In a scientific analysis it is seldom possible to proceed directly to complex cases. We begin with the simple and build up to the complex, step by step. In its early years any science is vulnerable to the charge that it neglects important instances. Boyle’s Law, relating the volume of a gas to its pressure, was a significant advance in knowledge, but a contemporary critic could easily have denounced it as a flagrant oversimplification. It was only necessary to vary the temperature to show that volume was not simply a function of pressure. When the temperature was specified in a new version of the law, more precise measurements could still show that there were discrepancies between different gases, and a “gas constant” had to be added to the equation. There is nothing wrong with this sort of patchwork; it is the way scientific knowledge grows.
In a science of behavior we begin in the simplest way. We study relatively simple organisms with relatively simple histories and under relatively simple environmental conditions. In this way we obtain the degree of rigor necessary for a scientific analysis. Our data are as uniform and reproducible as, say, the data of modern biology. It is true that the simplicity is to some extent artificial. We do not often find anything like it outside the laboratory—especially in the field of human behavior, which is of primary interest. As a result those who are impatient to get on to bigger issues are inclined to object to the “oversimplified” formulations of the laboratory. Their objections take the form, as in the example of Boyle’s Law, of setting forth apparent exceptions to the rule. Such criticism is useful if it points to facts which have been unseen or ignored. But very often the exceptions are only apparent; the existing formulation is capable of giving a good account of them when properly applied.
A common source of misunderstanding is the neglect of what happens when variables are combined in different ways. Although a functional analysis begins with relatively isolated relations, an important part of its task is to show how its variables interact. Several important cases will be discussed in the present chapter.
A given event may have two or more kinds of effects upon behavior at the same time. In the analysis of punishment in Chapter XII it was seen that a single aversive stimulus contingent upon a response has at least four effects. (1) It elicits reflexes, often of an emotional nature. (2) It alters emotional predispositions to act in various ways. (3) It serves as a reinforcing stimulus in respondent conditioning when paired with stimuli which precede or accompany it; these stimuli eventually evoke the responses and predispositions of (1) and (2), and any avoidance behavior which brings the stimuli to an end is reinforced. (4) It makes possible the reinforcement of any escape behavior which brings the punishing stimulus itself to an end. In this example, then, a single event serves as an eliciting stimulus, an emotional operation, a reinforcing stimulus in respondent conditioning, and a negative reinforcer in operant conditioning.
It would be meaningless to say that an event has two or more effects if we could not separate them. When the effects are felt at different times, this is not difficult. For example, a reinforcement may be of such magnitude that considerable satiation takes place. The strengthening effect of the reinforcement may be temporarily concealed by the weakening effect of satiation. Thus a single, relatively large payment of wages may produce such a degree of satiation that the worker does not work again for some time, but the reinforcing effect of the wage will become evident when a sufficient deprivation again arises.
A common objection to the Law of Effect provides another example. The doctrine of “need-gratification” in psychotherapy is based upon the fact that behavior which has been strengthened by deprivation is weakened by satiation. Satiation thus becomes a clinical procedure. For example, behavior which is strong because it has been reinforced with personal attention may be weakened if the individual receives attention or if the primary deprivations responsible for the reinforcing power of attention are reduced. Similarly, behavior which is strong because it gets affection can be weakened by giving affection or appropriate primary reinforcers. It has been argued that these results contradict the Law of Effect, which appears to predict that the behavior should be strengthened rather than weakened. But the case is easily formulated in terms of the multiple effects of giving attention or affection. A child who is behaving in an asocial fashion to draw attention to himself may be “cured” by a sizable measure of attention if satiation takes precedence over reinforcement, as it may. But what will happen when deprivation again arises? If the “cure” sends the patient back for more attention or affection, a reinforcing effect is obvious. (This can be avoided. If a certain “need-gratification” is prescribed, it should be given when the patient is not misbehaving. This will produce satiation without reinforcing undesirable behavior.)
