An Examination of Different Exlanations for the Mere Exosure Effect XIANG FANG SURENDRA SINGH ROHINI AHLUWALIA* This article investigates two cometing exlanations of the mere exosure effect the cognition-based ercetual fluency/misattribution theory (PF/M) and the affect-based hedonic fluency model (HFM) under incidental exosure conditions. In two studies, the classical mere exosure effect is relicated in the context of banner advertising. The findings rule out the cognition-based PF/M and suggest that the sontaneous affective reaction resulting from ercetual fluency is a crucial link between fluency and evaluation. The studies rovide strong evidence that the sontaneous affect influences evaluative judgments through a more comlex rocess, likely by coloring the interretation of the fluency exerience and the nature of resulting metacognitions relating fluency to liking. Theoretical and managerial imlications of the findings are mentioned. The majority of ad exosure occurs under incidental conditions where the audiences attention is focused elsewhere such as reading a magazine or browsing a Web age, aroximating the mere exosure condition, in which the target stimulus is made just accessible to the individual s ercetion (Zajonc 1968, 1). Reeated exosures under these conditions result in enhanced liking for a neutral stimulus. The two major exlanations for this effect are both based on the concet of fluency, or the ease with which information is rocessed, but they differ in the underlying mechanisms. The ercetual fluency/misattribution model, or PF/M, utilizes a cognitive ersective and suggests that eole generate inferences or metacognitions based uon their fluency exerience (e.g., inferring that stimuli that come to mind more readily are better liked), thus misattributing their fluency exerience to evaluations of the stimuli *Xiang Fang is assistant rofessor of marketing, Sears School of Business, Oklahoma State University, Tulsa, OK 74106 (xifang@okstate.edu). Surendra Singh is Southwestern Bell Chair in Marketing, School of Business, University of Kansas, Lawrence, KS 66045 (ssingh@ku.edu). Rohini Ahluwalia is associate rofessor of marketing, Carlson School of Management, University of Minnesota, Minneaolis, MN 55455 (rahluwalia@ csom.umn.edu). This article is based on the dissertation of the first author. The authors acknowledge the helful inut of the editor, associate editor, and the three reviewers. They also thank Jack Brehm, Sanjay Mishra, Dennis L. Rosen, Tracy Suter, Marlys J. Mason, and Julie Ruth for comments on earlier drafts. Corresonding author: Xiang Fang. Dawn Iacobucci served as editor and Mary Frances Luce served as associate editor for this article. Electronically ublished March 13, 2007 000 (Bornstein and D Agostino 1992, 1994). Most marketing studies have utilized PF/M as the exlanatory mechanism (e.g., Janiszewski and Meyvis 2001; Nordhielm 2002). The hedonic fluency model, HFM (Winkielman and Cacioo 2001), resents an alternative affect-based exlanation, arguing that ositive affect generated by the dynamics of fluent rocessing enhances stimuli evaluations (Winkielman et al. 2003). It is not clear which model (cognitive or affective) better exlains mere exosure effect (MEE) under a articular set of conditions (Lee 2004; Schwarz 2004). We test the different rocess exlanations for the MEE in two studies using banner ads as stimuli. Since most viewers ay minimal attention to banner ads, they not only are ideal stimuli for examining the MEE but also hel us accomlish a secondary goal: gaining a deeer understanding of how banner advertising works. Given the relative newness of the Web as a medium, our knowledge of how these ads influence viewers is woefully limited. The first study examines whether the MEE is able to account for the ersuasion effects of banner advertising. The second study examines the underlying rocess issues in the context of banner advertising. EXPLANATIONS FOR THE MERE EXPOSURE EFFECT Percetual Fluency/Misattribution Model: The Cognitive Persective The PF/M contends that reeated exosures to a stimulus enhance its ercetual fluency. When an individual is un- 2007 by JOURNAL OF CONSUMER RESEARCH, Inc. Vol. 34 June 2007 All rights reserved. 0093-5301/2007/3401-0010$10.00
