Journal of Experimental Psychology: Generalgbower/1981/selectivity...Journal of Experimental...

Journal of Experimental Psychology:General

VOL. 110, No. 4 DECEMBER 1981

Selectivity of Learning Caused by Affective States

Gordon H. Bower, Stephen G. Gilligan, and Kenneth P. MonteiroStanford University

SUMMARY

We investigated how emotional states influence learning and memory. Specifically,we asked whether people's remembering of a text varies with their emotional mood atthe time they read or recall a text, A theoretical framework is proposed that representsan emotion as a unit within a semantic network that encodes memories. It also assumesthat by spreading activation, a dominant emotion will enhance the availability ofemotion-congruent interpretations and the salience of congruent stimulus materials forlearning. To collect relevant observations, powerful moods were induced by posthyp-notic suggestions. Experiment 1 found that happy or sad readers identified with, andrecalled more facts about, a character who is in the same mood as they are. In Ex-periment 2, this selective recall by character could not be produced by inducing themood at recall after subjects had read the story in a neutral mood. In Experiment 3,subjects read a text wherein one character described many unrelated happy and sadincidents from his life. Readers were made to feel happy or sad while reading and,independently, while recalling this text. Mood during reading caused selective learningof mood-congruent incidents, but mood during recall had little effect. Experiment 4replicated with this one-character narrative the finding that inducing the mood duringrecall only produced no selective recall of its happy versus sad incidents. Experiment5 pitted the happy-sad nature of the incidents against the mood of the characternarrating them. Readers learned more mood-congruent than mood-incongruent inci-dents, but did not learn more about the mood-congruent character. Thus, rather thanidentifying exclusively with the same-mood character, subjects selectively learned what-ever affective material was congruent with their emotional state. The mood-congruityeffect is consistent with the network theory of emotion and memory. Several morespecific hypotheses were proposed. One is that mood-congruent material is more mem-orable because it elevates the intensity of the subject's feelings, whereas mood-incon-gruent material diminishes mood intensity. A second is that subjects focus on mood-congruent material in order to explain and justify their hypnotically instructed emotion.But further results did not support this attribution hypothesis. A third hypothesis isthat mood-congruent material may be more likely to remind the reader of a similarexperience, and this promotes learning.

Copyright 1981 by the American Psychological Association, Inc. 0096-3445/81/1004-0451$00.75

451

452 G. BOWER, S. GILLIGAN, AND K. MONTEIRO

How does emotion influence human learn-ing and remembering? Do the emotions fa-cilitate or hinder learning? Are there learn-ing differences among materials that arerelevant versus irrelevant to the subject'semotional state? When people are in a spe-cific emotional state, are they especially sen-sitized to, or primed in readiness for, stim-ulus material relevant to their feelings? Ifpeople show selective remembering of emo-tion-relevant material, is this effect locatedat "input" (learning), at "output" (test per-formance), or both?

The experiments reported here attempt toanswer these questions. They are part of aresearch program investigating emotionalinfluences on diverse cognitive processes thatwas reviewed in a previous article (Bower,1981). A theoretical framework was alsoadvanced to rationalize the findings. Thatframework will now be reviewed as back-ground to the current studies.

The theory supposes that memory can bemodeled by an associative network of se-mantic concepts or schemas that are usedto describe events. An event is representedin memory by a cluster of propositions (out-put by a perceptual analyzer); these propo-sitions are recorded in memory by establish-ing new associative connections amonginstances of the concepts used in the descrip-tion of the event. Such semantic networktheories of long-term memory have been in-troduced by Collins and Quillian (1972),Anderson and Bower (1973), Kintsch (1974),Rumelhart, Lindsay, and Norman (1972),Collins and Loftus (1975), Anderson (1976),and others. In such theories, the basic unitof thought is a single proposition or a prep-ositional compound (e.g., p causes q); thebasic process of thought is activation of aproposition and its concepts. The contentsof consciousness are the sensations, concepts,and propositions whose current activationlevels exceed some threshold. The activationpresumably spreads from one concept to an-other, or from one proposition to another,

This research was supported by National Institute ofMental Health Grant MH-13905 to the first author.

Requests for reprints should be sent to Gordon H.Bower, Department of Psychology, Stanford University,Stanford, California 94305.

(EMOTION)© - - - - Q (EMOTION ]~ 6

Figure I . Schematic diagram of an emotion unit con-nected to a number of its indicators and to memoriesof events that aroused that emotion. (From "Mood andMemory" by G. H. Bower, American Psychologist,1981, 36, 129-148. Copyright 1981 by the AmericanPsychological Association. Reprinted by permission.)

according to their associative linkages. Ananalogy is an electrical network with elec-trons (activation) flowing along links be-tween terminal junctions (concept nodes).

To relate this framework to emotion, thetheory supposes that

each distinct emotion such as joy, depression, or fearhas a specific node or unit in memory that collects to-gether many other aspects of the emotion that are con-nected to it by associative pointers.. . . Figure 1 showsa schematic for a small fragment of the many connec-tions to a given emotional node—say sadness for Emo-tion 3. Collected around this emotion node are its as-sociated autonomic reactions, standard role andexpressive behaviors (that is, the way we display sad-ness) and descriptions of standard evocative situationswhich when appraised lead to sadness. Also included arethe verbal labels commonly assigned to this emotionsuch as sadness, depression, and the blues. Some of thesevarious linkages are innate, while others are learned andelaborated throughout acculturation. (Bower, 1981, p.135)

Of particular relevance to memory studiesis the assumption that each emotion node islinked into memorized descriptions of lifeevents during which that emotion wasaroused. Thus, the grief one feels at a par-ent's funeral becomes associated with an in-ternal description of events at that funeral.These emotion nodes can be activated byvarious stimuli—for example, by physiolog-ical or symbolic ones. When activated abovea threshold, the emotion node transmits ex-citation to those nodes that generate theemotional response pattern. Activation alsospreads throughout the memory structures

AFFECTIVE STATES AND LEARNING 453

to which the emotion is connected, creatingsubthreshold excitation at those event nodes.Thus, if one is feeling sad and is promptedwith a weak cue like "parent," the activationfrom the emotion node may combine withthat from the weak cue so as to raise thetotal activation of a relevant event memoryabove a threshold of consciousness. Thus, thesad person may become conscious of think-ing about the funeral of a parent.

This network view of emotion and memoryhas several implications. First, it implies amood-dependent retrieval effect. Becausememories become associated with the moodprevailing during learning, their recall is im-proved by reinstating the same mood dur-ing recall testing. Supporting evidence formood-dependent retrieval was presented inBower (1981; also Bartlett & Santrock,1977; Bower, Monteiro, & Gilligan, 1978).Through hypnosis, subjects were induced tofeel happy while learning one word list andsad while learning a second list; later, whenhappy, subjects recalled the happy list betterthan the sad list; when sad, they recalled thesad list better. The network theory explainsthis by supposing that the present mood actslike on automatic cue activating the mem-ories associated to it, and these memories areeither the same as the targets to be recalled("same mood" condition) or different fromthem ("changed mood" condition). Thus,mood-dependent retrieval is hypothesized toresult from the memory-search advantageconferred by having an additional rele-vant cue.

A second implication of the network the-ory is that emotion should enhance the "sa-lience" of emotional stimuli in the environ-ment that agree with the perceiver's state.Thus, happy people should selectively en-hance pleasant events in their environment,whereas sad people should enhance sadevents. These enhancement effects shouldarise for two reasons. First, the active emo-tion node sends activation to perceptual cat-egories that are linked to it associatively sothese percepts are in readiness to be used.Similarly, schemas for interpreting actionsor social interactions as pleasant or unpleas-ant are brought into readiness by a corre-sponding emotion. These factors then oper-ate like a "cognitive set" to bias the way

perceivers interpret social messages theyhear.

Second, events that evoke pleasant eval-uations will enhance an ongoing positivemood but decrease an ongoing negativemood, and vice versa for unpleasant events.We will suppose that the amount of attentionand processing allocated to an incomingevent will be greater the greater is the emo-tional arousal (intensity) coincident withthat event. By this hypothesis, pleasantevents will receive more processing whenpeople are in a pleasant mood, and unpleas-ant events will receive more processing whenthey are in an unpleasant mood. As a result,subjects should learn to a greater degreeevents that are congruent with their currentemotion.

The experiments reported here were car-ried out to check for this predicted effect onlearning of mood-to-material congruity. Themoods investigated were happiness and sad-ness induced by hypnotic suggestions. Ourearlier research found hypnosis and post-hypnotic suggestions to be a valuable toolfor induction of different emotional states(see Bower, 1981, for a discussion of theadvantages and disadvantages of the hyp-notic procedures). The experiments manip-ulate congruity by variations in the narrativematerial the subject reads. In the initial ex-periments, the narrative describes two char-acters, one happy and one sad—their lifecircumstances, feelings, interactions. Thequestion is whether readers who are happywill identify more with the happy characterand learn more about him, and sad readerswill identify with and learn more about thesad character. In the later experiments, thenarrative describes a series of happy and sadincidents of a single character. The congru-ity question there is whether readers who arehappy learn more about the happy incidents,whereas sad readers learn more about thesad incidents.

The congruity prediction may be con-trasted with an alternative hypothesis thatsupposes that, given a choice, readers willselect material whose usual effect is to re-duce negative (unpleasant) moods and tomaintain or enhance positive (pleasant)moods. We will call this the "Pollyanna"(or "goody-goody") hypothesis. The congru-


ity and Pollyanna hypotheses both predictthat happy material will be especially salientfor happy readers; however, their predictionsfor sad readers differ. Sad readers shouldfind sad materials particularly salient, ac-cording to the mood-congruity hypothesis;but they should avoid sad materials and em-phasize or dwell on happy materials if thePollyanna hypothesis is correct. The exper-iments described below investigated theseand other hypotheses.

Experiment 1:Character Identification and Memory

We are primarily concerned here withwhether, and how, the mood of readers in-fluences their selection of which material ina narrative they find to be "interesting andmemorable." Before discussing mood ef-fects, however, let us briefly review as back-ground some of the subjective determinantsof interest and memorability of text mate-rial.

It is widely supposed that the meaningreaders extract from a story varies with theirinterests, the perspective from which theyread it, and the characters with whom theyidentify (see, e.g., Bower, 1978). Interestvalue of text elements partly determineswhat readers select and remember. Interestvalue in turn is determined by many factors,including basic dramatic themes (conflict,romance, revenge, death), information rel-evant to the welfare of oneself or loved onesor adversaries, unconventional surprises, andmyriad goals idiosyncratic to the reader(news of sports, gardening, finances). Thereader's interests often stem from constantgoals—for instance, to invest profitably inthe stock market—and information relevantto those goals will be noticed and selectedfor deeper processing. Interest often leads toexpertise (and vice versa); the expert'sknowledge of a topic is then triggered by itsmention in a text so that the material canbe elaborated and embedded into familiarschemata. For such reasons, interest-rele-vant material would be better remembered.