An objection which has been raised to the principle of satiation supplies an example of a different set of multiple effects. Suppose we approach a child who is playing happily by himself and give him a small piece of candy. We may observe the sudden emergence of a great deal of objectionable behavior—asking and teasing for more candy, then crying, and perhaps even a temper tantrum. We appeal to have increased his candy-hunger, although our definition of satiation implies that we have decreased it, at least by a small amount. The explanation is that the candy has had a second effect. The sight and taste of candy are discriminative stimuli under which the behavior of asking or reaching for candy is frequently effective. There is no likelier occasion for the reinforcement of such behavior than the immediate presence of candy. By giving the child a small amount of candy, we establish a common situation in which powerful behavior under the control of candy-deprivation is usually effective and hence strong. We have not made the child any hungrier in terms of deprivation. With a given history of deprivation the behavior of begging for candy shows two levels of strength under the control of two stimuli. In our experiment we change from the stimulus which controls the low level to that which controls the high. Another result then follows. A small piece of candy, as a discriminative stimulus, evokes behavior which is usually reinforced, but we have specified that it is not further reinforced in the present case. Not only does the child ask for candy; he asks unsuccessfully. This is the condition for an emotional reaction of “frustration,” in which the child begins to cry and perhaps ends with a temper tantrum (Chapter X). It is obvious that the child was free of these behaviors before seeing the candy, but this does not mean that he was not hungry. If we were to define hunger in terms of strength of behavior regardless of the presence or absence of discriminative stimuli, we should have to agree that a small amount of food increases it. But the case is not an exception to the present formulation.
We can separate the discriminating and satiating effects of the candy in several ways. For example, a regimen in which a child is never given more than a single piece of candy at a time will eventually extinguish the behavior of asking for more. As a result, the condition responsible for crying or a temper tantrum will not arise. A single piece of candy will have none of the disturbing effects described in this example, and it should be possible to demonstrate a small measure of satiation.
A somewhat more important parallel also shows how easily “drive” is identified with probability of response rather than with a probability due to deprivation. An individual in whom sexual behavior is at the moment not conspicuous may be aroused by exciting conversations, pictures, performances, and so on. It is not correct to say that his sex drive has then been strengthened. Sexual behavior has been strengthened, but by the presentation of stimuli appropriate to such behavior rather than by deprivation.
An operation may have two effects which change the probability of behavior in the same direction. For example, when a response has been reinforced consistently with food but now goes unreinforced for the first time, the probability due to previous reinforcement is decreased and emotional changes in behavior characteristic of frustration are generated. Since the latter include the weakening of any behavior reinforced with food, the first few responses in extinction will be followed by a reduction in rate for two reasons. For a time very few responses will be emitted and hence very few will go unreinforced. The emotional effect will therefore not be sustained, and the rate will rise, only to fall again as further responses go unreinforced. The result is, as we have seen, an oscillation in rate which gives the extinction curve a wavelike character.
At first blush it may seem difficult to separate these effects experimentally. We may, however, demonstrate the emotional effect by frustrating the organism in some other connection. We may also make use of the fact that emotional reactions eventually “adapt out.” By repeatedly extinguishing and reconditioning a response, particularly on a schedule of intermittent reinforcement, we obtain extinction curves with little or no interference from emotional effects. We may also use the fact that an emotional effect involves the whole repertoire of the organism, while extinction is fairly narrowly localized in the response not reinforced. It is possible to record the frequency of emission of two responses in the same organism at the same time. If the responses do not use the same musculature to any great extent, their changes in rate may show a surprising independence. In the pigeon experiment pecking a key and stepping on a pedal satisfy these conditions reasonably well. A somewhat more convenient arrangement is to suspend the pigeon in a harness with one leg free; the pecking response and the flexion of the leg can then be separately but simultaneously studied. When these two responses have been conditioned, they can be extinguished at the same time except for a slight delay in one process. The extinction curves, recorded separately, are slightly displaced in time, but the major oscillations occur simultaneously. This suggests that the rise and fall of frustration is a single process in the whole organism, while the change due to extinction is separately determined in each response.
Another way in which independent variables may interact is of greater importance. Two or more operations may combine in a common effect. We have already discussed several examples. An operant may be reinforced in more than one way, with the result that it varies with more than one deprivation. This is, in fact, the effect of a generalized reinforcer. A response so conditioned is not only more likely to be strong at any given time, because at least, one state of deprivation is likely to prevail, but it may have an especially high probability of emission if two or more states of deprivation prevail at the same time. A similar result is achieved if two or more reinforcements are directly applied to a single operant. The principle is used when attendance at the business meeting of a club is encouraged by the serving of refreshments. Although a member may not attend because of the refreshments alone or because of participation in the business meeting alone, he will be more likely to attend if the probabilities due to both of these reinforcements are combined.
