Journal of Consumer Research, Inc. Increasing the Persuasiveness of Fear Appeals: The Effect of Arousal and Elaboration Author(s): Punam Anand Keller and Lauren Goldberg Block Source: Journal of Consumer Research, Vol. 22, No. 4 (Mar., 1996), pp. 448-459 Published by: The University of Chicago Press Stable URL: http://www.jstor.org/stable/2489793 . Accessed: 01/04/2013 15:39 Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp . JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected]. . The University of Chicago Press and Journal of Consumer Research, Inc. are collaborating with JSTOR to digitize, preserve and extend access to Journal of Consumer Research. http://www.jstor.org This content downloaded from 150.210.231.20 on Mon, 1 Apr 2013 15:39:53 PM All use subject to JSTOR Terms and Conditions Increasing the Persuasivenessof Fear Appeals: The Effect of Arousal and Elaboration PUNAM ANAND KELLER LAURENG OLDBERG BLOCK* We investigate the conditions under which messages that prompt low and high levels of fear are likely to be effective. Our premise is that when a low level of fear is ineffective, it is because there is insufficient elaboration ,f the harmful conse- quences of engaging in the destructive behavior. By contrast, when appeals arousing high levels of fear are ineffective, it is because too much elaboration on the harmful consequences interferes with processing of the recommended change in behavior. We find support for these expectations in the context of a communication advocating that people stop smoking. The elaboration-enhancing interventions used, self-ref- erence and imagery processing, increased the persuasiveness of a low-fear appeal by prompting elaboration on the harmful consequences of smoking, whereas the use of two elaboration-suppressing interventions, reference to others and objective processing, increased the persuasiveness of a high-fear appeal by decreasing the extent to which consumers deny harmful consequences. W hat is the effect of fear arousal on attitude Two distinctf unctionso f this messagec an be discerned. Al VTchange? Do scare tactics enhance or inhibit One is to state the problem by presentingi nformation message persuasiveness? Despite more than 40 years of about the harmful consequences of the behavior. The research on this issue, an unequivocal answer is not second function is to offera solution comprisedo f rec- possible. In some instances, fear arousal prompts a de- ommended actions that one might take to avoid the crease in persuasion (Hovland, Janis, and Kelley 1953); negative consequences( Hovland et al. 1953). The per- in others, fear arousal enhances persuasion (King and suasivei mpacto f such fear-evokingm essagesi s typically Reid 1990) or results in an increase and then a decrease assessedb y the level of elaborationa nd agreementw ith in persuasion (Janis 1967; Miller 1951). Furthermore, the advocated solution. as Eagly and Chaiken (1993) noted, there is no com- Following this approach, several conditions have pelling account of when and why these outcomes occur. been found to underminee laborationo n a solution for As a starting point for gaining insight about how fear both low and high levels of fear. When the problem is arousal affects persuasion, consider the following mes- not perceived as serious, as is often the case when an sage, which is characteristic of those used in investi- appeal involves a low level of fear, an individual is un- gations of the effect of fear. likely to elaborateo n the solution. For instance, if an accidenti nvolvinga lcohol merelyc auses stiffnessi n the Drinking and driving is dangeroust o your health. You joints and tearing of the skin on one's knees, one is can be paralyzedi nstantaneously.A ll it takes is one sud- unlikely to be motivated to elaborate on the solution den lurchf orwarda nd a quick snappingb acko f the upper (e.g., designate a person to serve as driver). When the torso to sever your spinal cord. You can stop drinkinga nd drivingi n just threee asy steps: harmful consequences are perceived as too severe, as (1) designatea driverb efore you starto ut, (2) hand over in the case of the high-feara ppeali llustrateda bove, one the keys to the designated driver, and (3) include the may engagei n defensived enial of the messageb y deny- designatedd riveri n the festivities.D esignatedd riversc an ing either the existence of a problem( e.g., "I only have have as much fun as the rest of the group. two drinks if I'm driving") or its importance (e.g., "Drinking is not the only cause of traffica ccidents"). *Punam Anand Keller is associatep rofessoro f marketingi n the Again, there is little elaboration on the steps recom- Kenan-FlaglerB usiness School, University of North Carolina at mended in the solution. Chapel Hill, ChapelH ill, NC 27514. LaurenG oldbergB lock is as- The above analysiss uggestst hat an appealg enerating sistant professoro f marketing,S tern School of Business,N ew York low levels of fearw ould benefit by more elaborationo n University,N ew York, NY 10012. the harmful consequences so that recipients are moti- 448 X 1996 by JOURNALO F CONSUMERR ESEARCHI,n c. 0 Vol. 22 * March1 996 All rightsr eserved0. 