| C. D. Broad, Perception, Physics, and Reality, 1914 |
CHAPTER III
ON PHENOMENALISM1
BEFORE coming to the chapter on the Causal Theory of Perception and its effects on our beliefs in the reality of the objects of our perceptions, I propose to devote a short space to the discussion of Phenomenalism.
No apology is needed for discussing this theory somewhere in any essay which deals with the question of the information, if any, that perception can give us about reality. And when this question is raised with particular reference to the philosophical position of the truths of natural science, such a discussion is essential in view of the fact that Mach and his school are phenomenalists, and hold that phenomenalism is the philosophic theory that is best suited to be the basis for physics. The only preliminary point that does call for some explanation is why I should discuss phenomenalism here rather than after the chapter on the Causal Theory of Perception. It will be said that it is arguments based on the causal theory that chiefly undermine naïf realism, and that it would be a more reasonable order to discuss phenomenalism as an alternative when the difficulties of naïf realism were becoming insuperable than here, where they have hardly fairly begun.
We may anticipate the results of the next chapter so far as to agree that the main difficulties of naïf realism do spring from the causal account of perception; but we may still defend the order that we have adopted. For we shall try to show in the present chapter that causal arguments that refute naïf realism cannot be used to support phenomenalism. Moreover, the main problem about phenomenalism from the point of view of the philosophy of natural science is to be found in the question of its relation to causal laws, and so there is good reason for discussing it directly after we have finished with Causality. I shall therefore proceed to discuss phenomenalism without further apology.
First of all, what precisely is meant by phenomenalism? In the sense in which we propose to discuss it it is the theory about the reality of the world with which we come in contact in perception which is diametrically opposite to that of naïf realism. It holds, not merely that the objects of all our perceptions exist only when they are perceived, but also that there are no permanent real things with laws of their own that cause these perceptions and in some measure resemble their objects. The laws of science, stated in terms of such supposed realities and their states, are for it mere transcriptions of laws connecting the perceptions that people actually have, and these perceptions and their laws are all that we can hope to make the objects of science.
Phenomenalism, unlike naïf realism, is a position that needs proof. Every man is a realist except in so far as experience and reflexion force him away from that position. But nobody becomes a phenomenalist except by argument. Nor is phenomenalism the position at which we naturally arrive on leaving naïf realism2. As soon as the average man is forced away from naïf realism at any point he always assumes that, in the case of every characteristic that does not raise some special difficulty, he perceives the real, and that events in that real cause the perception of that characteristic which he now has to believe to be appearance. What, then, are the arguments for phenomenalism?
Clearly it has to refute both naïf realism and the modified form of realism that is put in its place. We have already discussed the arguments against naïf realism that are independent of causality, and seen that most of them have very little weight. And the arguments that remain to be discussed in the next chapter from the relativity of perceptions to an organ will not really prove phenomenalism. The argument is a little complicated, and we had better put it formally, in order to avoid all chance of error. Let p, be the proposition 'phenomenalism is true.' Let q be the proposition that the objects of our perceptions depend on the structure of our organs. Can we prove p from this, i.e. can we at the same time assert q and q --> p? In the first place, if phenomenalism be true, our perceptions cannot depend on the permanent structure of our organs, for they will have no permanent structure. They exist when somebody perceives them, but not otherwise. Hence, unless you can be sure that, e.g., somebody always perceives your eye when you perceive a colour, there is no truth in the statement that what you perceive by sight depends on your eye and its structure. Every time you perceive a colour when no one does perceive your eye, you have a perception of colour which does not depend on the existence of an organ and its structure. Hence we can assert the proposition p --> q. Taking this together with q --> p, we get q --> ~q. But q --> ~q .-->. ~q. Hence we reach the conclusion that q --> p .-->. ~q Thus to assert both q --> p and q would involve the assertion of both q and ~q. You cannot do this, and therefore you cannot prove p from the argument. It is true that Berkeley, whose argument is properly phenomenalistic, is so shocked at this result that he introduces God either -- for the point is uncertain --to perceive your eye when no one else does, or to be a permanent cause which can make people perceive an eye whenever someone else perceives a colour. But either alternative is a departure from pure phenomenalism. The first alternative is ridiculous, unless there be other grounds for believing in, the existence of God. Taken as an argument for God, the position might be stated as follows: 'I have produced a theory about the unreality of the objects of our perceptions which is intrinsically so contrary to what people generally believe that it needs powerful proofs. What would be a strong proof, if it were consistent both with known facts and with my theory, is unfortunately inconsistent with them. It would cease to be so if I introduced a new fact, viz. a percipient God. Therefore it is obvious that such a God must exist.' The second alternative takes us away from phenomenalism to a form of idealism, for it now holds that our perceptions have permanent causes common to all of us under like circumstances; but it goes beyond this by supposing that these causes are to be found in the volition of a single person whom it identifies with the God of theology. The first step takes us from phenomenalism to a form of realism, the second to a form of idealism.
