Learned Helplessness: Theory and Evidence PDF

\, Journal ol Experimental Psychology: General 1976, Vol. 105, No. 1, 3-46 Learned Helplessness: Theory and Evidence Steven F. Maier Ma...

\, Journal ol Experimental Psychology: General 1976, Vol. 105, No. 1, 3-46 Learned Helplessness: Theory and Evidence Steven F. Maier Martin E. P. Seligman University of Colorado University of Pennsylvania SUMMARY In 1967, Overmier and Seligman found that dogs exposed to inescapable and unavoidable electric shocks in one situation later failed to learn to escape shock in a different situation where escape was possible. Shortly thereafter Seligman and Maier (1967) demonstrated that this effect was caused by the uncontrollability of the original shocks. In this article we review the effects of exposing organisms to aversive events which they cannot control, and we review the explanations which have been offered. There seem to be motivational, cognitive, and emotional effects of uncontrol- lability. (a) Motivation. Dogs that have been exposed to inescapable shocks do not subsequently initiate escape response in the presence of shock. We review parallel phenomena in cats, fish, rats, and man. Of particular interest is the discussion of learned helplessness in rats and man. Rats are of interest because learned helplessness has been difficult to demonstrate in rats. However, we show that inescapably shocked rats do fail to learn to escape if the escape task is rea- sonably difficult. With regard to man, we review a variety of studies using in- escapable noise and unsolvable problems as agents which produce learned helpless- ness effects on both instrumental and cognitive tasks, (b) Cognition. We argue that exposure to uncontrollable events interferes with the organism's tendency to perceive contingent relationships between its behavior and outcomes. Here we review a variety of studies showing such a cognitive set. (c) Emotion. We re- view a variety of experiments which show that uncontrollable aversive events pro- duce greater emotional disruption than do controllable aversive events. We have proposed an explanation for these effects, which we call the learned helplessness hypothesis. It argues that when events are uncontrollable the organism learns that its behavior and outcomes are independent, and that this learning pro- duces the motivational, cognitive, and emotional effects of uncontrollability. We describe the learned helplessness hypothesis and research which supports it. Finally, we describe and discuss in detail alternative hypotheses which have been offered as accounts of the learned helplessness effect. One set of hypotheses argues that organisms learn motor responses during exposure to uncontrollable shock that compete with the response required in the test task. Another explanation holds that uncontrollable shock is a severe stressor and depletes a neurochemical necessary for the mediation of movement. We examine the logical structure of these explanations and present a variety of evidence which bears on them directly.  STEVEN F. MAIER AND MARTIN E. P. SELIGMAN What follows are three instances of the fails to solve the second one also. The third phenomenon to be explained: anagram, a relatively easy one, he solves, after about 60 sec. He fails to solve the 1. When placed in a shuttle box an ex- next eight, appearing to give up after about perimentally naive dog, at the onset of the 60 sec with, each one. He then solves six first electric shock, runs frantically about, in a row, but very slowly, and finally sees the until it accidentally scrambles over the bar- pattern. He now solves the final three im- rier and escapes the shock. On the next mediately. trial, the dog, running frantically, crosses the 3. A naive rat is placed in a shuttle box barrier more quickly than on the preceding and trained to escape from shock. Shock trial. Within a few trials the animal be- terminates immediately upon his running to comes very efficient at escaping and soon the opposite side; he learns readily, and learns to avoid shock altogether. After about escapes efficiently. A second rat who had 50 trials the dog becomes nonchalant and received inescapable shock earlier in another stands in front of the barrier. At the onset apparatus learns just as well as the first rat of the signal for shock, the dog leaps grace- to escape in the shuttle box. Now, how- fully across and rarely gets shocked again. ever, the contingency between crossing the But dogs first given inescapable shock in a shuttle box and shock termination is ob- Pavlovian hammock show a strikingly dif- fuscated ; shock does not terminate im- ferent pattern. Such a dog's first reactions mediately upon crossing, but only after 3 sec to shock in the shuttle box are much the elapse. The first rat continues to escape same as those of a naive dog. He runs readily, learning to bridge a 3-sec delay of around frantically for about 30 sec, but then reinforcement. The second rat, however, stops moving, lies down, and quietly whines. fails to respond; on other trials, he runs After 1 min. of this, shock terminates auto- across during shock, but overall he shows no matically. The dog fails to cross the barrier learning curve. and escape from shock. On the next trial, We believe these three phenomena are all the dog again fails to escape. At first he instances of "learned helplessness," instances struggles a bit and then, after a few seconds, in which an organism has learned that out- seems to give up and passively accept the comes are uncontrollable; by his responses shock. On all succeeding trials, the dog and is seriously debilitated by this knowl- continues to fail to escape. edge. This article explores the evidence for 2. A college student is confronted with a the phenomenon of learned helplessness, and series of 25 letter anagrams, each with the discusses a variety of theoretical interpreta- same pattern, 34251. He has a little trouble tions. Since the phenomenon results from with the first one, taking about 45 sec to experience with uncontrollable outcomes, we solve it. He solves each of the next three begin by defining uncontrollability. in about 30 sec, and now he sees the pattern. Each of the last 16 anagrams is solved im- TJJNCONTROLLABILITY AND THE INSTRU- mediately. In striking contrast is the college ! MENTAL TRAINING SPACE student who has first faced a series of dis- llearning theorists have usually viewed crimination problems which are unsolvable the I relations between instrumental respond- or a series of loud tones which are inescap- ing! and outcomes to which organisms are able. He works hard on the first anagram, sensitive in terms of the _ conditional prob- trying many rearrangements of letters, but ability of an outcome or reinforcer following fails to solve it in the 100 sec allowed. He a response />(RF/R), which can have values ranging from 0 to 1.0. At 1.0, every re- The preparation of this manuscript was supported sponse produces a reinforcer or outcome by Grant MH26827-01 to Steven F. Maier and (continuous reinforcement). At- 0, a re- Grant MH19604-01 to Martin E. P. Seligman. sponse never produces a reinforcer (extinc- Requests for reprints should be sent to Steven F. Maier, Department of Psychology, University tion). Intermediate values represent vari- of Colorado, Boulder, Colorado 80302. ous degrees of partial reinforcement.  LEARNED HELPLESSNESS: THEORY AND EVIDENCE One conditional probability, however, is an inadequate description of the relations be- tween response and outcomes about which an organism may learn. Important events can sometimes occur when no specific re- sponse has been made, and it would be a woefully maladaptive organism that was in- p (RF/R) sensitive to such a contingency. Rather than representing environmental contingen-.40 - cies as occurring along a single dimension, we think instrumental training can be better.20 - described using a two-dimensional space, as shown in Figure 1. The.ar-axis />(RF/R) represents the traditional dimension, the (.00 conditional probability of an outcome follow- ing a response. Orthogonal to the condi- tional probability of an outcome, given a FIGURE 1. The response-reinforced contingency response, is the conditional probability of an space. p( RF/R) = conditional probability of an outcome occurring in the absence of that re- outcome following a response, />(RF/R) = condi- tional probability of an outcome occurring in the sponse />(RF/R). This dimension is repre- absence of that response. sented along the y-axis. We assume that organisms are sensitive to variations along is reinforced 50% of the times that it makes both dimensions conjointly, and the empiri- R, but even if it fails to make R, it is rein- cal meaning of this assumption is that sys- forced 20% of the time. tematic changes in behavior should occur The traditional training procedures ar- with systematic changes along both dimen- rayed along the #-axis have been thoroughly sions. There is considerable convergence of explored by many experimenters (e.g., Fer- opinion and evidence among learning theo- ster & Skinner, 1957; Honig, 1966). The rists today that organisms can indeed learn points in the training space which do not about the contingencies within this instru- fall along the #-axis have not, however, been mental training space, including the crucial systematically investigated. Consider the 45° line (e.g., Catania, 1971; Church, 1969;; points that lie along the 45° line (x, y, Gibbon, Berryman, & Thompson, 1974; where x = y). Whether or not the subject Maier, Seligman, & Solomon, 1969; Res- responds, the density of reinforcement is the corla, 1967, 1968; Seligman, Maier, & Solo- same. The conditional probability of an mon, 1971; Wagner, 1969; Weiss, 1968). outcome, given a specific response, does not Thus an organism may learn the extent to differ from the conditional probability of re- which food occurs when it does not make a inforcement in the absence of that response. specific response along with learning the The outcome is independent of responding. extent to which food occurs when it does The concepts of controllability and uncon- make a specific response. trollability are defined within this instru- Consider a few examples. In Figure 1, mental training space. Any time there is Point a (1.0,0) is a case of continuous re- something the subject can do or refrain inforcement : The subject is always rein- from doing that changes what it gets, it has forced for response R, and is never