Retrieval of Attitude-Relevant Information From Memory: Effects on

Journal of Personality and Social Psychology Copyright 1982 by the American Psychological Association, Inc.1982, Vol. 42, No. 5, 798-810 0022-3514/82/4205-0798$00.75

Retrieval of Attitude-Relevant Information FromMemory: Effects on Susceptibility to Persuasion

and on Intrinsic Motivation

Wendy WoodUniversity of Wisconsin—Milwaukee

A distinction was drawn between (a) classic views of attitudes as stable dispo-sitions based on beliefs and prior experiences accessed from memory and (b) theself-perception analysis of attitudes as relatively transitory responses dependenton current contextual cues. Access to relevant information in memory was as-sessed by the number of issue-related beliefs and prior experiences subjects couldretrieve. To the extent that subjects could not retrieve information, they wereexpected to change their opinions to be consistent with new information andexperiences concerning the attitude topic. In the first experiment, subjects reada counterattitudinal message. As expected, subjects with little access to beliefsand prior experiences, in comparison to those with greater access, changed theiropinions to be more consistent with the message position. In the second exper-iment, subjects agreed to advocate a proattitudinal position for either a rewardor no reward. As expected, subjects with access to little relevant information inmemory inferred their attitudes from their decision to proselytize: Rewarded(vs. unrewarded) subjects explained their decision less in terms of their ownbelief in the advocated position and therefore inferred they were less in favorof this topic. In contrast, subjects who had greater access to information indi-cating that they were in favor of the stance taken in the message tended toattribute their decision to a belief in the topic, regardless of the presence orabsence of the reward, and thus remained relatively favorable.

Social psychological research on attitudes analyses (e.g., Fishbein & Ajzen, 1975) as-appears to stem from at least two traditions, sume that attitude judgments are a functionOne of these uses the attitude construct in of one's learned response to the attitude ob-a manner analogous to traits, emphasizing ject. A number of such approaches (e.g.,the enduring dispositions that underlie an Krech, Crutchfield, & Ballachey, 1962)attitude judgment. For example, McGuire's have focused on how attitudes are derived(1969) well-known review employed All- from various informational components thatport's (1935) definition of an attitude as "a comprise affective, cognitive, and conativemental and neural state of readiness to re- dimensions. Relevant information from eachspond, organized through experience, exert- of these domains is presumably stored ining a directive and/or dynamic influence on memory and accessed during the judgmentbehavior" (McGuire, 1969, p. 142). Consis- process.tent with this view, information processing According to the second tradition of re-

search, exemplified by self-perception theory(Bern, 1972), attitude judgments can rep-

This article is based on a dissertation submitted in resent on-the-spot assessments of one's re-partial fulfillment of the requirements for the doctoral cgnt behavior toward the attitude object anddegree at the University of Massachusetts—Amherst. ., . . . i_- i. ...i. «_ i_ • j

The author would like to thank Henry Bates, Shelly the context m. whlch the behavior occurred.Chaiken, Marshall Dermer, Lorraine Dorfman, and It is assumed that people often do not haveCarl Kallgren for their comments on an earlier draft of direct access to internal dispositions (e.g.,the manuscript, as well as chairperson Alice H. Eagly attitudes, emotions) and, to the extent thatfor her help in all aspects of this study. th , . u acce,s must infer jnternal

Requests for reprints should be sent to Wendy Wood, mev lacK sucn acc*ss> ""*« inter internalDepartment of Psychology, University of Wisconsin, states from external cues. Whether attitudesMilwaukee, Wisconsin 53201. derived from current cues affect future judg-

RETRIEVAL OF INFORMATION 799

ments and behavior presumably depends onthe accessibility of the initial attitude andthe information on which it was based. Thetransitory nature of such judgments is sug-gested by subjects' tendency to recall priorattitudes as similar to their current opinions,even when they have recently shifted to newattitude positions (Bern & McConnell, 1970).Advocates of self-perception theory, then,tend to view attitudes as self-descriptivestatements reflecting currently available ex-ternal cues, rather than as stable dispositionsdrawn from such internal data as one's be-liefs and prior experiences.

These two conceptualizations of attitudescan be incorporated into a comprehensivetheoretical framework that recognizes theimpact of both internal and external cue in-formation on attitude judgments (Chaiken& Baldwin, 1981; Eagly & Himmelfarb,1978).1 Consistent with this approach, it isassumed that attitudes reflecting primarilyinternal data are derived from a relativelyconstant, stable source of information thatessentially does not change with the situa-tion. Certainly aspects of these data maybecome more or less salient (see Fishbein& Ajzen's [1975] discussion of salient be-liefs) given certain situational constraints,but situationally dependent cues are not theprimary source of information on whichthese attitudes are based. In contrast, whenlittle attitude-relevant information can beretrieved from memory, an attitude judg-ment must be derived from other factors,such as perceived social desirability, recentlyreceived arguments supporting a particularposition, or a recent behavior and the contextin which it occurred. To the extent, then,that relevant internal data cannot be re-trieved, attitudes should be relatively re-sponsive to current cues concerning the ap-propriate position to take with respect to theattitude object.

The Present Research

The present research examined the rela-tion between subjects' retrieval of attitude-relevant data from memory and their opinionchange in response to new information con-cerning the attitude issue. The research firstassessed subjects' retrieval of beliefs and

prior experiences relevant to the attitudetopic, preservation of the environment. Tomeasure access to beliefs, subjects weregiven two minutes to list characteristics andattributes of preservation. Similarly accessto previous experiences was measured by theextent to which subjects could list relevantprior actions. The time limit restricted sub-jects to indicating only easily accessible be-liefs and behaviors, not the extent of theirknowledge concerning preservation. This as-sessment technique is compatible with recenttheorizing in the area of social cognition(e.g., Taylor & Crocker, 1981; Wyer &Carlston, 1979). According to this perspec-tive, access to social information in memoryis a function of integrated knowledge struc-tures, or schemata, which organize the datain a way that facilitates retrieval.

In the second stage of this research, sub-jects' responses to new information from oneof two sources were examined: (a) a persua-sive message or (b) a recent behavior andthe context in which it occurred.

