3UHDWWHQWLYH0HUHSRVXUHIIHFWV XWKRU Vf 5HYLHZHGZRUN Vf 6RXUFHRXUQDORIRQVXPHU5HVHDUF ID: 121318
Download Pdf The PPT/PDF document "Journal of Consumer Research, Inc." is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Journal of Consumer Research, Inc. 3 U H D W W H Q W L Y H 0 H U H ( [ S R V X U H ( I I H F W V $ X W K R U V \f 5 H Y L H Z H G Z R U N V \f 6 R X U F H - R X U Q D O R I & R Q V X P H U 5 H V H D U F K 9 R O 1 R ' H F \f S S 3 X E O L V K H G E \ 6 W D E O H 8 5 / $ F F H V V H G Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range ofcontent in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new formsof scholarship. For more information about JSTOR, please contact support@jstor.org. The University of Chicago Press and Journal of Consumer Research, Inc. are collaborating with JSTOR todigitize, preserve and extend access to Journal of Consumer Research. http://www.jstor.org This content downloaded on Wed, 26 Dec 2012 13:00:08 PMAll use subject to Preattentive Mere Exposure Effects CHRIS JANISZEWSKI* Mere exposure to a brand name or product package can encourage a consumer to have a more favorable attitude toward the brand, even when the consumer cannot recollect the initial exposure. This article provides evidence that mere exposure effects persist when initial exposures to brand names and product packages are incidental, devoid of any intentional effort to process the brand information. These unintentional mere exposure effects are attributed to preattentive processes and are explained through hemispheric processing theory. T he last two decades of research on the influence of repeated stimulus exposures has demonstrated that preferences can be formed without an accompa- nying awareness of the preference formation process (Bornstein 1989). Contexts that provide impoverished exposures or limited processing resources can make it difficult for people to recognize a previously presented stimulus, yet these exposures are sufficient to generate a feeling of familiarity that is later interpreted as a pref- erence for the stimulus (Mandler, Nakamura, and van Zandt 1987; Zajonc 1980). These findings have resulted in considerable discussion of the origins of affective re- sponses (Lazarus 1982, 1984; Zajonc 1980, 1984) as well as research designed to provide evidence for re- solving this debate (LeDoux 1986, 1989; Mandler et al. 1987). The compelling evidence that people have positive affective responses toward stimuli they cannot remem- ber having seen suggests that ads receiving limited at- tention may nonetheless be effective agents of attitude change (Anand, Holbrook, and Stephens 1988; Jani- szewski 1988, 1990a). This conclusion may be prema- ture, given that most demonstrations of the influence of mere exposure on affective responses occur in con- texts that encourage intentional processing of the target stimuli-the processing is goal directed and involved, albeit resource limited (Bornstein 1989). In contrast, the incidental exposure contexts to which consumer re- searchers hope to generalize mere exposure findings rarely involve intentional processing of the target stim- ulus and are often limited to a preattentive analysis, of the brand name or product package (Krugman 1986). Thus, it is would seem important to investigate whether preattentive processes are instrumental in the formation of affective responses and, if so, to provide an expla- nation of how these preattentive processes might op- erate. The primary goal of this article is to provide evidence that the affective responses associated with incidental exposure to brand names and packages may depend not only on the goal-directed mechanisms associated with intentional, attentive processes but also on the non- goal-directed mechanisms associated with preattentive processing. Attentive and preattentive processes can be differentiated by documenting the unique response of each set of mechanisms to manipulations of the con- sumer's environment. This effort begins with a discus- sion of two potential sources of the affective responses that result from exposure: (1) attentive-resource allo- cation associated with intentional information acqui- sition and (2) hemispheric activation associated with preattentive processing. Experiment 1 demonstrates that preattentive processes uniquely influence affective responses. Affective responses to a nonfocal stimulus are shown to be sensitive to other nonfocal stimuli in the environment. Experiment 2 differentiates between attentive and preattentive processes by demonstrating that increasing the processing requirements associated with an attended focal task can enhance, rather than degrade, the affective response toward a nonfocal stim- ulus. Experiment 3 illustrates that a preattentive feature analysis and a preattentive semantic analysis can influ- ence the affective response that results from an inci- dental exposure. BACKGROUND. Affective Responses Hoffman (1986) has identified three potential sources of affective responses, each sensitive to a unique set of variables. The evaluation of a stimulus can be attributed to (1) a direct affective response to the stimulus (i.e., hardwired emotions, conditioned responses, mimicry), *Chris Janiszewski is associate professor of marketing at the Uni- versity of Florida, Gainesville, FL 3261 1. I would like to thank Rich- ard Lutz, John Lynch, Alan Sawyer, and Luk Warlop for their helpful comments on earlier versions of this article, John Lynch and Luk Warlop for their suggestions for the analysis of experiment 2 data, Sharon Brackett for her work on experimental materials, and my laboratory staff for their assistance in data collection and analysis. Il 'lA C 1993 by JOURNAL OF CONSUMER RESEARCH, Inc. * Vol. 20 * December 1993 All rights reserved. 0093-5301/94/2003-0003$2.00 This content downloaded on Wed, 26 Dec 2012 13:00:08 PMAll use subject to PREATTENTIVE MERE EXPOSURE EFFECTS 377 (2) an affective response to the match between the stim- ulus and a stored cognitive representation or schema (i.e., the feeling of familiarity that results when pro- cessing a previously seen stimulus), or (3) an affective response to the meaning of the stimulus (i.e., the affec- tive tone of associations in memory). Of these three potential sources of affective responses, we will use the second in our account of affective re- sponses that result from incidental exposure to brand names and product packages. When brand names and product packages are part of the secondary information within the environment, or simply information that is not being pursued, they are being processed in a pre- attentive manner. In these situations, the best possible outcome of repeated exposure to a product package or brand name may be that the individual forms a sub- conscious mental representation or memory that can influence the subsequent perception, evaluation, and interpretation of the stimulus (Jacoby, Lindsay, and Toth 1992; Janiszewski 1990a; Mandler et al. 1987). For example, a subconsciously formed mental repre- sentation of a brand name makes subsequent perception of the name easier (e.g., perceptual fluency), providing a feeling of familiarity with the name. In the absence of recognition that could explain the familiarity, the viewer will interpret the familiarity as a positive feeling toward the brand (Bornstein and D'Agostino 1992; Janiszewski 1990a; Mandler et al. 1987). Researchers have identified two general strategies for enhancing affective responses to incidentally viewed brand names and product packages. The first strategy is to increase the opportunity for intentional processing of the stimulus, thus increasing the availability of at- tentive resources. A common method of encouraging intentional processing is through repetition (Obermiller 1985; Sawyer 1981). Repeated exposure to a stimulus provides an individual with repeated opportunities to attend to the stimulus, thus increasing the opportunity to establish a memory trace that can result in the fa- miliarity that biases a subsequent conscious appraisal. Bornstein (1989) points out that repeated exposure is an effective method of encouraging a positive affective response toward a stimulus, provided the exposures do not encourage recognition of the stimulus. A person's awareness of previous exposures to the stimulus can provide an explanation of his/her familiarity with the stimulus and discourage the person from interpreting the familiarity as liking. A second method of increasing the availability of at- tentive resources is to increase the duration of the initial stimulus exposures (see, e.g., Hamid 1973; Seamon, Marsh, and Brody 1984). As the length of an exposure increases, opportunities for attending to the stimulus are enhanced, as are the resources that can be devoted to creating a mental representation of the stimulus. Like repetition, duration of exposure is an effective agent for promoting affective responses, provided that increased duration does not result in recognition of the stimulus. Recognition of a stimulus can be used to explain fa- miliarity with a stimulus and lessen the bias of this cue on evaluation. When exposure durations become too long, repeated exposure to a stimulus will result in ex- tensive processing of the stimulus and will eventually decrease subsequent liking ratings (e.g., wearout). A second strategy for enhancing exposure effects in- volves increasing the availability of preattentive re- sources (see, e.g., Anand et al. 1988; Janiszewski 1988, 1990a). Although people engage in a preattentive feature analysis of their environment automatically, the ability to establish a memory trace of this analysis may be re- source dependent. Hemispheric resource theory predicts that the availability of resources to form a memory trace of the outputs of a feature analysis may be sensitive to activation created by the feature analysis itself (Jani- szewski 1990a). Thus, it is postulated that the auto- matic, preattentive analysis of the environment can en- courage the allocation of resources for forming memories of the outputs of this analysis. Hemispheric resource theory assumes that there are at least two unique resource pools that can support the formation of a mental representation based on the preattentive feature analysis (Allen 1983). One resource pool is better suited for holistic, inferential processing and is associated with the right hemisphere, while a sec- ond resource pool is better suited for analytic, sequen- tial, repetitive processing and is associated