The Impact of Familiarity with a Communicator on the Persuasive Effectiveness of Pandemic-Related Fear Appeals Explained Through Parasocial Relationships

ABSTRACT

Parasocial relationships can improve the effectiveness of persuasive messages such as advertising. However, little is known about the role of parasocial relationships in the processing of fear appeals—a communication strategy that, despite its popularity, often evokes unwanted responses such as reactance. Perceived self-efficacy is one key variable that determines whether a fear appeal improves attitudes/behavior or has unwanted boomerang effects. In a two-level between-subjects experiment (N = 91), we show that a COVID-19-related fear appeal promoting anti-coronavirus measures from a familiar communicator (compared to an unfamiliar communicator) evokes more perceived self-efficacy explained by media users’ parasocial relationship with the familiar communicator. A second two-level between-subjects experiment (N = 239) replicates these findings and shows that perceived self-efficacy inhibits reactance responses (message derogation and perceived threat to freedom) and fosters positive attitudes and behavioral intentions toward anti-coronavirus measures.

The COVID-19 pandemic has now been keeping us in suspense for three years. Overcoming this major global challenge not only requires medical progress but also a willingness on the part of the world’s population to accept and behave in line with anti-coronavirus measures, such as social distancing and wearing protective masks. Fear appeals that emphasize the severe threat of coronavirus infection and address the “fundamental fear of asphyxia” are typical vehicles that governments and health organizations used to enhance this willingness in the population (e.g., Australian Government, 2021). However, such fear appeals are not always successful, as they often have unwanted effects, such as evoking reactance responses (e.g., protests against the obligation to wear a protective mask) instead of the intended adaptive changes. Unintended reactance responses are a well-known problem with the use of fear appeals (e.g., Shen & Coles, 2015). With regard to pandemics, these boomerang effects of well-intended fear appeals are dangerous and can have grave consequences, even leading to the deaths of many people if they cause a derogation of anti-coronavirus measures.

Previous research on fear appeals has already identified several factors that may inhibit reactance responses, such as using moderate- instead of high-fear appeals (e.g., Chen, 2016). However, in some situations (such as a severe pandemic), it is nearly impossible to use this knowledge to communicate substantial personal restrictions like social distancing to prevent infection with a potentially deadly virus without mentioning severe consequences and, thus, evoking fear in individuals. We need to identify other ways to foster adaptive changes and inhibit reactance responses to fear appeals for such cases. Parasocial relationships (PSRs; i.e., one-sided relationships between a media personality and a media user) are one factor that has been neglected in research on fear appeals but that has been shown to enhance intended outcomes while inhibiting reactance in other persuasive contexts such as advertising (e.g., Breves, Liebers, et al., 2021). Both theoretical (e.g., Moyer-Gusé & Nabi, 2010) and empirical (e.g., Breves, Amrehn, et al., 2021) findings on the persuasive impact of PSRs suggest that persuasive messages from a familiar communicator with whom an individual has a PSR increase the individual’s involvement with the message and lead to a positive bias in processing (see also Breves & Liebers, 2023). Assuming that these affect mechanisms occur in similar ways in the processing of fear appeals from familiar communicators, the familiarity of a communicator and, in turn, the PSR with the communicator should be able to enhance perceived self-efficacy—a variable that has been shown to determine whether a fear appeal evokes wanted danger control processes (e.g., adaptive behavior) or unwanted fear control processes (e.g., message derogation; e.g., Ruiter et al., 2014). In this article, we investigate how PSRs with familiar communicators can improve the persuasive effectiveness of fear appeals.

Theoretical background

Fear appeals: Definition, impact, and the current state of research

Fear appeals are defined as messages that emphasize severe or dreadful consequences due to a specific threat that will happen if the reader/listener/viewer does not carry out a recommended behavior (e.g., Rogers, 1975; Shen, 2015; Witte, 1992). As fearing serious consequences is thought to be a critical motivator for adaptive changes in one’s behavior, using fear appeals as a communication strategy is particularly popular in health communication (e.g., Shi & Smith, 2016). Examples include anti-smoking campaigns that caution against lung cancer if one does not stop smoking and, more recently, campaigns that explain the serious health risks that accompany coronavirus infection if one refuses to get vaccinated.

When examining the processing of fear appeals, it is essential to consider three variables: perceived threat, perceived efficacy, and fear itself. The extent to which a threat is perceived depends on two aspects—the perceived severity (the perceived seriousness of the threat) and the perceived susceptibility or vulnerability (the perceived chances of experiencing the threat; e.g., Rogers, 1983). Only when an individual evaluates a threat as serious (perceived severity; e.g., life-endangering consequences of a disease) and as personally relevant (perceived susceptibility; e.g., being at high risk of contracting the disease) is the perceived threat high for that individual (e.g., Witte, 1992). Perceived efficacy, in turn, is based on the perceived response efficacy (perceived effectiveness of the recommended behavior) and the perceived self-efficacy (perceived ability to perform the recommended behavior; e.g., Rogers, 1975). Only when an individual believes that following a recommended behavior is effective against the threat (perceived response efficacy; e.g., wearing a protective mask prevents infection with the disease) and thinks that they are able to perform the behavior (perceived self-efficacy; e.g., buying a protective mask is easy, and the individual feels capable of wearing it) does the individual perceive high efficacy (Witte, 1992). Finally, fear can be understood as a negative emotion that gets triggered by a perceived threat accompanied by strong physical arousal (e.g., Easterling & Leventhal, 1989). When processing a fear appeal, the interplay of these three variables (perceived threat, perceived efficacy, and fear) determines the impact of the fear appeal. Following the Extended Parallel Process Model proposed by Witte (1992), the first step in processing a fear appeal is evaluating the threat. If the perceived threat is low, the individual is not motivated to reflect further on the fear appeal, and there is no further processing of it. However, if the perceived threat is high, the individual evaluates the efficacy of the recommended behavior. Depending on the interplay of perceived threat and efficacy, the individual may experience fear, with the fear being particularly high if the perceived threat is high and the perceived efficacy is low. It is mainly the perceived efficacy and corresponding fear that determine whether an individual engages in danger control processes (intended outcomes such as acceptance of the message or adaptive changes in attitudes or behavior) or fear control processes (unintended outcome such as derogation of the message or maladaptive changes in attitudes or behavior; see also Ruiter et al., 2014).