000 JOURNAL OF CONSUMER RESEARCH aware of the true source (rior exosure) of the enhanced ercetual fluency, (s)he may misattribute it to liking for the stimulus or, deending uon the contextual cues rovided by the task, to other relevant judgments regarding the stimulus, such as fame, truth, duration, loudness, stimulus brightness and darkness, and narrow/broad roduct distribution (e.g., Mandler, Nakamura, and Van Zandt 1987; Nordhielm 2002). If the individual realizes having seen the stimulus earlier, (s)he will engage in an automatic and effortless correction rocess, revising the initial interretation of fluency and discounting the stimulus evaluation. The PF/M assumes that (1) rocess facilitation is an affectively neutral, arousal-like exerience and (2) this exerience can influence extremity of the evaluation by leading to either a more ositive or a more negative evaluation of the stimulus, deending on the contextual factors (Winkielman and Cacioo 2001). Thus, if the evaluation involves darkness (brightness) judgment, more extreme evaluation of darkness (brightness) should occur. Hedonic Fluency Model: The Affective Persective The HFM focuses secifically on reference judgments (as oosed to other judgments, such as brightness, fame, and truth) and argues that the dynamics of information rocessing itself (i.e., rocessing fluency) lead to ositive affective resonses, which in turn lead to a more ositive evaluation. High fluency may generate ositive affect artly because it indicates stimulus familiarity, which signals a harmless situation. It may also indicate rogress toward the goal of successful recognition and coherent interretation of the target, which is a rewarding exerience (Winkielman and Cacioo 2001). Winkielman and Cacioo (2001) demonstrated that ercetual fluency increases were associated with affective resonses, as measured by facial electromyograhy (EMG). The rocess imlied in this model is akin to the affect-asinformation model (Schwarz and Clore 1983), where articiants infer their evaluations from how they feel (e.g., if I feel good, I must like it ). Note that this rocess is also metacognitive in nature but differs from the PF/M in that the inferences are based on the affective exerience and not the exerience of ercetual fluency. These data, however, are also consistent with another ossibility that affect influences the interretation of the fluency exerience and the resultant metacognitions relating fluency to liking. In some situations affect can mobilize subsequent thoughts, influencing the direction of cognitions as well as interretation of evidence (e.g., Pham et al. 2001). Therefore, it is ossible that affect may influence evaluations via its effect on interretation of the fluency exerience. More imortant, the EMG data also do not rule out the ossibility that a arallel rocess, in which individuals generate metacognitions based on their fluency exerience (PF/M), might also occur and influence evaluations. Therefore, these data leave the door oen for an array of alternative rocesses by which ercetual fluency could influence evaluation. Lee (2001, 2004) makes an insightful suggestion for testing the rocesses driving the effects of ercetual fluency on evaluations. She argues that if the liking resonses were based on naive theories or metacognitions of the inferential nature based on the ercetual fluency exerience (PF/M), then individuals would correct their judgments when they become aware of the source of ercetual fluency (e.g., noticed the reetition). However, if these resonses were based on affect, then they would not be likely to exhibit this correction effect. To test this idea, Lee (2001) asked articiants to count the number of triangles in abstract atterns and varied their exosure frequency (zero to six). Later they were resented with airs and asked to make either cognitive (choosing the attern with more triangles) or affective judgments (choosing the attern that they liked more). Although a monotonic relationshi between choice and number of exosures was obtained with affective judgments, it failed to emerge with cognitive judgments after six exosures. Lee interreted these data as suggesting that correction rocesses set in for the cognitive judgments at six exosures; however, no correction occurred with affective judgments. The lack of correction effects in the affective judgment condition could ossibly have emerged due to similarity between the initial task (counting the number of triangles) and the cognitive judgment (which attern has more triangles) but not the affective judgment (which attern is referred). The mismatch of tasks in the affective judgment condition could resumably have lowered recognition likelihood in this condition. Therefore, the imlications of this study for the rocess underlying MEE are not conclusive. In sum, although the ast decade of research has clearly established that ercetual fluency lays a key role in influencing evaluations, the jury is still out on how this effect occurs. Past literature suggests at least three otential rocesses: (i) fluency-based metacognitions (the cognitive route), (ii) affect-based metacognitions (affect as information), and (iii) affect influences the interretation of the fluency exerience and metacognitions generated from it. Our goals are to reconcile the evidence resented from different ersectives, revise the existing theories of the MEE, and generate a more integrative model for understanding how ercetual fluency influences evaluations. Most rior work on MEE in marketing is based on focal exosure to target stimuli, and none has used a direct measure of ercetual fluency. The first study tests the MEE using incidental exosure to banner ads and directly measures ercetual fluency using the ercetual identification measure (e.g., Jacoby and Dallas 1981) to establish the validity of MEE in the chosen substantive context (banner advertising). The second exeriment, using a misattribution aradigm, tests the different exlanations based on the rationale that if individuals are made aware of the source (of either fluency or affect), they are likely to correct their evaluations.
DIFFERENT EXPLANATIONS FOR THE MERE EXPOSURE EFFECT 000 EXPERIMENT 1: MERE EXPOSURE EFFECTS AND BANNER ADVERTISING Design and Procedure Using a 3 (exosure frequency: zero vs. five vs. 20) # 2 (evaluation orientation: ositive vs. negative resonses) between-subjects design, 232 students at the University of Kansas were randomly assigned to the six cells; 13 failed to comlete all arts, leaving 219 usable resonses. Both ositive and negative evaluation orientations were measured because generation of ositive affect, associated with ercetual fluency, is exected to change the ositive but not negative evaluations (Winkielman and Caccioo 2001). Uon arrival in a comuter lab, articiants were instructed to carefully read a five-age article on online education with the exectation of later answering questions about the article. The article s mood neutrality was ascertained via a retest. Banner ads rotated (for 5 seconds each) on to of the article on the comuter screen. In the fiveexosure condition, the target ad aeared once on each of the five ages, with five filler ads in the following order: F 1 F 2 TF 3 F 4 F 5 (where T target ad and F filler ad). In the 20-exosure condition, the target ad aeared four times on each age (with the same five fillers) as follows: F 1 T F 2 TF 3 TF 4 TF 5. In the zero-exosure (control) condition, only filler ads were resented. Since articiants read the article at their own ace, each age was set to aear on the screen for a minimum of 45 seconds (i.e., the resondents could not move to the next age before 45 seconds) to afford exosure oortunity for all ads (fillers and the target) on each age. After reading the article, articiants resonded to a few questions related to the article, followed by the key deendent measures. Stimuli Target Banner Ad. A full-size (468 # 60 ixels) banner ad for a digital camera with a novel and neutral brand name Pretec was created (retest, n 20: Mliking 3.88, Mfam. 1.60, seven-oint scales). The ad simly featured the brand name (Pretec digital camera), ictures of digital cameras, and a click here button. Layout of Web Pages. Using the E-rime 1.1 software, an interface similar to a student online magazine was develoed. Banner ads were laced on the to of the article (outside of articiants focal view) to create incidental exosure, while the article on online education was laced in articiants focal view. Banner ads rotated on to, with one ad aearing at a time. Deendent Variables Percetual Identification Task. A direct, objective measure of rocess fluency, the ercetual identification task, was administered for both the target ad and filler ads. For each banner ad, articiants saw 15 visually degraded banner ad images, each aearing for 1 second. Though initially comletely masked by noise, over rogressive trials the ad became clearer. Particiants were instructed to ress