with the left hemisphere (Alwitt 1981; Dimond 1972; Hansen 1981). As might be expected, these two pools vary in their abil- ity to store the representation resulting from a feature analysis, the right hemisphere better able to form a rep- resentation of a pictorial stimulus, the left hemisphere better able to form a representation of a verbal stimulus (Ellis and Miller 1981; Janiszewski 1988, 1990a; Mos- covitch and Klein 1980).1 The differential suitability of the left and right hemi- sphere for forming a mental representation of a stimulus during preattentive processing has a direct implication for our understanding of one potential benefit of the Incidental processing of ads. The visual field in which a stimulus appears can constrain preattentive processing to a hemisphere and, as a consequence, influence the formation of a mental representation of the stimulus (Young 1982). For example, Janiszewski (1988) found that pictorial ads were more liked when placed in the left, as opposed to the right, visual field because this layout encouraged the viewer to use the holistic pro- cessing resources of the right hemisphere to initially 'The preattentive processing system can be thought of as a parallel processing system responsible for parsing the environment into meaningful units, selecting the appropriate subset of units to receive further attention, and keeping a record of both used and unused units for further reference. It is the relative strength of these records that serve as inputs into feelings of familiarity. Thus, it may be the ability of a hemisphere to create and record the same unit during preattentive processing that provides for a hemispheric resource influence of sub- sequent affective responses. This content downloaded on Wed, 26 Dec 2012 13:00:08 PMAll use subject to 378 JOURNAL OF CONSUMER RESEARCH perceive the ad. The right hemisphere is more capable of extracting feature information from a pictorial ad; hence, viewers should have formed an accessible mental representation during exposure. During the subsequent evaluation of the pictorial ad, this representation al- lowed for a more fluent perception Qf the stimulus, making it seem more familiar and encouraging a more positive evaluation of the ad. A similar pattern of in- fluence was obtained for verbal stimuli placed in the right visual field. There is also evidence that the quantity of resources of a particular composition can influence an affective response to a preattentively processed stimulus (Jani- szewski 1990a). Janiszewski demonstrated that simple, unfamiliar brand names were more liked when sent to the left hemisphere while the viewer was engaged in the attentive processing of a pictorial stimulus and when sent to the right hemisphere while the viewer was en- gaged in the attentive processing of a verbal stimulus. Janiszewski hypothesized that matching activation, the increased availability of resources in one hemisphere because of an increased processing load in the opposing hemisphere, was responsible for the effect (see also Allen 1983; Friedman and Polson 1981). This increased readiness of the less activated hemisphere to form a representation of the brand name was thought to result from its "expectations" of processing responsibilities in an interactive, two-pool processing system (Friedman and Polson 198 1). Thus, as the preattentive processing associated with the attended task activated one hemi- sphere to a greater extent, resource availability in the opposing hemisphere increased in anticipation of co- operative processing and was allocated to creating a mental representation of the brand name. A more ac- cessible mental representation of the brand name was formed, and the fluency experienced during subsequent perception of the brand name encouraged the individual to evaluate it more favorably. The matching-activation hypothesis is interesting because it suggests that there may be a way of manip- ulating the amount of energy consumers devote to an incidental task-in this case creating a memory trace of the outputs of a preattentive feature analysis-when there is no goal or objective associated with the pro- cessing of the information. Yet, such evidence does not yet exist. In Janiszewski's (1 990a) investigation, matching activation could have resulted from the re- source availability associated with the intentional pro- cessing of the attended task or from the preattentive feature analysis of information that was a necessary precursor to the attentive processing. There are two strategies one could use to demonstrate that the incidental processing of brand names and product packages that results in affective responses to- ward the stimuli can be attributed to preattentive-re- source availability. The first strategy involves creating manipulations that rule out the possibility that the re- sidual activation from intentional processing could in- fluence the preattentive-resource availability. In short, it is necessary to show that the preattentive resources available for the processing of incidental information are sensitive to the manipulation of other incidental information. The second strategy involves demonstrat- ing a dissociation between attentive- and preattentive- resource pools. For example, it could be shown that the attentive-resource pool is a limited-capacity pool that exhibits interference in multitask situations, whereas the preattentive-resource pools are expanding-capacity pools that exhibit facilitation in similar multitask sit- uations. Experiments 1 and 2 provide these types of evidence. EXPERIMENT 1 Janiszewski (1 990a) manipulated the hemisphere re- sponsible for performing the initial analysis of a brand name by placing the name to the left or the right of a text passage or pictorial display. The brand name was more preferred when it was placed to the left of an at- tended verbal task or to the right of an attended pictorial task. Although this procedure did not encourage sub- jects to view the brand name, it did require subjects to attend to a verbal or pictorial task. Thus, it is possible that the intentional processing associated with the focal task was instrumental in making resources available to' form an implicit memory representation of the nonfocal brand name. To rule out this possibility, one could ma- nipulate the resource requirements of the stimulus in the visual field opposite the one containing the brand name instead of manipulating the resource require- ments of the attended task. In this way, any differences in evaluation of the brand name would be the result of processing associated with a secondary, inconsequential stimulus. Method Subjects. Subjects were 56 male and female under- graduate liberal arts students from an introductory speech course who were given extra credit for their par- ticipation in the experiment. Subjects were run in three groups ranging in size from 11 to 31 during a two-week period. Materials. The instrument was a 10-page student newspaper consisting of articles, ads, and experimental instructions similar to that used in Janiszewski (1990a), experiment 2 (see Fig. 1). In all conditions, the target task was a pictorial search task that required subjects to find hidden objects in a line drawing of a Thanks- giving celebration, a task presumed to require relatively more right-hemisphere resources. The target brand name was the letter triplet "NEO" and was written in 1-inch uppercase roman letters. It was placed either to the right or to the left of the attended target task, de- pending on the condition. The secondary stimuli oc- cupied a space on the opposite side of the attended task. This content downloaded on Wed, 26 Dec 2012 13:00:08 PMAll use subject to PREATTENTIVE MERE EXPOSURE EFFECTS 379 The pictorial stimulus was an ad for Air France Airlines. It pictured a couple window-shopping in an Italian square. The verbal stimulus was the phrase "COKE IS IT," one word per line. Again, letters were 1-inch up- percase roman. The nonsense brand name and pictorial and verbal stimuli were all vertically and horizontally centered in a 7?/2-inch-high bordered space, to make them appear to be ads. The stimuli occupied a visual field ranging from 3.5? to 270 from the possible points of focus on the attended task, assuming a reading dis- tance of 18 inches. Procedure. After being seated, subjects were given a cover story to reduce anticipated suspicions about the experimental procedure and dependent measures. Sub- jects were told that the experiment investigated lexical priming effects. The experimenter explained that prim- ing occurs when an initial task biases one's performance on a subsequent task. To investigate these biases, sub- jects would be asked to perform three tasks. Subjects were told that the first task involved the reading of an article that contained a set of key words, the second task was an activity that either enhanced or degraded the influence of the words, and the third task measured the effect. Subjects were also told the effect was facili- tated by a leisurely approach to processing, hence the embedding of the tasks in the newspapers. Unbeknownst to the subjects, the tasks were identical for all conditions. All newspapers had a human interest story about a student hostel placed on page 2, the back of the cover page. The article was marked with an orange dot and subjects were allowed to read it at their own pace. Page 5 of the newspaper was blank except for a set of instructions that explained the second, pictorial search task to the subjects. All subjects began the pic- torial search task at the same time and were given two and one-half minutes to complete the task. This time constraint was insufficient for the completion of the picture search task and resulted in an equivalent length of exposure to the brand name across conditions. After being told to stop task 2, subjects closed their newspapers and were given a dependent-measure booklet. The booklet contained a set of six stimuli. In accordance with the cover story, the instructions ex- plained that a lexical priming effect was most prevalent when people responded to stimuli using their first impressions, and subjects were encouraged to respond in this fashion. Subjects were then asked to evaluate each of the stimuli using five nine-point, bipolar adjec- tive items: unappealing/appealing, unattractive/attrac- tive, bad/good, unlikable/likable, unpleasant/pleasant. The target stimulus "NEO" was in the fourth position, the Air France ad in the fifth position, and