BiasedReasoning BR

http://psp.sagepub.com/Bulletin

Personality and Social Psychology

http://psp.sagepub.com/content/30/3/384The online version of this article can be found at:

DOI: 10.1177/0146167203261296 2004 30: 384Pers Soc Psychol Bull

Britta RennerBiased Reasoning: Adaptive Responses to Health Risk Feedback

Published by:

http://www.sagepublications.com

On behalf of:

Society for Personality and Social Psychology

can be found at:Personality and Social Psychology BulletinAdditional services and information for

http://psp.sagepub.com/cgi/alertsEmail Alerts:

http://psp.sagepub.com/subscriptionsSubscriptions:

http://www.sagepub.com/journalsReprints.navReprints:

http://www.sagepub.com/journalsPermissions.navPermissions:

http://psp.sagepub.com/content/30/3/384.refs.htmlCitations:

What is This?

- Mar 1, 2004Version of Record >>

at Universitaet Konstanz on November 9, 2011psp.sagepub.comDownloaded from

10.1177/0146167203261296 ARTICLEPERSONALITY AND SOCIAL PSYCHOLOGY BULLETINRenner / ADAPTIVE RESPONSES TO RISK FEEDBACK

Biased Reasoning:Adaptive Responses to Health Risk Feedback

Britta RennerUniversity of Greifswald, Germany

The present study examined reactions toward repeated self-relevant feedback. Participants in a community health screen-ing received feedback about their cholesterol level on two separateoccasions. Reactions to the first feedback were examined withregard to feedback valence and expectedness. The findingsshowed that negative feedback was devalued, but only when itwas unexpected. Feedback consistency was incorporated intoanalyses of the second feedback. Again, results showed that neg-ative feedback was not always devaluedonly when it wasinconsistent with the first feedback. Furthermore, positive feed-back was not unconditionally accepted. When receiving unex-pected positive feedback of low consistency, recipients were doubt-ful about its accuracy. Conversely, expected positive feedbackwas accepted regardless of its consistency. These results suggestthat negative or unexpected positive feedbacks evoke greater sen-sitivity to feedback consistency, indicating elaborate cognitiveprocessing. Theoretical accounts of these findings are discussed.

Keywords: risk perception; expectations; motivation; feedback;reasoning

The present study examined the reception of self-relevant feedback in relation to consequential and per-sonally relevant information, extending the work ofexperimental studies that have demonstrated differen-tial acceptance of feedback in dependence of itspositivity and expectedness. Furthermore, the presentstudy compared predictions derived from four theoreti-cal perspectives that assume that differential feedbackacceptance reflects either motivational biased reasoningcaused by positivity or consistency strivings, or reflects anasymmetrical allocation of processing resources.

FEEDBACK VALENCE: POSITIVITY STRIVINGS

VERSUS ALLOCATION OF PROCESSING RESOURCES

Experimental studies in various contexts have shownthat individuals receiving self-relevant negative feedbackoften question its validity and accept it less readily than

positive feedback (for reviews, see Campbell &Sedikides, 1999; Kunda, 1990; Pyszczynski & Greenberg,1987; Taylor & Brown, 1988). The phenomenon of dif-ferential acceptance also can be observed after the pro-vision of health-related feedback (Croyle, Sun, & Hart,1997). For instance, participants who believe they sufferfrom fictitious thioamine acetylase (TAA) enzyme defi-ciency perceived their test result as less accurate andrated TAA deficiency as a less serious health threat thanparticipants who believed that they showed no TAA defi-ciency (e.g., Jemmott, Ditto, & Croyle, 1986). Similarresults were found in experimental studies of appraisalsof blood pressure and cholesterol test results (Croyle,1990; Croyle, Sun, & Louie, 1993, Study 1), gum diseasetest results (McCaul, Thiesse-Duffy, & Wilson, 1992), anda hypothetical bacterial condition (Cioffi, 1991).

Differential feedback acceptance is commonly inter-preted as evidence for motivational biased reasoningthat primarily serves the desire to achieve or maintain apositive sense of self. Depending on the feedback val-ence, different self-defensive processing strategies areinvoked: Whereas positive feedback elicits reasoningthat supports the validity of the given information, nega-tive feedback leads to strategies that undermine it (e.g.,Baumeister, Bratslavsky, Finkenauer, & Vohs, 2001;Croyle et al., 1997; Dawson, Gilovich, & Regan, 2002;Kunda, 1990; Pyszczynski & Greenberg, 1987; Taylor &Brown, 1988). Thus, inherent in the motivated reason-

384

Authors Note: This research was supported by the Deutsche For-schungsgemeinschaft (Grants Re 1583/2-1 and Schw 208/11-01-03)and the Techniker Krankenkasse Berlin-Brandenburg. I would like tothank Harald Schupp for numerous helpful comments and sugges-tions. I also gratefully acknowledge helpful comments by HanneloreWeber, Judith Bler, and Tony Arthur. Correspondence concerningthis article should be addressed to Britta Renner, University ofGreifswald, Psychology, Franz-Mehring-Str. 47, 17487 Greifswald, Ger-many; e-mail: [email protected].

PSPB, Vol. 30 No. 3, March 2004 384-396DOI: 10.1177/0146167203261296 2004 by the Society for Personality and Social Psychology, Inc.


ing conception is the notion that information from theenvironment is molded according to motivationalneeds, that is, self-defensive positivity strivings. However,a theoretical alternative to the positivity strivingperspective has recently been proposed.

According to the quantity of processing view (QoP)(Ditto & Lopez, 1992; Ditto, Scepansky, Munro,Apanovitch, & Lockhart, 1998), feedback valence doesnot determine qualitatively different processing strate-gies, as assumed by the positivity striving conception, butreflects an asymmetrical allocation of processingresources. Whereas negative feedback serves as a strongcue for attention and elaborated cognitive processing,positive information generates superficial processing incomparison. Thus, people scrutinize negative informa-tion carefully, showing sensitivity to details of the giveninformation. However, if cognitive analysis reveals thatthe negative feedback is of rather dubious quality, it willprobably be rejected. Conversely, positive feedback isprocessed superficially and, therefore, people are lesssensitive to the details of the information and are likelyto accept feedback of low quality. According to this view,differential acceptance appears as a byproduct of thequantity of processing.

The assumption that negative information receivesmore attention and effortful cognitive analysis than posi-tive information is supported in various domains ofsocial psychology and health psychology (Baumeisteret al., 2001; Ditto & Lopez, 1992; Pratto & John, 1991;Taylor, 1991). For instance, Liberman and Chaiken(1992) reported that individuals receiving health-threatening information invested more effort in read-ing the message than did individuals who received lesshealth-threatening information. A more stringent test ofthe QoP perspective is provided by a study that manipu-lated information quality (Ditto et al., 1998). A similarmethodological strategy has been frequently used to testfor shallow or elaborate processing in persuasionresearch and attributional inferences (Gilbert &Malone, 1995; Petty & Cacioppo, 1986). Consistent withthe assumption of shallow processing, participantsreceiving positive feedback were insensitive to TAA feed-back quality. Participants receiving negative feedbackwere highly sensitive to this detail of feedback informa-tion, presumably as a reflection of elaborate feedbackprocessing. Thus, negative feedback of low quality wasrelatively less accepted than negative feedback of highquality. Moreover, negative feedback of high quality wassimilarly accepted as positive feedback.

