A Terror Management Perspective on the Persuasiveness of Fear Appeals

Transcription

1 TERROR SHEHRYAR MANAGEMENT AND HUNT THEORY AND FEAR APPEALS JOURNAL OF CONSUMER PSYCHOLOGY, 15(4), Copyright 2005, Lawrence Erlbaum Associates, Inc. A Terror Management Perspective on the Persuasiveness of Fear Appeals Omar Shehryar Montana State University David M. Hunt University of Wyoming We use terror management theory (TMT) to show that the nature of the threatening consequences included in fear-appeal communications influences the responses to the messages promoted in such communications. On the basis of differences between death-related consequences and non-death-related consequences, they provide an explanation for maladaptive responses to fear appeals. Results from 2 experiments indicate that participants who were highly committed to a worldview of drinking alcohol rejected socially acceptable attitudes toward drinking and driving when the message containing such attitudes was accompanied by a fear appeal that contained death as a consequence, but not when fear appeals contained the fear of arrest or serious injury. Participants perceived their level of experienced fear to be the same across the experimental conditions. The results have implications for considering the qualitative nature of the threatening consequence in fear-appeal research and for using TMT to understand maladaptive responses to fear-appeal communications. The dominant paradigm in fear-appeal research asserts that differences in level of fear lead to differences in the persuasiveness of the communication (Keller & Block, 1996; Rogers, 1985; Witte, 1994). However, researchers have raised concerns about using level of fear as a stand-alone measure of the efficacy of fear-appeal communications (LaTour & Rotfeld, 1997). Results observed in high-fear conditions in past studies complicate the debate on the level-of-fear construct. Specifically, reacting to a high-fear-arousing communication, some message recipients exhibit increased levels of risky behavior, a response opposite to the intent of the message (Rippetoe & Rogers, 1987; Rogers, 1983; Witte, 1994). Evidence that high-fear messages can result in increased risky behavior poses a serious threat to the efficacy of fear-appeal communications. Unfortunately, such maladaptive responses have not generated sufficient research. Pechmann, Zhao, Goldberg, and Reibling (2003) acknowledged that researchers have focused on factors that strengthen intentions to refrain from risky behavior at the expense of understanding when fear-arousing communication Requests for reprints should be sent to Omar Shehryar, College of Business, Montana State University, 449 Reid Hall, Bozeman, MT omar@montana.edu fails. Witte (1992) suggested that a thorough understanding of fear appeals can be achieved only by studying cases in which fear appeals work as well as cases in which fear appeals do not achieve their desired objectives. In this research, we address the concern over the unexpected failure of high-fear appeals. We draw on terror management theory (TMT; Greenberg, Pyszczynski, & Solomon, 1986) to propose that the qualitative nature of a perceived threat influences the elicited response to the advocated message accompanying the threat. We distinguish messages that use death as a threatening consequence from messages that use threats of nonfatal consequences. We use this distinction to explain and predict increases in maladaptive responses to fear-appeal communications. We suggest that a TMT perspective allows for a meaningful comparison of fear-appeal efficacy on a dimension other than level of fear. Moreover, a TMT approach augments our current understanding of maladaptive responses to fear-appeal communications. Based on theoretical distinctions between the psychodynamics evoked by a fear of death versus those evoked by non-death-related fears, a TMT perspective helps explain and predict maladaptive responses to fear appeals. Our approach to differentiate fear appeals on the basis of the nature of the threatening consequence deviates from the

2 276 SHEHRYAR AND HUNT dominant paradigm that differentiates fear appeals on the basis of the level of evoked fear (Block & Keller, 1998; Hovland, Janis, & Kelley, 1953; Rogers, 1983; Witte, 1992). However, we demonstrate that this alternative theoretical perspective has the potential to supplement existing theory and provide a new direction for research on fear appeals. BACKGROUND Fear-appeal communications are predominantly used to curb undesirable social behaviors such as smoking, drug use, and drinking and driving. Recently, fear appeals have also been used to sell insurance, safety devices, over-the-counter diet programs, and prescription drugs, to name just a few examples. The sustained interest in the use of fear as a persuasive tool has fueled substantial research in the area of fear-appeal communication. Three theories (a) drive theory (Janis & Feshbach, 1953), (b) protection motivation theory (PMT; Rogers, 1983), and (c) parallel response theory (Tanner, Hunt, & Eppright, 1991) have guided much of the extant research on fear appeals. More recently, fear-appeal researchers have favored PMT because of its comprehensive nature (Pechmann et al., 2003), and in some cases researchers have extended PMT to improve its explanatory power (Block & Keller, 1998; Rogers, 1983; Tanner et al., 1991; Witte, 1994; Witte & Allen, 2000). Despite these theoretical advances, the extant research has been criticized for its inability to explain or predict when fear appeals will not work (Witte, 1994; Witte & Allen, 2000). In some fear-appeal studies, message recipients who were exposed to high-fear-arousing messages exhibited more favorable attitudes toward the undesirable behaviors, a response that is opposite to the intent of the message (Arnett, 2000). Keller (1999) and Keller and Block (1996) suggested that high-fear messages are more likely to be resisted because of motivated reasoning and message discounting. Although this explanation accounts for the avoidance of a message, it remains unclear why resistance to high-fear messages is sometimes followed by greater affinity for the behavior that the message intends to curb. A discussion of TMT offers insight into this issue. TMT TMT is based on the notion that human beings understanding of their impending death combined with the instinctive drive for self-preservation engenders a tremendous potential for terror. Without a mechanism to cognitively manage the understanding of our own mortality, the resulting terror can be debilitating. The theory posits that cultural worldviews are effective in protecting an individual from existential terror evoked by awareness of one s finitude (Greenberg et al., 1986). Cultural worldviews are humanly created and transmitted beliefs about the nature of reality shared by groups of individuals (Greenberg, Solomon, & Pyszczynski, 1997, p. 65). One important function of cultural worldviews is to contribute to self-esteem by providing the satisfaction of adhering to and upholding the shared values of one s society. In this manner, cultural worldviews act as anxiety buffers by weaving a web of order and meaning, by providing standards of value, and by offering symbolic death transcendence to those who strive to meet these standards (Greenberg et al., 1997). Cultural worldviews vary across cultures and individuals and may include religious and social values, political and nationalistic beliefs, and moral codes (Greenberg et al., 1997). However, worldviews are not restricted to these major life domains. Mundane tasks such as evaluating a work of art or playing a game of tennis do not necessarily constitute worldviews. However, if a person derives