Blog

The Wisdom of Some: Do We Always Need HighConsensus to Shape Consumer Behavior?Michael R. Sciandra, Cait Lamberton, andRebecca Walker ReczekFrom the Food and Drug Administration’s efforts to prompt healthier eating to the Environmental ProtectionAgency’s desire to prompt people to engage in environmentally friendly behaviors, a wide range of policymakers aim to persuade consumers. To do so, they must decide how and whether to use information aboutthe behavior of other consumers as part of their persuasive message. In four experimental studies, theauthors demonstrate that the persuasive advantage of high- versus low-consensus information dependson the target consumer’s trait level of susceptibility to interpersonal influence (SII). Low-SII consumersdifferentiate between low- and high-consensus information, such that they are more persuaded byhigh-consensus information. In contrast, high-SII consumers find any cue about the behavior of otherspersuasive, regardless of whether it is high or low consensus. Importantly, this finding suggests that policymakers may find success motivating behavioral change even in low-consensus situations. The authors closeby reporting data from two broadscale correlational surveys that identify behavioral, psychographic, anddemographic characteristics related to consumer SII as well as domains in which low consensus currentlyexists, so that policy makers can identify and target these individuals and related issues.Keywords: persuasion, susceptibility to interpersonal influence, social norms marketing, consensus, social proofPolicy makers often want to persuade people to change their thoughts and behaviors. For example, politicians frequently stump for votes, advocate for change, and seeksupport for new policies or legislation. Similarly, state and localgovernmental agencies and nonprofit nongovernmental organizations (NGOs) continually encourage consumers to engagein a multitude of desirable behaviors, including eating healthyfoods, staying physically active, recycling, and conserving energy. As such, policy makers frequently employ persuasivemessages designed to encourage specific behaviors. For example, in May 2015, New York City launched a campaign designed to reduce consumer waste and generated awarenessthrough persuasive messages on buses and subways and indigital communications (Gerlat 2015).Prior research in psychology and marketing has indicated thatthe success of such persuasive messages is largely contingent ontwo factors: (1) the credibility of the message source (i.e., thepolicy maker or organization; Petty and Briñol 2008) and (2) thestrength or quality of the message argument (i.e., how stronga case is made for the advocated attitude or behavior; Petty andCacioppo 1984). In general, most policy makers are endowedwith source credibility by virtue of political power (in the case ofelected officials), legal authority (e.g., state and local government agencies), or expertise (attributable to people within agovernmental or nongovernmental agency as a result of formaleducation or experience with a particular societal issue). Furthermore, once established, source credibility cannot be easilyaltered within a given persuasive message.However, policy makers do have considerable control overthe nature and strength of the arguments advanced in thepersuasive communications they employ. One aspect of message strength has to do with the behavior of other people. Thatis, messages can provide high-consensus information, whichstates that a majority of people engage in a given behavior, orlow-consensus information, which states that a minority ofpeople do so. In general, high-consensus messages are perceived as providing a stronger argument. The normative endorsement of a majority of consumers presents a compellingargument that other people should behave in the same waybecause it is either objectively superior or socially desirable(Aarts and Dijksterhuis 2003; Cialdini, Kallgren, and Reno1990; Naylor, Lamberton, and West 2012). For example, highlighting the fact thatmost students do not engage in binge-drinkingbehavior effectively curbed the consumption of alcohol amongcollege students (Haines and Spear 1996). Similarly, informinghotel patrons that the majority of guests reuse their towels increased reuse of towels in a hotel field study (Goldstein, Cialdini,and Griskevicius 2008).However, what can be done when policy makers want toprompt behaviors that are presently low consensus—for example,Michael R. Sciandra is Assistant Professor of Marketing, Dolan Schoolof Business, Fairfield University (e-mail: [email protected]). CaitLamberton is Fryrear Faculty Fellow and Associate Professor of Marketing, Katz Graduate School of Business, University of Pittsburgh(e-mail: [email protected]). Rebecca Walker Reczek is AssociateProfessor of Marketing, Fisher College of Business, Ohio State University (e-mail: [email protected]). Barbara Bickart served as associateeditor for this article.