the information processing of pictures in print advertisments

Upload: cosmo0624

Post on 29-May-2018

228 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/8/2019 The Information Processing of Pictures in Print Advertisments

    1/18

    The Information Processing ofPictures in Print AdvertisementsJULIE A. EDELLRICHARD STAELIN*

    This study postulates that the presence of a dominant picture in a print ad canalter a consumer's cognitive activity while viewing the ad. Using the findings o(previous research, we develop a model of the process by which a viewer evalu-ates, encodes, and stores infomiation found in the ad. This model is tested in alaboratory se tting. The results clearly indicate that the way a messag e is conveyed(pictorially or verbally), whether the viewer is given some reference point for en-coding the picture (framed or not framed), and the type of claim made (objective,subjective, or characterization) have a significant effect on Ihe processing of theinformation presented in the advertisements. These differences in processing arealso shown to affect the viewer s brand attitudes and purchase intentions.

    E xtensive use of pictures in print advertisements is nota recent phenomenon. Even 20 years ago, over 75percent of all print advertisements had pictures covering atleast half of the available space (Baker 1961) and advertis-ing researchers were predicting that print ads would even-tually be all picture (Plisken 1963). Although pictures areused extensively to convey information about ihe brand, toshow its users and uses, and/or to create an image or per-sonality for tbe brand (Runyo n 1979). only recently has thepicture component of print advertisements become an objectof research interest. Tbis research has focused primarily onpictorial effects on the memorability of advertisements andthe cognitive structure (belief strengths, attitudes, inten-tions) resulting from exposure to these advertisements.

    Most memorability research is concemed with how wellconsumers recognize pictorial versus verbal stimuli or whyone form of stimulus is better tban the other in terms ofrecall. Starcb (1966), using his now famous advertisingeffectiveness technique, found that when a print advertise-ment contained a picture, people were more likely to re-member seeing it tban wben it did not. In a laboratory

    'Julie A. Edell is an Assistani Professor of Business and Richard Staelinis Edward and Rose Donnell Professor of Business Administration, bothai the Fuqua School of Business, Duke Universily, Durham. NC 27706.The authors wish lo acknowledge the financial support provided by theAmerican Marketing Association Dociorat Dissenation Grant and theAmerican Association of University Women Educational Foundation Fel-lowship. We also wish to thank Andre^^ A. Mitchell, J. Edward Russo.and Eric J. Johnson for their valuable guidance in stimulating, developing,and presenting these ideas, as well as the JCR reviewers for their com-ments on previous drafts of ihis paper.

    Study, Shepard (1967) showed subjects 612 pictures, manyof wbich were photographs taken from advertisements.Subjects were tben shown 68 pairs of pictures containinga previously seen photo and a novel one. In an immediaterecognition test, 96.7 percent of the stimulus items werecorTectly identified. This was contrasted with 89 percentcorrect recognition of sentences in a similar test environ-ment, a difference found to be statistically significant. Evenafter a delay of 97 days between the initial exposure to thepictures and the test, subjects correctly recognized 87 per-cent of the previously shown pictures.

    Recently, researchers have expanded their investigationsto test hypotheses concemed with why pictures are moreeasily recalled. Lutz and Lutz (1977) published a studyexamining two types of pictures in Yellow Page advertise-ments. These pictures were either interactive (integratingthe brand name and the product class in tbe picture) ornoninteractive {depicting the brand name or product sepa-rately from the written fonri of the other member of thepair). They found that recall of brand name was facilitatedonly by tbe interactive picture. Thus they concluded thatin the paired-associate leaming task, it is the integration ofthe information into the single image tbat increases the re-call of the corresponding item when given the cue.More recently, Kisielius (1982) postulated that infor-mation presented pictorially stimulates more cognitive elab-oration and thus results in the development of more storagelocations and pathways in memory, which in tum increasestbe likelihood of the information being retrieved in laterrecall tasks. To test her theory, Kisielius presented subjectswith brand information either by sentences alone (low im-age condition) or by sentences and drawings in combination

    45C JOURN AL OF CONSUMER RESEARCH V ol. 10 June

  • 8/8/2019 The Information Processing of Pictures in Print Advertisments

    2/18

    46 THE JOURNAL OF CONSUMER RESEARCH

    {high image condition). In a subsequent recall task, subjectsseeing tbe high image information recalled more brand in-formation than those who saw the sentences alone, sup-porting her hypothesis. Kisielius also examined the brandattitude formed as a result of viewing the stimuli. Her the-ory predicts that the addition of pictures to the sentencesenhances or reduces the brand attitude, depending on therelative favorability of the information conveyed by thepictures. In her study, the high image condition resulted inless positive brand attitudes.

