Journal oj Personality and Social Psychology1975, Vol. 31, No. 4, 674-685

The Role of Expectations and Attributions in the Alleviationof Learned Helplessness

Carol S. DweckUniversity of Illinois at Urbana-Champaign

The purpose of the investigation was to determine whether altering attribu-tions for failure would enable learned helpless children to deal more effectivelywith failure in an experimental problem-solving situation. Twelve childrenwith extreme reactions to failure were identified and were given intensive,relatively long-term experience with one of two training procedures. It washypothesized that a procedure which taught the helpless children to take re-sponsibility for failure and to attribute it to lack of effort would result inunimpaired performance following failure in the criterion situation, but that aprocedure which provided success experiences only (as in many programmedlearning and behavior modification programs) would lead to changes of alesser magnitude. The results revealed that following training, the subjects inthe Success Only Treatment continued to evidence a severe deterioration inperformance after failure, while subjects in the Attribution Retraining Treat-ment maintained or improved their performance. In addition, the subjects inthe latter condition showed an increase in the degree to which they empha-sized insufficient motivation versus ability as a determinant of failure.

The present study of the alleviation oflearned helplessness in children has grown outof several relatively independent lines of in-vestigation: contingency learning in experi-mental animal psychology, attribution theoryin personality-social psychology, and cogni-tive therapy in clinical psychology.

The phenomenon of learned helplessness,conceptually related to the earlier notion of"hopelessness" advanced by Mowrer (1960),was first studied systematically by Seligmanand Maier (1967). The latter found thatanimals pretreated with unavoidable andinescapable shock later failed to avoid andescape traumatic shock in another situationin which shock was avoidable and escapableby performing a simple response. In describ-ing the phenomenon, the investigators used

This study is based on a doctoral dissertation sub-mitted to the Graduate School of Yale University(1972). The author would like to thank the membersof the dissertation committee (Robert Abelson, DavidFeldman, Arnold Lazarus, Dick Reppucci, and FredSheffield) as well as Ellen Langer and Steven Reiss,for their useful suggestions. Thanks are also ex-tended to the administrators and teachers of theAnsonia School System for their support and coop-eration.

Requests for reprints should be addressed toCarol S. Dweck, Department of Psychology, Uni-versity of Illinois, Champaign, Illinois 61820.

the term "learned helplessness" to refer tothe learning or perception of independencebetween one's behavior and the presentationand/or withdrawal of aversive events.

In a study linking this phenomenon to re-inforcement responsibility (i.e., internal vs.external attributions for outcomes), Dweckand Reppucci (1973) demonstrated that fol-lowing failure, a certain group of children donot perform the response required to succeed,even though they are motivated to and arefully capable of doing so. An analysis of thereinforcement responsibility profiles of thesubjects revealed that those children mostlikely to give up in the face of failure whencompared to the more persevering subjects(a) took less personal responsibility for thesuccesses and failures they met with and (b)to the extent that they did take responsibility,tended to attribute the outcomes of their be-havior to ability rather than to effort. Overall,the more helpless subjects placed significantlyless emphasis on the amount of effort exertedas a determinant of success and failure thandid the more persevering subjects. Theseresults imply that since helpless children seethemselves as less instrumental in determin-ing outcomes, they would be less likely toview adverse circumstances as surmountable;

674

ALLEVIATION OF LEARNED HELPLESSNESS 675

since they tend to attribute failure to lack ofability, they would be less likely to respond tofailure with increased effort or perseverance.1

A number of studies investigating the relation-ship between reinforcement responsibility andachievement motivation (e.g., Rotter, 1966;Weiner & Kukla, 1970; Odell, Note 1) orfacilitating anxiety (Butterfield, 1964) haveyielded results consistent with these findings.

In learned helplessness the important varia-ble is not the occurrence of the aversive event,but the perception of the relationship betweenone's behavior and the occurrence of thatevent. As shown in the Dweck and Reppuccistudy, two children may receive exactly thesame number and sequence of success andfailure trials yet react quite differently as afunction of whether they interpret the failureto mean that the situation is beyond or withintheir control. Many recent studies have shownthat an individual's reaction to events islargely determined by his attribution of thecauses of that event. Nisbett and Schachter(1966), Ross, Rodin, and Zimbardo (1969),Storms and Nisbett (1970), and Valins andNisbett (1971) have discussed the possibilityof "attribution therapy," a procedure wherebythe teaching of a new attribution for certainsymptoms might lead to the lessening of thedebilitating or undesirable effects of thosesymptoms. Moreover, several clinicians (Da-vison, 1966; Ellis, 1962; Meichenbaum, Note1) have reported favorable results with pa-tients treated by "cognitive restructuring,""rational-emotive therapy," or by modifica-tion of "what they say to themselves aboutevents."

1 It is interesting to note that these patterns ofreinforcement responsibility need not correspond todifferences in actual ability. In the Dweck and Rep-pucci (1973) study those children who gave up inthe face of failure performed just as well on thetask presented as those who persisted. In fact, forthe female subjects, the helpless children showedsuperior initial performance. It should also be notedthat an attribution of failure to lack of ability hasdifferent implications depending on whether thatlack is perceived as a temporary deficit in obtainableskills or as a more general intellectual deficit (al-though both might indicate that persistence in theimmediate situation would be fruitless). Helplesschildren tend to key on the latter as an explanationfor failure, but it might prove beneficial to encour-age a distinction between the two.

Given the fact that the performance of thehelpless child tends to deteriorate in the faceof failure, the purpose of the present studywas to determine whether altering the help-less child's perception of the relationshipbetween his behavior and the occurrence offailure (i.e., his attributions for failure)would result in a change in his maladaptiveresponse to failure in the experimental situ-ation. If the helpless child tends to perceiveindependence between his responses and fail-ure, then one step would involve demonstrat-ing the dependence of failure upon his re-sponses and another would involve explainingto which aspects of the situation failureshould be attributed.