Despite the long tradition of fear appeals in practice, empirical studies on the persuasive effectiveness of fear appeals report mixed findings (e.g., Ruiter et al., 2014). Although previous research has suggested that perceived threat and perceived efficacy are important (e.g., Witte & Allen, 2000), and that the interplay of these variables operates as hypothesized in the Extended Parallel Process Model (Peters et al., 2013), the general effectiveness of fear appeals is often doubted. The main reason for this skeptical view is the high risk of unintended outcomes in the form of fear control processes such as message derogation or even reactance, resulting in maladaptive changes in attitudes and behaviors (Ruiter et al., 2014). Risking unintended reactions when using fear appeals in health communication is critical, as this may lead to boomerang effects and harmful consequences for individuals and the broader society. Therefore, research on fostering danger control and inhibiting fear control in individuals in the processing of fear appeals is particularly important. Reflecting on the reality that the type of reaction depends mainly on the perceived efficacy (Witte, 1992), investigating ways to increase perceived response efficacy and self-efficacy is a promising research focus. In line with this, Shi and Smith (2016), for example, found evidence that repeated contact with a fear appeal enhances perceived efficacy. Another factor that has been neglected in research on fear appeals so far but that has been linked to perceived (self-) efficacy in the context of health-related entertainment broadcasts is the (parasocial) connection between a media character/communicator and the reader/listener/viewer (e.g., Phua, 2016; Siegenthaler et al., 2021). In the current article, we address this research gap by first introducing PSRs as an amplifier of the effectiveness of fear appeals and then investigating the impact of PSRs in the processing of fear appeals using two empirical studies.

Introducing parasocial relationships as an amplifier of the effectiveness of fear appeals

The term parasocial was first introduced by Horton and Wohl (1956) to describe one-sided mediated forms of social interactions and relationships between a media character (e.g., a television-series character) and a media user (e.g., a television viewer). Despite their lack of mutuality, parasocial interactions (PSIs; situational interactions with a media character during reception, such as watching television or reading a book) and PSRs (overarching relationships with a media character that are not limited to a specific situation) are considered to be similar to real-life social interactions and relationships. This similarity manifests not only through similar predictors such as attractiveness (e.g., Liebers & Schramm, 2017) and impacts such as an enhanced feeling of social relatedness (e.g., Derrick et al., 2009) but also through parallels in the development of parasocial phenomena. In line with real-life interactions and relationships, PSIs and PSRs are interdependent: PSIs shape the cross-situational relationship with a media character in the form of the PSR; PSRs influence the extent to which an individual engages in subsequent media reception situations (Gleich, 1997). In addition to this interdependence, it is essential to note that parasocial phenomena develop in stages similar to real-life relationship stages (Knapp, 1978; Tukachinsky & Stever, 2019). This means that parasocial phenomena get more intense over time, with increasing quantity (and quality) of media contact and a corresponding growing familiarity with a media character being key predictors of the strength of PSIs and PSRs.

We can assume that parasocial phenomena can develop in any kind of media ranging from books to social media—as long as it provides a chance to get to know a media character. In line with the diverse media settings in current work, a wide range of research exists on the antecedents and implications of parasocial phenomena (Liebers & Schramm, 2019), with the persuasive impact of parasocial phenomena being one of the most prominent research foci (particularly since the rise of social media influencers; e.g., Breves et al., 2019). In brief, the vast majority of previous research finds support for parasocial phenomena enhancing persuasive effectiveness (see also the meta-analysis by Tukachinsky et al., 2020). Due to the diversity in research and topics, there are several different explanations for this effect. One explanation is the simultaneous increase of message involvement and biased processing because of parasocial phenomena (Breves & Liebers, 2023; Tukachinsky & Sangalang, 2016). This mechanism is explained in the following paragraphs.

In brief, an individual’s involvement (i.e., an individual’s motivation to attend to and elaborate on a message) determines how many cognitive resources are allocated to the elaboration of a message. More involvement leads to the allocation of more cognitive resources, which can be beneficial for the persuasive effectiveness of a message (Petty & Cacioppo, 1986). Having a PSI with the communicator of a persuasive message can be seen as one way to enhance an individual‘s involvement in a persuasive message (Tukachinsky & Tokunaga, 2013) and has already been linked to the enhancement of elaboration on a persuasive message (e.g., Schartel Dunn, 2018). However, under certain circumstances, such as low argument quality, an enhanced PSI and the resulting increased involvement can also lead to backfire effects, as noted by Tukachinsky and Sangalang (2016). At this point, the overarching relationship with the communicator, and therefore the PSR, becomes relevant. PSRs play a key role in an individual’s involvement with a persuasive message because of two different factors. First, a previously existing PSR with a communicator is one of the main predictors of how intensely an individual engages in a PSI during the processing of a persuasive message (Gleich, 1997; Tukachinsky & Stever, 2019) and, therefore, indirectly determines the individual’s involvement (Breves & Liebers, 2023). Second, the PSR, which can usually be described as a positive relationship similar to a friendship (Liebers & Schramm, 2019), comes with biased processing through the enhancement of favorable interpretations of the communicator and the rejection of negative information about the communicator because of a well-established character scheme (Breves, Liebers, et al., 2021). PSRs thus enhance involvement while inhibiting backfire effects through biased processing (for a more detailed explanation, see Breves & Liebers, 2023).