    Two other advertising studies involving pictorial-verbaldistinctions have examined the consumer's attitudes andbeliefs about the brand subsequent to being exposed to theadvertisement. Rossiter and Percy (1978) presented a theorywhich predicts that both the visual and verbal componentsof advertisements are processed for product information. Inaddition, the visual portion of the ad is hypothesized todirectly affect the consumer's attitude toward the brand.This advertisement attitude effect is above any effect thead has via the consumer's brand beliefs.To test their theory , Ro ssiter and Percy exposed subjectsto advertisements with a constant portion that containedthree copy points together with a large or small drawing.If the picture was small, the copy was presented in a largetype face; if the picture was large, the copy was in smalltype face. Attitude toward the brand was found to be sig-nificantly more positive when the picture was large. Ros-siter and Percy inferred that this difference was a result ofa direct transfer of affect. Unfortunately, no measures ofproduct beliefs were taken, so they could not say anythingabout the alternative explanation that the attitude differ-ences were caused by differences in the belief structurebrought about by the different ad formats.In a second study reported by Mitcbell and Olson (1981),each subject saw four advertisements, each for a differentbrand of a fictitious facial tissue. The advertisements con-tained only one element (a picture or verbal claim) and tbebrand name. Subjects saw tbe advertisements between twoand eigbt times. The verbal claim was "Brand I FacialTissues are Soft." The pictures were of a "fiuffy ki tte n,"a "spectacular sunset over an ocean," and "a presumablyneutral picture ofan abstract painting." In this instance thead containing the drawing of the kitten (but no verbal mes-sage to frame the drawing) was found to yield more favor-able beliefs that the brand was very soft and came in moreattractive colors than was the verbal message without apicture. Moreover, the overall brand attitude for the facialtissue advertised using the kitten picture was found to bemore positive. Mitchell and Olson concluded that the morefavorable attitude was caused by the difference in beliefsbrought about by the pictorial infonnation as compared tothe verbal infonnation.These studies are compatible with the premise that pic-torial advertising stimuli can yield different outcomes thanverbal advertising stimuli. In general, pictorial stimuli arebetter recalled. Also, more pictorial advertisements havebeen found to result in more favorable product beliefs andbrand attitudes, as well as less favorable ones. Although

    several authors have hypothesized why these differencesmight occur, none of the empirical studies have ruled outother plausible hypotheses or examined in depth the me-diating cognitive processes which bring about these differ-ences in outcomes.

    RESEARCH PURPOSEThe intent of this study is to develop and test a processmodel useful in explaining the abovementioned differencesin consumer responses to verbal and pictorial advertisingstimuli. Since our interest includes gaining a better under-standing of wby we observe differences in cognitive struc-ture, our model incorporates measures of the consumer'scognitive activities while viewing the ad, as well as themore standard input and output measures. Such processmeasures have been shown to be associated with (1) themeaning and evaluation assigned to the stimulus, (2) howthe infonnation is stored in memory, and (3) the likelihoodthat and the speed with which the information will subse-quently be recalled (e.g.. Chase and Clark 1972; Craik andLockhart 1972; Greenwald 1968; Hyde and Jenkins 1973;Johnson and Russo 1978; Lutz and Swasy 1977; Paivio1971; Wright 1973, 1980).

    Our model is based in part on published material (thatdealing with both theory development and empirical obser-vations) and in part on our intuition. A major componentof our model is our conceptualization of how infonnationprovided in the verbal portion of the advertisement affectsthe processing of the information in the pictorial ponion,and vice versa. We have previously cited a number of stud-ies wbich show that inserting pictures in ads causes differ-ences in attitudes and beliefs. H owever, Taylor and Thom p-son's review article (1982) cites three studies that examinedthe effect of adding photographs containing the same in-formation to a verbal message. In none of tbese studies didthe information conveyed in the combined verbal and pic-torial condition result in significantly different persuasioneffects than when the information was conveyed only ver-bally.We take these results to indicate that the presence of apicture in an advertisement is not a sufficient condition toguaran tee differences in cognitive proc essing of the mes-sage. Moreover, our intuition leads us to believe that ad-vertisements' pictorial and verbal components are usuallytreated differently by consumers when viewing an ad. Pic-tures in general are more attention-getting, pleasant, andeasier to process tban is verbal text. Consequently, viewersattend more to the picture and tend to expend less process-ing resourc es on the more effortful verbal text. In addi tion,the cognitive activities directed at the picture, u nless guidedby the verbal message, often concem thoughts less relatedto brand evaluation. For example, take a Belair cigarettead which uses V*. of the page to show a couple playing ithe surf followed by text "Now 30% less 'tar'!" In thiscase the picture may serve to distract viewers by encour-aging them to think about their last trip to the beach, theman in his swimming suit, and so on. Such a distraction

  • 8/8/2019 The Information Processing of Pictures in Print Advertisments

    3/18

    PICTURES IN PRINT AD S 47can lead consumers to spend less effort critically thinkingabout attributes they normally would use to evaluate thebrand's tar claim.If we assume that the inclusion ofan "unframed" picture( i .e. , an ad in which the verbal message does not relate thepicture to the brand) results in the consumer allocating cog-nitive resources to items not normally related to brand eval-uation, it follows that this unframed picture also diminishesthe chance that the consumer will activate from memoryany stored infonnation about the product class or productbeing advertised. This results in the viewer generatingfewer evaluative thoughts during the encoding process.Moreover, since product information stored in memory isnot activated, any evaluative thoughts generated are morelikely to be about attributes mentioned in the ad than aboutattributes which the subject normally uses to evaluate thebrand.

