the process of selective exposure: why confirmatory information search weakens over time

12
The process of selective exposure: Why confirmatory information search weakens over time Peter Fischer a,, Stephen Lea b , Andreas Kastenmüller c , Tobias Greitemeyer d , Julia Fischer e , Dieter Frey e a Institute of Psychology, University of Graz, Universitaetsplatz 2, 8010 Graz, Austria b University of Exeter, UK c John-Moores-University, Liverpool, UK d University of Innsbruck, Austria e Ludwig-Maximilians-University, Munich, Germany article info Article history: Received 24 July 2007 Accepted 7 September 2010 Available online 16 October 2010 Accepted by William Bottom Keywords: Selective exposure Decision making Confirmatory information search Biased assimilation Sequence of information search abstract The present research investigated whether the tendency to prefer decision-consistent to decision-incon- sistent information after making a preliminary choice would vary during the sequential process of searching for additional pieces of decision-relevant information. Specifically, it was tested whether deci- sion makers would be more confirmatory in their information evaluation and search at the commence- ment rather than end of an information search process. In fact, five studies revealed that participants exhibited stronger confirmatory tendencies in both information evaluation (Studies 2 and 5) and search (Studies 1, 3, and 4) immediately after making a preliminary decision compared to during the later stages of an information search process. With regard to the underlying mechanism, results further revealed that individuals appear to be more motivated to detect the best decision alternative at the beginning (as opposed to the end) of an information search process, which leads to increases in confirmatory informa- tion processing during these stages. Ó 2010 Elsevier Inc. All rights reserved. Introduction When searching for new information, individuals often preferen- tially seek pieces of information which are consistent with their a priori decisions, attitudes, or viewpoints – a tendency that is called ‘confirmatory information search’ or ‘selective exposure’ (Festinger, 1957; Frey, 1986; Jonas, Schulz-Hardt, Frey, & Thelen, 2001). A wide range of studies have shown that selective exposure occurs in con- text of both individual (Frey, 1986; Jonas et al., 2001) and group deci- sion making (Schulz-Hardt, Frey, Lüthgens, & Moscovici, 2000). It has been also observed in context of the stereotypes people hold (John- ston, 1996); self-serving conclusions they make (Holton & Pyszczyn- ski, 1989); attitudes they hold (Lundgren & Prislin, 1998); and advices they give (Jonas & Frey, 2003). This tendency is an important psychological phenomenon in research on decision making, because it is often responsible for poor decision outcomes (e.g., Kray & Galin- sky, 2003). An additional factor that is strongly related to the selec- tive exposure effect (indeed being one of its most important determinants; Fischer, Jonas, Frey, & Schulz-Hardt, 2005; Fischer, Schulz-Hardt, & Frey, 2008; Schulz-Hardt, Fischer, & Frey, 2010) is that individuals evaluate information that is consistent with their own decision preferences more positively than that which is incon- sistent with them (Ditto & Lopez, 1992; Greitemeyer, Fischer, Frey, & Schulz-Hardt, 2009). This can also prove a barrier to the revision of incorrect initial judgements, in both individual (Greitemeyer & Schulz-Hardt, 2003) and group settings (Greitemeyer, Schulz-Hardt, Brodbeck, & Frey, 2006). Although there has been a great deal of research into selective exposure, knowledge about what happens within a specific infor- mation search sequence (incorporating the processes of informa- tion search and evaluation) is limited. Confirmatory information tendencies have mostly been treated as constant within a given information search process, and as a consequence, research has mostly focused on its final indicator, confirmation bias; regardless of whether these tendencies occurred early or late in the decision making process. The possibility of variations between different temporal stages of information evaluation and search has been only sparsely discussed and investigated. The present research pro- poses that the extent to which decision makers exhibit confirma- tory information evaluation and search is stronger in the initial (rather than later) stages of a decision-relevant information search process. The effect of search stage upon the level of selectivity in information search is anticipated by both the classic defense- motivational (dissonance) approach (Frey, 1986) and the more re- cent, evaluation-based accuracy-motivational perspective (Fischer et al., 2005; Schulz-Hardt et al., 2010). Five studies tested this 0749-5978/$ - see front matter Ó 2010 Elsevier Inc. All rights reserved. doi:10.1016/j.obhdp.2010.09.001 Corresponding author. Address: Karl-Franzens-University, Department Psychol- ogy, Social Psychology, Universitaetsplatz 2, 8010 Graz, Austria. E-mail address: peter.fi[email protected] (P. Fischer). URL: http://www.uni-graz.at/socialpsychology (P. Fischer). Organizational Behavior and Human Decision Processes 114 (2011) 37–48 Contents lists available at ScienceDirect Organizational Behavior and Human Decision Processes journal homepage: www.elsevier.com/locate/obhdp

Upload: peter-fischer

Post on 11-Sep-2016

213 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: The process of selective exposure: Why confirmatory information search weakens over time

Organizational Behavior and Human Decision Processes 114 (2011) 37–48

Contents lists available at ScienceDirect

Organizational Behavior and Human Decision Processes

journal homepage: www.elsevier .com/ locate/obhdp

The process of selective exposure: Why confirmatory information searchweakens over time

Peter Fischer a,⇑, Stephen Lea b, Andreas Kastenmüller c, Tobias Greitemeyer d, Julia Fischer e, Dieter Frey e

a Institute of Psychology, University of Graz, Universitaetsplatz 2, 8010 Graz, Austriab University of Exeter, UKc John-Moores-University, Liverpool, UKd University of Innsbruck, Austriae Ludwig-Maximilians-University, Munich, Germany

a r t i c l e i n f o a b s t r a c t

Article history:Received 24 July 2007Accepted 7 September 2010Available online 16 October 2010Accepted by William Bottom

Keywords:Selective exposureDecision makingConfirmatory information searchBiased assimilationSequence of information search

0749-5978/$ - see front matter � 2010 Elsevier Inc. Adoi:10.1016/j.obhdp.2010.09.001

⇑ Corresponding author. Address: Karl-Franzens-Unogy, Social Psychology, Universitaetsplatz 2, 8010 Gra

E-mail address: [email protected] (P. FischURL: http://www.uni-graz.at/socialpsychology (P.

The present research investigated whether the tendency to prefer decision-consistent to decision-incon-sistent information after making a preliminary choice would vary during the sequential process ofsearching for additional pieces of decision-relevant information. Specifically, it was tested whether deci-sion makers would be more confirmatory in their information evaluation and search at the commence-ment rather than end of an information search process. In fact, five studies revealed that participantsexhibited stronger confirmatory tendencies in both information evaluation (Studies 2 and 5) and search(Studies 1, 3, and 4) immediately after making a preliminary decision compared to during the later stagesof an information search process. With regard to the underlying mechanism, results further revealed thatindividuals appear to be more motivated to detect the best decision alternative at the beginning (asopposed to the end) of an information search process, which leads to increases in confirmatory informa-tion processing during these stages.

� 2010 Elsevier Inc. All rights reserved.

Introduction that individuals evaluate information that is consistent with their

When searching for new information, individuals often preferen-tially seek pieces of information which are consistent with their apriori decisions, attitudes, or viewpoints – a tendency that is called‘confirmatory information search’ or ‘selective exposure’ (Festinger,1957; Frey, 1986; Jonas, Schulz-Hardt, Frey, & Thelen, 2001). A widerange of studies have shown that selective exposure occurs in con-text of both individual (Frey, 1986; Jonas et al., 2001) and group deci-sion making (Schulz-Hardt, Frey, Lüthgens, & Moscovici, 2000). It hasbeen also observed in context of the stereotypes people hold (John-ston, 1996); self-serving conclusions they make (Holton & Pyszczyn-ski, 1989); attitudes they hold (Lundgren & Prislin, 1998); andadvices they give (Jonas & Frey, 2003). This tendency is an importantpsychological phenomenon in research on decision making, becauseit is often responsible for poor decision outcomes (e.g., Kray & Galin-sky, 2003). An additional factor that is strongly related to the selec-tive exposure effect (indeed being one of its most importantdeterminants; Fischer, Jonas, Frey, & Schulz-Hardt, 2005; Fischer,Schulz-Hardt, & Frey, 2008; Schulz-Hardt, Fischer, & Frey, 2010) is

ll rights reserved.

iversity, Department Psychol-z, Austria.

er).Fischer).

own decision preferences more positively than that which is incon-sistent with them (Ditto & Lopez, 1992; Greitemeyer, Fischer, Frey, &Schulz-Hardt, 2009). This can also prove a barrier to the revision ofincorrect initial judgements, in both individual (Greitemeyer &Schulz-Hardt, 2003) and group settings (Greitemeyer, Schulz-Hardt,Brodbeck, & Frey, 2006).

Although there has been a great deal of research into selectiveexposure, knowledge about what happens within a specific infor-mation search sequence (incorporating the processes of informa-tion search and evaluation) is limited. Confirmatory informationtendencies have mostly been treated as constant within a giveninformation search process, and as a consequence, research hasmostly focused on its final indicator, confirmation bias; regardlessof whether these tendencies occurred early or late in the decisionmaking process. The possibility of variations between differenttemporal stages of information evaluation and search has beenonly sparsely discussed and investigated. The present research pro-poses that the extent to which decision makers exhibit confirma-tory information evaluation and search is stronger in the initial(rather than later) stages of a decision-relevant information searchprocess. The effect of search stage upon the level of selectivity ininformation search is anticipated by both the classic defense-motivational (dissonance) approach (Frey, 1986) and the more re-cent, evaluation-based accuracy-motivational perspective (Fischeret al., 2005; Schulz-Hardt et al., 2010). Five studies tested this

Page 2: The process of selective exposure: Why confirmatory information search weakens over time

38 P. Fischer et al. / Organizational Behavior and Human Decision Processes 114 (2011) 37–48

proposal: Studies 1, 3, and 4 looked at whether the effect could befound for confirmatory information search (selective exposure),whereas Studies 2 and 5 examined whether it would also holdfor confirmatory information evaluation. In addition, Studies 2, 3,4, and 5 additionally tested the motivational basis of the expectedeffect; that is, whether it is based on different levels of defense(Studies 2–4) or accuracy motivation (Study 5).