Although the impact of PSRs on persuasive effectiveness has been shown in other contexts, such as advertising (e.g., Breves et al., 2019) and entertainment-education (e.g., Siegenthaler et al., 2021), research on preexisting PSRs with communicators and their impact on the processing of a fear appeal is limited. Despite this lack of prior investigations, the inclusion of PSRs in research on fear appeals seems promising. Reflecting on the reality that PSRs enhance both involvement and the positively biased processing of messages, PSRs should lead to an increased allocation of cognitive resources toward a fear appeal and its components as well as to processing them less critically. Perceived self-efficacy is high when individuals are confident that they can conduct the behavior that the communicator of a fear appeal recommends in order to protect against the threat (Witte, 1992). For this, the individual has not only to trust the communicator that he/she does know what the individual is capable of doing but also be highly involved as the individual has to actively process the information to link it to his/her own daily life. Reflecting on the reality that PSRs enhance the allocation of cognitive resources to the processing of a message (Breves & Liebers, 2023), individuals should be more motivated to cognitively respond to the fear appeal by linking the recommended protective behavior to their own lives when receiving a fear appeal from a communicator with whom they already established a PSR. Together with their bias in processing and the perceived trust that accompanies a PSR with a communicator, this should result in enhanced perceived self-efficacy. Self-efficacy, in turn, is, as already mentioned, one of the most critical variables determining whether the intended danger control processes or unwanted fear control processes are activated in an individual in response to a fear appeal (Witte, 1992).

In the following sections, we present two empirical studies that address the gap in existing research by investigating the impact of the PSR with a communicator on the processing of fear appeals related to the COVID-19 pandemic. We believe that our assumptions (and later results) are transferable to other kinds of health-related fear appeals as well (see the Discussion section); however, because of the current global crisis, we wanted to explicitly investigate fear appeals related to the coronavirus so we could offer empirically tested advice for communicating in this important context.

Study I

Hypotheses

In the first study, we focus on the key assumption that a preexisting PSR with the communicator of a fear appeal can enhance perceived self-efficacy about the recommended anti-coronavirus measures (e.g., social distancing). We introduce our hypotheses in the following paragraphs.

In line with the parasocial relationship stages model (Tukachinsky & Stever, 2019), we assume that PSRs develop in stages and over time. Whereas PSRs established during the initial few contacts with a media character are rather weak and superficial, PSRs that have developed over time and with a growing knowledge of the communicator are more intense. Reflecting on this reality, prior exposure to and familiarity with a communicator should enhance the PSR intensity, leading to our first hypothesis:

H1:

Individuals have a more intense PSR with a familiar communicator than with an unfamiliar communicator.

A preexisting PSR with a communicator should translate, in turn, into enhanced involvement with the communicator’s persuasive message and positively biased processing of this message (Breves & Liebers, 2023). A PSR should thus enhance an individual’s perceived self-efficacy about the recommended anti-coronavirus measures. Hence, we assume the following:

H2:

A more intense PSR between an individual and the communicator of a COVID-19-related fear appeal corresponds to higher perceived self-efficacy regarding the recommended anti-coronavirus measures.

Method

Design and procedure

We conducted an online study with a two-level between-subjects experimental design in Germany to test our hypotheses. The two experimental groups varied in their familiarity with the communicator of a fear appeal (Group 1: high familiarity, Group 2: low familiarity). After the participants were welcomed and had given informed consent, they were presented with three different scientists with high media presence in the context of the COVID-19 pandemic in Germany.¹ Along with the presentation of the three scientists, the participants were asked via an instruction on the top of the page to choose one of these scientists. The instruction varied depending on which experimental group the participant was randomly assigned to; this was the manipulation in the study. Thus, participants in the high-familiarity group were told to select a scientist they knew through the media, and participants in the low-familiarity group were instructed to select a scientist with whom they were not familiar. This design followed the methodological approach used by Breves, Amrehn, et al. (2021) and Breves, Liebers, et al. (2021), who experimentally manipulated the familiarity of media personalities (social media influencers) by presenting several social media influencers and letting the participants choose one via a randomly assigned instruction. The presentation of existing media personalities and the random assignment of the participants to an instruction about which of the media personalities to choose offer several advantages, such as facilitating high external validity without risking low internal validity because of different groups that are difficult to compare, as may occur with a quasi-experimental approach (see also the abovementioned studies conducted by Breves and colleagues about this methodological approach and its advantages/limitations). After the participants had chosen a scientist in line with the instruction they were given, they answered questions about their PSR with the chosen scientist. We then presented an interview with the selected scientist that contained a fear appeal and coronavirus-related behavioral recommendations (see Materials). Afterwards, participants indicated their perceived self-efficacy concerning the behavioral recommendations. Finally, the participants provided some information on their demographic characteristics and were debriefed carefully, which included explaining the study’s hypotheses, the creation of the interviews, and a contact address in case the participants had any further questions.

Materials

The study took place in July and August 2020 in Germany. During this time, Germany had mostly recovered from the first wave of the COVID-19 pandemic, but the second wave was already in sight. Knowledge of these contextual factors is important to better understand the dynamics of the pandemic at this time and the materials we created and used in our study. Again following the procedure used by Breves, Amrehn, et al. (2021) and Breves, Liebers, et al. (2021), and with the aim of ensuring high internal validity, we used the same interview for each scientist. Because the type of interview necessary did not exist for all scientists, we created our own interview, including a photograph of the chosen scientist. As a cover story, we told the participants that the interview had been published in the Deutsches Ärzteblatt (in English: German Medicine Newspaper), an existing popular website that often features interviews with scientists on medical topics. The presented interview contained a photograph of the scientist and a fear appeal, which included both the threat of the coronavirus (e.g., COVID-19 is a severe illness that might be harmful to everyone and that can, in the worst case, lead to a person’s death by lung failure) and behavioral recommendations on how to reduce the risk of being infected with the virus (e.g., social distancing). Except for the photograph and the name and age of the scientist, all elements (e.g., the content of the interview) were kept constant (see Figure 1).

Figure 1. Stimuli of Study I* .

