LITERATURE REVIEW (12-16 pages) This essay is basically synthesizing 10 sources to support the new research im proposing that needs to be done. The sources needs to be categorized into 3/4 subheadings

Influencing light versus heavy engagers of harmful behavior to curb their habits through positive and negative ad imagery JENNIFER L. BURTON 1*, DONNA J. HILL 2and AYSEN BAKIR 3 1 Phillips School of Business, High Point University, 833 Montlieu Avenue, High Point, NC 27262, USA2Foster College of Business, Bradley University, 1501 West Bradley Avenue, Peoria, IL 61625, USA3College of Business, Illinois State University, Campus Box 5590, Normal, IL 61790-5500, USA ABSTRACT Mass media advertising has increasingly been seen as an important vehicle to influence attitudes and behaviors on issues designed to en- hance the well-being of society overall. However, prior research shows mixed results on the ability of advertising to encourage substantial or long-term changes in self-destructive behavior. The current research provides a framework to reconcile previousfindings and demon- strates that different psychological processes are occurring when trying to convince light versus heavy engagers to curb their potentially harmful behaviors.

Validated across two contexts (texting while driving and excessive gambling), the present study demonstrates that ads containing pos- itive imagery (as opposed to negative imagery) are more effective at influencing heavy engagers to limit their harmful behaviors. Heavy engagers in the harmful behavior seem to resist ads containing negative imagery and do notfind them to be more credible or involving.

Instead, they are more influenced to curb their harmful behaviors after seeing positive imagery containing characters and situations to which they can relate.

On the other hand, light engagers of harmful behavior are more likely to be persuaded to limit their behaviors after exposure to adver- tising containing negative imagery. The negative imagery evokes perceptions of ad credibility and encourages individuals to think about the advertised problem and consequences. Both credibility and ad involvement are shown to be important mediators of ad effectiveness for light engagers, whereas relatability was not a significant mediator for this group. Implications for the design of effective public service announce- ments targeted toward these separate groups are discussed. Copyright © 2015 John Wiley & Sons, Ltd. INTRODUCTION Mass media advertising (e.g., print, radio, and television) has increasingly been seen as an important vehicle to influence attitudes and behaviors on issues designed to enhance the well-being of society overall. Because of their demonstrated propensity to broaden awareness of important social issues and encourage pro-social behaviors, research on the impact of public service announcements has increased considerably in recent years (Keller and Lehmann, 2008). While public service announcements typically have honorable and socially beneficial objectives, prior research shows that they are often ineffective at encouraging substantial or long-term changes in self-destructive behavior (Midford and McBride, 2001; Foxcroftet al., 2003) and have the greatest impact on people less likely to engage in the harmful behavior (Kellyet al., 1996). One reason given for the ineffectiveness of advertis- ing to influence individuals engaging in harmful behavior is the third-person effect, where individuals feel that depicted consequences of a harmful behavior are more likely to happen to other people rather than themselves (Golan and Banning, 2008; Grier and Brumbaugh, 2007; Raghubir and Menon, 1998). Another reason for the ineffectiveness of ad- vertising to influence people potentially engaging in harmful behavior is the boomerang effect. The boomerang effect is when public service announcements cause an increased like- lihood of engaging in socially destructive behavior becausethe ad sensationalizes the experience for target audiences looking for ways to seek stimulation (Fishbeinet al., 2002a, 2002b; Zhaoet al., 2006).

Our study reconciles some of the conflictingfindings on the effectiveness of ads designed to curb potentially harmful behaviors by suggesting that different psychological pro- cesses are occurring for individuals who are heavy versus light engagers in harmful behaviors. The current study uti- lizes a meditational approach to suggest that ads containing positive imagery will be more effective for individuals who are heavily engaged in the harmful behavior and ads contain- ing negative imagery will be more effective for individuals who are light engagers in the harmful behavior. We study three important mediators of ad effectiveness and test the no- tion that individuals who do not frequently engage in the harmful behavior experience higher levels of persuasion when negative ad imagery evokes perceptions of ad credibil- ity and encourages individuals to think about the issue and the associated consequences. On the other hand, we test the idea that positive ad imagery evokes high levels of relatability, which is an important factor in influencing heavy engagers of the harmful behavior to curb their bad habits. We examine these hypotheses in two separate contexts for gener- alizability: texting while driving (Study 1) and excessive gambling (Study 2).

The results of this study have important implications for managers who use advertising to encourage positive social behaviors. The goals of many public service campaigns are often twofold: (1) to encourage heavy engagers to obtain help and curb their habit and (2) to encourage light and non-engagers from developing a bad habit in thefirst place. * Correspondence to: Jennifer L. Burton, Phillips School of Business, High Point University, 833 Montlieu Avenue, High Point, NC 27262, USA.

E-mail: [email protected] Copyright © 2015 John Wiley & Sons, Ltd.

Journal of Consumer Behaviour,J. Consumer Behav.14: 237–247 (2015) Published online 7 April 2015 in Wiley Online Library (wileyonlinelibrary.com)DOI:10.1002/cb.1514 Oftentimes, these two goals are pursued simultaneously with the same advertising stimuli. We hypothesize that heavy ver- sus light engagers do respond differently to positive versus negative stimuli and exhibit different message processing strategies that influence persuasion. If this proves to be the case, advertisers would be more likely to affect social change by having separate campaigns targeted toward heavy versus light engagers of harmful behaviors.

HYPOTHESES AND THEORETICAL DEVELOPMENT A main premise of this study is that individuals who fre- quently engage in a harmful behavior are more likely to curb their habit after viewing an ad containing positive imagery, while individuals who are light engagers of a harmful behav- ior are more likely to curb their habit after viewing an ad containing negative imagery. Numerous researchers have found differential effects of positive and negative stimuli (see Baumeisteret al., 2001; Rozin and Royzman, 2001 for reviews). For instance, meat products are described more favorably by consumers when they are advertised with a positive frame of reference (i.e., containing 75% lean beef) as opposed to a negative frame (i.e., containing 25% fat) (Levin and Gaeth, 1988). On the other hand, Meyerowitz and Chaiken (1987) report that public service announce- ments promoting breast self-exams were more effective with a negatively framed message. In the context of public service announcements designed to influence individuals to check their risk for heart disease, Maheswaran and Meyers-Levy (1990)find that negative messages are considered to be more informative and effective when individuals have the motiva- tion to engage in higher levels of message processing.