2 Other (non-dissonance) lines of research have investigated confirmation biasesfrom alternative theoretical perspectives, including positive test strategies (e.g.Klayman & Ha, 1987; Snyder & Swann, 1978); option- vs. attribute-orientedinformation search (e.g. Payne, Bettman, & Johnson, 1993); and inadequate integra-tion of base rates into judgment processes (e.g. Fiedler, Brinkmann, Betsch, & Wild,2000). Although there is a great deal of non-dissonance theoretical research on biasedinformation search, most of the cognitive paradigms differ significantly from theclassic selective exposure paradigm (which is used in the present studies). The mostimportant difference is that in the classic selective exposure paradigm, participantsknow in advance whether a specific piece of information will support or contradicttheir viewpoint or decision. This is generally not true of experiments within the more

Research on selective exposure

Theoretical perspectives

Empirical research on selective exposure has mainly been per-formed within the framework of dissonance theory (Festinger,1957; Frey, 1986), which explains it through assuming processesrelated to defense motivation. Concretely, dissonance theory pro-poses that after making a decision, individuals experience cogni-tive dissonance due to the potential negative implications of thechosen decision alternative, and the potential positive implicationsof the non-chosen one. Because dissonance is experienced as anunpleasant and self-threatening state of aversive arousal, peopleare motivated to reduce it. One way to do so is through selectiveexposure; preferring information that supports one’s made deci-sion, and thus neglecting conflicting information (Fischer, Jonas,Frey, & Kastenmüller, 2008; Frey, 1986). In other words, disso-nance theory proposes that decision makers exhibit confirmatoryinformation tendencies in order to defend their position and reachthe goal of dissonance reduction.

Other lines of research have revealed that under specific cir-cumstances (such as when no additional threats to decision qualityare induced and decision-makers experience a medium to high le-vel of decision certainty; for a discussion, see Fischer et al., inpress; Fischer, Jonas, et al., 2008), the selective exposure effectcan also occur due to increased accuracy motivation (Fischeret al., 2005; Fischer, Schulz-Hardt, et al., 2008; Schulz-Hardtet al., 2010). This account is based on the assumptions that: (a)decision makers generally try to find the qualitatively best piecesof information (accuracy motivation) and (b) that there seem tobe testing asymmetries between decision-consistent and -incon-sistent information, so that decision makers are generally morecritical of inconsistent than consistent information, and thus sys-tematically devalue it (Ditto & Lopez, 1992; Ditto, Scepansky, Mun-ro, Apanovitch, & Lockhart, 1998; Fischer et al., 2005; Russo,Medvec, & Meloy, 1996). In this context, Schulz-Hardt et al.(2010) have shown that perceived information quality is indeed astrong predictor of confirmatory information search. By searchingfor the ‘best’ information, decision makers tend to systematicallyprefer decision-consistent pieces of information, as it is seen tobe qualitatively better thanks to asymmetrical testing (see alsoFischer et al., 2005; Fischer, Schulz-Hardt, et al., 2008). If decisionmakers are highly accuracy motivated and experience a high levelof decision importance (and thus invest more cognitive energy intothe processing of decision-relevant information), these functionaltesting asymmetries can become stronger, and thus counter-intui-tively increase confirmatory information evaluation and selectiveexposure.1 In sum, confirmatory information evaluation (biasedassimilation) and search (selective exposure) can occur from in-creases in defense motivation and, in certain circumstances, accu-racy motivation. The present studies examine whether theexpected impact of stage of information search upon selectivity of

1 Please note that increased cognitive elaboration does not always lead to increasedlevels of confirmatory information evaluation and search. For example, Fischer, Jonas,et al. (2008) found that increased cognitive elaboration reduces confirmatoryinformation search when specific decision cues (in this case, loss decision frames)lead to experiences of low decision certainty.

information evaluation and search is based more on processes of de-fense motivation or accuracy motivation.

The classic selective exposure paradigm and typical findings

The extent to which individuals prefer consistent to inconsistentinformation (determined through subtracting the number of incon-sistent pieces of information chosen from the number of consistentpieces selected) is called confirmation bias, and is a typical indicatorof the strength of the selective exposure effect (Frey, 1986; Jonaset al., 2001).2 In the classic selective exposure research paradigm, par-ticipants are asked to work on a decision problem, which requiresthem to decide between two different decision alternatives (such aswhether the contract of a manager should be extended or not; whichof two consumer products should be bought; or which investmentstrategy should be chosen). Participants are then asked to make a pre-liminary decision on the basis of the available decision-relevant infor-mation. Afterwards, participants have the opportunity to search foradditional information, which is normally received as short state-ments indicating the perspectives of newspaper articles, experts, orformer participants. From these statements, the information seekerlearns both (a) how the article’s author argues in general and (b) thedecision outcome they recommend (e.g., ‘‘The manager was successfulin setting up new products. Thus, his/her contract should be pro-longed”). The participants are asked to mark the pieces of informationthat they would like to read in more detail later on. Regardless of theparticipant’s initial decision, half of the additional information is con-sistent with it (arguments for the chosen alternative and argumentsagainst the non-chosen alternative), and half are inconsistent (argu-ments against the chosen and for the non-chosen). Selective exposureto supporting information occurs when participants systematicallyselect more consistent than inconsistent pieces of decision-relevantinformation.3

In prior selective exposure research, multiple situational vari-ables have been identified that increase the tendency for confirma-tory information search: restricted access to additionalinformation (Fischer et al., 2005); high commitment to a position(Schwarz, Frey, & Kumpf, 1980); gain-framed (compared to loss-framed) decision problems (Fischer, Jonas, et al., 2008); depletedself-regulation resources (Fischer, Greitemeyer, & Frey, 2008); in-creased levels of dissonance arousal (Rhine, 1967); and negativeaffective states (Jonas, Graupmann, & Frey, 2006). Most of these ef-fects have been explained by a defense-motivational perspective(dissonance reduction; e.g., Frey, 1986; Jonas et al., 2006); butsome recent findings have also been explained by processes ofaccuracy motivation (e.g., Fischer et al., 2005; Fischer, Schulz-Hardt, et al., 2008). The present studies test whether the strengthof the selective exposure effect varies with the stage (initial com-pared to later) of information search.

cognitive paradigms on biased information search.3 Since most selective exposure paradigms provide the possibility of revising an

initial decision via a final one to be made later (cf. Schulz-Hardt et al., 2000); one cansay from a normative perspective that preferring consistent to inconsistent informa-tion after a preliminary decision can indeed be regarded as a ‘‘bias”, which provesdysfunctional for the decision maker’s willingness to revise a potentially wrongpreliminary decision.

Page 3: The process of selective exposure: Why confirmatory information search weakens over time

P. Fischer et al. / Organizational Behavior and Human Decision Processes 114 (2011) 37–48 39

The present research: the process of information seeking in theclassic selective exposure paradigm

Although many determinants of confirmatory information pro-cessing and search have been identified, there is still little knowl-edge about what happens within the sequential information searchprocess in context of the classic dissonance theoretical informationsearch paradigm. In other words, the process of information searchhas been handled like a black box where the final outcome (theoverall intensity of the confirmation bias) has been assumed to oc-cur at consistent strength throughout the information searchprocess.4 However, from a defense-motivational (dissonance) per-spective, increased confirmatory information processing and searchin the early stages of information search should be particularly effec-tive, as it will affirm and bolster the self-concept, thus reducing ini-tial post-decision dissonance. This will prepare the decision makerfor any upcoming, self-threatening inconsistent information that islikely to increase dissonance arousal. After the self has been suffi-ciently affirmed, the cognitive system is stable enough to be con-fronted with any inconsistent information that may emerge duringthe later stages of information search (see Frey, 1986), which helpsthe decision maker to appear unbiased and fair (illusion of objectiv-ity; Pyszczynski & Greenberg, 1987).

Moreover, from an evaluation-based accuracy-motivationalperspective, one would also expect increased levels of selectiveexposure in the initial rather than later stages of informationsearch. At the beginning of an information search process (rightafter the preliminary decision is made), decision makers shouldexperience comparatively high levels of decision salience and thusdecision importance (motivation to make a good decision), whichshould lead to strong intentions of finding the qualitatively bestpieces of decision-relevant information especially in this earlystage of decision-relevant information search. However, becauseconsistent information is generally perceived to be of higher qual-ity than inconsistent information (see Ditto & Lopez, 1992; Dittoet al., 1998), decision-consistent information should receive a sys-tematic selection advantage during this early stage. In contrast,decision salience (and thus decision importance) should be rela-tively reduced in the later stages of an information search process,which might be reflected in a reduced desire to find the best piecesof decision-relevant information, and thus reduced levels of confir-matory information evaluation and search (selective exposure).

In summary, increased levels of confirmatory informationevaluation and search in the initial (compared to later) stages ofan information search process are predicted by both the defense-motivational and accuracy-motivational accounts. The followingfive studies attempt to demonstrate that confirmatory informationprocessing tendencies are indeed more pronounced in the initialstages of the information evaluation and search process. Moreover,it will be tested whether the expected effect is best explainedby defense-motivational (dissonance) or accuracy-motivationalprocesses.