*Translation of the German interviews: “Prof. Dr. [XY]—In his actual interview with the German Medicine Newspaper, he applauds the Germans for their behavior but finds warning words for the upcoming months, too: ’[…] even though the latest number of infections is on the decline, there is still a high health risk for the German population. We have seen that this concerns not only those who have been categorized as high-risk groups. Even young, athletic, or perfectly healthy individuals without any pre-existing disease sometimes endure severe courses of the disease and have to have artificial respiration. In the worst case, these individuals experience a painful death due to lung failure. The threat of COVID-19 is and will continue to be of high relevance during the next few months.’

Thus, the most critical need of the hour is that everyone continues to do his/her part to keep the spread of the coronavirus as low as possible. This can be implemented through diverse hygienic measures or simply by keeping your distance from others. Those who still have to leave their homes should continue to wear masks, even if this changes from an obligation to a recommendation in the future. The effectiveness of all these measures have already been validated by several studies. They surely help to inhibit the spread of the coronavirus and a second wave,’ explains the [XY]-year-old.”

Participants

The participants² were recruited online via social media. In doing so, the researchers used their own accounts to create posts on Facebook and Instagram and asked their friends/followers to participate and share the study resulting in a snowball system. In total, 97 people completed the questionnaire. However, several participants were excluded because they did not follow the instruction they were given in choosing a scientist (for instance, being asked to select an unfamiliar scientist but indicating at the end of the questionnaire that he*she is familiar with the scientist through the media; n = 6). In doing so, we aimed at having higher data quality by excluding those who did not pay attention while participating in the study. This resulted in a final sample of N = 91 volunteers. Most of the participants (73.63%) were female (25.27% male), and the mean age of the final sample was 27.99 (SD = 9.66) years, with a range of 16 to 64 years (median = 25 years). Most of the participants were students at the time of their participation (67.03%). The appropriate tests indicated no differences between the two experimental groups in the distributions of gender (χ² [2, N = 91] = 0.77, p = .682) or current occupation (χ² [5, N = 91] = 4.68, p = .457). However, the participants differed in terms of age, with those in the high-familiarity condition being older (M = 31.89, SD = 13.11) than those in the low-familiarity condition (M = 25.31, SD = 4.84), on average (t[42.79] = 2.92, p = .006). The difference in age between the two experimental groups might be a sign of systematic dropout between the experimental conditions, and its implications should be kept in mind when interpreting differences in the dependent variables between the experimental groups.

Measurement

All ratings were made on a seven-point Likert scale ranging from 1 (strongly disagree) to 7 (strongly agree).

Parasocial relationship

PSR was captured using the 13 items of the established Multiple Parasocial Relationships Scale created by Tukachinsky (2010), that present a positive, amicable bond between a media characters and a user. Example items include “I think X could be a friend of mine.” Cronbach’s α for the 13-item measure of PSR was .91, indicating excellent reliability (M = 3.66; SD = 1.18).

Perceived self-efficacy

To measure perceived self-efficacy, we drew on three items developed by Witte et al. (1996), adapting them slightly to match the COVID-19 context. One example is “The recommended behavior is easy to use to prevent infection with the coronavirus.” Cronbach’s α for the three-item measure of perceived self-efficacy was .77, indicating satisfactory reliability (M = 6.01; SD = 0.87).

Results

An overview of zero-order correlations between key demographics and outcome variables, as well as an overview of means according to the experimental groups, can be found in the Appendix (Tables A1 and A2).

In line with our first hypothesis, the participants in the experimental group with a familiar communicator reported more intense PSRs (M = 4.28, SD = 1.08) than those in the experimental group with a non-familiar communicator (M = 3.23, SD = 1.06; t[89] = 4.59, p < .001; Cohen’s d = 0.979).

To test the second hypothesis, which predicted that the enhanced PSR with a familiar communicator increases perceived self-efficacy, we performed a mediation analysis using the PROCESS path analysis macro for SPSS developed by Hayes (2018). The analysis was based on bootstrapping samples (n = 10,000) with bias-corrected 95% confidence intervals (see Figure 2). The analysis revealed a positive effect of the familiarity of a communicator on the PSR, again supporting H1, and a positive effect of PSR on perceived self-efficacy, in line with H2. Accordingly, familiarity has a significant indirect effect on self-efficacy that operates through PSR. As the direct effect of familiarity on perceived self-efficacy did not reach significance, this is a complete mediation, operating entirely through PSR.

Figure 2. Mediation analysis of the influence of familiarity with a communicator on perceived self-efficacy explained by parasocial relationship (N = 91)*.

*The mediation analysis is based on Hayes (2018, Model 4) with unstandardized regression coefficients and bootstrapping (m = 10,000).

Discussion and interim conclusion

Our first study empirically supports the assumption that individuals develop more intense PSRs with familiar communicators and that these intense PSRs, in turn, lead to an increase in perceived self-efficacy. These findings provide first important insights, leading to the conclusion that the prior media presence of a communicator can be an amplifier of the effectiveness of a fear appeal by modifying perceived self-efficacy through the PSR between the media user and the communicator.

However, Study I also has serious limitations. One limitation is that the sample of participants was very homogeneous. One reason for this might be the recruiting via the researchers’ social media accounts as, for example, there were more female friends and followers among those who participated and shared the study. This bias in the sample composition is potentially problematic because previous studies have found differences in attitudes toward anti-coronavirus measures by demographic characteristics (e.g., Mallinas et al., 2021). Moreover, it is unclear to what extent the findings of enhanced self-efficacy might translate to other critical variables in the context of the COVID-19 pandemic, such as reactance or actual behavioral intentions regarding anti-coronavirus measures. Finally, the COVID-19 pandemic is highly dynamic, and it is essential to know whether these findings can be replicated at another time during the pandemic. To address these limitations and gain further knowledge on the influence of PSRs with familiar communicators in the context of fear appeals during the COVID-19 pandemic, we conducted a second study.