We may then, I think, agree that no arguments based on relativity of perception to an organ can by themselves legitimately lead to the proposition that all that we ever perceive is appearance, and that there corresponds nothing permanent in the real to the objects that we time and again perceive, and that common-sense takes to be relatively permanent realities. I do not, of course, wish to deny at this stage that the propositions about relativity to an organ could be stated in a roundabout way in terms of phenomenalism. I only want to show that they could not when so stated be consistently made into an argument to prove the truth of phenomenalism in general.
There is, in fact, no direct argument for phenomenalism that can make any claim on us. The doctrine, if held at all, can only be held reasonably on some such grounds as the following: Suppose it were found that naïf realism could not be maintained, and that the causal view that is substituted for it leaves us in complete agnosticism about the real, then the cry might well arise: Why not drop all reference to the real and state everything in terms of perceptions and the laws of their connexion? To this question the answer must be: Either you do not intend to attempt to find any causal laws that will tell us what perceptions to expect or you do. In the former case you must remember that your theory is less well off than the one that boldly assumes real causes of our perceptions rather like their objects, and assumes that they obey certain laws. For these assumptions do account for a good many of the perceptions that we have, and there is no reason to suppose that there is any a priori objection to making them. In the latter case, if you are to keep to pure phenomenalism you will have to account for present perceptions by causal laws that bring in no data beside other perceptions. And it may certainly be questioned whether you will be able to do this. If you bring into your laws anything like 'possible perceptions,' or 'perceptions beneath the threshold of consciousness,' you have, however, left pure phenomenalism. For you are now assuming the existence of something that is not, and never has been, the object of a direct awareness, and that is precisely what the ordinary scientist does in his assumptions about the real world. The question will then merely be (a) whether his assumption that the real world is on the whole very much like the objects that he perceives is a priori less probable than yours, and (b) whether your theory can explain the occurrence of the perceptions that we actually have as well as the rival theory.
I think it is perfectly clear that an absolutely pure phenomenalism that wishes to explain and anticipate our perceptions can be ruled out of court. We will suppose that it is allowed to assume present perceptions and those that it can remember. It is quite clear that with these alone there are no causal laws possible that will account for the perceptions that we may expect to have anything like as well the assumptions which science makes will do. To make such laws possible we shall certainly have to take into account the perceptions that other people have, that we and they might have, and those that we have had but have forgotten. The question is whether the processes by which the phenomenalist -- who has now ceased to be a pure phenomenalist --arrives at his beliefs in all these other perceptions would not equally justify the plain man's assumption of a real world more or less like what he perceives. We will consider, then, the phenomenalist's answer to the two questions proposed to him on the last page.
(a) Why should it be held to be a priori more probable that what is real is perceptions than that it is something like the objects of our perceptions? To this the phenomenalist would answer: Everyone has perceptions, and perceptions at least must certainly exist whether they themselves are objects of direct awareness or not. On the other hand, the objects of our perceptions are clearly only known to exist in the relation of being perceived. What right, then, have you to suppose that they could exist out of this relation? Thus the argument is that in assuming other perceptions as the real causes of our present ones we are only assuming that the real consists of what we know on other grounds must be capable of existing unperceived; whilst in assuming that it is like the objects of our perceptions, we are assuming that what exists unperceived is like that which is clearly only experienced as perceived, and cannot be proved independently to exist in any other state.
