In this chapter we review, and contribute to, the intense debate which has raged concerning the appropriate notion of content for psychology (both folk and scientific). Our position is that the case for wide content (that is, content individuated in terms of its relations to worldly objects and properties) in any form of psychology is weak; and that the case for narrow content (that is, content individuated in abstraction from relations to the world) is correspondingly strong. But we also think that for some common-sense purposes a notion of wide content is perfectly appropriate.
The main reasons why this debate is important have to do with the implications for folk and scientific psychology, and the relations between them. (But it will also turn out, in Chapter 9, that the defensibility of narrow content is crucial to the naturalisation of consciousness.) For if, as some suggest, the notion of content employed by folk psychology is wide, whereas the notion which must be employed in scientific psychology is narrow, then there is scope here for conflict. Are we to say that science shows folk psychology to be false? Or can the two co-exist? And what if the very idea of narrow content is incoherent, as some suggest? Can scientific psychology employ a notion of content which is externally individuated? Or would this undermine the very possibility of content-involving psychology?
Some wide-content theorists, such as McDowell (1986, 1994), believe that the debate has profound implications for philosophy generally, particularly for epistemology. McDowell maintains that narrow-content theorists place an intermediary between the mind and the world, somewhat in the way that Cartesians and sense-data theorists did, making sceptical worries especially pressing. We believe that this is a muddle. The debate is about the individuation-conditions for content, not about referential semantics, or about the phenomenology of thinking. Narrow-content theorists should agree that each token thought will have truth-conditions, and those truth-conditions will standardly involve worldly items and states of affairs. Equally, a narrow-content theorist should agree that when one thinks a token thought, the whole focus of ones attention may be on the worldly items which the thought concerns. But narrow-content theorists deny what wide-content theorists assert, namely that the truth-conditions of thoughts are essential to their identity. A narrow-content theorist will say that the very same thought could have been entertained, in different circumstances, with different truth-conditions.
The theory of content which derives from Frege (1892), and which has dominated philosophical thinking for much of this century, distinguishes two different aspects of thought-content and sentence-meaning there is reference, which is constituted by the states of affairs and objects in the world which our thoughts concern; and there is sense, which is the mode of presentation of or the manner of thinking of reference. The terms Venus and The Evening Star share the same reference but differ in sense. And the thoughts expressed by Venus has set and The Evening Star has set share the same truth-conditions, but differ in the manner in which those conditions are presented in thought. So one might, for example, believe the first thought to be true while denying that the second was, and vice versa.
As is suggested by this last remark, Fregean sense is to be individuated in accordance with the intuitive criterion of difference two senses are distinct if it is possible for someone rationally to take differing epistemic attitudes to thoughts which differ only in that the one contains the one sense while the other contains the other (as in the example of Venus and The Evening Star just given).
On the Fregean account, sense is supposed to determine reference. It is supposed to be impossible that any term or thought-component should share the very same sense as the term Venus and yet differ in reference. (Reference, on the other hand, does not determine sense there are many different ways of referring to, or thinking of, the planet Venus.) So it is sufficient to individuate the content of a thought, or the meaning of a sentence, that one should specify its sense, since the reference will thereby have been fixed.
Difficulties for the Fregean system began to arise when it was noticed that there are many terms which do not appear to differ in the manner in which they present their referents (from a subjective point of view, at least), and yet which refer to different things. For example, the indexical term I seems to have the same sense for each one of us, but picks out a different person in each case. So either (1) we have to say that sense does not determine reference, or (2) we have to say that the actual reference belongs amongst the individuation-conditions of a sense.
Defenders of narrow, or internalist, content take the first option. They say that the thought, I am cold has the same sense (the same narrow content) for each one of us. But those senses are about different things, and different tokens of the very same (narrow) thought can have different worldly truth-conditions. Defenders of wide, or externalist, content take the second option. They say that since the token thoughts expressed by I am cold have different truth-conditions in the case of each one of us (and can in some cases be true while in other cases being false), those thoughts belong to different types, with different contents. So we do not think the same thing when we each of us thinks I am cold. The thoughts are distinct because the referents are.
2 Arguments for wide content
In this section of the chapter we shall consider some of the arguments which have been offered in defence of wide content, concluding with the argument that narrow content is actually incoherent. Then in Section 3 we respond to this challenge, arguing that it is at least possible that psychology should be narrow; and in Section 4 we shall argue that explanatory psychology is narrow.
2.1 Externalist intuitions
Putnam (1975a) devised a new type of philosophical thought-experiment to demonstrate that meanings aint in the head. We are to imagine that there is, or could be, an exact duplicate of Earth (Twin Earth, often written Twearth), where everything is exactly as it is on Earth, except for some minor respect which can be varied depending upon the type of example. Imagine, in particular, that everything on Twearth is exactly as it is on Earth, down to the smallest detail, except that on Twearth water H2O. Rather, on Twearth water XYZ, where the two substances can only be distinguished from one another in a chemistry laboratory. Putnam argues that if a person on Earth, Petere, asserts Water is wet and his twin on Twearth Petertw makes the same utterance, then their thoughts differ in content (and their sentences differ in meaning) because the substances they respectively refer to are different.
Suppose that neither Petere nor Petertw initially knows the composition of water/twater. Then each of them is told in similar circumstances Water is H2O, and believes it. Surely, Putnam argues, they cannot have formed the very same belief, since Peteres belief is true while Petertws belief is false. And how can the very same thought-content be both true and false at one and the same time? Yet every aspect of their brains and of their internal (non-relationally described) psychology is, by hypothesis, exactly the same. Conclusion: the contents of thoughts about natural kinds (and the meanings of sentences referring to natural kinds) depend upon the actual internal constitution of the kinds in question. Where that actual internal constitution differs, then so does the content of the thought. That is to say: thought-content is wide in its individuation-conditions, involving properties (often unknown) in the thinkers environment.
Similar arguments have been developed by Burge (1979, 1986a), in connection with non-natural kinds, this time turning on linguistic division of labour. (This latter phenomenon was also discovered by Putnam I can say There is an elm in the garden and mean that it is an elm, even though I personally cannot tell elms from beeches, because I speak with the intention of deferring to the judgements of those who can distinguish them.) Burges well-known arthritis example is designed to show that thought-content depends for its identity partly on social facts about ones linguistic community, which can differ even when nothing internal to two thinkers is different. So, again, the moral is that meaning aint in the head.
The example (slightly adapted) is this: Petere and Petertw are identical in all physical and non-relationally-described respects; and each believes that arthritis is a painful condition effecting the joints and bones. The difference between them is that Petere lives in a community where people use the term arthritis just to designate a certain kind of inflammation of the joints (his false belief results from some sort of misinformation or confusion); whereas Petertw lives in a community where people use the term arthritis rather more broadly, to refer to a range of painful conditions (by hypothesis, Petertw formed his belief through a causal route exactly mirroring the way in which Petere formed his). But now when each of them asserts, I have arthritis in my thigh, one of them (Petere) says something false, whereas the other (Petertw) expresses a belief which is true. This is then supposed to motivate us to think that Petere and Petertw entertain beliefs with different contents, merely by virtue of living in different linguistic communities. So, it is argued, those beliefs must be externally individuated.
Yet another set of externalist intuitions is invoked by Evans (1982) and McDowell (1984, 1986, 1994), who focus especially on singular thoughts. They maintain that singular thoughts are both Russellian, in that they involve, as constituents, the actual individual things thought about, and Fregean, in that they also involve a mode of presentation of those things. Bertrand Russell had maintained that thoughts are relations between persons and propositions, where a proposition is a complex consisting of the actual objects of thought themselves (individuals and properties). So if I entertain the thought Pavarotti is fat, this consists of a relation between me, the singer Pavarotti himself, and the property of fatness. Such a view at least has the virtue of simplicity.
