C. D. Broad, Mind and Its Place in Nature , 1925


The Nature of Traces and Dispositions

      In the last Chapter we saw that the question whether there are unconscious processes which are literally mental involves the question: "What is the nature of those processes which manifest themselves partially through more or less discontinuous series of interrelated conscious experiences, and which presumably fill the temporal gaps between the conscious experiences of such series?" We have already suggested several alternative possible theories on this point; and it is now time to consider them in greater detail with a view to deciding, if possible, between them.

Analogous Facts About Material Substances.

      It will be wise to begin by considering the analogies and differences between the mental facts under discussion and certain facts about material substances. Let us take any material substance, e.g., a circular ring of elastic steel wire. We notice that there are certain characteristics which, strictly speaking, belong to its states rather than to it; and that there are other characteristics, which, strictly speaking, belong to it rather than to any of its states. The wire ring throughout its history always has some shape or other, and it may have the same shape for long periods of its history. If we squeeze it between our fingers it assumes various elliptical shapes; and, when we let go of it, it goes back to what we call its "natural shape", which is circular. We should commonly express these facts by saying that "the shape of the ring is sometimes circular, sometimes elliptical and of eccentricity e, sometimes elliptical and of eccentricity e', and so on". lf we divide up the history of this ring into successive adjoined slices there is a certain determinable characteristic which belongs to all these slices, viz., the characteristic of "having some shape" or even the more determinate characteristic of "having a shape which is some conic section". This can be said to belong to the thing; in the sense that it belongs to all the successive states of the thing, and that, if an event did not have this determinable characteristic, it would not be counted as a state of this thing. But the completely determinate forms of this determinable characteristic, e.g, circularity, ellipticity of eccentricity e, etc., belong strictly to the states of the thing and not to the thing. In the first place, one state may have one of these determinate characteristics, and another state of the same thing may have another of them. And, secondly, even if it should happen that all the states of the thing have precisely the same determinate shape, this is regarded as contingent. An event which had a different determinate shape would not eo ipso be denied to be a state of this thing.

      We have so far distinguished

  1. certain determinable characteristics which may be said to belong to a thing, in so far as they belong to every state of the thing and in so far as any event to which they did not belong would not be counted as a state of this thing. And
  2. completely determinate forms of these determinable characteristics.
Every state of the thing has one or other of these, and it is possible that all its states may have the same determinate form of a certain determinable characteristic. But this is not necessary; an event may have a different determinate form of this determinable characteristic without thereby failing to be a state of this thing.

      We have now to notice a quite different kind of characteristic, which can be said to belong to the thing but not to any of its states. The ring in our example has an inherent tendency to assume an elliptical state of such and such eccentricity when squeezed in such and such a way; and it has an inherent tendency to assume the circular form when left alone. This might be called a "causal characteristic" of the ring. The characteristic of having some shape or of having a shape which is some conic section is not a causal characteristic. And the determinate forms which this determinable characteristic assumes in various states of the thing are not causal, though they are caused by the causal characteristic and the external circumstances. Now causal characteristics of a things may change without the thing being thereby destroyed. If I heated the ring to a certain temperature and then cooled it in a certain way, it would lose its elasticity. After this it would stay in any shape that I squeezed it into, and would have no tendency to pass into the circular shape when I ceased to squeeze it. Such facts as these show that we must distinguish causal properties of various orders in a material thing. We may say that it is a "first order" causal characteristic of the ring to pass into a certain elliptical shape when squeezed and to pass back into the circular shape when released. And we may say that, after being heated, it has lost this first-order causal characteristic and has acquired the first-order causal characteristic of staying in any shape into which I may squeeze it. But it is also a causal characteristic of the ring that, when it is heated and cooled in a certain way, it loses the former first-order causal characteristic and gains the latter. This may be called a "second-order" causal characteristic of the ring. If we denote the two first-order characteristics by p, and p2 respectively, we might denote by p12 the second-order causal characteristic that the substance loses p1 and gains p2 under such and such conditions.

      Now sometimes such changes as these can be reversed either by reversing the original process or by some other means. E.g., by heating up the ring again, hammering it, and cooling it suitably I could make it lose p2 and acquire p1. If this be so we may say that it has a second-order causal characteristic p21 as well as p12. In other cases a change of first-order causal characteristics is not, so far as we know, reversible. It is a second-order causal characteristic of an organism that, if you give it a good dose of arsenic, it loses the first-order causal characteristic of "vital response". And we do not know of any way in which this first-order characteristic can be restored. We must of course recognise the possibility of causal characteristics of the third and higher orders; but there is no need to go into detail about them. All that we need say is that the lower the order of a causal characteristic the more is it possible for this characteristic to change without our saying that the original substance has ceased to exist. Provided that the causal characteristics of higher order remain unaltered, and especially if the changes in first-onder causal characteristics be reversible, we tend to hold that the same substance is still existing.

      We must next notice the connexion between causal characteristics and internal structure. So long as a material substance has a certain causal characteristic we are inclined to believe that it must have a certain characteristic internal structure. The word "structure" must here be taken in a wide sense to include both purely spatial and spatio-temporal structure. It is a first-order causal characteristic of a pillar-box to look red when illuminated by white light and viewed by a normal eye. We ascribe this to a certain persistent spatial structure of the minute particles of the pigment, in virtue of which the surface selectively reflects the red constituent of the white light. If the pillar-box be heated to a high enough temperature it will henceforth appear brown or black under similar conditions of illumination to a normal eye. Thus this first-order causal characteristic will have changed into a different one. We ascribe this change to a change in the minute spatial structure of the pigment on the surface of the pillar-box. Other causal characteristics are ascribed, not to the mere persistence of a certain spatial structure, but to the fact that the minute particles of the body are continuing to move in certain orbits with certain characteristic velocities. Magnetic properties are an example in point. Here the persistence of the causal characteristic is correlated with the persistence of a certain spatio-temporal structure. It is characteristic of modern science as contrasted with mediaeval science to correlate causal properties with minute spatial or spatio-temporal structure, and not to take them as ultimate facts. And there is no doubt that, in the case of material substances, this hypothesis has led to great advances in knowledge and has been "verified" as completely as any such hypothesis well could be.

      Let us consider the essential meaning of this procedure. It comes to this. We correlate a causal characteristic of a substance with a certain non-causal characteristic of its successive states. Let us suppose that the substance has the first-order causal characteristic p1, up to and including the moment t1 and that it then loses p1 and gains p2 instead. We assume that there is a certain determinable non-casual characteristic p which belongs to all the states of the substance both before and after t. And we assume that all the states of the substance up to and including the moment t have this determinable characteristic in the determinate form p1, whilst all the states of the substance after the moment t have this determinable characteristic in the different determinate form p2. In the case of material substances the non-causal determinable p is always supposed to be some general type of internal spatio-temporal or purely spatial structure; and the determinates p1 and p2 are supposed to be different specific forms of this determinable spatial or spatio-temporal structure. Thus there are really two independent assumptions, viz.

  1. that each causal property of a substance depends upon a certain non-causal characteristic of its successive states, that so long as a causal property of a substance remains unchanged its successive states have the same determinate form of this non-causal characteristic, and that when the causal property changes the later states of the substance have this non-causal characteristic in a different determinate form; and
  2. that the non-causal characteristics on which the causal characteristics of substances depend are always certain types of internal spatial or spatio-temporal structure.

      There are two important points to notice before leaving this subject. (1) If the non-causal characteristics on which the causal characteristics of a material substance depend be forms of internal spatial or spatio-temporal structure, they are of the same general nature as non-causal characteristics which we can actually observe. For we can observe that material substances have shapes and sizes, that several such substances can be arranged in various spatial patterns, and that they can move about in various ways as wholes. We are thus merely ascribing to the inside of a material substance and to its minute parts characteristics which are analogous to those which we can perceive in its outside and on a large scale. And we assume that the minute structure and the minute movements which we cannot observe are subject to the same laws of geometry and mechanics as the gross structure and the gross movements which we can observe. (If the Quantum Theory be correct we are probably witnessing a partial breakdown of the latter assumption.)

      (2) The second point to notice is that we cannot wholly reduce causal characteristics to non-causal characteristics by correlating the former with the persistence of a certain type of internal structure. When we say that the movements of the minute internal parts obey the laws of mechanics we are ascribing a certain causal characteristic to them. And when, e.g., we say that a certain minute spatial structure of a body causes it to select and reflect the red component of white light we are ascribing a causal characteristic to this structure. The most that we can do by this means is to reduce a number of characteristics which seem at first sight to be independent and disconnected to a comparatively few fundamental causal characteristics which are familiar on the large scale and are very general and pervasive.

