future event schemas and certainty about the future

7/29/2019 Future Event Schemas and Certainty About the Future

1/13

Journal of Personality and Social Psychology1992, Vol. 63, No. 5, 711-723Copyright 1992 by the American Psychological Association, Inc.0022-3514/92/S3.00

Future-Event Schemas and Certainty About the Future:Automaticity in Depressives' Future-Event PredictionsSusan M . Andersen, Lisa A. Spielman, an d John A . Bargh

New "York UniversityThe proposition was tested that depressives make predictions about the future based on a pessimis-tic future-event schema. Participants varying in depression predicted whether positive and negativeevents would happe n to them (or to an average person) in the future by pressing yes or no at acomputer terminal as quickly as possible, either under a concurrent attentional load or under nosuch load. As hypothesized, depressives predicted more negative events and fewer positive eventsthan did mild depressives or nondepressives and showed greater automaticity in their predictions.That is, the attentional load did not increase depressives' response latencies for either negative orpositive events, even though it did so reliably for both mildly depressed and nondepressed individ-uals. Depressives may thus possess a highly developed future-event schem a tha t operates efficientlyin enabling future-event predictions.

Hopeless expectations about the future represent a funda-mental com ponent of depression, at least according to cognitivetheories of this d isorder (e.g., Abram son, Seligman, & Teasdale,1978; Beck, 1967). Learned helplessness theory, for example, isnow conceptualized as a hopelessness model of depression(Abramson, Metalsky, & Alloy, 1989; Abramson et al., 1978).Beck's theory of depression, as well, holds tha t hopeless expec-tations about the future play an important role (Beck, 1967,1976). Although central to cognitive the ories, the hopelessnessconstruct has received relatively little empirical attentionbeyond demonstrating that, as a psychometric instrument, it isassociated with depression (Beck, Brown, Steer, Eidelson, &Riskind, 1987; Beck, Riskind, Brown, & Steer, 1988; Beck,Weissman, Lester, & Trexler, 1974; Minkoff, Bergman, Beck, &Beck, 1973; although see, e.g., Andersen, 1990; Rholes, Ris-kind , & Neville, 1985).

Predictive Certainty and HopelessnessIt has been suggested that hopelessness is best conceptual-ized as the point at which an individual begins to experienceEditor's Note. One of the authors of this article is anassociate editorof the Journal of Personality an d Social Psychology: Attitudes an d So -cial Cognition section. Therefore, this article was subjected to a specialreview process to avoid any conflict of interest. Eliot Smith handledthe article; he solicited the reviews and made the recommendation that

the article be published. I am grateful to him for taking on t his chore.Portions of this article were presented at the E astern PsychologicalAssociation's Annual Convention, Boston, April 1989, and at theWorld Congress of Cognitive Therapy, Toronto, On tario, Ca nada , June1992. We thank Irene Zombeck and Kathleen G allagher for serving asexperimenters in this research and four anonymous reviewers for in-sightful comments on a previous version of the article.Lisa A. Spielman is now at the Payne Whitney Clinic, New 'York.Correspondence concerning this article should be addressed to Su-san M. A ndersen, De partm ent of Psychology, New 'York University, 6Washington Plac e, 4th Floor, New York, New York 10003.

and treat dreaded events (or a single dread ed event) as inevitablerather tha n simply likely. W hen dread ed events seem certain tooccur (Andersen & Lyon, 1987; Garber, Miller, & Abramson,1980), they are treated as if they have already occurred andgiven a reality that is functionally equivalent to their havingalready transpired (Andersen, 1990; Beck, 1976; see also John-son & Raye, 1981; Riskind, Rholes, Brannon, & Burdick,1987). One is certain enough tha t one ceases to expend effort;one gives up.By contrast, when events are not seen as inevitable, one maybe able to retain some hop e that things might transp ire differ-ently. In this case, efforts may even be renewed to stave off thedreaded events (see Taylor & Brown, 1988; Wortman & Brehm,

1975). The notion that loss of hope occurs when an individualmakes the subjective judgmen t th at dreaded events are inevita-ble is implicit in the concept of hopelessness. In this sense,hopelessness is best conceptualized as depressive predictive cer-tainty, the point at which dreaded future events are treated ascertain to occur or that desired future events are treated ascertain not to o ccur (Andersen, 1990; Andersen & Lyon, 1987;Garbe r et al., 1980; Stotland, 1969).Predictive Certainty and the PessimisticFuture-Event Schema

In this sense, the onset of depression can be viewed as atrajectory or course that begins with increasingly pessimisticpredictions (e.g., Andersen, 1990; Pyszczynski, Holt, & Green-berg, 1987; Riskind et al., 1987) that are initially relatively un -certain in nature (see Garber et al., 1980; Wortman & Brehm,1975). When people become depressed, by contrast, theyclearly show depressive predictive certainty relative to those w hoare only mildly depressed (Andersen, 1990). Mild depressives,although pessim istic, lack this type of certainty in the ir future-event predictions; m oreover, individuals who are not depressedat all ("normals") show no predictive certainty of any sort (noteven about the occurrence of positive events). Depressive pre-dictive ce rtainty increases with de pth of depression. Similarly,711


2/13

712 S. ANDERSEN, L. SPIELMAN, AND J. BARGHin terms of self-schematic thinking, there is considerable evi-dence for negative self-schematic thinking among depressives(eg., Bargh & Tota, 1988; Segal, H ood, Shaw, & Higgins, 1988)and, by contrast, considerable evidence for a lack of such nega-tive self-schematic thinking among mild depressives (e.g., Bargh& Tota, 1988; Kuiper & MacDonald, 1982; Kuiper, Olinger,MacDonald, & Shaw, 1985; see also MacDonald & Kuiper,1984; MacDonald, Kuiper, & Olinger, 1985). Hence, differ-ences between depressives and mild depressives in self-sche-matic thinking appear to co-occur with comparable differ-ences in depressive predictive certainty.

These parallels are interesting because predictive certaintymay also involve a form of schematicity. Tha t is, it is conceivablethat depressive predictive certainty results in pa rt from the in-creasing tendency among mild depressives and depressives tothink about the future and to do so in increasingly pessimisticwaysthat is, to rum inate about possible positive and negativefuture outcom es, about the various unknown s, and about whatmight happen (e.g., Moretti & Shaw, 1989). T his m ay constitutea kind of mental simulation or rehearsal (see also Taylor &Schneider, 1989) that is increasingly pessimistic and that leadsto increased facility and ease in making future-event predic-tions, so that these p redictions come to be made relatively auto-matically among people who are truly depressed, with minim alinvestment of cognitive effort.Relative effortlessness or efficiency is one of several featuresthought to be associated with automaticity. Others include u n-intentionality, uncontrollability, and lack of awareness. Al-though th e various features associated with automaticity d o notcovary together in an all-or-none sense, in tha t a given processcould possess one or more, bu t not other, features of autom atic-ity (Bargh, 1989; Logan, 1989; Logan & Cowan, 1984), the termautomaticity can be used if the exact defining feature used isspecified. In the present work, we concern ourselves with ef-

fortlessness or efficiency, that is, with the notion that the pro-cess requires little cognitive effort.Our basic proposition is that, through experience and rehear-sal, depressives may acquire a highly efficient knowledge struc -ture or schema for predicting future events, one that operatesrelatively effortlessly. To the extent that processing efficiency orfluenc y exists in anticipating future events, this greater easemight be experienced by the individual as greater truth value,as something that feels unquestionably correct or certain , and isthus treated as such (Bargh, 1989; Jacoby & Kelley, 1987,1990).The notion that depressives make pessimistic predictionsabout future events and do so with certainty is consistent withthe notion that they possess a well-elaborated future-eventschema, containing beliefs about what the future will be like,whereas mildly depressed and nondepressed individuals donot. This is, in fact, suggested by Beck's (1967) conce ptualiza-tion of the cognitive triad in d epression. In Beck's (1967) model,there are thre e com ponents to the depressive's system of nega-tive beliefs that are theorized to lead to depressed mood andother affective and motivational concom itants of depression: Anegative view of the self, of the world, and of the future (Beck,1967, pp . 256-2 61). These are considered to be separate thoughinterrelated patterns of belief that contribute to depressivesymptom s. That is, the depressive's views of the future a re notidentical to his or her views of the self, nor do they simply spring