Readers' perspectives on the material alsoinfluence what parts of it they find interest-ing and memorable. Such a selectivity wasdemonstrated in an experiment by Pichert

and Anderson (1977); they had readers de-liberately adopt the perspective of someonereading a passage with certain interests inmind. Specifically, a passage about a boycoming home with a friend and giving thefriend a tour of his house was to be readfrom the perspective either of a prospectiveburglar of this house or a prospective buyerof it. For subjects reading from the perspec-tive of a burglar, the relevant schema for thesituation provides them with a purpose (e.g.,to rob without being caught) and a standardplan (e.g., to gain entry, snatch the valu-ables, and exit unseen), and it identifies in-formation relevant to that plan (e.g., notethe location of unlocked doors and portablevaluables). Information relevant to this per-spective should receive special attention andprocessing in memory. Similarly, subjectsreading from the perspective of the homebuyer should be more likely to notice andremember details relevant to the structuralquality of the house (e.g., leaky roofs andcracked plaster ceilings). Indeed, this wasthe case. Pichert and Anderson found thattheir subjects did recall more informationrelated to their respective points of view.

The perspective phenomenon studied byPichert and Anderson involved a frameworkexternal to the story itself. That is, the bur-glar and home buyer were not characters inthe story. Their perspective-taking instruc-tions were a veiled form of orienting instruc-tions, like those used by Rothkopf (1972)explicitly directing the reader to look for andretain prescribed information. A differentaspect of perspective in narrative compre-hension, studied by Abelson (1975), Owens,Dafoe, and Bower (Note 1), and Bower(1978), is "character identification." Read-ers were subtly induced to empathize oridentify with one of the characters in a bal-anced story they read; how that identifica-tion affected the readers' interpretation andrecall of events was then examined. Owenset al. found, for instance, that memory be-:came distorted in the direction sympatheticto the character with whom the reader iden-tified. Slip-ups and mishaps of this characterwere attributed to the difficulty of his ex-ternal conditions. In contrast, the errors ofhis adversary were attributed to the adver-sary's incompetence.


Our initial study extends this work oncharacter identification in text comprehen-sion and memory. The question at issue iswhether the emotional mood of the readerwill affect the story character he or she iden-tifies with and, consequently, affect encodingand recall of the story. Our two hypothesespredict different outcomes for this experi-ment. The mood-congruity hypothesis sup-poses that, other things being equal, thereader will identify with the storybook char-acter who is most similar to him or her onimportant dimensions and that emotionalmood will be a salient dimension of similar-ity. Thus, the happy subject will identifywith a happy character, and the sad subjectwill identify with a sad character. In con-trast, the Pollyanna hypothesis supposes thatpeople generally prefer reading and thinkingabout pleasant, enjoyable events. The pref-erence might be acute for sad readers whomay want "cheering up"; hence, they mayavoid the bad news in the text in favor of thegood news. This view predicts that all read-ers will identify more with the happy char-acter in a balanced narrative.

After our subjects read the story, theyanswered some questions to assess which oftwo characters they had identified with. Thesubjects returned the next day and while ina neutral ("normal") mood recalled every-thing they could about the text they had readthe day before. The specific issue is whethersubjects recall more facts about a story char-acter having the same mood as they hadwhen they read the story. The following de-scription of methods is slightly complex be-cause subjects were run in two experimentsin an interleaved fashion: The story recallexperiment is reported here; the word listexperiment, which investigated mood-statedependent learning, was reported as Exper-iment 2 in Bower, Monteiro, and Gilligan(1978). The studies were combined due toa shortage of funds and of highly hypnotiz-able subjects.

Method

Subjects. The subjects were 16 Stanford under-graduates from the "highly hypnotizable" pool main-tained by the Stanford Hypnosis Laboratory. They hadbeen screened as scoring 10-12 (the upper 15% of the

college population) on the Stanford Hypnotic Suscep-tibility Scale, Form C (Weitzenhoffer & Hilgard,1962), having passed the amnesia test item. Highly hyp-notizable subjects were used in order to obtain reliablystrong mood states and to increase the likelihood of post-hypnotic amnesia for the mood arousal that the subjectexperienced while reading the story. Most subjects hadserved in one or more experimental studies of hypnosis(none involving mood states or learning). The subjectswere paid upon completion of their second experimentalsession.

Materials. The story written for these experimentswas about 1,000 words long and was divisible into 121"idea units" (roughly, simple clauses) for scoring pur-poses. It is reproduced in the Appendix. The story wasabout two college men, Jack and Andre, as they met forand played a friendly afternoon game of tennis. Thestory narrates their driving to and arriving at the gym,getting dressed in their tennis clothes, warming up andplaying a game of tennis, then showering and gettingdressed in their street clothes again. Andre is very happyand is enjoying himself; he and his girlfriend are gettingalong well and he is anticipating a fabulous concert anddinner date with her that evening. He is in a buoyantunworried mood, is joking, singing, and enjoying thespring sunshine, and he wins the tennis match easily.In contrast, his friend Jack is in a particularly glum anddepressed mood on this day. His girlfriend has just jiltedhim and he broods over that loss. He is also worriedover his upcoming exams. Everything goes wrong todepress him; his car needs repairs, his gym locker jams,he breaks a shoelace, feels scorched by the sun, losesat tennis, and feels miserable. Each sentence of the storywas written to be clearly about either Andre (54 ideaunits) or Jack (57) or neither of them (10). In this way,we were able to score each recall unit as being aboutAndre, Jack, or neither.

Procedure. The subjects were assigned in randomalternation to the happy or sad reading conditions andwere tested singly. The experiment was described tothem vaguely as a study of the influence of mood onexpressive style. On Day 1, subjects were hypnotizedusing a standard relaxation eye-closure induction (seeWeitzenhoffer & Hilgard, 1962). Being very suscepti-ble, all subjects rapidly entered a "deep trance" butwere able to talk and indicate their depth of hypnosis(1-10 scale). After induction, the experimenter askedthe subjects to choose a situation that had made themintensely happy (or sad) or to make up one that woulddo so. They were to adjust the intensity of the emotionalfeeling to be intense but not overwhelming or unbear-able. After thoroughly familiarizing themselves withtheir happy or sad mood, subjects were told that afterawakening they would re-experience this same emo-tional feeling (but not the imaged situation) when theybegan to read a story that the experimenter would handthem. The emotional state was to be maintained steadilyas they read the story twice, filled out a qusetionnaireabout it, and then learned a list of 20 words (for anunrelated experiment not reported here). The mood wasto be lifted finally and they were to revert to a normalmood when the experimenter pointedly said, "That'sfine. It's over." Furthermore, the experimenter told sub-jects that the posthypnotic cues would trigger the be-ginning or end of the mood without their being aware


of the source of their mood. They would remain unawareuntil the end of the second session when the experi-menter said, "Now you can remember everything aboutthis study." At that point, the posthypnotic suggestionwas to be cancelled. We did not erase it at the end ofthe first session because we wished to rearouse the moodduring the second session for use with the word list ex-periment that occurred after the story was recalled.

After receiving the posthypnotic suggestion on Day1, the experimenter awakened subjects from hypnosisand engaged in light conversation. The mood was rein-stated as they were handed the story (the cue) to readthrough twice at their own pace. When they finishedreading and were still in their mood, subjects filled outa five-item questionnaire that asked: "Which characterdid you identify with? Which character was stressed(emphasized) the most? Who had the most details as-sociated to him? How comprehensible (on a 10-pointscale) was the story? Describe any images you may havehad as you read."

After completing the questionnaire, the subjects forpurposes of the unrelated experiment received twostudy-free recall cycles with a 20-word list while in theemotional mood. After that they were told, "You're fin-ished. That's fine. It's over." After insuring normali-zation of their mood, the experimenter sent the subjectshome with instructions to return to the same room 24hours later.

At the beginning of the second session, the subjectwas hypnotized for the word list experiment and askedto experience either a happy or sad mood (half of eachmood group from Day 1); it was also suggested that thesubject would feel this emotion later when he or sheread recall instructions (for the word list) printed ongreen paper. The "green paper" cue was particularlystressed, since we wanted to ensure that the subject didnot access the experimentally suggested mood (i.e., wasin a neutral mood) when he or she recalled the story ofthis experiment on white paper.

After the mood induction and posthypnotic sugges-tion, subjects were awakened from trance and thenhanded the story recall instructions. These were printedon white paper (so presumably no mood was induced).They emphasized recall (in writing) of all story detailsas completely as possible and as close to the originaltext as possible. Subjects were allowed as much time asneeded for recall. After finishing, subjects were handeda green sheet with instructions to recall the word list forthe other experiment. The green paper supposedly rein-stated the emotional mood established at the beginningof the second session. When word list recall was finished,the experimenter said, "That's fine. It's over," whichwas to remove the mood and return the subject to aneutral mood. He next asked subjects to remember theirmoods during both sessions and to speculate about theaims of the experiment. They were then told, "Now youcan remember everything about this experiment," tocancel the amnesia suggestion. Following this, he askedthem again about their mood changes during the twosessions and about any hypotheses concerning the aimof the experiment. After ensuring mood normalization,the experimenter thoroughly debriefed, paid, and thankedthe subjects for their participation.

Results

Success of hypnotic suggestions. Beinghighly susceptible, the subjects rapidly wentinto deep trance and were adept at "gettinginto" the moods under hypnosis, typicallyimagining a success or a vacation scene forthe happy mood and a personal failure, loss,or funeral scene for the sad mood. Informalquestioning indicated that they followed theposthypnotic suggestions exactly, feelinghappy or sad on the appropriate occasions.Also, their behavior was appropriate; happysubjects often laughed over the story, sadsubjects appeared glum and on the verge oftears as they read the story. When ques-tioned at the end of the second session, mostcould remember that they had experienceda mood shift while recalling the word list onDay 2, and many recalled feeling in a moodwhile reading the story on Day 1 and whilelearning the word list on Day 1. However,when probed about the purpose of the ex-periment, they either said they had no ideaor they reiterated the vague description theyhad been given when initially contacted.

Questionnaire. The questionnaire askedthree questions relevant to which story char-acter the subject identified with: "Who didyou mostly identify with? Who was stressed?Who had the most details associated tohim?" Twelve of the 16 subjects answeredall three questions with the same character;four subjects gave the same character as theanswer to two items and replied with theother character to the third question. As-signing these four subjects to identificationwith the character mentioned in two of thethree questions, we find a perfect correlationbetween mood state and character identifiedwith; all eight happy subjects identified withhappy Andre, and all eight sad subjects iden-tified with sad Jack. So our mood manipu-lation was completely effective in determin-ing the character with whom the subjectidentified.

