Pyllis Young can you assist?

1'sychological Review 1974, Vol. 81, No. 2 , 119-145

AN OPPONENT-PROCESS THEORY OF MOTIVATION:

I. TEMPORAL DYNAMICS OF AFFECT 1

RICHARD L. SOLOMON *

University of Pennsylvania

JOHN D. CORBIT

Brown University

A new theory of motivation is described along with its applications to addiction and aversion. The theory assumes that many hedonic, affective, or emotional states are automatically opposed by central nervous system mechanisms which reduce the intensity of hedonic feelings, both pleasant and aversive. The opponent processes for most hedonic states are strength- ened by use and are weakened by disuse. These simple assumptions lead to deductions of many known facts about acquired motivation. In addition, the theory suggests several new lines of research on motivation. It argues that the establishment of some types of acquired motivation does not de- pend on conditioning and is nonassociative in nature. The relationships between conditioning processes and postulated opponent processes are dis- cussed. Finally, it is argued that the data on several types of acquired motivation, arising from either pleasurable or aversive stimulation, can be fruitfully reorganized and understood within the framework provided by the opponent-process model.

First, we describe the kind of phenome- non which has caught our attention. Two fictitious examples will suffice. In the first, a woman at work discovers a lump in her breast and immediately is terrified. She sits still, intermittently weeping, or she paces the floor. After a few hours, she slowly regains her composure, stops cry- ing, and begins to work. At this point, she is still tense and disturbed, but no longer terrified and distracted. She manifests the symptoms usually associated with intense anxiety. While in this state she calls her doctor for an appointment. A few hours later she is in his office, still tense, still frightened: She is obviously a very unhappy woman. The doctor makes his examination.

1 This research was supported by U. S. Public Health Service Grant MH-04202 to the first au- thor and Grant MH-16608 to the second author. We are grateful to Burton S. Rosner, Francis W. Irwin, and Martin E. P. Seligman for their pains- taking and helpful editing of an earlier draft of this paper. Finally, we are indebted to Dorothea Jameson Hurvich and Leo M. Hurvich, whose development of the Hering theory into their co- herent, opponent-process color vision theory first suggested to us a new way of thinking about af- fect and hedonic process.

2 Requests for reprints should be sent to Richard L. Solomon, Department of Psychology, Univer- sity of Pennsylvania, Philadelphia, Pennsylvania 19104.

He then informs her that there is no possi- bility of cancer, that there is nothing to worry about, and that her problem is just a clogged sebaceous gland requiring no medical attention.

A few minutes later, the woman leaves the doctor's office, smiling, greeting strang- ers, and walking with an unusually buoyant stride. Her euphoric mood permeates all her activities as she resumes her normal duties. She exudes joy, which is not in character for her. A few hours later, how- ever, she is working in her normal, per- functory way. Her emotional expression is back to normal. She once more has the personality immediately recognizable by all of her friends. Gone is the euphoria, and there is no hint of the earlier terrifying ex- perience of that day.

In the second example, a couple have just begun sexual foreplay, and it is quite plea- surable. After a few moments of a constant level of mutual stimulation, the pleasure de- creases somewhat. Normally, this decline would elicit behavior calculated to increase the intensity of mutual stimulation and to maintain the high level of pleasure. Un- fortunately, at that moment a telephone rings. One partner leaves and goes into another room to answer it, and the other partner lies alone in the bed. The aban-

119

120 R J C J J A R I J L. SOLOMON AND J O H N D. COKBIT

+100 INTENSITY OF

PRIMARY AFFECT

INTENSITY OF AFFECTIVE

AFTER-REACTION

+100

PEAK OF PRIMARY AFFECTIVE REACTION

ADAPTATION PHASE

^STEADY LEVEL

DECAY OF "AFTER-RE ACTION

PEAK OF AFFECTIVE AFTER-REACTION

ON

TIME

Standard Pattern of Affective Dynamics

FIGURE 1. The standard pattern of affective dynamics, showing the five distinctive features: the peak of the primary affective reaction, the adaptation phase, the steady level, the peak of the affective after-reaction, and, finally, the decay of the after-reaction. (The heavy black bar repre- sents the time during which the affect-arousing stimulus is present. The ordinate represents two hedonic scales, each departing from neutrality, one for the primary affect, the other for the affective after-reaction.)

cloned partner experiences a quick decline of the pleasure, then becomes tense and ir- ritated, and strongly craves a resumption of the sexual stimulation. Time goes by, how- ever, and the other partner does not return. Finally, the abandoned partner gets out of bed, absentmindedly turns on the television set, and becomes absorbed in a news broad- cast. Gone is the irritability and intense craving. There is no hint, in overt beha- vior, of the pleasurable sexual experience of a few minutes ago. A type of dispas- sionate normality now pervades.

We can distill from these two examples some important empirical features common to many hedonic, emotional, or affective ex- periences. First, following the sudden in- troduction of either a pleasurable or aversive stimulus, an affective or hedonic reaction begins and quickly rises to a peak. It then slowly declines to a steady level where it remains if the stimulus quality and intensity is maintained. Then, at the sudden termi- nation of the stimulus, the affective reaction quickly disappears and gives way to a quali- tatively very different type of affective re- action which reaches its own peak of inten- sity and then slowly disappears with time.

Figure 1 diagrams these changes in he- donic or affective state and illustrates what

we call the standard pattern of affective dy- namics. The pattern has five distinctive features: (a) the peak of the primary he- donic process or state, precipitated by stim- ulus onset; (6) a period of hedonic or affec- tive adaptation during which the intensity of the hedonic state declines, even though stimulus intensity is maintained; (c) a steady level of the hedonic process which continues as long as stimulus intensity is maintained; (d) a peak of affective after- reaction, which quickly follows stimulus ter- mination, and whose quality is hedonically very different from that of the primary he- donic state; and ( e ) finally, the afterstate decays and subsequently disappears.

This standard pattern describes both fic- titious examples. In the first, the initial, primary hedonic or affective process was un- pleasant and the after-reaction was pleasant. In the second, the primary process was pleasant and the after-reaction was unpleas- ant. The data of psychology contain liter- ally dozens of examples of this kind. For brevity we will first describe some of the more interesting or important cases chosen from very different areas of psychological research. We can therefore demonstrate the great generality of the standard pattern of affective dynamics. Then we will de- scribe a theoretical model for the underlying mechanism.

EXAMPLES OF EMPIRICAL P H E N O M E N A To BE EXPLAINED

Table 1 presents seven examples of emo- tional, affective, hedonic, or motivational phenomena. Some are "behavioral," some "experiential." Some are experimental, others are observational, derived from com- mon everyday experiences. Finally, some are precipitated by pleasant, and some by unpleasant stimuli.

Example 1 describes behavior changes seen in a dog subjected to intense aversive stimulation. The example draws on selected parts of studies reported by Katcher, Solo- mon, Turner, LoLordo, Overmier, and Res- corla (1969) and Church, LoLordo, Over- mier, Solomon, and Turner (1966). A dog in a Pavlov harness was stimulated by sev-

OPPONENT-PROCESS THEORY OF MOTIVATION 121

TABLE 1

SELECTED EXAMPLES OF HEDONIC-AFFECTIVE P H E N O M E N A

Example

Dogs in Pavlov harness, 10-second shocks, gross behavior

Dogs in Pavlov harness, 10-second shocks, electro- cardiograph re- sponses

Epstein's para- chutists, free f a l l , gross behavior, physiology

Opiate users, intra- venous injection, moods and feelings

Dogs and M & Ms, gross behavior

Love, interpersonal stimulation, moods, feelings

Imprinting, the at- tachment of creatures to t h e i r "mothers"

First few stimulations

State A (input present)

terror, panic

large cardiac acceleration

terror, autonomic nervous system arousal

euphoria, rush, pleasure

pleasure, tail wag- ging, chewing

ecstasy, excite- ment, happiness

pleasure, cessation of fear, no distress

State B (input gone)

stealth (subdued, cautious, inac- tive, hesitant)

slow deceleration, small overshoot

stunned, stony- faced

craving, aversive withdrawal signs, short duration

tenseness, motion- less

loneliness

loneliness, distress cries, short duration

A f t e r many stimulations

State A' ( i n p u t present)

unhappy (an- noyed, anxious, afraid)

small acceleration or none

tense, eager, expectant

loss of euphoria, normal feeling, relief

—

normal, comfort- able, content

pleasure, no cries

State B' (input gone)

joy (euphoric, active, social), happy

quick deceleration, large overshoot

exhilaration, jubilation

intense craving, abstinence agony, long d u r a t i o n

—

grief, separation syndrome, long duration

loneliness, intense cries, long duration

eral 10-second shocks. The dog appeared to be terrified during the first few shocks. It screeched and thrashed about, its pupils dilated, its eyes bulged, its hair stood on end, its ears lay back, its tail curled between its legs. Expulsive defecation and urina- tion, along with many other symptoms of intense autonomic nervous system activity, were seen. At this point, the dog was freed from the harness, it moved slowly about ( l i e room, appeared to be stealthy, hesitant, and u n f r i e n d l y . Its "state" had suddenly changed from terror to stealthiness.

We now arbitrarily label the state during shock (the terror state or the peak of the primary reaction to shock) State "A." The stealthy state, right after shocks were ter-

minated (the after-reaction), will be called State "B." In that way, we can temporarily ignore whether we have correctly labeled the states. We know that State A was not State B. Indeed, State A was very differ- ent from State B, judging by the many be- havioral changes observed when the dog was suddenly released. Furthermore, State B gradually disappeared. I n a few minutes, the clog appeared to he normal, like its pre- vious, preshock, natural, self: active, alert, and socially responsive. When this hap- pened, it was impossible to tell by looking at the dog that either State A or B had just transpired. The evidence was gone. The dog had progressed from "normalcy" to A to B, and back to "normalcy." This sequen-

122 RICHARD L. SOLOMON AND JOHN D. CORBTT

250

200

150-

BASELINE

0 2.5 5.0 75 10.0

Time After Shock Onset (Sec)

100-

90-

80

BASELINESLINE 4 ma P ^>,---° «>.- -™O-»Vi(

8 ma

0 10 20 30 40 50

Time After Shock Termination (Sees)

FIGURE 2. Heart rate changes as a function of shock onset, maintenance, and termination. (There is a decline following the initial peak reaction to shock onset. There is a deceleratory "overshoot" following shock termination, and then the heart rate slowly returns to baseline rate. Note that the eight-milliampere shock produces a bigger heart rate increase and a bigger deceleration than does the four-milliampere shock.) (Adapted f r o m an article by Russell M. Church, Vincent LoLordo, J. Bruce Overmier, Richard L. Solomon, and Lu- cille H. Turner appearing in the August 1966 Journal of Comparative and Physiological Psy- chology. Copyrighted by the American Psycho- logical Association, Inc., 1966.)

tial pattern will be seen in all of the ex- amples given in Table 1. ft is, we believe, the basic pattern for the dynamics of affect.