from these self-constructs (see also Janoff-Bulman, 1989).Thoughts abou t the future are of independ ent relevance to de-pression, even though they clearly must take into account be-liefs about the self and about the world as well (e.g., otherpeople).Beck's (1967) description of the cognitive phenomenology ofdepression is consistent with the present conceptualization ofdepressive predictive certainty, in that depressive expectationsabou t the future involve an inability to conceive of any improve-ment in the current state of affairs. "If [the depressed person]regards himself as currently deprived, immobilized, or re-jected, he visualizes a future in which he is continually de-prived, imm obilized, or rejected" (Beck, 1967, p. 260). In fact,"when [the depressive] awakens in the morning , he anticipatesthat every experience during the day will pose great difficul-ties. . . . When a suggestion is made to engage in an activitythat he ordinarily enjoys when not depressed, he automatically[emphasis added ] assumes he will not have a good time " (Beck,1967; p. 260). Hence, depressives ought to make predictionsthat are considerably more pessimistic than are those made bynormals (as has been shown, e.g., Andersen, 1990) and ought tomake them more automatically.

Evidence for the Role of PredictiveCertainty in DepressionThe relevance of predictive certainty to depression has beendemonstrated both experimentally and correlationally. An ex-perimental study showed, for example, that when an individualgives up in relation to a dreaded future event, this leads to asharp increase in depressive affect (Andersen & Lyon, 1987). Inthis work, negative-outcome certainty was defined as th e beliefthat a dreaded event was 100% likely as compared with otherincreasingly high outcome likelihoods. More specifically, the

perceived probability that an extremely aversive event wouldoccur later in the experiment was manipulated am ong normalcollege students, who w ere told that the negative event was ei-ther 0%, 25%, 50%, 75%, or 100% likely (without their being ableto exercise any control over this likelihoo d). The discontinuityhypothesis (pitting certainty against likelihood) was confirmedby precipitous increases in depressive affect in the 100% condi-tion, significantly greater than in any other condition; further-more, there was no linear increase in depressive affect acrossthe manipulated probability conditions, only the discon-tinuous increase. Hence, anticipating a highly aversive eventthat is certain to occur can have a causal impact on depressiveaffect (Anderson & Lyon, 1987).Of course, it is one thing to say that the c ertainty with whichnegative events are anticipated can cause depressive affect andanother to show that depressed people actually experience suchnegative-outcome certainty. If negative-outcome certainty isimportant in depression in ways not reflected simply by contin-uous likelihood ratings (which have been term ed pessimism), itshould also be associated with depression independently of pes-simism in samples in which all of these factors are allowed tovary freely. That is, people who make the judgment thatdreaded events will definitely oc cur should have reached a kindof set point that is associated with depression in a manner notreflected by mere pessimism. A study examining this hypothe-


3/13

FUTURE-EVENT SCHEMAS 71 3sis demonstrated that the relation between negative-outcomecertainty and depressive affect, supported in experimentalwork (Andersen & Lyon, 1987), also holds as a correlationalpattern among a random sample of college studen ts (Andersen,1990). Tha t is, increases in depression (as assessed by the BeckDepression Inventory; BDI) are associated bo th w ith depressivepessimism (estimating that negative events are likely and thatpositive events are not) and with certainty in mak ing these pre-dictions. Moreover, depressive certainty is uniquely associatedwith depression, even when the simple perceived likelihood ofthe events are considered. These findings extend other workshowing that depressed individuals often perceive negativeevents as likely and positive events as unlikely (Alloy & Ahrens,1987; Hollon & Garber, 1980; Lobitz & Post, 1979; Pietromo n-aco & M arkus, 1985; Pyszczynski et a l, 1987; Riskind et al ,1987) and show that the certainty with which these prediction sare made is uniquely associated with depression. Such ce rtaintyis less prevalent in m ild depression.1

The Relevance of Automatic Processing ofSelf-Information Among Depressives

The concept of predictive certainty in depression dovetailsnicely with the information-processing literature on depres-sion. As indicated, this literature suggests that depressed ind i-viduals possess highly consolidated "depressive" self-schemata,that is, well-organized self-representations that are negativelytoned compared with the self-schemata of nondepressives(Kuiper & Derry, 1981; Kuiper, MacDonald, & Derry, 1983).Evidence for such a depressive self-schema clearly exists amo ngdepressives (Bargh & Tota, 1988; Segal et a l, 1988).Most work on the information-processing consequences ofdepression derives from Beck's (1967,1976) notion that a cen-tral characteristic of depression is autom atic negative self-refer-ential thought. Automatic thinking is seen as tied to the activa-tion of negatively toned cognitive structures, typically negativeself-schemata. When such structures are activated, automaticthoughts follow. If we define automaticity in terms of process-ing efficiency, depressives should show efficiency in processingnegative information based on the negative self-schema (Bargh& Tota, 1988; Kuiper, Olinger, & Mac Donald , 1988).

To test the hypothesis that depressed people process negativeinformation about the self more automatically tha n do nonde-pressives, recent research has used a con current mem ory loadparadigm (Bargh & Tota, 1988). The principle behind this par-adigm is that attentional capacity is limited; only so much at-tentional capacity is available at any one tim e (e.g, Kahneman,1973; Miller, 1956). For this reason, a task tha t is particularlycapacity demanding ought to interfere with another cognitivetask that is being done simultaneously to the extent that thislatter task also requires cognitive capacity (Kantowitz, 1974;Logan, 1979). Such a capacity-demanding task should not in-terfere much with another task that requires little capacity.Hence, the less effort or attention required by a task (i.e, themore automated the task), the less it should be affected by con-straints on the availability of attentional resources.There is general agreement that efficiency-based automatic-ity should be measured using this kind of dual-task, conc urrentmemory-load paradigm. In such a paradigm, two tasks com-

pete for one's attention, an d t he consequenc es for the effectiveperformance of one of the two tasks are assessed (Logan, 1988).The procedure is based on the modal view of automaticity,which assumes a single, limited pool of attentional capacity tha tcan be drained by a capacity-demanding task (see reviews byJohnson & H asher, 1987; Logan, 1988; Posner & Snyder, 1975;Shiffrin & Schneider, 1977). Alternative conceptions (of bothautomaticity and the nature of processing capacity) do, ofcourse, exist, some of which do not require processing-capacityassumptions (e.g, Logan, 1988). Nonetheless, the preferredmethod for assessing efficiency-based automaticity continuesto be the dual-task paradigm; moreover, the modal view ofautomaticity con tinues to be the limited-resource view.

In the present study, we use a dual-task paradigm to studyautomaticity and assume a general limited pool of attentionalcapacity (as did Bargh & To ta, 1988). Two major objections havebeen raised in relation to this assumption. One is that atten-tional capacity is no t a singular entity. Instead, multiple pools ofresources exist so that n o single pool can be draine d by a givenattentional load manipulation (e.g, Navon & Gopher, 1979;Wickens, 1984). This criticism has been directed at claims ofautomaticity based on a lack of interference from a second task(in a dual-task paradigm) when the re is no way of knowing thatboth tasks share the same poo l of resources (Navon & Gopher,1979, p. 250). This objection, however, canno t be m ade, when,as in the present study, two or m ore groups of subjects performthe same two tasks, and the prediction is that one group willshow interference and another will not. Under these condi-tions, the dual-task paradigm is on firm ground with respect tothe single-resource assump tion.The second objection to th e single-resource view is that auto-maticity may n ot be due to a n ever more efficient use of cogni-tive capacity but instead to the accrual, with frequency andconsistency of experience in a domain, of specific event in-

stances in memory, as proposed by the in stance theory of auto-maticity (Logan, 1988). Even in this m odel, however, less inter-ference by a concurrent load task (i.e, dual-task interference) ispredicted when components of the relevant task are automa-tized (Logan, 1988, p. 513). It offers a different explanation forwhy automatic processes show less interference from a conc ur-rent task than do nonautomatic ones, but this lesser interfer-ence is nonetheless the exact phenomenon for which instancetheory attem pts to account. In short, conclusions can be drawnabout autom aticity from dual-task paradigm s when they assess

1 In this research, pessimism was operationalized as the subjects'average likelihood rating of a series of negative future events on a scaleranging from - 5 to + 5, minus the ir average likelihood rating of a seriesof positive events. Depressive predictive certainty was operationalizedas the number of+5s endorsed on the same scale for the negative eventsplus the number o f- 5s endorsed for the positive events. It was assumedthat "if subjects treated the othe r 10 points of th e 11-point scale differ-ently from th e predicte d endpoin t of the scale, by circling the predictedendpoint even after having had the opportunity to express unc ertaintyby choosing any other point on the scale, this reflects a decision in thefavor of certainty. The issue is that dep ressed people m ay feel and act asif the event were certain , which circling any endpo int implies." (Ander-sen, 1990, p. 210). Sim ple likelihoo d or confidence ratings do not ade-quately index the certainty cons truct; they measure pessimism.