FEEDBACK EXPECTEDNESS: CONSISTENCY STRIVINGS

VERSUS ALLOCATION OF PROCESSING RESOURCES

Differential acceptance might not only be conse-quent on the valence of the feedback information but

also may arise where information is inconsistent withpreexisting expectancies. Information that is unex-pected is generally perceived as less trustworthy anddiagnostically accurate than information that is concor-dant with preexisting expectancies (e.g., Edwards &Smith, 1996; Shrauger, 1975; Swann, Griffin, Predmore,& Gaines, 1987). Although comparatively few studieshave explored this phenomenon in the context of healthpsychology, some observations suggest that expectanciesmoderate feedback processing. For instance, a studywith cancer patients undergoing chemotherapyrevealed that unexpected positive health information(rapid tumor shrinking), as opposed to expected infor-mation (gradual tumor shrinking), can elicit serious dis-tress and negative effects (Nerenz, Leventhal, Love, &Ringler, 1984; but see Shepperd & McNulty, 2002). As forthe reception of negative feedback, the differentialacceptance of expectancy-consistent as opposed toexpectancy-inconsistent information is commonly con-sidered from a motivational biased reasoning perspec-tive. The preference for consistent information and thedevaluation of inconsistent feedback are assumed toreflect the striving for consistency in cognitions aboutthe self, which enables feelings of control and predic-tability (Swann, 1983).

However, the principal logic underlying the negativefeedbackdriven QoP view also may hold for expectancy-inconsistent feedback information. Already, several linesof research suggest that expectancy-inconsistent infor-mation is subjected to an elaborate stimulus analysis(e.g., Edwards & Smith, 1996; Hilton, Klein, & vonHippel, 1991; for review, see Stangor & McMillan, 1992).For instance, research on argument evaluation showedthat belief-incompatible arguments induce a longerreading time and more thought and are judged asweaker than belief-compatible arguments (Edwards &Smith, 1996; see also Lord, Ross, & Lepper, 1979; Petty &Cacioppo, 1986). The QoP approach has consequentlybeen extended to incorporate these findings. For brev-ity, this conception is denoted as the cue adaptive reason-ing account (CARA). The model assumes that both neg-ative feedback and unexpected feedback serve as cuesthat draw attentional resources for elaborate stimulusprocessing. Following more general conceptions on theaffect system and self-regulation (cf. Baumeister et al.,2001; Taylor, 1991), the preferential allocation of pro-cessing resources to negative or unexpected infor-mation is considered an adaptive response. In a worldwhere many stimuli and varying demands competefor processing resources, investment of processingresources to self- and survival relevant stimuli fosters suc-cessful adaptations to environmental challenges anddemands (cf. Baumeister et al., 2001; Ditto et al., 1998).The reasoning triggered by feedback information varies

Renner / ADAPTIVE RESPONSES TO RISK FEEDBACK 385


theoretically on a continuum ranging from shallow toelaborate information processing, with negative andunexpected feedback triggering more elaborate pro-cessing. If, as CARA suggests, negative or unexpectedinformation is processed in a more detail-oriented man-ner, individuals receiving unexpected negative, ex-pected negative, or unexpected positive feedbackshould be more likely to accept high quality feedbackthan low quality feedback. Conversely, expected positivefeedback should initiate little cognitive analysis andindividuals should therefore demonstrate relativeinsensitivity to feedback quality.

THE PRESENT STUDY

Participants received cholesterol feedback on twooccasions, which were 6 months apart. The first choles-terol feedback provided the opportunity to assess feed-back reception as a function of Feedback Expectancyand Feedback Valence. Accordingly, analysis of Time 1(T1) primarily addressed the question of whether feed-back reception varies as a function of either FeedbackValence, as predicted by the self-defensive positivity striv-ing account, or as a function of Feedback Expectedness,as predicted by the self-consistency account. By also con-sidering Feedback Consistency, the reception of the sec-ond feedback allowed the examination of motivationalbiased reasoning perspectives (positivity and self-consistency strivings) and the allocation of processing

resources perspective (QoP and CARA). Previousresearch experimentally manipulated the quality ofinformation by providing bogus information about thereliability of the feedback (Ditto et al., 1998, Study 3). Inthis study, it was assumed that people in principle con-sider consistent repeated feedback as more reliable thaninconsistent feedback. Hence, the sensitivity to thisaspect of the feedback information served as a tool toprobe the quantity of information processing.

To facilitate comparison across the models, Figure 1provides the hypothetical means of accuracy ratings as afunction of Feedback Expectancy, Feedback Valence,and Feedback Consistency for each model. Predictionsregarding the positivity and self-consistency strivingviews were identical for the first screening.1 The self-defensive positivity striving view predicts only a signifi-cant main effect for Feedback Valence (Panel A). Asshown in Panel B, the primary prediction of the self-consistency model is that participants will consider feed-back information as less accurate when it conflicts withtheir expectancies, irrespective of Feedback Valence orFeedback Consistency (resulting in a Feedback Expec-tancy Feedback Valence interaction).

Focusing on the allocation of processing resourcesperspective (QoP and CARA), more complex result pat-terns emerge due to the critical significance of FeedbackConsistency. Following the QoP view (Panel C), Feed-back Valence and Feedback Consistency interact in that

386 PERSONALITY AND SOCIAL PSYCHOLOGY BULLETIN

Unexpected Negative Expected Positive Feedback

Unexpected Positive Feedback

Expected Negative Feedback Feedback

High

Low

Perceived Test AccuracyHigh

Expected Positive Feedback


Expected Negative Feedback

Unexpected Negative Feedback

Low

HighLow

Feedback ConsistencyA

C Perceived Test Accuracy





High

Low





High

Low

B

D

Perceived Test Accuracy


Figure 1 Hypothetical means of perceived test accuracy used to illustrate the predictions of the positivity striving (Panel A), the self-consistency(Panel B), the QoP (Panel C), and the CARA (Panel D) accounts.

NOTE: Qop = quantity of processing, CARA = cue adaptive reasoning account.


negative feedback of low consistency is considered to beless accurate than negative feedback of high consistencyor positive feedback of either consistency. As shown inPanel D, CARA extends the QoP approach by assumingthat not just negative feedback but also unexpected feed-back triggers more effortful cognitive analysis. Thus,CARA is the only model predicting a triple interactionamong Feedback Expectancy, Feedback Valence, andFeedback Consistency. As shown, decomposing the tri-ple interaction for positive and negative feedback sepa-rately, a main effect of Feedback Consistency is expectedfor negative feedback, whereas an interaction of Feed-back Valence and Feedback Consistency is predicted forpositive feedback. Alternatively, when decomposing thetriple interaction for low and high consistent feedbackseparately, a significant interaction Feedback Expec-tancy Feedback Valence effect is only predicted for lowconsistent information. Both methods of decomposingthe triple interaction should reflect the predictions that(a) unexpected positive feedback, expected negativefeedback, and unexpected negative feedback shouldlead to deeper processing and, therefore, high consis-tent feedback should, on average, be viewed as moreaccurate than low consistent feedback; and (b) par-ticipants receiving expected positive feedback shouldreadily accept the feedback irrespective of FeedbackConsistency.