Study 1

Study 1 was designed to provide evidence that decision makersexhibit a stronger preference for decision-consistent informationin the initial rather than later stages of an information search

4 For an exception, see the literature on Svenson’s (1992) ‘differentiation–consol-idation theory’, which postulates that decision makers search for information in abalanced way (weighing the pros and cons of a decision alternative) before making adecision, but are confirmatory after they have made their decision in order toconsolidate it. However, this research does not address different stages of informationsearch after decision makers have made their (preliminary) decision (as is the case inthe present research paradigm).

process. Participants made a preliminary consumer choice regard-ing purchasing one of two high-end cars, and were subsequentlygiven the opportunity to read eight additional pieces of informa-tion, which could either support or undermine their selection.These eight pieces were presented in two blocks, with four piecesof information representing the initial stage of a decision-relevantinformation seeking process (block 1), and four pieces representingthe later stage (block 2). It was expected that participants wouldexhibit a stronger preference for decision-consistent informationin the first rather than second block of information search.

Method

Participants and design

Fifty-seven students (38 female and 18 male; one participantdid not indicate her/his gender) aged between 19 and 34 yearsold (M = 22.54, SD = 2.85) from the University of Munich partici-pated in the study exchange for course credit. Two participantswere excluded from the analyses due to providing incomplete data.The study had a 2 (type of information: consistent vs. inconsis-tent) � 2 (stage of information search process: block 1 vs. block 2)design with repeated measures on both factors.

Material and procedure

As in previous research on selective exposure (cf. Frey, 1986), aconsumer decision paradigm was employed. Participants wereasked to make a preliminary decision regarding purchasing aBMW- or Mercedes-brand car, on the basis of a given cover story.In context, participants were instructed to imagine working as anengineer for a construction company, which provided the optionof utilizing a company car. The company was described as havingcontracts with two vehicle manufacturers, resulting in the poten-tial acquisition of either a BMW X5 4.4 or Mercedes ML 430. Bothcars were cross country vehicles, and very similar in features suchas power, speed, weight, and price.5 Participants were providedwith the most important performance features of both cars, includ-ing two pictures. Subsequently, they were asked to make a prelimin-ary decision about which one they would select (participants werealso informed that they would have to make a final decision later).

After making their preliminary choice, participants were toldthat additional information concerning the vehicles was available,and that it would help them come to a final decision. The addi-tional information consisted of eight one-page statements purport-edly written by car experts, each of which had been summarized ina two-sentence thesis (though in reality, the extended versions ofthe articles did not exist). Each summary had been pre-tested withregard to maintaining equal levels of perceived information qualitythroughout the set (i.e., credibility, reliability, and importance; seeKastenmüller, Greitemeyer, Jonas, Fischer, & Frey, in press), andmade the expert’s opinion of the given vehicle (be it positive ornegative) clear. An example of a main thesis in favor of the BMWwas: ‘‘Most of the control units of the BMW X5 4.4i are located directlyon the steering wheel, so that the driver’s hands can stay on the steer-ing wheel while driving. This is a big advantage for the BMW.” Anexample of a main thesis favoring the Mercedes was: ‘‘The Merce-des ML 430 has a very big trunk, affording it a very high trunk loadcapacity. So, the Mercedes can be recommended.” Critical opinionsof the vehicles included theses such as: ‘‘The BMW X5 4.4i can beused only for small loads. This is a big disadvantage of the BMW”,and ‘‘It is difficult to transport big things with the Mercedes ML 430because one cannot simultaneously use the trunk and the backseat

5 Both vehicles represent typical ‘‘Premium Segment” German cars.

Page 4: The process of selective exposure: Why confirmatory information search weakens over time

Number of selected articles

Fig. 1. The number of selected consistent and inconsistent pieces of information asa function of stage of information search process in Study 1. Block 1 represents the

40 P. Fischer et al. / Organizational Behavior and Human Decision Processes 114 (2011) 37–48

area. Hence, this car cannot be recommended.” Overall, both carswere subject to four positive appraisals and four negative critiques,meaning that participants were exposed to equal amounts of deci-sion-consistent and -inconsistent information, regardless ofwhether they favored the BMW or the Mercedes when makingtheir preliminary decision.

Furthermore, participants were told that they would receive theinformation in two separate blocks of four pieces of informationeach; supposedly because recent research had shown that present-ing additional material in blocks improved decision making qual-ity. In each block, participants were allowed to select as manypieces of information (from one to four) as they wished. After par-ticipants had made their selection(s), they were asked to make a fi-nal decision about choosing the BMW or the Mercedes as theircompany car. Participants were then debriefed (with particular re-gard to the fact that there were not actually any extended versionsof the additional pieces of information) and thanked for taking partin the study.

initial stage of information search; block 2 represents the later stage.

Results and discussion

Check for effects of decision making

Initially, 38 participants preferred the BMW and 17 the Merce-des; in the final decision, 33 participants preferred the BMW and22 the Mercedes. Eleven participants changed their initial decisionpreference during the information search process, with eightswitching from BMW to Mercedes, and three from Mercedes toBMW. When the initial preference was included in the analysisof the factorial design, a significant three-way interaction wasfound between type of information, type of decision stage, anddecision preference, F(1, 53) = 9.88, p < .01, indicating that the ef-fect reported below was mainly due to participants who selectedthe BMW instead of the Mercedes. That is, participants who optedfor the BMW became more balanced with regard to informationsearch during block 2, whereas participants who selected the Mer-cedes exhibited more confirmatory information search in block 2.However, this specific pattern did not affect the overall effect re-ported below.

Information search

For an overview, see Fig. 1. Consistent articles were classified asthose in favor of the chosen vehicle or opposed to the non-chosenone, and inconsistent articles consisted of those either opposed tothe chosen, or those favoring the non-chosen vehicle. A 2 (type ofinformation: consistent vs. inconsistent) � 2 (stage of informationsearch process: block 1 vs. block 2) ANOVA revealed a significantinteraction, F(1, 54) = 4.61, p < .04, g2 = .08. No significant main ef-fects were found, both Fs < 1.

Follow-up analyses conducted separately for block 1 and block2 revealed that in block 1 (the initial stage of the information seek-ing process), participants selected significantly more decision-con-sistent (M = 1.33, SD = 0.67) than decision-inconsistent pieces ofinformation (M = 1.09, SD = 0.78), F(1, 54) = 4.16, p < .05, g2 = .07.In contrast, no significant difference between selected consistent(M = 1.09, SD = 0.82) and inconsistent pieces of information(M = 1.25, SD = 0.82) was obtained for block 2 (the final stage ofthe information seeking process), F(1, 54) = 1.00, p > .32, g2 = .02.

Study 1 thus provided initial evidence that decision makers ex-hibit stronger confirmatory information search tendencies at thebeginning (block 1) rather than later stages (block 2) of an informa-tion search process. Because this effect was mainly due to partici-pants who opted for the BMW, an alternative decision case waschosen for the next study.

Study 2

Using a different decision case, Study 2 investigated whether asimilar effect would also occur for information evaluation, which isa crucial determinant of actual information search (Fischer et al.,2005; Fischer, Schulz-Hardt, et al., 2008; Schulz-Hardt et al.,2010). As in Study 1, participants were asked to make an initialdecision, and were consequently exposed to four decision-consis-tent and four decision-inconsistent pieces of additional informa-tion. The information was presented in two separate blocks,consisting of two decision-consistent and two decision-inconsis-tent pieces each. However, in contrast to Study 1, participants werethen asked to evaluate the quality of each piece instead of simplysearching for information.

In order to gather evidence as to whether the effect found inStudy 1 had defense-motivational (dissonance) roots, measuresof (a) the perceived pleasantness of the consistent and inconsistentinformation, and (b) experienced self-esteem threat were em-ployed; both of which were administered during the two stagesof information evaluation. If the defense-motivational account isvalid (thus making decision-consistent pieces of information moreself-serving in the initial stage of information search), they shouldbe assessed as more pleasant in the initial stage of informationevaluation (block 1), in comparison to the later stage (block 2). Inaddition, from the defense-motivational point of view, self-esteemthreat should be higher in the initial (block 1) rather than laterstage of information evaluation (block 2).

Method

Participants and design

Twenty-nine participants (18 female and 11 male, ranging from17 to 33 years of age; M = 24.52, SD = 4.31) from the University ofMunich participated in the experiment, which had a 2 (type ofinformation: consistent vs. inconsistent) � 2 (stage of informationseeking process: block 1 vs. block 2) factorial design with repeatedmeasures on both factors.

Material and procedure

After participants individually arrived at the experimental lab,they were asked to imagine the following situation: ‘‘You havehad the idea of opening a new kind of snack stall. You intend to startwith a single stall. But you have one further decision problem: You’ve

Page 5: The process of selective exposure: Why confirmatory information search weakens over time

P. Fischer et al. / Organizational Behavior and Human Decision Processes 114 (2011) 37–48 41

had two good and innovative ideas and you have to decide on one con-cept. The goods you offer on your stall could consist either of diet prod-ucts (e.g., low fat and low carb products, etc.) or organic products (e.g.,vegetables grown without pesticides or genetic manipulation, etc.).Both the organic and diet industries seem to be very trendy inGermany.”

After making their decision, participants were confronted witheight pieces of additional information, half of which supportedthe organic business, and half of which supported the diet productidea – meaning that 50% of the additional information supportedthe participant’s preliminary decision, and 50% conflicted with it.As in Study 1, the information provided consisted of two-sentencesummaries of purported longer articles, which did not actually ex-ist. An example of a piece of information supporting the organicbusiness was: ‘‘Organic products are healthier than diet products.Thus, an organic stand can be recommended.” An example of a pieceof information supporting the diet product business was: ‘‘Organicproducts are more expensive than diet products. Hence, a diet businessshould be preferred.”