The questionnaire also asked for reportsof imagery during reading. Five of the 16subjects wrote no answer to this question.For the remaining 11, an average of 3.5 im-ages were reported in which the characteridentified with was central, but only .4 im-


ages in which the other character was cen-tral. Nine of the 11 subjects reported moreimages relevant to the character they hadidentified with (/> < .03 by a sign test). Thus,we may conclude that in identifying with acharacter, the reader tends to have imageryabout that character and the events in whichhe participates, according him "center stage."

Recall. The recall protocols were scoredfor propositions or idea units recalled, andthese were classified as facts about happyAndre, sad Jack, or neither character. First,the happy and sad subjects did not differ inthe total number of idea units recalled; therecall means were 23.6 for happy and 20.8for sad subjects, t(l4) = .56. Also, the twogroups did not differ in the number of neu-tral idea units recalled (3.5 for happy vs. 1.0for sad).

The major difference between the groupswas in the proportions of happy versus sadfacts recalled. The proportion of sad factsrecalled out of happy-plus-sad facts recalledwas calculated for each subject and thenaveraged. This sad proportion averaged 45%for the happy subjects and 80% for the sadsubjects. The difference in proportions (arcsine transformation) is statistically reliable,K14) = 3.02, p < .005. All eight sad subjectsrecalled more facts about sad Jack, com-pared to only three of eight happy subjects.Setting aside one subject who recalled anequal number of facts about the two char-acters, 12 of the other 15 subjects recalledmore facts about the character who felt thesame as they did when they read the story.This recall bias is significant by the sign test(p < .018).

Although the recall advantage for con-gruent material is consistent, the effect issmall, but that is to be expected. The happyand sad statements were integral parts of astory line with much interactive dialogue, sothere were strong contextual linkages be-tween the two categories of statements.Thus, recall of a happy fact would very likelycue recall of a sad fact that it was linked toin the story (and vice versa). To the extentthese dialogue-determined interconnectionswere formed and used, the mood-congruitybias in recall was reduced in magnitude.

We wondered why the effect in favor of

recalling happy facts by happy subjects wasnot nearly so pronounced (relative to a 50%baseline) as was recall of sad facts by sadsubjects. An answer was forthcoming uponre-examining the story; in retrospect, thestory seemed subjectively to be somewhatmore about sad Jack than happy Andre. Wehad another 22 nonhypnotized subjects (anintroductory psychology section) read thestory and indicate which character, if either,they thought was more prominent or salient.Thirteen (59%) picked sad Jack, seven saidemphasis was equal, and two (9%) pickedhappy Andre. Thus, the story is frequentlyviewed as being somewhat more about sadJack, so one would expect a normal-moodcontrol group to recall fewer facts abouthappy Andre than about sad Jack. In fact,a later control experiment found 60% recallof sad to happy-plus-sad facts for subjectswho identified with neither Andre nor Jack.Thus, the 45% sad recall percentage ob-served for happy subjects in Experiment 1indicates an effect of mood identificationthat probably is close in magnitude to thatreflected in the 80% sad recall by sad sub-jects.

Discussion

We found that readers identify with thestory character who is experiencing a moodsimilar to the one they are experiencingwhile they read the story. Thus, mood sim-ilarity seems to be a determinant of identi-fication. The character with whom readersidentify influences their perception and re-call of the story. Readers believe the storywas mainly about the character they iden-tified with and believe that more details weregiven about him; he holds center stage intheir dramatic imagery, and more factsabout him are recalled later.

The results decisively reject the Pollyannahypothesis in this application: Readersshowed no overall favoritism toward thehappy character. Thus, rather than focusingon cheerful aspects of the text, our sad sub-jects supported the adage "misery likes com-pany."

The results support the mood-congruityhypothesis that material relevant to the


reader's mood will be especially salient andwill receive special attention, processing, andelaboration during encoding into memory.Two separable aspects of this account war-rant discussion. First, the effect may or maynot be mediated specifically by readers iden-tifying with a specific character as opposedto their emphasizing and making salient spe-cific kinds of affective material. Second, themood effect is here alleged to occur at thetime of encoding (learning), not at the timeof recall.

The first issue is whether the mood-con-gruity effect is mediated specifically by char-acter identification as opposed to special sa-lience of material matching the reader'smood. These two factors are confounded inthe Jack-Andre story, since the only happyevents are those associated with Andre, andthe only sad events are those associated withJack. Perhaps the reader's selectivity isbased on happy versus sad events, ratherthan happy versus sad characters. This issuewas of sufficient interest that Experi-ments 3 and 4 were undertaken later toexamine it.

If the mood effect is based on characteridentification, then one should be able toproduce a similar recall bias in neutral-moodreaders by instructing them explicitly to takethe point of view of one or the other selectedcharacter. A brief attempt in this directionfailed completely. A class of 30 HarvardUniversity summer students was given theAndre-Jack story to read; a third were in-structed to adopt Jack's point of view, a thirdto adopt Andre's point of view, and a thirdreceived neutral reading instructions. Theinstructions to the first two groups askedthem to "read the story from Andre's (orJack's) point of view, as though you were inhis shoes." The subjects read the story for5 min. and then recalled it 20 min. later.

The results were quite disappointing. Thethree groups recalled about the same numberof idea units, and their ratios of sad to happyunits recalled were similar. The percentagesof recalled sad units out of sad-plus-happyunits recalled were 54% and 60% for subjectswho reported identifying with the happy andthe sad character, respectively. The happyrecall percentage was 55% for subjects whoreported identifying equally with both char-

acters or with neither character. This studyrevealed to us the ambiguity of our "iden-tification" instructions and questionnaire,since subjects admitted to confusing identi-fication ("Which character did I projectmyself onto?") with attractiveness ("Whois the more likable character?") and simi-larity judgments ("Whose personality ismore like mine?"). These differing bases forjudgment led to there being no systematicinfluence of our instructions on subjects' re-ports of "identifying" with the two charac-ters. If the results of this pilot experimentare valid, they suggest that our emotionalmood manipulation was far more powerfulthan explicit instructions in causing readersto identify with a selected character. An al-ternative interpretation, mentioned previ-ously, is that the recall bias in Experiment1 may not depend on character identificationbut on differential salience of pleasant versusunpleasant material for happy versus sadreaders. In such a case, one would not expectrecall differences when neutral-mood sub-jects are told to identify with one or anothercharacter.

Taking up the second issue, we claim thatthe mood effect here results from differentiallearning (input) of happy versus sad mate-rial. Could the result instead be an outputeffect? A plausible alternative is that char-acter focus during reading affects characterfocus during later remembering. That is, thesubject might organize his or her reconstruc-tion of the story around a subjective "maincharacter," and prompt or cue himself orherself to recall facts about that characterto the detriment of recalling facts about theother characters. Or perhaps the priority inrecall of facts about the central charactercreates "output interference" for later recallof facts about secondary characters.

The "priority in recall order" explanationcan be discarded immediately because inrecalling, subjects overwhelmingly followedthe narrative order of events. Thus, factsabout each character are not recalled en blocbut rather in a sequential mix. This must beso, since an assertion about one character isoften linked narratively (and associatively)to an assertion about the other character.Putting aside the output-order explanation,the remaining selective output hypothesis


could be tested either by using recognitionmemory tests for text sentences (thus min-imizing effects due to reconstructive strat-egies) or by inducing the emotional moodand character identification after the personhas read the text (in a neutral state) butbefore he or she recalls it. Although we findthe encoding hypothesis more plausible, theoutput reconstruction hypothesis is suffi-ciently plausible that we carried out Exper-iment 2 to test it.

Experiment 2

The next experiment asked whether in-ducing the emotional mood only at the timeof recall would influence the recall of happyversus sad facts. The reader read the storywhile in a normal (neutral) mood. Later amood was induced by hypnosis and the sub-ject recalled the story while in that mood.

If an effect of mood on recall is observed,a second issue is whether new memorieswould become available if the subject's moodwere shifted before a second recall. Ander-son and Pichert (1978) reported a small in-crease in recall when the perspective of re-callers on a story was shifted before a secondrecall; the recall increment came primarilyfrom parts of the story relevant to the secondperspective. We wondered whether a moodshift from happiness to sadness (or viceversa) between recalls would make mood-congruent memories more recallable. Ac-cordingly, after a first recall in one mood,subjects were shifted to the opposite moodand recalled the narrative a second time.

MethodSubjects. The subjects were 16 adult volunteers from

a self-hypnosis training class conducted by the secondauthor. All were mental health professionals (doctorsor graduate students in clinical psychology or psychia-try). All were excellent hypnotic subjects.

Procedure. All subjects read the Jack-Andre storyduring the final morning of a S-day workshop on self-hypnosis. They were told to read the story (for 6 min.)to be able to answer some questions about it later. Fivehours later, half the subjects were placed hypnoticallyin a happy mood and half in a sad mood using the sameprocedure as in Experiment 1. They were then given 15min. to recall the story while in this mood and askedto write down every detail they could remember as closeto verbatim as possible. After completing the first recall,the subjects' moods were removed. They were thenplaced in the mood opposite the one they had first been

given and asked to recall the story again, including theold facts they had recalled and any new facts that cameto mind. Again subjects had 15 min. to complete thetask. While maintaining their second mood, they thenfilled out a questionnaire containing 14 questions. Twoquestions asked subjects to rate on a 7-point scale howmuch they identified with Andre and with Jack whilereading the story and at the current time. Completionof the questionnaire ended the experiment.

Results and Discussion

Questionnaire. A first question is whetherthe subjects' current moods during recall in-fluenced their report of the character withwhom they identified. We found no effectof current mood on subjects' retrospectivereports of which character they had identi-fied with earlier when they read the story(while in a neutral mood). However, sub-jects' current moods did influence their re-ports of which character they were currentlyidentifying with. Subjects who were sad atthe time they answered the questionnairesaid that at that moment they identifiedmore with sad Jack (5.25 on a 7-point scale)than with happy Andre (2.87), and theseratings differ reliably, t(l4) = 3.35, p < .01.Similarly, subjects who were happy said theyidentified at that moment more with happyAndre (5.37) than with sad Jack (3.37)—areliable difference, t( 14)i = 2.75, p< .05.Comparison of subjects' identification rat-ings for the two characters shows that 13of the 16 subjects reported identifying morewith the character whose mood matchedtheirs at the time of the rating. This split issignificant by a sign test (p < .01). We mayconclude that our subjects' moods at the timeof recall were effective in causing them toreport identifying at that time with the char-acter of the same mood. We next examinewhether this character identification or moodduring recall had any influence on recall.

