Chapter VIII: Part 8

WITHOUT the observation of direct experience we should be in great danger of constructing an artificially simplified system of psychology, as is that of current behaviorism. On the other hand, it seems absolutely impossible to me to develop psychology as a science of direct experience or of “consciousness.” For the development of that science the field of actual experience alone is inadequate, especially when compared with the totality of those processes in the nervous system, a few of which are at each moment accompanied by experience, but _all_ of which seem to be more or less interdependent. How can one pretend to construct an adequate theory of psychological events using experience alone, if the processes underlying experience are merely a dynamical province of a much larger functional whole? It is almost impossible to deny this fact nowadays. Therefore, we cannot hope to understand experience itself exclusively from its own aspect, any more than one could hope to learn and understand the game of chess through watching only the moves in one corner of the board the whole time.

In this latter case the observer of the game would soon become aware of the fact that something important was going on beyond the narrow field of his observation, that evidently the moves in it referred to something beyond, since they even seemed to come from there and to disappear again into the unseen. Exactly the same is true of experience. When reading or talking about things which are not present, we do not usually develop adequate “images” of them. Sometimes scarcely anything at all seems to be experienced of those absent things. When asked about my profession I answer that I am a psychologist. But all the _actual_ experience connected with this word is a certain feeling of familiarity, of being sure about it, and of a certain direction in which I might start at once in order to furnish more concrete and detailed data. This “readiness for transition” in a definite direction, the mark of which is not given explicitly in actual experience, has been described excellently by William James. It is one of the most frequent and common things to be found in experience. And its most conspicuous character consists in its being _experienced as_ pointing beyond the actual field of experience towards something particular and ready outside. So experience tells us about its own functional incompleteness. Nor should we be too much astonished by the fact; nothing else can be expected if, of a larger functional whole, only a restricted field has experience as a correlate; to the dynamical dependence of this field upon definite physiological states beyond itself corresponds the felt definite direction in experience. Though not experienced actually, those physiological states “beyond” seem to be definite enough; for in general our reading and talking go the right way, sufficiently determined by those “states beyond.”

Perhaps the simplest example of this is “successive comparison,” in its different forms. After a few years of traveling I meet a friend, and my first idea is “How old he looks!” That does not at all mean that he looks particularly old on an absolute scale. I see older-looking men every day. Nor does it mean that I reproduce an image of him, as I knew him before, and now compare the two. Nevertheless, what is meant refers to the past and represents an extreme form of what occurs in most cases of successive comparison. If five seconds after a first noise I hear a second one, similar but sufficiently louder, I am able to judge about the relation of their intensities without any difficulty, though I do _not_ reproduce the first at the moment of the second. (Personally, I am not able to experience even a somewhat inadequate “image” of one noise when I hear a second one which is slightly different.) In this point all observers at present seem to agree.

But how do we judge about the relation, if only one of the noises is actually experienced? We may answer that we do not experience the second noise as something isolated and discrete, but that, when it occurs, it seems to come with a definite direction “from something in the past.”[49] Even if we take this for granted, however, our problem is not yet solved. Our judgments are usually rather accurate. Hence, what is left of the past, i.e., of the first sound, which now determines the direction with which the second noise appears, must be a fairly definite representative of the first noise. On the other hand, this _trace_ of the first sound is an effect not comparable to that process which, five seconds before, was accompanied by our experience of the first sound. If it were still of the same kind, it should be accompanied by a similar experience throughout, which it certainly is not. Hence, the only possibility seems to be that a physiological product of that first process remains after that process itself has ceased, that such a product represents sufficiently well that process by which it was produced, and that the definite direction with which the second noise appears depends upon the “level” of that product in its relation to the intensity of the second noise.

The concentration of those molecules which have taken part in the first process may be considered as a product of that process, left behind as its trace, with corresponding properties. There are other hypotheses. But no theory will be competent which does not assume some trace corresponding more or less to the actual properties of the first process. Some years ago I developed a more detailed theory of successive comparison, primarily concerned with the way in which the trace of a first sound determines the direction with which the second sound appears.[50] I concluded that by experimentation on successive comparison we are able to find out what happens to the trace of the first process after this process itself has ceased. The final result of these considerations was that such a trace is preserved for a long time and that, probably, it is identical with what has to be assumed as the physiological basis of enduring memory and of reproduction.