The eight pieces of information were presented in two blocks offour, with each block featuring two pieces of information support-ing the organic business idea, and two pieces favoring the dietproduct one. Participants were asked to evaluate each piece ofinformation with regard to its perceived expected reliability(0 = not reliable; 10 = very reliable) and importance for making agood decision (0 = not important; 10 = very important). In addition,they also indicated the extents to which the individual pieces ofinformation threatened them (0 = not at all; 10 = extremely) andwere pleasant to read (0 = not at all pleasant; 10 = very pleasant).After participants finished evaluating the information presentedin block 1, their responses were collected by the experimenter,the materials for block 2 distributed, and evaluations for the sec-ond set of information made. In order to control for potential mate-rial order effects, the information items featured in blocks 1 and 2were counterbalanced. Participants made no second decision inthis study. Upon finishing the second set of evaluations, partici-pants were debriefed, with particular regard to the fact that no ex-tended versions of the articles existed.

Quality assessment

Fig. 2. The quality evaluation of consistent and inconsistent pieces of informationas a function of stage of information evaluation in Study 2. Block 1 represents theinitial stage of information search; block 2 represents the later stage.

Results and discussion

Due to high inter-correlation between participants’ assessmentsof reliability and importance (r = .68, p < .001), these two evalua-tions were collapsed to form a new scale, entitled ‘perceived infor-mation quality’. This procedure is in line with previous research onbiased information evaluation in context of the classic selectiveexposure research paradigm (see Fischer et al., 2005).

Check for material effects

A 2 (type of information: consistent vs. inconsistent) � 2 (stage ofinformation seeking process: block 1 vs. block 2) � 2 (order of pre-sented information: order 1 vs. order 2) ANOVA revealed that orderwas not significantly associated with either information evaluationor the stage of decision making, all Fs < 1. The interaction betweentype of information and stage of evaluation (see below) was stillmarginally significant (p = .07) when order was controlled. Thus,it can be concluded that the results of the following analyses can-not be due to presentation order effects.

Check for effects of decision making

When making a preliminary decision, 18 participants opted forthe organic and 11 for the diet food snack stall. A 2 (type of infor-mation) � 2 (stage of information seeking process) � 2 (type of

decision) ANOVA with repeated measures on the first two factorsrevealed no significant three-way interaction, F < 1. However, thebasic two-way interaction between type of information and stageof information evaluation reported below was still significant(p = .04). Thus, type of decision is unlikely to confound the resultsreported below.

Information evaluation

For an overview, see Fig. 2. A 2 (type of information) � 2 (stage ofinformation seeking process) ANOVA with repeated measures on thelast factor revealed a significant main effect for type of information,F(1, 28) = 12.38, p = .002, g2 = .31, indicating that decision-consis-tent pieces of information (M = 5.81, SD = 2.01) were ascribed high-er quality than decision-inconsistent pieces of information(M = 4.30, SD = 1.66).

More importantly, this main effect was qualified by the pre-dicted two-way interaction, F(1, 28) = 4.44, p = .044, g2 = .14. Fol-low-up analyses revealed that in the first block, participantsascribed higher quality to consistent (M = 6.15) than inconsistentpieces of information (M = 4.16), F(1, 28) = 20.37, p < .001, g2 = .42,but this effect declined to a marginal difference between consistent(M = 5.48) and inconsistent pieces of information (M = 4.45) in thesecond block of information evaluation, F(1, 28) = 3.84, p = .06,g2 = .12.

Perceived threat to self-esteem

A 2 (type of information) � 2 (stage of decision process) ANOVArevealed a significant main effect for type of information,F(1, 28) = 12.44, p < .01, g2 = .31, indicating that decision-inconsis-tent pieces of information (M = 2.91, SD = 2.52) were perceived asmore threatening than decision-consistent pieces of information(M = 1.31, SD = 1.73). However, no significant interaction occurredbetween stage of decision process and type of information, F < 1.

Perceived pleasantness of information

A 2 (type of information) � 2 (stage of decision process) ANOVArevealed a significant main effect for type of information,F(1, 28) = 31.68, p < .001, g2 = .53, indicating that decision-consis-tent pieces of information (M = 5.51, SD = 2.19) were reported tobe more pleasant to read than decision-inconsistent pieces ofinformation (M = 2.66, SD = 1.65). However, no significant

Page 6: The process of selective exposure: Why confirmatory information search weakens over time

42 P. Fischer et al. / Organizational Behavior and Human Decision Processes 114 (2011) 37–48

interaction occurred between stage of decision process and type ofinformation, F(1, 28) = 2.40, p > .13, g2 = .08.

In summary, Study 2 revealed that the quality of decision-con-sistent pieces of information is perceived to be higher during theearlier (block 1) rather than later (block 2) stage of an informationevaluation process. However, no differences were found with re-gard to dissonance relevant processes (pleasantness and threat toself-esteem; see Frey, 1986). This result is not in line with theassumption that defense-motivational (dissonance) processes ac-count for the prior effect. However, it must be noted that this con-clusion is limited, since it is based on null results on the defense-motivational (dissonance-related) measures of pleasantness ofdecision-relevant information and self-esteem threat. These mea-sures may be sensitive to demand or impression managementeffects. In addition, it was not shown that these measures areindeed sensitive to the defense-motivational constructs they wereintended to assess. Because of these limitations, a third study wasconducted, which provided a more direct test of whether the pro-posed effect is driven by defense motivation.

Study 3

The aim of Study 3 was twofold. Firstly, an accountabilitymanipulation (derived from Simonson & Nye, 1992) was used inorder to further clarify whether the observed effect was based ondefense-motivational processes. Participants were assigned toone of two accountability conditions, with those in the highaccountability condition being told both that they would need toexplain their decision to the experimenter; and that their decisionwould be the basis for subsequent seminars at the University ofMunich. In contrast, participants in the low accountability conditionwere not given any such expectations. Secondly, the number ofavailable decision-relevant pieces of information was increased;providing two blocks of eight pieces of decision-consistent and -inconsistent information (double those of the two previous stud-ies). This was done in order to circumvent the possibility of ceilingeffects due to the low number of available pieces of information inthe preceding studies.

As in Study 1, it was expected that participants would exhibithigher levels of confirmatory information search in the first ofthe two information search blocks. From a defense-motivationalperspective, the high accountability manipulation should increaseparticipants’ motivation to defend their viewpoints, leading themto search for additional decision-consistent information in allstages of the information search process (that is, in both blocks 1and 2; cf. Jonas, Schulz-Hardt, & Frey, 2005). Moreover, this effectshould be increased in the final stage of information search, asthe psychological distance to being accountable for one’s decisionis gradually reduced (as the time at which the decision must be ex-plained to the experimenter draws closer). If the observed effect isbased on defense-motivational processes, the difference in confir-matory information search between blocks 1 and 2 should vanishin the high accountability group, since high accountability shouldmake participants equally motivated to defend their decision pref-erence at all stages of the information search, thus leading toequally high levels of exhibited confirmation bias in both searchstages.

Method

Participants and design

Forty-three students (33 female; 10 male) at the University ofMunich (aged between 19 and 30 years old; M = 23.86, SD = 2.80)took part in exchange for course credit. The study consisted of a

2 (stage of information search process: block 1 vs. block 2) � 2 (typeof information: decision-consistent vs. -inconsistent) � 2 (account-ability: low vs. high) factorial design with repeated measures onthe first and second factors.

Material and procedure

The study utilized the same decision case employed in Study 2;where participants were presented with the scenario of opening asnack-food stall and had to decide whether they would sell organicproducts or diet foodstuffs. Participants’ levels of accountabilitywere manipulated (cf. Simonson & Nye, 1992) through instructionsgiven to them immediately after learning about the decision sce-nario. Participants in the high accountability condition receivedthe following instruction: ‘‘Please consider the following point whenyou make your decision: after you finish this study you will have topersonally explain and justify your decision to the experimenter. Yourdecision outcome will then be the basis for discussion in student sem-inars on economic psychology carried out in the psychology depart-ment of the University of Munich. Thus, you will also be asked toleave your e-mail address in case the experimenter and seminar mod-ule convener have any questions for you about your decision”. In con-trast, participants in the low accountability condition received nosuch information.

After the accountability manipulation was provided and preli-minary decisions made, participants were confronted with 16pieces of additional information, half of which supported the or-ganic business idea, with the remainder supporting the diet prod-uct proposal. Thus, half of the additional pieces of informationsupported and half conflicted with the participants’ preliminarydecision. These 16 pieces of information were presented in twoblocks, each containing eight pieces of information (four pro-or-ganic, and four pro-diet). Participants were asked to indicatewhether or not they wanted to read more detailed information(one-page statements written by business idea experts; as in Stud-ies 1 and 2, these did not actually exist) about each individualargument after the experiment. In order to control for material pre-sentation effects, the information items included in block 1 andblock 2 were counterbalanced.

Results and discussion

Check for material effects and manipulation

A 2 (type of information: consistent vs. inconsistent) � 2 (stage ofinformation search process: block 1 vs. block 2) � 2 (order of pre-sented information: order 1 vs. order 2) ANOVA revealed that orderwas not significantly associated with searching for consistent orinconsistent pieces of information, nor with the stage of decisionmaking, all Fs < 1. The basic sequence effect reported below wasstill significant when order was controlled for in the overall analy-sis (p < .05). Thus, the results of the following analyses are unlikelyto be due to material effects. Moreover, in order to prevent demandeffects, accountability was not measured directly after the manip-ulation; instead being assessed during the debriefing session, byasking participants whether they believed that they would beinterviewed about their decision by the experimenter. All partici-pants in the high accountability condition believed the cover story;with none voicing suspicion.