Recall. The recalls were scored for thenumber of idea units each subject recalledassociated with happy Andre, sad Jack, orneither or both characters (neutral facts).The group averages of those scores areshown in Table 1. A first observation is thatthe sad-then-happy subjects were slightlybetter recallers overall than the happy-then-sad subjects, a difference of no importance.Second, there was no evidence whatsoeverfor a selective effect of mood on recall of


Table 1Mean Number of Happy, Sad, Neutral, and Total Idea Units Recalled During the First andSecond Recalls by Subjects in Experiment 2

Group

Happy then sadSad then happy

Sad

13.7516.50

First

Happy

13.7514.62

recall

Neutral

.501.00

Second recall

Total

28.0032.10

Sad

17.3719.75

Happy

15.0015.75

Neutral

0.871.62

Total

33.2537.12

happy versus sad material. For none of thecolumns in Table 1 does the difference be-tween the two groups' means even approachsignificance. Indexing selectivity by the ratioof sad to sad-plus-happy facts recalled, wefind that the indices on the first recall are.50 for the happy recallers and .53 for thesad recallers; for the second recall the pro-portionate sad-fact recall was .54 for the sadrecallers and .56 for the happy recallers.None of these proportions differs signifi-cantly from the .50 baseline of no bias inrecall of facts about the happy versus sadcharacter. Moreover, a change in recallmood caused no reliable shift in the recallratio for either group. Inspection of Table1 reveals that more idea units occurred onthe second recall than on the first recall, aneffect explicable simply by a longer memory-search time. We had wondered whether amood shift would cause relatively more newfacts congruent with that new mood to ap-pear in the second recall, but there was noevidence for a mood-congruent increment:Averaged over the two groups, the mean in-crement in recall for shifted mood-congruentidea units was 2.37, and for shifted mood-incongruent idea units it was 2.25, an insig-nificant difference.

Thus, Experiment 2 found no influence ofthe subjects' current mood on selective recallof happy versus sad material read earlierwhile in a neutral mood. The current mooddefinitely influenced the character with whomrecallers said they identified at that moment;however, that report of identification was notcorrelated with the number of facts recalledabout the happy or sad character. Further,inspection of the recall protocols revealed noconsistent, obvious qualitative differences inthe subjective focus or story-telling perspec-tive of sad versus happy subjects. Thus, al-though recallers may be sympathetic to one

character, they nonetheless retell the storysubstantially in the balanced third-personperspective in which it was written and read.

The conclusion from Experiment 2 is es-sentially negative; selectivity of recall ofmaterial encoded while in a neutral moodcould not be induced by manipulating mood(and character identification) at output. Thissuggests that the recall selectivity obtainedin Experiment 1 was an input (encoding)effect, not an output (retrieval) effect. Thatis, the readers' moods during reading influ-ence selectively the way they process andlearn mood-congruent versus incongruentmaterial. Their later differential recalllargely reflects differing degrees of learningof the two types of materials,

The null result in Experiment 2 also dis-credits the idea that our earlier results onlyreflect the hypnotic demand characteristicsof moods for the suggestible subjects. Thathypothesis supposes that suggestible subjectsshould comply with whatever role behaviorthey believe is being implicitly demanded bythe hypnotist. Thus, happy or sad subjectsmay feel that in concentrating on happy orsad material, respectively, they are followingthe implicit demands of the situation. Butthe demand hypothesis expects a strong se-lective effect of mood on output of happy orunhappy events, and none was forthcomingin Experiment 2 (or later in Experiment 4),thus weakening the demand hypothesis.

Experiment 3

Experiment 1 left us with the question ofwhether the selective recall effect was spe-cifically due to the reader identifying withthe mood-congruent character or focusingon and elaborating statements that werehappy or sad. As noted, the two factors areconfounded in the Jack-Andre story. Our


next experiment used a narrative about asingle character in a neutral mood describinga series of happy incidents and sad incidentsfrom his life. If the selective learning in Ex-periment 1 had been caused by characteridentification, then happy and sad readersshould not differ in what they recall of thisone-character narrative. On the other hand,if the selective learning had been due to moreelaborative processing of any mood-con-gruent material, then happy readers shouldrecall more happy incidents and sad readersmore sad incidents.

A second question is whether subjects'moods at the time of recall will have a maineffect on recall or will interact with theirmoods during learning. Although Experi-ment 2 found no recall difference due tooutput mood, that experiment confoundedcharacter with type of material, and thestory was read in a neutral mood. Perhapsoutput mood affects recall only if the ma-terial was acquired in a specific mood. Amood-state-dependent retrieval hypothesiswould suppose that recall would be bestwhen the mood during recall matches thesubject's mood during original learning. Toexplore these possibilities, different subjectslearned the story in a happy or sad mood,then half of each subgroup was tested forrecall while happy and half while sad.

Along with recall, subjects also estimatedfrom memory the relative frequency of sadand happy incidents in the narrative. A hy-pothesis of Tversky and Kahneman (1973)suggests that relative frequency judgmentsof alternative events are biased according tothe "availability" of those events in memory.This hypothesis expects subjects' estimatesof the proportion of happy versus sad inci-dents in the narrative to follow their pro-portionate recall of these incidents. Thus, ifmood at input or output influences recall ofhappy versus sad incidents, similar effectsshould occur for relative frequency estimatesof happy versus sad incidents.

Method

Design. The experiment used four groups arrangedin a 2 X 2 factorial with input mood (happy or sad) asone factor, and output mood (happy or sad) as a secondfactor. After reading the story, subjects were tested forrecall and gave their estimates of the relative frequency

of happy to sad incidents in the story they had read.Subjects also answered a retrospective questionnaireabout their subjective moods and experiences whilereading or recalling the story.

Subjects, There were 32 subjects, 19 females and13 males. Sixteen were Stanford undergraduates se-lected for high scores (10-12) and for passing the am-nesia test on the Stanford Hypnotic Susceptibility Scale,Form C (Weitzenhoffer & Hilgard, 1962). The other16 subjects were selected from a pool of former studentsin hypnosis workshops conducted by the second author.All were excellent hypnotic subjects who scored in the10-12 range on the Form C test and showed amnesia.Eight subjects at random were assigned to each of thefour experimental groups. All subjects were paid fortheir participation in the experiment.

Materials. All subjects read a 660-word narrativethat described four different age-regression sessions in-volving a male patient (Paul Smith) undergoing hyp-notherapy with a psychiatrist. Except for brief state-ments giving the setting and the framework of beginningand ending the four therapeutic sessions, the narrativemainly depicted Paul recalling and relating a carefullybalanced variety of happy, neutral, and sad incidentsfrom his life. The text is reproduced in the Appendix.It was divided into 78 basic idea units, of which 26 camefrom descriptions of happy incidents, 26 from sad in-cidents, and 26 from neutral incidents. Examples ofhappy items were memories of riding piggyback on hisfather's back, jokes he had been told, festive family gath-erings, good grades in school, winning a football game,and his first true love. Sad memories included the deathsof his grandfather and his dog, being cut from his base-ball squad, the breakup of the Beatles music group, andhis sister's injuries in an auto accident. Neutral itemsusually referred to the setting information, the begin-ning trance induction or ending trance termination ofPaul at each session, and remarks between the doctorand Paul about his progress.

After recalling this narrative, subjects filled out threesections of a questionnaire. One part asked their impres-sion of the narrative, its emotional content, and imagesthey had while reading and while recalling. A secondsection asked them to estimate how much of the story'semotional content was composed of happy incidents andhow much was composed of sad incidents. The finalsection of the questionnaire probed their memory forany moods they experienced while reading and recalling,fluctuations in mood, and whatever hypotheses they hadto explain these moods.

Procedure. Subjects were assigned randomly to thefour experimental groups and were run in subgroups offour. After establishing rapport, the experimenter in-duced a deep trance by the eye-closure method. Subjectswere then asked to recall a personal experience that hadinvolved a happy or sad mood (depending on their con-dition). They were to re-experience that emotionalmood. Further, it was suggested that once out of trance,they would fully reexperience this mood state (thoughnot the remembered situation) when they were handeda story to be read. This mood would remain throughoutthe reading of the story until a cue was given. Amnesiainstructions for all hypnotic suggestions were then given,along with posthypnotic cues for trance re-entry ("Okay,


Table 2Average Number of Idea Units (out of 26) Recalled by Subjects in the Four Groups

Mood

Learn

HappyHappySadSad

Recall

HappySadHappySad

Happy

8.637.006.755.00

Sad

6.875.756.639.25

Type of fact

Neutral

7.254.253.753.12

Sum

22.7517.0017.1317.37

New

3.621.621.872.25

Frequencyestimate(% sad)

42.559.545.665.9

now") and for amnesia removal ("All right, that's allfor now"). Subjects were then removed from trance.This entire first portion took about 15 min.

Once out of trance, subjects were handed the PaulSmith narrative and instructed to read it carefully for5 min. During this reading, the subjects should have felthappy or sad according to their posthypnotic suggestion.After 5 min. they handed back the story sheet.

Next the cue for trance re-entry was given followedby a 10-min. induction. The subjects were then askedto recall a different situation from their past that hadmade them intensely happy (or sad, depending on theirexperimental condition). After reviving that emotionalfeeling, they were told that after awakening from trancethey would re-experience this mood (but not the situ-ation) while engaging in recall and while filling out aquestionnaire. Furthermore, amnesia for these tranceinstructions was suggested except for cues for the liftingof the amnesia and for their return to a normal mood.Following these instructions, subjects were taken out oftrance. This second trance procedure required about 20min, so that was the interval between story reading andrecall.

Next, each subject received the recall sheet. It in-structed subjects to recall everything they could aboutthe story, including all details they could remember nomatter how trivial they seemed. They had 10 min. towrite their recall. The recall sheets were then collectedand the questionnaires were handed out. After comple-tion of the first parts of the questionnaire, the cue forremoval of amnesia and mood normalization was given.After recovery of memory for trance events and moodnormalization had been carefully insured, the subjectscompleted the final section of the questionnaire. Thisasked subjects' impressions of mood fluctuations duringreading and recall and asked their hypotheses aboutthese changes. Then subjects were thoroughly debriefedabout the experiment and dismissed.

Results

Recall. The recall protocols were scoredfor idea units recalled; each was classifiedas a happy, sad, or neutral unit, and newunits (recall intrusions) were noted. Thegroup averages for these categories areshown in Table 2.

Several conclusions are warranted by theresults in Table 2. First, happy and sad items

were recalled better overall than neutralitems, but this comparison is not balancedfor the type of material (e.g., settings vs.Paul's emotional incidents) and so should beignored. Second, there is a slight mood-state-dependent retrieval effect; the groups whoserecall mood matched their learning moodrecalled slightly more in total than thegroups whose recall mood mismatched theirlearning mood. However, this interaction isnot reliable with the total recall scores, F(l,28) = 2.48, p > .05. A second test for thisInput X Output interaction was carried outon recalls of only the emotional incidents,but that too was not significant. Thus, thesedata reveal no mood-state-dependent re-trieval effect. This failure duplicates thatfound for one-list recall experiments byBower, Monteiro, and Gilligan (1978); theyobserved mood-state retrieval effects onlywhen their subjects were placed in a list-dis-crimination problem, having learned twoconfusable lists, each in a different mood.