It was therefore hypothesized that a long-term training procedure which taught thechild (a) to take responsibility for failure and(b) to attribute it to insufficient effort wouldlead to increased persistence on the experi-mental task in the face of failure. For thosechildren receiving this retraining of attribu-tions, it was expected that failure on the taskwould no longer serve as a cue for continuedfailure (but as a cue to do something differentor something additional) and that it wouldtherefore no longer have deleterious effects onthe child's performance.

Since the literature suggests that higherexpectations of success lead to greater persis-tence in the face of difficulty (Battle, 1965;Feather, 1966; Tombaugh & Tombaugh,1963; Tyler, 19S8), it was hypothesized thata procedure which provided only success ex-periences would lead to some positive changesin the helpless child's reactions to failure.However, to the extent that failure retainedits meaning as a cue for continued failure,the subjects in this group were not expectedto show as much improvement as those sub-jects who were given attribution retraining.A success only procedure is one recommendedby many behavior modifiers (e.g., Bigelow,1972; Hart & Risley, 1968; Meacham &Wiesen, 1969) and advocates of programmedlearning (e.g., Holland, 1960; Skinner,1968), especially for children who have diffi-culty dealing with failure. These investigatorsrecommend the elimination of errors in thebelief that errors serve only to elicit negative

676 CAROL S. DWECK

emotions on the part of the child and to makethe materials and/or the situation aversive,while they do little or nothing to enhancelearning. Although there is ample evidencethat errors per se do indeed have adverseeffects on the performance of these childrenand that success works to motivate them, thequestion is whether the most effective way ofovercoming such reactions to failure is toeliminate it from the situation or to teach thechild how to deal with it.

In summary, the purpose of the study wasto determine whether a treatment that alteredattributions for failure would alter responsesto failure to a greater degree than a com-monly advocated procedure that did not alterattributions. The Success Only Treatmentwas expected to raise the child's expectationof success, thereby enabling him to sustainhis performance despite failure. It was ex-pected, however, that the Attribution Re-training Treatment would prove superior sinceit provided a new interpretation for failure byteaching the child to attribute it to insufficienteffort.

METHODOverview

Twelve extremely helpless children were identifiedand compared to persistent children, of the sameage and sex and in the same classroom, on severalmeasures. Baselines for speed and accuracy of per-formance on math problems were obtained for eachof the helpless subjects and the effects of failure onperformance following failure were assessed. Follow-ing this initial baseline and interpolation of failure,the helpless subjects were given one of two treat-ments in a different situation with different problems

2 While the simultaneous manipulation of thesetwo variables is entirely compatible with the purposeof the study (i.e., to assess the effects of a pro-cedure that altered attributions vs. a commonly advo-cated procedure that did not), the differences be-tween the two treatments should be kept in mind incomparing their effects.

3 Three other children, originally selected as help-less, were not used. Two of them did not performconsistently enough to establish stable baselinesagainst which the effect of failure on thair perform-ance could be assessed. One was eliminated becausehe did not meet the criteria for helplessness on thepremcasures and interpolated failure; it appearedthat he did not perform well in school because therewards of the classroom meant little to him andnot because he found the threat of failure, or failureitself, debilitating.

for 2S sessions. In the Success Only Treatment, trialswere programmed such that the subject could suc-cessfully complete the number of problems requiredwell within the time limit on every trial. In thistreatment, while success was attributed to subject'sresponses, failure was avoided or glossed over whenit occurred. In the Attribution Retraining Treat-ment, the procedure differed in two respects: theprogrammed occurrence of failure and the attributionof failure.2 The success trials were identical to thoseof the Success Only Treatment. However, on ap-proximately 20% of the trials, the requirementslightly exceeded the number the subject was cur-rently able to solve within the time limit. On thesetrials, failure was clearly attributed to insufficienteffort (i.e., experimenter verbalized the attributionto subject). The effects of failure on performancein the original situation were again assessed at themiddle and end of training.

SubjectsThere were 12 subjects,8 5 females and 7 males,

between the ages of 8 and 13 years. The femalesranged in age from 10 to 13 years (10, 11, 11, 11,and 13); 3 were white and 2 were black. The malesranged in age from 8 to 13 years (8, 10, 10, 11, 12,12, and 13); 4 were white and 3 were black. Twoof the females and 6 of the males were in specialclasses for children performing below grade level,while the remaining children were in regular classes.The subjects attended two public elementary schoolsin a lower-middle-class town outside of New Haven,Connecticut.

The subjects were identified as "helpless" (i.e.,characterized by expectation of failure and deteri-oration of performance in the face of failure) bytheir school psychologist, their principal, and theirclassroom teacher, independently. They were chosenas the most extreme cases out of the 750 children inthe two schools. To ensure that the teachers' criteriamatched those of the investigator, teachers weregiven a helplessness rating scale for each of thechildren designated. The scale described variousreactions to a stressful academic situation in whichfailure was possible or present. The teacher wasasked to rate, on a scale of 1 to S, the degree towhich each child evidenced these reactions in similarsituations.

In an attempt to avoid any possible stigma asso-ciated with being singled out for participation inthe study, subjects were told that the investigatorswere studying how children do math problems andthat they had been specially selected to help out onthis project.

To determine how and how much the helplesschildren differed from persistent children, 10 per-sistent comparison subjects of the same age andsex, and from the same classrooms as the helplesssubjects were selected. Teachers were asked to choosechildren they believed to be of equal-ability to thehelpless children but who were characterized bytheir persistence in the face of failure or difficulty.Teacher selection, rather than IQ tests, was used as

ALLEVIATION OP LEARNED HELPLESSNESS 677

a means ol matching subjects on intelligence, sincefor various reasons (e.g., the fact that IQ testsinevitably involve failure at some point, the factthat the groups may differ in test anxiety, etc.) IQtests do not accurately reflect the abilities of thehelpless child. These comparison subjects were giventhree of the same measures that were administeredto the helpless children (but not the training ortesting procedures). Because of the small size of theclasses from which the subjects were drawn, one ofthese persistent subjects served as the comparisonfor three helpless subjects.

