Causal Explanations As a Risk Factor for Depression: Theory and Evidence Christopher Peterson Martin E

Causal Explanations as a Risk Factor for Depression: Theory and Evidence Christopher Peterson Martin E. P. Seligman Virginia Polytechnic Institute and University of Pennsylvania State University

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

After experiencing uncontrollable events, a Our primary purpose here is to describe the variety of organisms show cognitive, motiva- empirical investigations of this theory con- tional, and emotional deficits (Maier & Selig- ducted by our research group, and to evaluate man, 1976; Seligman, 1975). This learned the theoretical status of the attributional helplessness phenomenon has parallels with model. The juxtaposition of conceptual clar- depression in people and has been proposed ification with evidence from several levels of as a model of this psychopathology (Seligman, analysis may allow a valid assessment of the 1974). When helplessness theory proved un- theory. able to account for the generality and chronicity of depressive symptoms or for self-es- Theory teem loss in depression, it was revised along Helplessness Theory: Old and New attributional lines (Abramson, Seligman, & Teasdale, 1978). According to this reformu- The learned helplessness phenomenon was lation, the boundary conditions of depression first described systematically by animal learn- following bad events are determined by causal ing researchers at the University of Pennsyl- attributions about the events (for similar for- vania (Overmier & Seligman, 1967; Seligman mulations, see Miller & Norman, 1979, and & Maier, 1967). Mongrel dogs, following ex- Roth, 1980). The central prediction of the re- posure to inescapable electric shock, showed formulation is that individuals who have an striking deficits 24 hours later when placed in explanatory style that invokes internal, stable, a shuttlebox in which the simple act of crossing and global causes for bad events tend to be- a barrier would terminate shock. Unlike dogs come depressed when bad events occur. not previously exposed to uncontrollable shock, these animals seemed helpless. They initiated few attempts to escape the shock This research was supported by U.S. Public Health Ser- (motivational deficit). They were not likely to vice Grant MH-19604 to Martin E. P. Seligman. follow an occasionally successful response with The authors thank the many individuals at the University another (learning or cognitive deficit). They of Pennsylvania and elsewhere who contributed to this research, the anonymous reviewers for their comments on did not evidence much overt emotionality an earlier version of this article, and James Austin, Peter while being shocked (emotional deficit). Villanova, and Steve Zaccaro for their help in polishing These deficits were interpreted in cognitive the final draft. terms (Maier, Seligman, & Solomon, 1969; Se- Requests for reprints should be sent to Martin E. P. ligman, Maier, & Solomon, 1971). During ex- Seligman, Department of Psychology, University of Penn- sylvania, 3815 Walnut Street, Philadelphia, Pennsylvania posure to the electric shocks, the dogs learned 19104. that shocks were independent of responses.

347 348 CHRISTOPHER PETERSON AND MARTIN E. P. SELIGMAN

Regardless of what the dogs did or did not do, depression and about the paradox of self-es- the shocks occurred. This learning was rep- teem loss following helplessness. resented as an expectation of future response- To address these shortcomings, Abramson, outcome independence (i.e., uncontrollability) Seligman, and Teasdale (1978) revised help- that was generalized to new situations to pro- lessness theory to include the individual's duce the observed deficits. causal explanations of the original bad events. The cognitive interpretation of animal According to this revision, when people face helplessness has been controversial. Some re- uncontrollable bad events, they ask why. Their searchers have attempted to explain the phe- answer affects how they react to the events. nomenon in peripheral or biological terms, Abramson et al. (1978) argued that three ex- whereas other researchers have argued for the planatory1 dimensions are relevant. First, the cognitive explanation (for reviews of this con- cause may be something about the person (in- troversy, see Maier & Jackson, 1979; Maier & ternal explanation), or it may be something Seligman, 1976; Seligman & Weiss, 1980). Re- about the situation or circumstances (external gardless of this controversy, psychologists in- explanation). Second, the cause may be a factor terested in human adaptation were quick to that persists across time (stable explanation), sense the possible pertinence of learned help- or it may be transient (unstable explanation). lessness to failures of human action. Two par- Third, the cause may affect a variety of out- allel lines of research with human subjects took comes (global explanation), or it may be lim- place. ited just to the event of concern (specific ex- In the first, the basic helplessness phenom- planation). Table 1 presents examples of these enon was investigated in the laboratory with types of explanations as they might be made human subjects (for a review, see Wortman & about the bad event "My checking account is Brehm, 1975). In the second, helplessness the- overdrawn." ory was used to explain a variety of human The reformulation assigns particular roles difficulties (see Garber & Seligman, 1980). to each of these three dimensions. Internality Perhaps the best known of these applications of causal beliefs affects self-esteem loss follow- has been Seligman's (1972, 1974, 1975) sug- ing bad events. If the person explains a bad gestion that learned helplessness may model event by an internal factor, then self-esteem depression with respect to symptoms, causes, loss is more likely to occur. If a person explains preventions, and cures. the event by an external factor, then self-esteem The main problem with the original help- loss is less likely to occur. Stability of causal lessness model, applied both to human help- beliefs affects the chronicity of helplessness lessness in the laboratory and to natural hu- and depression following bad events. If a bad man depression, is its failure to account for event is explained by a cause that persists, boundary conditions. Sometimes laboratory depressive reactions to that event tend to per- helplessness is general (e.g., Hiroto & Selig- sist. If the event is explained by a transient man, 1975), and sometimes it is circumscribed factor, then depressive reactions tend to be (e.g., Cole & Coyne, 1977). Sometimes bad short lived. Finally, globality of causal beliefs events precipitate depressive reactions (occa- influences the pervasiveness of deficits follow- sionally transient, occasionally long lasting), ing bad events. If one believes that a global and sometimes they do not (e.g., Brown & factor has caused a bad event, then helplessness Harris, 1978; Lloyd, 1980). What determines the chronicity and generality of helplessness and depression? 1 Because of the numerous and sometimes contradictory Similarly, the original model does not ex- senses in which the terms attribution and altributional style are used, we here refer to causal explanation and plain the self-esteem loss frequently observed explanatory style. In particular, attribution by dictionary among depressives (Beck, 1967; Freud, 1917/ definition is a broader term than is explanation, and in- 1957). Why should individuals blame them- volves imputing any property to an object. Because the selves for events over which they perceive no only property that concerns us here is causal characteristics, we prefer explanation to attribution. Although this shift control (Abramson & Sackeim, 1977)? The in terminology may be confusing in the short run, we simple helplessness model is silent about the believe that clarity will eventually be served by the use of chronicity and generality of helplessness and these more precise terms. CAUSAL EXPLANATIONS AND DEPRESSION 349

Table 1 and an increase in susceptibility to disease. In Examples of Causal Explanations for the Event addition, the symptom of self-esteem loss is "My Checking Account is Overdrawn" sometimes one of the symptoms of helplessness. These symptoms, taken together, look Explanation very much like the syndrome of depression. Style Internal External How do these symptoms come about? In learned helplessness theory, both original Stable and reformulated, the expectation that no ac- Global "I'm incapable of "All institutions tion will control outcomes in the future is a doing anything chronically make right" mistakes" sufficient condition for the production of all Specific "I always have "This bank has always of these symptoms of helplessness except self- trouble figuring used antiquated esteem loss. This expectation is represented my balance" techniques" in the center of Figure 1. Whenever and wher- Unstable Global "I've had the flu "Holiday shopping ever this expectation occurs, the symptoms for a few weeks, demands that one will develop. Thus, our main concern is how and I've let throw oneself into other processes and events conspire to bring everything it" about this expectation, resulting in the symp- slide" toms of helplessness. This expectation is usu- Specific "The one time I "I'm surprised — my didn't enter a bank has never ally triggered when bad events are perceived check is the made an error as uncontrollable. The reality of the bad un- one time my before" controllable events influences the content of account gets the expectation. So, for example, if the bad overdrawn" event is blindness, and one's job is proofread- ing, the range of outcomes to which the expectation applies includes work. deficits tend to occur in a variety of different Explanations and explanatory style also in- situations. If one believes that a more specific fluence the expectation that no action will factor is the cause, the deficits tend to be cir- control outcomes in the future (see Figure 1). cumscribed. Explanations involving global causes tend to produce the expectation that action will not Conceptual Status of Explanation control many outcomes, which in turn pro- and Explanatory Style duces the symptoms of helplessness in exactly Causal explanations: Sufficient condition or that large range of situations. In parallel, if the risk factor for depression? We used qualified cause of a bad event is explained by stable language above in stating the central predic- factors, the expectation tends to occur for a tions of the reformulation. So, an internal ex- long time into the future, and therefore, the planation for a bad event is said to make self- symptoms of helplessness are long-lasting. If esteem loss more likely, but not to cause self- the explanation for a bad event is internal, esteem loss. It is important to realize that then the symptom of lowered self-esteem tends explanations and their precursor, explanatory to be displayed. Thus, the particular expla- style, are not sufficient to produce depressive nation an individual makes for the bad event deficits but rather are risk factors for such def- influences the generality and time course of icits (Abramson et al., 1978). the symptoms of helplessness, as well as the The relationship among variables in the loss of self-esteem. theory is diagrammed in Figure 1. Let us begin In the model, there are two influences on by looking at the extreme right-hand side of the particular explanation chosen. The first is this figure. What Figure 1 attempts to explain the reality of the bad events themselves. If the is the general process by which the symptoms bad event that sets off the expectation of help- of helplessness are produced. These symptoms lessness is the death of one's spouse, this is a are passivity; cognitive deficits; emotional def- stable and global loss. The spouse will not re- icits including sadness, anxiety, and hostility; turn, and many of the activities in which one a lowering of aggression; a lowering of appe- has customarily engaged will be undermined. titive drives; a set of neurochemical deficits; The second influence on what particular ex- 350 CHRISTOPHER PETERSON AND MARTIN E. P. SELIGMAN