Check for effects of decision making

Thirty-four participants initially opted for the organic snackstall, and nine for the diet one. Five participants changed their pref-erence (four from diet to organic, and one from organic to diet), so

Page 7: The process of selective exposure: Why confirmatory information search weakens over time

P. Fischer et al. / Organizational Behavior and Human Decision Processes 114 (2011) 37–48 43

that in their final decisions 37 participants decided for the organicand six for the diet snack stall. No significant three-way interactionbetween type of decision, type of information, and stage of infor-mation search process, F < 1, was found. In addition, when typeof decision was controlled in the full factorial design, the basic ef-fect of differential confirmatory information search in block 1 vs.block 2 (see below) was still marginally significant (p = .08). Thus,type of decision was not a confounding variable.

Information search

For an overview, see Fig. 3. A 2 (type of information: consistent vs.inconsistent) � 2 (stage of information search process: block 1 vs.block 2) � 2 (accountability: low vs. high) ANOVA with repeatedmeasures on the first two factors revealed a significant main effectfor type of information, F(1, 41) = 4.05, p = .05, g2 = .09, indicatingthat overall, participants selected more decision-consistent(M = 0.85, SE = 0.18) than decision-inconsistent pieces of informa-tion (M = 0.53, SE = 0.13). This main effect was qualified by a signif-icant two-way interaction between type of information and stage ofinformation seeking process, F(1, 41) = 4.46, p = .04, g2 = .10. How-ever, no significant three-way interaction between type of informa-tion, stage of information search, and accountability occurred,F(1, 41) = 0.83, p > .36, g2 = .02.

In terms of the two-way interaction, follow-up analyses sepa-rately conducted for block 1 and block 2 revealed that in block 1, par-ticipants exhibited a stronger preference for consistent (M = 1.00,SD = 1.36) than inconsistent pieces of information (M = 0.49,SD = 0.94), F(1, 42) = 5.86, p = .02, g2 = .12. In contrast, no differencebetween the number of consistent (M = 0.70, SD = 1.21) and incon-sistent (M = 0.58, SD = 1.03) pieces of information selected wasfound in the second information search block, F < 1.

Study 3 thus replicated the basic effect of decision makersexhibiting more confirmatory information search tendencies inthe initial (block 1) rather than later stages (block 2) of a decisionand information search process. This effect did not vanish whenthe motivation to search for decision-consistent information wasincreased by a classic high accountability manipulation (which iswhat would have been expected if the effect were due to de-fense-motivational forces). Some unexpected effects in this studyhave to be discussed: for example, the high accountability manip-ulation tended to increase participants’ preferences for consistent

Number of selected articles

Fig. 3. Confirmation bias (the number of selected consistent minus inconsistentpieces of information) as a function of accountability and stage of informationsearch process in Study 3. Block 1 represents the initial stage of information search;block 2 represents the later stage.

information in block 1, but not in block 2. It may be that the study’slow sample size did not yield enough statistical power to detect thethree-way interaction. It would be interesting for future researchto replicate this study with a higher sample size and test whetherthis is a random effect or not. Also, there seems to be a trend thatthe accountability manipulation reduced the total amount of infor-mation searched in block 1, but not in block 2. Although this pat-tern is not significant, one explanation for this finding might bethat high accountability participants tend to wait for the piecesof information offered in block 2, which they might find easier toremember in anticipation of the subsequent discussion with theexperimenter, due to their shorter temporal distance to the inter-view. Another shortcoming of Study 3 is that it did not incorporatea manipulation check for accountability in order to prevent poten-tial demand effects. Therefore, to increase the reliability of this testof whether defense-motivational processes are involved in the ef-fect or not, the next study also manipulated defense-motivationalinfluences, eliminating motivation to defend the made decisionand further testing the defense-motivational explanation of theeffect.

Study 4

One problem in detecting the underlying psychological mecha-nism for the effect is that the available pieces of decision-relevantinformation are classified as decision-consistent or -inconsistenton the basis of the participants’ prior decision. Therefore, it cannotbe said whether participants perceive decision-consistent informa-tion to be of higher quality because it is in line with their priordecision; whether they chose their decision alternative becausethey thought particularly deeply about those specific pieces ofinformation in advance; or whether there exists an unknown thirdfactor that determines both the initial choice and the response tothe subsequent pieces of decision-relevant information.6 In orderto solve this problem, another experiment was conducted, which re-moved any defense-motivational reason to exhibit a confirmatorysearch effect in one of the two information search blocks. Specifi-cally, it was manipulated whether participants had free choice inopting for a specific decision alternative, or whether they were as-signed to a particular decision alternative (thus eliminating freechoice and any associated potential processes based on defensemotivation; see Frey, 1986; Zanna & Cooper, 1974). If the effect isdue to defense-motivational (dissonance) processes, eliminating freechoice (and consequently, any motivation to defend one’s decision)should completely eliminate differential selectivity in both the initial(block 1) and later stages (block 2) of information search.

Method

Participants and design

In this study, 33 (21 female; 12 male) students at the Universityof Munich participated in exchange for course credit (ages rangedfrom 20 to 42 years; M = 25.12, SD = 4.81). The study consisted of a2 (stage of information search process: block 1 vs. block 2) � 2 (typeof information: consistent vs. inconsistent) � 2 (free choice: yes vs.no) factorial design with repeated measures on the first twofactors.

Material and procedure

The study employed the same decision case as in Studies 2 and3, which dealt with opening a diet or organic food snack stall. As in

6 We thank an anonymous reviewer for this important proposition.

Page 8: The process of selective exposure: Why confirmatory information search weakens over time

Number of selected articles

44 P. Fischer et al. / Organizational Behavior and Human Decision Processes 114 (2011) 37–48

previous studies, participants in the free choice condition could se-lect between the organic and diet food ideas. In contrast, partici-pants in the no-choice condition were all assigned to the organicfood idea.7 They were then told that due to the experimental design,it was important for them to take the perspective of someone whohad decided for the organic food alternative. No participant in thelow choice condition questioned this fixed-position assignment.

As in Study 3, participants were then confronted with 16 piecesof additional information presented in two blocks; with four piecesof information supporting the organic product s idea and four sup-porting the diet product one within each block. Participants wereasked to indicate which pieces of information they wanted to readin more detail after the experiment.

Fig. 4. Confirmation bias (the number of selected consistent minus inconsistentpieces of information) as a function of choice and stage of information searchprocess in Study 4. Block 1 represents the initial stage of information search; block 2represents the later stage.

Results and discussion

Check for effects of decision making

In the free choice condition, eleven participants initially chosethe organic food idea; and 10 participants chose the diet one. Fourparticipants changed their preference in the final decision (threefrom diet to organic, and one from organic to diet). No significantthree-way interaction was found between type of initial decision,type of information, and stage of information search process,F < 1. The interaction between type of information and stage ofinformation search was still marginally significant when type ofdecision was statistically controlled for, p < .10. Therefore, it is un-likely that the results reported below are confounded by type ofdecision.

Information search

For an overview, see Fig. 4. A 2 (stage of information search:block 1 vs. block 2) � 2 (type of information: consistent vs. inconsis-tent) � 2 (free choice: yes vs. no) ANOVA with repeated measureson the first two factors revealed a marginally significant main ef-fect for type of information, F(1, 31) = 3.47, p = .07, g2 = .10, indicat-ing that participants selected more consistent (M = 1.14, SE = 0.25)than inconsistent pieces of information (M = 0.84, SE = 0.21). Inaddition, a marginally significant two-way interaction betweentype of information and stage of information seeking process wasobserved, F(1, 31) = 4.43, p = .07, g2 = .10. Importantly for the clar-ification of the underlying psychological process, no two-way orthree-way interactions with choice occurred, all Fs < 1.

Follow-up analyses conducted separately for the two informa-tion search blocks revealed that in block 1, participants exhibiteda stronger preference for consistent (M = 1.45, SD = 1.58) thaninconsistent pieces of information (M = 0.94, SD = 1.34),F(1, 32) = 5.58, p = .02, g2 = .15. In contrast, no difference betweenthe number of consistent (M = 0.85, SD = 1.33) and inconsistent(M = 0.72, SD = 1.10) pieces of information selected was found inthe second block, F < 1.

To conclude, Study 4 revealed that increased levels of confirma-tory information search at the beginning (block 1) of the informa-tion search process occurred even when motivational incentivesfor confirmatory information search were completely removed.So far, evidence that confirmatory information processing is morepronounced in the initial (block 1) rather than later (block 2) stagesof decision-relevant information processing has been accumulated.However, what is still missing is a psychological explanation forthis finding. This investigation examined the assumption that peo-ple experience greater levels of arousal in the early stages of the

7 Due to a research assistant error, participants were unfortunately assigned only tothe organic food idea.

information search process (block 1) compared to later ones (block2), and thus reduce dissonance via exhibiting confirmatory infor-mation processing tendencies during this early stage (block 1). Incontrast to this expectation, Studies 2–4 failed to provide evidencefor a defense-motivational (dissonance) account underlying theeffect.

What else, then, could explain these findings? Confirmatoryinformation processing tendencies are not only due to defense-motivational (dissonance) processes, but also to accuracy-motiva-tional ones. From the accuracy-motivational perspective, decisionmakers are generally expected to search for the ‘best’ pieces of addi-tional decision-relevant information (cf. Schulz-Hard et al., 2010;see also Fischer et al., 2005; Fischer, Schulz-Hardt, et al., 2008).However, there seem to be general functional testing asymmetriesbetween decision-consistent and -inconsistent pieces of informa-tion, so that decision makers are more critical of inconsistent thanconsistent information; which is reflected in the systematic deval-uation of inconsistent compared to consistent information (Ditto &Lopez, 1992; Ditto et al., 1998; Fischer et al., 2005; Russo et al.,1996). In this regard, Schulz-Hardt et al. (2010) have shown thatperceived information quality is indeed a strong predictor of confir-matory information search; that is, by seeking the best pieces ofinformation, decision makers tend to systematically prefer deci-sion-consistent information (see also Fischer et al., 2005; Fischer,Schulz-Hardt, et al., 2008). Based on these considerations, it is con-ceivable that individuals are more concerned with seeking the bestpieces of information in initial (block 1) rather than later stages(block 2) of the information search process, and that it is thus morelikely that confirmatory information processes will occur duringthese initial stages. This reasoning will be tested in Study 5.