Prior research provides some evidence to suggest that light engagers of a harmful behavior may be more likely to engage in higher levels of message processing. For example, Mandel and Johnson (2002) suggest that individuals with less experience and knowledge in a particular context (e.g., light engagers of harmful behavior) are more likely to rely on external sources of information in forming their overall evaluations. Additionally, these individuals are more likely to carefully assess and examine all presented informa- tion in making their decisions (Chiet al., 1982; Mitchell and Dacin, 1996) where they are prone to the bias of weighing negative information more than equally valenced positive information (Fiske, 1980; Maheswaran and Meyers-Levy, 1990). In the context of drugs, individuals who had the least experience with and intention to do drugs were more likely to believe in the negative consequences of drug use and less likely to believe that drug use had any positive consequences compared with heavy users (Fishbeinet al., 2002a, 2002b).

Because of the importance of aligning the message with the attitudes and beliefs of the target audience, we predict that in- dividuals who do not frequently engage in harmful behaviors will be more persuaded by ads containing negative imagery.

On the other hand, consumers with extensive prior knowl- edge and experience (e.g., heavy engagers of harmful behavior) are more selective in their information search (Bettman and Park, 1980) and focus on information that is consistent withtheir goals (Lewandowsky and Kirsner, 2000). This group is morelikelytorelyoninternalversusexternalinformation search in decision making (Mandel and Johnson, 2002), and exhibit a greater tendency to make quick decisions (Thunholm, 2005). Thesefindings suggest that heavy engagers in harmful behavior are more likely to use their prior knowledge and experience in forming judgments about their future behavior.

They are also more likely to respond positively to imagery that represents their overall goals (i.e., the positive reasons why they engage in the harmful behavior). Indeed, prior research illus- trates that individuals who frequently use drugs are more likely to believe in the positive outcomes of drug use and less likely to believe in the negative outcomes associated with drug use compared with those non-users or infrequent users (Fishbein et al., 2002a, 2002b). Additionally, Wolburg (2006) shows that while non-smokers generally respond well to anti-smoking public service announcements, smokers generally responded with denial and even defiance. Research suggests that the best approach for targeting smokers is a message that is non- judgmental (McKenna and Williams, 1993) and includes pleasing elements such as humor or music (Grubeet al., 1996). Therefore, wepredict that ads featuring positive imagery will be more effective at persuading heavy engagers of harmful behavior to curb their habits. In summary, we predict that there will be a significant interaction between ad imagery (positive versus negative) and the extent of prior harmful behavior on the likelihood the ad influences individuals to limit their harmful behavior. Formally, H1A: The ad containing positive imagery will be more effective at influencing heavy engagers to limit their behaviors; and H1B: The ad containing negative imagery will be more effec- tive at influencing light engagers to limit their behaviors.

The next set of hypotheses are based on prior research on the “ dual mediation”hypothesis that suggests that attitude toward the ad has both a direct relationship with attitude to- ward the brand and an indirect relationship through brand cognitions (MacKenzie and Lutz, 1989; MacKenzieet al., 1986). The only difference is that, in our case, we are not interested in attitude toward the brand but rather the extent to which the ad persuaded individuals to limit a potentially harmful behavior (i.e., texting while driving and excessive gambling). Therefore, we decided to concentrate on three as- pects of attitude toward the ad that a meta-analysis on public service announcements has shown to be important in gener- ating compliance with health recommendations: credibility, ad involvement, and relatability (Keller and Lehmann, 2008). These mediating variables are also expected to work differently for individuals who have high versus low levels of participation in the harmful behavior.

We define credibility as the extent to which individuals find an advertisement to be believable, informative, persua- sive, and trustworthy. As previously mentioned, ads evoking negative imagery are generally considered to be more credible because individuals tend to weigh negative information more than equally valenced positive information in forming evalua- tions (Fiske, 1980). This tendency has also been shown in the 238J. L. Burtonet al. Copyright © 2015 John Wiley & Sons, Ltd.J. Consumer Behav.14: 237–247 (2015) DOI: 10.1002/cb context of public service announcements promoting choles- terol screening (Maheswaran and Meyers-Levy, 1990), breast self-examinations (Meyerowitz and Chaiken, 1987), and smoking cessation (Agostinelli and Grube, 2003). Addition- ally, prior research has shown credibility to be an important predictor of attitude toward the brand (MacKenzie and Lutz, 1989) and positive health behaviors in public service announcements (Keller and Lehmann, 2008). Therefore, we predict that the extent to which individualsfind the ad to be credible mediates the relationship between ad imagery (positive versus negative) and the likelihood the ad influences individuals to limit their harmful behaviors. Formally, H2: Individuals will perceive the ad with negative imagery to be more credible, influencing them to limit their harmful behaviors.

In addition to credibility, the extent to which an ad generates issue involvement is also predicted to be an important deter- minant of an individual’s likelihood to curb harmful behav- iors. We define issue involvement as the extent to which an ad motivates an individual to think about the featured prob- lem (e.g., texting while driving and excessive gambling) and associated consequences. Prior literature suggests that negative content in public service announcements mobilizes and attracts physiological, cognitive, and social resources to a greater extent than does positive content in anti-piracy messages (Levinet al., 2007), as well as anti-verbal abuse, anti-smoking, anti-terrorism and, acquired immune deficiency syndrome prevention messages (Parryet al., 2013). The ability of these messages to mobilize individuals’physiolog- ical, cognitive, and social resources results in higher levels of issue involvement (Pratto and John, 1991). Research on persuasion suggests that higher levels of involvement lead to the formation of attitudes that are stronger, more stable, and persistent over time (Petty and Cacioppo, 1986). Addition- ally, a recent meta-analysis by Keller and Lehmann (2008) illustrates that emphasizing the consequences of harmful behaviors in public service announcements is an important aspect in generating compliance with health recommenda- tions. Therefore, we predict that the extent to which individ- ualsfind the ad to generate issue involvement mediates the relationship between ad imagery (positive versus negative) and the likelihood the ad influences individuals to limit their harmful behaviors. Formally, H3: Individuals will perceive the ad containing negative imagery as creating more issue involvement, thus influencing them to limit their harmful behaviors.

Another important factor in generating persuasion is for the ad to contain characters and situations individuals can relate to (i.e., relatability). However, in this case, we believe that the ad containing positive imagery will be more relatable to individuals when it comes to persuading them to curb their texting while driving or excessive gambling habits. This pre- diction is based on research on public service announcements promoting safe sex where participants described the most persuasive messages as those featuring characters andsituations to which they could relate (Van Steeet al., 2012). Additionally, research on attitudes of smokers show that heavy smokers are much more likely to believe that their behavior is associated with positive outcomes and less likely to believe that their behavior could be associated with nega- tive outcomes (Fishbeinet al., 2002a, 2002b). Such tenden- cies lead to what is known as the third-person effect (Golan and Banning, 2008; Grier and Brumbaugh, 2007) that states that when individuals are confronted with negative situations in ads, they are more likely to think the negative conse- quences will apply to other people more than themselves.