The problem for Russells account is that it is too austere to do all the work that we need a notion of thought to perform. In particular, we surely think that there can be many different thoughts about the singer Pavarotti and the property of fatness. Whereas on Russells account there is only one (not involving any other elements, such as negation Russell of course allows that the thought Pavarotti is not fat is different). Thus the thoughts Pavarotti is fat and That man is fat where the that is a demonstrative element, picking out a particular person seen on TV or on the stage are surely different. For if I do not know what Pavarotti looks like, I might believe the one to be false while believing the other to be true. And so those thoughts would guide my behaviour differently too I would say No in response to a question about the first, but would say Yes in answer to the same question about the second. Yet they both involve the very same Russellian proposition: both ascribe the property of being fat to the very same man.
Evans and McDowell believe that singular thoughts are individuated, in part, by the objects they concern. But they allow that thoughts may also differ by differing in the way in which one and the same object is presented. (It is a consequence of this view that in the absence of an appropriate individual, there is no singular thought there to be had. So someone who is merely hallucinating the presence of an individual is incapable of thinking any singular thought about that putative thing. We shall return to this consequence below.) On this account, then, although a singular thought contains two different aspects (the object in the world, and its mode of presentation), these are not supposed to be fully separable. In particular, there is supposed to be no possibility of the singular mode of presentation either existing, or being characterisable, independently of the object presented.
In what follows, we shall focus on the case of singular thought in particular, for two reasons. The first is that all the issues which concern us arise here in their sharpest relief. Our main conclusions should generalise from this relatively simple case to thought about both natural and non-natural kinds. The second is that the externalist intuitions stimulated by Twin Earth thought-experiments are usually felt most strongly in the case of natural kinds like water. Resisting those intuitions is liable to involve arguments sketching elaborate scenarios transportation between Earth and Twin Earth, migrant interplanetary plumbers who may or may not suffer from amnesia, a Twix Earth where the liquid in the Atlantic is XYZ while that in the Pacific is H2O, and such like. Diverse variations on the original thought-experiment may dilute the strength of the intuitions, but usually allow some way out for the externalist who stubbornly clings to Putnams model of the indexical introduction of natural kind concepts, according to which one means by water (say) something like: Stuff which is the same as this in its basic composition. So the most incisive line of attack will be directed against singular thoughts such as those about ostended samples, on which indexical concepts must ultimately depend.
2.2 Arguments for externalist folk psychology
What reason is there to believe that singular thoughts are Russellian (constitutively involving the object thought about) as well as Fregean? That is, why should one believe (roughly) that same content = same reference + same mode of presentation? One bad argument, which is nevertheless frequently to be found in the literature, is that we routinely describe such thoughts in terms of their objects. We say John thinks that that cat is dangerous; Mary thinks that John is a coward; and so on.
Now, this certainly does not show anything very much by itself. For the fact that we individuate, describe, or pick out, something by its relation to another does not show that the latter figures amongst the identity conditions of the thing in question. Thus, I might pick out Big Ben tower in London for you by saying, It is the clock-tower which stands next to the Houses of Parliament. But this does not make the Parliament building constitutive of the identity of Big Ben. On the contrary, we think that the former could be destroyed, for example, while leaving the latter the very same as it was. So, the fact that I individuate Johns thought for you by indicating the particular cat he is thinking about, does not show that the cat in question is a constitutive part of, or essential to the existence and identity of, his thought.
There are, however, cases in which we might be tempted to insist that distinct thoughts are entertained, where the only available distinguishing feature is the difference in their objects. For example, suppose that Mary and Joan walk into different but exactly similar burger bars, and sit down at identical tables in the corner. Each then thinks, This table is greasy. Since one of these thoughts might be true while the other is false (in Joans case the table may only be wet), it looks as if we need to insist that the thoughts belong to distinct types. But there is nothing in the mode(s) of presentation of the tables to distinguish them. The difference must then lie with their objects that is to say, in the numerical difference between the two tables. So, in contrast with the Big Ben example above, it might be held that it is no mere accident that we would describe these thoughts by indicating which table is in question by saying, for example, Mary is thinking that that table (the one in front of her) is greasy, whereas Joan is thinking that that table (the one in front of her) is greasy.
But this argument just assumes, without defence, that it is thought-types rather than thought-tokens which are the primary bearers of truth-values. For recall that narrow-content theorists do not deny that thoughts have truth-conditions; they just deny that thoughts (as types) are to be individuated in terms of their truth-conditions. So if it is thought-tokens which are the bearers of truth-values, then we can say that Mary and Joan are both thinking thoughts of the very same type, with the very same (narrow) content; but since they entertain distinct tokens of that type, the one can be true while the other is false.
Another more powerful argument picks up on, and defends, the Russellian consequence that a singular thought must fail to exist in the absence of an appropriate object. (If singular thoughts are individuated by their relation to the object referred to, in such a way that the object is part of the identity of the thought, then if there is no object, there is no thought either.) Thus suppose that I hallucinate the presence of a cat, and think to myself, That cat is lost. On the Russellian view, I here attempt, but fail, to think a singular thought. It merely seems to me that I have thought a demonstrative thought, when I have not. Now if we wish to reject the Russellian view, we shall need to avoid this consequence. And that means finding a way of saying what thought I succeed in thinking in the case of the hallucinated cat. A further argument for externalism, then, is that none of the available alternatives seems successful.
In particular, the content of the (putative) singular thought, That cat is lost is not the same as the content of any descriptive thought, which would be available for me to entertain (but which would then be false) in the case of the non-existent cat. (For these purposes we assume the truth of Russells Theory of Definite Descriptions, according to which a statement of the form The F is G should be analysed as saying There exists one and only one [relevant] F and it is G.)
E.g. (1): the thought The cat in my office is lost (supposing that I am in my office at the time). For I may doubt this, while continuing to believe that that cat is lost, if I forget where I am. So the thoughts are distinct by the Fregean intuitive criterion of difference. Alternatively, I might believe that there are two cats in my office, and so deny that the cat in my office is lost, while continuing to believe that that cat is lost.
E.g. (2): the thought The cat over there is lost. For again, since in a hall of mirrors I might wonder, Is that cat over there?, I might doubt whether the cat over there is lost while continuing to believe that that cat is lost.
E.g. (3): the thought The cat now causing these very experiences is lost (contra Searle, 1983, ch.8). For while it seems implausible that I could ever doubt whether that cat is causing these experiences (that is, the experiences which now ground my demonstrative reference to that cat), it seems quite wrong to make all demonstrative thoughts involve reference to ones current experiences. For my experiences are not, normally, an object of attention in such cases. And indeed, it surely seems possible for someone (a young child, or an autistic person, say) to entertain the thought, That cat is lost who does not yet have the concept of experience.
These points give rise to an argument against the very coherence of narrow content. For narrow contents are supposed to be available to be thought, whether or not their putative worldly objects exist or are present. But if it turns out that in the absence of an object there is no way of stating the content of the putative singular thought, then it seems that there can be no such world-independent content.
3 The coherence of narrow content
In this section we take up the challenge presented by the argument outlined above. Note that it is a suppressed premise of the argument that if a thought-content exists at all, then it can be specified by means of a that-clause. That is to say, it is assumed that, if a singular thought really is entertained in the hallucination case, then it must be possible for us to say what thought is entertained by means of a phrase of the form, He is thinking that such-and-such. This assumption is tacitly rejected in the alternative proposals for specifying narrow content considered in Section 3.1 below. It will then be explicitly examined and criticised in Section 3.2.