Bad Events Perceived As Uncontrollable Passivity

Cognitive Expectation That No Action Deficits Will Control

Figure 1. The process of learned helplessness. planation is made is the habitual tendency to Because there is usually similarity between choose certain kinds of explanations for bad causal explanation and expectation of con- versus good events. We have been able to iden- sequences, knowing an individual's explana- tify individual patterns in the selection of tion and explanatory style will usually predict causes over a variety of events, and we call helplessness deficits. Because there can some- these patterns explanatory style. The particular times be dissimilarity, an individual's causal style that most concerns us is the depressive explanation and explanatory style will not explanatory style, in which one tends to give cause the deficits. And so we speak of these internal, stable, and global explanations for variables as risk factors (just as smoking is a bad events (it's me; it's going to last forever; risk factor for lung cancer), rather than suf- and it's going to affect everything I do). ficient conditions for helplessness and depres- It should now be apparent why a particular sion. explanation or explanatory style is not suffi- Our research focused on measuring ex- cient for the symptoms of helplessness to ap- planatory style, because a methodological pear. These variables influence the expectation, framework already existed for doing so (e.g., but it is the expectation which is sufficient. Harvey, Ickes, &Kidd, 1976, 1978, 1981). We Usually, causal explanations for an event and chose not to focus on the expectation itself, expectations about the consequences of an even though it is more proximate to the symp- event have the same properties. For example, toms, because we do not believe that a valid if the explanation for blindness is a progressive means of measuring expectations yet exists. brain disease, this cause has stable and global Also, in most cases, we chose not to focus on properties, as do the consequences of blind- particular explanations because explanatory ness. But sometimes the properties of a cause style of necessity is a more reliable individual and its consequences can be dissimilar. If, for difference (Epstein, 1980). example, the cause of blindness was a freak Causal explanations and explanatory style: accident, the cause is unstable and specific, Hypothetical constructs. The term attribution but the consequences are stable and global. has been used in a variety of ways. Some the- CAUSAL EXPLANATIONS AND DEPRESSION 351 orists regard attributions as the stated beliefs point scales corresponding to the internality, that compose an individual's naive psychology stability, and globality dimensions. The ASQ (Heider, 1958), whereas other theorists suggest does not constrain or create the causal expla- that attributions have tacit components (Lan- nations provided by the subject, but at the ger, 1978; Wortman & Dintzer, 1978). Con- same time it allows simple and objective quan- troversy over how best to assess attributions tification of responses by asking the subject to has ensued (Nisbett & Wilson, 1977). We wish rate the internality, stability, and globality of to make explicit that we regard an attribution the causes. as a hypothetical construct, a way for the the- The questionnaire is group administered, orist to explain observable behaviors. We do and the following directions appear on the first not use the notion to refer to an intervening page of the booklet: variable (cf. MacCorquodale & Meehl, 1948). As a hypothetical construct, causal expla- Please try to vividly imagine yourself in the situations that follow. If such a situation happened to you, what nation does not have a single definition nor would you feel would have caused it? While events may is it exhausted by any one operation. An in- have many causes, we want you to pick only one—the tervening variable is exhausted in its meaning major cause if this event happened to you. Please write by one class of input operations and one class this cause in the blank provided after each event. Next we of output operations. Our view of attributions want you to answer some questions about the cause. To summarize we want you to: contrasts with Heider's (1958) phenomeno- 1. Read each situation and vividly imagine it happening logical view of attributions, in which the abil- to you. ities to introspect upon and to report verbally 2. Decide what you feel would be the major cause of one's causal beliefs are necessary and sufficient the situation if it happened to you. 3. Write one cause in the blank provided. conditions for them. Heider (1958) regarded 4. Answer three questions about the cause. attributions as intervening variables. For us, 5. Go on to the next situation. on the other hand, an attribution is not a real thing, like a microphone or a typewriter. Because we are interested in style—cross-sit- Rather, it is akin to concepts like natural se- uational explanations—we describe 12 differ- lection, life, reward, preference, or—perhaps— ent hypothetical events. Half are good events the atom. These are all hypothetical constructs, (e.g., you meet a friend who compliments you measured with a number of different con- on your appearance), and half are bad events verging operations, no one of which defines (e.g., you go out on a date, and it goes badly). or exhausts the construct of interest. By this After each event are parallel questions. First, logic, not only introspections and answers to the subject is asked to write down the one questionnaires are relevant to knowing about major cause of the event. Then the subject is one's causal explanations, but so too are a asked to rate the cause along the three ex- variety of behavioral observations. In the planatory dimensions. The wording of the studies we report, we attempt to use converging various questions reflects the specific event to ways to infer the presence of causal expla- be explained, and the example in Table 2 il- nations. lustrates the nature of these questions. Attributional Style Questionnaire. One way The three ratings of each cause are scored in which we measure explanations and ex- in the directions of increasing internality, sta- planatory style is with a questionnaire that we bility, and globality. Total scores are formed developed for this purpose: the Attributional separately for the bad and good events simply Style Questionnaire (ASQ; Peterson, Semmel, by summing the appropriate items and divid- et al., 1982). This self-report instrument yields ing the sum by 6. Table 3 reports the internal scores for the explanation of bad and good consistencies, intercorrelations, and test-retest events with internal versus external, stable reliabilities of these scales for a sample of un- versus unstable, and global versus specific dergraduates at the State University of New causes. The format reflects the fact that we York at Stony Brook (Peterson, Semmel, et wanted questions to assess how much respon- al., 1982). dents used particular values of the three di- Several things should be noted about these mensions. We ask subjects to generate their data. First, the individual scales have modest own cause for each of a number of events, and reliabilities, ranging between .44 and .69. One then to rate that cause themselves along 7- reason for these low reliabilities is the small 352 CHRISTOPHER PETERSON AND MARTIN E. P. SELIGMAN number of items—six—in each scale.2 Second, answers on the ASQ reflect the natural, un- the scales are substantially intercorrelated constrained causal explanations an individual within good events and within bad events. For makes? this reason, we usually combine them into To test if the scores obtained with the ASQ overall composites for good events and bad are more than questionnaire artifacts, we asked events, thereby bolstering reliability to more 66 adults (31 males, 35 females) on the Uni- acceptable levels (alphas of .75 and .72, re- versity of Pennsylvania campus to describe in spectively). Third, causal explanations for good writing, using 250 to 300 words, the two worst events are independent of causal explanations events involving themselves during the last for bad events. Conceptions that confound the year (Peterson, Bettes, & Seligman, 1982). In- two (e.g., Rotter, 1966) incorrectly average over structions stressed descriptive accuracy, and two quite different phenomena. Fourth, the there was no mention of causal explanation. stability of explanatory style scores is respect- The subjects then completed the ASQ, which ably high, as we had hoped in assuming that had not been mentioned prior to this point. they operationalize a characteristic style of This procedure allowed us to see (a) if causal causal explanation. accounts are spontaneously included in de- Criterion validity of the ASQ. What is the scriptions; (b) if individuals are consistent validity evidence for this questionnaire? We across different explanations in terms of their present support for the construct validity be- internality, stability, and globality; and (c) if low. However, two recently completed studies these spontaneous explanations converged speak to the criterion validity of the ASQ: the with scores of the ASQ. degree to which the questionnaire predicts We had first developed a procedure by which naturally occurring causal explanations. Do spontaneous explanations in written material could be extracted and scored for internality, stability, and globality (Peterson, Luborsky, & Seligman, 1983). A researcher reads the ma- Table 2 terial and notes descriptions of bad events in- Sample Question From the Anributional volving the writer for which a causal expla- Style Questionnaire nation is made (identified by such key phrases Event: You go out on a date, and it goes badly. as "because of," "as a result of," and so on). Next, each event description and the accom- 1. Write down the one major cause of this event. panying explanation are transcribed verbatim and are shown singly to independent and blind judges who rate the explanation with 7-point scales corresponding to its internality, stability, and globality. These judges use the definitions 2. Is the cause of the date going badly due to something about you or something about other people or from the ASQ (Peterson, Semmel, et al., 1982). circumstances? Finally, the ratings by the different judges are Due to Due combined into a profile for the individual. other to For this particular study, there were four people 1234567 me judges. Coding reliabilities, estimated by (Circle one number.) Cronbach's (1951) alpha, were satisfactory: 3. In the future when dating, will this cause again be .93, .89, and .90, for the internal, stable, and present? global ratings, respectively. Table 4 presents Never Always be again present the ratings made by the judges for this excerpt present 1234567 from a written description (emphasis added): (Circle one number.) About four months ago, he called me on the telephone 4. Is the cause something that just influences dating or from , where he's been working. He told me that does it also influence other areas of your life? our relationship was over, that he didn't want to see me Just this All situa- situa- tion 1234567 tions (Circle one number.) 2 For a newly revised version of the ASQ with 24 bad events, reliabilities are substantially improved: .66 for in- Note. From Peterson, Semmel, et al., (1982). ternality, .85 for stability, and .88 for globality. CAUSAL EXPLANATIONS AND DEPRESSION 353

Table 3 Reliabilities and Intercorrelations of the Attributional Style Questionnaire

Dimension 1

Good events 1. Internality (.50) 2. Stability .62* (.58) 3. Globality .38* .59* (.44) 4. Composite (.75) Bad events 5. Internality .11 .01 -.03 (.46) 6. Stability -.17 -.07 .03 .18* (.59) 7. Globality -.15 .04 .24* .28* .45* (.69) 8. Composite .02 (.72) M 5.26 5.36 5.11 5.25 4.29 4.14 3.87 4.12 SD .79 .68 .80 .62 .84 .71 1.07 .64 Test-retest correlations .58* .65* .59* .70* .64* .69* .57* .64*

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

anymore. I felt devastated. I tried to argue, but what could two questions of the SADS interview ask the I say? I'm still flipped out. I guess I'm just no good at patient (a) why treatment was sought and (b) relationships. I've never been able to keep a man interested in me. I talked to my roommate about it all, right after the origin of their symptoms. We performed he called me. She knows me real well, plus she went to a blind content analysis—exactly as is de- high school with She told me to forget about scribed above—of the causal explanations of- him and find someone else. She said I made a mistake fered in answer to the first two questions of and the best thing to do was move on. There was never a chance for a lasting thing. But how could I misjudge things the SADS interview. so badly? When I get stars in my eyes I get carried away. Content analysis seems valid for a severely depressed population. The composite score of All of the descriptions contained at least the ASQ for bad events correlated significantly one causal explanation. We concluded that with the composite score for bad events ob- causal explanations are spontaneously offered, tained by this content analysis (r - .38, p < even when they are not explicitly prompted. .02). Furthermore, explanatory style, mea- Ratings of the first explanation offered by an sured in both ways, correlated with severity individual were correlated with those of the second, and significant correlations were obtained. Furthermore, when these ratings were combined across explanations by the same in- Table 4 dividual, they correlated respectably with the Example of Content Analysis corresponding dimensions of the ASQ. Exact Event explanation Judge Int Sta Glo values are reported in Table 5. We carried out a second study to see if the Our relationship was over. A 7 7 2 (because) I'm just no B 7 7 3 content analysis of explanatory style was valid good at relationships. C 7 7 3 in patient populations as well as in normal I've never been able to D 6 7 1 populations (Castellon, Ollove, & Seligman, keep a man interested 1982). Toward this end, we gave 40 outpatients in me. the Schedule for Affective Disorders and How could I misjudge A 5 6 3 Schizophrenia (SADS) diagnostic interview things so badly? B 5 6 2 (Spitzer & Endicott, 1977) in the same session (because) When I get C 5 7 2 as we gave them the ASQ and the Beck stars in my eyes I get D 4 7 3 Depression Inventory (BDI; Beck, 1967). carried away. These patients were diagnosed unipolar de- Note. From Peterson, Bettes, and Seligman (1982). Int: pressive episode by this interview. The first internality, Sta = stability, and Glo = globality. 354 CHRISTOPHER PETERSON AND MARTIN E. P. SELIGMAN

Table 5 Spontaneous Explanations

r with corresponding Dimension M SD Consistency ASQ score

Internality 3.71 1.54 .25** .41*** Stability 3.78 1.62 .37** .19* Globality 3.13 1.43 .33** .23* Composite 3.54 1.27 .36** .30**

Note, n ~ 66. ASQ = Attributional Style Questionnaire. *p < .10. **p < .05. ***p < .001. of depressive symptoms, as measured by in terms of task difficulty. Subjects given the BDI. "easy" unsolvable problems showed later These two studies have several important problem-solving deficits, whereas subjects implications. First, the important constructs given "difficult" unsolvable problems showed of the helplessness reformulation—explana- facilitation. tions and explanatory style—are more than Situational factors also affect explanations responses to questions about hypothetical and depressive reactions to events outside the events. Individuals offer unsolicited causal ex- laboratory. Peterson and Conn (1982) inves- planations about important bad events, and tigated the reactions of people who had just these explanations are consistent with each ended a romance. Although most subjects had other along the dimensions of internality, sta- depressive symptoms and sad affect following bility, and globality. Second, the ASQ has cri- the breakup, the most severe reactions were terion validity. Internality, stability, and glo- by those whose partner left them for someone bality ratings of spontaneous explanations else. In contrast, the least severe reactions were converged with the corresponding scales of the by those whose romance ended because their questionnaire. partner was attending a different college. Explanatory style versus reality. As we ar- If one's partner leaves for someone else, the gued above, explanations are determined by answer to the why question tends to be in terms both situational and dispositional factors. That of stable and global characteristics of the per- the reality of an event can determine expla- son left behind. Many external explanations nations is undeniable. Also, we prefer the term (e.g., my partner is not interested in romance) explanation to attribution because attribution are implausible, as are interpretations in terms connotes projection too strongly. In saying that of transient and circumscribed causes. In con- reality partly determines explanations, we are trast, if romance ends because of distance, one not saying that causal explanations are there- need not see oneself as causing the breakup. fore accurate (Langer, 1978; Nisbett & Wilson, Furthermore, one need not believe that the 1977). Rather, we are saying that causal beliefs causes will persist, or will be pervasive. Indeed, are not entertained in a vacuum, and appro- subjects whose romance ended because of dis- priate investigations of the helplessness refor- tance were more optimistic about achieving mulation must not look only at dispositional future relationships than were subjects whose determinants of explanations. partner left for someone else. A number of studies have shown manipu- Individual dispositions in the explanations lations of reality to affect causal explanations, are needed to explain why different individuals which in turn affect laboratory helplessness have different reactions to the same events. (see reviews by Abramson et al., 1978; Miller Why do some people become helpless following & Norman, 1979; and Roth, 1980). A rep- unsolvable problems and others not (Alloy, resentative study is that of Tennen and Eller Peterson, Abramson, & Seligman, 1984; (1977), who instructed subjects given unsolv- Dweck & Licht, 1980)? Why do some people able problems that the task was either easy or become depressed following bad events and difficult. Failure at an easy task encourages others not (Lloyd, 1980; Metalsky, Abramson, explanations in terms of lack of ability, whereas Seligman, Semmel, & Peterson, 1982)? The failure at a difficult task results in explanations reformulation claims that people susceptible CAUSAL EXPLANATIONS AND DEPRESSION 355 to helplessness and depression interpret these Baeyer (1979) administered the ASQ to a bad events in internal, stable, and global terms. sample of 143 college students at the University If reality is ambiguous enough, a person of Pennsylvania, along with the short form of may project and impose habitual explanations. the BDI (Beck & Beck, 1972), a 13-item self- According to the reformulation, individuals report instrument that assesses the severity of have a characteristic way of explaining events. common depressive symptoms (for validity They are consistent in the way they use internal evidence among undergraduates, see Bum- (vs. external), stable (vs. unstable), and global berry, Oliver, & McClure, 1978). As predicted (vs. specific) causes to explain bad events. by the helplessness reformulation, depressive Thus, when reality is ambiguous, the ASQ symptoms among undergraduates correlated works as a projective test and may be used to with internal (r = .41, p < .001), stable (r = measure a person's characteristic explanatory .34, p < .001), and global (r = .35, p < .001) style. If this style invokes internal, stable, and explanations for bad events (composite r = .48, global causes, then the person tends to become p < .001). depressed when bad events occur. This is the Study 2: Spontaneous Explanations central prediction of the reformulation, and we now describe converging evidence in sup- and Depression port of this prediction. We have already described our investigation Research strategies. We use a variety of of spontaneous explanations (Peterson, Bettes, research strategies: (a) cross-sectional corre- & Seligman, 1982). Sixty-six adults wrote es- lational studies, (b) causal modeling with lon- says from which causal explanations of bad gitudinal data, (c) experiments of nature, (d) events were extracted and rated for internality, laboratory experiments, and (e) case studies. stability, and globality. These explanations Each class of investigation addresses a more were consistent and converged with the cor- stringent prediction. The cross-sectional stud- responding scales of the ASQ. These subjects ies seek merely to correlate explanatory style also completed the short form of the BDI. and depressive symptoms. The longitudinal ASQ scores correlated with depressive studies investigate whether a depressive style symptoms as predicted by the reformulation: precedes depression, consistent with a predis- internality (r = .44, p < .001), stability (r = posing role. The experiments of nature look .42, p < .001), globality (r = .29, p < .02, and at how naturally occurring bad events and ex- composite (r = .45, p < .001), replicating the planatory style interact to produce a subse- results of Study 1. In addition, when ratings quent depressive reaction. The laboratory ex- of the extracted explanations were averaged periments see if experimentally manipulated across different causes offered by the same in- bad events interact with explanatory style to dividual, these scores also correlated with de- result in laboratory helplessness, a depressive pressive symptoms: internality (r = .36, p < analogue. Finally, the case studies ascertain .005), stability (r = .29, p < .02), globality the applicability of the reformulation to real (r = .33, p < .01), and composite (r= .39, lives. p< .001). Evidence The studies so far described support the predictions. However, these studies employed Cross-Sectional Correlational Investigations somewhat restricted samples: upper middle- In this section, we describe several inves- class college students who were on the whole tigations of the hypothesis that depressive not seriously depressed. To extend the em- symptoms correlate with the use of internal, pirical support for the depressive explanatory stable, and global causes to explain bad events. style, we conducted parallel investigations with These studies are cross-sectional, and employ three different samples: lower-class women, a variety of populations, using the ASQ. De- children, and psychiatric inpatients. pressive symptoms were assessed by self-report Study 3: Lower-Class Women questionnaires and by psychiatric diagnosis. Depressive disorders among the lower class Study 1: ASQ and Depression have sometimes been regarded as mainly so- In the first investigation of the reformula- matic (Schwab, Bialow, Brown, Holzer, & tion, Seligman, Abramson, Semmel, and von Stevenson, 1967). Accordingly, the relevance 356 CHRISTOPHER PETERSON AND MARTIN E. P. SELIGMAN