The measure of perceived accuracy, which is alsoreferred to as perceived fact, is complemented by mea-sures of perceived implications for the self (Croyle et al.,1993). In general, a similar pattern is predicted by thedifferent accounts, except that negative feedback shouldgenerate more perceived threat for the self and pressureto change than positive feedback. Thus, according themotivational biased perspective, participants shouldshow relative insensitivity to Feedback Consistency. Con-versely, the allocation of processing resources view pre-dicts that participants receiving unexpected or negativefeedback are sensitive to Feedback Consistency.

METHOD

Participants

A large proportion of the participants (66%) wererecruited for a health screening conducted by the FreeUniversity of Berlin and the Technicians Health Insur-ance Agency (Techniker Krankenkasse) through adver-tisements placed in local newspapers in Berlin, Ger-many. The remaining participants (34%) were recruitedby a letter that was sent to people insured with the Tech-nicians Health Insurance Agency who lived near thefour study locations (two universities and two city halls).In total, 1,487 individuals were recruited for the first cho-lesterol screening and, of these, 604 participants also

took part in the second screening. From these 604 par-ticipants, 14 participants (2%) had to be excluded fromthe data set because they failed to complete the ques-tionnaires. In the data analyses, only participants whoprovided complete data sets for the first and secondscreening were included (study sample n = 590). Themean age of this sample was 45 years (SD = 15), and 51%were male. The average cholesterol level was 225 mg/dl(SD = 45) and 218 mg/dl (SD = 46) at the first and sec-ond measurement, respectively, which is below the meanGerman population cholesterol level of 237 mg/dl(Troschke, Klaes, Maschewsky-Schneider, & Scheuer-mann, 1998).

Control analyses showed that the study sample was, onaverage, 7 years older; had higher cholesterol levels (M =225 mg/dl vs. M = 214 mg/dl), ts(1,471) > 4.7, ps < .001;and received more frequent expected negative feedbackand less frequent expected positive feedback than thedropout group, 2(1) = 20.52, p < .001. Analyses of thereactions toward the first cholesterol feedback showedthat the study sample and the dropouts did not differ sys-tematically with respect to perceived accuracy and per-ceived threat, Fs < 1, ns, respectively. However, the studysample felt more pressure to change than the dropouts(M = 2.9 vs. M = 2.6), F(1, 1465) = 6.31, p = .012. No inter-action between Feedback Expectancy, FeedbackValence, and the Sample Group (study sample vs. drop-outs) was significant, Fs < 2.5, ns.

Measures

Feedback expectancy. Individuals completed an initialquestionnaire, which included a measure of the ex-pected cholesterol test result. Specifically, participantswere asked, Immediately after completing this ques-tionnaire your cholesterol level will be measured. Whatcholesterol level do you expect? Participants rated theirexpected cholesterol test result on a scale of 1 (very low)through 4 (optimal) to 7 (very high). Participants weredivided according to whether they expected an optimalor lower cholesterol test result (positive expectancy) oran elevated reading (negative expectancy).

Perceived feedback accuracy. Two questions were askedregarding participants beliefs about the accuracy oftheir cholesterol test result. First, How likely do youthink it is that your cholesterol test result is false or inac-curate? (1 = very likely and 7 = very unlikely). Second,How likely do you think it is that your cholesterol mea-sure represents a temporary fluctuation? (1 = very likelyand 7 = very unlikely). The two measures were signifi-cantly correlated (first feedback, r = .69, p < .001; secondfeedback, r = .64, p < .001) and thus they were averaged togenerate a single measure of the perceived accuracy ofthe test result.



Perceived threat. Two items served to assess perceivedthreat. Participants were asked to rate how serious ahealth threat their cholesterol level was on a 7-pointscale, anchored by 1 (very low) through 4 (moderatelyhigh) to 7 (very high). They also were asked to rate howworried they felt due to their cholesterol test result. Rat-ings were made on a scale of 1 (absolutely not worried)through 4 (worried) to 7 (very worried). These two mea-sures were also highly correlated (first feedback, r = .83,p < .001; second feedback, r = .81, p < .001) and weretherefore averaged to create an overall perceived threatscore.

Perceived pressure to change. Pressure to change reflectsthe extent to which a person feels pressured to lowertheir cholesterol level and change their behavior (cf.Fuchs, 1996). Participants were given the following state-ment: It is necessary for me to do something to lowermy cholesterol level. The responses were given on a 4-point scale ranging from 1 (strongly disagree) to 4 (stronglyagree).

Perceived changes in nutrition behavior. At the secondscreening, before the feedback was given, participantswere asked whether they had adopted a more healthynutrition since the first screening. The general stem wasas follows: Have you changed your nutrition since thelast screening half a year ago? (a) I have lowered mycholesterol intake, (b) I have lowered my calorieintake, and (c) I have lowered my fat intake.Responses were made on a dichotomous scale where 1 =yes and 2 = no. When people indicated that they hadchanged at least one of these three behaviors, it wascoded as perceived change in behavior.

Feedback valence. Participants were divided accordingto whether they had received positive feedback (totalcholesterol < 201 mg/dl) or negative feedback (totalcholesterol > 200 mg/dl).

Feedback consistency. When the valence of the secondfeedback was in concurrence with the valence of the first,it was coded as being of high consistency (n = 479). Con-versely, when the second feedback was discrepant to thefirst feedback, it was coded as being of low consistency(n = 111). Because feedback was based on actual feed-back, a comparable low prevalence of inconsistent feed-back has to be expected.

Procedure

After arriving at the screening site, participantsreceived a brief description of the study and signed aconsent form. Participants then answered a question-naire that included a measure of the expected choles-terol test result. Afterward, participants weight andheight were measured. Trained laboratory assistantsthen measured the total cholesterol level using a

fingerstick blood draw and a Reflotron desktop analyzer.Following the cholesterol measurement, participantswere provided with their exact actual cholesterol level.Furthermore, participants received feedback on theircholesterol level risk category according to internationalstandards (National Heart, Lung, and Blood Institute,1995). Participants with a cholesterol level of 200 mg/dlor less were told that their cholesterol level was optimaland did not pose a risk for cardiovascular diseases. Indi-viduals with either borderline high cholesterol levels(between 201 mg/dl and 249 mg/dl) or high cholesterollevels (above 249 mg/dl) were informed about thepotential risks of borderline and high cholesterol levelsfor cardiovascular diseases.2 Shortly after receiving thecholesterol feedback, participants were given a secondquestionnaire. Among the filler questions, participantswere asked to report on the results of their cholesteroltest. After completing the second questionnaire, partici-pants received individualized follow-up recommenda-tions, were thanked for their participation, and receivedan invitation for the second screening, which took placehalf a year later. The second screening followed a similarprocedure except that participants also were askedwhether they had changed nutrition-related behaviorssince receiving the first feedback.

RESULTS

Feedback Expectancy and Feedback Valence

At the first and second cholesterol screening, 227 and324 (39% and 55%) participants expected positive cho-lesterol feedback, whereas 363 and 266 (61% and 45%)expected a negative test result. Based on the actual cho-lesterol reading, 177 and 218 (30% and 37%) individu-als received positive feedback and 413 and 372 (70% and63%) were confronted with negative feedback.