the Coke ad was in the sixth position. After evaluating the stimuli, subjects indicated their gender and handedness and were asked to review the stimuli and indicate whether they remembered seeing the stimulus in the newspaper (yes/ no) and whether they had seen the stimulus prior to that day (yes/no). Results The five items measuring the subjects' evaluative re- sponses to the brand name were summed for each sub- ject and used as an indicator of a subject's evaluation of the brand name. A Bartlett-box homogeneity test comparing variances across conditions was not signif- icant. A plot of the residuals supported the assumption of the independence of errors. It was predicted that the evaluation of the target brand name would be greater when it was placed in th'e right visual field and the stimulus in the left visual field was pictorial or when it was placed in the left visual field and the stimulus in the right visual field was verbal. As can be seen in Figure 1, the means are consistent with the predictions. An ANOVA performed on the evaluations of the brand name revealed that there was a significant inter- action between placement and type of opposing stim- ulus (F(1,52) = 22.0, p .05, co2 = .27). When the brand name was placed in the right visual field, it was evalu- ated more positively when the stimulus in the left visual field was a pictorial, as opposed to a verbal, stimulus (F(1,52) = 6.4, p .05, a,2 = .07). When the brand name was placed in the left visual field, it was evaluated more positively when the stimulus in the right visual field was a verbal, as opposed to a pictorial, stimulus (F(1,52) = 16.8, p .05, a,2 = .20). When the brand name appeared with a pictorial stimulus on the opposite side of the attended task, it was more preferred when initially sent to the left, as opposed to the right, hemi- sphere (F(1,52) = 17.2, p .05, c2 = .2 1). When the brand name appeared with a verbal stimulus on the opposite side of the attended task, it was more preferred when initially sent to the right, as opposed to the left, hemisphere (F(1,52) = 6.4, p .05, co2 = .07). There were no main effects of handedness or gender, and these factors did not interact with the independent variables (all F's 1.0). Recognition measures were used to determine whether or not differential levels of attention to the pe- ripherally placed stimuli might have been responsible for the differences in affective response. Evaluation of the target brand name blocked by the subject's report of recognition is shown in Table 1. When recognition was treated as an independent variable, evaluation of the brand name did not differ by level of recognition (F(1,48) = 0.0). In addition, a test was run to determine whether treatment variables interacted with recognition to influence subsequent af- fective response. Recognition that the critical stimulus had been in the newspaper did not interact with place- ment, opposing stimulus type, or a combination of the two (F(1,48) = 0.05, F(1,48) = 0.34, F(1,48) = 0.51). Nonparametric indexes of sensitivity (equivalent to d' This content downloaded on Wed, 26 Dec 2012 13:00:08 PMAll use subject to 380 JOURNAL OF CONSUMER RESEARCH FIGURE 1 EVALUATION OF TARGET SYLLABLE IN EXPERIMENT 1 Oa Shipon m POW , _LAW. Mfd t u 14 aay c Up nifty Yew Qf 3 ~~~~~~~~ ~~~~NEO NEO COKE VatmI 32- 13 : 8 kam".. I ~~~~~~~~~~~~~~~~~~~~~~~IS T .. _ 29.3 (&4) 29.4 (89) Verbal Stimulus in 26 Opposing Visual Field 22.1 (6.9)\ 20Thftw YLaw.mll Pictorial Stimulus in \Opposing Visual Field wwi V-'-I 1 COE 78 (8.3) COKEI I NE 0o ain La. iM a m y LA" Upht~Yy 14- NEO LEFT HEMISPHERE RIGHT HEMISPHERE Target Syllable Initially Sent To in signal-detection analysis) and bias (equivalent to c in signal-detection analysis) were computed for each of the four groups (Grier 1971). As shown in Table 1, sen- sitivity (A') is similar across conditions, but the positive B" coefficients suggest that there was a bias (B") to say no to recognition questions in all conditions. The small number of recognition trials for each subject (n = 6) required that sensitivity and bias coefficients be com- puted with group data, so tests of significance could not be performed (see Macmillan and Kaplan 1985). If the processing of the peripherally placed pictorial and verbal stimuli were indirectly responsible for the differences 'in affective response to the peripherally placed target brand name, then there should also be differences in the affective responses to the pictorial and verbal stimuli. The evaluations of the stimuli, blocked by recognition, are shown in Table 2. Comparison of opposing stimuli means revealed a significant placement effect on the evaluation of the verbal stimulus (F(l,25) = 3.22, p .1O, w 2 = .05) but not on the pictorial stimulus (F( 1,25) = 1. 1). The lack of a placement effect on pictorial stimulus evaluations may be a. consequence of using a pictorial attended task. Evaluations of both opposing stimuli were not influ- enced by handedness or gender (verbal F's 1.0; pic- torial F's 2.0). When recognition was used as an in- dependent variable, there was no effect of recognition on evaluation of the verbal stimulus (F( 1,23) = 0. 15). Recognition did not interact with placement (F(1,23) = 0.20). Discussion The results of experiment 1 provide evidence that affective responses to a brand name can be influenced by preattentive processes active during incidental ex- posure. The affective response to an unfamiliar brand name presented in one visual field was influenced by This content downloaded on Wed, 26 Dec 2012 13:00:08 PMAll use subject to PREATTENTIVE MERE EXPOSURE EFFECTS 381 TABLE 1 EVALUATION OF TARGET BRAND NAME BY REPORTED LEVEL OF RECOGNITION IN EXPERIMENT 1 Initial destination of target brand name Left hemisphere Right hemisphere Opposing stimulus n M SD n M SD Pictorial stimulus 14 29.3 5.4 15 17.8 8.3 No recognition 11 30.0 5.0 7 17.3 10.3 Recognition 3a 26.7 7.1 8b 18.3 6.7 Verbal stimulus 14 22.1 6.9 13 29.4 8.9 No recognition 6 20.7 6.6 10 29.3 8.9 Recognition 8c 23.3 7.2 3d 29.7 10.8 a False alarm rate = 6%; sensitivity A' = .86; bias B" = .63. b False alarm rate = 2%; sensitivity A' = .65; bias B" = .28. c False alarm rate = 6%; sensitivity A' = .70; bias B' = .45. d False alarm rate = 10%; sensitivity A'= .83; bias B" = .85. the preattentive processing of a stimulus presented in the opposing visual field. The brand name was evaluated more favorably when it was sent to the right hemisphere while a verbal stimulus was being instantiated in the left hemisphere. The brand name was also evaluated more positively when it was sent to the left hemisphere while a pictorial stimulus was being instantiated in the right hemisphere. The verbal stimulus placed opposite the target brand name was evaluated more positively when instantiated in the left hemisphere. Evaluation of the pictorial stimulus was not sensitive to placement, although it did have to compete for holistic processing resources with the attended pictorial task, lessening the likelihood an implicit memory trace would form. The results of experiment 1 provide indirect evidence of the influence of preattentive processes in generating the affective responses that result from incidental ex- posure. In experiment 1, the verbal and pictorial stimuli responsible for increasing the quantity of preattentive resources available for forming a memory trace of the brand name were placed in peripheral visual fields; thus, they received a limited amount of directed attention. Recognition of the incidental material did not allow subjects to explain their familiarity with the material and was unrelated to the evaluation of the target brand name, pictorial, and verbal ads. More important, the crossover interaction created by the experimental ma- nipulation makes it difficult to advance a hypothesis that attributes the results to directed attention. The af- fective responses seem to be attributable to the hemi- spheric resources associated with preattentive process- ing of incidental information. The results of experiment 1 also provide a replication of Janiszewski's (1 990a) matching-activation demon- stration. The primary difference is that Janiszewski's matching-activation -demonstration depended on the consumer's processing of an attended stimulus, whereas experiment 1 manipulated a secondary stimulus located TABLE 2 EVALUATION OF OPPOSING STIMULI BY REPORTED LEVEL OF RECOGNITION IN EXPERIMENT 1 Initial destination of opposing stimulus Left hemisphere Right hemisphere Opposing stimulus n M SD n M SD Pictorial stimulus 15 28.9 10.3 14 32.6 8.3 No recognition 11 30.9 8.2 12 32.3 8.3 Recognition 4 23.5 14.6 2 34.0 11.3 Verbal stimulus 13 28.3 6.9 14 23.9 5.8 No recognition 11 27.9 7.0 8 24.0 6.5 Recognition 2 30.5 7.8 6 23.8 5.3 in a peripheral visual field in order to vary the hemi- spheric resources available to form an implicit memory trace of the target brand. Thus, it seems that goal-di- rected attentive processes are not necessary for mere exposure-driven affective responses. More definitive evidence that hemispheric resource availability can influence affective responses would come from a demonstration that attentive and preat- tentive exposure effects are independent. Such an opportunity is provided by the assumptions about at- tentive- and preattentive-resource pools. Attentive- resource theory explanations of task performance have emphasized the zero-sum nature of a single resource pool: increases in the resource requirements of a target task will decrease the resources available for an inci- dental task (Navon 1984; Navon and Gopher 1979).:In contrast, the hemispheric resource theory account of the matching activation suggests that increases in the resource requirements of a target task may increase the availability of resources for an incidental task. It is this expected difference in the operation of attentive and preattentive processes that forms the foundation for a demonstration of their independence. EXPERIMENT 2 Prior research has identified two general strategies for enhancing affective responses to incidentally viewed brand names and product packages: increasing atten- tive-resource availability and increasing preattentive- resource availability. Attentive-resource availability is directly related to goal-directed resource allocation, whereas preattentive-resource availability is related to the noncontrolled resource activation associated with a stimulus environment. Differences in assumptions about how these resource systems operate suggests