The conspicuous effect in Table 2 is thatpreferential recall of happy versus sad in-cidents differs according to the subject'smood during learning. To reduce intersub-ject variability in levels of recall, we indexedpreferential recall by the ratio of sad itemsrecalled to the sum of happy and sad itemsrecalled. The average ratios across groupswere .43 for happy-happy, .44 for happy-sad, .50 for sad-happy, and .66 for sad-sad.An analysis of variance was carried out usingarc sine transformations of these ratios. Thisrevealed a significant main effect of the sub-jects' mood during reading, with F(l,28) = 4.60, p < .05. However, subjects' moodduring recall had no significant effect, nordid the interaction of learning and recallmoods. To summarize, happy learners re-called more happy incidents and sad learners


tended to recall more sad incidents, but therecall mood had no significant effect on pref-erential recall. The glaring exception to thisgeneral "storage" effect was the sad-happysubjects (line 3 in Table 2) who recallednearly equal numbers of sad and happy in-cidents. We have no explanation for this ex-ception, although it is in a direction pre-dicted by an output effect working againsta storage effect of mood.

Frequency estimates. After recall, sub-jects estimated from memory the propor-tions of the narrative sentences that had beenabout happy incidents, and, separately, aboutsad incidents. We formed the ratio of sad tohappy-plus-sad estimated proportions to ob-tain a "percent sad" estimate for each sub-ject. The group averages of these ratios arepresented in the last column of Table 2. In-spection of these relative frequency esti-mates suggests an effect due to the subjects'mood at the time they made the estimate(just after recall). Analysis of variance onthe arc sine transformations of these per-centage estimates revealed a significant ef-fect of mood during testing, with F(l,28) = 16.9, p < .01, but no effect of moodduring learning or of the interaction betweenlearning and testing mood. Currently happysubjects estimated that the narrative hadcontained more happy incidents, whereassad subjects estimated that it had containedmore sad incidents. This test effect of moodon frequency estimation contrasts with thestorage effect of mood on selective recall.

For three of the groups the estimated per-centages of sad to happy-plus-sad incidentswere very close to the percentage ratios ofrecalled sad and happy incidents. Thus, thehappy-happy group had a sad recall ratioof 43.6 and an estimated sad percentage of42.5. The sad-sad group had a sad recallratio of 66.2 and an estimated sad percent-age of 65.9. The sad-happy group estimateda sad percentage of 45.6 and recalled in asad ratio of 49.5. The closeness of the recallratios and percentage estimates for thesethree groups supports the "availability" hy-pothesis of how people make relative fre-quency judgments. But the happy-sad groupis quite discrepant from this hypothesis; theirsad recall ratio was 43.9 compared to anestimated sad percentage of 59.5. These sub-

jects, therefore, were giving estimates of sad-fact proportions considerably higher thanthe proportion of sad to happy facts they hadjust been able to recall. In fact, this dis-crepancy largely accounts for the differinginfluences of storage versus test mood onselective recall versus frequency estimates.We cannot explain this discrepant result.

Questionnaire. A few interesting find-ings were provided by the questionnaire.Thirty of the 32 subjects reported that in-cidents of the narrative often reminded themof certain incidents from their own life. Fif-teen of the subjects gave only a general re-port of reminding, but another 15 reportedspecific story incidents that had remindedthem of events from their life. Of 33 reportedspecific incidents, 24 (73%) agreed in emo-tional quality with the reader's mood. Forinstance, sad readers were more likely to bereminded of an (invariably sad) incidentfrom their life while reading a sad incidentabout Paul, whereas happy incidents ofPaul's life were less likely to remind themof any incident from their life. Happy read-ers showed the reverse selectivity of remind-ing. This selective reminding probably co-varies with the personal elaboration of mood-congruent material. In the Discussion sec-tion we return to the question of whetherthis selective reminding process might beresponsible for the selective learning ofmood-congruent material in these experi-ments.

Finally, 21 subjects reported that theirmood fluctuated somewhat during readingand that this was most likely to happen asthey read an incident of an emotional qualityopposite to the one they were feeling. Thus,happy readers might have found their eu-phoria subsiding briefly when they read ofsome sad incident of Paul's life. In retro-spect, this effect seems entirely plausible;however, it has implications for the mood-congruity hypothesis that we will note at theend of this article.

Discussion

The major finding of Experiment 3 wasthat selective learning occurs with respectto mood-congruent incidents. Thus, the se-lective effect of mood on learning does not


require the reader to identify with a selectedcharacter, as appeared to be the case in Ex-periment 1. Rather, the selective-learningeffect of Experiment 1 can now be reinter-preted as showing an effect due to enhancedsalience of happy versus sad materials ratherthan an effect due to identification with thehappy versus sad character. Experiment 5attempts to further separate these two fac-tors.

A second major result is that mood led toselective learning but not selective recall ofhappy versus sad incidents in the narrative.Happy readers tended to learn more happythan sad incidents; sad readers did the re-verse. But mood during recall had no effecton what was recalled. An explanation for thebetter learning of mood-congruent materialwas suggested by responses to the question-naire. Although not designed for this specificpurpose, the questionnaire did elicit reportsthat subjects were frequently reminded ofincidents from their past by incidents inPaul's narrative. Incidents that evoked rem-iniscence usually matched the emotionalquality the reader was feeling; that is, sadreaders were more likely to be reminded bya sad incident in the narrative, and happyreaders by a happy incident. It seems likelythat these reminding experiences themselvespromote very good learning, perhaps becausethe remembered event can be used to elab-orate and embroider the current input, per-haps because the reminding lends distinctivepersonal and affective tone to the narrativeevent. In related work, Bower and Gilligan(1979) reported that brief event or objectdescriptions (e.g., "a dying dog" "a softchair") were better recalled later by subjectswho were asked whether each descriptionreminded them of a personal experience thenby subjects who judged the meaning of eachdescriptive phrase. Moreover, descriptionsthat successfully reminded the former sub-jects of a personal experience were betterrecalled later then descriptions that did not.Several hypotheses are plausible to explainthe reminding effect on memory. Ratherthen pursuing these here, we will merely ob-serve that such memory enhancement byreminding may underlie the mood-congruityeffect we have found for materials. We re-

turn to this explanation at the end of thisarticle.

The disturbing element of the remindingexplanation is that it assumes that the sub-jects' mood should facilitate the likelihoodthat an emotional narrative will remind themof a similar personal incident. But this ex-periment has shown no selective effect ofrecall mood on retrieval; that is, sad recallerswere not more likely to retrieve sad thanhappy parts of the narrative. However, theinteraction between input mood and outputmood was suggestively in the predicted di-rection, so perhaps one should not reject theselective retrieval hypothesis until more ev-idence has accumulated.

Experiment 4

This experiment was run at the same timeas Experiment 3 to search for an influenceof recall mood on selective retrieval. Its re-sults are worth reporting briefly to indicatethe reliability of our results. The basic ques-tion was whether a selective retrieval effectcould be demonstrated with the one-char-acter (Paul) narrative used in Experiment3. Therefore, we essentially replicated Ex-periment 2 using the Paul narrative. Sub-jects read the story while in a neutral mood,later recalled it while in a happy or sadmood, then switched moods and recalled ita second time. The issue is whether subjectsselectively retrieved more incidents withemotional qualities that agreed with theirown while recalling.

MethodSixteen highly hypnotizable subjects from the same

sources as the preceding experiments were randomlyassigned to one of two recall conditions (happy or sad).All subjects first read the Paul Smith story for 5 min.while awake and in a neutral mood. They were in-structed to remember it. They were asked to return laterin the day (4-6 hr. later). Upon returning, they werehypnotized, placed into a happy or sad mood (eight sub-jects per mood), and instructed to maintain this moodlater while recalling the previously read story. The sub-jects were then aroused from trance, given several sheetsof paper, and instructed to write a detailed free recallof the Paul Smith story. After 10 min. of this task,subjects were rehypnotized, put into the alternate mood,and told to maintain this second mood while again re-calling the story. They were then taken out of tranceand given 10 min. to write their second recall of the


Table 3Number of Idea Units Recalled in Experiment 4 Categorized by Type Of Unit

Mood First recall type of fact Second recall type of fact

First Second Happy Sad Neutral Total Happy Sad Neutral Total

HappySad

SadHappy

6.126.25

6.377.50

6.006.25

18.5020.00

6.126.75

6.877.12

4.756.00

17.7519.87

Paul narrative (while in the second mood). Approxi-mately 20 min. elapsed between the end of the first recalland the beginning of the second. The hypnotic proce-dures of Experiment 2 were followed closely. The recallprotocols were scored as in Experiment 3.

Results

The mood manipulations were effective inthat subjects reported feeling happy or sadas instructed and they behaved appropri-ately. However, their emotional mood duringrecall had practically no influence on whatthey recalled. The average recall scores aredisplayed in Table 3. The mood of the re-caller did not significantly affect either totalrecall or the ratio of happy to sad incidentsrecalled. Sad incidents were recalled slightlymore than happy incidents under all circum-stances. Moreover, the shift in mood betweenthe two recalls did not permit subjects toaccess reliably more facts congruent with thesecond mood.

These negative results essentially reaf-firmed those of Experiment 2 with differentlearning materials; material read and learnedunder a neutral mood was not differentiallyretrievable later according to the mood ofthe recaller. Perhaps output mood will affectrecall only if the subject feels that mood in-tensely during learning, so that the materialis encoded in the context of that emotion.That is, output mood might affect recall onlyby virtue of mood-state-dependent retrieval.

Experiment 5

Experiment 1 demonstrated that subjectsrecalled more about a character who was inthe same mood as they were. But statementsabout him described either all pleasant orall unpleasant events and feelings. Experi-ment 3 found that subjects recalled moreevents (about a single character) for which

the predominant emotion matched the moodof the reader. The mood-congruity effect inExperiment 3 raised the question of whetherthe result in Experiment 1 resulted from sub-jects' identifying with a happy (or sad) char-acter or from the fact that all happy (or sad)experiences were associated only with happyAndre (or sad Jack). Conceivably, the en-hanced salience of material congruent withthe reader's mood suffices to explain the re-sults of Experiment 1, with no need to pos-tulate an additional effect of character iden-tification. Alternatively, if mood promotesthe reader's identification with a character,this factor could operate along with the ma-terials-congruity factor to promote betterlearning of congruent material associatedwith a same-mood character.