In total, 65% and 67% of the study sample showed amatch between the expected feedback and the actualfeedback at the first and second screening. In particular,99 and 173 (17% and 29%) participants received posi-tive feedback expectantly and 285 and 218 (48% and37%) were confronted with negative feedback expec-tantly; 128 and 154 (22% and 26%) expected positivefeedback but received negative feedback and thereforedemonstrated an optimistic bias. Conversely, 78 and 45(13% and 8%) expected negative feedback but receivedpositive feedback, demonstrating a pessimistic bias.Hence, if participants made an inaccurate estimation oftheir actual feedback, they were more likely to make anunrealistically optimistic estimation than an unrealis-tically pessimistic one, 2(1) = 12.14 and 61.11, p < .001,for T1 and Time 2 (T2), respectively.




Reactions toward the first cholesterol feedback (T1). Theanalyses of the perceived test accuracy included bothparticipants prior expectancies (positive vs. negative)and the received valence of the cholesterol feedback(positive vs. negative), which were analyzed in a 2 2ANOVA design with additional post hoc Bonferronicontrasts.

Results indicated a significant main effect for Feed-back Valence, F(1, 586) = 4.88, p = .028. However, thismain effect was further qualified by a significant Feed-back Expectancy Feedback Valence interaction, F(1,586) = 15.48, p < .001. As shown in Figure 2, participantsreceiving unexpected negative cholesterol feedback(M = 5.2, SD = 1.5) gave significantly lower accuracy esti-mates compared to the other three groups, ts > 2.5, ps =.01, which did not differ from each other (expected neg-ative M = 5.9, SD = 1.3; expected positive M = 6.0, SD = 1.2;unexpected positive M = 5.8, SD = 1.2; ts < 1.2, ns).

Reactions toward the second cholesterol feedback (T2). Theanalyses of the perceived test accuracy at T2 included notonly Feedback Expectancy (positive vs. negative) andFeedback Valence (positive vs. negative) but also theadditional variable Feedback Consistency (low vs. high),which were analyzed in a 2 2 2 ANOVA design andpost hoc Bonferroni contrasts.

The ANOVA revealed the triple interaction amongFeedback Expectancy Feedback Valence FeedbackConsistency, F(1, 582) = 5.13, p = .024, which was pre-dicted by the CARA account. Accordingly, the tripleinteraction was followed up by analyzing the FeedbackExpectancy Feedback Consistency interactions andtheir corresponding main effects for the positive andnegative feedback group, respectively.

Negative feedback. Within the negative feedback group,the main effect for Feedback Consistency reached statis-tical significance, F(1, 582) = 5.62, p = .018. As Figure 3demonstrates, participants receiving consistent feed-back (M = 5.8, SD = 1.4) showed, on average, higheracceptance than did participants receiving inconsistentfeedback (M = 4.9, SD = 1.6). Neither the main effectFeedback Expectancy nor the interaction FeedbackExpectancy Feedback Consistency were significant, Fs< 1, ns.

Positive feedback. Analyses within the positive feedbackgroup yielded a significant Feedback Expectancy Feed-back Consistency interaction, F(1, 582) = 4.82, p = .029,indicating that Feedback Consistency mattered only forparticipants receiving unexpected positive health feed-back. Thus, unexpected positive feedback of low consis-tency was rated, on average, as significantly less accuratecompared to unexpected positive feedback of high con-sistency or expected positive feedback of either high orlow consistency, ts > 4.2, p < .001. In contrast, expectedpositive feedback was accepted equally whether it was ofhigh or low consistency, F < 1, ns.

Low versus high consistency feedback. A second approachto complement the significant triple interaction Feed-back ExpectancyFeedback ValenceFeedback Consis-tency is to consider the low and high consistency feed-back groups separately. For participants receiving highlyconsistent information, neither Feedback Expectancynor Feedback Valence had any impact on reported testaccuracy, Fs < 1, ns. Thus, highly consistent informationwas generally accepted as accurate independently ofwhether it was unexpected, negative, or even both. Con-versely, those receiving information of low consistencyrevealed differential accuracy ratings as a function ofboth Feedback Expectancy and Feedback Valence, F(1,582) = 7.11, p = .008. Although risk feedback of low con-sistency was equally devalued by participants receiving


1

2

3

4

5

6

7Perceived Test Accuracy (t2)

Negative FeedbackExpected

Negative FeedbackUnexpected

Positive FeedbackExpected

Positive FeedbackUnexpected

Feedback Consistency High Low

Figure 2 Reactions to the first feedback as a function of Feedback Ex-pectancy and Feedback Valence.

NOTE: t2 = Time 2.

1

2

3

4

5

6

7Perceived Test Accuracy (t1)

PositiveFeedback

NegativeFeedback

1

2

3

4

5

6

7Perceived Threat (t1)

ExpectedUnexpected

PositiveFeedback

NegativeFeedback

1

2

3

4Perceived Pressure to Change (t1)

PositiveFeedback

NegativeFeedback

Feedback

Figure 3 Perceived accuracy of the second feedback as a function ofFeedback Expectancy, Feedback Valence, and FeedbackConsistency.

NOTE: t1 = Time 1.


unexpected positive feedback, expected negative feed-back, or unexpected negative feedback, ts > 1, it wasaccepted as highly valid by participants receivingexpected positive feedback, ts > 2.7, ps < .05.

Perceived Implications

Reactions toward the first cholesterol feedback (T1). Per-ceived threat and pressure to change elicited by the firstcholesterol feedback were analyzed with a 2 2 ANOVAincluding Feedback Expectancy and Feedback Valenceas between-subjects variables. Both measures convergein the findings. As expected, negative feedback elicitedhigher perceived threat and pressure to change than didpositive feedback, Fs(1, 586) = 156.24 and 317.30, ps 22.32, ps < .001.

Reactions toward the second cholesterol feedback (T2). Per-ceived threat and pressure to change elicited by the sec-ond cholesterol feedback were analyzed with a 2 2 2ANOVA including Feedback Expectancy, FeedbackValence, and Feedback Consistency as between-subjectsvariables.

Again, as expected, ANOVAs for perceived threat andpressure to change yielded a significant main effect forFeedback Valence, Fs(1, 582) = 24.33 and 99.36, ps 4, ps < .05. In contrast, individuals receivingexpected positive feedback were not sensitive to Feed-back Consistency for either perceived threat or pressureto change, Fs < 1.

Control Analyses

Changes in expectancy. The analysis of the reception ofthe second cholesterol feedback provided empirical sup-port for the view that feedback reception varies as a func-tion of Feedback Consistency. However, alternatively,one might assume that the consistency of the FeedbackExpectancy might have influenced the reception of thesecond feedback. Overall, 69% of the participants hadstable expectancies across both feedback sessions. Asexpected, positive feedback for T1 was more likely tochange participants expectancies than was negativefeedback for T1 (39% vs. 28%), 2(1) = 6.22, p = .013.However, for participants receiving positive feedback atT1, change of expectancy varied as a function of Feed-back Expectancy at T1. Specifically, only 4% of theexpected positive feedback group, but 81% of the unex-pected positive feedback group, changed their expec-tancy. In contrast, expectancy change was similarly pro-nounced for participants receiving expected andunexpected negative feedback (30% and 27%).