Study 5

The final study investigated the evaluation-based accuracy-motivational account. From this theoretical perspective, it wouldbe expected that differential levels of selectivity in informationsearch as a function of stage of information search are mediatedby varying levels of experienced decision importance. In the initialstages of an information search process, decision makers mayexperience higher levels of decision salience and thus decisionimportance (an indicator of accuracy motivation) than they do inlater stages. This should lead to the effect that decision makersare more inclined to search for subjectively assumed high qualityinformation in the initial rather than later stages of information

Page 9: The process of selective exposure: Why confirmatory information search weakens over time

P. Fischer et al. / Organizational Behavior and Human Decision Processes 114 (2011) 37–48 45

search. Since decision-consistent information is generally per-ceived to be of higher quality than inconsistent information (Ditto& Lopez, 1992; Ditto et al., 1998) decision-consistent informationshould receive a stronger, systematic selection advantage in thebeginning rather than end of an information search process. In con-trast, in the later stages of a search process, decision salience (andthus decision importance) should be relatively reduced, whichmight be reflected in reduced intentions of finding the qualitativelybest pieces of decision-relevant information, and thus reduced lev-els of selective exposure.

To test this argument, Study 5 measured decision importancebefore the first block of information evaluation, and then a secondtime prior to the presentation of the second block. According to theevaluation-based accuracy-motivational account, the effect of theblocks (initial vs. later stage) should be mediated by the differencein experienced decision importance between blocks 1 and 2. Inaddition, fatigue in blocks 1 and 2 was measured in order to testwhether the effect could be alternatively explained by the fact thatparticipants might lose interest in decision-relevant informationprocessing over time.

Method

Participants and design

Twenty-seven (21 female; 6 male) students of the University ofMunich participated in this study in exchange for course credit(ages ranged between 19 and 27 years; M = 21.96, SD = 1.79).Two participants had to be excluded from further analyses becauseof considerable amounts of missing data. The study consisted of a 2(stage of information search process: block 1 vs. block 2) � 2 (type ofinformation: consistent vs. inconsistent) factorial design with re-peated measures on both factors.

Material and procedure

To increase external validity, an alternative classic selectiveexposure decision paradigm was employed, in which participantswere asked to decide whether the contract of a manager namedMr. Miller should be extended or not (Fischer et al., 2005; Frey,1986). Mr. Miller’s job was to manage a fashion store for 1 year,with participants being asked to imagine that they were the store-owner. Participants were informed that Mr. Miller’s job perfor-mance had been mixed. For example, he was successful ingaining new customers by introducing new product lines, butother regular customers had been lost because they were notdrawn to the new fashion collections. In sum, positive and negativeaspects of Mr. Miller’s work were balanced. After reading the back-ground information, participants were asked to make a preliminarydecision as to whether Mr. Miller’s contract should be prolonged ornot. Subsequently, they were asked to indicate how much cogni-tive energy they wished to invest in making a good decision (deci-sion importance), on a scale from 0 (not at all) to 10 (extremely). Tomeasure fatigue, participants were asked to indicate the extent towhich they felt exhausted on a scale from 0 (not at all) to 10 (extre-mely), doing so before the administration of each of the two blocksof information (and thus twice overall).

Next, participants were informed that additional informationregarding the decision problem was available. The additional infor-mation consisted of 12 one-page statements produced by Mr. Mill-er’s former colleagues, and were presented in two blocks of sixpieces of information each. Each main thesis was summarized intwo to three sentences (as before, the full-length statements didnot exist). The main thesis contained the key argument of thecorresponding statement and indicated whether the colleague

was for or against a prolongation of Mr. Miller’s contract. An exam-ple of a main thesis in favor of an extension was: ‘‘Mr. Miller’s workhas been satisfactory; therefore, his contract should be extended”. Anexample of a main thesis against an extension of the contract was:‘‘Mr. Miller hasn’t fulfilled our expectations; therefore, his contractshouldn’t be extended”. Regardless of the participants’ initial deci-sion preference, three statements in each block supported theirdecision (decision-consistent information), and three statementsconflicted with it (decision-inconsistent information).

Within each block, participants were instructed to evaluate theexpected quality of all the available pieces of information with re-gard to their credibility (‘‘How credible do you expect this informa-tion to be?”; 0 = not at all, 10 = extremely) and importance (‘‘Howimportant will this information be for making a good decision?”;0 = not at all, 10 = extremely). Because the credibility and impor-tance assessments were highly correlated (a = .84), they were col-lapsed into an index of confirmatory information evaluation – acrucial indicator of confirmatory information processing (e.g.,Fischer, Greitemeyer, et al., 2008). Between blocks 1 and 2, partic-ipants again indicated how much cognitive energy they wanted toinvest in making a good decision (decision importance). When par-ticipants had completed their evaluation of the information inblock 2, they were asked to make a final decision as to whetherto extend Mr. Miller’s contract.

Results and discussion

Check for effects of decision making

Initially, seven participants decided for a prolongation of thecontract, and 18 against it. No participant changed his/her decisionduring the course of the experiment. However, a significant three-way interaction between type of decision, type of information, andstage of information evaluation was found, F(1, 23) = 11.06, p < .01,indicating that the effect was mainly due to the participants whodecided against extending the contract.

Check for effects of fatigue

A paired samples t-test revealed that participants did not expe-rience more fatigue in block 2 (M = 3.96, SD = 2.99) than block 1(M = 3.40, SD = 2.93), t(25) = �1.20, p = .24, d = 0.19. Thus, it is un-likely that the effect is due to differing levels of fatigue betweenblocks 1 and 2.

Information evaluation

A 2 (stage of information search: block 1 vs. block 2) � 2 (type ofinformation: consistent vs. inconsistent) ANOVA with repeated mea-sures on both factors revealed a significant main effect for type ofinformation, F(1, 24) = 19.42, p < .001, g2 = .45, indicating that par-ticipants evaluated consistent information (M = 6.89, SD = 0.97)more positively than inconsistent information (M = 5.30, SD = 1.43).

Moreover, a significant two-way interaction between type ofinformation and stage of information processing was observed,F(1, 24) = 5.14, p = .03, g2 = .18. Follow-up analyses conducted sep-arately for block 1 and block 2 revealed that in block 1, participantsevaluated consistent information (M = 6.95, SD = 1.18) more posi-tively than inconsistent information (M = 4.88, SD = 1.48),F(1, 24) = 23.40, p < .001, g2 = .49. This evaluation bias betweenconsistent (M = 6.84, SD = 1.24) and inconsistent information(M = 5.73, SD = 1.69) was less pronounced in block 2 of the infor-mation evaluation process F(1, 24) = 7.46, p = .01, g2 = .24. An addi-tional t-test revealed that the evaluation bias (evaluation ofconsistent minus inconsistent information) was significantly

Page 10: The process of selective exposure: Why confirmatory information search weakens over time

46 P. Fischer et al. / Organizational Behavior and Human Decision Processes 114 (2011) 37–48

stronger in block 1 (M = 2.07, SD = 2.14) than in block 2 (M = 1.11,SD = 2.04), t(24) = 2.27, p = .03, d = 0.46. This result replicated thebasic findings of the previous studies: people are more confirma-tory in the initial (block 1) rather than later (block 2) stages ofadditional decision-relevant information processing.

Decision importance (cognitive investment)

A paired samples t-test revealed that experienced decisionimportance was significantly stronger in block 1 (M = 6.32,SD = 3.13) than block 2 (M = 5.84, SD = 3.16), t(24) = 2.07, p < .05,d = 0.15. This result is in line with the assumption that decisionmakers exhibit increased levels of confirmatory information pro-cessing in the initial (block 1) rather than later stages (block 2)of decision-relevant information processing, as they experiencean increased level of decision importance in the initial (as opposedto later) processing stages.

Mediation analysis

Because the independent, dependent, and mediating variableswere within-factors, an analysis of covariance (ANCOVA) was usedto test for mediation. To examine whether the different levels ofdecision importance between blocks 1 and 2 mediated the effectof stage of information search on the evaluation of consistent vs.inconsistent information, a 2 (stage of information search: block1 vs. 2) � 2 (type of information: consistent vs. inconsistent) AN-COVA was conducted using the difference value of decision cer-tainty in block 1 minus decision certainty in block 2 as acovariate. This analysis revealed that the significant interaction be-tween stage of information search and type of information was nolonger significant when the potential mediator ‘decision certainty’was statistically controlled, F(1, 23) = 1.85, p = .19, g2 = .07,whereas decision certainty had a significant effect on confirmatoryinformation evaluation, F(1, 23) = 5.61, p = .027, g2 = .20.8 In linewith the mediation criteria set forth by Baron and Kenny (1986), thisanalysis shows that, in context of a full within-subjects design, deci-sion certainty mediates the effect of stage of information search onconfirmatory information evaluation.