In some degree all sound theories of memory, of habit, and so forth, will be hypotheses about physiological traces. Furthermore, each of these theories will have to assume that the properties of the traces correspond more or less to those of the processes by which they were produced. Otherwise we should not be able to explain the accuracy of reproduction which in a great many cases is very high. In _gestalt_ psychology we state at once the assumption that, as far as the original process (and the experience which may have accompanied it) has been organized in a definite manner, its trace must have similar properties of organization, in general. Take the examples given in Chapter VI where the concept of “real form” was considered. If certain segregated wholes with definite forms are experienced in our field of vision, the aspect of one of them later on may reproduce other experiences which had been present at the same time. But if, for some reason, instead of that particular organization, grouping and form, we experience other wholes with their real forms, no reproduction will occur, though the same, or almost the same, stimuli may be given us. So the number 4 will reproduce its name when written in a certain environment (cf. p. 209), but when Fig. 10 is shown to an unprejudiced subject, this name will not occur at all. On the other hand, if the subject has once found the 4 in that figure, which means that 4 has now become a segregated whole, he will consequently find it again easily, and reproduce the name. From this it follows that the traces of past experiences are neither an indifferent continuum nor a mosaic of independent points; rather they must be pictures of past organization. As such they take part in processes of reproduction.

This same property of the traces may be deduced from recognition. When Rubin induced his subjects at the first presentation to see certain pictures in a definite distribution of figure and ground, they recognized them very well if, afterwards, the same organization was favored by the conditions of the experiment. If, however, in the second presentation what had been figure before became ground, and vice versa, the subjects found unknown forms in the field, i.e., failed to recognize, though the constellation of stimuli was the same throughout. Here again the traces are shown to correspond entirely to past organization, and not to the aggregate of past stimuli. We may go a little further: In most cases of reproduction the reproduced material itself appears as organized. This is true not only of “images,” but also of motor “melodies” with which we are well acquainted, as Michotte and van der Veldt have shown most vividly in a recent book.[51] Readers who have visual images will easily find that their image of a definite individual tree is detached from the dim environment, as a figure-whole possessing a real form. Of course, in “free imagination” and in dreaming the field given us may be very different from any sensory experience we have ever had before. Nevertheless, even the oddest creations of dreaming will remain figures, groups or events, which, whether very definite in shape or not, exhibit all the essential properties of organization.

So decisive will organization be in a tremendous number of cases that we may change the stimuli in the most radical manner and still get the same recognition or the same reproduction, if only the essential traits of the organization in question remain the same. A melody is recognized in a key which may not contain a single tone of its first presentation. Or again, some days after we have heard it the first time we find ourselves humming it spontaneously in a changed key. Here all the factors excepting organization seem to be indifferent for reproduction. I hardly need mention the fact that an unknown figure seen to-day in red color somewhat to the left of the point of fixation and in a definite size, will be recognized without difficulty if to-morrow it appears as green or yellow, somewhat to the right and in quite another size.[52] Evidently, then, recognition and reproduction deal with the organization of past events as much or rather more than with local effects of stimulation, which are assumed as the parts of past events in mosaic theory. We shall come back to this point later on.

From this viewpoint some facts can be explained which would remain perplexing if we did not recognize the reality of organization. When we experiment on delayed reaction in animals we find that some of them are able to choose the right object, let us say among three equal ones before them, after a delay of many seconds or even some minutes; though at the time when the delay is over and we release the animal, the right object does not exhibit any peculiar traits, as it did before the delay. If, during the delay, the animal remains directed towards the right object, his correct choice is not very astonishing. A real problem arises in the case in which the animal moves freely in his cage during the delay and still assumes the right direction after we release it. It has been said that in such a case the animal’s reaction depends upon some internal cue. This is true insofar as there would be no reason at all for the right reaction after the delay without some after-effect of previous stimulation (for instance, the aspect of food) which at that time made one object a particular thing. That after-effect is undoubtedly an internal cue. But when, after the delay and after accidental movements of the animal, the internal cue works the right way, there must be some characteristic of the right object by which it can be discriminated. This something cannot be discovered when we examine each object by itself, because in exact experimentation they all have the same properties. Still there is some difference between them, and this difference is the rôle which each object plays in the group. One is the left end, one the right end, and one the middle or interior part of the group. If, with that after-effect as an internal cue, the animal reacts after the delay, the only property (of the right object) which can make the animal refer its cue to the right object is _the place of that object in the group of three objects_.[53] The particular stimulus (the food, for instance, or a light, as in Hunter’s famous experiments) which was given at the beginning of the delay has operated as a distinguishing mark of one definite and well characterized member of the group (let us say, of the first to the right in a single case). Therefore, after the delay, the group, and that definite rôle which just one member of it plays in sensory process, is able to work upon the cue and to produce the right reaction. In this sense delayed reaction to one among several objects, having identical properties as isolated objects, involves the principles of organization and cannot be understood without them. This becomes even more obvious, perhaps, when we consider the multiple choice method which Yerkes has used with so much success. There it is evidently the definite rôle of one object in a group which has become connected with a definite reaction and consequently will reproduce that reaction whatever may be the position of the group.