self-esteem from the belief that he or she is a connoisseur of art or a skilled tennis player, then such beliefs do constitute worldviews. TMT posits that reminders of mortality increase the need to defend and strengthen one s worldview to sustain its anxiety-buffering function. Support for this hypothesis exists in more than 175 empirical studies (Arndt, Solomon, Kasser, & Sheldon, 2004). Findings show that belief in one s worldviews is stronger following mortality-salient manipulations than following non-mortality-salient manipulations (Pyszczynski et al., 1996; Rosenblatt, Greenberg, Solomon, Pyszczynski, & Lyon, 1989). Treatments considered equally noxious, such as threatening someone with serious injury, or intense pain, do not evoke mortality salience and do not lead to a defense of cultural worldviews (Greenberg, Pyszczynski, Solomon, Simon, & Breus, 1994). Moreover, the salience of dental pain, a common threat used in fear-appeal research, does not replicate the results obtained from the salience of mortality (McGregor et al., 1998). TERROR MANAGEMENT AND FEAR APPEALS The distinction between individuals response to the fear of death versus their response to other noxious outcomes has important implications for fear-appeal research. If fear-appeal communications that threaten audiences with death make mortality salient, then it is likely that message recipients will invest in their worldviews to assuage the fear aroused by a reminder of their mortality. Past research has found that the specific worldview used to buffer the fear of death is the worldview made salient by the context (Wischusen, Nelson, & Pollini, 2003). According to TMT, following mortality salience, if a message s advocated worldview admonishes the worldview held by a message recipient, the message recipient will reject the advocated worldview and defend the preexisting worldview relevant to the context. Moreover, a message recipient is likely to defend

3 TERROR MANAGEMENT THEORY AND FEAR APPEALS 277 his or her worldview, even if it is distorted and delinquent (Arndt, Goldenberg, Greenberg, Pyszczynski, & Solomon, 2000). Those who are highly committed to a worldview, and thus derive self-esteem from that commitment, are labeled here as having a high precommitment to that worldview. In contrast, individuals who do not derive self-esteem from commitment to a specific worldview are labeled here as having low precommitment to that worldview. We predicted that when individuals with high precommitment to a worldview are exposed to a message that conflicts with that worldview and uses the threat of death as a consequence, they will reject the recommendation of the message. We also predicted that message rejection will not be observed when a message using the threat of death is targeted to individuals with low precommitment to a worldview. Moreover, consistent with past research, we expected that fear appeals that do not make mortality salient will be equally effective for audiences with high or low precommitment to a specific worldview. Hypothesis 1 is based on this discussion. H1: The qualitative nature of a threatening consequence will moderate the relation between degree of precommitment to an undesirable behavior and persuasiveness of a message such that the following may occur: H1a: When mortality is salient, high precommitment to an undesirable behavior will lead to message rejection, but low precommitment will lead to message acceptance. H1b: When mortality is not salient, both high and low precommitment to an undesirable behavior will lead to message acceptance. TMT offers a second distinction between responses to a fear of death and responses to other types of fears. Specifically, TMT research outlines two kinds of defensive responses to mortality salience: (a) proximal and (b) distal (Arndt, Greenberg, Solomon, Pyszczynski, & Simon, 1997). A proximal defense involves suppression or defensive avoidance of a message and thus mirrors fear-control processes outlined in past fear-appeal research (Witte, 1994; Witte & Allen, 2000). Fear-control processes involve immediate suppression of a message to avoid the message altogether (Keller, 1999; Liberman & Chaiken, 1992). A distal defense, however, is unique to the contemplation of mortality. A distal defense is used after the proximal defense has been invoked. TMT suggests that it is the distal defense that encourages adherence to a salient worldview, thereby restoring its anxiety-buffering function and assuaging the fear aroused by thoughts of one s mortality (Arndt et al., 1997). Analogous to the distal defense suggested by TMT, fear-appeal research suggests that once a fear-control process such as suppression is accounted for, a danger-control process is activated. Danger-control processes encourage message acceptance in order to mitigate the danger posed by an imminent threat (Witte, 1994). On the surface, a distal defense seems identical to a danger-control process; however, if one s salient worldview is distorted, then a distal defense to the fear of death that includes adherence to such a worldview will likely lead to message rejection. In contrast, the predicted outcome from a danger-control process would be message acceptance. Thus, maladaptive responses that persist even after suppression has been accounted for pose a challenge to fear-appeal research. We suggest that by introducing a delay before measuring participants responses to fear appeals, initial suppression can be mitigated. Past research indicates that two parallel processes ensue when a thought is suppressed. One process attempts to suppress the thought, and a second process attempts to remember what needs to be suppressed (Wegner & Erber, 1992). Hence, a suppressed message is still processed subconsciously. In one study, Arndt et al. (1997) exposed participants to a threat but imposed a cognitive load soon after exposure to the threat. The presence of the cognitive load prevented suppression and resulted in conscious processing instead of defensive avoidance. We predicted that by introducing a delay before measuring the audience s response to viewing a fear appeal, it is possible to capture the distal response to the fear of death. We expected that, if responses are measured both with and without a delay, a distal defense, characteristic of terror management functions, will be observed after the delay but will not be observed immediately following the threat. Such a defense will be evoked only in response to a fear of death resulting in message rejection by individuals with a high precommitment to an undesirable behavior. In contrast, danger-control processes that encourage message acceptance will also ensue after a delay but only if the threatening consequence is of a nonfatal nature. Thus, individuals with a high precommitment to an undesirable behavior are more likely to accept a recommended message when their response to a nonfatal consequence is measured after a delay as opposed to immediately following a threat. Hypotheses 2 and 3 are based on this discussion. H2: When response to a message accompanying a fear appeal is measured without a delay, individuals with a high precommitment to an undesirable behavior are less likely to be persuaded by the message than individuals with a low precommitment to an undesirable behavior. H3: When response to a message accompanying a fear appeal is measured with a delay, individuals with a high precommitment to an undesirable behavior are more likely to reject a message if the accompanying fear appeal is death related than if it is not death related. In Study 1, we tested H1. In Study 2, we tested H2 and H3. Together, results from the two experiments provide converging evidence for the TMT perspective on fear appeals.