We must consider both sides of this argument carefully. We must keep separate the two distinctions of conscious and unconscious perceptions, and of perceptions which are, and those which are not, reflected upon. These two differences are often confused. Let us begin with that between those that are and those that are not reflected upon. It is quite clear that when I perceive a tree or any of the objects of ordinary life I do not generally reflect upon this perception and say that I know that I perceive a tree3. And it is generally considered that the opposite view would involve an infinite regress in knowledge that is psychologically impossible. On the other hand, my awareness of the tree, whether reflected upon or not, would be called a 'conscious awareness,' as against the so- called unconscious perceptions that I am supposed to have by Leibniz, or believers in the sub-liminal self, when I hear the roaring of the waves. These are supposed to consist of perceptions of the numberless little noises due to the rolling of the separate stones, and they are supposed to differ from any perception like that of the tree in a definite way which would perhaps best be described as a difference of intensive magnitude. Now, it is clear that the distinction between perceptions that are and those that are not reflected upon is a valid one, and can be witnessed by introspection, but there is much more doubt whether the distinction between conscious and unconscious perceptions is valid. Unconscious perceptions have often been introduced where either there was no need to assume any event in the brain or the mind, or where all that was needed was a persistent state of the brain or the mind, or both, which alone does not produce or constitute a perception, but in company with other such states, or under new bodily or mental conditions is capable of giving rise to perception. But there is no more reason for calling a psychical state of this kind an unconscious perception than for calling a match-box an unlighted bonfire. Similarly if we grant that at a given moment there may be unanalysed detail in the object of a perception which attention can discover we do not assume any but conscious perceptions. Before it was discovered it was part of the object of a total conscious perception; afterwards it forms the objects of several new conscious perceptions. Perceptions, then, are either conscious and like our perceptions of trees and chairs, or there is no reason for calling them perceptions at all. And it is certainly not essential to the existence of perceptions that they should be reflected upon.
Hence, to be plausible, the phenomenalist position has to be stated as follows. The assumption that the real causes of our perceptions are perceptions that we have had and have forgotten (i.e. on which we have ceased to be able to reflect) or are perceptions in other people (i.e. perceptions on which we never could have reflected) is so far more probable than any other alternative a priori in that we know that perceptions can and do exist unreflected upon, whilst we do not know that the objects of those perceptions, or anything like them, exists unperceived.
Before discussing this argument in its present form we must say a word in explanation of the second half of it. This is the argument that, since the objects of our perceptions are clearly only known to exist when perceived, it is a greater assumption to suppose that the unknown real causes of our perceptions are like their objects than that they are other perceptions, since we have now seen that other perceptions are able to exist unreflected upon. This does not, of course, mean that when I perceive a tree I perceive it as something perceived by me, for this would be just to deny the true assertion of the other part of the argument that perceptions can and do exist unreflected upon. What it means is that a thing is only directly known to exist while it is actually an object of perception, whether, as a matter of fact, we did or did not reflect at the moment of perceiving it that this was the case. The point, then, is that, whilst there is no ground independent of the success with which the assumption meets in accounting for the occurrence of our perceptions which makes it necessary to suppose that the objects of perceptions, or anything like them, exists unperceived, it is certain that perceptions can and do exist unreflected upon. The assumption that the unknown real causes of our perceptions are other perceptions that are unreflected upon by us is therefore a priori more probable or a less improbable assumption than that they are like the objects of our perceptions.
Now that we have stated this argument as fairly as possible, there are two criticisms to be made upon it, one from each side. From the side of the argument that perceptions can certainly exist unreflected upon, we must ask whether the sense in which this is true is the sense in which it will make the phenomenalist assumption of forgotten perceptions and perceptions in other people as the real causes of our perceptions more probable a priori than the rival assumption. The phenomenalist wants to be able to assume forgotten perceptions and perceptions in other people. He argues that these are just perceptions that are not reflected upon by the person who assumes them, and that, since it is known that such can exist, the assumption of them is a priori the most probable one that can be made about the nature of the real. I think that this argument loses most of its weight when we examine a little more carefully the meaning of reflexion. We shall see, in fact, that, if the assumption of forgotten states of mind in oneself is a priori slightly more probable than that of a real world like the objects of our perception, the assumption of perceptions in other people -- which is certainly necessary for phenomenalism even more than for other views -- is not more probable.
When we say that it is certain that perceptions can and do exist unreflected upon, we mean that we can now remember to have had many perceptions of which we can also remember that at the time at which we had them we did not think or say to ourselves: 'I have now a perception of X.' This, it is to be noted, involves memory. I am directly aware now of the perception that I had some time ago, and I am directly aware that I did not at the moment at which I had the perception have the sort of experience that enables me to say, 'I have such and such a perception.' Such a proposition would, of course, have been true; but I did not then have, as I do now, that direct awareness of my perception at that time which would have enabled me to assert it. You can only show that you have perceptions of which you are not directly aware when you have them by becoming directly aware of them later, and not by the alleged infinite regress that accompanies the opposite view. That regress only applies to the assertion that you cannot know without knowing that you know, and knowing that you know that you know, and so on. But it certainly does not disprove the possibility that every perception might be accompanied by a coexistent awareness of it. We only discover that this is not the case by becoming directly aware through memory of a past perception, and also of the fact that there was no awareness of that perception contemporary with its occurrence4 .