The third-person effect has been illustrated in public service announcement research in many contexts such as those soliciting charitable donations (Golan and Banning, 2008), promoting safe sex (Raghubir and Menon, 1998; Van Stee et al., 2012), and encouraging individuals to quit alcohol, tobacco, and gambling (Banning, 2001). Additionally, re- search has shown the third-person effect to be less pro- nounced when a message is positively framed (Golan and Banning, 2008) and the characters and situations are similar to the viewer and his/her experiences (Raghubir and Menon, 1998). For these reasons, we predict the following:

H4A: Individuals will perceive the ad containing positive imagery to be more relatable than the ad containing negative imagery.

H4B: The extent to which individualsfind the ad to be relat- able will predict the extent to which the ad influences them to limit their harmful behaviors.

Previous research has shown that individuals display variation in their emotional responses to affect laden stimuli whether they are images (Larsen and Diener, 1987) or advertisements themselves (Moore and Harris, 1996). In general, advertising appeals that are self-relevant and activate an individual’sself- schema are considered more effective, and this effect is even stronger for advertisements featuring negative messages (Meyers-Levy and Peracchio, 1996). This is consistent with re- search by Wolburg (2006) that suggests that anti-smoking pub- lic service announcements are better received by non-smokers than smokers because the messages are more consistent with their current ideals and beliefs. It is also consistent with research by Van Steeet al. (2012) that illustrates that public service announcements featuring the negative consequences of unsafe sex practices are more effective when they feature characters and situations to which the viewers can relate.

Public service announcement research across many con- texts such as smoking (Van Steeet al., 2012) and alcohol, tobacco, and gambling (Banning, 2001)finds that individuals may respond with denial or defiance when they are threatened by a message. In the context of drinking and driving public service announcements, Block (2005) demonstrates that indi- viduals react differently depending on whether they feel guilt or resonate with the message. This research has identified two paths explaining how individuals might deal with guilt. These two paths are described as“guilt reduction”and“guilt avoid- ance.”For example, we would expect the light engagers of harmful behavior to pursue a guilt-reduction strategy where an ad would generate feelings of credibility and involvementGenerating positive change through ad imagery239 Copyright © 2015 John Wiley & Sons, Ltd.J. Consumer Behav.14: 237–247 (2015) DOI: 10.1002/cb to motivate them to limit their texting while driving or gam- bling behaviors. This strategy is reflected in a number of public service announcement studies that show individuals tak- ing positive measures to prevent harm in the contexts of choles- terol screening (Maheswaran and Meyers-Levy, 1990), breast self-examinations (Meyerowitz and Chaiken, 1987), and smoking cessation (Agostinelli and Grube, 2003). On the other hand, we would expect the heavy engagers of harmful behavior to pursue a guilt avoidance strategy where individuals con- sciously try to avoid thinking about guilt-evoking incidents (Kubany and Watson, 2003) upon exposure to the ad featuring negative imagery. Thus, these individuals would not likely ex- perience the same levels of perceived credibility or ad involve- ment as the light engagers. The guilt avoidance strategy is one that is reflected in a number of studies that show denial or defi- ance when faced with guilt-provoking content in public service announcements dealing with quitting smoking (Van Steeet al., 2012) and alcohol, tobacco and gambling (Banning, 2001).

Prior literature suggests that negative or positive appeals can be successful at achieving effective attitude change depending on the individual’s aroused level of emotional tension (Wheatley and Oshikawa, 1970). For example, if the individual’s emotional tension is stimulated by a negative appeal, the individual who is exposed to the message will try to reduce his or her anxiety level by looking for some type of comfort. Then, if the message in the ad is perceived to be offering some comfort, it may decrease the emotional tension for the individual and persuade the individual to follow a rec- ommended course of action (Wheatley and Oshikawa, 1970).

This is similar to what Keller (1999) found with regard to safe-sex public service announcements. He found that threatening content in advertisements has facilitating effects of making the person wonder if he/she is vulnerable to the threat and think about whether the recommendations would protect that person. He also found that too much threat in advertisements has debilitating effects where the person doubts the effectiveness of the health recommendations and his/her ability to follow them (Keller, 1999). There- fore, individuals who are frequently engaged in harmful behavior might look for cues in the message that decrease the emotional tension for themselves. This, in turn, might de- crease their level of involvement in the ad and how credible they think the message is when the ad contains negative imag- ery. On the other hand, individuals who do not frequently engage in harmful behavior might not feel any guilt leading to higher perceptions of credibility and involvement because they do not need to reduce their emotional tension because of their exposure to the negative imagery. Therefore, it is hypothesized that H5: Individuals who frequently engage in the harmful behav- ior do not consider the ad containing negative imagery to be more credible or involving. However, they do consider the ad containing positive imagery to be more relatable.

H6: Individuals who do not frequently engage in the harmful behavior consider the ad containing negative imagery to be more credible and involving. However, they do not consider the ad containing positive imagery to be more relatable.STUDY 1 Participants and procedure Study 1 was designed to test the aforementioned hypotheses by exposing individuals to advertisements designed to curb texting while driving behaviors. One-hundred eighty students from two Midwestern universities participated in this study in exchange for credit in their introductory business courses. A between-subjects design was used to test the degree to which positive and negative imagery used in the advertisements influenced individuals to limit their texting while driving behaviors.

Subjects were randomly assigned to one of two conditions where we manipulated the extent to which the public service announcement featured positive or negative imagery. The ad featuring positive imagery showed a smiling young man and a smiling young lady driving their car while attempting to type out a text message on their cell phone. The positive im- agery was meant to express the positive reasons that people engage in the harmful behavior. The reason the positive imagery reflected the positive reasons people engage in the behavior as opposed to the positive benefits of not engaging in the behavior is because previous literature emphasizes the importance of highlighting situations and characters to which viewers can relate (Grubeet al., 1996; Van Steeet al., 2012).

Also, the benefits of not engaging in harmful behaviors are similar to double negatives (e.g., not getting cancer and not getting into an accident) than actual positives. Finally, the main positive benefits of living long, healthy, and injury-free lives are often arbitrary and vary greatly by individual. The ad featuring negative imagery showed a dangerous crash with a person being run over and blood on the windshield.

The imagery also showed a hand holding a cell phone to con- note that the driver was texting while driving to cause the accident. The negative imagery chosen was meant to express the potential negative consequences of texting while driving.

We held the text of the two advertisements constant. The headline asked,“Do you text while driving?” The ad then proceeded to explain that over 500 000 people per year are killed or seriously injured because of texting while driving.

