JOU R N AL OF

PersonalitySocial Psychology

Volume 36 January 1978 Number 1

Affect, Accessibility of Material in Memory, and Behavior:A Cognitive Loop?

Alice M. IsenUniversity of Maryland

Baltimore County

Margaret ClarkUniversity of Maryland

College Park

Thomas E. ShalkerUniversity of Maryland

Baltimore County

Lynn KarpBrandeis University

Two studies investigated the effect of good mood on cognitive processes. In thefirst study, conducted in a shopping mall, a positive feeling state was inducedby giving subjects a free gift, and good mood, thus induced, was found to im-prove subjects' evaluations of the performance and service records of productsthey owned. In the second study, in which affect was induced by having sub-jects win or lose a computer game in a laboratory setting, subjects who hadwon the game were found to be better able to recall positive material in mem-ory. The results of the two studies are discussed in terms of the effect thatfeelings have on accessibility of cognitions. In addition, the nature of affect andthe relationship between good mood and behavior (such as helping) are dis-cussed in terms of this proposed cognitive process.

Recent work on the relationship betweengood mood and helping has begun to focus onthe nature of that relationship—on why and

The authors wish to thank Fred Polner for his as-sistance in conducting Study 1 and Aron Siegmanand Tom Trabasso for their comments on the manu-script.

Margaret Clark is now at Carnegie-Mellon Uni-versity.

Requests for reprints should be sent to Alice M.Isen, Department of Psychology, University of Mary-land Baltimore County, Baltimore, Maryland 21228.

by what process being in a good mood shouldlead one to do kind things for others (Isen,Clark, & Schwartz, 1976; Levin & Isen, 1975;Isen, Note 1). In this context, it has beenproposed that the relationship between feel-ing good and helping is cognitively mediated,mediated by mood-dependent changes in theinformation accessible to a person who isfeeling good (Isen, Note 1) .

In accord with the cognitive social-learn-

ing approach, which stresses the role of ex-pectation in governing behavior (Mischel,

Copyright 1978 by the American Psychological Association, Inc. All rights of reproduction in any form reserved.

1

1SEN, SHALKKR, CLARK, AND KARP

1973), several authors have suggested thathelping is the result of a decision-makingprocess in which the person considers thecost-reward matrix associated with helping ornot helping in the particular case at hand(Latane & J)arley, 1970; Piliavin, Rodin, &1'iliavin, 1969). I t has subsequently been pro-posed that mood plays a role in this processby affecting the person's perception or weight-ing of those costs and rewards (Levin & Jsen,197S; I sen, Note 1). There are several waysin which this might occur. One, for example,is as follows: When a person is confrontedwith a situation in which he or she can help,presumably, cognitions concerning both theadvantages and disadvantages of helping areavailable in memory from past experience insimilar situations. These advantages and dis-advantages, however, may not all be equallyaccessible or retrievable to the person at themoment, and thus, they may not all come tomind. What we are suggesting is that moodplays a role in what comes to mind.

The Accessibility and "Cognitive, Loop"Hypotheses

We propose that a person in a good moodis more likely to retrieve positive than nega-tive material from memory and that this im-proved access to positive material affects thedecision-making process with regard to be-havior (especially behavior such as helping).Thus, we are suggesting that thoughts as-sociated with or responsible for the goodmood or mood-inducing event may serve tocue other positive material in memory in muchthe same way that a category name has beenfound to cue material of that category thatwas previously learned (Tulving & Pearlstone,1966). In one study, Tulving & Pearlstonefound that subjects could recall more words(for example, "engineer," "lawyer") from alist of learned words when presented with thename of the category ( for example, "profes-sions'') appropriate to those words, thanwhen not given this cue. The authors inter-preted this finding as evidence for the differ-ence between availability of material in mem-ory storage and accessibility of that same ma-terial under particular conditions of recall.Their study illustrates the way in which a

single cue during recall can affect the amountand content of the material recalled, andthey point out that more powerful cues mightproduce an even greater facilitation of recall(Tulving & Pearlstone, 1966, p. 389). Goodmood, we suggest, at least in certain contexts,serves as such a cue, facilitating the recall ofpositive material, thereby affecting one's esti-mate or expectation with regard to the be-havior under consideration, and thus influenc-ing the decision-making process with regardto a behavior such as helping.

One might say, then, that a person who isin a good mood may tend to be in a cogni-tive "loop" that is characterized by accessibil-ity of positive thoughts, memories, and as-sociations. We speak of a loop because thepositive material that is accessed may, inturn, tend to keep the person thinking aboutpositive things. Good mood, pleasant materialfrom memory, and positive activities such ashelping may tend to make each other morelikely. The mood may cue positive thoughts;these may provide access to still other posi-tive material in memory and may contributeto the maintenance of the mood state; andthe consideration of a positively toned activ-ity may itself contribute to the process.1 Ifthe behavior is performed and reacted to asexpected, of course, this would provide stillanother source of continued good feeling, an-other component to the positive loop.

Evidence oj the. Positive, Cognitive Loop

If positive material is thus more salientwhen one is in a good mood and influencesthe person's decision-making process, weshould find evidence of such a cognitivechange in other opinions, expectations andactions that persons express when they arein a good mood. Not only might they be morehelpful than at another time, but they shouldalso, as the saying goes, see the brighter sideof things; they should be generally more op-timistic, and they should act accordingly.