    To support this unframed picture premise, we point tothe findings of Bither and Wright (1973), who studied asomewhat analogous situation of showing consumers a TVcommercial with incongnient video and audio messages.As the video message became more distracting, the subjectswere able to recall significantly fewer points from the audiomessage. Moreover, the group of subjects identified a priorias most likely to counterargue with the audio message werefound to be more influenced by the audio message when itwas accompanied by distracting video. This latter findingis compatible with our premise that "d istrac ted" consumersdo not retrieve from memory the criteria normally used toevaluate the advertised brand and thus do less critical think-ing.

    Our model predicts a different cognitive process whenthe pictorial information is restated in a verbal form (i.e.,"framed"). Much research has examined the influence ofverbal labeling of objects upon the way the object is en-coded (e.g., Carmichael, Hogan, and Walter 1932; Cohen1966; Coop er 1974; Daniel 1972). This work found thatrecall of ambiguous objects is greater when the object islabeled prior to its encoding than when it is not labeled orlabeled only prior to recall. In addition, Jorg and Hormann(1978) found that even for clearly recognizable objects, thespecificity of the label (tulip versus flower) affects the en-coding and recognition ofthe object. Although these studiesdid not use a print advertisement as a stimulus, they arecompatible with the hypothesis that when a picture in anadvertisement is framed, the viewer uses this label as aframework for encoding the picture in relationship to theadvertised brand. As such, the picture functions more likean illustration of the verbal message than as a potentialsource of distraction. Thus a framed picture is processedmore like the verbal material in the absence of the picturethan the picture is in the absence of the verbal label.

    This discussion forms the basis for the following processmodel (see Figure A). When consumers attend to a printadvertisement for the purpose of gaining information aboutthe brand and the ad is not primarily pictorial, they retrieverelevant information from memory in order to encode theinformation in the ad. In addition, the consumer engages

    in brand evaluation by comparing the new information withpreviously stored criteria normally used for evaluating al-ternatives within a product class. The brand evaluationalong with the encoded brand facts are integrated with thepreviously activated information. Also, a trace of the ad isstored in memory.A similar process is postulated for consumers viewing aprimarily pictorial print advertisement when the major pic-ture is framed. However, the process differs if the dominantpicture in a primarily pictorial ad is not framed. In thissituation consumers tend not to retrieve relevant informa-tion from memory. This modifies how the information pro-vided in the ad is encoded and results in the consumersgenerating fewer evaluative thoughts. More specifically,consumers are less likely to use their normal criteria (whichare stored in memory) to form brand evaluations. Instead,if the ad results in any brand evaluations, these are ondimensions suggested by the ad and not by the consumer'snormal evaluation criteria. Moreover, since little prior in-formation is retrieved, any brand evaluations formed are

    less likely to be integrated with prior relevant information.Instead, these brand concepts are stored in memory alongwith the trace of the ad.In summary, our model postulates different cognitiveactivities between (1) print ads where either the messageis conveyed verbally or the dominant picture is framed, and(2) print ads where the dominant picture is not framed. Wetest specific aspects of this model using a variety of processmeasures in addition to the more standard outcome mea-sures. This testing is done by postulating hypotheses thatrelate our model to specific process and outcome measures.

    M E T H O DExperimental DesignTo provide a generalizable setting for testing specificaspects of our process model, we decided to manipulate(wo factors in addition to the structure of the advertise-mentthe products advertised, and the content of the pic-ture and/or verbal mes sage. This resulted in a 3 x 3 x 3after-only Latin-square design with the following factors:

    structure of the varied portion of the ad (verbal, pictorialunframed, and pictorial framed); content of the varied mes-sage (objective, subjective, and characterization); and prod-uct class advertised (car, camera, and calculator). The de-sign used was based on Winer's repeated measures plan 11(1971, pp. 739-745). Repeated measures on each subjectwere obtained in order to separate individual differencesknown to influence cognitive activity (e.g., verbalness,self-confidence, intelligence) from the hypothesized differ-ences in processing due to the structure of the ad. In sucha design all the treatment combinations are used, allowingfor complete within-subject information on all the maineffects. In addition, partial information can be obtained forthe interaction effects. As shown in Table 1, each subjectsaw an ad from each structure condition, as well as onefrom each content and product category, As used here.

  • 8/8/2019 The Information Processing of Pictures in Print Advertisments

    4/18

    48 THE JOURNAL OF CONSUMER RESEARCHRGURE A

    PROCESS MODEL OFHOW PEOPLE VIEW A PRINT AD

    Look at layoutof ad

    Is adprimarilypictorial?

    the dominantpictureframed?

    N o

    Yes

    Do no brandevaluation on

    a priori importantbrand attributes

    Retrieve relevantinformation fronn

    memory

    Process informationfrom the ad using theretrieved information

    Store ad tracebut not withproduct classinformation

    Store brandevaluations, brand

    Information, and adtrace with recalled

    product information

    "group" refers to the particular set of three advertisementsseen by the subject.Products. The three product classes used in the exper-iment were cars, cameras, and calculators. Two generalprinciples guided us in selecting these product classes.First, it was desirable tohave products about which subjectswould normally use information from ads in the purchaseprocess. Second, the products had to be ones about which

    subjects had some prior information. To select the threeproducts, a questionnaire concerning 20 products felt tomeet these criteria was administered to a sample of 50 sub-jects selected from the population of consumers designatedfor subsequent testing. Information from this questionnairewas used to ascertain their purchase and usage experienceand to determine how likely it was that they would usemagazine advertisements as a major source of informationin brand search and selection. Cars, cameras, and calcu-

  • 8/8/2019 The Information Processing of Pictures in Print Advertisments

    5/18

    PICTURES IN PRINT ADS 49TABLE 1

    EXPERIMENTAL DESIGNGroup

    123456789

    PictorialSubjectiveObjectiveCharacterizationCharacterizationSubjectiveObjectiveObjectiveCharacterizationSubjective

    unframed adCameraCalculatorCa rCalculatorCa rCameraCa rCameraCalculator

    Verbal adCharacterizationSubjectiveObjectiveObjectiveCharacterizationSubjectiveSubjectiveObjectiveCharacterization

    CalculatorCarCameraCarCameraCalculatorCameraCalculatorCa r

    Pictorial framedObjectiveCharacterizationSubjectiveSubjectiveObjectiveCharacterizationCharacterizationSubjectiveObjective

    adCa rCameraCalculatorCameraCalculatorCa rCalculatorCa rCamera

    lators were the three products that ranked highest on bothusage experience and the use of magazine advertisementsas a major source of information.Content. The second design factor, content of the mes-

    sage in the experimental portion of the ad, consisted ofthree levels: objective, subjective, and characterization.The objective message described physical properties of thebrand by giving factual, verifiable informationi.e., in-formation conveyed using a scale not subject to individualinterpretation. An example of objective content is a state-ment of the tar and nicotine level of a brand of cigarettes.In contrast, subjective product infomiation is open to in-dividual interpretation, primarily because there is no uni-versal agreement on the scale used to measure the productperformance attribute mentioned. A subjective message ina cigarette advertisement might tell how smoking the ad-vertised brand results in satisfaction and a rich, full-bodiedtaste. The third category, characterization, attempts totransfer positive im pressions to the brand by associating thebrand with a favorably evaluated object, situation, or per-son. An ad for a cigarette showing a rugged he-man smok-ing a cigarette as he rides on horseback through a beautifulmountain valley is one such example.The three content levels were based in part on the workof Holbrook (1978) and in part on common sense. Holbrookfound that when the message was objective (factual), theresulting brand beliefs and attitude were more positive thanwhen the message was subjective (evaluative). We tookthese findings to imply that content is also likely to affectthe intervening cognitive activities under investigation here.Since characterization content is commonly used in highlypictorial advertising , w e included it so that the results couldbe generalized to a broader set of advertisements.The attributes used in the experiment and the contentcategories to which they belong are given in Table 2. Manyof these attributes could have been conveyed so that theywould fall in any of the content categories. To determinewhether subjects would perceive the message used lo bethe type of content intended, a pretest was conducted.Twenty subjects were asked to classify the content of thepictorial or verbal form of the experimental ad segment

    according to the definitions already given. All nine mes-sages used in both the pictorial and verbal forms were cor-rectly classified by all 20 subjects.Structure. Three structures for conveying information

    in the experimental portion of the ad were an unlabeledpicture (pictorial-unframed), a paragraph (verbal), and boththe picture and paragraph in combination (pictorial-framed). Since the pictorial structure and the verbal struc-ture were not physically identical, great care was taken toensure that the verbal and pictorial segments conveyed anequivalent message. Only then could we attribute differ-ences in processing to the differences in structure. W e de-veloped equivalent messages as follows.First, 36 pictures (four from each of the three contentcategories for each product) were selected from either ad-vertisements from other product classes or from print adswhich were more than two years old. These pictures were

    shown to 20 subjects who were asked to write down every-thing the picture conveyed to them about a brand from theintended product class. Half of the pictures were then elim-inated for containing too many messages or for conveyingvery different things to different people. The 18 remainingpictures (two for each content category for each product)were given to a second sample of 10 subjects. Their in-suuctions were: "On each of the following pages there isa picture which is part of a magazine advertisement. Thead is to be written for radio, so the picture must be con-verted into a verbal message. If you were given the assign-ment of conveying the message of the picture in words,what would you say?" These responses were used as thebasis for writing a paragraph conveying the same messageas the picture. Each paragraph contained exactly 38 words.

    A third group of 20 subjects was then given the 18 adsegments. Ten were given the messages conveyed via pic-tures, and 10 were given the messages conveyed via para-graphs. All were asked lo indicate the main message beingconveyed in each of the 18 ad segments. We selected be-tween the two remaining ad segments for each of the ninecells by choosing the segment that had the greater messageconsistency within the cell for the two formats. These 20subjects also rated each segment on two seven-point scales

  • 8/8/2019 The Information Processing of Pictures in Print Advertisments

    6/18

    50 THE JOURNAL OF CONSUMER RESEARCHTABLE 2

    ATTRIBUTES USED IN EXPERIMENTAL STIMULI

    ProductsCar

    Calculator

    Camera

    Objectivetype of seats

    equipped witha card readerequipped withan electronicflash

    Content categoriesSubjectivehandling

    durability

    takes greataction shots

    Characterizationused by richpeopleused by businesspeopleused byprofessionalphotographers

    indicating how difficult or easy the message was to under-stand and how believable it was. For the nine segmentsused in the study, r-tests indicated no differences betweenthe verbal and pictorial presentations for ease of under-standing (largest t = 1.35) or believability (largest / =1.15).The nine ad segments selected were subjected to onefinal test. A list of all the responses was compiled from thefirst pretest for the pictures and the second pretest for theverbal segments. A fourth group of 20 subjects was shownthe nine ad segments in either the pictorial or verbal formand asked to select from the list of possible messages forthat product category all messages conveyed by the seg-ment. In every case the intended message was selected byall 20 subjects. For eight of the nine segments tested, mes-sages other than the intended message were also selectedby both groups. A chi-square test of homogeneity indicatedno significant difference in the number of messages se-lec ted be tween s t ruc tures ( la rges t x^ = 2.67,p > 0.3). Based upon these data we concluded that nosystematic differences existed between the pictorial or ver-bal presentation with respect to (1) the number of messagesconveyed by the ad segments, (2) the major message con-veyed, (3) the ease of understanding the message, or (4) itsbelievability.Stimuli

    The stimuli were full-page, color print ads with a con-stant and an experimental component. The constant pmrtionconsisted of a headline, the (hypothetical) brand name, acolor photograph of the product, and two verbal claims.This portion of the ad was the same for all forms of the adfor a product and took up about 60 percent of the nonblankspace. The variation in content and structure across ads wasaccomplished via the nine experimental ad segments (de-scribed in Table 2). These nine messages were presentedfor each of the three product categories, yielding 27 dif-ferent ads. In addition to these 27 ads, two other ads forcameras were shown to all subjects. These advertisements

    had the same format as the experimental ads and were in-cluded as part of a second study run simultaneously.Subjects

    The subjects used throughout all phases of the study var-ied in age between 20 and 28. A total of 27 subjects par-ticipated in the main experiment, which lasted one and ahalf" hours. The num ber of subjects required was determinedfrom pretest results using the power approach. The a levelused to control for the risk of a type I error was chosen tobe 0.05 . The power of the test was set at 0.80 (Cohen 1969,p. 54). The variables used in the calculations were thoseabout which major hypotheses had been made. These were:number of counterarguments, number of suppon argu-ments, and attitude toward the brand. (See the subsequentdiscussion of hypotheses.) Using the means and variancesof the unframed pictorial and verbal conditions from a pre-test of eight subjects, the effect sizei.e., the standardizeddifference scorewas calculated and was always greaterthan 0.7 (Cohen 1969, pp. 25, 42). Using the power func-tions tabulated in Cohen, the sample size needed neverexceeded 27 subjects. As is discussed later, in some in-stances fewer than 27 subjects were used in obtaining somemeasures. In these cases we were less sure of the power ofour tests. However, our plan of analysis for most measuresincorporated tests more powerful than the straight f-testsassumed in the sample size calculations. Thus we believedthe power calculations would prove to be conservative.

    ProcedureAn overview of the experimental session is provided inthe Exhibit. Subjects were telephoned three to seven days

    before participating in the main experiment. Under theguise of a new product study conducted by an independentsurvey research company, they were asked what informa-tion they would need before selecting a brand in sevendifferent p roduct classe s, including cars, cam eras, and cal-culators. These responses were used to determine the attri-butes the subjects normally used to evaluate a particularbrand within each of the three product classes studied. Nosubject mentioned anything about the telephone survey dur-ing or after the actual experiment, although no specificquestion conceming the telephone survey was asked of therespondents.Each subject performed the main task individually. Upon

    entering the laboratory, subjects were told they were partof a test marketing study for several new brands of prod-ucts. The subject was asked to view each of the five ad-vertisements as if s/he were considering purchasing theproduct and to say whatever came into his/her mind whileviewing the advertisement. Eye movements were also col-lected during the viewing of the advertisements for ninesubjects. Following the exposure to the five advertisements,all the subjects were given a five-minute distraction task toensure that any information recalled was from long-termmemory.

  • 8/8/2019 The Information Processing of Pictures in Print Advertisments

    7/18

    PICTURES IN PRINT ADS SIEXHIBIT

    OVERVIEW OF EXPERIMENT

    Task Time ofoccurrence Measures Number ofsubjects1 . Name important attributes on 7 product

    classes.2 . View 3 experimental ads plus 2 otherads and give thoughts aloud.

    3-7 days beforemain experiment

    Mainexperiment3. Nonsense task (5 minutes).4 . State beliefs about each brand.5. Either:Write down everything recalled about

    the brand and the ad.Or:Answer 3 questions about each brandfrom the constant ad segments.6. Indicate brand beliefs, brand attitudes,and purchase intentions.7. Indicate product class knowledge,interest, and purchase experience.6. Manipulation checks.

    Responses used to determine a prioriimportant attributesCognitive responsesNumber of counterargumentsNumber of support argumentsNumber of other thoughts

    Eye movementsTime spent viewing the 3 different ad segmentsNo measure takenUsed responses to determine what attributes tobe included in the belief questionnaireRecallNumber of brand items recalled, number of ad

    items recalledResponse time and whether response wascorrect

    ,. e,, . PIKnowledge, interest, and number of previouspurchasesType of content and m essage conveyed.believability, and ease of understanding themessage

    2727

    2727

    1413

    272727

    Subjects were then asked to tell the interviewer every-thing they believed to be true about each advertised brand.This information was used along with the infonnation ob-tained during the "market survey" telephone interview todetermine each subject's salient attributes for each product.Fourteen of the 27 subjects (balanced across cells) werethen asked to complete a written recall of everything theyremembered about each of the products and each of theadvertisements, regardless of whether they believed it ornot. The remaining 13 subjects responded to a series ofquestions about information presented in the advertisement.All subjects were asked to complete a number of scalesindicating their beliefs, attitudes, and purchase intentionsfor each of the advertised brands. They then provided in-formation about their purchase experience and indicatedtheir interest in and knowledge of each product class. Fi-nally, several manipulation checks were completed.Measures

    An overriding consideration in designing our experimentwas to obtain multiple measures using different measure-ment techniques. This led us to utilize a number of process-

    tracing techniques. (For a good discussion of the merits ofeach of these techn iques, see Gardner, Mitchell, and Russo1978; Jacoby et al. 1978; Payne and Ragsdale 1978; Russo1978.) The measures we used are as follows:Verbal protocols. Verbal protocols were collected byasking subjects to "say whatever thoughts come to mindas you look at the ad. Please speak freely about yourthoughts conceming each ad and new brand. 1 would likeall of your thoughts as they come to mind." The verbalprotocols analyzed and reported here were from the entireadvertisement, i .e., from both the experimental and the

    constant portions.Three judges (not the authors) classified all thoughtsabout the brand into one of three categoriessupport ar-guments, counterarguments, and neutral thoughtsusingthe definitions of Wright (1973). The basis for the finalclassification was the modal rating of the three judges.There were no cases in which all three judges disagreed.Interjudge reliability was 0.93 for support arguments, 0.95for counterarguments, and 0.91 for neutral thoughts. Foreach support and counterargument, the judges also indi-cated the attribute about which the subject was thinking and

  • 8/8/2019 The Information Processing of Pictures in Print Advertisments

    8/18

    52 THE JOURNAL OF CONSUMER RESEARCHwhether it was on the list of attributes considered a priorias being important. A total of 99 percent of the thoughtswere unanimously classified either as "a priori important"attributes or as "n ot a priori im portan t" by all three judges.

    Eye movements. Eye movement data were collected atthe Consumer Behavior Laboratory of the University ofChicago. In this study, the eye movements were used pri-marily to ascertain how much of the total time was spentlooking at the various components of the advertisements.Specifically, four measures were derived: total time spentlooking at the advertisement; time spent on the experimentalportion of the advertisement; and times spent on the con-stant verbal and the constant pictorial portions of the ad-vertisements.

    Recall. Two types of recall data were collected. First,subjects were asked to write down everything they couldrecall about the product advertised, whether they believedit or not, and about the advertisement itselfits layout,tone, how it conveyed the information, and so on. Recallwas scored as the number of items correctly indicated. Sec-ond, aided recall measures were obtained by asking ques-tions on specific brand facts. Response times were mea-sured for the brand facts.

    Cognitive structure measures. In addition to the pro-cess-tracing variables listed above, data were collected ona number of cognitive structure variablesnamely, brandattribute beliefs (b^) and evaluation (ej, brand anitude (AJ,attitude toward the act of purchasing the brand (A^^), andpurchase intentions (PI). Instead of directly using the in-dividual belief and evaluation measures, an aggregate-weighted-average brand-belief measure (Sf^ \bieJN) wasformed. Unlike most previous applications, the attributesused in our brand-belief measure were unique to each sub-ject. Rather than being composed of a fixed set of attributes,the measure was based on those attributes which were either(1) mentioned by the subject as a priori important in eval-uating a brand within the product class, (2) listed by thesubject as true for the advertised brand during the elicitationtask, or (3) one of the three attributes conveyed in theexperimental portion of the ad.The b, measures were obtained by asking the subjects toindicate how likely it was that each brand possessed eachofthe abovementioned attributes on a seven-point scale (1= not at all likely, 7 = very likely ). The e, measures came

    from another seven-point scale ( - 3 = very bad, + 3 =very good ). Since the number of attributes mentioned dif-fered across subjects and product classes, an averagingmodel was used (^b^e^lN).Attitude toward the brand (A J was measured by the sumof four five-point evaluative scales (go od-b ad, dislike verymu ch-like very much, pleasant-unpleasant, poor quality-highquality), while A^^, was the sum of three five-point scales(good-bad, foolish-wise, beneficial-harmful; Mitchell andOlson 1981), The coefficient alpha scores for A^ and A^^,were 0.91 and 0.89, respectively. Purchase intention (PI)

    toward each brand was measured on a single bipolar 10-point scale anchored by the phrases "not at all likely tobuy" and "very likely to buy."Covariates. Several other measures were collected andused as covariates in the analyses. These were: the impor-tance of the experimental attribute, as measured on a seven-

    point scale; product class know ledge, as measured on a 15-item multiple-choice quiz; number of previous purchasesof each product category; and product class interest, asmeasured on a five-point scale.Each of these variables has been shown or hypothesizedto have an impact on how information from advertisementsis processed (Edell and Mitchell 1978; Wright 1980). Thesecovariates, along with the order in which the advertisementswere viewed, were incorporated in most of the subsequentanalyses to account for variation that might otherwise havebeen inconectly attributed to the variables of interest dueto spurious correlation.