      I think that I have now said enough about the characteristics of material substances, though it would be necessary to go into many more details and to draw many more subtle distinctions if I were professing to give a complete account of this subject. I propose now to consider the analogies and differences between minds and material substances in these respects. In the first place, it is evident that we ascribe causal characteristics to minds as well as to material substances. First of all there are certain general "mental powers", which we regard as characteristic of all minds. For instance there is the power of cognising, the power of being affected by past experiences, the power of association, and so on. These may be compared to the most fundamental causal characteristics of matter, such as inertia, gravitational attraction, etc. Secondly, there are "mental dispositions" which differ from one mind to another and are fairly general in their effects. One man, e.g., is "born good-tempered" and another man is "born irritable". This means that the mind of the former has a causal characteristic such that very few conditions will put it into an angry state, whilst the mind of the latter has a causal characteristic such that very many conditions will put it into an angry state. Now under certain conditions a mind will lose the characteristic of being good-tempered and will acquire the characteristic of being bad-tempered. This may happen suddenly through a wound in the head or gradually through disease or continually irritating surroundings. Here we have an instance of a causal characteristic of higher order belonging to a mind. Some of these changes in the causal characteristics of minds are reversible. A man who has become irritable may be restored to a good-tempered disposition by regulating his liver or by treating him psycho-analytically. "Being irritable" may be compared to "looking red to most people in most lights", and "being good-tempered" may be compared to "looking brown to most people in most lights"; and the characteristic of losing the former and gaining the latter when wounded in the head may be compared to the characteristic of losing one colour and permanently acquiring the other on being heated to a high enough temperature. So far there is obviously a great deal of analogy between the characteristics of mental and of material substances. Let us now consider some of the differences.

  1. Mental substances seem to start mainly with powers to acquire other and more determinate powers. A baby does not at first have the power to talk or to reason, but it has the power to acquire these powers if proper stimuli are applied. And in most cases these stimuli are applied if the baby lives, so that these more determinate powers generally are in fact acquired. I do not think that there is much analogy to this in the case of material substances. And this is not surprising. For a power to acquire some specific power, such as that of talking rationally, could hardly come into action if it were not for the general powers of association and "memory" in its widest sense; and these are common and peculiar to minds.
  2. Although some changes in which a mind loses one causal characteristic and gains another are reversible, most of them, so far as we know, are not. On the other hand, most changes in the causal characteristics of material substances are reversible. Of course this difference may simply be due to the fact that we know much more about matter than about mind, and have much greater practical control of the former than of the latter in consequence.
  3. A mind in the course of its history is continually acquiring new and extremely determinate powers. E.g., it is continually acquiring the power to remember a certain definite event, which it did not have before this event happened. Although it also loses many of these determinate memory-powers with efflux of time, yet, on the whole, through many years of its life the number of determinate memory-powers which it possesses is probably steadily increasing. I do not think that there is anything analogous to this in the case of matter.
It must of course be pointed out that these contrasts which I have been indicating are at their sharpest when we compare a highly developed mind with a bit of inorganic matter. Organised bodies certainly form a half-way house between mind and matter in these respects.

      So far I have been considering differences between the causal characteristics of mental and material substances which are immediately obvious and involve no special theories about these characteristics. I have said that, in the case of material substances, we make two hypotheses which have been amply verified. The first is that, so long as a material substance has a certain causal characteristic, there is a certain correlated non-causal characteristic which belongs to each of its successive states and determines this causal characteristic. The second is that this non-causal characteristic is always a certain type of internal spatial or spatiotemporal structure. Now we do, no doubt, generally assume something analogous to the first proposition in dealing with the causal characteristics of minds. We do assume that, so long as a mind has a certain causal characteristic, there is a certain non-causal characteristic of something which determines this causal characteristic of the mind. But it is very difficult to maintain anything like the second proposition. In the first place, a mind, as such, does not seem to be a spatio-temporal whole; we can, therefore, hardly talk of its spatio-temporal structure. If we want to talk of spatio-temporal structure in this connexion we have to desert the mind and start talking about the brain and nervous system. Again, the spatio-temporal parts of a material substance are themselves material substances; e.g., the molecules of a gas are as good material substances as the gas itself. And the relations of the parts of a material substance within it are analogous to the relations of this material substance to another which is outside it. E.g., the relations of the molecules of a bit of dust to each other are geometrically and mechanically analogous to the relations of a number of bits of dust dancing about in the air. Now a theory of mental structure analogous to this would have to be of the following kind. We should have to suppose that the observable minds of ourselves and our friends are composed of unobservable minds. And we should have to suppose that the unobservable minds which compose my mind are related to each other within it in the same kind of way as my mind is related to other observable minds to form a society. Now, of course, this might have been true, but it seems pretty evident that it is not in fact true. So far as one can judge the unity of an individual mind is not in the least like the unitv of a society of minds.

      The difficulty then is this. If we try to correlate the causal characteristics of minds with minute spatio-temporal structure we are forced to ascribe this structure to the brain and nervous system and not to the mind itself. In that case I think we shall have to admit that we can hardly talk of a purely mental substance. The mind, in abstraction from the brain and nervous system, will be a mere set of mental events with many gaps and very imperfect internal unity. It might be called an "incomplete substance." The only complete mental substance will be not merely mental but also material; it will be the "mind-brain", if I may use that expression. This, of course, is by no means a revolutionary view. St Thomas, though he did not mean exactly what I mean and did not use the arguments which I am using, held that a human soul is in an "unnatural" state when separated from its body, and used this as an argument for the resurrection of the body. If, on the other hand, we try to assign a purely mental structure to our minds but otherwise to follow the analogy of material substances as closely as possible, we land in a different kind of trouble. We then have to regard each observable mind as a society of unobservable minds; and this hypothesis seems not to fit the facts. The result of this is that we cannot get away from the much decried "faculty-psychology". We must remember that before Descartes' time there was a "faculty-physics", and that Descartes' greatest achievement was to show that the various causal characteristics of physical things can be connected with each other by correlating them all with characteristic forms of spatio-temporal structure and a few very general and pervasive causal characteristics. No one has succeeded in connecting the various mental "powers" in any analogous way, and that is why psychology at present hardly deserves the name of a "science". It is, I think, quite certain that psychology will remain in this unsatisfactory state unless and until someone succeeds in doing for it what Galileo, Descartes, and Newton did for physics. And the difficulty of doing anything of the kind is obvious when we remember how difficult it is to conceive of purely mental "structure" or to imagine what can be the few fundamental causal characteristics which, together with differences of "mental structure", will explain and connect the various observable mental powers.

The Theory of Mnernic Causation.

      This is a theory which abandons all attempts to correlate the causal characteristics of minds with any non-causal characteristic. A fortiori it refuses to correlate them with purely mental "structure" or with the spatio-temporal structure of the brain and nervous system. This theory has been tentatively suggested by Mr Russel in Lectures IV and V of his Analysis of Mind. He rather unhappily associates it with the name of Semon, who, so far as I can see, never thought of anything of the kind. Semon was anxious to show that mnemic phenomena are much commoner and more important than has generally been thought. But his theory of their causation seems to be quite commonplace; it is just a theory of traces thinly disguised under the name of "engrams".

      I shall have to draw certain distinctions and go into certain details which will not be found in Mr Russell's book. I will begin by trying to give definitions of "mnemic" and "non-mnemic" events, such that there shall be no doubt that there are mnemic events.

Definition of "Mnemic Events ".

      I shall begin with rough definitions, and shall gradually polish them under the friction of criticism. A gas-explosion might be taken as a typical example of a non-mnemic event, and a memory of a past visit to a certain town as a typical example of a mnemic event. What is the relevant difference between them? The gas-explosion could be fully accounted for by reference to the state of affairs which immediately preceded it. In this we should find a certain mixture of gas and oxygen and the striking of a light; and these together are enough to account for the happening of the explosion there and then. No doubt, if we like, we can go further back. We could predict the presence of this particular mixture by knowing that a gas-tap had been turned on for so long; and we could predict the striking of the light from the fact that a man had filled his pipe and had just taken out his match-box to light it. But there is no need for us to go back to these earlier events. The explosion is directly determined by what immediately precedes it; and it is determined by earlier events only in so far as these determine its immediate antecedents. We will define a "non-mnemic event" as one whose independently necessary conditions are all contained in the state of affairs which immediately precedes it. No one questions that many events are non-mnemic, in this sense.