4/13

714 S. ANDERSEN, L. SPIELMAN, AND J. BARGHdifferences in automaticity between groups of people (e.g., de-pressed and nondepressed) that are performing the same tasks(as in Bargh & Tota, 1988).Using such a paradigm , researchers examining self-judgmentamong depressives recently asked both depressed and nonde-pressed subjects to jud ge the descriptiveness of a series of posi-tive and negative trait adjectivesin relation to themselves orin relation to another person (Bargh & Tota, 1988). Half ofthese subjects also performed a concurrent memory load taskwhile making these judgments by holding random sets of digitsin memory. The load condition clearly increased depressives'response latencies in making self-judgments about negativetraits less than nondepressives' latencies in self-judging thesame traits and less than their own latencies in self-judgingpositive tra its.

Hence, for depressed people, deciding w hether negative traitsare self-descriptive, on seeing them, is a relatively automatictask, whereas making these same decisions about positive traitsis not. Negative self-constructs show greater automaticityamong depressives than am ong nondepressives and greater au-tomaticity than do positive self-constructs. Interestingly, in thecase of positive traits and nondepressives, the findings wereliterally reversed. The load manipulation interfered less withnondepressives' self-judgments about positive traits than withthose of depressives and less than with their own judgmentsabout negative traits. Hence, automaticity also exists amongnondepressives but in judging the self-descriptiveness of posi-tive traits.Overall, these data imply that a negative self-schema (or set ofnegative self-constructs) is more accessible for depressives tha nfor nondepressives. That is, when depressed people think ab outthe self, negative constructs come to mind and are used morereadily, tha t is, with less cognitive effort an d a ttention, th an areother con structs. The opposite is true for nondepressives.

Predictive Certainty and Automaticity in PredictingFuture EventsConceptualizing the processes associated with certainty indepression in information-processing terms, it is apparent thatnegative associations are readily activated am ong depressed in-dividuals and that this ease of retrieval or immediacy maymake these associations feel more "real" and "true" to the de-pressed individual than other associations (Andersen, 1990;Bargh & Tota, 1988) because of the relative lack of felt effort ordeliberate thought (see Bargh, 1989; Jacoby & Kelley, 1987,1990). In other words, the amount of mental effort needed tointerpret or categorize an event may be used as a cue as to how

closely the even tfits hat interpretation; if little effort is neededand the interpretation is made relatively automatically, the p er-son feels it is valid.The Certainty-as-Efficiency Hypothesis

Hence, on e way of conceptualizing depressive predictive cer-tainty is in terms of efficiency in particular types of informa-tion processing, that is, as ease in considering th e possibility offuture suffering and in making effortless judgments about itsoccurrence (Andersen, 1990). If mild depression commences

with increasingly frequent tho ught ab out th e future that is in-creasingly pessimistic as compared with nondepressives (e.g.,Andersen, 1990; Riskind et al, 1987), this process may ulti-mately become automatized. That is, depressed individualsmay come to possess a pessimistic and highly efficient future-event schema that enables relatively effortless predictions aboutfuture events and that operates in accordance with a depressivepredictive certainty that is not apparent among mildly de-pressed or nondepressed individuals.Accordingly, we predicted that depressed individuals wouldmake predictions about the future more pessimistically thanwould mildly depressed or nondepressed individuals and thatthey would show automaticity in m aking their future-event pre-dictions. A depressive future-event schema should contain pre-dictions that are relatively m ore pessimistic than those made byother subjects, tha t is, relatively more yes-to-negative and no -to-positive predictions, and should operate automatically. Hence,depressives' response latencies in ma king future-event predic-tions were not expected to increase reliably on the basis of aload on the ir attentional resources, whereas a reliable increasewas expected among mildly depressed and nondepressed indi-viduals, who were assumed not to possess a future-eventschema and thus were not expected to show automaticity intheir predictions. Supporting this notion, other research hasshown no t only pessimism in depressives' predictions (Alloy &Ahrens, 1987; Andersen, 1990; Pyszczynski et al., 1987; Ris-kind et al., 1987) but certainty as well. Such certainty existsconcerning both positive and negative future events, is less prev-alent among m ild depressives, and does not exist among nonde-pressives (who are not even optimistically certain about futureevents; Andersen, 1990).

Depressive certainty has most recently been operationalizedas a particular type of extreme respon ding th at takes bo th eventvalence and yes-no response into account (Andersen, 1990),that is, as the selection of the most extreme subjective likeli-hood rating of negative events occurring and for positive eventsnot occurring (Andersen, 1990). Depressives clearly do thismore often than do m ild depressives and nondepressives, eventhough they may do so for only a small num ber of future events.Operationalizing certainty in this way is important in that ittaps into those expectancies around which depressives mayhave given up hope. Focusing exclusively on such events, how-ever, m akes it impossible to identify any other more generalizeddifferences in predictive certainty between depressives andnondepressives th at may exist. For example, if depressives pos-sess a well-developed future-event schema th at is relatively pe s-simistic, as we propose, all the predictions deriving from itought to be made more automaticallywith greater certaintythan are the predictions made by people with no suchschema.In a similar vein, research on self-schematic processing indepression (as applied to trait adjectives) has suggested thatautomatic processes cannot be clearly indexed using self-reportdecisions, such as yes or no responses, or their associated laten-cies. The problem is that response type te nds to tap into delib-erative processes in addition to construct activation and flu-ency, such as self-presentational concerns (Bargh & Tota, 1988;Wyer & Gordon, 1984). Moreover, response type (yes or no)rarely interacts with relevant variables to predict response la-


5/13

FUTURE-EVENT SCHEMAS 715tency (Bargh & Tota, 1988; Rogers, 1974). Hence, it was in partfor this reason that the certainty-as-efficiency hypothesis in-cluded no interaction with response type, although greater pes-sismism a mong depressives than among o ther subjects is a nec-essary componen t of our definition of the future-event schema.