The narrative used in Experiment 5 variedindependently the two factors of characteridentification and mood of the incidents de-scribed by the characters. While happy orsad, our subjects read and learned a narra-tive that described two characters undergo-ing psychiatric treatment, one character al-ways depressed, the other always euphoric.During the therapeutic interview, each char-acter recounted in mixed order many happyand sad incidents from his life. Subjects laterrecalled as much of the narrative as theycould while they were in a normal, neutralmood.

The hypotheses yield differing predictionsfor the recall results. The character-identi-fication hypothesis predicts that readers willselectively learn the facts about the char-acter most resembling their mood. Thus,happy readers should remember more aboutthe happy than the sad character, whereassad readers should show the reverse orderingin recall. The character-identification hy-pothesis predicts no difference in recall ofhappy versus sad incidents. On the other


hand, the materials-salience hypothesis sup-poses that a sad mood should enhance thesalience and memorability of any sad inci-dent, and a happy mood should enhance thememorability of any happy incident, regard-less of which character recounts the incident.Experiment 5 tested these alternative pre-dictions.

MethodSubjects. The sixteen subjects, solicited from the

same sources as in the previous experiments, were highlyhypnotizable (Form C scores of 10-12). A random eightsubjects read and learned the narrative while happy,and eight while sad. They were run in groups of foursame-mood subjects at one time.

Materials. The narrative of about 1,200 words de-scribed two men, Mike and Joe, having a final jointpsychotherapy meeting with their psychiatrist. He hadbeen engaging them in hypnotic age regression; in thissession he asked each patient to share with the other theset of boyhood incidents he had described earlier duringage regression. Thus, the narrative consisted of the twopatients describing to each other (and the psychiatrist)a series of brief incidents from their lives. Half the in-cidents described by each character were happy or pleas-ant, and half were sad or unpleasant. The passage con-tained 135 idea units (simple clauses). Of these, 1S unitswere neutral, and 60 were attributed to Mike, and 60to Joe; half of each character's 60 units related happyincidents and half related sad incidents.

Before reading the narrative, subjects read one para-graph describing the setting and patients in the narra-tive. This background mentioned that Mike sufferedfrom long-standing, severe depression, whereas Joe wasa happy-go-lucky manic who entered therapy becausehis continual euphoria and practical joking were inter-fering with his job and social adjustment. Throughoutthe narrative, Mike was described as a morose, miser-able, tearful man who spoke sadly; Joe was describedas happy, even giddy, laughing frequently. Nonetheless,each character described equal numbers of happy andsad incidents from his life.

Procedure. The subjects were hypnotized in groupsof four using the eye-closure method as before. Aftertrance induction, subjects were made happy or sad bythe usual procedure. They were to maintain this emotionat level intensity as they read a narrative handed themby the experimenter. They were not given amnesia forthis instruction. While in trance and experiencing theinstructed mood, each subject received the typed nar-rative and was told to read it in order to answer somelater questions about it. They had 10 min. to read andstudy the narrative. Then the mood was removed, andthey were awakened from hypnosis. Subjects then en-gaged in a distracting task (adding columns of numbers)for 10 min. Following this, they were handed severalsheets of paper and asked to recall everything they couldabout the Mike and Joe narrative they had read 20 min.earlier. They were told to recall as many details as theycould, as near to the original wording as possible, but

to recall the gist of any incidents if they could not recallthe original wording. Subjects had unlimited time towrite their recall, but all finished within IS min. Afterreading the narrative, subjects were asked which char-acter they had identified with and who was the "maincharacter" in the narrative. The recall protocols werescored for basic units (clauses) recalled; these were clas-sified according to the central character (Mike or Joe)who related the event about himself and whether it wasa happy or sad incident.

Results

The mood manipulation proved effectiveonce again, as indicated by subjects' reportsand observed emotional behaviors whilereading. The recall results are displayed inTable 4. This displays the average numberof idea units recalled out of 30, classified bythe happy and sad incidents of each char-acter. The rows indicate whether subjectswere happy or sad while they were readingthe narrative. Happy and sad readers did notdiffer significantly in their overall level ofrecall, F(l, 14) = 2.49, p< .05, and themain effects due to type of incident andmood of character were not significant. Anobvious effect in Table 4 is that regardlessof their mood, readers recalled more sad in-cidents about the sad character and morehappy incidents about the happy character,F(l, 14) = 109, p < .01. Thus, selective re-call of incidents was biased in the directionof the predominant mood of the story char-acter relating them. A second effect is thathappy readers recalled more happy than sadincidents, whereas sad readers recalled moresad than happy idea units, F(l, 14) = 7.3,p < .05. Thus, readers recalled more inci-dents with emotional quality that matchedtheir own—the mood congruity effect thatwas predicted. On the other hand, there wasnot a mood-congruity effect with respect tocharacters; that is, recall of facts about agiven character was not significantly greaterfor readers whose mood matched that char-acter's. The test for this interaction betweenreader's mood and character's mood was in-significant. Finally, the triple interaction be-tween type of mood, character, and incidentwas not significant.

After recalling in a neutral mood, subjectsanswered the questionnaire asking (on a 7-point scale) how much they had identifiedwith each character when they were reading


Table 4Average Number of Idea Units Recalled by Happy and Sad Readers, the Units Classified asNarrated by the Happy or Sad Character and About a Happy or Sad Incident

Incidents

Reader mood

HappySad

Total

Happy

4.875.25

10.12

Sad character

Sad

5.629.12

14.74

Happy character

Total

10.4914.3724.86

Happy

6.506.75

13.25

Sad

3.375.508.87

Total

9.8712.2522.12

the story. These retrospective reports pro-duced no differences in reported identifica-tion of happy and sad readers. In a com-parison of the 7-point identification ratingsgiven to the two characters, four of our sub-jects gave tied ratings, seven said they hadidentified more with the same-mood char-acter during reading, and five said they hadidentified more with the opposite-mood char-acter. This split does not differ from chance.Further, the mean identification for thesame-mood character was 3.00 (3 means"cannot identify very much with this char-acter") and for the opposite-mood characterwas 3.12; these means do not differ signifi-cantly. According to this retrospective ques-tionnaire, then, the readers' emotional moodswere not influencing their identification withone or the other character.

The questionnaire also asked subjects toestimate from memory the percentage of sadincidents among the total incidents describedin the narrative by each character. Overallthere was a bias to estimate that the sadcharacter related more sad incidents (55%)and the happy character fewer sad incidents(43%). However, these estimates were notinfluenced by the reader's mood. The esti-mates closely parallel the fact that readersrecalled more sad incidents about the sadcharacter (59%) and fewer sad incidentsabout the happy character (40%). Hence,these percentage estimates support the mem-ory availability heuristic of subjective fre-quency.

To summarize, Experiment 5 found thatmood selectivity due to happy versus sad in-cidents was stronger than that due to happyversus sad characters. Sad readers recalledmore sad than happy incidents, whereashappy readers recalled more happy than sad

incidents. Readers did not recall more over-all about the character who matched theirmood. There was a "consistency" effect ofrecalling more incidents consonant with themood of the character, but this occurred in-dependently of the mood of the reader. Ourquestionnaire results, if they are to be cred-ited, suggest that identification with a same-mood character is not automatic or guar-anteed. Perhaps the conflicting emotions ofthe incidents being described by each char-acter prevented clear identification; perhapsthe narrative dwelt too much on the incidentsthemselves rather than on the current moodstate of the two characters; perhaps askingfor the identification reports retrospectivelyafter the mood was lifted caused the failure.The failure of mood-congruent characteridentification is a discrepancy from Experi-ment 1 that remains a puzzle to be investi-gated. A simple inference, and possibly thecorrect one, is that the reader's mood influ-ences the salience of mood-congruent eventsbut not character identification per se, andthat the effect in Experiment 1 was spurious,due to the Andre-Jack story confoundingthe type of event with the character.

General Discussion

Let us summarize the sequence of ques-tions and findings. First, we asked whetherthe reader's mood would cause selectivelearning of a narrative about a happy andsad character interacting. In Experiment 1readers reported identifying with the char-acter who was in the same mood as theywere. Moreover, they remembered more ac-tions of that same-mood character. WhereasExperiment 1 showed a clear effect on learn-ing of mood during reading, Experiment 2


found that mood during recall had no selec-tive influence on recall of the two-characternarrative when it had been read in a normalmood. Surprisingly, readers did not intrudeor import confabulations congruent withtheir mood while reconstructing the narra-tive.

Experiment 3 showed that similar selec-tive recall occurred for happy versus sad in-cidents about a single character; emotionalreaders recalled more about those incidentswhose emotional quality matched their own.Thus, character identification was not nec-essary to produce selective learning. Therewas a slight mood-dependent retrieval effect;recall was somewhat better when mood dur-ing recall matched that present during learn-ing. Experiment 4 showed again, with theone-character narrative, that mood duringrecall (after neutral-mood learning) did notselectively influence the pleasantness of theincidents recalled. Experiment 5 pitted thecharacter identification factor against thehappy versus sad quality of the incidents.The narrative presented a happy and a sadcharacter each describing happy and sad in-cidents. We found that readers tended torecall more about incidents whose emotionalquality matched their own (summing overthe characters narrating the incident),whereas they did not recall more about thecharacter whose mood matched theirs. Also,regardless of their mood, all subjects recalledmore sad incidents about the sad characterand more happy incidents about the happycharacter.

The upshot of these experiments is thatmood during reading causes selective learn-ing of mood-congruent material in opposi-tion to mood-incongruent material. The ef-fect of mood during recall is small tononexistent; if the material was input whilethe subject was in a neutral mood, then out-put mood had no influence; if material waslearned while in a strong mood, there wasa slight tendency for recall to be better whenthe subjects' output mood matched their in-put mood. Earlier experiments by Bower,Monteiro, and Gilligan (1978) found thatsuch mood-state-dependent retrieval effectscan be enhanced by having the subject learnseveral sets of materials under differentmoods.

Our conclusion about the absence of an

effect due to output mood obviously dependson an experimental context in which, despitethe mood, subjects were highly motivated torecall everything they could to please theexperimenter. Clearly, mood might affectrecall in conditions having lower motivationfor performance. Thus, in real life the de-pressed person may not feel motivated to tryto recall very much; or the angry person maybe so distracted by his or her rage that itinterferes with the memory search. Our con-ditions have avoided these motivational orinterference effects due to output moodalone.

Apparently our results are not due to oursubjects' complying with the demand char-acteristics of the mood manipulations. First,our actual purpose was masked, so few sub-jects would have thought of the connectionof mood with the target behaviors of interest.Second, Experiments 2 and 4 showed no ef-fect of output mood on selective recall ofhappy versus sad material encoded in a neu-tral mood; yet a simple compliance hypoth-esis would have expected an effect in thosestudies.

We conclude, then, that mood causes se-lective learning of material congruent withthe reader's mood. Presumably the sameoutcome would occur with movies or audio-tapes used as the narrative input. The nextissue is to explain this enhancement of mood-congruent learning.