4 278 SHEHRYAR AND HUNT STUDY 1 Sample We tested H1 in a drinking-and-driving context, for several reasons. For convenience in regard to data collection, we intended to recruit college students for our sample. The relevance of drinking and driving for college-age adults improves the internal validity and the external generalizability of the design. Evidence for relevance of drinking and driving to college students comes from several studies that have noted the prevalence of alcohol-related problems on college campuses (Engs, Diebold, & Hanson, 1996; Wechsler, Kuh, & Davenport, 1996). Converging evidence from secondary data indicates that the highest intoxication rates in fatal crashes in 2003 were recorded for drivers aged 21 to 24 (39%; National Highway Traffic Safety Administration, 2003). Furthermore, the drinking-and-driving context is especially suitable for research on message rejection because message rejection is prevalent among young adults (Arnett, 2000). Studies show that despite being aware of the ill effects of excessive alcohol consumption, college-age adults indulge in binge drinking (Engs & Hanson, 1988; Substance Abuse and Mental Health Service Administration, 2002; Wechsler, Davenport, Dowdall, Moeykens, & Castillo, 1994). Because we are studying responses to threatening consequences in the context of worldview defense, it is important to first establish that drinking alcohol is considered a worldview among college students. According to TMT, to be considered a worldview a behavior should be shared, and it should provide self-esteemtobufferanxietyintheeventofareminderofone s death(rosenblatt et al., 1989). Past research provides substantial evidence that alcohol consumption among college students fulfills both conditions. College students are disproportionately involved in drinking and, as such, they represent a population within which a large number of people share homogeneous beliefs regarding alcohol consumption. Wechsler et al. (1994; Wechsler, Molnar, Davenport, & Baer, Sample 1999) reported that 44% of college students were classified as binge drinkers and that binge drinkers consumed 68% of all alcohol consumed by college students. Indeed, a variety of research suggests that college students engage in drinking to enhance their self-worth and strengthen their sense of belonging (Glindemann, Geller, & Fortney, 1999; McCormack, Laybold, Dickerson-Nelson, & Budd, 1993; Workman & Beer, 1989).Asampleofcollegestudentsislikelytoprovidearepresentative cross-section of levels of precommitment to alcohol consumption, thereby allowing us to capture a variety of responses to different types of threatening consequences. Therefore, using a college student sample in a drinking-and-driving context is both pragmatic and theoretically justified. Pretest To test TMT hypotheses in a fear-appeal domain, it is important to determine whether fear-appeal advertisements that use the threat of death as a consequence evoke mortality salience. TMT researchers have evoked mortality salience through both conscious and subconscious manipulations. For instance, Rosenblatt et al. (1989) evoked mortality salience by having participants write a short essay about their own death. In contrast, Pyszczynski et al. (1996) measured the mortality salience of pedestrians who had just walked past a funeral home. Results of a standard measure of mortality salience indicated that both approaches evoked mortality salience among participants. On the basis of studies that have evoked mortality salience through normal, everyday reminders of death (e.g., walking past a funeral home), we expected that threats of death-related consequences in fear-appeal advertisements also will evoke mortality salience. To verify this, we conducted a pretest of the stimuli. Forty-nine undergraduate students in a marketing class in a U.S. university participated in the study for course credit. The average age of the participants was 21 years and 7 months. Seventy-five percent were female. Fear appeals for the pretest consisted of two types of anti-drinking-and-driving ads. Participants were randomly assigned to the two conditions (mortality salience vs. fear of arrest). The ad for the mortality salience condition showed a family around a casket with statistics on drunk driving deaths and the persuasive slogan Don t let this be the occasion for your next family gathering: Do not drink and drive. The fear-of-arrest condition showed a person being handcuffed by a police officer, and statistics including threats such as attorney fees, license revocation, and fines, along with the same slogan, presented in the mortality salience condition. Statistics used in the advertisements contained actual figures taken from the National Highway Traffic Safety Administration (2003). There were 25 participants in the fear-of-death condition and 24 in the fear-of-arrest condition. Each group of participants was brought into a room equipped with an overhead projection device used to show the print advertisement. A questionnaire titled Advertising Appeals Survey was distributed among the participants. Participants were instructed to give their opinions on an advertisement projected onto the screen. The ad was displayed for 20 sec. To test the notion that mere exposure to ads makes mortality salient, we asked participants to fill out a word fragment completion task. This instrument has been used in previous TMT studies to test for mortality salience (Bassili & Smith, 1986; Greenberg et al., 1994). The task requires participants to fill in blanks to complete 25 incomplete words. Five of the 25 word fragments are such that they can be completed by either death-related words or by neutral words. For example, de canbecompleted as dead, desk, ordebt. Similarly gra canbecompleted as grave, grace,orgrape, and ki e d may be completed as kissed or killed. The two other fragments that could be completed as death-related or neutral words were bur d and coff. Word fragments that cannot be completed with death-related words

5 TERROR MANAGEMENT THEORY AND FEAR APPEALS 279 included tr, ok, and fl_w_r. Thenumber of death-related words completed by each participant is summed, and the means are compared across the experimental conditions. Higher means suggest greater accessibility of death-related words and provide evidence for mortality salience. In the pretest, the mean number of death-related words completed in the mortality-salience condition was 2.3, compared with 1.3 for the fear-of-arrest condition. This difference was significant, t(47) = 2.94, p =.005, providing evidence that a mere 20-sec exposure to an advertisement can evoke mortality salience. Experimental Design After ensuring that mere exposure to an advertisement can make mortality salient, we tested H1 in an experiment involving 178 students enrolled in a U.S. university. The study hada4 2between-subjects design and included three types of fear appeals and a control condition (fear of arrest, fear of death, fear of serious injury, and control with no fear) and two levels of precommitment (high and low). The condition involving fear of serious injury was added because of a concern regarding the qualitative nature of the consequences used in the pretest. Death is a physical danger, whereas arrest, as used in the pretest, is a social threat. Research empirically supports the distinction between social and physical threats and the responses that each evokes (Ho, 1998). Schoenbachler and Whittler (1996) found that social threats that warned teenagers of negative consequences of drug use were more effective than physical threats. Their findings are consistent with the ordered protection motivation model (Tanner et al., 1991), which suggests that, following fear appeals, rather than choosing a response that minimizes the threat, recipients may choose a suggested alternative that maximizes social acceptance. If this is the case, then it is no surprise that the threat of arrest, which evokes embarrassment, will likely lead to greater message acceptance than the threat of death. Unless responses to death-related consequences are established as qualitatively different from responses to both social consequences and other physical consequences, we cannot conclusively assert that death-related consequences are unique in the responses they evoke. Therefore, to test a purely physical threat we included an advertisement involving threat of serious injury. As a control, we also included a condition involving no threat at all. In the serious-injury advertisement, a young man was shown in a wheelchair with statistics regarding serious injuries due to drinking and driving. The control condition featured an advertisement for Jell-O brand dessert. The advertisements for fear of death and fear of arrest were the same as those used in the pretest. Participants were randomly assigned to one of four fear-appeal conditions. Each participant was given a booklet containing dependent measures, control measures, and the stimuli. The booklet included distraction tasks and filler measures designed to disguise the intent of the