The Matching HypothesisOur conceptualization of depressive predictive certainty asthe efficient operation of a future-event schema is rather differ-ent from the framework typically invoked to account for find-ings in the self-schema literature. In self-judgment research,depressives are thought to judge the self-relevance of negativetraits more quickly than that of positive traits (regardless ofwhether they respond with a yes or with a no judgment) becausethere is a match between the ir chronically accessible self-con-structs (negative ones) and the valence of the potentially self-de-scriptive adjective (also negative). When there is a match be-tween the valence of the trait adjective and the cognitive con-struct, the matching hypothesis of schema theory holds that thestimulus should be processed more automatically (Bargh &Tota, 1988; see also Spielman & Bargh, 1990).Hence, just as is the case in judging the potential self-de-scriptiveness of trait adjectives, when depressed individualsconfront negative information about the future (a negative fu-ture event) and judge it in relation to the self, the matchinghypothesis suggests that their prediction be made automati-cally, due to the stimulus-self-relevant construct match. Thus,depressives should show autom aticity in the ir predictions aboutnegative events but not about positive events (as has been shownfor trait judg men ts, Bargh & Tota, 1988). Moreover, because itis the self-schema that is pessimistic in nature, this pessimisticmatching should take place only for self-judgments and not forjudgments about others (unless of course depressives possess

pessimistic, negative construc ts in a variety of domains otherthan the self, in which case such negative matching could bemore widespread).The matching hypothesis, however, m ay well be less applica-ble to future-event predictions than to making current judg-ments, such as trait judgments. Its central tenet is that an under-lying construc t in memory corresponds directly to the currentjudgment task, in that the individual possesses an existingstructure in mem ory directly relevant to the self-judgment: theself-schema. On presentation of a specific to-be-judged trait,the individual accesses the existing self-schema to determinewhether a match exists. The more accessible the traits, thestronger the match between the self and the trait stimuli andthe more automatic the judgment.For predicting what will ha ppen over time, however, the self-schema may be less effective, given its essential focus on thecurrent attributes of the individual (e.g., Derry & Kuiper, 1981;Markus, 1977). Future-event pred ictions m ust, logically speak-ing, rely not only on one's notions about who one is at presentbu t on the state of the world (which includes other people) andhow it will evolve over time in relation to the self (see also Beck,1967; Janoff-Bulman, 1989). Hence, although predictionsabout the future may involve reference to the self, considerableadditional information is also required, suggesting that self-schema m atching processes may not drive future-event predic-

tions. If they did, the hypothesis would be that future-eventpredictions would show automaticity only when they concernnegative events considered in relation to the self, that is, whenthere is a match to the depressive self-schema.In a different vein, the matching hypothesis would suggestthat nondepressed individuals should show automaticity forpositive future events, as has been found for trait judgments(Bargh & Tota, 1988). By contrast, the certainty frameworkholds that because people cannot know exactly what their fu-tures hold for them, nondepressives should not possess an elabo-rated future-event schema, and little automaticity should thusbe found in their future-event predictions. In short, the frame-work suggests that som ething beyond self-schemata a nd pe rti-nent matching processes is likely to be required in accountingfor future-event predictions among depressives and, specifi-cally, that whereas depressives possess a (pessimistic) future-event schema, nondepressives do not.

On th e basis of our conce ptualization of predictive certaintyas efficiency in the operation of a relatively pessimistic future-event schema, we predicted that automaticity would exist indepressives' predictions abou t bo th positive and negative futureevents relative to both mildly depressed an d nond epressed per-sons. We also examined the extent to which the obtained effectswould be limited to the context in which people predict theirown future experiences or, by contrast, would extend to m akingjudgments about other individuals as well. If depressive predic-tive certainty is linked exclusively with self-constructs and notwith other personal constructs used by the individual, the pre-dicted effects should be observed only when th e self is the targetabout whom the prediction is made, which would demonstrateself-concept mediation.

MethodOverview

Conceptualizing depressive predictive ce rtainty in terms of auto-matic future-event predictio ns, we designed a study in which depressedand nondepressed subjects were asked to make predictions about thefuture, using either themselves or the average student as the target ofthe prediction. Seated at a computer screen, they pressed yes or no asquickly as they could on a response box in front of them to indicatewhether each of a series of positive and negative future e vents was likelyto occur in the future, either while under a concurrent attentional loador under no such cognitive load. This was a modified version of theBargh and Tota (1988) paradigm, different only in that positive andnegative future events served as stimuli (similar to those used by An-dersen, 1990) instead of positive and negative descriptive traits.

SubjectsSubjects were 68 undergraduate students enrolled in introductorypsychology at New \brk University, who participated in partial fulfill-ment of a course requirem ent. Subjects were preselected based on theirscores on the BDI (Beck, Ward, Mendelson, Mock, & Erbaugh, 1961),administered as part of a mass testing session and again at the end ofthe experim ent. Subjects were classified as nondepressed if they scored4 or below on th e BDI at the experimental session, as mildly depressedif they scored between 9 and 13, and as moderately depressed if theyscored 14 or greater (as in Andersen, 1990). Ultimately, the study in-


6/13

71 6 S. ANDERSEN, L. SPIELMAN, AND J. BARGHeluded 36 nondepressed, 15 mildly depressed, and 17 moderately de-pressed subjects.Apparatus and Materials

All experimental instructions and stimuli w ere presented to subjectsusing a cathode ray tube (CRT) monitor, under the control of an Apple11+ microcomputer located in a separate room (as in Bargh & Tota,1988). Subjects responded to the presented stimuli by pressing one oftwo buttons (marked yesand no) located on a response box on the tablein front of them. In the attentional load condition, a sm all microphonewas placed on the table as well, near the response box, ostensibly torecord subjects as they repeated aloud each of the random six-digitnumbers to which they were exposed in this c ondition.Fifty-two positive a nd negative future events were drawn from a poolthat combined events used in Andersen (1990) and events developedfor this study. The original events varied co nsiderably in term s of theirnormative perceived likelihood for the average college student, and weadded still others that seemed, based on face validity, to be eitherextremely likely or extremely unlikely to occur. These items were in-cluded to ensure that all subjects would m ake at least some yes andsome no responses for both the positive and the negative events, per-mitting us to examine response latencies as a function of yes or noresponse. (In Bargh & Tota, 1988, this was not possible because nonde-pressed subjects rarely responded yes to the negative adjectives.) Tenpractice trials of neutrally valenced items were included as buffers atthe beginning of each session. All items are presented in Table 1.

ProcedureSubjects were run individually. After being greeted by the experi-menter, subjects read a brief description of the basic task, which theywere to perform entirely by computer, and signed a consent form. Theythen read detailed instructions, presented on the computer monitor.The instructions informed them that they would be asked to indicateas quickly as possible, by pressing yes or no on the response box,whether each of a series of events to be presented on the screen waslikely to happ en either to th em or to the average New York Universityundergraduate, depending on the prompt that would precede everytrial. In addition, subjects in the cognitive load condition were in-formed that, before each trial, a six-digit number would appear on thescreen that they would be asked to hold in mind while making theirjudgment and that they would be asked to say this number aloud intothe microphone after each trial.After indicating that they understood the instructions, subjectscompleted a series of 10 practice trials with the experimenter present tomake sure they understood the procedures. The experimenter then leftthe room, and the experimental trials began. Each future event waspresented only once to each subject.In each trial, subjects were first presented with the target personlabel and then, followed by a 1-s delay, with the future event to bejudged. Their response terminated the display and thus enabled themto self-pace the experimental trials. The future events were presentedin one of two random orders, within which the order of the targetperson (self or average person) was randomized, so as to prevent sub-

Table 1Life Events That Subjects Judged W ould or Would Not Happen at Some Timein the Future by the Event's Face-Value Likelihood

Negative events Positive eventsExceptionally likely events

Die (eventually)Accidentally break somethingCatch a coldGo to a concertGet into an argumentHave a headacheHave an unexpected expense

Have a warm shower or bathListen to musicGet a letter from a family memberBe out on a sunny day

Events varying in likelihoodBe stuck in an unfulfillingjobRegret a m ajor life decisionWork with people I don't likeExperience the death of a spouseGet a fatal diseaseSuffer a great financial lossBe unhappy in long-term relationshipsBe comm itted to a mental institutionFail to contribute to societyBe very lonely when oldEventually become senileCause someone to sufferBe divorced

Cope successfully with job stressAchieve life goalsLive a healthy and active lifeHave an ideal life-styleLive in an ideal homeDo enjoyable thingsHav e a sexually fulfilled lifeHave plenty of moneyHave an ideal jo bBe satisfied with major life decisionsGain the respect of my co-workersBe able to retire at age 40

Exceptionally unlikely eventsBe murderedGo blindBe kidnappedHave a home destroyed by a tornadoBe convicted of a felonyLose an armBe struck by lightning

Win the lotteryWin a N obel prizeBecome a film starMake 10 million dollarsBe elected to CongressWrite a best-seller


7/13

FUTURE-EVENT SCHEMAS 717jects from developing any target expectancies that might artifactuallyaffect response tim es (see Bargh & Tota, 1988). Each future event to berated was either negative or positive and was either very likely, variablein likelihoo d, or not very likely. Equal numbers of events from eachcategory were randomly paired with the self or the average-otherprompt.In the attentional load c ondition, a six-digit number was presentedat the beginning of each trial for 2 s, followed by a 1-s blank screenpause and then the target (self or average other) and the future event,both presented as in the no-load condition. After subjects made the irjudgmen t, and before the next trial, they were prompted to repeat thesix-digit number alo ud. They w ere given 5 s to do this before the onsetof the next trial. The experimenter kept track of the accuracy of thedigit responses throughout the exp eriment.For each of the 52 trials, accuracy scores in digit task performancefor subjects in the load condition ranged from 6 (all six digits recalledcorrectly) to 0 (no digit recalled correctly). An analysis of the averagenumber of errors subjects made across all trials showed that there wereno significant differences in the number of errors made by nonde-pressed (M = 1.25), mildly depressed (M = 1.67), and depressed sub-jects (M = 1.83), F(2, 20) < 1.Following the event trials, subjects completed th e BDI to verify the irdepression group classification (made on th e basis of their e arlier BDIscores; see Kendall, Hollon, Beck, Ha mm en, & Ingram, 1987; Spiel-man & Bargh, 1990). The experimenter was blind to subjects' depres-sion status throughout th e experim ent. Finally, subjects were debriefedand thanked for their participation.