© 2017, American Marketing Association Journal of Public Policy & MarketingISSN: 0743-9156 (print) Vol. 36 (1) Spring 2017, 15–351547-7207 (electronic) 15 DOI: 10.1509/jppm.14.123to make prosocial but atypical choices (e.g., considering laborpractices of manufacturers as an important determinant ofwhich brand to buy, carrying reusable grocery bags, donatingto charities) or act in ways that may improve well-being evenwhen others are not engaging in the behavior (e.g., gettinga yearly flu shot, eating the daily recommended servingsof fruits and vegetables)? To the extent that low-consensusmessages are viewed as weaker arguments for a given behavior,some research has argued that they may be unsuccessful atpersuading consumers. For instance, Gerber and Rogers (2009)note that persuasive messages indicating that voter turnout iscurrently low actually depressed, rather than increased, voting.In light of findings such as these, policy makers might concludethat they should only use high-consensus information in persuasive messages.However, we argue that all hope is not lost in cases with lowconsensus. In this article, we identify consumers who do notneed high-consensus information to shape their behavior. Inparticular, we explore the role of consumers’ susceptibility tointerpersonal influence (SII; Bearden, Netemeyer, and Teel1989; McGuire 1968)—a psychographic measure with reliabledemographic correlates—in determining sensitivity to persuasivemessages containing high- versus low-consensus information.Across four experiments, we demonstrate that whether highconsensus information is more effective than low-consensus information at persuading consumers depends on the consumers’level of SII. To do so, we hold the persuasion goal and the sourceof the persuasive message constant. We manipulate only argument strength by varying consensus information to be high orlow. We demonstrate that for low-SII consumers, the degree ofpersuasion is contingent on consensus level (i.e., whether it ishigh or low consensus), whereas high-SII consumers find anyconsensus information persuasive, regardless of its level. Asa result, high-SII consumers can even be persuaded by thepresence of low-consensus information.We close by reporting results from two broadscale correlational surveys that identify behavioral, psychographic, anddemographic characteristics related to consumer SII, as wellas behaviors likely to be desirable to policy makers that arecurrently low consensus. These results offer targeting directionfor the promotion of new or less popular behaviors, consistentwith previous work in the marketing and public policy domainsuggesting that psychographic differences can be used to informinterventions (Rose, Bearden, and Manning 1996; Wood 2012).As such, our work can provide hope in situations in whichdesirable behaviors are low consensus, prompting some segment of consumers to begin to build the consensus that maypersuade others to conform.Theoretical DevelopmentProcessing Persuasive MessagesPetty and Cacioppo’s (1986) elaboration likelihood model hasarguably become the dominant model of persuasion in marketing (Petty, Cacioppo, and Schumann 1983; Shrum et al.2012) and public policy (Angst and Agarwal 2009; Rucker andPetty 2006) research. One of the central arguments of Petty andCacioppo’s theory is that messages can be processed centrally(i.e., carefully and with effort) or peripherally (i.e., with littlecare). Elaboration likelihood model research has suggested thatpeople are likely to centrally process most policy-related messages for two reasons. First, public policy messages are oftenrelated to important issues (e.g., health, financial security,safety) that people find personally relevant, a key determinant ofinvolvement (Celsi and Olson 1988), and prior research hasshown that increased involvement leads to more central routeprocessing (Petty and Cacioppo 1979; Petty, Cacioppo, andSchumann 1983). Second, many public service announcements(PSAs) from policy makers and NGOs use emotion-basedappeals (e.g., an anti–texting and driving campaign that usesa fear appeal), and emotions have also been shown to increasemotivation to process (Tiedens and Linton 2001).When consumers are processing centrally, both sourcecredibility (Petty and Briñol 2008) and argument strength (Pettyand Cacioppo 1984) have been identified