any key as soon as they thought they could identify the brand name shown in the ad. Once articiants ressed a key, the ad image was relaced by a question age asking them to tye the brand name in a blank box. Particiants were instructed to resond as quickly and as accurately as ossible. The rogram recorded the resonse time (from the beginning of the resentation to the time when articiants ressed the key) and brand name. Particiants comleted two ractice trials before the actual task. Ad Evaluation. Following Winkielman and Cacioo (2001), articiants in the ositive (negative) resonse grou were asked to rate the degree of their ositive (negative) reaction to the banner ad on a nine-oint scale. The scale range was one (no ositive/negative reaction) to nine (very ositive/negative reaction). Recognition. In a two-alternative forced-choice recognition test, articiants were shown a air of banner ads, the target ad for Pretec and a decoy (Relisys). They were asked to select the ad that was shown on the Web ages reviously. Results and Discussion Recognition. Consistent with the MEE, recognition scores in both treatment grous were not different from the control grou (all s 1.25). The scores were not different from the.50 chance level (all s 1.40), confirming the incidental nature of the ad exosures. Percetual Fluency. After excluding 12% of outliers and missing/incorrect resonses (evenly distributed across conditions), a two-way ANOVA with exosure frequency and evaluative orientation as two indeendent variables was erformed on the remaining data. A significant main effect of exosure (ex.) frequency emerged ( M0ex. 9,045 milliseconds, 8,281 milliseconds, M20ex. 7,895 milliseconds; F(2, 186) 5.97,.003). Planned comarisons revealed that articiants in both treatment conditions had shorter resonse times than those in the control condition (both s!.015, one-tailed), although the means between two treatment conditions did not differ statistically ( t(186) 1.13,!.13, one-tailed). A trend test on exo- sure frequency revealed a significant linear trend in resonse time ( F(1, 189) 11.75,.001), indicating that reeated incidental exosures increase ercetual fluency of the target banner ad. See table 1. Evaluation of the Target Banner Ad (Pretec). Oneway ANOVAs were conducted on both ositive and negative evaluations of the target banner ad (Pretec). Consistent with ast research, there was a significant main effect of exosure frequency in the ositive evaluation condition (M0ex. 3.91, 4.73, M20ex. 5.11; F(2, 104) 3.24,!.05). Particiants in both treatment grous evaluated the
000 JOURNAL OF CONSUMER RESEARCH TABLE 1 MEANS AND STANDARD DEVIATIONS OF DEPENDENT VARIABLES IN STUDY 1 Positive evaluation Negative evaluation 0 ex. (n 34) 5 ex. ( n 37) 20 ex. ( n 36) 0 ex. (n 39) 5 ex. (n 37) 20 ex. ( n 36) Percetual fluency (n samle size) a 8,960 (1,768) 8,246 (1,666) 7,856 (1,711) 9,118 (2,257) 8,320 (1,793) 7,936 (2,112) n 31 34 31 36 31 29 A ad 3.91 (2.08) 4.73 (2.04) 5.11 (1.91) 3.31 (1.85) 3.05 (2.24) 3.39 (2.02) Recognition.53.51.56.46.57.58 NOTE. ex. exosures. a The smaller samle size in each grou is due to missing resonses, incorrect resonses, and outliers. target ad more ositively than those in the control grou (both s!.05, one-tailed), although there was no difference between the two treatment grous ( t! 1). In contrast, a sig- nificant main effect of exosure frequency did not emerge in the negative evaluation condition ( M0ex. 3.31, 3.05, M20 ex. 3.39; F! 1). A trend test revealed a significant linear trend in ositive evaluations as exosure frequency increased ( F(1, 104) 6.23,!.015) but not with negative evaluations ( F! 1). Discussion. The attern of obtained results indicates that reeated incidental exosures to banner ads resulted in increased ercetual fluency without increasing recognition. Consistent with ast research, we found increased ercetual fluency to be accomanied with more ositive evaluations of the ad but not with negative evaluations, suggestive of a ositive affect. Note, however, that the ercetual identification task used in this study is an objective measure of ercetual fluency that rovides direct evidence of fluency with reeated exosures to stimuli (suorting the MEE). Assessments of subjective fluency (the conscious exerience of rocessing ease that may become disassociated from objective assessments of fluency; Winkielman et al. 2003, 193), though, are needed