Experimenters

The two experimenters were women, 24 years ofage, who had no knowledge of the experimentalhypothesis.4 Lest they infer that the subjects receiv-ing one treatment were "supposed" to performbetter, or that one treatment was thought to besuperior to the other, they were told that it was nota matter of "better-poorer" but that a different pat-tern of results might follow from each treatment.The two experimenters administered all the measuresthat were given prior to training as well as all sub-sequent training and testing, with each experimenterworking with the same subjects every day. The ex-perimenters were given extensive training in the pro-cedure by the author and their performances weremonitored daily.

Questionnaire measures given after the completionof training were administered by two teacher's aides,employed by the schools, who were trained in theadministration of these measures by the author.This was done to ensure that possible changes inthese measures from before to after training werenot due to familiarity with the experimenter.

Comparison Measures

Measures on which the helpless subjects were com-pared to their persistent classmates consisted of theIntellectual Achievement Responsibility Scale, twosubscales of the Test Anxiety Scale for Children,and a Repetition Choice task. They were all indi-vidually and orally administered.

The Intellectual Achievement Responsibility (IAR)Scale (Crandall, Katkovsky, & Crandall, 196S) wasgiven to all the helpless and persistent subjects. TheIAR is designed to determine the degree to which achild believes that the intellectual failures and suc-cesses he encounters are a result of his own be-havior versus the behavior of important others inhis environment (e.g., teachers, parents, friends). Itconsists of 34 forced-choice items, each depicting apositive or negative achievement situation and pre-senting two alternative attributions: (a) an internalattribution, in which responsibility for the outcomeis assumed by subject, and (b) an external attribu-

4 The author wishes to thank Karen Gould andSuella Wallace for their extremely competent workas experimenters.

tion, in which responsibility for the outcome is rele-gated to some property of the situation or the otherperson. The 1+ score represents the number ofpositive items for which subject takes responsibility;the I— score represents the number of negative itemsfor which he takes responsibility; and the total Iscore represents the sum of the subscores. The in-ternal alternatives can be further categorized intothose which attribute the outcome to the ability ofthe subject versus those which attribute the outcometo his motivation. Thus the 1+ score can be sub-divided into I+B (effort) and I+A (ability); the I-score can be subdivided into I—B (effort) and I—A(ability).

Two subscales of the Test Anxiety Scale for Chil-dren (TASC) (Sarason, Davidson, Lighthall, Waite,& Ruebush, 1960) were administered. The scale hasbeen factor analyzed and divided into four subscalesby Feld and Lewis (1969), of which the test anxietysubscale and the poor self-evaluation subscale wereused in the present study. Since the TASC has beenused as an indirect measure of the motive to avoidfailure (e.g., Atkinson & Feather, 1966), it wasthought to be a measure which might distinguishbetween the helpless and persistent subjects.

Finally, a repetition-choice task, similar to that ofBialer and Cromwell (1960) was administered to allsubjects to test for the relative strength of thetendency to avoid failure versus the tendency tostrive for success. Two 24-piece jigsaw puzzles, Hiin. X 14i in. (.29 X .37 m), were used. They wereof approximately the same difficulty and interestvalue and were cut from the same die. All subjectswere allowed to complete the first puzzle, but werestopped 9 pieces short of completion on the second.They were then given the choice of reconstructingeither puzzle at their leisure. The decision to recon-struct the already successfully completed one wastaken as an indication of the tendency to avoidfailure, while the decision to reconstruct the failedone was taken as an indication of the tendency tostrive for success.

In addition to these measures, an Effort versusAbility Failure Attribution Scale was developed andadministered to the helpless subjects only. Since theIAR fails to give the child a choice between twointernal alternatives, that is, one attributing failure toinsufficient effort and the other attributing failure toinsufficient ability, it is difficult to assess the relativeemphasis which an individual child places upon thesetwo factors. Therefore, the present investigator de-vised a simple S-question scale which presents Sfailure situations involving arithmetic and asks sub-ject to choose between the ability and effort al-ternatives.

All of the measures above were readministered tothe helpless subjects following the completion oftraining, largely for exploratory purposes. The Effortversus Ability Failure Attribution Scale was ex-pected to yield the greatest changes as a result ofattribution retraining, since it was most closelyrelated to what this procedure endeavored to teachthese children.

678 CAROL S. DWECK.

Establishment oj Baselines and Interpolationof Failure

After the administration of the premeasures andprior to the experimental treatment, the experimentalsubjects were given sheets of arithmetic problems todo. They were given three or four sheets per sessioneach containing 25 or 30 problems. The sheets wereadapted from the children's workbooks and werechosen for each subject to reflect the range of hisarithmetic abilities from moderately easy to mod-erately difficult. The problems were done in groupsof five and a star was placed after every fifth prob-lem so that the subject would know he had reachedthe end of the group. After every group of 5 prob-lems, the experimenter recorded the time and gradedthem. If the subject got at least 4 of the 5 correcthe was given a token. At the end of the session hewas permitted to trade in his tokens for a smallprize or save them for a larger one.

The subject was given the same sheets every day,but in a different order for 10 days. Then to testfor the effects of failure on rate and accuracy ofperformance, special sheets were made. They wereidentical to the moderately difficult arithmetic sheetson which the subjects had been working, except thatin the second and fourth group of five problems,two problems were replaced with ones that werebeyond subject's abilities. Thus, he would fail onthat group of problems and could not earn a token.The interpolated failure sheet was placed second orthird among the other sheets. When subject cameto a star on a failure set of five problems, experi-menter corrected them, marking the two (or more)wrong, and said, "You got too many wrong thattime. You don't get a token." If subject simplysat there and stared at the problems without at-tempting to do them and without indicating that hewas finished with that set, the experimenter waited3 minutes and said, "Time's up." She then cor-rected them, marking the incorrect and unattemptedproblems wrong and said, "You got too many wrongthat time. You don't get a token." If, during afailure set, subject asked how to do the problems,experimenter said, "Just do the best you can."