093-5301/96/2204-0007$2.00 This content downloaded from 150.210.231.20 on Mon, 1 Apr 2013 15:39:53 PM All use subject to JSTOR Terms and Conditions INCREASINGT HE PERSUASIVENESS OF FEAR APPEALS 449 vated to seek a solution. By contrast, a message that high-fear appeal may have prompted an increase in evokes high levels of fearw ould benefitf rom a decrease elaboration of the problems, an assertion that is sup- in elaborationo f the harmfulc onsequencesa nd a focus ported by work in the effects of both self-referencea nd on the solution. To examine these predictions, inter- vividness (Bower and Gilligan 1979; Kisielius and ventionsa re neededt hat can enhance or underminet he Sternthal 1984; Rogers, Kuiper, and Kirker 1977). In level of elaborationo n the harmful consequences and turn, the elaborationo f problemsm ay have limited the thus determinet he effectivenesso f a fear appeal. Rel- processingo f the solutions offered by the message, an evant literaturei s reviewedw ith this goal in mind. assertiont hat is consistentw ith the absenceo f references to the original appeal found in subjects' open-ended responsesa nd the greaters usceptibilityt o counterpro- EFFECTS OF FEAR AROUSAL paganda. Thus, Janis and Feshbach's (1953) findings The impact of high fear arousal appearst o be influ- for the high-fearc ondition suggestt hat high feara rousal encedb y the level of defensivem aneuversp eople engage accompaniedb y substantialp roblem elaborationm ay in when faced with a threateningm essage. These de- undermine persuasion because it interferesw ith elab- fensive techniques may include avoiding the message, oration on the recommendeds olutions. minimizing the severity of the threat, selectively at- Janis and Feshbach's( 1953) results for the low-fear tending the message,d iscountingt he threat, and deny- conditiona lso seem congenialw ith the notion that some ing its personal relevance (Eagly and Chaiken 1993; elaborationo n the problems contained in the message Rogers 1983). Along these lines, Janis and Terwillinger enhances its persuasive impact. Along these lines, it (1962) found that subjects generatedm ore counterar- might be noted that Janis and Feshbach'sl ow-fearm es- guments and had poorer recall of the harmful conse- sage only included three types of problems: cavities, quences of smoking when high rather than low levels tooth fillingsa nd decay, and mouth and gum infections, of fearw eree vokedb y an appeal.T hese findingss uggest whereast he high-fearm essage contained referencest o that "the more strongly fear is aroused by a warning seven types of problems:t oothaches, tooth extractions communication,t he more stronglym otivated the per- and other painful dental work, tooth fillings, mouth son will become to avoid symbolic responses and and gum infections,d iscoloredt eeth, decayedt eeth, and thought sequencesw hich lead him to recall or to focus cavities. Because subjectsw ere given the same amount his attention on the essential content of the arguments of time to processe ach message( 15 minutes), they may and conclusions"( Janisa nd Terwillinger1 962, p. 409). have had the opportunity to elaborate on each of the If this observation is correct and a high level of fear problems described in the low-fear appeal. Janis and arousali nterferesw ith messagep rocessingb y prompting Feshbach'sl ow-fearc ondition also included pictureso f messagea voidance,t he presenceo f an interventiont hat healthy teeth that may have encouragede laborationo f reduces the desire to avoid a high-fear appeal should the solution. This premise is supportedb y the obser- enhance its impact. vation that in relationt o a control messaget hat did not Additionali nsight about how feara rousala ffectsp er- include any health consequences, subjects in the low- suasion is offered in a study conducted by Janis and fear condition were able to provide significantlym ore Feshbach (1953). Three messages were designed to explicit references to the problems and the solutions evoke low, medium, and high levels of fear. This en- listed in the original fear messagea week after message tailed varyingt he numbera nd severityo f problemst hat exposure. Thus, it appearst hat a low-fear appeal may were describeda s arisingf rom poor dental hygiene and be more persuasivei f it is accompaniedb y the oppor- either using or not using personal language and vivid tunity or ability to elaborate on the message-related pictures to describe the problems. Conformity to the problemsa nd solutions. message's recommended toothbrushingp ractices was Although Janis and Feshbach (1953) provide evi- assessedb y having subjectsr eporto n their toothbrush- dence on how an increase in problem elaborationd e- ing behavioro ne week priort o and one week after mes- creases message processing for a high-feara ppeal and sagee xposure.T he findingsi ndicatedt hat only the low- facilitatesm essagep rocessingf or a low-feara ppeal,t hey fear appeal produceda significantg ain in conformity. do not examine the effects of reduced problem elabo- The medium appealg enerateda marginallys ignificant ration for low-and high-feara ppeals. Specifically,t hey increase,w hereast he high-feara ppealp rovedi neffective do not include conditions that reduced problem elab- in increasing conformity. orationf or the high-andl ow-feara ppealst o test whether From our earlier discussion, we suggest that Janis a decrease in problem elaborationw ould enhance the and Feshbach's( 1953) findingsw ere a function of their persuasivenesso f a high-feara ppeala nd reducet he per- appeals'v aryingt he level of probleme laboration.S pe- suasivenesso f the low-feara ppeal.T hese conditionsa re cifically, their high-fear appeal featured personalized tested in the present study. language( e.g., "This can happen to you") and was ac- Supportf or our account of the effects of fear appeal companied by vivid photographso f tooth decay and would be of both theoreticala nd practicali nterest.F rom mouth infections. Our view is that the use of self-ref- a theoretical perspective, confirming evidence would erence and vivid pictures of the consequences in the specify the relationship between problem elaboration This content downloaded from 150.210.231.20 on Mon, 1 Apr 2013 15:39:53 PM All use subject to