Another interesting aspect of these explanations is that rather than representing entirely new formulations, each attempt builds on, and includes elements from, earlier explanations. This is an interesting choice as the explication of each new explanation generally involves a scholarly thrashing of the predecessors.
The Drive Model
As with several other topics, the social scientific study of fear appeals largely begins with Hovland et al.’s (1953) Communication and Persuasion. Their fear appeal chapter offered considerable speculation (in part because there were very few data beyond Janis & Feshbach, 1953) concerning the persuasive effects of fear appeals. The chapter introduced many concepts that remain at the fore-front of fear appeal scholarship today. Moreover, their drive model dominated the fear appeals conversation for over 20 years. Although its predictions are inconsistent with the accumulated data, it remains important bedrock on which the foundations of other explanations are built.
Working from the classical conditioning paradigm, Hovland et al.’s (1953) drive model, depicts fear as a drive; that is, a negatively valenced state that an individual is motivated to avoid or (once experienced) eliminate. The explanation claims that the threat presented in a persuasive message initiates fear as a drive state. The emotional arousal generates a search for, and effort to work through, several strategies to reduce the drive. Whatever strategy is successful in reducing the drive state is considered a reward, thereby increasing the probability that it will be repeated in the future (Hovland et al., 1953). If rehearsing and advocating message recommendations reduces fear, they will be integrated into the receiver’s cognitive structure (i.e., attitude and behavior change occur) and this strategy will represent the preferred option for processing future messages. If, on the other hand, rehearsing message recommendations fails to reduce the drive, audience members will attempt alternative strategies. For example, Hovland et al. suggest that the receiver might attempt to disregard the message (i.e., inattentiveness), derogate the message source, or minimize or ignore the threat. Of course, if receivers take this route, attitude and behavior change seems unlikely. Again, if any of these alternative strategies are successful in reducing drive, it is rewarding and represents the response of choice in similar future situations.
From the drive model, Hovland et al. (1953) predicted a curvilinear (i.e., inverted-U shaped) relationship between the strength of a fear appeal and attitude and behavior change. Specifically, “from zero to some moderate level, acceptance tends to increase, but as emotional tension mounts to higher levels, acceptance tends to decrease” (Hovland et al., 1953, pp. 83–84). In short, the drive model predicts that greatest levels of attitude and behavior change will occur when fear appeals are moderate (when compared to when they are either low or high).
Two noteworthy extensions of the drive model are Janis’s (1967) “family of curves” and McGuire’s (1968, 1969) two-factor explanation. Janis argues that fear arousal creates competing forces of vigilance to the threat, on the one hand, and hypervigilance (i.e., interference) on the other. Increasing fear arousal facilitates audience members’ motivation and ability to process the persuasive message, however, as fear arousal increases further, facilitating factors are countered by interference (e.g., defensive avoidance through counterarguing in a search for loopholes; Janis, 1967, p. 176). The point on the arousal continuum at which inhibiting factors outweigh facilitating factors varies across messages and audiences, resulting in a family of curves where the function’s peak falls at different points on an emotional arousal continuum.
McGuire (1968, 1969), on the other hand, argued that when a persuasive message creates fear, the emotion can act as a cue or as a drive. As a cue, receivers depend on learned responses that interfere with message reception and acceptance. On the other hand, as a drive, fear motivates receivers to avoid the threat and is thought to facilitate message acceptance. Both drive and cue functions increase monotonically as fear arousal increases, but at differing rates. This combination of competing forces leads to the predicted inverted U-shaped relationship.
Evaluating the Drive Model
One important criticism of the drive model focuses on the representation of fear as a drive. Equating these terms essentially places emotion in the same category as hunger or thirst. Thus, although it refers to emotional arousal, does little more than wave its hand toward emotion as it is considered in more recent theory and research.
What is more, the curvilinear prediction between fear appeal strength and responses derived from the drive model was doomed from the beginning. Although social science theorizing at the day was replete with inverted-U-shaped functions (Leventhal, 1970), pairing the drive model with a curvilinear relationship was a curious choice. The most complete data that Hovland et al. (1953) reported was the negative linear relationship on behavior change from Janis and Feshbach (1953). It is difficult to shoehorn these data into the drive model’s inverted U-shaped function. Hovland et al. claim that the negative linear relationship represents the downward slope in the inverted-U (see also Janis, 1967). For this to be the case, however, Janis and Feshbach’s minimal appeal would have to be at least moderate on some objective fear arousal continuum. This supposition is handicapped because Hovland et al. also went into great detail describing the truly mundane nature of the minimal appeal. In summary, evidence consistent with the curvilinear relationship in subsequent research is virtually nonexistent (Boster & Mongeau, 1984; Witte & Allen, 2000), although I will argue later that the methods typically used in fear appeal studies do not fully test the model.
