Confounding of controllability in the triadic design for demonstrating learned helplessness

11
Journal of Personality and Social Psychology 1981, Vol. 41, No. 3, 458-468 Copyright 1981 by the American Psychological Association, Inc. 0022-3514/81/4103-0458100.75 Confounding of Controllability in the Triadic Design for Demonstrating Learned Helplessness R. E. Lubow, R. Rosenblatt, and I. Weiner Tel Aviv University, Tel Aviv, Israel The triadic design for producing "learned helplessness" was analyzed and found to be logically deficient for defining the theoretical construct of controllability. Two experiments with adult human subjects were performed. In Experiment 1, three groups constituted the traditional learned helplessness triadic design. A fourth group, yoked to the escapable group, was instructed to press a button when the tone terminated. A fifth group received a flash of light upon tone offset. Groups 4 and 5, which had no controllability over tone offset, but which did have a second event immediately following the tone, both showed significantly better performance on a subsequent tone-escape task than Group 1. Experiment 2 replicated these findings. The data indicate that controllability may be a sufficient condition but not a necessary condition in the triadic design for preventing sub- sequent learned helplessness. It follows logically that uncontrollability is not the appropriate designation of the process or processes involved in producing such effects. It has been more than 10 years since the publication of the germinal experiments that gave rise to the mushrooming research on the effects of preexposure to shock on shock- escape behaviors (Overmier & Seligman, 1967; Seligman & Maier, 1967). In these experiments, dogs who received inescapable shock, as compared to dogs who could ter- minate the shock, subsequently failed to ac- quire the appropriate escape-from-shock re- sponse. Since those first studies, this decremental effect has been replicated with a variety of preexposure and test conditions and with a number of different species. It has also been 10 years since this effect was first treated within a general theoretical framework and labeled "learned helpless- ness." (For recent general reviews, see Abramson, Seligman, & Teasdale, 1978; Maier & Seligman, 1976; Seligman, 1975). The learned helplessness theory was derived primarily from the conceptualization and the data of the triadic design that was consis- tently employed in the early experimental This research was supported by grants from the Israel Academy of Sciences and Humanities Basic Research Foundation and the Schwartz Foundation. Experiments 1 and 2 were conductedby the second author, under the direction of the first author, in partial fulfillment of the requirements for the master's degree. Requests for reprints should be sent to R. E. Lubow, Department of Psychology, Tel-Aviv University, Ra- mat-Aviv, 69978 Israel. studies. In this design, three groups of ani- mals are used. One group receives a series of noxious stimuli (e.g., shock) that it can control by a response. The second, or yoked group, receives exactly the same series of stimuli, but there is no response these ani- mals can make to terminate the stimuli. The third group does not receive any expo- sure. Many authors have labeled Group 1 the control, or escapable shock group; Group 2, the yoked, no control, or inescapable shock group; and Group 3, the no shock group. However, it should be readily apparent that the names of Groups 1 and 2 prejudge the identification of the critical dimension of the preexposure phase that affects performance in the test phase. Seligman and Maier have repeatedly affirmed this intent with their theory of learned helplessness. Simply, the theory states that the repeated absence of control over the termination of a noxious, stressful event leads to the acquisition of the cognition that any responding will be inde- pendent of any situational outcome. During the preexposure phase the subject learns that responses and consequences are independent of each other and during the test phase the subject performs relatively poorly because in the recent past, response and consequence were independent of each other. The major question that is addressed by this article concerns the assumption and the 458

Transcript of Confounding of controllability in the triadic design for demonstrating learned helplessness

Journal of Personality and Social Psychology1981, Vol. 41, No. 3, 458-468

Copyright 1981 by the American Psychological Association, Inc.0022-3514/81/4103-0458100.75

Confounding of Controllability in the Triadic Designfor Demonstrating Learned Helplessness

R. E. Lubow, R. Rosenblatt, and I. WeinerTel Aviv University, Tel Aviv, Israel

The triadic design for producing "learned helplessness" was analyzed and foundto be logically deficient for defining the theoretical construct of controllability.Two experiments with adult human subjects were performed. In Experiment 1,three groups constituted the traditional learned helplessness triadic design. Afourth group, yoked to the escapable group, was instructed to press a buttonwhen the tone terminated. A fifth group received a flash of light upon tone offset.Groups 4 and 5, which had no controllability over tone offset, but which did havea second event immediately following the tone, both showed significantly betterperformance on a subsequent tone-escape task than Group 1. Experiment 2replicated these findings. The data indicate that controllability may be a sufficientcondition but not a necessary condition in the triadic design for preventing sub-sequent learned helplessness. It follows logically that uncontrollability is not theappropriate designation of the process or processes involved in producing sucheffects.

It has been more than 10 years since thepublication of the germinal experiments thatgave rise to the mushrooming research onthe effects of preexposure to shock on shock-escape behaviors (Overmier & Seligman,1967; Seligman & Maier, 1967). In theseexperiments, dogs who received inescapableshock, as compared to dogs who could ter-minate the shock, subsequently failed to ac-quire the appropriate escape-from-shock re-sponse. Since those first studies, thisdecremental effect has been replicated witha variety of preexposure and test conditionsand with a number of different species. Ithas also been 10 years since this effect wasfirst treated within a general theoreticalframework and labeled "learned helpless-ness." (For recent general reviews, seeAbramson, Seligman, & Teasdale, 1978;Maier & Seligman, 1976; Seligman, 1975).The learned helplessness theory was derivedprimarily from the conceptualization and thedata of the triadic design that was consis-tently employed in the early experimental

This research was supported by grants from the IsraelAcademy of Sciences and Humanities Basic ResearchFoundation and the Schwartz Foundation. Experiments1 and 2 were conducted by the second author, under thedirection of the first author, in partial fulfillment of therequirements for the master's degree.

Requests for reprints should be sent to R. E. Lubow,Department of Psychology, Tel-Aviv University, Ra-mat-Aviv, 69978 Israel.

studies. In this design, three groups of ani-mals are used. One group receives a seriesof noxious stimuli (e.g., shock) that it cancontrol by a response. The second, or yokedgroup, receives exactly the same series ofstimuli, but there is no response these ani-mals can make to terminate the stimuli.The third group does not receive any expo-sure.

Many authors have labeled Group 1 thecontrol, or escapable shock group; Group 2,the yoked, no control, or inescapable shockgroup; and Group 3, the no shock group.However, it should be readily apparent thatthe names of Groups 1 and 2 prejudge theidentification of the critical dimension of thepreexposure phase that affects performancein the test phase. Seligman and Maier haverepeatedly affirmed this intent with theirtheory of learned helplessness. Simply, thetheory states that the repeated absence ofcontrol over the termination of a noxious,stressful event leads to the acquisition of thecognition that any responding will be inde-pendent of any situational outcome. Duringthe preexposure phase the subject learns thatresponses and consequences are independentof each other and during the test phase thesubject performs relatively poorly becausein the recent past, response and consequencewere independent of each other.

The major question that is addressed bythis article concerns the assumption and the

458

CONFOUNDING OF CONTROLLABILITY 459

requirement of the learned helplessness the-ory that response-outcome independence, asoperationally defined in the paradigmatictriadic design, is a valid illustration of un-controllability. Presenting shock by itself, inthe absence of a contingent response, alsois an operation for producing habituationand stress as well as uncontrollability. It isan operational procedure for inducing anumber of different intervening psycholog-ical states. Why, of all the possibilities forconceptualizing the procedure of repeatedshock presentation, is the feature of uncon-trollability chosen for special explanatorystatus? The reason, of course, is that thereis another group in which subjects' responsescan terminate shock, and it is both com-fortable and acceptable to speak of such ananimal as controlling the shock offset. Inoperational terms, this characterization ofthe situation is quite appropriate. That theanimal's response terminates the shock is thesame as saying that the animal controlsshock offset. Nevertheless, neither of thesedescriptive statements can be used to inferthat, at the more conceptual level, "con-trollability" is the general theoretical vari-able that mediates the shock-responseoperation and the subsequent superiorperformance as compared to animals whocannot respond to escape shock. What isbeing asserted here is that the concept ofcontrollability, as operationally defined, isconfounded with operations that also char-acterize other concepts. More specifically,we would like to emphasize the operationalfact that escapable and inescapable groupsdiffer not only in their having control overthe termination of an aversive event, but alsoin their having some event follow the shock.Consequently, we propose an alternative in-terpretation of shock-response effects, basedon the analysis of the shock-response-ter-minates-shock sequence in terms of a broaderconceptualization of that relationship—namely that of the contingency between twoevents, E, and E2, any one of which may bea traditional stimulus, a response, or somecombination of the two.

The proposed alternative description ofthe necessary condition for producing learnedhelplessness effect is that of a lack of con-tingency between a repeatedly presented tar-get event, Et (e.g., shock), and any other

event in the environment. The result of sucha manipulation is reduced associability of E,with other potentially contingent events ina subsequent test, such as a response to es-cape from shock. On the other hand, theoccurrence of an event, E2, which is reliablycorrelated with the target event, EI, will pre-vent the loss of associability of Et. Thus,when subjects are given the opportunity torespond to EI (e.g., to terminate shock) dur-ing preexposure, no subsequent decrementaleffects should be found. In terms of the pres-ent approach, the escape response during thelearned helplessness (LH) preexposure phaseconstitutes an event of consequence, E2, inthe environment that is correlated with EI.However, there are many other such eventsthat can serve the same purpose.

Similarly to the theory proposed by Maierand Seligman (1976), we emphasize a learn-ing process during the stage of preexposureand the transfer of this learning to the stageof subsequent acquisition. However, thepresent approach and learned helplessnesstheory differ in the identification of what islearned and in the specification of which the-oretical construct best describes the acquisi-tion of such learning. Whereas learned help-lessness theory emphasizes the lack ofcontingency between response and outcome,and consequently the development of a cog-nitive set of uncontrollability during shockpreexposure, this approach addresses itselfto lack of contingency between a repeatedlypresented stimulus (e.g., shock) and anyother event in the environment, and the de-velopment of an associative deficit to thepreexposed stimulus due to such lack of con-tingency. We are suggesting that response-outcome contingency is but one instance ofa broader concept of contingency amongevents, and can be analyzed in terms of gen-eral properties of such contingencies, ratherthan in terms of a controllability-uncon-trollability dimension.

The above approach has been fruitful inexplaining another stimulus preexposurephenomenon, that of latent inhibition (LI).In the LI paradigm, repeated preexposureto the to-be-conditioned stimulus (CS) re-tards its subsequent conditioning when it ispaired with an unconditioned stimulus orreinforcing event (see Lubow, 1973). Theapparent similarity between the two phe-

460 R. E. LUBOW, R. ROSENBLATT, AND I. WEINER

nomena, latent inhibition and learned help-lessness, has been acknowledged by a num-ber of authors (Baker, 1976; Mackintosh,1974; Wickens, Tuber, Nield, & Wickens,1977). Among these similarities are the fol-lowing: (a) both use a stimulus preexposuremanipulation followed by a transfer test toassess the effects of such preexposure. In onecase, latent inhibition, a typical CS is preex-posed for a number of trials and the effectsare measured in a learning task that usesthat same CS. Likewise, in the learned help-lessness paradigm a typical US is preexposedfor a number of trials and the effects aremeasured in a learning task that uses thesame US. (b) The major effects of both LIand LH procedures are poor performance ascompared to groups that have not receivedstimulus preexposure. (c) When E2 is added,so that it immediately follows El5 the dec-remental effects (learned helplessness or la-tent inhibition) are abolished as comparedto an EI-alone group. This, of course, is seenin the triadic design in which the group thatcan terminate shock with a response (E2)performs similarly to the no-shock preex-posed group on the subsequent test. Like-wise, this effect of E2 is present in those la-tent inhibition studies that have usedpreexposure procedures in which §! is fol-lowed by a second stimulus such as a lightor tone (Lubow, Alek, & Arzy, 1975; Lu-bow, Schnur, & Rifkin, 1976; Szakmary,1977; Szakmary, Note 1). Si-S2 paired, ascompared to S, alone or to Si and S2 un-paired, sharply attenuates the latent inhi-bition effect. The present studies will providea further test for the proposition that the twostimulus preexposure phenomena, LI andLH, are governed by similar processes.

The test of the applicability of the presentapproach to learned helplessness is straight-forward. If the decremental effects of shockpreexposure are due to lack of contingencybetween E, and any other event in the en-vironment, then any event correlated withE) should attenuate the learned helplessnesseffect, even though it does not terminateshock presentation (i.e., does not establishresponse-outcome contingency and does notincrease controllability). Thus, if one canshow that the effect of the response that ter-minates shock in the so-called controlling

group can be duplicated by nonresponseevents, such as following the shock with a"neutral" S2 (as in the analogous latent in-hibition experiments described above) or withresponses that are contingent to the shock butdo not control its offset, then it becomes likelythat Seligman and Maier's theoretical inter-pretation of the role of response-terminatedshock in terms of control is not appropriate.Therefore, for reasons argued earlier, the char-acterization of the deleterious effects of shockalone as due to the learning of uncontrolla-bility would be questionable. Control or lackof control, as used by Seligman and Maier,would be a sufficient condition for demonstrat-ing learned helplessness effects, but not a nec-essary condition. Control would be one aspectof a more general variable that might betterbe described as contingency among events.

In the following section, two experimentsare presented that support the propositionthat the Maier and Seligman theory indeedhas provided an inaccurate description of thecritical variable. The rationale for these ex-periments, as already presented, suggeststhat the response controlling shock offsetreduces the amount of subsequent behaviordecrement, as compared to a passive shockgroup, because of its general properties asan E2. It therefore follows that any E2,whether a stimulus or a response that doesnot control shock offset, should serve to re-duce the learned helplessness effect.

There were six groups of adult humans inthe first experiment. The groups were dif-ferentiated by their treatment during thepreexposure phase. The first three groupsrepresented the usual triadic design for dem-onstrating learned helplessness. Group 1subjects received a noxious tone that theycould terminate with a button press re-sponse. Group 2 subjects were yoked to thoseof Group 1 and received the same durationand pattern of tones as their active partnersbut had no control over tone termination.Subjects of Group 3 were not preexposed tothe noxious tone. The novel and theoreticallycritical Groups 4 and 5 were yoked to Group1. In both groups, subjects had no controlover tone offset. Subjects in Group 4 wereinstructed to press the button after the tonestopped. Subjects in Group 5 were treatedsimilarly to subjects in Group 4 except that

CONFOUNDING OF CONTROLLABILITY 461

E2, instead of a response, was a flash of lightthat followed shock offset. All subjects werethen treated in an avoidance-escape shuttle-box. The predictions were that the groupsof the traditional triadic design should showthe typical effects as repeatedly demon-strated in many previous experiments. Thatis, the passive Group 2 should show pooreravoidance-escape performance than the ac-tive control group and the no-tone preex-posure group, the latter two of which shouldnot differ from each other; Groups 4 and 5(tone followed by response and tone followedby light) should likewise show significantlybetter performance than Group 2. No pre-dictions were made as to the relative per-formances of Groups 1, 4, and 5. AlthoughE2 is operative in all cases, either as a re-sponse following tone or as a light followingtone, there was no independent predictionof their relative effectiveness. Subjects inGroup 6 were treated similarly to those ofGroup 2 and were exposed to inescapablenoxious tones. However, unlike Group 2 sub-jects, who received instructions suggestingthat escape was possible, subjects of Group6 were given neutral, non-frustration-pro-ducing information. This was added becauseof the suggestion in the literature that theuncontrollability groups of early learnedhelplessness studies (e.g., Hiroto, 1974; Hir-oto & Seligman, 1975) may have performedpoorly in the test phase as a result of theinduction of frustration during the preex-posure phase. It should be noted here thatalthough the original versions of the learnedhelplessness theory (Maier & Seligman,1976; Seligman, 1975; Seligman, Maier, &Solomon, 1971) might have been uncom-fortable with a frustration effect, the mostrecent version that emphasizes attributionformulations would not find such an effectembarrassing (Abramson et al., 1978). Onthe other hand, the alternative approach thatwas discussed earlier would find it difficultto encompass such a frustration effect. Inaddition, Group 6 serves to control for theabsence of frustration instructions in Groups4 and 5. Without such a group, the com-parisons with Group 2 would be confoundedin that Group 2 has no E2 and frustration-instructions, whereas Groups 4 and 5 haveE2 and no frustration-instructions.

The results of Experiment 1 strongly sup-ported the contention that lack of control-lability is not the sine qua non of the learnedhelplessness effect. The theoretical impli-cations of this conclusion are of such im-portance that it was deemed prudent to rep-licate the experiment. For Experiment 2,Groups 1 to 5 were rerun exactly as in Ex-periment 1. However, in addition to the shut-tle response test for tone escape, a secondtask that did not employ the preexposedstimulus was added. The purpose of this task(anagrams) was to determine whether thegeneralization of learned helplessness, sowidely reported in the literature (see Maier& Seligman, 1976, for a review), would alsobe affected by the E2 manipulations.

Experiment 1

Method

Subjects

The subjects were 40 females and 20 males from in-troductory psychology courses at Tel Aviv University.The mean age was approximately 22 years. Subjectsparticipated in the experiment to fulfill a course re-quirement. They were randomly assigned to one of sixgroups, with two restrictions: (a) the number from eachsex was kept the same across groups; (b) for each of the10 replications the first assignment was made to theescape group.

All subjects were informed that the study would in-volve listening to "some loud noises which have beenjudged to be somewhat unpleasant but are not harmfulor dangerous to you." They were then presented witha sample tone and given the opportunity to withdrawfrom the experiment. On this basis three subjects refusedto continue in the study. A fourth subject left in themiddle of the preexposure session, and a fifth wasdropped because of an apparatus malfunction. All fivesubjects were replaced.

Apparatus

The noxious preexposure stimulus, a 110 dB, 2000Hz tone, was delivered to the subject via stereo ear-phones. For each replication the escape group subjectwas tested first. All subjects from the other groups were"yoked" to the escape subject. That is, the escape sub-ject, who could press a button to terminate the tone,provided the tone durations for all other members ofthat replication. This was accomplished by taping thesession and then using the tape to present the subsequenttone preexposures to subjects of the other groups.

The response device was a standard, spring-loaded,red doorbell button placed 75 cm from the edge of atable. All subjects were required to sit next to the tablewith their hands resting on their knees. This procedure

462 R. E. LUBOW, R. ROSENBLATT, AND I. WEINER

ensured that the escape group would generate tones thathad a minimum duration of 1 sec. Two small, uncapped3-watt bulbs were located 10 cm to either side of thebutton. These lights provided the "passive 82" condition.

In the test phase, the apparatus was a modificationof the Turner and Solomon (1962) two-way shuttle box.The shuttle-box 62 X 12.75 X 15.24 cm, had a 4 cmplastic doorknob protruding from the top. The knobcould slide along the length of a 49.3 cm straight chan-nel. On alternate trials either the right or left switchterminated the tone. The tone was the same frequencyand intensity as that used in the preexposure phase. Asmall red jewel lamp was centered on the top surfaceof the shuttle-box, 5 cm from the channel. Standardtimers and counters were used to record the data.

Procedure

Six groups were differentiated on the basis of theirtreatment during the preexposure phase. All subjectswere brought into the room individually, seated in frontof the table, and given preliminary instructions (seeSubjects section). They were then fitted with earphones.The groups were treated as follows:

Group 1—escape. Subjects received the followinginstructions:

Listen to these instructions carefully. I am not allowedto give you additional information other than whatis given to you now, so please listen and do not askme questions. From time to time a loud tone will bepresented to you. When that tone comes on, there issomething you can do to stop it. Between tones placeyour hands on your knees.

The above instructions, with the exception of the lastsentence, were identical to those used by Hiroto (1974).The button for controlling tone-offset was placed on thetable after the instructions had been read to the subject.

The tones were presented immediately following theinstructions. Each subject received 30 tones with a max-imum duration of 5 sec. Tones could be terminated bya single press of the red button. The intertrial intervalranged from 20 to 55 sec, with a mean duration of40 sec.

Group 2—inescapable. Subjects received the sameinstructions as those of Group 1. They were treated inan identical manner to Group 1 subjects, with the ex-ception that tone presentations were yoked, and buttonpresses, if they occurred, had no effect on tone duration.

Group 3—no-tone. Subjects did not receive anypreexposure treatment. They were given the preliminaryinstructions and then proceeded directly to the testphase.

The first three groups constitute the characteristictreatments of the triadic design of learned helplessnessexperiments. Three additional groups were added to thisdesign.

Group 4—inescapable tone with response. Subjectsin this group received the same tone presentations asthose in Group 1, but were required to press the redbutton as soon as the tone ended. Thus they had to makea response that had no effect on the tone. The instruc-tions to these subjects were the same as to those of Group1 except that the penultimate sentence was changed

to: "Immediately after the tone stops, you are to pressthe button in front of you."

Group 5—inescapable tone with light. Subjects inthis group received the same tone presentations as thosein Group 1. However, immediately following tone offset,the two lights next to the button were illuminated for3.5 sec. Subjects in this group received the followinginstructions: "You are about to hear a series of loudtones. Please remain seated during the entire series."

Group 6—inescapable tone with non-frustrating in-structions. This group was identical to Group 2 exceptfor the instructions, which were replaced by those ofGroup 5.

The test phase was initiated immediately followingthe preexposure phase. All subjects were tested in thesame manner. The hand shuttle-box was placed on thetable in front of the subject. The following instructionswere read to the subject.

You will be given some trials in which a relativelyloud tone will be presented. Now here is the importantpart, and I want you to listen carefully. You can stopthe tone by sliding the knob to the sides in a certainpattern. It's up to you to find the pattern.

These instructions were almost identical to those usedby Hiroto (1974), with the exception that in our in-structions the penultimate sentence was more specific.

The test phase consisted of 18 signaled tone presen-tations. Tones were preceded by a red light that lastedfor 5 sec. The maximum tone duration was 10 sec. Theintertrial interval ranged from 20 to 55 sec, with a meanof 40 sec. On the first trial, the control knob was locatedat the center point of the box. The correct response(which when made during the light would terminate thelight and prevent tone onset, or when made during thetone would terminate the tone) was to slide the knob toone side of the box on one trial and to the opposite sideof the box on the subsequent trial. In other words, afterinitially finding which side was correct, a simple alter-nation of right and left sides would provide optimumperformance scores. If on any trial the subjects did notterminate the tone, a 10-sec escape latency was re-corded.

Three typical dependent measures of learned help-lessness were recorded: (a) the overall mean responselatency for the 18 trials; (b) the number of failures toescape, defined as the number of trials with escape la-tency of 10 sec; and (c) trials to criterion of escapelearning, defined as three consecutive escapes in whichthe first response terminated the tone.

Results and Discussion

The median scores for the three dependentmeasures—number of failures to escape,number of trials to reach the learning cri-terion, and the response latency—for eachof the six groups are presented in Figure 1.As can be seen in Figure 1, all of the specificpredictions, at least in terms of direction ofthe results, were supported. Table 1 presentsthese predictions together with the results

CONFOUNDING OF CONTROLLABILITY 463

510J

8-

6-

4-

2-

__

^

s14:K>

6-

2- |-_

__

1 2 3 4 5 6 1 2 3 4 5 6GROUPS

1 2 3 4 5

Figure 1. Median latency to escape tone, median number of failures to escape tone, and median numberof trials to criterion for learning escape response on the hand shuttle box, Experiment 1. (Groups receivedthe following preexposure conditions: 1 = escapable tone; 2 = inescapable tone, frustrating instructions;3 = no tone; 4 = inescapable tone plus response; 5 = inescapable tone plus light; 6 = inescapable tone,non-frustrating instructions.)

of the tests of reliability. The comparisonsare made against Group 2, the standard in-escapable group of the triadic design, andagainst Group 6, the inescapable group with-out potentially frustrating instructions. Allstatistical comparisons were made with theWilcoxon matched-pair signed-rank test.Matching was done on the basis of subjectsappearing within the same replication, thusequating, within preexposure pairs, for thesame stimulus duration as developed throughthe yoking procedure.

The lack of significant differences between

Table 1Predicted Direction and Significance Levels forthe Three Dependent Measures of the EscapeTask (Experiment I)

Measuref ItfUlVlllMi

of groupperformance

1 <21 <63 < 23 <64 < 24 < 65 <25 <66 = 2

Latency

.05

.05

.025

.05

.025

.025nsnsns

(p>.10)

Failuresto escape

.005

.005

.005

.01

.005

.005

.05

.01ns

(p> .10)

Trials tocriterion

.005

.05

.005

.025

.025ns

.05nsns

(p > .10)

Note. Significance levels are one-tailed.

the two inescapable tone groups (2 and 6)for any of the three dependent measures dis-qualifies those suggestions that learnedhelplessness is, at least in part, a result offrustration-producing instructions. The in-escapable group subjects that were given theinstructions that there was something thatcould be done to escape the tone when infact no solution existed, performed no morepoorly than the group receiving inescapabletone with non-frustrating instructions. Theseresults are not consistent with the conclu-sions of a number of other studies that didfind that the "failure instructions" were nec-essary for producing subsequent interferenceeffects (for a recent review of these studiessee Buchwald, Coyne, & Cole, 1978, pp. 84-86). However, the results are consistent withthose found in the animal literature, where,because of the particular subjects used, theymust remain uninstructed, but still show thesubsequent interference effect. Likewise, theresults are compatible with the theoreticalapproach advanced in the introduction tothis study.

The major point of the present argument,which was presented as an alternative touncontrollability, is that the interference ef-fect is due to the absence of E2. This inter-pretation follows from the formal and logicalanalysis of the triadic design and is irre-spective of the particular mechanism that ispostulated to account for the workings of E2.

464 R. E. LUBOW, R. ROSENBLATT, AND I. WEINER

Clearly, E2 is absent in both the inescapablegroup with frustration instructions and theinescapable group with non-frustration in-structions. As predicted, the simple absenceof E2 is sufficient to produce an interferenceeffect, or conversely, the presence of E2 willattenuate the interference effect.

The results of major importance fromGroups 1, 2, 3, 4, and 5 will be treated inthe general discussion following Experiment2, which replicates Experiment 1.

Experiment 2

Method

SubjectsSixty subjects, 31 female and 29 male, served in the

experiment. Subjects were students from the faculty ofsocial sciences at Tel Aviv University. They were paid20 Israeli shekels for participating in the experiment(approximately 2 US dollars). Five subjects were re-placed because of apparatus failure or aversion to thenoxious tone.

ApparatusThe apparatus for both the preexposure and test

phases was the same as in Experiment 1, with the ad-dition that the test phase included a task with 20 Hebrewanagrams. The words, of approximately equal fre-quency, were chosen from a Hebrew word frequencytable (Balagur, 1968) and were typed individually onindex cards. The scrambled letters were arranged in acommon order, 3-4-2-5-1, so thai in addition to specificsolutions for individual anagrams, a general solutioncould be obtained for all anagrams.

ProcedureThe procedure was similar to that of Experiment 1,

with two exceptions. Group 6 of Experiment 1, whichreceived the non-frustrating instructions and was foundnot to be significantly different from Group 2, wasdropped from the experiment. The five groups in Ex-periment 2, then, were the same as the first five groupsin Experiment 1.

An additional difference was that all subjects weretested with both the shuttle-box and anagram tests. Halfof the subjects of each group were tested first on theshuttle-box and then on the anagrams. The order oftesting was reversed for the other half of the subjects.The instructions for the anagram test were as follows:

You will be asked to solve some anagrams. As youknow, anagrams are words with the letters scrambled.The problem for you is to unscramble the letters sothat they form a word. When you've found this word,tell me what it is. There could be a pattern or principleby which you can solve the anagrams, but that's upto you to figure out. I can't answer any questions now.

These instructions were almost identical to those ofHiroto and Seligman (1975).

All 20 anagrams were soluble and had the same lettersequence. In cases in which the subject gave an incorrectanswer, the experimenter replied "That's not it; pleasetry again."

The dependent variables for the shuttle-box test werethe same as for Experiment 1. Three related measureswere used for the anagram test; (a) response latency;(b) number of failures to solve, defined as the numberof trials with latencies of 100 sec, the time at which thetrial was terminated; (c) trials to criterion, defined asthree consecutive solutions each with less than a 15 seclatency.

Results and Discussion

The median scores for the three dependentmeasures of the escape task, for each of thesix groups, are presented in Figure 2. As inExperiment 1, the results, as displayed inFigure 2, are all in the direction predictedby the specific hypotheses. The data weretreated in an identical manner to those ofExperiment 1. Table 2 presents the predic-tions together with the results of tests of re-liability.

Similarly, Figure 3 presents the medianscores for each of the five groups on the threedependent measures of the anagram test.Unlike the escape task comparisons, the onlyreliable difference was that between Groups2 and 3. The inescapable tone group hadsignificantly more failures to solve the ana-gram problems (p < .025), took more trialsto reach the learning criterion (p < .01), andhad longer latencies to solution (p < .001)than the group that was not preexposed tothe noxious tone. All of the other criticalcomparisons were not reliably different(ps > .05).

A comparison of Figures 1 and 2 and ofTables 1 and 2 indicates that the results ofExperiment 1 were successfully replicated.In both experiments, the typical learnedhelplessness effect was obtained. This isshown by the fact that preexposure to theinescapable tone, as compared to preexpo-sure to the escapable tone or to no-tonepreexposure, resulted in poorer performanceon the subsequent escape task. This was truefor all three dependent measures and sup-ports the findings of other studies that haveused these procedures (e.g., Hiroto, 1974).

Likewise, in both experiments, it wasshown that the abolition of the decrementin the escape task as a result of preexposureto escapable tone was not necessarily dueto the subjects' ability to control tone offset.

CONFOUNDING OF CONTROLLABILITY 465

fcK>-

7_

1 2 3 4 5

gifr

"" 8-2

Q KUJ D'

4-

2-

0

1 2 3 4 5G R O U P S

1 2 3 4 5

Figure 2. Median latency to escape tone, median number of failures to escape tone, and median numberof trials to criterion for learning escape response on the hand shuttle box, Experiment 2. (Groups receivedthe following preexposure conditions: 1 = escapable tone; 2 = inescapable tone; 3 = no tone; 4 =inescapable tone plus response; 5 = inescapable tone plus light.)

Subjects who were told to press the buttonafter the tone was terminated (Group 4) alsoshowed significantly less disruption on theescape test than the group preexposed to in-escapable tone. Again, this was true of allthree measures. Thus, a comparison of thegroups exposed to the escapable tone and thegroups exposed to the tones followed by thenon-controlling response indicated no signif-icant differences (ps > .10), confirming thatthe important property of the response in theescapable group is not one of control of toneoffset.

It was suggested, for reasons stated in theintroduction, that any event following thetone during preexposure would serve to pre-vent the learned helplessness effect. A com-parison of Group 5, in which a light followedtone offset, and the inescapable tone group(Group 2) only partially supports this con-

Table 2Predicted Direction and Significance Levels forthe Three Dependent Measures of the EscapeTask {Experiment 2)

Predictionof group

performance

1 <23 < 24 < 25 <2

Latency

.005

.025

.025ns

Measure

Failuresto escape

.01

.005

.005

.025

Trials tocriterion

.10'

.005

.005

.025

tention. As can be seen in Figures 1 and 2,in both experiments and for all three depen-dent measures, the results were in the pre-dicted direction. However, only two of thethree comparisons (number of failures toescape and trials to criterion) were reliablysignificant.1 Nevertheless, because of theconsistency in direction and the fact that wehave reliably obtained such an effect withrats (Lubow, Weiner, Lindenberg, & Mar-golit, Note 2), we suspect that the effect isindeed real. One possible reason for the rel-atively weak appearance of the effect wasthe nature of the stimulus that followed toneoffset—a rather small, dim light that waspresented in a normally lit room. In the par-allel study with rats, we used a very brightlight in a dark chamber.

The data from the anagram test in Ex-periment 2 are somewhat more difficult tointerpret. On the one hand, the comparisonof the preexposure to inescapable tone groupwith the no-tone group shows the relativelypoor performance that has been previouslyreported for the inescapable group (e.g.,Hiroto & Seligman, 1975; Miller & Selig-man, 1975). However, even though the datawere in the appropriate direction, there wereno significant differences between the escap-able and inescapable groups on any of themeasures. Since these critical differences

' See Footnote 1.

1 While 9 out of 12 subjects responded in the appro-priate direction the failure to obtain a more reliabledifference was due to the fact that one of the threedeviant subjects never reached criterion.

466 R. E. LUBOW, R. ROSENBLATT, AND I. WEINER

I<2-

1 2 3 4 5 1 2 3 4 5G R O U P S

1 2 3 4 5

Figure 3. Median latency, median number of failures, and median trials to criterion for anagram prob-lems, Experiment 2. (Groups received the following preexposure conditions: 1 = escapable tone; 2 =inescapable tone; 3 = no tone; 4 = inescapable tone plus response; 5 = inescapable tone plus light.)

were not obtained, there is no reason on thebasis of our theorizing to expect either thegroups preexposed to tone-plus-response orto tone-plus-light to show significantly betterperformance than the inescapable tone group.Indeed, they did not perform reliably dif-ferent from the inescapable tone group norfrom the escapable tone group. The datafrom the anagram test, then, cannot be usedone way or the other in answering the ques-tion about the role of uncontrollability forproducing generalized learned helplessness.

As to why Experiment 2 failed to providethe often reported differences betweenpreexposure to escapable and inescapabletone on the subsequent anagram test, onecan only speculate. The subjects (Hebrew-speaking Israelis), seemed very apt at solvingthe Hebrew anagrams, as compared to, forinstance, Hiroto and Seligman's Americansubjects who solved English anagrams. Themean time to solution across the three basicgroups of the triadic design in the presentstudy was 18.8 sec, whereas in the Hirotoand Seligman study it was 28 sec. The dif-ference might well be accounted for by thethree-letter root structure of Hebrew words,which would tend to facilitate anagram so-lutions. This suggests that, as a result of theanagram test being too easy, otherwise re-liable group differences were obscured. Itshould be noted that another learned help-lessness study also failed to find significantdifferences with Hebrew anagrams (Rosen-baum & Schichman, 1979).

In general, the conditions under whichhelplessness effects generalize to situations

that are different from the one in which theywere induced, and what determines whetherthese effects are transient or persist overtime, are not clear (Wortman & Dintzer,1978; Abramson et al., 1978). Abramson etal. proposed that the chronicity of helpless-ness would depend on the attribution thatsubjects made about the cause of failure: Aglobal attribution implies that helplessnesswill occur across situations, whereas a spe-cific attribution implies helplessness only inthe original situation. In their discussion ofstudies that obtained generalization of help-lessness, Abramson et al. (1978) stated that"the expectation for the inescapability of thenoise . . . must have been global enough totransfer across situations" (p. 60). Appar-ently, in the present experiment, such ex-pectation was not global enough. The reasonfor such a difference is not clear. Maybe wetapped some cultural aspects of helplessness(i.e., Israeli students may be less prone togeneralization of helplessness than theirAmerican counterparts).

General Discussion

The conclusions that follow from the ob-tained results are, quite simply, that con-trollability as represented by a response thatterminates a noxious stimulus is a sufficientbut not a necessary condition for attenuatingthe interference effects produced by the ab-sence of such controllability. Therefore, log-ically, the absence of controllability (uncon-trollability) is not the critical variable forproducing the interference effects. Our ex-

CONFOUNDING OF CONTROLLABILITY 467

periments demonstrate that interference ef-fects may be reduced by either (a) a responsethat controls tone-offset, (b) a response thatoccurs after tone-offset but is in close tem-poral proximity to the tone, (c) a light thatoccurs after tone-offset but is in close tem-poral proximity to the tone. Since the lasttwo conditions are not related to control, itfollows that the absence of control relativeto condition (a) is not the variable that isprimarily responsible for the interferenceeffect.

One might argue that there are two orthree distinctly different variables—as forexample uncontrollability for (a) and un-predictability for (c)—that account for theinterference effect. However, in addition tolacking parsimony, this would seem to bequite unreasonable, as it is the one conditionof tone-alone that produces the subsequentinterference effect. The demonstration thatthere are at least three different conditionsattenuating the effect suggests that thesethree have some property in common, andthat the absence of this common property(rather than the absence of any one of theoperational conditions, such as control) pro-duces the interference effect. Lack of con-trollability is a sufficient but not a necessarycondition, and therefore does not provide agenerally valid description of the causal con-ditions for producing the decremental effectsassociated with the induction procedure ofthe learned helplessness paradigm.

What then is the broader descriptive cat-egory that has as exemplars stimulus-con-trolling responses, responses following targetstimuli that do not control them, and stimulithat simply follow target stimuli? At onelevel of description they all represent thegeneral class of E,-E2: the target event, EI,being followed by any other event, E2 (whichmay be a controlling response, instructedresponse, or stimulus). Thus, at one descrip-tive level, we have a theory of interferenceeffects based on the absence of E2. On an-other, deeper, level one might ask how E2exerts its effect, or conversely, how the ab-sence of E2 exerts its effect. In the intro-duction to this article we offered one possibleapproach based on concepts of presence ver-sus absence of contingency among events.This approach deals with the generalized EI-E2 episode (disregarding distinctions be-

tween controlling responses, or other re-sponses and stimuli that are contingent toEI) in describing the common effective E2characteristics.

Two mechanisms have been proposed toexplain the effects of E2: memorial and at-tentional. According to the former, E2 servesto maintain the Et representation in short-term memory active, and thus to enhance itsrehearsal (Wagner, 1978). According to thelatter, E2 serves to maintain attention to EIthat would otherwise be ignored (Lubow,Weiner, & Schnur, 1981; Dickinson, Hall,& Mackintosh, 1976). Both the memorialand the attentional mechanisms are stimu-lus-specific (i.e., they refer to the effects ofE2 on EI and only EI). Therefore, the me-morial and attentional models would notpredict that the E!-E2 episode would atten-uate LH in a new situation.

The generalization test (using anagrams)might have provided a further understandingof E2 effect had it yielded the basic differ-ence between the group controlling tone off-set and the group not controlling tone offset.To further clarify these issues, a replicationof Experiment 2 is required that is sensitiveenough to detect the interference effect inthe generalization test when the response-contingent group is compared to the groupwith no contingency. The predictions fromthe two above approaches (memorial andattentional) are unequivocal. The additionof E2 in a conditioning relationship to E,should attenuate the interference effect ina test employing the same EI (as in Exper-iments 1 and 2), but should have no effecton a dissimilar test (as suggested in Exper-iment 2), Such findings would encourage theproposal that the typical "learned helpless-ness" results derive from two separate pro-cesses: one which is associative (stimulusspecific) and can be treated in the samemanner as latent inhibition; and the other,which is non-associative (generalizable tostimuli other than those that are preex-posed). Such a proposal would contend thatthose "learned helplessness" studies that usethe same E, in preexposure and test are ob-taining results from the combined effects ofthe associative and non-associative pro-cesses, whereas those generalization teststhat use a different EI in preexposure andtest are only tapping the effects of the non-

468 R.E. LUBOW, R. ROSENBLATT, AND I. WEINER

associative process. Generalization tests maythus bypass the attentional and/or memorialprocesses that have been proposed in thepreceding paragraph as possible mediatorsof learned helplessness and latent inhibition.

As a final point, it has been suggested thatcontrollability, as defined by the operationof response-outcome contingency, is an in-stance of a broader category of event con-tingency. However, the reverse might alsobe the case: the subject's exposure to event-contingency may constitute a condition forthe acquisition of perceived control. Recentanimal theories of learning conceive of theorganism as a detector and seeker of event-contingencies or inter-event relationships.The existence of inter-event relationshipsappears by itself to be reinforcing. This rein-forcing property may relate to the organ-ism's perception of control when faced withcontingency among events.

Reference Notes

1. Szakmary, G. A. Latent inhibition with sequentiallyexposed material: More data. Paper presented at themeeting of the Psychonomic Society, San Antonio,Texas, November 1978.

2. Lubow, R. E., Weiner, I., Lindenberg, Y., & Mar-golit, C. The confounding of controllability inlearned helplessness studies with rats. Unpublishedmanuscript, 1979. Available from R. E. Lubow, TelAviv University, Ramat Aviv, Tel Aviv, Israel 420111.

ReferencesAbramson, L. Y., Seligman, M. E. P., & Teasdale,

J. D. Learned helplessness in humans: Critique andreformulation. Journal of Abnormal Psychology,1978,57,49-74.

Baker, A. G. Learned irrelevance and learned helpless-ness: Rats learn that stimuli, reinforcers, and re-sponses are uncorrelated. Journal of ExperimentalPsychology: Animal Behavior Processes, 1976, 2,130-141.

Balagur, R. List of basic words for school children. Tel-Aviv: Otzar Ha'Moreh, 1968.

Buchwald, A. M., Coyne, J. C., & Cole, C. S. A criticalevaluation of the learned helplessness model ofdepression. Journal of Abnormal Psychology, 1978,87, 180-193.

Dickinson, A., Hall, G., & Mackintosh, N. J. Surpriseand the attenuation of blocking. Journal of Experi-mental Psychology: Animal Behavior Processes, 1976,2. 313-322.

Hiroto, D. S. Learned helplessness and locus of control.Journal of Experimental Psychology, 1974, 102,187-193.

Hiroto, D. S., & Seligman, M. E. P. Generality oflearned helplessness in man. Journal of Personalityand Social Psychology, 1975, 31, 311-327.

Lubow, R. E. Latent inhibition. Psychological Bulletin,1973, 79, 398-407.

Lubow, R. E., Alek, M., & Arzy, J. Behavioral dec-rement following stimulus preexposure: The effectsof number of preexposures, presence of a second stim-ulus, and interstimulus interval in children and adults.Journal of Experimental Psychology: Animal Be-havior Processes, 1975, /, 178-188.

Lubow, R. E., Schnur, P., & Rifkin, B. Latent inhibitionand conditioned attention theory. Journal of Exper-imental Psychology: Animal Behavior Processes,1976, 2, 163-174.

Lubow, R. E., Weiner, E., & Schnur, P. Conditionedattention theory. In G. H. Bower (Ed.), The psy-chology of learning and motivation. New York: Ac-ademic Press, 1981.

Mackintosh, N. J. The psychology of animal learning.London: Academic Press, 1974.

Maier, S. F., & Seligman, M. E. P. Learned helpless-ness: Theory and evidence. Journal of ExperimentalPsychology: General, 1976, 105, 3-46.

Miller, W. R., & Seligman, M. E. P. Depression andlearned helplessness in man. Journal of AbnormalPsychology, 1975, 84, 228-238.

Overmier, J. B., & Seligman, M. E. P. Effects of in-escapable shock upon subsequent escape and avoid-ance learning. Journal of Comparative and Physio-logical Psychology, 1967, 63, 28-33.

Rosenbaum, M., & Schichman, S. Learned helplessnessand depression among Israeli women. Journal of Clin-ical Psychology, 1979, 55, 395-400.

Seligman, M. E. P. Helplessness: On depression, de-velopment, and death. San Francisco: Freeman, 1975.

Seligman, M. E. P., & Maier, S. F. Failure to escapetraumatic shock. Journal of Experimental Psychol-ogy, 1967, 74, 1-9.

Seligman, M. E. P., Maier, S. F., & Solomon, R. L.Unpredictable and uncontrollable aversive events. InF. R. Brush (Ed.), Aversive conditioning and learn-ing. New York: Academic Press, 1971.

Szakmary, G. A. Some observations concerning condi-tioned attention theory. Bulletin of the PsychonomicSociety, 1977, 9, 142-144.

Turner, L. H., & Solomon, R. L. Human traumaticavoidance learning: Theory and experiments on theoperant-respondent distinction and failures to learn.Psychological Monographs, 1962, 76 (40, Whole No.559).

Wagner, A. R. Expectancies and the priming of STM.In S. H. Hulse, H. Fowler, & W. K. Honig (Eds.),Cognitive processes in animal behavior. Hillsdale,N.J.: Erlbaum, 1978.

Wickens, D. D., Tuber, D. S., Nield, A. F., & Wickens,C. Memory for the conditioned response: The effectsof potential interference induced before and afteroriginal conditioning. Journal of Experimental Psy-chology: General, 1977, 106, 47-70.

Wortman, C. B., & Dintzer, L. Is an attributional anal-ysis of the learned helplessness phenomenon viable?A critique of the Abramson-Seligman-Teasdale re-formulation. Journal of Abnormal Psychology, 1978,87, 75-90.

Received August 12, 1980Revision received March 12, 1981