to clearly distinguish between the different exlanations and to test mediation, since these mechanisms are based on metacognitions resulting from the subjective exerience of fluency. (Interestingly, a mediation analysis using the objective measure from this study did not suort a significant mediation of exosure frequency on evaluation by fluency. Therefore, subjective fluency is measured in study 2.) EXPERIMENT 2: MISATTRIBUTION EFFECTS Objective and Predictions The underlying logic of this exeriment is that eole use an inut (affect or cognition) for judgments only as long as it is informative (Schwarz and Clore 1983), or else they tend to discount its effects on judgments. If articiants misattribute the basis of their evaluations (fluency or affect) to another source (e.g., music), then they would correct their judgments. We rovided articiants with the oortunity to misattribute either their fluency exerience or their felt affect to another source (background music). If consistent with PF/M, articiants liking judgments are solely based on the fluency exerience and its resultant metacognitions; correction effects would emerge only in the fluency-misattribution condition. However, if their evaluations are driven by the affect-as-information model, they would likely exhibit a correction effect in the affect-misattribution condition (affect attributed to the music and, therefore, no longer informative about the ad) but not in the fluency-misattribution condition, as their evaluation is based on the generated affect and not the fluency exerience er se (see Lee [2001] for a similar argument). However, if the generated affect lays a more comlex role by influencing the articiant s interretation of the fluency exerience and mobilization of related thoughts, then the ad evaluations would be corrected in both misattribution conditions, since both affect and the fluency exerience are involved in an interrelated manner in these metacognitions. Lee (2004) argues that the more sontaneous and automatic the affect (e.g., affect as information), the less likely it is to undergo correction; however, the higher the likelihood of cognitions, the greater the ossibility of correction. If Lee s argument holds, then correction effects under the affect-misattribution condition would be less likely if articiants were using affect solely as information and not utilizing it to interret the fluency exerience. Design and Particiants In a 2 (exosure frequency: zero vs. five) # 4 (misattribution: fluency vs. affect vs. general caution vs. no attribution) design, 304 undergraduate students were randomly assigned to eight cells. Misattribution maniulation was adated from Fazendeiro et al. (2005). In the fluency-misattribution grou, articiants were told that the background music might make it easier to rocess the information they were exosed to, which might give them a sense of fluency (things coming to their mind faster or easier). They were asked to make the next set of evaluations ignoring the feeling
DIFFERENT EXPLANATIONS FOR THE MERE EXPOSURE EFFECT 000 TABLE 2 MEANS AND STANDARD DEVIATIONS OF DEPENDENT VARIABLES IN STUDY 2 0 ex. ( n 37) Misattribution instruction No instruction General caution Fluency Affect 5 ex. ( n 40) 0 ex. ( n 38) 5 ex. ( n 36) 0 ex. ( n 39) 5 ex. ( n 38) 0 ex. ( n 39) 5 ex. ( n 37) A ad 4.75 (1.59) 5.92 (1.90) 4.76 (1.47) 5.77 (2.03) 4.81 (1.22) 4.72 (1.86) 4.83 (1.72) 4.74 (1.82) Ad-induced feeling 3.77 (1.86) 4.46 (1.67) 3.52 (1.45) 4.39 (1.76) 3.58 (1.59) 3.33 (1.64) 4.01 (1.48) 3.61 (1.43) Subjective fluency 3.22 (2.24) 3.98 (2.39) 2.63 (2.02) 4.11 (2.69) 3.10 (2.07) 2.82 (2.14) 4.00 (2.49) 3.62 (2.52) Recognition.51.48.50.53.54.55.56.51 NOTE. ex. exosures. of fluency. In the affect-misattribution grou, articiants were told that the music might induce different tyes of feelings. They were instructed to make the evaluations ignoring their feelings. In addition, a general-caution condition was included to hel assess whether any caution, irresective of the focus on affect or fluency, might reduce evaluations. In this condition, articiants were told that the background music might influence their judgments and were instructed to make the evaluations ignoring the effects of background music. No music-related instructions were given in the control condition. Procedure and Measures Uon arrival in the comuter lab, articiants were instructed to carefully read the same five-age article on online education from study 1 and to exect to answer several questions about the article. Exosure frequency was maniulated as in study 1. During the entire study 2, articiants were exosed, through headhones, to a iece of soft meditation music. 