Given the drive model’s inability to explain the accumulated data, subsequent explanations attempted to explain both positive linear and negative linear relationships between the strength of fear appeals and responses. Specifically, similar to McGuire’s (1968) two-factor approach, these explanations present multiple ways of processing fear-arousing messages (e.g., Das, de Wit, & Stroebe, 2003; Leventhal, 1970; Witte, 1993). Typically, one way of dealing with fear appeals facilitates message acceptance while the other inhibits it. Moreover, over time, the next generation of explanations focused increasingly on the cognitive, rather than affective, processes.
Parallel Response Model
Although other explanations appeared in the interim (e.g., G. R. Miller, 1963), the second historically important fear appeal explanation is the parallel response model (Leventhal, 1970, 1971). This explanation acts as a bridge between the largely affectively focused drive model and the more recent, cognitively oriented, explanations. Specifically, the parallel response model suggests that audience members will deal with the depicted threat in one of two ways: fear control or danger control.
Fear Control
In fear control, internal affective responses (e.g., physiological arousal) are interpreted as a sign of fear. When further cognitive work suggests that there is nothing to be done to avoid the threat, audience members will look internally to reduce emotional arousal. Leventhal (1970) argues that a wide variety of responses might occur during fear control (e.g., defensive reactions, eating or drinking to interfere with emotional responses, or reinterpreting the arousal). Audience members’ internal focus likely inhibits focusing on, elaborating about, and accepting message recommendations. When audience members engage in fear control, then, the strength of fear appeals should be inversely related to attitude and behavior change.
Danger Control
Danger control represents a problem-solving process whereby recipients consider how to deal with the message threat. Information guiding problem-solving typically comes from the message itself, though other sources of information (e.g., memory, coping abilities, and emotional responses) are also important, particularly as the problem-solving process develops (Leventhal, 1970). In other words, danger control increases the probability that receivers will attend to, elaborate on, and accept message recommendations. Therefore, when audience members engage in danger control, the strength of fear appeals should be positively correlated w
ith attitude and behavior change.
As the model’s label suggests, danger control’s problem solving process works parallel to, and independently from, fear control emotional processes. Thus, emotional arousal (i.e., fear) should be positively correlated with attitude and behavior change, however, this relationship is spurious as both responses are caused by the message (Leventhal, 1970). Thus, fear appeals generate both fear and adaptive responses; however, they are not causally related.
Evaluating the Parallel Processing Model
The primary value of the parallel processing model is historical, as it was the first explanation to clearly and explicitly separate cognitive from affective responses to a fear appeal. Therefore, the parallel processing explanation provided a logical explanation for the inconsistencies in fear appeal studies of the time. Specifically, studies generating negative relationships between fear appeals and attitude change (e.g., Janis & Feshbach, 1953) produced fear control while studies producing positive correlations (e.g., G. R. Miller & Hewgill, 1966) produced danger control.
Although this explanation facilitated thinking about fear appeals’ persuasive successes and failures, it had one serious flaw. Specifically, it lacked the specificity necessary to generate clear predictions and tests (Beck & Frankel, 1981; Boster & Mongeau, 1984; Rogers, 1975). Specifically, Leventhal (1970, 1971) neither developed operationalizations fear control and danger control, nor specified the conditions under which each would operate. (Ironically, Leventhal criticized both Janis’ family of curves explanation and the drive model on the same grounds.) Leventhal predicted that the strength of fear appeals would generally be positively related to attitude and behavior change. Sternthal and Craig (1974), however, used the same explanation to predict an inverted-U-shaped relationship between fear appeal strength and attitude and behavior change. Specifically, danger control would predominate from low to some moderate level of emotional arousal after which fear control would take over. Given the lack of specificity in construct specification and predictions, it is impossible to differentiate, or test, these predictions.
Therefore, although historically important for extending thinking about fear appeal processing, the parallel response model is not testable. Following the trend sweeping the social sciences, the next family of explanations (e.g., Beck & Frankel, 1981; Rogers, 1975) posited that attitude and behavior change was an exclusive function of the cognitive processing of fear appeals and virtually ignored emotional arousal (Witte, 1992). Although fear and perceptions of threat are “intricately and reciprocally related” (Witte & Allen, 2000, p. 592), the emotion of fear went AWOL from fear appeals for nearly two decades.
Protection Motivation Explanation
Rogers (1975) attributed the inconsistent fear appeal results to the haphazard manner in which fear appeals had been conceptualized and operationalized. Thus, the primary contribution of Rogers’ (1975, 1983) protection motivation explanation (both original and revised) is the specification of fear appeal components. Although these components had been previously described several times (e.g., Hovland et al., 1953; Janis, 1967; Leventhal, 1970), Rogers’s systematic descriptions generated greater consistency in the development of fear appeals and, perhaps, study results.
As previously noted, Rogers (1975, 1983) described fear appeal as containing four components (severity, susceptibility, self-efficacy, and response efficacy). Working from an expectancy value perspective (cf., chapter 8 in this volume), the protection motivation explanation posits that receivers evaluate messages along each message component (i.e., perceived severity, perceived susceptibility, perceived response efficacy, and perceived self-efficacy). In the initial formulation, these perceptual variables combined multiplicatively to create protection motivation, an intervening variable that “arouses, sustains, and directs activity” (Rogers, 1975, p. 94). Protection motivation, in turn, directly influences behavior change. Thus, protection motivation (akin to behavioral intention) replaces fear as the mediating variable between fear appeals and responses. “Our emphasis on ‘protection motivation’ rather than ‘fear’ is designed to emphasize the importance of cognitive processes rather than visceral ones” (Rogers, 1983, p. 169).