questionnaire and reduce potential demand effects. The main stimulus was one of the anti-drinking-and-driving advertisements or the Jell-O advertisement. It was preceded by a page from a lifestyle and entertainment magazine as well as a full-page filler advertisement for a hybrid gasoline electric automobile. The questionnaire measured level of fear arousal, perceived susceptibility to the threat, mortality salience evoked by the stimuli, acceptance or rejection of the promoted attitude toward drinking and driving, and demographic variables. Manipulation Check: Mortality Salience To assess mortality salience and distinguish it from other noxious fears, we had participants fill out the word fragment completion task used in the pretest. The mean number of death-related words was calculated and compared across the experimental conditions. Higher accessibility of death-related words provides evidence for mortality salience. Independent Variable: Precommitment to Drinking To measure precommitment to drinking alcohol, we asked participants to complete a scale adapted from Neese and Taylor (1994). Participants were asked to indicate their agreement on a 7-point scale ranging from 1 (strongly disagree)to 7(strongly agree; α =.80) with such items as Others view me favorably because I drink alcohol, When I have a couple of drinks with my friends, I feel that I can express the real me, I like being known as a person who drinks alcohol, I like being known as a person who has a high tolerance for alcohol, and Drinking alcohol is an important part of who I am. To reduce potential demand effects, we collected data on precommitment to drinking alcohol 1 month before the main data collection in the guise of a personality survey. Two hundred participants completed this stage. Because of attrition 178 participants took part in the main study. There were 101 women (56.7%) in the final sample. The average age of the participants was 22.0 years (SD = 2.23). Dependent Variable: Message Acceptance and Rejection Participants rated their agreement with the following items on 7-point scale ranging from 1 (strongly disagree) to7 (strongly agree; reverse coded; α =.85) to assess their attitudes toward drinking and driving: I think more time and effort should be spent patrolling for drunk drivers, I think drunk driving is a very serious problem, I think people who drink and drive have very low moral standards, I think penalties for drunk driving should be stiffened, and I think drunk driving is made out to be a bigger problem than it is. The advocated message in both conditions recommended that participants not drink and drive. Thus, a high score on the

6 280 SHEHRYAR AND HUNT scale indicates that participants agreed with the recommendation, resulting in message acceptance. Confound Check To support our argument that the nature of the threatening consequence contained in a fear-arousing message influences the elicited response to that message, it is necessary to account for the level of fear evoked by the four stimulus ads. We measured level of fear using adjectives describing fearful states. The adjectives were selected from the Positive and Negative Affect Schedule (Watson, Clark, & Tellegen, 1988), and participants were instructed to rate their feelings directly in response to the stimulus advertisements. The scales included such adjectives as frightened, interested, inspired, excited, and nervous, among others and were rated on a 7-point scale ranging from 1 (strongly disagree) to 7 (strongly agree). Several past studies have used identical or similar adjectives when measuring fear arousal (see Block & Keller, 1998; Gleicher & Petty, 1992; Keller & Block, 1996; Maddux & Rogers, 1983). The complete Positive and Negative Affect Schedule consists of 22 positive and negative states. We used principal-component analysis to arrive at a factor labeled Level of Fear, which included the affective states labeled distressed, frightened, afraid, upset, anxious, scared, and nervous (α =.89). A second potential confound could arise if participants do not consider themselves susceptible to the consequences depicted in the stimulus advertisements. In fact, one explanation for message rejection is that young adults do not consider themselves vulnerable to the consequences depicted in fear-arousing communications (Pechmann & Shih, 1999). If, despite random assignment, participants are not homogeneous in terms of perceived susceptibility to threat, then highly susceptible participants could exhibit greater message acceptance, and vice versa. To account for this, we measured participants perceived susceptibility to the depicted threat with the following two questions: How likely is it that you will be involved in a drunk driving accident that is your fault? and How likely is it that you could be involved in an alcohol-related traffic violation? The responses were scored on a 6-point scale ranging from 1 (impossible),2(highly unlikely),3(unlikely),4(likely),5(highly likely), to 6 (inevitable; α =.86). Procedure Participants were instructed to look through the booklet as they would normally look through a magazine. We screened all filler material and found no content that could potentially confound the study. Participants answered questions about the filler advertisement for the hybrid automobile as well as the issue of protecting the environment. A page containing the fear appeal or Jell-O advertisement followed these questions. Immediately following the main stimulus, participants were given a word search task, to introduce a time delay between the stimulus and participants responses. By introducing a time delay after message exposure we attempted to rule out immediate suppression or defensive avoidance that may otherwise influence the results. The word search task itself consisted of a search for neutral words related to television viewing, such as soap opera, actor, and drama. Next, participants completed the word fragment completion task and the items to measure level of fear and answered questions measuring perceived susceptibility to threat and message acceptance or rejection. Finally, participants recorded their thoughts on the purpose of the study. None of the participants were able to guess the intent of the study. Analysis and Results Manipulation check. To test whether the advertisement that threatened participants with death as a consequence made mortality salient, it is important to determine whether participants in the mortality-salient condition completed more death-related words on the word fragment completion task than participants in the other conditions. A one-way analysis of variance (ANOVA) indicated that the effect of nature of threatening consequence was significant, F(3, 177) = 31.8, p <.001. A post hoc comparison with Bonferroni correction indicated that participants who viewed the funeral advertisement completed more death-related word fragments than participants in any other condition. Means are as follows: death, M = 2.45 (SD = 0.59); serious injury, M = 1.22 (SD = 0.94); arrest, M = 1.35 (SD = 0.70); and control, M = 0.95 (SD = 0.81). All paired comparisons testing the difference in mean numbers of death-related words completed in the mortality-salient condition versus the other conditions were statistically significant (p <.001). The same post hoc Bonferroni test showed that paired comparisons testing the difference in numbers of death-related words among the other three conditions were not significantly different at p <.1. Thus, the mortality-salience manipulation was successful. We used a median split to create the low and high conditions for precommitment to consuming alcohol (median value of 3.0 on a 7-point scale; M = 2.57, minimum = 1.0, maximum = 6.6). Confound check. It is critical to test for any difference in level of fear between the mortality-salience condition and the other three conditions. Unless the level of fear is similar, results cannot be confidently attributed to the nature of threatening consequences in the advertisement. Means for level of fear are as follows: death, M = 3.95 (SD = 1.42); serious injury, M = 3.62 (SD = 1.77); and arrest, M = 3.27 (SD = 1.41). An ANOVA comparing the participants self-rated level of fear across the three fear appeals was not significant, F(2, 133) = Planned contrasts confirmed that there was no significant difference in participants self-rated level of