Study II

Hypotheses

In contrast to Study I, the second study has a broader focus, investigating not only the influence of PSRs on perceived self-efficacy but also how this translates to other critical variables in the context of anti-coronavirus measures. Here, we begin with the same hypotheses as in Study I, assuming that individuals have more intense PSRs with familiar communicators than with unfamiliar communicators (H1) and that enhanced PSRs, in turn, lead to an increase in perceived self-efficacy (H2). In Study II, based on theoretical hypotheses drawn from the Extended Parallel Process Model (Witte, 1992) and previous empirical research (Ruiter et al., 2014), we assume that enhanced perceived self-efficacy helps individuals to cope with the potential threat and corresponding fear that is evoked while processing the fear appeal. Engaging in functional coping that can lower the perceived negative emotions (Witte, 1992), in turn, should foster adaptive responses toward the fear appeal and higher agreeing to the recommended behavior resulting in better attitudes toward the recommended behavior. Accordingly, concerning the COVID-19 pandemic, we assume that perceived self-efficacy regarding the recommended anti-coronavirus measures improves attitudes toward these measures:

H3:

Higher levels of perceived self-efficacy toward the recommended anti-coronavirus measures correspond to more favorable attitudes toward these measures.

Besides the direct positive effect of enhanced self-efficacy on attitudes toward the recommended behavior, perceived self-efficacy can also have an indirect impact as it lowers the chance for unwanted fear control processes. Individuals that experience a significant threat but do not feel that they can implement the recommended behavior to protect them against the threat (in other words, have a low perceived self-efficacy) are prone to engage in thoughts and behavior to control the fear, not the threat itself (Witte, 1992). One way of doing this is by derogating a message (e.g., perceiving the message as distorted or unbalanced; Shen & Coles, 2015). Message derogation is a relevant factor here, as it as a form of reactance should, in turn, also lead to a worsening of attitudes toward the recommended behavior and ultimately even lead to maladaptive behavioral changes (Witte, 1992). Briefly summarized, we assume:

H4A:

Higher levels of perceived self-efficacy toward the recommended anti-coronavirus measures correspond to less message derogation.

H4B:

Less message derogation corresponds to more favorable attitudes toward the recommended anti-coronavirus measures.

However, message derogation is not the only way an individual tries to control the unwanted fear evoked by a fear appeal when not experiencing high self-efficacy. Another unwanted facet of fear control processes and type of reactance is an increase in perceived threat to freedom (i.e., the feeling that a message is trying to manipulate the recipient or is threatening their freedom to choose; Shen & Coles, 2015). This became particularly important in the COVID-19 pandemic, as the governments of several nations established formal COVID-19 regulations (e.g., wearing protective masks in public spaces). In the context of the perceived threat to freedom, individuals concentrate rather on the fact that the communicator of the fear appeal wants to manipulate them and limit their freedom rather than controlling the threat by adopting a recommended behavior. This leads to an urge within the individual not to agree with the communicator resulting in worsened attitudes toward a recommended behavior. Accordingly, we assume that an increase in perceived self-efficacy will be beneficial for the persuasive effectiveness of a coronavirus-related fear appeal, as it inhibits perceptions of a threat to one’s freedom and the resulting consequences:

H5A:

Higher levels of perceived self-efficacy regarding the recommended anti-coronavirus measures correspond to lower perceptions of a threat to freedom.

H5B:

Lower perceptions of a threat to freedom correspond to more favorable attitudes toward the recommended anti-coronavirus measures.

Finally, it is important to investigate not only attitudes but also behavioral intentions toward recommended anti-coronavirus measures. On the basis of the theory of planned behavior (Ajzen, 1991), we assume that more favorable attitudes toward the recommended anti-coronavirus measures translate into more positive behavioral intentions toward these measures:

H6:

More favorable attitudes toward the recommended anti-coronavirus measure

Method

Design and procedure

The design and procedure of Study II are similar to those used in Study I and will, therefore, only be described in brief here. Again, we conducted an online study in Germany with a two-level between-subjects experimental design with the two experimental groups varying in their familiarity with the communicator of a fear appeal (Group 1: high familiarity, Group 2: low familiarity). Like Study I, participants were welcomed, and they provided informed consent. They were then presented with four different scientists who had high media presence in the most recent months of the COVID-19 pandemic in Germany.³ As a methodological improvement compared with Study I and with the aim of preventing systematic dropout between the experimental groups, participants were then asked to indicate whether they knew at least one of the presented scientists, and only those who indicated that they were familiar with at least one of the scientists were allowed to proceed with the study. The others were carefully debriefed, which included naming the hypotheses of the study, explaining why they could not proceed with the study, and giving contact details in case someone had further questions. Those who proceeded with the study were then presented with the four scientists again. Similar to the approach used in Study I, these participants were instructed at random to choose either a familiar or a non-familiar scientist.

After the participants had chosen a scientist according to this instruction, they answered questions about their PSR with the chosen scientist. Then, as in Study I, we presented an interview with the chosen scientist (see Materials). The participants were then asked to indicate their level of perceived threat to freedom and message derogation. Afterward, we captured perceived self-efficacy, attitudes, and behavioral intentions related to the behavioral recommendations given by the scientist in the fear appeal. Finally, the participants provided some information on their demographic characteristics and were debriefed carefully similar to Study 1.

Materials

Study II took place in Germany during April–June 2021. During this time, Germany was just recovering from the third wave of the COVID-19 pandemic. We created similar interviews to those used in Study 1; the only differences were that we adapted the content slightly to suit the current phase of the pandemic and used a partly different list of scientists as communicators. For more information on the materials, please see the corresponding section in Study I.