rein- control. Specifically, a response R stands in forced if it fails to make R. Point b (0,1.0) a relation of control to a reinforcer RF if is a case in which the subject is never rein- and only if forced for making the designated R, and is /» (RF/R) ^ # (RF/R). (1) always reinforced for refraining from R (dif- ferential reinforcement of other behavior). Furthermore, when a response will not Consider Point c (.5,.2): Here the subject change what the subject gets, the response  STEVEN F. MAIER AND MARTIN E. P. SELIGMAN, and reinforcer are independent. Specifically, a response R stands in relation of independ- ence to a reinforcer RF if and only if />(RF/R) = (2) When this is true of all emitted responses (as in Pavlovian conditioning) the subject cannot control the reinforcer, and the rein- forcer is defined as uncontrollable. How can we tell that the phenomena we will discuss result from experiencing uncon- 4 5 6 trollable outcomes as opposed to merely ex- TRIALS periencing the outcome itself? To put it FIGURE 2. Median response latency in a shuttle another way, how can we tell whether help- box for dogs given escapable, yoked inescapable, lessness is a psychological phenomenon as or no shock in a harness. (The yoked group did not learn to escape.) (From "Failure to Escape opposed to merely being the result of physi- Traumatic Shock" by Martin E. P. Seligman and cal stimulation? Steven F. Maier, Journal of Experimental Psy- There is a simple and elegant experimental chology, 1967, 74, 1-9, Copyright 1967 by the design which isolates the effects of con- American Psychological Association. Reprinted by permission.) trollability from the effects of the outcome being controlled. In this "triadic" design, trol which it had over shdck; while pressing three groups are used: One group receives the panel did not affect the programmed as its pretreatment an outcome that it can shocks in the yoked group, panel pressing control by some response. A second group terminated shock in the escape group. A is "yoked"—it receives exactly the same naive control group received no shock in the : physical outcome as its counterpart in the hammock. - first group, but there is no response the Twenty-four hours after the hammock yoked subject can make which modifies these treatment, all three groups received escape/ outcomes. A third group receives no pre- avoidance training in a shuttle box. Figure treatment. Later, all groups are tested on a 2 shows the results of this experiment. The new task. escape group and the naive control group Helplessness does not result from trauma performed well in the shuttle box. They per se: In the studies we discuss, deficits jumped the barrier readily. In contrast, the do not occur in the groups that control yoked group was significantly slower to re- shock, but only in the yoked group (Hiroto spond than the escape group and the naive & Seligman, 1975; Maier, 1970; Maier, control group. Six of the eight subjects in Anderson, & Lieberman, 1972; Seligman & the yoked group failed completely to escape Beagley, 1975; Seligman & Maier, 1967).1 shock. So it was not the shock itself, but The triadic design is a direct test of the the inability to control the shock, that pro- hypothesis that learning that shock is uncon- duced failure to respond.1 trollable, and not shock per se, causes help- lessness. Here is an example of how the EFFECTS OF UNCONTEOLLABILITY triadic design is used: Seligman and Maier Having defined the objective conditions (1967) used three groups of eight dogs. An under which uncontrollability occurs and escape group was trained in a hammock to turn off shock by pressing a panel with its lit should be mentioned : that Church (1963) nose. A yoked group received shocks identi- has argued against the use of the yoked group cal in number, duration, and pattern to the as a control group for instrumental learning. This shocks delivered to the escape group. The argument is not relevant to helplessness experi- ments in which the yoked ;group is the experi- yoked group differed from the escape group mental group, and the other groups are each con- only with respect to the instrumental con- trol groups.  LEARNED HELPLESSNESS: THEORY AND EVIDENCE delineated the kind of experimental design of the signal and the shock is 10 sec. If the which isolates the effects of uncontrollability dog jumps the shoulder-high barrier during from the effects of stimulation per se, we now this interval, the signal terminates and shock review the deficits produced by uncontrol- is prevented. Failure to jump during the lable outcomes. In general, when an organ- signal-shock interval leads to a 4.5-mA ism experiences uncontrollable events, three shock which remains until.the dog jumps the deficits often ensue: motivational, cognitive, barrier. If the dog fails to jump the barrier and emotional. within 60 sec after signal onset, the trial a) The motivation to respond in the face automatically ends. of later aversive events seems to wane, b) Between the years 1965 and 1969 the be- Moreover, even if the subject does respond havior of about 150 dogs that received prior and the response succeeds in producing re- inescapable shock was studied. Of these,- lief, the subject often has difficulty learning two thirds (about 100) did not learn to that the response worked, c) Finally, emo- escape and went through the striking se- tional balance may be distributed; depression quence of behaviors that we described. The and anxiety, measured in a variety of ways, other one third seemed completely normal. may predominate. The motivational deficits Like naive dogs, they escaped efficiently. produced by helplessness are in many ways There was no intermediate outcome. In- the most striking, so we turn to them first terestingly enough, of the several hundred for review and analysis. naive dogs who had been given shuttle box training, about 95% have been efficient re- Motivational Deficits sponders. The other 5% failed to learn Dogs. These sets of experiments form the even with no prior inescapable shock. We base of the pyramid on which we construct believe that the prior history of these dogs a theory of learned helplessness, so these before they arrived at the laboratory may studies will be examined thoroughly. The determine whether a naive dog will show a behavior of dogs exposed to inescapable learned helplessness effect and whether a shock seems to typify what many species do dog given inescapable shock will be immune when they are faced with uncontrollability. to this effect. When we discuss the theory Here is the typical procedure that pro- and prevention of the learned helplessness duces learned helplessness in the dog (Over- effect below, we will be more explicit about mier, 1968; Overmier & Seligman, 1967; how to immunize against failure to escape. Seligman & Groves, 1970; Seligman & Since dogs exposed to inescapable shock Maier, 1967; Seligman, Maier, & Geer, seem to be physically capable of jumping the 1968). On the first day, the subject is barrier, the behavioral deficit must have a strapped into a hammock and given 64 in- psychological base. Occasionally, they jump escapable electric shocks, each 5.0 sec long the barrier between trials. Further, if a dog and of 6.0 mA (moderately painful) in- has been sitting and taking shock after shock tensity. The shocks are not predicted by any on the left side of the box, and the door on signal and they occur randomly in time. the right side is opened at the end of the Twenty-four hours later, the subject is given session, he will often come bounding across 10 trials of signalized escape/avoidance train- to escape from the box altogether ing in a two-way shuttle box. The dog The learned helplessness effect in the dog must jump over the barrier from one com- occurs in a variety of situations and is easily partment into the other to escape or avoid produced. Within limits, it does not de- shock. Shocks can occur in either com- pend on the use of any particular shock partment, so there is no place that is always parameters. We have varied the frequency, safe, but the response of shuttling or jump- intensity, duration, and temporal pattern of ing always leads to shock termination. shocks, and still produced the effect. Further- The onset of a signal (light dimming) be- more, it does not matter if the inescapable gins each trial, and the signal stays on until shock is or is not preceded by a signal. the trial ends. The interval between the start Finally, it does not matter what apparatus  STEVEN F. MAIER AND MARTIN E. P. SELIGMAN the inescapable shocks are given in or where goldfish and then tested them in an aquatic the escape/avoidance training takes place. shuttle box. These fish were slower to The shuttle box and hammock are inter- avoid than naive controls. (See Behrend changeable. If the dog is given inescapable & Bitterman, 1963; Bintz, 1971; Frumkin shock in the shuttle box and then required & Brookshire, 1969; Padilla, 1973, for re- to press panels with its head to escape in lated goldfish data.) the hammock, he is still helpless. Further, Rats. Until recently, it has proven dif- after uncontrollable shock, dogs are not only ficult to produce a learned helplessness effect debilitated at fleeing from the shock itself in rats. A substantial number of experi- (escape), but they also seem to be unable ments were performed involving inescapable to prevent or "avoid" shock. Overmier shock; by and large, however, these showed (1968) gave dogs inescapable shock in the rather small, if any, effects on later response hammock and then tested them in a shuttle initiation. (Maier et al., 1969, and Selig- box. If the dog jumped after the signal man et al., 1971, reviewed the complicated went on, but before the shock came on, he literature, and the interested reader is re- could avoid the shock. But no escape was ferred there for details. See also Anderson, allowed, for if the dog failed to jump in the Cole, & McVaugh, 1968; de Toledo & Black, signal-shock interval, the barrier was closed 1967; Dinsmoor & Campbell, 1965a, 1965b; and inescapable shock occurred. The help- Looney & Cohen, 1972; Mullin & Mogen- less dogs failed to avoid, just as they had son, 1963; Weiss, Krieckhaus, & Conte, failed to escape. So, dogs previously ex- 1968, for representative experimental posed to inescapable shocks failed to cope studies.) Unlike dogs, a rat given prior adaptively with signals for shock as well as inescapable shock was typically only a bit with shock. slower to escape shock on the first few trials, Debilitation of response initiation as a con- or slower to acquire avoidance—it did not sequence of uncontrollable outcomes has fail to learn. been reported in cats, rats, mice, birds, pri- After intensive experimentation, however, mates, fish, and man, as well as in dogs. several investigators have :now independently The learned helplessness effect seems rather produced substantial learned helplessness ef- general among species that learn. fects in rats. (Maier, Albin, & Testa, 1973; Cats. Thomas and Baiter (in press) re- Maier & Testa, 1975; Seligman & Beagley, ported an effect in cats which seems identical 1975; Seligman, Rosellini, & Kozak, 1975.)