1 After reading the article, articiants answered a few questions about it. Before making their evaluative judgments, they received different misattribution instructions as discussed above. Attitude toward the ad was measured with a traditional biolar attitude measure (three nine-oint items: dislike/like, unleasant/leasant, and bad/good; Cronbach s alha.93). To assess ad-induced feelings, following Madden, Allen, and Twible (1988), articiants resonded to the romt, How does this ad make you feel? followed by seven adjectives (good, leased, cheerful, joyous, amused, layful, and stimulated; Cronbach s alha.94) on nineoint scales (not at all/very much so). Subjective fluency was assessed by asking articiants to indicate how easy it was for them to rocess the ad (nine-oint scale: very easy to rocess/very difficult to rocess). Finally, articiants resonded to maniulation check questions (described subsequently), a recognition test, and demograhic items. 1 The music can be requested from the authors. Results and Discussion Maniulation Checks. More than 80% of resondents in each condition resonded ositively to a general maniulation check question ( Did you take the information in instructions into account when making your judgments? ) adoted from Fazendeiro et al. (2005). Resonses to the two secific questions relating to misattributions (extent to which they ignored the effects of [i] ease of rocessing the stimuli and [ii] feelings aroused on nine-oint scales) are discussed next. Particiants in the fluency-misattribution condition ( Mflu. 5.60) were more likely to reort ignoring the fluency effects than those in other instruction conditions (Mno ins. 4.51, Mgen. cau. 4.96, and Maffect 4.96; all s!.02, onetailed). Similarly, resondents in the affect-misattribution condition ( Maffect 5.47) were more likely to ignore the effects of feelings aroused than those in other conditions (Mno ins. 4.69, t(300) 2.17,!.02; Mgen. cau. 5.18, t(300)! 1; and Mflu. 4.92, t(300) 1.52,!.07; all one-tailed). These results indicate that articiants followed the instructions when making their evaluations. Recognition scores in all grous did not differ from the.50 chance level (all s 1.40), verifying the incidental nature of ad exosure. See table 2. Attitude toward the Target Banner Ad. A two-way ANOVA on attitude toward the ad revealed a significant interaction effect of exosure frequency and misattribution ( F(3, 296) 3.02,.03). Particiants in the five-exosure condition evaluated the ad more ositively than those in the control in both the no-misattribution ( M0ex. 4.75, 5.92; F(1, 296) 8.91,!.002, one-tailed) and general-caution conditions ( M0ex. 4.76, 5.77; F(1, 296) 6.34,!.01, one-tailed). Correction effects, however, emerged in both the fluency- and the affect-misattribution conditions, so that the difference between the control and five-exosure conditions was not significant (fluency misattribution: M0ex. 4.81, 4.72; affect misattribution: M0ex. 4.83, 4.74; both F s! 1). By demonstrating that misattribution of either the fluency
000 JOURNAL OF CONSUMER RESEARCH signal or the affective exerience eliminates the fluency s effect on judgments, these results clearly suort the more comlex model in which affect influences the interretation of fluency. Ad-Induced Feelings. A marginally significant main effect of misattribution ( F(3, 296) 2.33,!.08) qualified by a significant interaction of exosure frequency and misattribution ( F(3, 296) 3.04,!.03) emerged. Planned comarisons revealed that in both the no-misattribution and general-caution conditions, articiants in the five-exosure condition reorted more ositive feelings than those in the control grou (no misattribution: M0ex. 3.77, 4.46; F(1, 296) 3.53,!.035, one-tailed; general caution: M0ex. 3.52, 4.39; F(1, 296) 5.42,!.02, one-tailed). Significant differences in the ad-induced feelings between the five-exosure and control grous did not emerge in the fluency- ( M0ex. 3.58, 3.33; F! 1) and the affect-misattribution conditions ( M0ex. 4.01, 3.61; F(1, 296) 1.22, 1.13, one-tailed), where subjects exhibited correction effects. Subjective Fluency. A two-way ANOVA on