      Now contrast this with the memory which I now have of a visit which I paid to a certain town last year. In order to account for the occurrence of this memory it is not, on the face of it, enough to refer to what immediately preceded it. We should no doubt find there the stimulus which called forth the memory just then. But it is obvious that precisely the same stimulus might have acted and would have called forth no such memory if I had not visited the town at all. Thus, in order to account completely for the occurrence of the memory now, it seems necessary to go back to the event of last year, viz., my visit to the town. It might, therefore, seem plausible to define a "mnemic event" as one whose independently necessary conditions are not all contained in the state of affairs which immediately precedes it.

      This definition, however, would be unsatisfactory for the following reason. It would leave it uncertain whether there are any mnemic events, and the decision would depend on whether we did or did not accept mnemic causation. This is just what we do not want. On the usual form of the trace-theory all the independently necessary conditions of the memory do immediately precede its occurrence. For this event is supposed to be completely determined by the present stimulus and the present trace which this stimulus excites. On this view the actual visit to the town last year is at most a dependently necessary condition; i.e., the most that we can say is that the trace would not have existed unless the visit had taken place. So that, if the trace-theory be true, the memory is not really a mnemic event in the sense defined above. In order to avoid this complication I shall have to introduce the words "macroscopic" and "microscopic" into my definitions.

These terms were introduced into physics by Lorentz in his Theory of Electrons; and I propose to borrow them.

      I will now define a "macroscopically mnemic event". This will be an event whose independently necessary macroscopic conditions are not all contained in the state of affairs which immediately precedes its occurrence. There is no doubt of the existence of macroscopically mnemic events, in the sense defined. If we make no hypotheses about traces, unconscious mental processes, etc., i.e., about microscopic events, and confine ourselves wholly to what we can perceive and introspect, it is quite certain that events in the remote past are independently necessary conditions of memory-experiences. Thc question at issue between those who accept and those who reject Mnemic Causation can now be stated clearly.

"Are those events which are macroscopically mnemic microscopically non-mnemic, or are they not?"
The trace-theory says "Yes", and the theory of Mnemic Causation says "No".

Causal and Epistemological Conditions.

      We must now notice a distinction which Mr Russell does not explicitly draw. On the ordinary form of the trace-theory, if a similar trace could have existed without the visit to the town having taken place, and if the same stimulus had acted, I should have had a similar experience though I had never visited the town. And the case of hallucinatory memory experiences might be quoted in support of this view. But it would be possible to hold a trace-theory, and yet to hold that the existence of the trace and the occurrence of the stimulus are not sufficient conditions for the occurrence of the memory experience. A person who holds the realistic view that, in all memory-situations, there is direct acquaintance with the actual past event remembered would have to regard the exlstence of the past event as an independently necessary condition of the occurrence of the memory-experience. In order to deal with these possibilities it is necessary to distinguish between the "causal" and the "epistemological" conditions of a macroscopically mnemic event. The causal conditions of a memory-experience are those which would make this kind of situation arise at a certain moment provided that a suitable object exists to be its objective constituent. The existence of such an object, ready to be the objective constituent of the situation if it arises, is what I mean by the "epistemological condition" of the event. We can now consider the attitude which various possible theories would take up towards a macroscopically mnemic event, such as a memory-experience.

      (i) The ordinary form of the trace-theory would hold that both the causal and the epistemological conditions of a memory-situation are non-mnemic when we consider microscopic events and objects. Let us illustrate this, and the alternative theories, by diagrams. Let us represent momentary events by dots, memory-images by circles, persistent traces by crosses, the causal relation by a full arrow, and the cognitive relation by a dotted arrow. Then the ordinary trace-theory of memory is represented by the diagram below.

Here e, a past event, produces a trace t which persists. In course of time a stimulus s excites this trace and produces the awareness of a memory-image i which resembles e. The whole situation "being aware of the image i which resembles the past event e" is the memory of this past event. Here the causal and the epistemological conditions are both ultimately non-mnemic.

      (ii) Let us next consider the trace-theory combined with a realistic view of memory. The diagram is given below.

Here, as before, the past event e produces the trace t which persists and is eventually excited by the stimulus s. But here the result is not to make me aware of a present image resembling the past event. The result is to make me cognise directly the past event e which left the trace. The cognitive relation jumps over the time-gap between stimulus and past event, though the causal relation does not. Thus the causal conditions are here ultimately non-mnemic, but the epistemological conditions are mnemic.

      (iii) We will next consider Mr Russell's form of the Mnemic Causation Theory. The diagram is as follows:

Here the past event e and the present stimulus s together produce by mnemic causation the awareness of a memory-image i which in fact resembles e and is accompanied by a "feeling of familiarity". This constitutes the memory m of the event e. The causal conditions are here irreducibly mnemic, whilst the epistemological conditions are non-mnemic.

      (iv) It would obviously be possible to imagine a still more radically mnemic theory, by combining the theory of Mnemic Causation with the realistic view of memory. This is illustrated in the appended diagram.

Here the causal conditions are as in (iii); but the result is not to produce the awareness of a present image which feels familiar and in fact resembles the past event. The result is to produce a direct awareness of the past event e itself, i.e., a cognitive situation of which e itself is the objective constituent. Here then both the causal and the epistemological conditions would be irreducibly mnemic.

      Of course there are plenty of macroscopically mnemic events where there is no need to introduce the distinction between causal and epistemological conditions. For many such events are not cognitions at all; and many which are cognitions do not have past events as their epistemological objects. Still, memories are the most striking example of macroscopically mnemic events, and with them it is necessary to introduce the distinction. Perhaps the distinction can be made clear to anyone who finds it obscure, by means of the following analogy. Suppose we take the second possible theory, viz., the trace-theory combined with the realistic view of memory. We might compare the past event, on this view, to a lock; the trace to a key made to fit the lock; and remembering the past event to undoing the lock with the key. The existence of the lock plays two different parts in determining the event of unfastening the lock.

  1. The lock was originally made; then someone took a wax model of it; and then someone cut a key from this model to fit the lock, and this key is used from time to time to unfasten the lock. (Those of my readers who are either professional burglars or fellows of colleges with a weakness for losing their fellow's keys will be familiar with the causal sequence which I have been describing.) This corresponds to the original event as an immediate causal condition of the trace and a remote causal condition of the memory.
  2. The act of undoing the lock with the key cannot occur unless the lock exists in its original form to put the key into. This corresponds to the original event as an independently necessary epistemological condition of the memory."

"Temporal Separation" and "Immediate Precedence".

      There is one other notion which is involved in our definitions of mnemic and non-mnemic events, which needs to be cleared up. Real events are not momentary, but have a finite duration; they are like lines and not like points. This implies that "momentary events" are not literally constituents of finite events, as short events are of longer ones that overlap them. They are complicated functions of finite events, and have to be defined by Extensive Abstraction in the way which Whitehead has shown us. Again, the time-series is supposed to be continuous; and this implies that, when we "analyse" finite events into "momentary events", no two of these momentary events will be next to each other, as two successive railings of a fence are. When we say that all the independently necessary conditions of an event are contained in the state of affairs which immediately precedes it, we seem to imply

  1. that all these conditions are "momentary",
  2. that they all belong to the same moment, and
  3. that this moment "immediately precedes" the moment at which the event begins.
These statements are all Pickwickian, and we must now interpret them.

      Let us represent the event e, which is to be the effect, by a short line bc. Let the various conditions which are severally necessary and jointly sufficient to produce e be contained in various earlier slices of the world's history. These may be represented by a number of lines ab, a'b, a"b, which all end at b the beginning of e. Thus --

      Now take short slices along ba, ba', and ba"; e.g., bx, bx', and bx''. Suppose we find that, no matter how short we make these stretches, they still contain all the independently necessary conditions of e. Then we can sum this up by saying that all the conditions of e are "momentary" and that they all "immediately precede e."

      If we did not find this to be true, several alternatives would be possible.