1 The type of behavior under consideration by theelated person, as well as the extent of the person'sexperience with such behavior, may serve as mod-erators of the relationship between mood and be-havior. Negative activities, for example, may not bemade more likely by good mood.

A COGNITIVE LOOP?

There is evidence that, in fact, persons ina good mood do tend to see the brighter sideof things and respond more positively thanothers to a variety of stimuli. Work by sev-eral authors indicates that persons who suc-ceed tend to expect future success and to re-ward themselves more generously than doothers (e.g., Feather, 1966; Mischel, Coates,& Raskoff, 1968). Children who are inducedto feel happy also have been found to reportan expectation of other kinds of positiveevents, such as free gifts (Masters & Fur-man, 1976). Postman and Brown (19S2)found lower tachistoscopic thresholds for suc-cess-related words following success (and forfailure-related words following failure), anda recent study has reported that relative toothers, successful subjects tend to look atpositive, rather than negative, self-relevantinformation (Mischel, Ebbesen, & Zeiss,1973). Success and other experiences thatmight make one feel happy have been foundto have interpersonal effects as well. As men-tioned earlier, those who have been exposedto any of a wide variety of good-mood-induc-ing experiences are more likely to help others(Aderman, 1972; Aderman & Berkowitz,1970; Isen, 1970; Isen et al., 1976; Isen &Levin, J 9 7 2 ; Moore, Underwood, & Rosen-han, 1973), and, as might be expected fromthis good mood-helping relationship, suchsubjects also report greater liking for astranger (Clore, 1975; Clore & Byrne, 1974;Gouaux, 1971; Veitch & Griffitt, 1976).Finally, a paper currently in preparationprovides some additional details about the ef-fect of mood on ratings. It reports that rat-ings of positive, negative, and ambiguousslides can be affected by procedures designedto induce mood, but that at the same time,subjects retain the ability to differentiate be-tween the slide types (Isen & Shalker, Note2). Subjects noticed the details of the slidesand responded appropriately, even thoughtheir mood states tended to affect their gen-eral level of rating.

Thus, there is evidence that a person whohas reason to be in a good mood, although heor she does not distort material grossly, doestend to estimate situations more positivelythan at another time, and to behave accord-ingly. Generally, theoretical models proposed

to account for this relationship between posi-tive affect and evaluations have stressed thereinforcement or reward value of the affect-inducing event and the extension of this re-inforcement value to surrounding stimuli (e.g.,Byrne & Clore, 1970; Lott & Lott, 1974;Masters & Furman, 1976). These same data,however, arc also compatible with the acces-sibility and cognitive loop hypotheses that weare suggesting, but several other aspects ofour formulation need to be tested. The stud-ies presented in this paper attempt to examinethe most basic of the many questions thatremain about the cognitive processes in-volved in the effects of mood state on judg-ment and behavior.

The Present Studies

Specifically, we begin by asking whetherjudgments made about familiar items, frommemory, will be affected by feeling state, ashas been found for those made about un-familiar items with which the person is con-fronted. This is crucial for the accessibilityinterpretation of such improved judgments.Second, if good mood is serving as a cue bywhich positive material in memory is ac-cessed, as proposed, then we should find thatmemory for positive material is enhancedwhen the person "feels good," and in Study 2we ask this question directly. We leave tofuture papers specification of the types of be-havior, including the types of helping, thatmight be more likely when one is in a goodmood.

Study 1

In the first study, conducted at a shoppingmall, subjects were unaware that they wereparticipating in an experiment but believedthat they were part of a "consumer opinionsurvey." The study examined their judgmentsof the performance and service records oftheir automobiles and television sets, as afunction of a positive mood-inducing event.

Method

Subjects

Men and women who were walking alone in asuburban mall between 6:30 and 9:30 p.m. week-

1SEN, SHALKER, CLARK, AND KARP

Table 1Mean Product Ratings by Subjects in IlacliCondition und Proportion of Persons in 1'MchCondition Who Agreed to Participate WhenApproached (Study I)

Measure

Mean

ProportionPercentage

Receivedgift

6.01

25/3181

Did notreceive gift

5.13

20/4367

nights, or between 10:00 a.m. and 3:00 p.m. week-days, served as subjects in this experiment. All pass-ersby who approached the confederate when theexperimenter was free became subjects in this study,

Procedure

Subjects were randomly assigned to one of twoconditions by a female confederate who then handedout notepads (to females) or nail clippers (to males)to those subjects assigned to the experimental group.-The confederate posed as a representative of thecompany that made the notepad or nail clipper andexplained that the sample was free and was meantto introduce the subject to the company's product.Both objects, priced at 29 cents in a nearby store,were small enough to fit easily into a pocket orhandbag. Subjects in the control condition were notapproached by the confederate.

When a subject (experimental or control) hadpassed the confederate, the latter turned around andthus signaled to a coordinator, standing, about 50yards (46 m) away, with a female experimenter whowas waiting to approach subjects with a "consumersurvey" questionnaire. The coordinator stood facingthe confederate, while the experimenter stood withher back to the action, so that she would not knowa subject's condition when later interacting with himor her. At the confederate's signal, the coordinatordescribed the subject to the experimenter, whereuponthe experimenter, not knowing whether the personhad received a free gift or not, approached the desig-nated subject. She introduced herself, explained thatshe was taking a consumer opinion survey, and askedwhether the subject would be willing to answer afew questions. If a subject agreed to participate, theexperimenter presented the questionnaire orally, ask-ing subjects the make, model, year, and age of theirautomobiles (if any) and then their television sets(if any). The final two questions with regard toeach possession were requests to rale, on a scale of1 (low) to 7 (high), the present performance of theitem and its overall service record. The dependentmeasure, then, for each subject was his/her averagescore on these four questions.

