8 Bertrand Russell “Analogy” ‘he postulates hitherto considered have been such 4s ate required for knowledge of the physical ‘world. Broadly speaking, they have led us to admit {certain degree of knowledge as o the space-time Structure of the physical world, while leaving us Completely agnostic as regards its qualitative char Scter_ But where other human beings are con ferned, we feel that we know more than this; we tre convinced that other people have thoughts and feelings that are qualitatively faisly similar to our own, We are not content to think that we know fnly the space-time structure of our friends minds, oF their capacity for initiating causal chains that end in sensations of our own. A philosopher ‘might pretend to think that he knew only this, but Tet him get cross with his wife and you will see that he does not regard her as a mete spatio-temporal cilfice of whieh he knows the logical properties but not a glimmer of the intsinsic character. We are therefore justified in inferring that his skepticism is professional rather than sincere The problem with which we are concerned is, the following. We observe in ourselves sch occur rences as remembering, feasoning, feeling plea sure, and feeling pain. We think that sticks and stones do not have these experiences, but that ther people do, Most of us have no doubt thatthe higher animals fee pleasure and pain, shough I was once assured by a fisherman that “Fish have no sense nor feeling.” failed to find out how he had acquired this knowledge. Most people would dis {agree with him, but would be doubtful about oys ters and starfish. However this may be, common sense admits an increasing doubtfulness as we de scend in the animal kingdom, but as regards ‘human beings it admits no dob From human Knowledge: ts Scape and ints, Part VI chapter 8, "Alogy.” Lendon: George Ak & Unwin, 1948, 9p. It is clear that belief in the minds of others re quires some postulate that is not required in phys ics, since physics can be content with a knowledge Cc. structure, My present purpose i to suggest what is further postulate may be is clear that we must appeal to something that may be vaguely called “analogy.” The behav ior of other people is in many ways analogous to four own, and we suppose that it must have anal ‘ogous causes. What people say is what we should say if we had certain thoughts, and so we infer that they probably have these thoughts. They give us information which we can sometimes subse {quently verity. They behave in ways in which we behave when we are pleased (or displeased) in ci cumstances in which we should be pleased (or dis pleased). We may talk over with a fiend some in. ident which we have both experienced, and find {hat his reminiscences dovetail with our oven; this is particularly convincing when he remembers something that we have forgotten but that he re éalls to our thoughts. Or again: you set your boy a problem in arithmetic, and with luck he gets the Fight answer; this persuades you that he is capable of arithmetical reasoning. There are in short, very ‘many ways in which my responses to stimuli differ from those of “dead” matter, and in all thee ways ‘other people resemble me, As itis clear to me that the causal laws governing my behavior have to do with “thoughts,” itis natural to infer that the same is true ofthe analogous behavior of my friends. The inference with which we ae at present con cerned is not merely that which takes us beyond solipsism, by maintaining that sensations have Causes about which somerhing can be known, This kind of inference, which sulhces for physics, has {5 Copyrgt ©) 1948 by Brand Rural! Renewed @ TOYS by te Estate of Bertrand Russel Repeted with persion ofalready been considered. We are concerned nov witha much more specific kind of inference, the ind that is involved in our knowledge of the thoughts and feelings of others—assuming that we hhave such knowledge. Its of course obvious that Wuch knowledge is more oF les doubtful. There hot only the general argument that we may be ‘reaming, there is aso the possibility of ingenious ‘utomata, There ae calculating machines that do gums much better than our schoolboy sons; there are gramophone records that remember impecc biy what So-and-s0 said on such-and-such an oc tation; there are people in the cinema who, though copies of real people, are not themselves alive There is no theoretical limit to what ingenuity could achieve in the way of producing the illusion fife wherein fact life is absent Bat, you wil say, in all such cases it was the thoughts of human beings that produced the in fenious mechanism. Yes, but how do you know this? And how do you know thatthe gramophone does not think"? There is, in the fist place, a difference in the causal laws of observable behavior. IT say toast dient, "Write me a paper on Descares’ reasons for believing in the existence of rvattr," I shall, ihe is industrious, cause a certain response. A giamo- Phone record might be s0 constructed as 10 Spond to this stimulus, perhaps better then the st dent, but if so it would be incapable of telling me anything about any other philosopher, even if | threatened to