the subjective fluency measure revealed a significant interaction effect of exosure frequency and misattribution (F(3, 296) 2.74,!.05). Planned comarisons revealed that articiants in the five-exosure grou erceived the ad easier to rocess than those in the control grou, in both the no-misattribution (marginally significant: M0ex. 3.22, 3.98; F(1, 296) 2.04,!.08, one-tailed) and the general-caution conditions ( M0ex. 2.63, 4.11; F(1, 296) 7.46,!.005, one-tailed). No significant differences in subjective fluency were found between the five-exosure and the control grous in both the fluency- ( M0ex. 3.10, 2.82; F! 1) and the affect-misattribution conditions ( M0ex. 4.00, 3.62; F! 1). 2 Discussion. The results of exeriment 2 clearly rule out a urely cognitive exlanation of fluency s effects on evaluations and suggest that not only does the fluency exerience generate affect but that the generated affect is also likely to influence subsequent cognitions as well as interretation of the fluency exerience. GENERAL DISCUSSION We conducted two studies to understand the rocesses underlying MEE in incidental ad exosure conditions. In study 1, the classical MEE was relicated in the domain of banner ads: reeated incidental exosures to banner ads, just under the level of ercetion, increased ercetual fluency and resulted in more ositive evaluations. In study 2, we 2 A series of mediation analyses (based on Baron and Kenny 1986) suggest that both subjective fluency and ad-induced feelings mediate the effect of exosure frequency and misattribution on A ad. In addition, the ad-induced feelings artially mediate the effect of subjective fluency on A ad. In other words, fluency exerience and the resulting affect rovide a crucial link between exosure frequency and ad evaluations. used the misattribution aradigm to test the underlying rocesses and found that misattribution of either the fluency signal or the affective exerience eliminated the MEE, which suorted the more comlex rocess whereby the fluency exerience generates affect and the generated affect influences subsequent cognitions and the interretation of the fluency exerience. Although our studies focus on affective (reference) judgments, our findings may also enhance understanding of cognitive judgments (e.g., truth, fame, and brightness). When exosure to a stimulus occurs under conditions of mere exosure, resondents are likely to have at least two available inuts at the time of judgment: fluency exerience and affect. If affect is erceived to be diagnostic (as with reference judgments), it is likely to be used as an inut in decision making (affect as information) and may be instrumental in interreting other inuts (e.g., the fluency exerience). However, when affect is not diagnostic for the judgment at hand (e.g., fame, brightness, and truth), it is not likely to be used as an inut in decision making (e.g., Feldman and Lynch 1988). Instead, individuals might generate other tyes of naive theories or metacognitions (e.g., easy to retrieve, so must be brighter) to interret the fluency exerience. Imortantly, these effects are likely to aear only when no other diagnostic information is available, and the judgment is based on whatever little information is available in the evaluation setting. One caveat: our findings may be limited to low-involvement, incidental ad rocessing conditions. One ractical imlication of this research is that online advertisers might be lacing excessive emhasis on the click-through rates the rimary metric for measuring the effectiveness of online ads. Our results suggest that even when there is no overt sign of effectiveness, such as recognition or click through, the banner ads may still imact ad liking. Past research suggests that reliance on affect tends to increase when cognitive resources are severely constrained (e.g., Shiv and Fedorikhin 1999), given the more automatic and less resource-intensive nature of affective rocesses. Our finding that affect is a key rocess underlying the mere exosure effect imlies that these effects are exected to strengthen under resource-constrained decision-making situations (e.g., limited cognitive caacity, lack of motivation to rocess information, and time ressure), which are revalent in marketlace environments. Additionally, findings from our study 1 suggest that consumers tend to have a relatively high level of tolerance for reeated exosure to banner ads the wear-out effects of banner ads did not kick in even after 20 exosures in this exeriment. A future research issue might involve examining the exosure levels at which wear-out effects are likely to emerge. Another issue that would benefit from future research is the generalizability of these effects to brand evaluations. Reetition is just one of many variables that may influence fluency and hence evaluations. Variables, such as symmetry,