7 TERROR MANAGEMENT THEORY AND FEAR APPEALS 281 fear between either the fear-of-death and fear-of-arrest conditions, t(86) = 1.18, ns, or the fear-of-death and fear-of-serious-injury conditions, t(84) < 1, ns. Thus, level of fear can be ruled out as a potential confound. It must be noted that another ANOVA and a post hoc Scheffé test across all four conditions, including the control condition, showed that the mean level of fear in the control condition was significantly different from the levels of fear in the three experimental conditions, F(3, 171) = 19.06, p <.001 (control: M = 1.72, SD = 0.71). This significant difference is to be expected because the control advertisement is not a fear-arousing advertisement. This further confirms that, unlike a neutral advertisement, all fear appeals indeed aroused fear among participants. Test of hypothesis. We predicted that message acceptance or rejection would be influenced by an interaction between type of threat and precommitment. We conducted an ANOVA with type of threat and precommitment as the two between-subjects factors and message acceptance or rejection as the dependent variable. The omnibus F test was significant, F(7, 178) = 6.43, p <.001. The means are presented in Figure 1. The predicted interaction between type of threat and precommitment was statistically significant, F(3, 178) = 7.43, p <.001. One can observe in Figure 1 that when mortality was salient, participants who had high precommitment rejected the message, whereas participants who had low precommitment accepted the message. Planned contrasts indicated that the difference between the two groups was statistically significant (Ms = 4.01 vs. 5.23), t(42) = 5.02, p <.001. Planned contrasts also indicated that high precommitment alone did not influence responses because participants with high precommitment were not different from participants with low precommitment in message acceptance or rejection in either the serious injury condition (Ms = 5.00 vs. 5.16), t(44) < 1, ns, or the arrest condition (Ms = 5.19 vs. 5.13), t(46) < 1, ns. As predicted, message rejection occurred only when fear of death was the threatening consequence and a worldview was challenged, but not when a qualitatively different threat, the threat of arrest or serious injury, was used. It is worth noting that a planned contrast indicated that in the control condition there was a significant difference in message acceptance between high- and low-precommitment participants (Ms = 4.79 vs. 5.22), t(38) = 1.94, p =.054. However, compared with the mean for message acceptance for high-precommitment participants in the arrest (M = 5.19) and serious injury (M = 5.00) conditions, the difference was not significant, t(168) < 1. This supports the finding that adherence to a worldview occurred only when mortality was made salient for participants with a high precommitment to that worldview. These results provide support for H1. Perceived susceptibility and message acceptance or rejection. To study the role of perceived susceptibility to a threatening consequence, we substituted precommitment to drinking with perceived susceptibility as an independent variable and conducted another ANOVA with type of threat as the other independent variable and message acceptance or rejection as the dependent variable. This analysis can help determine whether precommitment provides a vantage point that is different than one offered by perceived susceptibility. The omnibus F test was not significant, F(7, 178) = The main effect of susceptibility on message acceptance or rejection also was not significant, F(1, 177) < 1. Finally, the interaction between perceived susceptibility and type of fear also did not achieve statistical significance, F(3, 177) < 1. Overall, perceived susceptibility did not affect persuasion. This suggests that when studying responses to fear appeals, precommitment to an undesirable behavior manifests itself differently than perceived susceptibility to a threat. From a TMT perspective, precommitment to a behavior is a gauge of self-esteem that accrues when one engages in that behavior under the duress of an imminent threat of death. In contrast, susceptibility to a threat neither offers death transcendence nor protects one from the lingering anxiety of annihilation. Hence, it is important to distinguish between the role of susceptibility and precommitment and to recognize the significance of precommitment when extending TMT into the fear-appeal domain. FIGURE 1 Message acceptance or rejection as a function of type of threat and precommitment to drinking alcohol. Discussion We proposed that the qualitative nature of the threatening consequence of a fear-appeal message influences the elicited response to that message. We used TMT to explain message rejection and predicted that message rejection would be ob-

8 282 SHEHRYAR AND HUNT served if the threat consisted of death as a consequence but not if the threat consisted of a qualitatively different type of consequence, such as physical harm or social embarrassment. Finally, we proposed that it is possible to predict message rejection by using the target audience s level of precommitment to a specific worldview as a segmentation variable. We predicted that death-related fear appeals would be persuasive only when precommitment to consuming alcohol is low, whereas non-death-related fear appeals would be equally persuasive when precommitment is low or high. Our data support this hypothesis. In the control condition, high-precommitment participants exhibited significantly lower message acceptance than did low-precommitment participants. This suggests that in the absence of any threat, people with a greater affinity for an undesirable behavior are predisposed to weaker intentions of giving up that behavior. This finding is consistent with results of past research (Keller, 1999). However, in our study we found that even when individuals are highly precommitted to behaviors that are socially undesirable, compliance with an advocated message can occur if the threatening consequences are nonfatal. It is this finding that differentiates our research from past research on fear appeals. We discuss this finding in detail next, and we explain when message rejection may occur by showing that threat of fatality will elicit stronger adherence to undesirable behaviors among audience members who derive self-esteem from the focal behavior. Nature of fear versus level of fear. According to TMT, reminders of mortality elicit different responses than reminders of equally noxious but non-death-related consequences. It may be argued that death is the ultimate consequence and is inherently more fear arousing than any other threat. However, getting arrested and being put in jail for drinking and driving poses a significant threat because of potentially embarrassing and troublesome consequences. In fact, fear of arrest may be a more realistic threat to young adults than fear of death because young adults consider themselves invulnerable to potential physical harm (Arnett, 2000; Pechmann & Shih, 1999). The fact that we did not find any differences in level of fear across the three types of threat supports this notion and suggests that young adults felt equally threatened by all three consequences. However, the nature of the threat was different because participants completed more death-related words in the mortality-salience condition than in all other conditions. Mortality salience and message rejection. We predicted and found that participants rejected the persuasive message when death was the threatening consequence but not when arrest or physical harm constituted the threat. This finding has implications for classification of fear appeals based on the purpose of the communication. Specifically, the nature of the threatening consequence is critical to consider depending on whether the intent of the message is behavior prevention (i.e., target recipients are not a priori committed to the focal behavior) or behavior cessation (i.e., target recipients are already committed to the focal behavior). It should be noted that a classification of individuals based on our notion of precommitment to a self-esteem-enhancing behavior is essentially different than a classification based on behavioral labels such as unconverted and adherents as defined by Keller (1999). According to Keller, the unconverted are those who partake in the undesirable behavior. The unconverted are persuaded by low-fear-arousing messages because they tend not to discount such messages, whereas they suppress high-fear-arousing messages (Keller, 1999; Liberman & Chaiken, 1992). We found that if initial suppression is accounted for by a delay, then unconverted individuals indeed accept fear-arousing messages. However, the converted unconverted dichotomy does not explain why after a delay only mortality salience results in message rejection. This result is consistent with the TMT explanation that precommitment to a worldview does not lead to worldview defense unless mortality is made salient. Study 1 was designed to test the central hypothesis that the qualitative nature of the fear appeals influences the responses they evoke among audiences with low and high precommitment to an undesirable behavior. In Study 2 we addressed the notion that responses measured with a delay are different than responses measured without a delay. Testing audience response both with and without a delay can provide converging evidence for the results of Study 1 and test the premise that the psychodynamics of coping with the fear of death are different than those of coping with non-death-related fears. In Study 1, we used the word completion task as a manipulation check for mortality salience. It is probable that while working on this task, participants in the experimental conditions, including the mortality-salient condition, may fill out death-related words by chance alone and in doing so evoke mortality salience. To rule out this potential confound, participants responses must also be recorded without the death-related word completion task. The design of Study 2 addressed this concern. STUDY 2 Experimental Design The second study involved 240 undergraduate students enrolled in business courses in a U.S. university. Similar to the method adopted in Study 1, precommitment to consuming alcohol as a source of self-esteem was measured 1 month before the main study. Because of attrition, 236 participants remained in the main study. The measures and stimuli were identical to the ones used in Study 1, with the exception of two notable changes. First, to address the issue of immediate suppression following a threat, the Study 1 design was supplemented with a third factor, called delay between stimulus