  1. We might find that all the independently necessary conditions of e were momentary, but that they did not all immediately precede e. For instance, we might find that the stretch bx fails to contain a certain necessary condition of e, but that the stretch xx contains this missing condition, no matter how near x be to x. Then we could say that e has a momentary condition which is separated from it by the time-gap bx.
  2. We might find that some of e's conditions are neither momentary nor immediately precedent to e. For instance, it may be that if xx be made too short it will fail to contain a certain necessary condition of e. There may in fact be certain characteristics which determine by their occurrence the occurrence of e, and need a certain minimum stretch of duration to inhere in. E.g., if e were partly determined by a certain characteristic rate of vibration, it would seem that this could not inhere in a smaller duration than that taken by one complete period of the vibration.
  3. It might happen that, whilst some of e's conditions are not momentary, yet the non-momentary conditions are continuous in time with e. This would mean, e.g., that one of the necessary conditions of e is the pervasion of a certain finite stretch such as bx by a certain nonuniform characteristic, but that this stretch ends at the same moment as e begins.

      We can now give a more accurate definition of non-mnemic and of mnemic events. A non-mnemic event would be one whose "momentary" conditions (if it has any) all "immediately precede" it, and whose non-momentary conditions (if it has any) are all continuous with it. A mnemic event would be one which has at least one independently necessary condition which is separated from it by a finite gap in time. It does not follow that this gap may not also contain conditions which are necessary for e's occurrence; the point is that they are not sufficient without the condition which precedes the gap. We must also recognise the theoretical possibility that a remote condition of an event might be both an independently and a dependently necessary condition of it. Suppose, e.g., that a "momentary" event at x determines the filling of the stretch xb, and that e is a function both of this event and of the filling of xb. Then this remote event will be an independently necessary condition of e; but it will also be a dependently necessary condition of e, in so far as it determines the filling of xb which in turn partially determines e.

      The important point for us to notice is that what is characteristic of mnemic causation is the time-gap between an event and some of its independently necessary conditions. The question whether the conditions are or are not "momentary" is not the distinguishing mark. And the question whether this time-gap does or does not contain other necessary conditions is not relevant, so long as it does not contain all the independently necessary conditions. It is important to see this. For it is almost certain that there is causation in which the conditions are not "momentary", in the sense defined above. Therefore, if we confuse mnemic causation with causation in which some of the conditions are non-momentary, we shall be liable to accept the former on grounds which are relevant only to the latter.

Criticism of Mnemic Causation.

      We may sum up the differences between the trace-theory and the theory of mnemic causation as follows. Whenever we have a macroscopically mnemic event there is a time-gap between the event and some of its independently necessary macroscopic conditions. The trace-theory holds that such a gap cannot be an ultimate fact; all the independently necessary conditions of an event must be continuous with it, if they be non-momentary, and must "immediately precede" it, if they be "momentary". We have explained the Pickwickian phrases in inverted commas. Hence the trace-theory holds that the past experience is not an independently necessary causal condition of the memory of it; and it has to fill the gap by postulating hypothetical microscopic entities, viz., traces, which are produced by the past experience and persist into the present. The mnemic theory, on the other hand, is prepared to accept as an ultimate fact that some of the independently necessary conditions of an event are neither continuous with it nor "immediately precede" it. It is prepared to bridge the temporal gap by postulating a special kind of causal relation.

      On a first inspection each theory is seen to have its characteristic merits and defects. One keeps to the familiar kind of causal relation, but has to postulate purely hypothetical persistent entities; the other keeps to events which can actually be observed by introspection, but has to postulate an unfamiliar kind of causal relation. In dealing with matter I do not think that we should hesitate for a moment between the two. The notion of a hidden minute structure in matter is perfectly familiar to us; we know that our unaided senses cannot distinguish the finer divisions of matter, and that microscopes often reveal a highly differentiated structure in what seems quite homogeneous to the naked eye. Moreover, we know that the details thus discovered very often explain the behaviour of the bodies in which we discover them. Hence the idea of material structure and states too minute for us to perceive even with the microcsope is almost forced upon us; and it is reasonable to suppose that the not very numerous macroscopically mnemic phenomena which are observed in the inorganic realm are microscopically non-mnemic. It is much more harder to conceive of a microscopic mental structure, and this is the only reason why we are tempted to introduce the theory of mnemic causation into mental phenomena. Since the mind and its states are not in any obvious sense extended, the idea of a structure which cannot be observed because of its spatial minuteness fails us here. The notion of non-introspectible and perhaps unowned series of mental events, which otherwise resemble our conscious experiences, is not easy for us to grasp; and it is hard for us to give a meaning to mental events which do not resemble our conscious experiences. Hence the main motive for considering favourably a mnemic-causation-theory for mental phenomena is simply the difficulty which we have in conceiving the intrinsic nature of traces and dispositions unless we are prepared to regard them as purely material modifications of the brain and nervous system. Against this we might put two considerations.

  1. Although the mental events which we can introspect do not seem to be extended they do seem to have different degrees of intensity, and it does seem to be harder to discriminate them introspectively as their intensity decreases. Thus it might be possible to substitute low intensity for small extension, and to conceive of traces and dispositions as being of the nature of ordinary conscious mental events but of very low intensity.
  2. Some people would certainly hold that there are a priori objections to the whole notion of mnemic causation, and that we must therefore adopt a trace-theory, no matter what difficulties we may have in picturing to ourselves the intrinsic nature of traces and dispositions.

      I propose now to say something about this last point. In order to discuss it adequately it would be necessary to enter in great detail into the nature of causation in general. This would be out of place in the present connexion. I shall, therefore, confine myself to a few remarks which seem relevant and do not carry us too far afield.

/First Remark/

      (1) In answering objections against the possibility of mnemic causation Mr Russell assumes that causal laws are merely assertions of regular sequence, and that any true assertion of regular sequence is a causal law. He supposes an objector to say that, in mnemlc causation, some of the independently necessary conditions of an event have ceased to exist long before the event begins to happen. And then the objector is supposed to raise the difficulty: "How can anything act after it has ceased to exist for a finite time?" To put the objection in a concrete form: --

"According to the theory of mnemic causation my perception of a town which I visited last year literally produces a memory of this event whenever a suitable stimulus acts on me. But the perception is long past and is in no sense continued into the present. It has ceased to exist itself, and nothing now exists which can be regarded as a continuation of it. How then can it do anything now?"
Mr Russell answers that this objection presupposes the activity theory of causation, which is now rejected by most philosophers. And he goes on to say that causation simply means regular sequence; and that, with this interpretation, there is no a priori objection to mnemic causation. By saying that C causes E, on this view, we simply mean that C is a set of conditions c1, c2, . . . cn, such that
  1. whenever they are all fulfilled E happens, and
  2. whenever E happens they have all been fulfilled.
This says nothing about c1 . . . cn being all of the same date and all "immediately preceding" E. Hence, if this be all that we ever mean by saying that C causes E, mnemic causation is antecedently quite as possible as non-mnemis causation; and it becomes a mere question of fact (which could never be conclusively settled) whether there is mnemic causation .

      For a complete discussion it would be necessary, first, to consider whether causation does mean nothing but regular sequence. If we found that it did involve something more, the next question would be whether this extra factor would be inconsistent with the possibility of mnemic causal laws. And, lastly, we might ask whether, even if causation is simply regular sequence, it is true to say that past experience and present stimulus are suffcient by themselves to cause a memory of the past event.

      (a) It is, of course, impossible for me to give an adequate discussion of the meaning of Causation here. I will simply say that, even if all causation involves regular sequence, I very much doubt whether all regular sequence would be counted as a causal law. I should say that there are many cases where we should admit regular sequence and unhesitatingly deny causation; though there are perhaps no cases where we can unhesitatingly assert causation in addition to regular sequence. I do not propose to do more here than to show that the line of argument by which the doctrine that causation is simply regular sequence is commonly supported is a very weak one. The argument generally takes the following form. The plain man starts by believing that there is something in causation beside regular sequence. His oppanent then asks him to state clearly what this extra factor is. The plain man is then inclined to say that causation involves "activity" or "necessity" or both, in addition to regular sequence. His opponent then tries to show that the notion of "activity" is just an illegitimate extension to all cases of causation of certain characteristics which accompany the very special experience of voluntarily initiating an action. He also argues, on the lines of Mr Hume, that no causal law is found to be necessary on careful reflection. Thus the plain man finds that the two marks by which he proposed to distinguish causation from mere de facto regularity of sequence vanish under his opponent's criticisms; and he has to admit that he cannot state any factor which differentiates a causal law from a mere statement of regular sequence. His opponent then argues that this failure to state the difference is due to there being no difference to state; and the plain man is reduced to silence, though not altogether to conviction.