Two potential subjects were eliminated from thestudy because they refused the free sample; one was

eliminated because she was carrying her notepad inplain view of the experimenter when the experi-menter approached her with the survey; and two(one in each condition) were not approached by theexperimenter because the coordinator saw that theyhad interacted with someone after leaving/passingthe confederate.

Results

Table 1 presents the mean product per-formance and service record ratings by condi-tion, along with the proportions of subjectsagreeing to participate in the survey. A two-way analysis of variance of the rating data re-vealed a significant effect of condition, F(l,50) = 8.22, p = .006, but no effect due tosex, F(\, 50) < I , p> .1, nor to the inter-action between sex and condition, F(l, 50) —2.S1, p > .1. Subjects who had received afree gift gave higher ratings of the perform-ance and service records of products theyowned than did subjects who had not re-ceived a free gift.

A chi-square analysis of the proportion ofsubjects agreeing to participate indicated thatdespite a tendency for experimental subjectsto be more likely to participate, these per-centages were not significantly different inthe two conditions, x"(T) = 1.59, p > .1.Thus, there is no indication that the ratingdata were influenced by self-selected par-ticipation on the part of the subjects.

Discussion

The experimental conditions did not differsignificantly in number of subjects agreeingto participate. This may seem surprising ifone views the inquiry about the consumersurvey as a request for help and therefore thesubject's agreement to participate as a help-ing response. On the other hand, we did notexpect the experimental manipulation to af-fect subjects' willingness to take part in thesurvey, because of the circumstances of therequest and because of the nature of the task:

- A pilot study in which notepads were distributedto both men and women in the experimental condi-tion had revealed that although notepads appeared tobe effective in inducing good mood among women,they did not seem to be effective among male sub-jects.

A COGNITIVE LOOP?

Subjects were directly approached and, in aface-to-face encounter, asked to participatein a task that seemed harmless, interesting,and even beneficial to the subjects themselves,who were consumers. Under these circum-stances of relatively high motivation to par-ticipate and low reason to refuse participa-tion, we expected any effect of the free gift tobe obscured by a high overall rate of par-ticipation. (And, in fact, the rate of 67% inthe control group is higher than that usuallyobserved when control subjects are given theopportunity to help an experimenter in simi-lar field settings.)

The results of the study indicate that in afield setting, as has been found in laboratorysettings, subjects who have reason to be ingood moods make more positive judgmentsthan do members of a control group. More-over, this effect holds despite the fact thatthe items are rated from memory and arevery familiar to the subjects. Thus, we haveevidence that as predicted by our hypothesis,persons who are likely to be feeling good havea more positive outlook—they do seem to"see the brighter side of things," and evalua-tions made from memory reflect this positivebias. Moreover, this optimism seems to affectsubjects' decision-making processes, at leastas these are reflected in their evaluations oftheir consumer goods for purposes of con-sumer action.

Thus, in conjunction with the diverse lit-erature indicating a positive inclination fol-lowing procedures designed to induce goodfeeling, we have demonstrated that personswho have undergone such procedures arelikely to interpret situations more positivelythan they would at other times and to actaccordingly. Our formulation goes on to sug-gest that this process, as well as other mood-related behavior such as the observed rela-tionship between good mood and helping, isa by-product of an increased accessibility forpositive material during the good-mood state.Study 2 examines this question directly, mea-suring the accessibility of various types ofmaterial as a function of the subject's ex-posure to the mood-related events of victoryand defeat.

Study 2

Method

Subjects

Subjects were 47 male and female students in anintroductory psychology class.

Stimuli

Eighteen words of equal frequency of usage wereselected from a list of SSS personality trait wordsthat have been rated for likeablencss by Anderson(1968) ; 6 words were selected from among the 100most-liked words; 6 from among the 100 least-likedwords; and 6 from among words ranked between227 and 327. In addition, 18 neutral, nontrait wordswere selected from a list of moderate frequencywords compiled by Tiattig and Montague (1969).These 36 words were then recorded on a cassettetape to he played to subjects. The order of presenta-tion of the words was as follows: 6 of the nontraitwords taken from Battig and Montague (1969)were placed at the beginning of the list, and 6 at theend of the list. The order of the remaining 24 wordswas counterbalanced, between positions 7 and 30,in the 36-word list.

Procedure

Subjects were seated at a teletype and were readthe following instructions by a male experimenter:

There are several components to this learningstudy. The first thing I would like you to do isto play a computer game called "Star Trek." You,in the Starship "Enterprise," will he in a battlewith a Klingon Battle Cruiser. You have sixmoves that you can make: (1) fire forwardphasers, (2) fire rear phasers, (3) fire forwardphoton torpedoes, (4) fire rear photon torpedoes,(5) decrease range, (6) increase range.

You move by pressing the appropriate numberof the teletype and then pushing "return." Thecomputer will then tell you the results of yourmove and what the Klingon did. Then it will giveyou the next situation.