JSTOR Terms and Conditions 450 JOURNAL OF CONSUMER RESEARCH and persuasion.A lthought here is evidence document- intervention to reduce vividness was sought. Research ing the relationshipb etweene laborationo f the solution reportedb y McGill and Anand (19 89a, 19 89b) suggests and persuasion (Janis and Feshbach 1953; Leventhal that an instruction to image is likely to -enhancee lab- 1970), the relationshipsb etween problem elaboration oration, whereas instructions to process a message in and persuasiona nd between problem elaborationa nd an objective manner and not use imagerya re likely to solutione laborationh ave not been reported.T o provide encourage subjects to be literal and suppresse labora- a test for the mediatingr ole of probleme laboration,w e tion. Theirs tudiess uggestt hat such objectivep rocessing comparet he effect of increasinga nd reducingt he level instructionsp rompt subjectst o rely on the messagei n- of problem elaboration on the persuasivenesso f low- formation itself rather than relate the information to and high-feara ppeals.I n addition, we conduct a series their prior feelings and beliefs about the message con- of mediationt ests to supportt he causald irectiono f the tent. Consistent with this view is the premise that per- relationship from problem elaboration to solution suasion under objectivep rocessingi s determinedm ore elaborationt o persuasion.F rom a practicalp erspective, by learning the message content than the associations our investigationw ill identify elaborationi nterventions connected to the messagec ontent. Thus, we expect de- that are likely to enhancet he persuasivenesso f low-and creased problem elaboration when recipients are in- high-feara ppeals. structedt o think objectivelya bout the problema s being someone else's. We applied these findings to our hypotheses on the VARYING PROBLEM ELABORATION relationshipf rom fear arousal to problem elaboration To test our view, a suitablew ay to manipulatep rob- to persuasion.R ecall that we expected conditions that lem elaborationa nd feara rousalw as needed. The study increasep robleme laborationt o increases olution elab- by Janis and Feshbach (1953) just described suggests oration and thus persuasionf or a low-fear appeal and that self-referencei s a means of varyingp roblem elab- to decrease solution elaboration and persuasion for a oration. The emergingv iew is that information about high-feara ppeal. Thus, the high-feara ppeal should be the self includesa vast arrayo f knowledge( e.g., physical less persuasivet han the low-feara ppeal in the imagery appearance,p ast experiences, behavior patterns, atti- and/or self-referencec onditions. By contrast, we pre- tudinal likes and dislikes,a nd relationshipst owardo th- dicted that a decrease in problem elaboration should ers), that renderst he self a source of one of the richest increasee laborationo f the messages olution and, thus, and most elaboraten etworksi n memory. Becausep eo- persuasionf or a high-feara ppeal and should decrease ple have more knowledge about themselves than they solution elaborationa nd persuasionf or a low-feara p- have about others, events encoded with respect to the peal. Thus, the high-feara ppeal should be more per- self can be made more elaboratet han events encoded suasive than the low-fear appeal in the objective pro- with respect to others (Bower and Gilligan 1979; cessing and/or other-referencec onditions. Burnkranta nd Unnava 1995;G reenwalda nd Pratkanis 1984; Kuiper and Rogers 1979; Rogers et al. 1977). METHOD Thus, we expected greatere laborationw hen the prob- lem was directly relatedt o the self (reader)r athert han Subjects someone else. The experimentalt reatmentsw ere administeredi n a The Janis and Feshbach (1963) study also implies 2 (low fear/high fear) X 2 (self-reference/referenceto that varying the vividness of the stimulus is a viable others) X 2 (imagery/objectivep rocessing) between- means of varying problem elaboration. We chose im- subjectsd esign. Ninety-seven smokers, recruitedf rom agery processing for this purpose because it prompts a pool of students at a large eastern university,p artic- more or less elaborationo f a messagew hile allowingu s ipated in the study, which requiredt hem to reada pam- to hold the content of the messagec onstant (MacInnis phlet on smoking, put it aside, and answer questions and Price 1987). in a booklet. The sample group had smoked for, on The effectso f imageryp rocessingo n elaborationa re average,4 .2 years;7 3 percent of the subjectsh ad tried well documented (Kisielius and Sternthal 1984; to quit at least once, 62 percentc laimedt o have suffered Maclnnis and Price 1987). For example, Brown, from the side effects of smoking, and all of them had Keenan, and Potts (1986) show that imaging oneself heard about a nicotine patch as a possible method to producesg reatere laborationt han imagings omeone else reduce the incidence of smoking. (Bone and Ellen [1992]; for an exception, see Lord [1980]). Thus, we expect even greater elaboration on Stimulus the problem when recipients are asked to imagine themselves sufferingt he consequences of noncompli- In accordancew ith the design to study fear appeal ance. most commonly used in academic research( Eaglya nd To test whether a decrease in problem elaboration Chaiken 1993;H ovland et al. 1953), the firstp arto f the would increase persuasion for a high-feara ppeal and appeal was devoted to the consequences of smoking. reducep ersuasionf or a low-feara ppeal,a n appropriate To manipulate the point of reference, 51 subjects in This content downloaded from 150.210.231.20 on Mon, 1 Apr 2013 15:39:53 PM All use subject to JSTOR Terms and Conditions INCREASINGT HE PERSUASIVENESS OF FEAR