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

Dimension

Item Valence Varied Held constant

A good friend tells you that he hates you. Bad Stability; globality a. My friend was in a bad mood that day. External b. I wasn't nice to my friend that day. Internal You get all the toys you want on your birthday. Good Internality; globality a. People always guess what toys to get me for my birthday. Stable b. This birthday people guessed right as to what toys I wanted. Unstable You get an "A" on a test. Good Internality; stability a. I am smart. Global b. I am smart in the subject that the test was in. Specific

Note. From Seligman et al. (1984).

of a cognitive model like the helplessness re- and sixth grades participated. The subjects formulation may be questioned. For this rea- completed the Children's Depression Inven- son, we asked women from the lower socio- tory (CDI; Kovacs & Beck, 1977) and a economic class to respond to the ASQ and the Children's Attributional Style Questionnaire BDI (Navarra, 1981). Forty-one women (av- (CASQ), which we developed. During class- erage age = 38 years) on welfare in Philadel- time, the researcher read the questionnaires phia were recruited through newspaper ad- aloud while the children silently read their vertisements as research participants. In an copies. The questionnaires were completed interview format, they answered the questions twice by the same children, at an interval of of the ASQ and of the long form (21 items) 6 months. of the BDI (Beck, 1967). The CDI is a 27-item questionnaire that As in the other samples, causal explanations measures the severity of emotional, motiva- for bad events correlated with depression, al- tional, cognitive, and somatic symptoms of though the stability dimension did not reach depression. Each item consists of three self- statistical significance: internality (r = .42, p < report statements graded in severity from 0 to .01), stability (r = .12, ns), globality (r = .50, 2. The child is instructed to choose the option p < .001), and composite (r = .50, p < .001). corresponding to how he or she has been feel- ing during the preceding 2 weeks. The CASQ is a forced-choice instrument. Study 4: Children We found in pilot work that young children had trouble with the adult ASQ, particularly The helplessness reformulation purports to globality. Hence, we used a force-choice format be a general theory that should explain in which hypothetical good or bad events in- depression across the life span (Seligman & volving the child were followed by two possible Peterson, in press). Explanatory style should explanations, which varied one of the explan- correlate with depressive symptoms among atory dimensions while holding the other two children in the same pattern as among adults. constant. Sixteen questions pertain to each of Accordingly, we conducted an investigation to the three dimensions; half refer to good events, see if the depressive explanatory style char- and half to bad events. Examples of items from acterized depressed children (Seligman et al., the CASQ are shown in Table 6. 1984). The CASQ is scored by assigning a 1 to each We tested 96 children from two Philadelphia internal or stable or global response, and a 0 elementary schools that consisted predomi- to each external or unstable or specific re- nantly of white, middle-class children. Ap- sponse. Scales are formed by summing these proximately equal numbers of boys (« = 50) scores across the appropriate questions for each and girls (n = 46) from the third, fourth, fifth, of the three dimensions, separately for good CAUSAL EXPLANATIONS AND DEPRESSION 357

Table 7 Children's Attributional Style Questionnaire Scores and Depression

r with same measure r with Dimension M SD fu 6 months later CDI

Good events Internality 4.61 1.48 .32 .53** -.34** Stability 4.21 1.91 .55 .61** -.47** Globality 4.67 1.58 .40 .54** -.35** Composite 13.49 3.72 .66 .71** -.53** Bad events Internality 2.30 1.57 .43 .63** .45** Stability 2.40 1.40 .42 .52** .31* Globality 1.88 1.27 .31 .64** .21* Composite 6.58 2.77 .50 .66** .51** Children's Depression Inventory (CDI) 7.71 6.28 .86 .80** —

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 of among adults, not just at the symptom level the scales range from 0 to 8. but also in terms of explanatory style (Schul- Table 7 presents the reliabilities and sta- terbrandt & Raskin, 1977). bilities of the questionnaires completed by the children at Time 1. Statistics at Time 2 were Study 5: Depressed Patients essentially the same, and are not shown. CASQ subscales were only modestly reliable. More These four studies have two shortcomings. satisfactory reliabilities were obtained by First, few of the individuals were seriously de- combining the subscales (separately for good pressed. Although the helplessness reformu- events and for bad events). Alpha for the good lation regards mild and severe unipolar events composite was .66, and for bad events depression as continuous, at least with regard was .50 (see Table 7). The CASQ scales and to cognitive characteristics (Seligman, 1978), composites were consistent over the 6-month this is an empirical issue. Second, the research interval, showing explanatory style to be a so far described does not show the "depressive" somewhat stable individual difference among style to be specific to depression. Perhaps it is children, just as it is among adults. a general characteristic of psychopathology. CDI scores were highly reliable and highly We therefore conducted a study in which consistent over the 6 months, implying that formally diagnosed unipolar depressed inpa- childhood depression, at least as measured tients completed the ASQ (Raps, Peterson, here, is not a transient phase (Tesiny & Lef- Reinhard, Abramson, & Seligman, 1982). We kowitz, 1982). As predicted, explanatory style compared their scores with those of two other correlated with depressive symptoms (see Ta- patient groups. The first comparison group ble 7). The explanation of bad events with was medical and surgical patients, to control internal, stable, and global causes covaried for being hospitalized (Taylor, 1979). The sec- with CDI scores, as did the reverse style for ond was nondepressed schizophrenics, to con- good events. Considering the modest reliabil- trol for severity of psychopathology and length ities of the CASQ, these correlations are sub- of hospitalization (Raps, Peterson, Jonas, & stantial; depression among school children is Seligman, 1982). Although a schizophrenic closely tied to the causal explanations they group is not an exhaustive control for psy- pick. chopathology, it provides a first step toward This study broadens the empirical base of assessing specificity. the depressive explanatory style by showing it We administered the ASQ to male veterans among 9- to 13-year-old children. It also adds hospitalized at the Northport Veterans Ad- to the evidence that childhood depression is ministration Medical Center: 30 unipolar de- a coherent syndrome similar to depression pressed patients, 15 nondepressed schizo- 358 CHRISTOPHER PETERSON AND MARTIN E. P. SELIGMAN

Table 8 Attributional Style Questionnaire: Patients' Scores and Depression Medical and Depressives Schizophrenics surgical patients (n = 30) (n = 15) (« = 61)

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

Internality 4.90, 1.12 3.51y 1.18 4.30, 1.12 7.34* Stability 4.89S 1.16 4.01y 1.01 4.06y 0.90 7.67* Globality 4.84X 1.29 4.10Ky 1.19 3.65S 1.37 8.25* Composite 4.88X 2.92 3.87y 2.26 4.00y 2.73 10.69* Note. From Raps, Peterson, Reinhard, Abramson, & Seligman (1982). Means in a row are different (p < .05) by planned comparisons (Winer, 1971, pp. 210-215) if they have different subscripts. *p < .001. phrenic patients, and 61 nondepressed medical The effects were striking. Depressed patients, and surgical patients. As shown in Table 8, whether endogenous, nonendogenous, or un- depressives explained bad events with more remitted, all differed from the control group internal, stable, and global causes. These re- in terms of explanations for bad events (all sults confirm, in a sample of unipolar de- ps < .0001). Table 9 presents these data. pressed patients, the association between ex- For these patients, sensitivity of the ASQ, planatory style for bad events and depressive denned as the probability of a correct diagnosis symptoms predicted by the helplessness re- of depression by the test given that the patient formulation. Such a style is not a general char- is depressed (i.e., the percentage of true pos- acteristic of psychopathology, because schizo- itives; Lusted, 1968, pp. 108-109), was 61% phrenic patients did not show it. for internality for bad events, 58% for stability Two other studies have also correlated the for bad events, and 77% for globality for bad ASQ with depressive symptoms in severe uni- events. Specificity of the ASQ, denned as the polar depressed patients. Eaves and Rush probability of a correct diagnosis of nonde- (1984) compared 31 depressed female psy- pressed by the test given that the patient is chiatric patients, diagnosed as unipolar, major nondepressed (i.e., the percentage of true neg- aifective disorder, with an age-, sex- and ed- atives; Lusted, 1968, pp. 108-109), was 94% ucation-matched control group of 17 people. for internality for bad events, 94% for stability Table 9 Attributional Style Questionnaire Scores and Depression: Patients Over Time