However, Example 1 has not yet been fully described. When the same dog was brought back for the same treatment day after day, its behavior gradually changed. During shocks, the signs of terror disap-

peared. Instead, the dog appeared pained, annoyed, or anxious, but not terrified. For example, it whined rather than shrieked, and showed no further urination, defecation, or struggling. Then, when released suddenly at the end of the session, the dog rushed about, jumped up on people, wagged its tail, in what we called at the time "a fit of joy." Finally, several minutes later, the dog was its normal self: friendly, but not racing about. Here again, the sequence was nor- mal —» State A —> State B —» normal, where State A accompanied the arousing input, and State B directly followed the sudden termination of that input, and then slowly died out.

We wish to emphasize that the qualita- tive and quantitative features of States A and B during later shocks were not the same as those for States A and B during the first few shocks. This pattern of changes, occur- ring as a consequence of repeated exposures to the stimulus input which causes A, also is,. we believe, typical of the basic pattern of affective dynamics. Because the later States A and B were not identical to the earlier states, we label the later ones A' and B', respectively.

Look at Example 2 in Table 1. A dog was in an experiment in which its heartbeat was measured by an electrocardiograph dur- ing repeated shocks to its hind feet. The observations are from Church et al. (1966). During the first few shocks, there was a large cardiac acceleration; in some dogs it was an increase of 150 beats per minute. At shock termination, the rate suddenly de- creased, and within 5 seconds it descended below the baseline rate. It often fell as low as 20 to 30 beats per minute below baseline rate. Then, it slowly "recovered" to base- line rate, over a period of as long as 30 to 60 seconds. The below-baseline excursion has been called "vagal overshoot." It is a well-studied phenomenon. Figure 2 is taken from Church et al. (1966) and demonstrates these dynamic events. The figure also sug- gests that there may be a relationship be- tween intensity of the A state and the mag- nitude and duration of the B state. Now we can define baseline, State A, State B, and the return to baseline using an electro-

OPPONENT-PROCESS THEORY OF MOTIVATION 123

cardiograph measurement instead of gross observations of emotional behavior.

States A and B changed markedly after several sessions. Shock onset now caused little, if any, increase in heart rate. Any in- crease was momentary. The rate often de- creased even while the shock was still on. However, when shock was suddenly termi- nated, the "overshoot" was much larger than it was on early shock trials. Heart rate in some individual cases descended to 50 to 60 beats per minute (sometimes more than 50 beats per minute below baseline rate), and recovery to baseline took as long as two to five minutes. As Katcher et al. (1969) put it, "The deceleratory heart-rate overshoot produced by stimulus termination shows shortened latencies . . . and greater magnitude over trials [p. 172]." Thus, States A and B have changed: A' seems to be weaker than A, B' stronger than B, and B' longer lasting than B. But, as in the previous examples, we can still identify the sequence: baseline —> A —> B —» baseline.

Example 3 comes from Epstein's (1967) report of physiological, emotional-expres- sive, and social reactions of parachutists. When parachutists make their first jump, they are often terrified, judging by tele- metered autonomic responses and photo- graphed facial expressions. When they land safely, they look stony-faced or stunned for several minutes, then gradually resume nor- mal composure. After the parachutists have made several jumps and are experts, their responses are different. When jumping, they are no longer terrified. They may be anxious, tense, or even eager. After they land safely, they feel exuberant, exhilarated, and good. They like the feeling, and the mood lasts sometimes for hours. Such parachutists love to jump because of this after-feeling. Again, we see that the quali- tative and quantitative attributes of States A and B have changed with the repetition of eliciting conditions. However, two very different states are still observable, and with each stimulation input, we can identify the sequence: baseline —> A —» B —» baseline.

Example 4 represents states in opiate use (Jaffa, 1965; Maurer & Vogel, 1967). Early in a history of opiate use, the user

experiences the "rush" (an intensely plea- surable feeling) directly after the opiate in- jection, followed by a period of less intense euphoria. Then, with further passage of time, the user suffers aversive, painful, and frightening somatic withdrawal symptoms, together with a feeling of craving. Here again, we see a baseline —> A —> B —> base- line sequence. However, with opiates, B may last a long time, sometimes for several days.

After repeated dosages of opiates over several weeks, State A begins to weaken, and at the same time State B begins to in- tensify and takes longer to return to base- line. State A' is called "normal" rather than euphoric. The rush is no longer ex- perienced. Yet, State B' is more physio- logically disturbing than B was, and lasts much longer than did State B. The craving aspect of State B' is now extremely intense, aversive, and enduring. It is called absti- nence agony. It can be months before B' returns to baseline. Perhaps it never really does. It is a ghastly experience.

The repeated use of some drugs results in the behavioral phenomenon of addiction. People find themselves craving a substance in which they previously had little interest. It is the most vivid instance of acquired motivation, because of its intensity and duration. It also provides a fruitful, em- pirical model for analyzing many kinds of acquired motivation. Indeed, we later show, with examples taken from love and affection, social attachment, and imprinting in birds, that addiction does not differ in principle from any acquired motivational system. We can easily describe opiate, alcohol, barbi- turate, amphetamine, or cigarette addiction (see Solomon & Corbit, 1973) within the empirical framework of the analysis we have proposed. They all have four attributes: (a) The B' state lasts a long time; ( & ) the B' state is intensely aversive; (e) the elici- tation of State A or A' is effective in caus- ing immediate removal of State B or B'; and (rf) the user learns to employ the drug which elicits States A and A' in order to get rid of State B or B'.

A lasting cycle of addiction will not arise, even though A and B are repeatedly experi-

124 R I C H A R D L. SOLOMON AND J O H N D. COKUIT

enced, if the properties of affective response to a drug are such that B fades out to base- line very quickly. This is true because an- other dosage of the drug is never needed in order to get rid of the aversive B state. It quickly gets rid of itself. This is the case in Example 5, a common type of pleasurable situation. A laboratory dog is sitting de- jectedly in a cage. It is suddenly handed one M & M candy. It wags its tail, moves about, chews, and swallows. This reflects State A. It smacks its lips a few times, curls its tongue across its lips, then becomes motionless and tense. It orients toward the experimenter and focuses its eyes on the ex- perimenter's hand (the one which held the single M & M). Assume this to reflect State B. Then, if another M & M is not forthcoming, the B state dies out after about 15 to 30 seconds. The clog moves away, walks around the cage, sniffs here and there, begins to "ignore" the experimenter, then sits again. It has returned to its origi- nal state. Here again, termination of one hedonic event has precipitated another state not present prior to the onset of stimula- tion, and the second state disappeared by itself merely with the passage of time. In this case, A is pleasant, so we infer B to be qualitatively different, and probably craving is the best term for it. The so-called "pea- nut phenomenon" in humans is comparable. Once you start eating peanuts, it is hard to stop unless the cycle is interrupted for a period longer than the time required for the peanut craving, or B state, to die out. This has been called a mini-addiction.3 Indeed, the case of the couple interrupted during mutual sexual stimulation, described in the Introduction, is certainly similar.

In Example 6, the pleasurable input gen- erates a condition in which the B state typi- cally lasts a lot longer than that for an M & M or a peanut, and so the favorable conditions for addiction are present. A boy and girl "fall in love." This S t a t e A is characterized by pleasurable excitement, f r e - quent sexual feelings, a prevailing mood of ecstasy, happiness, and good feelings. When the lovers, whose multimodal mutual stimu-

3 This term was first suggested by Eliot Stellar at a cocktail party.

lation will cause State A, are separated from each other, they will feel lonely, sad, and de- pressed (State B). Even with anticipations of reunion (symbolic, conditioned arousers of State A) loneliness may prevail. Actual reunion will simultaneously erase B and re- instate A just as described in A t t r i b u t e c of addiction to opiates.

After several years of repeated mutual stimulation, the qualitative and quantitative changes in A and B are a matter of public lore. State A' is characterized (if all has gone well) as contentment, normalcy, and comfort. But State B' is now potentially of high intensity and long duration. If it should occur, it is often called grief or, as Bowlby (1952) has described it in children, the "separation syndrome." It requires a lot of time for this B' state to decay. The partners have become addicted to one an- other, and when separated they experience w i t h d r a w a l symptoms. As in all the pre- vious five examples, note that the sudden termination of the stimulus that arouses A or A' leads to the occurrence of B or B' before the eventual return to emotional base- line or normalcy. In this case, the termina- tion of A', even though A' does not manifest itself as intensely as did A, is followed by a more powerful and much more protracted B' state. This is the same pattern seen in all the previous examples, whether the A state is pleasurable or aversive.

Example 7 is one of imprinting. F f we take the duckling as our subject, it shows the same patterns of affect revealed in opiate addiction. First, right after hatching there are very few distress cries. Indeed, the duckling may appear to be quite happy with its new environment. Then, if the duckling is exposed to a white, moving object, it looks intently at it. Also, if there were any distress vocalizations, they tend to disap- pear. However, if the moving object is then removed from view ( H o f f m a n , 1968; Hoff- m a n , S t r a t t o n , Ncwby, & B a r r e t t , 1970), there is a burst of distress crying which may last for several minutes and then dis- appear. With successive presentations and removals of the i m p r i n t i n g s t i m u l u s , the frequency and i n t e n s i t y of distress c r y i n g will increase.

OPPONENT-PROCESS THEORY OF MOTIVATION 125

The efficacy of the presentation of the moving stimulus in eliminating the distress cries appears to be optimal at the outset. Then the duckling can be shaped to push a pole in order to present itself with the im- printing stimulus. At that point, of course, the duckling is exhibiting all criteria for addictive behavior. It is "hooked" on the imprinting object, the presence of which is a positive reinforcer and the absence of which is an aversive event.

Note that the ethological description of imprinting as the sudden establishment of "following behavior," released by an ade- quate imprinting stimulus, is utterly inap- propriate in the light of our analysis. The moving object releases some affective State A, presumably an unconditioned, pleasant emotional reaction to the moving stimulus. The removal of the stimulus then precipi- tates State B, which is an aversive event. State B intensifies with repeated stimula- tions. Ducklings will then work on an avoid- ance schedule to prevent the disappearance of the imprinting object, just as the opiate addict will develop anticipatory behavior which prevents the occurrence of at least the more intense withdrawal symptoms.

EMPIRICAL GENERALIZATIONS

First, in all seven empirical examples, as well as in the fictitious ones in our Introduc- tion, the sudden onset of some new stimulus aroused an affect or hedonic state not pres- ent prior to onset. The state terminated when the stimulus terminated. Then, a new state appeared, qualitatively unlike either the prestimulation state or the state pro- duced by the onset and maintenance of the stimulus. Finally, this new poststimulus state persisted for a while and died out. The baseline state eventually returned. In none of the examples did the subject's af- fective state return directly to baseline upon cessation of stimulation. Baseline was re- gained via some new state which became manifest at stimulus termination, and then slowly died away.

Second, in some cases the states changed in their quality and intensity with successive, repeated stimulations. Whenever this oc-

curred, the A states became weaker and the B states stronger and longer lasting.

These two phenomena, the dynamic he- donic response pattern, and its modification with repeated experience, were seen whether the A state was pleasurable or aversive.

THE EXPLANATORY MODEL

In our opinion, the simplest theoretical model that organizes these typical motiva- tional phenomena (and countless other phe- nomena of innate and acquired motivation) is an opponent-process system. It was bor- rowed from previously developed accounts of sensory dynamics (Hurvich & Jameson, 1957). The primary a process for a given hedonic state is aroused by its adequate stimulus. We then imagine a single oppo- nent loop generating the secondary b process and having an hedonic sign opposite to that of the state aroused by the input. The loop generating the b process is activated when- ever any input evokes a sufficient hedonic consequence. The b process is sluggish, so it has a relatively long latency, recruits slowly, and dies out slowly. Finally, the b process is strengthened by use and weak- ened by disuse.

Because many formal properties of the patterns of both sensory and affective phe- nomena seem so similar, we must show at the outset that they are not identical and that the standard pattern of affective dy- namics is not a direct consequence of the pattern of sensory dynamics. Otherwise, one could explain all hedonic or affective dy- namics in terms of sensory events. For example, in the case of the woman who dis- covered the lump in her breast, one might argue that her steady-level anxiety state was less intense than her initial peak of terror at the discovery of the lump because of sensory adaptation; that is, the perceived presence of and magnitude of the lump decreased. We can refute such an interpre- tation, leaning heavily on the fact that the time course of typical sensory dynamics is of a completely different order of magnitude from that of the standard pattern of affec- tive dynamics reflected in Figure 1 and Table 1. Sensory changes usually occur in

126 R I C H A R D L. SOLOMON AND J O H N D. CORBIT

a matter of milliseconds, seconds, and min- occur in minutes, hours, days, weeks, and utes, whereas the emotional changes usually months. There is, therefore, a theoretical

Pbnel A. THE PROCESSING SYSTEM

nodoUL J L

^Environmental stimulus: a dog

yesf

Cognitive - Perceptual

categorical, all-or-none

detector network

J L

,̂ Cognitive - r

Perceptual signal; yes -no detection

Affective or Hedonic

0 1

^Affective signal

Panel B. THE AFFECTIVE STAGE OF THE SYSTEM

1 1 FIRST nJ L COMPONENT °

^Cognitive - Perceptual signal

Process a : affect emotion hedonic tone feeling mood

I I s^

,

a -signal

-t-

0

>

SECOND COMPONENT

Process b =

THIRD COMPONENT

Summing device a + b, where b is(-)

/

-1

opponent - affect

opponent - emotion

opponent - hedonic tone opponent -

feeling opponent -

mood

<^r~ b- signal

0 7 1

.

Affective signal g - b

Reinforcer quality and intensity;

^ Hedonic output standard pattern

FIGURE 3. Panel A : The detection of an environmental stimulus by a cognitive-perceptual mecha- nism, and the arousal of an affective stage yielding an affective signal which shows the standard pat- tern of Figure 1. Panel B: An analysis of the three components of the affective mechanism. (In the first component the a process is aroused. The second component, the b process, is aroused via the arousal of a. Then the third component, a summing device, combines the a and /; signals to generate the standard pattern of affective dynamics.)

OPPONENT-PROCESS THEORY OF MOTIVATION 127

necessity for a distinct opponent-process mechanism for affect and motivation in- dependent of mechanisms for sensory dy- namics.

Figure 3 illustrates the type of system that we have in mind. Panel A shows two stages of information processing, a cogni- tive-perceptual stage that converts the stim- ulus to an informational signal, and an affective or hedonic stage that converts the informational signal to an affective signal. The affective system in Figure 3 receives a square wave input, and follows it with a dy- namic affective response of the standard pattern shown in Figure 1.

Perhaps an example will clarify the point of Figure 3. The sight of a dog is a fear- arousing stimulus for a cat. The dog repre- sents a complex, multidimensional display, and the perception of the dog is categorical, all-or-none. The stimulus sequence is as follows: The dog enters the cat's environ- ment, remains for a while, and then leaves. The cognitive-perceptual sequence is: The dog is detected, continues to he detected as long as it remains, and then ceases to be detected when it leaves. There is no adap- tation (i.e., the dog does not become less doglike), nor is there any appreciable sen- sory after-reaction at stimulus removal (i.e., no negative [antidog?] afterimage). The affective sequence, in contrast, will show the primary reaction and after-reaction compo- nents : intense fear at first, subsiding to a steady level, and then, when the dog is gone, the appearance of another state, very dif- ferent from the first. The after-reaction (relief?) then gradually dies out.

Figure 3, Panel A, shows that the affec- tive system follows its square wave input with an output having the complex dynamic features of Figure 1. How can we account for this behavior? The affective system in Panel A is shown as a single stage. Now, we will open the "black box" for this stage and look inside. Our proposal for the mechanism responsible for affective dynam- ics is shown in Figure 3, Panel B. Here, the affective stage is analyzed into its three component parts. The cognitive-perceptual stage has acted as a categorical detector. The informational signal enters the affec-

tive system as input to the first component, the a process, which has a short time con- stant. The signal from the a process acti- vates the second component or b process, which responds with a slow rise and a slow decay. The third component is a summing device that adds the a and b signals, and it generates as its output the affective signal, which shows the sequence of the peak pri- mary reaction A, adaptation, steady level, after-reaction B, and decay of B. Thus, we see that the input from the perceptual- cognitive stage has an affective, hedonic side effect, Process a. When it does, the oppo- nent loop is activated, calling into play the opponent process (which has a hedonic quality in some way opposite to, and very different from, that of Process a). The op- ponent process, which we call Process b, reduces the hedonic intensity of the state which the input initially aroused. When the perceptual-cognitive input ceases, the op- ponent process reveals itself as "pure" State B, because the b process takes a while to decay.

The opponent process is a slave process: It is activated indirectly via the activation of the a process. Presumably, the slave process has an evocation threshold, a la- tency, a recruitment or augmentation time, and a decay function, all characteristic of a given opponent-process system. We will also see later that the opponent process can, under proper conditions, be activated by events in memory, as a consequence of Pav- lovian conditioning procedures.

The block diagram of the affect-control system in Figure 3 yields the temporal dy- namics of affect shown in Figure 4, Panel A. There we first see a baseline state. Then the affect-arousing stimulus is pre- sented and it stays on for 10 seconds. Next, it is suddenly terminated. This simple event sequence activates the underlying opponent processes. First, there is a quick rise of Process a to a peak intensity. Shortly after- ward, there is a slow recruitment of Process b. When the stimulus is terminated, Pro- cess a quickly goes to zero, but Process b, having a sluggish decay property, persever- ates and dies out slowly. The resultant manifest dynamics of affect are a conse-

128 R I C H A R D L. SOLOMON AND J O H N D. CORBIT

A

MANIFEST AFFECTIVE 0

RESPONSE B

UNDERLYING

OPPONENT

PROCESSES

STIMULUS EVENT

Panel A.

FIRST FEW STIMULATIONS

Panel B.

AFTER MANY STIMULATIONS

g - b

TIME

FIGURE 4. Panel A: The operation of the summing device for the first few stimulations. (The summation of the underlying opponent processes, a and b, yields the manifest affective response.) Panel B : The operation of the summing device after many repeated stimulations.

quence of subtracting the b process from the a process. The subtraction of the two quan- tities yields Manifest State A when a > b and yields Manifest State B when b > a. The intensity of the manifest state is given by the quantity |a — b\.

When we subtract the underlying pro- cesses, a and b from each other, we obtain

the manifest affective response pattern shown in Figure 5. Immediately after stimulus onset, a is large and b is zero, thus yielding a peak of State A. As b is slowly recruited, the quantity \a — b\ decreases, yielding a decline in the magnitude of State A. This is the adaptation phase shown in Figure 1, the standard pattern of affective

+100

to

Neutral o ca

J5 to

+100

First Few Stimulations

PEAK OF A

ADAPTATION ^STEADY LEVEL OF A

4 OF B PEAK OF B

ON

TIME FIGURE 5. The manifest temporal dynamics generated by the opponent-

process system during the first few stimulations. (The five features of t h e affective response are labeled.)

OPPONENT-PROCESS THEORY OF MOTIVATION 129

dynamics. When the b process reaches an asymptote, there will be a steady level of State A. Then, when the stimulus is ter- minated, the intensity of a goes quickly to zero, but the b process dies away slowly. At this moment, b > a, yielding the peak and subsequent decay of State B. At this point, State B is pure b.

The theory represented in Figures 3 and 4 gives a rough, qualitative account of all of the data of Table 1, for the first few stim- ulations only. First, it explains peaks of intensity of affect or hedonic quality at stim- ulus onset: The opponent process has not yet had enough time to get into action. So we see phenomena like terror, the rush, ecstasy, etc. Or, rate the painfulness of a 30-second shock. The peak painfulness is at onset. For example, see the peak of heart rate in Figure 2.

The other major event explained by the theory is the emergence of the after-reaction, postulated to be a function of the opponent process, which becomes manifest after the termination of stimulus input. This emer- gence is due to pure b perseverating in time after the a has quickly disappeared. The B state slowly decays, and baseline is even- tually attained.

In each of the seven empirical examples of Table 1, the manifest B state was, in some unspecified way, related to A, but not the same as A. The model designates the rela- tionship as oppositeness. But in what way can we say, for example, that loneliness is opposite to the pleasure produced by the presence of a loved one? Surely, they are hedonically opposed. This concept is built into the theoretical model. If A is pleasant, then B must be unpleasant. If A is a posi- tive reinforcer, then B is a negative rein- forcer. Other affective attributes of A and B remain an empirical question. The model tells us how to identify the attributes of oppositeness. All one has to do, for any given A state, is observe the attributes of affect which are revealed at the peak of B. Therefore, the model puts a constraint on what we call oppositeness. For example, the opposite of love must be the attributes of grief. To say that, "Hate is the opposite of love" is simply wrong, when at the death

of a loved one, one experiences grief, not hate. The kinds of oppositeness which are generated by each A state are still, in many cases, waiting to be studied and named. Here is a vast, neglected area of investiga- tion.

We postulated that the /; process is a slave process. That means that at first it cannot be aroused directly by ordinary sensory in- puts, but instead can arise only indirectly via the arousal of an a process (see Figure 3) and the subsequent activation of the op- ponent loop. This fits common sense. Try to imagine being grief-stricken without hav- ing loved someone. Try to imagine craving and abstinence agony without drug use. Try to imagine the exhilaration of the para- chutist without any jump. On the other hand, we will point out later that it may be possible to arouse a b process directly by electrical or chemical stimulation of the brain, or to eliminate it by surgery, or to condition it by Pavlovian procedures. In- itially, however, it is a slave process, inac- cessible to direct environmental inputs, but indirectly arousable via hedonic and affec- tive processes elicited by environmental inputs.

STRENGTHENING OF OPPONENT PROCESSES BY REPEATED STIMULATION

We have not yet explained the changes in hedonic dynamics brought about by repeated affect-arousing stimulations over a relatively long period of time. There are important differences between the A and B states on the left side of Table 1 as compared to the right side. Fortunately, one postulate brings order into the data: The opponent process is strengthened through use and weakened through disuse, but the primary affective process is not seriously affected by use. A b process will acquire more power if fre- quently elicited. It will show a shorter latency of response to a, a quicker rise, a higher asymptote, and a longer decay time. In contrast, an a process is a relatively stable, unconditioned reaction. This seems reasonable for a system which is designed to minimize deviations from affective neu- trality. Why should an opponent process not act like a defensive or immunization pro-

130 R I C H A R D L. SOLOMON AND J O H N D. CORBIT

After Many Stimulations^nn +JUU

I CO

Neutral ° eo

+1001

STEADY LEVEL OF A' -=^>- .- --_,-_._„ Baseline

PXEAK OF B'

ON

TIME FIGURE 6. The manifest temporal dynamics generated by the opponent-

process system a f t e r many repeated stimulations. (The major features of the modified pattern are labeled.)

cess, which produces antibodies more effi- ciently and in larger numbers in the face of repeated challenge? In the same vein, disuse should weaken a b process, and it should slowly return to its original magni- tude whenever its a process has not occurred for a long time.

Figure 4, Panel B, shows how the oppo- nent processes will interact after the b pro- cess has been strengthened by repeated use. A comparison of Panels A and B explains why the left-hand portion of Table 1, de- scribing "early stimulation," differs from the right-hand portion of Table 1, describing "later stimulations."

The effect of repeated experiences on the hedonic response is confined to strengthen- ing the b process. During later stimulations the b process increases at a faster rate at stimulus onset and reaches a greater asymp- totic intensity. In addition, the b process requires a much longer time to decay after the later stimulus terminations. As a con- sequence of these changes in the b process, the resultant pattern of the affective reaction changes so that the peak of A is consider- ably attenuated, and the peak of B becomes much greater and longer lasting. Figure 4, Panel B, shows these simple algebraic sum-

mations of the a and b processes after many repeated stimulations.

Figure 6 shows the pattern of emotional dynamics after many repeated stimulations and should be contrasted with the pattern shown in Figure 5. Three phenomena are corollaries of the use postulate. First, the peak of A' will be less intense because the latency of the b process is decreased and its intensity is increased. Second, the steady level of A', |a — b\, during maintained stimulation will be close to baseline and per- haps even below it in some cases. Third, the peak of B' should be intense and long lasting, compared to what it was during early stimulation (Figure 5 ) .

The events in Figure 6 do not follow from a simple affective contrast model. If a prin- ciple of simple hedonic or affective contrast were operating, then the peak of B in Figure 5 should be greater in intensity than it is in Figure 6, because the intensity of A directly prior to stimulus termination is greater in Figure 5 than it is in Figure 6. However, if we run down the descriptive adjectives for States A, B, A', and B' in Table 1, even as loosely defined as they are, the opponent-process model works well, and a simple contrast model fails.

OPPONENT-PROCESS THEORY OF MOTIVATION 131

More convincingly, the objective data from the electrocardiograph experiments with dogs fit the opponent-process model. In Table 1 we saw the following:

A, large acceleration; B, small overshoot, short duration; A', small or absent acceleration; B', large overshoot, longer lasting.

These findings fit the postulate that repe- titions of affect-arousing stimulation, in this case shocks, will strengthen the opponent process which dampens the primary cardiac acceleration process. Furthermore, there are many signs of qualitative changes; e.g., signs of pleasure and euphoria appear in the period following termination of the shock session, although they were absent in earlier sessions. Our model cannot yet deal pre- cisely with these qualitative changes which are produced by many repetitions of the same stimulus. A' is less intense than A, and B' is more intense and longer lasting than B. The model deduces these quantita- tive changes (Figure 4). In many cases, however, A' is qualitatively different from A, even though they have the same hedonic sign. Similarly B' and B are sometimes qualitatively different, even though they have the same hedonic sign. When the model is fully developed, it must contain a rationale for these qualitative changes.

The observations on the right-hand side of Table 1 also support the assumption that many repetitions of pleasures will strengthen their aversive opponent processes. Aversive states, manifesting themselves after the sud- den termination of pleasurable inputs, be- come more intense with repeated experi- ences. Mild loneliness later becomes grief. Mild craving later becomes abstinence agony and intense craving. In addition, after many repetitions the steady level of plea- sure produced by the continued presence of the pleasurable stimulus input has decreased. The confirmed opiate user experiences a "loss of euphoria," and the rush is gone. The pleasure-affective systems seem even- tually to yield to opponent processes which keep departures from hedonic equilibrium relatively small. The aversive opponent process, when it is manifest, is more intense

and longer lasting than it once was. So fre- quently repeated pleasure has its costs, psy- chologically, in an increased potentiality for displeasure.

Similarly, for A' states aroused by aver- sive input, there is a "cost," but this cost is an increased potentiality for pleasure. Table 1 lists in the column under B' the opponent affects for the aversive A' states. Words like "joy," "exhilaration," and "good feeling" appear. These are the emotional costs of aversiveness. The model thus re- quires that any prolonged or repeated de- partures from hedonic or affective neutrality, regardless of hedonic sign, have a cost. Any significant departure from hedonic or affective neutrality should have correlates in increased autonomic and central nervous system activity aimed at reducing that de- parture. The cost of this activity will not only be psychological, but also will be physi- ological (metabolic, hormonal, and neural).

We are assuming that prolonged exercise of an opponent-process system, whether it be pleasurable or aversive, might cause physiological stress in the same sense that Selye (1950) uses the term stress. That is, many physiological resources might be re- quired in order to keep the opponent pro- cess strong. If we follow Selye's argument, we would expect that such a constant de- mand might lead to the exhaustion of a par- ticular overworked opponent-process system or to the debilitation of other defensive systems.

Thus we come to a major implication of our opponent-process model: There proba- bly are stresses caused by pleasurable stim- ulation just as there are stresses caused by aversive stimulation. Therefore, there should be adaptational costs as a conse- quence of both kinds of stresses. Further- more, if we look for them, we should find diseases of adaptation jor both, due to the correlated, physiological side effects of long- duration often-elicited intense b processes.

In the past, psychologists have identified psychological stress with aversion, pain, and unpleasantness. Theories of mental disease, psychosomatic disease, and behavior disor- ders usually emphasize that aversiveness means stress, stress means aversiveness, and

132 RICHARD L. SOLOMON AND J O H N D. CORBIT

both cause emotional disorders, psychoso- matic illnesses, and behavioral malfunctions. Our opponent-process theory of motivation puts all this to question. From our point of view, stresses caused by aversive stimu- lation are only one half of the story. There should be emotional disorders, psychoso- matic diseases, and behavior disorders caused by long-lasting, repeated, and intense b pro- cesses in general, whether these opponent processes are pleasurable or aversive.

Take the case of the parachutists. Their operant behavior, when B' is strong, will be reinforced by the pleasurable experience of B'. But the cost of neutralizing the innate aversiveness of a fall through space might be high, both physiologically and emotion- ally, just as Epstein (1967) has postulated. Even so, one might imagine, if there were no other pleasurable sensory inputs else- where available to the individual, that he might put himself through repeated aversive stimulation in order to experience the plea- surable B' state which would be both intense and lasting. From our point of view, this type of apparent masochistic behavior is not a mental disorder but is, rather, a reflection of the normal functioning of a healthy, auto- matic, affect-control system. Thus, also, from this point of view, there is nothing ab- normal or strange about addiction. It is only a socially vivid example of the normal hedonic and motivational functionings of an efficiently operating affect-control system.

Motivational systems involving pleasur- able A states and aversive A states are simi- lar. In both cases the onset, maintenance, and termination of the stimulus results in a certain amount of pleasure and a certain amount of displeasure. They differ mainly in whether pleasure or displeasure comes first. In the case of the pleasurable A states, we can assume that the subsequent aversive B state functions as a drive that energizes the performance of operants, and that the pleasurable A states may positively reinforce these operants. Electrical self- stimulation of rewarding brain sites, chemi- cal self-stimulation with opiates, and love relationships should work this way. In con- trast, when the A state is aversive, and when the A state stimulation is absent, noth-

ing functions as a drive to energize an oper- ant upon which the A state is contingent. So we have the problem of how to get the behavior started. The B state for an aver- sive A state is not an energizer. Instead, it is a positive reinforcer. There is thus an important asymmetry between motivational systems for pleasurable and aversive A states. Some outside energizing influence is needed to get behavior started when the operant is followed by an aversive A state. Examples of such behaviors include "thrill- seeking" behaviors, such as parachute jump- ing, mountain climbing, automobile racing, etc., all of which involve an aversive com- ponent, followed by a pleasurable feeling of exhilaration. Why should one initiate an activity when its immediate effect is aver- sive, i.e., when a punishment contingency exists? Some competing outside influence, such as social pressure from peers, is re- quired. However, after many repeated stimulations, such outside influences may not be needed. Because the aversive A state is then weak and the subsequent, positively reinforcing B state is strong, the A state will function as a positive reinforcer. We know that this can happen when a weak shock signals the onset of food for cats (Masserman, 1943). Note that an out- side influence is not needed when the A state is pleasurable. These operants are energized by the aversive B state.

RELATION OF THE THEORY TO OTHER CONCEPTS

To the reader well versed in the history of theories of learning, two aspects of our model should now be apparent. First, the phenomena of acquired motivation produced merely by the repetition of affect-arousing stimuli are nonassociative in nature. For example, the person repeatedly dosed with morphine does not have to know anything and is not required to be subjected to Pav- lovian stimulus contingencies, nor to con- tingencies between operants and outcomes, in order to develop an increasing tendency to suffer when the morphine is withdrawn. The model is therefore very different from previous theories of acquired motivation,

OPPONENT-PROCESS THEORY OF MOTIVATION 133

all of which have emphasized associational processes.

Second, the model, in relating its hypo- thetical mechanism to the phenomena of operant conditioning, explicitly assumes that operants are energized only by aversive states and that they can be reinforced either by the onset of pleasurable states or by the termination of aversive states. This is the case whether the pleasure or aversiveness comes from A states or B states. For ex- ample, the reinforcing effects of shock ter- mination will not only be due to the elimi- nation of the aversive A state but also, and perhaps more importantly, it will be due to the subsequent pleasurable B state. Woodworth and Schlosberg's (1954) con- cept of "safety," Mowrer's (1960) con- cept of "relief," and Denny's (1971) concept of "relaxation" are all emphasizing this con- cept of reinforcement in aversive situations.

The assumption that operants are ener- gized only by aversive states fits most com- fortably with the theoretical position taken by Hull (1943, 1952) and elaborated by Mowrer (1947), Miller (1948), Spence (1956), and Brown (1961). The assump- tion that operants can be reinforced by the onset of pleasurable states fits easily with the position taken by Young (1955) and by Pfaffman (1960). Of course, the assump- tion that operants are reinforced by the termination of aversive states is most like the drive-reduction position maintained es- pecially by Hull (1943) and by Mowrer (1947). These assumptions are really not at issue for us. We take all three to be axiomatic, a point of departure, and we go on from there to spell out the dynamics of pleasurable and aversive states.

CONDITIONABILITY OF A STATES AND B STATES

The opponent-process theory would en- compass an even larger array of data on acquired motivation if it could be safely assumed that A states, or B states, or both could be brought under the control of pre- viously neutral stimuli as a consequence of experience. A Pavlovian conditioning pro- cedure then would result in the establish-

ment of conditioned stimuli which could evoke either state as a conditioned response.

Assume that we have a valid measure- ment of the intensity of affect produced by the onset and maintenance of a 10-second shock in dogs and that we can measure the opponent process when it reveals itself di- rectly after shock termination. Now we re- peatedly pair a conditioned stimulus (tone) with shock onsets. If a Pavlovian condi- tioning process is effective, then the condi- tioned stimulus would then become an A- state elicitor. We call such a conditioned stimulus a CSA. (In the Pavlovian frame- work such a conditioned stimulus would be called a CS+.) If we present a test trial with CSA, with no shock presented, we should see CSA arouse a conditioned State A (or a weak relative). When CSA is sud- denly terminated, we should see State B appear, peak, and then decay to baseline in time. The hedonic recovery from such a test with CSA alone should therefore ap- pear to be biphasic.

Contrast this with what should occur if we establish a conditioned stimulus by pair- ing a signal with the peak of State B, which will occur directly following shock termi- nation. That paradigm is Pavlovian back- ward conditioning: The conditioned stim- ulus always follows termination of the unconditioned stimulus closely in time. Such a conditioned stimulus should become a Pav- lovian CS—, or in our terminology, a CSn. It should be able to elicit a conditioned state very similar to State B. A test with CSs (without a shock preceding it) should result in the conditioned State B appearing at the onset of CSa. Then, after CSu termination, State B should decay monotonically, and baseline should be reestablished. In other words, a monophasic recovery following CSn termination should occur, whereas a biphasic recovery ought to occur following CSA termination. Why a monophasic re- covery for B ? Because, if the b process is a slave process with respect to a, and yet the resultant states are also conditionable, then we ought to be able to arouse the b process directly by onset of a CSB, but then, after CSB has terminated, all State B can do is die away. But a CSA, if successful in

134 RICHARD L. SOLOMON AND JOHN D. CORBIT

20

16-

12-

8 -

4 -

«-CS+

+10 +15 +2'0 +2T+30,

Pos»-CSPre-CS CS

Successive 5 -Second Periods FIGURE 7. Data from Rescorla and LoLordo

(1965). (The solid lines and dots show the bi- phasic response following termination of the fear- arousing conditioned stimulus [CS]. In this graph the CS+ is a CSA. The dotted lines and open circles show the monotonic recovery following the termination of the fear-suppressing conditioned stimulus. In this graph the CS— is a CSu.)

eliciting the conditioned a process, will also indirectly arouse its opponent process. Therefore, State B will be seen immediately after CSA is suddenly terminated, yielding a biphasic recovery. But a peak of State A will never be seen after a CSB is terminated : Only a return of State B to baseline will occur.

In comtnonsense terms, what the oppo- nent-process theory is saying for the case of conditioned fear is peculiar. The sudden termination of a danger signal (CSA) should cause the conditioned subject to act as though it had received a safety signal. But the sudden termination of a safety sig- nal (CSn) should not cause the conditioned subject to act as though it had received a clanger signal. An analogous asymmetry should of course exist for appetitively con- ditioned stimuli, though we are not aware of the existence of experimental data on this point.

Some Pavlovian conditioning data fit these expectations very well for aversive stimulation. Rescorla and LoLordo (1965) trained dogs to avoid shocks on a Sidman nonsignalized avoidance schedule (shock- shock interval = 10 seconds, response- shock interval = 30 seconds) . The avoid- ance response rate was used as a measure of fear intensity. When the dogs' avoid-

ance response rate had stabilized, they were tested during avoidance performance with a CSA which had previously been paired with shock onsets during separate Pavlovian fear-conditioning sessions. The onset of the CSA during the shock-free tests resulted in a large increase in the avoidance response rate. This was interpreted to mean that the avoidance response was energized by the in- crement of conditioned fear caused by the presentation of the CSA. The most inter- esting phenomenon to us is the fact that recovery from a CSA test was biphasic. The avoidance response rate, when CSA was ter- minated, dropped quickly below baseline, then recovered slowly to baseline in 30 to 45 seconds (see Figure 7, solid line with solid circles).

Rescorla and LoLordo (1965) used a dis- criminative conditioning feature, with a conditioned stimulus paired with a long shock-free intertrial interval. When the conditioned stimulus was tested during the clogs' avoidance performance, it suppressed the avoidance response rate. Then, when the conditioned stimulus was terminated, the rate slowly returned to baseline in 30 to 45 seconds (see Figure 7, dotted lines with hollow circles). The recovery from the conditioned stimulus was exactly what we should expect if the conditioned stimulus had been conditioned to the B state. In- tuitively, one might think that the termi- nation of the conditioned stimulus, which is, after all, a safety signal, ought to cause a sudden intensification of fear above the original fear baseline. This does not hap- pen, and adds strength to the theoretical analysis we have presented. However, there is a serious problem in applying the model to these data on Pavlovian differential con- ditioning. The conditioned stimulus was always presented following an intertrial in- terval of 1| minutes or longer. Therefore, by no stretch of the imagination could we consider the conditioned stimulus to have been regularly paired with the peak of B, because the peak of B would have occurred very shortly after the previous shock termi- nation. How, then, can we manage this conceptual problem? If we are to stay within the confines of our model, we must

OPPONENT-PROCESS THEORY OF MOTIVATION 135

make two assumptions. First, after many conditioning trials during which hundreds of shocks have been sustained, the b process would be very strong and last a long time (perhaps for many minutes, or even for hours). Second, we must assume contem- poraneous conditioning, an event—state con- ditioning process, where the conditioned stimulus is the event and the state is B. This is an assumption for which we have no empirical support.

The findings of Moscovitch and LoLordo (1968) add more unambiguous support for the conditioning of a B state. Using the same experimental situation as did Rescorla and LoLordo (1965), they employed a Pav- lovian backward-conditioning paradigm to establish a CSB. According to our analy- sis, this should pair the CSB onset with the peak of B and create a maximally powerful CSB. Recovery from a short-duration test with such a CSB should be monotonic, be- cause there is no mechanism whereby the b process could engender the a process. This is indeed what happens, as shown in Figure 8. Tests with CSB caused a reduction in avoidance response rate to a level about 1/5 of that during the baseline performance. Recovery to baseline after CSs termination took more than 30 seconds (see Moscovitch & LoLordo, 1968, p. 675, Figure 1), and it was monotonic. Moscovitch and LoLordo referred to the backward conditioned stimu- lus as a "safety signal"; they reported their dogs to be relaxed; "there were no indi- cations of freezing or fright [p. 675]." This relaxation occurred in spite of the fact that the dogs were still in the avoidance-training situation, which was frightening enough to motivate a steady rate of avoidance re- sponses (about six per minute). So their backward conditioned stimulus met all the criteria for a conditioned stimulus for a B state.

To convince oneself that B-state condi- tioning has indeed occurred, it might not be sufficient merely to show that the back- ward conditioned stimulus will suppress an ongoing Sidman avoidance response. The decrease could reflect a reduction of fear, a decreased expectation of shock, or both rather than some pleasant hedonic state with

3 -

2

1-

-J5 -10 -5

Pre-CS +15 +20 +25 +30,

Post-CS FIGURE 8. Data from Moscovitch and LoLordo

(1968). (Here we see the response to a Pavlo- vian backward conditioned stimulus [CS]. The recovery from the conditioned stimulus presenta- tion is monotonic and is a C5n.)

positive reinforcement properties. Two lines of evidence would help. First, the gross be- havior of the subjects, such as tail wagging in the presence of the backward conditioned stimulus, would suggest a qualitatively dif- ferent state from that which the avoidance situation normally evokes. Second, the demonstration that a backward conditioned stimulus would subsequently reinforce the development of a new operant in a non- fearful situation might strengthen the con- viction that the conditioned stimulus was really a CSB of a pleasurable sort.

We have seen that Pavlovian laws seem to apply well to the conditioning of both A states and B states precipitated by aver- sive stimulation. We should then expect the same laws to hold for A states and B states precipitated by pleasurable sensory events. First, the conditioned stimulus events for pleasurable A states should be positive reinforcers, should be able to rein- force operants, and finally should be able to counter and sometimes even temporarily eliminate the B state. Therefore, in opiate addiction, presentation of a drug container, a syringe, a needle prick, or a room full of satisfied addicts should all tend to function as CSAs which would oppose the B state by arousing a conditioned A state. Such conditioned opposition would of course be weaker than that produced by the opiate itself, but it should be detectable. Many

136 RICHARD L. SOLOMON AND J O H N D. CORBIT

opiate addicts actually give themselves "sham" injections. Even the familiar social surroundings where the drugs are ob- tained and used should have CSA proper- ties. They should temporarily reduce crav- ing and abstinence panics. Clinical evidence suggests that this is indeed the case (Mau- rer & Vogel, 1967; Wikler, 1971).

The action of €83 events should be even more interesting in addiction, because they should augment the B state and produce more intense craving. Such CSnS would be those paired with the peak of B, the most intense craving state. Thus, we would ex- pect a variety of social and personal events to become conditioned arousers of craving. In general, they should be events and places associated with lack of the addictive drug. A jail cell, confinement, lack of money, all of these should be able to arouse a condi- tioned B state to augment an existing un- conditioned B state. Perhaps the occur- rence of the conditioned B state might precede the actual emergence of the uncon- ditioned B state. This conditioning phe- nomenon should lead to increased frequency of dosage. For someone trying to "kick a habit," the careful avoidance of these CSn events, which elicit B states, should be very helpful.

The conditioning process results in the addict being "hemmed in" to a great extent. If we assume that the events in Figures 7 and 8 occur also for the addictions, then in- creased craving would be a consequence of conditioned stimulus events, no matter whether a CSA or a CSu were presented. In the case of the CSA, the craving should occur following the termination of the CSA (see the biphasic recovery in Figure 7). In the case of CSB, the craving should occur during and after the CSs presentation (see the monotonic recovery in Figure 8). Thus, the Pavlovian conditioning of A states and B states, during the acquisition of an ad- dictive cycle, overdetermines the craving for the drug. The addict, even if he tries to reinforce himself symbolically, i.e., with a variety of CSA onsets, will experience an increased aversive B state afterward.

Conditioned b processes can become very powerful relative to a processes, just as un-

conditioned b processes can become very powerful. Take the work of Kimmel (1971), who used the galvanic skin re- sponse as an anxiety index. In analyzing the phenomena of galvanic skin response conditioning with aversive unconditioned stimuli, he found that the magnitude of an- ticipatory galvanic skin responses in re- sponse to CSAS often increased on the very early trials of conditioning. However, on later trials they decreased, even though the same unconditioned stimulus reinforcer was being regularly presented. Kimmel called this phenomenon the inhibition of emotional behavior, a manifestation of the organisms' adaptive adjustment. His idea is that the fear of the unconditioned stimuli has be- come manageable by the action of a fear- inhibition mechanism. Kimmel's concepts seem to be close to our own, if we change the inhibition concept to an hedonic oppo- nent-process concept. We do not know experimentally whether the phenomena dem- onstrated by Kimmel also manifest them- selves in appetitive conditioning. We sus- pect that they do.

The observations of Kimmel (1971), taken in conjunction with an opponent- process theory, suggest a variant of the ex- periment by Rescorla and LoLordo (1965) which would be a very strong test of the opponent-process theory. One would train dogs to perform a regular Sidman avoid- ance response. Then, during the condition- ing phase of the experiment, a long series of trials with the unconditioned stimulus alone could be presented. Theoretically, this should strengthen the b process for fear. Then the discriminative conditioning phase of the experiment could be carried out just as Rescorla and LoLordo did. When the CSA is presented on a test trial while the dogs are performing this avoid- ance response, the CSA now, in contrast to what Rescorla and LoLordo found, should be relatively impotent in energizing the avoidance response. In other words, the CSA onset would be a weak fear elicitor (Kimmel, 1971). In contrast, the termi- nation of CSA as well as the presentation of the CSB, when presented during avoid- ance responding, should powerfully suppress

OPPONENT-PROCESS THEORY OF MOTIVATION 137

avoidance behavior. This suppressive ef- fect should persist for a relatively long time after termination of the test with CSB. These deductions stem from the assumption that the long series of pretreatment trials with the unconditioned stimulus alone has strengthened the b process, so that when the conditioned stimuli are finally introduced in the later discriminative conditioning, CSA will be paired with a shock which elicits very little fear (b is large, so a minus b is small), but CSB will be paired with the intense peak of the B state which is the opponent of fear (joy, euphoria).

A further deduction from opponent-pro- cess theory suggests a way to reverse what has heretofore been a strong empirical gen- eralization. In most discriminative-condi- tioning experiments the elicitation power of the CS+ grows quickly as a function of trials of conditioning, but the inhibitory properties of the CS— require many more trials to manifest themselves. In the vari- ant of the Rescorla and LoLordo (1965) experiment which we suggested as a strong test of our opponent-process model, the re- verse should be true if discriminative con- ditioning is started after a long series of trials on which the unconditioned stimulus alone has been presented. The effective- ness of the unconditioned stimulus should be greatly attenuated when conditioning is started, and a conditioned stimulus paired with an ineffective unconditioned stimulus should acquire excitatory properties very slowly. In contrast, a conditioned stimulus coming directly after unconditioned stimu- lus termination would be associated with an intense B state. Such a conditioned stim- ulus should quickly acquire CSB properties.

The postulates of an opponent-process theory of affect, combined with Pavlovian laws for the conditioning of A states and B states, give us a new way of looking at the phenomena of acquired motivation.

ANALYSIS OF SELECTED MOTIVATIONAL P H E N O M E N A IN LIGHT OF THE

OPPONENT-PROCESS THEORY

There are many areas of psychological research which can be profitably reanalyzed

within the confines of the opponent-process model. For example, we have found the model useful in increasing our understand- ing of addiction, childhood attachments, love and affection, imprinting in precocial birds, rewarding and punishing electrical stimula- tion of the brain, so-called thrill-seeking behavior, the dynamics of fear and avoidance behavior, the pleasures and displeasures as- sociated with thermoregulatory behavior, some aspects of masochistic behavior, de- pression and euphoria, and taste cravings and obesity, to mention a few. We have selected only two of these phenomena for intensive and detailed analysis in order to demonstrate the applicability and use of the opponent-process model to deduce new phe- nomena and experiments. In the first, the A state is pleasurable, and in the second it is aversive.

Addiction and Stimulation of A and B States with Drugs

We have discussed the main features of an opponent-process theory of affect. These are as follows: (a) Process a; (b) Oppo- nent Process b, a slave process; (c) the strengthening of the b process as a function of repetitions of the a process; and (d) the conditioning of A and B states by Pavlovian contingencies.

We believe any addiction can be better understood by reference to these features. First, the addictive substance must be capa- ble of giving pleasure, at least sometime during its early use. Actually, there may be some aversive effects from the first few doses, but these are overshadowed by the pleasure effects coming either from the drug itself or from other available reinforcers. If not, the dosage will not be repeated. If dosage is repeated, the opponent process will begin to strengthen. Withdrawal symp- toms and craving will intensify and become longer lasting. They will be aversive enough and persistent enough so that the user will try ways of getting rid of them.

It is here that we have to qualify the theory with a specificity assumption. We assume that, because the b process is the opponent of a, the quickest and most effec-

138 RICHARD L. SOLOMON AND JOHN D. CORBIT

tive way of getting rid of the B state is to use the substance which directly produces the A state. We do not know, however, why this should be so, but it is the way ad- dictive drugs work. Thus, behavior re- sulting in the obtaining and use of the A-arousing substance will be strongly re- inforced because it produces A and it simul- taneously terminates B. This will lead to further strengthening of the b process. Therefore, amounts of the substance will have to be increased in order for the sum, (a— b ) , to be greater than zero, or above baseline. This is the drug tolerance phe- nomenon (Jaffe, 1965; Wilder, 1953). In- crease of dosage then will reinforce all the behavior upon which it is contingent, and the b process will be further strengthened by more frequent use. This is the addictive cycle.

During early doses, when the quantity |a — b\ is large, previously neutral stimuli should become conditioned to A if they di- rectly precede the onset of State A. They will become CSA stimuli, Pavlovian elicitors of the A state or components of it. These stimuli, as CSAS for a pleasurable Drug State A, will be positive secondary reinforc- ers, capable of reinforcing new operants.

As the b process becomes stronger, the Pavlovian conditioning of previously neutral stimuli which directly follow a-process ter- mination, and which also directly precede the peak of B, should readily occur. We will then see the emergence of strong CSBs, aversive in quality. They will be the con- ditioned stimuli for the craving and with- drawal symptoms. Thus they will be sec- ondary negative reinforcers and will energize escape and avoidance behaviors. As the CSBs become stronger they will become more anticipatory, and so redosage and re- stimulation will start to occur at shorter intervals. The addict then will be using the drug partly to remove conditioned B- state arousals precipitated by the presence of CSBs. Therefore, increased dosage fre- quency, until some asymptote is reached, should occur whenever: (a) The b process becomes stronger, and (b) CSsS are estab- lished by Pavlovian conditioning contingen- cies. Increased frequency of dosage should

occur even if drug amounts per dosage do not substantially change. Finally, if the ad- dict, under the influence of other motivation, stops using the drug, the b process will weaken gradually with the passage of time.

The dynamics of behavior change which we have outlined are typical of most addic- tions (see Solomon & Corbit, 1973). Natu- ralistic observations appear to fit the deduc- tions from the opponent-process model. Could the opponent-process model for ac- quired motivation help us to understand bet- ter the problems of the addict who wants to quit? Could we devise a regime to help him q u i t ? In a well-addicted drug user, the major behavioral engineering task would seem to be to weaken B'. This is the state which energizes the strong escape and avoid- ance behaviors which terminate craving each time it occurs. The well-addicted drug user is exhibiting avoidance behavior much of the time rather than escape behavior. He would rather not experience the B state; he indulges so frequently that he rarely lets it occur, and if it occurs, he quickly gets rid of it with another dosage. However, in order to weaken B', the addict must lessen the constant, repeated challenges to the op- ponent process. One way is to stop arous- ing A, i.e., cold turkey withdrawal, total cessation all at once. Another way is to decrease the intensity and frequency of A inputs. In other words, fade out A by grad- ually cutting down on drug use.

There is an analogy in color vision. Take a saturated red, and very gradually diminish the saturation until it is gray. A green afterimage does not occur, but the time of fade-out has to be equal to, or longer than, the duration of the green afterimage when red is suddenly terminated rather than faded out. If the time of fade-out is shorter than that, there will be a green afterimage be- cause some finite quantity of the opponent process will still be operative due to its own intrinsic decay time. Therefore, reasoning by analogy, in order to weaken the b process in addiction by a fade-out technique, we must first know how long the craving would have lasted had the addict gone through cold turkey withdrawal until the craving had dis- appeared. This could be a very long time,

OPPONENT-PROCESS THEORY OF MOTIVATION 139

and it must be estimated. If the fade-out technique involves a gradual decrease of drug use frequency and amount, which lasts for a shorter period than the cold turkey period, the technique is bound to fail. That is because, after the last small dose is taken and terminated, the B' state would still come in strongly enough to energize behavior leading to another redosage.

The opponent-process model warns of still another problem for the addict who wants to quit. Even if the drug user were successful in reducing the amount of B, there is still the positive reinforcing effect of A to contend with. Together with so- cial reinforcers, it was probably important initially in reinforcing the user's behavior. Now the a process again has a chance to re- inforce drug use, because the quantity a — b\ gets larger as b gets smaller. Thus

a strong positive reinforcer (A) for drug use will again become available for any ex- perienced user late in the cold turkey with- drawal or the fade-out regimen. This trade-off, deduced by the model, makes the "kicking" of an addiction very difficult. Late in addiction, the aversive craving state energizes the behavior more forcefully, but after partial withdrawal, the positively rein- forcing state reinforces the self-administra- tion operants.

How does one handle the problem of countering a positive reinforcer? First, by the action of punishment. To weaken the drug-use behavior at a time when |a — b\ is large, punishment is one indicated alter- native. Punishment probably should not be used early in withdrawal, or early in fade- out, because the drug-use behavior is com- posed of avoidance and escape operants driven by the aversive B' (the withdrawal agony and craving). Punishment at that time might intensify the operants, as it fre- quently does for any operants controlled by aversive states (the vicious circle phenome- non). Perhaps late in withdrawal, strong rewards for abstinence might be used to- gether with punishment for drug taking. But such rewards would have to be hedoni- cally stronger than is the drug-produced reward A at the time when a is strong and b is weak.

Another technique suggested by the op- ponent-process theory of motivation is the use of antagonistic drugs. It should be pos- sible to break the addictive cycle with a drug which antagonizes or eliminates the aversive b process. The addict medicated with such a b process antagonist would still experience the pleasurable A state, but the withdrawal agony and craving would be gone. He might, under the impetus of other motivational influences, become a casual user rather than a driven addict. An even more effective therapy could be developed if we had two drugs, an o-process antago- nist as well as a ^-process antagonist. The former could be relatively short acting, but the latter would have to be long acting in order to counteract the long-lasting aversive b process.