These differences in expectancy change also arereflected in the analyses of the second cholesterol feed-back, which was based on the feedback given at T2 andthe expectancy of the second test. Thus, participantswho expected positive feedback for T2 were signifi-cantly more likely to have changed their expectancyafter the first feedback than were participants whoexpected negative feedback for T2 (43% vs. 21% expec-tancy change), 2(1) = 50.49, p < .001. More specifically,38% and 28% of the expected and unexpected positivefeedback group and 13% and 48% of the expected andunexpected negative feedback group had changed theirexpectations.

Finally, whether participants did or did not changetheir expectancies might systematically influence thereception of the second feedback. Because only 13% of

the expected negative feedback group changed theirexpectancy between T1 and T2, this group had to beexcluded from the analyses. Accordingly, the three mea-sures of feedback reception were analyzed in a 3 2 2ANOVA design with the three factors Feedback Group atT2 (expected positive, unexpected positive, and unex-pected negative feedback), Feedback Consistency (highvs. low), and Expectancy Change (yes vs. no). Neitherperceived accuracy nor the two measures of perceivedimplications yielded a significant main effect or a sig-nificant interaction effect involving the factor Expec-tancy Change, Fs < 2.6, ns. Hence, whether participantschanged their expectancy between T1 and T2 or not didnot significantly influence the reception of the feedbackat T2.

Perceived changes in diet behaviors. Forty-three percentof the participants reported that they had adopted ahealthier diet after the first feedback, supporting thenotion that the cholesterol feedback was perceived asconsequential and personally relevant information. Asexpected, negative feedback was significantly morelikely to induce (self-reported) behavior change thanpositive feedback (50% vs. 26%), 2(1) = 26.98, p < .001.More specifically, 54% and 43% of the expected andunexpected negative feedback group and 24% and 28%of the expected and unexpected positive feedbackgroup stated that they had changed their behavior.

Examining the frequency of self-reported changes innutrition from the perspective of the second feedbackshowed that the negative feedback group was more likelyto have changed their nutrition than the positive feed-back group (49% vs. 32%), 2(1) = 14.79, p < .001. Morespecifically, 52% and 44% of the expected and unex-pected negative feedback group and 31% and 39% ofthe expected and unexpected positive feedback groupreported that they had changed their nutrition.

Finally, to explore whether participants reportedbehavioral change affected the reception of the feed-back at T2, the three measures of feedback receptionwere analyzed in a 2 2 2 2 ANOVA design with thefactors Feedback Expectancy at T2, Feedback Valence atT2, Feedback Consistency, and perceived BehaviorChange (yes vs. no). The analyses of perceived test accu-racy and perceived threat yielded neither a significantmain effect nor a significant interaction effect for thefactor Behavior Change, (Fs < 2, ns). For perceived pres-sure to change, the analysis yielded the effects reportedpreviously for Feedback Expectancy, Feedback Valence,and Feedback Valence Feedback Consistency (Fs > 5,p < .05). However, in addition, the main effect for Behav-ior Change, F(1, 574) = 6.24, p = .013, and the FeedbackValence Behavior Change interaction, F(1, 574) = 7.95,p = .005, were significant. The Feedback Valence Behavior Change interaction indicates that participants



who received positive feedback felt equally low per-ceived pressure to change independently of self-reported behavioral changes (M = 1.5 vs. M = 1.7; F < 1,ns). Conversely, those who received negative feedbackreported more pressure to change when they reportedthat they had already adopted a more healthy diet thanwhen they had not (M = 2.7 vs. M = 3.3), F(1, 574) = 19.21,p < .001.

DISCUSSION

The present study addressed the question of how peo-ple respond to feedback that is personally consequen-tial. The main goal of the present study was to examinewhether differential feedback acceptance reflects anasymmetrical allocation of processing resources. Theallocation of processing resources was probed by testingwhether participants showed sensitivity to the consis-tency of cholesterol feedback given on two occasions.The analysis of the feedback accuracy ratings revealed asignificant triple interaction Feedback Expectancy Feedback Valence Feedback Consistency, which wasconsistent with the predictions made by CARA.

The QoP view (Ditto et al., 1998; Ditto & Lopez, 1992)and CARA suggest that both expected and unexpectednegative feedback trigger elaborate feedback processingand, therefore, sensitivity to information consistency.Consistent with these assumptions, participants receiv-ing negative feedback of high consistency accepted thefeedback as being more valid than did participantsreceiving negative feedback of low consistency. More-over, and in line with the predictions by CARA, partici-pants receiving unexpected positive feedback alsoshowed sensitivity to feedback consistency: When thefeedback was of high consistency it was rated as moreaccurate than when it was of low consistency. This con-trasts clearly with the insensitivity to feedback consis-tency shown by participants receiving expected positiveinformation.

Extending the QoP view, CARA assumes that unex-pected positive feedback also serves as a cue for system-atic processing. Specifically, in the context of personallyconsequential feedback, people might examine unex-pected positive information carefully to prevent termi-nating protective actions erroneously due to potentiallyfalse-negative health information, which might causesevere harm in the future. A complementary explana-tion emerges from considering that individuals brace forpossible negative outcomes by lowering their expecta-tions strategically beforehand (Shepperd, Findley-Klein,Kwavnick, Walker, & Perez, 2000). Presumably, partici-pants securitized unexpected positive feedback morecarefully to avoid disappointments in the future. How-ever, people might examine unexpected positive infor-mation carefully only when the issue at hand is impor-

tant and the potential cost of erroneous acceptance ofthe feedback is high (cf. Michie et al., 2002).

A number of important control analyses ensured thatthe sensitivity to feedback consistency primarilyreflected Feedback Valence and Feedback Expectancyrather than the consistency of expectancies across bothfeedback sessions or perceived behavior changes. Oneinteresting result of these control analyses was that nega-tive feedback is more likely to induce self-reported pre-ventive behaviors than is positive feedback. Further-more, unexpected positive feedback was highly effectivein changing participants expectancies. However, it isparticularly relevant that perceived feedback accuracydid not vary systematically as a function of expectancy orself-reported behavior change.

Taken together, the analysis of feedback accuracyreveals the pattern of results predicted by CARA, anextension of the QoP view (Ditto et al., 1998; Ditto &Lopez, 1992). If one accepts the contention that sensitiv-ity to feedback consistency probes the amount ofeffortful cognitive processing of the given feedbackinformation, the data suggest that unexpected positivefeedback, expected negative feedback, and unexpectednegative feedback serve as cues for the increased alloca-tion of processing resources. However, although sensitiv-ity to details of the information has served as a measureof elaborate cognitive processing in numerous studies inpersuasive communication and attribution research(Gilbert & Malone, 1995; Petty & Cacioppo, 1986), fur-ther research is necessary to provide direct evidence forthe elaborate processing of information after receivingnegative or unexpected information.

