it seems to me that there is in fact an alternative account which treats the NICS self-ascriptions as knowledge, but which is not a purely reliabilist account. This alternative account says that an NICS self-ascription of (say) a belief that p is knowledge only if it is made in circumstances in which the thinker is also willing to make the first-order judgement that p. We can call the requirement appealed to in this alternative account the requirement of first-order ratifiability.16 There is considerable plausibility in the claim that it is the holding of first-order ratifiability which makes an NICS second-order self-ascription of belief a case of knowledge (makes the ascriber entitled to his second-order judgement, if you will). Suppose a thinker were to make an NICS self-ascription of a belief, but that first-order ratifiability failed - when he reflects on it, the thinker is not prepared to make the first-order judgement that p. In these circumstances, the second-order judgement would, other things equal, be unstable. The second-order judgement would in those circumstances normally be withdrawn. If it were the case that our subject had been told that his NICS self-ascriptions were a way of tapping into his unconscious beliefs, perhaps the second-order self-ascription would not be withdrawn. That would, though, then clearly be a case of inferential knowledge, very different from the normal case in which NICS self-ascriptions constitute non-inferential self-knowledge.
A requirement of first-order ratifiability can also help to explain the evident rationality of the Evans procedure for self-ascription. Necessarily, someone self-ascribing a belief by Evans's procedure meets the condition that he self-ascribes the belief that p in circumstances in which he is also willing to make the first-order judgement that p.
If first-order ratifiability is the correct explanation of how NICS self-ascriptions can constitute knowledge, then in order of philosophical explanation - as opposed to frequency of examples - the intermediate-conscious-state cases are philosophically more fundamental than their NICS counterparts. If first-order ratifiability is required for these case to be knowledge, as I am inclined to believe, then NICS cases count as knowledge (when they do) because of the relation in which they stand to conscious first-order attitudes, and to the rational basis those conscious first-order attitudes provide for self-ascribing attitudes. On this approach, then, the existence of NICS self-ascriptions which constitute knowledge can be squarely acknowledged without embracing a purely reliabilist epistemology.17
(c) The present approach must incline one to scepticism about the view that the consciousness of an occurrent propositional attitude consists in its massive inferential integration. Let us fix on one reading of an example of Dennett's, that of the father who is self-deceived in his belief that his son is a good painter (1979, essay 3). This belief may well not be strongly inferentially integrated. All the evidence which, in the case of anyone else, would convince the father that some third person is not a good painter is ineffective in the case of his son's painting. The belief that his son a good painter is conscious all the same, and such thoughts as "My son is a good painter" may pass through his mind as an objection when he hears someone assert "None of the good painters have living parents".
The conscious belief will be inferentially integrated in a somewhat weaker sense. The father will respond to offered evidence that his belief is false by trying to undermine or discount the evidence. But that weaker sort of integration can also be present when there is no conscious belief. The mother who does not know whether her soldier son has survived a battle may have no conscious opinion on the matter, but may still try to undermine or discount any evidence that he has not. (This could be caused by a desire that he be alive.) This falls short of amounting to a conscious belief that her son is alive, belief which she may sincerely deny she has. If one consciously accepts that p, there will indeed normally be extensive inferential integration of the belief that p with one's other conscious attitudes, practical reasoning and emotions. It seems to me that this is a normal - though as the examples show, not an invariable - consequence of the consciousness of a belief that p. It is not what its conscious status consists in. The mother's unconscious belief that her son is alive may have as extensive ramifications for her other attitudes and her reaction to apparent counterevidence as does the father's conscious belief that his son is a good painter. If this is possible, then the consciousness of an attitude cannot be elucidated in terms of extent of influence. This point also seems to count against approaches to consciousness in terms of "cerebral celebrity", as Dennett felicitously calls it (1993, pp.929-31). It looks as if the father's conscious belief and the mother's unconscious belief may be on a par in respect of cerebral celebrity.
A different manoeuvre might be attempted. For instance, it is also true in the imagined examples that the father, unlike the mother, is under a rational obligation to explain away the contrary evidence, to show it is not contrary, or to change his views, whereas the mother is under no such rational obligation. This too, though, seems to me to be a consequence of the conscious status of the father's belief. A conscious state or event with content imposes rational requirements on a thinkers, for it gives reasons for thinking or doing. If it gives reason for accepting something the thinker also has other reasons for rejecting, the thinker is under a rational obligation to revise his beliefs.