The main interest of experimental psychology has not been directed toward the nature of the traces which are left in the nervous system after the original processes have faded away. When we say that the traces of organized processes must be organized themselves, we do not seem to have mentioned the most important feature of memory, habit and so forth, namely, that those traces are _connected_ in the manner which is generally called their “association” and which leads to the reproduction of the second when the first is activated by a repetition of its process. That there are traces is one basic factor of habit, etc. But if two processes, A and B, have occurred frequently together, reactivation of the trace of A is said to reactivate B in consequence of a curious bond between them created by those past events. This is the second basic factor of habit, and almost all the classical research on memory has dealt with the formation of that bond and its career in the course of time. There is almost no other phase of our science of which we are prouder than this, because here methods have been developed and laws have been found which may almost be compared with those of the natural sciences.

In some respects this pride seems fully justified to me, though at the present time we have begun to realize that we have used those excellent methods to investigate inconsiderately a special type of memory and habit and that we should not apply the results obtained to everyday memory and habit. There is one problem, however, which was not made an object of those experiments because it did not seem at first to be a problem. Indeed, it was presupposed as a fact that the mere repetition of two processes, A and B, having occurred frequently together, would create that curious bond, the association, between them. This concept of association shall be discussed in the following paragraphs.

The law of association just mentioned has been considered particularly satisfactory, due to the purely mechanical character of its content, which does not in the least imply mystical forces or anything else foreign to the ideals of natural science. I must confess, however, that just from the viewpoint of natural science the law of contiguity appears rather strange to me. A and B happen to occur together and, whatever the nature of A and B may be, the bond is produced between them! I do not know a single law or rule in physics or chemistry which might be compared in this respect with the concept of association. I repeat what was said in Chapter IV: Wherever an A and B have anything to do with each other in physics, the effect is found to depend upon the properties of A and B in their relations to each other. This is the case in astronomy where acceleration depends upon the masses of A and B. It is the case in electrostatics where the occurrence of attraction or repulsion depends upon the nature of electric charges. In chemistry we find the violent reaction or indifference of atoms determined by the relation of their properties in the given case. But there is not a single example of an effect produced by the interaction of two things or processes quite independently of their properties. Nevertheless this is the character of the classical law of association, as we find it stated in most textbooks.

At this point we return to that trait of mosaic theory which we have found so characteristic of its treatment of sensory processes. If the distribution and direction of events in the central nervous system are determined dynamically by interaction in the field, they must depend upon “the relative properties” of those processes which exert influences upon each other. This is excluded in principle by the mosaic theory of the sensory field. Sensation processes are assumed to be indifferent to each other in general, so that, with regard to them, any pattern may exist equally well in the field. Exactly the same assumption of discrete bits of process, altogether indifferent to each other, underlies the classical concept of association for, without even mentioning the actual properties of A and B in their relations to each other, it states that _any_ A and _any_ B may acquire that neutral bond, if it is supposed only that they occur together a number of times.

It seems to me that, even at the present time and before any experimental research has been specially directed upon this question, we may say confidently that, in this point at least, the concept of association is opposed to a large number of well-known facts. In order to corroborate this claim, we have only to gather statements scattered in papers, the authors of which have worked with the classical methods and the classical nonsense syllables, which their subjects had to commit to their memories.

We see at once that in a series of these syllables the A and B (i.e., two syllables following each other) cannot remain altogether indifferent to each other, since by a simple experiment we can prove that A and B are not even indifferent to F and G and H and so forth (i.e., to members remote in the series). When a subject is told to write down six syllables which we read to him quickly, he is usually able to comply. But if, instead of six, we give him a series of twelve syllables, the result is that on the average less than six are written down. This argues a mutual disturbance which indicates that even syllables rather distant in time are not indifferent to each other. If after some repetitions we test the subject, giving him single syllables and instructing him to name the following one in each case, our procedure is based upon the assumption that there are a number of independent associations and reproductions, the results admitting a statistical treatment. This does not seem to be altogether correct insofar as it neglects a more general interdependence. It may be neglected whenever this interdependence is the same statistically in all the series and when the problems under investigation are the usual ones. However, as soon as the nature of association itself becomes a problem, we must be more careful.