      If we reflect we shall see that this is a very poor argument for the purpose. Suppose that causation did involve an unique and not further analysable relation. It might be that regular sequence was not even part of what we mean by causation, but was merely a sign (though by no means an infallible one) by which the presence of this other relation is indicated. If this relation be unique and unanalysable, like the relation of inside and outside in space, for instance, it will be impossible to define it in any but tautological terms. Thus the failure to define anything in causation except regular sequence may be due, not to the absence of an extra factor, but to its being ultimate and unanalysable. Our extreme unwillingness to admit that causation as nothing but regular sequence, and the extreme paradoxes to which any such views lead (cf. Mr Russell's examples about the hooters at two distant factories which both sound at the same time and therefore are both equally causes of either set of workmen going to their work) suggest strongly that there is something in causation beside merely regularity of sequence. On the other hand, I think we must admit that, if there be a peculiar relation involved in causation, we are seldom if ever directly acquainted with it as we often are directly acquainted with certain spatial and temporal relations. I cannot perceive the causal relation by any of my senses, as I can perceive that one thing in my field of view is to the right of another thing in my field of view. And I cannot, I think, ever be perfectly certain on reflection that A causes B in addition to regularly preceding it. (If there be any exception to this, I think it is in my voluntary initiation of certain changes.) But I think that I can be absolutely certain that I do not mean the same thing by "A causes B" and "A is regularly followed by B". And I think that I can often be quite certain that A does not cause B in spite of complete regularity of sequence between the two. E.g., I am quite sure that the hooter of a factory in Manchester does not cause the workmen of a factory in London to go to their work, even though the Manchester hooter does always blow just before the London workmen start to wend their way to the London factory.

      (b) So far I have suggested that regular sequence may be no part of what we mean by causation, but that it is one of the signs by which we judge with more or less conviction that the causal relation is present. It may be, however, that regular sequence by itself is not an adequate sign of the presence of the causal relation. It may be that only certain kinds of regular sequence are trustworthy as signs of the causal relation. And, again, even if it be held that all causation is regular sequence and that there is no specific and unanalysable factor in causation, it might still he held that only certain kinds of regular seguence are cases of causation. I think that this must be admitted in view of our refusal to regard the sequence of the blowing of the Manchester hooter and the movement of the London workmen as an instance of causation. Now the missing factor seems to be a certain spatio-temooral continuity between the sequent events. I am inclined to think that it is the absence of such continuity between the blowing of the Manchester hooter and the movements of the London workmen which makes me so certain that the former is not a cause of the latter. I think that it is the absence of the required temporal continuity between "cause" and "effect" which is the real basis of the objection to mnemic causation which Mr Russell has dismissed as due to the ghost of the activity theory of causation. If such continuity be essential to the notion of causation then "mnemic causation" can be dismissed, even if we admit that there is no specific and unanalysable causal relation. For only certain kinds of regular sequences will count as "causal"; and the sequence of past experience and present memory is not of the required kind. My own view is that I do not mean by "causation" any kind of regular sequence; but that certain kinds of regular sequence are fairly trustworthy signs of the presence of the causal relation. But the final result is the same. For the sequence of past experience and present memory is not of the kind which I regard as a trustworthy sign of the presence of a direct causal relation between the two.

      (c) Let us now ask ourselves the question: "Suppose that causation were simply regular sequence, and that any and every kind of regular sequence were causation, should we be justified in holding that a past experience and a present stimulus are the complete cause of a present memory?" Suppose that I visited a certain town two years ago, and that last year someone mentioned its name to me and I thereupon remembered my visit to it. Suppose that some time after this I had a bad illness or an accident; it might well happen that, if the name were now mentioned to me, I should not remember my visit to the town. Such cases are of course very common. It follows at once that the two conditions, past experience and present stimulus, are not jointly sufficient, though they may he severally necessary, to cause a memory even on the most extreme form of the regularity-theory of causation. For the memory does not regularly follow on the fulfilment of these conditions and of these alone. In fact, we have made the common mistake of ignoring a condition which is just as necessary as the rest, but is unexciting and is much more often fulfilled than not. We talk carelessly of a gas-escape and a spark as the cause of an explosion. But the presence of Oxygen is equally necessary and much more likely to be forgotten, because this condition is nearly always fulfilled, whilst it is much less common for gas to be escaping or for sparks to be flying about. In the same way something which may vaguely be called "the general integrity of the brain and nervous system" is at least as necessary as the past experience and the present stimulus if the memory is to arise. We forget this condition because it is generally fulfilled while we are alive. Now this condition cannot be given a definite date. If it breaks down anywhere between the original experience and the subsequent stimulus, the memory is liable not to arise. It seems to me extremely unlikely that there is any such thing as mnemic causation, even on the extreme regularity-theory which Mr Russell assumes, if by this you mean that a number of conditions separated in time from each other and from the event which they are supposed to cause are jointly sufficient to cause this event. And it is perfectly certain that memories are not completely determined by past experience and present stimulus in the sense that they regularly follow on the fulfilment of these two conditions alone. To put the matter generally, I should say that even on the regularity-theory of causation the complete cause of any event involves persistent as well as transient conditions. Even if the transient conditions be separated from each other by temporal gaps these gaps must be filled with persistent conditions which stretch right up to the beginning of the effect. Mr Russell in his account of mnemic causation seems to have made the common mistake of mentioning the transient and forgetting the persistent conditions.

      It is, however, quite easy to rectify this oversight, and then we can see precisely where the difference between mnemic causation and ordinary causation would lie. The difference remains considerable. On Mr Russell's theory, as modified to meet the above criticism, there is a persistent condition involved in memory, but it is general and not special. This persistent condition is just the general integrity of the brain and nervous system, which existed before as well as after the past experience and was in no way modified by it. On the trace-theory there is a special persistent condition, which was started by the past experience and would not have existed without it. The two independently necessary conditions of the memory are this special persistent and the stimulus. The difference can be seen most clearly as follows. On the trace-theory, if you were to take a cross-section of the history of the experient's body and mind anywhere between the past experience and the stimulus you would find something, viz., the trace, which corresponds to and may be regarded as the representative of the past experience. On Mr Russell's theory, even when modified to meet the above criticisms, these intermediate slices, though relevant and necessary, would contain nothing which corresponds to and represents the past experience. In mnemic causation we should have the following situation. Although there is continuity between the total cause and the effect (since one esseutial part of the cause is a general persistent condition which fills the gap between its earlier and later transient parts), yet there is not continuity between the effect and each independently necessary factor in the cause. The original experience is not joined on to the memory either directly; or by transmission of a disturbance through a medium, as in the case of light or sound; or by some special persistent which represents it, as on the trace-theory. If the possibility of mnemic causation is to be denied, it must be on the ground that one or other of these special kinds of continuity is needed in addition to the merely general continuity which the integrity of the brain and nervous system provides. I am not prepared to assert that this additional dose of continuity is needed; and, therefore, I am not prepared to deny the possibility of mnemic causation, as modified by us in the course of the discussion.

/Second Remark/

      (2) The second remark which I wish to make about mnemic causation is the following. Suppose that c1 . . . cn are independently necessary and jointly sufficient transient conditions for the happening of an event e. We will not now insist that they must all be contemporary with each other, or that they shall "immediately precede" e or be continuous with it. But I do think that we should expect there to be some characteristic time-relation between them. Surely we should expect the law to be at least of somewhat the following form:

"Whenever c1 is followed by c2 after the interval t12, and by c3 after the interval t13, . . . and by cn, after the interval t1n, e follows; and e does not happen except when c1 . . . cn have all happened with these characteristic intervals between them."
No doubt in any causal law the absolute dates of the various factors are variable; but one would expect the relative dates to be constant and characteristic of the law. Now let us apply this to the case of past experience, present stimulus and memory. Assuming the general persistent condition to be fulfilled, we find that the memory arises whenever the stimulus is given, so long as it is after the original experience; i.e., there is no characteristic interval between the two transient conditions in the supposed mnemic causal law. Now I should be inclined to suppose that, whenever this is so, we have not got an ultimate causal law but only an empirical generalisation of the very crudest kind.