Motivational Biased Reasoning and theReception of Consequential Health Risk Feedback

The differential acceptance of negative health feed-back as a function of feedback consistency is difficult toexplain from a motivational biased reasoning perspec-tive. Considering positivity and self-consistency strivingssimultaneously, as suggested by multiple motives con-ceptions, unexpected negative feedback constitutes themost aversive information (e.g., Jussim, Yen, & Aiello,1995; Sedikides, 1993; Shrauger, 1975; Stahlberg,Petersen, & Dauenheimer, 1999; Swann & Schroeder,1995; Taylor, Neter, & Wayment, 1995). When both self-consistency and self-defensive strivings combine againstaccepting the information, individuals should be highlymotivated to undermine its validity. Conversely, bothmotives support the acceptance of expected positivefeedback. However, in contrast to these predictions, neg-ative feedback of high consistency, whether unexpectedor expected, was accepted to a similar degree as ex-pected positive feedback. This is remarkable because,theoretically, estimations by the expected positive feed-



back group should be unbiased or should even reflectoverestimated accuracy ratings.

A further challenge for positivity striving perspectivesis the sensitivity to feedback consistency of the unex-pected positive feedback group. If people predomi-nately strive to attain or maintain a positive view of theself, this group should readily accept the feedback asvalid information. However, participants receiving unex-pected positive feedback of low consistency were as rig-orous in rejecting the given feedback as participantsreceiving negative feedback of low consistency. Similarly,from the self-consistency perspective, one might arguethat participants may have been reluctant to accept un-expected positive feedback because it conflicted withself-consistency needs and the desire to protect feelingsof control. In contrast, both unexpected and expectedpositive feedback of high consistency was highlyaccepted.

Sensitivity to Feedback Consistencyand Perceived Implications

Providing further support for the CARA perspective,measures of perceived implications (perceived threatand pressure to change) also revealed sensitivity to feed-back consistency: Participants receiving expected nega-tive feedback, unexpected negative feedback, or unex-pected positive feedback showed sensitivity to feedbackconsistency, whereas participants receiving expectedpositive feedback were insensitive to its consistency.Despite these overall similarities of measures of per-ceived fact and implications, they did not mirror eachother completely. Specifically, differences emerged forthe reception of negative feedback. Whereas perceivedaccuracy varied only as a function of feedback consis-tency, perceived implications varied as a function of twoindependent effectsfeedback consistency and expec-tedness. Thus, unexpected negative feedback was lessthreatening than expected negative feedback, irrespec-tive of the consistency of the feedback.

These differences presumably reflect that perceivedfact and implication tap into different aspects of feed-back processing. The CARA and the QoP approach onlyspecify conditions under which more effortful and elab-orate feedback processing might occur, but not whatkinds of specific information are considered for apprais-ing different aspects of the feedback. From a normativeperspective, the valence of feedback is of great conse-quence for self-related implications (e.g., threat for theself), but it is not informative for appraising its generalaspects (e.g., feedback accuracy, general threat, andimplications). Of interest, perceived threat apparentlynot only reflects feedback valence but participants calcu-late the danger they potentially face as a result of the

given feedback in conjunction with their expectancy andperceived previous behavior changes.

The finding that unexpected negative feedback wasperceived as less threatening than expected negativefeedback might be considered as evidence for motiva-tional biased reasoning. Positivity and consistencystrivings combine for this feedback group; thus, themotivation to downplay information is most pro-nounced. According to this perspective, the first line ofdefense as indexed by perceived accuracy might havebeen more difficult to derogate, whereas the second lineof defense indexed by measures of perceived threatallowed more leeway to downplay unwanted informa-tion (Croyle et al., 1993). However, from this perspectiveit is difficult to explain why participants receiving posi-tive feedback of low consistency felt substantially morethreatened when the feedback was unexpected thanwhen it was expected.

The Reception of Health Feedback:The Need to Consider Expectancies and Valence

The typical finding in health psychology is that peo-ple derogate negative in comparison to positive healthfeedback (cf. Croyle et al., 1997). However, the presentstudy observed that feedback reception varied as a func-tion of both Feedback Valence and Feedback Expec-tancy. Although an interaction of Feedback Valence andFeedback Expectancy emerged for both feedback ses-sions, the first screening is of particular relevance be-cause it employed a setting comparable to previous stud-ies. Considering the first health feedback, participantsreceiving unexpected negative feedback considered thetest result to be less accurate than did participants re-ceiving expected negative feedback.

Previous experimental studies providing health feed-back (for a review, see Croyle et al., 1997) might haveinadvertently confounded Feedback Expectancy andFeedback Valence. In these studies, differential accep-tance of health feedback was presumably observedbecause people who received positive feedback receivedit expectantly, whereas negative feedback probably tookthem by surprise. Support for this notion is derived fromthe many studies that demonstrate that individuals har-bor unrealistic positive expectancies about their healthand their future (e.g., Renner, in press; Weinstein, 1980,in press).

This reasoning is further supported by studies ofunrealistic optimism, which suggest that people whounderestimate their risk are prone to defensiveness(Davidson & Prkachin, 1997; Radcliffe & Klein, 2002;Weinstein & Klein, 1995; Wiebe & Black, 1997). Forexample, similar to the present study, Radcliffe andKlein (2002) found that unrealistically optimistic indi-viduals worried less about their risk than did others (who



were either accurate or pessimistic). In a similar vein,Avis, Smith, and McKinlay (1989) reported that optimis-tically biased individuals were rather resistant in chang-ing their risk perception of having a heart attack afterreceiving unexpected negative feedback (66% demon-strated stable risk perception). These findings convergewith the present study. Seventy percent of the unrealisticoptimists maintained the expectancy of a positive choles-terol feedback after receiving negative feedback at T1,whereas only 30% of the unrealistic pessimists main-tained a negative expectancy. A similar asymmetry wasobserved for the expected feedback groups. Takentogether, individuals might not only harbor positiveunrealistic positive expectancies about their health butalso may show resistance to negative feedback. In partic-ular, the finding of asymmetrical patterns of expectancychange for unrealistic optimists and unrealistic pessi-mists might be considered as evidence for defensiveness(Irle & Krolage, 1973; Weinstein & Klein, 1995).

However, a more cautionary perspective on this rea-soning is suggested by considering the self-reportedbehavioral changes that were induced by the feedback.Specifically, although 70% of the unrealistic optimistsmaintained their expectancy, 42% of them reported thatthey had changed their behavior after receiving unex-pected negative feedback. The behavioral changesreported within the unexpected negative group mightbe due to an increase in perceived risk immediately afterreceiving negative feedback at T1, which in turn mighthave motivated them to change their behavior. However,once participants believe that they have modified theirrisk behavior, expectancies might reflect these behaviorchanges because they have removed or reduced thesource of the risk (Kreuter & Strecher, 1995). Thus,maintaining a positive expectancy despite receiving neg-ative feedback might reflect either defensiveness or thebelief that the modification of risk behaviors was effec-tive in reducing health risk. Furthermore, the measureof perceived behavior change probably underestimatesthe probability of people being motivated to actuallychange their behavior because health behavior changedepends on additional variables, for example, outcomeexpectancies or perceived self-efficacy (cf. Renner &Schwarzer, in press; Schwarzer & Renner, 2000).