Susceptibility to Persuasion

According to a cognitive response analysis(e.g., Greenwald, 1968; Petty & Cacioppo,1979), message persuasiveness is a functionof the nature of thoughts recipients generatein response to the communication. Attitudesderived from one's own supporting beliefsand experiences may be little affected by acounterattitudinal message because infor-mation is available for the effective gener-ation of counterarguments to the materialpresented. Individuals who have little accessto such attitude-relevant information in

1 Differences between the self-perception analysis andmore traditional analyses of attitudes cannot be ex-plained by assuming that they address different stagesof the attitude formation process. Kelley (in Harvey,Ickes, & Kidd, 1978) suggested that participants in self-perception studies may have had little opportunity toformulate previously an opinion about the attitude ob-ject, and thus they are affected by the context in whichthe first encounter occurs. Self-perception research,however, that has been conducted with familiar attitudeobjects, such as drawing pictures (Lepper, Greene, &Nisbett, 1973), and research that has provided partic-ipants with prior experiences with the attitude object(Greene, Sternberg, & Lepper, 1976) have still foundattitudes to be a function of contexual cues.

800 WENDY WOOD

memory may be less resistant to persuasionbecause they cannot effectively counterarguethe message and, instead, may generate pri-marily favorable responses. Thus to the ex-tent subjects could retrieve relevant datafrom memory, they were expected to gen-erate more counterarguments and fewerthoughts favorable to the message and there-fore to show less opinion change.

Intrinsic Motivation

Research on intrinsic motivation has com-monly found that attitudes are affected byrecent behavior and the context in which itoccurs (e.g., Condry, 1977; Deci, 1971).Typically in this research, subjects performan attractive task for either a reward or noreward, and rewarded (vs. unrewarded) sub-jects are subsequently found to be less fa-vorable toward the task. These findings arecommonly interpreted in terms of the dis-counting principle (Kelley, 1972), whichsuggests that unrewarded subjects indicatea relatively favorable opinion because theyhave only their liking for the task as a plau-sible explanation for task performance. Re-warded subjects, however, have two poten-tial explanations, a favorable opinion and thereward. The presence of the reward lessensthe plausibility of a favorable attitude as thecause, and thus rewarded subjects indicateless favorable opinions. Interpretation of theopinion change findings in terms of subjects'attributions for their behavior has remainedtentative, however, because researchers haverarely assessed the supposed attributionalmediators. One study that directly measuredsubjects' attributions did not obtain clearlinkage among the presence or absence ofthe reward, attributions, and opinion change(Kruglanski, Alon, & Lewis, 1972).

It was anticipated that people with accessto attitude-relevant data in memory, in com-parison to those with little access, would beless affected by current cues such as a recentbehavior and the context in which it oc-curred. In support of this analysis, rewardingsubjects for task performance does not ap-pear to reduce their favorability toward thetask if subjects are reminded that they pre-viously performed the task for no reward(Fazio, 1981).

To further explore the relation betweenaccess to internal cues and intrinsic moti-vation, the present study used a procedureadapted from Kiesler, Nisbett, and Zanna(1969), in which subjects were rewarded ornot rewarded for deciding to deliver proat-titudinal arguments. Subjects who could re-trieve data indicating that they were in favorof the advocated position were expected toattribute their decision to a belief in the is-sue, regardless of whether they received areward. Explanations in terms of belief wereexpected to lead both rewarded and unre-warded subjects to maintain their initiallyfavorable opinions on the message topic.However, subjects with little access to in-formation indicating that they were in favorof the advocated position were expected toexplain their behavior in terms of currentlyavailable cues. These subjects were expectedto exhibit the classic intrinsic motivationeffect: Rewarded, compared with unre-warded, subjects were expected to attributetheir behavior more to the reward and lessto their belief in the issue and thus to showgreater opinion change away from the ad-vocated position.

Method

OverviewSubjects participated in a two-session study ostensibly

concerned with attitude assessment. In the preliminarysession subjects completed questionnaires assessing theiropinions and other reactions to seven social issues, in-cluding the message topic, preservation of the environ-ment. One to 2 weeks later, subjects returned to par-ticipate in one of the following experiments. In thepersuasion study, subjects read a counterattitudinalmessage arguing against preservation and then indicatedtheir final opinions on this topic. In the intrinsic moti-vation study, subjects agreed to present a proattitudinalmessage arguing in favor of preservation for a $5 rewardor no reward and then indicated their opinions.

Preliminary SessionSubjects

One hundred sixty-six University of Massachusettspsychology students participated for extra course credit.Seven of these were eliminated because they did notcomplete the second half of the experiment.

ProcedureSubjects participated in the first session in groups of

about 12. They completed questionnaires assessing their


opinions and other responses to seven social issues, in-cluding preservation of the environment (see below). Inaddition, subjects provided background informationabout themselves (e.g., sex, class). They were then as-signed to participate in one of the experiments.

Measuring Instruments

Opinions. Subjects indicated their initial opinions onthe topic "preservation of the environment" on a 15-point scale anchored by "very favorable" and "very un-favorable."

Self-perception of previous reactions. On 15-pointscales, subjects indicated how frequently in the past fewyears they had thought about preservation of the en-vironment, taken some action in regard to it, and hadpositive or negative feelings about it.

Self-perception of knowledge. Subjects rated on 15-point scales (a) how well-informed they were concerningpreservation and (b) how frequently in the past fewyears they had engaged in specific information gatheringbehaviors, such as talking with others about the topic,reading articles and books on it, taking relevant courses,and watching TV programs on it. Ratings of the specificbehaviors were averaged, and each subject was assigneda mean score. Subjects' ratings of how well informedthey were proved to be highly correlated with this meanscore (r = .71), and the two measures were averagedinto an index representing subjects' knowledge aboutthe topic.

Self-perception of involvement. Subjects rated ontwo 15-point scales how important preservation was tothem and how involved they were in the topic. Responsesto these two items were highly correlated (/• = .62) andwere averaged into an index representing degree of in-volvement.