APPEALS 451 the self-reference condition were exposed to an ad that We do requestt hat you read the ad carefully,b eing well- employed second-person wording ("Cigarette smoking reasoned and logical to help you get a sense of what is is dangerous to your health"). The remaining subjects beingp resented.D on't let your imaginationg et the better of you. Rather,t ry to objectivelyu nderstandt he content were assigned to the other-reference condition, in which of the ad. the message used third-person wording ("Cigarette smoking is dangerous to the health of those close to Procedure you"). This operationalization departs from the tradi- tional operationalization of reference to others. Rather Subjects were run in large groups with all eight con- than use a manipulation of reference to others that in- ditions randomized within each group. The experiment volves directing the message recipient to change some- took approximately half an hour to complete. The one else's-smoking behavior (Bower and Gilligan 1979; questionnaire was administered after subjects read the Rogers et al. 1977), we sought to change the recipient's pamphlet on smoking. own smoking behavior because of concern for others. To check the level of elaboration, subjects were asked Our operationalization also seemed appropriate because to list their thoughts immediately after message expo- it is more consistent with the other-reference campaigns sure. They were told not to describe what was in the used by marketing practitioners (Hinsberg 1990; Stute- message itself but rather what was going through their ville 1970). minds as they read the message. A person who was blind The manipulation of outcomes in the low- and high- to the experimental conditions classified the thoughts fear appeals was based on those used by the American as related to the problem (e.g., "I didn't know smoking Cancer Association. The low-fear appeal included out- made eyelids droop") or as related to the patch or the comes such as coughing, wheezing, fever, weakness, and solution (e.g., "I don't have any non-hairy areas on my weight loss (see App. A for samples of appeals) that body"). could result in a predisposition for shortness of breath, Subjects next indicated the level of message persua- especially during physical exercise. The high-fear con- sion on the basis of their agreement with the message's dition included more dire outcomes: drooping eyelids, solution. Six seven-point semantic differential attitu- shoulder and arm pain, difficulty in swallowing, and dinal scales were used to measure subjects' estimates of swollen lymph nodes in the neck, resulting in athero- whether they were likely to follow the recommendations sclerosis, which predisposes the sufferer to heart attacks to stop smoking by using the patch, how likely they and possible death. This design, which includes multiple were to buy the patch, whether they would recommend symptoms and consequences, is similar to the one used the patch to a friend, how interested they would be in by Janis and Feshbach (1953) in their dental hygiene learning more about the patch, how likely they were to study. The appeals were also designed to be of equal discuss the patch with a friend, and whether they wanted length (approximately seven health consequences and to receive an additional information sheet on the patch. 112 words in each message). Several additional measures were included as possible The harmful outcomes were followed by recommen- covariates: gender, frequency of behavior, familiarity dations to reduce the incidence of smoking. Again, this with the effects of smoking, familiarity with the actions ordering of problems and recommendations was based one could take to stop smoking, familiarity with other on Hovland et al.'s (1953) design. All the appeals had patches, whether the subject had suffered any health the same set of recommendations: they advocated trial effects from smoking or if anyone they knew had suf- of the Wonder patch (a fictitious name). Our infor- fered health effects from smoking, whether they felt mation about how to apply the Wonder patch was ob- vulnerable to the dangerous effects caused by smoking, tained from companies marketing other nicotine if they had tried to quit before, and whether any mem- patches. Recommendations on how to use and remove bers of their family or close friends smoke. There were the patch were included in the second half of the mes- no significant main effects of the covariates. Although sage (see App. B for the recommendations). there were some significant interactions between the The third between-subjects variable was the type of covariates and the independent variables, these results processing instructions given to subjects. The manip- were not interpretable and thus will not be discussed ulation of imagery and objective processing was adopted further. from McGill and Anand (1989b). Half the subjects re- To assess the adequacy of the fear arousal manipu- ceived the following instructions to engage in imagery lation, subjects indicated the degree to which the ad processing. made them feel very unafraid/very afraid, relaxed/tense, calm/agitated, and restful/excited on a seven-point se- We do requestt hat you read the ad carefully,u sing your mantic differential scale. We also included several ad- imaginationt o help you get a sense of the content of the ditional measures to check the potential effects of the ad. Utilize the power of your imagination to help you level of fear arousal on various affective and cognitive visualize this situation. responses identified in previous research. Affective re- The other half of the subjects received the following sponses were examined by having subjects indicate their objective