9 TERROR MANAGEMENT THEORY AND FEAR APPEALS 283 and response (delay), with two levels: (a) delay and (b) no delay. Thus, the study had a between-subjects design with four types of advertisements, including three types of threat and a Jell-O advertisement as control, two levels of precommitment, and two levels of delay. To manipulate the delay factor, the neutral word search task requiring a television-related word search was retained in the delay condition but removed from the no-delay condition. Analysis and Results Confound check. We conducted an ANOVA to test whether there were any significant differences in the level of fear perceived across the four types of advertisements and the delay conditions. A Nature of Threat Delay interaction, with level of fear as the dependent variable, was not significant, F(3, 236) = The main effect of delay was not significant either, F(1, 236) < 1. Therefore, the main effect of type of fear was analyzed next. Results of an ANOVA indicated that there was a significant difference in the level of fear experienced across the conditions, F(3, 236) = 92.40, p <.001. A post hoc test with Bonferroni correction indicated that the level of fear experienced in the control condition was significantly lower than that experienced across the other conditions. Because a fear appeal was not used in the control condition, this result is expected. The paired comparisons for the level of fear between the control condition and each of the three fear-appeal conditions were statistically significant (p <.001). More important, paired comparisons indicated no significant difference in the self-reported level of experienced fear among participants exposed to the death, serious injury, and arrest advertisements. Means for level of fear are as follows: death, M = 3.99 (SD = 1.34); serious injury, M = 4.19 (SD = 1.36); arrest, M = 4.21 (SD = 1.29); and control, M = 1.19 (SD = 0.28). Thus, level of fear can be ruled out as a potential confound. Test of hypotheses. We conducted an ANOVA with type of threat, precommitment, and delay as the three between-subjects factors and message acceptance or rejection as the dependent variable. The omnibus F test was significant, F(15, 237) = 6.08, p <.001. The three-way interaction among the factors was statistically significant, F(3, 237) = 2.66, p <.05. A two-way interaction between nature of fear and delay also was statistically significant, F(3, 237) = 3.12, p <.05. We predicted in H2 that if an audience s responses to fear appeals are measured without a delay, then participants with a high precommitment to an undesirable behavior are likely to exhibit lower levels of persuasion than those with a low precommitment to an undesirable behavior. A planned contrast in the no-delay condition indicated that there was a marginally significant difference in message acceptance or rejection between the low- and high-precommitment conditions across the three fear appeals, F(1, 221) = 2.87, p =.092. To pinpoint the source of this effect, we carried out specific pairwise comparisons. Pairwise comparisons indicated that in the no-delay condition there was a significant difference in persuasion between high- and low-precommitment participants in the control, arrest, and serious injury conditions but not in the death condition; high-precommitment participants had significantly lower message acceptance than low-precommitment participants in the control condition, t(28) = 3.46, p <.05 (Ms = 5.78 for low precommitment and 4.85 for high precommitment); the fear-of-arrest condition, t(28) = 2.00, p =.054 (Ms = 5.57 for low precommitment and 4.96 for high precommitment); and the serious-injury condition, t(34) = 2.75, p =.006 (Ms = 5.20 for low precommitment and 4.57 for high precommitment). For the threat of death, however, when measured in the no-delay condition, there was no significant difference in message acceptance or rejection between participants with high and low precommitment to consuming alcohol, t(22) < 1, ns (Ms = 5.23 for low precommitment and 4.93 for high precommitment). This is counter to the expectation that respondents with high precommitment will show less persuasion compared with the respondents with low precommitment. However, a post hoc contrast showed that when means of message acceptance were compared only among the high-precommitment participants, the mean in the mortality-salience condition was not significantly different than the means in the other three conditions, t(168) < 1, ns. Thus, if suppression occurred immediately after a delay, all respondents with high precommitment toward drinking alcohol expressed it at a statistically similar level across the three types of fear appeals. Overall, as expected, in the no-delay condition the means for persuasion in the high precommitment conditions were lower than the means for persuasion in the low precommitment conditions, indicating that participants with high precommitment were more likely to suppress the message than were those with low precommitment. The means are presented in Figure 2. Hypothesis 3 predicted that when responses are measured with a delay a distal response will be evoked and the message will be rejected by participants with high precommitment, but only when mortality is made salient. In contrast, when the threat is not death related and responses are measured with a delay, a danger-control mechanism will be engaged and the message will be accepted even if the respondents have high precommitment to the undesirable behavior. To test H3, we compared the means for message acceptance or rejection for respondents with high precommitment to drinking alcohol between the delay and no-delay conditions across the three fear appeals. Planned contrasts indicated that for high-precommitment participants in the arrest condition there was no significant difference in message acceptance between the delay and no-delay condition, t(30) < 1. However, the means for message acceptance or rejection were in a direction consistent with the prediction in H3. When message acceptance or rejection was measured with a delay, the mean was higher than in the no-delay condition (arrest: Ms = 5.22 in the delay condition vs in the no-delay condi-

10 284 SHEHRYAR AND HUNT FIGURE 2 Message acceptance or rejection as a function of type of threat and precommitment to drinking alcohol in the no-delay condition. FIGURE 3 Message acceptance or rejection for high-precommitment respondents as a function of type of threat and delay. tion). A similar planned contrast for the serious-injury condition revealed that message acceptance for high-precommitment respondents was significantly higher in the delay condition than the no-delay condition, t(35) = 3.42, p <.001. The mean for message acceptance was higher when measured with a delay, indicating that danger-control processes led to message acceptance after initial suppression was accounted for (serious injury: Ms = 5.33 in the delay condition vs in the no-delay condition). Finally, a planned contrast within the death condition revealed a significant difference in persuasion between high-precommitment participants in the delay and no-delay conditions, t(22) = 2.79, p =.005. Contrary to the arrest condition and the serious-injury condition, the direction of the mean difference indicated that the message was rejected in the delay condition, because the mean was significantly lower than the mean in the no-delay condition (death: Ms = 4.19 in the delay condition vs in the no-delay condition). This finding underscores the difference in cognitive responses to qualitatively different threats. Taken together, results from the tests of H2 and H3 indicate that, following fear appeals, the message s recommendation is suppressed across all types of fears. However, after a delay, non-death-related fear appeals result in compliance with the message even for participants with high precommitment. On the other hand, a fear appeal that makes mortality salient results in the activation of a distal defense that leads to message rejection and continued maladaptive behavior. The means for message acceptance for