ED NE UD C Dimension (n= 11) (n = 13) (n = 17) Internality Time 1 5.40X 5.38X 5.23, 3.94y Time 2 5.35X 5.13X 5.41, 4.14y Stability Time 1 5.44X 4.81X 5.07, 3.92y Time 2 4.76* 4.28, 5.01, 3.49y Globality Time 1 5.44X 4.92, 5.11, 2.95y Time 2 4.54* 4.19, 5.27X 2.96y Composite Time 1 16.27X 15.10X 15.41, 10.80y Time 2 14.65* 13.61* 15.70X 10.59y Note. From Eaves and Rush (1984). ED = endogenous ; NE = nonendogenous; UD = unremitted; C = control; see text. Means in a row are different (p < .05) if they have different subscripts. * Time 1 different than Time 2 (p < .05). CAUSAL EXPLANATIONS AND DEPRESSION 359 for bad events, and 88% for globality for bad and/or global explanations for bad events and events. These compare favorably to the best obtained nonsignificant correlations (e.g., biological tests of depression (Eaves & Rush, Hammen & deMayo, 1982). Some have con- 1982). cluded that the helplessness reformulation is In addition, explanatory style for bad events therefore wrong (e.g., Coyne & Gotlib, 1983). significantly correlated with the total amount How do we reconcile this with the substantial of time the individual was depressed, the av- support found for a depressive explanatory erage length of episode, and the length of the style? current episode for nonendogenously de- Conclusions based on accepting the null hy- pressed individuals. This was particularly the pothesis are always problematic, because non- case for the stability measure, which correlated significant results may alternatively reflect an .79 with the total time of depression, .76 with insensitive or inappropriate method. This may the average length of episode, and ,71 with the be the case for studies that have failed to find length of the current episode (all ps < .001). a depressive explanatory style. Peterson and Theoretically, these data suggest that stability Raps (1983) observed that these studies tend for bad events reflects, and may predict, how not to distinguish explanations per se from long a depressive episode will last. explanatory style. As a result, they operation- Another study of unipolar depressed pa- alize style with single-item questionnaires, tients was conducted by Persons and Rao running the risk that the particular real event (1981). Forty-nine patients completed the about which respondents offer explanations is ASQ, as well as other measures. Upon ad- the overriding determinant of that explanation. mission, the internality score for bad events This technique minimizes the role of the per- correlated .60 with the BDI. The globality son's cognitive style. If respondents are asked score for bad events correlated .53, whereas to offer explanations about several (hypothet- the stability score for bad events did not cor- ical) bad events, then the chances are greater relate significantly with the BDI. Thirty-four that the average of these explanations will re- of these patients were followed through dis- flect a characteristic style. charge, and changes in explanatory style and This is really a logical argument, another BDI scores were measured. Explanatory style way of phrasing the statistical truism that re- change over therapy (largely chemical) ac- liability and thus validity are served by in- counted for 49% of total change in depression. creasing the number of observations above one This was the most powerful variable correlated (Epstein, 1980). To illustrate this empirically, with change in depression. Peterson and Raps (1983) classified a number Overall, then, these three studies show that of cross-sectional investigations of the expla- severely depressed patients have an insidious nation-depression link, noting the numbers explanatory style characterized by internal, of events about which explanations were made stable, and global explanations for bad events. in each study. As Table 10 shows, confirming This style is continuous with the findings in studies were more likely than were discon- mild depression and not a general character- firming studies to ask about several events. istic of psychopathology, inasmuch as schizo- Although the disconfirming studies may be phrenic patients did not show it. In addition, correct in concluding that contextual char- explanatory style is an index that is both sen- acteristics are importantly related to depres- sitive and specific. Finally, changing explan- sion, they are incorrect in further concluding atory style is a good correlate and may be a that explanatory style is unrelated, because good predictor of changing depression. This they usually did not assess style in a psycho- last finding is particularly suggestive for ther- metrically or theoretically reasonable way. apy, because techniques that encourage an individual to change his or her explanatory style Conclusions may relieve depression (Seligman, 1981). Severity of depressive symptoms is often correlated with the habitual use of internal, Disconfirming Studies stable, and global causes to explain bad events Several researchers have correlated depres- involving the self. This depressive explanatory sive symptoms with the use of internal, stable, style has been demonstrated with a variety of 360 CHRISTOPHER PETERSON AND MARTIN E. P. SELIGMAN

procedures in a variety of populations. Some swers that count toward the self-report of evidence suggests that it may be specific to depression are phrased in terms of internal, depression. Studies failing to find the depres- stable, and global causes, for example, "I will sive explanatory style tend to use inappropriate always fail at whatever I do." Perhaps a bad methods. So, there is a clear relationship be- explanatory style is just a special symptom of tween a particular explanatory style and depression. This is not the case. If we eliminate depression. But does this style put one at risk such items from the BDI and correlate ex- for later depression? Several other possibilities planatory style with the remaining items, the are compatible with this correlation. correlation does not go down (Peterson, First, there is the possibility of a tautological Schwartz, & Seligman, 1981). In addition, as relationship of depressive explanatory style and we report below, such a style precedes and depressive symptoms. For example, many an- predicts who, among people not depressed, will later become depressed. Second, there is the possibility that depres- Table 10 sion wholly causes the bad explanatory style. Cross-Sectional Investigations of the On this view, when one becomes depressed, Reformulation and the Number of Events for the depressive explanatory style is activated as Which Explanations Were Made a result. A study by Mukherji, Abramson, and Martin (1982) argues against this possibility. Studies No. of They induced a depressed mood in under- events graduates with the Velten (1968) procedure. Although mood induction was successful, ex- Disconfirming planatory style as measured by the ASQ was Danker-Brown and Baucom (1982) 1 unaffected. Feather and Davenport (1981) 1 Third, there is the possibility that some third Garber and Hollon (1980) 1 Gong-Guy and Hammen (1980) 5 variable, such as biochemical deficits or a Hammen and Cochran (1981) 5 preexisting style of turning anger inward, Hammen and deMayo (1982) 1 causes both depression and the depressive ex- Peterson and Conn (1982) 1 planatory style. The remaining lines of re- Average 2.14 search attempt to untangle these various possibilities from the process predicted by the re- Confirming formulation. Blaney, Behar, and Head (1980) - One 6 In the next section, we begin this untangling - Two 6 by describing evidence for the next most strin- Eaves and Rush (1984) 6 gent prediction of the reformulation: that de- Janoff-Bulman(1979) 4 pressive explanatory style precedes depression. Peterson, Bettes, and Seligman (1982) - One 2 - Two 6 Peterson, Schwartz, and Seligman (1981) 18 Longitudinal Investigations Raps, Peterson, Reinhard, Abramson, and 6 Seligman (1982) Seligman, Abramson, Semmel, and von Baeyer 6 According to the reformulation, explanatory (1979) style is a risk factor for depression. It must Seligman et al. (1984) 8 precede the occurrence of depressive symp- Sweeney, Shaeffer, and Golin (1982) 6 toms. Because cross-sectional studies are mo- Zemore and Johansen (1980) 1 mentary, they cannot address this prediction. Average 6.25 However, several longitudinal investigations have been conducted in which explanatory Note. From Peterson and Raps (1983). All studies cor- style was assessed at one time and conse- related depressive symptoms with the use of internal, stable, quences at a later time. and/or global explanations for bad events involving the self. Disconfirming studies were those that found no significant correlations (p > .05); confirming studies were Study 6: Children those that found significant correlations (p < .05). Not all studies investigated all three explanatory dimensions. Not included are studies that found mixed support for We have already described our research with the reformulation. grade-school children, who completed mea- CAUSAL EXPLANATIONS AND DEPRESSION 361 sures of explanatory style and depressive O'Hara et al. (1982) and also showed that ex- symptoms at two times, 6 months apart (Se- planatory style predicted speed of recovery ligman et al., 1984). The cross-sectional data from postpartum depression. from this research supported the reformulation. Do the longitudinal data? Conclusion The question of interest is whether explan- These studies mostly support the prediction atory style at Time 1 predicts depressive that depressive explanatory style precedes de- symptoms at Time 2, above and beyond the pressive symptoms. Explanatory style may be level of depression at Time 1. As predicted by a risk factor for later depression, allowing the reformulation, the depressive explanatory identification of individuals at risk by the way style for bad events was correlated with sub- they explain bad events. Primary prevention sequent depressive symptoms, even when ini- of depression should then be focused on such tial CDI scores were partialed out (partial r = individuals (but see Lewinsohn, Steinmetz, .26, p < .02). So, explanatory style for bad Larson, & Franklin, 1981, and Peterson et al., events may be a risk factor for depression 1981). among children. The general flaw of these studies as tests of Nolen-Hoeksema (1983) performed a sim- the reformulation is that they do not manip- ilar study using 168 children from Grades 3- ulate or assess bad life events. According to 5. Children who were both nondepressed in Abramson et al. (1978), a depressive explan- January and had a nondepressive explanatory atory style per se is not sufficient for depres- style remained nondepressed in March. In sion. It is only when bad events occur and are contrast, children who were nondepressed in interpreted in terms of internal, stable, and January but had a depressive explanatory style global causes that depressive symptoms are were likely to be depressed in March (p < .01). more likely to ensue. Because these studies did Several recently completed studies using not look at bad events, they did not test this adults address the same question. Golin, prediction. The studies in the next section test Sweeney, and Shaeifer (1981) and Firth and the next strongest proposition that a preex- Brewin (1982) administered measures of both isting bad explanatory style followed by bad explanatory style and depressive symptoms at events makes depression more likely. two times. Both studies provided support for the strong claim that explanations affect Experiments of Nature depression to a greater degree than depression affects explanations. In the Golin et al. (1981) The ideal way to test the helplessness re- study, causal priority of stability and globality formulation is to measure the explanatory style for bad events was demonstrated for 180 col- of individuals, and then to choose at random lege students over a 1-month period. In the half of these individuals to experience some Firth and Brewin (1982) study, causal priority important bad event. The most severe depres- for stability for bad events was demonstrated sion is predicted to ensue for subjects with the for 16 female patients over a 5-week period. preexisting depressive explanatory style who In another longitudinal study, O'Hara, also then experienced the bad event. The ob- Rehm, and Campbell (1982) administered the vious ethical dilemma can be partly solved by ASQ to 170 women in their second trimester a quasi-experimental method in which natu- of pregnancy, along with several other cog- rally occurring bad events are the manipula- nitive-behavioral assessments. The strongest tion. We report two studies conducted by our predictor of level of depression, 3 months research group. In the first, the bad event was postpartum, was explanatory style for bad an unsatisfactory grade on a midterm ex- events, which significantly correlated with BDI amination. In the second, the bad event was scores even when prepartum level of depres- imprisonment in a penitentiary. We also de- sion was held constant. scribe several other experiments of nature that In similar studies, Manly, McMahon, Brad- are pertinent to the reformulation. ley, and Davidson (1982) found no support Study 7: Midterm Experiment for the hypothesis that explanatory style predicts depression following childbirth, whereas This investigation was a prospective study Cutrona (1983) replicated the findings of of college students and their reactions to a low 362 CHRISTOPHER PETERSON AND MARTIN E. P. SELIGMAN grade on a midterm examination (Metalsky pressive explanatory style is a constitutional et al., 1982). According to the helplessness re- weakness (diathesis) and the low midterm formulation, students who habitually explain grade is the environmental stressor (stress), bad events in terms of internal, stable, and which interact to predispose depressed mood global factors are more likely to feel depressed (Metalsky et al., 1982). The practical signifi- on learning that they received a low grade than cance lies in the possibility of identifying, in are students who tend to explain bad events advance, individuals at risk for depressive re- in terms of external, unstable, and specific fac- action to bad life events from explanatory style. tors. The participants were undergraduates in an Study 8: Prison Experiment introductory psychology course at the State Does explanatory style predict depression University of New York at Stony Brook, who following imprisonment (Bukstel & Kilmann, completed questionnaires for this research 1980)? For most individuals, imprisonment is during classtime. At Time 1, they took the inarguably a bad event. Like any total insti- ASQ and a questionnaire about their aspira- tution, prisons deny individuals control over tions for the class midterm. Students indicated even the most mundane aspects of their lives the grades with which they would be happy (Goffman, 1961; Taylor, 1979). We would ex- and unhappy. At Time 2, 11 days later, just pect, then, that a common reaction to im- prior to the midterm examination, students' prisonment is depression (Seligman, 1975). level of depressed mood was assessed with Furthermore, we predicted that prisoners with the Multiple Affect Adjective Check List the depressive explanatory style are the most (MAACL; Zuckerman & Lubin, 1965), a likely to become depressed after internment measure of transient depressed mood. At Time (Abramson et al., 1978). We report a prelim- 3, another 5 days later, immediately following inary study of this (Peterson, Nutter, & Selig- receipt of the midterm grade, students again man, 1982). Upon imprisonment, individuals completed the MAACL. completed the ASQ, and shortly before release, Subjects were considered to have received these same individuals completed the BDI. a low grade if their midterm grade was less Within 1 week following imprisonment at than or equal to the grade with which they one of four maximum security prisons in New said they would be unhappy (« = 53). Students York, Ohio, and Pennsylvania, 245 adult males were considered to have received a high grade (ranging in age from 17 to 64 years; average if their midterm grade was equal to or greater age = 27 years) completed the ASQ. Within than the grade with which they said they would 1 week before their release, which varied from be happy (n = 28). To test our predictions, 1 month to 1 year, they completed the BDI. explanatory style scores for bad events were For 28 prisoners, BDI scores from the time of correlated with standardized residual gain initial imprisonment were available. These av- scores on the MAACL depression scale from eraged only 1.68, suggesting that subjects were Time 2 to Time 3, separately for the two not at all depressed at the time of imprison- groups. Explanatory style for bad events pre- ment. dicted increases in depressed mood for stu- At the end of their imprisonment, prisoners dents who received low grades—internality scored an average of 17.7 on the BDI, placing (r=.34,p< .02), stability (r = .04, ns), and them in the moderately to severely depressed globality (r = .32, p < .02)—but not for stu- range (Beck, 1967). Furthermore, their de- dents who received high grades—internality pressive symptoms at the end of imprisonment (r = .12, ns), stability (r = .36, p < .10), and were strongly associated with their explanatory globality (r = .22, ns). Additional correlational style at the beginning of imprisonment. As analyses revealed that explanatory style was expected, explanations for bad events were independent of the actual grade received and positively correlated with depressive symp- of grade aspirations. Thus, several possible toms: internality (r = .34, p < .001), stability confounds are ruled out. (r = .36, p < .001), globality (r = .35, p < The theoretical significance of these findings .001), and composite (r = .35, p < .001). Sur- is that they support the reformulation's claim prisingly, explanations for good events were to be a diathesis-stress model, where the de- similarly correlated with depression: inter- CAUSAL EXPLANATIONS AND DEPRESSION 363 nality (r = .39, p < .001), stability (r = .38, cifically, women who explained their cancer p < .001), globality (r = .37, p < .001), and in terms of behavior (i.e., internal, unstable, composite (r = .39, p < .001). Unlike results and specific causes) believed that they were in our other research, explanatory styles for cancer free; they were likely to be nonde- bad versus good events did not have opposite pressed. In contrast, women who explained effects on depression. Furthermore, unlike re- their cancer in terms of personality (i.e., in- sults in our other research, explanatory styles ternal, stable, and global causes) believed that for bad versus good events were not indepen- they were not cancer free; they were likely to dent. Instead, they were substantially inter- be depressed. correlated (all rs > .60; all ps < .001). Not all studies that have looked at the link between causal explanations and reactions to Other Experiments of Nature serious disease have supported the reformulation (e.g., Bard & Dyk, 1956). Factors such Several other experiments of nature deserve as the seriousness and chronicity of the disease mention. Although not designed to test the may affect the relationship between explana- helplessness reformulation, they are intriguing tion and reaction, perhaps by creating expec- and illustrate how investigations of reactions tations about the future course of the disease to naturally occurring bad events might be incongruent with its causal explanation. used to test the reformulation. Tennen, Affleck, Allen, McGrade, and Ratzan (1983) studied Life Events and Depression 32 children who had diabetes. Children who explained the onset of their disease in terms The large research literature on life events of their behavior (i.e., internally, unstably, and and psychological distress is also relevant to specifically; see Peterson, Schwartz, & Selig- the helplessness reformulation, although few man, 1981) were rated by their physicians as theorists have drawn explicit parallels. How- coping better with diabetes than were children ever, in a recent review of this literature, Thoits who explained the onset in terms of genetic (1983) proposed that many of its consistent inheritance (i.e., internally, stably, and glob- findings are compatible with a helplessness in- ally). Although a third variable such as parental terpretation. The modest relationship typically attitude may underlie these data, they are con- observed between the experience of life events sistent with the reformulation. and subsequent distress is increased when (a) Similarly, Affleck, Allen, McGrade, and the psychological consequence is depression McQueeney (1982) interviewed the mothers as opposed to some other psychopathology and of 43 infants with severe perinatal complica- (b) the events themselves are relatively un- tions. Mothers who explained the infant's con- desirable, uncontrollable, unexpected, impor- dition in terms of some behavior during preg- tant, and clustered together in time. nancy (i.e., unstably and specifically) reported Although life events research has typically less mood disturbance and anticipated fewer not assessed causal explanations, the types of caretaking problems than did mothers who events most frequently linked to psychological explained the complications in terms of other distress lend themselves to internal, stable, and people (i.e., stably and globally). global explanations. This possibility is consistent with the helplessness reformulation. Fur- Timko and Janoff-Bulman (1983) interviewed 42 women who had undergone a mas- thermore, that life events are more apt to result in depression than in other psychological dis- tectomy following a diagnosis of breast cancer, asking them to offer a causal explanation for orders adds to the evidence for the specificity their cancer and to complete the BDI. Also of the helplessness model as an account of assessed were other beliefs about the success depression (cf. Raps, Peterson, Reinhard, of the mastectomy and the avoidability of fu- Abramson, & Seligman, 1982). ture cancer. The helplessness reformulation Conclusions was strongly supported. Causal explanations did not directly affect depression, but they did These experiments of nature suggest that predict expectations about future cancer, the helplessness model may be useful in ex- which in turn predicted BDI scores. More spe- plaining reactions to major life events. In par- 364 CHRISTOPHER PETERSON AND MARTIN E. P. SELIGMAN ticular, the studies by our research group sug- Table 11 gest that the depressive explanatory style tends Explanations and Laboratory Helplessness: to produce depression when bad events are Mean Latencies (in s) at Test Task encountered. Other studies show that particular explanations of bad events are associated Style with depression and poor coping. Each of these Test task Global Specific studies is open to criticism, because experimental control over the bad events was not Similar possible, and possible confounds might un- Escapable noise 2.26, 1.98. derlie the obtained correlations. However, be- Inescapable noise 3.43b 3.43b cause the results of these studies converge, our No noise control 2.54a 2.58b Dissimilar confidence increases that the conjunction of Escapable noise 16.4, 17.4, depressive explanatory style and bad events Inescapable noise 41.5b 20.0. predisposes subsequent depression. No noise control 20.7. 19.0. Experimental studies, unlike experiments of nature, allow the bad event to be presented Note. From Alloy, Peterson, Abramson, & Seligman (1984). to randomly selected subjects. We now turn For similar test task, maximum latency = 5.00 s. For each to a set of laboratory experiments that suggest condition, n = 18. For dissimilar test task, maximum latency = 100.0 s. For each condition, n = 10. Means in a that helplessness symptoms occur as predicted row or in a column of three are different (p < .05) by when uncontrollable bad events are imposed NeWman-Keuls procedure if they have different subscripts. on people who differ in explanations and explanatory style.