con- flicting strategies for enhancing affective responses to incidentally viewed stimuli. The attentive-resource strategy advocates reducing the attentive resources al- located to target information, thus increasing the like- lihood that resources will be allocated to the processing This content downloaded on Wed, 26 Dec 2012 13:00:08 PMAll use subject to 382 JOURNAL OF CONSUMER RESEARCH of incidental information (e.g., target brand name). In contrast, the preattentive-resource strategy advocates increasing the preattentive processing requirements of target information, thus increasing the hemispheric re- sources available for processing the target brand name in the opposing hemisphere. These qualitatively differ- ent predictions provide the basis for an experiment de- signed to illustrate that preattentively generated affec- tive responses are independent of attentive operations. Method Design and Materials. The experiment used com- plete ads for six brands that consisted of a headline and a pictorial display on the top half of the ad, verbal claims in the lower right-hand corner of the ad, and a target brand name consisting of a brand name and a simple line drawing in the lower left-hand corner of the ad. The layout of these ads allowed for the target brand name to be instantiated in the right hemisphere while the subject was attending to the verbal text in the lower right-hand corner of the ad. The experiment consisted of a factorial manipulation of the processing load of the verbal claim information (low, moderate, and high) and the availability of atten- tive resources for the processing of the target brand logo (limited, extended) with a stimulus set replication. Pro- cessing load was manipulated by varying the length of verbal claims, with the low-load condition consisting of an average of 3.5 meaningful words, the moderate- load condition an average of 7.5 words, and the high-load condition an average of 12.5 words. The attentive-resource factor was a between-subjects vari- able consisting of a processing time and instruction manipulation. The subject was given 3.5 (6.5) seconds to skim (read) the ad in the limited (extended) attentive- resource conditions. Pretesting had shown that the lim- ited attentive-resource manipulation would result in a limited, but equivalent, level of attention to the brand logo across the processing-load conditions. In contrast, the extended attentive-resource manipulation would encourage increased attention to the brand logo as the processing load of the verbal claims decreased. The stimulus replication used unknown brands from six product categories: aspirin-Dots; office supply-Blos- som; landscaping-Dr. Oak; insurance-Covington; limousine service-Larry's; travel-Merlin. Ads were created with commercial computer soft- ware. Twelve ads were created for each of the six brands-there were four ads for each of the three in- formation-load conditions. Each low-information-load ad for a brand had a unique headline, pictorial display, and verbal claim-only the brand logo was constant. Low-information-load ads were transformed into mod- erate- or high-information-load ads by adding text to the basic verbal claim. For example, the verbal claims for the three levels of one of the Dot's headache remedy ads were "Headache Relief," "Headache Relief after a Day of Crisis," and "Headache Relief after a Day of Crisis. We'll take care of your problem." For all ads, the additional claim information was clarifying or re- dundant, so as not to make the brand more appealing. Ads were arranged into three pods. Each pod included the low-information-load ads for two of the brands, the moderate-information-load ads for two of the brands, and the high-information-load ads for two of the brands. The four ads for the six brands were rotated. Each pod of 24 ads was preceded by three filler ads for three un- related brands. Subjects. Subjects were 168 male and female stu- dents recruited from an introductory marketing course. Subjects were given extra credit for their participation in the experiment and were processed one at a time over a five-week period. The experiment was part of a much larger data collection effort. Procedure. In order to gain insight into the influ- ence of differing levels of attention to the text claims and the brand logo on affective responses to the logo, subjects' eye movements were monitored while they viewed the ads. Upon arriving at the laboratory, the subject was seated in front of a 20-inch Multisync 5d computer monitor. The monitor was located approxi- mately 34 inches from the eyes of the subject with the eyes level with a point somewhere near the bottom of the display; the level of the eyes varied with the subject's height. An eye-tracking camera was located directly be- low the catalog display area and was also 34 inches from the subject's eyes. The distance and angle from the eye to the catalog and the eye to the camera were con- strained by a factory calibration of the eye-tracking system. The eye-tracking system was explained to the subject, and the system was calibrated to the subject's eyes, a process that took approximately five minutes. Then the subject was told that s/he would see a series of ads on the computer monitor and that there would be multiple ads for some products. Depending on the resource availability condition, subjects were told to either skim or read the ads. The ads were presented with Microsoft's Powerpoint Presentation software. All ad presentations were sepa- rated by a 3.5-second (6.5-second) presentation of a blank fixation slide that allowed the experimenter to make adjustments in the calibration of the eye-tracking system. Afterward, subjects were asked to fill out a de- pendent-measure questionnaire similar to that used in experiment 1. Eye-Movement Measurement and Data Preparation Procedure. Eye movements were measured with an Applied Science Laboratory 1996 Eye View Monitoring System that used a corneal-reflection procedure (see Young and Sheena [1975] for review). A corneal-re- flection device reflected an infrared light off the surface of the eye while measuring the position of the pupil. This content downloaded on Wed, 26 Dec 2012 13:00:08 PMAll use subject to PREATTENTIVE MERE EXPOSURE EFFECTS 383 During calibration, the distance and the angle between the infrared reflection and the center of the pupil were measured while subjects looked at predefined areas. Each of these areas was also identified in visual space with x,y location coordinates. Thus, the distance be- tween a subject's corneal reflection and pupil center could be directly related to x,y locations in a scene. During recording, the eye-tracking system recorded the distance between the center of the pupil and the corneal reflection every sixtieth of a second, simulta- neously converting this distance into an x,y coordinate in the viewing area. These locations were reduced to a series of fixations through an algorithm that identified sequences of data points that represent less than .5? of movement within a 1 17-millisecond time period, less than 10 of movement in 234 milliseconds, or less than 1.50 of movement in 351 milliseconds. This manufac- turer-defined algorithm is commonly used in fixation research (Stark and Ellis 1981). To determine where a subject looked in a given scene, corners of rectangular areas within the scene were de- fined with a set of four x,y coordinates. For this exper- iment, the areas of interest were the top of the ad, the verbal claims, and the brand logo. Files containing the fixations for a subject were combined with files con- taining the coordinates of rectangles that defined these areas of interest in the ads. After these files had been combined, statistics about how long a subject looked at a particular area were computed, then averaged across subjects. Predictions It was predicted that increasing the processing load of the verbal claims would increase matching activation, thus increasing the hemispheric resources available for forming a mental representation of the target brand logo in the opposing hemisphere. In the limited attentive- resource condition, it was expected that attention to the brand logo would be equivalent across the three processing-load conditions; thus, only increases in the preattentive resources associated with matching acti- vation could result in a more positive evaluation of brand logo. In the extended attentive-resource condi- tions, it was expected that the attention to the brand logo would increase as the processing load of the verbal claims decreased. This attentive processing exposure ef- fect should overwhelm the matching-activation effect and reverse the trend of brand logo evaluations observed in the limited attentive-resource conditions. Assump- tions about the allocation of directed attention were tested first. Manipulation Checks The attentive-resource manipulation was designed to limit the level of attention to the target brand name in the limited attention conditions but to allow for variable attention to the text and target brand name in the ex- tended attention conditions. The data were analyzed as a counterbalanced design; within-subject factors were processing load nested within brand pairs and brands nested within brand pairs; the between-subjects factor was attentive-resource availability (see Kenny and Smith [1980] for discussion). The tests for an interaction of the processing-load and attentive-resource manipu- lations on the duration of attention to the text (F(2, 321) = 124.85, p .05) and target brand logo (F(2, 321) = 20.24, p .05) were significant (see Table 3 for uni- variate stimulus means).2 A linear-trend test of the in- fluence of the processing load in the limited attention conditions on the duration of attention to the target brand logo was not significant (F(1,72) = 1.04, p &#x 000; .05). Linear-trend tests of the influence of the processing load in the extended attention conditions on the duration of attention to the text (F(1,90) = 126.33, p .05) and target brand logo (F(1,90) = 26.75, p .05) were sig- nificant, and, as expected, the trends were in opposite directions. Thus, the manipulations influenced atten- tion as expected. Results The pattern of means was as predicted (see Fig. 2). A test for an interaction of the processing-load and at- tentive-resource manipulations on the evaluation of the target brand logo was significant (F(2, 321) = 15.03, p .05, w2 = .08). In the limited attentive-resource con- dition, evaluation of the brand