We can now see that phenomenalism has just as much and just as little right to assume other minds as common-sense and science have to assume that the causes of our perceptions are like their objects in general character. It is true that we know directly that perceptions have existed at moments when we were not directly aware of them. But it is also clear that the only perceptions of which we know this are those of which we have been directly aware at some time. Now, the perceptions of other people just are perceptions of which we never can be directly aware. Hence the passage from perceptions in us of which we were not aware when they existed to the assumption of perceptions in other people is not a mere assumption of more of the same kind, but an assumption that perceptions of which we can never be directly aware exist, whilst all that we know directly is that perceptions of which we are sometimes, but not always, aware exist. For all that direct experience can tell us, it might be the case that the only perceptions that can exist are those of which we are sometimes directly aware. Of course, I do not use this as an argument to show that we make a mistake in assuming perceptions in ourselves and others that are never the objects of direct awareness, but merely to show that the phenomenalist who assumes perceptions of which he can never ex hypothesi be directly aware, makes a jump just as much as the man who assumes realities like the objects of his perceptions but which are not perceived. To be strictly fair, however, we must grant that the phenomenalist's jump from solipsism is not as great as that of the man who holds that the real in its general character resembles the objects of his perceptions. To assume that there exist perceptions of which I am never directly aware when I know that there are perceptions of which I am very rarely aware, and which exist when I am not aware of them, is undoubtedly a less assumption than to suppose that something like that which I can never know to exist except as an object of perception exists unperceived.
And in the matter of assuming forgotten perceptions of his own the phenomenalist's position is still less open to cavil. A forgotten perception is one of which I believe that I could at one time have been directly aware, but of which I can not at the present be directly aware. Now, of course, it is true that the only perceptions of which the phenomenalist can be immediately certain that they existed when be was not directly aware of them are those of which he can be directly aware when he makes the judgment, and ex hypothesi perceptions that he has forgotten are not in that position. So that there is a jump into a slightly different territory when I assume perceptions of which I am no longer able to be directly aware. But the jump is less in this case than that which takes the phenomenalist out of solipsism. There he started with perceptions of which he was immediately certain that they existed when he was not aware of them, but could only be thus certain because he could be aware of them when he made the judgment, and assumed perceptions of which he never could be aware. Here he starts from the same set of perceptions and assumes others which only differ from them by the fact that he can no longer be aware of them at will, although he believes that he could have been aware of them at some time.
We must now consider from the other side what is to be said about the phenomenalistic argument that its assumption about the nature of the real is a priori more probable than that of the scientific realist. Is it true that there are no other considerations that ought to be noticed in trying to find the relative a priori probabilities of the rival assumptions about the real beside those mentioned in the phenomenalistic argument? It is clear that in considering the a priori probabilities we ought to take into account all relevant knowledge and belief as well as that about the comparative success with which the two assumptions account for our perceptions. Now, as we have constantly pointed out, people invariably begin with the conviction that what they perceive is real. It is only by arguments that they can be led out of this position, and when they pass out of it to the belief that what they perceive is appearances, the perception of which is caused by a reality, they still do not maintain a perfectly open mind as to the nature of that reality. It is quite true that purely on the grounds of the law of causation there is no reason why a cause should resemble its effect; some causes do so and others do not. But people do not decide here solely on the grounds of the law of causality. They suppose that in this particular case the cause does resemble -- not indeed the effect, which is a psychical event, and is therefore more like the cause on the phenomenalistic theory -- but the object of the perceptions that are the effects. If we started with perfectly open minds from the position that our perceptions as psychical events have causes which exist but cannot be perceived, then the phenomenalistic argument that it is a priori more probable that those real causes should be of the same general nature as perceptions than that they should be of the same general nature as the objects of perceptions would, as we have seen, have a certain though not a great weight. But when we are considering the a priori probabilities in this subject we must take into account, not merely the law of causality and the amount of knowledge and ignorance that it allows us, but also any other relevant judgments about probability that may be made on this point. Now it seems to me clear that people do think that it is more likely that the real shall be like the objects of their perceptions than like their perceptions as psychical events. This is not a belief, as we have seen and shall see, to which the standard arguments for subjective idealism5 can bring any valid objection. Of course there are many beliefs which are very generally held, and the opposite of which it is very difficult to believe, which are as a matter of fact false. But I am not suggesting that it is more probable a priori that the real world should be like the objects of our perception than like the perceptions themselves, because most or all people think so. I merely wish to suggest that when a man recognises that there is no reason why the real world should not resemble in character the objects of his perceptions he will undoubtedly hold that it is more probable that it does than that it does not, and that this belief must be taken into account by every man who has it when he considers the relative probability a priori of phenomenalism and scientific realism. Our beliefs may be wrong, but so long as we have them and no valid argument can be found against them, it is our duty to judge in accordance with them, and not to pretend that they are non-existent.