In a more striking manner the same warning may be deduced from what happens to a series of syllables during learning. The subjects read them in a definite and rhythmical distribution which, usually, consists of larger wholes, containing sub-units. At the same time the reading will assume the character of a melody, the pitch of the voice going up and down, following that grouping exactly.[54] Evidently, a process of organization occurs during learning and most of all during the first readings of the series. If that be so, the syllables ought to acquire a definite “flavor” which belongs to them insofar as they have a particular place in the larger whole. This consequence is verified perfectly in those cases in which, after learning a series as a whole, the subjects are shown the same syllables in another sequence. In the new order _they look or sound new or strange_. A similar implication is to be found in the highly important concept of “association with place,” which means that a syllable is learnt as having a definite place, not absolutely but with respect to the whole series in which it occurs. Objectively, the rôle of organization is most convincingly proved if, after learning the whole series completely so that it is repeated as a whole without hesitation, the subjects are shown single syllables in order to reproduce the following one in each case. It has been found by Nagel that scarcely one-third of the reproductions was achieved under these circumstances.[55] When given alone, a syllable is obviously not the same thing as it is in the stream of the organized series.

Nonsense syllables were chosen by Ebbinghaus and his successors as the best material for investigating associations, because they wished to keep their experiments free from the older, already-formed associations which, if “meaningful” material were employed, would influence the results in an uncontrolled manner. Furthermore, nonsense syllables seemed to be a more uniform material than any other. It would be altogether unjust if we should deny the value of the great impulse which psychology received from that method. It seems, however, that the early investigators used this method in a somewhat one-sided manner. The most valuable observations, perhaps, were made after the one-sidedness of the original point of view was discovered.

Some psychologists have criticized the method of Ebbinghaus in that it does not at all investigate automatic associations, whereas the results of the method are formulated as though that were the case. Indeed, if the mere neighborhood of the syllables is supposed to produce their association, most experiments using this material are far from testing such a pure association, since the subject does not simply receive a succession of syllables, but rather _is asked to learn them_. If he follows this instruction, some factor takes part in what is built up during learning, and this factor is not even mentioned when the results are formulated in terms of automatic associations. Undoubtedly, here is a flaw in the scientific procedure. The mistake is a grave one for it has been shown that without that factor the learning of series of nonsense syllables would be almost impossible.[56]

But what are the subjects doing when they try to learn a series intentionally? No one is better authorized to give an answer to this question than G. E. Müller, who has spent a large part of his scientific life studying the rules of association and reproduction. His answer is: “A series of figures, consonants, syllables and so forth, is learned essentially by our combining the members of the series in an attitude of synthesis, so that they become solid groups.”[57] We have seen in an earlier chapter that by a corresponding attitude one may favor definite forms of grouping, and that the products of this procedure may be just as real as any spontaneous organization is. Following the statement of Müller, we may say, therefore, that intentional learning essentially means intentional organizing.

Though such an attitude seems to be almost necessary in the case of nonsense material and, most of all, of syllables, it is not needed evidently where the material has other properties. Time and again we find ourselves reproducing events from our past experience, where the incentive to reproduction has certainly not been _intentionally_ combined with what we are now able to reproduce. If that be so, we must draw the conclusion that the nonsense material used in classical investigations may be excellent in some respects from the viewpoint of exactness, but that it cannot teach us the whole truth about so-called associations. These are not formed in the same manner when, instead of the classical material, we have to deal with the more natural experiences of everyday life.

If now we ask ourselves, whether _all_ our experiences outside the laboratory associate themselves spontaneously, so that the evocation of one may reproduce its neighbors, we must confess that this is not the case. We may hear a telephone number dozens of times together with a name and still remain unable to reproduce it when the name occurs later on. In this case conditions seem to be similar to those present in the case of nonsense syllables. The name and the number are as indifferent to each other as are those syllables usually. Thus the suspicion arises that we may have spontaneous association where organization is spontaneous, and that intentional combining is needed where the material in itself is but slightly organized.