Participants

Again, the participants were recruited online via social media. However, we predominantly used Facebook this time, and a new account solely for recruiting participants was created. Hence, the posts were not shared with personal friends and followers, but were shared the study in diverse Facebook groups to obtain a more diverse sample, including wider ranges in age and educational background. Compared to Study 1, no participants had to be excluded due to not following the introduction regarding the selection of a communicator. In total, N = 239 people completed the questionnaire. Most of the participants (70.29%) were female (27.20% male; 2.51% did not indicate a binary gender identity), and the mean age of the final sample was 46.33 (SD = 13.08) years, with a range of 20 to 76 years (median = 47 years). Most of the participants were employees at the time of their participation (46.03%), followed by independent entrepreneurs (15.48%), retired persons (10.46%), and students (9.21%). The appropriate tests indicated no differences between the two experimental groups in the distributions of gender (χ² [3, N = 239] = 3.07, p = .382), age (t[237] = 0.68, p = .495), or current occupation (χ² [8, N = 239] = 3.94, p = .862).

Measurement

As in Study I, all ratings were made on a seven-point Likert scale ranging from 1 (strongly disagree) to 7 (strongly agree).

Parasocial relationship

PSR was again captured with the 13 items of ‘s (2010) Multiple Parasocial Relationships Scale. Cronbach’s α for the 13-item measure of PSR in Study II was .96, indicating excellent reliability (M = 3.04; SD = 1.54).

Perceived self-efficacy

To measure perceived self-efficacy, we used the same adaption of three items by Witte et al. (1996) as in Study I. Cronbach’s α for the three-item measure of perceived self-efficacy was .88, indicating good reliability (M = 5.48; SD = 1.71).

Message derogation

Message derogation was measured with six items presented by Shen and Coles (2015). The items were: “The article is biased.” “The article is convincing.” “The article is sensible.” “The article is distorted.” “The article is fair.” “The article is balanced.” Cronbach’s α for the six-item measure of message derogation was .95, indicating excellent reliability (M = 3.21; SD = 1.97).

Perceived threat to freedom

Along with message derogation, perceived threat to freedom was measured with four items presented by Shen and Coles (2015). The items were: “The article tried to make a decision for me,” “The article tried to manipulate me,” “The article threatened my freedom to choose,” and “The article tried to pressure me.” Cronbach’s α for the four-item measure of the perceived threat to freedom was .95, indicating excellent reliability (M = 3.02; SD = 2.15).

Attitudes toward anti-coronavirus measures

To measure attitudes toward the recommended anti-coronavirus measures, we formulated six items—two for each recommended measure (i.e., being hygiene conscious, social distancing, and wearing a protective mask). Examples include “I think it is important to wear an adequate protective mask these days.” Cronbach’s α for the six-item measure was .94, indicating excellent reliability (M = 5.29; SD = 1.77).

Behavioral intentions toward anti-coronavirus measures

As we did to assess attitudes toward anti-coronavirus measures, we formulated six items to capture behavioral intentions toward anti-coronavirus measures, with two items for each recommended behavior. Examples include “Whenever I leave my home over the next few months, I will have an adequate protective mask with me.” Cronbach’s α for the six-item measure of behavioral intentions toward anti-coronavirus measures was .92, indicating excellent reliability (M = 4.82; SD = 1.73).

Results

An overview of zero-order correlations between key demographics and outcome variables, as well as an overview of means according to the experimental groups, can be found in the Appendix (Tables A3 and A4).

In line with the findings of Study I and our first hypothesis, the participants in the experimental group with a familiar communicator reported more intense PSRs (M = 3.37, SD = 1.66) compared with those in the experimental group with a non-familiar communicator (M = 2.66, SD = 1.28; Welch’s t_(232.86) = 3.70, p < .001; d_Cohen = 0.472).

In the next step, our hypotheses were tested using a path model (see Figure 3). As we assumed individuals’ reactance in the forms of message derogation and perceived threat to freedom to be highly correlated (Shen & Coles, 2015), we allowed the error terms of these variables to be correlated in the model. The path model was computed with the AMOS 27.0 statistical package using the maximum likelihood method. Model fit was assessed using established fit indices and indicated an adequate overall fit (χ2₍₁₂₎ = 45.30, p < .001, CMIN/df = 3.78, CFI = .979, RMSEA = .108, GFI = .951). Confirming H2, PSR was positively related to perceived self-efficacy (β = .23, p < .001). In line with H3, perceived self-efficacy improved attitudes toward anti-coronavirus measures (β = .51, p < .001) and, as predicted by H4, decreased message derogation (β = −.83, p < .001), which, in turn, was found to worsen attitudes toward anti-coronavirus measures (β = −.34, p < .001). H5 was only partly supported, as self-efficacy decreased the perceived threat to freedom (β = −.79, p < .001), but, surprisingly, perceived threat to freedom was not a significant predictor of attitudes toward anti-coronavirus measures in the current study (β = −.08, p = .192). Finally, attitudes toward anti-coronavirus measures were related to behavioral intentions toward these measures, as predicted in our last hypothesis (β = .92, p < .001).

Figure 3. Path model of the influence of familiarity with a communicator on the persuasive effectiveness of a fear appeal concerning the COVID-19 pandemic (N = 239)*.

*Reported values represent standardized path coefficients at a significance level of p < .001, if not declared otherwise.

General discussion

The studies presented here investigated the influence of preexisting PSRs on the persuasive impact of COVID-19-related fear appeals. In this investigation, we assumed that being familiar with a communicator and, in turn, having a PSR with him/her increases an individual’s involvement with the message of the communicator and also leads to positive bias in the processing of this message (see also Breves & Liebers, 2023). As a result, we hypothesized that having a PSR with the communicator of a fear appeal enhances perceived self-efficacy—a variable that has been found to determine the persuasive success of a fear appeal (Ruiter et al., 2014). In two empirical studies, we were able to show the positive effect of familiarity with a communicator on perceived self-efficacy, operating through PSR. In these studies, we found additional empirical support for the persuasive impact of familiarity and, in turn, preexisting PSRs with communicators, which had previously been investigated mainly in the context of product advertising in social media (e.g., Breves, Liebers, et al., 2021). Moreover, these findings extend the current state of knowledge on the role of parasocial phenomena in health communication. Beyond fear appeals, the impact of parasocial phenomena has previously been studied sporadically in the context of health communication and perceived self-efficacy (e.g., entertainment education; e.g., Phua, 2016). However, none of this previous work implemented an experimental manipulation of the PSR with the communicator (e.g., Tian & Yoo, 2015). Reflecting on this reality, we find that our studies add to the current state of research on parasocial phenomena by providing internally valid evidence for the importance of PSRs in evoking self-efficacy.

Moreover, by introducing PSRs as an amplifier of the persuasive effectiveness of fear appeals, drawing on the Extended Parallel Process Model (Witte, 1992), the current studies were able to gain insight into the processing and impact of fear appeals in response to communicators. As already mentioned, because of higher levels of PSR, individuals responded to a COVID-19-related fear appeal with more perceived self-efficacy concerning the recommended anti-coronavirus measures when the fear appeal came from a familiar communicator than when it came from an unfamiliar communicator. Perceived self-efficacy, in turn, decreased unintended fear control processes by inhibiting message derogation, reducing the perceived threat to freedom, and improving attitudes toward the recommended anti-coronavirus measures. As predicted, attitudes toward the recommended anti-coronavirus measures were strongly correlated with behavioral intentions toward these measures. Both perceived self-efficacy’s positive influence on attitude and its negative influence on reactance are consistent with the theoretical assumptions of the Extended parallel Process Model and with previous empirical findings (Shen & Coles, 2015; Witte, 1992). However, surprisingly, there was no significant effect of the perceived threat to freedom on attitudes toward the recommended anti-coronavirus measures. This finding may be explained by the operationalization of attitudes toward the anti-coronavirus measures and the setting of the current studies. To keep the two studies comparable, we focused on recommendations on hygiene, social distancing, and wearing a mask—not vaccination (there was no vaccine in sight when we conducted Study 1). However, since the introduction of the anti-coronavirus vaccines, vaccination and the possibility of a vaccine mandate are much-discussed topics in Germany, with one of the most widespread arguments being that mandatory vaccination threatens one’s personal freedom to choose. Being such a focus within public debates around the threat of freedom, other measures, such as wearing a protective mask, were (comparably) no longer related to harming individuals’ freedom. However, as we did not include vaccination in our fear appeal, we also did not include it in our measurement of attitudes toward the recommended measures. The missing of the in some individuals’ opinion freedom threatening vaccination from the attitude measures might be the reason for the missing significant link between perceived threat to freedom and attitudes in the current study.

To summarize, the current studies highlight the importance of PSRs with communicators in processing fear appeals. The studies not only advance the current state of research on parasocial phenomena and fear appeals, but also provide practical implications for communicators in health communication in general and especially during the COVID-19 pandemic. The results of the current studies lead us to conclude that communicators with whom individuals engage in PSRs are particularly effective in communicating fear appeals during the current pandemic. This is knowledge that can be used in multiple ways in the planning of further communication measures. First, we can recommend the use of communicators known for their strong parasocial bonding. This means that campaigns on anti-coronavirus measures would be best built around famous communicators such as actors and musicians rather than unknown testimonials. In this context, one group of famous individuals that seems particularly promising as effective communicators are social media influencers. Social media influencers are known for the strong parasocial engagement of their followers (e.g., Hudders et al., 2021), which they should be able to use to promote not only products but also fear appeals by enhancing perceived self-efficacy. Concerning set communicators such as the head of states or the Minister of Health, we recommend that they present themselves in a way that facilitates parasocial engagement with them. Other possible ways to establish strong PSRs, in addition to high media coverage (familiarity as a predictor; Tukachinsky & Stever, 2019) include the use of physical and verbal addressing of the audience (addressing as a predictor; Hartmann & Goldhoorn, 2011) or stressing similarity with the audience (perceived similarity as a predictor; Liebers & Schramm, 2017), to name a few examples.

Limitations, future research, and open questions

In the current studies, we investigated the role of PSRs in the processing of fear appeals on the COVID-19 pandemic. This meant investigating a very prominent and much-discussed topic with which nearly everyone has their own experiences through travel restrictions, psychological issues due to lockdowns, illness, job loss, or death of loved ones. One of the methodological issues in the current studies is that we did not capture these prior experiences or the existing attitudes toward COVID-19 associated with these experiences before study participation. As we used an experimental setting rather than a quasi-experimental approach and because the participants should not have varied between the two experimental groups in terms of prior experiences or attitudes because of randomization, we assume that prior experiences and attitudes did not systematically distort our findings. However, this favorable assumption should only be made with caution, reflecting on the reality that age varied systematically between the experimental groups in Study 1. Moreover, the samples were both convenient samples with a large number of females and well-educated participants (again, particularly Study 1). Along with that, we cannot rule out a systematic distortion of our findings despite our experimental approach. As a result, future studies, particularly in the context of a pandemic, should include measures of prior experiences with the health issue, as former studies have found that personal concern but also prior attitudes may change the processing of fear appeals (Van ‘t Riet & Ruiter, 2013) and it might come to systematic dropouts despite randomized experiments.

Moreover, we concluded that, based on our findings, communicators of fear appeals during the COVID-19 pandemic should present themselves in a way that facilitates parasocial engagement with them by, for example, directly addressing their audience—i.e., directly looking into the camera and speaking to the audience. This conclusion makes sense, as our findings suggest that PSRs enhance the persuasive impact of fear appeals, and fostering PSIs, the most important antecedent of PSRs (e.g., Liebers & Straub, 2020), should be a promising means of achieving similar positive effects. However, this is only speculation and is not based strictly on empirical findings and should, therefore, be interpreted with caution—especially because former findings report mixed results with regard to the persuasive effects of PSIs (without preexisting positive PSRs; Tukachinsky & Sangalang, 2016). Hence, future research is needed to investigate the persuasive impact of a broader spectrum of parasocial phenomena to fully understand the role of individuals’ engagement with communicators in the processing of fear appeals.

Another aspect that requires future research is the impact of PSRs on other critical variables in the processing of fear appeals. In the current studies, we focused on perceived self-efficacy because this variable has been shown to be particularly critical for the persuasive effectiveness of fear appeals (e.g., Ruiter et al., 2014). However, assuming that the persuasive impact of PSRs operates through an enhanced, positively biased involvement with the communicator’s message, which comes along with an increase in the cognitive resources allocated toward the fear appeal, PSRs should enhance not only perceived self-efficacy, but also perceived response efficacy, as well as perceived severity and vulnerability. Reflecting on this reality, future research on the persuasive impact of PSRs on other aspects in the processing of fear appeals is needed.

Moreover, to create internally valid stimuli, we decided to include communicators that were as similar as possible regarding background, age, gender, attractiveness, and ethnicity. Although this made our stimuli more comparable, it however decreased their external validity. Reflecting on the reality that not all communicators of fear appeals are middle-aged white men, it is critical for future research to investigate the persuasive impact of communicators with more diverse demographics.

To conclude, again, the current studies focused on the specific context of the COVID-19 pandemic. This focus was selected so that the current studies could provide valuable insight into the processing of COVID-19-related fear appeals and, thus, help to develop effective communication of anti-coronavirus measures during this period. However, the hypotheses tested and effect mechanisms assumed in this article are not necessarily limited to the context of the current pandemic. Although future research is needed to validate the role of the parasocial bonding with the communicator in other contexts, PSRs should enhance perceived self-efficacy, which, in turn, facilitates danger control processes while inhibiting fear control processes, regardless of the investigated issue. Reflecting on this reality, future research on a wide range of other topics, including diverse health-related issues (e.g., HIV prevention) could adopt this line of inquiry and investigate how communicators can foster the persuasive effectiveness of fear appeals while inhibiting unwanted fear control processes. We already knew this, but the last two years have reinforced how difficult it is to persuade individuals to act in their own and society’s best interests. Neglecting the communicator or, to be more precise, the PSR with the communicator, as a potential factor that can facilitate an individual’s adaptive changes in fearful scenarios such as pandemics is a missed chance that we should make better use of in the future.

Research transparency statement

The authors are willing to share their data, analytic methods, and study materials with other researchers. The material will be made available upon request.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Notes on contributors

Nicole Liebers

Nicole Liebers (PhD, University of Würzburg) is assistant professor of Persuasive Communication at the Amsterdam School of Communication Research. She conducts research in the fields of entertainment and persuasive communication, as well as audience and effects studies with a focus on parasocial interactions and relationships.

Achim Vogel

Achim Vogel (MSc, University of Würzburg) studied Media Communication at the University of Würzburg, Germany. He conducted research on the impact of parasocial relationships in the context of fear appeals in his master thesis.

Priska Breves

Priska Breves (PhD, University of Würzburg) is assistant professor of Persuasive Communication at the Amsterdam School of Communication Research. She conducts research in the fields of persuasive communication, with a focus on advertising, environmental and health communication.

Holger Schramm

Holger Schramm (PhD, Hanover University of Music, Drama, andMedia) is professor of Communication at the University of Wurzburg, Germany.His research fields include music in the media, sports communication, entertainment, and advertising/brand communication with a focus on audience/reception and effects studies(parasocial interactions and relationships, emotions, and non-mediationphenomena like presence and flow).

Notes

¹ The scientists were Christian Drosten, Lothar Wieler, and Hendrik Streeck.

² Please note that it is not mandatory (and rather unusual) to consult an ethical committee for conducting studies at German Universities unless potential physical or psychological harm of participants is expected. Hence, there is no ethics approval for this study – however, we conducted our research in line with the Declaration of Helsinki.

³ The scientists were Christian Drosten, Lothar Wieler, Alexander Kekulé, and Gérard Krause.

Appendix

Table A1. Zero-order correlation matrix of key demographics and outcome variables of Study 1.

Variable	1	2	3	4
1. Age		−.13	.05	.16
2. Gender***			−.18	−.18
3. PSR				.36**
4. Perceived self-efficacy

*p < .05; **p < .01; *** 1=male, 2=female.

Table A2. Table of means of Study 1 according to the experimental groups.

Variable	Unfamiliar communicator M (SD)	Familiar communicator M (SD)
PSR**	3.23 (1.08)	4.28 (1.06)
Perceived self-efficacy*	5.86 (0.89)	6.23 (0.80)

Difference between the experiments significant at level: *p < .05; **p < .01.

Table A3. Zero-order correlation matrix of key demographics and outcome variables of Study 2.

Variable	1	2	3	4	5	6	7	8
1. Age		.13*	−.10	−.01	−.03	−.01	.01	.04
2. Gender***			−.19**	−.02	.01	.03	−.02	−.01
3. PSR				.41**	−.50**	−.40**	.40**	.39**
4. Perceived self- efficacy					−.84**	−.79**	.87**	.80**
5. Message derogation						.89**	−.85**	−.81**
6. Perceived threat to freedom							−.80**	−.76**
7. Attitudes								.92**
8. Behavioral intentions

*p < .05; **p < .01; *** 1 = male, 2 = female.

Table A4. Table of means of Study 2 according to the experimental groups.

Variable	Unfamiliar communicator M (SD)	Familiar communicator M (SD)
PSR**	2.66 (1.28)	3.37 (1.66)
Perceived self-efficacy	5.63 (1.67)	5.34 (1.75)
Message derogation	3.08 (1.97)	3.33 (1.98)
Perceived threat to freedom	2.87 (2.16)	3.16 (2.15)
Attitudes	5.44 (1.72)	5.15 (1.80)
Behavioral intentions	5.02 (1.74)	4.65 (1.71)

Difference between the experiments significant at level: *p < .05; **p < .01.