I think it is fair then to conclude that the phenomenalistic argument, when superposed on the belief that our perceptions have causes, does not succeed in making it a priori more probable that the nature of those realities should be like that of our perceptions rather than that it should be like that of the objects of our perceptions. We will pass then to the relative final probabilities of the two theories, i.e. the probabilities that they gain by the agreement of the perceptions that they predict with those which we actually have. This takes us to the question which on p. 168 I called (b). If two hypotheses be equally probable a priori and explain known facts, their relative final probabilities are dependent on how well they explain the facts. By explaining the facts is meant making that probable which is actually found in most cases to be true. In the present reference there are two points to be noted: (1) Whether it is necessary at all to go outside the perceptions that we now have and can remember; and (2) whether, if we do so, laws entirely in terms of perceptions will explain better than laws in terms of realities whose general nature is like that of the objects of our perceptions.
It is quite clear that if we keep entirely to what we can remember we shall be able to find but few causal laws among our perceptions. We must therefore give up all hope of being able to predict what perceptions we are likely to have in given circumstances or else assume something beside those perceptions that we can at a given moment reflect upon directly. The assumption that we make will have its initial probability increased in proportion to the success which it has in accounting for what is found to be true. The question that remains therefore is what assumption does this best.
Phenomenalism has a plausible argument to prove that the assumption of real causes of our perceptions like their objects in general character cannot explain better than the assumption of perceptions, because the hypotheses are really equivalent. This argument runs as follows: Suppose you do assume that your perceptions are caused by realities that obey certain laws among themselves. The only evidence for the particular kinds of realities and for the laws that they obey are the regularities that can be observed among your perceptions. Surely, then, it would be just as well to say that these regularities justify causal laws among perceptions of the form 'the occurrence of the perceptions p, q, r at t makes that of s, t, u at t + T probable,' as to say that the first set were caused by X, that X makes Y probable and that Y causes the second set. The argument then is that you do not directly need to assume anything fresh. You say that the observed regularities among perceptions are evidence for certain causal laws among them. Then no doubt you will find that these sometimes inexplicably break down unless you assume that you had perceptions on which you cannot now reflect, i.e. forgotten perceptions. The first point to notice about this argument is that, as we have already seen, the amount of regularity among the perceptions that can be remembered at any given time is very small indeed. As a matter of fact, we all assume on no grounds at all that what we perceive is more or less the same as what is real, and when this assumption is made as it is, it enables us to discover and make probable very many regularities which without it would not have been noticed or judged probable. The suggested laws of the form mentioned above connecting the occurrence of perceptions with each other could certainly never have been discovered if it had not been the case that by nature we make the assumption that the real world resembles the objects that we perceive. The phenomenalist then will have to regard our incorrigible tendency to make this assumption as a 'felix culpa,' like that which led to the Redemption by way of the Fall. Still I do not think that we have any right to hold that because we should not have discovered certain propositions if we had not made certain assumptions, therefore the assumptions must be held to be more probable than the propositions which would not have been discovered without them, The essence of the phenomenalist argument is that, in as far as the assumption of the real world with its laws could ever give a verifiable result, it could be replaced by hypothetical propositions about the occurrence of perceptions, and that as far as it could not give a verifiable result it was useless.