ResultsYes-No Responses

Although our central hypotheses concern the effect of mem -ory load on response latencies, such latencies are theoreticallymeaningful only if depressives show pessimism in the ir positiveand negative future-event predictions. Subjects' yes-no re-sponses to predict whether or not th e future events would hap-pen to them or to the average person were analyzed by calculat-ing the percentage of yes responses to each type of event foreach subject and examining the scores in a 3 (depression) X 2

(load) X 2 (future-event type) X 2 (self-other target) analysis ofvariance (ANOVA), with two repeated measures.As predicted, the analysis yielded a highly reliable interac-tion between depression and future-event ty pe, F(2,64 ) = 9.83,p < .0005. These endorsem ent rates are presented in Figure 1.Planned contrasts indica ted th at depressed subjects said yes toreliably more negative events than did mildly depressed or non-depressed subjects, f(64) = 6.46, p < .001, and said no to reliablymore positive events, z(64) = 3.29, p < .001. In addition, w hereasnondepressed subjects said yes to reliably more positive eventsthan negative events, t(64) = 7.00, p < .001, as did mildly de-pressed subjects, t(64) = 1.62, p < .05, depressed subjects saidyes to an approxim ately equal num ber of positive and negativeevents (/ < 1). (All planned contrasts are one-tailed / tests basedon the error term from the relevant interaction.) Hence, thesedata clearly show greater pessimism among depressives thanamong other subjects.

This analysis also yielded thre e other reliablefindings.First,it yielded a main effect for future-event type, F (l, 64) = 28.52,p< .0001, such that subjects, on the average, endorsed morepositive events (M = 0.62) than negative events (M = 0.47),although th is was qualified by the interaction reported earlier.Second, a two-way interaction between self-other target andfuture-event type also emerged, F(l, 64) = 19.24, p < .0001,indicating th at subjects, on the average, predicted more positiveevents for themselves (M = 0.67) than for others (M = 0.57),t(64) = 3.48, p < .001, and fewer negative events (self, M = 0.42;others, M = 0.52 ), f(64) = 3.62, p < .001. Thisfinding s consis-tent with p revious studies of self-other differences in trait ad-jective endorsement, which typically have found that depres-sives and nondepressives alike endorse more positive traits forself tha n for o ther (e.g., Bargh & Tota, 1988; Pietromona co &Markus, 1985; see review by C oyne & Go tlib, 1983). No otherreliable effects emerged for the self-other target variable.Finally, the analysis produced an unexpected two-way inter-action between depression and load, F(2, 64) = 3.27, p < .05,indicating a linear trend of increasing yes responses with in-

% of "yes" responses

nondepressed mildly depressed depressedpositive negative

Figure 1. Mean endorsement rates for possible future life eventsby valence of the event and depression group.


8/13

718 S. ANDERSEN, L. SP1ELMAN, AND J. BARGHcreasing depression under no load (M s = 0.528, 0.552, and0.614) and a curvilinear trend under load (Ms = 0.533, 0.580,and 0.509). There was also a trend for depressed subjects to sayyes to somewhat fewer events in the load condition (M = 0.509)than in the no-load condition (M = 0.614), whereas the rem ain-ing subjects did n ot. Although none of these trends was signifi-cant (t < 1.21, ns), the latter is one that is potentially problem-atic.When yes-no responses change under load, it poses prob-lems for interpreting response latency differences in term s ofautomaticity, even though it is not uncommon for people's re-sponses to change as they adapt to the need for quick decisions(e.g., Bargh & Tota, 1988; Wyer & Gordon, 1984). For example,when subjects deliberate strategically in producing th eir y es-noresponses in the no-load condition an d this is no longer permit-ted in the load condition, the deliberative basis for their yes -noresponse is prevented and may thus leave only responses thatare m ore automatic. This is, in principle, why a load co nditionis a better indicator of nonstrategic responding than a no-loadcondition.

On the other hand, such afindingcreates an interpretationalambiguity for predicted differences in response latency be-tween the load and no-load conditions. That is, one cannotunequivocally interpret such differences in term s of automatic-ity when subjects change what they are doing across the condi-tions. If some subjects respond almost as quickly in the loadcondition as in the no-load con dition, while others show inter-ference effects, but the former subjects also change their actualresponses, it may be their change in response th at prevents thedual-task interference rather than automaticity per se. Al-though th e relevant contrast was nonsignificant in our data, thesuggestion of such a difference poses an interpretational diffi-culty and indicates that our response latency data should betreated with some caution.In a different vein, it is also relevant to note that our problem -atic interaction between yes-n o response and load actually sug-gests only that depressed subjects said no somewhat more oftenacross the board when strategic effort was not available. Be-cause our response latency prediction s do not involve any inter-action with ye s-no response and one might expect to find suchan interaction if a change in yes-no response is what is responsi-ble for differential effects among depressives versus nondepres-sives, it is at least somewhat less likely that this across-the-boardchange in depressives' responses could account for our pre-dicted response latency findings.

Response LatenciesThe response latency data were examined in a 3 (depres-sion) X 2 (load) x 2 (future-event type) X 2 (yes-no response) X 2(self-other target) ANOVA, in which the latter three factorswere repeated measures. O ur principal prediction was a greaterautomaticity (smaller increases due to load) in depressives' pre-dictions about future events than in those made by mild depres-sives or nondepressives. If event valence plays no role, thisshould amount to a two-way interaction between depressionand load; if event valence is relevant, it should result in a three-way interaction involving depression, load, and future-eventtype. The analysis, in fact, yielded the three-way interaction,

F(2,62 ) = 4.20, p < .05, and not the two-way interaction (F < 1).The four-way interaction between subjects' yes-no responsesand these three factors was not reliable (Fs < 1), nor was thefour-way interaction involving the self-other target factor andthese three factors (Fs < 1 ).2To test our specific hypotheses, we conducted plan ned con-trasts that compared the load and no-load conditions for eachtype of future event (positive or negative) for each depressiongroup. For ease of presentation, Figure 2 portrays the load -no-load difference scores that directly reflect the magnitude of theresponse delay difference tested by our contrast. These con-trasts clearly showed that the difference between the load andthe no-load conditions was reliable for both negative and posi-tive events among nondepressed subjects, positive events,t(62) = 2.72, p < .005; negative events, t(62) = 1.61, p = .06, aswell as among mild depressives, positive events, ;(62) = 2.59,p < .01; negative events, /(62) = 4.03, p < .001. As predicted,however, the load condition had no reliable effect on depressedsubjects' response latencies; that is, it did not reliably increasetheir response latencies in makin g predictions about either neg-ative or positive future events (ts < 1), as in the certainty-as-fluency hypothesis. Cell means an d marginals are presented inTable 2.