refuse to give i a degree. One of the most notable peculiarities of human behavior is change of response 10 a given stimulus. Am inge hious person could construct an automation hich would always laugh at his jokes, however bften it heard them: but a human being, after Taughinga few times, will yawn, and end by saying, ‘How I laughed the fist time I heard that joke But the difference in observable behavior be tween living and dead matter do not suffice to prove that there are “thoughts” connected. with Tiving bodies other than my own. It is probably possible theoretically to account for the behavior Df living bodies by purely physical causal laws, and itis probably impossible to refute materialism by external observation alone, Ifwe are to believe that there are thoughts and felings other than our own, that must be in virtue of some inference in which ‘our own thoughts and feelings ate relevant, and Such an inference must go beyond what is needed physics, Tam, of course, not discussing the history of in other minds. W ‘ourselves believing in them when we frst begin to feflet; the thought that Mother may be angry or pleased is one which rises in eatly infancy. What I am discussing is the possibility of a postulate which shall establish a rational connection be tescen this belief and data, eg. betwoen the beiet Mother is angry” and the hearing ofa loud voice The abstract schema seems to be as follows. We know. from observation of ourselves, a causal law of the form "A causes B.” where Ais a “thought “ind B a physical occurrence, We sometimes ob serve a B when we cannot observe any A: we then infer an unobserved A. For example: | know that ‘when I say, "Fm thirsty,” [say o, usually, because Tram thirsty, and therefore, when T hear the sen tence “I'm thirsty” ata time when Lam not thirsty Tassume that someone ese is thirsty. assume this, the more readily if Lee before me a hot, drooping body which goes on to say, "I have walked twenty desert_miles in this heat with never a drop 10 drink.” Tis evident that my confidence inthe “in ference” is increased by increased complenity in the datum and also by increased certainty of the Causal law derived from subjective observation, nrovided the causal law is such as to account for ihe complexities of the datum. Tes clear that in so far as plurality of causes is tobe suspected, the kind of inference we have been considering is not valid. We are supposed to know SA causes B,” and also to know that B has ec curred; if this is to justify us in inferring A, we must know that only’ causes B. Or, if we are con tent to infer that Ais probable, it will sufice if we can know that in most cases it is A that causes B If you hear thunder without having, sen lightning, You confidently infer that there was lightning, be Cause you are convinced thatthe sort of noise you heard #8 seldom caused by anything except light hing. AS this example shows, our principle isnot Only employed to establish the existence of other ‘minds but is habitually assumed, though in a less Concrete form, in physics I say’"a less concre form” because unseen lightning is only abstract similar to seen lightning, whereas we suppose the similarity of other minds to our own to be by no means purely abstract ‘Complexity in the observed behavior of an: ‘other person, when this ean all be accounted for by 2 simple cause such as thirst, increases the proba bulity ofthe inference by diminishing the probabil ity of some other cause. I think that in ideally fa vorable circumstances the argument would be Formally as follows From subjective observation 1 know that A, whieh is a thought or feeling, causes B, which isa bodily act, ¢2, a statement. 1 know also that whenever B isan act of my own body, A is ils Cause: Fnow observe am act of the kind Bina body rot my own, and Tam having no thought or feeling OF the kind A. But Ishill believe, on the basis of selFobser fore infer though it this grou proportio In pra sure that, cnly caus Bin our holds out that we stHOW OTHER MINS serfobseevation, that only A can cause By 1 there sisine that there was an A which caused B, ‘hough it was not an A that I could observe. On thi ground F infer that other people's bodies are ins Evted with minds, which resemble mine in jroportion as their bodily behavior resembles my Tn practice, the exactness and certainty of the shove stalement must be softened. We cannot be Sore that, in oUF Subjective experience, Ais the aly cause of B, And even iA isthe only cause of B in our experience, how can we know that this holds outside our experience? It is not necessary that we should know this with any certainty tis ‘enough itis highly probable. Ii the assumption of probability in such cases that is our postulate The postulate may therefore be stated as follows whenever we can observe whether A and 8 are resent or absent, we fd that every case of B has fn A asa causal antecedent, thon tf probable that most B's have A's as causal antecedents, even ft ‘cases where observation does nt enable us te hrow This postulate, if accepted, justifies the infer cence 10 other minds, a8 well as many other infer ‘ences that are made unreflectingly by common