I think, however, that the position of the phenomenalist is open to criticism. It unquestionably follows from the laws of probability that if a proposition or set of propositions p strengthen the probability of a proposition q, then if q be found to be true the probability of p is strengthened. Now it is quite clear that the assumption of a real world with certain laws is a set of propositions that strengthen the probability of what is actually found to happen. And it is also clear that it is a different assumption from that of laws connecting perceptions, even though it always has to be verified, if at all, by noting the occurrence of certain perceptions under certain circumstances which must also appear as perceptions. Hence it is quite certain that this particular assumption, like any other that makes probable what is as a matter of fact true, has its probability increased, Of course the phenomenalist would be perfectly right if he said that the discoverable order in our perceptions does not by itself prove the existence of a regular world of reals that cause them. For if any regular system led us to another regular system as probably existing, we should have no right to stop at the first external world. Its regularity would make it very probable that there was another that caused it, and so on. The fact is that the strengthening of probability takes place multiplicatively, and, if the a priori probability of the assumption be small, the final probability of the existence of such an assumed world will be negligible. Hence you cannot prove to a man who thinks it very unlikely that anything exists except perceptions, that it is probable that a real world like the objects of perceptions exists on the ground that it explains so well what actually takes place.
But, granted that no one really is in the position that he thinks it positively improbable that there should be a real world like the objects of his perceptions, except through listening to erroneous arguments, we must compare the strengthening of the probability of such an external world with the strengthening of the probability of laws connecting our perceptions, which are equivalent to the former in all cases capable of verification. Let us take some simple scientific law and consider it -- say the proposition that 'heat causes metals to expand.' The ordinary theory would be that the metals, the temperature, and the length exist whether we perceive them or not; that, under suitable circumstances, we can perceive them; and that, when we do so, we are actually able to perceive that connexion which we believe to hold between change of temperature and change of length, whether we perceive it or not. The phenomenalist position is that it is absurd to assume any more than the law about perceptions, that whenever we have a perception of a metal with a certain degree of temperature we can perceive a certain length, and that this will increase as the felt temperature increases. This, the phenomenalist would say, is all that ever could be verified in the realist's law, and therefore all in his assumption that any verification can render probable. Now if all laws were as simple as this the phenomenalist might be right; but this is not the case. The actual position that we have to face is the following one: In general it is not a case of finding regular conjunctions of perceptions and arguing from this that there is a causal law of the form that the occurrence of p1 strengthens the probability of the occurrence of p2 and makes it practically certain. There are some such regularities among our perceptions, and if we confined ourselves to them, I think it would be true to say that such hypothetical propositions about perceptions were better justified than the assumption of real connexions of events that take place alike when we do and when we do not perceive them. But the propositions at which we arrive as probable when we assume a real world like the objects of our perceptions, are not propositions at which we could have arrived from any actual recurrent series of connected perceptions, for the excellent reason that such a recurrent series will not in general have taken place. Thus we must not merely say that the assumption of real causes of our perceptions obeying certain laws enables us to discover regularities in our perceptions that we should otherwise miss. What we must say is that the assumption actually renders probable propositions about the connexion of our perceptions that would not be rendered probable by our actual perceptions alone, and which are verified. Propositions of the form 'if p occurs it is practically certain that q will occur, where p and q are perceptions,' can only be rendered probable by our perceptions when p and q actually do constantly occur together. But in general both p and q scarcely ever occur at all, and therefore it is true to say that our perceptions alone do not render such propositions probable. But this just means that unless the assumption is probably true these propositions are not probable. Now the less probable a proposition is antecedently, i.e. without the assumption of a given proposition, the more probable does that assumption become if it makes the antecedently improbable proposition soluble, and the latter is as a matter of fact found to be true.