« to learned helplessness in dogs. (See also In doing so, one crucial factor emerged—the Masserman, 1943, 1971; Seward & Hum- response used in the test for learned help- phrey, 1967; Zielinski & Soltysik, 1964, for lessness must be difficult, and not something other reports of debilitation in cats caused by the rat does very readily. So, for example, inescapable shock). They designed a ham- if rats are first exposed to inescapable shock mock for cats and gave them inescapable and then tested on a simple escape response shock. When later placed in a cat shuttle like pressing a bar once (FR-1), or fleeing box, those cats failed to escape. Like dogs, to the other side of a shuttle box, no deficits they sat and took the shock. Thomas and are found. If, however, the response re- Baiter also reported that blocking activity of quirement is increased—so the bar must be the septal area of the brain breaks up help- pressed three times in order for shock to lessness, and directly stimulating the septum end (FR-3 lever pressing) or the rat has to electrically produces a learned helplessness- run across the shuttle box and back (FR-2 like effect. When we discuss the theory and shuttling)—then the rat that has experienced therapy for helplessness, we will examine inescapable shock fails to escape. In con- these physiological data in more detail. Fish. Following inescapable shock, fish 2 also show poor escape and avoidance re- In passing it should be mentioned that mice have been reported to show response deficits fol- sponding. Padilla, Padilla, Ketterer, and lowing inescapable shock (Braud, Wepman, & Giacolone (1970) gave inescapable shock to Russo, 1969).  LEARNED HELPLESSNESS: THEORY AND EVIDENCE 9 60 were then taken to a hand shuttle box; in FR-2 FR-3 order to escape noise the individual had only 50 to move his hand from one side to the other. Preshocked Both the no-noise and escape group learned 40 readily to shuttle with their hands. Like other species, however, the human inescap- 30 able group failed to escape and avoid; rather, most sat passively and took the aver- 20 sive noise. (For other learned helplessness 10 experiments in man, see Fosco & Geer, 1971; Glass & Singer, 1972; Hiroto & Selig- man, 1975; Klein, Fencil-Morse, & Selig- 1 2 3 4 1 2 3 4 1 2 3 4 Blocks of 10 Trials man, in press, Krantz, Glass, & Snyder, 1974; Miller & Seligman, 1975a, Note 1; FIGURE 3. Mean latency to press a lever to Racinskas, 1971; Rodin, 1975, Roth, 1973; escape shock in an FR-1, FR-2, and FR-3 sched- ule, for rats given inescapable or no shocks. (The Roth & Bootzin, 1974; Roth & Kubal 1975; yoked subjects escaped very poorly only on the Thornton & Jacobs, 1971.) FR-3.) (From "Learned Helplessness in the Rat" Hiroto's design was actually more com- by Martin E. P. Seligman and Gwyneth Beagley, plicated and had two other important factors. Journal of Comparative and Physiological Psy- chology, 1975, 88, 534-541. Copyright 1975 by the Half the subjects in each of the three groups American Psychological Association. Reprinted were told that how they did in the shuttle by permission.) box was a test of skill; the other half were told that how they did was governed by trast, rats that have had prior escapable chance. Those who received chance instruc- shock or no shock learn even more difficult tions tended to respond poorly in all groups. responses without difficulty (see Figure 3). Finally, the personality dimension of "ex- Hannum, Rosellini, & Seligman (in press) ternal vs. internal locus of control of rein- recently extended these findings to develop- forcement" was also varied in Hiroto's de- ment. Three groups of rats received four sign (see Lefcourt, 1966; Rotter, 1966; sessions of escapable, inescapable, or no James, Note 2), with half of all the students shock shortly after they were weaned. At in each group being "externals" and half 90 days of age they were tested on an FR-3 lever press escape task. Rats that had re- 60 r ceived inescapable shock at weaning failed to escape shock. Rats that had received - SO escapable shock or no shock escaped well vu Yoked (see Figure 4). 40 Man. What are the laboratory effects of inescapable trauma in man? Like the dog, 30 cat, rat, and fish, when a human being is Control faced with noxious events that it cannot 20 control, its motivation to respond seems to be reduced. Hiroto's study (1974) is representative 10 Escapable and it replicated in college students the find- ings on dogs. In a triadic design, subjects 1 2 3 4 in his escape group received loud noise which Blocks of 10 Trials they learned to turn off by button pushing. The subjects in the inescapable group re- FIGURE 4. Mean escape response latency for ceived the same noise, but the noise was rats given either escapable, yoked inescapable, or no shocks as infants (Hannum, Rosellini, & Selig- independent of their responding. A third man, in press). The rats given inescapable shocks group received no noise. All the groups as infants did not learn to escape.) 10 STEVEN F. MAIER AND MARTIN E. P. SELIGMAN "internals." An external is a person who believes, as shown by his answers on a per- sonality inventory, that reinforcements occur in his life by chance or luck and are beyond his control. An internal believes that he controls his own reinforcers and that skill will out. Hiroto found that externals be- came helpless in his experiment more easily than did internals (see also Dweck, 1975). So three independent factors, the laboratory procedure of uncontrollability, the cognitive set induced by chance instructions, and ex- 1 3 5 7 13 IS 16 8 2O 22 2« 26 28 30 ternality, each produced a learned helpless- Trials ness effect. Given this convergence, Hiroto FIGURE S. Mean speed to escape a nonrewarded concluded that these three factors all erode goalbox. (The group designated by E previously motivation to escape by contributing to the received escapable shocks, the group designated by expectation that responding and relief are I received inescapable shocks, and the groups independent. designated by C received no shocks. The groups designated by IS received rewarded runway ac- This concludes the survey of motivational quisition training, whereas the group designated by deficits produced by exposure to uncontrol- 0 did not. The rats.given inescapable shocks lable aversive events in different species. It [15-1] showed no tendency to escape the frustrat- seems to be generally true that uncontrol- ing goalbox.) (From "Frustration and Learned Helplessness" by Robert A. Rosellini and Martin lability produces deterioration in the readi- E. P. Seligman, Journal of Experimental Psy- ness of dogs, cats, rats, fish, monkeys, and chology: Animal Behavior Processes, 197S, 1, men to respond actively to trauma. 149-157. Copyright 1975 by the American Psy- Generality. Does uncontrollability pro- chological Association. Reprinted by permission.) duce a habit limited to situations very like the ones in which uncontrollability is experi- in the rat, also for the earliest apparent enced, or does it produce a more general ef- learned helplessness study in the literature.) fect ? The question we are really asking here Rosellini and Seligman (1975) performed is "Is helplessness just an isolated set of an experiment in which shock-induced habits or does it involve a more basic learned helplessness may have transferred change?" We believe that what is learned to frustration. Three groups of rats received when the environment is uncontrollable can escapable shock, inescapable shock, or no have consequences for a wide range of be- shock. After this pretreatment, they learned havior. to run down an alleyway in order to get At the lowest level of generality, helpless- food in a goal box. Food was present on ness transfers from some types of apparatus every trial in the goal box. Once they had to others, as long as shock occurs in both learned, food was no longer placed in the.situations; dogs given inescapable shock in goal box. So during this extinction pro- a hammock fail to escape later in a shuttle cedure, the rats ran down the alleyway into box. But does what is learned transfer to the goal box where they expected food, but aversive situations not involving shock? found none. This has been shown to be a Braud et al. (1969) used a triadic design "frustrating" and aversive experience for a with mice. One group could escape shock rat (Amsel, Rashotte, & MacKinnon, 1966). by climbing up a pole, a second group was The rats were then given an opportunity to yoked, and a third group received no shock. jump out of the goal box and escape this All groups were then placed in an alley frustration. Rats that had experienced flooded with water and had to swim out in escapable shock or no shock escaped the goal order to escape. The yoked group was box readily. Rats that had experienced in- poorest at escaping from water, (See Mc- escapable shock did not escape from the goal Culloch & Bruner, 1939, for similar findings box. (See Figure 5.) The effect of ex-  LEARNED HELPLESSNESS: THEORY AND EVIDENCE 11 gressive as well as defensive responses. (See 40-1 Brookshire, Littman, & Stewart, 1961, Ex- periment 6; and Anderson & Paden, 1966, for other evidence on intra-aversive trans- fer.) X O Does experiencing uncontrollable aversive 30H events have effects on nonaversively moti- vated behavior ? Recently, Hiroto and Selig- man (1975) and Miller and Seligman (1975) systematically explored the transfer CONFINED^-* of helplessness between instrumental tasks u 20" ESCAPED = and cognitive tasks. Three groups of col- YOKED* lege students received escapable, inescap- able, or no loud noise. They were then switched to a nonaversive anagrams test, 1 2 3 4 5 and time to solve a series of 20 anagrams like IATOP was recorded. Students who SESSIONS