as key elements thatdetermine the effectiveness of the persuasive message. In general, sources perceived as credible are more persuasive than lesscredible sources (Lirtzman and Shuv-Ami 1986; Maddux andRogers 1980; Ohanian 1991; Watts and McGuire 1964), andstrong arguments are more persuasive than weak arguments(Batra and Stayman 1990; Petty and Cacioppo 1984). Thus, wewould anticipate that strong arguments yield strong persuasion.However, we propose that this is not always true. Specifically,we argue that a psychographic characteristic, SII, influences howpeople integrate consensus information into their judgments ofargument strength; thus, SII affects an argument’s persuasiveness (Petty and Wegener 1998). That is, even when consumersare processing a persuasive message centrally, individual differences may direct different amounts of attention to variousmessage components, making them more or less important inpersuasion. In general, our proposition is consistent with researchsuggesting that individual difference measures can influencepersuasion (Cacioppo and Petty 1982; Cacioppo, Petty, andMorris 1983) and with previous research on the power ofconsensus showing that high consensus sometimes mattersand sometimes does not (Maass and Clark 1984; Moscovici1985; Wood et al. 1994). We detail our predictions in thefollowing section.SII and the Power of ConsensusSusceptibility to interpersonal influence is “the need to identifyor enhance one’s image with significant others through acquisition and use of products and brands, the willingness toconform to the expectations of others regarding purchasingdecisions, and/or the tendency to learn about products andservices by observing others and/or seeking information fromothers” (Bearden, Netemeyer, and Teel 1989, p. 474). Susceptibility to interpersonal influence has been shown to havea wide range of effects on consumer product preferences andmessage responses. For example, people high in SII prefer products that induce positive attributions from others (Netemeyer,Bearden, and Teel 1992) and that are socially visible (Batra,Homer, and Kahle 2001).Not surprisingly, prior work has argued that the persuasivepower of information about the behavior of others varies withSII (Bearden, Netemeyer, and Teel 1989; Martin, Wentzel, andTomczak 2008; McGuire 1968; Mourali, Laroche, and Pons2005). In general, people low in SII are motivated to makedecisions on the basis of what they believe to be “correct” andtherefore do not make decisions solely to fit in with others16 The Wisdom of Some(Batra, Homer, and Kahle 2001; Bearden, Netemeyer, and Teel1989; Netemeyer, Bearden, and Teel 1992; Wooten and Reed2004). Therefore, we predict that people low in SII will use boththe message source and the argument strength when evaluatingpolicy-relevant messages. Furthermore, we believe these consumers will differentiate between low- and high-consensusinformation when making a decision. This is due to the informational social influence of consensus information; lowerSII people are more likely to conform to high-consensus thanlow-consensus information because people assume that theactions of a majority of others reflect the more “correct”behavior (Thibaut and Kelley 1959). In general, social prooftheory acknowledges that people determine what is correct ina certain situation by looking to the behavior of others (Cialdini2009; Lun et al. 2007). Greater consensus will therefore beinterpreted as a stronger message argument on the basis of thesocial proof it provides (Cialdini, Kallgren, and Reno 1990,1991). Thus, we predict that, all else being equal, lower-SII peoplewill be more persuaded by information indicating that theadvocated behavior is a high- (vs. low-) consensus behavior.In contrast, high-SII people are primarily motivated to makesocially safe decisions (Wooten and Reed 2004). Those high inSII have been shown to easily trust the judgments and behaviorsof a single interpersonal source and to view this information asa reliable foundation of reality and sufficient for making decisions (Deutsch and Gerard 1955; Mourali, Laroche, and Pons2005). Therefore, we argue that even though they are processingcentrally, people high in SII will primarily attend to whetheranyone has engaged in the action rather than to the exact level ofconsensus provided. As a result, unlike low-SII people, wepredict that high-SII people do not need high-consensus information to be persuaded. The presence of even a small numberor proportion of supporters may be a strong argument to thisgroup. As such, they will be equally persuaded by either low- orhigh-consensus messages. This prediction is consistent with