This assumption is corroborated by the well-known fact that meaningful nouns form associations much more easily than nonsense material. In this case, of course, the nouns appear as imbued with their meanings by a process of learning accomplished in our childhood. When as adult subjects we learn a series of nouns we find these meanings ready-made, and obviously now it is these _meanings_ which are so easily associated. But why should they be? At the present time most psychologists will answer that this is a consequence of their having been previously associated, so that the actual process of learning has only to strengthen some particular bond between the meanings of a pair of nouns which existed a long time before. But at this point the difference between _gestalt_ psychology and associationism becomes striking. Let somebody read the following pairs of nouns a few times, with an attentive attitude: lake-sugar, boot-plate, girl-kangaroo, pencil-gasoline, palace-bicycle, railroad-elephant, book-toothpaste. Learning will be considerably easier here than if the same number of nonsense syllables were given. Can one really say that between lake and sugar, palace and bicycle, and so forth, there are old associations which just need some slight exercise and which thereupon facilitate the learning? It seems to me that we cannot, mainly because the same words have occurred thousands of times in other much more regular connections, which ought absolutely to inhibit the _assumed_ associations in this particular case. But there is another explanation. When I read those words I may imagine as a (very strange) picture, how a lump of sugar might dissolve in a lake, how a boot might rest on a plate, how a girl might feed a kangaroo, and so forth. If this happens during reading I shall have experienced some well organized though quite unusual wholes, and it may be that here learning is so easy, because _organizing_ is so much easier in this case than in that of indifferent nonsense syllables. In order to exclude the possibility of frequent similar combinations in the past, I have chosen strange pairs of nouns, the meanings of which may be organized into larger wholes (pictures), but do not do so quite spontaneously. It seems to me that those combinations and sequences which become even more easily associated in everyday life are simply cases of strong spontaneous organization.

We may say, then, that in one respect nonsense syllables constitute the worst material which might be chosen for discovering essentially what association is. As they do not spontaneously form well-characterized organizations, the nature of spontaneous association will not become apparent to the psychologist who uses them. Furthermore, as the series of syllables are built up in a haphazard manner, one cannot even examine the way in which learning depends upon what may be called the structure of a series. Even though it be composed of nonsense material throughout, a series may be constructed in a great many different ways. Syllables may be chosen which fit together phonetically in a very characteristic way, or the contrary may be done. Some pairs may be constructed according to one principle, some according to another. The whole series may exhibit an extraordinarily striking principle of structure or it may be an indifferent series, as are the usual ones. All these variations of material ought to be examined in order to see whether or not organization is the essential fact underlying so-called association. From the trend of the foregoing discussion, we should be inclined to say that it is.

As a last argument in favor of this thesis we may mention the fact that, sometime after learning a series of members by combining them into pairs, for instance, the subjects will be able to reproduce the second members of these pairs rather easily, if the first members are given them as stimuli to reproduction, whereas, on the other hand, the number of right reproductions will be small if we present the second members of pairs and ask for the following serial items which are the first members of the next successive pairs or groups. If we suppose that during learning the members of the whole series objectively formed a uniform sequence as to spatial and temporal intervals, this result makes it evident that the old rule of association is superficial and that the conditions of association are ambiguous as long as we do not take into account the organization of the material. We get strong association only in those parts of a series where organization is also strong. It is true that neighborhood or proximity in space and time is a factor of great importance in association. But it does not seem to operate directly. In an earlier chapter we saw that the factor of neighborhood is one of the conditions upon which depend grouping and the segregation of wholes. From what we have just stated it would follow, then, _that neighborhood in space and time influences association only insofar as it determines organization_. Since this condition is just one among a great many others determining organization, and since organization seems to be the really decisive condition of association, the rule of association has to be reformulated accordingly.

I repeat. Where organization is naturally strong we have spontaneous association; where there is practically no organization association does not occur until some organization is created intentionally. Also, after being associated, the members of a series have definite properties which depend upon their position in the surrounding whole, as tones do in an organized melody. And finally, if in a series we determine those sequences which are the _solid_ groups in the total series, and if, then, we also determine where association is strongest we shall find that both coincide.

After these preliminaries we can discuss the nature of that particular bond which is said to originate between the traces of two processes when these processes are associated. The prevailing opinion is that during learning the nerve fibers uniting the place of one process with that of the other acquire an increased conductivity, because with each repetition some particular current of nervous energy is assumed to pass from one to the other. In consequence of this change of conductivity, whenever the process corresponding to the first trace activates this trace the current spreads along those fibers to the second trace, and reactivates it also. With this hypothesis one may understand, perhaps, why after some repetitions the current should flow in that particular direction and thereby increase its conductivity; but we do not at all see why it should flow in just that direction on the _first_ occasion. And this is a serious difficulty, particularly in those cases where after one single experience we find the association well established. One can scarcely avoid the inference that in this case the processes underlying the experience _as such_ contain all that is implied in the concept of association, at once and originally, no special bond being necessary for it.