    HYPOTHESESTbe previously discussed process model (Figure A) pos-tulated a series oi causal relationships. In this section, weoperationalize six of these relationships using the above-mentioned measures. They are as follows:H I : When the dominant picture in a print advertise-ment is not framed, no brand evaluative thoughts(counterarguments and support arguments) on apriori important attributes will be given. Whenthe dominant picture is framed or its message isconveyed verbally, brand evaluative thoughts ona priori important attributes will be given.

    Hypo thesis I operational izes our premise that the cog-nitive activity differs for ads composed primarily of an un-framed picture compared to ads composed of a dominantframed picture or of primarily verbal messages.We limit our hypothesis to a priori important attributesbecause our theory predicts that subjects will retrieve rel-evant information from memory when viewing the verbalor framed pictorial ads. It is this retrieved information thatthey will use for evaluating the incoming information. Sub-jects could make comments about the attributes mentionedin any of the advertisementsespecially since they wereinstructed to view ads as if they were considering purchas-ing the product.H 2: When the dominant picture of a print advertise-ment is not framed, free recall of the brand In-formation presented in the ad will be less than ifthe dominant picture is framed or its message isconveyed verbally. However, the number of ad-vertisement items recalled will not vary due tothe structure ofthe advertisement.

    Hypothesis 2 follows from our premise that processingdiffers by ad structure and therefore that information storage(and retrieval) also differ. Specifically, the cognitive activ-

  • 8/8/2019 The Information Processing of Pictures in Print Advertisments

    9/18

    PICTURES IN PRINT ADS

    ity associated with evaluating the brand on a priori impor-tant attributes will lead to complex memory structures ofthe brand and, consequently, to better recall of the brandmessages. Since our model says nothing about how struc-ture affects thoughts about the ad, we hypothesize no dif-ferences in these ad thoughts across structures.H 3 : When the dominant picture of a print advertise-ment is not framed, the time to correctly respondto brand statements will be longer than if thedominant picture is framed or its message is con-veyed verbally.

    Hypothesis 3, like Hypothesis 2, follows from our prem-ise that subjects engage in different cognitive activitieswhen viewing the different ad structures. Subjects whoevaluated the brand on a priori important attributes shouldhave stored the newly acquired information in the readilyaccessible product class network. Thus their response timesshould be faster. However, the response rates (in terms ofcorrectness) may not differat least in the short run sincesubjects can probably access the memory trace of the ad ifneed be.

    H 4: When the dominant picture of a print advertise-ment is framed or its message is conveyed ver-bally, the content of the message will influencethe number and type of cognit ive responses.When the content is primarily objective, the num-ber of support arguments will be greater thanwhen it is not objective. The converse is true fora subjective content, i.e., the number of coun-terarguments will be greater than when it is notsubject ive. When the message is primari lycharacterization, few brand evaluative thoughts,either support or counterarguments, are gener-ated.

    The portion of Hypothesis 4 regarding objective and sub-jective content is based on the previously discussed findingsof Holbrook (1978). Since cognitive responses have beenshown to influence attitude (Edell and Mitchell 1978;Wright 1980) and ad content has been found to infiuenceattitude (Holbrook 1978), we postulate more support ar-guments when the content is objective and more counter-arguments when the content is subjective. Our premise forcharacterization is based on the belief that the viewer is notled to question (or agree with) the implicit linking of theproduct to the particular person, place, or situation. In otherwords, any liriking of the product with the characterizationis accomplished without any substantial cognitive activityon the part of the viewer, thus leading to fewer brandevaluative thoughts:

    H 5 : The number and type of cognitive responses thatoccur during the viewing of the advertisementwill impact the cognitive structure oft he subject.Specifically, counterarguments will result inmore negative brand attitudes, while support ar-guments will result in more positive brand atti-

    tudes. Lack of counterarguments and support ar-guments will result in an intermediate brandattitude.Hypothesis 5 is not unique to our model; rather, it isbased on prior findings (e.g., Edell and Mitchell 1978;Wright 1973) and is included to show the ramifications ofHypotheses 1 through 4 on the cognitive structure outputmeasures. In this regard, we also state the following hy-pothesis:

    H 6: A more positive brand attitude and attitude to-ward purchasing the brand will lead to greaterpurchase intentions.RESULTS