thedefinition of SII, which does not provide any guidance on howmany “others” are necessary to persuade or influence the behavior of those high in the SII trait. Thus, we predict that, ceterisparibus, high-SII people are just as persuaded by informationthat the behavior in the persuasive message is high consensus asthey are when informed the behavior is low consensus.While this prediction is consistent with a broader literaturebase acknowledging the potential for small groups or singleindividuals to wield considerable persuasive power (Maass andClark 1984; Moscovici 1985; Wood et al. 1994), our researchdiffers from this prior work in that we assess the influence of anunidentified minority (low consensus) or majority (high consensus) of people. In particular, prior research has found thatinformation on minorities is most influential when the identityof the minority is known. For example, information froma single individual, such as a market maven (Feick and Price1987) or opinion leader (Iyengar, Van den Bulte, and Valente2011; King and Summers 1970) can have a significant influenceon opinions and product adoption decisions. However, opinionleaders are often sought out for their expertise or influentialposition in a network (Feick and Price 1987); their identity isknown and part of the reason for their influence from a minorityposition. Similarly, the theory of reasoned action (Ajzen andFishbein 1980; Fishbein and Ajzen 1975) recognizes the role ofnorms in influencing behavioral intentions, noting that a singleindividual or small group can be particularly persuasive if thebeliefs of the minority are highly valued. However, again, fora minority to be influential, information about the identity of theminority individual or group is paramount. In contrast, we investigate situations in which consumers do not have any information about the identity of the minority or majority being usedin the persuasive message. Formally, we therefore propose:H1: When evaluating persuasive messages, as SII increases, therelative advantage of high-consensus information over lowconsensus information in generating persuasion decreases.We further predict that this failure to distinguish betweenpersuasive messages containing information about high- andlow-consensus behaviors is because high-SII people are primarily vigilant in detecting the

presence or absence of anyconsensus information but have a weaker focus on the level ofconsensus provided. That is, even though high-SII consumersare processing policy-related messages centrally, they treat consensus information more as a peripheral cue, focusing only onwhether it is there and not on the detail of whether it is high orlow. Note that this prediction is congruent with dual processtheories of persuasion, which acknowledge that peripheral cuescan be used in central route processing (Chaiken, Liberman, andEagly 1989; Eagly and Chaiken 1993). Thus, we are proposingthat attentional differences in the processing of consensus information drive the differences we predict for high- and low-SIIconsumers. This effect should therefore be moderated as follows:H2: When evaluating persuasive messages, attentional cues highlighting whether information about others’ behavior is highversus low consensus moderate the effect of SII on persuasionsuch that (a) when no attentional cue is given, as SII increases,the relative advantage of high-consensus information over lowconsensus information in generating persuasion decreases, and(b) when an attentional cue is given, the relative advantage ofhigh-consensus information over low-consensus information ingenerating persuasion is preserved.Overview of StudiesNext, we report the results of four experiments and twobroadscale surveys designed to test our predictions about SIIand the (un)importance of high consensus in certain segments;we then highlight the practical utility of our findings. Consistentwith recent recommendations (Murayama, Pekrun, and Fiedler2014; Schmidt 2009), we test our hypotheses in multiple contexts that may be relevant to policy makers and marketers,using both lexical and quantitative manipulations of consensuslevel and testing for replication in projected and real behavior.In Study 1a, we probe the interplay between consumer SII andconsensus information using an ethically based purchase decision. In Study 1b, we investigate the impact of consensusinformation on healthy eating habits. Study 2 assesses theimpact of consensus information and SII on environmentallyfriendly behaviors. Finally, Study 3 evaluates charitable contribution decisions and examines the impact of attentional cuesemphasizing high-consensus and low-consensus