The experimenter then started the game and advisedthe subject on the first move:

The way you tell whether to use your forwardor rear weapons is by the vector heading given toyou by the computer. If it is 0-90 degrees, thenyou want to use your forward weapons. If it is90.1-480 degrees, then you want to use the rear.Phasers can only be used at a range of 0-300,000km; photon torpedoes can only be used between200,000-600,000 km. The computer gives you thevector heading and range prior to each move. Theonly time it is to your advantage to increase or

JSK.N, SHALKKR, CI.ARK, AM) KARP

PleasantNeutralUnpleasantNontra i t

Success at Time 2

Success Failureat Time 1 at Time 1 Combined

1.57.0-1

1.07

1.00

Failure at Time 2

Success Failureat Time 1 at Time 1 Combined

1 .36.36.70.03

1.46

.50

.03

.96

.87

.44

.78

.80

.50

.291.00

1.19

.68

.37

.801.05

decrease the range is (1) if your phascrs arc de-stroyed and yon have to increase the range to useyour photon torpedoes, or (2) i f the Klingon isbeyond 600,000 km, you must move closer to himto order attack.

All oi the informat ion concerning appropriatemoves in various situations was summarized on asheet above the teletype, so that subjects would notneed f u r t h e r aid from the experimenter .

The outcome of each move was determined by theexperimenter, using a table of random numbers todetermine success or fai lure. At the end of the game,a message of "Congratulations: Klingon vessel de-stroyed" was typed back to the subject if five ormore hits were made on the Klingon vessel. Jf theKnterprise was hit five times, however, the message"Enterprise destroyed" was returned.

After the subject had completed his/her game andhad received feedback as to victory or defeat, the ex-per imenter asked the subject lo listen to the 36-wordtape (described above) and to try to remember asmany words as possible. Following this, subjectsagain played "Star Trek" and either won or lost thegame. They were then asked to recall the wordswhich they had learned after the first game andwere given S minutes to write down as many wordsas they could remember.

Thus, the study was composed of f o u r groups ofrandomly assigned subjects: those who succeededtwice, those who failed twice, those who succeededfirst but later failed, and those who failed first butsucceeded on their second game. Learning of theword list occurred after the first game; recall of thelist was solicited after the second game. Thus, assum-ing that outcome in the game was related to sub-jects' moods, we could see the relative importance(for memory for various types of words) of moodat time of encoding, of mood at time of retrieval,and of the congrucncy of encoding and retrievalmoods (the "stale-dependent learning" ;) paradigm).

Results and Discussion

Table 2 presents the mean number of wordsof each type recalled in each condition. A3-way mixed analysis of variance (Victory or

Defeat at Time I X Victory or Defeat atTime 2 X Word Typo) revealed that whilerecall did not appear to be affected by out-come at time of learning, F(l, 44) — .49, p> ..SI, it did seem to be affected by victoryor defeat at time of recall, /''(I, 44) = 3.78,p < .055. There was no significant interactionbetween outcome at Time 1 and outcome atTime 2. (Such interaction would have beenin keeping with a state-dependent learninghypothesis.) In addition, there was a signifi-cant effect of word type, f f ( 3 , 176) ~ 159.85,p < .0001, and a significant interaction be-tween word type and mood at time of recall,7<'(3, 176) -3.58, p < .008. Subjects whohad experienced a positive outcome at thetime of recall recalled significantly morepositive words.

These findings can be interpreted to indi-cate that recall seems to be affected by good-mood state at the time of recall. Specifically,those subjects who had been victorious re-called more positive words, but not more ofthe other words, than subjects who had beendefeated. These findings fit our accessibilityhypothesis in two ways. First, since there wasno recall difference due to the outcome ofthe first game, when words were learned, but

a significant difference attributable to theoutcome of the second game, we can say that

11 The "siate-clcpendcnt-learning" effect refers tothe tendency for material learned when a subject isin a specific state (say, mania or alcoholic intoxica-tion) to be better recalled when he or she is again inthat state than at another t ime (Henry, Wein-gartner, & Murphy, 1973; Weingartner & Faillace,1971). This effect would be compatible with, thoughnot necessarily implied by, Tulving and Thomson's(1973) encoding specificity hypothesis as well.

A COGNITIVE LOOP? 7

the recall difference seems to be a function ofthe retrieval process, rather than of factorsassociated with initial attention or learning.Differences in recall did not appear related tovictory or defeat at the time that the list waslearned. Second, since victorious subjects dif-fered from defeated subjects only in memoryfor positive words, it is reasonable to suggestthat the difference is attributable to in-creased accessibility for positive materialwhen in a good mood.

It should be noted that our data do notsupport a similar interpretation of the effectof negative mood on negative material inmemory. Negative mood is probably a morecomplex event than positive mood, however,and an expectation of symmetry between thetwo may be unwarranted. We will have moreto say on this below. It is worth pointing outhere, however, that Study 2 did not contain acontrol group that neither succeeded norfailed at time of list learning and list recall.