Methodological Limitations

The phenomenon of biased reasoning was exploredhere in a field study, and limitations of the internal andexternal validity of the present study must therefore beacknowledged. People who choose to be tested are bydefinition self-selected and may, in part, be psychologi-cally and behaviorally prepared for dealing with badnews. Consequently, the degree to which the findings

generalize to people who refrained from testing is lim-ited. Although typical for public health screeningstudies with volunteers (cf. Glanz & Gilboy, 1995), a clearrestriction for the external validity might be that theattrition rate between the first and second screening ledto a systematic sample bias. There are a number of vari-ables (e.g., education, age) that might possibly contrib-ute to the attrition rate that cannot be completely ruledout in this study. However, control analyses showed thatthe dropouts and the study sample did not differ system-atically in their reception of the first feedback (i.e., per-ceived accuracy, perceived threat), except that the drop-outs felt less pressure to change than the study sample. Inaddition, the dropouts had a lower total cholesterol levelthan did the study sample. Thus, participants probablydid not abstain from retesting because they were es-pecially threatened by the first feedback or because theywere more defensive.

A further limitation of the present study is that thecholesterol feedback was not randomly assigned to therecipients but was based on their actual cholesterol testresults. The advantage of giving actual feedback is that itis naturalistic and personally important. Moreover, itappears that random assignment to experimental condi-tions is only ethically feasible for studying short-termeffects because negative health feedback is emotionallyupsetting for the recipients (cf. Baumann, Cameron,Zimmerman, & Leventhal, 1989; Croyle et al., 1997).Conversely, without any question, a priori differencesbetween the two feedback groups might have impairedinternal validity. Although previous studies have shownno direct relationship between risk factor appraisalsand individual difference variables such as self-esteem,monitoring versus blunting coping style, repression-sensitization, or dispositional optimism (Croyle et al.,1993; Ditto, Jemmott, & Darley, 1988; Radcliffe & Klein,2002), dispositions might influence feedback expectan-cies and, consequently, risk feedback reception.

However, the observed pattern might primarily applyto personally consequential settings that are at leastpartly under behavioral control. Dunning (1995), forexample, observed that feedback about a stable (non-controllable) aspect of personality induced self-defensive reactions, whereas feedback about a malleableaspect generated more unbiased reactions (see alsoDitto et al., 1988). Considering these findings in con-junction with the present study leads to the conclusionthat perceived controllability might be an importantmediator that needs further investigation.

NOTES

1. According to the motivational biased reasoning perspective, pro-cessing of wanted (positive or expected) information elicits reason-



ing that supports the validity of the information, whereas unwanted(negative or unexpected) information leads to reasoning that under-mines it. Thus, processing of both wanted and unwanted informationis equally vigorous but directed toward different ends (cf. Ditto,Scepansky, Munro, Apanovitch, & Lockhart, 1998). In addition, manyresearchers assume that positivity and self-consistency strivings are lim-ited by the desire to maintain an illusion of objectivity or by realityconstrains (e.g., Kunda, 1990; Pyszczynski & Greenberg, 1987; Taylor &Brown, 1988). However, the desire to maintain an illusion of objectivityshould be operative whether the received feedback is wanted orunwanted. According to this reasoning, one would expect an addi-tional main effect Feedback Consistency. It should be noted that nointeraction involving Feedback Consistency is derived by thisadditional reasoning.

2. Control analyses revealed that borderline and high cholesterolfeedback groups did not differ significantly in their accuracy ratings atTimes 1 and 2, Fs < 1.6, ns. As expected, the high cholesterol feedbackgroup felt somewhat more threatened and more pressure to changethan did the borderline cholesterol feedback group, Fs < 53, p < .001. Inaddition, the borderline high cholesterol feedback reported, on aver-age, higher threat and pressure to change than did the optimal choles-terol feedback group, Fs > 28, p < .001. Given these findings, borderlineand high cholesterol feedback groups were combined in the analyses.

REFERENCES

Avis, N. E., Smith, K. W., & McKinlay, J. B. (1989). Accuracy of percep-tions of heart attack risk: What influences perceptions and canthey be changed? American Journal of Public Health, 79, 1608-1611.

Baumann, L. J., Cameron, L. D., Zimmerman, R. S., & Leventhal, H.(1989). Illness representations and matching symptoms. HealthPsychology, 8, 449-469.

Baumeister, R. F., Bratslavsky, E., Finkenauer, C., & Vohs, K. D.(2001). Bad is stronger than good. Review of General Psychology, 5,323-370.

Campbell, W. K., & Sedikides, C. (1999). Self-threat magnifies theself-serving bias: A meta-analytic integration. Review of General Psy-chology, 3, 23-43.

Cioffi, D. (1991). Asymmetry of doubt in medical self-diagnosis: Theambiguity of uncertain wellness. Journal of Personality and SocialPsychology, 61, 969-980.

Croyle, R. T. (1990). Biased appraisal of high blood pressure. Preven-tive Medicine, 19, 40-44.

Croyle, R. T., Sun, Y. C., & Hart, M. (1997). Processing risk factorinformation: Defensive biases in health-related judgments andmemory. In K. L. Petrie & J. A. Weinman (Eds.), Perceptions of healthand illness (pp. 267-290). Amsterdam: Harwood Academic.

Croyle, R. T., Sun, Y. C., & Louie, D. H. (1993). Psychologicalminimization of cholesterol test results: Moderators of appraisalin college students and community residents. Health Psychology, 12,503-507.

Davidson, K., & Prkachin, K. (1997). Optimism and unrealistic opti-mism have an interacting impact on health-promoting behaviorand knowledge changes. Personality and Social Psychology Bulletin,23, 617-625.

Dawson, E., Gilovich, T., & Regan, D. T. (2002). Motivated reasoningand performance on the Wason Selection Task. Personality andSocial Psychology Bulletin, 28, 1379-1387.

Ditto, P. H., Jemmott, J. B., & Darley, J. M. (1988). Appraising thethreat of illness: A mental representational approach. Health Psy-chology, 7, 183-201.

Ditto, P. H., & Lopez, D. F. (1992). Motivated skepticism: Use of differ-ential decision criteria for preferred and nonpreferred conclu-sions. Journal of Personality and Social Psychology, 63, 568-584.

Ditto, P. H., Scepansky, J. A., Munro, G. D., Apanovitch, A. M., & Lock-hart, L. K. (1998). Motivated sensitivity to preference-inconsistentinformation. Journal of Personality and Social Psychology, 75, 53-69.

Dunning, D. (1995). Trait importance and modifiability as factorsinfluencing self-assessment and self-enhancement motives. Per-sonality and Social Psychology Bulletin, 21, 1297-1306.

Edwards, K., & Smith, E. E. (1996). A disconfirmation bias in the eval-uation of arguments. Journal of Personality and Social Psychology, 71,5-24.

Fuchs, R. (1996). Causal models of physical exercise participation:Testing the predictive power of the construct pressure tochange. Journal of Applied Social Psychology, 26, 1931-1960.

Gilbert, D. T., & Malone, P. S. (1995). The correspondence bias. Psy-chological Bulletin, 117, 21-38.

Glanz, K., & Gilboy, M. B. (1995). Psychosocial impact of cholesterolscreening and management. In R. T. Croyle (Ed.), Psychosocialeffects of screening for disease prevention and detection (pp. 39-64). Lon-don: Oxford University Press.