Fodor (1987, ch.2) acknowledges that we cannot express a narrow content directly, using a that-clause; because any such clause will automatically take on one or another wide content (that is, truth-condition). But he thinks we can (as he puts it) sneak up on narrow contents, providing such contents with an indirect characterisation. Fodor maintains, in fact, that narrow contents are functions from contexts to truth-conditions. Thus the narrow content which I and my twin share when each of us says, Water is wet, is that unique content which, when anchored on Earth has the truth-condition, H2O is wet, and when anchored on Twearth has the truth-condition, XYZ is wet. (Most of Fodors discussion concerns natural-kind examples.) Similarly, the narrow content which we share when each of us says, That cat is dangerous, is the unique content which, when anchored in the context of Tiddles has the truth-condition, Tiddles is dangerous, and when anchored in the context of Twiddles has the truth-condition, Twiddles is dangerous.
Notice that Fodors approach makes narrow contents entirely parasitic upon wide content indeed, upon wide content conceived of purely in terms of truth-conditions, or worldly states of affairs. For Fodor will have no truck with Fregean senses, or modes of presentation of truth-conditions. In fact, there is nothing more to any given narrow content than its being that state which, when embedded in one context yields one truth-condition, and when embedded in another context yields another. Indeed, as we shall see in the next chapter, Fodors project is to offer a naturalistic wide-content semantics, characterising meaning and reference in purely causal terms, and then to construct a notion of narrow content to ride piggy-back on that. Why is Fodor so minimalist about the nature of narrow content? In part because of an obsessive fear of holism, as we shall see in due course. If one said more, of an intra-cranial sort, about what makes any given narrow content the content that it is, this would presumably have to work by relating that content to others (what further beliefs that content may lead the thinker to by inference, for example). But then there may be no way of stopping short of saying that the (narrow) content of any one belief will implicate all the subjects other beliefs. This is a consequence Fodor is keen to avoid.
The main problem with Fodors minimalist approach, however, is that it fails to give us a notion of content which satisfies the Fregean intuitive criterion of difference. Thus, let us return to an earlier example, comparing the two thoughts, Pavarotti is fat and That man is fat. These are plainly distinct, by the intuitive criterion, since I might doubt the one while believing the other, or vice versa. But they come out as possessing the same narrow content, on Fodors account (and so as having the same wide content too, of course). For, of each of these thoughts you can say that it is the thought which, when embedded in a context containing the singer Pavarotti, in such a way that the referential element of the thought is causally connected to that person, it has the truth-condition, Pavarotti is fat. Both thoughts end up with a truth-condition which attributes fatness to one and the same man. Similarly, consider the two thoughts one might express by saying, That man is well paid (both, again, involving reference to the singer Pavarotti), where the one is grounded in vision and the other in hearing. By the intuitive criterion these should come out as distinct, since one might of course doubt whether that man (seen) is that man (heard). But by Fodors account they will come out as the very same, since the same function from contexts to truth-conditions is instantiated. The two thoughts are such that, whenever the demonstrative elements are caused by one and the same person, then they have the same truth-condition. Bad news for Fodor, we say.
Carruthers (1987a) proposal is initially somewhat similar. It is that we can describe the narrow content entertained, in the case of hallucination (for example, of a dangerous cat), by exploiting the alleged identity of narrow content across contexts. We can say, in fact: He is entertaining a thought with the very same (narrow) content as he would have had if there had been a real cat there causing his experiences, and if he had entertained a demonstrative thought, concerning the cat, that it is dangerous. And we can say what is common to Petere and Petertw when each entertains a thought they would express with the utterance, That cat is dangerous, by saying: Each is entertaining the very thought they would have whenever there is a cat in front of them, causing their experiences in such a way as to ground a demonstrative thought, and they think, of the perceptually presented cat, that it is dangerous. Notice that these accounts are not reductive they do not attempt to reduce demonstrative thoughts to something else. Rather, they just describe the content of one token demonstrative thought by specifying it as being identical to the (narrow) content of another.
Now in one way, of course, this proposal can seem like a cheat. It merely uses a claimed identity of narrow content in order to describe the content of a target thought, without attempting to tell us what narrow content is, or what the conditions of narrow-content identity are. Nevertheless, the proposal is, we claim, sufficient to rebut the charge of incoherence levelled against narrow content the charge that, in a case of hallucination, there is no way to describe the (putative) content of the singular thought entertained. On the contrary, there is such a way, and we have just given it. Moreover, the proposal leaves open the possibility of a more substantive account of narrow content (in a way that Fodors proposal does not). It might be said, for example, that the narrow content of the demonstrative element that man, when grounded in a visual presentation, is given by the location in egocentric space at which the man is represented. So all tokenings of the thought, That man is fat, provided they represent the man in question in the same position in the thinkers egocentric space, will count as having the very same narrow content, irrespective of any further differences between the men and their circumstances. Of course, this is just one highly debatable proposal. But it illustrates how the proposed approach to narrow content might admit of further supplementation.
Note, too, that on both of the above ways of characterising narrow content (Fodors and Carruthers), narrow content is not really a kind of content at all, if by that you mean something which has to have a unique semantic value (true or false). For narrow contents do not, in themselves, have truth-conditions, and are not, in themselves, about anything. And it makes no sense to ask whether a narrow content, as such, is true or false. Only when embedded in a particular context do narrow contents come to have truth-conditions. Nevertheless, an individual tokening of a given narrow content will normally have some particular truth-condition. Hallucinatory cases aside, every time someone thinks, That cat is dangerous, their (narrow) thought comes to have some or other truth-condition, through its embedding in a particular context. So one way to put the point is that it is, properly, narrow content tokens, rather than narrow content types, which have truth-conditions, and which are the bearers of truth-values.
The assumption which was implicit in the argument for the Russellian (object-involving) status of singular thought which we discussed in Section 2.2, is that any genuine content has to be specifiable in a that-clause. The responses of Fodor and Carruthers considered above take it for granted that this assumption is false. Here we shall argue that it is. But note, first of all, that the assumption is quite widespread in philosophy. Thus, you will find people arguing that dogs and cats do not really have beliefs (Davidson, 1975), on the grounds that we cannot describe their beliefs using our concepts we cannot say, for example, The cat believes that the bird is edible, since the concept bird has many conceptual connections (for example, to living thing) which we might be loath to attribute to an animal. Yet the assumption in question appears to be wholly unmotivated.
Obviously this is so if (like us) you are a realist about propositional attitudes, thinking that beliefs and desires are what they are independently of our descriptions of them. But the same is surely true even if, like Davidson, you are an interpretationalist about the attitudes, maintaining that there is nothing more to being a believer/desirer than being a creature whose behaviour can be successfully predicted and explained by suitable attributions of beliefs and desires. For why insist that the descriptions used to generate predictions and explanations have to be couched in the form of a that-clause? We do agree that being able to specify the contents of others intentional states in this way greatly enhances our predictive and explanatory powers. For, as we noted above in Chapter 4 (Section 3), attribution of a belief-content allows us to run a simulation on the content and so, by inferential enrichment, attribute many other beliefs and we can only run a simulation when we have a complete content to insert into our own (off-line) inferential systems. By contrast, our mapping of the beliefs of animals and very young children is patchy and incomplete. But even if we cannot predict what they will infer, we can still predict and explain some of their actions on the basis of such a description of their beliefs and desires.
When it is claimed that any genuine thought must have a content specifiable in the form of a that-clause, there are three things which this might mean.
(1) It might mean specifiable by someone, sometime. But then this would be a principle without teeth, unless we simply beg the question against narrow content. For of course the people on Twearth can express their thought in a that-clause, by saying, We think that water is wet. And the person hallucinating the presence of a cat can similarly say, I think that that cat is dangerous. Unless we just assume that narrow contents do not exist, it is hard to see why this should not count as a genuine description of the content of their thought.
(2) It might mean specifiable by us, now. But then this conflicts with the obvious truth that there are people who entertain thoughts whose contents I cannot now share (and so whose contents I cannot now express in a that-clause), because I lack some of the requisite concepts. It is surely obvious that there will now exist many perfectly genuine thoughts whose contents I cannot now express, entertained, for example, by scientists in disciplines of which I am ignorant.
(3) It might mean specifiable by us, in principle. This answers the point about scientists in (2) above, since I can presumably learn their theories and acquire their concepts, and then I could describe their thoughts using a that-clause. But it still runs into trouble in connection with the thoughts of animals, since it seems likely that I cannot, even in principle (while retaining my status as a sophisticated thought-attributer) acquire the concepts which a cat uses to categorise its world. Since the common-sense assumption that cats do have thoughts works pretty well, there had better be some powerful independent argument if we are to give it up. But in fact there is none.
So none of the available proposals is at all attractive. In fact, as Fodor (1987) points out, the reason why we cannot describe the singular thought of our cat-hallucinator using a that-clause is simple, and trivial. It is that, since we do not believe in the existence of the cat, we cannot describe the hallucinators thought by using a demonstrative within the scope of a that-clause (nor, indeed, by using any singular concept). We cannot say, He thinks that that cat is dangerous, since this would require us to entertain, ourselves, a demonstrative thought about the putative particular cat. Similarly, the reason why we cannot use the term water within the scope of a that-clause to describe the thoughts of Petertw, is that the reference of that term, in our mouths, is of course tied to the constitution of the stuff on Earth. There is, surely, nothing of deep significance about the nature of content to be derived from these facts and certainly not a refutation of the coherence of narrow content.
4 Explanation and causation
In this section we examine the respective roles of wide and of narrow content in psychological explanation, asking whether either or both might be causally relevant to behaviour, and concluding that only narrow content is genuinely causally explanatory. But we begin with an argument against wide content, from its failure adequately to explain the behaviour of a hallucinator.
The thoughts of our cat-hallucinator turn out to give rise to a powerful argument against the ubiquity of wide content. Recall that it is a consequence of the wide-content theory, as applied to the case of singular thought, that in a case where there exists no actual object of thought (for example, through hallucination or misinformation), then there exists no singular thought either. For singular thoughts are supposed to be Russellian, being partly individuated in terms of the objects thought about. In such cases people are said to essay, or attempt to entertain, a singular thought of a certain type, but to fail.
Thus, compare two examples: in the one case I am really confronted by a cat, which I perceive and believe to be vicious; I think, That cat is dangerous, and lash out at it with my foot. In the other case everything is, from my subjective perspective, exactly the same, and issues in an exactly similar bodily movement, except that there is really no cat there; I am merely hallucinating. Wide-content theorists will say that in the first case I do entertain, and act on, a singular thought, but in the second case I do not; it merely seems to me that I have done so. Narrow-content theorists will say, in contrast, that in each case I entertain the very same type of thought, which explains my action in virtue of instantiating the same psychological (content-involving) law the law, namely, that whenever people take themselves to be confronted with something dangerous, then they will, ceteris paribus, take action to deflect or avoid that threat. This certainly accords well with the intuition that the two cases are, psychologically speaking, alike.
The immediate problem for the Russellian is to explain how, in the hallucination example, my movement is genuinely an intentional action, which admits of a rationalising (that is, content-involving) explanation. For how can an action which is not done for a reason (not caused by a thought) really be intentional? Yet, in the case in question, it surely is. My lashing out with my foot was certainly not a mere reflex, like a knee-jerk, but an attempt to achieve something. Now, the response usually made by Russellians is that there are plenty of other (non-singular) thoughts still available to me, grounded in my hallucination, which can still serve to rationalise my action. Thus, I will still have such general beliefs as, I am confronted by a dangerous cat, There is a cat over there, and so on. And I can then be said to act because of these beliefs, in order to deflect a believed threat.
There are two problems with this response. The first is that it rides rough-shod over the (putative) distinction between actual (or core) and merely dispositional beliefs. This distinction may be needed to explain how we can have infinitely many beliefs, consistent with our finite cognitive space. (I say something true of you when I say that you believe that 1 is less than 2, that 1 is less than 3, that 1 is less than 4, and so on indefinitely.) What may really be the case is that we have a finite number of actually existing beliefs, represented and stored in some fashion in the brain; and from these beliefs we are immediately disposed to deduce any number of further beliefs, as the situation demands. Now, in a case where I see a cat and think, That cat is dangerous, it seems perfectly possible that beliefs such as, I am confronted by a dangerous cat are merely dispositional. That is, I would immediately assent to them if asked, but have not actually computed and stored them. The Russellian, however, must deny this. For a belief which remains merely dispositional cannot be a cause. If the general (non-singular) belief, I am confronted by a dangerous cat is to explain my behaviour, then it must first have become actual. So the Russellian must maintain that we routinely actualise a great many more beliefs than we appear to which, although possible, is otherwise unmotivated.
The second and stronger objection to the Russellian response is this. Even if the belief, I am confronted by a dangerous cat was in some way activated, it certainly did not figure as a conscious judgement. The only (putative) thought which I consciously entertained was the singular one, That cat is dangerous. So, if the Russellian is right, my act of kicking was caused by non-conscious thoughts only. And now (quite apart from the intuitive implausibility of this suggestion) the Russellian has a real problem. For it must then be said that in the veridical case too, where there really is a cat present, my action is caused by non-conscious thoughts only. (Either that, or it is causally overdetermined). And then it is hard to see how we can avoid the consequence that my actions are never caused by conscious singular judgements, but only ever by non-conscious general ones. And that, surely, would be absurd.
The only other option, for the Russellian, is to claim that it is not really thoughts, but rather thought-signs (sentences, or sentence-like objects) which cause actions. And then the pattern of causation in the two cases can be the same. For in the hallucination case it need not be in doubt that I do entertain a thought-sign of some sort. For example, I might entertain in auditory imagination the English words, That cat is dangerous (see Chapter 8). Or as Fodor has claimed in another context (1994), it may be that it is signs of Mentalese which are the only intra-cranial components of (widely individuated) thoughts. The Russellian merely claims that, in context, these signs do not express any complete content. But the trouble with this is that it pitches the explanation of my actions at the wrong level. Even if, at some level of description, our actions are caused by sentence-processings (as the computational model of the mind maintains, indeed; see Fodor, 1980), we also think that they are caused, at a higher level of description, by thoughts psychological states with intentional content. And it is this that Russellians cannot accommodate, if they take this final option.
(Although Fodor was once a champion of narrow content, in his 1994 he proposes to use wide content plus modes of presentation in the shape of sentences of Mentalese to do the explanatory work he had previously assigned to narrow content. But, despite the case in favour of a language of thought for which see Chapter 8 below this manoeuvre fails to preserve the right sort of psychological explanation. Oedipus was not horrified that he had made love to Jocasta, but now he is horrified that he has made love to his mother. The explanation for why he puts out his eyes must surely advert to the fact that he has realised that Jocasta is his mother, and thus realised that he has committed incest not just that he has come to have some new sentences of Mentalese which refer to his mother tokened in his brain.)
Thus far in this chapter we have defended the coherence of the notion of narrow content, and have argued that the examples of singular thought attempted in cases of hallucination present a powerful challenge to a wide-content theorist. We now turn explicitly to questions concerning the respective roles of wide and narrow content in psychological explanation (both folk and scientific).
Return to the Twin Earth examples. Someone might argue thus: since, by hypothesis, the behaviours of Petere and Petertw are exactly the same, we should look for the same explanations of those behaviours too that is, we should ascribe to both Peters the very same behaviour-determining thoughts. Thus suppose that each of the two Peters is confronted by a glass containing a colourless liquid, and that each thinks a thought they would express with the words, There is still some water left in that glass, and consequently lifts the glass to drink from it. Since the behaviour is the same in each case, we might think that the explanations we advance of that behaviour should be the same too which means not individuating the thoughts in terms of the inner structure of the natural kinds in question, but rather narrowly, independently of the actual environment.
Of course the background principle appealed to here is not a hard-and-fast one. For we know that there can be cases of convergent causation. That is, there can be cases where instances of the very same event-types are caused by quite different routes. This is especially familiar in the case of human action, since examples where people behave similarly but for very different reasons are rife. Thus, consider the variety of reasons people might have for writing to apply for a particular job one because he needs a job, and any job would do; another because she wants that particular job; another because he wants to please his mother; and so on. Yet the behaviour in each case is of an identical type (in some respects).
All the same, whenever two systems are changing and evolving in such a way as to follow exactly similar trajectories, we surely have powerful reason to believe that the underlying causal processes must be the same. Thus imagine two ropes being tested in a companys testing laboratory: each begins to fray in the same place after exactly the same amount of time, and then each snaps in the same place, again at the same time. Surely these facts would give us reason to believe that the intrinsic properties of the two ropes were the same, and that they were subjected to the same forces throughout. Otherwise we would have to believe that the similar effects were a mere coincidence. Moreover, the more complex the effects in a pair of parallel sequences, the more unlikely the coincidence. And remember that in the Twin Earth examples, all the behaviours of the two Peters are the same over an indefinite time-span!
There is an obvious reply that defenders of wide content can make to the above argument. They can deny that the behaviours of Petere and Petertw are the same (under an intentional description). And if they do not really behave in the same way, then there need be no presumption that their behaviours should be caused by thoughts of identical types. Thus, consider what it is that they do when they lift and drink from the glass: while Petere drinks water (H2O), Petertw drinks twater (XYZ). And these can be counted as belonging to two different action-types. So it can, in effect, be objected that the argument above presupposes content-neutral (non-intentionally described) descriptions of behaviour arm-movings, glass-liftings, and so on, but not water/twater-drinkings. In which case that argument seems just to beg the question at issue. For if thoughts are widely individuated, then so too will a persons intentions be; and then so will their intentional behaviour.
A similar point holds in connection with singular thought. Consider the case where Mary and Joan think, That table is greasy, and each reaches for a tissue to wipe it. Or consider the case where I and my twin each sees a cat, thinks, That cat is dangerous, and lashes out at it with a foot. While it might initially seem that in the two types of case we are dealing with two instances of the same behaviour (table-wiping and cat-kicking respectively), which should then receive the same (narrow) explanations, in fact the behaviours can be categorised as different. For Mary wipes this table while Joan wipes that one. And Petere kicks Tiddles while Petertw kicks Twiddles. So if singular thought is relationally individuated, in such a way as to embrace the actual objects thought about, then actions guided by such thoughts, under an intentional description, will come out as relationally individuated too. And then the argument above collapses.
Notice, however, that this response by the externalist places questions of sameness and difference of behaviour, and of sameness and difference of psychological explanation, in hock to scientific discovery, in a way which may seem unpalatable. For suppose it had turned out that water (somewhat like jade) is differently constituted in different parts of the globe. In that case Mary, in England, and Kylie, in Australia, might have been engaged in different behaviours when reaching for a glass of water, even in advance of the discovery of the difference. (We assume that the externalist must say that if water in England is H2O, but in Australia is XYZ, then the word water refers to different substances when used by Mary and Kylie respectively.) And when we explain those behaviours by saying, She wanted a drink of water, the explanations would have been different too, attributing thoughts of a different type. So the question of how many types of psychological explanation there are depends on the question of how many types of water (and other natural kinds) there are which is possible, perhaps, but somewhat hard to swallow!
People sometimes assume that if content is individuated narrowly, then we would have to resort to individuating behaviour in terms of the bodily movements involved. But this is not so. There are a variety of ways of classifying behaviour, depending upon purpose and context. Sometimes we need to classify behaviour in terms of agents intentions, narrowly construed as, for example, to distinguish the pursuits of knights in search of the Grail, alchemists in search of the Philosophers Stone, and contemporary hunters of the Loch Ness monster, even if all alike are on a wild-goose chase. Our cat-hallucinator and someone actually confronted by a fearsome feline may equally run away out of fear of a dangerous cat, for all that one cannot distance oneself from something which is not there. Parts of our vocabulary for describing actions have wide commitments, other parts do not. So you cannot mine gold unless gold gets mined; although you can try to mine gold in a place where there is nothing but fools gold, and you could prospect for gold in a world in which there was no such stuff at all. The vocabulary appropriate for classification of behaviour will depend upon whether our interest is focused on the agents or on their environment, on psychological explanation and prediction, or on the acquisition and communication of other facts (see Section 5 below, for the related distinction between explanatory and semantic content).
Let us try another tack. Consider the way in which physicalism about the mental is often expressed: by claiming that mental states supervene upon brain states. It is often said that there can be no differences at the level of the mental, without some corresponding differences in the brain. If two people have distinct mental states, then there must it is alleged be some other (physical) difference between them (presumably in their brains) which explains the difference. In contrast with mental/physical dualism, we no longer accept that mental facts can float free of physical facts. On the contrary, almost everyone today is a physicalist.
This now gives rise to an argument for narrow content. For brain states are surely not individuated relationally. No one would want to maintain that Petere and Petertw are in two different brain states, merely on the grounds that the one has water in his environment whereas the other has twater in his. Equally, no one would want to say that Mary and Joan must be in distinct brain states, merely on the grounds that the tables confronting them are numerically distinct. So, if brain states are non-relationally (that is, narrowly) individuated, and mental states supervene on brain states, then mental states must be narrowly individuated as well. For otherwise there would be (relational) mental differences without any corresponding brain difference.
On reflection, however, this argument, too, just begs the question in favour of narrow content. For if mental states aint (entirely) in the head, as wide-content theorists maintain, then, plainly, mental states will not supervene on brain states alone. Rather, they will supervene on brain states together with relational facts. This can still be fully consistent with physicalism, provided that those relational facts are themselves physical ones (as, indeed, they are).
A more promising strategy for a narrow-content theorist is to appeal to the thought that mental states should supervene upon causal powers. For from the standpoint of explanatory psychology, we are only going to be interested in differences amongst mental states which reflect differences in their causal powers. And where the causal powers of two token mental states are identical, we shall therefore want to regard them as being of the very same type. (We understand causal powers here to include the potential causes as well as the potential effects of the state in question. We also assume that psychology is not unlike geology, for example an historical science; that is, it does not individuate the kinds with which it deals in terms of their actual causal history. See Chapter 7 for further discussion.)
Now the notion of a causal power is counterfactual-involving. To talk of the causal powers of state S is to talk, not just of what S actually causes, but also of what S would cause (or be caused by) in various hypothetical and counterfactual circumstances. Seen in this light, it is obvious that the causal powers of the states of the two twins are the same. For if Petere were to be on Twin Earth, then he would behave exactly as Petertw does (even under an intentional description); and if Petertw were on Earth, he would behave exactly as Petere does. Similarly, if Mary were to be sitting where Joan is, then she would be behaving as Joan does, and vice versa. In fact, it is the causal powers of mental states which supervene on (non-relationally described) brain states. Then if we insist that mental state types should supervene on causal powers, it will follow that content is narrow. For otherwise there would be (relational) differences between mental states (widely individuated) which would not reflect differences in their causal powers.
This looks as if it might become a powerful argument in support of narrow content. But why should one accept that mental states are only distinct where their causal powers are? This will follow if we think that mental states are, basically, the theoretical posits of an explanatory proto-science (that is, if we accept some or other version of the theory-theory of mental states, as we argued in Chapter 4 we should). For science, in general, types entities and states by their causal powers, taking no interest in differences between states not reflected in differences in their causal powers. (At least, this is true of sciences which are a-historical.)
No doubt folk-psychology may be more than a proto-science, and may also take an interest in (merely relational) differences amongst mental states not reflected in their causal potential. (Indeed, we will argue as much in Section 5 below.) But to the extent that folk psychology is at least attempting to do the work of a scientific theory typing states by their causal powers, and explaining events as caused by the states so distinguished to that extent we have reason to categorise thoughts narrowly, in terms of a non-relationally individuated notion of content. Moreover, if we are to extract from folk psychology a notion of content fit to subserve a content-based scientific psychology, then it would appear, from the arguments above, that the notion extracted had better be a narrow one.
Supposing that there are some psychological (content involving) laws (or nomic tendencies, at least), what can be concluded about the notion of content likely to figure in those laws (or tendencies)? Much, of course, depends upon what kinds of law may be in question. Some putative psychological laws operate by quantifying over content, for example; in which case nothing much can be concluded about the nature of such content. Thus, consider the practical reasoning syllogism:
(x)(P)(Q)(if x wants that P, and x believes that, by bringing it about that Q, x can succeed in bringing it about that P, and x believes that it is now within xs power to bring it about that Q, then ceteris paribus x will act in such a way as to try to bring it about that Q).
It does not seem as if this can throw any light on the nature of the contents P and Q. But some putative psychological laws will involve particular contents or types of content such as the law that the moon looks bigger near the horizon; or that people have an aversion to mother-son incest; or that people will (ceteris paribus) act so as to avoid or deflect a perceived threat. Here, plainly, the contents involved had better be typed narrowly, if the laws are to achieve the requisite generality. For example, if the law of threats involves a content such as, That is a threat to me, then this content will need to be individuated non-relationally, so that many different thinkers, entertaining demonstrative thoughts about many different things, can nevertheless be encompassed by the law. So to re-iterate: if there is a notion of content which gains its life and significance from the way in which it figures in (putative) psychological laws, then there is good reason to expect that notion to involve narrow (non-relational) principles of individuation.
It may be objected that at least some scientific laws appeal to properties which are individuated relationally. Consider the scientific discovery that malaria is caused by mosquito bites, for example. Here we have a law (or nomic tendency) relating the property of suffering from malaria, on the one hand, to the property of having a bite which was caused by a mosquito, on the other that is to say, a property which is individuated by its relation to another thing (the mosquito). So why should not psychology, similarly, formulate its laws in terms of properties of the agent individuated by relation to things external to the agent? But in fact there is no nomic connection between mosquito bites and malaria. The relevant law will relate malaria to the presence of a certain sort of parasite in the bloodstream. And it just so happens that the normal causal route by means of which those parasites enter the bloodstream is a bite from a mosquito. But it is a real possibility that in the course of evolution those parasites might come to be transmitted by other secondary hosts, apart from mosquitoes. The relationally-formulated law is not really a law at all, but rather a generalisation which lines up more-or-less usefully with the genuinely (and non-relationally-individuated) nomically connected properties.
It is deeply embedded in our common-sense, or folk, psychology that our reasons are causes of our actions. We think we normally act as we do because we believe this and desire that, or because we intend to achieve the other. But reasons, of course, are propositional attitudes with content, partly individuated in terms of their content. A belief is always a belief that P, and a desire is (arguably there is an issue as to whether desires for particular objects can always be analysed as desires for the truth of some corresponding proposition) a desire that Q. So when we believe that reasons are causes we believe that states individuated in terms of their content are causes.
But now the problem for the wide-content theorist is this: if contents, in turn, are relationally individuated, in terms of objects and properties external to the subject, then how can the content of a mental state be a causally relevant feature of it? For surely causation is, in general, local, mediated by intrinsic (non-relational) properties of the events and states in question. How can the fact that a state stands in a certain relation to something, which may be distant from that state in space and time, be a causally relevant feature of it, partly determining its causal powers? Admittedly, there do exist examples of relationally-individuated properties which are causally relevant. We have just discussed the causal relevance of the relationally-individuated property, being a mosquito-bite. Now consider the property, being a planet. This is, plainly, a relational property: to be a planet is to stand in a certain relation to a sun. Yet standing in that relation is one of the determinants of the causal powers of planets. This case is easy to understand, since the relation in question is correlated with the existence of a causal force (namely gravity) which acts on any planet qua massive object. There is nothing similar to help us in connection with widely-individuated mental states. How can the mere fact that the stuff in the lakes and rivers in my environment is composed of H2O rather than XYZ, for example, make any difference to the causal powers of my belief that water is wet?
Some naturalistic accounts of wide content (particularly the version of informational semantics due to Dretske, 1988) are designed, in part, to overcome this problem. On Dretskes account, mental states have the contents which they do in virtue of the information that they carry about the environment (where information is a causal notion). Content can then come to be of causal relevance, provided that the mental state in question becomes harnessed to the control of a particular type of behaviour (either through evolution or through learning) because of the information that it carries. But it is obvious that this solution to the present problem must be unsuccessful. One reason is that the behavioural success of my water-thoughts, for example, has nothing to do with the fact that they carry information about H2O (as opposed to XYZ), but rather with the fact that they carry information about waters properties of potability, solvency, and so on properties, note, which are equally shared with XYZ. Another reason is that many singular thoughts (widely individuated) are one-off. If I think That cat is dangerous and act accordingly, then there can be no historical explanation of the causal powers of the thought in terms of the information which it carries about that particular cat, since I may never before have encountered that cat, nor entertained that thought.
Some wide-content theorists have replied that there is really no problem here for them to answer (Klein, 1996). For on most accounts of wide content, the relation in question, in terms of which the content of a state is partly individuated, is itself a causal one. (This is true in connection with all varieties of what McGinn calls strong externalism. See his 1989.) Thus, when I think, That cat is dangerous, for example, my thought comes to have the wide content which it does in virtue of the causal relationship which obtains between my tokening of the thought and that particular cat. In fact, to individuate thoughts widely is to individuate them in terms of their causes, on most accounts. And then, it may be said, the (wide) content of a thought must, after all, be causally relevant. For the cause of a cause must be causally relevant to the latters effects. If my thought about the cat explains my attempt to kick it, and my thought is caused by the presence of a particular cat, then that particular cat is causally relevant to my kicking. And then to individuate my thought in terms of its causation by that particular cat is to individuate it in a way which must be causally relevant to the kicking.
This reply fails, however. For it does not show that wide content is relevant to the distinctive causal powers of (as opposed to the mere existence of) a thought. Individuating states in terms of their causes does not automatically mean (indeed, will normally not mean) individuating them in a manner relevant to their causal powers. (Note that by a causal power, here, we mean a capacity to bring about certain effects. Our topic now is whether reasons, as such, are causes; not whether reasons have causes.) Consider, for comparison, the concept chair. To oversimplify somewhat, this, too, individuates items in terms of their causal history: to be a chair is to be an object which was caused to exist by someones intention to produce something for sitting upon. Does this mean that chairhood is causally relevant to the effects which any given chair has? Surely not. If I trip over a chair in the dark and break my leg, then it is not because it is a chair that I break my leg it is not because the object I trip over was created with a certain intention in mind. Rather, the cause of the break is that I caught my foot on an object of a certain mass, rigidity, and shape. The fact that it was a chair which had those properties is causally irrelevant. (The point, here, is essentially the same as the one made above concerning the causal relevance of the mosquito in relation to malaria the causal powers of the bite depend upon its intrinsic properties, not its causation by the mosquito.) Then so too, it seems to us, in connection with wide content. The fact that my thought was caused by one particular cat rather than another, or by a sample of H2O rather than of XYZ, is irrelevant to its causal powers. And then to individuate mental states widely, in terms of their extra-cranial causes, is to individuate them in a way which is irrelevant to the causal status of the mental.
Peacocke (1993) has argued that what wide contents (contents relationally described) explain, are relational properties of movements. Thus, one and the same movement of my hand can be both a movement towards someone in the garden, and a movement northwards. But only the former is explained by saying that I wanted to draw your attention to that persons presence. For different counterfactuals are sustained. If that person had been in a different position in the garden then I would still have pointed towards him (given that I perceive him), but I would no longer have pointed northwards. Peacocke claims that only wide contents can give us this pattern of explanation of relationally-described movements by relationally-described mental states; only wide contents give us the right set of counterfactuals.
We have two points to make. The first is that sustaining counterfactuals is not the same thing as being a cause. For example, imagine a wave breaking on the seashore, destroying as it does so a particular sand-castle. And suppose that breaking waves always produce surf (a film of bubbles on their breaking edge). Then the following counterfactuals are true: (a) if the surf had not been present, then the sand-castle would not have broken; (b) if the sand-castle had not been broken, then the surf would not have been present. But the surf is not the cause of the destruction. Rather, the sand-castle is destroyed by the wave, which also causes the surf. So, the fact that wide contents sustain counterfactuals does not show that such contents are causes. Rather, the wide content may just supervene in a law-like way on what really does do the causing (that is, a narrow content which happens to have a particular worldly cause).
Our second point (see also Segal, 1989a) is that a narrow-content explanation, supplemented by relational facts, can sustain the same set of counterfactuals. Set out in more detail, the Peacocke example is this:
(1) I see a person in the garden;
(2) I want to draw your attention to him;
(3) so I move my hand in his direction.
And his point is that this explanation works whether or not anything more is known of the spatial relationship between myself and the person in question (for example, whether or not you know where I was in the room at the time). And it is true that if he had been in a different position in the garden then, provided that (1) and (2) remain true, I would have moved my hand in that direction instead.
Yet these features of the explanation can, surely, be replicated in a narrow-content account. A narrow-content explanation of the case would be this:
(i) I experience a person-as-represented in a particular direction in egocentric space (a content I can entertain whether or not it is really that particular person, or indeed anyone, who is there);
(ii) I want to draw your attention to the presence of the person I represent;
(iii) a particular person is in fact the veridical cause of the experience in (i);
(iv) so I move my hand in his direction.
Here, too, the explanation works whether or not anything further is known about where the person and I are. And here, too, the right counterfactuals are sustained: if he had been in a different position then, provided that (i), (ii) and (iii) remain true, I would have moved my hand in that direction.
The advantage, indeed, is firmly with the narrow-content explanation of the case. For Peacocke will be forced to postulate three psychologically distinct explanations for the cases where (a) I perceive the person in the garden, (b) I perceive not him but his identical twin brother, and (c) where I am hallucinating; for I shall be said to entertain different thoughts in each case. But the narrow-content theorist can advance exactly the same form of explanation for (a) and (b) the only difference being that a different person will be picked out in clause (iii). Moreover, explanation (c) will only differ in that clause (iii) is dropped altogether which gives us just the right counterfactuals, since all that is then relevant is where my experience represents a person as being. So on a narrow-content account the psychological aspect of the explanation will be the same for all three cases.
A wide-content theorist may reply to the difficulties we have been raising, by saying that there is no special problem in explaining how wide contents, qua contents, can be causes, since essentially the same problem will arise in connection with all conceptions of content. For after all, as physicalists we must believe that all bodily movements will have sufficient causes at a neurological level brain events causing brain events, causing muscles to contract, causing arms to move in certain directions, and so on. So how can there be any space for reasons to be causes as well, unless reason-descriptions are just alternative ways of describing brain events? In which case it will not be qua reason that a given brain event is a cause.
There are a number of different possibilities for responding to the allegation that all content must be epiphenomenal. The most direct (and, to our minds, the most convincing) is to point out (1) that reasons will be causes in virtue of their content if they figure in a distinctive set of content-involving causal laws; and (2) to claim that there are, indeed, such laws. The first part of this response is relatively uncontroversial. For there are few who are prepared to claim that the only real causes which exist are at the level of sub-atomic physics. Yet there is exactly the same sort of reason to claim that all processes of whatever level, and whatever degree of complexity must be realised in sub-atomic physical ones. At any rate, the second part of the above response, if true, would show that reasons have the same sort of causal status as genes, or H2O molecules, or any other natural kind above the level of basic physics. So: is claim (2) true? It would certainly appear that there are many content-involving laws, ranging from the highly particular (The moon looks bigger near the horizon; People have an aversion to the thought of mother/son incest) to the general (People try, ceteris paribus, to get what they want and note that all laws above the level of basic physics are ceteris paribus). And if the arguments above are sound, such laws will only employ, and vindicate, contents which are narrowly individuated.
Suppose it had turned out that folk-psychological content is wide, whereas scientific-psychological content would have to be narrow; and suppose, too, that it had turned out that narrow content is actually incoherent; then this would have meant that the prospects for an intentional scientific psychology were bleak. It would have meant that folk psychology is the only kind of intentional psychology we could ever have. Although we have, indeed, argued that scientific-psychological content should be narrow, we have denied that the notion of narrow content is incoherent. So there is no threat to scientific psychology from this quarter. But is there, now, an eliminativist threat to folk psychology? If folk psychology is committed to wide content, but science tells us that content is narrow, does that mean that folk psychology is in error, and should be replaced? That all depends, plainly, on what it is that folk psychology is trying to do.
What, then, is our common-sense notion of content? Many of those who defend narrow content do not think that it is, or is a component of, our common-sense conception. Thus Fodor (1987) thinks that the folk-psychological notion is wide indeed purely referential, dealing only in worldly properties and individuals. (This is then a notion of content which Russell himself would have been entirely at home with.) But he also thought (he is no longer so sure; see his 1994) that it is imperative that we should be able to construct a notion of narrow content to serve as the basis for a scientific psychology. Others (Burge, 1991) think that a notion of narrow content is indeed legitimate, but forms no part of our actual common-sense psychology maintaining that our actual notion of content is a Russellian hybrid of reference plus Fregean mode of presentation. We think, in contrast, that narrow content does (or should) form one strand in our common-sense notion (a strand which Carruthers has elsewhere labelled cognitive content see his 1989; here we adopt the terminology of explanatory content), the other being purely referential (labelled semantic content).
5.1 Two kinds of content
We claim that there are two very different perspectives which we can, and regularly do, take towards the contents of peoples thoughts that there are two distinct kinds of interest which we can and do take in descriptions of thought-content, each of which motivates a different set of identity constraints. Sometimes our interest in thoughts and thought-ascriptions is either explanatory or predictive. Often our main interest in the thoughts of other people is to use them in such a way as to explain what those people have done, or to predict what they will do. And often, from this perspective, it will be crucial to know the precise way in which the thinker conceptualises the subject matter it can make all the difference in explaining Oedipus remorse whether the content of his thought is described as I am married to Mother or as I am married to Jocasta. So our principles of individuation will at least need to be Fregean, requiring identity of mode of presentation for identity of thought-content. But equally, from this explanatory and predictive perspective we are generally not interested in the truth or falsity of the thoughts ascribed. So the principles of individuation can abstract away from the actual worldly referents of the component concepts of the thought such content can be narrow.