information,demonstrating process through moderation as advocated bySpencer, Zanna, and Fong (2005). To isolate the effect of SIIon susceptibility to persuasion by messages containing highversus low-consensus information, in all of our studies wemanipulate only argument strength, as reflected in consensusinformation. Within each study, we keep both the persuasiongoal of the message (i.e., the desired or advocated behavior) andJournal of Public Policy & Marketing 17the source of the persuasive message constant. Finally, we report the results of two correlational studies that isolate demographic and psychographic characteristics associated with SIIand identify low consensus issues for which our findings maybe most helpful. These findings can help policy makers segmentthe market using SII and choose consensus-based messages appropriately for segments with high versus low levels of SII.Study 1aStudy 1a examines consumers’ reactions to lexically describedhigh- and low-consensus information in a car-buying scenario.We adapt the conjoint procedure and ethical decision-makingstimulus employed by Irwin and Naylor (2009) to quantify reactions to high- versus low-consensus information for respondents at different levels of SII. In this study, we focus onbehavior that is prosocial and policy relevant: taking the laborpractices of the manufacturer into account when making a carpurchase decision.MethodSeventy-six participants (43% female; Mage = 36.5 years,age range 19–69 years) recruited using Amazon MechanicalTurk (MTurk) participated in Study 1a in exchange for a smallmonetary incentive. Study 1a employed an SII × consensus information (high vs. low) between-subjects design, whereby SIIwas a measured continuous variable and consensus informationwas manipulated. Before beginning the study, participantscompleted an instructional manipulation check (IMC) toidentify people who did not follow directions (Oppenheimer,Meyvis, and Davidenko 2009).1 All participants passed theIMC and were included in the analysis.Participants imagined that they were in the market for a newautomobile and were asked to share their opinions on a varietyof cars that differed on three main attributes: price, performance,and an ethical labor attribute. Participants were informed thatthe cars they would be evaluating did not differ in any waysother than these three attributes:• Price: The final negotiated cost of the car.• Performance: Performance ratings for the car, from a leadingconsumer magazine. The performance ratings range from 1 to10, with 10 being the highest rating.• Labor practices of the car manufacturer: The manufacturersdiffer in the treatment of their workers. The best measure of thistreatment is the number of lawsuits brought by employeesagainst the management.The explanation of the labor practices of the car manufacturer clearly stated that the number of lawsuits against management had no bearing on the quality of the car, only thetreatment of employees. Participants were then provided witha recommendation from a casual acquaintance on how toproceed with their car search. In the high-consensus information condition, participants were provided with the followingrecommendation:Most people consider the labor practices of car manufacturers whenpurchasing a car since that provides a good indication of how ethicalthe organization is. Like everyone else, you should consider laborpractices as an important factor in your decision.In the low-consensus information condition, participants sawthe following recommendation:Most people don’t consider the labor practices of car manufacturerswhen purchasing a car. However, unlike everyone else, you shouldconsider labor practices as an important factor in your decision sincethat provides a good indication of how ethical the organization is.After participants saw this recommendation, they viewed andrated all possible car combinations that could be formed usingthe aforementioned attributes. Because each attribute had threelevels, participants viewed and evaluated 27 different cars(labeled from car A through car AA). The three levels of theprice attribute were $15,977, $18,385, and $20,793. The threelevels of the performance attribute (on a ten-point scale) were6.0, 7.75, and 9.5. Consistent with Irwin and Naylor (2009), thethree levels of the ethical labor attribute were “fewer thanaverage,” “average,” and “more than average.” These categoriescorresponded with the following descriptions: “one or twolawsuits every few years,” “five to ten lawsuits per year,” and“many complaints, including assault charges.” Finally, aftercompleting a short filler task, participants completed Bearden,Netemeyer, and