When an attentional load manipulation has little impact onthe response latencies of a given set of subjects for a given set ofitems, even though the m anipulation drains cognitive capacityenough to significantly increase response latencies for othersubjects (or for the sam e set of subjects using different items), itsuggests that the target judgments were made relatively auto-matically by these subjects. Hence, the lack of dual-task interfer-ence among depressives implies greater automaticity in theirpredictions abou t both negative and positive future events, al-though some caution is warranted due to their slightly differentyes-no responses under load.In interpreting these data, the argument might also be raisedthat depressives may have shown no load effects simply becausethey for some reason do not show load effects in this kind ofparadigm. This can be ruled out by the fact that depressiveshave shown load effects when judg ing th e self-relevance of posi-tive trait adjectives (and not negative trait adjectives; Bargh &Tota, 1988). Ideally, of course, we would have included a controlcondition in which subjects made some altogether differenttype of prediction as well for which interference effects couldbe demonstrated, ensuring that they do show load effects withpredictions. Nonetheless, the existing data clearly favor the no-tion tha t depressives show interference in this para digm .Considering the data slightly differently, we examined notautomaticity in predictions, but absolute speed of processing

2 The analysis did yield two highly reliable main effects. That is,predictions about the self were made more quickly (M = 1,548) thanthose about another person (M = 1,662), F(l, 62) = 28.11, p < .001, afinding that is now commonplace (see Bargh & Tota, 1988; Spielman &Bargh, 1990, for a review). In addition, predictions about positiveevents tended to be made more quickly (M =1,571) than predictionsabout negative events (M = 1,647), F(l, 62) = 17.99, p < .001. It shouldagain b e noted , however, that in the absenc e of differential load effectson response latency conclusions about automaticity cannot be drawnfrom such data.


9/13

4 0 0

300 -

200 -

100 -

FUTURE-EVENT SCHEMASload/noload difference score (in ms)

719

- 100 nondepressed mildly depressed depressed

V/////M posi t ive negativeFigure 2. Mean difference scores between th e load and the no-loadconditions by valence of the event an d depression group.

under load across the depression groups. The data showed thatunder attentional load, depressed subjects made predictionsabout the negative events significantly more quickly (M = 1,601)than did mildly depressed subjects (M = 1,771), r(62) = 2.12,p < .025, or nondepressed subjects {M = 1,738), f(62) = 2.05,

p < .025. On the other h and, providing some supp ort for differ-ences in how negative versus positive events are anticipated,depressed subjects in the load condition did not differ signifi-cantly from the other subjects in th e latency of their positive-event predictions (both ts < 1), although the means were in the

Table 2Response Latency (in Milliseconds) to Predict That Positive or Negative Future Events Will orWill Not Happen to Self or to Another Person by Nondepressed, Mildly Depressed, andDepressed Individuals as a Function of Attentional Load Condition

Measure

SelfYesN oMeanOtherYesN oMeanNegative mean

SelfYesN oMeanOtherY esN oMeanPositive meanAll events

Nondepressives_,oad No load

,661,638,650,821,829,825,738

,603,576,590,744,736,740,665,702

Negative,596,558,577,668,739,704,641

Mild depressivesLoad N o load

events1,6411,8721,7571,7771,7911,7841,771

Positive events,460,544,502,633,509,571,537,589

1,5601,5461,5531,7121,8481,7801,6671,719

,432,426,429,443,436,440,435

,490,315,403,4851,433,459,431,433

DepressivesLoad N o load

1,4911,5741,5331,6831,6551,6691,601

1,6261,5491,5881,6701,6191,6451,6171,609

,683,536,610,634,763,699,655

,449,366,4081,7761,488,632,520,588


10/13

720 S. ANDERSEN, L. SPIELMAN, AND J. BARGHsame direction. T hus, under a ttentional load, depressives' pre-dictions about negative events were made more rapidly thanwere those of mildly depressed or nondepressed subjects,whereas their predictions about positive events were not. Thegreater automaticity with which depressives make their future-event predictions appears to extend to both negative and posi-tive events, whereas absolute speed differences across depres-sion groups (under load) hold only for negative events.3Figure 2 also makes it obvious that the notion of uncertaintyin the future-event predictions of mildly depressed subjects isstrongly supported by these data. Nothing at all close to auto-matic responding can be seen in these latency differences. Aspredicted, the load-no-load differences for mildly depressedsubjects were unusually large for both positive, /(62) = 2.59, p


11/13

FUTURE-EVENT SCHEMAS 721negative and fewer positive future events than did other sub-jects. Combined with the apparent efficiency with which de-pressives made th eir prediction s, these data p rovide support forthe no tion tha t depressive predictive certainty can be conceptu-alized in information-processing terms, as the certainty-as-effi-ciency hypothesis suggests. Th at is, these findings can be ac-counted for by assuming that depressives possess a well-devel-oped knowledge structure about the future, includingpessimistic expectancies, that m ild depressives and nondepres-sives do not, one that enables automatic future-event predic-tions. This future-event schema includes both positive and nega-tive events and relatively pessimistic expectations abo ut the seevents (as compared with the expectations of mildly depressedand nondepressed people). When such conceptions of the fu-ture come so automatically and effortlessly to mind , they maytend to be viewed as particularly unequivocal and inevitable(see Jacoby & Kelley, 1987,1990).

The data are thus readily handled by the conceptualizationof depressive predictive certainty as efficiency in relation to awell-developed future-event schema held by depressives. Al-though the present findings yielded no greater automaticityamong depressives for yes responses to negative events and noresponses to positive events than for other types of responses,there is reason to expect all of the p redictions within th e future-event schema to be made more automatically. As anticipated,depressed subjects predicted more negative events and fewerpositive events than did other subjects, showing pessimism intheir expectations, and generally made their predictions moreautomatically as well.The matching hypothesis, by contrast, derived from self-schema theory, suggests that automaticity should be observedonly in depressives' predictions ab out negative events and not intheir predictions about positive ones. In fact, research on self-judgments has shown that depressives make automatic self-

judgments about negative traits but not about positive ones(Bargh & Tota, 1988) presumably because negative self-con-structs become activated automatically for depressives. Yet, inthe present study, both negative and positive future-event pre-dictions appear to have been made more automatically by de-pressives. This is consistent with previous work on depressivepredictive certainty (Andersen, 1990), which has dem onstratedsuch certainty for both types of events, and is also consistentwith the certainty-as-efficiency hypothesis.On a different level, the matching hypothesis also predictsautomaticity among nondepressed individuals about positiveevents but not about negative events, as in present self-judg-ments (Bargh & Tota, 1988). Our data, however, show th at non-depressives do not m ake future-event predictions automatically.

This is consistent with research on depressive predictive cer-tainty among nondepressives, in which nondepressives appearto show little certainty about th e future (Andersen, 1990). Simi-larly, Beck (1967) discussed schematized processing of future-event possibilities only among depressives. Hence, our d ata a reconsistent with the general notion that maintaining a view ofthe future that is relatively flexible or nonschematized, that is,that leaves at least some room for thoughtful analysis or foruncertainty, may be adaptive in conceptualizing future out-comes (see also Andersen & Schwartz, 1992). Th at is, it may notbe very useful to think that one knows with certainty exactly

what the future will hold. If one were to think this, it mightminimally result in disappointment, and more seriously, in fail-ing to initiate and choreograph desired outcomes in one's life.Finally, although the matching hypothesis can apply to per-sonal con structs, broadly speaking, as well as to self-schemata,in particular, most relevant research has focused on self-sche-mata and on judgm ents about the current self. In such research,the negative trait automaticity found among depressives is lim-ited to self-judgments (Bargh & Tota, 1988). Depressives auto-matically think about the self in negative terms and actuallyautomatically thin k about others in positive terms. The presentdata show, by contrast, that depressives' certainty about thefuture extends not only to b oth positive and negative events butto both the self and others; our findings were not limited toself-predictions. Although these data cannot definitively ruleout the role of self-construct activation in depressives' future-event predictions, they clearly provide no evidence for self-con-cept mediation.