Belief retrieval. To determine the ease with whichsubjects could retrieve attitude-relevant beliefs, theywere asked to list on a questionnaire the characteristicsand facts they believed to be true about several issues,including preservation of the environment.2 One opiniontopic was listed at the top of each page, and six boxeswere provided underneath. Subjects were told to writeonly one belief in each box and to leave blank boxes ifthey had less than six beliefs to list. Several examplesof beliefs about noncrttical topics were provided. Sub-jects were then given 2 min. to list their beliefs abouteach topic. The number of discrete beliefs each subjectlisted about preservation of the environment was judgedby two independent raters (r = .91); disagreements wereresolved by discussion. In addition, to allow an analysisof the relation between access to beliefs concerning pres-ervation and access to beliefs on other topics, the numberof discrete beliefs that subjects listed concerning twoadditional issues, psychological research and right toabortion, were judged by two raters (rs = .94 and .88,respectively).

Behavior retrieval. Subjects' retrieval of attitude-relevant experiences was assessed in a manner similarto the belief retrieval task. Subjects were asked to listspecific instances of times when they had engaged inactions related to each topic. The number of discretebehaviors that each subject listed about preservation ofthe environment, psychological research, and right to

abortion were judged by two raters (rs = .89, .90, and.91, respectively).

Group membership. Subjects indicated whether theycurrently belonged to a number of organizations, in-cluding environmental groups such as the Sierra Clubor the Audubon Society.

First Experiment: Susceptibilityto Persuasion

SubjectsSixty-five subjects returned in groups of about 12 to

participate in this experiment.

Procedure

Subjects again expected to indicate their opinions ina variety of formats. The rationale, adapted from Jonesand Brehm (1967), for preceding the opinion question-naire by a persuasive message was that being exposedto someone else's opinion and the arguments he or sheuses to support this opinion gets people in the "rightframe of mind to be critical and careful about evaluatingtheir own opinions" and, therefore, makes it possible tomeasure their opinions more accurately.

The experimenter distributed a handout explainingthat each subject would read a transcript of an interview(actually hypothetical) that had been tape-recorded aspart of an opinion survey conducted on campus. Par-ticipants in this survey, including students, faculty, staff,and visitors, had (supposedly) been asked to give anopinion on an issue and then to support that opinionwith evidence. The handout stated that over 100 dif-ferent interviews covering 10 topics were available andthat by random selection, almost everyone would get adifferent interview to read. In addition, participantswere told that since the interviews represented a widesampling of opinions, they might read an interview inwhich the opinion expressed was different fromtheir own.

The experimenter then gave each subject the inter-view transcript, which contained the persuasive message.The transcript began with an interviewer asking an in-terviewee (source), Jim H., for some background infor-mation. Jim H. was portrayed as a graduate student inbiology who was interested in the issue of environmentalpreservation. In response to the interviewer's question,

2 It is important to consider the effect that completingthe retrieval tasks may have had on subjects' opinions.For example, subjects could have changed their opinions(after the preliminary assessment) to be consistent withthe numbers of beliefs and behaviors they listed in theretrieval tasks. Subjects were unlikely, however, to haveconducted this kind of detailed analysis of their re-sponses to preservation of the environment. Items con-cerning preservation were embedded in questionnairesmeasuring responses to six other social issues (some ofwhich were sensitive in nature, such as abortion), andthe 1- to 2-week delay between sessions further mini-mized any impact of the retrieval tasks on subjects'opinions in the second session.

802 WENDY WOOD

Jim stated that, "I am not very strongly in favor ofcurrent efforts to preserve our environment. . . we haveto recognize that preservation has negative effects." Jimthen went on to state the following arguments againstpreserving the environment: (a) preservation has a neg-ative impact on the economy; (b) the energy problemjustifies lowering environmental standards to allow theburning of coal; (c) preserved land is needed for housingand farming; (d) preservation is unnecessary becauseit is possible to reclaim polluted areas.

After allowing about 6 min. for reading the tran-scripts, the experimenter distributed a questionnaire onwhich subjects stated their opinions on a variety of socialissues. Two of these issues, including preservation of theenvironment, were identical to the ones subjects ratedin the preliminary session. The experimenter then ex-plained that she was also interested in subjects' reactionsto the interviews. Subjects completed a questionnairethat elicited their thoughts about the interview tran-script along with other responses (see below). Finally,subjects were debriefed and excused.

Measuring InstrumentsOpinions. Subjects' final opinions concerning pres-

ervation were assessed on the same scale used to assesstheir original opinions.

Cognitive response measures. Subjects were given2.5 min. to list their thoughts about what the commu-nicator said in the message. The questionnaire, similarto that used by Petty and Cacioppo (1979), listed theinstructions at the top of the page, with seven boxesunderneath. Two independent raters judged the numberof positive (r - .93), negative (r = .80), and neutral (r =.53) thoughts each subject produced.

Perceptions of the communicator. Subjects rated thecommunicator on 10 15-point bipolar scales. Positivepoles were consistent, honest, sincere, nonopportunistic,nonmanipulative, noncompliant, open-minded, unbi-ased, objective, and likeable.

Other measures. Subjects were asked to summarizeeach argument that the communicator used to supporthis position, and two independent judges determined thenumber correctly recalled (r = .88). No significant ef-fects were obtained on this variable, so it will not bediscussed further. Subjects were also asked to writedown the overall position the communicator took in theinterview, and all but two subjects were able to recallcorrectly this position. At the end of the experiment,subjects wrote down their interpretations of the study.Based on these interpretations, seven subjects (retainedin the analysis) were assessed as being suspicious of aninfluence attempt.

ProcedureThis session was supposedly concerned with attitude

change. Its apparent purpose was to determine the op-timal number of arguments to use in a persuasive mes-sage.

The procedure was adapted from an experiment byKiesler et al. (1969). Subjects were asked to deliver totwo people on campus a number of arguments previouslyprepared by the experimenter. It was explained that anumber of students were needed as communicators sothat the specific personality characteristics of a singlecommunicator did not affect the results. Subjects be-lieved that after presenting the arguments, they wouldask the message recipients whether they were willingto sign a petition in favor of the advocated position.Subjects expected to rehearse before leaving to conductthe task.

After describing the purpose and procedure of thestudy, the experimenter (ostensibly) randomly assignedto subjects one of the topics they had rated in the firstsession. In reality everyone was asked to argue in favorof environmental preservation.

At this point it was mentioned to half of the subjects(reward condition) that they would receive $5 for agree-ing to participate in this phase of the experiment. Themoney was placed in front of these subjects and re-mained visible throughout the rest of the session. Theother half of the subjects were not offered a monetaryreward for their participation (no reward condition).Subjects were then asked if they would present the per-suasive message.