Division into Treatment Groups

The subjects were matched as closely as possibleon the degree to which the interpolation of failuredisrupted their performance. This was done by cal-culating the percentage of decrease in the number ofcorrect problems per minute from the day beforeinterpolated failure to the day of interpolated failureon the same problems (i.e., those which followedfailure). The subjects were randomly assigned to theAttribution Retraining (AR) Treatment or the Suc-cess Only (SO) Treatment, the second member ofthe pair being automatically assigned to the othergroup. There were three females and three males inthe AR Treatment and two females and four malesin the SO Treatment. Each experimenter workedwith three subjects in each treatment group, with

approximately the same range of subjects from themost disrupted by failure to the least.

Experimental TrainingTwo pieces of equipment especially designed for

the training were used. The casing was made ofwood and measured 23 X 6i X 7J in. (.58 X .17 X.18 m). The front face contained two horizontalparallel rows of 20 small lights, the top row consist-ing of green lights, the bottom row of red lights.There were numbers from 1 to 20 across the faceto correspond to each of the lights. The back of theapparatus contained an on-off switch and a dialwhich could be set to any number between 1 and 20(according to the criterian number of problems ona given trial). The lights in the top row could beilluminated one at a time via a handswitch operatedby the experimenter. When the number of lightsilluminated in the top row matched the criterionnumber, the same number of lights in the bottomrow was illuminated automatically. However, theexperimenter could illuminate the criterion numberof lights in the bottom row by means of a hand-switch before that number was reached in the toprow.

Training was carried out for 25 daily sessions.During these sessions, the subject did problems oneat a time from a stack of 3 X 2 in. (.08 X .05 m)papers each containing one problem. For each sub-ject, 48 problems were taken from sheets of prob-lems that had been easy for him and were presentedin random order. On each trial, the experimenterpreset the criterion number of problems for that trialby means of the dial on the back of the apparatus.When experimenter told subject to begin, experi-menter started the stopwatch and the subject tookone problem from the top of the stack and com-pleted it. If he did it correctly, experimenter said,"Correct," and illuminated the first light in the toprow of lights. If he did the problem incorrectly,experimenter said, "No, go on to the next one," anddid not illuminate a light. The subject continuedworking the problems in this manner until eitherhe had reached the criterion number of problems,whereupon the bottom row of lights illuminated tomatch the number of lights in the top row, or untilexperimenter stopped him and told him his timehad ended.

Each subject was given 15 such trials in everysession. For the Success Only Treatment, the cri-terion number of problems on each trial was set ator below the number he was able to complete withinthe time limit; thus the members of this treatmentgroup could succeed on every trial. The procedurewas the same for the Attribution Retraining Treat-ment on 12 or 13 of the 15 trials. However, on theremaining 2 or 3 trials the criterion number was setone above the number he was generally able tocomplete within the time limit. On these trials, hewas stopped one or two problems short of criterion,his performance was compared to the criterionnumber required, and experimenter verbally attrib-uted the failure to insufficient effort. Thus the

ALLEVIATION or LEARNED HELPLESSNESS 679

members of this treatment group received 2 or 3failure trials in the course of each session.

On the first day of training, subjects were giventhe following instructions. Instructions in parentheseswere given to subjects in the AR Treatment onlyand were omitted for subjects in the SO Treatment.

Today, and from now on, we're going to do some-thing a little different. See the problems on thesepapers? You will do one at a time and you willwork for 1 minute. [The experimenter showedsubject how long it takes to time 1 minute on thestopwatch.]Every time you finish a problem and it is correct,one of these lights will light up [the experimenterindicated lights in the top row]. If you work hard,get enough problems right, and light up enoughlights, you will get a token.Here's how you'll know if you did enough for atoken. If you finish in time, these lights will lightup and tell you [the experimenter indicated lightsin the bottom row]. If you do not finish in time,I will stop you (and these lights will tell youhow hard you should have tried to earn a tokenthat time [the experimenter indicated lights inthe bottom row]).OK, let's try one and see how it works. [Theexperimenter set criterion to 2.] Begin. [The sub-ject took the first paper and did the problem.]Correct. [The experimenter illuminated the firstlight. The subject took the second paper and didthe problem.] Correct. [The experimenter illumi-nated the second light and the two lights in thebottom row lit up.] That was very good. Youfinished in time. See, those lights mean you hadto get 2 right [Experimenter pointed to the bot-tom row] and you did [the experimenter pointedto the top row]. Here's your token.Now let's see what happens when you don't finishin time. Suppose you did 1 problem and it tookthe whole minute. You got it right so the lightlit up [the experimenter illuminated the firstlight]. But then I said, "Stop, time is up! Youdidn't finish in time." (Then these lights will tellyou [experimenter lit two lights of the bottomrow] you had to do 2 problems to earn a tokenthat time. You did only one. That means youshould have tried harder.)

The experimenter then began the actual trials. Itwas determined in advance that if any subject ineither group took longer than 14 minutes tocomplete the criterion number of problems, he wasto fail on that trial. In summary, experimenter re-sponded to the success or failure trials of the twogroups in the following manner:

1. On success trials (Treatments AR and SO):"You finished in time. You needed , and yougot - You get a token."

2. On programmed failure trials (Treatment ARonly): "Stop, time's up. You didn't finish in time.Let's see how many you needed to earn a token.[The experimenter illuminated the criterion numberof lights in the bottom row by means of the hand-

switch.] You needed , and you got only. That means you should have tried harder."