In summary, Study 5 revealed that decision makers are moreconfirmatory in the initial stages of the information evaluationand search process because they experience higher levels of sub-jective decision importance than they do during the later stagesof the process. This finding is in line with the assumptions of theevaluation-based accuracy-motivational account of selective expo-sure introduced by Schulz-Hardt et al. (2010), which assumes thatdecision makers become more confirmatory in processing deci-sion-relevant information when accuracy motivation is increased,and pre-existing evaluation differences between consistent andinconsistent information are more pronounced (see also Fischeret al., 2005). Transferred to the present results, the increasedimportance of decision-relevant information processing in block1 leads to increased bias in information evaluation, because deci-sion makers are even more critical of inconsistent information. Incontrast, in block 2 decision importance is reduced, and the sys-tematic evaluation difference between consistent and inconsistentinformation decreases with it. In sum, increased decision impor-tance in block 1 motivates people to find the best decision-relevantpieces of information – and these are a priori the consistent onesdue to the processes assumed by Schulz-Hardt et al. (2010; see alsoDitto & Lopez, 1992; Ditto et al., 1998). The study also revealed that

8 Due to the fact that the experimental design was fully within-subjects, no Sobeltest could be performed.

the reported effect is not due to different levels of fatigue betweenblocks 1 and 2.

General discussion

The present paper began by questioning whether individuals’tendencies to prefer supporting to conflicting information aftermaking decisions vary within the sequential process of informationseeking. More specifically, it was asked whether decision makerswould be more selective in their information search and evaluationat the beginning rather than end of an information search process.The results of five studies suggest that this expectation was justi-fied: in the initial stages of information search (block 1) partici-pants exhibited stronger selectivity in both their informationsearch (Studies 1, 3, 4) and evaluation (Studies 2 and 5) than theydid in later stages (block 2). Furthermore, it was clarified that de-fense-motivational processes are not necessary to explain this ef-fect. Instead, different levels of experienced decision importancecan explain the effect from an accuracy-motivational perspective.Participants experience higher levels of decision importance inthe initial (block 1) rather than later (block 2) stages of the infor-mation evaluation and search process, and therefore attempt tofind the highest quality information at the earliest possible oppor-tunity. An analysis of covariance revealed that differing levels ofexperienced decision importance mediated the varying quantitiesof confirmatory information processing observed between the ini-tial (block 1) and later stages (block 2) of the information evalua-tion and search process.

Theoretical and practical implications

On a theoretical level, the present studies provide further in-sight into the post-decision processing of decision-relevant infor-mation; with particular regard to whether the selectivity ofinformation evaluation and search varies across the different tem-poral stages of its process. It has been shown that confirmatoryinformation processing is especially strong at the beginning of apost-decision information evaluation and search process. In addi-tion, it appears that defense-motivational (dissonance) processesare not able to explain this effect–instead, it is more related toaccuracy-motivational processes. Immediately after formulatingan initial decision (block 1), decision-makers experience a greatersense of decision importance, and are consequently more moti-vated to find the best pieces of decision-relevant information thanthey are in later stages of the information evaluation and searchprocess (block 2). In turn, this difference in decision importanceleads to decision-inconsistent information being tested more crit-ically in the initial stage, which makes it appear qualitativelyweaker in block 1 than in block 2 of the information search process.This finding supports the predictions of the evaluation model ofselective exposure set forth by Schulz-Hardt et al. (2010), whichproposes that selective exposure results from asymmetric testingbetween decision-consistent and inconsistent information. Whileinconsistent information is tested rather critically, consistent infor-mation is tested only at face value (Ditto & Lopez, 1992; Ditto et al.,1998; Fischer et al., 2005; Schulz-Hardt et al., 2010). As accuracymotivation and cognitive elaboration are increased, inconsistentinformation is tested even more critically, which leads to an in-creased evaluation and selection disadvantage for decision-incon-sistent information, and/or a systematic evaluation and selectionadvantage for decision-consistent information. The current re-search shows that this process is more pronounced right after aninitial decision, compared to during the later stages of the decisionand information search process.

Page 11: The process of selective exposure: Why confirmatory information search weakens over time

P. Fischer et al. / Organizational Behavior and Human Decision Processes 114 (2011) 37–48 47

Moreover, it is important to discuss these results in light of thefindings of Jonas et al. (2001), which (to the best of the authors’knowledge) is the only set of studies that has addressed theprocess of confirmatory information search. Jonas et al. (2001)exposed decision makers to additional decision-consistent and -inconsistent pieces of information, doing so either sequentiallyor simultaneously. Their principal finding was that presentinginformation sequentially led to a stronger confirmation bias insubsequent information search than did presenting it simulta-neously. This effect was mediated by increased commitment tothe initial decision, which developed in the sequential informa-tion presentation procedure. In this condition, participants com-pared the information with their initial decision, and thisincreased their focus and commitment towards their decision;in contrast, participants in the simultaneous information presen-tation condition compared additional information with otheravailable pieces of information, which led to an increased focuson the available pieces of information (information focus), butnot on the initially made decision. Interestingly, and in contrastto the findings of Jonas et al. (2001), the reported effect seemsto be based not on defense-motivational processes, but on moreaccuracy-motivational ones.

However, in this context it is important to note that the exper-imental procedure utilized in the current studies is considerablydifferent from that of Jonas et al. (2001). Whereas the currentparticipants were exposed to two different blocks of four to eightpieces of decision-relevant information each, participants in theJonas et al. (2001) studies were confronted either with all ofthe available pieces of information at once (simultaneous infor-mation presentation) or with one piece of information at a time(sequential information presentation). It appears that the processof information search is highly sensitive to contextual factors,such as differing information presentation and search procedures.In addition, the underlying psychological processes can varystrongly between different types of information presentationand search procedures. Whereas the information presentationprocedure employed in the Jonas et al. (2001) study induced de-fense-motivational processes (commitment to the initial deci-sion), the current effect does not necessarily require defense-motivational processes to be explained; and seems to be basedon accuracy-motivational processes that are related to increaseddecision importance and associated a priori testing differences be-tween consistent and inconsistent information (Ditto & Lopez,1992). Future research is needed to further disentangle the moti-vational and cognitive processes underlying effects in informationsearch processes.

From a practical perspective, the present research suggeststhat de-biasing techniques are particularly needed at the begin-ning of a decision and information search process; especiallythose that reduce the evaluation and search bias during thisstage of information search. For example, decision makers couldemploy the ‘consider the opposite’ technique (Lord, Lepper, &Preston, 1984), which requires individuals to process argumentsnot only from their own point of view, but also from the oppo-site perspective. This technique has been shown to reduce the‘‘prior belief effect,” with the assimilation of new informationbeing less biased in the direction of prior opinion when partici-pants ‘‘consider the opposite” than when they do not. The pres-ent findings suggest that decision makers should employ this de-biasing technique (particularly at the beginning of informationsearch) in order to reduce selective information evaluation andsearch. Future research examining the effectiveness of this tech-nique at different stages of the information search process wouldbe fruitful.

Limitations and future perspectives

On the basis of the present results, it can be reasonably confi-dently stated that the underlying psychological process is based ondifferences in accuracy motivation (i.e., perceived decision impor-tance). However, possible alternative explanations for this effect re-main. For example, in the later stages of information search,participants might experience a significant demand for consideringcounter-arguments (particularly if they expect to be asked furtherquestions after the search). In short, participants might feel thatthe experimenter expects them to be more balanced by the end ofthe information search sequence. Another alternative explanationcould be reduced attention and an associated lessening of discrimi-native processing, which might decrease during the experimentalprocedure (because participants are exhausted, lose interest in theexperimental task, or simply habituate to it). Although the resultsof Studies 3 and 4 (where the motivational incentives to be balancedwere systematically varied), as well as Study 5 (which found no dif-ferences between blocks 1 and 2 in terms of reported fatigue) speakagainst most of these alternative accounts, future research shouldaddress them. For example, pieces of information whose headlinesdo not obviously reveal whether a specific argument is consistentor inconsistent with the participant’s prior decision could be em-ployed (subtler cues could be integrated into the additional piecesof information used, which would still correlate with the advocateddecision preference or attitude). In addition, one could also use avariety of decision problems and associated pieces of informationin a single experiment, so that unbalanced information search with-in a specific decision topic is less obvious than in the present studies.

Moreover, it has to be asked whether the present set of studiessufficiently distinguished between defense and accuracy-motiva-tional processes underlying the stronger selectivity at the beginningof the information search process. Firstly, it is possible that the em-ployed measures of defense-motivational processes were simplynot sensitive to the psychological processes that may have oc-curred. This is a problem that is well known in recent decision mak-ing research (e.g., Johnson, Häubl, & Keinan, in press; Payne et al.,1993; Weber et al., in press), since measures of decision outcomesare often insufficient to distinguish between two contradicting psy-chological processes. Nevertheless, future research on biased infor-mation search and underlying psychological processes in early vs.late decision stages should broaden the variety of decision out-comes and process measures used, including the complementaryuse of both implicit and explicit measures as well as more sophisti-cated scale items (e.g., to measure decision importance). Future re-search should also directly manipulate the expected psychologicalprocess based on decision certainty. For example, a study could bedesigned that externally induces high levels of decision importancein both the initial and later stages of the information search process.If decision certainty is indeed the crucial mediator of the effect, thenthis type of manipulation should increase confirmatory informationevaluation and search in both the initial and later stages of the infor-mation search process. Finally, further clarifying the underlyingcausal psychological processes should help in the design of real-lifeenvironments for practical decision makers. The aim should be toreduce the impact of situational factors (such as the early stage ofinformation evaluation and search immediately after decision mak-ing) that systematically increase bias in decision-relevant informa-tion processing and search.

Furthermore, it would be an interesting endeavor for future re-search to test whether the effect would change if participants havea chance to read actual, extended versions of the additional piecesof information they are given. In deriving its methodology fromprevious classic selective exposure research, the current studies

Page 12: The process of selective exposure: Why confirmatory information search weakens over time

48 P. Fischer et al. / Organizational Behavior and Human Decision Processes 114 (2011) 37–48

simply used short main arguments that gave decision makers anidea of the content they could expect if they chose a specific pieceof information. Extended versions of the additional decision-rele-vant pieces of information might lead the confirmation bias to lastfor a longer period of time, as reading consistent information(especially in an initial stage of information search) may keep par-ticipants in a mindset that increases the perceived quality differ-ences between consistent and inconsistent information. Inaddition, using extended versions of articles would meet real-lifecircumstances of decision making (and associated informationsearch) more accurately than the short argument version used byclassic selective exposure research.