processing instructions. level of agreement with the extent to which the ad made This content downloaded from 150.210.231.20 on Mon, 1 Apr 2013 15:39:53 PM All use subject to JSTOR Terms and Conditions 452 JOURNAL OF CONSUMER RESEARCH them feel guilty, shameful, angry, remorseful,a nd sad should prompt more problem and solution thoughts on seven-points cales. As none of these emotions were than a high-fear appeal when the message encourages significantlyd ifferenta cross treatments,t hey were not problem elaboration (self-reference, imagery process- considered in subsequent analyses. In the same way, ing). By contrast, a high-fear appeal should be more analysis of several potential cognitive mediators, such persuasive than a low-fear appeal when the message as perceptionsa bout the amount of information con- discourages problem elaboration. This effect should oc- tained in the message, message efficacy,a nd perceived cur because a decrease in focus on the problem should susceptibility,b elievability,a nd credibilityo f the mes- reduce the defensive tendency to avoid the message and sage, indicatedt hat the two levels of fear and the other should increase message elaboration and thus increase independentv ariablesd id not vary on these dimensions persuasion. This prediction would be supported by more (F's< 1).. problem and solution thoughts in the high-fear condi- Point of referencew as examined by means of several tion than the low-fear condition when the message dis- indicators. Self-referencew as inferred on the basis of courages problem elaboration (with reference to others subjects' responses to questions about the extent to and/or objective processing). which the ad helped them to lead a more healthful life and helped prevent them from sufferingb ecause they Manipulation Checks smoke. The other-referencei ndicators were helping others to lead a more healthfull ife and helping to pre- Fear Arousal. The four items used to measure vent others from sufferingb ecause they were in close arousal were averaged and summed to form a fear index proximityt o the smoker. (a = .89). An ANOVA indicated that the treatments Three additional measuresw ere then administered. were effective in creating two different levels of fear Elaboration was assessed by the number of related arousal (X = 1.94, SD = .94 vs. X = 2.59, SD = 1.46; scenes and events generatedw hile subjectsw ere reading F(1,91) = 5.89, p < .05). Thus, despite the fact that the ad (e.g., "I rememberh ow difficulti t was for me to each message included recommendations that were de- breatheb efore my first cigarettee very morning"). Im- signed to reduce fear arousal, there was a significant ageryw as examinedb y askings ubjectso n a seven-point difference in fear arousal for the low- and high-fear ap- scale how easily the messagew as picturedo r imagined. peals. The main effect of point of reference (F(1,91) The extent to which subjectse ngagedi n literal message = 1.20, p = .27), type of processing, and the two- and processingw as determinedb y asking subjectst o recall three-way interactions among fear, type of processing, what they rememberedf rom the smoking pamphlet.A and reference on the fear index were not significant (F's recall score was computed by assigning one point for < 1). rememberinge ach of the four steps outlined in the Reference. An ANOVA indicated that those in the messager ecommendations. self-reference_condition perceived the ad as directed to- Finally,t o evaluatet he possibilityt hat individuald if- ward them (X = 5.08, SD = 1.95) more than those in ferences in ability to image may have influenced the the other-reference condition (X = 4.02, SD = 1.71; imagerye voked by the communication, we asked sub- F(1,94) = 7.29, p < .01) on the self index; those in the jects to complete a shortened version of Bett's Ques- other-reference_condition saw the target of the ad as tionnaire upon Mental Imagery Scale (Bett 1909). someone else (X = 4.33, SD = 1.11) rather than them- Twelvei tems wereu sed to estimatew hetheri ndividuals selves (X = 3.42, SD = 2.14; F(1,94) = 4.16, p < .05) could imagine positive (e.g., "Can you see the car with on the other index. In the same way, there were more a handsome couple inside?"), neutral (e.g., "Can you thoughts about others in the other-reference condition see a car standingi n front of a house?"),a nd negative (X = .48, SD = .65) than the self-reference condition events (e.g., "Cany ou see it get out of control and crash (X = .21, SD = .50; F(1,96) = 5.36, p < .01). The num- through the house?").A s there were no differencesi n ber of thoughts about the self did not differ by treatment subjects'o veralla bility to image acrosst reatments,t his (F < 1). The main and higher-order effects of fear, type measurei s not included in any of the subsequenta nal- of processing, and the interaction between fear, type of yses (F's < 1). processing, and point of reference were not significant on the reference checks (F's < 1). RESULTS Type of Processing. An ANOVA indicated that Overview those in the imagery condition generated more scenes and events (X = 4.24, SD = 1.49) than those in the We predictedt hat the low-feara ppeals houldb e more objective condition (X = 3.29, SD = 1.95; F(1,94) persuasivet han the high-feara ppeal when the message = 4.50, p < .05). By contrast, those in the objective encouragesp robleme laboration.T his effect should oc- condition recalled more (X = 3.00, SD = 1.42) than cur because an increasei n problem elaborations hould those in the imagery condition (X = 1.63, SD = 1.12; enhance elaborationo n the solution and thus persua- F(1,94) = 3.94, p < .05). Subjects in the imagery con- siveness in the low-fear condition. A low-fear appeal dition also