causal explanations as a risk factor for depression: theory ......depression with respect to...

Psychological Review1984, Vol. 91, No. 3, 347-374

Copyright 1984 by theAmerican Psychological Association, Inc.

Causal Explanations as a Risk Factor for Depression:Theory and Evidence

Christopher PetersonVirginia Polytechnic Institute and

State University

Martin E. P. SeligmanUniversity of Pennsylvania

The attributional reformulation of the learned helplessness model claims that anexplanatory style in which bad events are explained by internal, stable, and globalcauses is associated with depressive symptoms. Furthermore, this style is claimedto be a risk factor for subsequent depression when bad events are encountered.We describe a variety of new investigations of the helplessness reformulation thatemploy five research strategies: (a) cross-sectional correlational studies, (b) longi-tudinal studies, (c) experiments of nature, (d) laboratory experiments, and (e) casestudies. Taken together, these studies converge in their support for the learnedhelplessness reformulation.

After experiencing uncontrollable events, avariety of organisms show cognitive, motiva-tional, and emotional deficits (Maier & Selig-man, 1976; Seligman, 1975). This learnedhelplessness phenomenon has parallels withdepression in people and has been proposedas a model of this psychopathology (Seligman,1974). When helplessness theory proved un-able to account for the generality and chro-nicity of depressive symptoms or for self-es-teem loss in depression, it was revised alongattributional lines (Abramson, Seligman, &Teasdale, 1978). According to this reformu-lation, the boundary conditions of depressionfollowing bad events are determined by causalattributions about the events (for similar for-mulations, see Miller & Norman, 1979, andRoth, 1980). The central prediction of the re-formulation is that individuals who have anexplanatory style that invokes internal, stable,and global causes for bad events tend to be-come depressed when bad events occur.

This research was supported by U.S. Public Health Ser-vice Grant MH-19604 to Martin E. P. Seligman.

The authors thank the many individuals at the Universityof Pennsylvania and elsewhere who contributed to thisresearch, the anonymous reviewers for their comments onan earlier version of this article, and James Austin, PeterVillanova, and Steve Zaccaro for their help in polishingthe final draft.

Requests for reprints should be sent to Martin E. P.Seligman, Department of Psychology, University of Penn-sylvania, 3815 Walnut Street, Philadelphia, Pennsylvania19104.

Our primary purpose here is to describe theempirical investigations of this theory con-ducted by our research group, and to evaluatethe theoretical status of the attributionalmodel. The juxtaposition of conceptual clar-ification with evidence from several levels ofanalysis may allow a valid assessment of thetheory.

Theory

Helplessness Theory: Old and New

The learned helplessness phenomenon wasfirst described systematically by animal learn-ing researchers at the University of Pennsyl-vania (Overmier & Seligman, 1967; Seligman& Maier, 1967). Mongrel dogs, following ex-posure to inescapable electric shock, showedstriking deficits 24 hours later when placed ina shuttlebox in which the simple act of crossinga barrier would terminate shock. Unlike dogsnot previously exposed to uncontrollableshock, these animals seemed helpless. Theyinitiated few attempts to escape the shock(motivational deficit). They were not likely tofollow an occasionally successful response withanother (learning or cognitive deficit). Theydid not evidence much overt emotionalitywhile being shocked (emotional deficit).

These deficits were interpreted in cognitiveterms (Maier, Seligman, & Solomon, 1969; Se-ligman, Maier, & Solomon, 1971). During ex-posure to the electric shocks, the dogs learnedthat shocks were independent of responses.

347

348 CHRISTOPHER PETERSON AND MARTIN E. P. SELIGMAN

Regardless of what the dogs did or did not do,the shocks occurred. This learning was rep-resented as an expectation of future response-outcome independence (i.e., uncontrollability)that was generalized to new situations to pro-duce the observed deficits.

The cognitive interpretation of animalhelplessness has been controversial. Some re-searchers have attempted to explain the phe-nomenon in peripheral or biological terms,whereas other researchers have argued for thecognitive explanation (for reviews of this con-troversy, see Maier & Jackson, 1979; Maier &Seligman, 1976; Seligman & Weiss, 1980). Re-gardless of this controversy, psychologists in-terested in human adaptation were quick tosense the possible pertinence of learned help-lessness to failures of human action. Two par-allel lines of research with human subjects tookplace.

In the first, the basic helplessness phenom-enon was investigated in the laboratory withhuman subjects (for a review, see Wortman &Brehm, 1975). In the second, helplessness the-ory was used to explain a variety of humandifficulties (see Garber & Seligman, 1980).Perhaps the best known of these applicationshas been Seligman's (1972, 1974, 1975) sug-gestion that learned helplessness may modeldepression with respect to symptoms, causes,preventions, and cures.

The main problem with the original help-lessness model, applied both to human help-lessness in the laboratory and to natural hu-man depression, is its failure to account forboundary conditions. Sometimes laboratoryhelplessness is general (e.g., Hiroto & Selig-man, 1975), and sometimes it is circumscribed(e.g., Cole & Coyne, 1977). Sometimes badevents precipitate depressive reactions (occa-sionally transient, occasionally long lasting),and sometimes they do not (e.g., Brown &Harris, 1978; Lloyd, 1980). What determinesthe chronicity and generality of helplessnessand depression?

Similarly, the original model does not ex-plain the self-esteem loss frequently observedamong depressives (Beck, 1967; Freud, 1917/1957). Why should individuals blame them-selves for events over which they perceive nocontrol (Abramson & Sackeim, 1977)? Thesimple helplessness model is silent about thechronicity and generality of helplessness and

depression and about the paradox of self-es-teem loss following helplessness.

To address these shortcomings, Abramson,Seligman, and Teasdale (1978) revised help-lessness theory to include the individual'scausal explanations of the original bad events.According to this revision, when people faceuncontrollable bad events, they ask why. Theiranswer affects how they react to the events.Abramson et al. (1978) argued that three ex-planatory1 dimensions are relevant. First, thecause may be something about the person (in-ternal explanation), or it may be somethingabout the situation or circumstances (externalexplanation). Second, the cause may be a factorthat persists across time (stable explanation),or it may be transient (unstable explanation).Third, the cause may affect a variety of out-comes (global explanation), or it may be lim-ited just to the event of concern (specific ex-planation). Table 1 presents examples of thesetypes of explanations as they might be madeabout the bad event "My checking account isoverdrawn."

The reformulation assigns particular rolesto each of these three dimensions. Internalityof causal beliefs affects self-esteem loss follow-ing bad events. If the person explains a badevent by an internal factor, then self-esteemloss is more likely to occur. If a person explainsthe event by an external factor, then self-esteemloss is less likely to occur. Stability of causalbeliefs affects the chronicity of helplessnessand depression following bad events. If a badevent is explained by a cause that persists,depressive reactions to that event tend to per-sist. If the event is explained by a transientfactor, then depressive reactions tend to beshort lived. Finally, globality of causal beliefsinfluences the pervasiveness of deficits follow-ing bad events. If one believes that a globalfactor has caused a bad event, then helplessness

1 Because of the numerous and sometimes contradictorysenses in which the terms attribution and altributionalstyle are used, we here refer to causal explanation andexplanatory style. In particular, attribution by dictionarydefinition is a broader term than is explanation, and in-volves imputing any property to an object. Because theonly property that concerns us here is causal characteristics,we prefer explanation to attribution. Although this shiftin terminology may be confusing in the short run, webelieve that clarity will eventually be served by the use ofthese more precise terms.

CAUSAL EXPLANATIONS AND DEPRESSION 349

Table 1Examples of Causal Explanations for the Event"My Checking Account is Overdrawn"

Explanation

Style Internal External

StableGlobal

Specific

UnstableGlobal

Specific

"I'm incapable ofdoing anythingright"

"I always havetrouble figuringmy balance"

"I've had the flufor a few weeks,and I've leteverythingslide"

"The one time Ididn't enter acheck is theone time myaccount getsoverdrawn"

"All institutionschronically makemistakes"

"This bank has alwaysused antiquatedtechniques"

"Holiday shoppingdemands that onethrow oneself intoit"

"I'm surprised — mybank has nevermade an errorbefore"

deficits tend to occur in a variety of differentsituations. If one believes that a more specificfactor is the cause, the deficits tend to be cir-cumscribed.

Conceptual Status of Explanationand Explanatory Style

Causal explanations: Sufficient condition orrisk factor for depression? We used qualifiedlanguage above in stating the central predic-tions of the reformulation. So, an internal ex-planation for a bad event is said to make self-esteem loss more likely, but not to cause self-esteem loss. It is important to realize thatexplanations and their precursor, explanatorystyle, are not sufficient to produce depressivedeficits but rather are risk factors for such def-icits (Abramson et al., 1978).

The relationship among variables in thetheory is diagrammed in Figure 1. Let us beginby looking at the extreme right-hand side ofthis figure. What Figure 1 attempts to explainis the general process by which the symptomsof helplessness are produced. These symptomsare passivity; cognitive deficits; emotional def-icits including sadness, anxiety, and hostility;a lowering of aggression; a lowering of appe-titive drives; a set of neurochemical deficits;

and an increase in susceptibility to disease. Inaddition, the symptom of self-esteem loss issometimes one of the symptoms of helpless-ness. These symptoms, taken together, lookvery much like the syndrome of depression.How do these symptoms come about?

In learned helplessness theory, both originaland reformulated, the expectation that no ac-tion will control outcomes in the future is asufficient condition for the production of allof these symptoms of helplessness except self-esteem loss. This expectation is representedin the center of Figure 1. Whenever and wher-ever this expectation occurs, the symptomswill develop. Thus, our main concern is howother processes and events conspire to bringabout this expectation, resulting in the symp-toms of helplessness. This expectation is usu-ally triggered when bad events are perceivedas uncontrollable. The reality of the bad un-controllable events influences the content ofthe expectation. So, for example, if the badevent is blindness, and one's job is proofread-ing, the range of outcomes to which the ex-pectation applies includes work.

Explanations and explanatory style also in-fluence the expectation that no action willcontrol outcomes in the future (see Figure 1).Explanations involving global causes tend toproduce the expectation that action will notcontrol many outcomes, which in turn pro-duces the symptoms of helplessness in exactlythat large range of situations. In parallel, if thecause of a bad event is explained by stablefactors, the expectation tends to occur for along time into the future, and therefore, thesymptoms of helplessness are long-lasting. Ifthe explanation for a bad event is internal,then the symptom of lowered self-esteem tendsto be displayed. Thus, the particular expla-nation an individual makes for the bad eventinfluences the generality and time course ofthe symptoms of helplessness, as well as theloss of self-esteem.

In the model, there are two influences onthe particular explanation chosen. The first isthe reality of the bad events themselves. If thebad event that sets off the expectation of help-lessness is the death of one's spouse, this is astable and global loss. The spouse will not re-turn, and many of the activities in which onehas customarily engaged will be undermined.The second influence on what particular ex-


Passivity

CognitiveDeficits

Bad Events Perceived As Uncontrollable

Expectation That No ActionWill Control

Figure 1. The process of learned helplessness.

planation is made is the habitual tendency tochoose certain kinds of explanations for badversus good events. We have been able to iden-tify individual patterns in the selection ofcauses over a variety of events, and we callthese patterns explanatory style. The particularstyle that most concerns us is the depressiveexplanatory style, in which one tends to giveinternal, stable, and global explanations forbad events (it's me; it's going to last forever;and it's going to affect everything I do).

It should now be apparent why a particularexplanation or explanatory style is not suffi-cient for the symptoms of helplessness to ap-pear. These variables influence the expectation,but it is the expectation which is sufficient.Usually, causal explanations for an event andexpectations about the consequences of anevent have the same properties. For example,if the explanation for blindness is a progressivebrain disease, this cause has stable and globalproperties, as do the consequences of blind-ness. But sometimes the properties of a causeand its consequences can be dissimilar. If, forexample, the cause of blindness was a freakaccident, the cause is unstable and specific,but the consequences are stable and global.

Because there is usually similarity betweencausal explanation and expectation of con-sequences, knowing an individual's explana-tion and explanatory style will usually predicthelplessness deficits. Because there can some-times be dissimilarity, an individual's causalexplanation and explanatory style will notcause the deficits. And so we speak of thesevariables as risk factors (just as smoking is arisk factor for lung cancer), rather than suf-ficient conditions for helplessness and depres-sion.

Our research focused on measuring ex-planatory style, because a methodologicalframework already existed for doing so (e.g.,Harvey, Ickes, &Kidd, 1976, 1978, 1981). Wechose not to focus on the expectation itself,even though it is more proximate to the symp-toms, because we do not believe that a validmeans of measuring expectations yet exists.Also, in most cases, we chose not to focus onparticular explanations because explanatorystyle of necessity is a more reliable individualdifference (Epstein, 1980).

Causal explanations and explanatory style:Hypothetical constructs. The term attributionhas been used in a variety of ways. Some the-


orists regard attributions as the stated beliefsthat compose an individual's naive psychology(Heider, 1958), whereas other theorists suggestthat attributions have tacit components (Lan-ger, 1978; Wortman & Dintzer, 1978). Con-troversy over how best to assess attributionshas ensued (Nisbett & Wilson, 1977). We wishto make explicit that we regard an attributionas a hypothetical construct, a way for the the-orist to explain observable behaviors. We donot use the notion to refer to an interveningvariable (cf. MacCorquodale & Meehl, 1948).