After the subject had agreed to the task, the exper-imenter remarked that it would probably be a good ideato get a measure of how the subject felt right then aboutthe message topic. Subjects responded to a questionnaireon which they gave their opinions concerning preser-vation and other responses (see below).

After completing the questionnaire, subjects werequestioned for suspicion, then debriefed and excused.Those in the reward condition received $5.

Measuring Instruments

Opinions. Subjects indicated their opinions on pres-ervation of the environment on the opinion scale de-scribed previously.

Attributions. On 15-point scales subjects rated theimportance of several reasons for agreeing to persuadeothers to sign the petition: (a) receiving experimentalcredit or payment, (b) convincing others about a topicthe subject really believed in, and (e) any other' reasonhe or she cared to mention.

Second Experiment: Intrinsic Motivation

SubjectsNinety-four subjects returned individually to partic-

ipate in this experiment. One of these was eliminatedbecause she declined to deliver the persuasive message.Three more were eliminated because they did not believethey would actually deliver the persuasive arguments.

Results and Discussion

The number of beliefs subjects indicatedconcerning preservation of the environmentranged from 2 to 7, with a mean of 3.75, andthe number of behaviors ranged from 0 to6, with a mean of 2.83. Median splits wereperformed on both variables (medians =


3.70 and 2.73, for beliefs and behaviors, re-spectively) and Number of Beliefs Retrievedfrom Memory (few vs. many) X Number ofBehaviors Retrieved (few vs. many) analysesof variance were calculated, along with ap-propriate contrasts.3

Self-Perception of Past Experiences,Knowledge, and Involvement

Subjects' perceptions of their past ex-periences, knowledge, and involvement con-cerning preservation of the environment cor-responded to the number of behaviors theylisted. As shown in Table 1, subjects wholisted many behaviors, compared to thosewho listed few, rated that they had thoughtmore about preservation, F(l, 157) = 14.31,p < .001; had engaged in more action, F(l,157) = 12.26, p < .001; and had experiencedmore feelings, F(l, 157) = 15.67, p < .001.In addition, subjects who listed many be-haviors, compared to those who listed few,indicated that they knew more about thetopic, F(l, 157) = 20.15, p< .001, and per-ceived themselves to be more involved, F( 1,157) = 21.80, p < .001. Although no signif-icant effects on these variables were obtainedfor belief retrieval, differences between sub-jects who listed few beliefs and those wholisted many were consistently in the pre-dicted direction.

The obtained relations between retrievaland the self-perceptions are informative con-cerning the validity of the retrieval tasks.Self-perceptions were related to behaviorretrieval (and to a lesser extent to belief re-trieval) presumably because all of thesemeasures reflect, at least partially, the extentto which attitude-relevant information is or-ganized in memory in an easily accessibleform. The relation between retrieval and in-volvement will be elaborated further in thegeneral discussion.

Group Membership

Further evidence for the validity of be-havior retrieval was provided by the findingthat participants who Were members of en-vironmental groups such as the Sierra Cluband the Audubon Society listed more be*haviors (19 listed many behaviors vs. 6 listed

Table 1Preliminary Session: Mean Self-Perception ofThoughts, Behaviors, Feelings, Knowledge,and Involvement as a Function ofBehavior Retrieval

Few behaviors Many behaviorsretrieved retrieved

Measure from memory from memory

Frequency ofprevious thought

Frequency ofprevious action

Frequency ofprevious feelings

Degree of knowledgeExtent of

involvement

10.16

7.69

10.468.25

9.58

11.93

9.88

12.0410.11

11.43

Note. Higher numbers indicate more frequent thought,action, and feelings; greater knowledge; and more in-volvement concerning preservation of the environment.

few) than nonmembers (70 listed many be-haviors vs. 62 listed few), x2(0 = 4.52, p <.05. Parallel results were obtained for beliefretrieval, although they were not significant.

First Experiment: Susceptibilityto Persuasion

The hypotheses were explored by a 2 (fewvs. many beliefs) X 2 (few vs. many behav-iors) design.4

Opinions

Analysis of covariance was conducted onthe postopinions, with preopinions as the co-variate. A test for homogeneity of the co-variate regression coefficients indicated thatthe coefficients did not differ across experi-mental conditions. Analysis of variance in-dicated that preopinions were more proen-vironment in the many (M = 14.15) thanfew behaviors conditions (M - 13.22, p <

3 Due to the relatively moderate correspondence be-tween belief and behavior retrieval (r = .30), these vari-ables were employed separately in the reported analyses.

4 Analyses including subject sex as an additional vari-able in both studies yielded no differences betweenmales' and females' persuasibility and no systematicdifferences on other measures. This variable was, there-fore, not included in the reported analyses.

804 WENDY WOOD

Table 2Persuasion Experiment: Mean Postopinions asa Function of Belief and Behavior Retrieval

Number ofbeliefs retrieved

from memory

Few beliefsMany beliefs

Number of behaviors retrievedfrom memory

Few behaviors Many behaviors

5.68 3.924.33 3.26

Note. Means are adjusted postopinion scores on a 15-point scale on which higher numbers indicate greateropinion change.

.05),5 and that preopinions (M = 13.71) dif-fered significantly from postopinions (M =11.85, p< .01).

Opinion means, which are the postopinionscores adjusted on the basis of the analysisof covariance, appear in Table 2. Analysesof these data indicated that as predicted,subjects who listed many behaviors changedtheir opinions less in response to the persua-sive message than those who listed few be-haviors, F(l, 61) = 7.91,p < .01. Those wholisted a large number of beliefs changed theiropinions less than those who indicated fewbeliefs, F(l, 61) = 4.43, p< .05.

Cognitive Responses

Consistent with a cognitive response anal-ysis of persuasion (Petty & Cacioppo, 1979),subjects' cognitive responses appeared tomediate the effects of belief and behaviorretrieval on opinion change. In the first stepof this mediation, the cognitive responsesthat subjects produced were a function ofthe extent to which they could retrieve rel-evant experiences from memory. Fewer fa-vorable thoughts were generated by subjectswho listed many behaviors (M = .81), com-pared to those who listed few (M = 1.54),F(l, 61) = 5.86, p<.05. In addition, agreater number of counterarguments wereproduced by subjects who listed many be-haviors (M = 2.62) than by those who listedfew (M = 1.43), F(l, 61) - 8.10 p< .01.Behavior retrieval was not related to thenumber of neutral thoughts subjects pro-duced, and belief retrieval was not a predic-tor of any cognitive responses. Evidence forthe second step in the mediational link was

provided by the average within-cell corre-lations between favorable thoughts and ad-justed postopinions (/• = —.23, p < .10) andbetween counterarguments and postopinions(/• = .41,/><.01).