Emotional variables are frequently combined with variables in the fields of motivation and conditioning. Contrary to several well-established views there is no fundamental opposition between emotion and the “intellectual” behavior of the discriminated operant. Behavior is often most vigorous and effective when an emotional predisposition works in the same direction as a contingency of reinforcement. This is implied when we say that “a man’s heart is in his work,” where “heart” refers to emotional variables and “work” to contingencies of reinforcement. The individual in whom aggressive or brutal behavior is particularly strong may work especially well in certain kinds of employment—for example, in certain kinds of police or military work. An actress whose role required her to slap another person in a play slapped with unusual force when she became angry with him for extraneous reasons. The individual with an “affectionate” disposition may be especially successful at jobs which are concerned with helping other people.
In an important application of this principle, one discriminative stimulus is combined with another discriminative stimulus or with other variables. The effects are of various sorts. Some are commonly called “suggestion,” others are dealt with as “projective techniques,” while still others are important in the field of perception. Verbal behavior supplies particularly good examples.1 A single verbal response is especially likely to be a function of more than one variable because it may be part of several different repertoires. In simple imitative or echoic behavior the response is controlled by a verbal stimulus of similar form—the verbal stimulus “house” evokes the verbal response “house.” When the verbal stimulus is of different form—as in the word-association experiment—we may speak of an intraverbal repertoire—the stimulus “home” evokes the response “house.” In reading, the stimulus is a text—the printed stimulus “HOUSE” evokes the vocal response “house.” A great deal of verbal behavior is controlled by nonverbal stimuli, as when we name or describe objects and the properties of objects—an actual house evokes the response “house.” Since a single verbal response usually comes under the control of variables in all these fields, in addition to its relation to emotional and motivational conditions, it is likely to be a function of more than one variable at a time.
The presence of more than one stimulus variable in verbal behavior is sometimes dealt with as “multiple meaning.” The term is too narrow for our present purposes, for we must include contributions of strength from variables which are usually not included in the “meaning” of a response—for example, in the echoic response or the textual response to a printed word. A newspaper article about a convention of dentists reported that, in order to improve their profession, the dentists were urging the passage of certain laws “with teeth in them.” The circumstances under which this was written might have led to alternative responses such as “laws with appropriate penalties” or “laws which could be enforced.” These responses might have been equally probable if another profession had been under discussion. The response “with teeth in them” probably emerged because of the additional strength of the response “teeth”; a particular synonym had taken precedence over equivalent forms because of a multiple causation. Similarly, when a writer discussing a man who had been in China hunting for pandas reported that his plans had not “panned out,” the expression appears to have taken precedence over such synonyms as “worked out,” “come to anything,” or “materialized,” because of a contribution of strength from the variables responsible for “panda.”
The multiple determination of verbal behavior is the basis of much wit. The witty response differs from the unconsciously amusing to the extent that the speaker is able to respond to the multiple sources of strength and to point them up by a proper elaboration. We are concerned here merely with the multiple sources of the witty element, not with the complete joke. An example is a story told of Dean Briggs of Harvard. The Dean was speaking at a dinner on an uncomfortably hot evening. The chairs had recently been varnished, and when the Dean rose to speak, he found his coat stuck to the chair. There was a good deal of laughter as he pulled it loose. When he was at last able to speak, he began, “I had expected to deliver to you a round unvarnished tale, but circumstances make it impossible to fulfill my expectations.” The multiple sources of “unvarnished tale” are essentially the same as those in the preceding examples, but the Dean was able to construct a sentence which made the multiple causation of the response clear to everyone.
All sustained verbal behavior is multiply determined. When a man begins to speak or write, he creates an elaborate set of stimuli which alter the strength of other responses in his repertoire. It is impossible to resist these supplementary sources of strength. We cannot, for example, call out a random series of numbers. Various sequences of numbers are reinforced as we learn to count by ones, twos, threes, or fives, to recite multiplication tables, to give telephone numbers, and so on. When we call out a first number, therefore, we alter the probabilities determining the next call. When a series of some length has been emitted, later, numbers may be extremely powerfully determined.
In the same way, any sustained sample of verbal behavior establishes strong predispositions among the responses still to come. Our imitative or echoic repertoire produces rhyme, rhythm, assonance, and alliteration, which may appear simply as a disturbing singsong or, as in the parallel case of wit, may be elaborated into poetry. Verbal material we have memorized and familiar collocations of words in everyday use establish intraverbal tendencies which add other supplementary strengths. The literary artist exploits these when he fashions a poem or constructs a convincing argument. He builds multiple tendencies in the reader by virtue of which the reader finds himself unaccountably predisposed to “chime in” with the rhyming word of a poem or the clinching word of an argument.