3. On unscheduled failure trials (Treatments ARand SO): "Stop, time's up. You didn't finish intime. Let's go on to the next."

The criterion numbers for the success trials werespecified in advance and consisted of the three num-bers at the upper limit of what a given subjectwas generally able to complete within 1 minute.The experimenter recorded the time and errors oneach trial; every day the mean number of problemsper minute was calculated for each subject and thecriterion numbers for the following day were ad-justed accordingly. Thus, the subjects were alwaysworking to capacity. The criterion number on theprogrammed failure trials for Treatment AR was onenumber above the highest criterion number on thesuccess trials. (The two or three failure trials forTreatment AR in every session were randomly dis-persed among the trials with the constraint that therebe no more than two consecutive failures and thatneither the first nor the last trial be a failure.) Onthe last trial of every session, the subjects in bothtreatments were allowed to reach that number (i.e.,one above the highest criterion number on the othersuccess trials).

Midtraining and Posttraining InterpolatedFailure Tests

For both groups, training was interrupted afterthe thirteenth day for the midtraining interpolatedfailure test. Baselines on the sheets of math prob-lems were reestablished for 3 days, an interpolatedfailure sheet corresponding to the difficult problemswas inserted on the fourth day and baselines werereestablished for the following 3 days. Training wasthen resumed for 12 additional sessions after whichthe interpolated failure procedure was repeated.

RESULTS

Helpless versus Persistent Subjects

The helpless subjects differed from thepersistent subjects on all of the comparisonmeasures, thus verifying the selection proce-dure. The results from the IAR are in agree-ment with those of Dweck and Reppucci(1973) in that the helpless children took lesspersonal responsibility for the outcomes oftheir behavior and tended to place less em-phasis on the role of effort in determiningsuccess and failure than did the persistentchildren. The Total I scores of the helplesssubjects were significantly lower than thoseof the persistent subjects, X = 19.5 and 24.S,respectively, t(20) = 3.11, p < .DOS.5 Whilethey did not differ reliably in the degree towhich they attributed outcomes to ability,

6 All probability values are for one-tailed tests.

680 CAROL S. DWECK

TABLE 1

AND POSTJAILURE PERFORMANCE ON THE PRETRAINING, MlDTRAINING, AND POSTTRAININGTESTS EOR SUBJECTS IN THE ATTRIBUTION RETRAINING AND SUCCESS ONLY TREATMENTS

Test

PretrainingPrefailure

Total time (sec.)Number correct

Postf allureTotal time (sec.)Number correct

MidtrainingPrefailure

Total time (sec.)Number correct

PostfailureTotal time (sec.)Number correct

PosttrainingPrefailure

Total time (sec.)Number correct

PostfailureTotal time (sec.)Number correct

Attribution Retraining Treatment

AN

18010

2559

7310

969

SO10

4810

HA

28512

SIS3

1SSIS

21214

14814

15514

ID

110IS

13615

7415

8115

8515

8115

LA

12515

17214

14814

17215

10714

10815

TE

17012

3463

9513

11112

9913

9113

TO

208IS

36314

14515

15915

17815

13115

X

179.6713.17

297.839.67

115.0013.67

138.5013.33

111.1713.50

102.3313.67

Success Only Treatment

AL

27312

4779

2489

3403

14914

2549

CI

11710

2810

8810

179S

7710

1050

OD

23115

35614

13615

19714

9515

15415

OS

22510

35710

24910

27910

15110

13610

JO

210IS

28315

15015

16612

14215

20211

RO

16115

22712

12415

1279

12815

19314

X

202.8312.83

330.1710.0

165.8312.33

214.678.83

123.6713.17

174.09.83

Note. The number of problems for each subject were as follows: AN = 10, HA = 15, ID = 15, LA = 15, TE = 15, TO = 15(j? = 14 17)' AL = 15 CI = 10 OD = 15, OS = 10, JO = 15, RO = 15 (A = 13.33). Subjects working at different levels ofdifficulty' received sheets containing different numbers of problems. Prefailure and postfailure problems were identical and ad-ministered on consecutive days.

that is, (!+A) + (!-A), # = 8-6 and 10.3,respectively, t(2Q) = 1.3, p < .10, the differ-ence in the degree to which they attributedoutcomes to effort, that is, (!+E) + (!~E),was highly reliable, X = 10.9 and 14.1, re-spectively, *(20) = 3.20, p < .005.

On both of the TASC subscales adminis-tered (test anxiety and poor self-evaluation),6

the helpless subjects proved to be more anx-ious than the comparison subjects. For thetest anxiety subscale, the means for the help-less and persistent subjects were 5.3 and 4.2,respectively, t ( l & ) = 2.80, p < .01, and forthe poor self-evaluation subscale, the meanswere 3.9 and 3.1, respectively, <(18) = 2.47,p < .025.

On the Repetition Choice task, 9 of the12 helpless subjects, when given the choice,chose to reconstruct the puzzle they had al-ready successfully completed rather than theone on which they had been interrupted,while only 1 of the persistent subjects choseto do so, x2(l) =6.86, p<.Ql. Thus, on

•These data were missing for two of the per-sistent comparison subjects.

this measure, the helpless subjects evidenceda clear tendency to avoid failure, while thepersistent children showed a tendency tostrive for success.

Effects of Training

Table 1 contains the absolute levels of pre-failure and postfailure performance (time tocompletion and number correct) for eachsubject on the pretraining, midtraining, andposttraining tests. To obtain a measure thatwould take account of the effects of failureon both speed and accuracy of performance,the number of correct problems per minutewas calculated. The percentage decrease inthis measure was used to compare prefailure(baseline) to postfailure performance on theidentical problems.