As a hypothetical construct, causal expla-nation does not have a single definition noris it exhausted by any one operation. An in-tervening variable is exhausted in its meaningby one class of input operations and one classof output operations. Our view of attributionscontrasts with Heider's (1958) phenomeno-logical view of attributions, in which the abil-ities to introspect upon and to report verballyone's causal beliefs are necessary and sufficientconditions for them. Heider (1958) regardedattributions as intervening variables. For us,on the other hand, an attribution is not a realthing, like a microphone or a typewriter.Rather, it is akin to concepts like natural se-lection, life, reward, preference, or—perhaps—the atom. These are all hypothetical constructs,measured with a number of different con-verging operations, no one of which definesor exhausts the construct of interest. By thislogic, not only introspections and answers toquestionnaires are relevant to knowing aboutone's causal explanations, but so too are avariety of behavioral observations. In thestudies we report, we attempt to use convergingways to infer the presence of causal expla-nations.

Attributional Style Questionnaire. One wayin which we measure explanations and ex-planatory style is with a questionnaire that wedeveloped for this purpose: the AttributionalStyle Questionnaire (ASQ; Peterson, Semmel,et al., 1982). This self-report instrument yieldsscores for the explanation of bad and goodevents with internal versus external, stableversus unstable, and global versus specificcauses. The format reflects the fact that wewanted questions to assess how much respon-dents used particular values of the three di-mensions. We ask subjects to generate theirown cause for each of a number of events, andthen to rate that cause themselves along 7-

point scales corresponding to the internality,stability, and globality dimensions. The ASQdoes not constrain or create the causal expla-nations provided by the subject, but at thesame time it allows simple and objective quan-tification of responses by asking the subject torate the internality, stability, and globality ofthe causes.

The questionnaire is group administered,and the following directions appear on the firstpage of the booklet:

Please try to vividly imagine yourself in the situationsthat follow. If such a situation happened to you, whatwould you feel would have caused it? While events mayhave many causes, we want you to pick only one—themajor cause if this event happened to you. Please writethis cause in the blank provided after each event. Next wewant you to answer some questions about the cause. Tosummarize we want you to:

1. Read each situation and vividly imagine it happeningto you.

2. Decide what you feel would be the major cause ofthe situation if it happened to you.

3. Write one cause in the blank provided.4. Answer three questions about the cause.5. Go on to the next situation.

Because we are interested in style—cross-sit-uational explanations—we describe 12 differ-ent hypothetical events. Half are good events(e.g., you meet a friend who compliments youon your appearance), and half are bad events(e.g., you go out on a date, and it goes badly).After each event are parallel questions. First,the subject is asked to write down the onemajor cause of the event. Then the subject isasked to rate the cause along the three ex-planatory dimensions. The wording of thevarious questions reflects the specific event tobe explained, and the example in Table 2 il-lustrates the nature of these questions.

The three ratings of each cause are scoredin the directions of increasing internality, sta-bility, and globality. Total scores are formedseparately for the bad and good events simplyby summing the appropriate items and divid-ing the sum by 6. Table 3 reports the internalconsistencies, intercorrelations, and test-retestreliabilities of these scales for a sample of un-dergraduates at the State University of NewYork at Stony Brook (Peterson, Semmel, etal., 1982).

Several things should be noted about thesedata. First, the individual scales have modestreliabilities, ranging between .44 and .69. Onereason for these low reliabilities is the small


number of items—six—in each scale.2 Second,the scales are substantially intercorrelatedwithin good events and within bad events. Forthis reason, we usually combine them intooverall composites for good events and badevents, thereby bolstering reliability to moreacceptable levels (alphas of .75 and .72, re-spectively). Third, causal explanations for goodevents are independent of causal explanationsfor bad events. Conceptions that confound thetwo (e.g., Rotter, 1966) incorrectly average overtwo quite different phenomena. Fourth, thestability of explanatory style scores is respect-ably high, as we had hoped in assuming thatthey operationalize a characteristic style ofcausal explanation.

Criterion validity of the ASQ. What is thevalidity evidence for this questionnaire? Wepresent support for the construct validity be-low. However, two recently completed studiesspeak to the criterion validity of the ASQ: thedegree to which the questionnaire predictsnaturally occurring causal explanations. Do

Table 2Sample Question From the AnributionalStyle Questionnaire

Event: You go out on a date, and it goes badly.

1. Write down the one major cause of this event.

2. Is the cause of the date going badly due to somethingabout you or something about other people orcircumstances?Due to Due

other topeople 1 2 3 4 5 6 7 m e

(Circle one number.)

3. In the future when dating, will this cause again bepresent?Never Always be

again presentpresent 1 2 3 4 5 6 7


4. Is the cause something that just influences dating ordoes it also influence other areas of your life?Just this All

situa- situa-tion 1 2 3 4 5 6 7 tions


Note. From Peterson, Semmel, et al., (1982).

answers on the ASQ reflect the natural, un-constrained causal explanations an individualmakes?

To test if the scores obtained with the ASQare more than questionnaire artifacts, we asked66 adults (31 males, 35 females) on the Uni-versity of Pennsylvania campus to describe inwriting, using 250 to 300 words, the two worstevents involving themselves during the lastyear (Peterson, Bettes, & Seligman, 1982). In-structions stressed descriptive accuracy, andthere was no mention of causal explanation.The subjects then completed the ASQ, whichhad not been mentioned prior to this point.This procedure allowed us to see (a) if causalaccounts are spontaneously included in de-scriptions; (b) if individuals are consistentacross different explanations in terms of theirinternality, stability, and globality; and (c) ifthese spontaneous explanations convergedwith scores of the ASQ.

We had first developed a procedure by whichspontaneous explanations in written materialcould be extracted and scored for internality,stability, and globality (Peterson, Luborsky, &Seligman, 1983). A researcher reads the ma-terial and notes descriptions of bad events in-volving the writer for which a causal expla-nation is made (identified by such key phrasesas "because of," "as a result of," and so on).Next, each event description and the accom-panying explanation are transcribed verbatimand are shown singly to independent and blindjudges who rate the explanation with 7-pointscales corresponding to its internality, stability,and globality. These judges use the definitionsfrom the ASQ (Peterson, Semmel, et al., 1982).Finally, the ratings by the different judges arecombined into a profile for the individual.

For this particular study, there were fourjudges. Coding reliabilities, estimated byCronbach's (1951) alpha, were satisfactory:.93, .89, and .90, for the internal, stable, andglobal ratings, respectively. Table 4 presentsthe ratings made by the judges for this excerptfrom a written description (emphasis added):

About four months ago, he called me on the telephonefrom , where he's been working. He told me thatour relationship was over, that he didn't want to see me

2 For a newly revised version of the ASQ with 24 badevents, reliabilities are substantially improved: .66 for in-ternality, .85 for stability, and .88 for globality.


Table 3Reliabilities and Intercorrelations of the Attributional Style Questionnaire

Dimension 1

Good events1. Internality2. Stability3. Globality4. Composite

Bad events5. Internality6. Stability7. Globality8. Composite

MSDTest-retest correlations

(.50).62*.38*

.11-.17-.15

5.26.79.58*

(.58).59*

.01-.07

.04

5.36.68.65*

(.44)

-.03.03.24*

5.11.80.59*

(.75)

.02

5.25.62.70*

(.46).18*.28*

4.29.84.64*

(.59).45*

4.14.71.69*

(.69)

3.871.07.57*

(.72)

4.12.64.64*

Note. N = 130. From Peterson, Semmel, et al., (1982). Figures in parentheses, on diagonal, are reliabilities estimatedby Cronbach's (1951) alpha. Test-retest correlations over 5 weeks (n = 100).* p < .05.

anymore. I felt devastated. I tried to argue, but what couldI say? I'm still flipped out. I guess I'm just no good atrelationships. I've never been able to keep a man interestedin me. I talked to my roommate about it all, right afterhe called me. She knows me real well, plus she went tohigh school with She told me to forget abouthim and find someone else. She said I made a mistakeand the best thing to do was move on. There was never achance for a lasting thing. But how could I misjudge thingsso badly? When I get stars in my eyes I get carried away.

All of the descriptions contained at leastone causal explanation. We concluded thatcausal explanations are spontaneously offered,even when they are not explicitly prompted.Ratings of the first explanation offered by anindividual were correlated with those of thesecond, and significant correlations were ob-tained. Furthermore, when these ratings werecombined across explanations by the same in-dividual, they correlated respectably with thecorresponding dimensions of the ASQ. Exactvalues are reported in Table 5.

We carried out a second study to see if thecontent analysis of explanatory style was validin patient populations as well as in normalpopulations (Castellon, Ollove, & Seligman,1982). Toward this end, we gave 40 outpatientsthe Schedule for Affective Disorders andSchizophrenia (SADS) diagnostic interview(Spitzer & Endicott, 1977) in the same sessionas we gave them the ASQ and the BeckDepression Inventory (BDI; Beck, 1967).These patients were diagnosed unipolar de-pressive episode by this interview. The first

two questions of the SADS interview ask thepatient (a) why treatment was sought and (b)the origin of their symptoms. We performeda blind content analysis—exactly as is de-scribed above—of the causal explanations of-fered in answer to the first two questions ofthe SADS interview.

Content analysis seems valid for a severelydepressed population. The composite score ofthe ASQ for bad events correlated significantlywith the composite score for bad events ob-tained by this content analysis (r - .38, p <.02). Furthermore, explanatory style, mea-sured in both ways, correlated with severity

Table 4Example of Content Analysis

Event explanation Judge Int Sta Glo

Our relationship was over.(because) I'm just nogood at relationships.I've never been able tokeep a man interestedin me.

How could I misjudgethings so badly?(because) When I getstars in my eyes I getcarried away.

ABCD

ABCD

7776

5554

7777

6677

2331

3223

Note. From Peterson, Bettes, and Seligman (1982). Int :

internality, Sta = stability, and Glo = globality.


Table 5Spontaneous Explanations

Dimension M SD Consistencyr with corresponding

ASQ score

InternalityStabilityGlobalityComposite

3.713.783.133.54

1.541.621.431.27

.25**

.37**

.33**

.36**

.41***

.19*

.23*

.30**

Note, n ~ 66. ASQ = Attributional Style Questionnaire.*p < .10. **p < .05. ***p < .001.

of depressive symptoms, as measured bythe BDI.

These two studies have several importantimplications. First, the important constructsof the helplessness reformulation—explana-tions and explanatory style—are more thanresponses to questions about hypotheticalevents. Individuals offer unsolicited causal ex-planations about important bad events, andthese explanations are consistent with eachother along the dimensions of internality, sta-bility, and globality. Second, the ASQ has cri-terion validity. Internality, stability, and glo-bality ratings of spontaneous explanationsconverged with the corresponding scales of thequestionnaire.

Explanatory style versus reality. As we ar-gued above, explanations are determined byboth situational and dispositional factors. Thatthe reality of an event can determine expla-nations is undeniable. Also, we prefer the termexplanation to attribution because attributionconnotes projection too strongly. In saying thatreality partly determines explanations, we arenot saying that causal explanations are there-fore accurate (Langer, 1978; Nisbett & Wilson,1977). Rather, we are saying that causal beliefsare not entertained in a vacuum, and appro-priate investigations of the helplessness refor-mulation must not look only at dispositionaldeterminants of explanations.

A number of studies have shown manipu-lations of reality to affect causal explanations,which in turn affect laboratory helplessness(see reviews by Abramson et al., 1978; Miller& Norman, 1979; and Roth, 1980). A rep-resentative study is that of Tennen and Eller(1977), who instructed subjects given unsolv-able problems that the task was either easy ordifficult. Failure at an easy task encouragesexplanations in terms of lack of ability, whereasfailure at a difficult task results in explanations

in terms of task difficulty. Subjects given"easy" unsolvable problems showed laterproblem-solving deficits, whereas subjectsgiven "difficult" unsolvable problems showedfacilitation.

Situational factors also affect explanationsand depressive reactions to events outside thelaboratory. Peterson and Conn (1982) inves-tigated the reactions of people who had justended a romance. Although most subjects haddepressive symptoms and sad affect followingthe breakup, the most severe reactions wereby those whose partner left them for someoneelse. In contrast, the least severe reactions wereby those whose romance ended because theirpartner was attending a different college.

If one's partner leaves for someone else, theanswer to the why question tends to be in termsof stable and global characteristics of the per-son left behind. Many external explanations(e.g., my partner is not interested in romance)are implausible, as are interpretations in termsof transient and circumscribed causes. In con-trast, if romance ends because of distance, oneneed not see oneself as causing the breakup.Furthermore, one need not believe that thecauses will persist, or will be pervasive. Indeed,subjects whose romance ended because of dis-tance were more optimistic about achievingfuture relationships than were subjects whosepartner left for someone else.