logo became more pos- itive as the processing load of the verbal claims increased (F(2,72) = 3.45, p .05, w2 = .03). In the extended attentive-resource condition, evaluation of the brand logo became less positive as the processing load of the verbal claims increased (F(2,90) = 5.66, p .05, w2 = ).3 Correlations of the evaluation of the brand logo with attention to the verbal claim text and logo were com- puted for each brand in each attentive-resource con- dition (Table 3). It was predicted that there would be no correlation between the amount of attention to the logo or the text and evaluation of the logo in the limited attentive-resource condition. Nine of the 12 correlations were nonsignificant. It was also predicted that there 2Table 3 provides the summed evaluation scores for each brand logo and the average amount of seconds spent looking at the lower right-hand corner text and the lower left-hand corner brand logo in each ad. The limited attention condition is the skim instruction and the extended attention condition is the read instruction. "Low," "medium", and "high" refer to the length of the verbal claims. Brands 1- 6 are the brand replications. The correlations are of individual subject's brand evaluation scores with time of attention to the brand logo and the verbal text. 3Treating the brands manipulation as a random factor leads to the same conclusions. The only deviation is that the manipulation check test for the interactive influence of processing load and attentive- resource availability on attention to the target brand logo becomes insignificant. This content downloaded on Wed, 26 Dec 2012 13:00:08 PMAll use subject to 384 JOURNAL OF CONSUMER RESEARCH TABLE 3 EVALUATION OF BRAND NAMES WITH ACCOMPANYING LEVELS OF ATTENTION-EXPERIMENT 2 Verbal-claim-processing load Low Medium High Correlation Time Time Time Time Evaluation (sec.) Evaluation (sec.) Evaluation (sec.) (sec.) Limited attentive resources: Brand 1 33.4 2.7 34.8 2.1 35.7 2.3 -.10 Text 3.4 3.0 4.7 .11 Brand 2 32.0 2.0 33.5 1.8 34.6 1.8 -.09 Text 4.0 3.9 5.1 ,.27* Brand 3 25.0 2.3 27.3 2.2 28.7 2.2 -.01 Text 3.1 3.9 3.1 .07 Brand 4 24.5 2.7 26.7 2.4 28.1 1.8 .07 Text 2.7 4.2 3.3 .01 Brand 5 24.3 2.4 26.3 2.4 27.9 2.3 .20* Text 2.4 4.0 4.2 .08 Brand 6 17.1 2.2 18.6 2.2 20.0 2.1 -.05 Text 2.0 4.5 5.0 .20* Brand average 26.1 2.4 27.9 2.2 29.2 2.1 Extended attentive resources: Brand 1 37.2 4.7 35.0 3.3 33.7 3.4 .20* Text 4.9 8.1 10.4 -.20* Brand 2 36.2 3.8 35.1 2.5 34.0 1.8 .20* Text 7.2 11.8 12.1 -.11 Brand 3 30.3 3.9 28.6 3.4 26.9 2.5 -.04 Text 6.3 9.7 12.1 -.07 Brand 4 27.9 4.2 27.3 4.0 25.2 2.8 -.01 Text 4.3 8.4 10.3 .00 Brand 5 29.4 3.7 28.0 3.5 26.9 3.4 -.02 Text 5.8 8.4 11.9 -.16* Brand 6 22.1 4.2 18.9 3.5 17.8 3.2 .12 Text 4.9 8.0 10.6 -.32* Brand average 30.5 4.1 28.8 3.4 27.4 2.9 Significant at p .05, one-tail test. would be a positive correlation between evaluation of the brand name and attention to the brand logo but a negative correlation between evaluation of the brand name and attention to the verbal text claims in.the ex- tended attentive-resource condition. Five of the 12 cor- relations reached significance and were in the predicted direction. Discussion The results of experiment 2 suggest that the affective responses that result from incidental exposure to a brand name may be attributable to attentive processes associated with a controlled, fixed resource pool but also may be attributable to preattentive processes as- sociated with matching resource pools. When subjects skimmed the verbal claims in the ads, increases in the processing load of the attended claims resulted in in- creases in the evaluation of the brand names located to the left of the text. The increased evaluation of the brand logos can be attributed to the matching hemispheric activation resulting from the preattentive processing of the verbal claims. Attention to the brand logos remained constant across processing-load manipulations, and correlations between evaluation of the brand logos and attention to the brand logos were nonsignificant in five of six cases, which suggests that the evaluation of the brand logos was not a consequence of attention to the logos. When subjects were provided with more attentive ca- pacity for the processing of the ads, decreases in the resource requirements of the verbal target tasks resulted in increases in the availability of attentive resources for processing the brand logos (see Table 3). As a conse- quence, lower verbal-claim-processing loads were as- sociated with more positive evaluation of the brand lo- gos (i.e., an attentive-resource exposure effect). Correlations between evaluation of the brand logos and attention to the brand logos were partially supportive of a relationship between the duration of attention and evaluation. The apparent independence of preattentive and at- tentive contributions to affective responses resulting from exposure to a brand name reemphasizes the rel- This content downloaded on Wed, 26 Dec 2012 13:00:08 PMAll use subject to PREATTENTIVE MERE EXPOSURE EFFECTS 385 FIGURE 2 AVERAGE EVALUATION OF THE TARGET STIMULI AND AVERAGE ATTENTION TO THE TARGET STIMULI IN EXPERIMENT 2 Evaluation of the Target Stimuli 31.0 30.0 Limited 29.0 Resources 28.0 Extended Evaluation 27.0 Resources 26.0 25.0 24.0 23.0 LOW MEDIUM HIGH Information Load Attention to the Target Stimuli 4.5 4.0 ? 3.5- 3.0 Extended Resources Attention 2.5 Eoc in Seconds 2.0 Limited Resources 1.5 1.0 0.5 0.0 LOW MEDIUM HIGH Information Load evance of the two general strategies for enhancing preattentively driven exposure effects: resource com- patibility and increased resource availability via matching activation. Unfortunately, our understanding of the source of the matching-activation effect is quite limited. One possibility is that matching activation is a consequence of the feature analysis supporting atten- tion-control mechanisms. Shifts in attention depend on a gross feature analysis of parafoveal information (i.e., outside of the area to which the eyes are directed) that results in the parsing and categorization of the data as well as an assessment of its importance (Levy-Schoen 1981; Wolfe and Cave 1990). To the extent one hemi- sphere is more efficient at this feature analysis for certain types of stimulus environments (i.e., the right hemi- sphere is better at performing a feature analysis of a pictorial environment), the opposing hemisphere will provide impoverished information on items of potential importance that are located in the opposing visual field. Thus, from an attention-control standpoint, a noncon- trolled processing system would need to allocate addi- tional resources to the less activated hemisphere to en- sure it could contribute equally to the preattentive analysis of the environment. Matching activation may also be a consequence of the semantic analysis that results from interhemispheric attempts to perceive the environment. The information- processing system can engage in a semantic analysis of the environment using both controlled and noncon- trolled resources (Greenwald 1992). In controlled, at- tentive processing, foveal stimuli are instantiated in each hemisphere and the outputs of intrahemispheric This content downloaded on Wed, 26 Dec 2012 13:00:08 PMAll use subject to 386 JOURNAL OF CONSUMER RESEARCH activities are continually shared so as to update the analysis in the opposing hemisphere (Allen 1983). In anticipation of this cooperative processing, each hemi- sphere is allocated resources in accordance with the more activated of the two hemispheres (Friedman and Polson 1981). Noncontrolled semantic processing may occur in a manner similar to that of controlled processing. In noncontrolled processing, parafoveal stimuli are ini- tially feature analyzed, and, provided the features can be perceived, an elementary semantic analysis will ensue (Greenwald 1992; Marcel 1983). Because this elemen- tary semantic analysis relies on the same mechanisms that perform the analysis of foveal information in con- trolled resource processing, the greater activation of one hemisphere may increase the availability of resources in the opposing hemisphere. If feature analysis resources are responsible for the effects observed in experiments 1 and 2, then increasing the quantity of information available for preattentive processing in the nontarget hemisphere should increase the availability of resources for processing of the brand name in the target hemisphere. These additional re- sources should increase the accessibility of the stored feature representation of the brand name and the fa- vorableness of a response toward it. On the other hand, if semantic analysis resources are responsible for the effects observed in experiments 1 and 2, then increasing the meaningfulness of the information to be preatten- tively processed in the nontarget hemisphere should in- crease the availability of resources for processing of the brand name in the target hemisphere. Again, these ad- ditional resources should increase the accessibility of the stored feature representation of the brand name and the favorableness of a response toward it. Experiment 3 manipulated the quantity and the meaning of para- foveal information to assess whether the feature analysis activation or the semantic analysis activation associated with a stimulus instantiated in the nontarget hemisphere had an impact on the processing of the brand name instantiated in the opposing hemisphere. EXPERIMENT 3 Method Subjects. Subjects were 146 male and female stu- dents enrolled in an introductory marketing course, who were given extra credit for their participation in the experiment. Subjects were processed in groups ranging in size from four to 22 over a two-day period. Design and Materials. The experiment assessed the influence of a secondary stimulus placed in the right visual field on the affective response to a target stimulus ("NEO") placed in the left visual field using a context and target task similar to that used in experiment 1. The design consisted of a factorial manipulation of the quantity of secondary stimulus information (low, mod- erate, and high) and the meaningfulness of secondary stimulus information (meaningless, meaningful) plus a control group. The quantity-of-information manipu- lation involved increasing the number of words in the secondary stimulus. For example, in the semantic meaning conditions, the low-information-quantity sec- ondary stimulus was the word "COKE," the moderate- information-quantity stimulus was