(From R. Butler, ed. Analytical Philosophy: Second Series, pp. 1-19. Blackwell, 1968. notes omitted)
Once upon a time there was a tough-minded philosopher who said, 'What is all this talk about "minds," "ideas," and "sensations"? Really--and I mean really in the real world--there is nothing to these so-called "mental" events and entities but certain processes in our all-too material heads.'
And once upon a time there was a philosopher who retorted, 'What a masterpiece of confusion! Even if, say, pain were perfectly correlated with any particular event in my brain (which I doubt) that event would obviously have certain properties--say, a certain numerical intensity measured in volts--which it would be senseless to ascribe to the feeling of pain. Thus, it is two things that are correlated, not one-and to call two things one thing is worse than being mistaken; it is utter contradiction.'
For a long time dualism and materialism appeared to exhaust the alternatives. Compromises were attempted ('double aspect' theories), but they never won many converts and practically no one found them intelligible. Then, in the mid-1930s, a seeming third possibility was discovered. This third possibility has been called logical behaviourism. To state the nature of this third possibility briefly, it is necessary to recall the treatment of the natural numbers (i.e., zero, one, two, three ... ) in modern logic. Numbers are identified with sets, in various ways, depending on which authority one follows. For instance, Whitehead and Russell identified zero with the set of all empty sets, one with the set of all one-membered sets, two with the set of all two-membered sets, three with the set of all three-membered sets, and so on. (This has the appearance of circularity, but they were able to dispel this appearance by defining 'one-membered set,' 'two-membered set,' 'three-membered set,' &c., without using 'one,' 'two,' 'three,' &c.) In short, numbers are treated as logical constructions out of sets. The number theorist is doing set theory without knowing it, according to this interpretation.
What was novel about this was the idea of getting rid of certain philosophically unwanted or embarrassing entities (numbers) without failing to do justice to the appropriate body of discourse (number theory) by treating the entities in question as logical constructions. Russell was quick to hold up this 'success' as a model to all future philosophers. And certain of those future philosophers--the Vienna positivists, in their 'physicalist' phase (about 1930)--took Russell's advice so seriously as to produce the doctrine that we are calling logical behaviourism--the doctrine that, just as numbers are (allegedly) logical constructions out of sets, so mental events are logical constructions out of actual and possible behaviour events.
In the set theoretic case, the 'reduction' of number theory to the appropriate part of set theory was carried out in detail and with indisputable technical success. One may dispute the philosophical significance of the reduction, but one knows exactly what one is talking about when one disputes it. In the mind-body case, the reduction was never carried out in even one possible way, so that it is not possible to be clear on just how mental entities or events are to be (identified with) logical constructions out of behaviour events. But, broadly speaking, it is clear what the view implies: it implies that all talk about mental events is translatable into talk about actual or potential overt behaviour.
It is easy to see in what way this view differs from both dualism and classical materialism. The logical behaviourist agrees with the dualist that what goes on in our brains has no connection whatsoever with what we mean when we say that someone is in pain. He can even take over the dualist's entire stock of arguments against the materialist position. Yet, at the same time, he can be as 'tough-minded' as the materialist in denying that ordinary talk of 'pains,' 'thoughts,' and 'feelings' involves reference to 'Mind' as a Cartesian substance.
Thus it is not surprising that logical behaviourism attracted enormous attention--both pro and con-during the next thirty years. Without doubt, this alternative proved to be a fruitful one to inject into the debate. Here, however, my intention is not to talk about the fruitfulness of the investigations to which logical behaviourism has led, but to see if there was any upshot to those investigations. Can we, after thirty years, say anything about the rightness or wrongness of logical behaviourism? Or must we say that a third alternative has been added to the old two; that we cannot decide between three any more easily than we could decide between two; and that our discussion is thus half as difficult again as it was before?
One conclusion emerged very quickly from the discussion pro and con logical behaviourism: that the extreme thesis of logical behaviourism, as we just stated it (that all talk about 'mental events' is translatable into talk about overt behaviour) is false. But, in a sense, this is not very interesting. An extreme thesis may be false, although there is 'something to' the way of thinking that it represents. And the more interesting question is this: what, if anything, can be 'saved' of the way of thinking that logical behaviourism represents?