    Manipulation ChecksAnalyses were performed to ensure that subjects per-ceived Uie same message conveyed across structures. Usingthe responses to the posnest manipulation-check questionsdesigned to measure the subjects' ability to determine themessage conveyed in the experimental portion of the ad,chi-square tests of homogeneity showed no differences inthe message for a panicular content-product combinationacross the three structures. No r were any significant differ-ences found in believability of the experimental segmentacross structure levels {F = 0.23). The framed-pictorialsegment was found to be significantly easier to understandthan the unframed pictorial segment it = 2.12, p < 0.05).Thus the ease of understanding score was included as acovariate in the subsequent analyses.'Since an aitemative explanation for the hypothesized ef-fects is that the framed pictorial and the verbal conditions

    contained explicit information on an additional attribute notcontained in the unframed pictorial condition, we examinedthe b, measures for each experimental attribute. These mea-sures indicate how strongly that attribute was associatedwith the advertised brand. We found that the differencesacrossstructures were not significant (f = 2. 3 6, p > 0 .10).This indicates that regardless of which structure was usedto convey the a ttribute, it was equally likely to be asso ciatedwith the brand. We take this finding as further evidencethat the differences found in the study are not due to dif-ferent infonnation being conveyed by the different struc-tures.A manipulation check was also performed to see whether

    subjects could correctly classify the contents of the exper-imental segment when given the three content definitions.Only three of the 81 responses indicated that subjects per-ceived the content type to be different than that intended.'If the framed pictorial segment caused the message to be more under-standable and this affects some other variable of interest, then by includingunderstandability as a covariate we arc implicitly reducing the main effectof structure. We do include the covariate, bowever, because our majorinterest is not the effect of undcRtandability but the direct effect of struc-ture, (in reality it doesn't matter in our case, since the covariaie is neverstatistically significant).

  • 8/8/2019 The Information Processing of Pictures in Print Advertisments

    10/18

    54 THE JOURNAL OF CONSUMER RESEARCHThese misclassifications made no difference of any conse-quence in the interpretation of the results. The analysesreponed here used the intended content categories.Finally, we were interested in whether or not the attributefeatured in the experimental portion ofthe ad was perceivedto have different evaluations across the different contentcategories. Using the e, measures from the 18 respondentswho did not see that particular ad content message, averagee^ measures for each attribute featured for each product classwere calculated. All attributes were positively evaluated;the overall averages were 0.92, 0.74, and 2.56 for objec-tive, characterization, and subjective content, respectively.Even within product classes, differences between the firsttwo averages were not significant. However, the subjectivemessages were significantly {p < 0.05) more positivelyevaluated than either the objective or characterization mes-sages for all products. Our intent was to have the attributesfeatured in each content message evaluated equally favor-ably so that we could rule out the hypothesis that the dif-ference in favorability causes different cognitive activity.Unfortunately, our treatments do not allow us to do this.Other Measures

    We calculated two different types of descriptive statisticsfor each structure. One was the amount of time subjectsspent looking at each segment of the ad. Overall, subjectsaveraged approximately 86 seconds. The specific times foreach portion were: experimental portion, 18 seconds; con-stant picture, 20 seconds; and constant paragraph, 48 sec-onds. In each case no significant differences were observedby structure,A second statistic we calculated was the ability of thesubject to answer three true-false questions concerning

    facts associated with the constant portions of the ad. Again,no differences were found amo ng the structures, with theaverage error rate being 12 percent. Taken together, thesetwo measures provide no evidence that the subjects paidmore or less attention to an ad or a specific portion of anad for a specific structure.Test of Hypotheses

    Ana l ys i s o f va r i ance (ANOVA) or cova r i ance(ANCO VA) was used to test Hypotheses 1 through 4 , whileOLS regression analysis was used for Hypotheses 5 and 6.'Since the equations estimated included variables not nec-essarily orthogonal to each other, common variance existedamong the variables. Instead of attempting to partition thiscommon variance, we take a more conservative strategyand report only the variance uniquely explained by a vari-able. In this way all our tests of significance are conser-vative.

    hypotheses 5 and 6 are not stated in terms of the treatment variables(smicture, content, and product), but in terms of cognitive responses andcognitive structure. For this reason we use OLS instead of an ANOVAframework. How ever, the effects of the treatment variables can be derivedby following the links as set forth in Hypotheses 1 through 6.

    TABLE 3SUBJECTS GIVING ANY BRAND EVALUATIVE THOUGHTON A PRIORI IMPORTANT ATTITUDES

    ANCOVA

    Between subjectsGroupsSubjects within groupsWithin subjectsProductContentStructureProd X CntProd X StrCnt X StrProd X Cnt x Str

    CovariatesOrderImportanceKnowledgePrevious purchaseInterestUnderstandability

    df

    8182222226111111

    Mean square

    .08.18

    .00,632.37.04.03.09.03

    .08.02.07.12.01.00

    F

    .32

    .036 .2 1 '23 20".34,33.84.22JB 2.22.691.20.08.01

    Error 30

    Structure Adjustedmeans

    .10

    Content AdjustedmeansPictorial untramed .20a' Objective .77aPictotial Iramed .83b Subjective .72aVerbal ,69b Charac terization .44b

    pp < 0.001.^w o means loltowd by the um e lettef are not signilicanDy diflefsnl from each other.Two means Wtowed by diHefont letters are significanDy difiereni Irom each otfw [p Subjective