Occasionally, verbal behavior is actually distorted by this sort of multiple determination. We may be able to give a plausible account of the variables responsible, but the speech itself is not always effective. Many years ago a young woman was asked to speak at a dinner advocating the repeal of the Prohibition Amendment. It was her first public appearance, and she was extremely ill at ease. As she rose to speak, someone placed a microphone in front of her. It was an unfamiliar and frightening instrument. She decided to throw herself on the mercy of the audience and plead her inexperience. Her first words were, “This is the first time I have ever faced a speakeasy.” The intruding “speakeasy,” which was as much a surprise to the speaker as to her delighted audience, may be traced to several contributing variables: her subject was in part the evils of the speakeasy, she was concerned with her own ability to speak easily, and a microphone could be called a speakeasy in the sense that it enables one to speak to many people with little effort. We shall see later that the intruding response may also have reduced aversive stimulation from the incipient response “microphone.” We could presumably have shown that the stimulus “microphone” would elicit some of the emotional reflexes which, as in the case of the lie detector, are typical of aversive stimuli. We do not say that the response “microphone” had a tendency “not to be emitted,” but rather that any response which displaced it would be strong for that reason. Because of this overwhelming strength, the response broke into the speech in progress. In spite of the disruption the sources of strength were so obvious that the total response was not without an effect, and it was accepted as wit.
A different kind of distortion arises when two or more fairly similar forms of response are strengthened. One may prevail as the result of both sources of strength, or a combined form may be generated. Folk etymologies (“sparrow grass” for “asparagus”) and blends (“smog” for “smoke” and “fog”) or the portmanteau words of Lewis Carroll (“frumious” for “furious” and “fuming”) are examples. Some distortions are sufficiently effective to survive in the verbal behavior of the community, but others (such as “urving” for “urge” and “craving” or “heritage” for “heresy” and “sacrilege”) suffer a sadder fate.
Supplementary variables are often used in controlling behavior. A familiar case is “suggestion,” which may be defined as the use of a stimulus to raise the probability of a response already assumed to exist at some low value. Verbal suggestions may be classified according to the kind of supplementary stimulation. In the imitative or echoic case, we strengthen a response by supplying stimulation of the same form. We may call this formal suggestion. When we strengthen a response with nonverbal stimuli or verbal stimuli of different form, the suggestion is thematic. A cross-classification may be set up according to whether the response can or cannot be identified in advance. If we call the first a “prompt” and the second a “probe,” then we have to consider formal prompts, formal probes, thematic prompts, and thematic probes.
The formal prompt is the common practice in the theater. A word whispered in the wings strengthens the verbal behavior of the actor by setting up an echoic response which combines with the imperfectly memorized behavior. If the part has not been memorized at all, the actor repeats what he hears from the prompter merely as an echoic response. Since there is then only one source of strength, it is not prompting in the present sense. It is difficult to be sure of multiple sources if the prompter supplies the whole passage, but two variables are obviously at work if he does not. The relative strength of learned material is shown by how much of a prompt is required: if the passage has been fairly well memorized, a very small echoic contribution will suffice. Radio and television quiz programs use a kind of concealed formal prompt. The contestant who finds it difficult to answer a question may be helped if the master of ceremonies makes a remark containing a word which is similar to the answer. If the answer is, say, “Washington,” the concealed prompt might contain the word “washing.”
A thematic prompt having the same effect would be a remark containing the words, “Father of his Country.” When we acquire intraverbal behavior like “Washington was the Father of his Country,” we show an increased tendency to say “Washington” when “the Father of his Country” is heard. Neither the formal nor the thematic prompt will be effective if the response “Washington” does not already exist in some strength. If the contestant is simply told the answer and says “Washington,” this is echoic behavior, and no prompting in the present sense has taken place. The thematic prompt is ordinarily called a “hint.” Hinting, as a type of suggestion, always involves the use of a supplementary variable in rendering a given response more probable.