Figures 1 and 2 show the percentage de-crease in correct problems per minute follow-ing failure on the pretraining, midtraining,and posttraining interpolated failure tests forthe individual subjects in the AR and SOtreatments, respectively. Inspection of theindividual graphs reveals that the perform-

ALLEVIATION OF LEARNED HELPLESSNESS 681

ances of all subjects underwent markeddeterioration on the interpolated failure dayin the pretreatment period. That this decreasein rate is not a result of being more carefulis reflected by the increase in the number oferrors on the part of the subjects and by theobservation of other behavior in the periodfollowing failure, such as deterioration ofhandwriting, staring into space, etc. By mid-training, all the subjects in the AR Treat-ment snowed some improvement, althoughall still decreased their score over the previ-ous day. By the end of training, however, allof the subjects in this treatment showed eithernegligible impairment following failure orimprovement in the number of correct prob-lems per minute compared with their per-formance on the day prior to interpolatedfailure. In the SO treatment, on the otherhand, all but one of the subjects (OS) con-tinued to show substantial decrements in per-formance at both the midtraining and post-training tests, with two of the subjectsshowing increased impairment from pretrain-ing to posttraining. Thus, even though most

100

CCUJ0.

mocc

50

-50

100

50

AL V Cl OD

oi-

-50

OS JO RO

PHASE OF TRAINING

FIGURE 2. Percentage decrease in correct problemsper minute at pretraining, midtraining, and post-training for the subjects in the Success Only Treat-ment.

100

50K.UJO.

-50

AN HA

.

ID

oraa

UJo

100

50

0 -

-50

LA

\

TE

\

TO

Prl Mid Post

PHASE OF TRAINING

FIGURE 1. Percentage decrease in correct problemsper minute at pretraining, midtraining, and post-training for the subjects in the Attribution Retrain-ing Treatment.

of the subjects in both treatments improvedin their general performance on these prob-lems on nonfailure days, only the subjects inthe AR Treatment showed consistent andsubstantial decreases in their maladaptivereaction to failure.

Figure 3 shows the mean percentage de-crease in correct problems per minute for thetwo treatment groups at pretraining, mid-training, and posttraining. The t tests on thechanges from pretraining to midtraining andfrom midtraining to posttraining for the sub-jects in the AR Treatment showed a signifi-cant difference in percentage decrease incorrect problems per minute from pretrainingto midtraining, X= 51.4 and 18.2, t ( 5 ) -3.32, p < .025, and from midtraining to post-training, £=18.2 and -9.2, t(5) = 5.51,p < .005. A t test comparing the changescores from pretraining to posttraining of thesubjects in the AR Treatment to those of thesubjects in the SO Treatment revealed a sig-nificantly greater improvement for the ARTreatment, £AR = 51.4 and -9.2, £so = 50.4and 46.0, *(10) = 3.64, p < .005. Thus, it is

682 CAROL S. DWECK

2llJ_lCDOtra.h-oLUirtroo

UJa:o

2 - 2 0

60

40

20

TREATMENT

0—0 A R

PRE MID

T R A I N I N G

POST

FIGURE 3. Mean percentage decrease in correctproblems per minute at pretraining, midtraining,and posttraining for the Attribution RetrainingTreatment and the Success Only Treatment.

clear that while the subjects in the Attribu-tion Retraining Treatment were able to handlefailure more adaptively in the test situationfollowing treatment, some of the subjects inthe Success Only Treatment seem to havebecome more sensitive to failure in the testsituation following exclusive success duringtraining.

Readministered Measures

On the Effort versus Ability Failure Attri-bution scale, all of the subjects in the ARTreatment showed an increase in the choiceof the effort alternatives from pretraining toposttraining, X = 1.3 and 3.0, f(5) =3.39,p < .01. The subjects in the SO Treatmentshowed no such increase (^ — 1.3 and 1.5).A comparison of the posttraining scores ofthe subjects in the two treatments shows areliably greater tendency for the subjects inthe AR treatment to emphasize the role ofeffort in determining failure, £(10) =3.0,p < .01. The change in this measure providesevidence that the subjects in the AR Treat-ment, in addition to changing their reactionsto failure in the experimental situation,altered their attributions for failure in situa-tions involving mathematics in general.

There were no consistent changes in thescores of either treatment group on the IARscale as of the time this scale was readmin-istered. For the subjects in the AR Treatmentthe means before and after training were19.0 and 19.S; for those in the SO Treatmentthe means were 20.0 and 20.8.

On the Test Anxiety Subscale of theTASC, the subjects in the AR Treatmentshowed a decrease in test anxiety (^pHE =5.7 and Xp0sT = 3.7), although the changedid not reach significance, t ( 5 ) = 1.89, p <.10, while the subjects in the SO Treatmentshowed a slight increase (^PRE =5.0 and-?posT=5.5). The difference between thechanges in test anxiety for the two treat-ments, however, was not significant, t ( W ) =1.58, p < .10. On the Poor Self-EvaluationSubscale of the TASC, the subjects in bothtreatments showed decreases, but these de-creases were not reliable. For the AR Treat-ment, ^PHE = 4.5 and ^POST = 2.2, t ( 5 ) =1.71, p< .10; for the SO Treatment, XFRK

= 4.0 and XPOST = 3.0, t ( 5 ) = 1.59, p < .10,Initially, on the Repetition Choice task,

one subject in the AR Treatment and twosubjects in the SO Treatment had chosen toreconstruct the failed puzzle rather than theone which they had successfully completed.After training, two additional subjects in theAR Treatment chose to do the puzzle onwhich they had previously failed, while noneof the subjects in the SO Treatment alteredtheir choices. However, this change was notstatistically reliable.

In summary, the subjects in the AR Treat-ment showed large changes in their recogni-tion of effort as a determinant of failure asreflected in their change scores on the Effortversus Ability Failure Attribution scale, themeasure most closely related to the attribu-tion retraining treatment, but failed to showreliable changes on the other, more globalmeasures. The subjects in the Success OnlyTreatment, on the other hand, did not showsignificant changes on any of the measureswhich were readministered after training.