The public service announcement then encouraged people to“learn the facts to prevent yourself from becoming one of the statistics.”The ad concluded by providing a website and toll-free number people could call for more information.

Samples of the advertising stimuli used are provided in Appendix A.

Prior to being exposed to the advertisement, individuals answered questions about the extent of their prior texting- while-driving behaviors. We measured frequency of texting while driving by asking how many times per month individ- uals engage in texting while driving behavior. Individuals could enter anywhere between 0 and 16 or more times per month. We also measured quantity of texting while driving behavior by asking how many texts they send while engaging in this behavior. Individuals could enter anywhere between 0 and 10 or more texts. We multiplied these measures together to obtain an estimated monthly quantity of texts sent while driving that ranged from 0 (14% of the sample) to 160 or more texts (7% of the sample). The median 240J. L. Burtonet al. Copyright © 2015 John Wiley & Sons, Ltd.J. Consumer Behav.14: 237–247 (2015) DOI: 10.1002/cb number of texts sent per month in our sample was 32.5 texts.

Afterwards, individuals were exposed to the advertisement featuring positive or negative imagery depending on their condition. Finally, individuals answered questions regarding their assessments of the ad. Assessment measures included manipulation checks (e.g., extent to which the ad emphasized the positive and negative aspects of the behavior), mediation variables (e.g., measures of ad credibility, ad involvement, and ad relatability), and effectiveness variables (e.g., extent to which the ad influenced them to limit their behavior). Indi- viduals rated all assessment measures using 7-point scales.

RESULTS First, we performed manipulation checks to ensure that our manipulation of positive versus negative imagery worked in the texting while driving context. Individuals rated the ad containing positive imagery as“emphasizing the positive as- pects of texting while driving”(M= 2.76) more than the ad containing negative imagery (M= 2.15;t= 2.15,p<0.05).

Likewise, individuals rated the ad containing negative imag- ery as“emphasizing the negative aspects of texting while driving”(M= 6.34) more than the ad containing positive im- agery (M= 4.65;t= 6.79,p<0.01).

Thefirst hypothesis predicted a significant interaction be- tween ad imagery and prior behavior where individuals who did not engage in much texting while driving would be more influenced by the ad containing negative imagery, and heavy engagers in texting while driving would be more persuaded by the ad containing positive imagery. To test thefirst hy- pothesis, we regressed the extent to which the ad influenced individuals to curb their texting while driving against ad imagery (positive versus negative), extent of prior behavior, and an interaction term (ad imagery x prior behavior). The regression equation was significant (F= 2.72,p<0.05) and showed a positive main effect of negative ad imagery on likelihood to curb texting while driving (t= 2.85,p<0.01). However, as predicted in H1A and H1B, the inter- action was significant (t= 1.82,p<0.07). Spotlight analysis was performed to test the slopes of the two contrasts as rec- ommended by Irwin and McClelland (2001) and Aiken and West (1991) when the moderating variable is continuous.

Figure 1 shows the likelihood of individuals curbing their texting while driving behaviors after seeing the ad featuring positive and negative imagery for all levels of texting while driving behavior. Confirming H1A, individuals with the highest levels of texting and driving (160 or more texts per month) are more likely to curb their behavior after viewing the ad containing positive imagery (M= 4.41) than the ad containing negative imagery (M= 3.87). Individuals with the lowest levels of texting and driving (0 texts per month) are more likely to curb their behavior after viewing the ad with the negative imagery (M= 4.66) than the ad with the positive imagery (M= 3.73), confirming H1B.

H2 predicted that perceived ad credibility would mediate the relationship between ad imagery and individuals’likeli- hood to curb their texting while driving. H2 was confirmed as full mediation was present. Results showed that the ad with negative imagery was considered to be more credible (i.e., believable, informative, persuasive, and trustworthy) (t= 2.08,p<0.05). This credibility led to increased likeli- hood to curb future texting while driving behaviors (t= 10.81,p<0.01). While the ad with negative imagery does predict likelihood to curb future behaviors (t= 2.08, p<0.05), this coefficient is no longer significant (t= 1.14, p= n.s.) with credibility in the regression equation indicating that full mediation has occurred. Indeed, a Sobel test (Baron and Kenny, 1986) confirms the presence of mediation (t= 2.04,p<0.05).

As predicted in H3, ad involvement (i.e., the extent to which an ad encourages individuals to think about the issue and its consequences) also completely mediates the relation- ship between ad imagery and the likelihood individuals will limit their texting while driving behaviors. Ad imagery was shown to influence the level of involvement one feels while Figure 1. Interaction between ad imaginary and extent of prior behavior on individuals’likelihood to limit texting while driving. Generating positive change through ad imagery241 Copyright © 2015 John Wiley & Sons, Ltd.J. Consumer Behav.14: 237–247 (2015) DOI: 10.1002/cb viewing the ad (t= 3.07,p<0.01), which then leads to a greater likelihood to curb texting while driving behaviors (t= 16.47,p<0.01). While the ad containing negative imagery increases the likelihood individuals will limit their texting while driving behaviors (t= 2.39,p<0.05), this coefficient is no longer significant (t= 0.26,p= n.s.) when ad involvement is included in the regression equation. This complete mediation was also confirmed by a Sobel test (t= 3.02,p<0.01).

H4A and H4B predicted that the ads containing positive imagery would be seen as more relatable (i.e., containing characters and situations individuals can easily relate to), and that relatability was also important in influencing indi- viduals to curb their harmful behaviors. These relationships were supported in the texting while driving context.

Confirming H4A, individuals found the ad containing positive imagery to be much more relatable than the ad containing negative imagery (t= 2.23,p<0.05). In turn, the more individuals could relate to the ad, the greater the likeli- hood of them limiting their texting while driving behaviors (t= 7.62,p<0.01), confirming H4B. It is important to note that relatability did not mediate the relationship between ad imagery and likelihood to limit texting while driving. In- stead, the addition of relatability to the regression equation made the relationship between ad imagery and likelihood of limiting behavior stronger (t= 3.90,p<0.01), indicating that both influences are operating separately and simulta- neously. However, because the positive ad was generally seen to be more relatable, this relationship reduced the over- all effectiveness of the negative ad at influencing people to limit their harmful behaviors.