/Third Remark/

      (3) There is one other point rather closely connected with the above. Suppose we ask ourselves: "What is our usual test for the persistence of anything which is not under continuous observation?" I think that we should have to answer somewhat as follows. Suppose we find that throughout a long period of time whenever a certain condition C is fulfilled a certain result E immediately follows. And suppose we know that C by itself is not sufficient to produce E. Then we always assume that there is another persrstent factor P with which the variable factor C co-operates to give the result E. E.g., one of my main reasons for believing that there is a persistent something, called "my table", in my room is that throughout a long period of time whenever I look in a certain direction I become aware of an appearance of the table. I know that the mere fact of looking in this direction is not a sufficient condition of sensing this particular kind of appearance, and I assume that the other necessary condition is the persistent something which I call "my table". But this is almost exactly parallel to remembering my past visit to a certain town whenever the proper stimulus is applied. I know that neither the stimulus nor the mere general integrity of brain and nervous system is enough to account for the occurrence of this particular memory at this particular time; and I assume that there must be some other condition which is persistent. The plain fact is then that we have precisely the same kind of reason for believing in persistent traces as we have for believing in the persistence of tables when they are not under direct observation. If this test for persistence be a valid one, we ougrht to apply it to memory as well as to perception; and in that case we shall have to accept something like the trace-theorv if we refuse to apply the argument to traces we ought not to apply it to tables. We ought presumably to hold that later table-sensations are mnemically caused by the first table-sensation and the subsequent acts of looking in a certain direction; and we ought to reject the notion of a persistent physical object as a superstition unworthy of the "Free Man". No one (not even Mr Russell in any of his published works) does in fact take this alternative about the table; and it seems scarcely consistent to take it about memory and then to refuse to extend it to the precisely parallel case of successive perceptions of what we call "the same thing".


      I will now sum up the results of this discussion about mnemic causation. I have tried to explain clearly the difference between a mnemic and a non-mnemic event, and between mnemic and non- mnemic causation. In so doing I have stated the literal meaning of certain Pickwickian phrases which are used in the definitions. I have distinguished between the causal and the epistemological conditions of memory, and have explained and illustrated the four possible types of theory which arise when we allow the two kinds of condition to be either mnemic or non-mnemic. I then considered the arguments for and against mnemic causation in psychology. The only argument that I could find for it was the difficulty of conceiving the intrinsic nature of traces unless we take them to be purely material and thus pass outside the sphere of pure psychology. On the other side I argued that it is very doubtful whether causation can be reduced to mere regular sequence, and quite certain that not all kinds of regular sequence would be counted as instances of causation. If, then, the objections to mnemic causation can be answered only on an extreme fonn of the regularity-theory of causation, it is doubtful whether they can be answered at all. I then showed that, even on a pure regularity-theory of causation, it is certain that the past experience and the present stimulus are not jointly sufficient to cause a memory. At the very least a general persistent condition, which fills the gap between the two, is needed also. If this be granted, the difference between the trace-thcory and the theory of mnemic causation depends on whether a general persistent condition is enough or whether a special persistent condition, which depends on and "represents" the past experience, is also needed. I did not profess to be able to give a conclusive answer to this last question. But I pointed out that we might fairly expect a genuine mnemic law to involve characteristic time-intervals between the various independently necessary and non-contemporary transient conditions of an event. And we do not find this to be so in the case of memory. On the contrary we have here exactly the kind of situation which anywhere else would make us postulate a special persistent condition.

      I do not pretend to have absolutely refuted the possibility of mnemic causation as an ultimate fact in mental life. But I do think that I have shown that we have very little ground for accepting it, and that we have exactly the same kind of evidence for the existence of traces and dispositions as we have for the persistence of physical objects when they are not under continuous observation. Under these circumstances I think we shall do well to accept some form of the trace-theory until some philosopher has successfully applied the theory of mnemic causation, not only to the special case of mental phenomena, but also to all cases where we assume the existence of special persistents in spite of their not being under continuous observation. Henceforth, then, I shall assume that there really are such things as traces and dispositions, i.e., speclal "mnemic persistents," to revert to a name which we have already introduced. We must now consider the question: What is the intrinsic nature of mnemic persistents?"

The Nature of Mnemic Persistents.

      Let us begin by considering once more the case of material substances. We have agreed that here the causal characteristics are correlated with and dependent upon certain persistent non-causal characteristics. But, if we inquire more closely into the nature of the latter, we find that two different cases arise.

  1. We may have both identity of stuff and persistence of structure. Take, e.g., the pillar-box which looks red whenever it is viewed in white light by a normal eye. Here we have the same particles persisting, and at each moment they have the same spatial structure.
  2. Contrast this with the case of an organism. Here we have persistence of structure with continual change of stuff. Matter is contmually passing into and out of the organism; and the causal characteristics of the oganism depend on the fact that, as new matter comes in, it is continually organised and arranged in the same characteristic way and thus replaces the matter which is continually going out.

A purely Mental Theory of Traces.

      Now we might have a theory of purely mental traces analogous to the first of these possibilities if we adopted the Pure Ego theory of the self. The Pure Ego itself would be the persistent identical "stuff". And the causal characteristics of the mind might be correlated with various persistent states of the Pure Ego. On this view the existence of a trace would be the fact that the Pure Ego has a certain determinate non-causal characteristic at every moment within a certain period of time. Of course the analogy to the first possibility about material substances is only partial. There we had persistence of stuff and persistence of structure. Here, so far as we know, there would be no question of structure. There is persistence of stuff and an identical determinate quality possessed by this stuff for a certain period of time. But there is no reason to suppose this determinate quality in the possession of a certain internal structure; for the whole notion of internal structure may be nonsense as applied to the Pure Ego.

      Could we conceive of a theory of purely mental traces without assuming the Pure Ego theory of the self? I think that we could, and that it would be partially, though not exactly, analogous to the second possibility about material substances. I will begin by pointing out that even a purely physiological theory of traces is incompatible with persistence of stuff. Suppose we compare a scar, due to a burn, with a dent in a leaden ball, due to a blow. The dent is simply a persistent spatial rearrangement of the same particles of matter as were present in the leaden ball before the blow dented it. The scar, on the other hand, may persist years after every particle of matter which was in the body when it was burnt has left it and been replaced by other matter. What happens is that the new matter as it comes in is continually arranged so that we still have the scar. Even if traces be purely physiological they must be of the same nature as the scar and not of the same nature as the dent; i.e., they persist througll the same form being continually imposed on fresh matter, and not through the same matter retaining a certain form wllich has once been imposed on it.

      Now I have already said that, in many respects, an organism is a kind of half-way house between an inorganic material substance and a mind. It is, therefore, tempting to see whether we could not conceive of purely mental traces as analogous to scars in organic bodies. I will first point out where it seems to me that the analogy does not hold. Matter enters organisms from outsidc, is elaborated and arranged within them, remains there for some time, then gradually breaks down and is ejected. If we regard a mind as a complex whole of interrelated mental states, we can hardly suppose that new mental states come into the mind from elsewhere and pass from it after a while. If we take the "parts" of a mind to be its states, we must admit that the "parts" seem to be so dependent on the whole that it is doubtful whether they could have existed before they became parts of this whole or could exist after they have ceased to be parts of this whole. Another difference between a mind and an organism is the following. An organism exists continuously from birth to death, and at any two moments which are reasonably near together a great deal of the matter which composes it is the same. A mind, on the other hand, seems to cease to exist for considerable spells during dreamless sleep or fainting-fits, and then to take up its existence again at the point at which it left it.