In the basic, personal-level case in which something is done for a reason, whether it be in thought or bodily action, the reason-giving state must be either conscious, or it could become conscious for the thinker. A reason-giving state need not be actually conscious. If you decide to fly to Paris, you may call one airline rather than another. There need not be any conscious state, one contributing to what it's like for you, just before or after your decision, which is the reason-giving state which rationally explains your calling that airline. But if this was a minimally rational action, your reason could become conscious if the question arose. In a case in which no reason becomes conscious, when the question arises, and the thinker consequently cannot explain why he chose to call that airline, we have a much-diminished sense of the rationality of the action. The requirement that the reason could become conscious is reminiscent of a Kantian position: "It must be possible for the `I think' to accompany all my representations; for otherwise...the representation...would be nothing to me" (1929, B131; my emphasis). The requirement that the reason-giving state is one which is or could become conscious is intimately related to our conception of an agent as someone with a point of view, and whose rational actions make sense to the subject himself (and not just to other experts) given that point of view. For an alleged reason-giving state which could not even become conscious, this condition would not be met. Any action produced by it would not make sense even to the subject himself.
There is more than one way of regarding this principle linking conscious states and reasons. One approach regards this connection as part of a positive elucidation of consciousness in terms which do not, according to this first approach, presuppose consciousness. I do not find this first approach plausible. It certainly does not promise any explication of the important distinction between what is actually conscious and what is only potentially so. The approach would also apparently rule out the existence of more primitive creatures who enjoy what are, in some general but univocal sense, conscious states, but to which the full apparatus of reason-based explanation is inapplicable. A more plausible approach does not aim to treat the connection as contributing to some kind of reduction of consciousness, but rather sees it as linking reason-involving explanation with some general notion of consciousness which has a life outside the principle. This second approach allows for the possibility of the existence of conscious states prior to that of reason-based explanation. In any case, whatever the correct attitude to the connection, its mere existence would suffice to explain why rational subjects must be capable of enjoying conscious states.
Boghossian, P. (1989). "Content and Self-Knowledge." Philosophical Topics XVII: 5-26.
Block, N. (1993). Review of Dennett's Consciousness Explained. Journal of Philosophy 90: 181-93.
Block, N. (1995a). "On a confusion about a function of consciousness". Behavioral and Brain Sciences 18 227-247
Block, N. (1995b). "How many concepts of consciousness?". Behavioral and Brain Sciences 18 272-287.
Bonjour, L. (1985). The Structure of Empirical Knowledge. Cambridge, Mass., Harvard University Press.
Burge, T. (1988). "Individualism and Self-Knowledge". Journal of Philosophy 85 649-663.
Burge, T. (1996). "Our Entitlement to Self-Knowledge". Proceedings of the Aristotelian Society XCVI 91-116.
Davies, M. and G. Humphreys (1993). Introduction. Consciousness: Psychological and Philosophical Essays. Oxford, Blackwell.
Davidson, D. (1987). "Knowing One's Own Mind". Proceedings and Addresses of the American Philosophical Association. 60 441-58.
Dennett, D. (1972). Content and Consciousness. London: Routledge and Kegan Paul.
Dennett. D. (1978). Brainstorms. Vermont: Bradford Books.
Dennett, D. (1993). "The Message is: There is no Medium." Philosophy and Phenomenological Research LIII: 919-31.
Dennett. D. (1995). "The path not taken." Behavioral and Brain Sciences 18 252-253.
Evans, G. (1982). The Varieties of Reference. Oxford, Oxford University Press.
Flanagan, O. (1992). Consciousness Regained. Cambridge, Mass., MIT Press.
Goldman, A. (1993). "The psychology of folk psychology." Behavioral and Brain Sciences 16: 15-28.
James, W. (1983). The Principles of Psychology. Cambridge, Mass., Harvard University Press.
Kant, I. (1929). Critique of Pure Reason (tr. Kemp Smith). London, Macmillan.
Kaplan, D. (1989). "Demonstratives". In Themes from Kaplan ed. J. Almog, J. Perry and H. Wettstein. New York: Oxford University Press.
McDowell, J. (1984). "De Re Senses." Philosophical Quarterly xxxiv: 283-94.
McDowell, J. (1994). Mind and World. Cambridge, Mass., Harvard University Press.
Moran, R. (1988). "Making up your Mind: Self-Interpretation and Self-Constitution." Ratio (New Series) I: 135-151.
Morton, A. (1995) "Phenomenal and attentional consciousness may be inextricable." Behavioral and Brain Sciences 18 263-264
Peacocke, C. (1985). "Imagination, Possibility and Experience." In Essays on Berkeley, ed. J. Foster and H.Robinson. Oxford, Oxford University Press.
Peacocke, C. (1992). A Study of Concepts. Cambridge, Mass., MIT Press.
Peacocke, C. (1993). "Externalist Explanation." Proceedings of the Aristotelian Society XCIII: 203-230.
Peacocke, C. (1994). "Content, Computation and Externalism." Mind and Language. 9: 303-35
Peacocke, C. (1996). "Entitlement, Self-Knowledge and Conceptual Redeployment." Proceedings of the Aristotelian Society XCVI: 117-158.
Putnam, H. (1970). On Properties. Essays in Honor of Carl Hempel, ed. N. Rescher. Dordrecht, Reidel.