Further evidence that the cognitive re-sponses mediated acceptance of the persua-sive message was suggested by a hierarchicalregression analysis predicting opinion change.When the counterarguments variable wasentered into the analysis before belief andbehavior retrieval, belief retrieval becameonly a marginally significant predictor ofopinion change, F(l, 59) = 3.04, p<.10,and behavior retrieval became a less effec-tive predictor, F(l, 59) = 4.56, p < .05. Thefavorable thoughts measure was not a sig-nificant predictor of opinion change in theregression analysis. It thus appears that at-titudes derived from beliefs and behaviorsretrieved from memory were little affectedby the persuasive message, largely becauseinformation was available for the effectiveproduction of counterarguments to the ma-terial presented. In contrast, attitudes withlittle underlying cognitive support may havebeen influenced more by the message be-cause few counterarguments were generated.

Perception of the Communicator

A factor analysis (varimax rotation) of thesource ratings yielded three rotated factors.The factors, which accounted for 28.9%,18.2%, and 10.8% of the variance, were la-beled open-minded (open-minded, unbi-ased), sincere (honest, sincere, likeable), andobjective (objective, unbiased), respectively.Factor scores were computed for each sub-ject and then treated by analysis of variance.The source was judged more open-mindedby subjects who listed few, compared withmany, beliefs (p < .01) and as more sincereby subjects who listed few, compared with

5 To eliminate the relation in the persuasion studybetween extremity of initial opinions and behavior re-trieval, a median split was conducted on the number ofbehaviors listed at each point on the attitude scale. Incomparison to the median split procedure reported inthe text, this alternate analysis resulted in the reclas-sification of only four subjects. The alternate procedure,then, was not used in the analyses because it did notreduce the relation between extremity and retrieval.


many, behaviors (p < .01). No effects wereobtained in the analysis on the objectivefactor.

The source ratings that failed to loadhighly on any of these factors (i.e., consis-tent, nonopportunistic, nonmanipulative, andnoncompliant) were analyzed separately,and only the analysis of the nonmanipulativescale yielded significant effects. Subjectswho listed few, compared with many, beliefsperceived the source to be less manipulative(p < .05).

Subjects tended to change their opinionsto the extent that they perceived the sourceas open-minded and nonmanipulative (rs =.24 and .26, respectively, ps<.10). How-ever, because these perceptions were not sig-nificant predictors of opinion change whenentered into a hierarchical regression anal-ysis before belief and behavior retrieval, per-ceptions did not mediate the effects of re-trieval on opinion change.

differed significantly from postopinions (M =12.81, p<. 05).

Opinion means, which are the postopinionscores adjusted on the basis of the analysisof covariance, appear in Table 3. The resultsreplicated the findings typically obtained inself-perception research: Subjects who re-ceived a reward changed their opinions tobe less favorable toward preservation of theenvironment than subjects in the no-rewardconditions, F(l, 79) = 7.66, p < .01. Further,the findings supported the present hy-potheses. Subjects who indicated few behav-iors changed their opinions more than thosewho indicated many, F(l, 79) = 17.22,/x.OOl. More importantly, a Reward XNumber of Behaviors interaction, F( 1,79) =6.87, p < .05, indicated that the differencebetween the reward and the no-reward con-ditions was significant for subjects who listedfew behaviors, F(l, 79) = 17.59, p < .001,but not for those who listed many (F < 1).

Second Experiment: Intrinsic Motivation Attributions

The hypotheses were explored by a 2 (fewvs. many beliefs) X 2 (few vs. many behav-iors) X 2 (reward vs. no reward) design.

Opinions

Similar to the persuasion experiment,analysis of covariance was conducted on thepostopinions with preopinions as the covari-ate. A test for homogeneity of the covariateregression coefficients indicated that thecoefficients did not differ across experimen-tal conditions, Analysis of variance revealedthat preopinions did not vary across exper-imental conditions and that they (M = 14.03)

Rewarded subjects attributed their deci-sion to deliver the persuasive argumentsmarginally more to experimental credit orpayment (M = 11.26) than unrewarded sub-jects (M = 10.48), F(l, 79) = 3.08, p < .10.

Consistent with an attribution analysis ofself-perception (Nisbett & Valins, 1972),subjects' attributions to belief appeared tomediate the effects of the reward manipu-lation and retrieval on opinion change. Inthe first step of this mediation, subjects' at-tributions to belief were a function of thereward manipulation and the extent of theirretrieval (see Table 3). Subjects who listedmany, rather than few, beliefs attributed the

Table 3Intrinsic Motivation Experiment: Mean Postopinions and, Attributions to Belief as a Function ofBehavior Retrieval and Reward

Few behaviors retrievedfrom memory

Many behaviors retrievedfrom memory

Measure Reward No reward Reward No reward

Adjusted postopinionsAttribution of decision to belief

4.549.67

3.0511.05

2.7611.84

2.7710.60

Note. Higher numbers indicate greater opinion change and greater attribution to belief.

806 WENDY WOOD

decision to deliver the arguments marginallymore to belief in preservation (p < .10).More importantly, a Reward X Number ofBehaviors interaction, F(l, 79) = 8.20, p<.01, revealed that consistent with the opinionchange findings, in the few-behaviors groups,subjects who were not rewarded attributedthe decision more to their belief in the issuethan those who received a reward, F(l, 79)= 3.81, p< .05. In the many-behaviorsgroups, however, rewarded subjects unex-pectedly made a marginally stronger attri-bution to their belief than unrewarded sub-jects, F(l, 79) = 3.62, p < .10. This resultis not reflected in the opinion change find-ings. Evidence for the second step in themediational link was provided by a small butsignificant average within-cell correlationbetween attribution to belief and adjustedpostopinions (r = .23, p < .05).