Laboratory Experiments According to the reformulation, all subjects should show deficits following inescapable Does explanatory style affect laboratory-in- noise when tested on the similar noise task. duced learned helplessness in the same way However, only subjects with a global explan- that it affects naturally occurring depressive atory style for bad events should show deficits symptoms? Laboratory studies give us what following inescapable noise when tested on the experiments of nature lack. We can impose a dissimilar cognitive task. Table 11 shows these bad event on randomly selected subjects and predictions to be supported. Furthermore, the look at the effects of preexisting or induced differences remain even when internality and causal explanations. Several studies of this sort stability scores are covariates. have been done. This study experimentally manipulated bad events and also looked at a specific role as- Study 9: Global Versus signed to one of the explanatory dimensions. Specific Explanations In a related investigation (Peterson & Selig- man, 1981), we obtained some preliminary Alloy et al. (1984) gave subjects one of three evidence that stability for bad events is related pretreatments as described by Hiroto and Se- to the time course of laboratory-induced help- ligman (1975): a condition in which button- lessness as predicted by the reformulation. pushing terminated an aversive noise delivered through headphones (escapable), a condition Other Laboratory Experiments in which the noise could not be terminated (inescapable), and a condition in which no Several other laboratory studies concern noise—escapable or inescapable—was expe- both the manipulation of causal explanations rienced (no pretreatment). Then, subjects were and the manipulation of bad events. Pasahow tested at one of two tasks: a task similar to (1980) manipulated the global-specific dimen- pretreatment—a solvable hand shuttle-box in sion and imposed bad events on subjects in which success moving the hand stopped an the learned helplessness triadic design. Subjects aversive noise, and a task dissimilar to pre- instilled with a global explanation were given treatment—a series of solvable anagrams ad- written instructions that told them that the ministered in a different room by a different concept-identification task they were about to experimenter. take correlated very highly with how people CAUSAL EXPLANATIONS AND DEPRESSION 365