We do not know what happens in reproduction. The only thing we seem compelled to assume is some physiological communication between the traces of two processes, A and B, so that an event activated in A will spread to B and not to other traces, with the processes of which it had nothing whatever to do in the past. Now, two hypotheses are possible: If we believe A and B to be two processes indifferent to each other, which accidentally occur near to each other in space and time, then some special bond, some change of conductivity in fibers between them may be needed as a basis of the association. In full contrast to this older view, however, we may make quite a different assumption: Whenever we “associate” A and B we experience them, not as two indifferent things, but as members of one organized group. This certainly may be taken for granted, after the foregoing analysis. In this case the process underlying the experience cannot consist of two indifferent parts, one belonging to A and one to B. It must be one functional whole in which A and B possess only a _relative_ independence. If that be so, there will not be two separate traces left behind after the process itself has faded, but _one_ trace of the functional whole, with the after-effects of A and B in it as local regions of that single trace. In this way A and B, or their after-effects, would be as well connected in the single trace of the whole organized process as they might ever be by means of the special bond which united them, according to the older point of view. If such a bond determined “the spreading of nervous activity” from A to B, the segregated trace of one process-in-extension might well do precisely the same. Unfortunately, in neither case do we know as yet what are the functional properties of that “spreading” and what sort of nervous communication would be capable of determining its course in the observed manner of reproduction.

It will be convenient to give our assumption a somewhat radical formulation so that it will be easier to distinguish it from the usual one. From our viewpoint, association is given up as a special and independent theoretical concept. It is not more than a name for the fact that organized processes leave a trace picturing their organization and that in consequence of it reproductions are possible. I do not deny that repetition will make the association stronger, but as long as the process remains the same, repetition does not involve a change in a special bond; it means that the whole trace of the organized process becomes more enduring and stable. Nor shall I deny that sometimes, as in the case of nonsense material, some special attitude is needed for producing an association. But, as we have seen before, such an attitude consists in favoring definite organizations. When these are experienced, the processes corresponding to them will also be organized, and so will their traces. The only problem which is new in such a case is that of “intentional” organization. But this problem shall be treated quite apart from association.

* * * * *

Some psychologists will be inclined to say that it does not matter very much whether we accept one theory of association or the other, since we are not able to look into the brain to decide which one is right. This indicates a complete misunderstanding, however, of the value of a hypothesis. If there is anything concrete in a hypothesis, it must be possible to deduce concrete consequences from it, and to examine them in further experimentation. It is easy enough to do that in this case.

In the old rule about association nothing was said about the properties of the A and B which are associated because, in principle, association was believed to be an indifferent bond between processes indifferent to each other and to the bond. Organization, however, is not at all an aggregation of indifferent material. On the contrary, in sensory experience we have found organization to depend most decidedly upon “the relative properties” of stimulation. Therefore, if association is a consequence of organization, it must also depend upon the mutually relative properties of what is or shall be associated. In some degree this is verified by what we have reported above. But much more has to be done in the way of varying the nature of the material according to the principles of _gestalt_ theory, so that we may see more definitely whether, other circumstances remaining the same, an association will be the stronger, the better organized is the group or the whole, according to those principles. As yet we do not seem to have any experiments planned specially and deliberately in order to answer this question. As we know a great deal about the conditions upon which sensory organization depends, the path is laid and it will be the more interesting to follow in this direction since we should be able to deduce from all the rules known in the field of sensory organization, corresponding rules for association. Of course, we cannot hope to find all the properties of association in this manner, because the rules of sensory experience are in no way instructive with respect to the formation of _traces_ as such; nor can we from those rules deduce the career of a trace in the course of time, and under the influence of other processes and traces which develop after it in the same region of the nervous system. On the other hand, when once the identification of association with the traces of functional wholes is verified sufficiently, further discoveries with regard to association may give us hints about properties of organization itself which, for some reason, we may fail to detect directly in sensory experience.