high-precommitment participants across the four types of advertisements and the two levels of delay are presented in Figure 3. Discussion We conducted Study 2 to provide converging evidence that, unlike non-death-related threats, the threat of death evokes unique responses after accounting for initial suppression. In the no-delay condition, suppression of the message did indeed occur. This finding validates the usefulness of the delay task used in Study 1 to contain initial suppression. As expected, when responses were measured after a delay, participants with high precommitment rejected the message following a mortality-salient fear appeal. This finding is characteristic of the distal defense mechanism espoused by TMT. Moreover, message acceptance following a threat of nonfatal consequences also provides evidence that a danger-control mechanism yields a different response than distal defense and that the nature of the threatening consequence is responsible for this difference. The addition of the delay condition in Study 2 allowed for a comparison between delay and no-delay conditions, thereby providing converging evidence for the presence of unique responses to qualitatively different threats. Finally, in Study 2 we successfully replicated the results of Study 1 without any potential confounding effects of the manipulation check involving the death-related word completion task, which was excluded. GENERAL DISCUSSION We found that messages accompanying fear appeals composed of threats that make mortality salient elicit unique responses among individuals who derive self-esteem from

11 TERROR MANAGEMENT THEORY AND FEAR APPEALS 285 participating in the targeted behavior. Specifically, when threatened with death as a consequence, individuals who reported deriving self-esteem from drinking alcohol espoused more maladaptive responses than individuals who did not report deriving self-esteem from drinking alcohol. Our results suggest that differentiating fear appeals that use the threat of death as a consequence from those that do not is important for understanding the efficacy of fear-appeal communication and predicting when fear appeals do not work. The extant research on the efficacy of fear appeals has explored the persuasiveness of fear-appeal communications in response to manipulations of the level of fear evoked by a message. Although recent evidence suggests that high-fear appeals are more effective than low-fear appeals at persuading audiences (Witte & Allen, 2000), research has not shown with certainty what constitutes a high-fear appeal and what differentiates a high-fear appeal from a moderate-fear appeal (Keller, 1999). This has led researchers to suggest that levels of fear aroused in the laboratory may not mimic levels of fear experienced in real life (Keller, 1999). Moreover, LaTour and Rotfeld (1997) lamented that level of fear is a loosely defined concept, and they called it intuitive and ad hoc (p. 46). We found differences in fear appeal persuasiveness by holding constant the level of fear and manipulating the mortality salience evoked by an advertisement. Thus, we provide an alternative approach that avoids relying solely on level of fear as a determinant of the persuasiveness of fear appeals. Although our research was designed primarily to supplement extant fear appeal research, we also make a contribution to TMT. In a TMT domain, Taubman Ben-Ari, Florian, and Mikulincer (1999) demonstrated that mortality salience led to increased reckless driving among participants who indicated that driving was highly relevant to their self-esteem but not among participants who reported that driving was less relevant to their self-esteem. Because Taubman Ben-Ari et al. studied risky behaviors exclusively in a TMT context, they did not compare salience of death-related consequences to non-death-related consequences. In our research, we extend their work into the fear-appeal domain and explicitly compared qualitatively different types of threatening consequences used in fear appeals. In so doing, we examine a specific worldview heretofore unstudied in the TMT domain and offer validation for the significance of relatively distorted beliefs in the constitution of cultural worldviews. The issue of distorted worldviews is especially relevant to youth. For youth, the fear aroused by an advertisement s threatening consequences may be less threatening than the fear of following the message s recommendations. If their worldview is distorted, the consequence of following the advertisement s recommendation may include being cast out of a network of peers that offers support and lends meaning to their existence. As we found in our study, defense of worldviews occurs even if worldviews are distorted. Past research explains why high fear is suppressed in reaction to fear appeals. For instance, Keller and Block (1996) found that when high-fear messages are ineffective it is because too much elaboration on the harmful consequences interferes with processing of the recommended change in behavior. Similarly, Keller (1999) found that people who did not already follow a message s advocated recommendation were more likely to discount the message when the level of fear was high than when the level of fear was low. These studies provide important insights into the reason why different levels of fear result in differing degrees of persuasion. However, they do not explain why fear-appeal communications sometimes elicit maladaptive responses. Based on the unique psychodynamics evoked by the fear of death, the TMT perspective adopted in this study fills this gap by offering an a priori explanation for when fear-appeal messages may produce message rejection. Thus, it is likely that individuals who reject messages may accept them after an initial delay. However, if the message makes mortality salient, then a delay may not render the message effective. Previous studies have demonstrated that a recipient s perceived susceptibility to a threatening consequence influences the persuasiveness of the message. In our research, we account for susceptibility to threat. Our results indicate that susceptibility is conceptually different from commitment to a worldview. The results provide empirical support for a TMT-based view suggesting that the differences observed in level of persuasion can be attributed to cognitive mechanisms activated by defense of one s worldview in the wake of mortality salience. Current theoretical approaches suggest examining fear-appeal efficacy along dimensions of level of fear and converted versus unconverted recipients. Our results suggest that researchers should add to that list the dimension of nature of the threatening consequence and distinguish between advertisements that evoke mortality salience and those that do not. LIMITATIONS AND CONCLUSION Like all scientific research, our study has limitations. Because of the sensitive nature of drinking and driving, participants who expressed low precommitment may be prone to expressing inflated estimates of persuasion. By temporally separating the measurement of precommitment from the measurement of the study s main variables, we attempted to reduce the influence of this potential bias. Nevertheless, our results should be interpreted with this in mind. Alcohol consumption, for some individuals, is an addictive behavior. Our study did not account for the influence of alcohol addiction. The relation of precommitment to an addictive behavior is less clear. For instance, an alcoholic person likely does not derive self-esteem from her or his addiction. In fact, shame inflicted by family and friends may erode the self-esteem of an addict. Given the common use of fear appeals to curb addictive behaviors, it would be interesting to explore which worldview would be used by addicts as an