indicated on a seven-point scale that the This content downloaded from 150.210.231.20 on Mon, 1 Apr 2013 15:39:53 PM All use subject to JSTOR Terms and Conditions INCREASING THE PERSUASIVENESS OF FEAR APPEALS 453 TABLE 1 EFFECTS OF FEAR AROUSAL AND POINT OF REFERENCE ON PERSUASION, MESSAGE ELABORATION,A ND RECALL FOR IMAGERYA ND OBJECTIVE-PROCESSINGC ONDITIONS Imagery processing Objective processing Reference to others Self-reference Reference to others Self-reference Low fear High fear Low fear High fear Low fear High fear Low fear High fear Persuasion 2.80 3.82 *4.51 3.67 3.98 4.95 3.17 3.94 (1.06) (.99) (1.52) (.97) (1.31) (1.32) (1.66) (1.12) Problem elaboration .73 1.78 2.67 .69 .69 .77 .62 1.08 (.79) (1.09) (.87) (.75) (.86) (1.63) (.65) (1.13) Solution elaboration .18 1.33 2.44 .77 .46 2.18 .31 .77 (.41) (1.58) (1.13) (.73) (.52) (1.29) (.48) (1.07) Solution recall 1.64 1.77 1.33 1.77 2.46 1.71 1.77 1.85 (1.1 0) (1.83) (1.23) (1.03) (1.33) (1.1 1) (.83) (1.41) n 11 10 12 13 13 13 13 13 NOTE.-Values shown are means (?SD). message was more easily pictured or imagined (X = these predictions,t he interactionb etweenr eferencea nd 4.5 1, SD = 2.19) than subjectsi n the objectivec ondition fear was significanto n persuasion( F(1,90) = 10.26, p (X = 3.69, SD = 2.12; F(1,94) = 3.42, p < .05). An < .01), problem thoughts (F(l,90) = 10.94, p < .01), ANOVA indicated no other significantd ifferenceso n and solution thoughts (F(l,90) = 20.40, p < .01). No these two imagery measures or on the message recall significante ffects were obtained for the recall measure measurea cross treatments( all F's < 1, except for the (F's < 1). type of processing X reference interaction, for which Inspectiono f the attitudea nd thoughtsd atai ndicated F(1,94) = 1.28, p = .24). that subjectsi n the other-referencceo ndition responded differentlyt o the high- and low-fear appeals. As the graph of the imagery condition in Figure 1 (panel A, Hypotheses Tests left) indicates, in the presence of the other-reference interventiont o reduce problem elaboration,t he high- The level of persuasion, the number of problem fear appeal was more persuasivet han the low-fear ap- thoughtsa nd solution thoughts,a nd the extent of recall peal'(F(1,90) = 6. 10, p < .05). In the same way, panels categorizedb y treatment are shown in Table 1. The B and C indicatet herew erem ore problema nd solution six items relatedt o subjects'a ttitudest owardt he patch thoughts for the high- than the low-fear appeal in the were averagedf or each subjectb ecause they loaded on other-reference condition (for problem thoughts, one factor and formed a reliable persuasion index (a F(l,90) = 6.81, p < .05; for solution thoughts, F(l,90) = .90). An ANOVA indicatedt hat the three-wayi nter- = 7.13, p < .05). action betweenf ear,p oint of reference,a nd type of pro- A differentp attern of outcomes was observedi n the cessing was significant on the persuasion measure self-referencec ondition,w hich was expectedt o enhance (F(1,90) = 5.10, p < .05), problem thoughts (F(1,90) problem elaboration.A n increase in problem elabora- = 10.67, p < .01), and solution thoughts (F(1,90) tion was expected to increase the desire to avoid the = 3.95, p < .05). These outcomes are shown in Fig- high-fearm essage and stimulate the processingo f the ure 1. low-feara ppeal. The data depicted in Figure 1 are con- sistentw ith these expectations.T he low-feara ppealw as Imagery Processing. Our theorizing suggests that marginallym ore persuasivet han the high-feara ppeal for a high-feara ppeal, the presence of an intervention in the self-referencec ondition,w hich enhancesp roblem that underminedp roblem elaboration, such as other- elaboration( F( 1,90) = 2.19, p < .07). In addition, there reference,w ould enhance messagep rocessinga nd per- were more problem and solution thoughts for the low- suasion because subjects would be less likely to avoid feara ppealt han the high-feara ppeali n the self-reference the message.B y contrast,a reduction in problem elab- condition (for problem thoughts, F(1,90) = 6.30, p oration for the low-feara ppeal would undermine mes- < .05; for solution thoughts, F(l,90) = 16.16, p < .05). sage processinga nd favorablea ttitudes because an in- consequential message would be rendered even more Objective Processing. We expected the same pattern inconsequential.T his premise would be supportedb y of results in the objective processingc ondition as was more favorablea ttitudes and more problem and solu- obtained in the imagery condition. Specifically, the tion thoughtsf or the high-feara ppealt han the low-fear other-referencei ntervention was expected to enhance appeali n the other-referencceo ndition. Consistentw ith messagep rocessinga nd persuasionf or the high-feara p- This content downloaded from 150.210.231.20 on Mon, 1 Apr 2013 15:39:53 PM All use subject to JSTOR Terms and Conditions FIGURE 1 INTERACTIONB ETWEEN FEAR AND POINT OF REFERENCE ON PERSUASION, PROBLEM, AND SOLUTION ELABORATION IN THE IMAGERYA ND OBJECTIVEP ROCESSING CONDITIONS Imagery Objective PANELA 5.0 LowF ear 5.0 4.0 4.0 HighF ear Persuasion HighF ear 3.0 3.0 LowF ear 2.0 2.0 1.0 1.0 Other Self Other Self Reference Reference PANELB 3.00 3.00 LowF ear 2.50 2.50 2.00 2.00 ProblemE laboration 1.50 1.50 HighF ear 1.00 1.00 .50 .50 LowF ear HighF ear .00 . .00 Other Self Other Self Reference Reference PANELC 2.50 LowF ear 2.50 2.00 2/00 1.50 1.50 SolutionE laboration 1.00 1.00 .50 HighF ear .50 HighF ear LowF ear .00 .00 Other Self Other Self Reference Reference This content downloaded from 150.210.231.20 on Mon, 1 Apr 2013 15:39:53 PM All use subject to JSTOR Terms and Conditions INCREASINGT HE PERSUASIVENESS OF FEAR APPEALS 455 peal and reducem essagep