In short, the protection motivation explanation initially predicted a threat by efficacy interaction on attitude and behavior change. When either threat or efficacy levels are zero (e.g., either the threat is weak or if the coping response is ineffective), no attitude or behavior change is predicted to occur. Messages that present a severe threat that is likely to strike the audience member will be persuasive only if they are combined with recommended coping responses that are effective in avoiding the threat and that are in the receiver’s behavioral repertoire.
Evaluating the Protection Motivation Explanation
The primary strength of the protection motivation explanation is the specification of fear appeal components; particularly the bifurcation of threat and coping components. Although they were discussed earlier (e.g., Hovland et al., 1953), Rogers’s (1975, 1983) conceptualizations and operationalizations represent important grounding for subsequent theoretical frames (e.g., Stroebe, 2000; Witte, 1993).
The revised protection motivation explanation is relatively difficult to evaluate as it relates to fear appeals because it extends beyond persuasive messages to include other information sources (e.g., intrapersonal and social) relevant to coping. Thus, the revised explanation includes, but extends far beyond fear appeals. As a consequence, recent meta-analytic studies (Floyd, Prentice-Dunn, & Rogers, 2000; Milne, Sheeran, & Orbell, 2000) do not speak directly to the issue of fear appeals. One issue is clear, however. The multiplicative combination of cognitive meditational variables (i.e., perceived severity, perceived susceptibility, and perceived response efficacy, and perceived self efficacy) from the original protection motivation explanation has been rejected (Rogers, 1983). Factors influence responses separately rather than in combination (Floyd et al., 2000; Milne et al., 2000; Rogers, 1983).
Witte (1993) claims that an important limitation of the protection motivation explanation is that it foregrounds cognitive (especially subjective expected utility) processes while giving emotions only a tertiary role. For example, in the revised protection motivation model, fear influences protection motivation only as a function of perceptions of threat and efficacy (Rogers, 1983). In short, with the protection motivation explanations, the pendulum had swung entirely from an emotional position (where fear alone mediated the effectiveness of fear appeals) to nearly an entirely cognitive one (where the only cognitive concepts were of consequence).
Extended Parallel Processing Model
Witte (1992) lamented the erosion of fear and, more broadly, affective processes from fear appeal explanations. To reverse this trend, Witte created the extended parallel processing model (or EPPM), which attempted to balance cognitive and affective processes. This balance was attempted by combining elements from the drive, parallel processing, and protection motivation explanations. Specifically,
Leventhal’s model forms the basis of the theory. PMT explains the danger control side of the model (i.e., when and why fear appeals work), and portions of Janis and McGuire’s explanations can be accounted for under the fear control side of the model (i.e., when and why fear appeals fail). (Witte & Allen, 2000, p. 594)
In short, the parallel processing model (Leventhal, 1970) differentiated fear control from danger control, however, it failed to specify when each would operate. The EPPM (Witte, 1992) uses protection motivation components to fill this gap. Specifically, the EPPM predicts that when threat is weak, no attitude or behavior change will occur because receivers are unmotivated to engage in active message processing. If the message threat is either weak or remote (or both), there is no need to attend to, and/or heed, message recommendations (i.e., no threat, no sweat).
When the threat depicted in the message is both severe and likely to strike, receivers will experience fear (Witte, 1992). Fear creates further increases in perceived threat and
motivates receivers to look for ways of avoiding it. Whether audience members engage in fear control or danger control, then, depends on the message’s coping component. Specifically, if threat is high and efficacy is low (the recommended coping response is ineffective and/or the receiver is unable to perform it), fear increases greatly and participants are predicted to engage in fear control. Fear control processes include any of several forms of defensive avoidance, such as message avoidance and not thinking about the issue in the future (Witte, 1992). Under these conditions, attitude and behavior change in the direction of message recommendations is unlikely as participants focus on emotional reactions.
When both threat and efficacy are strong, on the other hand, receivers are predicted to engage in danger control. In doing so, receivers engage in unbiased cognitive processing of message content in choosing responses (Witte, 1992). Under these conditions, attitude, intention, and behavior change in the direction of message recommendations are predicted to occur.
In summary, in the EPPM, “threat-by-efficacy interactions are the fundamental determinants of study outcomes” (Witte, 1992, p. 330). Specifically, the model predicts a multiplicative interaction between threat and efficacy where both threat and efficacy have to be present for attitude, behavior, or intention change to occur. In any other combination of threat and efficacy, little or no change in attitude, behavior, or intention is expected. If threat is low, receivers are unmotivated to process the message. If threat is high and efficacy low, fear control processes predominate.
The SAGE Handbook of Persuasion Page 38