Teel’s (1989) 12-item measure of SII, whichwas indexed for analysis (a = .92; M = 3.02, SD = 1.17).ResultsTo be certain that our manipulation of consensus informationdid not influence our measure of SII, we first assessed SII withinboth the high- and low-consensus information conditions. Therewere no differences in SII between the high-consensus information condition (M = 2.94) and the low-consensus informationcondition (M = 3.07; F(1, 74) = .24, p > .60). Given this finding,we next discuss our focal analysis.We applied a sequential process to analyze the data, obtainingconjoint weights for each participant and then testing whetherthe weights were dependent on consensus information and SII.Following Irwin and Naylor (2009), we converted negativeslopes to zeros for the second part of our analysis.2 Given thatparticipants were advised to take labor practices into accountwhen evaluating the vehicles, we investigate labor practiceweights as a proxy for persuasion. Greater weights on the laborattribute indicated greater persuasion as a result of the recommendation provided. We conducted a regression analysiswith contrast-coded consensus information, consumer SII (meancentered for analysis), and the SII × consensus informationinteraction as predictors of weights for the labor attribute. Therewas no main effect of whether the recommendation was highconsensus or low consensus (t(1) = _1.68, p > .05). However,there was a main effect of SII (t(1) = 2.20, p .05). Figure 1, Panel A, provides a graphical representation of these results, and Table 1 captures the crossovervalues for all studies, beginning with this set of results. Thisfinding provides support for H1 and illustrates high-SII consumers’ tendency to be persuaded by a persuasive messagebacked by either high- or low-consensus information aboutothers’ behavior.DiscussionStudy 1a demonstrates that consumers’ response to highconsensus versus low-consensus information is dependent onSII. We found that lower-SII consumers were more persuadedby a recommendation backed by high-consensus information relative to one backed by low-consensus information.In contrast, higher-SII participants showed similar levelsof persuasion when a recommendation was accompanied byeither high- or low-consensus information. The results ofStudy 1a therefore suggest that the high-SII segment of consumers may be a particularly attractive segment for policymakers and marketers to target when advocating a new prosocial behavior (e.g., installing low-flow shower heads toconserve water) or attempting to build support for a newprosocial initiative (e.g., charitable or recycling programs).Targeting these people initially can be an effective way toultimately build the majority support needed to persuadelow-SII consumers.While Study 1a provides initial support for H1, these findingsare subject to an alternate explanation: psychological reactanceon the part of low-SII consumers. Psychological reactance occurs when a person feels threatened by a recommendation andis therefore motivated to do the opposite of the recommendation in a bid to regain freedom that has been lost or threatened(Brehm 1966). It is possible that low-SII people are particularlylikely to experience reactance to a low-consensus recommendation, viewing this communication as a strong threat to theirability to make a free choice, especially given that the lowconsensus information suggests a low-quality argument to thosehigh in SII. Therefore, the purpose of Study 1b is to provideadditional evidence for our demonstrated effect and dig deeperinto the process driving the result, including ruling out thisalternative explanation empirically. We also use a different operationalization of high and low consensus. The operationalization of low consensus used in Study 1a requires consumers tomake the inference that while “most people don’t” considerlabor practices, some people do (and, thus, there is low consensus). Given that we expect high-SII consumers to primarilyattend to whether any consensus is present and not to devoteadditional processing resources beyond simply noting whetherit is present, we use a simpler operationalization of low consensus (that does not require an inference) in Study 1b and allsubsequent studies.Figure 1. Consumer SII and Consensus Information forStudies 1a and 1bLabor Attribute Weight0 0.51.01.52.02.53.0+2 SDSIIHigh consensus Low consensus–2 SD 2.96A: Study 1aVegetables on Grocery List00.51.01.52.02.53.03.54.04.5–2 SD +2 SDSIIHigh consensus Low consensus2.60B: Study 1bJournal of Public Policy & Marketing 19Study 1bIn Study 1b, we aim to replicate the effect demonstrated inStudy 1a and provide additional support for H1 using a differentproduct category and consensus cue manipulation. In our firststudy, we used lexical cues (i.e., “most people”). In this study,we use quantitative consensus cues similar to the method usedby Goldstein, Cialdini, and Griskevicius (2008). An additionalgoal of this study is to provide initial evidence that attentionaldifferences are driving this observed effect.MethodIn Study 1b, 241 participants recruited through MTurk completed the study for a small monetary incentive. Study 1b usedan SII × consensus information (high vs. low) between-subjectsdesign, in which SII was a measured continuous variable andconsensus information was manipulated. Congruent with Study1a, participants completed an IMC to detect whether theyfollowed directions (Oppenheimer, Meyvis, and Davidenko2009). Eight participants failed the IMC three times and weredropped from the analysis. Thus, Study 1b had a final sample of233 participants (50% female; Mage = 36.7 years, age range18–83 years).To begin Study 1b, participants completed Bearden,Netemeyer, and Teel’s (1989) 12-item measure of SII (a = .92,M = 3.21, SD = 1.16). Measuring people’s SII before theycompleted the focal task helped ensure that our experimentalmanipulations did not influence this measure. After completing approximately ten minutes of unrelated filler tasks,participants were asked to imagine that they had just arrivedat their local grocery store to purchase some food items forthe week. On entering the grocery store, participants imaginedseeing a PSA produced by the Organization for Healthy Eating,a fictitious public health organization. In the high-consensusinformation condition, participants read the following recommendation from the organization:Almost 76% of American grocery shoppers eat five servings ofvegetables a day. We recommend you join your fellow shoppers andconsume at least five servings of vegetables a day.Conversely, in the low consensus information condition,participants read the following recommendation from theorganization:Almost 26% of American grocery shoppers eat five servings ofvegetables a day. We recommend you join your fellow shoppers andconsume at least five servings of vegetables a day.After reading the PSA, participants were asked to put together a grocery shopping list for their trip. Each participantcompiled a list of 15 items from a total of 60 available groceryitems. Ten of the 60 available grocery items were vegetables(potatoes, broccoli, carrots, tomatoes, celery, corn, cucumbers,lettuce, onions, and peppers).3 We used the number of vegetableitems included on the list as our focal dependent variable.In our conceptualization, we have argued that high-SIIconsumers attend to the fact that consensus information ispresent but place relatively less weight on the degree of consensus provided. If those high in SII indeed place less emphasison consensus information, we would anticipate that they wouldhave a more difficult time recalling the consensus informationpresent in the PSA when compared with those lower in SII.Therefore, after completing the grocery shopping task and several filler items unrelated to the current hypotheses, all participants were asked to recall the consensus information shown inthe PSA by typing in the percentage they saw in the PSA. Wemeasured memory for this information to serve as a proxy forparticipants’ attention given that greater attention should result ina stronger memory trace (Baddeley et al. 1984). Responses werecoded as either correct or incorrect.Finally, given that the results of Study 1a could be explainedby reactance (Brehm 1966) on the part of low-SII people, participants completed an eight-item measure of reactance to therecommendation from the Organization for Healthy Eatingadapted from Hong and Faedda (1996), which was indexed foranalysis (a = .90, M = 2.48, SD = 1.10; sample items: “I resistedthe attempt of the OHE to influence me”; “The recommendationfrom the OHE restricted my freedom of choice”; “I consideredthe recommendation from the OHE to be an intrusion”).ResultsWe first conducted a regression analysis with contrast-codedconsensus information, consumer SII (mean-centered forTable 1. Consumer SII Crossover Values for Floodlight AnalysesSignificant SII CrossoverValues from the Floodlight Analyses Average SII Value for StudyStudy 1a £2.96 3.02Study 1b £2.60 3.21Study 2: No control £3.44 3.28Study 3: Low attentional emphasis £2.18 3.38Study 3: High attentional emphasis ‡5.06 3.38Notes: The crossover values indicate the value of SII at which the floodlight test reached statistical significance for each study. The floodlight test highlights a rangeof valu