12 TERROR MANAGEMENT THEORY AND FEAR APPEALS 286 anxiety buffer to assuage the fear of death aroused by advertisements. It seems that unless the advocated message promises replacing the delinquent worldview with one that safeguards addicts equally well, converting addicts will be more challenging than converting those with high precommitment alone. We concur with Maheswaran and Agrawal s (2004) reasoning that the interaction of worldview defense and striving for self-esteem may invoke conflicting motivations in certain situations. Future research can integrate dual-process models of persuasion with TMT to better understand phenomena in which conflicting motivations underlie behavioral outcomes. Consumer research should also study mortality-salience issues outside of a fear-appeal domain. Many advertisements that do not follow the traditional fear-appeal format of presenting harmful consequences and offering a solution use themes of death in their messages. Often, themes of death are combined with humor. A TMT perspective may offer better insight into the attitudinal influence of evoking mortality salience in advertisements that incorporate themes other than fear. Moreover, everyday reminders of death are virtually unavoidable in news and entertainment media. Our study offers a perspective through which to explore how consumers are influenced by such incidental reminders of their mortality. ACKNOWLEDGMENTS We thank Jamie Arndt for his help with instrument development and for his valuable comments on an earlier version of this article. We also thank the two anonymous reviewers and the associate editor for their constructive comments and guidance. Both authors contributed equally to the study. REFERENCES Arndt, J., Goldenberg, J. L., Greenberg, J., Pyszczynski, T., & Solomon, S. (2000). Death can be hazardous to your health: Adaptive and ironic consequences of defenses against the terror of death. In P. R. Duberstein & J. M. Masling (Eds.), Psychodynamic perspectives on sickness and health (pp ). Washington, DC: American Psychological Association. Arndt, J., Greenberg, J., Solomon, S., Pyszczynski, T., & Simon, L. (1997). Suppression, accessibility of death-related thoughts, and cultural worldview defense: Exploring the psychodynamics of terror management. Journal of Personality and Social Psychology, 73, Arndt, J., Solomon, S., Kasser, T., & Sheldon, K. M. (2004). The urge to splurge: A terror management account of materialism and consumer behavior. Journal of Consumer Psychology, 14, Arnett, J. J. (2000). Optimistic bias in adolescent and adult smokers and nonsmokers. Addictive Behaviors, 25, Bassili, J., & Smith, M. C. (1986). On the spontaneity of trait attribution: Converging evidence for the role of cognitive strategy. Journal of Personality and Social Psychology, 50, Block, L. G., & Keller, P. A. (1998). Beyond protection motivation: An integrative theory of health appeals. Journal of Applied Social Psychology, 28, Engs, R. C., Diebold, B. A., & Hanson, D. J. (1996). The drinking patterns and problems of a national sample of college students. Journal of Alcohol and Drug Education, 41, Engs, R. C., & Hanson, D. J. (1988). University students drinking patterns and problems: Examining the effects of raising the purchase age. Public Health Reports, 103, Gleicher, F., & Petty, R. E. (1992). Expectations of reassurance influence the nature of fear-stimulated attitude change. Journal of Experimental Social Psychology, 28, Glindemann, K. E., Geller, E. S., & Fortney, J. N. (1999). Self-esteem and alcohol consumption: A study of college drinking behavior in a naturalistic setting. Journal of Alcohol and Drug Education, 45, Greenberg, J., Pyszczynski, T., & Solomon, S. (1986). The causes and consequences of a need for self-esteem: A terror management theory. In R. F. Baumeister (Ed.), Public self and private self (pp ). New York: Springer. Greenberg, J., Pyszczynski, T., Solomon, S., Simon, L., & Breus, M. (1994). The role of consciousness and accessibility of death-related thoughts in mortality salience effects. Journal of Personality and Social Psychology, 67, Greenberg, J., Solomon, S., & Pyszczynski, T. (1997). Terror management theory of self-esteem and cultural worldviews: Empirical assessments and conceptual refinements. Advances in Experimental Social Psychology, 29, Ho, R. (1998). The intention to give up smoking: Disease versus social dimension. Journal of Social Psychology, 138, Hovland, C. I., Janis, I. L., & Kelley, H. H. (1953). Communication and persuasion. New Haven, CT: Yale University Press. Janis, I. L., & Feshbach, S. (1953). Effects of fear-arousing communications. Journal of Abnormal and Social Psychology, 48, Keller, P. A. (1999). Converting the unconverted: The effect of inclination and opportunity to discount health-related fear appeals. Journal of Applied Psychology, 84, Keller, P. A., & Block, L. G. (1996). Increasing the persuasiveness of fear appeals: The effect of arousal and elaboration. Journal of Consumer Research, 22, LaTour, M. S., & Rotfeld, H. J. (1997). There are threats and (maybe) fear-caused arousal: Theory and confusion of appeals to fear and fear arousal itself. Journal of Advertising, 26, Liberman, A., & Chaiken, S. (1992). Defensive processing of personally relevant health messages. Personality and Social Psychology Bulletin, 18, Maddux, J., & Rogers, R. (1983). Protection motivation and self-efficacy: Revised theory of fear appeals and attitude change. Journal of Experimental Social Psychology, 19, Maheswaran, D., & Agrawal, N. (2004). Motivational and cultural variations in mortality salience effects: Contemplations on terror management theory and consumer behavior. Journal of Consumer Psychology, 14, McCormack, A. S., Laybold, A. M., Dickerson-Nelson, J., & Budd, C. F. (1993). Stress and substance use: Students attitudes toward alcohol, marijuana, and cocaine. College Student Journal, 27, McGregor, H. A., Lieberman, J. D., Greenberg, J., Solomon, S., Arndt, J., & Simon, L. (1998). Terror management and aggression: Evidence that mortality salience motivates aggression against worldview-threatening others. Journal of Personality and Social Psychology, 74, National Highway Traffic Safety Administration. (2003). Alcohol involvement in fatal motor vehicle traffic crashes. Retrieved April 10, 2005, from 30/NCSA/Rpts/2005/ pdf Neese, W. T., & Taylor, R. D. (1994). Verbal strategies for indirect comparative advertising. Journal of Advertising Research, 34, Pechmann, C., & Shih, C. (1999). Smoking scenes in movies and antismoking advertisements before movies: Effects on youth. Journal of Marketing, 63, Pechmann, C., Zhao, G., Goldberg, M. E., & Reibling, E. T. (2003). What to convey in antismoking advertisements for adolescents? The use of protec-

13 287 SHEHRYAR AND HUNT tion motivation theory to identify effective message themes. Journal of Marketing, 67, Pyszczynski, T., Wicklund, R., Floresku, A., Koch, H., Gauch, G., Solomon, S., et al. (1996). Whistling in the dark: Exaggerated estimates of social consensus in response to incidental reminders of mortality. Psychological Science, 7, Rippetoe, P. A., & Rogers, R. W. (1987). Effects of components of protection-motivation theory on adaptive and maladaptive coping with a health threat. Journal of Personality and Social Psychology, 52, Rogers, R. W. (1983). Cognitive and physiological processes in fear appeals and attitude change: A revised theory of protection motivation. In J. T. Cacioppo & R. E. Petty (Eds.), Social psycho-physiology (pp ). New York: Guilford. Rogers, R. W. (1985). Attitude change and information integration in fear appeals. Psychological Reports, 56, Rosenblatt, A., Greenberg, J., Solomon, S., Pyszczynski, T., & Lyon, D. (1989). Evidence for terror management theory: I. The effects of mortality salience on reactions to those who violate or uphold cultural values. Journal of Personality and Social Psychology, 57, Schoenbachler, D. D., & Whittler, T. E. (1996). Adolescent processing of social and physical threat communications. Journal of Advertising, 25, Substance Abuse and Mental Health Services Administration. (2002). Alcohol and tobacco use tables, Table H. 37. Retrieved June 21, 2004, from Tanner, J. F., Hunt, J. G., & Eppright, D. R. (1991). The protection motivation model: A normative model of fear appeals. Journal of Marketing, 55, Taubman Ben-Ari, O., Florian, V., & Mikulincer, M. (1999). The impact of mortality salience on reckless driving: A test of terror management mechanisms. Journal of Personality and Social Psychology, 76, Watson, D., Clark, L. A., & Tellegen, A. (1988). Development and validation of brief measures of positive and negative affect: The PANAS scales. Journal of Personality and Social Psychology, 54, Wechsler, H., Davenport, A. E., Dowdall, G., Moeykens, B., & Castillo, S. (1994). Health and behavioral consequences of binge drinking in college: A national survey of students at 140 campuses. Journal of the American Medical Association, 272, Wechsler, H., Kuh, G. D., & Davenport, A. E. (1996). Fraternities, sororities, and binge drinking: Results from a national study of American colleges. National Association of Student Personnel Administrators Journal, 33, Wechsler, H., Molnar, B. E., Davenport, A. E., & Baer. J. S. (1999). College alcohol use: A full or empty glass? Journal of American College Health, 47, Wegner, D. M., & Erber, R. (1992). The hyperaccessibility of suppressed thoughts. Journal of Personality and Social Psychology, 63, Wischusen, J., Nelson, L. J., & Pollini, N. (2003). Nationalism or racism in response to mortality salience: Which worldview is defended in a situation of multiple group identities. Journal of Psychology and the Behavioral Sciences, 17. Retrieved March 3, 2005, from psychweb/vol16/wischusen.pdf Witte, K. (1992). Putting the fear back into fear appeals: The extended parallel process model. Communication Monographs, 59, Witte, K. (1994). Fear control and danger control: A test of the extended parallel process model (EPPM). Communication Monographs, 61, Witte, K., & Allen, M. (2000). A meta-analysis of fear appeals: Implications for effective public health campaigns. Health Education and Behavior, 27, Workman, M., & Beer, J. (1989). Self-esteem, depression, and alcohol dependency among high school students. Psychological Reports, 65, Received: July 11, 2004 Accepted: April 17, 2005