DIFFERENT EXPLANATIONS FOR THE MERE EXPOSURE EFFECT 000 harmony, and comlexity, that govern aesthetic exerience also affect fluency exerience (Schwarz 2004). Therefore, our results may be tangentially relevant to logo design, oint-ofurchase dislay, or roduct design domains where aesthetics lay an imortant role. REFERENCES Baron, Reuben M. and David A. Kenny (1986), The Moderator- Mediator Variable Distinction in Social Psychological Research: Concetual, Strategic, and Statistical Considerations, Journal of Personality and Social Psychology, 31 (December), 1173 82. Bornstein, F. Robert and Paul R. D Agostino (1992), Stimulus Recognition and the Mere Exosure Effect, Journal of Personality and Social Psychology, 63 (October), 545 52. (1994), The Attribution and Discounting of Percetual Fluency: Preliminary Tests of a Percetual Fluency/Attributional Model of the Mere Exosure Effect, Social Cognition, 12 (Summer), 103 28. Fazendeiro, Tedra, Piotr Winkielman, Chun Luo, and Christoher Lorah (2005), False Recognition across Meaning, Language, and Stimulus Format: Concetual Relatedness and the Feeling of Familiarity, Memory and Cognition, 33 (2), 249 60. Feldman, Jack M. and John G. Lynch Jr. (1988), Self-Generated Validity and Other Effects of Measurement on Belief, Attitude, Intention and Behavior, Journal of Alied Psychology, 73 (August), 421 35. Jacoby, Larry L. and Mark Dallas (1981), On the Relationshi between Autobiograhical Memory and Percetual Learning, Journal of Exerimental Psychology: General, 110 (3), 306 40. Janiszewski, Chris and Tom Meyvis (2001), Effects of Brand Logo Comlexity, Reetition, and Sacing on Processing Fluency and Judgment, Journal of Consumer Research, 28 (June), 18 32. Lee, Angela Y. (2001), The Mere Exosure Effect: An Uncertainty Reduction Exlanation Revisited, Personality and Social Psychology Bulletin, 27 (10), 1255 66. (2004), The Prevalence of Metacognitive Routes to Judgment, Journal of Consumer Psychology, 14 (4), 349 55. Madden, Thomas J., Chris T. Allen, and Jacquelyn L. Twible (1988), Attitude toward the Ad: An Assessment of Diverse Measurement Indices under Different Processing Sets, Journal of Marketing Research, 25 (3), 242 52. Mandler, George, Yoshio Nakamura, and Billie J. Van Zandt (1987), Nonsecific Effects of Exosure on Stimuli That Cannot Be Recognized, Journal of Exerimental Psychology: Learning, Memory, and Cognition, 13 (4), 646 48. Nordhielm, Christie L. (2002), The Influence of Level of Processing on Advertising Reetition Effects, Journal of Consumer Research, 29 (December), 371 82. Pham, Michel Tuan, Joel B. Cohen, John W. Pracejus, and David G. Hughes (2001), Affect Monitoring and the Primacy of Feelings in Judgment, Journal of Consumer Research, 28 (Setember), 167 88. Schwarz, Norbert (2004), Metacognitive Exeriences in Consumer Judgment and Decision Making, Journal of Consumer Psychology, 14 (4), 332 48. Schwarz, Norbert and Gerald L. Clore (1983), Mood, Misattribution, and Judgments of Well-Being: Informative and Directive Functions of Affective States, Journal of Personality and Social Psychology, 45 (3), 513 23. Shiv, Baba and Alexander Fedorikhin (1999), Heart and Mind in Conflict: The Interlay of Affect and Cognition in Consumer Decision Making, Journal of Consumer Research, 26 (December), 278 92. Winkielman, Piotr and John T. Cacioo (2001), Mind at Ease Puts a Smile on the Face: Psychohysiological Evidence That Processing Facilitation Increases Positive Affect, Journal of Personality and Social Psychology, 81 (6), 989 1000. Winkielman, Piotr, Norbert Schwarz, Tedra A. Fazendeiro, and Rolf Reber (2003), The Hedonic Marking of Processing Fluency: Imlications for Evaluative Judgment, in The Psychology of Evaluation: Affective Processes in Cognition and Emotion, ed. Jochen Musch and Karl C. Klauer, Mahwah, NJ: Erlbaum, 189 217. Zajonc, Robert (1968), Attitudinal Effects of Mere Exosure, Journal of Personality and Social Psychology Monograhs, 9 (No. 2, Part 2), 1 27.