The mood-congruity effect is consistentwith the general theoretical framework out-lined in the introduction. There are severalmechanisms available within that frame-work to explain the mood-congruity effectin selective learning. These prospectivemechanisms are emotional intensity, moodattribution, and selective reminding. Thesemay be thought of as specialized mecha-nisms (consistent with the general theoreti-cal framework) that predict the mood-con-gruity effect.

The Emotional Intensity Explanation

A first hypothesis to explain the selectivelearning result relies on the assumption thatmemory for an event increases with theemotional intensity experienced at the timeof that event. This hypothesis has been pro-


posed for experiments by Kanungo andDutta (1966) and Dutta and Kanungo(1975), which examined people's memoryfor materials that evoked emotional reac-tions of differing intensities (see also Holmes,1970).

The relevance of the emotional intensityhypothesis to our results was suggested bysubjects' reports of mood fluctuations in Ex-periment 3. If, for example, subjects wereinstructed to maintain a steady happy mood,they experienced difficulty doing so whenthey read a sad incident. The general pointis that the experienced intensity of the moodis determined by the material being read aswell as by the posthypnotic suggestion. Ma-terial that agreed with the reader's moodmaintained or enhanced the intensity of thatmood; material that conflicted with thereader's mood evoked incompatible, oppos-ing tendencies, resulting in diminution of theintensity of the instructed mood.

The intensity hypothesis further assertsthat later memory largely reflects the en-hanced emotion experienced during encod-ing of mood-congruent material versus thediminished (or conflicted) emotion felt dur-ing encoding of mood-incongruent material.This hypothesis explains our findings, be-cause in our experiments it is plausible toassume that intensity of the felt emotion var-ied according to the congruence of the ma-terial with the subject's instructed mood.The hypothesis does not say why emotionalintensity of an event enhances its memora-bility; however, an explanation may be basedon differential rehearsal, consolidation rate,allocation of processing resources, or seman-tic elaboration. An interesting question iswhether the hypothesis requires that the sub-ject attribute his or her emotional state tothe learning material or whether the influ-ence of affect intensity occurs even if thelearning material is not viewed as causingthe emotion. Dutta and Kanungo (1975) didnot address that point.

The data needed to test the emotional in-tensity hypothesis were not collected in ourexperiments. One test would simply have ourmoody subjects rate the intensity of theiremotional reactions to the several incidentsas they read them; then we would checkwhether these intensity ratings correlatewith later recall of the incidents. An alter-

native test would use posthypnotic cues tosignal subjects to experience different levelsof emotional intensity as they are encodingblocks of text or learning materials; then onewould check whether later recall reflects thismanipulated intensity level during encoding.Research by Blum (1967) suggests this asa useful procedure. We are currently con-ducting experiments of both kinds to test theemotional intensity hypothesis of memory.

The Mood Attribution Explanation

The second hypothesis is really a refine-ment of the first one, since it proposes a par-ticular explanation (viz., motivation) of dif-ferential processing of congruent versusincongruent material. In Experiments 1 and3, given our posthypnotic suggestion, ourcompliant subjects began to feel increasinglysad (or happy) as they read the narrative.But their amnesia for this suggestion mayhave created a problem—the subjects nowhave to explain to themselves why they arefeeling so sad (or happy). In the absence ofa better explanation, suppose that they focuson the narrative itself as the source of theirsadness (or happiness). Moreover, supposethey attend selectively to that part of thenarrative that would justify the sad (orhappy) mood they are feeling. Thus, sadreaders would focus on sad material in thetext in order to justify their sadness, whichwould otherwise remain unexplained (andtherefore discomforting). Similarly, happyreaders would concentrate on happy mate-rial to justify their happiness. As a conse-quence of their selective attention to mood-congruent material, it is learned better.

This hypothesis is intriguing insofar as itlinks up with theories of self-perception andcognitive determinants of emotional feelings(see Maslach, 1979; Schachter & Singer,1962). In the Schachter and Singer experi-ment, for instance, subjects supposedly usedthe angry behavior of a stooge in responseto the frustrations of their common choresto interpret their own unexplained physio-logical arousal as the emotion of anger. Sim-ilarly, our subjects had to explain an inap-propriately exaggerated emotional reaction;this hypothesis supposes they solved thisproblem by attributing the cause of theirmood to the mood-congruent material in the


narrative, and concentrated more on it to thedetriment of the incongruent material.

This attribution hypothesis predicts thatno such selective effect of mood would occurif subjects had an acceptable explanation ofwhy they were feeling so sad (or happy)while reading the narrative. Thus, if theycould attribute their mood to a posthypnoticsuggestion or some alternative cause, therewould be no motive for selectively focusingon the mood-congruent material in the nar-rative.

We ran two conditions that provided apartial test of this prediction. First, in Ex-periment 5 all subjects read the narrativewhile hypnotized and while happy or sad,but no amnesia was suggested. Hence, allsubjects had a ready explanation for theirmood (viz., they had just been told to main-tain that mood while reading). Since theyshould have suffered no attribution problem,the attribution hypothesis would predict noselective learning in this case. However, asTable 4 showed, a substantial selective learn-ing effect was obtained nonetheless in Ex-periment 5; readers remembered more mood-congruent than mood-incongruent incidents.

A brief pilot study tested the attributionhypothesis directly. Nineteen subjects readthe Jack-Andre story, 12 made happy whilereading by a posthypnotic suggestion and 7made sad. Within each mood group, abouthalf the subjects received "source amnesia"suggestions regarding the cause of the emo-tion they were to feel while reading. Theremaining subjects were not given amnesiafor the suggestion. They were told to beaware that their emotion while readingstemmed from the earlier hypnotic sugges-tion, not from the material being read. Theexperiment replicated the basic result thatsubjects who reported identifying with happyAndre later recalled more facts about him(57%), and those identifying with sad Jacklater recalled more about him (63%); 79%of subjects recalled more facts about thecharacter with whom they identified. How-ever, this selective effect was not modulatedin any way by the presence or absence ofsource amnesia. The selective recall of factsabout the character with whom the readeridentified was of similar magnitude whetheror not the reader could explain the cause of

his or her mood. The percentage of recalledfacts favoring the identified character was62% for subjects aware of the cause of theirmood and 58% for unaware subjects. Thesefigures provide no support for the attributionhypothesis of selective learning of mood-con-gruent material.

The Selective Reminding Explanation

A third hypothesis to explain the selectivelearning results was suggested by subjects'frequent introspective reports that parts ofthe text reminded them of episodes fromtheir past. This happened particularly withthose narratives (in Experiments 3-5) thatwere basically descriptions of the characters'reminiscences of experiences of growing up.Although no systematic data were collectedon the issue, the. reports revealed a same-mood bias; sad subjects were more likely tohave a reminding experience triggered by asad rather than a happy incident in the nar-rative, and vice versa for happy subjects.Suppose that a reminding experience is anindex of effective memory processing of theinput—that is, for reasons of elaboration,vividness, or whatever, input events that re-mind us of episodes from our past will beremembered better (see, e.g., Schank, 1979).Results of Bower and Gilligan (1979) sup-port this view; their subjects recalled moreevent-descriptive phrases that reminded themof specific personal events of their lives thanphrases that failed to so remind them. Tocomplete the derivation from the selectivereminding hypothesis, because the readers'current mood causes them to have more re-minding experiences with same-mood inci-dents, they will recall more same-mood thanopposite-mood materials.

Tests of the selective reminding hypothesisare underway. One goal is to check the sup-position that the reader's current moodcauses him or her to be reminded of pastevents more often by same-mood events.Another is to check whether the past episodeaccessed by a sad (or happy) remindingevent in the text is itself sad (or happy).Third, we shall note whether, as before, textincidents that lead to reminding experiencesare themselves better recalled. Finally, thehypothesis predicts that within the subset of


text events that either did or did not remindthe moody subject of some personal episode,the effect of mood on recall will be consid-erably reduced, if not eliminated.1

We believe the mood intensity and theselective reminding hypotheses are bothlikely and complementary explanations.Should this be borne out by subsequent tests,there will still be many issues left. For ex-ample, in order for the selective remindinghypothesis to explain our data, it must becombined with the assumption of a moodbias in reminding. Since "reminding" islargely a cover term for either recognitionor cued recall of a past episodic memoryfrom a current input, the mood-bias featureof the theory amounts to assuming mood-state-dependent retrieval. Fortunately, mood-state-dependent retrieval seems well estab-lished in certain conditions (e.g., Bartlett& Santrock, 1977; Bower et al., 1978). An-other task is to explain why the intensity ofemotion aroused by a given input event in-creases memory for it. Perhaps intensity op-erates by controlling allocation of cognitiveresources to the input or causing it to beelaborated and rehearsed more. These aretopics on the research agenda.

1 Studies of reminiscence involve several methodolog-ical complications. First, "X reminds me of Y" is an ill-defined subjective judgment, since any concepts orevents are acceptable substitutes for X and Y, and sub-jects quickly learn this. Second, reminding is a "reac-tive" phenomenon; a query about one input sensitizessubjects to search memory for (or at least to report)reminiscences to later inputs. Third, with textual de-scriptions of events as inputs, the onsets of the "func-tional stimulus" and the "subjective reminiscence re-sponse" are ill defined and so latency measures havequestionable validity.

Reference Note1. Owens, J., Dafoe, J., & Bower, G. H. Taking a point

of view: Character identification and attributionalprocesses in story comprehension and memory.Speech given at the meeting of the American Psy-chological Association, San Francisco, August 1977.

ReferencesAbelson, R. P. Does a story understander need a point

of view? In R. C. Schank & B. Nash-Webber (Eds.),Theoretical issues in natural language processing(conference reports). Cambridge, Mass.: Bolt, Bera-nek, & Newman, 1975.

Anderson, J. R. Language, memory, and thought. Hills-dale, N.J.: Erlbaum, 1976.

Anderson, J. R., & Bower, G. H. Human associativememory. Washington, D.C.: V. H. Winston, 1973.

Anderson, R. C., & Pichert, J. W. Recall of previouslyunrecallable information following a shift in perspec-tive. Journal of Verbal Learning and Verbal Behav-ior, 1978, 17, 1-12.

Bartlett, J. C., & Santrock, J. W. Affect-dependentepisodic memory in young children. Child Develop-ment, 1977, 50, 513-518.

Blum, G. S. Hypnosis in psychodynamic research. InJ. E. Gordon (Ed.), Handbook of clinical and ex-perimental hypnosis. New York: Macmillan, 1967.

Bower, G. H. Experiments on story comprehension andrecall. Discourse Processes, 1978, /, 211-231.

Bower, G. H. Mood and memory. American Psychol-ogist, 1981, 36, 129-148.

Bower, G. H., & Gilligan, S. G. Remembering infor-mation related to one's self. Journal of Research inPersonality, 1979, 13, 420-432.