The opponent-process model deduces that there will be great difficulty in eliminating the addictive cycle. Even though all the conditions for weakening the b process are optimal, trouble can ensue. In addiction, State B is aversive. If aversive B states are not completely specific to their A states, then other aversive states might be reacted to as though they were the B state. A per- son who has reduced his craving by means of protracted abstinence might respond to some types of other aversive emotional states as though they were intense craving states, thus remotivating the drug-seeking beha- vior. We need to know a great deal more about the narrowness of specificity of A and B states before addiction is satisfactorily understood. The cross-tolerance phenome- non tells us that specificity in many cases is not narrow. For example, the alcohol addict develops a strong cross tolerance to barbiturates even though he does not use barbiturates. Therefore, he requires a larger barbiturate dose to produce a speci- fied behavioral effect than he would if he had not already acquired alcohol tolerance. The opponent-process model suggests that the b process elicited by the primary affec- tive consequences of alcohol is somehow similar to that elicited by barbiturates. The opponent-process model is concerned with affective and hedonic processes. Therefore, two drugs, no matter how chemically unre-

140 RICHARD L. SOLOMON AND J O H N D. CORBIT

lated they may be, if they arouse very simi- lar affective states, will be candidates for cross tolerance. Furthermore, this reason- ing deduces that drug-aided withdrawal treatments must fail if the drug used to an- tagonize the b process is one which itself generates an affect similar to that produced by the original addictive agent. Finally, an addict should be very quickly and easily ad- dicted to such a treatment drug. Thus, there are grave problems associated with the use of ^-process antagonists.

There is another difficulty. The action of CSA and CSs stimuli, whose properties were acquired over thousands of condition- ing events usually involved in long-standing addiction, must be considered in the process of withdrawal and "cure." The CSn stimuli will be conditioned arousers of B, the crav- ing. The CSA stimuli will be conditioned arousers of the positive reinforcing State A. A major problem during withdrawal ought to be CSn events, the conditioned elicitors of craving. They would be stimuli asso- ciated with the absence of the drug: no pushers, empty pockets, new surroundings, nonusers (they cannot offer you drugs), etc. Their presence should produce some degree of conditioned craving.

It appears possible, at first glance, to use CSA stimuli to enable an addict to endure intense craving episodes. For example, a nicotine addict might take out a pack of cigarettes, take out a match, put a cigarette in his mouth, then suck on it without light- ing it, going through all the usual acts, in- cluding inhaling, blowing out, etc. This would arouse a conditioned a process, and therefore should temporarily result in less craving than that felt before these acts. However, when the acts are terminated, their aftereffect should be the appearance of a conditioned B state which would be super- imposed on the existing unconditioned one. Then, of course, the craving should tem- porarily intensify. Once again, we see that the behavior of the addict is tightly hemmed in by processes, both conditioned and uncon- ditioned, which overdetermine the persis- tence of the behavior leading to a redose (Solomon & Corbit, 1973).

In conclusion, nature has devised a

powerful way of maintaining behavior which initially produces pleasure, even though eventually the pleasurable consequences of the behavior may become minimal while the aversive consequences of absence of that be- havior become more pronounced. We should not expect a quick cure for addic- tion, then, until we learn how to suppress or antagonize both opponent processes.

The powerful control of behavior result- ing from repeated pleasures is not limited to drug pleasures. Any intense pleasure, the termination of which is followed by an enduring aversive state, should lead to per- sistent, recurrent behavior that is extremely hard to eliminate. Addiction is, therefore, a possible consequence of any repeated pleasure.

Drug addiction is not the only behavioral phenomenon simply organized by an oppo- nent-process model. We cannot examine all of these cases in detail, but we are cer- tain that the reader in each case can de- velop the explanations imposed by the op- ponent-process model. One example is the reinforcing electrical stimulation of the brain. Assume only that the onset of such stimulation produces a pleasurable A state and that the sudden termination of the same stimulation reveals an aversive B state which is of relatively short duration (about 30 seconds). Another example is imprint- ing in precocial birds. Assume only that the sight of a moving white object arouses an unconditioned A state which is a strong positive reinforcer and that the removal of that object reveals a very aversive, long- lasting B state which intensifies rapidly with repeated elicitation. All of the known data on both distress calling in ducklings and following behavior (as well as other oper- ants) are simply explained (Hoffman & Solomon, 1974). Indeed, most of the major phenomena of human attachment behavior are subsumed in this analysis.

Aversive Learning and Performance

We have previously presented evidence that both A states and B states are condi- tionable. Stimuli directly preceding the peak of A will become CSAs for the elicita-

OPPONENT-PROCESS THEORY OF MOTIVATION 141

tion of a conditioned A state. Stimuli di- rectly preceding the peak of B will become CSsS for the elicitation of a conditioned B state. We have been referring to hedonic or affective processes, and so our condi- tioned A states, in the case of shocks, will be conditioned aversive states (fears). Our conditioned B states, in the case of termina- tion of shocks, will be conditioned pleasure states (happiness, relief, euphoria, exhilara- tion). The onset of the A state will nega- tively reinforce operants while the onset of the B state will positively reinforce operants.

The assumptions about the conditionabil- ity of opponent processes suggest a reinter- pretation of escape and avoidance learning and extinction. On the first few avoidance training trials, the aversive A state will be very intense, and its onset will energize a variety of escape behaviors. If signalized shocks are used, the signal will quickly be- come a conditioned A-state elicitor. If a Sidman, unsignalized, training procedure is used, the responses or acts of "staying still" or "doing the incorrect thing" will become conditioned A-state elicitors, and so long bursts of energetic activity will ensue. Early in training, shock termination will result in a weak b process, without much positively reinforcing effect, and it will fade quickly.

As shocks are repeated, two affective pro- cesses will evolve. First, the shocks will decrease in their aversiveness as the oppo- nent b process strengthens. Second, the positively reinforcing effect of the B state will increase, so that successful operants will have an opportunity to be reinforced by two events, the termination of the aver- sive A state and the onset of the pleasurable B state. These operants will be circum- scribed, as contrasted with the wild, disor- ganized, quickly altering bursts of behavior occurring during the first few shocks. Fur- thermore, if the b process strengthens very quickly, some subjects will not learn. They will be the failures so frequently referred to in footnotes (see Turner & Solomon, 1962). They will become "used to "shocks before B-state reinforcement can accurately select the successful avoidance response for

them. Their behavior will first be chaotic. Then they will appear to "give up" and be- come unresponsive to shocks. They will act as if shocks no longer "bother" them very much. Probably they do not.

Provided that the responses narrow down sufficiently to those successful in preventing shocks, a new stage of learning will ensue. Inactivity and other danger signals will con- trol a conditioned aversive A state. The correct avoidance response will control the appearance of the peak of B, the pure B state. The b process will be further strengthened by the repeated elicitation of conditioned A states even when the shocks are no longer experienced. The A state conditioning will then start to extinguish due to the nonoccurrence of the shocks. Meanwhile, the correct avoidance response will continue to be reinforced, differentiated, and perfected as an automatic act, as long as it will produce the pleasant B state. At this point the subject's behavior will have other attributes, too. Because the condi- tioned aversiveness of all stimuli in the training situation is now extinguishing, the subject may willingly enter the experimental room and apparatus. He will perform his avoidance behavior without signs of fear. Yet he will appear overjoyed when the training session is over. This reflects both the weakening of conditioned A and the continued strength of unconditioned and conditioned B. The subject has a new source of positive reinforcement in his en- vironment, the CSu (see Weisman & Lit- ner, 1969a). As Weisman and Litner (1969b) have shown, the CSBs have the power to reinforce new operants. They are, then, conditioned stimuli for a B state which is pleasurable. These stimuli must be serv- ing a function like that of the correct beha- vior of the ski-jumper which produces an exhilaration mood when the danger is over. Another example is that of the parachutists (Epstein, 1967).

The model predicts that avoidance beha- vior will eventually extinguish, but may take a long time. Extinction may, as Hullians claim, be motivated only by the aversiveness of effort, but the contrary aversive forces are also great. Should the subject fail to

142 RICHARD L. SOLOMON AND J O H N D. CORBIT

perform a correct avoidance response dur- ing an extinction series, he will reinstate a powerful C$A which has not been experi- enced since the early training trials. Arou- sal of the A state will then strengthen the b process further. Because no shocks oc- cur, the conditioned A state will weaken. The aversiveness of failing to perform the avoidance operant will consequently decrease.

Avoidance behavior will cease only when all response-produced and external danger signals no longer produce an aversive A state strong enough to energize the avoid- ance behavior. The more effortful the be- havior, the sooner this should happen. The sooner this happens, the more likely is the subject to appear to be frightened by his own inactivity even though he fails to re- spond. Toward the end of an extinction series, then, behavior should be variable. A long "vacation" away from the training situ- ation would weaken the b process, so (a — b) would increase and frequent "spon- taneous recoveries" should be seen, with avoidances occurring early in a new session. At the same time, the joyous behavior, which once characterized the end of a ses- sion, should disappear. Sources of positive affect thus dry up late in the extinction of avoidance responding, but sources of aver- sive affect are still lurking and are rela- tively weak. There should be no such phenomenon as extinction of avoidance behavior with equanimity.

Most of these phenomena deduced from an opponent-process model have occurred at one time or another in avoidance-training experiments, but they have tended to be overlooked or ignored by most writers, in- cluding one of us (Solomon & Brush, 1956; Solomon & Wynne, 1953; Turner & Solo- mon, 1962).

The analysis of escape and avoidance learning and performance in terms of op- ponent-process theory has generality be- yond the mere habituation to electrical shocks in the laboratory. The intense aver- siveness of the first free fall through space, the first exposure to the intense heat of a sauna bath, the first ski jump, or the first killing of another man in war might initially suggest that escape and avoidance behavior

would occur. Instead, the subject often perseveres in re-exposing himself to the aversive event. He looks like the rat which fails to learn to avoid shocks in the labora- tory and appears to be "unmoved" by those shocks. Why would these failures of es- cape and avoidance occur? There must be some outside motivational influence which keeps the subject in the aversive situation long enough to allow the strengthening of the b process. Once this has occurred, the quantity (a — b) is reduced and the B state after stimulus termination is more pleasur- able and long lasting. There is then a rea- sonable trade-off between the immediate oc- currence of the weakened, aversive A state and the somewhat delayed occurrence of the strengthened, pleasurable B state. Accord- ing to our theoretical model, many seem- ingly masochistic behaviors should have the etiology we have described.

A STATES W H I C H HAVE LITTLE OR No OPPONENT PROCESS

A few hedonic disequilibriums may cause very little corrective, oppositional reaction in the nervous system. A nonopposed sys- tem would manifest no peak of the A state, no adaptation, and no appearance of a B state after stimulus termination. One possi- ble example is the hedonic state engendered by marijuana. As a chemical stimulus it precipitates a mildly pleasurable A state. However, there is no reported peak or adap- tation, nor are there aversive withdrawal symptoms or craving. Furthermore, toler- ance often does not develop with repeated dosages. The concommitance of all these attributes would be exactly what the oppo- nent-process model would predict for a pleasure without an opponent. Nausea is an example of an aversive A state which may have no opponent B state. Perhaps some aesthetic pleasures have no opponent process.