Further evidence of the mediational na-ture of subjects' attributions to belief wassuggested by a hierarchical regression anal-ysis predicting opinion change. When attri-bution was entered into the equation beforebehavior retrieval, the reward manipulation,and the Reward X Behaviors interaction, theresults indicated that the interaction was nolonger a significant predictor of opinionchange, F(l, 82) = 2.63, p < .15, and thereward manipulation and behavior retrievalbecame less effective predictors, F(l, 82) =3.51, p < .10, and F(l, 82) = 6.62, p < .05,respectively.

In sum, the reward and the retrieval ofprior experiences appear to have affectedopinion change largely through the media-tion of subjects' explanations for their de-cisions. Subjects who had access to priorexperiences with preservation of the envi-ronment were able to draw on this infor-mation when constructing their explanationsand thus, regardless of the presence or ab-sence of the reward, explained their decisionin terms of a belief in the advocated position.Attribution to belief, then, tended to lead tothe maintenance of a favorable opinion. Incontrast, the mediation of opinion change forsubjects without access to relevant experi-ences was consistent with the attributionanalysis of attitude inference from behaviorand contextual cues (Nisbett & Valins,1972). Those who were not rewarded had

only one plausible explanation for their de-cision, belief in preservation, and they main-tained a relatively favorable opinion. Forrewarded subjects, however, the presence ofthe additional cause, the reward, moderatedexplanations in terms of belief in preserva-tion. These subjects, then, changed theiropinions to be relatively less favorable.

General Discussion

The two experiments support the hy-potheses. Individuals who had access to at-titude-relevant information in memory ap-parently used this data to assess howfavorable they were toward the attitude is-sue. In contrast, individuals who did not haveeasy access to such information in memoryrelied relatively more on the currently avail-able cues provided by the persuasive messageand the decision to proselytize.

Subjects' degree of access to relevant in-formation in memory proved not to have adirect impact on attitude change, but ratherto have been, at least in part, mediated byother processes. In the persuasion study, ac-cess to relevant information enabled subjectsto produce counterarguments to the persua-sive message, and the production of coun-terarguments then tended to lead to lessopinion change. In the intrinsic motivationstudy, retrieval of prior experiences enabledsubjects to explain a recent behavior in termsof their orientation toward the attitude ob-ject, rather than other factors. Explanationin terms of belief in the issue, then, tendedto lead to the maintenance of a favorableopinion.

In general, retrieval of beliefs and priorexperiences seemed to provide subjects withdata to evaluate new information concerningthe attitude issue, and it was the result ofthis analysis that determined whether sub-jects changed their opinions to be consistentwith the new information. The specific na-ture of subjects' evaluations differed in thetwo experimental settings, in one study in-volving the generation of cognitive responsesand in the other attributions for behavior.The manner in which subjects combine datafrom internal and external sources to arriveat an attitude judgment is likely to takemany forms, including potentially more for-


mal cognitive integration processes (e.g.,Wyer & Hartwick, 1980) in addition tothose identified in the present research. Yetit seems likely that the information utilizedin a variety of attitude judgment processeswill be a function of one's access to relevantinformation in memory.

Adequacy of Belief and BehaviorRetrieval Measures

Access to attitude-relevant informationwas operationalized in terms of the numberof beliefs and behaviors subjects indicatedin the listing tasks. The analysis consistentlyyielded the expected relations among behav-ior retrieval, opinion change, and the me-diators of opinion change, but only in a fewinstance's did belief retrieval yield the pre-dicted effects. There are several possible ex-planations for these results.

First, it may be that behavior retrieval isin general a more important contributor toopinions than belief retrieval. Indeed, it hasbeen argued that attitudes based on the in-formation obtained through direct experi-ences with the attitude object are relativelywell defined and held confidently (Fazio &Zanna, 1981; Zanna, Olson, & Fazio, 1980).Attitudes based primarily on informationobtained through less direct encounters (i.e.,information acquired second hand), how-ever, are thought to be less clear and lessconfidently held. The greater importance ofbehavior retrieval can also be understood interms of a general distinction between twotypes of memory stores. Long-term storageis sometimes thought to be composed of asemantic memory, which contains general-ized knowledge about objects and concepts,and an episodic memory, which consists ofa more concrete record of personal experi-ences (Tulving, 1972). It has been suggestedthat attitudes may commonly be derivedfrom episodic-type information, such as one'sprior experiences, rather than abstract cog-nitions, such as one's beliefs (Abelson, 1976).

The disappointing performance of beliefretrieval could also be explained in terms ofits being a less effective indicator of accessthan behavior retrieval. That subjects listed,on the average, a greater number of beliefsthan behaviors might suggest that the act of

retrieving beliefs spontaneously resulted innewly perceived relations between the atti-tude issue and other constructs stored inmemory. The retrieval of behaviors, how-ever, may be less likely to generate sponta-neously newly perceived instances becauserecall of specific examples of prior experi-ences is required. The measure of belief re-trieval may, thus, have contained a greaterdegree of error than behavior retrieval.

Finally, the relative effectiveness of beliefand behavior retrieval in predicting opinionscould depend on the attitude issue involved.Day-to-day life offers many opportunities toengage in actions toward the environment,such as decisions about whether to litter andwhether to recycle. Attitudes toward suchissues may, thus, be relatively dependent onbehavior retrieval. Other issues, however,such as tax reform, may offer little oppor-tunity for action, and attitudes toward thesetopics may depend relatively more on accessto beliefs.

Retrieval of Topic-Specific Information

The belief and behavior retrieval measureswere assumed to tap subjects' access totopic-specific information, not to assess ageneral cognitive style involving easy or dif-ficult access to information in memory. Thisassumption was explored by examining therelation between subjects' opinion change onpreservation of the environment and the ex-tent of their retrieval of information on twoadditional issues, psychological research andright to abortion.6 Opinion change on pres-ervation was not predicted from subjects'access to abortion-related information, andonly in the persuasion study was access toresearch-related information a predictor ofpreservation opinions: Although subjects whoindicated few research behaviors changedtheir opinions more than those who indicatedmany (p < .05), this effect was weaker thanthe prediction of opinion change from be-

6 To determine the relation between opinion changeon preservation and retrieval on other issues, mediansplits were conducted on the number of beliefs and thenumber of behaviors subjects listed about psychologicalresearch and right to abortion. For each issue, then,analyses were calculated with low versus high belief re-trieval and low versus high behavior retrieval.