We did not observe a state-dependentlearning effect: Victory or defeat at time oflist learning did not seem to interact withoutcome of the game at time of recall to pro-duce better memory when conditions werecongruent. Apparently, the material to belearned was encoded according to its mood-related semantic content but was not affectedby the mood state likely to have been experi-enced by a subject at time of encoding. Thus,the good-mood state at time of recall was ableto cue the positive material regardless of themood condition that had prevailed during itsencoding. This may have been the result ofour instructions to the subjects. Subjects inthis study had been instructed to rememberas many of the words as possible, and it hasbeen noted that subjects in such learning ex-periments most often utilize the semanticmeaning of the words in developing strategiesof memorization (Tulving & Thomson, 1973).Thus, in this instance, the subject's moodstate at the time of encoding may not havecontributed as much as did the semanticmeaning of the words to the overall way inwhich the stimulus words were experiencedand therefore encoded. This would explainwhy good-mood-related words could then becued by a good-mood state at the time of re-trieval, independent of the subject's mood at

the time of encoding. It also suggests that be-fore drawing any conclusions with regard tomood-state-dependent learning on the basisof Study 2, an experiment tailored specificallyto that issue should be undertaken.

General Discussion

The results of these studies indicate thatpeople who have been exposed to proceduresdesigned to induce a good mood are moregenerally positive than control subjects, evenin a realistic situation where their judgmentsof specific, familiar items are called for. Inaddition, there is evidence that exposure tosuch procedures results in improved recall ofpositive material, while not affecting recall ofother material. We interpret these findings asevidence for the accessibility hypothesis, pro-posing that mood state serves as a cue bywhich positive material in memory is accessedand that this material then plays a role in in-fluencing the person's decision-making pro-cess and, ultimately, behavior.

The, Positive Loop and the Processingof Negative Stimuli

Let us consider, for a moment, what hap-pens when a person in a good mood is con-fronted with a negative stimulus. We alreadyknow that he or she will be primed for posi-tive material and reactions, but that this pro-cess is limited (Tsen & Shalker, Note 2).Thus, we would expect that neither extremelynegative stimuli nor potentially harmful be-havior would be distorted into positives. This,then, helps us to understand why people'smoods do change. We would expect thatfocusing on clearly negative stimuli wouldresult, not in distortion of the negative situa-tion, but in mood change. The negative stimu-lus event, we would hypothesize, would ac-cess negative cognitions incompatible with thepositive ones associated with good mood, andthus would result in the termination of themood.

Thinking about incompatible material, whichinvolves shifting focus, is probably cogni-tively more difficult than processing informa-tion that is compatible with the existing cog-nitions or mood state (Naus, 1974), and forthis reason, we would expect that only the

8 ISEN, SHALKKR, CLARK, AND KARP

most urgent incompatible stimuli would beattended to. Nonessential incompatible stim-uli, we suggest, might be set aside [or themoment. Thus, we would expect that if aperson in a good mood were given the op-portunity to think about material that wasclearly incompatible with his or her good-mood state, the person might be less likely todo so than would one in whom good mood(the positive cognitive loop) had not been in-duced. The result of a recent study by Mischelct al. (1973), where subjects who had suc-ceeded chose to look at (think about) positivematerial about themselves, lends itself to thisinterpretation, as well as to the reasonablemood-maintenance interpretation suggestedby those authors and others (Masters & Fur-man, 1976; Isen, Note 1 ) .

A more recent study by Mischel, F.bbesen,and Zeiss (1976) is also compatible withboth the mood-maintenance interpretation andthe information-processing hypothesis outlinedabove. That study found that persons whohad been exposed to positive-expectation-in-ducing feedback were less able to recognizeinformation about their negative qualitiesthan about their positive qualities in a testof memory. Although the study found no ef-fect on positive material, this should not beinterpreted as evidence contradicting our ac-cessibility hypothesis, since Mischel et al.used a recognition test of memory, ratherthan a free-recall test. The accessibility hy-pothesis suggests that good mood serves as aretrieval in an otherwise free-recall situation;for this reason, it cannot be tested in the con-text of a recognition test of memory wherethe recognition test itself provides subjectswith the cues necessary to retrieve materialstored in memory.

In explaining their finding of apparent re-duced memory for negative material amonggood-mood subjects, Mischel ct al. (1976)suggested that these subjects may have spentless time and effort learning and rehearsingthe negative self-relevant information. Thisinterpretation seems especially likely, sincesubjects in that study were not asked tomemorize the material and were presentedwith the negative (incompatible) materialwhile they might still have been rehearsingor enjoying their positive-expectation-induc-

ing communication. (It has been suggestedthat the first few minutes following induc-tion of a good mood may be used by subjectsto rehearse or think about the mood-inducingevent (Isen et al., 1976).) This observationwould imply that a slight change in the situa-tion, such as a change in timing, or the pro-vision of an incentive for studying and learn-ing the negative self-relevant material, mightresult in success subjects attending to thenegative material and being able to rememberit. Thus, it is interesting that in our Study 2,where subjects were asked to try to remem-ber as many of the words as possible, success-ful subjects were not found to differ fromothers in memory for negative items. Such aneffect of incentive, for example, would cer-tainly lend support to the contention ofMischel et al. (1976) that failure to remem-ber the negative material was not due to anyirrepressible unconscious defense mechanismbut rather to a subject's decision, under a par-ticular set of circumstances, to attend to orrehearse certain material in preference toother material.

The Positive Loop and Behavior

In summary, we are suggesting that withinlimits, persons in a good mood will tend tothink about positive events or cognitions andthat their thoughts, feelings, or estimatesabout these cognitions will tend to be morepositive than they might be at another time.Behavior, too, is proposed as a component ofthis cognitive loop. Certain behavior will be-come more likely when one is feeling good,and it, in turn, will affect (through both itsassociations and its consequences) the per-son's mood state and cognitive processes.