Overall, it is clear that processes other than the simplematching of chronically accessible negative constructs andavailable stimulus information (the future events examined inthis study) must be invoked to account for our data. Depressivesare thought to be far more concerned about the future than arenondepressives, thinking about it constantly (Beck, 1967). AsBeck (1967) pointed ou t, "the depressed pa tient generally showsconsiderable preoccupation with ideas of the future. His expec-tations usually have a negative cast and m ay occur in the formof pictorial fantasies or as obsessive ruminations" (p. 259). Suchfrequent thought abou t a domain is a prerequisite for the devel-opm ent of an efficient schema (e.g., Hayes-Roth, 1977). That is,frequent and consistent thinking within a content domain,such as the future, is the precursor of efficient, automatizedthought processes (see Bargh, 1984,1989; Shiffrin & Schneider,1977). A depressive future-event schema would enable quick,relatively attention-free judgments to be made about futureevents and would contain beliefs about both positive and nega-tive eventsthat is, relatively pessimistic expectations aboutthese events (Beck, 1967, pp. 259-26 1).

Interestingly, we also found support for the hypothesis thatthe constructs that mildly depressed individuals use to makepredictions about future events are in a state offlux.The future-event predictions of mild depressives revealed no automaticitywhatsoever an d a very different pattern than shown by depres-sives or nondepressives. The data, in fact, suggest considerableuncertain ty in the future-event predictions of these individuals,both for positive and for negative events. The differences be-tween the load and the no-load conditions for these subjectswere an average of four times greater than those for depressedor nondepressed subjects. Hence, the data strongly suggest thatthere is actual uncertainty in the constructs with which milddepressives make their future-event predictions. Th is is consis-tent with the notion that predictive certainty is an index ofseverity in depression, such that it distinguishes mild depres-sives from those more seriously depressed (Andersen, 1990).The d ata also speak provocatively to the present conc eptual-ization of how efficiency or certainty in future-event predic-tions may develop among depressives. Th at is, mildly depressedindividuals should not yet have formed a future-event schemaand should thus show no efficiency, as we found. Moreover, we


12/13

722 S. ANDERSEN, L. SPIELMAN, AND J. BARGHassume that continued, strategic mental rehearsal or rumina-tion about future events in initial stages of dysphoria may ulti-mately become automatized in more serious depression (seeMoretti & Shaw, 1989; Tait & Silver, 1989; see also Taylor &Schneider, 1989). Although the present data cannot test thisrumination hypothesis, mild depressives did show some un-usual "strategic" responses. That is, when these subjects hadattentional capacity available to them (in the no-load condi-tion), they made their predictions about both negative and posi-tive events surprisingly rapidly and more rapidly than did anyother group. These extremely fast responses were clearly strate-gic, because when attentional capacity was drained, theirstored constructs showed no automaticity. Although one canonly speculate about the meaning of these findings, it is possi-ble that such strategic responsesquickly predicting futureeventsmay eventually become set into place as automatic pro-cess among moderately and severely depressed individuals. In asimilar vein, these responses may reflect a need to appearknowledgeable or certain about future outcomes among milddepressives, perhaps to counteract the disarray or uncertaintyof their stored constructs. These speculative accounts clearlywarrant continued research.

In conclusion, our findings demonstrate the relevance of theconcept of depressive predictive certainty to understanding thecognitive processes of depressed individuals and provide fur-ther explanation for why it is often so difficult to successfullydispute the negative interpretational and predictive patterns ofdepressed individuals (Andersen, 1990; Andersen & Lyon,1987; Bargh &Tota, 1988; Beck, 1967,1976). Although furtherresearch is needed to specify more precisely the exact mecha-nisms underlying depressive predictive certainty, it appears tohave its locus in the automatic cognitive processes throughwhich depressives anticipate future events, which may feel un-questionably real and true to the individual.

ReferencesAbramson, L. Y, Metalsky, G. I., & A lloy, L. B. (1989). Hopelessnessdepression: A theory-based subtype of depression. Psychological Re-view, 96, 358-372.Abramson, L. Y, Seligman, M. E. P., & Teasdale, J. D. (1978). Learnedhelplessness in humans: Critique and reformulation. Journal of Ab-normal Psychology, 87 , 49-74.Alloy, L. B., & Ahrens, A. H. (1987). Depression and pessimism for thefuture: Biased use of statistically relevant information in predictionsfor self versus others. Journal of Personality an d Social Psychology,52 , 366-378.Anderse n, S. M. (1990). The inevitability of future suffering: Th e roleof depressive predictive certainty in depression. Social Cognition, 8,

203-228.Andersen, S. M., & Lyon, J. E. (1987). Anticipating undesired out-comes: The role of outcome ce rtainty in the onset of depressive af-fect. Journal of Experimental Social Psychology, 23 , 428-443.Andersen, S. M., & Schwartz, A. H. (1992). Intolerance of ambiguityand depression: A cognitive vulnerability factor linked to hopeless-ness. Social Cognition, 10, 271-298.Bargh, J. A. (1984). A utomatic an d cognitive processing of social infor-mation. In R. S. Wyer & T. K. Srull (Eds.), Handbook ofsocial cogni-tion (Vol. 3, p p. 1-43). Hillsdale, NJ: Erlbaum.Bargh, J. A. (1989). Conditional autom aticity: Varieties of au tomaticinfluence in social perception and cognition. In J. S. Uleman & J. A.

Bargh (Eds.), Unintended thought (pp. 3-51). New York: GuilfordPress.Bargh, J. A., & Tota, M. E. (1988). Context-dependent automatic pro-cessing in depression: Accessibility of negative constructs with re-gard to self but not others. Journal ofPersonality and Social Psychol-ogy, 54, 925-939.Beck, A. T. (1967). Depression: Clinical, experimental, an d theoreticalaspects. New York: Hoeber.Beck, A. T. (1976). Cognitive therapy and the emotional disorders. Mad-ison, CT: International Universities Press.Beck, A. T, Brown, G, Steer, R. A., Eidelson, J. I., & Riskind, J. H.(1987). Differentiating anxiety and depression: A test of the cogni-tive content-specificity hypothes is. Journal of Abnormal Psychology,96,179-183.Beck, A. T., Riskind, R. H., Brown, G, & Steer, R. A. (1988). Levels ofhopelessness in DSM-III disorders: A partial test of content speci-ficity in depression. Cognitive Therapy and Research, 1 2, 459-470.Beck, A. X, Ward, C. H., Mendelson, M, Mock, J. E., & Erbaugh, J. K.(1961). An inventory for measuring depression. Archives of GeneralPsychology, 4, 561-571.

Beck, A. X, W eissman, A., Lester, D , &T rexler, L. (1974). The m easure-ment of pessimism: The hopelessness scale. Journal of Consultingan d Clinical Psychology, 42 , 861-865.Coyne, J. C , & Gotlib, I. H. (1983). The role of cognition in depression :A critical appraisal. Psychological Bulletin, 94, 472-505.Derry, EA.,& Kuiper, N. A. (1981). Schematic processing and self-ref-erence in clinical depression. Journal of Abnormal Psychology, 90,286-297.Garbe r, J., Miller, S. M , & Abramson, L. Y. (1980). On the distinctionbetween anxiety and depression: Perceived control, certainty, andprobability of goal attainment. In J. Garber & M. E. P. Seligman(Eds.), Human helplessness: Theory an d applications (pp. 131-169).San Diego, CA: Academic Press.Hayes-Roth, B. (1977). Evolution of cognitive structu res and proc esses.Psychological Review, 84 , 260-278.Hollon, S. D, & Garber, J. (1980). A cognitive-expectancy theory oftherapy. In J. Garber & M. E. P. Seligman (Eds.), Human helplessness:Theory and applications (pp. 173-195). San Diego, CA: AcademicPress.Jacoby, L. L., & Kelley, C. M. (1987). Unconscious influences of mem -ory for a prior event. Personality an d Social Psychology Bulletin, 13,314-336.Jacoby, L. L., & Kelley, C. M. (1990). An episodic view of m otivation:Unconscious influences of memory. In E. T. Higgins & R. M.Sorrentino (Eds.), Handbook ofmotivation an d cognition (Vol. 2, p p.451-481 ). New York: Guilford Press.Janoff-Bulman, R. (1989). Assumptive worlds and the stress of trau-matic events: A pplications of the schema construct. Social Cogni-tion, 7, 113-138.Johnson, M. K., & Hasher, L. (1987). Human learning and memory.Annual Review of Psychology, 38 , 631-668.