A second consequence of our hypothesis, which ought to be examined experimentally for the sake of that hypothesis, may have practical value as well. This consequence belongs to animal psychology. We have seen that nonsense syllables do not usually associate spontaneously because they do not form well-organized groups unless we combine them. In some cases series of them have been read passively hundreds of times without an appreciable effect in learning. But, as we have seen, the same subjects “have a good memory” in everyday life and can reproduce thousands of events which they have never intended to commit to their memories. This reminds one of a strange contrast which all animal psychologists have observed between animal learning, as it occurs during experimentation in the laboratory, and habit formation as it develops in the same animals outside the laboratory. I do not think that the concrete reason for this difference is contained in the reference to the “natural” circumstances present in the latter situation but not in the former. What does the term “natural” mean here? I suggest that it may mean “favorable” with respect to association in a very particular way, i.e., with respect to organization.

Indeed, it would seem in this connection that a great many experiments, most of all those on sensory discrimination, proceed in the wrong way, because, blinded by that old concept of association as a binding together of two indifferent things largely by many repetitions, we do not take account of organization at all when we are constructing the experimental situation. For instance, on the rear walls of two alleys we fix two objects which we wish the animal to learn to discriminate. On the floor, in no connection whatever with those objects, we put the wires for punishment in the case of a wrong choice, in such a way that an electric shock applied at this point will be more foreign to the wrong object even, perhaps, than one syllable is to the next in experimentation on human association. On the other hand, the animal will be fed after a right choice somewhere behind the scene, i.e., in a situation just as separate from the right object as the shock is from the wrong one. A young behaviorist once asked me whether, apart from vague concepts, _gestalt_ theory had anything new and concrete to offer science upon which it might work. It seems to me that if we had nothing to give behaviorism except our criticism of this method and some suggestions toward a better one, it would be quite sufficient. In man learning seems to depend upon organization. It is highly improbable that the same rule should not apply to habit formation in animals. Therefore, when we investigate sensory discrimination in animals, instead of separating those “stimuli” from each other, the coöperation of which most probably determines learning, we ought to make that coöperation as easy as we can. Some years ago I proposed the following procedure: If the wrong object made a sudden movement against the animal, whenever the latter made the wrong choice, we should certainly have a situation much more similar to the animal’s learning in common life, and a more efficient one, because then the negative stimulus would immediately imbue itself with “negativity.”[58] It is quite possible that a variation of procedure in this direction would save the experimenter much time. In addition to this practical reason, however, it seems to me a sound law of experimentation that conditions should be varied in _all_ respects. If behaviorists are interested in finding the conditions and the nature of learning, they may have great success some day if they vary the conditions of habit formation according to the principles of organization.

These remarks apply to the formation of so-called conditioned reflexes in the same way that they do to older methods used to study learning. From the viewpoint of machine theory it sounds a little nicer if we talk about conditioned reflexes instead of associations. I do not find, however, that the first concept is clearer than the second. One might even say that what we call a conditioned reflex now is just one special case of a certain type of association, because it is evident that the “stimulus,” which is to be artificially connected with a reflex, cannot be made to produce that reflex except by first becoming “connected” with an _adequate_ “stimulus” which naturally produces that reflex. This means an association of two sensory processes however. The association may become so strong that _via_ the mere trace of the adequate sensory process the new one will eventually arouse the reflex. But since that association is the thing to be “learned,” before a conditioned reflex can occur, and since the association of two processes is a direct effect of their organization into one whole, the same consequences have to be deduced here as in human association and in the old form of animal discrimination experiments. As yet no one seems to have examined the question whether a change in the presentation of the “artificial stimulus,” in its relation to the natural one, exerts an influence upon the building up of the conditioned reflex. A bell rings and food is shown or given, but no attention is paid to those experimental conditions upon which organization would depend. That is not at all astonishing since the old concept of “connecting two separate processes by learning” still prevails. On the other hand, animal psychology has an opportunity here to test the value of two assumptions at the same time, namely, that conditioning is practically the same thing as associating two sensory processes, and that associating depends upon organization.

When seen from the viewpoint we have reached in this chapter, certain earlier discussions will assume a somewhat surprising aspect. Our conclusion is that association depends upon organization because association is just an after-effect of an organized process. Now, when we first introduced the concept of organization, we were hampered at every step by those empiristic hypotheses which sought to explain by the influence of meaning or other affiliated factors whatever experiences did not fit into the scheme of mosaic or machine theory. In many of those cases I hope that I have shown that, fundamentally, the experiences in question cannot be derived from previous learning in any form; and that, therefore, organization has to be accepted as an original aspect of our experiences. Now we may go further and make the statement that, on the contrary, wherever the influences of past experience can really be discovered in actual experience, this secondary effect itself is a consequence of past organization, because, generally, meaning will accompany an actual experience or be contained in it by a process of reproduction, and this reproduction will be based upon the trace of a larger organized whole. Consequently, if we try to reduce organization in principle to associated meanings, we must explain actual organization by the traces of previous organization in a larger area. If our theory is right, the whole procedure would mean a vicious circle, for organization would be the more fundamental functional concept, and association or meaning would be ideas depending upon it. So much concerning the question of principle. I repeat, however, that very often our actual experiences _are_ really and thoroughly imbued with meaning. In these cases it should be acknowledged as a fact that the after-effects of previous organization determine the properties of actual experience. We may talk about an association under these circumstances if we keep in mind that even here it is the concept of organization which remains decisive functionally.