At this point, we have demonstrated that perceived ad credibility, ad involvement and relatability have a positive relationship with individuals’likelihood to curb texting- while-driving behaviors after ad exposure. H5 attempts to explain why individuals who frequently engage in the harm- ful behavior are more likely to curb their behavior after being exposed to the ad featuring positive imagery. H5 says that in- dividuals who frequently engage in the harmful behavior will notfind the negative ad to be more credible or involving but willfind the positive ad to be more relatable. A median split was performed for this analysis where individuals who sent more than 32.5 texts per month while driving were consid- ered as“frequent engagers.”Multivariate regression was used to determine the relationship between ad imagery (positive or negative) and key dependent variables (i.e., ad credibility, ad involvement, and relatability). Results confirmH5 and show that frequent engagers do notfind the negative ad to be more credible (F= 0.78,p= n.s.) or involving (F= 1.78,p= n.s.) but dofind the positive ad to be more relatable (F= 10.96,p<0.01) (Table 1). A similar analysis was performed to confirm H6 with respect to light engagers.

Results show that light engagers dofind the negative ad to be more credible (F= 3.80,p<0.05) and involving (F= 8.47, p<0.01) but do notfind the positive ad to be more relatable (F= 0.049,p= n.s.) (Table 1). Thesefindings explain why light engagers are more likely to curb their harmful behavior after being exposed to the ad with negative imagery.

STUDY 2 Participants and procedure A similar study was conducted to test the six hypotheses in the context of excessive gambling to see if the results were generalizable to public service announcements designed to curb harmful behaviors in other contexts. One hundred eighty-one students from two Midwestern universities partic- ipated in the gambling study for credit in their core business courses.

Once again, the text of the public service announcement remained constant and the experiment manipulated positive and negative imagery regarding the possible positive benefits and negative consequences of excessive gambling. The ads featuring positive imagery showed smiling people celebrat- ing and scooping up chips at a roulette game in a casino.

The ads featuring negative imagery showed two men alone at a casino table with their heads in their hands showing frus- tration at having lost plenty of money. The headline of the ad stated,“6 million Americans have serious problems with gambling.”The ad then encourages people to obtain help by visiting a website or calling a toll-free number if gambling is a problem for them. The exact ad stimuli used in this study are shown in Appendix A.

Similar to the texting while driving study, we collected information regarding the frequency and amount of gambling subjects perform in their daily lives. First, subjects indicated how many times they went gambling in a typical month (anywhere from 0 to 16 or more times per month). Next, subjects indicated how often they gambled on a typical gambling occasion (anywhere from 0 to 10 or more hours).

We created an index to reflect how many hours individuals spend gambling in a month by multiplying these two values.

Most individuals (73%) in our sample did not gamble.

Table 1. Extent to which ad tonality (positive versus negative) predicts relatability, credibility, and ad involvement for individuals with high versus low instances of texting while driving High levels of texting while driving Low levels of texting while driving tFSig. Adj.R 2 tFSig. Adj.R 2 DV = relatability; IV = ad tonality 3.311 10.962p<0.001 0.098 0.222 0.049p= n.s. 0.01 DV = credibility; IV = ad tonality 0.886 0.784p= n.s. 0.002 1.95 3.803p<0.054 0.029 DV = ad involvement; IV = ad tonality 1.333 1.776p= n.s. 0.008 2.911 8.472p<0.005 0.074 Wilks’lambdaF= 8.030,p<0.001F= 3.714,p<0.014 DV, dependent variable; IV, independent variable; Sig, significance; Adj, adjusted 242J. L. Burtonet al.

Copyright © 2015 John Wiley & Sons, Ltd.J. Consumer Behav.14: 237–247 (2015) DOI: 10.1002/cb Individuals who did gamble participated in the activity anywhere from 1.5 to 88 hours per month. Subsequently, participants were asked to view either the public service announcement featuring positive or negative imagery.

Finally, individuals provided their assessments of the ad on a number of mediating (ad credibility, ad involvement, and ad relatability) and effectiveness variables (e.g., the extent to which the ad influenced them to limit their gambling behavior).

RESULTS A manipulation check revealed that we had a successful ma- nipulation of positive versus negative imagery. Individuals rated the ad containing positive imagery as“emphasizing the positive aspects of gambling”(M= 4.78) more than the ad containing negative imagery (M= 1.64;t= 13.24, p<0.01). Similarly, the ad containing negative imagery was rated as portraying“the negative aspects of gambling” (M= 4.87) more than the ad containing positive imagery (M= 2.78;t= 7.59,p<0.01).

A regression examining how ad imagery, prior gambling behavior and an interaction term influences people’s likeli- hood to limit their gambling behavior was significant (F= 3.87,p<0.01). There was a positive main effect of ad imagery on people’s likelihood to limit their behavior (t= 2.91,p<0.01), as well as a significant interaction between imagery and prior gambling behavior (t= 2.28, p<0.05). Similar to the texting while driving study, we performed a spotlight analysis (Aiken and West, 1991; Irwin and McClelland, 2001) that examined the effectiveness of each ad at the various levels of gambling people exhibited in our study (Figure 2). As predicted in H1A, individuals with the highest levels of gambling (88 hours of gambling per month) are more likely to curb their behavior after viewing the ad with positive imagery (M= 6.23) than the ad with negative imagery (M= 1.76). Individuals with the lowest levels of gambling (i.e., non-gamblers) are more likely to curb their behavior after viewing the ad with negative imagery(M= 2.28) than the ad with positive imagery (M= 1.72), confirming H1B.

Full mediation was present when examining ad credibility as a mediator of the relationship between ad imagery and individuals’likelihood to curb their gambling behavior, confirming H2. The negative imagery in the excessive gambling ad led to stronger credibility perceptions (t= 4.02,p<0.01), which then led to an increased likelihood to curb gambling (t= 6.89,p<0.01). The negative relationship between ad imagery and likelihood to curb gambling behavior (t= 2.39,p<0.05) is no longer significant when credibility is added to the regression equation (t= 0.59,p= n.s.). Complete mediation for the gambling study was confirmed by a Sobel test (t= 3.47, p<0.01) (Baron and Kenny, 1986).

Ad involvement was also a full mediator of the relation- ship between ad imagery and subjects’likelihood to curb their gambling behaviors, confirming H3. The excessive gambling ad with negative imagery also led to increased ad involvement (t= 2.32,p<0.05), which then led to individ- uals’increased likelihood to curb their gambling habit (t= 9.23,p<0.01). The direct relationship between negative imagery and the likelihood to limit gambling (t= 2.39, p<0.05) is eliminated (t= 1.28, p = n.s.) when ad involve- ment is included in the regression equation indicating the presence of complete mediation (Sobel test:t= 2.27, p<0.05) (Baron and Kenny, 1986).