As has been found for evaluation of stimuli,however, potentially negative or harmful be-havior should not be grossly misperceived:Those behaviors that are compatible (i.e.,have positive cognitions available) with agood-mood state should be more likely whenone is in a good mood than they might be atanother time, and those behaviors that en-danger the mood state or that require think-ing about material incompatible with thepositive cognitions of the good-mood stateshould be less likely than at another time.

A COGNITIVE LOOP?

Helping, because it usually has positive ele-ments associated with it, should become morelikely during good mood; murder, except inunusual cases, should not. Even for helping,moreover, this should be true only if the help-ing can be associated with positive (mood-congruent) events. Not all kinds of helpingshould be more likely when one is in a goodmood (if a helpful act were clearly portrayedas involving the performance of an unpleasanttask or as incompatible with the good moodthat had been induced, we would expect thatit would be avoided by good-mood subjects).There is some support for this suggestion inthe results of a study by Tsen and Levin(1972), in which, relative to a control group,good-mood subjects were found to be morewilling to help, but less willing to annoy, an-other college student. Thus, in summary, wehave evidence that within limits, being in agood mood affects the material that peopleremember and their evaluation of their presentexperiences. We know that it affects their will-ingness to help others, and although the linkremains to be established, this behavioral ef-fect may be mediated by the cognitive changesobserved here.

Feeling States as Cognitive Categories

The process described and proposed in thispaper to understand the effects of mood onjudgment and behavior is based on principlesof information processing, and it also sug-gests something more general about the wayin which information is processed. Our re-sults indicate that mood state may serve asan effective retrieval cue. Given that the effec-tiveness of a retrieval cue may depend on theway in which the material was originally en-coded, as suggested by the encoding specificityhypothesis (Tulving & Thomson, 1973), thefact that mood can serve as an effective re-trieval cue implies that the affect generatedby, denoted by, or otherwise associated witha cognitive element may be one of the fac-tors that influence the way in which it is en-coded and stored in memory. This, of course,would not preclude cognitions being storedand accessible according to other appropri-ate characteristics as well (their being cross-filed, so to speak). But it does accord with the

contextualist position, expressed by Jenkins(1974), that people respond to or rememberthe quality of what they have experienced,rather than the individual stimuli. This sug-gests that memory for events and stimuli maybe affected by stimulus characteristics andaspects of the learning and remembering situ-ations that are not usually considered in stud-ies of cognitive processes.

Negative Moods

It should be noted that although our cogni-tive formulation theoretically might apply toboth positive and negative mood states, wehave focused only on positive mood. Evenwhere we have discussed the presentation ofnegative stimuli to persons who were feelinggood or the hypothetical consideration ofnegatively toned behavior by persons in goodmoods, we have considered these only up tothe point at which we think they would re-sult in mood change. We have not spokenabout, the cognitive activities of persons inbad moods, and some consideration of thistopic would be appropriate.

We would expect that persons who havebeen defeated, who have failed, or who aresad might indeed have more ready access tonegative material in memory, just as thosewho are happy are proposed to have betteraccess to positive. And there is some evidencefor this suggestion. Postman and Brown(1952), for example, found not only lowertachistoscopic thresholds for success-relatedwords among those who had succeeded, butalso lower tachistoscopic thresholds for fail-ure-related material among those who hadfailed. Likewise, in a slide-rating task de-scribed earlier (Isen & Shalker, Note 2), fail-ure was observed to reduce subjects' estimatesof the pleasantness of slides, just as goodfortune was seen to raise them.

On the other hand, the data of Study 2 donot support such a hypothesis with regard todefeat and negative material in memory, andmany other studies seem to suggest that suc-cess and failure, good and bad mood, are notsymmetrical in their effects. Mischel et al.(1976) make the point that it has repeatedlybeen found that success increases a person'sinclination toward the positive, relative to a

10 ISEN, SHALKER, CLARK, AND KARP

control group, while failure seems to have noeffect. Mischel et al. (1968), for example,found that successful subjects were more gen-erous with themselves than were control sub-jects but that failures were not less self-gen-erous; Masters and Furman (1976) foundthat positive affect induced expectation of fu-ture positive events but that negative affectdid not generate the opposite; Isen (1970)found that successful subjects were more help-ful than controls but that failure subjects didnot differ from the control group; and Sieg-man (in press) has noted similar findings inseveral studies in the area of interpersonal at-traction. Thus, although negative affectivestates may generate cognitive processes paral-lel to those generated by positive affect, thesestates or cognitions do not appear to havethe corresponding effects on behavior.

One interpretation of this incongruity isthat while a negatively oriented cognitiveprocess parallel to the one proposed for posi-tive states may be induced by the negativestate, the "negative cognitive loop" that itwould create if left unchecked would serve tomaintain the negative mood state. Mainte-nance of a negative mood state is probablynot a goal of most people. Thus, although theunhappy person may find negative thoughtsaccessible and positive cognitions more alienat the moment, such a person may also belikely to try to counter this process and changehis or her situation. Just as people are pro-posed to try to maintain their good-moodstates, it is reasonable to suppose that theymay try to break a negative mood loop andimprove their spirits. In fact, Cialdini, Darby,and Vincent (1973) suggest just this to ex-plain their finding that persons in bad moodssometimes help more than control subjects.I f helping is part of a positive loop, generat-ing positive associations and expectations (aswe propose), then doing something helpfulmight indeed be a good way to raise one'sown spirits. The cognitive or behavioral ac-tivities that a person chooses for improvingmood and his or her degree of success atmood improvement probably depend on sev-eral factors having to do with the particularsof the situation and the person's own historyof dealing with such problems. But in anycase, we are suggesting that two factors may

operate to account for the behavior of per-sons who arc feeling bad: Such persons may,in fact, have more easy access to negativematerial in memory, just as good-mood sub-jects have greater access to positive, but theymay actively try to counter this tendency,while the latter have no reason to do so.