Hilton, J. L., Klein, J. G., & von Hippel, W. (1991). Attention alloca-tion and impression formation. Personality and Social PsychologyBulletin, 17, 548-559.

Irle, M., & Krolage, J. (1973). Kognitive Konsequenzen irrtmlicherSelbsteinschtzungen [Cognitive consequences of erroneous self-judgments]. Zeitschrift fr Sozialpsychologie, 4, 36-50.

Jemmott, J. B., Ditto, P. H., & Croyle, R. T. (1986). Judging health sta-tus: Effects of perceived prevalence and personal relevance. Jour-nal of Personality and Social Psychology, 50, 899-905.

Jussim, L., Yen, H. J., & Aiello, J. R. (1995). Self-consistency, self-enhancement, and accuracy in reactions to feedback. Journal ofExperimental Social Psychology, 31, 322-356.

Kreuter, M. W., & Strecher, V. J. (1995). Changing inaccurate percep-tions of health risk: Results from a randomized trial. Health Psychol-ogy, 14, 56-63.

Kunda, Z. (1990). The case for motivated reasoning. Psychological Bul-letin, 108, 480-498.

Liberman, A., & Chaiken, S. (1992). Defensive processing of person-ally relevant health messages. Personality and Social Psychology Bulle-tin, 18, 669-679.

Lord, C. G., Ross, L., & Lepper, M. R. (1979). Biased assimilation andattitude polarization: The effects of prior theories on subse-quently considered evidence. Journal of Personality and Social Psy-chology, 37, 2098-2109.

McCaul, K. D., Thiesse-Duffy, E., & Wilson, P. (1992). Coping withmedical diagnosis: The effects of at-risk versus disease labels overtime. Journal of Applied Social Psychology, 22, 1340-1355.

Michie, S., Weinman, J., Miller, J., Collins, V., Halliday, J., & Marteau,T. M. (2002). Predictive genetic testing: High risk expectations inthe face of low risk information. Journal of Behavioural Medicine, 25,33-50.

National Heart, Lung, and Blood Institute. (1995). Recommendationsregarding public screening for measuring blood cholesterol (NIH Publica-tion No. 95-3045). Bethesda, MD: National Institutes of Health.

Nerenz, D. R., Leventhal, H., Love, R. R., & Ringler, K. E. (1984). Psy-chological aspects of cancer chemotherapy. International Review ofApplied Psychology, 33, 521-529.

Petty, R. E., & Cacioppo, J. T. (1986). Communication and persuasion:Central and peripheral routes to attitude changes. New York: AcademicPress.

Pratto, F., & John, O. P. (1991). Automatic vigilance: The attention-grabbing power of negative social information. Journal of Personal-ity and Social Psychology, 61, 380-391.

Pyszczynski, T., & Greenberg, J. (1987). Toward an integration of cog-nitive and motivational perspectives on social inference: A biasedhypothesis-testing model. In L. Berkowitz (Ed.), Advances in experi-mental social psychology (Vol. 20, pp. 297-340). San Diego, CA: Aca-demic Press.

Radcliffe, N. M., & Klein, W. M. P. (2002). Dispositional, unrealistic,and comparative optimism: Differential relations with the knowl-edge and processing of risk information and beliefs about per-sonal risk. Personality and Social Psychology Bulletin, 28, 836-846.

Renner, B. (in press). Hindsight bias after receiving self-relevanthealth risk information: A motivational perspective. Memory.

Renner, B., & Schwarzer, R. (in press). Social-cognitive factors pre-dicting health behavior change. In J. Suls & K. Wallston (Eds.),Social psychological foundations of health and illness. Oxford, UK:Blackwell.



Schwarzer, R., & Renner, B. (2000). Social-cognitive predictors ofhealth behavior: Action self-efficacy and coping self-efficacy.Health Psychology, 19, 487-495.

Sedikides, C. (1993). Assessment, enhancement, and verificationdeterminants of the self-evaluation process. Journal of Personalityand Social Psychology, 65, 317-338.

Shepperd, J. A., Findley-Klein, C., Kwavnick, K. D., Walker, D., &Perez, S. (2000). Bracing for loss. Journal of Personality and SocialPsychology, 78, 620-634.

Shepperd, J. A., & McNulty, J. K. (2002). The affective consequencesof expected and unexpected outcomes. Psychological Science, 13,85-88.

Shrauger, J. S. (1975). Responses to evaluation as a function of initialself-perceptions. Psychological Bulletin, 82, 581-596.

Stahlberg, D., Petersen, L. E., & Dauenheimer, D. (1999). Prefer-ences for and evaluation of self-relevant information dependingon the elaboration of the self-schemata involved. European Journalof Social Psychology, 29, 489-502.

Stangor, C., & McMillan, D. (1992). Memory for expectancy-congruent and expectancy-incongruent information: A review ofthe social and social developmental literatures. Psychological Bulle-tin, 111, 42-61.

Swann, W. B., Jr. (1983). Self-verification: Bringing social reality intoharmony with the self. In J. Suls & A. G. Greenwald (Eds.), Socialpsychological perspectives on the self (Vol. 2, pp. 33-66). Hillsdale, NJ:Lawrence Erlbaum.

Swann, W. B., Jr., Griffin, J. J., Predmore, S. C., & Gaines, B. (1987).The cognitive-affective crossfire: When self-consistency confrontsself-enhancement. Journal of Personality and Social Psychology, 52,881-889.

Swann, W. B., Jr., & Schroeder, D. G. (1995). The search for beautyand truth: A framework for understanding reactions to evalua-tions. Personality and Social Psychology Bulletin, 21, 1307-1318.

Taylor, S. E. (1991). Asymmetrical effects of positive and negativeevents: The mobilization minimization hypothesis. PsychologicalBulletin, 110, 67-85.

Taylor, S. E., & Brown, J. D. (1988). Illusion and well-being: A socialpsychological perspective on mental health. Psychological Bulletin,103, 193-210.

Taylor, S. E., Neter, E., & Wayment, H. A. (1995). Self-evaluation pro-cesses. Personality and Social Psychology Bulletin, 21, 1278-1287.

Troschke, J. von, Klaes, L., Maschewsky-Schneider, U., & Scheuer-mann, W. (1998). Die Deutsche Herz-Kreislauf-Prventionsstudie [TheGerman cardiovascular prevention study]. Bern, Germany: Huber.

Weinstein, N. D. (1980). Unrealistic optimism about future lifeevents. Journal of Personality and Social Psychology, 39, 806-820.

Weinstein, N. D. (in press). Exploring the links between risk percep-tions and preventive health behavior. In J. Suls & K. Wallston(Eds.), Social psychological foundations of health and illness. Oxford,UK: Blackwell.

Weinstein, N. D., & Klein, W. M. (1995). Resistance of personalrisk perceptions to debiasing interventions. Health Psychology, 14,132-140.

Wiebe, D. J., & Black, D. (1997). Illusional beliefs in the context ofrisky sexual behaviors. Journal of Applied Social Psychology, 27, 1727-1749.

Received December 23, 2002

Revision accepted August 11, 2003



Documents

BiasedReasoning BR