Participants found the ad containing positive imagery to be more relatable than the ad containing negative imagery (t= 2.53,p<0.05), which lead to increased likelihood to curb their gambling habit (t= 6.87,p< 0.01) confirming H4A and H4B. Consistent with Study 1, relatability did not mediate the relationship between ad imagery and the likelihood to limit excessive gambling. Adding relatability to the regression equation made the relationship between ad imagery and likelihood to limit gambling stronger (t= 4.17,p<0.01), indicating that both variables are independent and simultaneous predictors of subjects’inten- tions to limit their gambling. Because the positive ad was seen to be more relatable, this relationship reduces the Figure 2. Interaction between ad imaginary and extent of prior behavior on individuals’likelihood to limit gambling. Generating positive change through ad imagery243 Copyright © 2015 John Wiley & Sons, Ltd.J. Consumer Behav.14: 237–247 (2015) DOI: 10.1002/cb impact of the negative ad on subjects’likelihood to limit their gambling behaviors.

H5 says that individuals who frequently engage in the harmful behavior will notfind the negative ad to be more credible or involving but willfind the positive ad to be more relatable. A median split was performed for this analysis where individuals who gambled more than 1.5 hours per month were considered“frequent engagers.”Individuals who gambled less than 1.5 hours per month were considered “light engagers.”Multivariate regression was used to determine the relationship between ad imagery (positive or negative) and key dependent variables (i.e., ad credibility, ad involvement, and relatability). Results confirm H5 and show that frequent engagers do notfind the negative ad to be more credible (F= 2.83,p= n.s.) or involving (F= 0.02, p= n.s.) but dofind the positive ad to be more relatable (F= 3.70,p<0.05) (Table 2).

H6 posits that light gamblers are more likely tofind the negative ad to be more credible and involving but do not find the positive ad to be more relatable. H6 was confirmed, where light engagers dofind the negative ad to be more cred- ible (F= 13.74,p<0.01) and involving (F= 7.66,p<0.01) but do notfind the positive ad to be more relatable (F= 3.25,p= n.s.) (Table 2). Thefindings behind H5 and H6 explain why light gamblers are more likely to curb their behaviors after seeing the ad with negative imagery, and heavy gamblers are more likely to curb their behaviors after seeing the ad with positive imagery.

DISCUSSION The current study helps reconcile previousfindings that show mixed results on the ability of advertising to influence indi- viduals to curb potentially harmful behaviors. Our research demonstrates that different psychological processes are at work when trying to convince light versus heavy engagers to curb their potentially harmful behaviors. Validated across two contexts (texting while driving and excessive gambling), ourfindings show that positive imagery (as opposed to neg- ative imagery) is more effective at influencing heavy engagers to limit their harmful behaviors. Heavy engagers in the harmful behavior seem to resist the negative imagery and instead rely on the ability to relate to the situation and characters presented in the ad containing positive imagery.

This is consistent with research that suggests that individuals with more experience and knowledge about a domain aremore likely to be persuaded when the ad appeals to their higher level goals (Lewandowsky and Kirsner, 2000; Hong and Sternthal, 2010). Ourfindings show that persuasion is more likely to occur in heavy engagers even when the imag- ery is somewhat in conflict with the ultimate message of the ad, which is to encourage them to limit dangerous behaviors such as texting while driving and excessive gambling.

On the other hand, light engagers of harmful behavior are more likely to be persuaded to limit their behaviors after ex- posure to the ad containing negative imagery. The negative imagery evokes perceptions of ad credibility and encourages individuals to think about the advertised problem and conse- quences. Both credibility and ad involvement are shown to be important mediators of ad effectiveness for light engagers, whereas relatability was not a significant mediator for this group. As individuals with less knowledge and experience with the harmful behavior, they are more likely to rely on ex- ternal sources in determining their attitudes (Mandel and Johnson, 2002) and carefully scrutinize the information con- veyed in the ad and use it in forming their attitudes (Chi et al., 1982; Mitchell and Dacin, 1996).

Theoretical contributions A major contribution of this research is to reconcile prior research that shows mixed effectiveness of positive versus negative stimuli in public service announcements. For in- stance, Levin and Gaeth (1988) show that consumers respond better to messages regarding nutritional content when they have a positive frame of reference, but Meyerowitz and Chaiken (1987) show that consumers are more likely to per- form breast self-exams after exposure to messages with a neg- ative frame of reference. While Maheswaran and Meyers- Levy (1990) show that people are more likely to check their risk for heart disease with a negatively framed message, many other studies show that negatively framed public service an- nouncements lead to denial and defiance (Banning, 2001; Wolburg, 2006; Van Steeet al., 2012).

The reason for the inconsistentfindings regarding positive and negative messages in public service announcements is that prior research does not consider an important moderator of these effects—the amount of experience an individual has performing these harmful behaviors. Therefore, negatively framed messages are more likely to work for those with little experience performing harmful behaviors, while positively framed messages are more likely to work for those with little experience performing harmful behaviors.

Table 2. Extent to which ad tonality (positive versus negative) predicts relatability, credibility, and ad involvement for individuals with high versus low instances of excessive gambling High levels of gambling behavior Low levels of gambling behavior tFSignificance Adj.R 2 tFSignificance Adj.R 2 DV = relatability; IV = ad tonality 1.924 3.703p<0.06 0.053 1.803 3.252p= n.s. 0.017 DV = credibility; IV = ad tonality 1.682 2.828p= n.s. 0.037 3.711 13.744p<0.001 0.089 DV = ad involvement; IV = ad tonality 0.132 0.018p= n.s. 0.021 2.768 7.664p<0.006 0.048 Wilks’lambdaF= 2.457,p<0.075F= 9.741,p<0.001 DV, dependent variable; IV, independent variable.

244J. L. Burtonet al.

Copyright © 2015 John Wiley & Sons, Ltd.J. Consumer Behav.14: 237–247 (2015) DOI: 10.1002/cb The third-person (Golan and Banning, 2008; Grier and Brumbaugh, 2007) and boomerang (Fishbeinet al., 2002a, 2002b; Zhaoet al., 2006) effects are often blamed for the in- effectiveness of public service announcements. However, prior research shows that the third-person effect is greater when consumers have more experience performing the harm- ful behavior (Wolburg, 2006), and the message is negatively framed (Golan and Banning, 2008). Taken together, these results suggest and our research confirms that positively framed messages will work better for individuals heavily engaged in harmful behaviors.

Another major contribution of this study is to isolate the different psychological processes occurring in persuasion for these two important targets of public service announce- ments. While credibility, ad involvement, and relatability are generally perceived as important mediators of public service announcement effectiveness (see meta analysis by Keller and Lehmann, 2008), our research shows that these mediators work differently in determining the behavioral intentions of heavy versus light engagers of harmful behav- ior. Light engagers perceive the negative imagery as more credible and involving but do not perceive the positive im- agery as more relatable. On the other hand, heavy engagers perceive the positive imagery as more relatable but do not perceive the negative imagery as more believable or involving.

Practical implications The researchfindings have very important implications for advertisers designing campaigns to curb harmful behaviors in our society. These public service campaigns are often de- signed by non-profit agencies, governmental bodies, and even advertising agencies looking to make a difference in the community through theirpro bonowork. First, these agencies must determine what the primary objective of their public service announcement campaign should be: (1) to de- terfirst-time behavior, (2) to limit the behavior of heavy users, or (3) a combination of both. Many agencies pursue both goals simultaneously with a single campaign, which the present research suggests may be ineffective. If the goal is to deterfirst-time behavior, then a campaign should feature a strong message featuring the negative consequences of the harmful behavior along with a solution on how to avoid the bad outcomes depicted. If the goal is to limit the behavior of heavy users, the campaign should feature characters and situations to which the heavy user can relate, along with a call to action that will convince these users to limit their harmful behavior. If the objective is to accomplish both, an agency should consider creating two separate campaigns and use effective media planning strategies to reach the heavy engagers with one campaign and light engagers with the other campaign.

Limitations and future research As with any research, this study has some limitations that open the doors for potential future research. First, it is important to recognize that the current research focused on public service announcements designed to prevent or limit harmful behaviors. Therefore, it is important not togeneralize the results on the impact of positive and nega- tive imagery to other types of public service announce- ments such as those designed to solicit donations for important causes or encourage proactive health behaviors such as visiting a doctor or following a particular diet.

Future research can use a similar experimental design to test the generalizability of these results to other types of public service announcements.

Second, while the positive and negative imagery used in this study tried to emphasize positive and negative aspects of engaging in a harmful behavior, another interesting ave- nue would be to explore whether the positive imagery should depict the positive consequences of not engaging in a harm- ful behavior. We focused on the positive consequences of en- gaging in the behavior (which is a bit counter intuitive) mainly because prior research emphasizes the importance of characters and situations to which people can relate (Grubeet al., 1996; Van Steeet al., 2012). Examining which type of positive imagery would be more effective on light and heavy engagers of harmful behavior is an interesting di- rection for future research.

Finally, while we establish the extent to which an individ- ual engages in a harmful behavior as an important moderator of the effectiveness of positive and negative imagery in pub- lic service announcements, future research should explore other potential moderators. Other potential moderators of public service announcement effectiveness could be individ- ual personality characteristics such as reactance or risk seek- ing or even individual experiences such as being or knowing a victim of such harmful behaviors. BIOGRAPHICAL NOTES Dr Jennifer L. Burtonis in herfirst year as a faculty member in the Department of Marketing at High Point University. Her main re- search interests are marketing communications, consumer decision making, and consumer affect. She has a PhD in Marketing from the University of Texas at Austin, MBA in Brand Management from Wake Forest University, and BA in Spanish and Criminal Jus- tice from Indiana University. Prior to entering academia, she worked for 6 years in product and marketing management in the telecommunication and industrial equipment industries for global companies such as MCI and Harris Group Incorporated.

Dr Donna J. Hillis a Professor of Marketing, Emeritus at Bradley University where she taught Services Marketing, Marketing Research, Retail Management, and Professional Services Marketing. Her research has appeared in theJournal of Retailing,Journal of Service Research,Journal of Services Marketing,Journal of Consumer Policy, The Journal of Marketing Theory and Practice,The Journal of Applied Business Research,Journal of Marketing Management,Journal of Retailing and Consumer Services, andJournal of Consumer Satis- faction, Dissatisfaction and Complaining Behavior. She received her PhD in Marketing from Indiana University, MBA from Ball State University, and BA from Indiana University.

Aysen Bakiris a Professor of Marketing at Illinois State University.

She received her PhD from the University of Mississippi. Her primary research includes children’s advertising, gender roles, and cross-cultural consumer behavior. She has published articles in theJournal of Adver- tising,Journal of Advertising Research,Journal of Business Research, Journal of Marketing Theory and Practice,Journal of Consumer Marketing,Journal of Current Issues and Research in Advertising, Journal of Business Ethics, and other journals and proceedings. Generating positive change through ad imagery245 Copyright © 2015 John Wiley & Sons, Ltd.J. Consumer Behav.14: 237–247 (2015) DOI: 10.1002/cb REFERENCES Agostinelli G, Grube JW. 2003. Tobacco counter-advertising: a re- view of the literature and a conceptual model for understanding effects.Journal of Health Communication 8(2): 107–127.

Aiken LS, West SG. 1991.Multiple Regression: Testing and Interpreting Interactions. Sage Publications: Newbury Park, CA.

Banning SA. 2001. Do you see what I see? Third-person effects on public communication through self-esteem, social stigma, and product use.Mass Communication & Society 4(2): 127–147.

Baron RM, Kenny DA. 1986. The moderator-mediator variable dis- tinction in social psychological research: conceptual, strategic and statistical considerations.Journal of Personality and Social Psychology 51(6): 1173–1182.

Baumeister RDF, Bratslavsky E, Finkenauer C, Vohs K. 2001. Bad is stronger than good.Review of General Psychology 5(4):

323–370.

Bettman JR, Park CW. 1980. Effects of prior knowledge and expe- rience and phase of the choice process on consumer decision processes: a protocol analysis.The Journal of Consumer Research 7(3): 234–248.

Block LG. 2005. Self-referenced fearand guilt appeals: the moderating role of self-construal.Journal of Applied Social Psychology 35(11):

2290–2309.

Chi MTH, Glaser R, Rees E. 1982. Expertise in problem solving. In Sternberg RJ (ed.).Advances in the Psychology of Human Intel- ligence, Vol. 1. Lawrence Erlbaum Associates: Hillsdale, NJ; 11–76.

Fishbein M, Cappella J, Hornik R, Sayeed S, Yzer M, Ahern RK.

2002a. The role of theory in developing effective anti-drug public service announcements. In Crano WD, Burgoon M (eds).

Mass Media and Drug Prevention: Classic and Contemporary Theories and Research. Laurence Erlbaum Associates, Inc.:

Mahwah, NJ; 89–117.

Fishbein M, Hall-Jamieson K, Zimmer E, von Haeften I, Nabi R.

2002b. Avoiding the boomerang: testing the relative effective- ness of antidrug public service announcements before a national campaign.American Journal of Public Health 92(8): 238–245.

Fiske ST. 1980. Attention and weight in person perception: the im- pact of negative and extreme information.Journal of Personality and Social Psychology 38(6): 889–906.

Foxcroft DR, Ireland D, Lister-Sharp DJ, Lowe G, Breen R. 2003.