An alternative interpretation of the appar-ent asymmetry between positive and negativemoods is based on the suggestion that the twostates in fact may not be symmetrical—thatthe cognitive material supporting the twotypes of states may differ in some importantway. That is, there might be a larger or moredelineated or better interconnected networkof positive than negative associations, suchthat access to the positive material may bemore likely to have behavioral implicationsthan access to the negative network. It isalso interesting to note that persons maydiffer in the extent to which such a distinctionis true for them.

These cognitive processes may not only in-fluence social interaction, as described in thispaper, but they may apply as well to clinicalphenomena. Several authors and practitionershave adopted a cognitive approach to clini-cal problems such as depression, pointingout, as we would agree, that moods are notwholly affective but that they are associatedwith cognitions and behaviors (Beck, 1967;Becker, 1977; Kelly, 1955; Seligman, 1975).These authors stress the role that beliefs, at-titudes, and other cognitions can play in thedevelopment of depression, and some suggestdirect intervention in the cognitive process ofevaluation and interpretation, which, theypropose, results in the depressed mood (Beck,1967; 1976). Our data may shed additionallight on the cognitive implications of moodstates, once generated. The oft-noted clinicalobservation that depressed patients do notseem to see positive aspects of even normallypositive situations may reflect, in fact, thepositive material's relative inaccessibility un-der those conditions of recall (namely, de-pression). This interpretation suggests thatone important aspect of therapist interven-tion in the cognitive process may be to cuethe positive material that the depressed per-son has difficulty in recalling. In addition, itpoints to the circular nature of the "mood

A COGNITIVE LOOP? 11

loop" and implies that direct intervention inthe mood state itself might facilitate anyother attempted intervention, such as thatvia the supporting cognitions. One final pointto be made is that, as suggested above, in-dividuals may differ with regard to these cog-nitive processes. They may vary in the extentto which they use affect to categorize ma-terial, and there may be individual differencesin positive or negative networks. Thus, ex-ploration of a depressed individual's cogni-tive network or cognitive habits might be use-ful to the therapist in these endeavors. Like-wise, a consideration of the situations inwhich certain kinds of affect-related processesare generated might be helpful.

Affect and Cognition

The results of our studies indicate thatwhat we commonly call mood and most oftenthink of as an emotional state has a cognitivedimension and may even be more appropri-ately conceptualized as a cognitive state. Inaddition to having previously observed a re-lationship between good mood and behavior(specifically, helping), we have now observeda relationship between good mood and positiveinterpretation of st imuli , between good moodand satisfaction with possessions, and betweengood mood and memory for positive material.One may conceptualize all of this, as sug-gested in the introduction, as a relationshipbetween good mood and behavior that ismediated by cognitive processes or changes.On the other hand, one should consider thepossibility that mood itself consists in, or isa product of, those very cognitive changes.That is, a positive event may cue other posi-tive material in memory, and this may estab-lish both the cognitive loop that we have pro-posed as the mediator and the phcnomono-logically experienced emotional state of feel-ing good. Work by Schachtcr and Singer(1962) introduced the general idea, morethan a decade ago, that the mood experiencedmay be determined by the label applied to itor the label most appropriate in the situa-tion; and many studies have followed thislead. Still, more work is needed to provide athorough understanding of the relationshipbetween mood and cognition, and we hope

that the growing interaction between the sub-fields of psychology involved in these issues(cognitive, social, and clinical psychology inparticular) will continue. On the basis of ourstudies, we cannot say whether mood can beidentified separately from cognitive processesnor, if so, which precedes the other; what wecan say, however, is that the processes arenot independent of each other—that what isexperienced as emotion and what is seen ascognition are, if not identities, very much re-lated and interactive.

Reference Notes

1. Isen, A. M. Positive affec t , accessibility of cogni-tions and helping. In J. Piliavin (Chair) , Currentdirections in theory on helping behavior. Sym-posium presented at the meeting of the EasternPsychological Association, New York, 1975.

2. Isen, A. M., & Shalkcr, T. E. Do you "accentuatethe positive, eliminate the. negative" when yon arein a good mood? Unpublished manuscript, Uni-versity of Maryland Baltimore County, 1977.

References

Aderman, D. Elation, depression, and helping be-havior. Journal oj Personality and Social Psychol-ogy, 1972, 24, 91-101.

Aderman, D., & Berkowitz, L. Observational set,empathy, and helping. Journal oj Personality andSocial Psychology, 1970, 14, 141-148.

Anderson, N. II. Likableness ratings of 555 personal-i ty-trai t words. Journal oj Personality and SocialPsychology, 1968, 0, 272 279.

Baltig, W. 1''., & Montague, W. E. Category normsfor verbal items in 56 categories: A replication andextension of the Connecticut category norms. Jour-nal of Experimental Psychology Monograph, 1969,#0(3, pi. 2 ) .