rocessinga nd persuasionf or our findings in conditions that encourage elaboration the low-feara ppeal.B y contrast,t he self-referencec on- versust hose conditions that suppresse laboration.S pe- dition was expectedt o enhance messagep rocessinga nd cifically, seven of our eight conditions 'encourageda persuasionf or the low-feara ppeala nd underminem es- little or a lot of elaboration by evoking imagery pro- sagep rocessinga nd persuasionf or the high-feara ppeal. cessing, self-reference,a nd/or high fear. The patterno f The pattern of results obtained for the objective con- problem and solution thoughts is consistent with per- dition is presentedi n the right-handc olumn of Figure suasion in these conditions. By contrast, the low-fear, 1. An ANOVA indicated a significant interaction be- other-referenceo, bjectivep rocessingc ondition did not tweenr eferencea nd fearf or solutiont houghtsa nd recall include any elaborationm echanisms.I n the absenceo f (for solution thoughts, F( 1,90) = 6.30, p < .05; for recall, any interventiont o prompt elaboration,s ubjectss eem F(1,90) = 1.71, p < .05). The predicted interaction be- to have relied on their recall of the advocateds olution tween referencinga nd fear was not significantf or per- to form a persuasionj udgment. suasion or problem thoughts (F's < 1). Instead, there The pattern of correlationsb etween recall, solution weret wo main effectsf or persuasion;t he other-reference elaboration,a nd persuasioni s consistentw ith this view. message was more persuasive than the self-reference The correlationb etween recall and persuasioni s insig- message (F( 1,90) = 11 .97, p < .05), and high fear arousal nificant in all seven conditions that encouragea t least was more persuasive than low fear arousal (F(1,90) some elaboration (p's > .10) but not in the objective = 9.73, p < .05). There were no significant effects ob- processing,o ther-referencel,o w-fearc ondition (r = .64, tainedf or probleme laboration( for fear,F ( 1,90) = 1.43, p < .08). These outcomes are in contrastt o the signif- p > .10; all other F's < 1). icant correlations between solution elaboration Although the finding that the high-feara ppeal was thoughts and attitudes in the seven elaboration-en- more persuasivet han the low-feara ppeal in the other- hancing conditions (r = . 36, p < .05) but not in the referencec ondition was anticipated,t he lower persua- objectivep rocessing,o ther-referencel,o w-fearc ondition sion of the low-feara ppealt han the high-feara ppeal in (r = .05, p > .70). The implications of relyingo n recall, the self-referencec ondition was not. Our theorizing which is memoryb ased, versuse laboration,w hich may predictedt hat the low-feara ppeal would be more per- be consideredm ore as on-line processing,i s addressed suasive than the high-feara ppeal in the self-reference in more detail in the discussion section. condition. The significantr eferenceX fear interaction for solution thoughtsa nd recall enabled us to compare Mediation Analyses the high- and low-fear appeals in the self-and other- referencec onditions in order to understandw hy per- A series of path analyses were undertakent o deter- suasion was higher than anticipated for the low-fear mine whether problem elaboration mediates the rela- other-referencem essagea nd lower than anticipatedf or tionship between fear and persuasiona nd whether so- the low-fears elf-referencem essage. lution elaboration mediates the relationship between The differencesi n the pattern of solution thoughts problem elaboration and persuasion. In addition, we and recall offer some insight about the nature of per- tested an alternative model to examine whether a suasion in the objective processingc ondition. Specifi- change in problem elaboration had an effect on fear cally, we found that recipientse laboratedm ore on the and persuasion. Unlike the first model, in the latter solution in the other-referencec ondition when exposed model the effect of problem elaborationo n persuasion to the high-t han the low-fearm essage( F( 1,90) = 23.52, would be mediated by fear arousal. p < .05). By contrast,s ubjects's olution recallw as higher We used Sobel's procedurea s reportedi n Baron and when exposed to the low- than the high-fearm essage Kenny (1986) to test the models outlined above. Four (F(1,90) = 2.76, p < .05). In the same way, solution regressione quations are tested for each model. For ex- thoughts were higher for the high-fear appeal in the ample, the relationship from fear arousal to problem other-referencec ondition than the self-referencec on- elaborationt o persuasionw ould be supportedb y (1) a dition (F( 1,90) = 15.84, p < .05), whereas solution recall significante ffecto f feara rousalo n probleme laboration, was higherf or the low-feara ppeali n the other-reference (2) a significante ffect of problem elaboration on per- condition than the self-reference condition (F(1,90) suasion, (3) a significante ffect of fear arousal on per- = 2.04, p < .05). suasion, and (4) an insignificante ffect of fear arousal These contrastingp atternso f recall and elaboration on persuasion( full mediation)o r a less significante ffect indicate that persuasion is based on recall, not elabo- than the resultsi n Equation3 (partialm ediation)w hen ration, in conditions that encourage literal message the effect of problem elaborationi s partialedo ut. learninga nd discouragee laboration( i.e.; objectivep ro- Table 2 outlines the four equations. Column 1 pro- cessing, referencet o others, and low fear). By contrast, vides the standardizedc oefficientsa nd their significance elaboration determines persuasion under conditions values for the mediating effect of probleme laboration. that encourages ubjectst o