In the last thirty years, the original extreme thesis of logical behaviourism has gradually been weakened to something like this:
That there exist entailments between mind-statements and behaviour-statements; entailments that are not, perhaps, analytic in the way in which 'All bachelors are unmarried' is analytic, but that nevertheless follow (in some sense) from the meanings of mind words. I shall call these analytic entailments. That these entailments may not provide an actual translation of 'mind talk' into 'behaviour talk' (this 'talk' talk was introduced by Gilbert Ryle in his Concept of Mind), but that this is true for such superficial reasons as the greater ambiguity of mind talk, as compared with the relatively greater specificity of overt behaviour talk.
I believe that, although no philosopher would today subscribe to the older version of logical behaviourism, a great many philosophers would accept these two points, while admitting the unsatisfactory imprecision of the present statement of both of them. If these philosophers are right, then there is much work to be done (e.g., the notion of 'analyticity' has to be made clear), but the direction of work is laid out for us for some time to come.
I wish that I could share this happy point of view--if only for the comforting conclusion that first-rate philosophical research, continued for some time, will eventually lead to a solution to the mind-body problem which is independent of troublesome empirical facts about brains, central causation of behaviour, evidence for and against nonphysical causation of at least some behaviour, and the soundness or unsoundness of psychical research and parapsychology. But the fact is that I come to bury logical behaviourism, not to praise it. I feel that the time has come for us to admit that logical behaviourism is a mistake, and that even the weakened forms of the logical behaviourist doctrine are incorrect. I cannot hope to establish this in so short a paper as this one; but I hope to expose for your inspection at least the main lines of my thinking.
Logical Behaviourism
The logical behaviourist usually begins by pointing out what is perfectly true, that such words as 'pain' ('pain' will henceforth be our stock example of a mind word) are not taught by reference to standard examples in the way in which such words as 'red' are. One can point to a standard red thing, but one cannot point to a standard pain (that is, except by pointing to some piece of behaviour) and say: 'Compare the feeling you are having with this one (say, Jones's feeling at time t). If the two feelings have the identical quality, then your feeling is legitimately called a feeling of pain.' The difficulty, of course, is that I cannot have Jones's feeling at time t1--unless I am Jones, and the time is t1.
From this simple observation, certain things follow. For example, the account according to which the intension of the word 'pain' is a certain quality which 'I know from my own case' must be wrong. But this is not to refute dualism, since the dualist need not maintain that I know the intension of the English word 'pain' from my own case, but only that I experience the referent of the word.
What then is the intension of 'pain'? I am inclined to say that 'pain' is a cluster-concept. That is, the application of the word 'pain' is controlled by a whole cluster of criteria, all of which can be regarded as synthetic. As a consequence, there is no satisfactory way of answering the question 'What does "pain" mean?' except by giving an exact synonym (e.g., 'Schmerz'); but there are a million and one different ways of saying what pain is. One can, for example, say that pain is that feeling which is normally evinced by saying 'ouch,' or by wincing, or in a variety of other ways (or often not evinced at all).
All this is compatible with logical behaviourism. The logical behaviourist would reply: 'Exactly. "Pain" is a cluster-concept-that is to say, it stands for a cluster of phenomena.' But that is not what I mean. Let us look at another kind of cluster-concept (cluster-concepts, of course, are not a homogeneous class): names of diseases.
We observe that, when a virus origin was discovered for polio, doctors said that certain cases in which all the symptoms of polio had been present, but in which the virus had been absent, had turned out not to be cases of polio at all. Similarly, if a virus should be discovered which normally (almost invariably) is the cause of what we presently call 'multiple sclerosis,' the hypothesis that this virus is the cause of multiple sclerosis would not be falsified if, in some few exceptional circumstances, it was possible to have all the symptoms of multiple sclerosis for some other combination of reasons, or if this virus caused symptoms not presently recognized as symptoms of multiple sclerosis in some cases. These facts would certainly lead the lexicographer to reject the view that 'multiple sclerosis' means 'the simultaneous presence of such and such symptoms.' Rather he would say that ,multiple sclerosis' means 'that disease which is normally responsible for some or all of the following symptoms.....²Of course, he does not have to say this. Some philosophers would prefer to say that 'polio' used to mean 'the simultaneous presence of such-and-such symptoms.' And they would say that the decision to accept the presence or absence of a virus as a criterion for the presence or absence of polio represented a change of meaning. But this runs strongly counter to our common sense. For example, doctors used to say 'I believe polio is caused by a virus.' On the ,change of meaning' account, those doctors were wrong, not right. Polio, as the word was then used, was not always caused by a virus; it is only what we call polio that is always caused by a virus. And if a doctor ever said (many did) 'I believe this may not be a case of polio,' knowing that all of the text-book symptoms were present, that doctor must have been contradicting himself (even if we, today, would say that he was right) or, perhaps, 'making a disguised linguistic proposal.' Also, this account runs counter to good linguistic methodology. The definition we proposed a paragraph back-'multiple sclerosis' means 'the disease that is normally responsible for the following symptoms. . . .'-has an exact analogue in the case of polio. This kind of definition leaves open the question whether there is a single cause or several. It is consonant with such a definition to speak of 'discovering a single origin for polio (or two or three or four),' to speak of 'discovering X did not have polio' (although he exhibited all the symptoms of polio), and to speak of 'discovering X did have polio' (although he exhibited none of the 'textbook symptoms'). And, finally, such a definition does not require us to say that any 'change of meaning' took place. Thus, this is surely the definition that a good lexicographer would adopt. But this entails rejecting the ,change of meaning' account as a philosopher's invention.
Accepting that this is the correct account of the names of diseases, what follows? There may be analytic entailments connecting diseases and symptoms (although I shall argue against this). For example, it looks plausible to say that 'Normally people who have multiple sclerosis have some or all of the following symptoms . . .'
is a necessary ('analytic') truth. But it does not follow that 'disease talk' is translatable into ,symptom talk.' Rather the contrary follows (as is already indicated by the presence of the word 'normally'): statements about multiple sclerosis are not translatable into statements about the symptoms of multiple sclerosis, not because disease talk is 'systematically ambiguous' and symptom talk is 'specific,' but because causes are not logical constructions out of their effects.
In analogy with the foregoing, both the dualist and the materialist would want to argue that, although the meaning of 'pain' may be explained by reference to overt behaviour, what we mean by 'pain' is not the presence of a cluster of responses, but rather the presence of an event or condition that normally causes those responses. (Of course the pain is not the whole cause of the pain behaviour, but only a suitably invariant part of that cause; but, similarly, the virus-caused tissue damage is not the whole cause of the individual symptoms of polio in some individual case, but a suitably invariant part of the cause.) And they would want to argue further, that even if it were a necessary truth that 'Normally, when one says "ouch" one has a pain'
or a necessary truth that 'Normally, when one has a pain one says "ouch"'
this would be an interesting observation about what 'pain' means, but it would shed no metaphysical light on what pain is (or isn't). And it certainly would not follow that 'pain talk' is translatable into 'response talk', or that the failure of translatability is only a matter of the 'systematic ambiguity' of pain talk as opposed to the 'specificity' of response talk: quite the contrary. Just as before, causes (pains) are not logical constructions out of their effects (behaviour).
The traditional dualist would, however, want to go farther, and deny the necessity of the two propositions just listed. Moreover, the traditional dualist is right: there is nothing self-contradictory, as we shall see below, in talking of hypothetical worlds in which there are pains but no pain behaviour.²The analogy with names of diseases is still preserved at this point. Suppose I identify multiple sclerosis as the disease that normally produces certain symptoms. If it later turns out that a certain virus is the cause of multiple sclerosis, using this newly discovered criterion I may then go on to find out that multiple sclerosis has quite different symptoms when, say, the average temperature is lower. I can then perfectly well talk of a hypothetical world (with lower temperature levels) in which multiple sclerosis does not normally produce the usual symptoms. It is true that if the words 'multiple sclerosis' are used in any world in such a way that the above lexical definition is a good one, then many victims of the disease must have had some or all of the following symptoms ... And in the same way it is true that if the explanation suggested of the word 'pain' is a good one (i.e., 'pain is the feeling that is normally being evinced when someone says "ouch," or winces, or screams, &c.% then persons in pain must have at some time winced or screamed or said 'ouch'-but this does not imply that 'if someone ever had a pain, then someone must at some time have winced or screamed or said "ouch."' To conclude this would be to confuse preconditions for talking about pain as we talk about pain with preconditions for the existence of pain.
The analogy we have been developing is not an identity: linguistically speaking, mind words and names of diseases are different in a great many respects. In particular, first person uses are very different: a man may have a severe case of polio and not know it, even if he knows the word 'polio,' but one cannot have a severe pain and not know it. At first blush, this may look like a point in favour of logical behaviourism. The logical behaviourist may say: it is because the premisses 'John says he has a pain,' 'John knows English,' and 'John is speaking in all sincerity," entail 'John has a pain,' that pain reports have this sort of special status. But even if this is right, it does not follow that logical behaviourism is correct unless sincerity is a 'logical construction out of overt behaviour'! A far more reasonable account is this: one can have a 'pink elephant hat., lucination,' but one cannot have a 'pain hallucination,' or an 'absence of pain hallucination,' simply because any situation that a person cane not discriminate from a situation in which be himself has a pain counts as a situation in which he has a pain, whereas a situation that a person cannot distinguish from one in which a pink elephant is present does not necessarily count as the presence of a pink elephant.
To sum up: I believe that pains are not clusters of responses, but that they are (normally, is our experience to date) the causes of certain clusters of responses. Moreover, although this is an empirical fact, it underlies the possibility talking about pains in the particular way it which we do. However, it does not rule out any way the possibility of worlds in whirl (owing to a difference in the environmental and hereditary conditions) pains are not responsible for the usual responses, or even are not responsible for any responses at all.
Let us now engage in a little science fiction Let us try to describe some worlds in which pain are related to responses (and also to causes) quite a different way than they are in our world.
If we confine our attention to non-verbal responses by full grown persons, for a start, the matters are easy. Imagine a community of 'super-spartans' or 'super-stoics'-a community in which the adults have the ability to successfully suppress all involuntary pain behaviour. They may, on occasion, admit that they feel pain, but always in pleasant well-modulated voices--even if they are undergoing the agonies of the damned. They do not wince, scream, flinch, sob, grit their teeth, clench their fists, exhibit beads of sweat, or otherwise act like people in pain or people suppressing the unconditioned responses associated with pain. However, they do feel pain, and they dislike it (just as we do). They even admit that it takes a great effort of will to behave as they do. It is only that they have what they regard as important ideological reasons for behaving as they do, and they have, through years of training, learned to live up to their own exacting standards.
It may be contended that children and not fully mature members of this community will exhibit, to varying degrees, normal unconditioned pain behaviour, and that this is all that is necessary for the ascription of pain. On this view, the sine qua non for the significant ascription of pain to a species is that its immature members should exhibit unconditioned pain responses.
One might well stop to ask whether this statement has even a clear meaning. Supposing that there are Martians: do we have any criterion for something being an 'unconditioned pain response' for a Martian? Other things being equal, one avoids things with which one has had painful experiences: this would suggest that avoidance behaviour might be looked for as a universal unconditioned pain response. However, even if this were true, it would hardly be specific enough, since avoidance can also be an unconditioned response to many things that we do not associate with pain-to things that disgust us, or frighten us, or even merely bore us.
Let us put these difficulties aside, and see if we can devise an imaginary world in which there are not, even by lenient standards, any unconditioned pain responses. Specifically, let us take our 'super-spartans', and let us suppose that after millions of years they begin to have children who are born fully acculturated. They are born speaking the adult language, knowing the multiplication table, having opinions on political issues, and inter alia sharing the dominant spartan beliefs about the importance of not evincing pain (except by way of a verbal report, and even that in a tone of voice that suggests indifference). Then there would not be any 'unconditioned pain responses' in this community (although there might be unconditioned desires to make certain responses--desires which were, however, always suppressed by an effort of will). Yet there is a clear absurdity to the position that one cannot ascribe to these people a capacity for feeling pain.
To make this absurdity evident, let us imagine that we succeed in converting an adult 'superspartan' to our ideology. Let us suppose that he begins to evince pain in the normal way. Yet he reports that the pains he is feeling are not more intense than are the ones he experienced prior to conversion-indeed, he may say that giving expression to them makes them less intense. In this case, the logical behaviourist would have to say that, through the medium of this one member, we had demonstrated the existence of unconditioned pain responses in the whole species, and hence that ascription of pain to the species is 'logically proper.' But this is to say that had this one man never lived, and had it been possible to demonstrate only indirectly (via the use of theories) that these beings feel pain, then pain ascriptions would have been improper.
We have so far been constructing worlds in which the relation of pain to its non-verbal effects is altered. What about the relation of pain to causes? This is even more easy for the imagination to modify. Can one not imagine a species who feel pain only when a magnetic field is present (although the magnetic field causes no detectable damage to their bodies or nervous systems)? If we now let the members of such a species become converts to 'superspartanism,' we can depict to ourselves a world in which pains, in our sense, are clearly present, but in which they have neither e normal causes nor the normal effects (apart from verbal reports).
What about verbal reports? Some behaviourists have taken these as the characteristic form of pain behaviour. Of course, there is a difficulty here: If 'I am in pain' means 'I am disposed to utter this kind of verbal report' (to put matters crudely), then how do we tell that any particular report is 'this kind of verbal report'? The usual answer is in terms of the unconditioned pain responses and their assumed supplantation by the verbal reports in question. However, we have seen that there are no logical reasons for the existence of unconditioned pain responses in all species capable of feeling pain (there may be logical reasons for the existence of avoidance desires, but avoidance desires are not themselves behaviour any more than pains are).
Once again, let us be charitable to the extent of waving the first difficulty that comes to mind, and let us undertake the task of trying to imagine a world in which there are not even pain reports. I will call this world the 'X-world.' In the Xworld we have to deal with 'super-super-spartans.' These have been super-spartans for so long, that they have begun to suppress even talk of pain. Of course, each individual X-worlder may have his private way of thinking about pain. He may even have the word 'pain' (as before, I assume that these beings are born fully acculturated). He may think to himself: 'This pain is intolerable. If it goes on one minute longer I shall scream. Oh No! I mustn't do that! That would disgrace my whole family . . .' But X-worlders do not even admit to having pains. They pretend not to know either the word or the phenomenon to which it refers. In short, if pains are 'logical constructs out of behaviour', then our X-worlders behave so as not to have pains!-Only, of course, they do have pains, and they know perfectly well that they have pains.
If this last fantasy is not, in some disguised way, self-contradictory, then logical behaviourism is simply a mistake. Not only is the second thesis of logical behaviourism--the existence of a near-translation of pain talk into behaviour talk-false, but so is even the first thesis--the existence of 'analytic entailments.' Pains are responsible for certain kinds of behaviour--but only in the context of our beliefs, desires, ideological attitudes, and so forth. From the statement 'X has a pain' by itself no behavioural statement follows--not even a behavioural statement with a 'normally' or a 'probably' in it.
In our concluding section we shall consider the logical behaviourist's stock of countermoves to this sort of argument. If the logical behaviourist's positive views are inadequate owing to an oversimplified view of the nature of cluster words-amounting, in some instances, to an open denial that it is possible to have a word governed by a cluster of indicators, all of which are synthetic--his negative views are inadequate owing to an oversimplified view of empirical reasoning. It is unfortunately characteristic of modern philosophy that its problems should overlap three different areas--to speak roughly, the areas of linguistics, logic, and 'theory of theories' (scientific methodology)-and that many of its practitioners should try to get by with an inadequate knowledge of at least two out of the three.
Some Behaviourist Arguments
We have been talking of 'X-worlders' and 'super-spartans.' No one denies that, in some sense of the term, such fantasies are 'intelligible.' But 'intelligibility' can be a superficial thing. A fantasy may be 'intelligible,' at least at the level of 'surface grammar,' although we may come to see, on thinking about it for a while, that some absurdity is involved. Consider, for example, the supposition that last night, just on the stroke of midnight, all distances were instantaneously doubled. Of course, we did not notice the change, for we ourselves also doubled in size! This story may seem intelligible to us at first blush, at least as an amusing possibility. On reflection, however, we come to see that a logical contradiction is involved. For 'length' means nothing more nor less than a relation to a standard, and it is a contradiction to maintain that the length of everything doubled, while the relations to the standards remained unchanged.
What I have just said (speaking as a logical behaviourist might speak) is false, but not totally so. It is false (or at least the last part is false), because 'length' does not mean 'relation to a standard.' If it did (assuming a 'standard' has to be a macroscopic material object, or anyway a material object), it would make no sense to speak of distances in a world in which there were only gravitational and electromagnetic fields, but no material objects. Also, it would make no sense to speak of the standard (whatever it might be) as having changed its length. Consequences so counter-intuitive have led many physicists (and even a few philosophers of physics) to view 'length' not as something operationally defined, but as a theoretical magnitude (like electrical charge), which can be measured in a virtual infinity of ways, but which is not explicitly and exactly definable in terms of any of the ways of measuring it. Some of these physicists--the 'unified field' theorists--would even say that, far from it being the case that 'length' (and hence 'space') depends on the existence of suitably related material bodies, material bodies are best viewed as local variations in the curvature of space--that is to say, local variations in the intensity of a certain magnitude (the tensor g), one aspect of which we experience as 'length.'
Again, it is far from true that the hypothesis 'last night, on the stroke of midnight, everything doubled in length' has no testable consequences. For example, if last night everything did double in length, and the velocity of light did not also double, then this morning we would have experienced an apparent halving of the speed of light. Moreover, if g (the gravitational constant) did not double, then we would have experienced and apparent halving in the intensity of the gravitational field. And if h (Planck's constant) did not change, then.... In short, our world would have been bewilderingly different. And if we could survive at all, under so drastically altered conditions, no doubt some clever physicist would figure out what had happened.
I have gone into such detail just to make the point that in philosophy things are rarely so simple as they seem. The 'doubling universe' is a favourite classroom example of a 'pseudo-hypothesis'-yet it is the worst possible example if a 'clear case' is desired. In the first place, what is desired is a hypothesis with no testable consequences--yet this hypothesis, as it is always stated, does have testable consequences (perhaps some more complex hypothesis does not; but then we have to see this more complex hypothesis stated before we can be expected to discuss it). In the second place, the usual argument for the absurdity of this hypothesis rests on a simplistic theory of the meaning of 'length'-and a full discussion of that situation is hardly possible without bringing in considerations from unified field theory and quantum mechanics (the latter comes in in connection with the notion of a 'material standard'). But, the example aside, one can hardly challenge the point that a superficially coherent story may contain a hidden absurdity.
Or can one? Of course, a superficially coherent story may contain a hidden logical contradiction, but the whole point of the logical behaviourist's sneering reference to 'surface grammar' is that linguistic coherence, meaningfulness of the individual terms, and logical consistency, do not by themselves guarantee freedom from another kind of absurdity--there are 'depth absurdities' which can only be detected by more powerful techniques. It is fair to say that today, after thirty years of this sort of talk, we lack both a single convincing example of such a depth absurdity, and a technique of detection (or alleged technique of detection) which does not reduce to 'untestable, therefore nonsense'.
To come to the case at hand: the logical behaviourist is likely to say that our hypothesis about 'X-worlders' is untestable in principle (if there were 'X-worlders,' by hypothesis we couldn't distinguish them from people who really didn't know what pain is); and therefore meaningless (apart from a certain 'surface significance' which is of no real interest). If the logical behaviourist has learned a little from ,ordinary language philosophy,' he is likely to shy away from saying 'untestable, therefore meaningless,' but he is still likely to say or at least think: 'untestable, therefore in some sense absurd.' l shall try to meet this 'argument' not by challenging the premiss, be it overt or covert, that 'untestable synthetic statement' is some kind of contradiction in terms (although I believe that premiss to be mistaken), but simply by showing that, on any but the most naive view of testability, our hypothesis is testable.
Of course, I could not do this if it were true that 'by hypothesis, we couldn't distinguish X-worlders from people who really didn't know what pain is.' But that isn't true--at any rate, it isn't true 'by hypothesis. What is true by hypothesis is that we couldn't distinguish Xworlders from people who really didn't known what pain is on the basis of overt behaviour alone. But that still leaves many other ways in which we might determine what is going on 'inside' the X-worlders--in both the figurative and literal sense of 'inside.' For example, we might examine their brains.
It is a fact that when pain impulses are 'received' in the brain, suitable electrical detecting instruments record a characteristic 'spike' pattern. Let us express this briefly (and too simply) by saying that 'brain spikes' are one-to-one correlated with experiences of pain. If our X-worlders belong to the human species, then we can verify that they do feel pains, notwithstanding their claim that they don't have any idea what pain is, by applying our electrical instruments and detecting the tell-tale 'brain spikes.'
This reply to the logical behaviourist is far too simple to be convincing. 'It is true,' the logical behaviourist will object, 'that experiences of pain are one-to-one correlated with "brain spikes" in the case of normal human beings. But you don't know that the X-worlders are normal human beings, in this sense--in fact, you have every reason to suppose that they are not normal human beings.' This reply shows that no mere correlation, however carefully verified in the case of normal human beings, can be used to verify ascriptions of pain to X-worlders. Fortunately, we do not have to suppose that our knowledge will always be restricted to mere correlations, like the pain-'brain spike' correlation. At a more advanced level, considerations of simplicity and coherence can begin to play a role in a way in which they cannot when only crude observational regularities are available.
Let us suppose that we begin to detect waves of a new kind, emanating from human brainscall them 'V-waves.' Let us suppose we develop a way of 'decoding' V-waves so as to reveal people's unspoken thoughts. And, finally, let us suppose that our 'decoding' technique also works in the case of the V-waves emanating from the brains of X-worlders. How does this correlation differ from the pain-'brain spike' correlation?
Simply in this way: it is reasonable to say that 'spikes'--momentary peaks in the electrical intensity in certain parts of the brain--could have almost any cause. But waves which go over into coherent English (or any other language); under a relatively simple decoding scheme, could not have just any cause. The ,null hypothesis'-that this is just the operation of 'chance'-can be dismissed at once. And if, in the case of human beings, we verify that the decoded waves correspond to what we are in fact thinking, then the hypothesis that this same correlation holds in the case of X-worlders will be assigned an immensely high probability, simply because no other likely explanation readily suggests itself. But 'no other likely explanation readily suggests itself' isn't verification, the logical behaviourist may say. On the contrary. How, for example, have we verified that cadmium lines in the spectrographic analysis of sunlight indicate the presence of cadmium in the sun? Mimicking the logical behaviourist, we might say: 'We have verified that under normal circumstances, cadmium lines only occur when heated cadmium is present. But we don't know that circumstances on the sun are normal in this sense.' If we took this seriously, we would have to heat cadmium on the sun before we could say that the regularity upon which we base our spectrographic analysis of sunlight had been verified. In fact, we have verified the regularity under 'normal' circumstances, and we can show (deductively) that if many other laws, that have also been verified under 'normal' circumstances and only under 'normal' circumstances (i.e., never on the surface of the sun), hold on the sun, then this regularity holds also under 'abnormal' circumstances. And if someone says, 'But perhaps none of the usual laws of physics hold on the sun,' we reply that this is like supposing that a random process always produces coherent English. The fact is that the 'signals' (sunlight, radio waves, &c.) which we receive from the sun cohere with a vast body of theory. Perhaps there is some other explanation than that the sun obeys the usual laws of physics; but no other likely explanation suggests itself. This sort of reasoning is scientific verification; and if it is not reducible to simple Baconian induction--well, then, philosophers must learn to widen their notions of verification to embrace it.
The logical behaviourist might try to account for the decodability of the X-worlders' 'Vwaves' into coherent English (or the appropriate natural language) without invoking the absurd ,null hypothesis.' He might suggest, for example, that the 'X-worlders' are having fun at our expense-they are able, say, to produce misleading V -waves at will. If the X-worlders have brains quite unlike ours, this may even have some plausibility. But once again, in an advanced state of knowledge, considerations of coherence and simplicty may quite conceivably ,verify' that this is false. For example, the Xworlders may have brains quite like ours, rather than unlike ours. And we may have built up enough theory to say how the brain of a human being should 'look' if that human being were pretending not to be in pain when he was, in fact, in pain. Now consider what the 'misleading V-waves' story requires: it requires that the X-worlders produce V-waves in quite a different way than we do, without specifying what that different way is. Moreover, it requires that this be the case, although the reverse hypothesis that X-worlders' brains function exactly as human brains do-in fact, that they are human brains-fits all the data. Clearly, this story is in serious methodological difficulties, and any other 'counter-explanation' that the logical behaviourist tries to invoke will be in similar difficulties. In short, the logical behaviourist's argument reduces to this: 'You cannot verify "psycho-physical" correlations in the case of Xworlders (or at least, you can't verify ones having to do, directly or indirectly, with pain), because, by hypothesis, X-worlders won't tell you (or indicate behaviourally) when they are in pain. "Indirect verification"-verification using theories which have been "tested" only in the case of human beings-is not verification at all, because X-worlders may obey different laws than human beings. And it is not incumbent upon me (the logical behaviourist says) to suggest what those laws might be: it is incumbent upon you to rule out all other explanations.' And this is a silly argument. The scientist does not have to rule out all the ridiculous theories that someone might suggest; he only has to show that he has ruled out any reasonable alternative theories that one might put forward on the basis of present knowledge.