Beck, A. T. Depression: Clinical, experimental andtheoretical aspects. New York: Harper & Row,1967.

Beck, A. T. Cognitive therapy and the emotionaldisorders. New York: International UniversitiesPress, 1976.

Becker, J. Affective disorders. Morristovvn, N.J.:General Learning Press, 1977.

Byrne, D., & ("lore, G. L. Reinforcement model ofevaluative responses. Personality: An InternationalJournal, 1970, 1, 103-12,5.

Cialdini, R. B., Darby, B. L., & Vincent, J. E. Trans-gression and altruism: A case for hedonism. Jour-nal oj Experimental Social Psychology, 1973, 9,502-516.

Clore, G. L. Interpersonal attraction: An overview.Morrislown, X.J.: General Learning Press, 1975.

Clore, G. L., & Byrne, I). A reinforcement-affectmodel of attraction. In T. L. Huston (Ed.),

12 1SEN, SIIALKER, CLARK, AND KARP

Foundations of interpersonal attraction. NewYork: Academic Press, 1974.

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

Gouaux, C. Induced affective states and interper-sonal attraction. Journal of Personality and SocialPsychology, 1971, 20, 37-43.

Henry, G. M., Wcingartner, H., & Murphy, D. L.Influence of affective slates and psychoactivc drugson verbal learning and memory. American Jour-nal of Psychiatry, 1973, 130, 966-971.

Isen, A. M. Success, failure, attention, and reactionsto others: The warm glow of success. Journal ofPersonality and Social Psychology, 1970, 15, 294-301.

Isen, A. M., Clark, M., & Schwartz, M. F. Durationof the effect of good mood on helping: "Footprintson the sands of time." Journal of Personality andSocial Psychology, 1976, 34, 385-393.

Isen, A. M., & Levin, P. F. The effect of feeling goodon helping: Cookies and kindness. Journal of Per-sonality and Social Psychology, 1972, 21, 384-388.

Jenkins, J. J. Remember that old theory of memory?Well, forget it! American Psychologist, 1974, 29,785-795.

Kelly, G. The psychology of personal constructs.New York: Norton, 1955.

Latane, B., & Darley, J. M. The unresponsive by-stander: Why doesn't he help? New York: Apple -ton-Ccntury-Crofts, 1970.

Levin, P., & Isen, A. M. Further studies on the ef-fect of feeling good on helping. Sociometry, 1975,38, 141-147.

Lott, A. J., & Lott, B. 1C. On the role of reward inthe formulation of positive interpersonal atti-tudes. In T. L. Huston (Ed.), Foundations of in-terpersonal attraction. New York: Academic Press,1974.

Masters, J. C., & Furman, W. Effects of affect stateson noncontingcnt outcome expectancies and be-liefs in internal or external control. DevelopmentalPsychology, 1975, 12, 481-482.

Mischel, W. Toward a cognitive social learning rccon-ceptualization of personality. Psychological Re-view, 1973, 80, 252-283.

Mischel, W., Coatcs, B., & Raskoff, A. Effects of suc-cess and failure on self-gratification. Journal of

Personality and Social Psychology, 1968, 10, 381-390.

Mischel, W., Ebbescn, E., & Zeiss, A. Selective atten-tion to the self: Situational and dispositional de-terminants. Journal of Personality and Social Psy-chology, 1973, 27, 129 142.

Mischel, W., Ebbcsen, E., & /eiss, A. Determinantsof selective memory about the self. Journal ofConsulting and Clinical Psychology, 1976 44, 92-103.

Moore, B. S., Underwood, H., & Rosenhan, D. L.Affect and altruism. Developmental Psychology,1973, 8, 99-104.

Naus, M. J. Memory search of categorized lists: Aconsideration of alternative self-terminating searchstrategies. Journal of Experimental Psychology,1974, 102, 992-1000.

Piliavin, I. M., Rodin, ]., & Piliavin, J. A. GoodSamaritanism: An underground phenomenon?Journal of Personality and Social Psychology,1969, 13, 289-299.

Postman, L , & Brown, D. R. Perceptual consequencesof success and failure. Journal of Abnormal andSocial Psychology, 1952, 47, 213-221.

Schachter, S., & Singer, J. S. Cognitive, social andphysiological determinants of emotional state. Psy-chological Review, 1962, 69, 379-399.

Seligman, M. E. P. Helplessness. San Francisco: Free-man, 1975.

Siegman, A. W. The voice of attraction: Interper-sonal attraction and verbal behavior in the initialinterview. In H. Giles & R. St. Clair (Eds.), Lan-guage and social psychology. Oxford: Blackwell,in press.

Tulving, E., & Pcarlstone, Z. Availability versus ac-cessibility of information in memory for words.Journal of Verbal Learning and Verbal Behavior,1966, 5, 381-391.

Tulving, E., & Thomson, D. M. Encoding spe-cificity and retrieval processes in episodic memory.Psychological Review, 1973, 80, 352-373.

Veitch, R., & Griffitt, W. Good news—bad news: Af-fective and interpersonal effects. Journal oj Ap-plied Social Psychology, 1976, 6, 69-75.

Weingartncr, H., & Faillace, L. A. Alcohol state-de-pendent learning in man. Journal oj Nervous andMental Disease, 1971, 153, 395-406.

Received April 4, 1977 •