We have already talked about the varia- tions in the intensity and duration of B states as a function of the particular kind of A state elicited. For example, we men- tioned the intense and long-lasting aversive B states in the cases of grief and opiate

OPPONENT-PROCESS THEORY OF MOTIVATION 143

withdrawal. In contrast, we mentioned the relatively short duration of B states asso- ciated with the taste of M & Ms and pea- nuts, and rewarding electrical stimulation of the brain. In addition to these inherent variations in intensity and duration of B states elicited by different types of stimuli, repeated elicitation causes large changes in the intensity and duration of B states in some cases and not in others. Therefore, differences in the initial strength of b pro- cesses must reflect a parameter of our model, and the effects of repeated elicitation of A states must operate on this parameter. In the case of A states having no opponent process, the value of the parameter is zero.

We have been constantly amazed by the huge variations in the strengths and dura- tions of different b processes after repeated elicitations. Some b processes, even when well exercised, last but a few minutes. Some examples are taste cravings, the aver- sive b process for electrical stimulation of the brain, cardiac deceleration after shock termination, the "fit of joy" when the dog is released from a shock box, etc. Some last a few hours. Examples are the exhilaration following dangerous or endurance-challeng- ing exercise, or distress calling in precocial birds when the imprinting object is gone. Yet others last for months. Examples are loneliness and grief, craving following with- drawal from either opiates, alcohol, or bar- biturates.

At first glance it may appear that we have completely undercut our theory. We have speculated that some A states may not arouse B states, and we have provided no principles by which one could designate in advance whether or not a particular A state would be a part of an opponent-process pair. At present, we look on this as an empirical problem, and we use the following argu- ment. Opponent processes defend a hedonic equilibrium. They are part of the biologi- cal defense system mediated by the brain. In the realm of foreign body reactions to bacteria, viruses, and poisons we often can detect defense systems. Indeed, the actions of antigens and antibody formation bear many resemblances to the opponent-process systems we have described. However, not

all poisons and not all foreign bodies engen- der defense reactions. We can be defense- less. These substances are, therefore, called deadly poisons. We think the same situa- tion holds in the defense of hedonic equi- librium, and so we should not be surprised to discover that a given A state goes rela- tively unopposed. Perhaps one day we will have a theoretical rationale for the param- eter and operator involved here.

DISCUSSION

We have argued that there are certain systems in the brain, the business of which is to suppress or reduce all excursions from hedonic neutrality, whether those excursions be appetitive or aversive, pleasant or un- pleasant. The systems operate to decrease the intensity of subjective "hedonic qual- ity," "affect," "emotion," "arousal," or the objective reinforcing properties of stimuli. The systems function independently of oper- ants or instrumental acts. They are fully automatic. Thus, whereas operants tend to maximize positive reinforcement and to minimize negative reinforcement, the affect- controlling systems of the central nervous system minimize both. They are brought into play whenever significant departures from affective equilibrium occur as a con- sequence of stimulation onset and mainte- nance. When such a system is effective, it will reduce the intensity of the affective ex- perience even while the input is still there. This reduction will be manifest for both positive and negative reinforcers, and for pleasant and unpleasant stimuli. The theo- retical model refers to both subjective and objective psychological phenomena.

Reduction in affective or hedonic inten- sity is postulated to be brought about by the activation of an opponent loop, precipi- tated into action whenever affective, hedonic, or emotional states are aroused. The op- ponent loop opposes the stimulus-aroused affective state. Furthermore, the opponent process is postulated to be sluggish in its latency, recruitment, and decay (a heavily damped circuit, an inertia-laden system). The opponent loop itself is postulated to generate an hedonic process which is, in

144 RICHARD L. SOLOMON AND J O H N D. CORBIT

some abstract sense, the opposite to that precipitated by the stimulus input which has initially aroused affect. The opponent process will manifest its quality and inten- sity when the stimulus is suddenly termi- nated. The persistence of the opponent process will be seen for some time because of its sluggish decay property.

Furthermore, we have postulated that the opponent process is strengthened through use and weakened through disuse. These changes are nonassociative in nature. This makes the affect-control systems similar to some immunological mechanisms in their properties.

Even though the model is not yet as pre- cise as some mathematical models, from it, nevertheless, one can unambiguously de- duce many of the known phenomena of ac- quired motivation. As examples of this, we have illustrated in detail how the model or- ganizes information on aversive behavior control and the drug addictions.

The theory postulates no conditioning or learning mechanisms responsible for the oc- currence of the acquired motivations we have discussed. The acquisition is auto- matic, merely by virtue of the repeated oc- currence of affect-producing stimuli. How- ever, the interrelationships between the affect-control mechanisms and Pavlovian conditioning processes, and the interrela- tions of both with operant behavior, are discussed.

The novel feature of the opponent-pro- cess theory is that it sees the behavioral phenomenon of addiction as an empirical model for all acquired motivation. Addic- tion is not viewed as an abnormality. In- stead, it is the inevitable consequence of a normally functioning system which opposes affective or hedonic states. We assume, for example, that love is an addiction phenome- non characterized by habituation to the presence of the loved one and intensified aversion in the absence of the loved one. In the same vein, we assume that imprinting in precocial birds is an addiction phenome- non. In the case of aversive stimulation, we assume that masochistic phenomena are the consequence of a normally functioning system which opposes affect. These phe-

nomena are characterized by habituation to the presence of the feared or unpleasant event and intensified pleasure after the ter- mination of that event.

Finally, we have pointed out some new lines of empirical research suggested by the opponent-process theory of motivation.

REFERENCES

BOWLBY, J. Maternal care and mental health. Geneva: World Health Organization, 1952.

BROWN, J. S. The motivation of behavior. New York: McGraw-Hill, 1961.

CHURCH, R. M., LoLoRDO, V. M., OVERMIER, J. B., SOLOMON, R. L., & TURNER, L. H. Cardiac re- sponses to shock in curarized dogs. Journal of Comparative and Physiological Psychology, 1966, 62, 1-7.

DENNY, M. R. Relaxation theory and experi- ments. In F. R. Brush (Ed.), Aver five condi- tioning and learning. New York: Academic Press, 1971.

EPSTEIN, S. M. Toward a unified theory of anxi- ety. In B. A. Maher (Ed.), Progress in experi- mental personality research. Vol. 4. New York: Academic Press, 1967.

HOFFMAN, H. S. The control of distress vocali- zation by an imprinted stimulus. Behaviour, 1968, 30, 175-191.

HOFFMAN, H. S., & SOLOMON, R. L. An oppo- nent-process theory of motivation: III. Some affective dynamics in imprinting. Learning and Motivation, 1974, in press.

HOFFMAN, H. S., STHATTON, J. W., NEWBY, V., & BARRETT, J. E. Development of behavioral con- trol by an imprinting stimulus. Journal of Com- parative and Physiological Psychology, 1970, 71, 229-236.

HULL, C. L. Principles of behavior. New York: Appleton-Century, 1943.

HULL, C. L. A behavior system. New Haven: Yale University Press, 1952.

HURVICH, L. M., & J A M E S O N , D. An opponent- process theory of color vision. Psychological Review, 1957, 64, 384-404.

JAFFE, J. H. Drug addiction and drug abuse. In L. Goodman & A. Gilman (Eds.), The pharma- cological basis of therapeutics. (3rd ed.) New Y o r k : Macmillan, 1965.

KATCHER, A. H., SOLOMON, R. L., TURNER, L. H., LoLoRDO, V. M., OVERMIER, J. B., & RESCORLA, R. A. Heart-rate and blood pressure responses to signaled and unsignaled shocks: Effects of cardiac sympathectomy. Journal of Comparative and Physiological Psychology, 1969, 68, 163-174.

KIMMEL, H. D. Pathological inhibition of emo- tional behavior. In H. D. Kimmel (Ed.), Ex- perimental psychopathology. New York: Aca- demic Press, 1971.

MASSERMAN, J. H. Behavior and neurosis. Chi- cago : University of Chicago Press, 1943.

OPPONENT-PROCESS THEORY OF MOTIVATION 145

M A U R E R , D. W., & VOGEL, M. P. H. Narcotics and narcotic addiction. Springfield, 111.: Charles C Thomas, 1967.

MILLER, N. E. Studies of fear as an acquirable drive: I. Fear as motivation and fear-reduction as reinforcement in the learning of new re- sponses Journal of Experimental Psychology, 1948, 38, 89-101.

MOSCOVITCH, A., & LoLoRDO, V. M. The role of safety in the Pavlovian backward conditioning procedure. Journal of Comparative and Physio- logical Psychology, 1968, 66, 673-678.

MOWRER, O. H. On the dual nature of learning— A reinterpretation of "conditioning" and "prob- lem-solving." Harvard Educational Review, 1947, 17, 102-148.

MOWRER, O. H. Learning theory and behavior. New York: Wiley, 1960.

PFAFFMAN, C. The pleasures of sensation. Psy- chological Review, 1960, 67, 253-268.

RESCORLA, R. A., & LoLoRDO, V. M. Inhibition of avoidance behavior. Journal of Comparative and Physiological Psychology, 1965, 69, 406-412.

SELYE, H. The physiology and pathology of ex- posure to stress. Montreal: Acta, 1950.

SOLOMON, R. L., & BRUSH, E. S. Experimentally derived conceptions of anxiety and aversion. In M. R. Jones (Ed.), Nebraska Symposium on Motivation: 1956. Lincoln: University of Ne- braska Press, 1956.

SOLOMON, R. L., & CORBIT, J. D. An opponent- process theory of motivation: II. Cigarette ad- diction. Journal of Abnormal Psychology, 1973, 81, 158-171.

SOLOMON, R. L., & W Y N N E , L. C. Traumatic avoidance learning: Acquisition in normal dogs. Psychological Monographs, 1953, 67(4, Whole No. 354).

SPENCE, K. W. Behavior theory and condition- ing. New Haven: Yale University Press, 1956.

TURNER, L. H., & SOLOMON, R. L. Human trau- matic avoidance learning: Theory and experi- ments on the operant-respondent distinction and failures to learn. Psychological Monographs, 1962, 76(40, Whole No. 559).

WEISMAN, R. G., & LITNER, J. C. The course of Pavlovian excitation and inhibition of fear in rats. Journal of Comparative and Physiologi- cal Psychology, 1969, 69, 667-672. (a)

WEISMAN, R. G., & LITNER, J. C. Positive con- ditioned reinforcement of Sidman avoidance be- havior. Journal of Comparative and Physiologi- cal Psychology, 1969, 68, 597-603. (b)

WIKLER, A. Opiate addiction. Springfield, 111.: Charles C Thomas, 1953.

WIKLER, A. Some implications of conditioning theory for problems of drug abuse. Behavioral Science, 1971, 16, 92-97.

WOODWORTH, R. S., & SCHLOSBERG, H. Experi- mental psychology. New York: Holt, 1954.

YOUNG, P. T. The role of hedonic processes in motivation. In M. R. Jones (Ed.), Nebraska Symposium on Motivation: 1955. Lincoln: Uni- versity of Nebraska Press, 1955.

(Received May 29, 1973)