BIBLIOGRAPHY

M. Bentley: _The Field of Psychology_. 1924.

K. Koffka: _The Growth of the Mind_. 1924.

Michotte and van der Veldt: _L’Apprentissage du mouvement et
l’automatisme_. 1928.

R. M. Ogden: _Psychology and Education_. 1926.

O. Selz: _Die Gesetze des geordneten Denkverlaufs_. 1913.

IX

_Reproduction_

PSYCHOLOGY knows three main topics which are to be investigated in the field of memory and habit: (1) learning and those processes the traces of which make reproduction and recognition possible, (2) the career of these traces in the time between their formation and recall, and (3) the processes of reproduction and recognition themselves. In some degree reproduction or recognition will be needed in the investigation of _all_ these problems, to give us evidence of either the laws of learning, or those of retention, or those of recall itself. Nevertheless, in the first case we can keep conditions constant with respect to retention, and also with respect to the circumstances of reproduction, while the conditions of learning are varied; in the second case, the conditions of learning and those of reproduction will remain constant, those of the interval between learning and recall being varied; and in the last case, we vary only the circumstances of reproduction or recognition. So we are able practically to separate the three classes of problems. In this chapter our attention will be directed mainly to questions of retention and recall, though we shall also have to treat one highly important feature which may as well be regarded as belonging to the topic of learning and the formation of traces.

In the sixth chapter I mentioned certain experiments in which, after learning to choose one side of a pair, for instance the darker of two grays, animals had to react to a new pair, consisting of the “right” object of the learning period and a new object, which had more or less the same relation to the “right” object as this had had to the “wrong” object during learning. The result was that in the majority of the trials the animals chose that side which, in the new pair, played the same rôle which the right object had played during learning. This means that they chose the _new_ object which, in our example, was now “the dark side of the pair.”

This result is not quite general, however. It depends upon what interval of time elapses between the training with the old pair and the first trials with the new one. Occasionally, after the learning period, a chick was given the new pair in between choices on the old one; and this procedure was repeated until the single trials with the new pair, each of them distributed between trials with the old one, could be regarded as numerous enough. It was found that, under these circumstances, the animal chose the “right” object of the learning period quite as often as the new object, which would now be the right side _in the pair_. Obviously, this must be explained in the following manner: When the chick reacts to the group before him, the chick sees[59] it as a pair in which either gray plays a definite rôle as the dark or the bright side of it; but at the same time the first is seen as a more or less definite dark gray and the second as a bright one. Both ways of seeing them are altogether compatible with each other as long as the pair is not “transposed.” If during learning the chick reacts to the right object as one side of the pair and also as a more or less definite gray, there must be two products of training which will become opposed to each other as soon as we introduce the new pair, because now the first product of learning will favor the choice of one object, and the second the choice of the other. Let us suppose, now, that these different effects of learning do not fade equally fast with time; an increase of the interval between trials with the old pair and trials with the new one will favor those reactions which depend upon the more enduring product of training. From our experiment it follows, then, that the habit depending upon the pair as a whole is more enduring than the habit depending upon the nuance _as such_, since the pair as a whole is relatively less decisive when the animal reacts to the new pair immediately after some trials with the old one. But the two habits are alike with the exception only that one of them depends upon the _organization_ of a trace whereas the other depends upon the trace of a certain nuance of brightness. Hence, in the course of time, the traces of organization remain much more stable than do the traces of a more or less definite gray. This may be a more general fact. It has been remarked by several psychologists that we are frequently able to remember the general structure of something at a time when we cannot get the more particular content of it at all. It would be worth while to test this statement by appropriate experimentation, particularly since we might get some insight thereby into the nature of so-called “concepts.” In the case of the chick it was very easy to examine the hypothesis by having the same animal react to the new pair once more, after some minutes had elapsed since his last trials with the old one. The result was that the “relative” reactions predominated.