Granting, then, that we might discover a technique for 'reading' the unspoken thoughts of X-worlders: we would then be in the same position with respect to the X-worlders as we were with respect to the original 'super-spartans.' The super-Spartans were quite willing to tell us (and each other) about their pains; and we could see that their pain talk was linguistically coherent and situationally appropriate (e.g., a superspartan will tell you that he feels intense pain when you touch him with a red hot poker). On this basis, we were quite willing to grant that the super-spartans did, indeed, feel pain-all the more readily, since the deviancy in their behaviour had a perfectly convincing ideological explanation. (Note again the role played here by considerations of coherence and simplicity). But the X-worlders also 'tell' us (and, perhaps, each other), exactly the same things, albeit unwillingly (by the medium of the involuntarily produced 'V-waves'). Thus we have to say-at least, we have to say as long as the 'V-wave' theory has not broken down-that the X-worlders are what they, in fact, are just 'super-super-Spartans.
Let us now consider a quite different argument that a logical behaviourist might use. 'You are assuming,' he might say, 'the following principle:
'If someone's brain is in the same state as that of a human being in pain (not just at the moment of the pain, but before and after for a sufficient interval), then he is in pain.
'Moreover, this principle is one which it would never be reasonable to give up (on your conception of "methodology"). Thus, you have turned it into a tautology. But observe what turning this principle into a tautology involves: it involves changing the meaning of "pain." What "pain" means for you is: the presence of pain, in the colloquial sense of the term, or the presence of a brain state identical with the brain state of someone who feels pain. Of course, in that sense we can verify that your "X-worlders" experience "pain"-but that is not the sense of "pain" at issue.'
The reply to this argument is that the premiss is simply false. It is just not true that, on my conception of verification, it would never be reasonable to give up the principle stated. To show this, I have to beg your pardons for engaging in a little more science fiction. Let us suppose that scientists discover yet another kind of wavescall them 'W waves'. Let us suppose that W-waves do not emanate from human brains, but that they are detected emanating from the brains of X-worlders. And let us suppose that, once again, there exists a simple scheme for decoding W-waves into coherent English (or whatever language X-worlders speak), and that the 'decoded' waves 'read' like this: 'Ho, ho! are we fooling those Earthians! They think that the V-waves they detect represent our thoughts! If they only knew that instead of pretending not to have pains when we really have pains, we are really pretending to pretend not to have pains when we really do have pains when we really don't have pains!' Under these circumstances, we would 'doubt' (to put it mildly) that the same psycho-physical correlations held for normal humans and for X-worlders. Further investigations might lead us to quite a number of different hypotheses. For example, we might decide that X-worlders don't think with their brains at all-that the 'organ' of thought is not just the brain, in the case of X-worlders, but some larger structure--perhaps even a structure which is not 'physical' in the sense of consisting of elementary particles. The point is that what is necessarily true is not the principle stated two paragraphs back, but rather the principle: If someone (some organism) is in the same state as a human being in pain in all relevant respects, then he (that organism) is in pain.
-And this principle is a tautology by anybody's lights! The only a priori methodological restriction I am imposing here is this one:
If some organism is in the same state as a human being in pain in all respects known to be relevant, and there is no reason to suppose that there exist unknown relevant respects, then don't postulate any.
-But this principle is not a 'tautology'; in fact, it is not a statement at all, but a methodological directive. And deciding to conform to this directive is not (as hardly needs to be said) changing the meaning of the word 'pain', or of any word.
There are two things that the logical behaviourist can do: he can claim that ascribing pains to X-worlders, or even super-spartans, involves a 'change of meaning,' or he can claim that ascribing pains to super-spartans, or at least to X-worlders, is 'untestable.' The first thing is a piece of unreasonable linguistics; the second, a piece of unreasonable scientific method. The two are, not surprisingly, mutually supporting: the unreasonable scientific method makes the unreasonable linguistics appear more reasonable. Similarly, the normal ways of thinking and talking are mutually supporting: reasonable linguistic field techniques are, needless to say, in agreement with reasonable conceptions of scientific method. Madmen sometimes have consistent delusional systems; so madness and sanity can both have a 'circular' aspect. I may not have succeeded, in this paper, in breaking the 'delusional system' of a committed logical behaviourist; but I hope to have convinced the uncommitted that that system need not be taken seriously. If we have to choose between 'circles,' the circle of reason is to be preferred to any of the many circles of unreason.