DISCUSSION

The investigation began with the assump-tion that the manner in which a child viewsan aversive event, such as failure, determines,in large part, the way in which he reacts to

ALLEVIATION OF LEARNED HELPLESSNESS 683

that event. Specifically, if a child believesfailure to be a result of his lack of ability ora result of external factors beyond his con-trol, he is unlikely to persist in his efforts.On the other hand, if a child believes failureto be a result of his lack of motivation, heis likely to escalate his effort in an attemptto obtain the goal. It was further assumed(see Dweck & Reppucci, 1973) that thosechildren who are characterized as learnedhelpless tend to regard failure in the formerway, while those who are characterized aspersistent tend toward the latter view.

It was then hypothesized that by changingthe helpless child's attribution of failure inthe training situation (from fixed to variablefactors; see Weiner, 1972) one could therebychange his reaction to failure in the testsituation (from surrender to persistence).Training and testing differed in a numberof respects. During training, relatively easyproblems were completed by the subject andcorrected by the experimenter one at a time;during testing, more difficult problems werecompleted by the subject and corrected bythe experimenter in groups of five. In train-ing, responses were monitored on the appa-ratus; in testing, no apparatus was used. Intraining, failure meant that the child did notcomplete enough problems within the allottedtime; in testing, failure meant that the childgot the problems wrong. Despite these dif-ferences, it was expected that the new attribu-tion would generalize to the test situation andenable the child to persist following failure.The results of the investigation lend strongsupport to the hypothesis.

Prior to training, all subjects showedseverely deteriorated performance followingfailure; problems they had been solving dailybecame sources of great difficulty as evidencedby a decrease in rate of problem solving, anincrease in errors, and behavioral indicants ofwithdrawal. The children who were taught toattribute failure during training to insufficienteffort were able to persist after failure in thetest situation. That failure became a cue toescalate effort is supported by the findingthat five of the six subjects receiving theAttribution Retraining Treatment, in fact,showed superior performance following fail-ure (as compared to the day precedingfailure). Interestingly, three of the subjects

in the AR Treatment actually verbalized theattribution during testing.

Contrary to initial expectation, however,the subjects in the Success Only Treatmentdid not show any consistent improvement intheir response to failure, but rather continuedto display a marked impairment of perform-ance following failure. This occurred despitethe fact that their performance during train-ing and on nonfailure days during testingsteadily improved. While the successes mightwell have resulted in higher expectations ofsuccess and might have kept the situationfrom becoming aversive, this was not suffi-cient to prevent the adverse reaction to fail-ure. Two of the subjects in this group evenappeared to be slightly more sensitive tofailure after prolonged experience with con-tinued success. If their successes led them toform a new image of themselves, that of asucceeder, failure would indeed have a seriousimpact.

Investigators who advocate errorless learn-ing for children often cite Terrace's (1963a)work on errorless discrimination in thepigeon. Terrace demonstrated that pigeonsmaking unreinforced responses to the nega-tive stimulus (i.e., errors) during discrimina-tion training showed gross emotional re-sponses to that stimulus. In other words, itacquired aversive properties. When discrimi-nations were trained without responding tothe negative stimulus (i.e., were errorless), itremained a neutral stimulus. Skinner (1968)says of the pigeon, "It is, so to speak, 'notafraid of making a mistake.'" It is rarelypointed out, although it is explicitly statedby Terrace (1969), that in some cases "error-less discrimination is not an unmixed bless-ing." For example, Terrace (1963b) foundthat when an errorless discrimination wasfollowed by a different discrimination trainedwith errors, the original perfect discrimina-tion was disrupted: Similarly, Terrace (1969)reported that when a continuously reinforcederrorless discrimination was followed by ex-tinction, the previously errorless performancewas interfered with. It would seem that al-though errorless learning led to superior per-formance, it rendered the subjects less ableto deal with subsequent errors.

While a success only procedure for childrenis an effective approach for teaching a given

684 CAROL S. DWECK

body of material, the present findings suggestthat it might be a short-sighted approach. Theimplications for strategies of behavior changeor behavior building are rather straight-forward. An instructional program for chil-dren who have difficulty dealing with failurewould do well not to skirt the issue by tryingto ensure success or by glossing over failure.Instead it should include procedures for deal-ing with this problem directly. This is not tosuggest that failure should be included ingreat amounts or that failure per se is desir-able, but rather, that errors should be capi-talized upon as vehicles for teaching the childhow to handle failure.

The .training procedure in the present studyaddressed itself to a relatively narrow aspectof the student's achievement behavior—hisperformance in the experimental situation.The results for the children receiving attribu-tion retraining provide evidence for a changein behavior in that situation and for a greateremphasis on the role of motivation in deter-mining failure in arithmetic. It is not sur-prising that the less related, more global cog-nitive measures failed to show significantchanges, since (a) so little time elapsed be-tween the completion of training and thereadministration of these measures and (b)training was conducted in a limited context.It is possible that additional time would haveallowed the child to "test out" the new at-tribution in other settings and to begin to useit more generally. Unfortunately, the presentstudy was conducted at the end of the schoolyear (March through June) and there wastherefore no opportunity to test this possibil-ity. A safer assumption, however, is thatexplicit generalization training is required toassure changes in the global cognitive mea-sures. Such training would involve guidedexperience in applying the new attribution toa variety of situations.

Although direct measures of transfer of theeffects of training to the classroom were nottaken, verbal reports from the teachers (whodid not know which subject received eachtreatment) indicated that the subjects in theAttribution Retraining Treatment had begunto work harder and to develop a different atti-tude toward failure. This was evidenced bytheir increased persistence with new material

and their seeking help rather than withdraw-ing when unable to complete a task.