Sometimes, on the other hand, our interest in the thoughts of others is communicative, or belief-acquisitive. Often our perspective on the thoughts of others is that their thoughts may give us something which we ourselves may wish to believe or deny. And here Fregean modes of presentation are of no relevance, we maintain. All that matters is that we should get hold of which worldly objects and properties the persons thoughts concern. So such contents are purely Russellian. It seems to us that the popular view that our common-sense conception of content is a Russellian/Fregean hybrid comes from conflating these two perspectives, and/or from living on a diet of examples which vacillate ambiguously between them.
Would we have to maintain, then, that our common-sense notion of content is itself ambiguous? Are the identity conditions of the contents ascribed in statements of the form, A believes that P, sometimes purely referential, and sometimes narrow, depending upon the context? Not necessarily. It may be that we have an unambiguous, unequivocal, notion of content, but in such a way that the purpose-relativity of content is written into the notion itself. Thus, A believes that P, may mean something like, The content of As belief is similar enough to the content I would express by the assertion P for the purposes in hand. Where the purposes in hand are psychological (explanatory or predictive), the constraints imposed give us narrow content. But where the purposes are communicative, then the constraints give us a purely truth-conditional notion. But the content-sentence itself would mean the same both times.
It does seem to us quite likely that we do employ a notion of narrow content when our main interest is psychological; and that where examples are presented clearly in this light, they will evoke intuitions supporting narrow content. Imagine a case where two ticket-holders in a local lottery Peter and Paul are clamouring at the ticket-booth, shouting and hammering on the door. We ask, Why? Why are they both behaving like that? Answer: Each believes the very same thing: that he has won the lottery. Here we feel no compunction in attributing the same thought to them both, despite the fact that each of their thoughts (of course) concerns a different subject namely, himself. We are, therefore, apparently quite happy to individuate the thoughts narrowly, abstracting from the differences in subject matter for purposes of psychological explanation. Now extend the example in such a way that truth becomes relevant. We ask, Who has won? Have they both won, or only one of them? Answer: Only Peter has won; Paul misread his ticket number. Now, I think, our intuitions switch. We are inclined to insist that they thought different things, because one was right while the other was wrong. This is just as the position sketched above would predict.
It seems likely, then, that one strand in our folk-psychological notion of content is explanatory, conforming to narrow principles of individuation. This is just as it should be if, as we suggested in Chapter 2, our folk psychology embodies a set of more-or-less explicit psychological laws or nomic tendencies, which would then require a notion of narrow content for their proper formulation, if the arguments given above are correct. If this is right, then we do not have to revise or reconstruct folk psychology in order to get something which might serve as an appropriate basis for scientific psychology. On the contrary, it is already of the right form, and we might expect folk and scientific psychologies to merge seamlessly into one another.
It is worth stressing, however, that the identity-conditions of the thoughts we attribute is one thing, the surface form of the sentences we use in doing it may be quite another. We rely on a variety of conventions and dodges in communicating explanatory content, often leaving the latter to be garnered from the context. Thus in the example above we explained the behaviour of the lottery-ticket holders by saying, Each believes that he has won. But of course they themselves would not employ a third-person singular mode of presentation in their thought. Here we know which (narrow) thought is being entertained it is the same thought I would express by saying, I have won but we use an indirect means of describing it.
5.2 Semantic content
It seems to us likely that we also employ a notion of content which is purely referential, or Russellian; and that we do so when our interest in thoughts and thought-descriptions is basically belief-acquisitive. First, let us make a point about linguistic communication. It seems to us that successful communication, in many contexts, does not require mutual knowledge of modes of presentation, or of Fregean senses. All that is necessary is that there should be mutual knowledge of what is being said about what. Consider the following example. You are a security guard in a museum, to which a new sculpture has recently been delivered. You are sitting outside the room where the sculpture is the only work of art on display, but you have not, as yet, seen it yourself. You now hear a visitor in the room say, That sculpture wasnt worth what they paid for it. Do you understand this remark? It seems to us that you plainly do (contra Evans, 1982). You know which thing is being talked about, and you know what is being said about it. But you neither share with the speaker a mode of presentation of the referent of their demonstrative, nor know anything about what that mode of presentation may be like (after all, for all you know the speaker may be blind, and feeling the sculpture with their hands; this makes not a bit of difference to your success in understanding them).
Why is it that the conditions for successful communication, in general, are as they are, requiring only mutual knowledge of worldly truth-conditions? By way of answer, reflect on what communication is basically for. Communication is an important channel for the acquisition of new beliefs, second only to vision in our cognitive economy. When people tell me things, in general I believe them. This works because, when people believe things, in general they believe truly. And then all that really matters, in order to make linguistic communication a reliable method of belief-acquisition, is that the truth-conditions of the beliefs at either end of the process should be the same. When you assert something of the form, a is F, then provided that I know which thing you refer to by a and which property you designate by F, it doesnt matter how differently these things may be presented to you if your thought is true, then so too will mine be.
Very often communication takes place at one remove, by us being told what someone else believes. To continue with the museum example given above: suppose that Mary is a famous art-critic, and that you tell me, Mary thinks that the new sculpture wasnt worth what was paid for it. This gives me reason to believe what Mary believes, just as if I had heard her say, That sculpture wasnt worth it. And the same conditions for understanding apply. In order for your statement to serve as a reliable channel for the acquisition of a new belief, all that matters is that I should get hold of which sculpture Marys belief is about, and what she thought about it. Marys modes of presentation matter not at all. So where our interest is basically belief-acquisitive, the constraints on a correct description of Marys thought are simply that it should preserve the original truth-conditions. And this gives us a notion of thought-content which is purely truth-conditional.
This explains, we think, the strong pull of intuitions towards wide content. Since there is, indeed, a notion of content semantic content which is individuated by worldly truth-condition, we are apt, if we fail to notice the different perspectives on content-description, to think that the notion of content must be Russellian, or world-involving. But the truth is that we also employ a notion of content explanatory content which is narrowly individuated, where our interest is in psychological explanation.
Although it does seem to us likely that common sense employs both a notion of narrow content and a notion of wide content for different purposes, this is not really the crucial point. It would not matter to us if it should turn out to be indeterminate whether common sense employs two distinct notions of content, or just one hybrid notion. What is important is that, once we see that there are two quite different perspectives we can take on the notion of content, and two distinct purposes for which we employ that notion, we see that we should employ two distinct notions (or one context-sensitive one, with varying application-conditions). Here, as so often in philosophy, what matters is not what notion we actually do have; but rather what notion we should have, given our purposes (see Carruthers, 1987b; see also Craig, 1990).
6 Conclusion
In this chapter we have rebutted arguments for the ubiquity of wide content and against the coherence of narrow content. We have argued that narrow content is not only coherent, but is also the notion which should be employed for purposes of psychological explanation, whether folk or scientific. We have also allowed that we should employ a notion of content which is wide, in communicative contexts where our interests are belief-acquisitive.
Selected reading
· In defence of wide content: Putnam, 1975a; Burge, 1979, 1986a, 1986b; Evans, 1981, 1982; McDowell, 1986, 1994; McCulloch, 1989.
· In defence of narrow content: Fodor, 1980, 1987, 1991; Blackburn, 1984, ch.9; Block, 1986; Noonan, 1986, 1993; Segal, 1989a, 1989b, 1991.