      If we want to have a purely mental theory of traces we must first fill these gaps with something other than the persistence of a certain structure and the continuance of certain processes in the brain and nervous system. On the Pure Ego theory such gaps are filled by the continued existence of the Pure Ego and by the fact that the Pure Ego has certain determinate qualities throughout the whole period. But we are now trying to do without the Pure Ego theory. The only possible expedient is to suppose that the gaps are filled by literally unconscious and literally mental states and processes, which have to each other relations of the same kind as the conscious mental states and processes of waking life have to each other. These unconscious mental states and processes will not themselves be traces; but, if we are prepared to grant their existence, we can give a theory of purely mental traces without assuming the Pure Ego theory. This can be done as follows. The rclations between our ordinary conscious experiences, and the qualities of our ordinary conscious experiences, may justly be called "mental relations and qualities". But they are not themselves experiences, either conscious or unconscious. Now I would suggest the fpllowing as a possiblc theory about traces. Just before a certain moment my total state of mind consists of a set of mental events having certain qualities and standing in a certain characteristic relation to each other. Let us call these events e1, e2, . . . en, and let us denote the relation which binds them all together into a single state of my mind by R. Then the total state of my mind just before t may be symbolised by R (e1, e2, . . . en). Let us suppose that at t a "new" mental event happens and forms part of my total state of mind at t. We will call this event E. By calling it "new" I mean that it is not a "continuation" of any of the events e1 . . . en; it might, e.g., be a sensation due to someone suddenly sticking a pin into me. Most of the mental events which compose my total state of mind at t will be continuations of events which composed my total state of mind just before t, but probably some of these will not be continued. Let us suppose that e1 . . . em are continued as e'1 . . . e'm, whilst em+1 . . . en, are not continued. My total state of mind at t may then be symbolised by R(e'1 . . . e'm, E). Now I suggest that the presence of E modifies the qualities of e'1 . . . e'm, or of some of them, in a characteristic way, so that those of them which are continued into my total state of mind just after t are continued in the specially modified forms e''E1 . . . e''Em. It is also possible that there is a characteristic modification in the relation which binds them together, so that it is now RE instead of R. On this hypothesis my total state of mind just after t is of the form RE(e''E1 . . . e''Em), assuming for the sake of simplicity that no further "new" experience has taken place. We have now got our "trace" formed. We must next assume that this "E-quality" or this "E-relation" is henceforth imposed on the contents or the structure of each successive total state by the state that precedes it, very much as the scar is imposed on the new matter which comes into an organism from outside. On some such lines as these we can conceive of purely mental traces without needing to assume a Pure Ego; provided we are willing to admit that there are no real gaps in mental life and that the apparent gaps are filled up by non-introspectible mental events, which are of the same general nature and have the same kind of mutual relations as those which we can introspect. The trace is not itself a mental event, but is a characteristic modification in the qualities of mental events or in the relation which binds contemporary mental events into a single total state of mind. And this characteristic modification of quality or structure is imposed on each total state by the total state which immediately precedes it.

      I do not think that it would be necessary to suppose that all the events in any total state have this characteristic qualitative modification imposed on them. So long as some events in every total state after the occurrence of the "new" experience are modified in this way, we have as much as we need. Again, I do not think that it would be necessary to assume that the relation which is modified in a characteristic way is the relation which binds together all the events of a single mental state. It would suffice if some relation which binds together a sub-group of contemporary mental events were modified in this characteristic way and if this modified relation were handed on from this sub-group in any total state to the corresponding sub-group in its immediate successor.

      It is then possible to conceive of a purely "mental" theory of mind without assuming the Pure Ego theory. A mind will, on this view, be composed entirely of mental events. Some of these are introspectible and others are not. Again, certain of these mental events will be related to each other so as to form series of a characteristic kind. Such series will be mental processes. Some mental processes will be wholly imperceptible, i.e., none of the successive mental events which compose them will be introspectible. Others will be perceptible, i.e., some of the mental events which compose them will be introspectible and will be objects of simultaneous undiscriminating awareness. But there will be imperceptible parts of perceptible mental processes. On the present view, the imperceptible parts of perceptible mental processes, and the wholly imperceptible mental processes, will be in all other respects of precisely the same nature as the introspectible parts of perceptible mental processes. All mental processes will depend on traces and dispositions. But these traces and dispositions, though not themselves mental events or processes, will be purely mental; for they are just certain qualities of the mental events of one total state of mind, which are handed down to the mental events of the next total state of mind, and so on indefinitely. Or they are just certain relations between the mental events composing one total state of mind, which are impressed upon the mental events composing the next total state of mind, and so on indefinitely.

A purely Physiological theory of Traces.

      Let us now consider the alternative which lies at the opposite extreme to that which we have been treating above. On this view traces are simply modifications in the minute spatial or spatio-temporal structure of our brains and nervous systems, which are propagated from one state of the brain and nervous system to the next state in the way in which a scar on one's arm due to a burn is propagated for the rest of one's life. I think that the natural complement of such a theory of traces would be to hold that the non-introspectible "parts" of perceptible mental processes are not strictly mental at all, but are purely physiological events. And, in that case, of course they are not, strictly speaking, "parts" at all. The position will be that there are certain physiological processes, some parts of which are accompanied by mental events which depend on them, and other parts of which are not accompanied by mental events at all. And there will be other physiological processcs which are exactly like those which are accompanied by mental events except in the fact that they are not accompanied by any mental events. Thus "unconscious mental processes" will not really be mental at all: and the natural accompaniment of a purely physiological theory of traces is an epiphenomenalist theory of the nature of mind. The mind ceases to be a genuine substance theoretically capable of existing in its own right.

      I say that this would be the "natural" complement of a purely physiological thcory of traces and dispositions. I do not say that it would be an absolutely necessary consequence of such a theory. It is possible that those parts of a certain physiological process which are not accompanied by introspectiblc mental events are accompanied by non-introspectible mental events of a similar kind to the introspectible mental events wllich accompany other parts of the same physiological process. And it is possible that those physiological processes, no part of which is accompanied by introspectiblc mental events, are nevertheless accompanied by non-introspectible mental events. But, although this hypothesis would be possible on a purely physiological theory of traces and dispositions, it would seem to be quite unmotived. At most it could be supported only by arguments from analogy. Since some parts of a certain physiological process are accompanied by introspectible mental events, and since the later of these events seems to be an obvious development of the earlier, it might be argued that the part of the physiological process which fills the gap between two such introspectible mental events must be very much like those parts of the process which are accompanied by introspectible mental events. By analogy it might be argued that probably this part too is accompanied by mental events which fill the gap between those which we can introspect, but which, for some reason, are not introspectible by us. I do not see that this extra hypothesis would help us to explain anything that could not be explained without it; we must content ourselves with saying that it would be neither necessary nor impossible. The situation is, I think, quite different in the case of a purely mental theory of traces. Here we must postulate either a Pure Ego or a continuous series of total mental states; for, if we do not do this, we have nothing mental to carry the traces. And, since there are certainly gaps in our introspectible experiences, we must fill them with non-introspectible mental events if we want to keep to a purely mental theory of traces, and at the same time to avoid the hypothesis of a Pure Ego.

      Is there any conclusive objection to a purely physiological theory of traces and dispositions, and to the purely epiphenomenal theory of mind which seems to me to be its natural complement? At first sight there seem to be several objections, and the question is whether they are really conclusive.
(1) We have certain experiences in which it seems to us that our minds are acting on our bodies, and we have other experiences in which it seems to us that our bodies are acting on our minds. The voluntary initiation and control of bodily movements is an example of the first kind of experience, and the occurrence of a new sensation is an example of the second kind. Now, it might be said that this distinction between "active" and "passive" experiences could not exist, if epiphenomenalism were true; for in all cases our experiences would be merely idle accompaniments of certain physiological processes, and the latter would be the only real "agents". I do not thfink that this is the right way of putting the case. It is true that the interpretation which we put on this distinction would be mistaken, but it seems to me that the existence of the distinction could be explained perfectly well on the epiphenomenalist theory. Let us consider the observable differences between a volition which is followed by the desired bodily movement, and a sensation which arises when someone sticks a pin into me. The volition forms the end-point of a certain conscious mental process, viz, a process of deliberation, which has a characteristic kind of internal unity. It is no doubt succeeded by other mental events, but they do not form a continuation of the process of deliberation. The subsequent events which are specially closely connected with the volition are simply the sensations due to the bodily movement. Now contrast this with the new sensation. This is not a continuation of any conscious mental process which was going on before it happened, though it may form the starting point of a characteristic conscious mental process which succeeds it. The previous events with which it is most closely connected are events in my body which are unaccompanied by conscious mental events. We feel "passive" par exellence at those critical points where a physiological process which is not accompanied by consciousness passes into a physiological process which is accompanied by consciousness of a characteristic kind. We feel "active" par excellence at those critical points where a physiological process which has been accompanied by a series of mental events so related as to form a single conscious process passes into a physiological process which is either not accompanied by consciousness at all or is accompanied by mental events which are not continuations of the previous conscious mental process. Thus epiphenomenalism would seem to be quite capable of accounting for the existence of the distinction in question.