the phrase "COKE beats PEPSI," and the high-quantity stimulus was "COKE beats PEPSI. Independent taste tests prove Coke is preferred over Pepsi by 7 out of 10 consumers. Can't beat the feeling" (see Fig. 3). The meaningfulness manipulation was accomplished by rearranging the letters in the meaningful stimuli-to be orthographically regular (i.e., following English spelling rules), pronounceable pseudowords. For ex- ample, the phrase "COKE beats PEPSI" became the phrase "KECO teabs SIPEP." The control group stim- ulus had no secondary stimulus to the right of the target task, the task being located on the right border of the right-hand page, the target stimulus to the left. In all experimental conditions, the stimuli placed in the visual fields occupied a visual space ranging from 40 to 15.50 when the subject viewed the portion of the attended task closest to a peripherally placed stimulus and from 190 to 290 when the subject viewed the por- tion of the attended task farthest from a peripherally placed stimulus. The fonts used to present the Coke and Pepsi brand names were generic, to remove the possibility that the physical features of these well-known words would be responsible for observed effects of the semantic manipulation. The largest type size used was 1 inch, the smallest type size used was 1/4 inch. The stimuli were sized so as to inhibit conscious identifi- cation but to allow preattentive processing while in parafoveal vision (Antis 1973). Procedure. The procedure was similar to that used in experiment 1. The only differences were that subjects performed one picture search task within the newspaper (e.g., the target task) and that a postexperimental in- quiry was used to assess problems with hypothesis guessing. Changes in procedure were made because of time constraints for experimental sessions. The post- experimental inquiry was added because of the unusual nature of the no-semantic-meaning stimuli. Expected Results Three possible results would provide insight into the influence of preattentive processing on affective re- sponses to brand names. First, as the quantity of infor- mation instantiated in the nontarget hemisphere in- creased, the evaluation of a brand name instantiated in the opposing hemisphere could increase. This finding would suggest that a feature analysis in the nontarget hemisphere is increasing the availability of resources in the target hemisphere. Second, as the meaningfulness of information instantiated in the nontarget hemisphere This content downloaded on Wed, 26 Dec 2012 13:00:08 PMAll use subject to PREATTENTIVE MERE EXPOSURE EFFECTS 387 FIGURE 3 EVALUATION OF TARGET STIMULUS IN EXPERIMENT 3 Fw Y_o Fm SIU C "ar c.up Tin" Yws 25- _ _ _ _ ____1____Y Keom NEO -Cok =NEO . oJ~~~~~~~~~~~~~~~~~~~~~~~~~-Y Lr.W. NM_=- k -f 2175 =C:M ~hcJY11p 2S5 NEO ~ ~ ~ ~ ~ ~~~Maningless 20- W 17.79 n~~~~~~~~~~~~~~~~.,. Yr Y , b"." L..pniy Meaningful ' :i' = Z - - 15 -15.76 __ __ _ __ __ _NEO 15.00 \~~~~~~~~~~~~~~~~~~~~~~~~~~~~- NE0- Lam Mea pninglessd H-- Fcwt Y JO 1, )*F " 1 =, Up TWY d ~ ~ ~ ~ ~ ~ NE Pup _ - a _ ---: \~~~~~~~~a-- LOW MEDIUM HIGH Quant'ity of Information increased, the evaluation of a brand name could in- crease. This would suggest that a semantic analysis of information in the nontarget hemisphere is increasing the availability of resources in the target hemisphere. Third, we could observe an interaction between the quantity and meaningfulness of information. This might suggest that feature analysis resources and se- mantic analysis resources are interdependent or are channel outputs to a common operation (Friedman and Polson 1981). An a priori specification of the pattern of this interaction is not possible given our limited un- derstanding of how the feature and semantic analysis of one set of incidental information might influence the accessibility of memories associated with the processing of other incidental information. Results The changes in procedure and the characteristics of the experimental stimuli required that we initially check levels of recognition on the target stimulus. Mean eval- uations of the target brand name, blocked by levels of recognition, are shown in Table 4. Recognition of the target brand name was significantly greater than false recognition of the filler-dependent stimuli (z = 2.78, p .05), ranging from 10 to 33 percent. Levels of rec- ognition did not interact with experimental manipu- lations (recognition by meaningfulness: F(1, 118) = 0.47; recognition by quantity of secondary stimulus information: F(2, 118) = 0.64; recognition by mean- ingfulness by quantity of secondary stimulus infor- mation: F(2, 118) = 2.16), but recognition did have a differential impact on target brand name evaluation by condition (recognition by meaningfulness by quantity of secondary stimulus information: F(2, 1 11) = 4.96, p .05). Because this study was focused on preattentive processing, subjects that remembered having seen the target stimulus in the newspaper (n = 29), or indicated awareness on the postexperimental inquiry (n = 4), were removed from the analysis.4 The revised means are shown in Figure 3. 4It is possible that removing the subjects that recognized the target brand name introduces a selection bias into the data. For example, it may be that recognition scores are caused by extreme responses, instead of the opposite, and that the "recognition" subjects are not This content downloaded on Wed, 26 Dec 2012 13:00:08 PMAll use subject to 388 JOURNAL OF CONSUMER RESEARCH TABLE 4 EVALUATION OF TARGET BRAND NAME BY REPORTED LEVEL OF RECOGNITION IN EXPERIMENT 3 Secondary stimulus Semantically Semantically meaningful meaningless Secondary stimulus n M SD n M SD Low information quantity: Overall 21 20.6 7.9 20 18.3 8.3 No recognition 16 21.8 8.2 14 15.8 6.3 Recognition 5 17.0 4.9 6 24.2 4.8 Moderate information quantity: Overall 20 17.8 9.4 21 17.0 9.8 No recognition 18 18.2 9.4 14 17.7 10.0 Recognition 2 14.0 9.9 7 15.4 9.9 High information quantity: Overall. 20 17.3 8.3 21 18.8 10.4 No recognition 15 15.1 6.3 19 21.5 10.3 Recognition 5 23.8 12.6 2 12.0 8.5 Control group: 19 17.6 9.4 No recognition 17 15.8 7.9 Recognition 2 33.0 8.5 An analysis of the brand name evaluations revealed that there was not a main effect of the quantity-of-in- formation manipulation (F(2,90) = 0.06) nor was there a main effect of the meaningfulness manipulation (F(1,90) = 0.00). However, there was an interaction between the two factors (F(2,90) = 4.04, p .05, w2 = .03). Simple main-effect tests revealed that a mean- ingful, low-information-quantity stimulus was more beneficial to the evaluation of the target brand name than a meaningless, low-information-quantity stimulus (F(1,90) = 3.49, p .07, @2 = .03) and that a mean- ingless, high-information-quantity stimulus was more beneficial to the evaluation of the target brand name than a meaningful, high-information-quantity stimulus (F(1,90) = 4.60, p .05, w2 = .04). Evaluation of the target brand name by the meaningful, low-quantity stimulus group differed significantly from the control group's evaluation of the target stimulus (F(1, 114) = 4.10, p .05). Similarly, evaluation of the target brand name by the meaningless, high-quantity stimulus group differed significantly from the control group's evaluation of the target stimulus (F(1, 114) = 3.49, p .07). A simple main-effect test for the influence of increasing information quantity in the meaningless conditions was significant (F(2,90) - 3.59, p .06, c2 = .02), while the meaningful condition only approached significance (F(2,90) = 2.00, p .15, @2 = .01). Discussion The results of experiment 3 suggest that processes associated with the feature analysis and the semantic analysis of incidental information can each have an in- fluence on the degree to which other incidental infor- mation is preattentively processed. First, increasing the information quantity of the meaningless stimulus in the right visual field increased the evaluation of the tar- get brand name placed in the left visual field. The matching activation hypothesis predicted that increases in the feature analysis demands of the meaningless stimulus in the left hemisphere would result in the ac- tivation of additional resources in the right hemisphere. These additional resources were used to store a record of the feature analysis of the target brand name and increased the perceptual fluency associated with its rep- resentation. The increased perceptual fluency was in- terpreted as a positive feeling for the brand name. Thus, there is support for the feature analysis explanation of matching activation. There is also some support for the semantic analysis explanation of the source of matching activation. In the low-information-quantity conditions, increasing the meaningfulness of information in the right visual field increased the evaluation of the target brand name placed in the left visual field. The matching-activation hy- pothesis predicted that the semantic analysis of the Coke ad in the left hemisphere would provide additional re- sources for the processing of the brand name in the right hemisphere. These additional resources could be used to store a record of a feature analysis of the brand name so as to increase the accessibility of the represen- tation. The increased perceptual fluency was interpreted as a positive feeling for the brand name. The simple main-effect influences of the quantity-of- information and the meaningfulness-of-information manipulations are qualified by an interaction between the two variables. As the quantity of meaningful infor- mation in the right visual field was increased, evaluation of the target brand name in the opposing visual field decreased. This effect was not expected a priori, and it seems to be inconsistent with the information-load ma- nipulation in the skim condition of experiment 2. Yet, the resource-load manipulation in experiment 2 in- volved information that was processed foveally with a combination of preattentive and attentive resources, whereas the information-quantity manipulation in ex- periment 3 involved information that was processed with primarily preattentive resources. If attentive op- erations in some way assist in the selection of outputs from preattentive operations, then the interaction ob- served in experiment 3 may be explainable by Marcel's (1983) theory of perception. different from the other subjects. To test this hypothesis, the difference between each subject's evaluation of the target syllable and their av- erage evaluation