Bower, G. H., Monteiro, K. P., & Gilligan, S. G.Emotional mood as a context for learning and recall.Journal of Verbal Learning and Verbal Behavior,1978, 17, 573-585.

Collins, A. M., & Loftus, E. F. A spreading activationtheory of semantic processing. Psychological Review,1975, 82, 407-428.

Collins, A. M., & Quillian, M. R. Experiments on se-mantic memory and language comprehension. InL. W. Gregg (Ed.), Cognition in learning and mem-ory. New York: Wiley, 1972.

Dutta, F., & Kanungo, R. N. Affect and memory: Areformulation. Oxford, England: Pergamon Press,1975.

Holmes, D. S. Differential change in affective intensityand the forgetting of unpleasant personal experiences.Journal of Personality and Social Psychology, 1970,15, 234-239.

Kanungo, R. N., & Dutta, S. Retention of affectivematerial: Frame of reference or intensity? Journalof Personality and Social Psychology, 1966, 4, 27-35.

Kintsch, W. The representation of meaning in memory.Hillsdale, N.J.: Erlbaum, 1974.

Maslach, C. Negative emotional biasing of unexplainedarousal. Journal of Personality and Social Psychol-ogy, 1979, 37, 953-969.

Pichert, J. W., & Anderson, R. C. Taking differentperspectives on a story. Journal of Educational Psy-chology, 1977, 69, 309-315.

Rothkopf, E, Z. Structural text features and the controlof processes in learning from written materials. In J.Carroll & R. Freedle (Eds.), Language comprehen-sion and the acquisition of knowledge. Washington,D.C.: V. H. Winston, 1972.

Rumelhart, D. E., Lindsay, P. H., & Norman, D. A.A process model for long-term memory. In E. Tulving& W. Donaldson (Eds.), Organization of memory.New York: Academic Press, 1972.

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

Schank, R. C. Reminding and memory organization(Research Report #170). Computer Science Depart-ment, Yale University: Cambridge, Mass.: 1979.


Tversky, A., & Kahneman, D. Availability: A heuristicfor judging frequency and probability. Cognitive Psy-chology, 1973, 5, 207-232.

Weitzenhoffer, A. M., & Hilgard, E. R. Stanford Hyp-notic Susceptibility Scale, Form C. Palo Alto, Calif.:Consulting Psychologists Press, 1962.

Appendix

Story 1: Happy Andre and Sad Jack

The following story is about two friends, Andreand Jack. Andre is celebrating his first "anniver-sary" with Nancy and couldn't be happier. Jackhas just broken up with Lynne and, needless tosay, is miserable. "Ya gotta love it!" Andrescreamed to Jack over the sound of the car radio."Spring and women have got to be the best rea-sons for getting up in the morning." He thoughtabout the intimate celebration to come later to-night. Dinner and a movie, that's all they needed."Can you believe that Nancy and I have beentogether a whole year? By the way, Jack, Nancyand I are going to the City on Saturday to theEarth, Wind and Fire concert. It should really besomething. Afterwards we're going to 'DanceYour Ass Off.' You should ask Lynne if she wantsto go, and we could double. I know where youcould get some tickets for the concert."

"Nah," Jack whined as he drove the last mis-erable block to the gym wondering why he hadcommitted himself to play tennis with Andre. Hewas really at an all time low. Lynne had finallymade it clear that she wasn't interested in himany more—right before the Bio midterm! Now,he tried to ignore the knocking sound coming fromwithin the engine of his little Triumph warninghim that it was due for an expensive trip to thegarage. He wished he could just go home, turnon the stereo, lie back and escape. He groanedsilently as he spun the car into the parking lotcutting a little too close to a parked car. Squeezinginto the only remaining space, he swore under hisbreath as he made the car buck by absentmindedlyletting up on the clutch a split second before turn-ing off the engine. He pulled his racquet out frombehind the seat and locked the car. As he reluc-tantly approached the gym, Jack brooded over hisupsetting conversation with Lynne. He was get-ting terribly depressed.

Andre just caught a glimpse of Nancy as shedrove past on Galvez Street. He couldn't suppressthe broad grin and warm glow that came over himas he waved at her. He started towards the gymfeeling bouyant and a little giddy. He and Jackhadn't played tennis in a long time but he antic-ipated a good match. They usually played a prettyeven game and, since neither of them took thecompetition too seriously, they always enjoyed astrong match. Catching a stray softball, he loftedit back to an energetic little curly-haired boy who

was playing catch with his Dad. He watched thetwo for awhile, getting caught up by the child'senthusiasm for this simple game with his father.

Jack reached the locker room first and fumbledwith the jammed lock. "These lockers must havebeen installed at the turn of the century by drun-ken morons," he thought. "If just once this stupidtin can would open on the first try, it would bea miracle." Finally it opened and Jack draggedhis tennis gear out. As he yanked on his socks, histhoughts wandered back to this morning's blow-up with Lynne. He broke a shoelace and sworeat it. He had no replacement so he just knottedit at the break. Andre should already be outsidewarming up. Jack brushed hurriedly past hisfriends in the doorway and crossed the street tothe courts.

Andre was joking with Mike when Jack arrived.Jack muttered a hello to Mike and motioned toAndre. They moved onto the court and warmedup before volleying for the service. Andre won theservice and began. The sun was very warm, addingto his slight tan, but there was enough of a coolbreeze to keep the temperature pleasant. Andrefelt himself loosen and relax as the game contin-ued. This was a perfect way to spend the after-noon. He was glad that Jack had been so willingto play tennis. The games went quickly but thatwas all right for the first day back on the courts.Surprisingly, Andre had taken five of the sixgames, but he didn't think much of it.

Jack dragged himself off the court relieved thatit was over. He felt scorched by the sun, sore allover, and out of breath. The humidity had gottento him. He wished that Andre hadn't insisted onthis foolish game. All the while, he had been dis-tracted by the girl on the next court who lookedlike his girlfriend. Oh, that's right, Lynne wasn'this girlfriend anymore. Maybe he could call her?Or would that be too pushy? Go by her place?No, she had told him to stay away. He couldn'tshare this problem with Andre because he'd surelytell Nancy. Maybe he should just wait and seewhat Lynne would do.

Andre was whistling "Singing in the Rain" andcracking jokes, as usual, when Jack reached theshower room. The soap slipped from Jack's hand,and Andre laughingly began to sing "SlipperyWhen It's Wet." What a day! Andre thought ofthe great evening ahead. No math problems to do


and a great movie in town, it was perfect. Nancywas probably already at the apartment preparingtheir shrimp Creole dinner. He wondered what heshould put into the fruit salad he was going tomake. He'd definitely make fresh whipped cream.It was going to be a real feast, especially consid-ering that it was only a Wednesday night. Heturned to Jack, "Sure you don't want those tick-ets—great seats?"

"No, I said," Jack snapped. He dried himselfquickly and whipped the towel around his waistbefore storming out of the shower. He had tosee her!

Story 2: Paul Smith

Paul Smith had been seeing a psychiatrist forseveral sessions now. They had been making someprogress together, but Paul's ambivalence in cer-tain areas of his experience still remained prom-inent. The psychiatrist had suggested hypnotic ageregression sessions as a possible therapeutic mo-dality for him, so Paul had agreed.

The first trance experience uncovered some in-teresting childhood experiences. At the age of 4,Paul remembered the happiness of playing withhis family at home. He recalled the glee and fas-cination when he rode piggyback on his father'sback, seeing the broad smile glistening on hismother's face, and the joyous laugh of his father'svoice in the background. He also recalled somesad experiences at this early age. He recalled theoverwhelming sadness when his dog was run overby a car; his grief at his grandfather's death; andthe despair of watching a hard-earned quarter slipthrough the grating of the sewer. He also recalledthe giggles of his sibling as they lay awake atnight telling jokes, the happiness of his grand-mother's face at his birthday party, and the de-light and joy present at holiday family gatherings.

In the second age regression he went back tohis early years in grammar school. He recalledthe sadness in being cut from his baseball team,and the glumness encountered upon resignedlyrealizing the inevitability of a return visit to thedreaded dentist. He also remembered the joy andexcitement in learning to spell his name properly,and the jubilation of receiving the top grade inhis class. Additionally recalled was the dismal andcheerless memory of staying inside on a gloomy,rainy day, and his crestfallen stature upon re-ceiving news of his sister's auto accident. The ses-sion concluded with a termination of the trance,and a brief discussion of the significance of thepast events to the present situation.

With the arrival of the third hypnosis session,the psychiatrist informed Paul that he was rela-

tively happy with the therapeutic progress. Paulin return stated he could feel happy, and that hecould also feel sad. An age regression trance toadolescent years was then effected. In this trance,Paul recalled the happiness which enveloped himwhen he was blissfully with his first girlfriend; healso sadly remembered her family moving to an-other town. He re-experienced the elation in scor-ing high on his SAT test, and the fantastic joy ofbeing accepted in college. Other memories in-,eluded the sadness of departing from his highschool friend, and his sorrowful attendance at hisgrandmother's funeral. These experiences wereinterspersed with others that included a jubilantback-packing outing in the mountains, a fun-filledbeer party with his close friends, and a despair-filled evening that resulted from a sorrowful re-jection by a steady date. After several hours, thetrance session was terminated and Paul left witha mixture of feelings.

He arrived somewhat depressed for the nextsession, but reported he had been generally happyduring the week. After brief casual conversation,an age regression trance was once again utilized.The memories recalled in this session involvedPaul's early teenage years. He remembered theelation in hearing his first rock and roll albumand the high he felt in going to his first rock con-cert. He also recalled the despondency experi-enced in hearing rumors of the breakup of theBeatles, the despair of not obtaining tickets to aRolling Stones concert, and the sadness in dis-covering a warp in a newly purchased double al-bum. Other memories quickly flashed through hisawareness: the delight of meeting an old friend,the jubilation of a last-second victory in a footballgame, and the hilarious performance of a stagecomedian in a night club. The thoughts continuedat a quick but natural rate: the flunking of animportant final exam; the grief in his best friend'svoice when he informed Paul of his rejection bycollege admissions committees; the remorseful-ness after losing his allowance; and the over-whelming sorrow in hearing his mother had de-veloped cancer. Paul experienced himself as apassive but involved observer to these fleeting in-cidents. However, as the session ended, he knewa shift inside of him had occurred. He walked outof the office in deep internal processes.

When Paul returned to the office several dayslater, he was in the same state as when he left thelast session. They discussed the various outcomesof the age regression sessions, and decided thatthe therapeutic investigation would continue, butno longer using hypnosis.

Received February 9, 1981

Journal of Experimental Psychology: Generalgbower/1981/selectivity...Journal of Experimental...

Documents

Transcript of Journal of Experimental Psychology: Generalgbower/1981/selectivity...Journal of Experimental...