Johnson, M. K ., & Raye, C. L. (1981). Reality m onitoring. Psychologi-cal Review, 88 , 67-85.Kahneman, D. (1973). Attention and effort. Englewood Cliffs, NJ:Prentice-Hall.Kantowitz, B. H. (1974). Double stimulation. In B. H. K antowitz (Ed.),Human information processing: Tutorials in performance an d cogni-tion (pp. 83-132). Hillsdale, NJ: Erlbaum.Kendall, P. C, Hollon, S. D, Beck, A. X, Ham men , C. L., & Ingram,R. E. (1987). Issues and recom mendations regarding use of the BeckDepression Inventory. CognitiveTherapy and Research, 11,289-299.Kuiper, N. A., & Derry, P. A. (1981). The self as a cognitive prototyp e:An application to person perception and depression. In N. Cantor &


13/13

FUTURE-EVENT SCHEMAS 723J. F. Kihlstrom(Eds.), Personality, cognition, & social interaction (pp.215-232). Hillsdale, NJ: Erlbaum.Kuiper, N. A., & MacD onald, M. R. (1982). Self and other perceptionin mild depressives. Social Cognition, I, 223-239.Kuiper, N. A., MacDonald, M . R., & Derry, P. A. (1983). Parameters ofa depressive self-schema. In J. Suls & A. G. Greenwald (Eds.), Psycho-logical perspectives on the self: Vol. 2 (pp. 191-218). San Diego, CA:Academic Press.

Kuiper, N. A., Olinger, L. J., & MacDonald, M. R. (1988). Vulnerabilityand episodic cognitions in a self-worth contingency m odel of de-pression. In L. B. Alloy (Ed.), Cognitive processes in depression (pp.289-309). New York: Guilford Press.Kuiper, N. A., Olinger, L. 1, MacD onald, M . R., & Shaw, B. F. (1985).Self-schema processing of depressed and nondepressed content: Theeffects of vulnerability to depression. Social Cognition, 3, 77-93.Lobitz, W C, & Post, R. D. (1979). Parameters of self-reinforcementand depression. Journal of Abnormal Psychology, 88 , 33-41.Logan, G. D. (1979). On the use of a concurrent mem ory load to mea-sure attention and automaticity. Journal of Experimental Psychology:Human Perception an d Performance, 5,189-207.Logan, G. D. (1988). Toward an instance theory of automatization.Psychological Review, 95 , 492-527.Logan, G. D. (1989). Automaticity and cognitive control. In J. S. Ule-man & J. A. Bargh (Eds.), Unintended thought (pp. 52-74). New York:Guilford Press.Logan, G. D, & Cowan, W B. (1984). On the ability to inhibit thoughtand action: A theory of an act of control. PsychologicalReview, 91 ,295-327.MacDonald, M . R., & Kuiper, N. A. (1984). Self-schema decision con-sistency in clinical depressives. Journal of Social andC linicalPsychol-ogy, 2, 264-272.MacDonald, M . R., Kuiper, N. A., & Olinger, L. J. (1985). Vulnerabilityto depression, m ild depression, and deg ree of self-schema consolida-tion. Motivation and Emotion, 9 , 369-379.Markus, H. (1977). Self-schemata and processing information aboutthe self. Journal of Personality an d Social Psychology, 35 , 63-78.Miller, G. A. (1956). The magical number seven, plus or minus two:

Some limits on our capacity for processing information. Psychologi-cal Review, 65,81-97.Minkoff, K., Bergman, E., Beck, A. T, & Beck, R. (1973). Hopeless-ness, depression, and attempted suicide. American Journal of Psychi-atry, 130, 455-459.Mo retti, M . M., & Shaw, B. F. (1989). Autom atic and dysfunctionalcognitive processes in depression. In J. S. Uleman & J. A. Bargh(Eds.), Unintended thought (pp. 383-421). New York: Guilford Press.Navon, D., & Gopher, D. (1979). On the economy of the human process-ing system. Psychological Review, 86 , 214-255.Pagel, M., & Becker, J. (1987). Depressive thinking and depression:Relations with personality and social resources. Journal ofPersonal-ity and Social Psychology, 52 , 1043-1052.Pietromonaco, P. R., & Markus, H. (1985). The nature of negativethoughts in depression. Journal of Personality and Social Psychology,

48 , 799-807.Posner, M . I., & Snyder, C. R. R. (1975). Attention and cognitive con-trol. In R. L. Solso (Ed.), Information processing and cognition: TheLoyola Symposium (pp. 55-85). Hillsdale, NJ: Erlbaum.

Pyszczynski, X, Holt, K., & Greenberg, J. (1987). Depression, self-fo-cused attention, and expectancies for positive and negative futurelife events for self and others. Journal ofPersonality an d Social Psy-chology, 52 , 94-1001.Rhodewalt, F, & Agustsdottir, S. (1986). Effects of self-presentation onthe phenomenal self. Journal of Personality an d Social Psychology,50 , 47-55.Rholes, W S., Riskind, J. H., & Neville, B. (1985). The relationship of

cognitions and hopelessness to depression an d anxiety. Social Cogni-tion, 8, 36-49.Riskind, J.H., Rholes, W S., Brannon, A. M., & Burdick, C. A. (1987).Attributions and expectations: A confluence of vulnerabilities inmild depression in a college student population. Journal ofPersonal-ity and Social Psychology, 53 , 349-354.Rogers, T. B. (1974). An analysis of two central stages underlying re-sponding to personality items: The self-referent decision and re-sponse selection. Journal of Research in Personality, 8,128-138.Segal, Z. V, Hood, J. E., Shaw, B. E, & Higgins, E. T. (1988). A struc turalanalysis of the self-schema construct in major depression. CognitiveTherapy and Research, 12 , 471-485.Shiffrin, R. M., & Schneider, W (1977). Controlled and automatic hu-man information processing: II. Perceptual learning, automatic at-tending, and a general theory. Psychological Review, 84 ,127-190.Spielman, L. A., & Bargh, J. A. (1990). Does t he depressive self-schemareally exist? In C. D. McCann & N. S. Endler (Eds.), Depression: Newdirections in research, theory, an d practice (pp. 111 -126). Toronto,Canada: W all & Em erson.Stotland, E. (1969). The psychology of hope. San Francisco: Jossey-Bass.Taba chnik, N , Crocker, J., & Alloy, L. B. (1983). Depre ssion, socialcomparison, and the false-consensus effect. Journal of Personalityan d Social Psychology, 45 , 688-699.Tait, R., & Silver, R. C. (1989). Coming to terms with major negativelife even ts. In J. S. Ulem an & J.A. Bargh (Eds.), Unintended thought(pp. 351-382). New York: Guilford Press.Taylor, S. E., & Brow n, J. D. (1988). Illusion an d well-being: A socialpsychological perspective on mental health. Psychological Bulletin,103,193-210.Taylor, S. E., & Schneider, S. K. (1989). Coping and the simulation of

events. Social Cognition, 7,176-196.Weary, G., Elbi n, S., & Hill, M . G. (1987). Attribution and social com-parison processes in depression. Journal of Personality an d SocialPsychology, 52 , 605-610.Wickens, C. D. (1984). Processing resources in attention. In R. Parasur-aman & R. Davies (Eds.), Varieties of attention (pp. 63-102). SanDiego, CA: Academic Press.Wortman, C. M., & Brehm, J. W (1975). Responses to uncontrollableoutcomes: An integration of reactance theory and the learned help-lessness model. In L. Berkowitz (Ed.), Advances in experimental so-cial psychology (Vol. 8, pp. 227-336). San Diego, CA: AcademicPress.Wyer, R. S., Jr., & Gordon, S. E. (1984). The cognitive represen tation ofsocial information. In R. S. Wyer, Jr., & T. K. Srull (Eds.), Handbookof social cognition (Vol. 2, pp. 73-1 50). Hillsdale, NJ: Erlbaum.Received November 4,1991Revision received April 4,1992Accepted April 4,1992

future event schemas and certainty about the future

Documents