Putnam, H. (1975). The Meaning of "Meaning". Mind, Language and Reality. Cambridge, Cambridge University Press.
Ryle, G. (1971). A Puzzling Element in the Notion of Thinking. Collected Papers vol.II: Collected Essays 1929-1968. London, Hutchinson. 391-406.
Shoemaker, S. (1993). "Special access lies down with theory-theory." Behavioral and Brain Sciences 16: 78-9.
Shoemaker, S. (1994). "Self-Knowledge and "Inner Sense"." Philosophy and Phenomenological Research LIV: 249-314.
Stich. S. (1978). "Beliefs and sub-doxastic states". Philosophy of Science 45 499-58.
Wiggins, D. (1984). "The Sense and Reference of Predicates: A Running Repair to Frege's Doctrine and a Plea for the Copula". In Frege: Tradition and Influence, ed. C. Wright. Oxford, Blackwell.
Wright, C. (1989). "Wittgenstein's Later Philosophy of Mind: Sensations, Privacy and Intention." Journal of Philosophy LXXVI: 622-34.
1 In addition to the presentation at the 1995 St. Andrews conference on Self-Knowledge, versions of this material were given in lectures in Oxford in 1994, in a talk to a conference on Consciousness and Attention organised by the Institute for Advanced Studies at London University in 1995, and to seminars and a colloquium at New York University and the Graduate Center, CUNY, in spring of 1996. I am particularly grateful to Michael Martin for his detailed, sympathetic and very helpful comments delivered at the St. Andrews meeting. In particular, he raised the question which opens section 3 below. My thanks also to Paul Boghossian, Tyler Burge, Tony Atkinson, John Campbell, Martin Davies, John McDowell, Thomas Nagel, Stephen Schiffer, Barry Smith, Michael Tye and Crispin Wright for observations that have influenced the present version.
© Christopher Peacocke 1996
2 Here I differ from Goldman.
3 Second paragraph of the section "The Varieties of Attention" in the chapter "Attention" in volume I of The Principles of Psychology.
4 There is further discussion of the distinction in Peacocke 1985.
5 I should also emphasize that the procedure discussed in the example in the text is considered as a procedure for reaching beliefs, where belief is understood as a form of acceptance. We do, as Michael Martin emphasized to me, sometimes use `belief' for a feeling of conviction; the above procedure is not meant to apply to those interesting cases, which need a different treatment.
6 The case in which there is no intermediate conscious state is considered also in Peacocke 1996 (p.121). There are other kinds of cases as well. The self-ascription may, for instance, be contextually self-verifying - the case which Tyler Burge has investigated in detail (1988, 1996).
7 I have drawn elements of the no-reasons theory from Shoemaker's writings, but I should note explicitly that I have not found an endorsement of it by him in so many words. I should also note, without developing the point, that if introspection involves the occupation of attention, and the treatment of the occupation of attention in §1 above is roughly correct, then it is less tempting to elucidate introspection simply in terms of the production, in some specified way, of a certain kind of belief (as one of the sentences quoted from Shoemaker in the preceding paragraph suggests it can be).
8 This argument has certain affinities to that given in Peacocke 1994, in support of the conclusion that if a subpersonal psychology is to be capable of explaining externally individuated propositional attitudes, the explanatory states introduced by the psychology must also be externally individuated.
9 Their thoughts are relationally similar in the sense discussed in Peacocke 1996, pp.150-1.
10 The coincidence of the classifications is not exact, though. For instance, states of visual imagination are not states of receptivity in the Kantian sense. But I also do not think (though this might take some argument) that they are individuated by, or that their individuation has consequences for, what are good, non-instrumental, reasons for being in them.
11 Compare Moran, 1988.
12 On the distinction between properties and concepts, see Putnam 1970 and Wiggins 1984.
13 My own view is that judgements are in fact actions, a species of mental action. Judgements are made for reasons. Perhaps (though even this is debatable) they are not intentional under any description. I do not regard that as sufficient for them not to be actions. Forming an intention to do something seems equally to me to be an action (and can also be done for reasons). But in the sense in which judgements are not intentional under some description, forming an intention is not intentional under some description either.
14 For instance Dennett (1995), Morton (1995).
15 See, for instance, Bonjour (1985), pp.37-57.
16 One may want to strengthen the requirement somewhat, so that it does not merely talk of "the circumstances" in which one would be willing to make the corresponding first-order judgement. A plausible stronger requirement is that the mechanism which produces the second-order judgement in the NICS case must persist because it has the property of first-order ratifiability.
17 The points of this section also tell against the view that a full account of how self-ascriptions of belief are knowledge is given as follows: whenever a sentence s is stored in someone's "belief-box", the sentence I believe that s is also stored in his belief-box. Some such subpersonal mechanism may - perhaps must - exist. But some additional account has to be given if we are to explain why the belief realized by storing the sentence I believe that s in the belief-box is knowledge.