Longer-term primary prevention for alcohol misuse in young people: a systematic review.Addiction 98(4): 397–411.

Golan GJ, Banning SA. 2008. Exploring a link between the third- person effect and the theory of reasoned action: beneficial ads and social expectations.American Behavioral Scientist 52(2):

208–224.

Grier SA, Brumbaugh AM. 2007. Compared to whom? The impact of status on third person effects in advertising persuasion in a South African context.Journal of Consumer Behaviour 6 (1):

5–18.

Grube JW, Madden PA, Friese B. 1996. The effects of television alcohol advertising on adolescent drinking. Poster session presented at the annual meeting of the Research Society on Alcoholism, Washington, DC.

Hong J, Sternthal B. 2010. The effects of consumer prior knowledge and processing strategies on judgments.Journal of Marketing Research 47(2): 301–311.

Irwin JR, McClelland GH. 2001. Misleading heuristics and moder- ated multiple regression models.Journal of Marketing Research 38(1): 100–109.

Keller PA. 1999. Converting the unconverted: the effect of inclina- tion and opportunity to discount health-related fear appeals.

Journal of Applied Psychology 84(3): 403–415.

Keller PA, Lehmann DR. 2008. Designing effective health communi- cations: a meta-analysis.Journal of Public Policy and Marketing 27(2): 117–130.

Kelly KJ, Swaim RC, Wayman JC. 1996. The impact of a localized anti-drug media campaign on targeted variables associated withadolescent drug use.Journal of Public Policy and Marketing 15(2): 238–251.

Kubany ES, Watson SB. 2003. Guilt: elaboration of a multidimen- sional model.Psychological Record 53(1): 51–90.

Larsen RJ, Diener E. 1987. Affect intensity as an individual differ- ence characteristic: a review.Journal of Research in Personality 21(1): 1–39.

Levin AM, Dato-on MC, Manolis C. 2007. Deterring illegal downloading: the effects of threat appeals, past behavior, subjective norms, and attributions of harm.Journal of Consumer Behaviour 6(2/3): 111–122.

Levin IP, Gaeth GJ. 1988. Framing of attribute information before and after consuming the product.Journal of Consumer Research 15(3): 374–378.

Lewandowsky S, Kirsner K. 2000. Knowledge partitioning:

context-dependent use of expertise.Cognition & Memory 28 (2): 295–305.

MacKenzie SB, Lutz RJ. 1989. An empirical examination of the structural antecedents of attitude toward the ad in an advertising pretesting context.Journal of Marketing 53(2): 48–65.

MacKenzie SB, Lutz RJ, Belch GE. 1986. The role of attitude toward the ad as a mediator of advertising effectiveness; a test of competing explanations.Journal of Marketing Research 23(2): 130–143.

Maheswaran D, Meyers-Levy J. 1990. The influence of message framing and issue involvement.Journal of Marketing Research 27(3): 361–67.

Mandel N, Johnson EJ. 2002. When web pages influence choice: effects of visual primes on experts and novices.Journal of Consumer Research 29(2): 235–245.

McKenna JW, Williams KN. 1993. Crafting effective tobacco counter-advertisements: lessons from a failed campaign directed at teenagers.Public Health Reports 108(1): 85–89.

Meyerowitz BE, Chaiken S. 1987. The effect of message framing on breast self-examination attitudes, intentions and behavior.

Journal of Personality and Social Psychology 52(3): 500–510.

Meyers-Levy J, Peracchio LA. 1996. Moderators of the impact of self-reference on persuasion.Journal of Consumer Research 22(4): 408–423.

Midford R, McBride N. 2001. Alcohol education in schools. InN.

Heather and T. Stockwell International Handbook of Alcohol Dependence and Problems. John Wiley & Sons: Chichester.

Mitchell AA, Dacin PA. 1996. The assessment of alternative measures of consumer expertise.Journal of Consumer Research 2(3): 219–39.

Moore DJ, Harris WD. 1996. Affect intensity and the consumer’s attitude toward high impact emotional advertising appeals.

Journal of Advertising 25(2): 37–50.

Parry S, Jones R, Stern P, Robinson M. 2013.‘Shockvertising’:An exploratory investigation into attitudinal variations and emotional reactions to shock advertising.Journal of Consumer Behaviour 12(2): 157–177.

Pratto F, John OP. 1991. Automatic vigilance: the attention-grabbing power of negative social information.Journal of Personality and Social Psychology 61(3): 380–391.

Petty RE, Cacioppo JT. 1986.Communication and Persuasion:

Central and Peripheral Routes to Attitude Change. Springer- Verlag: New York.

Raghubir P, Menon G. 1998. AIDS and me, never the twain shall meet: the effects of information accessibility on judgments of risk and advertising effectiveness.Journal of Consumer Research 25(1): 52–63.

Rozin P, Royzman EB. 2001. Negativity bias, negativity dominance, and contagion.Personality and Social Psychology Review 5(4):

296–320.

Thunholm P. 2005. Planning under time pressure: an attempt to- ward a prescriptive model of military tactical decision making”.

In Montgomery H, Lipshitz R, Brehmer B (eds).How Experts Make Decisions. Lawrence Erlbaum Associates: Hillsdale, NJ; 43–56. 246J. L. Burtonet al.

Copyright © 2015 John Wiley & Sons, Ltd.J. Consumer Behav.14: 237–247 (2015) DOI: 10.1002/cb Van Stee SK, Noar SM, Allard S, Zimmerman R, Palmgreen P, McClanahan A. 2012. Reactions to safer-sex public service announcement message features: attention, perceptions of realism, and cognitive responses.Qualitative Health Research 22(11):

1568–1579.

Wheatley JJ, Oshikawa S. 1970. The relationship between anxiety and positive and negative advertising appeals.The Journal of Marketing Research 7(1): 85–59.Wolburg JM. 2006. College students’responses to antismoking messages: denial, defiance and other boomerang effects.The Journal of Consumer Affairs 40(2): 294–323.

Zhao X, Sayeed S, Capella JN, Hornik R, Fishbein M, Ahern MR.

2006. Targeting norm-related beliefs about marijuana use in an adolescent population.Health Communication 19(3):

187–196. APPENDIX A. ADVERTISING STIMULIGenerating positive change through ad imagery247 Copyright © 2015 John Wiley & Sons, Ltd.J. Consumer Behav.14: 237–247 (2015) DOI: 10.1002/cb Copyright ofJournal ofConsumer Behaviour isthe property ofJohn Wiley &Sons, Inc.and its content maynotbecopied oremailed tomultiple sitesorposted toalistserv without the copyright holder'sexpresswrittenpermission. However,usersmayprint, download, oremail articles forindividual use.