performed on all psychology experiment tasks. helpless subjects reported a less depressed Specific subjects were given written instruc- mood when induced to explain their success tions that described the task as a concept-for- internally. Similarly, both depressed and help- mation task that had little or no correlation less subjects performed better at a subsequent with other tasks used in psychology experi- anagram task when induced to explain their ments. success globally. All of the subjects were then given unsolv- Finally, Anderson (1983) manipulated able concept-identification problems and then causal explanations following failure for college tested on anagrams. Subjects induced to make students preselected for explanatory style. Us- global explanations for their failure performed ing his Attributional Style Assessment Test worse on the anagrams than did subjects en- (ASAT), Anderson (1983) obtained subjects couraged to give specific explanations. This who explained failure characterologically (i.e., suggests that the manipulation of explanations internally, stably, and globally) and subjects along the global-specific dimension for bad who explained failure behaviorally (i.e., in- events appropriately predicts deficits when bad ternally, unstably, and specifically). These sub- events are then imposed on subjects. jects then attempted to persuade other students Thus, we are given converging evidence to donate blood; explanations for failure were about the globality dimension for bad events, experimentally varied and included an ability/ In the Alloy et al. (1984) study, preexisting trait condition and a strategy/effort condition. explanatory style was merely measured and Consistent with the reformulation, subjects the bad event manipulated. Preexisting global who offered behavioral explanations, whether style for bad events predicted failure at new by preselection or manipulation, reported problems once the bad event occurred, greater expectations of success, displayed whereas specific style did not. In parallel, if higher motivation, and performed more suc- causal explanations are themselves manipu- cessfully following failure than did subjects lated along with the bad event, Pasahow (1980) who offered characterological explanations (see showed that the same results hold. also Anderson & Jennings, 1980). McFarland and Ross (1982) investigated the effect of explanations on self-esteem following failure. They assigned female college students Conclusions to success or failure feedback conditions fol- Laboratory investigations of learned help- lowing a social accuracy test. They also ma- lessness in humans have been controversial nipulated the causal explanation for perfor- since their inception (Wortman & Brehm, mance at this test: internal or external. 1975). In a recent review, Silver, Wortman, Consistent with the reformulation, subsequent and Klos (1982) criticized these studies on self-esteem was lowest for subjects who ex- three grounds. First, the relationship between perienced failure and were induced to explain experience with uncontrollable events and it internally. So, a manipulation designed to subsequent deficits may be mediated by a va- produce internal causal explanations lowered riety of factors, not just the expectation of self-esteem following a bad event, whereas a response-outcome independence (e.g., Frankel manipulation designed to produce external & Snyder, 1978; Peterson, 1978; Tennen, causal explanations did not. 1982). Second, laboratory studies of human In a laboratory experiment with depressed helplessness may be subject to demand char- inpatients, Miller and Norman (1981) told acteristics that trivialize obtained results subjects who were either acutely depressed or (Orne, 1962). Third, the laboratory setting is recently improved and made helpless by in- artificial, and generalization of results to hu- escapable noise, that they had done well at a man helplessness and depression outside of task measuring social intelligence. Subjects the laboratory may not be justified (e.g., Bui- were induced to explain their success in one man & Wortman, 1977; Coyne, Aldwin, & of four ways: with an internal-global cause, Lazarus, 1981). an internal-specific cause, an external-global These are reasonable criticisms, and the re- cause, or an external-specific cause. Consistent search program described here attempts to with the reformulation, both depressed and answer them. First, as we have already stated, 366 CHRISTOPHER PETERSON AND MARTIN E. P. SELIGMAN the expectation of response-outcome inde- whom measures of current psychological pendence is regarded as a sufficient condition characteristics are available. for depressive deficits, not a necessary one. That other factors may also determine deficits Study 10: Causal Explanations following uncontrollability is undoubtedly true and Psychotherapy and not contrary to this theory. However, converging research shows the importance of ex- In the first study (Peterson & Seligman, pectations in the development of such deficits. 1981), each causal explanation for a bad event Second, laboratory experiments may indeed was extracted from 300-word excerpts from be confounded by demand characteristics, but verbatim transcripts of individual psycho- other research strategies also have inherent therapy sessions. Transcripts were available flaws. Thus, we favor a multimethod research from the beginning, middle, and ending of strategy and a search for convergence among successful psychotherapy with four different results (Campbell & Fiske, 1959). Finally, the patients suffering depression following a loss. only way to ascertain whether laboratory find- We blindly rated these explanations for inler- ings are artificial is to attempt studies outside nality, stability, and globality, and averaged the laboratory. The out-of-the-laboratory re- these scores across explanations offered within sults described here are mostly consistent with the same excerpt. For each patient, explanatory the laboratory data. style perfectly ordered the three sessions; the These laboratory data indicate that the most internal, stable, and global explanations measurement and manipulation of explana- were offered in the beginning session, whereas tions and explanatory style, when accompa- the least internal, stable, and global explana- nied by the manipulation of bad events, yield tions were offered in the ending session. These what the theory predicts. The global-specific data supplement the Persons and Rao (1981) dimension, when manipulated and when study already described in suggesting that measured, governs the breadth of helplessness changes in explanatory style index changes in deficits. The internal-external dimension, depression during psychotherapy. when manipulated, appears to govern self-esteem deficits. Finally, tentative evidence sug- Study 11: Causal Explanations gests that the stable-unstable dimension gov- and Mood Swings erns the duration of helplessness deficits. In the second study (Peterson et al., 1983), Case Studies: Investigation of Real Lives we used the symptom-context method, a content-analytic approach developed by Luborsky The final line of evidence we wish to present (1964, 1970), to investigate the explanatory comes from the study of individual lives. It antecedents of mood swings by a single patient converges on the proposition that explanatory during psychotherapy sessions. With the style is a risk factor for depressive symptoms. symptom-context method, context is assessed One major barrier in trying to study the con- immediately before and after the onset of the sequences of explanatory style in the labora- symptoms of interest. In this case, the context tory or in natural settings is the obtrusive and was causal explanations for bad events, and reactive nature of questionnaires. The devel- the symptoms were mood swings—in and out opment of a valid method for assessing ex- of depression during therapy sessions. planatory style through the content analysis Our subject was a patient, Mr. Q, who dem- of written transcripts allows us to study, in a onstrated precipitous shifts in mood during wholly unobtrusive and nonreactive way, the psychotherapy sessions (conducted over 4 natural history of explanatory style and years). More than 200 sessions with Mr. Q depression. We now present several studies that were tape-recorded, and a sample was tran- use this technique. The first two studies analyze scribed, allowing a thorough content analysis verbatim transcripts of actual psychotherapy of his explanations before and after mood sessions with depressed patients. The third swings. Three types of sessions were analyzed: study analyzes verbatim interviews originally those in which Mr. Q became more depressed, conducted decades ago with individuals for those in which he became less depressed, and CAUSAL EXPLANATIONS AND DEPRESSION 367 those in which no shift at all occurred. Ex- planations for bad events were extracted from these sessions before and after the mood .4 I Depression swings, and were rated for internality, stability, (n=4) 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 mood just went down"), (b) the agreement of two independent judges reading the transcript of I No change the therapy session that a mood swing had just 6] (n=3) occurred, and (c) the fact that the swing did •i 4 not occur in close proximity to another swing (i.e., within 15 min of the same session). Ap- plying these three criteria resulted in a final set of four swings in which the patient became Depression more depressed and five swings in which he (n=5) became less depressed. These mood swings were validated by both the agreement of two independent judges listening to the taped therapy sessions that Mr. Q's voice showed a I change in affective quality and the scoring by Before shift After shift an independent judge using the depression Figure 2. Means of internal + stable + global ratings of component of the Gottschalk and Gleser explanations before and after mood shifts for the different (1969) hostility-inward scale. types of sessions. (Numbers of sessions on which means are based are in parentheses.) Explanations for bad events were extracted from the 400 words (spoken by Mr. Q) before the swing and from the 400 words following swings. Various content characteristics of these the swing. For comparison purposes, causal same transcripts, including Mr. Q's statements explanations were also extracted from ran- about anxiety, hopelessness, guilt, hostility to domly chosen 800-word segments of three ses- self, loss of self-esteem, and Oedipal conflict, sions in which no swing occurred. In all, 63 were examined. These were not redundant bad events linked to causal explanations were with the causal explanation scores, rising and rated for internality, stability, and globality by falling with level of mood in a given session. four independent judges blind to their source. In contrast, explanations were different before Ratings by the judges were combined, and swings. composites were formed. Means for the dif- This study shows that the reformulation may ferent types of sessions are shown in be applied powerfully to specific individuals Figure 2. with depressive symptoms. Explanation ratings The differences predicted by the reformu- for sessions in which swings to increased lation were present before swings in mood. depression occurred did not at all overlap with Highly internal, stable, and global causal ex- explanation ratings for sessions in which planations preceded increased depression, swings to decreased depression occurred, al- whereas much more external, unstable, and lowing perfect prediction. specific statements preceded decreased depression. There was no overlap between the Study 12: Individuals of the Berkeley- ratings of causal explanations before swings Oakland Study to more versus less depression. This is a fine-grained case study, but because We now describe a preliminary study using it is correlational, it is necessary to rule out what we have dubbed the time machine some plausible third-variable arguments to method. While conducting content analysis of conclude that explanations, and not some fac- causal explanations provided in essays and tor correlated with them, produced mood therapy transcripts, we realized that there was 368 CHRISTOPHER PETERSON AND MARTIN E. P. SELIGMAN no reason to limit ourselves to contemporary as if middle-class individuals were not so prone data. Content analysis can be used with any to react to economic difficulties with hope- written material, including newspaper stories, lessness, perhaps because they had available autobiographies, transcribed interviews, let- to them means for coping with difficulties. The ters, diaries, and so on (Simonton, 1981). So, helplessness model proposes that early mastery we decided to use content analysis to extract experience immunizes one against the bad ef- causal explanations from open-ended material fects of uncontrollability (Maier & Seligman, residing in already-existing data archives. The 1976); these data from the Berkeley-Oakland explanatory style of individuals years ago can Study may illustrate immunization. be assessed through content analysis, and the In an even more preliminary investigation relationship between causal explanations and using the time-machine methodology, we an- later characteristics of these individuals can alyzed interviews from the Grant Study, a lon- be ascertained. Prospective longitudinal re- gitudinal investigation of the Harvard classes search spanning decades is possible retro- of 1939-1942. Over the 40 years of this respectively. search, the material success, social success, Our first attempt to study individual lives mental health, and physical health of these over decades analyzed material in the Berke- men have been recorded, with special interest ley-Oakland Growth Study (Elder, Bettes, & in whether such psychological characteristics Seligman, 1982), an ongoing investigation of as defense mechanisms predict future life the effects of the Great Depression on family course (Vaillant, 1977). A great amount of and life patterns (Elder, 1974). Although eco- open-ended interview material is available for nomic hardships of the Great Depression hit each individual studied. We extracted causal both the lower class and the middle class, the explanations for bad events from a 1946 in- lower class was hit harder. Indeed, middle-class terview about war experiences conducted with individuals who suffered financial hardships 18 men. Explanations were rated for inter- during the Great Depression years later were nality, stability, and globality, as in our other happier, healthier, and more efficacious than studies. middle-class individuals who had not been These ratings were correlated with a global economically deprived (Elder, 1974). Does ex- rating of physical health in 1980: 1 = healthy; planatory style mediate this interaction be- 2 = minor health problems; 3 = chronic ill- tween social class and economic hardship? ness; 4 = disabled; and 5 = dead. Because of Causal explanations were extracted from the the small sample size, we report only the cor- open-ended responses of 28 women who de- relation with the composite explanatory style. scribed the worst events during the preceding It was positive (r = .40, p < .10). Depressive year in both 1943 and 1970. The focus of this explanatory style in 1946 predicted poor study was on hopelessness—the use of stable physical health 34 years later. The tentative and global causes to explain bad events—so nature of this study must be emphasized. these explanations were rated for stability and Sample size is small, and the obtained cor- globality, but not internality. A hopeless out- relation is of borderline statistical significance. look in 1943 was significantly correlated with The dependent measure assessed overall phys- a hopeless outlook in 1970 (r = .44, p < .05), ical health, not level of depression, and we do through which it affected overall psychological not know what the mechanism linking ex- health (assessed by a Q-sort procedure) in planatory style and health might be. 1970. Both of these investigations are preliminary, More detailed analyses suggested that hard- and we are currently analyzing the causal ex- ships in 1929 predicted hopelessness in 1943. planations in a greater number of interviews. Hardships per se had no direct effect on psy- Nevertheless, we have described these still-in- chological health in 1970. Only when hard- progress studies because they suggest that the ships gave rise to a hopeless outlook was psy- helplessness reformulation can be applied to chological health affected for the worse decades real people over the life span. later. Furthermore, these patterns seem more pronounced among the lower class than among General Conclusions the middle class. The small sample size renders We have described five interlocking types this conclusion highly tentative, but it appears of research investigating the attributional re- CAUSAL EXPLANATIONS AND DEPRESSION 369 formulation of the helplessness model of (1955) suggested that the function of thought depression. The cross-sectional studies showed is to allay doubt (see also Wong & Weiner, that a characteristic way of explaining bad 1981). It seems likely that we ask why more events with internal, stable, and global causes often when we are in doubt—when our path co-occurs with depressive symptoms. The lon- is suddenly obstructed—than when things are gitudinal studies showed that this explanatory going smoothly. Indeed, consciousness itself style preceded the development of depressive may emerge only when our routine actions fail symptoms. The experiments of nature indi- (cf. Langer, 1978). This implies that the ex- cated that this style resulted in depression once planatory analysis of failure among adults bad events were encountered. The laboratory would be much more articulate than would experiments showed that imposing uncon- the analysis of success. Children, on the other trollable bad events on individuals making hand, newer at the game, may analyze the particular explanations had the predicted ef- causes of both failure and success closely. Such fects on helplessness deficits. Finally, the case a developmental process would render chil- studies illustrated that the reformulation ap- dren's explanatory style for good events more plies predictively to the depressive symptoms useful than that of adults in predicting depres- of specific individuals. sion. Taken together, these five lines of research 2. The origins of explanatory style. The support the predictions of the attributional helplessness reformulation is silent about an- reformulation of the learned helplessness tecedents. We are turning our research atten- model of depression. Each type of research is tion to this issue, and it promises to be exciting. open to criticism on several grounds, but the Several preliminary findings are worth men- convergence of results across different strat- tioning. First, in our study of explanatory style egies of investigation, different operational- and depressive symptoms among children (Se- izations, and different populations argues ligman et al, 1984), we obtained correspond- strongly for the validity of the reformulation ing scores from the mothers of 47 children in (Campbell & Fiske, 1959). our sample and from the fathers of 36. We There are aspects of the reformulation still found that explanatory style for bad events in need of theoretical and empirical scrutiny, and depressive symptoms were correlated for and we briefly discuss these now. mothers and their children: explanatory style 1. The role in depression of causal expla- (r = .39, p < .01) and depression (r = .37, p < nations of good events. The helplessness re- .01). Mothers' explanatory style also correlated formulation does not explicitly address such with children's depression (r = .42, p < .005), causal beliefs, but they seem to have at least whereas children's explanatory style weakly a weak empirical relationship with depressive correlated with mothers' depression (r = .27, symptoms. We have assessed such beliefs and p < .10). Scores for fathers were unrelated to generally found them to have effects on those of their children and to those of their depression opposite to those of causal expla- mates. In our content analysis of the Berkeley- nations about bad events, but at less robust Oakland Study, explanatory style of mothers levels. However, some of the findings do not and sons (but not of mothers and daughters) fit these generalizations. In our longitudinal also converged. investigation of children, explanatory style for There are two other important hints about bad events predicted subsequent depression, the origins of explanatory style. In addition whereas style for bad events did not. In our to learning it from one's mother, one may learn study of prisoners, the explanation of good from the type of criticisms leveled by teachers. events with internal, stable, and global causes Dweck and Licht (1980) looked at the expla- was positively, not negatively, correlated with nations that teachers gave in the third grade later depression. A sound theoretical treatment when boys and girls failed in the classroom. is awaited. They found that teachers criticized girls with We speculate that our pattern of findings internal, stable, and global statements. Boys, with children may be related to the sorts of on the other hand, were criticized with more events that adults take for granted that children unstable and specific explanations, such as do not. When do we engage in explanatory "You're being rowdy", "You're not concen- analysis, and when do we forego it? Peirce trating." 370 CHRISTOPHER PETERSON AND MARTIN E. P. SELIGMAN

In other research, Dweck and Licht (1980) eralism in politics or vanity in interpersonal gave fourth grade boys and girls unsolvable relations. We believe this for three reasons. problems and showed that boys were less sus- First, the ASQ is designed to look at expla- ceptible to helplessness than were girls. Boys nations given across situations: half bad, half also tended to give less internal, less stable, good. An explanatory style is derived from and less global explanations for their failures cross-situational consistency, and this is part on these tasks than did girls. Boys said: "I of what is meant by trait. Second, some people wasn't trying hard", "I wasn't paying atten- have more or less consistency across situations. tion", and "I don't care about your test." Girls For example, some individuals average 5.0 on attributed their failure to incompetence and stability for bad events because they answer 5 stupidity. Dweck and Licht (1980) argued that to each of the six bad situations for stability, the negative explanatory style of the girls rel- whereas others average 5.0 because they chose ative to the boys may originate from learning three 7s and three 3s. These latter individuals their teachers' criticisms. That women show can hardly be said to have a trait with respect more depressive symptoms as adults than do to stability. When we calculate the standard men (Radloff, 1975) may indicate that they deviation of ratings for each dimension and have learned an explanatory style as girls that discard those individuals who have large stan- makes them vulnerable to depression as dard deviations, we find that correlations with women. depression and other outcomes go up. Third, Finally, the reality of one's first trauma may explanatory style shows fairly high stability play an important role in setting explanatory across time. style. Brown and Harris (1978) interviewed Although we believe explanatory style is lower- and middle-class women in South Lon- traitlike, we do not believe it is invariant— don and found that almost 20% of the lower- nor, incidentally, are traits like liberalism and class women showed severe symptoms of vanity. We have two reasons for taking the depression. An unusually high percentage of plasticity of explanatory style seriously. First, these depressed women had lost their mothers although we have argued that explanatory style by death before they were 11 years old. The affects depression, we also suspect that depres- death of mother for a young girl, as opposed sion affects explanatory style. Persons and Rao to the death of mother for a teenager, has more (1981) found that explanatory style changed stable consequences. In addition, for a young for the better among patients as their depres- girl, as opposed to a teenager, a large part of sion lifted. So also did Castellon et al. (1982) her repertoire (global) is wiped out when her and Hamilton and Abramson (1983). Therapy mother dies. It may be the case, then, that and other good or bad life events may change because a mother dying represents a stable explanatory style (Seligman, 1981). We do after and global loss for a young girl, and because all acquire it at some time in our lives. Second, her personal helplessness is thereby under- some individuals may display consistent ex- scored, that as an adult she imposes internal, planations of a given type at one time and stable, and global explanations on other bad consistent explanations of another type at an- events, rendering her vulnerable to depression. other time. Mr. Q's explanations, for example, In summary, then, there are at least three predicted his depressive mood shifts, but they means by which explanatory style may be ac- did not cohere into a style. On some days, he quired. First, one may learn it by imitating explained bad events internally, stably, and parents, particularly the primary care giver, in globally, whereas on other days, he explained our society the mother. Second, the criticisms them externally, unstably, and specifically. that are launched at one by teachers following Future investigations should adopt a more failure may teach an explanatory style. Finally, sophisticated view of explanatory style. It the reality of one's first traumatic loss—the should be treated as a dependent variable that extent to which it is actually internally, stably, can be modified by life events, as well as an and globally caused—may set one's explana- independent variable that modifies future tory style for life. events. Also, the likelihood that there is a bi- 3. Explanatory style as a trait. We view directional influence between depression and explanatory style as a trait, analogous to lib- explanatory style should be explored. And fi- CAUSAL EXPLANATIONS AND DEPRESSION 371 nally, the term style should be reserved for and of coping. Our speculations are testable, individuals whose causal explanations show and if true, they would be consistent with Hari low variability across time and situations. Seldon's program. 4. Seldonics. There is one implication of To conclude, we have described the empir- our research program that we believe to have ical support so far obtained for the helplessness uncommon promise. We call it Seldonics, after reformulation. Results to date support the Hari Seldon (11,988-12,069 G.E.), the fic- model. We have briefly sketched some areas tional psychohistorian who developed a science in need of further theoretical and empirical "that deals with the overall reactions of large scrutiny. The attributional reformulation groups of human beings to given stimuli under promises to be a useful account of depression. given conditions" (Asimov, 1982, p. 12). By Explanatory style, in conjunction with actual assuming a large enough group of people (and bad events, precede the development of de- a set of behavioral laws and mathematical re- pressive symptoms. Attention to them may be lations left to the reader's imagination), Seldon a practical means of predicting who is at risk founded a powerful psychohistory that pre- for depression. Assaying them may help di- dicts, for human conglomerates, crises and agnose depression with specificity and sensi- their resolutions far into the future. No such tivity. And interventions directed at changing claims are made here. them may be an effective means of combating However, the technique of verbatim analysis depression. has intriguing possibilities. First, it can be re- liably done. Second, it converges with a ques- References tionnaire that seems to predict which human Abramson, L. Y., & Sackeim, H. A. (1977). A paradox conglomerates will react to failure with pas- in depression: Uncontrollability and self-blame. Psy- chological Bulletin, 84, 835-851. sivity and depression and which will react with Abramson, L. Y., Seligman, M. E. P., & Teasdale, J. D. active coping attempts and good cheer. Third, (1978). Learned helplessness in humans: Critique and it may predict when a group reacts to failure reformulation. Journal of Abnormal Psychology, 87, 49- with helplessness (Berkeley-Oakland Study), 74. and it did predict with accuracy when one Affleck, G., Allen, D., McGrade, B. J., & McQueeney, M. (1982). Maternal causal attributions at hospital discharge individual (Mr. Q) became depressed and when of high risk infants. American Journal of Mental De- he became nondepressed. Finally, it is wholly ficiency, 86, 575-580. unobtrusive and nonreactive, able to be used Alloy, L. B., Peterson, C., Abramson, L. Y., & Seligman, with individuals quick or dead or otherwise M. E. P. (1984). Attributional style and the generality of learned helplessness. Journal of Personality and Social impractical to test with questionnaires. Psychology, 46, 681-687. Let us imagine that one wants to predict Anderson, C. A. (1983). Motivational and performance (or postdict) whether a terrorist will react to deficits in interpersonal settings: The effect of attribu- the bombing of his headquarters with redou- tional style. Journal of Personality and Social Psychology, bled terroristic acts or with passivity. Similarly, 45, 1136-1141. Anderson, C. A., & Jennings, D. L. (1980). When expe- imagine that one wants to predict (or postdict) riences of failure promote expectations of success: The how a human conglomerate like a given sales impact of attributing failure to ineffective strategies. force will react to a bad event like an unsuc- Journal of Personality, 48, 393-407. cessful product line. Will this sales force sell Asimov, I. (1982). Foundation's edge. Garden City, NY: Doubleday. more versus less of the established product Bard, M., & Dyk, R. B. (1956). The psychodynamic sig- lines than will a second sales force? nificance of beliefs regarding the cause of serious illness. AH that is needed to attempt these predic- Psychoanalytic Review, 43, 146-162. tions is (a) prior verbatim material (if the ASQ Beck, A. T. (1967). Depression: Clinical, experimental, and theoretical aspects. New York: Hoeber. has not been administered) such as diaries, Beck, A. T., & Beck, R. W. (1972). Screening depressed letters, high school compositions, therapy patients in family practice: A rapid technic. Postgraduate transcripts, and quotes from the newspaper, Medicine, 52, 81-85. in the case of individuals, or a composite of Blaney, P. H., Behar, V., & Head, R. (1980). Two measures such material, in the case of conglomerates; of depressive cognition: Their association with depression and with each other. Journal of Abnormal Psychology, (b) coding and analysis of the explanatory pro- 89, 678-682. file for failure contained in this material; and Brown, G. H., & Harris, T. (1978). Social origins of (c) reasonable operationalizations of failure depression. New York: Free Press. 372 CHRISTOPHER PETERSON AND MARTIN E. P. SELIGMAN

Bukstel, L. H., & Kilmann, P. R. (1980). Psychological chological works ofSigmund Freud (Vol. 14, pp. 243- effects of imprisonment on confined individuals. Psy- 258). London: Hogarth. (Originally published, 1917.) chological Bulletin, 88, 469-493. Garber, J., & Hollon, S. D. (1980). Universal versus personal Bulman, R. J., & Wortman, C. B. (1977). Attributions of helplessness in depression: Belief in uncontrollability blame and coping in the "real world": Severe accident or incompetence? Journal of Abnormal Psychology, 89, victims react to their lot. Journal of Personality and 56-66. Social Psychology, 35, 351-363. Garber, J., & Seligman, M. E. P. (Eds.). (1980). Human Bumberry, W., Oliver, J. M., & McClure, J. N. (1978). helplessness. New York: Academic Press. Validation of the Beck Depression Inventory in a uni- Goffman, E. (1961). Asylums. Garden City, NY: Anchor. versity population using psychiatric estimate as the cri- Golin, S., Sweeney, P. D., & Shaeffer, D. E. (1981). The terion. Journal of Consulting and Clinical Psychology, causality of causal attributions in depression: A cross- 46, 150-155. lagged panel correlational analysis. Journal of Abnormal Campbell, D. X, & Fiske, D. W. (1959). Convergent and Psychology, 90, 14-22. discriminant validation by the multitrait-multimethod Gong-Guy, E., & Hammen, C. (1980). Causal perceptions matrix. Psychological Bulletin, 56, 81-105. of stressful events in depressed and nondepressed out- Castellon, C., Ollove, M., & Seligman, M. E. P. (1982). patients. Journal of Abnormal Psychology, 89, 662-669. Unpublished data, University of Pennsylvania. Gottschalk, L. A., &Gleser, G. C. (1969). The measurement Cole, C. S., & Coyne, J. C. (1977). Situational specificity of psychological states through the content analysis of of laboratory induced learned helplessness. Journal of verbal behavior. Los Angeles: University of California Abnormal Psychology, 86, 615-623. Press. Coyne, J. C., Aldwin, C., & Lazarus, R. S. (1981). Depres- Hamilton, E. W., & Abramson, L. Y. (1983). Cognitive sion and coping in stressful episodes. Journal of Ab- patterns and major depressive disorder: A longitudinal normal Psychology, 90, 439-447. study in a hospital setting. Journal of Abnormal Psy- Coyne, J. C., & Gotlib, I. H. (1983). The role of cognition chology, 92, 173-184. in depression: A critical appraisal. Psychological Bulletin, Hammen, C., & Cochran, S. D. (1981). Cognitive correlates 94, 472-505. of life stress and depression in college students. Journal Cronbach, L. J. (1951). Coefficient alpha and the internal of Abnormal Psychology, 90, 23-27. structure of tests. Psychometrika, 16, 297-334. Hammen, C., & deMayo, R. (1982). Cognitive correlates Cutrona, C. E. (1983). Causal attributions and perinatal of teacher stress and depressive symptoms: Implications depression. Journal of Abnormal Psychology, 92, 161- for attributional models of depression. Journal of Ab- 172. normal Psychology, 91, 96-101. Danker-Brown, P., & Baucom, D. H. (1982). Cognitive Harvey, J. H., Ickes, W. J., & Kidd, R. F. (Eds.). (1976). influences on the development of learned helplessness. New directions in attribution research (Vol. 1). Hillsdale, Journal of Personality and Social Psychology, 43, 793- NJ: Erlbaum. 801. Harvey, J. H., Ickes, W. J., & Kidd, R. F. (Eds.). (1978). Dweck, C. S., & Licht, B. (1980). Learned helplessness New directions in attribution research (Vol. 2). Hillsdale, and intellectual achievement. In J. Garber & M. E. P. NJ: Erlbaum. Seligman (Eds.), Human helplessness (pp. 197-221). Harvey, J. H., Ickes, W. J., & Kidd, R. F. (Eds.). (1981). New York: Academic Press. New directions in attribution research (Vol. 3). Hillsdale, Eaves, G., & Rush, A. J. (1982). Cognitive patterns in NJ: Erlbaum. symptomatic and remitted unipolar depression. Un- Heider, F. (1958). The psychology of interpersonal relations. published manuscript, University of Texas. New York: Wiley. Eaves, G. & Rush, A. J. (1984). Cognitive patterns in Hiroto, D. S., & Seligman, M. E. P. (1975). Generality of symptomatic and remitted unipolar major depression. learned helplessness in man. Journal of Personality and Journal of Abnormal Psychology, 93, 31-40. Social Psychology, 31, 311-327. Elder, G. (1974). Children of the Great Depression. Chicago: Janoff-Bulman, R. (1979). Characterological versus be- University of Chicago Press. havioral self-blame: Inquiries into depression and rape. Elder, G., Bettes, B. A., & Seligman, M. E. P. (1982). Journal of Personality and Social Psychology, 37, 1798- Unpublished data, Cornell University. 1809. Epstein, S. (1980). The stability of behavior: Implications for psychological research. American Psychologist, 89, Kovacs, M., & Beck, A. T. (1977). An empirical-clinical 678-682. approach toward a definition of childhood depression. Feather, N. D., & Davenport, P. R. (1981). Unemployment In J. G. Schulterbrandt & A. Raskin (Eds.), Depression and depressive affect: An attributional analysis. Journal in childhood: Diagnosis, treatment, and conceptual of Personality and Social Psychology, 41, 422-436. models (pp. 1-25). New York: Raven Press. Firth, J., & Brewin, C. (1982). Attributions and recovery Langer, E. J. (1978). Rethinking the role of thought in from depression: A preliminary study using cross-lagged social interaction. In J. H. Harvey, W. J. Ickes, & correlational analysis. British Journal of Clinical Psy- R. F. Kidd (Eds.), New directions in attribution research chology, 21, 229-230. (Vol. 2, pp. 35-58). Hillsdale, NJ: Erlbaum. Frankel, A., & Snyder, M. L. (1978). Poor performance Lewinsohn, P. M., Steinmetz, J. L., Larson, D. W., & following unsolvable problems: Learned helplessness or Franklin, J. (1981). Depression-related cognitions: An- egotism? Journal of Personality and Social Psychology, tecedent or consequence. Journal of Abnormal Psy- 36, 1415-1423. chology, 90, 213-219. Freud, S. (1957). Mourning and melancholia. In J. Strachey Lloyd, C. (1980). Life events and depressive disorder re- (Ed. and Trans.), Standard edition of the complete psy- viewed: I. Events as predisposing factors. II. Events as CAUSAL EXPLANATIONS AND DEPRESSION 373

precipitating factors. Archives of General Psychiatry, 3 7, Overmier, J. B., & Seligman, M. E. P. (1967). Effects of 529-548. inescapable shock upon subsequent escape and avoid- Luborsky, L. (1964). A psychoanalytic research on mo- ance learning. Journal of Comparative and Physiological mentary forgetting during free association. Philadelphia Psychology, 63, 23-33. Association of Psychoanalysis Bulletin, 14, 119-137. Pasahow, R. J. (1980). The relation between an attributional Luborsky, L. (1970). New directions in research on neurotic dimension and learned helplessness. Journal of Abnor- and psychosomatic symptoms. American Scientist, 58, mal Psychology, 89, 358-367. 661-668. Peirce, C. S. (1955). The philosophical writings ofPeirce. Lusted, L. B. (1968). Introduction to medical decision J. Buchler (Ed.). New York: Dover. making. Springfield, IL: Charles C Thomas. Persons, J. B., & Rao, P, A. (1981). Cognitions and depres- MacCorquodale, K., & Meehl, P. E. (1948). On a distinction sion in psychiatric inpatients. Unpublished manuscript, between hypothetical constructs and intervening vari- University of Pennsylvania. ables. Psychological Review, 55, 95-107. Peterson, C. (1978). Learning impairment following in- Maier, S. R, & Jackson, R. L. (1979). Learned helplessness: soluble problems: Learned helplessness or altered hy- All of us were right (and wrong): Inescapable shock has pothesis pool? Journal of Experimental Social Psy- multiple effects. In G. H. Bower (Ed.), The psychology chology, 14, 53-68. of learning and motivation (Vol. 13, pp. 155-218). New Peterson, C., Bettes, B. A., & Seligman, M. E. P. (1982). Vbrk: Academic Press. Spontaneous attributions and depressive symptoms. Maier, S. E, & Seligman, M. E. P. (1976). Learned help- Unpublished manuscript, Virginia Polytechnic Institute lessness: Theory and evidence. Journal of Experimental and State University. Psychology: General, 105, 3-46. Peterson, C., & Conn, M. K. (1982). Depressive mood Maier, S. R, Seligman, M. E. P., & Solomon, R. L. (1969). reactions to breaking up: A learned helplessness per- Pavlovian fear conditioning and learned helplessness. spective. Unpublished manuscript, Virginia Polytechnic In B. A. Campbell & R. A. Church (Eds.), Punishment Institute and State University. and aversive behavior (pp. 229-243). New \brk: Ap- Peterson, C., Luborsky, L., & Seligman, M. E. P. (1983). pleton-Century-Crofts. Attributions and depressive mood shifts: A case study Manly, P. C., McMahon, R. J., Bradley, C. R, & Davidson, using the symptom-context method. Journal of Ab- P. O. (1982). Depressive attributional style and depression normal Psychology, 92, 96-103. following childbirth. Journal of Abnormal Psychology, Peterson, C., Nutter, J., & Seligman, M. E. P. (1982). 91, 245-254. Unpublished data, Virginia Polytechnic Institute and McFarland, C., & Ross, M. (1982). Impact of causal at- State University. tributions on affective reactions to success and failure. Peterson, C, & Raps, C. S. (1983). Depression, attributions, Journal of Personality and Social Psychology, 43, 612- and attributional style. Manuscript in preparation, Vir- 617. ginia Polytechnic Institute and State University. Metalsky, G. I., Abramson, L. Y., Seligman, M. E. P., Peterson, C., Schwartz, S. M., & Seligman, M. E. P. (1981). Semmel, A., & Peterson, C. (1982). Attributional styles Self-blame and depressive symptoms. Journal of Per- and life events in the classroom: Vulnerability and in- sonality and Social Psychology, 41, 253-259. vulnerability to depressive mood reactions. Journal of Peterson, C., & Seligman, M. E. P. (1981). Helplessness Personality and Social Psychology, 43, 612-617. and attributional style in depression. Tiddsskrift for Miller, I. W., & Norman, W. H. (1979). Learned help- Norsk Psykologforening, 18, 3-18; 53-59. lessness in humans: A review and attribution theory Peterson, C., Semmel, A., von Baeyer, C., Abramson, model. Psychological Bulletin, 86, 93-119. L. Y, Metalsky, G. I., & Seligman, M. E. P. (1982). Miller, I. W., & Norman, W. H. (1981). Effects of attri- The Attributional Style Questionnaire. Cognitive Ther- butions for success on the alleviation of learned help- apy and Research, 6, 287-299. lessness and depression. Journal of Abnormal Psychology, RadlofF, L. (1975). Sex differences in depression: The effects 90, 113-124. of occupational and marital status. Sex Roles, 1, 249- Mukherji, B. R., Abramson, L. Y., & Martin, D. J. (1982). 265. Induced depressive mood and attributional patterns. Raps, C. S., Peterson, C., Jonas, M., & Seligman, Cognitive Therapy and Research, 6, 15-21. M. E. P. (1982). Patient behavior in hospitals: Help- Navarra, A. (1981). Unpublished data, University of lessness, reactance, or both? Journal of Personality and Pennsylvania. Social Psychology, 42, 1036-1041. Nisbett, R., & Wilson, T. D. (1977). Telling more than we Raps, C. S., Peterson, C., Reinhard, K. E., Abramson, can know; Verbal reports on mental processes. Psycho- L. Y, & Seligman, M. E. P. (1982). Attributional style logical Review, 84, 231-259. among depressed patients. Journal of Abnormal Psy- Nolen-Hoeksema, S. (1983). Attributional styles, life events, chology, 91, 102-108. and depression in children. Unpublished master's thesis, Roth, S. (1980). A revised model of learned helplessness University of Pennsylvania, Philadelphia. in humans. Journal of Personality, 48, 103-133. O'Hara, M. W., Rehm, L. P., & Campbell, S. B. (1982). Rotter, J. B. (1966). Generalized expectancies for internal Predicting depressive symptomatology: Cognitive-be- versus external control of reinforcement. Psychological havioral models and postpartum depression. Journal of Monographs, 80(1, Whole No. 609). Abnormal Psychology, 91, 457-461. Schulterbrandt, J. G., & Raskin, A. (Eds.). (1977). Depres- Orne, M. T. (1962). On the social psychology of the psy- sion in children: Diagnosis, treatment and conceptual chology experiment: With particular reference to de- models. New York: Raven Press. mand characteristics and their implications. American Schwab, J. J., Bialow, M. R., Brown, J. M., Holzer, C. E., Psychologist, 17, 776-783. & Stevenson, B. E. (1967). Sociocultural aspects of 374 CHRISTOPHER PETERSON AND MARTIN E. P. SELIGMAN

depression in medical inpatients. Archives of General Taylor, S. E. (1979). Hospital patient behavior: Reactance, Psychiatry, 17, 539-543. helplessness, or control? Journal of Social Issues, 35, Seligman, M. E. P. (1972). Learned helplessness. Annual 156-184. Review of Medicine, 23, 407-412. Tennen, H. (1982). A re-view of cognitive mediators in Seligman, M. E. P. (1974). Depression and learned help- learned helplessness. Journal of Personality, 50, 526- lessness. In R. J. Friedman & M. M. Katz (Eds.), The 541. psychology of depression: Contemporary theory and re- Tennen, H., Affleck, G., Allen, D., McGrade, B. J., & search (pp. 83-113). Washington, DC: Winston. Ratzan, S. (1983). Causal attributions and coping in Seligman, M. E. P. (1975). Helplessness: On depression, juvenile diabetes. Unpublished manuscript, University development, and death. San Francisco: Freeman. of Connecticut. Seligman, M. E. P. (1978). Comment and integration. Tennen, H., & Eller, S. J. (1977). Attributional components Journal of Abnormal Psychology, 87, 165-179. of learned helplessness and facilitation. Journal of Per- Seligman, M. E. P. (1981). A learned helplessness point sonality and Social Psychology, 35, 265-271. of view. In L. P. Rehm (Ed.), Behavior therapy for Tesiny, E. P., & Lefkowitz, M. M. (1982). Childhood depression: Present status and future directions (pp. 123- depression: A 6-month follow-up study. Journal of Con- 141). New \ork: Academic Press. sulting and Clinical Psychology, 50, 778-780. Seligman, M. E. P., Abramson, L. Y., Semmel, A., & von Thoits, P. A. (1983). Dimensions of life events that influence Baeyer, C. (1979). Depressive attributional style. Journal psychological distress: An evaluation and synthesis of of Abnormal Psychology, 88, 242-247. the literature. In H. Kaplan (Ed.), Psychosocial stress: Seligman, M. E. P., & Maier, S. F. (1967). Failure to escape Trends in theory and research (pp. 33-103). New York: traumatic shock. Journal of Experimental Psychology, Academic Press. 74, 1-9. Timko, C., & JanofT-Bulman, R. (1983). Attributions, vul- Seligman, M. E. P., Maier, S. F., & Solomon, R. L. (1971). nerability, and coping: The case of breast cancer. Un- Unpredictable and uncontrollable avcrsive events. In published manuscript, University of Massachusetts. F. R. Brush (Ed.), Aversive conditioning and learning Vaillant, G. E. (1977). Adapatation to life. Boston: Little, (pp. 347-400). New York: Academic Press. Brown. Seligman, M. E. P., & Peterson, C. (in press). A learned Velten, E. (1968). A laboratory task for induction of mood helplessness perspective on childhood depression: Theory states. Behaviour Research and Therapy, 6, 473-482. and research. In M. Rutter, C. Izard, & P. Read (Eds.), Winer, B. J. (1971). Statistical principles in experimental Depression in childhood: Developmental perspectives. design (2nd ed.). New \brk: McGraw-Hill. New York: Guilford. Wong, P. T. P., & Weiner, B. (1981). When people ask Seligman, M. E. P., Peterson, C., Kaslow, N. J., Tanenbaum, "why" questions, and the heuristics of attribution search. R. L., Alloy, L. B., & Abramson, L. Y. (1984). Ex- Journal of Personality and Social Psychology, 40, 649- planatory style and depressive symptoms among chil- 663. dren. Journal of Abnormal Psychology, 93, 235-238. Wortman, C. B., & Brehm, J. W. (1975). Response to Seligman, M. E. P., & Weiss, J. (1980). Coping behavior: uncontrollable outcomes: An integration of reactance Learned helplessness, physiological activity, and learned theory and the learned helplessness model. In L. Ber- inactivity. Behaviour Research and Therapy, 18, 459- kowitz (Ed.), Advances in experimental social psychology 512. (Vol. 8, pp. 277-336). New York: Academic Press. Silver, R. L., Wortman, C. B., & Klos, D. S. (1982). Cog- Wortman, C. B., & Dintzer, L. (1978). Is an attributional nitions, affect, and behavior following uncontrollable analysis of the learned helplessness phenomenon viable?: outcomes: A response to current human helplessness A critique of the Abramson-Seligman-Teasdale refor- research. Journal of Personality, 50, 480-514. mulation. Journal of Abnormal Psychology, 87, 75-90. Simonton, D. K. (1981). The library laboratory: Archival Zemore, R., & Johansen, L. J. (1980). Depression, help- data in personality and social psychology. In L. Wheeler lessness, and failure attributions. Canadian Journal of (Ed.), Review of personality and social psychology (Vol. Behavioural Sciences, 12, 167-174. 2, pp. 217-243). Beverly Hills, CA: Sage. Zuckerman, M., & Lubin, B. (1965). Manual for the Mul- Spitzer, R. L., & Endicott, J. (1977). Schedule for affective tiple Affect Adjective Check List. San Diego, CA: Ed- disorders and schizophrenia—Lifetime version. New ucational and Industrial Testing Service. York: Biometrics Research, New York State Psychiatric Institute. Sweeney, P. D., Schaeffer, D., & Golin, S. (1982). Attri- bution about self and others in depression. Personality Received May 18, 1983 and Social Psychology Bulletin, 8, 37-42. Revision received December 15, 1983