A formal probe which supplements verbal behavior of unknown form utilizes a process which has long been familiar. We may be interested in the behavior which it reveals because of the light which is thrown on other variables. Ambitious young Dick Whittington, discouraged by his failure in London, leaves the city, but as he walks away he hears Bow Bells tolling the words, “Turn again, Whittington, thrice Lord Mayor of London town.” The stimulus from the bells must have been only vaguely similar to this response. No one else would have heard them saying the same thing. The words represent strong responses in the ambitious Whittington’s own behavior, to which the echoic supplement supplied by the sound of the bells gave the strength needed for emission. (The fact that Whittington heard the bells speak is a separate point to which we shall return later. The only speaker was Whittington himself.) The effect has often been used in literature: a young girl running away from home hears the click of the wheels of the train saying, “Why are you here? Why are you here?”; the lapping of water against the side of a boat whispers, “He speaks the truth. He speaks the truth.”
A device called the Verbal Summator, which is used experimentally and clinically to probe latent verbal behavior, uses the same process. Vague speech patterns—“eye-uh-ah-uh” or “oo-ee-uh-uh,” for example—are repeated by a phonograph so softly or against so noisy a background that they resemble barely audible speech. The subject is asked to listen to each repeated pattern until he hears “what is said.” The feeble echoic response generated by the repeated auditory stimulus combines with a verbal response already in some strength. The resulting response is often emitted with great confidence. A subject may respond to hundreds of different patterns while remaining convinced that they are genuine speech and that he is usually identifying them correctly. An extensive sample of latent verbal behavior may thus be collected which, since it bears little relation to the stimulating situation, must be the product of other variables in the behavior of the subject. The clinical use of the material is based upon the assumption that these variables—in the fields of reinforcement, motivation, or emotion—are probably important in interpreting other behavior of the individual.
A thematic probe is exemplified by the so-called word-association experiment. This is similar to the Verbal Summator except that the supplementary strength is derived from intraverbal responses. A stimulus word is spoken or shown to the subject, and he is asked to report “the first word he thinks of” or, as we should say here, to emit aloud the first verbal response which appears in his behavior. Many different responses are strengthened by an intraverbal stimulus. For example, the stimulus “house” may evoke “home,” “building,” “keeper,” and so on. Which of these is emitted at a particular time is presumably determined by a relatively effective additional source of strength. When verbal behavior is collected in this way, it is possible to infer some of the verbal history of the subject, as well as current variables responsible for his interests, his emotional predispositions, and so on. The clinical use of this material is based upon the assumption that these variables are relevant in interpreting other behavior. The supplementary strength of the thematic probe is not always intraverbal. We may strengthen verbal behavior simply by presenting pictures, objects, or events and asking our subject to talk about them.
By asking our subject to talk in a minimal stimulating situation we generate the condition for what is known as free association, which does not necessarily exemplify the present process. The verbal behavior obtained may be maximally controlled by variables in his history, and inferences about these variables may be of optimal value; but since no supplementary source of strength is used, the case is not classified as either a formal or thematic probe. A great deal of self-probing may go on, however, when parts of such a verbal production alter other parts through supplementary stimulation.
Formal and thematic probes are frequently called “projective tests,” but the word “projection” has a broader significance. Freud described the process to which it refers as a way in which repressed wishes work themselves out (Chapter XXIV). A similar mechanism is called “identification.” Quite apart from any analysis of wishes, we may classify the behavior in terms of its relevant variables; certain occasions for verbal or nonverbal behavior join forces with behavior already in some strength. When we “identify” ourselves with the hero of a novel, movie, or play, or “throw ourselves into a character,” we simply behave in the same way— that is, imitatively (Chapter VII). When our imitative behavior is so microscopic as to be wholly private, a special problem may arise, as we shall see in Chapter XVII. The imitative supplement may be either verbal or nonverbal, but the verbal behavior has several advantages. For example, in reading a novel we can more easily identify ourselves with the character who is speaking than with someone behaving nonverbally because the recorded speech provides a direct source of strength for verbal responses and because these responses can be executed in any environment. A widespread preference for conversation in novels seems to be due to this fact.
The behavior which is executed in identification must have some strength for other reasons. If the strength is considerable, we have to explain why the response is not emitted without supplementation. In a common case the behavior cannot be emitted in everyday life because the opportunity is lacking or because the behavior is restrained or punished. A tendency to identify oneself with, say, a fictional character may be clinically significant as evidence of the strength of the behavior. It is often the case, however, that a story simply builds up a tendency; the author forces a sort of identification, which is evident in the fact that interest in a character grows as the story unfolds. Such an identification may have little bearing upon variables operating elsewhere in the reader’s life.
We speak of projection, rather than identification, when the behavior is less specifically controlled by the supplementary stimulus. A classic example is the lover who accuses his beloved of coolness or unfaithfulness because he himself has grown cool or unfaithful. The lover has reacted with a response which is formally imitative of the behavior of the other person but which is controlled by quite different variables in his own behavior. For example, remaining silent for some trivial reason is imitated and combined with a gesture of boredom; a passing comment is echoed and combined with a critical remark. In what is sometimes called the “old maid’s neurosis,” a response which imitates the behavior of an innocent person is combined with a sexually aggressive response. The fact that the projector attributes similar aggressive behavior to the other person is an additional detail (Chapter XVII).
The possibility of identifying oneself with animals or even with inanimate objects offers an interesting opportunity to study the formal properties of behavior. In what way can a man’s behavior resemble the behavior of a cloud or wave or falling tree so that the imitative response will summate with other parts of his behavior?
It is only a short step to an issue of some importance in the field of perception. Our reactions are determined not only by stimuli, but by supplementary variables in the fields of emotion, motivation, and reinforcement. If we are expecting an important telephone call, we may rush to the phone at the faint sound of a doorbell. This is an example of stimulus generalization, which can easily be duplicated in the rat or pigeon. By increasing the deprivation we increase the range of effective stimuli or, to put it another way, reduce the importance of differences in stimuli. When a young man deeply in love mistakes a stranger passing in the street for his beloved, the strong motivation has made a wider range of stimuli effective in controlling the response of seeing his beloved. (We may report that the doorbell “sounded like” the telephone, and the lover may insist that the girl in the street “looked like” his beloved, just as Dick Whittington heard the bells speak rather than himself. What this means we shall see later.)
Two responses which use the same parts of the body in different ways cannot be emitted together. When two such responses are strong at the same time, the condition is often called “conflict.” When the incompatible responses are due to different kinds of deprivation, we speak of a conflict of motives; when they are due to different reinforcing contingencies, we speak of a conflict of goals; and so on. The term suggests an active struggle of some sort inside the organism—evidently between some of the hypothetical precursors of behavior. The conflict can scarcely be among the independent variables since these are physical events, and any conflict would be resolved at the physical level. From the present point of view, we must suppose that the conflict is between responses and that any “struggle” will be evident in the behavior. If we want to study conflict, then, we simply strengthen incompatible responses and observe the result.
Algebraic summation. When incompatible responses resemble each other in topography except for sign—when, in other words, they are diametrically opposed to each other—the result may be “algebraic summation.” Simple examples are observed in the postural reflexes. One reflex may call for the extension of a leg, another for its flexion. Under certain circumstances the occurrence of both stimuli at the same time produces an intermediate position of the leg. A similar opposition is possible in the discriminative behavior of the whole organism. A dog approaching a strange object, or a soldier going into battle, possesses diametrically opposed kinds of behavior—approach and withdrawal. If no other variables are to be taken into account, the resulting movement will be in one direction or the other but at a qualified speed: the individual will move cautiously forward or slowly retreat. The combination of variables may, of course, have other effects; the behavior may be poorly integrated, less skillfully executed, or, as is always the case with behavior of low strength, easily disturbed by extraneous variables.
If the resulting movement changes the relative strength of the variables, the behavior may oscillate. Thus if the stimulus which induces the dog to approach a strange object is stronger than that which controls withdrawal, the dog will approach slowly, but if this strengthens the variable controlling withdrawal, the direction may at some point be reversed. If withdrawal in turn weakens the variables controlling withdrawal or strengthens the variables controlling approach, a second reversal will occur—and so on. The oscillation will be slow or rapid depending upon the extent to which the variables are modified. The hand of the chess player reaching toward the piece to be moved may oscillate either slowly with a period of several seconds or almost as rapidly as in tremor, depending upon the pressure of the game.
The variables responsible for algebraic summation need not be stimuli. A man whose “heart is not in his work” exemplifies an opposition between reinforcing contingencies and variables in the field of motivation or emotion. Some of his behavior is due to reinforcement, possibly of an economic sort, which keeps him at his job. Opposed to this is behavior which is strong for different reasons. We see this in the tender-hearted thug, in the idealist caught up in a profession in which he must exploit or injure people, or in the pacifist drafted into military service.
Prepotency. Only rarely will the topography of incompatible responses permit algebraic summation since in general one response cannot simply be subtracted from another. In general, when two responses are strong at the same time, only one can be emitted. The appearance of one response is called “prepotency.” The term, like algebraic summation, is borrowed from the study of simple reflexes, but the principle applies to operant behavior as well. We appealed to this principle in noting, as an alternative to extinction or punishment, that we may prevent the occurrence of a response simply by creating circumstances which evoke an incompatible response which is prepotent over it.
The prepotent response does not, merely by virtue of its having been emitted, alter the strength of the dispossessed response. It may, however, change some of the variables controlling this response, and oscillation may then follow. This is all the more likely because the execution of the prepotent response usually weakens it—through partial station, for example. A simple instance is the selection of a necktie. The satiation which follows when a tie is worn is clearly evident when it reaches the point at which we are “tired of the tie,” but a smaller measure of satiation must be supposed to occur in a shorter time. In choosing between two ties, an oscillation may arise since putting on one tie increases the relative probability of putting on the other. The oscillation may under certain circumstances become pathological, as in folie du doute. More important examples are frequently dealt with in literary works. An ancient example is the conflict between behavior strengthened by “love” and behavior due to the ethical pressure which we speak of as “duty” (Chapter XXI). The execution of behavior appropriate to either variable changes the relative strength of the opposed behavior, which then becomes momentarily prepotent.
The oscillation is rapid if only a slight step in either direction makes a significant change in probability, as in the case of the individual who “cannot make up his mind” in ordering at a restaurant. A very slow oscillation is exemplified by the individual who turns from one field of interest to another and back again, perhaps remaining for years in one field. Sometimes a fairly acceptable solution to the problem of incompatible behavior is to engage primarily in one type of response but to interlace one’s activity with responses of another type. This is especially feasible when the latter are relatively independent of the external environment: torn between love and duty, one may do one’s duty while continuing to talk about love. The alternative response may also be executed “in fantasy,” as we shall see in Chapter XVII.
To do or not to do. We are often interested in whether a response will be emitted in competition with alternative behavior which is of no importance to us and which we dismiss as “doing nothing” or as “doing something else.” Such behavior (defined merely as incompatible with a specified response) appears in the analysis of punishment. Any response which interferes with punished behavior reduces a conditioned aversive stimulus and is reinforced for that reason, but we may have little interest in what the response is.
There are several kinds of conflict generated by punishment. An example of a response which is first reinforced and then punished is eating a delicious but indigestible food. The two consequences follow from the chemical properties of the food, which are positively reinforcing on contact with the tongue but eventually aversive in the stomach. In eating someone else’s food without his permission, aversive consequences may possibly be arranged by the owner of the food or by society. The aversive stimulus may precede the positive reinforcement—for example, when we swim in cold water for the effect of the invigorating glow which follows—but in both cases the aversive stimulus is avoided if the response is not emitted. The aversive stimulus may follow unless a response is emitted. When an individual takes steps to prepare for a bad storm, his behavior reduces the threat of strong aversive consequences or “avoids” the consequences of the storm, in the sense of Chapter XI; but a conflict will arise if the behavior has its own aversive consequences.
It is tempting to formulate these cases without mentioning the incompatible behavior. We are interested in whether the indigestible food is eaten, or the plunge taken, or the preparation for the storm made, not in what may be done instead. This may lead us to speak of a negative tendency to engage in the act which is supplanted. One variable increases the probability of a response while another appears to reduce it. But for both theoretical and practical purposes it is important to remember that we are always dealing with positive probabilities. Punishment, as we have seen, does not create a negative probability that a response will be made but rather a positive probability that incompatible behavior will occur.
Another example in which it is tempting to speak of negative probabilities is “Freudian forgetting.” The instances usually described involve punishment. Let us say that an aversive appointment—with the dentist, for example—is forgotten. The observed fact is simply that the behavior of keeping the appointment does not appear under appropriate circumstances. The theory of Freudian forgetting asserts that the aversive consequences of such appointments are relevant. Any step toward keeping the appointment generates conditioned aversive stimulation because of earlier painful stimulation in the dentist’s chair. Any behavior which reduces the aversive stimulation by displacing such a response is automatically reinforced in accordance with the analysis of Chapter XII. Two mutually exclusive kinds of behavior are therefore strong, and the issue is one of prepotency. We have no interest, however, in specifying the incompatible response. Hence we are likely to suppose that forgetting means that the probability of keeping the appointment has reached zero or has passed through zero to a negative value. But we need not deal with any behavior called “not keeping the appointment.” One response has simply lost out to another in the matching of probabilities. If the same result were achieved without “forgetting” by canceling the appointment, the action which supplanted the behavior would be clearly specified, and the principle of prepotency would be obvious. Forgetting is ordinarily attributed to an inner organism which “represses” the behavior of keeping the appointment, but the only repressing agent is the incompatible response.
Just as an additional source of strength may select one response from a group of responses otherwise all equally strong, so a sort of “negative selection” may arise from the strength of behavior which is incompatible with one response in a group. In the example described above, the intruding response “speakeasy” could be explained in part by its effect in displacing the aversive response “microphone.” When we are concerned simply with whether a single response will or will not be emitted, the incompatible behavior may remain unspecified. The basic process, emphasized by Freud, has long been recognized. In Barchester Towers, Anthony Trollope described the behavior of his hero Mr. Arabin as follows:
But he never could have loved the Signora Neroni as he felt that he now loved Eleanor! And so he flung stones into the brook, instead of flinging himself in, and sat down on its margin as sad a gentleman as you shall meet in a summer’s day.
We cannot account for suicide as a simple response. We cannot, for example, measure its frequency. No one jumps into a brook to bring his life to an end because the same behavior has had a similar consequence in the past. But the general behavior of throwing objects into water is another matter. It has a specifiable result: the objects disappear. This behavior is readily generalized; having thrown an old hat into a brook, we get rid of a pair of shoes in the same way. It is not impossible that throwing oneself into a brook may be merely a dramatic example of destroying oneself with the behavior which has destroyed other things. Fortunately we need not decide this issue to make the present point. Both Trollope and Freud agree that Mr. Arabin in flinging stones into the brook was to some extent flinging himself in. Circumstances had given rise to a strong tendency to “throw things into brooks,” but aversive consequences were also attached to some responses in this class. Mr. Arabin does not fling himself into the brook (or, with less aversive consequences, his watch or his pocketbook); he flings stones. This response may have only a tenuous membership in the strengthened group, but at least it has no aversive consequences and hence is emitted. (The same aversive consequences generate the familiar oscillation of the potential suicide, as Hamlet demonstrated.)
In these examples of incompatible behavior, we have considered the outcome when nothing intervenes. Obviously a sudden change in circumstances might yield a different result, and, as we shall see in a moment, the individual himself may effect such a change. Before analyzing how he does so, it is necessary to consider another way in which variables may be arranged.
A response may produce or alter some of the variables which control another response. The result is a “chain.” It may have little or no organization. When we go for a walk, roaming the countryside or wandering idly through a museum or store, one episode in our behavior generates conditions responsible for another. We look to one side and are stimulated by an object which causes us to move in its direction. In the course of this movement we receive aversive stimulation from which we beat a hasty retreat. This generates a condition of satiation or fatigue in which, once free of aversive stimulation, we sit down to rest. And so on. Chaining need not be the result of movement in space. We wander or roam verbally, for example, in a casual conversation or when we “speak our thoughts” in free association.
Some chains have a functional unity. The links have occurred in more or less the same order, and the whole chain has been affected by a single consequence. We often deal with a chain as a single “response.” When a cat pounces on a mouse, for example, this complicated act is an intricate network of postural reflexes, as the physiologist Magnus first showed. We often emphasize the initiating member (to jump or not to jump), overlooking the fact that it precedes by several stages the response which is actually reinforced by contact with the mouse. Long chains organized as simple sequences are exhibited as we pick our way through streets to a particular spot, or recite a poem, or play a piece of music. Other examples have been discussed in connection with conditioned reinforcement. Organized chains are not necessarily confined to the production of stimuli since other sorts of variables may be altered by behavior. In drinking a glass of water we change an important condition of deprivation which has the usual effect of making further drinking less probable, and behavior which has been suppressed by behavior which has led to drinking may then be released. A special kind of chaining is represented by behavior which alters the strength of other behavior and is reinforced because it does so. Such behavior could almost be said to distinguish the human organism from all others. In Section III we shall consider some of the more important problems which it raises.
1 For an extensive analysis of verbal behavior from this point of view see B. F. Skinner, Verbal Behavior, New York, Appleton-Century-Crofts, Inc., 1957.