relate message information Column 2 repeatst he analysisi n column 1 for the me- to prior feelings and beliefs (i.e., imagery processing, diatinge ffecto f total messagee laborationo n persuasion. self-referencea, nd high fear).T his view can account for Column 3 presentsa separatet est on the mediatingr ole This content downloaded from 150.210.231.20 on Mon, 1 Apr 2013 15:39:53 PM All use subject to JSTOR Terms and Conditions 456 JOURNAL OF CONSUMER RESEARCH TABLE 2 TABLE 3 PATH ANALYSIS OF THE EFFECT OF FEAR AROUSAL PATH ANALYSIS OF THE EFFECTS OF PROBLEM AND ON PERSUASION AND ITS MEDIATION BY PROBLEM, TOTAL MESSAGE ELABORATION ON PERSUASION TOTAL MESSAGE, AND SOLUTION ELABORATION AND THEIR MEDIATION BY FEAR AROUSAL Problem Total message Solution Problem Total message elaborationa elaboration elaboration elaborationa elaboration Fear elaboration .24 .21 .07* Elaboration fear .24 .21 Elaboration -- persuasion .23 .30 .23 Fear -- persuasion .16* .16* Fear persuasion .22 .22 .22 Elaboration persuasion .27 .33 Fear persuasion Elaboration persuasion (without elaboration) .16* .16* .20 (without fear) .23 .30 aValues shown are standardized regression coefficients. aValues shown are standardized regression coefficients. *p > .10 (all other p's < .05). *p > .10 (all other p's < .05). of solution elaboration.T his analysisw as conductedt o suasion. This model is based on the premise that an rule out the possibilityt hat therei s a directl ink between increasei n probleme laborationf or the low-feara ppeal fear arousala nd solution elaboration. may increase the level of fear arousal, which in turn The results indicate that both problem elaboration would increasep ersuasion.I t is also basedo n the notion and total message elaboration are significantf ull me- that a reductioni n probleme laborationw ould decrease diators of the relationship between fear arousal and fear arousal and thus increasep ersuasionf or the high- persuasion.F ear arousal does not have a direct effect fear appeal. For purposeso f comparisonw ith Table 2, odnp ersuasion. Rather, it causes elaboration on the we conducted an analysiso n probleme laboration( col. message,a nd particularlyt he problem part of the mes- 1) and total message elaboration( col. 2). These results sage,w hich in turn determinesp ersuasion.I n addition, are presentedi n Table 3. this analysisi ndicates that elaborationo n the solution The results in Table 3 provide support for the as- alone is not sufficientt o determinet he relationshipb e- sumption that a change in elaborationo f problem and tween fear and persuasion.T he results in row 1 of the total message elaborationd oes not affect fear arousal. solution elaboration column in Table 2 indicate that Two dimensions of the mediation tests remain unsat- fear arousal does not have a direct link with solution isfactory;t he effect of fear arousal on persuasioni s in- thoughts (p > .50), thus failing the first requirement significantf or problem elaborationa nd total message for the mediation test. These resultsa re also consistent elaboration( Eq. 2) and the effecto f probleme laboration with the view that problem elaboration must precede (# = .23, p < .03) and total messagee laboration( ,B= .30, solution elaboration. p < .01) on persuasionr emainss ignificant( Eq. 4). The Furthers upportf or the effecto f the ordero f problem fact that the relationship from elaboration to fear and solution elaboration on persuasion was obtained arousal to persuasion is not supportedp rovides addi- by testing whether solution elaboration mediates the tional support for the view that fear arousali s respon- relationshipb etween problem elaborationa nd persua- sible for generatingp roblem and solution elaboration, sion. The analysis indicated that solution elaboration which in turn determinesp ersuasion. partially mediates the relationship between problem elaborationa nd persuasion.A pplying Baron and Ken- DISCUSSION ny's (1986) procedure,w e found (1) a significante ffect of probleme laborationo n solutione laboration( f = .21, It is widely recognizedt hat persuasioni s a function p < .05), (2) a significante ffect of solution elaboration of fear arousala t the time a judgment is renderedf or a on persuasion( f = .23, p < .05), (3) a significante ffect threateningc ommunication. Our findings extend pre- of problem elaboration on persuasion (f = .27, p < vious work in this area by identifyingh ow fear arousal .01), and (4) a less significant (for partial mediation) affectsp ersuasion.S pecifically,w e identify the process effect of problem elaborationo n persuasionw hen the underlyingt he interferinge ffects of low and high fear effect of solution elaborationi s partialledo ut (f = .19, arousal on persuasion. p < .08 vs. A = .27, p < .0 1). These resultsi ndicate that Our results support the premise that a reduction in elaborationo n the problemd eterminese laborationo n the level of probleme laborationr educese laborationo f the solution, which in turn determinesp ersuasion. the problem,s olution,a nd the favorablenesso f attitudes A third mediatingt est for the relationshipf rom elab- for a low-fear appeal, whereas the same decrease in oration to fear arousal to persuasion was designed to probleme laborationi ncreasesm essagee laborationa nd rule out the possibility that an increase (decrease) in persuasionf or a high-feara ppeal.B y contrast,p roblem elaboration,e speciallyp roblem elaboration,w ould in- elaboration that increases elaboration of the solution crease (decrease)f ear arousal and thus determine per- and persuasive impact for a low-fear appeal reduces This content downloaded from 150.210.231.20 on Mon, 1 Apr 2013 15:39:53 PM All use subject to JSTOR Terms and Conditions
Description: