The Moderating Effect of Perceived Risk onConsumers' Evaluations of ProductIncongruity: Preference for the NormMARGARET C. CAMPBELLRONALD C. GOODSTEIN*Research supports the existence of a \"moderate incongruity effect\" such that anoption that is moderately inconsisteni with an evoked product category schemais sometimes preierred to a congruent option. We propose that perceived riskis an important siluationai factor that moderates the impact of congruity on eval-uations. Tnree studies show that the positive evaluation of moderately incDn-gruent products, relative to congruent ones, does not appear when there is riskassociated with product selection. When consumers perceive high risk associ-ated with a purchase, the moderate incongfuity effect is reversed such that thecongruent is preferred to the moderately incongruent product. Only in conditionswhere subjects perceived no real risk did ttie positive effect of moderate incon-gruity appear. The limiting effect of perceived risf( appears to be due to con-sumers' \"preferences for the norm\" under high-risk conditions The sal of findingsare discussed as they reiate to and extend current thinking about the effects otmoderate incongruity on evaluations.Evaluations of products arc sometime.s based not on theabsolute levels or values of attributes of ihe productbut, rather, on the di.screpancy between product attributesand consumers\" e.xpectatiotss for the type of product. Theorysuggests that alternatives that are viewed as moderately dis-crepant from expectations are preferred to either congruentor extremely incongruent alternatives (cf. Berlyne 1970;Mandler 1982). A series of consumer studies supports thispositive effect of moderate incongruity, demonstrating thatconsumers may evaluate moderately incongruent productsmore positively than congruent ones (e.g., Meyers-Levy andTybout 1989; Peracchio and Meyers-Levy 1994). This\"moderate incongruity effect\" has been demonstrated forincongruity arising from new product attributes (Meyers-Levy and Tybout 1989), brand extensions (Meyers-Levy.Louie, and Curren 1994), and taste (Stayman, Alden. andSmith 1992).*Margaret C. Campbell i.s ussistant prdfessor of martteting. Leeds Sciioolof Business. University of Colorado at Boulder. 419 I'CB. Boulder. CO8O3(W-()419; e-mail: mej.campbeitC'T'coloradoedu. Ronald C. Coodstein isassociate professor of marketing. Georgetown Lini\\'er.sit>\".'i McDonoughSchool of Business. Washington. DC 2m51: e-mail: goodj(*nisb.edu. Theauthors would like to thank Harold Ka.ssarjian. Alice Tybout. :uid Shi Zhangf'tir sheir commenis. and Mark Kruger nf Design by Mark for as.sislancewith .stimuli design. Ttie authors also thank editors Mick and Burnkruni.two iLssociate editors, and four reviewers for their insights about revisingthis article, and are gralefu! for tinancial assistance provided by the Uni-versity of Californin. Los Angeles.The rationale for the positive evaluation of moderatelyincongnient objects is based on theorizing by Mandler(1982). Mandler suggests that the level of congruity betweena stimulus and an evoked schema influences both processingand evaluation of the stimulus. He proposes that a congruent.stimulus is not arou.sing and resnlts in a mild positive eval-uation based on familiarity. However, moderate incongruityis believed to heighten processing, leading to the resolutionof the iticongruity and enjoyment of the product novelty. Inother words, consumers enjoy the stimulation of figijringout the no\\'el product and thus evaluate the novel productmore positix'ely than one that is typical—that is. the \"mod-erate incongruity effect.\"While this theory has been extended to predict consumers\"product evaluations under some conditions, a more completeframework would detail variables that moderate the rela-tionship between incongruity and evaluations. Identifyingboundary conditions that limit when the moderate incon-gruity effect occurs will provide a richer understanding ofthe phenomenon as it relates to consumer behavior. Recentresearch is consistent with this goal; for example, Meyers-Levy and Tybout (1989) demonstrate that the moderate in-congruity effect only occurs for nondogmatics. Subjects whodemonstrated the indi\\'idual trait of dogmatism did not showa higher evaluation for moderately incongruent products rel-ative to congruent ones. Additionally, Peracchio and Tybout(1996) show that prior product knowledge moderates the439' 2(MH by JOURNAL OF CONSUMEH RESEARCH, ini: • Vol. 2« • l^ccembeT 2(101440preference for moderately incongruent options; the moderateincongruity effect did not occur for people with high productknowledge. Thus, these studies identify two individual dif-ference variables that moderate the occurrence of the mod-erate incongruity effect.The purpose of the research presented here is to proposethat perceived risk is a very important situational variablethat serves as a boundary condition for the positive influenceof moderate incongruity on product evaluations. Productsand situations vary in tenns of the perceived risk consumersassociate with them; we propose that the moderate incon-gruity effect will not arise when a consumer perceives highrisk. Instead, when perceived risk is relatively high, con-sumers will have a preference for the norm over the novel,preferring an option that is congruent with an evoked prod-uct schema to one that is moderately inconsistent.Three experiments that manipulate perceived risk withsociai risk scenarios provide support for the important at-tenuating effect of risk on evaluations of moderately incon-gruent products. The first study demonstrates that risk mod-erates the relationship between congruity and evaluations.When perceived risk is high, evaluations of a congruentproduct are significantly higher than those associated witha moderately incongruent product, whereas no such pref-erence for the congruent product appears when risk is low.Because this initial study does not find supporting evidenceof the moderate incongniity effect even when risk is low,two additional studies investigate the conditions necessaryfor the effect to hold. These studies replicate the moderateincongruity effect found in earlier research, but only whenparticipants perceive there to be no real risk associated withthe product. These studies provide strong support for theimportant role of risk and begin investigation into the pro-cess underlying the effect of risk on evaluations of mod-erately incongruent products.PERCEIVED RISK AND CONSUMEREVALUATIONSRisk is an important construct in the social sciences witha rich and varied history of research. Here, we consider aportion of the research on risk that is directly relevant toconsumers' evaluations of products. Consumers' perceptionsof risk are considered to be central to their evaluations,choices, and behaviors (e.g., Dowling 1999). Consumer re-searchers define perceived risk in terms of uncertainty andconsequences; perceived risk increases with higher levels ofuncertainty and/or the chance of greater associated negativecon5»equences (Ogiethorpe and Monroe 1987). For example,if a consumer is considering buying an unfamiliar bottle ofwine for a dinner party, the perceived risk associated withthe purchase could arise because she does not know how thewine will taste (uncertainty) and is worried that her guestswill think poorly of her if it is not a good wine (consequences).Several sources of risk have been identified and measuredin consumer research, including financial, performance, so-JOURNAL OF CONSUMER RESEARCHcial, psychological, and physical (e.g., Kaplan, Szybillo, andJacoby 1974). Oftentimes one or more