808 WENDY WOOD

haviors relevant to preservation. In general,then, these results suggest that attitudechange was a function of the retrieval ofattitude-relevant data, rather than generalaccess to information in memory.

Retrieval and Involvement

The present research takes an informationprocessing approach to the often investigatedproblem of how attitude change varies withthe depth or intensity of the attitude. Earlierresearch indicated that resistance to coun-terattitudinal information is conferred byprior behavioral commitment to the attitudeposition (Kiesler, 1971), a high level of con-nectedness or centrality of the target beliefin relation to subjects' other beliefs (Ro-keach, 1968), and high ego-involvement inthe message topic (Sherif & Hovland, 1961).Although these analyses have been con-cerned primarily with factors that induceresistance to counterattitudinal persuasivemessages, it has been suggested that involve-ment also increases the probability of ac-cepting or assimilating proattitudinal mes-sages (Pallak, Mueller, Dollar, & Pallak,1972).

Given that in the data from the prelimi-nary session of the present research, sub-jects' ratings of their own involvement werecorrelated with extent of retrieval, it couldbe argued that the observed eifects of re-trieval on opinion change were actually dueto involvement. The attenuated opinionchange obtained in the persuasion experi-ment when subjects had access to relevantdata in memory is consistent with such aninterpretation. Yet subjects' responses toproattitudinal information in the intrinsicmotivation study are less clearly interpret-able in terms of involvement. Since subjectswith access to prior experiences tended tochange their opinions least, the results of thissecond study certainly cannot be interpretedin terms of any tendency for involvement toenhance the acceptance of proattitudinaldata. Traditional perspectives focusing onglobal concepts such as involvement are thusunable to account adequately for the presentresults.

More recent work has suggested that to

understand the relation between involvementand opinion change, it is fruitful to considerthe information processing consequences ofinvolvement (Chaiken, 1980). For example,an analysis that considers subjects' thinkingabout the content of the persuasive com-munication can account for involvement'senhancing or inhibiting persuasion. Sinceinvolvement increases thinking (as indicatedby cognitive responses), involvement can in-crease persuasion when a message containshigh quality argumentation (Petty & Ca-cioppo, 1979). Hence involvement has nonecessary relation to opinion change, ratherits effects seem to depend on such mediatingprocesses as subjects' extent of thought and,potentially, degree of access to relevant in-formation.

An information processing perspective,then, suggests that any relation between in-volvement and opinion change in the presentresearch is understandable only in terms ofinvolvement's effects on cognitive mediatorssuch as retrieval. Since involvement provedto be related to opinions in a manner similarto retrieval,7 it most likely enhanced accessto relevant data in memory and lessened re-liance on external cues. Indeed, high levelsof involvement have been found to facilitateprocessing of self-related data (Markus,1977). In general, then, the crucial infor-mation processing determinant of opinionchange, access to relevant information, wasmeasured directly in the present research.Global factors such as involvement are morecausally remote predictors of opinions and

7 To determine the relation between involvement andopinions, median splits were conducted on self-percep-tions of involvement to classify subjects according tohigh or low involvement. In the persuasion experiment,subjects who perceived themselves as involved changedtheir opinions less than those who perceived themselvesas uninvolved (p<.001). In the intrinsic motivationexperiment, involved subjects again changed their opin-ions less than those uninvolved (p<.001), and unre-warded subjects changed their opinions less than thoserewarded (p < .05). A Reward X Involvement interac-tion (p < .01) further indicated that for uninvolved sub-jects, presence (vs. absence) of the reward increasedopinion change away from the advocated position (p <.001), but the reward manipulation had little effect onthe opinions of involved subjects.


taken alone cannot account for the specificconfiguration of opinion change findings.

Conclusion

The findings of the two studies highlightthe importance of distinguishing betweentwo types of attitudes: (a) attitudes derivedfrom prior experiences and beliefs stored inmemory and (b) those derived from cur-rently available external cues such as per-suasive arguments or a recent behavioral in-cident. In both the persuasion and intrinsicmotivation experiments, subjects' opinionchange proved to be a function of the degreeto which they had access to relevant infor-mation in memory. This focus on informa-tion retrieval considers only the most im-mediate antecedents of opinion change. Theorganization and storage of information forfuture use, however, can be considered amultistage process, of which information re-trieval is only one aspect. A comprehensiveanalysis of the process of attitude judgmentswould consider information processing fromthe initial stage of attention to information,to subsequent stages of comprehension andstorage, and ultimately to retrieval. Futureresearch may profitably consider the impli-cations for attitude judgments of the infor-mation processed at each of these stages.

References

Abelson, R. P. A script theory of understanding, atti-tude, and behavior. In J. Carroll & T. Payne (Eds.),Cognition and social behavior. Hillsdale, N.J.: Erl-baum, 1976.

Allport, G. W. Attitudes. In C. Murchison (Ed.), Hand-book of social psychology. Worcester, Mass.: ClarkUniversity Press, 1935.

Bern, D. J. Self-perception theory. In L. Berkpwitz(Ed.), Advances in experimental social psychology(Vol. 6). New York: Academic Press, 1972.

Bern, D. J., & McConnell, H. K, Testing the self-per-ception explanation of dissonance phenomena: On thesalience of premanipulation attitudes. Journal of Per-sonality and Social Psychology, 1970, 14, 23-31.

Chaiken, S. Heuristic versus systematic informationprocessing and the use of source versus message cuesin persuasion. Journal of Personality and Social Psy-chology, 1980, 39, 752-766.

Chaiken, S., & Baldwin, M. W. Affective-cognitive con-sistency and the effect of salient behavioral infor-mation on the self-perception of attitudes. Journal ofPersonality and Social Psychology, 1981, 41, 1-12.

Condry, J. Enemies of exploration: Self-initiated versusother-initiated learning. Journal of Personality andSocial Psychology, 1977, 35, 459-477.

Deci, E. L. Effects of externally mediated rewards onintrinsic motivation. Journal of Personality and So-cial Psychology, 1971, 18, 105-115.

Eagly, A. H., & Himmelfarb, S. Attitudes and opinions.Annual Review of Psychology, 1978, 29, 517-554.