Individual dispositions in the explanationsare needed to explain why different individualshave different reactions to the same events.Why do some people become helpless followingunsolvable problems and others not (Alloy,Peterson, Abramson, & Seligman, 1984;Dweck & Licht, 1980)? Why do some peoplebecome depressed following bad events andothers not (Lloyd, 1980; Metalsky, Abramson,Seligman, Semmel, & Peterson, 1982)? Thereformulation claims that people susceptible


to helplessness and depression interpret thesebad events in internal, stable, and global terms.

If reality is ambiguous enough, a personmay project and impose habitual explanations.According to the reformulation, individualshave a characteristic way of explaining events.They are consistent in the way they use internal(vs. external), stable (vs. unstable), and global(vs. specific) causes to explain bad events.Thus, when reality is ambiguous, the ASQworks as a projective test and may be used tomeasure a person's characteristic explanatorystyle. If this style invokes internal, stable, andglobal causes, then the person tends to becomedepressed when bad events occur. This is thecentral prediction of the reformulation, andwe now describe converging evidence in sup-port of this prediction.

Research strategies. We use a variety ofresearch strategies: (a) cross-sectional corre-lational studies, (b) causal modeling with lon-gitudinal data, (c) experiments of nature, (d)laboratory experiments, and (e) case studies.Each class of investigation addresses a morestringent prediction. The cross-sectional stud-ies seek merely to correlate explanatory styleand depressive symptoms. The longitudinalstudies investigate whether a depressive styleprecedes depression, consistent with a predis-posing role. The experiments of nature lookat how naturally occurring bad events and ex-planatory style interact to produce a subse-quent depressive reaction. The laboratory ex-periments see if experimentally manipulatedbad events interact with explanatory style toresult in laboratory helplessness, a depressiveanalogue. Finally, the case studies ascertainthe applicability of the reformulation to reallives.

EvidenceCross-Sectional Correlational Investigations

In this section, we describe several inves-tigations of the hypothesis that depressivesymptoms correlate with the use of internal,stable, and global causes to explain bad events.These studies are cross-sectional, and employa variety of populations, using the ASQ. De-pressive symptoms were assessed by self-reportquestionnaires and by psychiatric diagnosis.

Study 1: ASQ and DepressionIn the first investigation of the reformula-

tion, Seligman, Abramson, Semmel, and von

Baeyer (1979) administered the ASQ to asample of 143 college students at the Universityof Pennsylvania, along with the short form ofthe BDI (Beck & Beck, 1972), a 13-item self-report instrument that assesses the severity ofcommon depressive symptoms (for validityevidence among undergraduates, see Bum-berry, Oliver, & McClure, 1978). As predictedby the helplessness reformulation, depressivesymptoms among undergraduates correlatedwith internal (r = .41, p < .001), stable (r =.34, p < .001), and global (r = .35, p < .001)explanations for bad events (composite r = .48,p < .001).

Study 2: Spontaneous Explanationsand Depression

We have already described our investigationof spontaneous explanations (Peterson, Bettes,& Seligman, 1982). Sixty-six adults wrote es-says from which causal explanations of badevents were extracted and rated for internality,stability, and globality. These explanationswere consistent and converged with the cor-responding scales of the ASQ. These subjectsalso completed the short form of the BDI.

ASQ scores correlated with depressivesymptoms as predicted by the reformulation:internality (r = .44, p < .001), stability (r =.42, p < .001), globality (r = .29, p < .02, andcomposite (r = .45, p < .001), replicating theresults of Study 1. In addition, when ratingsof the extracted explanations were averagedacross different causes offered by the same in-dividual, these scores also correlated with de-pressive symptoms: internality (r = .36, p <.005), stability (r = .29, p < .02), globality(r = .33, p < .01), and composite (r= .39,p< .001).

The studies so far described support thepredictions. However, these studies employedsomewhat restricted samples: upper middle-class college students who were on the wholenot seriously depressed. To extend the em-pirical support for the depressive explanatorystyle, we conducted parallel investigations withthree different samples: lower-class women,children, and psychiatric inpatients.

Study 3: Lower-Class WomenDepressive disorders among the lower class

have sometimes been regarded as mainly so-matic (Schwab, Bialow, Brown, Holzer, &Stevenson, 1967). Accordingly, the relevance


Table 6Sample Items From the Children's Attributional Style Questionnaire

Dimension

Item Valence Varied

Note. From Seligman et al. (1984).

Held constant

A good friend tells you that he hates you. Bada. My friend was in a bad mood that day.b. I wasn't nice to my friend that day.

You get all the toys you want on your birthday. Gooda. People always guess what toys to get me for

my birthday.b. This birthday people guessed right as to what

toys I wanted.You get an "A" on a test. Good

a. I am smart.b. I am smart in the subject that the test was in.

ExternalInternal

Stable

Unstable

GlobalSpecific

Stability; globality

Internality; globality

Internality; stability

of a cognitive model like the helplessness re-formulation may be questioned. For this rea-son, we asked women from the lower socio-economic class to respond to the ASQ and theBDI (Navarra, 1981). Forty-one women (av-erage age = 38 years) on welfare in Philadel-phia were recruited through newspaper ad-vertisements as research participants. In aninterview format, they answered the questionsof the ASQ and of the long form (21 items)of the BDI (Beck, 1967).

As in the other samples, causal explanationsfor bad events correlated with depression, al-though the stability dimension did not reachstatistical significance: internality (r = .42, p <.01), stability (r = .12, ns), globality (r = .50,p < .001), and composite (r = .50, p < .001).

Study 4: Children

The helplessness reformulation purports tobe a general theory that should explaindepression across the life span (Seligman &Peterson, in press). Explanatory style shouldcorrelate with depressive symptoms amongchildren in the same pattern as among adults.Accordingly, we conducted an investigation tosee if the depressive explanatory style char-acterized depressed children (Seligman et al.,1984).

We tested 96 children from two Philadelphiaelementary schools that consisted predomi-nantly of white, middle-class children. Ap-proximately equal numbers of boys (« = 50)and girls (n = 46) from the third, fourth, fifth,

and sixth grades participated. The subjectscompleted the Children's Depression Inven-tory (CDI; Kovacs & Beck, 1977) and aChildren's Attributional Style Questionnaire(CASQ), which we developed. During class-time, the researcher read the questionnairesaloud while the children silently read theircopies. The questionnaires were completedtwice by the same children, at an interval of6 months.

The CDI is a 27-item questionnaire thatmeasures the severity of emotional, motiva-tional, cognitive, and somatic symptoms ofdepression. Each item consists of three self-report statements graded in severity from 0 to2. The child is instructed to choose the optioncorresponding to how he or she has been feel-ing during the preceding 2 weeks.

The CASQ is a forced-choice instrument.We found in pilot work that young childrenhad trouble with the adult ASQ, particularlyglobality. Hence, we used a force-choice formatin which hypothetical good or bad events in-volving the child were followed by two possibleexplanations, which varied one of the explan-atory dimensions while holding the other twoconstant. Sixteen questions pertain to each ofthe three dimensions; half refer to good events,and half to bad events. Examples of items fromthe CASQ are shown in Table 6.

The CASQ is scored by assigning a 1 to eachinternal or stable or global response, and a 0to each external or unstable or specific re-sponse. Scales are formed by summing thesescores across the appropriate questions for eachof the three dimensions, separately for good


Table 7Children's Attributional Style Questionnaire Scores and Depression

Dimension

Good eventsInternalityStabilityGlobalityComposite

Bad eventsInternalityStabilityGlobalityComposite

Children's DepressionInventory (CDI)

M

4.614.214.67

13.49

2.302.401.886.58

7.71

SD

1.481.911.583.72

1.571.401.272.77

6.28

fu

.32

.55

.40

.66

.43

.42

.31

.50

.86

r with same measure6 months later

.53**

.61**

.54**

.71**

.63**

.52**

.64**

.66**

.80**

r withCDI

-.34**-.47**-.35**-.53**

.45**

.31*

.21*

.51**

—

Note. From Seligman et al. (1984). rn is internal consistency estimated by Cronbach's (1951) alpha.*p < .05. **p < .001.

events and for bad events. Scores for each ofthe scales range from 0 to 8.

Table 7 presents the reliabilities and sta-bilities of the questionnaires completed by thechildren at Time 1. Statistics at Time 2 wereessentially the same, and are not shown. CASQsubscales were only modestly reliable. Moresatisfactory reliabilities were obtained bycombining the subscales (separately for goodevents and for bad events). Alpha for the goodevents composite was .66, and for bad eventswas .50 (see Table 7). The CASQ scales andcomposites were consistent over the 6-monthinterval, showing explanatory style to be asomewhat stable individual difference amongchildren, just as it is among adults.

CDI scores were highly reliable and highlyconsistent over the 6 months, implying thatchildhood depression, at least as measuredhere, is not a transient phase (Tesiny & Lef-kowitz, 1982). As predicted, explanatory stylecorrelated with depressive symptoms (see Ta-ble 7). The explanation of bad events withinternal, stable, and global causes covariedwith CDI scores, as did the reverse style forgood events. Considering the modest reliabil-ities of the CASQ, these correlations are sub-stantial; depression among school children isclosely tied to the causal explanations theypick.

This study broadens the empirical base ofthe depressive explanatory style by showing itamong 9- to 13-year-old children. It also addsto the evidence that childhood depression isa coherent syndrome similar to depression

among adults, not just at the symptom levelbut also in terms of explanatory style (Schul-terbrandt & Raskin, 1977).

Study 5: Depressed Patients

These four studies have two shortcomings.First, few of the individuals were seriously de-pressed. Although the helplessness reformu-lation regards mild and severe unipolardepression as continuous, at least with regardto cognitive characteristics (Seligman, 1978),this is an empirical issue. Second, the researchso far described does not show the "depressive"style to be specific to depression. Perhaps it isa general characteristic of psychopathology.

We therefore conducted a study in whichformally diagnosed unipolar depressed inpa-tients completed the ASQ (Raps, Peterson,Reinhard, Abramson, & Seligman, 1982). Wecompared their scores with those of two otherpatient groups. The first comparison groupwas medical and surgical patients, to controlfor being hospitalized (Taylor, 1979). The sec-ond was nondepressed schizophrenics, to con-trol for severity of psychopathology and lengthof hospitalization (Raps, Peterson, Jonas, &Seligman, 1982). Although a schizophrenicgroup is not an exhaustive control for psy-chopathology, it provides a first step towardassessing specificity.

We administered the ASQ to male veteranshospitalized at the Northport Veterans Ad-ministration Medical Center: 30 unipolar de-pressed patients, 15 nondepressed schizo-


Table 8Attributional Style Questionnaire: Patients' Scores and Depression

Depressives(n = 30)

Schizophrenics(n = 15)

Medical andsurgical patients

(« = 61)

Dimension M SD M SD SD F(2, 103)

InternalityStabilityGlobalityComposite

4.90,4.89S4.84X4.88X

1.121.161.292.92

3.51y4.01y

4.10Ky3.87y

1.181.011.192.26

4.30,4.06y3.65S4.00y

1.120.901.372.73

7.34*7.67*8.25*

10.69*

Note. From Raps, Peterson, Reinhard, Abramson, & Seligman (1982). Means in a row are different (p < .05) byplanned comparisons (Winer, 1971, pp. 210-215) if they have different subscripts.*p < .001.

phrenic patients, and 61 nondepressed medicaland surgical patients. As shown in Table 8,depressives explained bad events with moreinternal, stable, and global causes. These re-sults confirm, in a sample of unipolar de-pressed patients, the association between ex-planatory style for bad events and depressivesymptoms predicted by the helplessness re-formulation. Such a style is not a general char-acteristic of psychopathology, because schizo-phrenic patients did not show it.

Two other studies have also correlated theASQ with depressive symptoms in severe uni-polar depressed patients. Eaves and Rush(1984) compared 31 depressed female psy-chiatric patients, diagnosed as unipolar, majoraifective disorder, with an age-, sex- and ed-ucation-matched control group of 17 people.

The effects were striking. Depressed patients,whether endogenous, nonendogenous, or un-remitted, all differed from the control groupin terms of explanations for bad events (allps < .0001). Table 9 presents these data.

For these patients, sensitivity of the ASQ,denned as the probability of a correct diagnosisof depression by the test given that the patientis depressed (i.e., the percentage of true pos-itives; Lusted, 1968, pp. 108-109), was 61%for internality for bad events, 58% for stabilityfor bad events, and 77% for globality for badevents. Specificity of the ASQ, denned as theprobability of a correct diagnosis of nonde-pressed by the test given that the patient isnondepressed (i.e., the percentage of true neg-atives; Lusted, 1968, pp. 108-109), was 94%for internality for bad events, 94% for stability

Table 9Attributional Style Questionnaire Scores and Depression: Patients Over Time

Dimension

InternalityTime 1Time 2

StabilityTime 1Time 2

GlobalityTime 1Time 2

CompositeTime 1Time 2

ED( n = 11)

5.40X5.35X

5.44X4.76*

5.44X4.54*

16.27X14.65*

NE(n = 13)

5.38X5.13X

4.81X4.28,

4.92,4.19,

15.10X13.61*

UD

5.23,5.41,

5.07,5.01,

5.11,5.27X

15.41,15.70X

C(n = 17)

3.94y4.14y

3.92y3.49y

2.95y2.96y

10.80y10.59y

Note. From Eaves and Rush (1984). ED = endogenous ; NE = nonendogenous; UD = unremitted; C = control; seetext. Means in a row are different (p < .05) if they have different subscripts.* Time 1 different than Time 2 (p < .05).


for bad events, and 88% for globality for badevents. These compare favorably to the bestbiological tests of depression (Eaves & Rush,1982).

In addition, explanatory style for bad eventssignificantly correlated with the total amountof time the individual was depressed, the av-erage length of episode, and the length of thecurrent episode for nonendogenously de-pressed individuals. This was particularly thecase for the stability measure, which correlated.79 with the total time of depression, .76 withthe average length of episode, and ,71 with thelength of the current episode (all ps < .001).Theoretically, these data suggest that stabilityfor bad events reflects, and may predict, howlong a depressive episode will last.

Another study of unipolar depressed pa-tients was conducted by Persons and Rao(1981). Forty-nine patients completed theASQ, as well as other measures. Upon ad-mission, the internality score for bad eventscorrelated .60 with the BDI. The globalityscore for bad events correlated .53, whereasthe stability score for bad events did not cor-relate significantly with the BDI. Thirty-fourof these patients were followed through dis-charge, and changes in explanatory style andBDI scores were measured. Explanatory stylechange over therapy (largely chemical) ac-counted for 49% of total change in depression.This was the most powerful variable correlatedwith change in depression.

Overall, then, these three studies show thatseverely depressed patients have an insidiousexplanatory style characterized by internal,stable, and global explanations for bad events.This style is continuous with the findings inmild depression and not a general character-istic of psychopathology, inasmuch as schizo-phrenic patients did not show it. In addition,explanatory style is an index that is both sen-sitive and specific. Finally, changing explan-atory style is a good correlate and may be agood predictor of changing depression. Thislast finding is particularly suggestive for ther-apy, because techniques that encourage an in-dividual to change his or her explanatory stylemay relieve depression (Seligman, 1981).

Disconfirming Studies

Several researchers have correlated depres-sive symptoms with the use of internal, stable,

and/or global explanations for bad events andobtained nonsignificant correlations (e.g.,Hammen & deMayo, 1982). Some have con-cluded that the helplessness reformulation istherefore wrong (e.g., Coyne & Gotlib, 1983).How do we reconcile this with the substantialsupport found for a depressive explanatorystyle?

Conclusions based on accepting the null hy-pothesis are always problematic, because non-significant results may alternatively reflect aninsensitive or inappropriate method. This maybe the case for studies that have failed to finda depressive explanatory style. Peterson andRaps (1983) observed that these studies tendnot to distinguish explanations per se fromexplanatory style. As a result, they operation-alize style with single-item questionnaires,running the risk that the particular real eventabout which respondents offer explanations isthe overriding determinant of that explanation.This technique minimizes the role of the per-son's cognitive style. If respondents are askedto offer explanations about several (hypothet-ical) bad events, then the chances are greaterthat the average of these explanations will re-flect a characteristic style.

This is really a logical argument, anotherway of phrasing the statistical truism that re-liability and thus validity are served by in-creasing the number of observations above one(Epstein, 1980). To illustrate this empirically,Peterson and Raps (1983) classified a numberof cross-sectional investigations of the expla-nation-depression link, noting the numbersof events about which explanations were madein each study. As Table 10 shows, confirmingstudies were more likely than were discon-firming studies to ask about several events.Although the disconfirming studies may becorrect in concluding that contextual char-acteristics are importantly related to depres-sion, they are incorrect in further concludingthat explanatory style is unrelated, becausethey usually did not assess style in a psycho-metrically or theoretically reasonable way.

Conclusions

Severity of depressive symptoms is oftencorrelated with the habitual use of internal,stable, and global causes to explain bad eventsinvolving the self. This depressive explanatorystyle has been demonstrated with a variety of


procedures in a variety of populations. Someevidence suggests that it may be specific todepression. Studies failing to find the depres-sive explanatory style tend to use inappropriatemethods. So, there is a clear relationship be-tween a particular explanatory style anddepression. But does this style put one at riskfor later depression? Several other possibilitiesare compatible with this correlation.

First, there is the possibility of a tautologicalrelationship of depressive explanatory style anddepressive symptoms. For example, many an-

Table 10Cross-Sectional Investigations of theReformulation and the Number of Events forWhich Explanations Were Made

Studies No. ofevents

Disconfirming

Danker-Brown and Baucom (1982)Feather and Davenport (1981)Garber and Hollon (1980)Gong-Guy and Hammen (1980)Hammen and Cochran (1981)Hammen and deMayo (1982)Peterson and Conn (1982)

Average

1115511

2.14

Confirming

Blaney, Behar, and Head (1980) - One 6- Two 6

Eaves and Rush (1984) 6Janoff-Bulman(1979) 4Peterson, Bettes, and Seligman (1982) - One 2

- Two 6Peterson, Schwartz, and Seligman (1981) 18Raps, Peterson, Reinhard, Abramson, and 6

Seligman (1982)Seligman, Abramson, Semmel, and von Baeyer 6

(1979)Seligman et al. (1984) 8Sweeney, Shaeffer, and Golin (1982) 6Zemore and Johansen (1980) 1

Average 6.25

Note. From Peterson and Raps (1983). All studies cor-related depressive symptoms with the use of internal, stable,and/or global explanations for bad events involving theself. Disconfirming studies were those that found no sig-nificant correlations (p > .05); confirming studies werethose that found significant correlations (p < .05). Notall studies investigated all three explanatory dimensions.Not included are studies that found mixed support forthe reformulation.

swers that count toward the self-report ofdepression are phrased in terms of internal,stable, and global causes, for example, "I willalways fail at whatever I do." Perhaps a badexplanatory style is just a special symptom ofdepression. This is not the case. If we eliminatesuch items from the BDI and correlate ex-planatory style with the remaining items, thecorrelation does not go down (Peterson,Schwartz, & Seligman, 1981). In addition, aswe report below, such a style precedes andpredicts who, among people not depressed,will later become depressed.

Second, there is the possibility that depres-sion wholly causes the bad explanatory style.On this view, when one becomes depressed,the depressive explanatory style is activated asa result. A study by Mukherji, Abramson, andMartin (1982) argues against this possibility.They induced a depressed mood in under-graduates with the Velten (1968) procedure.Although mood induction was successful, ex-planatory style as measured by the ASQ wasunaffected.

Third, there is the possibility that some thirdvariable, such as biochemical deficits or apreexisting style of turning anger inward,causes both depression and the depressive ex-planatory style. The remaining lines of re-search attempt to untangle these various pos-sibilities from the process predicted by the re-formulation.

In the next section, we begin this untanglingby describing evidence for the next most strin-gent prediction of the reformulation: that de-pressive explanatory style precedes depression.

Longitudinal Investigations

According to the reformulation, explanatorystyle is a risk factor for depression. It mustprecede the occurrence of depressive symp-toms. Because cross-sectional studies are mo-mentary, they cannot address this prediction.However, several longitudinal investigationshave been conducted in which explanatorystyle was assessed at one time and conse-quences at a later time.

Study 6: Children

We have already described our research withgrade-school children, who completed mea-


sures of explanatory style and depressivesymptoms at two times, 6 months apart (Se-ligman et al., 1984). The cross-sectional datafrom this research supported the reformula-tion. Do the longitudinal data?

The question of interest is whether explan-atory style at Time 1 predicts depressivesymptoms at Time 2, above and beyond thelevel of depression at Time 1. As predicted bythe reformulation, the depressive explanatorystyle for bad events was correlated with sub-sequent depressive symptoms, even when ini-tial CDI scores were partialed out (partial r =.26, p < .02). So, explanatory style for badevents may be a risk factor for depressionamong children.

Nolen-Hoeksema (1983) performed a sim-ilar study using 168 children from Grades 3-5. Children who were both nondepressed inJanuary and had a nondepressive explanatorystyle remained nondepressed in March. Incontrast, children who were nondepressed inJanuary but had a depressive explanatory stylewere likely to be depressed in March (p < .01).

Several recently completed studies usingadults address the same question. Golin,Sweeney, and Shaeifer (1981) and Firth andBrewin (1982) administered measures of bothexplanatory style and depressive symptoms attwo times. Both studies provided support forthe strong claim that explanations affectdepression to a greater degree than depressionaffects explanations. In the Golin et al. (1981)study, causal priority of stability and globalityfor bad events was demonstrated for 180 col-lege students over a 1-month period. In theFirth and Brewin (1982) study, causal priorityfor stability for bad events was demonstratedfor 16 female patients over a 5-week period.

In another longitudinal study, O'Hara,Rehm, and Campbell (1982) administered theASQ to 170 women in their second trimesterof pregnancy, along with several other cog-nitive-behavioral assessments. The strongestpredictor of level of depression, 3 monthspostpartum, was explanatory style for badevents, which significantly correlated with BDIscores even when prepartum level of depres-sion was held constant.

In similar studies, Manly, McMahon, Brad-ley, and Davidson (1982) found no supportfor the hypothesis that explanatory style pre-dicts depression following childbirth, whereasCutrona (1983) replicated the findings of

O'Hara et al. (1982) and also showed that ex-planatory style predicted speed of recoveryfrom postpartum depression.

Conclusion

These studies mostly support the predictionthat depressive explanatory style precedes de-pressive symptoms. Explanatory style may bea risk factor for later depression, allowingidentification of individuals at risk by the waythey explain bad events. Primary preventionof depression should then be focused on suchindividuals (but see Lewinsohn, Steinmetz,Larson, & Franklin, 1981, and Peterson et al.,1981).

The general flaw of these studies as tests ofthe reformulation is that they do not manip-ulate or assess bad life events. According toAbramson et al. (1978), a depressive explan-atory style per se is not sufficient for depres-sion. It is only when bad events occur and areinterpreted in terms of internal, stable, andglobal causes that depressive symptoms aremore likely to ensue. Because these studies didnot look at bad events, they did not test thisprediction. The studies in the next section testthe next strongest proposition that a preex-isting bad explanatory style followed by badevents makes depression more likely.

Experiments of Nature

The ideal way to test the helplessness re-formulation is to measure the explanatory styleof individuals, and then to choose at randomhalf of these individuals to experience someimportant bad event. The most severe depres-sion is predicted to ensue for subjects with thepreexisting depressive explanatory style whoalso then experienced the bad event. The ob-vious ethical dilemma can be partly solved bya quasi-experimental method in which natu-rally occurring bad events are the manipula-tion. We report two studies conducted by ourresearch group. In the first, the bad event wasan unsatisfactory grade on a midterm ex-amination. In the second, the bad event wasimprisonment in a penitentiary. We also de-scribe several other experiments of nature thatare pertinent to the reformulation.

Study 7: Midterm Experiment

This investigation was a prospective studyof college students and their reactions to a low


grade on a midterm examination (Metalskyet al., 1982). According to the helplessness re-formulation, students who habitually explainbad events in terms of internal, stable, andglobal factors are more likely to feel depressedon learning that they received a low grade thanare students who tend to explain bad eventsin terms of external, unstable, and specific fac-tors.

The participants were undergraduates in anintroductory psychology course at the StateUniversity of New York at Stony Brook, whocompleted questionnaires for this researchduring classtime. At Time 1, they took theASQ and a questionnaire about their aspira-tions for the class midterm. Students indicatedthe grades with which they would be happyand unhappy. At Time 2, 11 days later, justprior to the midterm examination, students'level of depressed mood was assessed withthe Multiple Affect Adjective Check List(MAACL; Zuckerman & Lubin, 1965), ameasure of transient depressed mood. At Time3, another 5 days later, immediately followingreceipt of the midterm grade, students againcompleted the MAACL.

Subjects were considered to have receiveda low grade if their midterm grade was lessthan or equal to the grade with which theysaid they would be unhappy (« = 53). Studentswere considered to have received a high gradeif their midterm grade was equal to or greaterthan the grade with which they said they wouldbe happy (n = 28). To test our predictions,explanatory style scores for bad events werecorrelated with standardized residual gainscores on the MAACL depression scale fromTime 2 to Time 3, separately for the twogroups. Explanatory style for bad events pre-dicted increases in depressed mood for stu-dents who received low grades—internality(r=.34,p< .02), stability (r = .04, ns), andglobality (r = .32, p < .02)—but not for stu-dents who received high grades—internality(r = .12, ns), stability (r = .36, p < .10), andglobality (r = .22, ns). Additional correlationalanalyses revealed that explanatory style wasindependent of the actual grade received andof grade aspirations. Thus, several possibleconfounds are ruled out.

The theoretical significance of these findingsis that they support the reformulation's claimto be a diathesis-stress model, where the de-

pressive explanatory style is a constitutionalweakness (diathesis) and the low midtermgrade is the environmental stressor (stress),which interact to predispose depressed mood(Metalsky et al., 1982). The practical signifi-cance lies in the possibility of identifying, inadvance, individuals at risk for depressive re-action to bad life events from explanatory style.

Study 8: Prison ExperimentDoes explanatory style predict depression

following imprisonment (Bukstel & Kilmann,1980)? For most individuals, imprisonment isinarguably a bad event. Like any total insti-tution, prisons deny individuals control overeven the most mundane aspects of their lives(Goffman, 1961; Taylor, 1979). We would ex-pect, then, that a common reaction to im-prisonment is depression (Seligman, 1975).Furthermore, we predicted that prisoners withthe depressive explanatory style are the mostlikely to become depressed after internment(Abramson et al., 1978). We report a prelim-inary study of this (Peterson, Nutter, & Selig-man, 1982). Upon imprisonment, individualscompleted the ASQ, and shortly before release,these same individuals completed the BDI.

Within 1 week following imprisonment atone of four maximum security prisons in NewYork, Ohio, and Pennsylvania, 245 adult males(ranging in age from 17 to 64 years; averageage = 27 years) completed the ASQ. Within1 week before their release, which varied from1 month to 1 year, they completed the BDI.For 28 prisoners, BDI scores from the time ofinitial imprisonment were available. These av-eraged only 1.68, suggesting that subjects werenot at all depressed at the time of imprison-ment.

At the end of their imprisonment, prisonersscored an average of 17.7 on the BDI, placingthem in the moderately to severely depressedrange (Beck, 1967). Furthermore, their de-pressive symptoms at the end of imprisonmentwere strongly associated with their explanatorystyle at the beginning of imprisonment. Asexpected, explanations for bad events werepositively correlated with depressive symp-toms: internality (r = .34, p < .001), stability(r = .36, p < .001), globality (r = .35, p <.001), and composite (r = .35, p < .001). Sur-prisingly, explanations for good events weresimilarly correlated with depression: inter-


nality (r = .39, p < .001), stability (r = .38,p < .001), globality (r = .37, p < .001), andcomposite (r = .39, p < .001). Unlike resultsin our other research, explanatory styles forbad versus good events did not have oppositeeffects on depression. Furthermore, unlike re-sults in our other research, explanatory stylesfor bad versus good events were not indepen-dent. Instead, they were substantially inter-correlated (all rs > .60; all ps < .001).

Other Experiments of Nature

Several other experiments of nature deservemention. Although not designed to test thehelplessness reformulation, they are intriguingand illustrate how investigations of reactionsto naturally occurring bad events might beused to test the reformulation. Tennen, Affleck,Allen, McGrade, and Ratzan (1983) studied32 children who had diabetes. Children whoexplained the onset of their disease in termsof their behavior (i.e., internally, unstably, andspecifically; see Peterson, Schwartz, & Selig-man, 1981) were rated by their physicians ascoping better with diabetes than were childrenwho explained the onset in terms of geneticinheritance (i.e., internally, stably, and glob-ally). Although a third variable such as parentalattitude may underlie these data, they are con-sistent with the reformulation.

Similarly, Affleck, Allen, McGrade, andMcQueeney (1982) interviewed the mothersof 43 infants with severe perinatal complica-tions. Mothers who explained the infant's con-dition in terms of some behavior during preg-nancy (i.e., unstably and specifically) reportedless mood disturbance and anticipated fewercaretaking problems than did mothers whoexplained the complications in terms of otherpeople (i.e., stably and globally).

Timko and Janoff-Bulman (1983) inter-viewed 42 women who had undergone a mas-tectomy following a diagnosis of breast cancer,asking them to offer a causal explanation fortheir cancer and to complete the BDI. Alsoassessed were other beliefs about the successof the mastectomy and the avoidability of fu-ture cancer. The helplessness reformulationwas strongly supported. Causal explanationsdid not directly affect depression, but they didpredict expectations about future cancer,which in turn predicted BDI scores. More spe-

cifically, women who explained their cancerin terms of behavior (i.e., internal, unstable,and specific causes) believed that they werecancer free; they were likely to be nonde-pressed. In contrast, women who explainedtheir cancer in terms of personality (i.e., in-ternal, stable, and global causes) believed thatthey were not cancer free; they were likely tobe depressed.

Not all studies that have looked at the linkbetween causal explanations and reactions toserious disease have supported the reformu-lation (e.g., Bard & Dyk, 1956). Factors suchas the seriousness and chronicity of the diseasemay affect the relationship between explana-tion and reaction, perhaps by creating expec-tations about the future course of the diseaseincongruent with its causal explanation.

Life Events and Depression

The large research literature on life eventsand psychological distress is also relevant tothe helplessness reformulation, although fewtheorists have drawn explicit parallels. How-ever, in a recent review of this literature, Thoits(1983) proposed that many of its consistentfindings are compatible with a helplessness in-terpretation. The modest relationship typicallyobserved between the experience of life eventsand subsequent distress is increased when (a)the psychological consequence is depressionas opposed to some other psychopathology and(b) the events themselves are relatively un-desirable, uncontrollable, unexpected, impor-tant, and clustered together in time.

Although life events research has typicallynot assessed causal explanations, the types ofevents most frequently linked to psychologicaldistress lend themselves to internal, stable, andglobal explanations. This possibility is consis-tent with the helplessness reformulation. Fur-thermore, that life events are more apt to resultin depression than in other psychological dis-orders adds to the evidence for the specificityof the helplessness model as an account ofdepression (cf. Raps, Peterson, Reinhard,Abramson, & Seligman, 1982).

Conclusions

These experiments of nature suggest thatthe helplessness model may be useful in ex-plaining reactions to major life events. In par-


ticular, the studies by our research group sug-gest that the depressive explanatory style tendsto produce depression when bad events areencountered. Other studies show that partic-ular explanations of bad events are associatedwith depression and poor coping. Each of thesestudies is open to criticism, because experi-mental control over the bad events was notpossible, and possible confounds might un-derlie the obtained correlations. However, be-cause the results of these studies converge, ourconfidence increases that the conjunction ofdepressive explanatory style and bad eventspredisposes subsequent depression.

Experimental studies, unlike experimentsof nature, allow the bad event to be presentedto randomly selected subjects. We now turnto a set of laboratory experiments that suggestthat helplessness symptoms occur as predictedwhen uncontrollable bad events are imposedon people who differ in explanations and ex-planatory style.

Laboratory Experiments

Does explanatory style affect laboratory-in-duced learned helplessness in the same waythat it affects naturally occurring depressivesymptoms? Laboratory studies give us whatexperiments of nature lack. We can impose abad event on randomly selected subjects andlook at the effects of preexisting or inducedcausal explanations. Several studies of this sorthave been done.

Study 9: Global VersusSpecific Explanations

Alloy et al. (1984) gave subjects one of threepretreatments as described by Hiroto and Se-ligman (1975): a condition in which button-pushing terminated an aversive noise deliveredthrough headphones (escapable), a conditionin which the noise could not be terminated(inescapable), and a condition in which nonoise—escapable or inescapable—was expe-rienced (no pretreatment). Then, subjects weretested at one of two tasks: a task similar topretreatment—a solvable hand shuttle-box inwhich success moving the hand stopped anaversive noise, and a task dissimilar to pre-treatment—a series of solvable anagrams ad-ministered in a different room by a differentexperimenter.

Table 11Explanations and Laboratory Helplessness:Mean Latencies (in s) at Test Task

Style

Test task Global Specific

SimilarEscapable noiseInescapable noiseNo noise control

DissimilarEscapable noiseInescapable noiseNo noise control

2.26,3.43b2.54a

16.4,41.5b20.7.

1.98.3.43b2.58b

17.4,20.0.19.0.

Note. From Alloy, Peterson, Abramson, & Seligman (1984).For similar test task, maximum latency = 5.00 s. For eachcondition, n = 18. For dissimilar test task, maximum la-tency = 100.0 s. For each condition, n = 10. Means in arow or in a column of three are different (p < .05) byNeWman-Keuls procedure if they have different subscripts.

According to the reformulation, all subjectsshould show deficits following inescapablenoise when tested on the similar noise task.However, only subjects with a global explan-atory style for bad events should show deficitsfollowing inescapable noise when tested on thedissimilar cognitive task. Table 11 shows thesepredictions to be supported. Furthermore, thedifferences remain even when internality andstability scores are covariates.

This study experimentally manipulated badevents and also looked at a specific role as-signed to one of the explanatory dimensions.In a related investigation (Peterson & Selig-man, 1981), we obtained some preliminaryevidence that stability for bad events is relatedto the time course of laboratory-induced help-lessness as predicted by the reformulation.

Other Laboratory Experiments

Several other laboratory studies concernboth the manipulation of causal explanationsand the manipulation of bad events. Pasahow(1980) manipulated the global-specific dimen-sion and imposed bad events on subjects inthe learned helplessness triadic design. Subjectsinstilled with a global explanation were givenwritten instructions that told them that theconcept-identification task they were about totake correlated very highly with how people


performed on all psychology experiment tasks.Specific subjects were given written instruc-tions that described the task as a concept-for-mation task that had little or no correlationwith other tasks used in psychology experi-ments.

All of the subjects were then given unsolv-able concept-identification problems and thentested on anagrams. Subjects induced to makeglobal explanations for their failure performedworse on the anagrams than did subjects en-couraged to give specific explanations. Thissuggests that the manipulation of explanationsalong the global-specific dimension for badevents appropriately predicts deficits when badevents are then imposed on subjects.

Thus, we are given converging evidenceabout the globality dimension for bad events,In the Alloy et al. (1984) study, preexistingexplanatory style was merely measured andthe bad event manipulated. Preexisting globalstyle for bad events predicted failure at newproblems once the bad event occurred,whereas specific style did not. In parallel, ifcausal explanations are themselves manipu-lated along with the bad event, Pasahow (1980)showed that the same results hold.

McFarland and Ross (1982) investigated theeffect of explanations on self-esteem followingfailure. They assigned female college studentsto success or failure feedback conditions fol-lowing a social accuracy test. They also ma-nipulated the causal explanation for perfor-mance at this test: internal or external.Consistent with the reformulation, subsequentself-esteem was lowest for subjects who ex-perienced failure and were induced to explainit internally. So, a manipulation designed toproduce internal causal explanations loweredself-esteem following a bad event, whereas amanipulation designed to produce externalcausal explanations did not.

In a laboratory experiment with depressedinpatients, Miller and Norman (1981) toldsubjects who were either acutely depressed orrecently improved and made helpless by in-escapable noise, that they had done well at atask measuring social intelligence. Subjectswere induced to explain their success in oneof four ways: with an internal-global cause,an internal-specific cause, an external-globalcause, or an external-specific cause. Consistentwith the reformulation, both depressed and

helpless subjects reported a less depressedmood when induced to explain their successinternally. Similarly, both depressed and help-less subjects performed better at a subsequentanagram task when induced to explain theirsuccess globally.

Finally, Anderson (1983) manipulatedcausal explanations following failure for collegestudents preselected for explanatory style. Us-ing his Attributional Style Assessment Test(ASAT), Anderson (1983) obtained subjectswho explained failure characterologically (i.e.,internally, stably, and globally) and subjectswho explained failure behaviorally (i.e., in-ternally, unstably, and specifically). These sub-jects then attempted to persuade other studentsto donate blood; explanations for failure wereexperimentally varied and included an ability/trait condition and a strategy/effort condition.Consistent with the reformulation, subjectswho offered behavioral explanations, whetherby preselection or manipulation, reportedgreater expectations of success, displayedhigher motivation, and performed more suc-cessfully following failure than did subjectswho offered characterological explanations (seealso Anderson & Jennings, 1980).

Conclusions

Laboratory investigations of learned help-lessness in humans have been controversialsince their inception (Wortman & Brehm,1975). In a recent review, Silver, Wortman,and Klos (1982) criticized these studies onthree grounds. First, the relationship betweenexperience with uncontrollable events andsubsequent deficits may be mediated by a va-riety of factors, not just the expectation ofresponse-outcome independence (e.g., Frankel& Snyder, 1978; Peterson, 1978; Tennen,1982). Second, laboratory studies of humanhelplessness may be subject to demand char-acteristics that trivialize obtained results(Orne, 1962). Third, the laboratory setting isartificial, and generalization of results to hu-man helplessness and depression outside ofthe laboratory may not be justified (e.g., Bui-man & Wortman, 1977; Coyne, Aldwin, &Lazarus, 1981).

These are reasonable criticisms, and the re-search program described here attempts toanswer them. First, as we have already stated,


the expectation of response-outcome inde-pendence is regarded as a sufficient conditionfor depressive deficits, not a necessary one.That other factors may also determine deficitsfollowing uncontrollability is undoubtedly trueand not contrary to this theory. However, con-verging research shows the importance of ex-pectations in the development of such deficits.Second, laboratory experiments may indeedbe confounded by demand characteristics, butother research strategies also have inherentflaws. Thus, we favor a multimethod researchstrategy and a search for convergence amongresults (Campbell & Fiske, 1959). Finally, theonly way to ascertain whether laboratory find-ings are artificial is to attempt studies outsidethe laboratory. The out-of-the-laboratory re-sults described here are mostly consistent withthe laboratory data.

These laboratory data indicate that themeasurement and manipulation of explana-tions and explanatory style, when accompa-nied by the manipulation of bad events, yieldwhat the theory predicts. The global-specificdimension, when manipulated and whenmeasured, governs the breadth of helplessnessdeficits. The internal-external dimension,when manipulated, appears to govern self-es-teem deficits. Finally, tentative evidence sug-gests that the stable-unstable dimension gov-erns the duration of helplessness deficits.

Case Studies: Investigation of Real Lives

The final line of evidence we wish to presentcomes from the study of individual lives. Itconverges on the proposition that explanatorystyle is a risk factor for depressive symptoms.One major barrier in trying to study the con-sequences of explanatory style in the labora-tory or in natural settings is the obtrusive andreactive nature of questionnaires. The devel-opment of a valid method for assessing ex-planatory style through the content analysisof written transcripts allows us to study, in awholly unobtrusive and nonreactive way, thenatural history of explanatory style anddepression. We now present several studies thatuse this technique. The first two studies analyzeverbatim transcripts of actual psychotherapysessions with depressed patients. The thirdstudy analyzes verbatim interviews originallyconducted decades ago with individuals for

whom measures of current psychologicalcharacteristics are available.

Study 10: Causal Explanationsand Psychotherapy

In the first study (Peterson & Seligman,1981), each causal explanation for a bad eventwas extracted from 300-word excerpts fromverbatim transcripts of individual psycho-therapy sessions. Transcripts were availablefrom the beginning, middle, and ending ofsuccessful psychotherapy with four differentpatients suffering depression following a loss.We blindly rated these explanations for inler-nality, stability, and globality, and averagedthese scores across explanations offered withinthe same excerpt. For each patient, explanatorystyle perfectly ordered the three sessions; themost internal, stable, and global explanationswere offered in the beginning session, whereasthe least internal, stable, and global explana-tions were offered in the ending session. Thesedata supplement the Persons and Rao (1981)study already described in suggesting thatchanges in explanatory style index changes indepression during psychotherapy.

Study 11: Causal Explanationsand Mood Swings

In the second study (Peterson et al., 1983),we used the symptom-context method, a con-tent-analytic approach developed by Luborsky(1964, 1970), to investigate the explanatoryantecedents of mood swings by a single patientduring psychotherapy sessions. With thesymptom-context method, context is assessedimmediately before and after the onset of thesymptoms of interest. In this case, the contextwas causal explanations for bad events, andthe symptoms were mood swings—in and outof depression during therapy sessions.

Our subject was a patient, Mr. Q, who dem-onstrated precipitous shifts in mood duringpsychotherapy sessions (conducted over 4years). More than 200 sessions with Mr. Qwere tape-recorded, and a sample was tran-scribed, allowing a thorough content analysisof his explanations before and after moodswings. Three types of sessions were analyzed:those in which Mr. Q became more depressed,those in which he became less depressed, and


those in which no shift at all occurred. Ex-planations for bad events were extracted fromthese sessions before and after the moodswings, and were rated for internality, stability,and globality, as described above.

Several criteria were used to identify Mr.Q's swings to and from depression: (a) Mr.Q's report of a shift in mood (e.g., "My moodjust went down"), (b) the agreement of twoindependent judges reading the transcript ofthe therapy session that a mood swing had justoccurred, and (c) the fact that the swing didnot occur in close proximity to another swing(i.e., within 15 min of the same session). Ap-plying these three criteria resulted in a finalset of four swings in which the patient becamemore depressed and five swings in which hebecame less depressed. These mood swingswere validated by both the agreement of twoindependent judges listening to the taped ther-apy sessions that Mr. Q's voice showed achange in affective quality and the scoring byan independent judge using the depressioncomponent of the Gottschalk and Gleser(1969) hostility-inward scale.

Explanations for bad events were extractedfrom the 400 words (spoken by Mr. Q) beforethe swing and from the 400 words followingthe swing. For comparison purposes, causalexplanations were also extracted from ran-domly chosen 800-word segments of three ses-sions in which no swing occurred. In all, 63bad events linked to causal explanations wererated for internality, stability, and globality byfour independent judges blind to their source.Ratings by the judges were combined, andcomposites were formed. Means for the dif-ferent types of sessions are shown inFigure 2.

The differences predicted by the reformu-lation were present before swings in mood.Highly internal, stable, and global causal ex-planations preceded increased depression,whereas much more external, unstable, andspecific statements preceded decreased de-pression. There was no overlap between theratings of causal explanations before swingsto more versus less depression.

This is a fine-grained case study, but becauseit is correlational, it is necessary to rule outsome plausible third-variable arguments toconclude that explanations, and not some fac-tor correlated with them, produced mood

I6]•i 4

.4 I Depression

(n=4)

No change

(n=3)

Depression(n=5)

I

Before shift After shiftFigure 2. Means of internal + stable + global ratings ofexplanations before and after mood shifts for the differenttypes of sessions. (Numbers of sessions on which meansare based are in parentheses.)

swings. Various content characteristics of thesesame transcripts, including Mr. Q's statementsabout anxiety, hopelessness, guilt, hostility toself, loss of self-esteem, and Oedipal conflict,were examined. These were not redundantwith the causal explanation scores, rising andfalling with level of mood in a given session.In contrast, explanations were different beforeswings.

This study shows that the reformulation maybe applied powerfully to specific individualswith depressive symptoms. Explanation ratingsfor sessions in which swings to increaseddepression occurred did not at all overlap withexplanation ratings for sessions in whichswings to decreased depression occurred, al-lowing perfect prediction.

Study 12: Individuals of the Berkeley-Oakland Study

We now describe a preliminary study usingwhat we have dubbed the time machinemethod. While conducting content analysis ofcausal explanations provided in essays andtherapy transcripts, we realized that there was


no reason to limit ourselves to contemporarydata. Content analysis can be used with anywritten material, including newspaper stories,autobiographies, transcribed interviews, let-ters, diaries, and so on (Simonton, 1981). So,we decided to use content analysis to extractcausal explanations from open-ended materialresiding in already-existing data archives. Theexplanatory style of individuals years ago canbe assessed through content analysis, and therelationship between causal explanations andlater characteristics of these individuals canbe ascertained. Prospective longitudinal re-search spanning decades is possible retro-spectively.

Our first attempt to study individual livesover decades analyzed material in the Berke-ley-Oakland Growth Study (Elder, Bettes, &Seligman, 1982), an ongoing investigation ofthe effects of the Great Depression on familyand life patterns (Elder, 1974). Although eco-nomic hardships of the Great Depression hitboth the lower class and the middle class, thelower class was hit harder. Indeed, middle-classindividuals who suffered financial hardshipsduring the Great Depression years later werehappier, healthier, and more efficacious thanmiddle-class individuals who had not beeneconomically deprived (Elder, 1974). Does ex-planatory style mediate this interaction be-tween social class and economic hardship?

Causal explanations were extracted from theopen-ended responses of 28 women who de-scribed the worst events during the precedingyear in both 1943 and 1970. The focus of thisstudy was on hopelessness—the use of stableand global causes to explain bad events—sothese explanations were rated for stability andglobality, but not internality. A hopeless out-look in 1943 was significantly correlated witha hopeless outlook in 1970 (r = .44, p < .05),through which it affected overall psychologicalhealth (assessed by a Q-sort procedure) in1970.

More detailed analyses suggested that hard-ships in 1929 predicted hopelessness in 1943.Hardships per se had no direct effect on psy-chological health in 1970. Only when hard-ships gave rise to a hopeless outlook was psy-chological health affected for the worse decadeslater. Furthermore, these patterns seem morepronounced among the lower class than amongthe middle class. The small sample size rendersthis conclusion highly tentative, but it appears

as if middle-class individuals were not so proneto react to economic difficulties with hope-lessness, perhaps because they had availableto them means for coping with difficulties. Thehelplessness model proposes that early masteryexperience immunizes one against the bad ef-fects of uncontrollability (Maier & Seligman,1976); these data from the Berkeley-OaklandStudy may illustrate immunization.

In an even more preliminary investigationusing the time-machine methodology, we an-alyzed interviews from the Grant Study, a lon-gitudinal investigation of the Harvard classesof 1939-1942. Over the 40 years of this re-search, the material success, social success,mental health, and physical health of thesemen have been recorded, with special interestin whether such psychological characteristicsas defense mechanisms predict future lifecourse (Vaillant, 1977). A great amount ofopen-ended interview material is available foreach individual studied. We extracted causalexplanations for bad events from a 1946 in-terview about war experiences conducted with18 men. Explanations were rated for inter-nality, stability, and globality, as in our otherstudies.

These ratings were correlated with a globalrating of physical health in 1980: 1 = healthy;2 = minor health problems; 3 = chronic ill-ness; 4 = disabled; and 5 = dead. Because ofthe small sample size, we report only the cor-relation with the composite explanatory style.It was positive (r = .40, p < .10). Depressiveexplanatory style in 1946 predicted poorphysical health 34 years later. The tentativenature of this study must be emphasized.Sample size is small, and the obtained cor-relation is of borderline statistical significance.The dependent measure assessed overall phys-ical health, not level of depression, and we donot know what the mechanism linking ex-planatory style and health might be.

Both of these investigations are preliminary,and we are currently analyzing the causal ex-planations in a greater number of interviews.Nevertheless, we have described these still-in-progress studies because they suggest that thehelplessness reformulation can be applied toreal people over the life span.

General ConclusionsWe have described five interlocking types

of research investigating the attributional re-


formulation of the helplessness model ofdepression. The cross-sectional studies showedthat a characteristic way of explaining badevents with internal, stable, and global causesco-occurs with depressive symptoms. The lon-gitudinal studies showed that this explanatorystyle preceded the development of depressivesymptoms. The experiments of nature indi-cated that this style resulted in depression oncebad events were encountered. The laboratoryexperiments showed that imposing uncon-trollable bad events on individuals makingparticular explanations had the predicted ef-fects on helplessness deficits. Finally, the casestudies illustrated that the reformulation ap-plies predictively to the depressive symptomsof specific individuals.

Taken together, these five lines of researchsupport the predictions of the attributionalreformulation of the learned helplessnessmodel of depression. Each type of research isopen to criticism on several grounds, but theconvergence of results across different strat-egies of investigation, different operational-izations, and different populations arguesstrongly for the validity of the reformulation(Campbell & Fiske, 1959).

There are aspects of the reformulation stillin need of theoretical and empirical scrutiny,and we briefly discuss these now.

1. The role in depression of causal expla-nations of good events. The helplessness re-formulation does not explicitly address suchcausal beliefs, but they seem to have at leasta weak empirical relationship with depressivesymptoms. We have assessed such beliefs andgenerally found them to have effects ondepression opposite to those of causal expla-nations about bad events, but at less robustlevels. However, some of the findings do notfit these generalizations. In our longitudinalinvestigation of children, explanatory style forbad events predicted subsequent depression,whereas style for bad events did not. In ourstudy of prisoners, the explanation of goodevents with internal, stable, and global causeswas positively, not negatively, correlated withlater depression. A sound theoretical treatmentis awaited.

We speculate that our pattern of findingswith children may be related to the sorts ofevents that adults take for granted that childrendo not. When do we engage in explanatoryanalysis, and when do we forego it? Peirce

(1955) suggested that the function of thoughtis to allay doubt (see also Wong & Weiner,1981). It seems likely that we ask why moreoften when we are in doubt—when our pathis suddenly obstructed—than when things aregoing smoothly. Indeed, consciousness itselfmay emerge only when our routine actions fail(cf. Langer, 1978). This implies that the ex-planatory analysis of failure among adultswould be much more articulate than wouldthe analysis of success. Children, on the otherhand, newer at the game, may analyze thecauses of both failure and success closely. Sucha developmental process would render chil-dren's explanatory style for good events moreuseful than that of adults in predicting depres-sion.

2. The origins of explanatory style. Thehelplessness reformulation is silent about an-tecedents. We are turning our research atten-tion to this issue, and it promises to be exciting.Several preliminary findings are worth men-tioning. First, in our study of explanatory styleand depressive symptoms among children (Se-ligman et al, 1984), we obtained correspond-ing scores from the mothers of 47 children inour sample and from the fathers of 36. Wefound that explanatory style for bad eventsand depressive symptoms were correlated formothers and their children: explanatory style(r = .39, p < .01) and depression (r = .37, p <.01). Mothers' explanatory style also correlatedwith children's depression (r = .42, p < .005),whereas children's explanatory style weaklycorrelated with mothers' depression (r = .27,p < .10). Scores for fathers were unrelated tothose of their children and to those of theirmates. In our content analysis of the Berkeley-Oakland Study, explanatory style of mothersand sons (but not of mothers and daughters)also converged.

There are two other important hints aboutthe origins of explanatory style. In additionto learning it from one's mother, one may learnfrom the type of criticisms leveled by teachers.Dweck and Licht (1980) looked at the expla-nations that teachers gave in the third gradewhen boys and girls failed in the classroom.They found that teachers criticized girls withinternal, stable, and global statements. Boys,on the other hand, were criticized with moreunstable and specific explanations, such as"You're being rowdy", "You're not concen-trating."


In other research, Dweck and Licht (1980)gave fourth grade boys and girls unsolvableproblems and showed that boys were less sus-ceptible to helplessness than were girls. Boysalso tended to give less internal, less stable,and less global explanations for their failureson these tasks than did girls. Boys said: "Iwasn't trying hard", "I wasn't paying atten-tion", and "I don't care about your test." Girlsattributed their failure to incompetence andstupidity. Dweck and Licht (1980) argued thatthe negative explanatory style of the girls rel-ative to the boys may originate from learningtheir teachers' criticisms. That women showmore depressive symptoms as adults than domen (Radloff, 1975) may indicate that theyhave learned an explanatory style as girls thatmakes them vulnerable to depression aswomen.

Finally, the reality of one's first trauma mayplay an important role in setting explanatorystyle. Brown and Harris (1978) interviewedlower- and middle-class women in South Lon-don and found that almost 20% of the lower-class women showed severe symptoms ofdepression. An unusually high percentage ofthese depressed women had lost their mothersby death before they were 11 years old. Thedeath of mother for a young girl, as opposedto the death of mother for a teenager, has morestable consequences. In addition, for a younggirl, as opposed to a teenager, a large part ofher repertoire (global) is wiped out when hermother dies. It may be the case, then, thatbecause a mother dying represents a stableand global loss for a young girl, and becauseher personal helplessness is thereby under-scored, that as an adult she imposes internal,stable, and global explanations on other badevents, rendering her vulnerable to depression.

In summary, then, there are at least threemeans by which explanatory style may be ac-quired. First, one may learn it by imitatingparents, particularly the primary care giver, inour society the mother. Second, the criticismsthat are launched at one by teachers followingfailure may teach an explanatory style. Finally,the reality of one's first traumatic loss—theextent to which it is actually internally, stably,and globally caused—may set one's explana-tory style for life.

3. Explanatory style as a trait. We viewexplanatory style as a trait, analogous to lib-

eralism in politics or vanity in interpersonalrelations. We believe this for three reasons.First, the ASQ is designed to look at expla-nations given across situations: half bad, halfgood. An explanatory style is derived fromcross-situational consistency, and this is partof what is meant by trait. Second, some peoplehave more or less consistency across situations.For example, some individuals average 5.0 onstability for bad events because they answer 5to each of the six bad situations for stability,whereas others average 5.0 because they chosethree 7s and three 3s. These latter individualscan hardly be said to have a trait with respectto stability. When we calculate the standarddeviation of ratings for each dimension anddiscard those individuals who have large stan-dard deviations, we find that correlations withdepression and other outcomes go up. Third,explanatory style shows fairly high stabilityacross time.

Although we believe explanatory style istraitlike, we do not believe it is invariant—nor, incidentally, are traits like liberalism andvanity. We have two reasons for taking theplasticity of explanatory style seriously. First,although we have argued that explanatory styleaffects depression, we also suspect that depres-sion affects explanatory style. Persons and Rao(1981) found that explanatory style changedfor the better among patients as their depres-sion lifted. So also did Castellon et al. (1982)and Hamilton and Abramson (1983). Therapyand other good or bad life events may changeexplanatory style (Seligman, 1981). We do afterall acquire it at some time in our lives. Second,some individuals may display consistent ex-planations of a given type at one time andconsistent explanations of another type at an-other time. Mr. Q's explanations, for example,predicted his depressive mood shifts, but theydid not cohere into a style. On some days, heexplained bad events internally, stably, andglobally, whereas on other days, he explainedthem externally, unstably, and specifically.

Future investigations should adopt a moresophisticated view of explanatory style. Itshould be treated as a dependent variable thatcan be modified by life events, as well as anindependent variable that modifies futureevents. Also, the likelihood that there is a bi-directional influence between depression andexplanatory style should be explored. And fi-


nally, the term style should be reserved forindividuals whose causal explanations showlow variability across time and situations.

4. Seldonics. There is one implication ofour research program that we believe to haveuncommon promise. We call it Seldonics, afterHari Seldon (11,988-12,069 G.E.), the fic-tional psychohistorian who developed a science"that deals with the overall reactions of largegroups of human beings to given stimuli undergiven conditions" (Asimov, 1982, p. 12). Byassuming a large enough group of people (anda set of behavioral laws and mathematical re-lations left to the reader's imagination), Seldonfounded a powerful psychohistory that pre-dicts, for human conglomerates, crises andtheir resolutions far into the future. No suchclaims are made here.

However, the technique of verbatim analysishas intriguing possibilities. First, it can be re-liably done. Second, it converges with a ques-tionnaire that seems to predict which humanconglomerates will react to failure with pas-sivity and depression and which will react withactive coping attempts and good cheer. Third,it may predict when a group reacts to failurewith helplessness (Berkeley-Oakland Study),and it did predict with accuracy when oneindividual (Mr. Q) became depressed and whenhe became nondepressed. Finally, it is whollyunobtrusive and nonreactive, able to be usedwith individuals quick or dead or otherwiseimpractical to test with questionnaires.

Let us imagine that one wants to predict(or postdict) whether a terrorist will react tothe bombing of his headquarters with redou-bled terroristic acts or with passivity. Similarly,imagine that one wants to predict (or postdict)how a human conglomerate like a given salesforce will react to a bad event like an unsuc-cessful product line. Will this sales force sellmore versus less of the established productlines than will a second sales force?

AH that is needed to attempt these predic-tions is (a) prior verbatim material (if the ASQhas not been administered) such as diaries,letters, high school compositions, therapytranscripts, and quotes from the newspaper,in the case of individuals, or a composite ofsuch material, in the case of conglomerates;(b) coding and analysis of the explanatory pro-file for failure contained in this material; and(c) reasonable operationalizations of failure

and of coping. Our speculations are testable,and if true, they would be consistent with HariSeldon's program.

To conclude, we have described the empir-ical support so far obtained for the helplessnessreformulation. Results to date support themodel. We have briefly sketched some areasin need of further theoretical and empiricalscrutiny. The attributional reformulationpromises to be a useful account of depression.Explanatory style, in conjunction with actualbad events, precede the development of de-pressive symptoms. Attention to them may bea practical means of predicting who is at riskfor depression. Assaying them may help di-agnose depression with specificity and sensi-tivity. And interventions directed at changingthem may be an effective means of combatingdepression.

ReferencesAbramson, L. Y., & Sackeim, H. A. (1977). A paradox

in depression: Uncontrollability and self-blame. Psy-chological Bulletin, 84, 835-851.

Abramson, L. Y., Seligman, M. E. P., & Teasdale, J. D.(1978). Learned helplessness in humans: Critique andreformulation. Journal of Abnormal Psychology, 87, 49-74.

Affleck, G., Allen, D., McGrade, B. J., & McQueeney, M.(1982). Maternal causal attributions at hospital dischargeof high risk infants. American Journal of Mental De-ficiency, 86, 575-580.

Alloy, L. B., Peterson, C., Abramson, L. Y., & Seligman,M. E. P. (1984). Attributional style and the generalityof learned helplessness. Journal of Personality and SocialPsychology, 46, 681-687.

Anderson, C. A. (1983). Motivational and performancedeficits in interpersonal settings: The effect of attribu-tional style. Journal of Personality and Social Psychology,45, 1136-1141.

Anderson, C. A., & Jennings, D. L. (1980). When expe-riences of failure promote expectations of success: Theimpact of attributing failure to ineffective strategies.Journal of Personality, 48, 393-407.

Asimov, I. (1982). Foundation's edge. Garden City, NY:Doubleday.

Bard, M., & Dyk, R. B. (1956). The psychodynamic sig-nificance of beliefs regarding the cause of serious illness.Psychoanalytic Review, 43, 146-162.

Beck, A. T. (1967). Depression: Clinical, experimental,and theoretical aspects. New York: Hoeber.

Beck, A. T., & Beck, R. W. (1972). Screening depressedpatients in family practice: A rapid technic. PostgraduateMedicine, 52, 81-85.

Blaney, P. H., Behar, V., & Head, R. (1980). Two measuresof depressive cognition: Their association with depressionand with each other. Journal of Abnormal Psychology,89, 678-682.

Brown, G. H., & Harris, T. (1978). Social origins ofdepression. New York: Free Press.


Bukstel, L. H., & Kilmann, P. R. (1980). Psychologicaleffects of imprisonment on confined individuals. Psy-chological Bulletin, 88, 469-493.

Bulman, R. J., & Wortman, C. B. (1977). Attributions ofblame and coping in the "real world": Severe accidentvictims react to their lot. Journal of Personality andSocial Psychology, 35, 351-363.

Bumberry, W., Oliver, J. M., & McClure, J. N. (1978).Validation of the Beck Depression Inventory in a uni-versity population using psychiatric estimate as the cri-terion. Journal of Consulting and Clinical Psychology,46, 150-155.

Campbell, D. X, & Fiske, D. W. (1959). Convergent anddiscriminant validation by the multitrait-multimethodmatrix. Psychological Bulletin, 56, 81-105.

Castellon, C., Ollove, M., & Seligman, M. E. P. (1982).Unpublished data, University of Pennsylvania.

Cole, C. S., & Coyne, J. C. (1977). Situational specificityof laboratory induced learned helplessness. Journal ofAbnormal Psychology, 86, 615-623.

Coyne, J. C., Aldwin, C., & Lazarus, R. S. (1981). Depres-sion and coping in stressful episodes. Journal of Ab-normal Psychology, 90, 439-447.

Coyne, J. C., & Gotlib, I. H. (1983). The role of cognitionin depression: A critical appraisal. Psychological Bulletin,94, 472-505.

Cronbach, L. J. (1951). Coefficient alpha and the internalstructure of tests. Psychometrika, 16, 297-334.

Cutrona, C. E. (1983). Causal attributions and perinataldepression. Journal of Abnormal Psychology, 92, 161-172.

Danker-Brown, P., & Baucom, D. H. (1982). Cognitiveinfluences on the development of learned helplessness.Journal of Personality and Social Psychology, 43, 793-801.

Dweck, C. S., & Licht, B. (1980). Learned helplessnessand intellectual achievement. In J. Garber & M. E. P.Seligman (Eds.), Human helplessness (pp. 197-221).New York: Academic Press.

Eaves, G., & Rush, A. J. (1982). Cognitive patterns insymptomatic and remitted unipolar depression. Un-published manuscript, University of Texas.

Eaves, G. & Rush, A. J. (1984). Cognitive patterns insymptomatic and remitted unipolar major depression.Journal of Abnormal Psychology, 93, 31-40.

Elder, G. (1974). Children of the Great Depression. Chicago:University of Chicago Press.

Elder, G., Bettes, B. A., & Seligman, M. E. P. (1982).Unpublished data, Cornell University.

Epstein, S. (1980). The stability of behavior: Implicationsfor psychological research. American Psychologist, 89,678-682.

Feather, N. D., & Davenport, P. R. (1981). Unemploymentand depressive affect: An attributional analysis. Journalof Personality and Social Psychology, 41, 422-436.

Firth, J., & Brewin, C. (1982). Attributions and recoveryfrom depression: A preliminary study using cross-laggedcorrelational analysis. British Journal of Clinical Psy-chology, 21, 229-230.

Frankel, A., & Snyder, M. L. (1978). Poor performancefollowing unsolvable problems: Learned helplessness oregotism? Journal of Personality and Social Psychology,36, 1415-1423.

Freud, S. (1957). Mourning and melancholia. In J. Strachey(Ed. and Trans.), Standard edition of the complete psy-

chological works ofSigmund Freud (Vol. 14, pp. 243-258). London: Hogarth. (Originally published, 1917.)

Garber, J., & Hollon, S. D. (1980). Universal versus personalhelplessness in depression: Belief in uncontrollabilityor incompetence? Journal of Abnormal Psychology, 89,56-66.

Garber, J., & Seligman, M. E. P. (Eds.). (1980). Humanhelplessness. New York: Academic Press.

Goffman, E. (1961). Asylums. Garden City, NY: Anchor.Golin, S., Sweeney, P. D., & Shaeffer, D. E. (1981). The

causality of causal attributions in depression: A cross-lagged panel correlational analysis. Journal of AbnormalPsychology, 90, 14-22.

Gong-Guy, E., & Hammen, C. (1980). Causal perceptionsof stressful events in depressed and nondepressed out-patients. Journal of Abnormal Psychology, 89, 662-669.

Gottschalk, L. A., &Gleser, G. C. (1969). The measurementof psychological states through the content analysis ofverbal behavior. Los Angeles: University of CaliforniaPress.

Hamilton, E. W., & Abramson, L. Y. (1983). Cognitivepatterns and major depressive disorder: A longitudinalstudy in a hospital setting. Journal of Abnormal Psy-chology, 92, 173-184.

Hammen, C., & Cochran, S. D. (1981). Cognitive correlatesof life stress and depression in college students. Journalof Abnormal Psychology, 90, 23-27.

Hammen, C., & deMayo, R. (1982). Cognitive correlatesof teacher stress and depressive symptoms: Implicationsfor attributional models of depression. Journal of Ab-normal Psychology, 91, 96-101.

Harvey, J. H., Ickes, W. J., & Kidd, R. F. (Eds.). (1976).New directions in attribution research (Vol. 1). Hillsdale,NJ: Erlbaum.



Heider, F. (1958). The psychology of interpersonal relations.New York: Wiley.

Hiroto, D. S., & Seligman, M. E. P. (1975). Generality oflearned helplessness in man. Journal of Personality andSocial Psychology, 31, 311-327.

Janoff-Bulman, R. (1979). Characterological versus be-havioral self-blame: Inquiries into depression and rape.Journal of Personality and Social Psychology, 37, 1798-1809.

Kovacs, M., & Beck, A. T. (1977). An empirical-clinicalapproach toward a definition of childhood depression.In J. G. Schulterbrandt & A. Raskin (Eds.), Depressionin childhood: Diagnosis, treatment, and conceptualmodels (pp. 1-25). New York: Raven Press.

Langer, E. J. (1978). Rethinking the role of thought insocial interaction. In J. H. Harvey, W. J. Ickes, &R. F. Kidd (Eds.), New directions in attribution research(Vol. 2, pp. 35-58). Hillsdale, NJ: Erlbaum.

Lewinsohn, P. M., Steinmetz, J. L., Larson, D. W., &Franklin, J. (1981). Depression-related cognitions: An-tecedent or consequence. Journal of Abnormal Psy-chology, 90, 213-219.

Lloyd, C. (1980). Life events and depressive disorder re-viewed: I. Events as predisposing factors. II. Events as


precipitating factors. Archives of General Psychiatry, 3 7,529-548.

Luborsky, L. (1964). A psychoanalytic research on mo-mentary forgetting during free association. PhiladelphiaAssociation of Psychoanalysis Bulletin, 14, 119-137.

Luborsky, L. (1970). New directions in research on neuroticand psychosomatic symptoms. American Scientist, 58,661-668.

Lusted, L. B. (1968). Introduction to medical decisionmaking. Springfield, IL: Charles C Thomas.

MacCorquodale, K., & Meehl, P. E. (1948). On a distinctionbetween hypothetical constructs and intervening vari-ables. Psychological Review, 55, 95-107.

Maier, S. R, & Jackson, R. L. (1979). Learned helplessness:All of us were right (and wrong): Inescapable shock hasmultiple effects. In G. H. Bower (Ed.), The psychologyof learning and motivation (Vol. 13, pp. 155-218). NewVbrk: Academic Press.

Maier, S. E, & Seligman, M. E. P. (1976). Learned help-lessness: Theory and evidence. Journal of ExperimentalPsychology: General, 105, 3-46.

Maier, S. R, Seligman, M. E. P., & Solomon, R. L. (1969).Pavlovian fear conditioning and learned helplessness.In B. A. Campbell & R. A. Church (Eds.), Punishmentand aversive behavior (pp. 229-243). New \brk: Ap-pleton-Century-Crofts.

Manly, P. C., McMahon, R. J., Bradley, C. R, & Davidson,P. O. (1982). Depressive attributional style and depressionfollowing childbirth. Journal of Abnormal Psychology,91, 245-254.

McFarland, C., & Ross, M. (1982). Impact of causal at-tributions on affective reactions to success and failure.Journal of Personality and Social Psychology, 43, 612-617.

Metalsky, G. I., Abramson, L. Y., Seligman, M. E. P.,Semmel, A., & Peterson, C. (1982). Attributional stylesand life events in the classroom: Vulnerability and in-vulnerability to depressive mood reactions. Journal ofPersonality and Social Psychology, 43, 612-617.

Miller, I. W., & Norman, W. H. (1979). Learned help-lessness in humans: A review and attribution theorymodel. Psychological Bulletin, 86, 93-119.

Miller, I. W., & Norman, W. H. (1981). Effects of attri-butions for success on the alleviation of learned help-lessness and depression. Journal of Abnormal Psychology,90, 113-124.

Mukherji, B. R., Abramson, L. Y., & Martin, D. J. (1982).Induced depressive mood and attributional patterns.Cognitive Therapy and Research, 6, 15-21.

Navarra, A. (1981). Unpublished data, University ofPennsylvania.

Nisbett, R., & Wilson, T. D. (1977). Telling more than wecan know; Verbal reports on mental processes. Psycho-logical Review, 84, 231-259.

Nolen-Hoeksema, S. (1983). Attributional styles, life events,and depression in children. Unpublished master's thesis,University of Pennsylvania, Philadelphia.

O'Hara, M. W., Rehm, L. P., & Campbell, S. B. (1982).Predicting depressive symptomatology: Cognitive-be-havioral models and postpartum depression. Journal ofAbnormal Psychology, 91, 457-461.

Orne, M. T. (1962). On the social psychology of the psy-chology experiment: With particular reference to de-mand characteristics and their implications. AmericanPsychologist, 17, 776-783.

Overmier, J. B., & Seligman, M. E. P. (1967). Effects ofinescapable shock upon subsequent escape and avoid-ance learning. Journal of Comparative and PhysiologicalPsychology, 63, 23-33.

Pasahow, R. J. (1980). The relation between an attributionaldimension and learned helplessness. Journal of Abnor-mal Psychology, 89, 358-367.

Peirce, C. S. (1955). The philosophical writings ofPeirce.J. Buchler (Ed.). New York: Dover.

Persons, J. B., & Rao, P, A. (1981). Cognitions and depres-sion in psychiatric inpatients. Unpublished manuscript,University of Pennsylvania.

Peterson, C. (1978). Learning impairment following in-soluble problems: Learned helplessness or altered hy-pothesis pool? Journal of Experimental Social Psy-chology, 14, 53-68.

Peterson, C., Bettes, B. A., & Seligman, M. E. P. (1982).Spontaneous attributions and depressive symptoms.Unpublished manuscript, Virginia Polytechnic Instituteand State University.

Peterson, C., & Conn, M. K. (1982). Depressive moodreactions to breaking up: A learned helplessness per-spective. Unpublished manuscript, Virginia PolytechnicInstitute and State University.

Peterson, C., Luborsky, L., & Seligman, M. E. P. (1983).Attributions and depressive mood shifts: A case studyusing the symptom-context method. Journal of Ab-normal Psychology, 92, 96-103.

Peterson, C., Nutter, J., & Seligman, M. E. P. (1982).Unpublished data, Virginia Polytechnic Institute andState University.

Peterson, C, & Raps, C. S. (1983). Depression, attributions,and attributional style. Manuscript in preparation, Vir-ginia Polytechnic Institute and State University.

Peterson, C., Schwartz, S. M., & Seligman, M. E. P. (1981).Self-blame and depressive symptoms. Journal of Per-sonality and Social Psychology, 41, 253-259.

Peterson, C., & Seligman, M. E. P. (1981). Helplessnessand attributional style in depression. Tiddsskrift forNorsk Psykologforening, 18, 3-18; 53-59.

Peterson, C., Semmel, A., von Baeyer, C., Abramson,L. Y, Metalsky, G. I., & Seligman, M. E. P. (1982).The Attributional Style Questionnaire. Cognitive Ther-apy and Research, 6, 287-299.

RadlofF, L. (1975). Sex differences in depression: The effectsof occupational and marital status. Sex Roles, 1, 249-265.

Raps, C. S., Peterson, C., Jonas, M., & Seligman,M. E. P. (1982). Patient behavior in hospitals: Help-lessness, reactance, or both? Journal of Personality andSocial Psychology, 42, 1036-1041.

Raps, C. S., Peterson, C., Reinhard, K. E., Abramson,L. Y, & Seligman, M. E. P. (1982). Attributional styleamong depressed patients. Journal of Abnormal Psy-chology, 91, 102-108.

Roth, S. (1980). A revised model of learned helplessnessin humans. Journal of Personality, 48, 103-133.

Rotter, J. B. (1966). Generalized expectancies for internalversus external control of reinforcement. PsychologicalMonographs, 80(1, Whole No. 609).

Schulterbrandt, J. G., & Raskin, A. (Eds.). (1977). Depres-sion in children: Diagnosis, treatment and conceptualmodels. New York: Raven Press.

Schwab, J. J., Bialow, M. R., Brown, J. M., Holzer, C. E.,& Stevenson, B. E. (1967). Sociocultural aspects of


depression in medical inpatients. Archives of GeneralPsychiatry, 17, 539-543.

Seligman, M. E. P. (1972). Learned helplessness. AnnualReview of Medicine, 23, 407-412.

Seligman, M. E. P. (1974). Depression and learned help-lessness. In R. J. Friedman & M. M. Katz (Eds.), Thepsychology of depression: Contemporary theory and re-search (pp. 83-113). Washington, DC: Winston.

Seligman, M. E. P. (1975). Helplessness: On depression,development, and death. San Francisco: Freeman.

Seligman, M. E. P. (1978). Comment and integration.Journal of Abnormal Psychology, 87, 165-179.

Seligman, M. E. P. (1981). A learned helplessness pointof view. In L. P. Rehm (Ed.), Behavior therapy fordepression: Present status and future directions (pp. 123-141). New \ork: Academic Press.

Seligman, M. E. P., Abramson, L. Y., Semmel, A., & vonBaeyer, C. (1979). Depressive attributional style. Journalof Abnormal Psychology, 88, 242-247.

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

Seligman, M. E. P., Maier, S. F., & Solomon, R. L. (1971).Unpredictable and uncontrollable avcrsive events. InF. R. Brush (Ed.), Aversive conditioning and learning(pp. 347-400). New York: Academic Press.

Seligman, M. E. P., & Peterson, C. (in press). A learnedhelplessness perspective on childhood depression: Theoryand research. In M. Rutter, C. Izard, & P. Read (Eds.),Depression in childhood: Developmental perspectives.New York: Guilford.

Seligman, M. E. P., Peterson, C., Kaslow, N. J., Tanenbaum,R. L., Alloy, L. B., & Abramson, L. Y. (1984). Ex-planatory style and depressive symptoms among chil-dren. Journal of Abnormal Psychology, 93, 235-238.

Seligman, M. E. P., & Weiss, J. (1980). Coping behavior:Learned helplessness, physiological activity, and learnedinactivity. Behaviour Research and Therapy, 18, 459-512.

Silver, R. L., Wortman, C. B., & Klos, D. S. (1982). Cog-nitions, affect, and behavior following uncontrollableoutcomes: A response to current human helplessnessresearch. Journal of Personality, 50, 480-514.

Simonton, D. K. (1981). The library laboratory: Archivaldata in personality and social psychology. In L. Wheeler(Ed.), Review of personality and social psychology (Vol.2, pp. 217-243). Beverly Hills, CA: Sage.

Spitzer, R. L., & Endicott, J. (1977). Schedule for affectivedisorders and schizophrenia—Lifetime version. NewYork: Biometrics Research, New York State PsychiatricInstitute.

Sweeney, P. D., Schaeffer, D., & Golin, S. (1982). Attri-bution about self and others in depression. Personalityand Social Psychology Bulletin, 8, 37-42.

Taylor, S. E. (1979). Hospital patient behavior: Reactance,helplessness, or control? Journal of Social Issues, 35,156-184.

Tennen, H. (1982). A re-view of cognitive mediators inlearned helplessness. Journal of Personality, 50, 526-541.

Tennen, H., Affleck, G., Allen, D., McGrade, B. J., &Ratzan, S. (1983). Causal attributions and coping injuvenile diabetes. Unpublished manuscript, Universityof Connecticut.

Tennen, H., & Eller, S. J. (1977). Attributional componentsof learned helplessness and facilitation. Journal of Per-sonality and Social Psychology, 35, 265-271.

Tesiny, E. P., & Lefkowitz, M. M. (1982). Childhooddepression: A 6-month follow-up study. Journal of Con-sulting and Clinical Psychology, 50, 778-780.

Thoits, P. A. (1983). Dimensions of life events that influencepsychological distress: An evaluation and synthesis ofthe literature. In H. Kaplan (Ed.), Psychosocial stress:Trends in theory and research (pp. 33-103). New York:Academic Press.

Timko, C., & JanofT-Bulman, R. (1983). Attributions, vul-nerability, and coping: The case of breast cancer. Un-published manuscript, University of Massachusetts.

Vaillant, G. E. (1977). Adapatation to life. Boston: Little,Brown.

Velten, E. (1968). A laboratory task for induction of moodstates. Behaviour Research and Therapy, 6, 473-482.

Winer, B. J. (1971). Statistical principles in experimentaldesign (2nd ed.). New \brk: McGraw-Hill.

Wong, P. T. P., & Weiner, B. (1981). When people ask"why" questions, and the heuristics of attribution search.Journal of Personality and Social Psychology, 40, 649-663.

Wortman, C. B., & Brehm, J. W. (1975). Response touncontrollable outcomes: An integration of reactancetheory and the learned helplessness model. In L. Ber-kowitz (Ed.), Advances in experimental social psychology(Vol. 8, pp. 277-336). New York: Academic Press.

Wortman, C. B., & Dintzer, L. (1978). Is an attributionalanalysis of the learned helplessness phenomenon viable?:A critique of the Abramson-Seligman-Teasdale refor-mulation. Journal of Abnormal Psychology, 87, 75-90.

Zemore, R., & Johansen, L. J. (1980). Depression, help-lessness, and failure attributions. Canadian Journal ofBehavioural Sciences, 12, 167-174.

Zuckerman, M., & Lubin, B. (1965). Manual for the Mul-tiple Affect Adjective Check List. San Diego, CA: Ed-ucational and Industrial Testing Service.

Received May 18, 1983Revision received December 15, 1983

causal explanations as a risk factor for depression: theory ......depression with respect to...

Documents