of these sources maydrive consumers' overall perceptions of risk; risk may bederived from just one or from a combination. In the es.ampleof the woman buying wine for her party, the source of therisk due to negative consequences could be performance (thewine will not taste good), psychological (her self-concept isthreatened), and/or social (her guests will think poorly of her).These sources of Hsk combine to form an overall level ofperceived risk that the consumer associates with the purchaseof the wine.Perceived risk produces wariness or risk aversion andoften leads to a variety of risk-handling activities (Bettman1973; Dowling 1999). More important for the re.search pre-sented here, perceived risk has been shown to have an impacton a variety of consumer behaviors, such as the use of word-of-mouth information, new product adoption, brand loyalty,and reliance on well-known brands (e.g., Erdem 199S).Montgomery and Wernerfelt (1992) theorize that umbrella-branding strategies are most effective for risk-averse con-sumers becau.se the strategy allows them to stick with knownbrands rather than to explore new brands. Erdem (1998)shows that when consumers perceive the risk of purchasinga new product to be high, they are more likely to choose aknown brand than a new brand. Overall, this stream of re-search suggests that higher risk inhibits exploratory ten-dencies and leads consumers to prefer familiar options tounfamiliar ones.Thus, when perceived risk is high, consumers becomemore wary and risk averse. Additionally, we propose thatunder high risk, consumers will prefer \"familiar\" options tomoderately incongruent alternatives. Schema congruent op-tions seem more familiar to consumers because of the highdegree of match between the product and the consumer'sproduct schema, whereas moderately incongruent optionsare novel and somewhat unfamiliar. We hypothesize thathigh perceived risk leads consumers to become more con-servative, as evidenced by preferring the norm to the novel.In contra.st, when the perception of risk is low, consumerswill enjoy the positive stimulation provided by novel prod-ucts and will evaluate them more positively. Thus, perceivedrisk moderates the effect of incongruity on evaluations, andpreferences for moderate incongruity will not appear whenrisk is high.Reexamination of existing research that supports the mod-erate incongruity effect reveals that the preference for mod-erate incongruity has been shown under very low-risk con-ditions. For example, the studies reviewed above (i.e.,Meyers-Levy and Tybout 1989; Peracchio and Tybout 1996;Stayman et al. 1992) were conducted under the guise ofmarket research. The cover stories involved simple, context-free evaluations of proposed new products, as opposed topersonal or social scenarios. Evaluating proposed new prod-ucts for market research probably presents very little riskto subjects because they are not considering the product foractual purchase or use.Consumers, however, seldom evaluate products withoutMODERATING EFFECT OF PERCEIVED RISK441in a 2 (low/high risk) x 2 (congruent/moderately incongru-ent) between-subjects design. Participants were 67 managersin a fully employed MBA program at a western university.A drawing for cash prizes motivated voluntary participation.Each participant was given a research bookJet that pro-vided random assignment to one of the four experimentalconditions. Participants imagined themselves in one of twowine purchase scenarios. In the low-risk scenario, partici-pants imagined themselves in a store purchasing a bottle ofwine to have around the house. In the high-risk scenario,participants imagined themselves purchasing a bottle ofwine to take to dinner at the home of a potential employer.considering both the product attributes and the usage situ-ation related to their purchasing goals {Dowling 1986;Dowling and Staelin 1994). Evaluations of a product foruse in a specific, high-risk situation are likely to differ fromevaluations made in a context-free environment (e.g., Payne,Bettman, and Johnson 1993). For example, the goal ofchoosing a good wine for a socially risky occasion is likelyto lead to a preference for a product that matches expec-tations rather than for a novel product. Even if the consumerfound the moderately incongruent alternative interesting andenjoyable in and of itself, she would likely evaluate it un-favorably for the high-risk occasion because, any enjoymentresulting from the incongruity would be overwhelmed bythe desire to make the \"right\" choice for the purchase oc-casion. When risk is high, the more positively evaluatedproduct is likely to be one that looks \"normal,\" that is, onethat matches the product category schema, and the stimu-lation provided by moderate incongruity is not likely to bepositively valued. Thus, we predict that perceived risk mod-erates the relationship between congruity and evaluationssuch that under high-risk conditions, consumers are likelyto prefer an alternative that is consistent with schema ex-pectations to one that is moderately incongruent.RESEARCH OVERVIEWThree separate experiments inve.stigate the hypothesis thatperceived risk moderates the relationship between incongruityand evaluations and begin to explore the process by whichthis occurs. In each laboratory study, subjects were asked toimagine themselves in a particular purchase situation. Thepurchase situations manipulated perceived risk by affectingthe degree to which subjects were likely to perceive socialrisk. They were then provided with a description and pictureof a proposed new product; the degree of incongruity wasmanipulated by the product packaging. We used packagingto manipulate congruity because the way a product looks inits package is an important compotient of the structured prod-uct category knowledge that consumers have stored in mem-ory (e.g., Bloch 1995; Meyers-Levy and Tybout 1989). Pack-aging is a way for managers to alter consumers' productperceptions without substantively changing the product itselfand is increasingly important to consumer decision makinga,s more decisions are made at the point of purchase (Bloch1995; Wansink 1996).STUDY 1Design and ProcedureStudy 1 wa.*; designed to test the basic hypothesis thatperceived risk moderates the effect of congruity on evalu-ations. In line with the particular interest in whether per-ceived risk affects the moderate incongruity effect, we ex-amined two levels of congruity (congruent and moderatelyincongruent) and two levels of social risk (low and high)A pretest indicated that these scenarios were perceived tohave significantly different levels of risk (on a four-item,nine-point scale; low - 4.56, high = 6.79; f(l,70) =20.44, p<.O(X)l).Next, participants read a description and saw a picture ofa bottle of white wine. The level of congruity was manip-ulated by the shape of the bottle. A pretest was conductedin which 57 MBA students viewed four different wine bot-tles that differed on shape and color, two important dimen-sions of package design (Bloch 1995). The pretest subjectsindicated on a two-item, seven-point typicality scale (Pear-son correlation = .72) that the typical white wine bottle (alight green, cylindrical bottle) was congruent (M = 1.77),and a bottle that varied on one dimension (a light green,triangular bottle) was moderately incongruent (M = 4.00).The moderately incongruent bottle was rated as significantlyless typical than the congruent (F(l,56) = 20.72, p<.0001) as well as significantly more congruent than an ex-tremely incongruent bottle that varied on two dimensions(a black, triangular bottle; M - 6.21; F(l,56) - 11.65,p < .(K)l). Thus, these two bottles were used in the mainstudy. (The pretest subjects each rated four bottles, and al-though the bottles were counterbalanced for order, this pro-cedure may have affected perceived typicality. Thus, in themain study, subjects viewed only a single package.)After seeing one of the two wine bottles described above,the managers completed a variety of measures at their ownpace. First, evaluations of the new product were measuredon four seven-point items anchored by very unfavorable/veryfavorable, bad^good, unappealing/appealing, and undesirable/desirable, with higher numbers indicating a more positiveevaluation (a = .95). Subjects then rated the perceived typ-icality of the product on a nine-point version of the two-itemscale used in the pretest (Pearson correlation = .71). Theythen indicated the perceived risk of their scenarios on fournine-point items anchored by the choices not at all risky/extremely risky, not at all concerned/highly concemed, veryunimportant/very important, and not at all worried/very wor-ried (a = .86). Finally, subjects answered covariate itemsabout gender, age, and wine purchase and usage.ResultsManipulation Checks. We conducted a 2x2ANCOVA (wine purchase and use were covariates) to verify442the success of the congruity and risk manipulations. The anal-ysis of perceived typicality showed a significant main effectof bottle (F(1.55)'= 60.70, /j<.0001). As predicted, thegreen, cylindrical bottle wa.s perceived as typical (M =4.05), and the green, triangular bottle wa,s perceived as mod-erately atypical [M — 7.40). This was the only effect thatreached significance. There was a directional main effect ofrisk scenario on perceived typicality {/^(1,55) = 3.22, p<.08) such that low risk led to slightly higher perceptions ofincongruity than high risk (low = 5.93, high = 5.37). Over-all, the analysis indicates that the congruent and moderatelyincongruent packages affected perceived typicality as desired.The perceived risk analysis revealed only a significant maineffect of the risk scenario (f (1, 55) = 28.45, p < .0001). Thelow-risk scenario resulted in lower perceived risk (low =4.48) than did the high-risk scenario (high = 6.57). indicat-ing fhat perceived risk was successfully manipulated. Meansfor this study appear in table 1.Evaluations. The same 2x2 ANCOVA on evaluations showed a main effect of risk (f(l.55) = 5.01. p<.05) and a main effect of congruity (F(l,55) = 19.84./3<.(X)01>, qualified by a significant interaction effect(F{\\.55) - 4.59, /7<.O5). As predicted, evaluations ofmoderate incongruity differed in the low-risk and high-riskscenarios. When social risk was high, the congruent productwas evaluated significantly more positively than was themoderately incongruent product (C = 4.66, MI = 2.97;F(l, 55) = 23.37, p < .0001). There was no significant dif-ference between evaluations of the congruent and moder-ately incongruent products when risk was low (C = 4.96,Ml = 3.95; f(I, 55) = 2.05-p> .15). This pattern supportsthe proposed attenuating effect of perceived risk on the re-lationship between evaluations and congruity but does notsupport the moderate incongruity effect, even in the low-risk condition. There was no significant difference betweenevaluations of the congruent wine across risk conditions(f < I, NS), but evaluation of the moderately incongruentwine was significantly lower in the high-risk than in thelow-nsk situation (f(l, 55) = 9.20, p < .01).DiscussionThis experiment supports the predicted moderating effectof perceived risk on the relationship between congruity andevaluations. As hypothesized, consumers appear to prefer acongruent versus moderately incongruent product when so-cial risk is high. This preference does not appear when riskis low. While the results support the moderating role of riskon evaluations of moderately incongruent products, thestudy did not find the moderate incongruity effect under lowrisk. To understand why, we returned to earlier re.search andnoted two ways in which our study differed. First, researchsupporting the moderate incongruity effect tends to useproduct categories such as soft drinks that are likely to havelow inherent risk (e.g., Meyers-Levy and Tybout 1989; Stay-man et al. 1992). It may be that the risk associated with thewine product category is enough to eliminate the moderateJOURNAL OF CONSUMER RESEARCHincongruity effect. In order to gain insight into this possi-bility, a focus group was conducted with eight self ratedwine \"experts.\" The 60-minute guided discussion revealedthat these consumers felt thai wine and wine purchases areinherently ri.sky. Even when directly asked, the only low-risk scenario they were able to identify was purchasing winefor cooking. This suggests that the wine product categorymay have obliterated the moderate incongruity effect be-cause of the risk associated with the category itself (cf.Mandler 1982). The 4.48 rating of perceived risk (on a nine-point scale) in the low-risk condition of the study is in linewith this conjecture.The second difference between our study and earlier workwas that the latter tended to be conducted under the guiseof market research. In these earlier studies, participantsnever evaluated products for potential purchase; thus, riskassociated with purchase was nonexistent. It is possible thaieven the mild level of risk associated with considering buy-ing a product to have at home is enough to attenuate themoderate incongruity effect.To examine these possibilities, we conducted a secondstudy with a lower-nsk product category and included threelevels of risk (no, low, and high). The primary goals forstudy 2 were to replicate the results of earlier research andthe moderating effect of perceived risk found in study 1.STUDY 2Design and ProcedureStudy 2 followed the same procedure as study 1, with twomajor changes. First, the product category studied was sofidrinki. Earlier research that finds the moderate incongruit>effect used this product, and a pretest (/; = 23) revealed thaisoft drinks were associated with low product category risk(M = 1.87, scale from 1 to 7). significantly lower than wine{M = 3.53; /(22) = 5.58, p<.0001). Second, a \"no-risk\"scenario, in which subjects were asked to evaluate a productfor marketing research, was included. Study 2 employed a3 (no/low/high risk) x 2 (congruent/moderately incongru-ent) between-subjects factorial design. Subjects were 171 un-dergraduate students from a major midwestern university. Re-search booklets were randomly distributed, such that eachsubject saw one experimental condition.Subjects were first asked to imagine themselves in one ofthree social risk scenarios. Subjects either evaluated the prod-uct for market research (no social risk), to purchase to haveat home (low risk), or to \"take to a picnic where you wantto make a good impression on your significant other's closefriends\" (high dsk). They then saw the same description ofa proposed new soft drink found in Meyers-Levy and Tybout(1989, pp. 44-45), with the exception that the la.st line of thedescription vaded the packaging of the soft drink. The con-gruent condition .stated that the new beverage would be avail-able \"in cans and bottles\" and displayed a graphic artist'ssketch of an aluminum can. The moderately incongruent con-dition stated that the beverage would be available \"in cansand sports bottles\" and was followed by a sketch of a sportsMODERATING EFFECT OF PERCEIVED RISKTABLE 1STUDY MEANS AND STANDARD DEVIATIONSLow risk (M =: 4.48)CongruenlStudy 1 {n = 67):TypicalityPerceived riskEvaluationModerately incongruent767(1,13)5.20(1.84)3.95(1.02)Norisk (M - 2.51)CongruentStudy 2{n= Typicality171):3.19(1.27)2.24(.93)4.39(1.11)5.08(1.94).11(.31).04(.19)CongruentStudy 3 (n = 147):TypicalityPerceived riskEvaluationPurchase intentPreference for normNOTE.—M - means443Higli risk(M = 6.57)Congruent3.53(t.45)6.53(1.53)4.66(1.08)Lowrisk (M = 4.68)Moderately InoongruentCongruent605(1.26)4.57(1.76)4.60(1.08)5.22(2.03).12(.28).02(.09)Congruenl3.77(1.45)5.19(1.59)3.64(1.57)4.26(2.38).36(1.73)3.66(1.15)6.02(1.24)3.91(1.26)3.58(2.19).00(.00).54(.46)Moderately incongruent4.54(2.09)3.91(1.46)4.96(.75)7.31(1.26)6.61(1.14)2.97(1.15)High risk (M = 6.10)Moderately incongruent5.44(1.21)6.19(1.37)3.35(1.09)3.37(1.94).09(.29).67(.38)Moderately incongruent5.30(1.29)5.48(1.78)2.97(.97)2.51(1.68)1.37(1.49)Moderately incongruent Congruent5.44(1.25)2.78(1.05)4.96(.92)6.07(1.94).21(38).00(.00)Nio nsk {M =3,10)Moderately incongruent4.39(1.42)3.1B(1.37)4.72(.93)5.24(1.62)-1.36(1.26)3.54(1.18)4.81(1.52)4.55(1.14)5.36(1.91)06(.21).05(.20)Perceived riskEvaluationPurchase intent% unresoived% preference tor normHigh risk(M = 5.34)256(1.36)3.02(1.51)4.20(.94)5.32(2.14)-1.47(1.04)bottle (i.e.. a plastic single-serve bottle with a drinking vaiveon the lid). A pretest (n = 80) indicated that purchasing softdrinks to have at home is significantly less risky {M =3.89) than to take to the picnic {M = 5..^9: F(i.78) =10.37. p < .002) and that the cans and bottles condition wasperceived as more typical (A/ = 2.9) than the cans and sportsbottles condition {M = 3.01; F(l,78) = 13.80, p < .001).Subjects then completed the four seven-point items as-.sessing product evaluation (a — .92). followed by a single-item, nine-point scale of purchase intent. Subjects next listedany thoughts or feelings that they had about the described.scenario and soft drink. After providing their thought.\";, sub-jects were asked to complete the four-item, nine-point scaleof perceived risk (a = .91). Congruity of the soft drinkswas assessed using three nine-point scale items (a = .57)asking .subjects how novel the soft drink was (not at al!novel/extremely novel), how likely it was that a soft drinkcame in this package (very unlikely/very likely), and howwell this soft drink matched their expectations for soft drinksin general (not at all well/very weiS).ResultsManipuiaxion Checks, Before testing hypothesized ef-fects, we verified that subjects perceived the manipulationsof risk and congruity as intended. A 3 x 2 ANOVA revealeda significant main effect of the risk scenarios on subjects'perceptions of risk {F(2. f 65) = 97.89. p < .0001). with noother significant effects. The market research scenario wasperceived as significantly less risky than the to-have-at-home scenario (on a 1-9 scale; no = 2.51. low = 4.68;F(], 165) = 72.86. /3<.OOO1). and the low-risk, to-have-at'home scenario was viewed as significantly less risky thanthe picnic scenario (high = 6.10; F(\\. 165) = 3J.95. p<444.(XX) 1). Analysis of perceived typicality indicated a signif-icant main effect of the packaging manipulation on typicalityratings (F(l, 165) = 135.94. p < .0001), with no other sig-nificant effects. The can condition was viewed as being moretypical (on a nine-point scale: C = 3.48). and the sportsbottle was rated as moderately incongruent (MI — 5.67).Thus, our manipulations were successful. (Means appear intable I.)Evaluations. The ANOVA revealed a significant maineffect of social risk on evaluations (F(2, 165) = 15.28, p <.0001) such that all products were evaluated less favorablyas risk increased. Planned contrasts showed no differencesbetween the no- and low-risk conditions (congruent: no —4.39, low = 4.55. F(l, 165) = 1.65, NS; moderate; no =4.96, low = 4.60, F <\\. NS). However, both the congruentand moderately incongruent soft drinks were evaluated lessfavorably under high- than under low-risk conditions (con-gruent; low = 4.55. high = 3.91, F(l, 165) = 5.05, p<.03; moderate; low =: 4.60. high = 3.35. f(l, 165) =19.00, p < .0001). This main effect was qualified by a sig-nificant risk by congruity interaction (F(2, 165) = 3.53,p < .03), Planned contrasts demonstrated that under the no-risk scenario, the moderately incongruent package was pre-ferred to the congruent one (C = 4.39, Ml = 4.96;F{1, 165) = 3.48, p < .06); this marginally significant resultsupports the moderate incongruity effect. When there waslow risk, there was no significant difference in the evaluationof the two drinks (C = 4.55. M! = 4.60; F < 1, NS). Whenthe risk was high, there was a marginal effect, such that themoderately incongruent package received lower evaluationsthan did the congruent one (C = 3.91, Ml = 3.36;F(l, 165) = 3.55, p < .06). Thus, the predicted moderatingeffect of perceived risk on evaluations of moderate incon-gruity was supported.Additional Analysis. The same ANOVA model wasused to analyze the effects of the independent variables onthe measure of purchase intentions. The analysis of purchaseintentions revealed only a main effect of risk {F(2, 165) =18.41, /> < .0001). Follow-up simple effects tests (cf. Iacob-ucci 2001; Keppel 1991) revealed that under no risk the mod-erately incongruent option was directionally more likely tobe purchased than was the congruent product (C = 5.08,MI = 6.07; F(l, 165) = 3.30, /> < .07). There were no sig-nificant differences in purchase intentions under low or highrisk (F's < 1).Subjects' thoughts were examined to explore the processunderlying the effect of risk on the congruity-evaluationrelationship. Thoughts were coded by two independentjudges (agreement = 93%, and disagreements were re-solved by discussion) as to whether they (1) indicated suc-cessful or unresolved categorization of the new product (e.g.,\"the drink was described a bit eccentric; don't quite get it\")and (2) suggested a preference for the norm or reluctancefor the novel (e.g., \"no matter how good the taste, there isvery little chance I would bring a new, unfamiliar drink toany party\"). If the explanation for the effect of perceivedJOURNAL OF CONSUMER RESEARCHrisk on the moderate incongruity effect is that subjects wereunable to successfully resolve the incongruity, then wewould expect to see a greater percentage of unresolvedthoughts for the moderately incongnient package, relativeto the congruent package, under high risk. Although therewere marginally more unresolved thoughts when the pack-age was moderately incongruent (C = .056, MI = .139;F{\\, 153) - 3.69, p< .06), there were no significant dif-ferences due to risk or its interaction with congruity (allF 's < 1). Further, within the high-risk condition, a plannedcontrast revealed that subjects were equally able to cate-gorize the congruent and the moderately incongruent prod-uct (F(l, 153) = 1.19, NS). Thus, this explanation will notbe explored further.If, as proposed, risk induces a preference for the norm,we would expect higher risk to lead to greater percentagesof thoughts about preference for the norm or reluctance forthe unusual. Analysis of the percentage of thoughts aboutpreference for the norm versus the unusual revealed only asignificant main effect for risk (F(2, 153) = 81.89, p<.0001). There was no difference between the no- and low-risk conditions (no == .03, low = .04; F < I), but subjectshad a significantly higher preference for the norm underhigh versus low risk (high = .57; F(l, 153) = 130.43.p < .0001). Simple effects tests revealed that under high Hsk.subjects indicated directionally more thoughts about pre-ferring the norm versus the unusual in reaction to the mod-erately incongruent package versus the congruent one(C = .54, MI = .67; F(l, 153) = 2.87, p < .Oh}. Further,we found that the moderately incongruent package evokedsignificantly greater numbers of thoughts about preferencefor the norm under high versus low risk (low = .02,high = .67; F(i, 153) = 80.66, p<.()001). Although thecognitive responses did not show the same omnibus inter-action effect as the evaluations, these results suggest thathigh risk may lead to preference for the norm, which maythen influence evaluations of moderate incongruity. The re-sults from this study indicate that this potential underlyingprocess warrants further investigation.DiscussionStudy 2 replicates and extends the results from study 1Evaluations of moderately incongruent products were sig-nificantly less positive than evaluations of congruent prod-uct.s under high-risk conditions. Importantly, evaluations ofthe moderately incongruent product were found to be mar-ginally higher than those of the congruent product underno-risk conditions similar to those used in earlier research.Being exposed to even the low risk associated with buyinga product for consumption at home eliminated the moderateincongruity effect This suggests that the positive evaluationof moderate incongruity may be limited to relatively risklessevaluations of products without consideration of purchase.This study also provides some initial evidence that perceivedrisk may affect purchase intentions toward moderately in-congruent products.Further, study 2 begins to explore the possible processesMODERATING EFFECT OF PERCEIVED RISK445underlying the relationship between perceived ri.sk and eval-uations of moderately incongruent products. The open-ended thought data showed that the ability to resolve mod-erate incongruity did not depend upon risk. Instead, subjects'thoughts indicated that they had higher preference for thenorm under high risk, particularly when considering a mod-erately incongruent option. While the overall interaction wasnot significant, this may be because subjects may have onlyspontaneously reported thoughts about preferring the normnr avoiding the unusual in the high-risk condition. That is,subjects may not have actively thought about the fact thatthey had a low preference for the norm or did not mind theto classify or categorize the product on a seven-point scaleanchored by strongly disagree/strongly agree.The next set of measures asses.sed the processes that mightunderlie the predicted effects. To test preferences for thenorm, participants indicated their degree of agreement on aseven-point scale (-3/+3) with the following two state-ments: (1) \"I would prefer something 'normal' (for the sit-uation)\" and (2) \"In this situation, it would be really inter-esting to explore a new product.\" The .second scale wasreverse scored, and the average of these two highly corre-lated items (Pearson correlation = .6H,p < .0001) providedthe measure of preference for the norm. Next, subjects com-unusual when there was low risk associated with the product.They may have only actively thought about their preferencefor the norm when there wa.s high risk. A close-ended scalemeasure may reveal low as well as high preference for thenorm.While the results of study 2 provide encouraging supportfor the proposed moderating role of perceived risk arisingfrom a risk-induced preference for the norm, there are twoconcerns that could be effectively addres.sed in an additionalstudy. First, although the risk X incongruity interaction effecton evaluations is highly significant, tbe moderate incongru-ity effect under low risk and the preference for congruityuiider high risk are both significant at thep < .06 level, ratherthan the traditional p < .05 level. Second, more data pro-viding insight to the proposed process would be beneficial.STUDY 3Procedure and DesignStudy 3 was designed to replicate the moderate incon-gruity effect under no risk and the reversed preference forcongruity under high risk at the traditional p < .05 level oJsignificance. Thus, this study includes only the no-risk andhigh-risk conditions in a 2 (risk: no vs. high) x 2 (congruity:congruent vs. moderately incongruent) between-subjects de-sign. The study also investigated the process underlying iheinteraction with close-ended process measures. Subjectswere 147 MBA students from an eastern university whowere randomly assigned to treatments; participation was mo-tivated by entry to win a cash prize.The study followed the same basic design as the two studiesdescribed earlier. Participants were asked to evaluate a newsoft drink as part of a market research study (no risk) or totake to an invited dinner at the home of the business schooldean (high risk). The manipulation of congruity was the sameas in study 2; a new soft drink was described and picturedas either available in cans and bottles (congruent) or cans andsports bottles (moderately incongruent). Subjects then com-pleted the product evaluations (a = .93), intent to purchase.perceived risk (a = .86). and four nine-point typicality scalesanchored by very typicaVextremely atypical, not ai all novel/extremely novel, not at all unusual/very unusual, and(matches) very well/not at all well (o = .86). Participantsalso answered' a question assessing whether it was difficultpleted the change seeking index (CSI) short form (Steen-kamp and Baumgartner 1995; a = .82). The CSI measuresindividuals' desire or need for variation and has been usedas a measure oi\" the individual difference variable of opti-mum stimulation level; this could be related to the proposedrisk induced preference for the norm. (As no significantresults were found using this variable, it will not be dis-cussed further.) Finally, .subjects indicated their gender, age,soft drink purchase, and usage inforaiation.ResultsManipulation Checks. A 2 x 2 ANOVA revealed amain effect of the congruity manipulation on perceived con-gruity such that the can was seen as significantly more con-gruent with expectations than was the plastic sports bottle (ona mne-point scale; Can = 3.16. SB = 4.85; F(l, 141) =53.54. p < .0001). There was also a main effect of risk suchthat participants perceived significantly more risk in pur-chasing a soft drink to take to the dean's dinner than inevaluating a soft drink for market research (nine-point scale:high = 534. no = 3.10; F(l, 142) = 73.90, p<.0001).The risk condition also influenced the perceived typicality ofthe package such that the product package (regardless ofwhether it was the can or the sports bottle) was perceived assianificantly more atypical when the product purchase wasrisky (no = 3.47. high = 4.56; (F(l, 141) = 21.53. p<.0001). The interaction between risk and package congruitywas not significant (F < 1. NS). Overall, the manipulationchecks indicate that the experimental manipulations of riskand congruity were successful, although risk also affectedrated typicality (see tabte 1).Evaluations. A 2 x 2 ANOVA revealed a significanteffect of risk on evaluations of the new product(F{1 143) = 38.05, p < .0001). as well as a significant in-teraction effect (F(l, 143) = 10.17.p< .002). Plannedcom^-parisons showed that both the congruent and moderately in-congruent options were evaluated more negatively when riskwas high than when there was no risk (congruent; no =4 7 high = 3.64. F(1.143) = 4.41, /? < .05; moderate:no'= 4.72. high = 2.97. F(K 143) = 42.71. /)< .0001).Importantly, planned comparisons showed that when therewas no risk, the moderately incongruent product was eval-uated significantly more positively than was the congruentproduct (C = 4.20, Ml = 4.72; F(l, 143) = 3.90. p< .05).446When risk was high, however, the moderately incongruentproduct was evaluated significantly less positively than thecongruent product (C = 3.64, Ml = 2.97; F(l, 143) =6.43. p < ,0\\), There were no other significant effects.Analysis of purchase intentions revealed main effects ofcongruity (F(l, 142) = 7.80, /:-< .006), nsk (F(l, 142) =33.50, p<.0001) and also a significant interaction(F( 1, 142) = 6.50, p < .01). There was a main effect suchthat purchase intentions were higher under no risk than underhigh risk (Research = 5.28. Dinner = 3.36). Further, underhigh risk, purchase intentions were higher for the congruentthan the moderately incongnient product (C = 4.26. MI —2.51; F(l. 142) =16.1,/?< .0001), though this effect dis-appeared under no risk (C = 5.32. Ml - 5.24; F< 1. NS).Process. Analy.sis of subjects' preferences for the normshowed a similar pattern of results as the evaluations measure.There was a significant main effect of risk on the preferencefor the norm measure (F(l, 141) = 96.21. p < .0001) suchthat preference for the norm was lower under no risk thanwhen risk was high (no = -1.42, high = .86). There wasaLso a significant effect of congruity (F(l, 141) = 5.76./7<.O2), qualified by an interaction effect (F( 1,142) =3.75, p < .05). As expected, the preference for the norm wasthe strongest when risk was high and the subject saw a mod-erately incongruent product. It is important to note that highrisk increased preference for the norm acro.ss both levels ofcongruity. and this preference was heightened by the mod-erately incongruent option; the contrast between preferencefor the norm in the low-risk, congruent condition and thehigh-risk, congruent condition is significant (no = -1.47.high = .36;F(1, 141) - 31.21,/? < .0001), as is the contra.stbetween the moderately incongruent option under no versushigh risk (no = -1.36, high = 1.37; F(l, 141) = 68.48,;?<.CIOOI). The preference for the norm was not differentbetween the congruent and moderately incongruent packagesundernori.sk(C - -1.47,A// = -1.36;F{1, 141)< l,NSjbut was different under high risk (C = .36. MI = 1.37;F(l,14i) = 9.21, ^<.003).Because the pattern of significant effects on the proposedmediator, preference for the norm, was .similar to that foundon the evaluation dependent variable, we followed the restof the three-step mediation testing procedure outlined inBaron and Kenny (1986). We next conducted an analysisof the experimental treatment variables on evaluations, in-cluding the preference for the norm measure as a covariate.Not only was preference for the norm a significant covanate(F(l, 140) = 15.73, p < .0001), but including it in the anal-ysis weakened the effect of risk (F(l,140) = 5.96. p<.02) and, importantly, the interaction (F(l,140) = 7.12,pK.Ol) on evaluations. According to Baron and Kenny(1986), the change in the F-value from 10.17 in the analysiswithout preference for the norm to 7.12 in the analysis in-cluding preference for the norm suggests that preference forthe norm partially mediates the effects of risk and congruityon product evaluations.We also examined the possibility that the mediation wasreversed; that is. we examined whether preference for theJOURNAL OF CONSUMER RESEARCHnorm influenced perceived typicality and perceived typi-cality influenced evaluations. Analysis revealed that, whenboth preference for the norm and perceived typicality wereincluded as predictors of evaluations, only the preferencefor the norm achieved significance (F(l, 142) = 40.16,p< .0001). Perceived typicality was not a significant vari-able in the presence of preference for the norm (F < I, NS).Together, these analyse.s support preference for the norm asa partial mediator of the effects of risk on the relationshipbetween congruity and evaluations.Alternatiie Analyxi,s, We also explored the possibilitythat an inability to resolve the moderate incongruity underhigh risk explains the moderating role of risk. In study 2.analysis of thought protocols suggested that there were norisk-induced differences in ability to resolve the incongruity.Study 3 included a direct, self-report measure of categori-zation difficulty. Analysis of this measure found a marginaleffect of package; it was somewhat easier for subjects tocategorize the new product when it was packaged in a canthan when it was packaged in a sports bottle (C ~ —.31.MI = .26;F(I. 143) = 2.99, p < .08). These results supportthe manipulation of moderate incongruity, but the mild na-ture of the effect also suggests that subjects fell that theywere able to successfully categorize both products. Impor-tantly, this was the only effect found. Neither risk nor therisk by congruity interaction had any effect on categorizationdifficulty. The lack of a simitar pattern of results as foundon the dependent variable (evaluations) suggests that cate-gorization ability does not mediate the effect of risk andcongruity on evaluations, especially when combined withthe analysis reported above.DiscussionStudy 3 replicates the results of the earlier studies andprovides initial support for the proposed process. When thereis no risk associated with the evaluated product, a moder-ately incongruent option is evaluated more positively thana congruent option. However, when risk is high, the mod-erately incongruenl option is evaluated less positively thanthe congruent option. Overall, analyses support the idea thairelatively high levels of risk lead consumers to have pref-erences for the norm, such that moderately incongruent op-tions are evaluated less highly than congruent options. Riskleads consumers to prefer what looks \"like it should,\" thatis, what matches category expectations.Further AnalysisTo provide more insight to the research presented here.we conducted analyses to obtain overal! levels of signifi-cance for the effects across all three studies. Following therecommendation in Rosenthal (1978) for combining prob-abilities across fewer than five studies, we used the methodof adding probabilities (from Edgington 1972) as well asthe method of adding Z's (see also Hunter and Schmidt1990). Both methods showed an overal! significant mainMODERATING EFFECT OF PERCEIVED RISK447effect of risk at /J<.0001, no significant main effect ofcongruity (p > .38), and a highly significant overall inter-action (p < .001). We also used these methods to derive theoverall probabilities for the within-risk contrasts of the con-gruent versus the moderately incongruent products. Acrossthe two studies with no risk conditions, the moderately in-congruent product was preferred to the congruent product(p < .006) when there was no risk; that is, the moderateincongruity effect appears across the two studies when thereIS no risk. There was no significant difference across thetwo studies between evaluations of the congruenl and mod-erately incongruent products within low dsk. Across themay provide a strong explanation for the effects of uncer-tainty on variety seeking. In fact, if high risk is seen as anextrinsic motivation for the norm, then we offer additionalsupport for the finding that extrinsic motivations often limitthe desire for variety (cf. Zuckerman 1991). An interestingexten.sion of this conjecture would be to test how differentrisk scenarios (e.g., purchase occasions) affect consumers'preferences for variety seeking and switching brands versusremaining loyal to a 4tnown brand.Our findings also hold interesting implications for schemaresearchers. Schemata are conceptualized to be relativelystable patterns of attributes and reactions to stimuli thatthree studies with high-risk conditions, the moderately in-tongruent product was evaluated significantly less positivelyIhan the congruent option when risk was high (/' < .0001);that is. across the three studies, high risk leads to a reversalof the moderate incongruity effect. Thus, we see that thesethree studies provide strong support for the moderating roleof risk in evaluations of product incongruity.GENERAL DISCUSSIONSutnmary of Findings and ImplicationsThe three studies presented here provide .strong evidencethat evaluations of moderately incongruent products are de-pendent upon the risk associated wish the product. Meyers-Levy and Tybout (1989) suggest that the positive effect ofmoderate incongruity may be cjualified by consumer behav-ior variables. We demonstrate that perceived risk is one suchmoderating variable. Our results show that high risk reversesthe moderate incongruity effect. Additionally, the researchshows that this effect can be eliminated by relatively lowrisk and that it may only occur when there is no risk as-sociated with product evaluation.This finding has important implications for our under-standing of consumer evaluations, as well as for the way inwhich research on preferences i.s conducted. At this tmie,additional research should verify these findings and. in par-ticular, should examine the extent to which purchase-basedevaluations are different from relatively risk-free evaluationsin terais of schema effects. Though several researchers havesuggested that goals will be a significant moderator ofschema effects (e.g., Fiske and Neuberg 1990; Goodstein1993). little empirical research has investigated this or theperceived risk associated with different purchase goals. Theresearch presented here sugge.sts that people have a greaterpreference for the norm, such that they become more con-servative in their evaluations, when goals are associated withhigher risk. These findings highlight the importance of stud-ying both risk and goals in consumer contexts.The ability of ri.sk to increase preference for the normshould be of interest to consumer re.searchers for severalreasons. First, although a conjecture in the variety-seekingdomain is that lack of variety seeking can arise because ofa similar process (e.g.. Van Trijp, Hoyer. and Inman 19%),to date this hypothesis has not been empirically tested. Ourresults indicate that the preference for the norm constructdevelop with category experience (e.g., Fiske 1982; Mandler1982). Our data suggest that product schemata evoked dur-ing goal-derived categorization (cf. Barsalou 1985) mayoverride those formed in the absence of goals. Ratneshwarand Shocker (1991) propose that most consumer decisionsinvolve categorization processes that combine experientialand goal-derived schemata. They state that identifying thefactors that determine these combinations and the resultingchanges in market reactions holds important implicationsfor consumer re.search. Based on the research reported here.we believe that perceived risk causes consumers to combinetheir general product schema with one related to a particularpurchase scenario in evaluating products for purchase andthat this alters the evaluative responses to levels of schemacongruity.Finally, this research sounds a note of caution for theapplication of the moderate incongruit}' effect. The literatureon this effect has been deemed as normative by many re-searchers. For example, it has been suggested that buildingin moderate incongruity could be effective in product design(e.g.. Btoch 1995). The research presented here suggeststha^t schemata and risk interact in determining when themoderate incongruity effect occurs and that this processmust be better understood before such prescriptions aremade.Limitations and Future ResearchAlthough the purchase scenarios tested are more descrip-tive of actual consumer experience than are simple evalu-ations (cf. Ratneshwar and Shocker 1991). our particularscenarios stitl represent a trade-off between experimentalcontrol and external validity. Consumers may have manygoal-derived rationales for picking particular products thatdiffer in their degree of risk, complexity, and other factors(Barsalou 1985). Therefore, before generalizing these find-ings across all contexts, we suggest testing a broader rangeot^consumer goals, time pressures, and so on. Likewise, riskwas operationalized in these three studies with primarilyone type of risk, that is, social risk. Other types of risk (e.g.,performance, financial) wart-ant study to explore whethervarious types of risk affect processing and evaluations inthe same manner. We suggest that performance risk wouldmoderate congruity effects in the .same way. For example,if a consumer has a terrible headache, she will prefer apainkiller that is congruenl with her schema over one that448JOURNAL OF CONSUMER RESEARCHis moderately incongruent. On the other hand, it could beargued that financial outlays could be used as a way ofhandling risk: a high price could be seen to signal the qualityof incongnient products, and paying a high price could thusreduce the perceived risk (Bettman 1973). If this were thecase, the moderate incongruity effect could occur even underhigh risk if the moderately incongruent option had a highprice. In other words, price could moderate the effect ofrisk. Futtire research could examine the role of price in therisk-evaluation relationship.Moderate incongruity was manipulated in all three of thestudies, with alterations in an extrinsic factor of the prod-stimulation of the product novelty are likely to attenuate thepositive effect of incongruity on evaluations and are worthyof future research. Research demonstrates that dogmatism,a variable that is likely to decrease the motivation to processincongruity, moderates the relationship between moderateincongruity and evaluations (Meyers-Levy and Tyboui1989). Likewise, product category knowledge, which affectsthe ability to accommodate or resolve incongruity, moderates this relationship (Peracchio and Tybout 1996). We sug-gest that risk, by increasing preference for the norm, influ-ences consumers' enjoyment of moderate incongruity.Arousal is an important part of the theoretical basis oiuct—packaging. Although the perceived typicality ratingscollected in the studies support these manipulations of con-gruity. they do not address a larger issue of what determinesincongruity in general. More specifically, one could askwhether all manipulations of incongruity are equivalent. Thatis to say. is manipulating moderate incongruity with the pack-age the same as a manipulation via the product itsetf or viasome other product element, such as the ingredients or dis-tribution channel? One possibility is that intrinsic sources ofincongruity would lead to different effect.s than extrinsicsources of incongruity. It remains to future research to attemptto more specifically pin down the general concept of schemaincongruity and the degree to which differences arise fromdifferent types of incongruity.An additional limitation of this research is that it doesnot provide unique insight to the underlying process bywhich the moderating effect of perceived ri.sk arises. Thethree studies provide strong exploratory evidence that highrisk leads to preference for the norm, which influences eval-uations of moderately incongruent products. This doe.s not,however, eliminate all other explanations. For example, thedata provide some support for the idea that the moderatingeffect of risk is not due to a change in perceptions of in-congruity. In spite of this, in study 3, we do see a significantinfluence of risk on perceived typicality. It is possible thathigh risk changes how consumers process levels of incon-gruity or even alters the schema \"'boundaries\" such that thereare smaller ranges within levels of congruity. TTie data fromthe three studies presented here do not allow a good un-derstanding of the relationship between risk and perceivedtypicality. Our research focused on evaluations; researchgeared to more completely understanding the relationshipbetween risk and perceptions of congruity would contributeto a more complete model of the phenomenon.An important area for research is to extend the study ofthe boundary conditions for the moderate incongruity effect.Examination of the theoretical reasoning that pioneered re-search on the moderate incongniity effect (e.g., Mandler1982) suggests that four conditions are necessary for thiseffect to occur. Consumers must (1) find the moderate in-congruity to be arousing, (2) be motivated to process theincongruity, (3) be able to resolve the incongniity, and (4)enjoy or positively value the stimulation provided by thenovel product. Factors that affect arousal, processing, theability to resolve the incongruity, or the enjoyment of thethe moderate incongruity effect but is not addressed in theresearch presented here. The moderating variables investi-gated thus far in the literature (e.g.. prior knowledge) maymake moderate incongruity less arousing than it would beotherwise. Similarly, the attenuating effect of risk on eval-uations of moderate incongruity may be due to the fact thatboth risk and moderate incongruity increase arousal. Thiscould result in a level of arousal that is overstimulating toa point that affects the evaluation of the mtxlerately incon-gruent option. This possibility should be explored to extendthe strong empirical support for the moderating role ol' riskon the relationship between congruity and evaluations inorder to further our understanding of the reasons for con-sumers' preferences.[Received October I99S. Revised February- 2001. DavidGlen Mick served as editor, and Punam Anand Keller.ser\\'ed as associute edilor for this article,]REFERENCESBaron. Reuben M. and David A. Kenny (1986). \"The Moderator-Mediator Variable Distinction in Social Psychological Re-search: Conceptuiil, Strategic and Statistical Considerations.\"Journal of Penwnality and Social P.sychology. 51 (Decenvt>er), 1173-1182.Barsalou. Lawrence W. {1985), \"Ideals. Central Tendency, and Fre-quency of Instantiation as Determinants of Graded Structure.\"Journal of Experimental P.svchology: Learning, Memory, andCognition, 11 (October), 627-654.Berlyne, Daniel E, (1970). \"Novelty, Complexity, and HedonitValue.\" Perception and Psychophysics, 8 (November).279-2S6.Bettman, James R. (1973). \"Perceived Risk and Its Components:A Model and Empirical lt^\\.\" Journal of Marketing Research.10 (May), 184-190.Bloch, Peter H. (1995), \"Seeking the Ideal Form: Product Designand Consumer Response,\" Journal of Marketing. 59 (Jtily).16-29.Dowting. Grahame R. (1986), \"Perceived Risk: The Concept andhs Measurement,\" P.sychology and Marketing, 3 (Fall).lW-210.(1999), \"Perceived Risk,\" in The Elgar Companion to Con-sumer Research and Economic Psychology, ed. Peter E. Earland Simon Kemp. Cheltenham. UK: Edward Elgar, 419-424.Dowling, Grahame R. and Richard Staelin (1994), \"A Model ofPerceived Risk and Intended Risk-Handling Activity,\" Jour-nal of Consumer Research, 21 (June), 119-134.MODERATING EFFECT OF PERCEIVED RISKEdgington, Eugene S. (1972), \"An Additive Method for CombiningProbability Values from Independent Experimems,\" Journalof Psychology, 80 (March), 351-363.Erdem, Tulin (1998). \"An Empirical Analysis of Umbrella Brand-ing,\" Journal of Marketing Research. 35 (August), 339-35!.Fiske, Susan T. (1982). \"Schema-Triggered Affect: Applicationsto Social Perception,\" in Affect and Cognition: The .Seven-teenth Annual Carnegie Symposium on Cognition, ed. Mar-garet S. Clark and Su.san T. Fiske. Hiltsdale. NJ: Erlbaum.55-78.Fiske. Susan T. and Steven L, Neuberg (19901, \"\"A Continuum of449Montgomery. Cynthia A. and Birger Werneri'elt (1992), \"Risk Re-duction and Umbrella Branding,\" Journal of Business, 65(January), 31-50.Ogiethorpe. Janet E. and Kent B. Monroe (1987). \"Risk Perceptionand Risk Acceptability in Consumer Behavior: ConceptualIssues and an Agenda for Future Research.\" in AMA WinterMarketers Educators' Conference, ed. Russell W. Belk et al.,Chicago: American Marketing Association. 255-260.Payne. John W., James R. Bettman. and Eric J. Johnson (1993).The Adaptive Decision Maker. Cambridge: Cambridge Uni-versity Press.Impression Formation from Category-Based to IndividuatingProcesses: Influences of Information and Motivation on At-tention and Interpretation.\" in Advances in Experimental So-cial Psychology., Vol. 23. ed. Mark P. Zanna. San Diego. CA:Academic Press, 1-74.Goodstein, Ronald C. (1993). \"Category-Based Applications andExtensions in Advertising: Motivating More Extensive Pro-cessing.\" Journal of Consumer Research, 20 (June). 87-99.Hunter, John E. atid Frank L. Schmidt (1990). Methods of Meta-analysis: Correcting Error and Bias in Research Findings,Newbury Park, CA: Sage.lacobucci. Dawn (2001). \"Analysis of Variance: Can I Test forSimple Effects in the Presence of an In.significant Inierac-tion?\" Journal of Consumer P.sychology. 10 (1-2). 5-9.Kaplan. Leon B., George J. Szybillo. and Jacob Jacoby 11974).\"Components of Perceived Risk in Producl Purcha.se,\" Jour-nal of Applied P.'iychology. 59 (June). 287-291.Keppel. Geoffrey (1991), De.vi^n and Analysis: A Researcher'sHandbook, Englewood Cliffs, NJ: Prentice-Hall.Mandier, George (1982). \"The Structure of Value: Accounting lorTaste.\" in Affect and Cognition: The Seventeenth Annual Car-negie Symposium on Cognition, ed. Margaret S. Clark andSusan T. Fiske. Hillsdale. NJ: Erlbaum, .3-36.Meyers-Levy. Joan. Therese A. Louie, and Mary T Curren (1994).\"How Does the Congruity of Brand Names Affect Evaluationsof Brand Name Extensions''\" Journal of Applied Psychology.79 (February). 46-53.Meyers-Levy, Joan and Alice M. Tybout (1989). \"Schema Con-\" gruity as a Basis for Product Evaluation,\" Journal of Con-.sumer Research, 16 (June). 39-54.Peracchio. Laura A. and Joan Meyers-Levy (1994), \"Ho* Am-biguous Cropped Objects in Ad Phoios Can Affect ProductEvaluations,\" Journal of Consumer Research, 21 (June),190-204.Peracchio. Laura A. and Alice M. Tybtiut (1996), \"The ModeratingRole of Prior Knowledge in Schema-Based Product Evalua-tion,\" Jouraa/^/Cwwumfr/fcvforr/j. 23 (December), 177-192.Raltieshwar. S. and Allan D. Shocker (1991), \"Substitution in Useand the Role of Usage Context in Product Category Struc-tures.\" Journal of Marketing Research. 28 (Augusi). 281-295.Rosenthal. Robert (1978), \"Combining Results of IndependentS\.\" P.'Tvchological Bulletin. S5 (January), 185-193.Stayman. Douglas M.. Dana L. Alden, and Karen H. Smith (1992),\"Some Effects of Schematic Processing on Consumer Ex-pectations and Disconlintiation Judgments.\" Journal of Con-sumer Re.search, 19 (September). 240-255.Steenkamp, Jan-Benedict E, M. and Hans Baumgartner (1995),•'Development and Cross-Cultural Validation of a Shon Formof CSI as a Measure of Optimum Stimulation Level,\" Inter-national Journal of Re.search in Marketing. 12, 97-104.Van Trijp. Hans C. M., Wayne D. Hoyer. and J. Jeffrey Inman(1996). \"Why Switch']' Product-Category Level Explanationsfor True Variety-Seeking Behavior.\" Journal of MarketingResearch. 33 (August). 281-292.Wansink. Brian (1996). \"Can Package Size Accelerate Usage Vol-ume\"^\" Journal of Marketing, 60 (July), 1-14.Zuckerman. Marvin (1991). \"One Person's Stress Is Another Per-son\\s Pleasure.\" in Stre.s.'i and Emotion, Anxiety. Anger, andCuriositx. Vol. 14, ed. Charles D. Spielberger and Irwin G.Sarason. New York: Hemisphere. 31-45.