Cognitive-personality variables, such as themanner in which a person perceives the rela-tionship between his behavior and the occur-rence of certain events, indeed appear to beimportant determinants of the way in whichpeople react to events. Many clinicians (e.g.,Rogers, 1951) have hoped for changes inthese variables to come about as byproductsof therapy, for example, increased responsibil-ity for one's behavior or increased self-esteem.This may be akin to the Success Only Treat-ment in the present experiment. Others (e.g.,Ellis, 1962; Meichenbaum, Note 1) havebegun to try to modify directly "what clientssay to themselves" about events. The presentinvestigation emphasizes the usefulness of con-sidering such cognitive-personality variablesin planning strategies of behavior change.

REFERENCE NOTES

1. Meichenbaum, D. H. Cognitive factors in behav-ior modification: Modifying what clients say tothemselves. Paper presented at the 5th AnnualMeeting of the Association for Advancement ofBehavior Therapy, Washington, D.C., September1971.

2. Odell, M. Personality correlates of independenceand conformity. Unpublished master's thesis, OhioState University, 1959.

REFERENCES

Atkinson, J. W., & Feather, N. T. (Eds.) A theoryof achievement motivation. New York: Wiley,1966.

Battle, E. Motivational determinants of academictask persistence. Journal of Personality and SocialPsychology, 1965, 2, 209-218.

Bialer, I., & Cromwell, R. L. Task repetition inmental defectives as a function of chronologicaland mental age. American Journal of MentalDeficiency, 1960, 65, 265-268.

Bigelow, G. The behavioral approach to retardation.In T. Thompson & J. Grabowski (Eds.), Behaviormodification of the mentally retarded. New York:Oxford University Press, 1972.

Butterfield, E. C. Locus of control, test anxiety,reactions to frustration and achievement attitudes.Journal of Personality, 1964, 32, 298-311.

Crandall, V. C., Katkovsky, W., & Crandall, V. J.Children's beliefs in their own control of reinforce-ments in intellectual-academic achievement situa-tions. Child Development, 1965, 36, 91-109.

Davison, G. Differential relaxation and cognitive re-structuring in therapy with a "paranoid schizo-phrenic" or "paranoid state." Proceedings of the

ALLEVIATION OF LEARNED HELPLESSNESS 685

74th Annual Convention of the American Psycho-logical Association, 1966, 1, 177-178. (Summary)

Dweck, C. S., & Reppucci, N. D. Learned helpless-ness and • reinforcement responsibility in children.Journal of Personality and Social Psychology,1973, 25, 109-116.

Ellis, A. Reason and emotion in psychotherapy. NewYork: Lyle-Stuart, 1962.

Feather, N. T. Effects of prior success and failureon expectations of success and subsequent perform-ance. Journal of Personality and Social Psychol-ogy, 1966, 3, 287-298.

Feld, S. C., & Lewis, J. The assessment of achieve-ment anxieties in children. In C. P. Smith (Ed.),Achievement-related motives in children. NewYork: Russell Sage Foundation, 1969.

Hart, B. M., & Risley, T. R. Establishing use ofdescriptive adjectives in the spontaneous speech ofdisadvantaged children. Journal of Applied Behav-ioral Analysis, 1968, 1, 109-120.

Holland, J. G. Teaching machines: An applicationof principles from the laboratory. Journal of theExperimental Analysis of Behavior, 1960, 3, 275-287.

Meacham, M. L., & Wiesen, A. E. Changing class-room behavior. Scranton, Pa.: International Text-book, 1969.

Mowrer, O. H. Learning theory and behavior. NewYork: Wiley, 1960,

Nisbett, R. E., & Schachter, S. Cognitive manipula-tion of pain. Journal of Experimental Social Psy-chology, 1966, 2, 227-236.

Rogers, C. R. Client-centered therapy. Boston:Houghton-Mifflin, 1951.

Ross, L. D., Rodin, J., & Zimbardo, P. G. Towardan attribution therapy: The reduction of fearthrough induced cognitive-emotional misattribu-tion. Journal of Personality and Social Psychology,1969, 12, 279-288.

Rotter, J. B. Generalized expectancies for internalversus external control of reinforcement. Psycho-

logical Monographs, 1966, 80(1, Whole No. 609).Sarason, S. B., Davidson, K. S., Lighthall, F. F.,

Waite, R. R., & Ruebush, B. K. Anxiety in ele-mentary school children. New York: Wiley, 1960.

Seligman, M., & Maier, S. Failure to escape trau-matic shock. Journal of Experimental Psychology,1967, 74, 1-9.

Skinner, B. F. The technology of teaching. NewYork: Appleton-Century-Crofts, 1968.

Storms, M. D., & Nisbett, R. E. Insomnia and theattribution process. Journal of Personality and So-cial Psychology, 1970, 16, 319-328.

Terrace, H. S. Discrimination learning with andwithout "errors." Journal of the ExperimentalAnalysis of Behavior, 1963, 6, 1-27. (a)

Terrace, H. S. Errorless transfer of a discriminationacross two continua. Journal of the ExperimentalAnalysis of Behavior, 1963, 6, 223-232. (b)

Terrace, H. S. Extinction of a discriminative operantfollowing discrimination learning with and withouterrors. Journal of the Experimental Analysis ofBehavior, 1969, 12, 571-582.

Tombaugh, T. N., & Tombaugh, J. Effects of appar-ent task solubility and reward upon the durationof problem-solving behavior. Psychological Record,1963, 13, 83-87.

Tyler, B. B. Expectancy for eventual success as afactor in problem-solving behavior. Journal ofEducational Psychology, 1958, 49, 166-172.

Valins, S., & Nisbett, R. E. Some implications ofattribution processes for the development andtreatment of emotional disorders. New York:General Learning Press, 1971.

Weiner, B. Theories of motivation. Chicago:Markham, 1972.

Weiner, B., & Kukla, A. An attributional analysis ofachievement motivation. Journal of Personalityand Social Psychology, 1970, S, 1-20.

(Received September 24, 1973)