It is important to note that when explaining selective exposureand confirmatory information processing effects, the defense andaccuracy-motivational processes are not necessarily mutuallyexclusive. For example, dissonance reduction effects may be muchmore important when people make real-life decisions, which canhave varying potential consequences. Real-life decision making in-volves higher levels of commitment and involvement than thehypothetical decision cases used in the present studies (cf. Frey,1986). Future selective exposure research should employ real-lifedecision contexts and investigate whether corresponding potentialeffects can also be explained in a more accuracy-motivational fash-ion. The present hypothetical decision cases at least suggest thatdefense-motivational processes are not necessarily required to ex-plain why confirmatory information processing is stronger in theinitial (block 1) than later stages (block 2) of information evalua-tion and search processes.

Finally, it needs to be asked whether similar findings would beobtained in other, related domains of confirmatory informationprocessing. For example, other frameworks subsume the term con-firmation bias under the concept of a ‘positive test strategy’ (Klay-man & Ha, 1987), which means that individuals are inclined todevalue information that contradicts a hypothesis or opinion theyhold (e.g., Nisbett & Ross, 1980; Ross & Lepper, 1980), and conductbiased tests that provide only a low risk of leading to disconfirmingresults (e.g., Snyder & Swann, 1978), or fail to falsify their currentdecision or hypothesis (Klayman & Ha, 1987). It would be benefi-cial for future research to investigate whether similar differencesbetween the initial and later stages of information processing alsooccur in these research paradigms’ studies of confirmation biases.

References

Baron, R. M., & Kenny, D. A. (1986). The moderator–mediator variable distinction insocial psychological research: Conceptual, strategic, and statisticalconsiderations. Journal of Personality and Social Psychology, 51, 1173–1182.

Ditto, P. H., & Lopez, D. F. (1992). Motivated skepticism: Use of differential decisioncriteria for preferred and nonpreferred conclusions. Journal of Personality andSocial Psychology, 63, 568–584.

Ditto, P. H., Scepansky, J. A., Munro, G. D., Apanovitch, A. M., & Lockhart, L. K. (1998).Motivated sensitivity to preference-inconsistent information. Journal ofPersonality and Social Psychology, 75, 53–69.

Festinger, L. (1957). A theory of cognitive dissonance. Stanford, CA: StanfordUniversity Press.

Fiedler, K., Brinkmann, B., Betsch, T., & Wild, B. (2000). Sampling approach to biasesin conditional probability judgments: Beyond base rate neglect and statisticalformat. Journal of Experimental Psychology: General, 129, 399–418.

Fischer, P., Kastenmueller, A., Greitemeyer, T., Fischer, J., Frey, D., & Crelley, D. (inpress). Threat and selective exposure: The moderating role of threat anddecision context on confirmatory information search after decisions. Journal ofExperimental Psychology: General.

Fischer, P., Greitemeyer, T., & Frey, D. (2008). Self-regulation and selective exposure:The impact of depleted self-regulation resources on confirmatory informationprocessing. Journal of Personality and Social Psychology, 94, 382–395.

Fischer, P., Jonas, E., Frey, D., & Kastenmüller, A. (2008). Selective exposure anddecision framing: The impact of gain and loss framing on biased informationseeking after decisions. Journal of Experimental Social Psychology.

Fischer, P., Jonas, E., Frey, D., & Schulz-Hardt, S. (2005). Selective exposure toinformation: The impact of information limits. European Journal of SocialPsychology, 35, 469–492.

Fischer, P., Schulz-Hardt, S., & Frey, D. (2008). Selective exposure and informationquantity: How different information quantities moderate decision makers’

preference for consistent and inconsistent information. Journal of Personalityand Social Psychology, 94, 231–244.

Frey, D. (1986). Recent research on selective exposure to information. In L.Berkowitz (Ed.). Advances in experimental social psychology (Vol. 19, pp. 41–80).New York: Academic Press.

Greitemeyer, T., Fischer, P., Frey, D., & Schulz-Hardt, S. (2009). Biased assimilation:The role of source position. European Journal of Social Psychology, 39, 22–39.

Greitemeyer, T., & Schulz-Hardt, S. (2003). Preference-consistent evaluation ofinformation in the hidden profile paradigm: Beyond group-level explanationsfor the dominance of shared information in group decisions. Journal ofPersonality and Social Psychology, 84, 322–339.

Greitemeyer, T., Schulz-Hardt, S., Brodbeck, F., & Frey, D. (2006). Informationsampling and group decision making: The effects of an advocacy procedure andtask experience. Journal of Experimental Psychology: Applied, 12, 31–42.

Holton, B., & Pyszczynski, T. (1989). Biased information search in the interpersonaldomain. Personality and Social Psychology Bulletin, 15, 42–51.

Johnson, E. J., Häubl, G., & Keinan, A. (in press). Aspects of endowment: A querytheory account of loss aversion for simple objects. Journal of ExperimentalPsychology: Learning, Memory, and Cognition.

Johnston, L. (1996). Resisting change: Information-seeking and stereotype change.European Journal of Social Psychology, 26, 799–825.

Jonas, E., & Frey, D. (2003). Information search and presentation in advisor–clientinteractions. Organizational Behavior and Human Decision Processes, 91, 154–168.

Jonas, E., Graupmann, V., & Frey, D. (2006). The influence of mood on the search forsupporting vs. conflicting information. Personality and Social Psychology Bulletin,32, 3–15.

Jonas, E., Schulz-Hardt, S., & Frey, D. (2005). Giving advice or making decisions insomeone else’s place – The influence of impression, defense and accuracymotivation on the search for new information. Personality and Social PsychologyBulletin, 31, 977–990.

Jonas, E., Schulz-Hardt, S., Frey, D., & Thelen, N. (2001). Confirmation bias insequential information search after preliminary decisions: An expansion ofdissonance theoretical research on ‘‘selective exposure to information”. Journalof Personality and Social Psychology, 80, 557–571.

Kastenmüller, A., Greitemeyer, T., Jonas, E., Fischer, P., & Frey, D. (in press). Selectiveexposure: The impact of collectivism and individualism. British Journal of SocialPsychology.

Klayman, J., & Ha, Y. W. (1987). Confirmation, disconfirmation, and information inhypothesis testing. Psychological Review, 94, 211–228.

Kray, L. J., & Galinsky, A. D. (2003). The debiasing effect of counterfactual mind-sets:Increasing the search for disconfirmatory information in group decisions.Organizational Behavior and Human Decision Processes, 91, 69–81.

Lord, C. G., Lepper, M. R., & Preston, E. (1984). Considering the opposite: A correctivestrategy for social judgment. Journal of Personality and Social Psychology, 47,1231–1243.

Lundgren, S. R., & Prislin, R. (1998). Motivated cognitive processing and attitudechange. Personality and Social Psychology Bulletin, 24, 715–726.

Nisbett, R., & Ross, L. (1980). Human inference: Strategies and shortcomings of socialjudgment. Englewood Cliffs, New Jersey: Prentice Hall.

Payne, J. W., Bettman, J. R., & Johnson, E. J. (1993). The adaptive decision maker.Cambridge: University Press.

Pyszczynski, T., & Greenberg, J. (1987). Toward an integration of cognitive andmotivational perspectives on social inference: A biased hypothesis-testingmodel. In L. Berkowitz (Ed.), Advances in experimental social psychology (20 ed.,pp. S297–S340). New York: Academic Press.

Rhine, R. J. (1967). The 1964 presidential election and curves of information seekingand avoiding. Journal of Personality and Social Psychology, 5, 416–423.

Ross, L., & Lepper, M. R. (1980). The perseverance of beliefs: Empirical andnormative considerations. New Directions for Methodology of Social andBehavioral Science, 4, 17–36.

Russo, J. E., Medvec, V. H., & Meloy, M. G. (1996). The distortion of product informationduring decisions. Organizational Behavior and Human Decision Processes, 66,102–110.

Schulz-Hardt, S., Fischer, P., & Frey, D. (2010). Confirmation bias in accuracy-motivated decision-making: A cognitive explanation for biased informationseeking. Unpublished manuscript.

Schulz-Hardt, S., Frey, D., Lüthgens, C., & Moscovici, S. (2000). Biased informationsearch in group decision making. Journal of Personality and Social Psychology, 78,655–669.

Schwarz, N., Frey, D., & Kumpf, M. (1980). Interactive effects of writing and readinga persuasive essay on attitude change and selective exposure. Journal ofExperimental Social Psychology, 16, 1–17.

Simonson, I., & Nye, P. (1992). The effect of accountability on susceptibility to decisionerrors. Organizational Behavior and Human Decision Processes, 51, 416–426.

Snyder, M., & Swann, W. B. (1978). Behavioral confirmation in social interaction: Fromsocial perception to social reality. Journal of Experimental Social Psychology, 14,148–162.

Svenson, O. (1992). Differentiation and consolidation theory of human decisionmaking: A frame of reference for the study of pre- and post-decision processes.Acta Psychologica, 80, 143–168.

Weber, E. U., Johnson, E. J., Milch, K. F., Chang, H., Brodscholl, J. C., & Goldstein, D. G.(in press). Asymmetric discounting in intertemporal choice: A query theoryaccount. Psychological Science, 18, 516–523.

Zanna, M. P., & Cooper, J. (1974). Dissonance and the pill: An attribution approach tostudying the arousal properties of dissonance. Journal of Personality and SocialPsychology, 29, 703–709.