had received inescapable loud noise failed FIGURE 6. Shock-induced fighting frequency, for to find the solutions more often than the each of five test sessions, for rats that had received escape group and the no-noise group. So escapable, yoked inescapable, or no shocks. (In- exposure to uncontrollable aversive events escapable shocks depressed fighting.) (From "In- fluence of Control of Shock on Subsequent Shock- retards solution of cognitive problems in a Elicited Aggression" by Steven F. Maier, Christine nonaversive situation. Anderson, and David A. Lieberman, Journal of Are the debilitating effects of uncontrol- Comparative and Physiological Psychology, 1972, lability only produced by uncontrollable SI, 94-100. Copyright 1972 by the American Psy- chological Association. Reprinted by permission.) trauma? What happens to response initia- tion when preceded by a history of uncon- posure to one kind of uncontrollable event trollable outcomes which are not traumatic? generalized to another—frustration. Hiroto and Seligman (1975) tried to pro- Another possible instance of the transfer duce a learned helplessness effect using un- of a learned helplessness effect is related to solvable discrimination problems. Just as a "shock-elicited aggression.'' If two rats are solvable discrimination problem is control- placed in a small enclosure and repeatedly lable in the same sense that an escapable shocked, the shocks frequently come to shock is controllable, an unsolvable discrimi- elicit attack and aggressive posturing. Maier nation problem is uncontrollable in the same et al., (1972) gave rats escapable shock, sense that an inescapable shock is uncon- inescapable shock, or no shock, and then trollable. The conditional probability of placed these rats in a shock-elicited aggres- success given any response (e.g., to the left) sion situation. As shown in Figure 6, the is the same as the probability of success if amount of shock-elicited aggression was the alternate response (e.g., to the right) is lower in the rats previously exposed to in- made. escapable shock than it was in the rats given With the formal similarity of unsolv- escapable or no shock (also see Payne, ability and inescapability in mind, Hiroto Anderson, & Murcurio, 1970; Powell & and Seligman (1975) and Klein et al. (in Creer, 1969). In a related study (unpub- press) gave three groups of college students lished data) we found that dogs that had four sets of solvable, unsolvable, or no- received inescapable shock as puppies lost discrimination problems. Then all groups out in competition for food (only one nose were given the hand shuttle box with loud fits into a coffee cup full of Alpo) with dogs noise to be escaped. Individuals who had that received no shock or escapable shock. solvable discrimination problems or no prior So it seems that exposure to uncontrollable problems escaped noise readily. The un- aversive events retards the initiation of ag- solvable group failed to escape the noise. So 12 STEVEN F. MAIER AND MARTIN E. P. SELIGMAN we believe that response initiation to con- berg et al. removed the treadle during the trol noxious events may be impaired by ex- autoshaping test, and so it could be argued perience with uncontrollable reward. that the stimulus controlling the competing Equally intriguing is the possibility that behavior was removed for the treadle group. delivery of appetitive events independent of The stimulus controlling any competing be- behavior may produce a failure to learn re- havior was not removed for the noncontin- sponses to procure such events, an effect gent food group, and this difference could analogous to learned helplessness. Engberg, account for the results. In response to Hansen, Welker, and Thomas (1973) these criticisms Welker (1974) conducted trained one group of pigeons to press a a further investigation of the phenomenon. treadle to obtain food. A second group of One group of pigeons was trained to peck pigeons received food delivered indepen- a key for food, a second was given non- dently of behavior, and a third group re- contingent food, and a third was given only ceived only a small amount of magazine a small amount of magazine training. All training. Following this pretreatment, all pigeons were then given a test in which pigeons were tested with a key peck auto- treadle pressing produced food on a con- shaping procedure. In autoshaping (Brown tinuous reinforcement schedule. The key & Jenkins, 1968) the pigeon key is periodi- was present during the treadle test, so it cally illuminated for a brief period of time, could not be argued that the stimulus con- and food is presented at the end of each key trolling the competing behavior was re- illumination period whether or not a peck moved for the contingent food group. The has occurred. Pigeons normally acquire results were the same as in Engberg et al.; key pecking under such an arrangement. noncontingent food presentation retarded ac- The result was that the group given previous quisition. Taking a hint from the helpless- treadle press training was fastest to acquire ness experiments in rats (see pp. 8-9), Wel- key pecking and the group given noncon- ker repeated his experiment with an FR-3 tingent food was the slowest to acquire key treadle response as the test task. That is, pecking. Engberg et al. interpreted their the treadle had to be pressed three times in results as constituting an appetitive analogue order to produce food. This magnified the to learned helplessness; food was delivered size of the observed effect. independently of behavior, and this retarded A potentially related finding was recently the acquisition of a response to procure food. reported by Bainbridge (1973). At SO days However, the Engberg et al. experiment of age, rats were given a solvable black- has been criticized on a number of grounds white discrimination problem, an unsolvable (Gamzu, Williams, & Schwartz, 1973). problem, or no problem. Twenty days later First, Gamzu et al. argued that autoshaping they were tested with the same tasks but is not a reasonable task to use in attempting different stimuli, or in a Hebb-Williams to establish an appetitive analogue to learned maze based on spatial cues. The rats given helplessness because autoshaped key pecks an initial unsolvable problem performed are largely controlled by Pavlovian con- very poorly in both the same and the differ- tingencies (Moore, 1973) rather than instru- ent apparatus. Thus exposure to appetitive mental contingencies; learned helplessness is events delivered independently of behavior a debilitation of instrumental responding. may interfere with the acquisition of ap- Second, Gamzu et al. noted that it is easy petitively motivated responses, just as ex- to account for the Engberg et al. experiment posure to noncontingent shock interferes in terms of competing motor responses. with the acquisition of shock-motivated re- Both the group given treadle training and sponses. the group given noncontingent food would So we believe that the psychological state have been expected to learn a motor re- produced by uncontrollability may under- sponse incompatible with key pecking mine response initiation quite generally. (treadle responding and standing by the Dogs, rats, cats, fish, and people make fewer grain hopper, respectively). However, Eng- responses to escape shock after receiving  LEARNED HELPLESSNESS: THEORY AND EVIDENCE 13 uncontrollable shock. Furthermore, these It appears as if one success is just not motivational deficits may not be limited to enough to make a dog that has experienced shock or even noxious events. Escape from inescapable shock learn that his responding frustration, aggressive behavior, and even now produces shock termination. the propensity to solve anagrams might be Miller and Seligman (1975) and Klein undermined by inescapable aversive events. et al. (in press) found that such a negative Conversely, uncontrollable rewards disrupt cognitive set results from uncontrollability escape from loud noise, learning to procure in man. Three groups of students received food, and competitiveness. However, it escapable, inescapable, or no loud noise. should be stressed that many of the experi- Then they confronted two new tasks, a task ments cited here are subject to alternative of skill and a task of chance. On each of 10 interpretation, and much more research will trials in the skill task, they sorted 15 cards be required before the true extent of these into 10 categories of shape, attempting to effects will be known. Clearly, there must complete it within IS sec. Unknown to be a limit to the degree of generalization of them, the experimenter arranged to have learned helplessness effects, and these limits them succeed or fail on any given trial by must be explored and delineated. We can- saying that time was up before they had not yet say what factors will prove import- finished or after. So they went through a ant. All that can be said now is that there prearranged run of successes and failures. appears to be some degree of transfer. At the end of each trial, the subjects rated. (on a 0-10 scale) what they thought their Cognitive Deficits chances of succeeding on the next trial We have argued that a major consequence would be. Subjects who were previously ex- of experience with uncontrollable events is posed to inescapable noise showed very little motivational; uncontrollable events under- change in their expectancy for successes mine the motivation to initiate voluntary re- after each new success and failure. They sponses to control future events. Experience had difficulty perceiving that their response with uncontrollable events may also have a would affect succeeding or failing. Control second consequence that is cognitive; ex- subjects and subjects who had escaped noise perience with uncontrollability may produce showed large expectancy changes following a difficulty in learning that responses have each.success and failure. This showed that succeeded, even when responding is actually they believed outcomes to be dependent on successful. Uncontrollability may retard the their actions. The three groups did not perception of control. differ in expectancy changes following suc- This phenomenon seems to appear in cess and failure in a "chance" task that they helpless dogs, rats, and men. Occasionally, perceived as a guessing game. So inescap- a naive dog sits and takes shock on the first able loud noise produced a cognitive set in three or four trials in the shuttle box, and which people believed that success and fail- then on the next trial jumps the barrier and ure is independent of their own skilled; escapes shock successfully for the first time. actions within the experimental situation, Once a naive dog makes one response that and they therefore had difficulty perceiving produces shock termination, it learns rapidly. that skilled responses were effective. On all further trials, the subject responds Hiroto and Seligman (1975), Miller and vigorously and goes on to learn to avoid Seligman (Note 1) and Klein et al. (in shock altogether. But dogs that first re- press) also reported this negative cognitive ceived inescapable shock were different in set in another form. After escapable, in- this respect also. About one third of them escapable, or no noise, as the reader will go through a similar pattern—sitting through recall, students had to solve anagrams (see shock on the first three or four trials and p. 11). Two kinds of deficits emerged. then escaping successfully on the next. These Inescapable noise interfered with their dogs, however, then revert to taking the ability to solve any given anagram. In shock, and they fail to escape on future trials. addition, there was a pattern to the 20 ana- 14 STEVEN F. MAIER AND MARTIN E. P. SELIGMAN TABLE 1 Recent investigation has shown that active MEAN NUMBER AND STANDARD DEVIATION or learning does indeed occur when condi- CONSECUTIVE SUCCESSES BEFORE tioned stimuli (CSs) and unconditioned PATTERN SOLUTION stimuli (USs) are independently presented. Noise Mellgren and Ost (1971) reported that a Subject Inescapable Escapable None group for which CSs were presented inde- pendently of food took longer to learn later Depressed that the CSs were associated with the food M 5.9 4.2 7.1 SD 2.53 1.36 1.22 than did naive rats (or even rats for whom Nondepressed the CSs predicted the opposite relationship M 6.9 4.5 3.7 with food). Kemler and Shepp (1971) SD 3.36 1.58.86 showed that irrelevant stimuli were most slowly learned about when they became the Note. Adapted from Miller and Seligman (Note 1), In- escapable noise increased the number of successes required in relevant stimuli for the solution of a discrimi- nondepressed subjects. nation problem for children. Thomas, Free- man, Svinicki, Burr, and Lyons (1970) grams to be solved; each was arranged showed that pigeons that had two colors with its letters in this order: 53124, for presented independently of food tended not example, ISOEN, DERRO, OURPG, etc. to discriminate later between two line tilts, As can be seen in Table 1, students who had one of which predicted the presence of food received inescapable noise required about and the other its absence. MacKintosh seven consecutive successes before catching (1973) also reported retardation of condi- on to the pattern; students who received tioning by prior CS-US independence. escapable noise or no noise needed about So independence between two stimuli three consecutive successes. Unsolvable dis- may produce active learning, and this learn- crimination problem's, incidentally, produced ing seems to retard the ability of rats, pi- the same disruption of anagram solution as geons, and men to learn later that the stimuli did inescapable noise. depend on one another. This evidence co- The existence of a negative cognitive set heres with the effects of response-outcome produced by independence between respond- independence on cognition and bolsters the ing and outcomes bears on an important hypothesis that such independence retards issue in learning theory; When two events the ability of the organism to perceive that are presented independently of each other, his responding has contingent consequences for example, a tone and a shock presented (also see the discussion of experiments by at random, does the subject learn anything Maier & Testa, 1975, pp. 24-25). at all about the tone or does he merely come to ignore it? A helplessness point of view Emotional Deficits holds that men and animals actively learn that responses and outcomes are independent Our first hint that uncontrollable aversive of each other, and one way the learning is events had emotional as well as motivational manifested is by the difficulty they later and cognitive consequences came when we have learning that the response produces the found that the motivational effects dissi- outcome when it does. Merely substitute pated in time under some circumstances. shock for outcome and tone for response. Overmier and Seligman (1967) found that This suggests that organisms should also after one session of inescapable shock in the actively learn when a tone and shock are hammock, dogs were helpless in the shuttle independent and that organisms show this box 24 hours later. If, however, the dogs by having trouble learning later that the tone were not tested in 24 hours, but at 48, 72, is followed by the shock when it actually is. or 168 hours after inescapable shock, they Rescorla (1967) held the contrary view. were normal. This is the only 'circumstance Independence between a tone and shock is a we know under which helplessness dissipates neutral condition in which nothing is learned. in time. If multiple sessions of inescapable  LEARNED HELPLESSNESS: THEORY AND EVIDENCE 15 shock are given to the dog in the hammock, mals. Moreover, the yoked rats lost more he will fail to escape a week later in the weight, defecated more, and drank less than shuttle box (Selignian & Groves, 1970). the executive. So rats given inescapable Further, if the dog is reared from birth in shock show more stress or emotional effects laboratory cages, having minimal experience when measured by stomach ulcers than rats with control over events, small amounts of who can control shock. inescapable shock produce nontransient help- There is further evidence in rats that un- lessness (Seligman & Groves, 1970). Finally, controllable shock produces more severe no time course has been observed in rats; emotional reactions than does controllable following one session of inescapable shock, shock. Mowrer and Viek (1948) shocked rats fail to escape 5 minutes, 1 hour, 24 two groups of rats while the rats were hours, and 168 hours later. It must be eating. One group could control the shock borne in mind however, that under one cir- by jumping in the air, and the other group cumstance—a single session of inescapable received uncontrollable shock. The rats shock in dogs of unknown past history— getting uncontrollable shock subsequently helplessness dissipates in time, and such a ate less than those controlling shock (see time course hints of a transient emotional Brimer & Kamin, 1963; Desiderate & New- disturbance. man, 1971; Lindner, 1968; and Payne, 1972, Other evidence indicates the emotional for an ongoing controversy surrounding these consequences of uncontrollability. One data). In an analogous human study by widely quoted study (Brady, Porter, Con- Hokanson, DeGood, Forrest, and Brittain rad, & Mason, 1958) is related to uncon- (1971), subjects performed a symbol match- trollability and helplessness, but appears to ing task while being shocked. The schedules show less emotionality with uncontrollability. were individually arranged so that each sub- Two groups of four monkeys were given ject received an average of one shock every shocks; half of them—the "executives"— 45 sec. Subjects in the controllability group had control over the shock and could avoid were allowed to take as many time-outs as them by bar pressing. The other four were they wished whenever they wanted. A yoked yoked, or helpless, since they could not control group received the same number of modify shock. The executives formed stom- time-outs at the same times. Measures of ach ulcers and died, but the helpless mon- blood pressure taken at 30-sec intervals in- keys did not. Unfortunately, this result dicated that yoked subjects showed con- may be an artifact of the way the monkeys sistently higher blood pressures. (See Averill were assigned to the two groups. All eight & Rosenn, 1972; Bandler, Madaras, & Bern, monkeys were placed on the executive sched- 1968; Corah & Boffa, 1970; Elliot, 1969; ule originally, and the first four to start and Stotland & Blumenthal, 1964, for re- the lever pressing became the executives. lated studies in humans using a variety of The last four became the yoked subjects. other measures of emotional arousal. This It is possible that the more emotional a mon- is a complex and inconsistent literature and key is, the more quickly it begins to bar is reviewed by Averill, 1973.) press when it is shocked (Sines, Cleeland, & Hearst (1965) found that the presenta- Adkins, 1963). So the four most emotional tion of uncontrollable shocks resulted in a animals may have become the executives, breakdown of a well-trained appetitive dis- and the four least emotional may have be- crimination in rats. During uncontrollable come the yoked subjects. shock his rats no longer discriminated be- Jay Weiss (1968, 1971a, 1971b, 1971c) tween two stimuli, one of which signaled the has recently repeated the executive monkey presence and the other the absence of food. design with rats. (See also Moot, Ceballa, During controllable shock the appetitive & Crabtree, 1970.) Three groups of rats discrimination was maintained. were randomly assigned to the triadic de- Such breakdown of appetitive discrimina- sign. The executive animals showed fewer tions is reminiscent of the classic work on and less severe ulcers than the yoked ani- "experimental neurosis." The concept of 16 STEVEN F. MAIER AND MARTIN E. P. SELIGMAN experimental neurosis is not a homogeneous poorly groomed; b'ehaviorally they performed un- one; neither is it well denned. Unlike the predictably if at all on the right hand lever prob- lems, pausing often for naps and rest. The be- other studies reviewed, controllability has haviors exhibited by these animals were clearly not been manipulated explicitly to produce nonadaptive in nature; for example, two subjects the "neurosis." Yet from an operational spent hours in catching "imaginary" flying insects, point of view, we can speculate that the one subject masturbated almost continuously, three subjects became almost compulsive hair lack or loss of control may be important in pullers, and all tended to show movement stereo- their etiology. Typically, an animal is re- typy alternating with an almost total lack of in- strained in some type of harness seriously terest in their external environment. (Stroebel, limiting what it can do. Often the procedure 1969, p. 97) is Pavlovian, and by definition the organism It is not clear whether there can be any has no control over the onset or the offset one theory which can account for "experi- of the stimuli presented. In Shenger-Krest- mental neuroses," nor is it clear whether all nikova's (in Pavlov, 1927) classic experi- these phenomena are closely related. But ment, an appetitive discrimination deterio- uncontrollability is prominently present, and rated and signs of 'distress were noted when emotional disruption is the frequent result. the dog could no longer tell the difference In summary, three types of disruption between the rewarded and nonrewarded stim- seem to be caused by uncontrollability of ulus. In the work of Liddell, James, and aversive events in the laboratory: The mo- Anderson (1934), sheep developed a range tivation to respond is reduced, the propensity of maladaptive behaviors following uncon- to perceive success is undermined, and emo- trollable electric shock. Masserman (1943) tionality is modified. These effects hold trained monkeys to feed in response to a across a variety of situations and species. We signal and made them "neurotic" by pre- need a theory to put all this together and we senting a fear-arousing stimulus during will now propose one. feeding. Without "therapy" these monkeys remained disturbed almost indefinitely. Ac- LEARNED HELPLESSNESS THEORY cording to Masserman (1943) : Markedly different, however, was the case of ani- The basic facts about learned helplessness mals that had been taught to manipulate various effects in the laboratory have now been ar- devices that actuated the signals and feeder be- rayed before the reader. - What must an cause in this way they could exert at least partial adequate theory of helplessness accomplish? control over their environment. This stood them in good stead even after they were made neurotic It must account for the three facets of the in as much as when their hunger increased they effects of uncontrollable aversive events:.gradually made hesitant, but spontaneous, attempts motivational, cognitive, and emotional. It to reexplore the operation of the switches, signals must be testable; there should be experi- and food boxes, and were bolder and more suc- ments that can be performed which would cessful as food began to reappear, (p. 82) confirm it if it is true, or disconfirm it if it is In a dramatic primate study, Stroebel (1969) false. Finally, it must be ; applicable outside trained a group of rhesus monkeys to air- the laboratory—it must be: useful in explain- condition their overheated chamber by press- ing any helplessness effects found in the real ing a lever, and also to control loud noise, world. This final requirement is beyond the annoying light, and mild shocks by pressing scope of this article, but the interested reader the same lever. He then retracted the lever should see Seligman (1975). so that it could still be seen, but could no The theory to be presented is straight- longer be pressed. No further physical stres- forward, and the groundwork for it has been sors were presented. Initial responding was prepared by the way we laid out the data frantic. Brain temperature became irregular above. It accounts directly for the motiva- and the circadian rhythm was disrupted. tional and cognitive deficits and, with an As rhythm disturbance developed, members of this additional premise, can account for the emo-... group of subjects began to show lassitude and tional disturbance. The j theory has been weakness; their fur became knotted, mottled and tested in several ways, and as a bonus it  LEARNED HELPLESSNESS: THEORY AND EVIDENCE 17 predicts ways to eliminate helplessness effects shock, while the dogs who could escape by and ways to prevent them. panel pressing and the dogs who were not When an organism is faced with an out- shocked did not fail to respond later. Clearly come that is independent of his responses, something different happened to the dogs he sometimes learns that the outcome is in- that received shock independently of their dependent of his responses. responses. We believe they learned that This is the cornerstone of our view and responding was futile and therefore expected probably seems obvious to all but the most future responding to shock to be futile. We sophisticated learning theorist. You will re- have gone to pains to give an objective call our discussion of the response contin- definition of uncontrollabtiity and response- gency space (Figure 1). Learning theorists outcome independence. This is because the would much prefer that the kinds of con- theory is basically a three-step affair. tingencies that.can be learned about be as Information about contingency —> simple as possible. First they believed that Cognitive representation of the contingency the most that could be learned about was a (learning, expectation, perception, belief)—* simple pairing of a response and an outcome, Behavior and the pairing of the response with the ab- sence of the outcome. But this had to be An organism must begin with the informa- broadened to include partial reinforcement, tion about the contingency between response with the subject integrating both kinds of and outcome. This information is a prop- pairings. So what could be learned about erty of the world out there, a set of stimuli, was broadened to the probability of an out- and not a property of the receiver. We have come given a response. Then, it was shown carefully defined the kind of information that that organisms could also learn about the can be called information that a response and probability of an outcome given that it did an outcome are independent. not make the indicated response. The added But the middle step in the chain is crucial step which our view makes is that the organ- and is easily overlooked. The information ism can learn about both these probabilities about the contingency must be processed and conjointly, that variation of experience cor- transformed into a cognitive representation responding to the points in the response of the contingency.3 Such a representation contingency space will produce systematic has been variously called "learning that out- particular, we have argued that exposing come and response are independent," "per- changes in behavior and cognition. In ceiving that response and outcome are in- organisms to the 45° line, in which the prob- dependent," or "believing that response and ability of the outcome is the same whether or outcome are independent." We prefer to not the response of interest occurs, produces call the representation "the expectation that learning. Behaviorally, this learning should responding and an outcome are independent." tend to produce lack of response initiation to A person or animal can be exposed to the control the outcome; cognitively it should environmental contingency in which an out- produce a belief in the inefficacy of respond- come response are independent, yet not ing and difficulty in learning that responding form such an expectation. Immunization, succeeds: and emotionally when the outcome as we shall see later in this article ';(refer to is traumatic it might produce emotional p. 25), is an example. Conversely, a per- changes. The basic triadic design employed in all 3 For attempts to spell out in detail the relation- the helplessness studies reviewed above is, of ship between the contingency information and its course, directly relevant to the premise that cognitive representation, the interested reader should men and animals learn about, and form ex- consult, for example, Kelly (1967, 1972), Weiner, pectations concerning, independence between Frieze, Kukla, Reed, Rest, and Rosenbaum (1971) outcome and responses. So, for example, from an attribution theory point of view; Irwin (1971) and Seligman and Johnston (1973) from in the Seligman and Maier (1967) study a cognitive-learning theory point of view; as well only the yoked dogs later failed to escape as Lazarus (1966) and Stotland (1969). 18 STEVEN F. MAIER AND MARTIN E. P. SELIGMAN son can show a helplessness effect without We assume that the incentive to initiate being exposed to the contingency as such; voluntary responses in a traumatic situa- he can merely be told that events are uncon- tion is partly produced by the expectation trollable. that responding produces relief. In the ab- Glass and Singer (1972) have done an sence of this incentive, voluntary responding extensive series of studies on the role of will decrease in likelihood. When a person controllability in reducing stress and have or animal has learned that relief is indepen- found that merely telling a human subject dent of responding, the expectation that re- about controllability duplicates the effects of sponding produces relief is negated, and actual controllability. They attempted to therefore response initiation is reduced. mimic the stress of the urban environment Most generally put, the incentive to initiate by having their subjects (college students) voluntary responses to control any outcome listen to a very loud melange of sound: two (e.g., food, sex, shock termination) comes, people speaking Spanish, one person speak- in part, from the expectation that responding ing Armenian, a mimeograph machine, a produces that outcome. (See Bolles, 1972, calculator, and a typewriter. When sub- for a discussion of expectational mechanisms jects could actually turn off the noise by and incentive motivation.)* In the absence button pushing, they were more presistent of this incentive, voluntary responding will in problem solving, they found the noise decrease in likelihood. When a person or less irritating, and they did better at proof- animal has learned the outcome is indepen- reading than yoked subjects. dent of responding, the expectation that re- Glass and Singer also presented another sponding will produce the outcome, and group of subjects with the same noise, but therefore response initiation, decreases. this time it was uncontrollable. However, The way in which this undermining of this group had a panic button and was told: motivation works has been seen with crystal "You can terminate the noise by pressing clarity in a human helplessness experiment the button. But we'd prefer you not do it." (Thornton & Jacobs, 1971). Following in- None of the subjects in fact tried to turn escapable shock, college students sat and off the noise. All they had was the false took the shock. When asked why they did information that they could control the noise not respond appropriately, 60% of the sub- if they had to. These subjects did just as jects reported they had no control over well as the subjects who actually controlled shocks, so why try. These subjective re- the noise. Actual controllability and actual ports suggest that a belief in uncontrollability uncontrollability may often produce cor- undermines the incentive to initiate re- responding expectations. But this type of sponses. experiment, in which the expectation is in- We also assume that learning that an out- valid, highlights the fact that it is the ex- come is independent of a response makes it pectation and not the objective conditions of more difficult later to learn that responses controllability that is the cornerstone of our produce that outcome. theory. (See also Geer, Davison, & Gatchel, Response^outcome independence is an ac- 1970; Geer & Maisel, 1972; Klein et al., in tive form of learning, and like any other press; Langer, 1974; Pervin, 1963; and active form of learning, it can proactively Stotland & Blumenthal, 1964, for related interfere with contravening forms of learn- effects of perceived control.) ing through an associative interference. At So the first step of the theory is that the organism acquires an expectation of re- sponse-outcome independence, when out- *It should also be mentioned that innately comes are uncontrollable. The second step elicited struggling is another source of responding in the theory is the means by which the ex- in a traumatic situation, but it is the waxing and pectation of response-outcome independence waning of voluntary (outcome-sensitive) responses which is our concern here. This does not deny produces the effects associated with helpless- that innate responses can be transformed into ness. voluntary responses (Schwartz & Williams, 1972).  LEARNED HELPLESSNESS: THEORY AND EVIDENCE 19 first the dog is not a passive recipient of some theoretical positions which have been shock but emits many responses, but each of advanced as alternatives to the helplessness the responses is unrelated to shock termina- hypothesis. It should be noted at the outset tion. The dog might, for example, turn his that these hypotheses have been offered as head and shock might happen to go off on explanations of only the fact that organisms the trial, but on other trials he might turn exposed to inescapable shock later fail to his head and shock would not go off. Shock learn to escape and avoid shock in a shuttle would also terminate when he had not turned box, rather than as explanations of the his head. Now when he is tested in the broad range of phenomena considered above. shuttle box and jumps the barrier, which in Thus we will describe these theoretical posi- reality causes shock termination, the dog has tions in the context of the basic learned trouble learning this. This is because he has helplessness effect. These alternatiye the- had many experiences in which a pairing be- ories can be divided into two groups, motiva- tween a response and shock termination tional accounts and motor accounts. proved spurious and, as for head turning, he still expects that shock will just as likely go Motivational Alternatives off if he fails to jump the barrier. Such a dog may return to taking shock even after Adaptation. The adaptation hypothesis he makes one or two "successful" jumps. maintains that a subject adapts to shock In contrast, a naive dog has no interfering during pretreatment with inescapable shocks expectation that shock termination is inde- and is therefore not sufficiently motivated to pendent of responding, so one experience escape from shock in the shuttle box. The with barrier jumping leading to shock term- hypothesis is inadequate: ination is sufficient for him to learn. For some evidence regarding such associative 1. Adaptation to repeated, intense shock interference see the discussion of Maier and has never been demonstrated (Church Lo- Testa (1975), pp. 24-25. Lordo, Overmier, Solomon, & Turner, This, then, is the theory of helplessness. 1966). The expectation that an outcome is inde- 2. It is unlikely that very much adapta- pendent of responding (a) reduces the mo- tion could persist for as long as 24 hours tivation to control that outcome and (b) and beyond. interferes with learning that responding con- 3. In experiments with dogs, the dogs do trols the outcome. not look as though they are adapted; they Seligman (1975) has recently extended howl, defecate, and urinate to the first this theory to changes in emotionality. He shock presentation in the shuttle box. On argued that when a traumatic event first later trials, the dogs are passive; but they occurs it causes a heightened state of emo- whimper and jerk with the shock. tionality, which has been called "fear." The 4. We have disconfirmed the adaptation fear continues until the subject learns that hypothesis experimentally. Raising the shock he can or cannot control the trauma. If the level in the shuttle box should increase subject learns he can control the trauma, motivation to escape. However, Overmier fear is reduced and may disappear altogether. and Seligman (1967) found that increasing If the subject learns that he cannot control the shock level from 4.5 mA to 6.5 mA does the traumatic event, fear decreases and is not eliminate the interference effect. replaced with depression. For a more de- 5. A series of escapable shocks in the tailed discussion see Seligman (1975). hammock does not produce failure to escape in the shuttle box (Seligman & Maier, ALTERNATIVE EXPLANATIONS 1967), while the same shocks, if inescapable, Now that we have described the learned do produce failure to escape. By this hy- helplessness hypothesis and the manner in pothesis, both conditions should lead to which it is able to account for the phe- equal adaptation to shock and to similar be- nomena under consideration, we will describe havior in the test situation, but they do not. 20 STEVEN F. MAIER AND MARTIN E. P. SELIGMAN 6. Seligman and Maier (1967) gave one spond on his own. Then the barrier was re- group of dogs 10 trjals of escape/avoidance placed, and the subject continued to escape training in a shuttle box before treatment and avoid. The recovery from helplessness with inescapable shock in the harness. Fol- was complete and lasting. This type of ex- lowing exposure to inescapable shocks in the posure to the escape and avoidance contin- harness, the dogs escaped and avoided shock gencies should not affect any adaptation to normally when returned to the shuttle box. shock that might have been present. Selig- Thus prior exposure to controllable shock man et al. (1975) replicated these findings "immunized" the dogs against the interfer- in rats. ing effects of exposure to inescapable shocks. This result follows from the helplessness hy- Sensitization. Perhaps the inescapable pothesis because prior experience with con- shocks received in the harness sensitize the trollable shock should proactively interfere subject to shock so that it is too motivated with the subject's learning that shock is un- to enable it to make organized responses in controllable and should also allow the sub- the shuttle box. This hypothesis is inade- ject to discriminate between the places where quate, (a) Sensitization explains the in- shocks are controllable and uncontrollable. efficiency of responding but not the absence However, this result is not consistent with of responding, (b) Lowering the shock the adaptation hypothesis. Prior exposure level in the shuttle box should permit the to escapable shock in the shuttle box should subject to make organized responses. How- not eliminate any adaptation produced by ever, Overmier and Seligman (unpublished inescapable shocks in the harness. data) found that the interference effect is 7. If the subject is tested and fails to not attenuated when shock in the shuttle box escape 24 hours after inescapable shocks, is reduced to 3.0 mA. (3) Arguments 5, the learned helplessness effect will persist in 6, and 7 in the previous paragraph, which chronic form; the subject will fail to escape invalidate the adaptation hypothesis, also in- on later opportunities. The helplessness validate the sensitization argument. hypothesis suggests a way to eliminate chro- nic failure to escape. By this hypothesis, Motor Activity Alternatives the dog does not try to escape because Within the last 2 years three hypotheses he does not expect that any instru- have been offered as explanations of the mental response will produce shock termina- learned helplessness effect which are more tion. By forcibly exposing the dog to the difficult to dismiss than are sensitization and escape and avoidance contingencies, this ex- adaptation. These explanations have the pectation might be altered. This type of advantage that they are able to account for training by "putting through" has been used the fact that the learned helplessness effect by others with mixed success (Loucks, is determined by the controllability of the 1935; Miller & Konorski, 1928; Woodbury, shock during the initial treatment. We have 1942). Seligman et al. (1968) reasoned classified them as motor activity hypotheses that forcibly dragging the dog from side to because they all maintain that exposure to side in the shuttle box, in such a way that inescapable shock interferes with subsequent the dog's changing compartments terminated shuttle box acquisition because it changes shock, might effectively expose the dog to motor activity. Two of these hypotheses the response-reinforcement contingency. This argue that exposure to inescapable shock was the case. The experimenter pulled three establishes a motor response that is incom- chronically helpless dogs back and forth patible with shuttling and therefore competes across the shuttle box with long leashes. with shuttling. Thus these two hypotheses This was done during CS and shock, while argue that the organism does not learn shut- the barrier was absent. After being pulled tling because it is performing a. motor re- across the center of the shuttle box (and thus sponse that prevents the occurrence of shut- terminating shock and CS) 20, 35, and 50 tling. The remaining hypothesis maintains times, respectively, each dog began to re- that inescapable shock is a powerful stressor  LEARNED HELPLESSNESS: THEORY AND EVIDENCE 21 and depletes a neurochemical necessary for this hypothesis can explain why escapable the occurrence of movement. Thus this view shock does not produce a subsequent decre- holds that inescapably shocked animals fail ment. If shock is escapable, the organism to acquire shuttling because they cannot learns a response that completely eliminates : move sufficiently. shock (the escape response) and so does Incompatible motor response theories. As not have to learn a response that only miti- already noted, two different incompatible gates shock (the putative incompatible re- motor response explanations of the learned sponse). It need only be argued that the helplessness effect have been proposed. The former is not incompatible with shuttling but first to be described was offered by Brace- the latter is. well and Black (1974) and was suggested We will present a large amount of data by three experiments. In their first experi- which we feel to be inconsistent with the ment they examined the effects of restraint Bracewell and Black (1974) hypothesis. and inescapable shock upon simple FR-1 However, these experiments also bear on the shuttlebox escape/avoidance acquisition. second competing motor response theory and They found that restraint in the absence of so they will not be presented until that theory any exposure to inescapable shock produced has been described. Nevertheless, a few a decrement in FR-1 shuttlebox acquisition, comments are appropriate here. First, the but that exposure to inescapable shock in the fact that explicit punishment of movement absence of restraint did not produce a reli- produces a decrement in subsequent escape/ able decrement. In their second experiment avodiance acquisition does not imply that Bracewell and Black found that exposure to other procedures which produced escape/ "high" intensity (1.0 mA) inescapable shock avoidance decrements do so because they produces a small decrement in FR-1 shuttle punish movement. There are undoubtedly box acquisition even when delivered in the many ways to produce poor escape acquisi- unrestrained condition. In their final ex- tion, and it is unlikely that all operate periment Bracewell and Black explicitly through a similar mechanism. Paralyzing punished movement by arranging a positive the animal with a drug would retard escape correlation between movement and shock acquisition, but this does not mean that the intensity. This procedure led to a decre- debilitating effect of inescapable shocks oc- ment in subsequent FR-1 shuttle box ac- curs because inescapable shock induces pa- quisition. ralysis. Thus the retarding effect of move- Bracewell and Black (1974) proposed the ment punishment does not imply that in- following interpretation of their data: escapable shock retards acquisition because One explanation that accounts for these results is it punishes movement. Moreover, the ex- the operant conditioning of responses incompatible plicit punishment of movement in the Brace- with the shuttle box escape response that occurred well and Black experiment led to a much in the third case [where movement was explicitly smaller escape acquisition decrement than punished], also occurred in the first two. That did inescapable shock delivered to restrained is, restrained rats may have been punished for struggling and consequently, learned to hold still, subjects. The punishment procedure in- and unrestrained rats may have learned to hold creased the subsequent mean escape latency still or to perform some other incompatible re- across the 10 test trials from the control sponse to reduce the high intensity shock, (p. 67) level of 2.4 sec to 9.1 sec. The inescapable Thus Bracewell and Black (1974) argued shock procedure resulted in a 21.5-sec mean that because explicit punishment of move- escape latency. So it can be argued that the ment produced a retardation in shuttle box explicit training of an incompatible response acquisition, the other instances of retarded did not duplicate the effects of exposure to acquisition might also be due to punishment inescapable shock in restrained subjects, and of movement during the pretreatment phase. so the argument that the movement punish- They went on to argue that the learned ment effect accounts for the inescapable shock helplessness effect might be explainable in effect seems questionable. It can, of course, the same terms. It should be noted that be argued that a more effective or different 22 STEVEN F. MAIER AND MARTIN E. P. SELIGMAN punishment procedure might yield a larger et al. (1975) reported large escape deficits escape decrement, but that remains to be produced by exposure tO; inescapable shock demonstrated. in unrestrained rats. Again there are many Further, it is not clear that Bracewell and differences between studies, but it is clear Black (1974) employed a test procedure that that restraint does not always produce a is generally sensitive to learned helplessness subsequent escape decrement nor does in- effects* In order for an effect to be called a escapable shock have to be delivered to re- learned helplessness effect, it must be demon- strained subjects for a large effect to occur. strated that the effect is caused by the con- A second incompatible motor response trollability of the events delivered during theory has been proposed by Anisman and pretreatment. There is no evidence that Waller (1973). They argued that exposure decrements in FR-1 shuttle box acquisition to shock induces response repertoire changes produced by prior exposure to shock are in the organism, and if no, coping response is caused by the inescapability of the shocks in available, "freezing" becomes the organism's rats. There is not a single study employing dominant reaction to shock. Freezing would, a triadic design in rats (see p. 6) that has of course, compete with ; the occurrence of found appropriate effects with FR-1 shuttling active responding in a subsequent escape/ as the dependent variable. In fact, there avoidance task. As evidence for their hy- have been both reported (Anderson, Schwen- pothesis, Anisman and Waller cite a variety diman & Payne, Note 3; Maier et al., of studies which show that manipulations 1973; Seligman & Beagley, 1975) and un- designed to increase freezing have a detri- reported (Carder, Leaf, Note 4) failures to mental effect on avoidance behavior, and find any effect of prior exposure to shock on manipulations designed to reduce freezing FR-1.shuttle box escape acquisition in the facilitate avoidance performance. So, for rat. It is to be noted that both Anderson example, strong shock, which causes more et al. and Maier et al. (1973) delievered in- freezing in rats, produced more interference escapable shock to restrained rats and still with shuttle avoidance than weak shock failed to observe any effects. There are a (Anisman & Waller, 1972) and scopalamine, number of differences between the various a drug that reduces freezing, makes rats FR-1 shuttle box experiments, and it is not better avoiders (Anisman, 1973). known which are responsible for the differ- A number of comments are appropriate. ence in results. Nevertheless, the dependent First, the experiments discussed by Anisman variable used by Bracewell and Black may and Waller are not directly relevant to the not be one that is sensitive to helplessness learned helplessness effect. They are all effects in rats and so the implications of studies of effects on avoidance, not escape. their results for learned helplessness is dif- There is no evidence that the manipulations ficult to assess. described by Anisman and Waller affect A further problem concerns the generality escape acquisition. In fact, escape latencies of the effects found by Bracewell and Black are not even reported in many of the experi- (1974). It will be recalled that they found ments (e.g., Anisman, 1973). The learned restraint in the absence of shock to produce helplessness effect is a failure to escape—to a subsequent escape deficit and inescapable flee from shock itself—a more profound shock in the absence of restraint to produce debilitation than failure to prevent shock. only a small escape decrement. However, Second, even if it could be shown that in- Maier et al. (1973) used an escape/avoid- ducing freezing debilitates escape, this does ance task known to be sensitive to helpless- not mean that inescapable shock interferes ness Affects in rats (see p. 8) and found with escape by inducing freezing. In addi- no effect of restraint alone. Cohen (1970) tion, dogs do not seem to freeze, people who found no effect of restraint on shuttle box receive unsolvable discrimination problems escape/avoidance acquisition in dogs. or inescapable noise do not freeze, and organ- Further, Looney and Cohen (1972), Selig- isms that receive noncontingent food do not man and Beagley (1975), and Seligman freeze, yet these conditions may all produce  LEARNED HELPLESSNESS: THEORY AND EVIDENCE 23 helplessness. Finally, why does inescapable exposed to inescapable shocks did not learn shock, but not escapable shock, produce the FR-2 escape response even though they freezing (if it does) in rats? We have not responded rapidly on the five FR-1 trials, been able to think of an answer that does not while all control rats learned the FR-2 imply that the rat has learned that shock is escape response. It should be noted that we inescapable, and this is the heart of the have replicated this effect several times, as helplessness view. have other laboratories (Seligman & Experimental evidence. We feel that there Beagley, 1975; Weisman, Note 5)i are six lines of evidence that bear on the in- The fact that rats exposed to inescapable compatible motor response theories, and shock will learn FR-1 shuttling but will not each will be described. learn FR-2 shuttling might be explained in a variety of ways. The helplessness hy- 1. It has already been noted that rats ex- pothesis asserts that prior exposure to in- posed to inescapable shock often do not fail escapable shock interferes with the subse- to learn an FR-1 shuttle box response. This quent formation of associations between has been investigated by Maier et al. (1973), responding and shock termination and, in and in agreement with a number of previous addition, reduces the organism's incentive to investigators they found that rats exposed attempt to escape. From the foregoing dis- to inescapable shock under a wide range of cussion it is clear that crossing the shuttle parameters subsequently responded as box once is a high probability ini