Let us take an example from the wave-theory of light. It can be shown to follow from this that there will be a bright spot in the middle of the shadow cast by a small circular object like a coin. Now the phenomenalist would say: 'Why suppose that this verifies the wave-theory in as far as that involves what cannot be perceived? Why not say at once that it makes probable the law that I shall always perceive a bright spot in the middle when I perceive a circular shadow. This is the only part of the law that can be verified. True I should never have thought of looking for a bright spot there if I had not made the assumption of the wave theory, but this merely shows that the assumption was a fortunate aid to the discovery of a hypothetical proposition connecting my perceptions.' In accordance with what we have said, the following is the right answer to this contention. Certainly it is perfectly true that what has been discovered involves a hypothetical law about perceptions. But if you keep to perceptions the law was not probable unless there was an actual coexistence in all or most cases of the shadow and the bright spot. Now, as a matter of fact, on your theory there was not such a coexistence. It is useless for you to tell us that there was a bright spot, but that you did not notice it, because you did not expect it to be there. If phenomenalism be correct it was not there when you did not notice it. Hence unless the assumption of the wave theory be true, or at least probable, your law about perceptions and their connexion does not merely remain undiscovered, it actually has practically no probability. On the other hand, when you have made the assumption of the wave-theory, and learnt from it that such a spot ought to be found, the fact that you find it and that antecedently to the assumption there was practically no probability of its being there strengthens the probability of the assumption enormously. In fact, the whole criticism of the phenomenalistic argument under discussion might be put as follows: Propositions asserting constant conjunctions hypothetically can only be rendered probable either by the actual observable recurrence of both terms in conjunction or by the fact that they can be inferred from probable hypotheses. In most of the hypothetical laws about the conjunctions of perceptions a constant recurrence of both terms in conjunction is not as a matter of fact experienced. It follows that such laws are not probable unless they follow from some probable hypothesis. But if such an hypothesis can be found, and the conjunction can be verified, even in a single case the very fact that, apart from the hypothesis, it had no particular probability, makes the probability of the hypothesis stronger.
We cannot therefore accept the phenomenalistic position that we may just as well assume hypothetical laws about our perceptions as causes of them obeying among themselves the laws laid down by physics. The laws of physics will indeed in general be able to be stated in terms of connexions between possible perceptions; but the laws so stated will have little or no probability apart from the truth or probability of the assumption about the real world and its laws. And the mere statement of the laws of physics in terms of possible perceptions would not be without difficulty. Atoms, molecules, and ether-waves could not be put directly into phenomenalistic laws. For they are supposed to be of such a nature that they could not possibly be perceived. Hence a phenomenalistic law about them would take the form: If I had the perception p which, ex-hypothesi, I cannot have, I should always have also the perception q which, ex-hypothesi, I cannot have. If, then, we are to be able to carry science to any pitch it is essential that we should be able to allow the existence of real causes of our perceptions as at least possible, and when this is done the most probable laws for them to obey among themselves are those which science finds it necessary to assume in order to account for what is perceived. Now these laws are not in the least like those which perceptions obey among themselves, although they are of course connected with the latter. They are in fact laws about the kind of changes that we can observe in the object of a single continuous perception; and the only common characteristics of the objects of our perceptions and the perceptions themselves is that both have temporal relations and can enter into causal laws. Hence, until any one can make up a theory in terms of laws like those that hold between perceptions which will explain our perceptions better than the theory of science, we shall be justified in holding that if there be a real world at all it probably resembles the objects of our perceptions. But this is subject to the difficulties which have to be dealt with in the next chapter, due to the fact that it has been held that the causal theory of knowledge, which begins by assuming a real world like the objects of our perceptions, ends by overthrowing all knowledge of the real. To the discussion of that theory we now turn.
Notes 1 It must be understood that in this chapter I am not discussing arguments for phenomenalism -- so far as I am aware -- that any phenomenalist uses. The average phenomenalist bases his position on the kind of considerations that Mr Moore overthrew in his Refutation of Idealism. For the purposes of the philosophy of science Mach is the most important phenomenalist. But he has no positive arguments for his position that are worth discussing. He is vitiated by the fallacy that Mr Moore overthrew, and it seems that his main reason for holding the doctrine is that he supposes --for some inscrutable reason -- that it is scientific and 'economical' to a preeminent degree. Anyhow he has now retired behind the formula "Es giebt keine Machschen Philosophie." I have therefore tried to invent the argument which a philosophical phenomenalist might reasonably use, and to criticise it.
2 As with naïf realism so with phenomenalism there is no room for appearance, and the plain man does soon distinguish appearance and reality.
3 This does not of course prove that I am not as a matter of fact directly aware of my perception when I have it, but merely that I do not make a certain judgment which I could not make unless I were directly aware of the perception.
4 How far memory can be taken to prove a negative on such a point I should hesitate to dogmatise.
5 It should be noted, however, that these do not exhaust the arguments on which Idealism has been based. The arguments for esse = percipi do not depend, as Mr Moore seemed to hold in his Refutation of Idealism, on the doctrine that esse = percipi. He forgot such arguments as the Hegelian Dialectic and Lotze's doctrine that substances must be selves. I do not think that these arguments prove idealism, but this is not the place to discuss them.
Contents -- Chapter 4