of filler ads was computed. The recognition subjects' difference scores were constant across conditions (F 1.0), while "no recognition" subjects' difference scores paralleled the results generated with the target syllable evaluation scores. Thus, the subjects that were retained in the analysis had their evaluation of the target syllable altered, whereas subjects that recognized the target syllable exhibited a response scale bias. This content downloaded on Wed, 26 Dec 2012 13:00:08 PMAll use subject to PREATTENTIVE MERE EXPOSURE EFFECTS 389 Marcel (1983) has shown that access to semantic in- formation precedes access to feature information, even though a feature analysis is a necessary prerequisite to a semantic analysis of the information. This implies that access to the record of the feature analysis that is responsible for perceptual fluency may be sensitive to the amount of concurrent semantic analysis that must be performed, especially because a semantic analysis of text encourages the interhemispheric coordination of multiple pieces of information (Brownell, Potter, and Bihrle 1986; Janiszewski 1990b). If the accessibility to feature analysis and semantic analysis outputs depends on a common mechanism, and this mechanism is re- source limited, then increases in the semantic integra- tion demands of a peripherally placed stimulus could reduce the availability of resources for forming an ac- cessible mental representation of the feature informa- tion, as was the case in experiment 3.5 When there are sufficient resources available for in- tegrating semantic information, as was the case with the controlled attentive processing of the verbal claims in the skim condition of experiment 2, there should also be sufficient resources for forming accessible feature representations of incidentally viewed information. In experiment 2, attention to the verbal claim stimulus was a recognition by the processing system that the se- mantic outputs of the preattentive analysis of the claims were relevant. The attention associated with the selec- tion of the semantic records allowed for the matching activation from the preattentive feature analysis to be used to form records of the preattentive analysis of other information within the environment. In summary, mechanisms that are responsible for the affective responses that result from incidental exposure can be thought of as consisting of preattentive and at- tentive processes. Preattentive processing mechanisms can engage in a feature analysis of single and multiword stimuli, as well as a semantic analysis of single word stimuli. This analysis encourages increased resource availability in the less activated hemisphere and, if a stimulus is present, a mental representation or record of the feature analysis is made. Yet, this record for- mation process is resource demanding. Attempts at in- formation integration will take precedence over infor- mation storage in limited resource situations, as was the case in experiment 3. DISCUSSION An affective response to a stimulus can be the result of direct affect transfer, the ease with which a stored representation of the stimulus is accessed, and valenced cognition associated with the stimulus (Hoffman 1986). The three experiments discussed above suggest that ac- cess to a stored representation of an incidentally viewed brand name is sensitive to the hemispheric resources available for storing the outputs of a preattentive pro- cessing of the name. The availability of hemispheric resources to engage in this preattentive formation of a memory representation can be enhanced by matching activation. Matching activation is an increase in the resources in a target hemisphere because of increased resource demand in an opposing hemisphere. In our experiments, matching activation was shown to be sen- sitive to the demands on feature analyzers, and possibly on low-level semantic analyzers, in the nontarget hemi- sphere. In general, the effect sizes declined as we progressed from experiment 1 to 2 to 3. These declines in effect sizes are directly linked to the strategy for manipulating preattentive-resource availability. In experiment 1, preattentive-resource availability was manipulated with a change in the form of a peripherally placed stimulus (e.g., pictorial vs. verbal). In experiment 2, preattentive- resource availability was manipulated by changing the amount of a particular form of stimulus (e.g., more vs. less verbal). The large effect sizes observed in experi- ment 1 suggest a form- or resource-type manipulation of resource availability is more effective than an amount-of-resource manipulation. Like experiment 2, experiment 3 used an amount-of-resource manipula- tion, but unlike experiment 2, this manipulation was accomplished with peripheral material. There are likely to be more preattentive operations associated with the processing of attended material, hence a preattentive- resource manipulation is likely to have a stronger in- fluence when initiated by focal (experiment 2) as op- posed to peripheral (experiment 3) information. The experiments can be viewed as direct evidence against a long-standing belief that preattentive opera- tions have no influence on learning and that attention is necessary for the formation of sensory traces and mere-exposure-influenced affective responses (Green- wald and Leavitt 1984). Preattentive processes have been assumed to be automatic and outside of the control of marketers wishing to influence consumer percep- tions. It is probably more accurate to say that a subset of preattentive processes is automatic. For example, the responses that result from incidental exposure are the consequence of a variety of subtasks, only some of which are automatic. During incidental exposure there is a feature analysis, memory access, implicit memory for- mation, and perceptual construction. Similar events occur every time a stimulus is reencountered. The ev- idence provided above suggests that the feature analysis of incidental information may be automatic, but the memory traces of these feature analyses is resource de- pendent and the availability of these resources is in- dependent of the operations associated with attentive processing. Hence, consumers can be influenced by in- cidental exposures to brand names and product pack- 'We are not claiming that a semantic integration of information from a parafoveally presented, multiword stimulus is possible, only that it is attempted. This content downloaded on Wed, 26 Dec 2012 13:00:08 PMAll use subject to 390 JOURNAL OF CONSUMER RESEARCH ages that are not motivated by goal-directed, attentive processing. Familiarity created via incidental exposure has the potential to be a multifaceted managerial tool. First, familiarity promotes liking. Brand managers that engage in event sponsorship, licensing, and promotion should maximize consumer exposure to their brand name. At- tention or motivated consideration of the brand name is not important to creating a benefit of exposure. In- cidental exposure is sufficient. Second, familiarity pro- motes attention. Familiarity created via incidental ex- posure makes a stimulus easier to perceive (e.g., perceptual fluency) and increases the likelihood it will be seen on the store shelf (Jacoby et al. 1992). Preat- tentive processes play a critical role in selecting infor- mation for the more intensive analysis associated with attention. Incidental exposure is a potential means of biasing the consumer's selection of information. Third, familiarity promotes memory illusions and faulty in- ferences. For example, perceptual fluency created via exposure has been related to a number of memory il- lusions, including assessments of the truth of a state- ment, the ease of solving a problem, and the popularity of a person (Jacoby et al. 1992). These illusions suggest that brand familiarity generated via incidental exposure may bias subjective judgments about the brand. Thus, it is possible that incidental exposure to a brand could influence consumer perceptions of brand-attribute per- formance in those situations in which consumers are uncertain about actual attribute performance. Further- more, because consumers are unlikely to be aware of prior exposures, it is unlikely that they will be able to negate judgment biases that arise from incidental ex- posure to a brand (Bornstein 1989). The consumer's inability to remember incidental ex- posures to a brand, or to know that these prior exposures are influencing their judgments, highlights the effec- tiveness of this persuasive technique. Incidental expo- sure has the potential to exert a bias on judgment and choice whenever the consumer lacks information. In the event a consumer has subjective advertising infor- mation and discounts it, familiarity becomes one of the few remaining cues for decision making. Hence, when- ever consumers do not actively use recognition or fre- quency information to discount the influence of famil- iarity, familiarity is likely to exert a bias. If preattentive operations are involved in goal-di- rected processes, then there should be situations in which the rules associated with a single-pool, limited- capacity processing system will be supplanted by the rules associated with a multipool, preattentive process- ing system. For example, studies of goal-directed search behavior have shown that search is more efficient when the search environment is less complex or when the search criteria are based on two, as opposed to three, cues (Treisman 1988). Each of these findings is quite consistent with the predictions of a single-pool, limited- capacity system. Yet, Sereno and Kosslyn (1991) have found that a search of peripherally placed stimuli was more efficient when these stimuli were presented si- multaneously in both visual fields and less efficient when they were presented sequentially or in a single visual field, a facilitation effect consistent with preattentive processing and the matching-activation hypothesis. Similarly, Wolfe, Cave, and Franzel (1989) find that visual search through a complex environment is often driven by a preattentive analysis of peripheral stimuli based on simple feature cues. They find that search based on triple conjunction (i.e., three cues) is more efficient than search based on a single conjunction (i.e., two cues). In each case, the simultaneous presentation of more information facilitates information processing, a finding clearly at odds with the assumptions of a single- pool, limited-capacity processing model. The appropriate organization of a multiple stimulus presentation would seem particularly relevant in the construction of television commercials, in which verbal, musical, and visual elements are combined to com- municate a message, especially when the goal is implicit learning of product benefits or consumption behaviors. For example, commercials that engage in a quick pre- sentation of a number of pictorial images (right hemi- sphere prime) prior to written presentation of product benefits (left hemisphere task) or a brand name are pro- viding a context for matching activation, as in Sereno and Kosslyn (1991). This format can encourage atten- tion and facilitate the processing of the product benefits or can simply make a brand name more familiar and encourage favorable affective responses. Commercials that have a musical beat (left hemisphere activating) or melody (right hemisphere activating) accompanying their visual and verbal presentation are also making use of the rules governing hemispheric resource availability (Gordon 1983). Background music with a beat (melody) can encourage increased preattentive processing of pic- torial (verbal) information in the right (left) hemisphere, increasing the likelihood that the person will actively attend to a specific type of information. If the primary advertising goal involves comprehension of the verbal information, then it would be beneficial to encourage attention to the verbal information by including mel- ody, allowing the pictorial information to have its im- pact implicitly. On the other hand, if the goal is implicit learning through pictorial images, then music consisting of a beat would be more effective. In all of these ex- amples, a strategic increase in the amount of infor- mation in a commercial (e.g., increasing the number of scenes, adding music) is predicted to increase the effectiveness of the commercial. This article has provided evidence that preattentive operations can have a meaningful influence on the pro- cessing of a stimulus. Incidental exposure to a brand name or a product package can result in a more favor- able feeling toward the brand. These affective responses are enhanced when an incidentally viewed brand name has access to efficient resource pools or when the com- This content downloaded on Wed, 26 Dec 2012 13:00:08 PMAll use subject to PREATTENTIVE MERE EXPOSURE EFFECTS 391 peting information within the environment encourages a greater availability of resources for the processing of the target stimulus. Hemispheric resource models pro- vide insight into the influence of preattentive processing and suggest that these processes may have the power to influence the comprehension of information and judg- ments about brands. [Received September 1992. Revised March 1993.] REFERENCES Allen, Max (1983), "Models of Hemispheric Specialization," Psychological Bulletin, 93 (January), 72-104. Alwitt, Linda F. (1981), "Two Neural Mechanisms Related to Models of Selective Attention," Journal of Experi- mental Psychology: Human Perception and Performance, 7 (April), 324-332. Anand, Punam, Morris B. Holbrook, and Debra Stephens (1988), "The Formation of Affective Judgments: The Cognitive-Affective Model versus the Independence Hy- pothesis," Journal of Consumer Research, 15 (Decem- ber), 386-391. Antis, S. M. (1973), "A Chart Demonstrating Variations in Acuity with Retinal Position," Vision Research, 14 (July), 589-592. Bornstein, Robert F. (1989), "Exposure and Affect: Overview 4nd Meta-analysis of Research, 1968-1987," Psycholog- ical Bulletin, 106 (September), 265-288. and Paul R. D'Agostino (1992), "Stimulus Recogni- tion and the Mere Exposure Effect," Journal of Person- ality and Social Psychology, 63 (October), 545-552. Brownell, Hiram H., Heather H. Potter, and Amy M. Bihrle (1986), "Inference Deficits in Right Brain-Damaged Pa- tients," Brain and Language, 27 (March), 310-321. Dimond, Stuart J. (1972), The Double Brain, New York: Hal- stead. Ellis, Andrew W. and Diane Miller (1981), "Left and Wrong in Adverts: Neuropsychological Correlates of Aesthetic Preference," British Journal of Psychology, 72 (May), 225-229. Friedman, Alinda and Martha C. Polson (1981), "Hemi- spheres as Independent Resource Systems: Limited-Ca- pacity Processing and Cerebral Specialization," Journal of Experimental Psychology: Human Perception and Performance, 7 (October), 1031-1058. Gordon, Harold W. (1983), "Music and the Right Hemi- sphere," in Functions of the Cerebral Right Hemisphere, ed. Andrew W. Young, London: Academic Press, 65- 86. Greenwald, Anthony G. (1992), "New Look 3: Unconscious Cognition Reclaimed," American Psychologist, 47 (June), 776-779. and Clark Leavitt (1984), "Audience Involvement in Advertising: Four Levels," Journal of Consumer Re- search, 11 (June), 581-592. Grier, J. Brown (1971), "Nonparametric Indexes for Sensi- tivity and Bias: Computing Formulas," Psychological Bulletin, 75 (June), 424-429. Hamid, Paul N. (1973), "Exposure Frequency and Stimulus Preference," British Journal of Psychology, 64 (Novem- ber), 569-577. Hansen, Flemming (1981), "Hemispherical Lateralization: Implications for Understanding Consumer Behavior," Journal of Consumer Research, 8 (June), 23-36. Hoffman, Martin L. (1986), "Affect, Cognition and Motiva- tion," in Handbook of Motivation and Cognition: Foun- dations of Social Behavior, ed. Richard M. Sorrentino and E. Tory Higgins, New York: Guilford, 244-280. Jacoby, Larry L., D. Stephen Lindsay, and Jeffrey P. Toth (1992), "Unconscious Influences Revealed: Attention, Awareness and Control," American Psychologist, 47 (June), 802-809. Janiszewski, Chris (1988), "Preconscious Processing Effect: The Independence of Attitude Formation and Conscious Thought," Journal of Consumer Research, 15 (Septem- ber), 199-209. - (1990a), "The Influence of Print Advertisement Or- ganization on Affect toward a Brand Name," Journal of Consumer Research, 17 (June), 53-65. (1990b), "The Influence of Nonattended Material on the Processing of Advertising Claims," Journal of Mar- keting Research, 27 (August), 263-278. Kenny, David L. and Eliot R. Smith (1980), "A Note on the Analysis of Designs in Which Subjects Receive Each Stimulus Only Once," Journal of Experimental Social Psychology, 16 (September), 497-507. Krugman, Herbert E. (1986), "Low Recall and High Rec- ognition of Advertising," Journal of Advertising Re- search, 26 (February/March), 79-86. Lazarus, Robert S. (1982), "Thoughts on the Relations be- tween Emotion and Cognition," American Psychologist, 37 (September), 1019-1024. (1984), "On the Primacy of Cognition," American Psychologist, 39 (February), 124-129. LeDoux, Joseph E. (1986), "Sensory Systems and Emotions: A Model of Affective Processing," Integrative Psychiatry, 4 (December), 237-243. (1989), "Cognitive-Emotional Interactions in the Brain-Special Issues: Development of Emotion-Cog- nition Relations," Cognition and Emotion, 3 (Novem- ber), 267-289. Levy-Schoen, Ariane (1981), "Flexible and/or Rigid Control of Oculomotor Scanning Behavior," in Eye Movements: Cognition and Visual Perception, ed. Dennis F. Fisher et al., Hillsdale, NJ: Erlbaum, 299-314. Macmillan, N. A. and H. L. Kaplan (1985), "Detection The- ory Analysis of Group Data: Estimating Sensitivity from Average Hit and False-Alarm Rates," Psychological Bul- letin, 98 (July), 185-199. Mandler, George, Yashio Nakamura, and Billi Shebo Van Zandt (1987), "Nonspecific Effects of Exposure on Stim- uli that Cannot Be Recognized," Journal of Experimental Psychology: Learning, Memory, and Cognition, 13 (Oc- tober), 646-648. Marcel, Anthony J. (1983), "Conscious and-Unconscious Perception: Experiments on Visual Masking and Word Recognition," Cognitive Psychology, 15 (April), 197-237. Moscovitch, Morris and Danny Klein (1980), "Material-spe- cific Perception Interference for Visual Words and Faces: Implications for Models of Capacity Limitations, Atten- tion, and Laterality, Journal of Experimental Psychology: Human Perception and Performance, 6 (July), 590-604. Navon, David (1984), "Resources-A Theoretical Soup Stone," Psychological Review, 91 (April), 216-234. This content downloaded on Wed, 26 Dec 2012 13:00:08 PMAll use subject to 392 JOURNAL OF CONSUMER RESEARCH and Daniel Gopher (1979), "On the Economy of the Human Information Processing System," Psychological Review, 86 (April), 214-255. Obermiller, Carl (1985), "Varieties of Mere Exposure: The Effects of Processing Style and Repetition on Affective Response," Journal of Consumer Research, 12 (June), 17-30. Sawyer, Alan G. (1981), "Repetition, Cognitive Responses, and Persuasion," in Cognitive Responses in Persuasion, ed. Richard E. Petty et al., Hillsdale, NJ: Erlbaum, 237- 261. Seamon, John G., Richard L. Marsh, and Nathan Brody (1984), "Critical Importance of Exposure Duration for Affective Discrimination of Stimuli That Are Not Rec- ognized," Journal of Experimental Psychology. Learning, Memory, and Cognition, 10 (September), 544-555. Sereno, Anne B. and Stephen B. Kosslyn (1991), "Discrim- ination within and between Hemifields: A New Con- straint on Theories of Attention," Neuropsychologia, 27 (July), 659-675. Stark, Lawrence and Stephen R. Ellis (1981), "Scanpaths Re- visited: Cognitive Models Direct Active Looking," in Eye- Movements: Cognition and Visual Perception, ed. Dennis F. Fisher et al., Hillsdale, NJ: Erlbaum, 193-226. Treisman, Ann M. (1988), "Features and Objects: The Four- teenth Bartlett Memorial Lecture," The Quarterly Jour- nal of Experimental Psychology, 40A (May), 201-237. Wolfe, Jeremy M. and Kyle R. Cave (1990), "Deploying Vi- sual Attention: The Guided Search Model," in AI and the Eye, ed. Andrew Blake and Tom Troscianko, Chi- chester: Wiley, 79-104. , Kyle R. Cave, and Susan L. Franzel (1989), "Guided Search: An Alternative to the Feature Integration Model for Visual Search," Journal of Experimental Psychology: Human Perception and Performance, 15 (August), 419- 433. Young, Andrew W. (1982), "Methodological Theoretical Bases," in Divided Visual Field Studies of Cerebral.Or- ganization, ed. J. Graham Baeumont, London: Academic Press, 11-29. Young, Laurence R. and David Sheena (1975), "Eye-Move- ment Measurement Techniques," American Psychologist, 30 (March), 315-330. Zajonc, Robert B. (1980), "Feeling and Thinking: Preferences Need No Inference," American Psychologist, 35 (Feb- ruary), 151-171. (1984), "On the Primacy of Affect," American Psy- chologist, 39 (February), 117-123. This content downloaded on Wed, 26 Dec 2012 13:00:08 PMAll use subject to