      (2) A rather similar difficulty could be raised over the distinction between mere passive association of ideas and active deliberate thinking, in which we select ideas and control the processes of association. It might be said that the former is compatible with epiphenomenalism and that the latter is not. I think that we can make a very similar answer to that which we made to the previous objection. If epiphenomenalism be true, the process of deliberate thinking is no doubt correlated with a peculiar physiological process, and the association of ideas is correlated with a different physiological process. And, doubtless, the two physiological processes are so connected that

  1. the latter can go in the absence of the former,
  2. the former depends for its possibility on the traces and dispositions which are involved in the latter, and
  3. when the former is going on it greatly and characteristically modifies the latter.
A process of the former kind may supervene at a certain stage in the course of a process of the latter kind, and at that stage we shall pass from mere day-dreaming to active and deliberate thinking. Thus the observable distinction seems to be quite capable of cxplanation by epiphenomenalism, though the interpretation which we commonly put upon it will not be strictly accurate if epiphenomenalism be true.

      (3) A third objection which might be made is the following.

"Does not epiphenomenalism amount to saying that conscious mcntal events and conscious mental processes are 'appearances of' certain physiological events and processes, just as colour and temperature are supposed to be 'appearances of' certain movements of molecules and electrons? And must not an appearance of snmething be an appearance to someone who is not an appearance? Does not the epiphenomenalist theory thus tacitly assume the existence of a mind in a sense in which it explicitly denies that minds exist? And is it not therefore radically inconsistent?"
This objection sounds plausible, but I do not think that there is anything in it. Epiphenomenalism may be taken to assert one of two thines.
  1. That certain events which have physiological characteristics have also mental characteristics, and that no events which lack physiological characterlstics have mental characteristics. That many events which have physiological characteristics are not known to have mental characteristics. And that an event which has mental characteristics never causes another event in virtue of its mental characteristics, but only m vlrtue of its physiological characteristics. Or
  2. that no event has both mental and physiological characteristics; but that the complete cause of any event which has mental characteristics is an event or set of events which has physiological characteristics. And that no event which has mental characteristics is a cause-factor in the causation of any other event whatever, whether mental or physiological.

      It seems plain that neither of these alternative statements of epiphenomenalism involves any tacit reference to the existence of a "mind" in some sense which is inconsistent with epiphenomenalism. "But," it might be said, "this is not the whole truth about the matter. Some events which have mental characteristics are states of knowing other things. And, again, some events which have mental characteristics do not merely exist but are themselves known by the mind which owns them. Does not this involve the existence of a 'mind' in some sense which epiphenomenalism cannot accept?" It seems to me that this is not an objection which applies specially and directly to epiphenomenalism. Epiphenomenalism asserts nothing positive about the qualities and relations of mental events, and it denies only one thing about them. It simply says that mental events either (a) do not function at all as cause-factors; or (b) that, if they do, they do so in virtue of their physiological characteristics and not in virtue of their mental characteristics. It has no need to deny that certain mental events stand in the cognitive relation to other things; for the relation of cognising is not, and does not involve, the relation of causation between the terms which it connects. And it has no need to deny that two mental events may be so related that one is cognised by the other. Of course epiphenomenalism does tacitly deny the Pure Ego theory, and it does explicitly deny that the unity of a mind is a direct causal unity. But it denies nothing else. In particular it has no need to deny that certain contemporary mental events are bound together by unique and very intimate relations, so that they together compose a single total mental state. And it has no need to deny that certain successive total mental states are bound together by unique and very intimate relations, so that they together form a single mind. The objection under discussion thereforc applies, not directly and specially to epiphenomenalism, but to the view that knowledge (and, in particular, self-knowledge) can be explained without a Pure Ego or without direct causal relations between mental events. And this is a question which we are not at present in a position to discuss.

      (4) The last objection which I propose to consider is the following. It might be said that the hypothesis that there are literally unconscious desires, emotions etc., and that they literally interact with each other in the way in which conscious desires and emotions appear to interact, is found to be practically useful by psycho-analysts and others. This kind of hypothesis does enable them to suggest methods of treatment, and often to effect cures. The hypothesis of purely physiological traces, dispositions, and processes is not found to be practically effective. Therefore, probably the former hypothesis is approximately correct, and the latter is probably wrong.

      I think that there is little or nothing in this argument. Two conscious desires appear to interact in certain characteristic ways. If the epiphenomenalist be right, they do not really interact. But each is correlated with a characteristic physiological process, and these physiological processes really do interact with each other, thus producing characteristic modifications in the series of observable mental events. Now the patient observes a certain series of mental events which is modified in this characteristic way, and he fails to find any other series of conscious mental events which seems to account for this modification in the former. Let us suppose that his psycho-analyst is an epiphenomenalist. Then he would simply postulate the existence of a physiological process which (a) is not accompanied by any mental events which the patient can introspect, but (b) is otherwise of the same general nature as the physiological processes which are accompanied by a conscious mental process of conation. If he chooses to assume that this is accompanied by a conative mental process which the patient cannot introspect, he is simply using an argument from analogy which may be good or bad but is quite irrelevant to any predictions that he may make or to any course of treatment that he may devise. It is no doubt also true that he is not helped in any way by assuming that this process is physiological; but he is also not hindered by this assumption. The point is that, even if the process be physiological, he does not know anything about its physiological details; he knows it only as "the sort of process which would generally be accompanied by a conscious conation or emotion of a certain kind". So long as it acts as such a process might be expected to act it makes no difference to his predictions whether its intrinsic nature be physiological or mental or both or neither. Thus the success of psycho-analytic treatment which assumes literally unconscious conations and emotions, and the ill-success of methods of treatment which assume a certain hypothetical structure and processes in the brain and nervous system, seems to me to have no bearing one way or the other on the truth of epiphenomenalism.

The Choice between the Two Theories.

      I have argued that it is perfectly possible to hold a purely mental theory of traces and dispositions, with or without the Pure Ego theory of the self. And I have argued that it is perfectly possible to hold a purely physiological theory of traces and dispositions. A purely mental theory of mnemic persistents, which does not accept the Pure Ego, requires the assumption of literally unconscious mental processes; but there seems to be no conclusive objection to this. A purely physiological theory of mnemic persistents permits the assumption of literally unconscious mental processes, but renders this assumption superfluous. There is nothing to choose between the two types of theory so far as concerns our ability to predict and control mental events. Whether traces and dispositions be purely mental, or purely physiological, or both, or neither, we know absolutely nothing about them in detail, and can predict nothing from one hypothesis about their intrinsic nature which we could not predict equally well from any other hypothesis. The fact is that we know what they are only from what they do, and our knowledge of what they do is equally compatible with either of the four possible theories about what they are. Is there then the slightest possibility of deciding even tentatively in favour of one rather than of another?

      If there be no phenomena which give us any reason to believe that minds can exist and operate after the destruction of the bodies which they have animated, I think that there will be some reason to prefer the epiphenomenalist theory on the ground that it involves fewer assumptions than the others. We know that our brains and nervous systems exist throughout our lives, and that they are very closely connected with our minds. We know that our conscious mental life is subject to great interruptions, and we do not know that these gaps are filled by literally mental processes which we cannot introspect or remember. Since the brain and nervous system are capable of carrying the necessary traces and dispositions, and since processes in the brain and nervous system are capable of filling the temporal gaps between our introspectible mental events, it would seem superfluous to postulate mental traces and dispositions or literally unconscious mental events and processes in addition. Thus, unless there be reason to believe that minds can survive the death of their bodies, I should consider that some form of epiphenomenalism was the most reasonable view to take of the nature of mind and its relation to the body. I have said, and I repeat, that all the arguments against interaction are invalid. I have said, and I repeat, that a purely mental theory of traces and dispositions is perfectly possible with or without the assumption of a Pure Ego. All that I assert here is that epiphenomenalism is also a possible theory; and that, if there be no reason to believe that the mind ever exists apart from the body, this theory is to be preferred as involving the minimum of assumptions.

      It is plain that we can get no further till we have considered the alleged evidence for the doctrine that minds can and do exist apart from bodies. If this evidence produces even a faint probability, it will be rash to accept epiphenomenalism. For epiphenomenalism would seem to be quite inconsistent with the very possibility of the independent existence of a mind. The very essence of this doctrine is that the mind by itself is not a genuine substance capable of independent existence, but either consists of events which are also bodily or is absolutely dependent for its existence on such events. Now we have no strong positive ground for accepting epiphenomenalism; the alternative theories are equally possible, and much more in accordance with common-sense. We have given a tentative preference to epiphenomenalism only on grounds of "economy"; a theory supported only in this way could be overthrown by a very light blow. In the next section of this book I propose to consider the alleged evidence for the existence of minds apart from bodies.