Fazio, R. H. On the self-perception explanation of theoverjustification effect: The role of the salience of ini-tial attitude. Journal of Experimental Social Psy-chology, 1981, / 7, 417-426.

Fazio, R. H., & Zanna, M. P. Direct experience andattitude-behavior consistency. In L. Berkowitz(Ed.),Advances in experimental social psychology (Vol.14). New York: Academic Press, 1981.

Fishbein, M., & Ajzen, I. Belief, attitude, intention, andbehavior. Reading, Mass.: Addison-Wesley, 1975.

Greene, D., Sternberg, B., & Lepper, M. R. Overjus-tification in a token economy. Journal of Personalityand Social Psychology, 1976, 34, 1219-1234.

Greenwald, A. G. Cognitive learning, cognitive responseto persuasion, and attitude change. In A. G. Green-wald, T. C. Brock, & T. M. Ostrom (Eds.), Psycho-logical foundations of attitudes. New York: Aca-demic Press, 1968.

Harvey, J. H., Ickes, W., & Kidd, R. F. A conversationwith Edward E. Jones and Harold H. Kelley. InJ. H. Harvey, W. Ickes, & R. F. Kidd (Eds.), Newdirections in attribution research (Vol. 2). Hillsdale,N.J.: Erlbaum, 1978.

Jones, R. A., & Brehm, J. W. Attitudinal effects ofcommunicator attractiveness when one chooses to lis-ten. Journal of Personality and Social Psychology,1967, 6, 64-70.

Kelley, H. H. Attribution in social interaction. In E. E.Jones, D. E. Kanouse, H. H. Kelly, R. E. Nisbett, S.Valins, & B. Weiner (Eds.), Attribution: Perceivingthe causes of behavior. Morristown, N.J.: GeneralLearning Press, 1972.

Kieslec, C. A. The psychology of commitment. NewYork: Academic Press, 1971.

Kiesler, C. A., Nisbett, R. E., & Zanna, M, P. Oninferring one's beliefs from one's behavior. Journalof Personality and Social Psychology, 1969,11, 321-327.

Krech,; D., Crutchfield, R. S., & Ballachey, E. L. In-dividual in society. New York: McGraw-Hill, 1962.

Kruglanski, A. W., Alon, S,, & Lewi;s, T. Retrospectivemisattribution and task enjoyment. Journal of Ex-perimental Social Psychology, 1972, 8, 493-501.

Lepper, M. R., Greene, D., & Nisbett, R. E. Under-mining children's intrinsic interest with extrinsic re-ward: A test of the overjustification hypothesis. Jour-nal of Personality and Social Psychology, 1973, 28,129-137.

Markus, H. Self-schemata and processing informationabout the self. Journal of Personality and SocialPsychology, 1977, 35, 63-78.

McGuire, W. J. The nature of attitudes and attitudechange. In G. Lindzey & E. Aronson (Eds.), Thehandbook of social psychology (2nd ed., Vol. 3).Reading, Mass.: Addison-Wesley, 1969.

810 WENDY WOOD

Nisbett, R. E., & Valins, S. Perceiving the causes ofone's own behavior. In E. E. Jones, D. E. Kanouse,H. H. Kelley, R. E. Nisbett, S. Valins, & B. Weiner(Eds.), Attribution: Perceiving the causes of behav-ior. Morristown, N.J.: General Learning Press, 1972.

Pallak, M. S., Mueller, M., Dollar, K., & Pallak, J.Effect of commitment on responsiveness to an extremeconsonant communication. Journal of Personalityand Social Psychology, 1972, 23, 429-436,

Petty, R. E., & Cacioppo, J. T. Issue involvement canincrease or decrease persuasion by enhancing mes-sage-relevant cognitive responses. Journal of Person-ality and Social Psychology, 1979, 37, 1915-1926.

Rokeach, M. Beliefs, attitudes, and values. San Fran-cisco: Jossey-Bass, 1968.

Sherif, M., & Hovland, C. I. Social judgment: Assim-ilation and contrast effects in communication andattitude change. New Haven, Conn.: Yale UniversityPress, 1961.

Taylor, S. E., & Crocker, J. Schematic bases of socialinformation processing. In E. T. Higgins, C. P. Her-

man, & M. P. Zanna (Eds.), Social cognition: TheOntario symposium. Hillsdale, N.J.: Erlbaum, 1981.

Tulving, E. Episodic and semantic memory. In E. Tulv-ing & W. Donaldson (Eds.), Organization of mem-ory. New York: Academic Press, 1972.

Wyer, R. S., & Carlston, D. E. Social cognition, in-ference, and attribution. Hillsdale, N.J.: Erlbaum,1979.

Wyer, R. S., & Hartwick, J. The role of informationretrieval and conditional inference process in beliefformation and change. In L. Berkowitz (Ed.), Ad-vances in experimental social psychology (Vol. 13).New York: Academic Press, 1980.

Zanna, M. P., Olson, J. M., & Fazio, R. H. Attitude-behavior consistency: An individual difference per-spective. Journal of Personality and Social Psychol-ogy, 1980, 38, 432-440.

Received January 28, 1981Revision received August 28, 1981

Manuscripts Accepted for Publication in the SectionAttitudes and Social Cognition

Self-Focused Attention and Helping Behavior. Frederick Gibbons (Department of Psychology, IowaState University, Ames, Iowa 50011) and Robert A. Wicklund.

Control and Information in the Intrapersonal Sphere: An Extension of Cognitive Evaluation Theory.Richard Ryan (Department of Psychology, University of Rochester, River Station, Rochester, NewYork 14627).

Automatic Information Processing and Social Perception: I. Detection of Self-Relevant Stimuli WithoutConscious Awareness. John A. Bargh (Department of Psychology, New York University, 6 WashingtonPlace, Seventh Floor, New York, New York 10003).

Perceived Dimensions and Consequences of Attributions for Loneliness. John Michela (Teachers College,Columbia University, New York, New York 10027), Anne Peplau, and David Weeks.

The Structure of Psychological Weil-Being: A Socio-Historical Analysis. Fred B. Bryant (Institute forSocial Research, University of Michigan, Ann Arbor, Michigan 48106) and Joseph Veroff.

Documents

Retrieval of Attitude-Relevant Information From Memory: Effects on