SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. Method
  4. Results
  5. Discussion
  6. References

This study tests whether the adaptation of a narrative's protagonist to be similar to readers increases narrative effects in the health domain. A between-subjects (N = 220) experiment was conducted that varied the similarity of the protagonist to the participants. Results showed that participants who read the version with a similar protagonist perceived themselves to be more at risk of the disease with which the protagonist was diagnosed and felt more efficacy to deal with the symptoms of this disease, than participants who read the version with a dissimilar protagonist. These effects were mediated by self-referencing, indicating that adaptation of a protagonist to be similar to readers makes readers relate the story to themselves, which in turn increases narrative impact.

Over the past decade, the power of narratives to influence health beliefs has been demonstrated by multiple studies (e.g., Green, 2006; Morgan, Movius, & Cody, 2009; Murphy, Frank, Moran, & Patnoe-Woodley, 2011). In health communication, narratives can be used to influence beliefs like perceived risk for a certain illness, or feelings of efficacy to deal with a medical condition (Kreuter et al., 2007; Kreuter et al., 2010). Because these beliefs underlie health behavior (Rimal & Real, 2003; Witte, 1994), narratives may be effective tools in promoting a healthy lifestyle and reducing unhealthy choices.

Many studies on narrative impact in the health domain have focused on potential underlying processes of narrative effects, such as transportation into a narrative world and identification with characters (e.g., Moyer-Gusé & Nabi, 2010; Murphy et al., 2011). In doing so, these studies have provided valuable insight into the mechanisms by which narratives have effects. However, another important issue in this domain has not yet received as much attention. There is a lack of research into the role of the content of a narrative in narrative processes and effects (Green, 2008). Narratives that include health messages can vary greatly in terms of content, such as the type of events they show, the setting in which they take place, and the characters they are about. For instance, features of characters can differ, like gender and background. Insight into the role of these content elements in narrative effects is important to the goal of predicting which narratives will most likely influence readers' beliefs. This is why Green (2008, p. 48) sees uncovering “the active ingredients of effective narratives” as one of the most important challenges in the research area of narrative impact.

This study takes up on this challenge to identify narrative elements that influence narrative processes and effects in the domain of health communication. An important determinant of message effectiveness that has emerged in research on health communication is the extent to which a message is adapted to characteristics of the recipients (Rimer & Kreuter, 2006). In this research area, a distinction is made between “targeting,” which refers to adapting a message to characteristics of a subgroup of the population and “tailoring,” which refers to adapting a message to characteristics of specific individuals (Kreuter & Wray, 2003). Both types of customization are expected to be effective because the information in a customized message is more relevant for recipients, which facilitates learning and behavioral effects (Kreuter & Wray, 2003; Rimer & Kreuter, 2006). This study connects research on message adaptation—starting with the more general targeting—to research on narrative impact by examining whether a story with a protagonist who has characteristics similar to a group of readers leads to more story-consistent beliefs than a story with a protagonist with dissimilar characteristics. In this way, whether the similarity of the protagonist is a narrative element that makes narratives more effective can be established.

Character-reader similarity

Similarity refers to the extent to which a reader is like a character (Cohen, 2001). Characters can be similar to readers on objective characteristics, such as age or nationality, but also on subjective characteristics, such as beliefs or opinions (Hoffner & Buchanan, 2005). This distinction between similarity on objective and subjective characteristics is important in a study of narrative impact because the effects studied in this research area involve cognitive constructs like beliefs and attitudes (Green, 2006). Thus, these effects entail that certain subjective characteristics of a reader are influenced. In the case of narrative impact, readers can become more similar to a narrative character on subjective characteristics. However, objective characteristics of the reader, such as age or gender, cannot be affected by a narrative. This means that similarity between a reader and a character on objective characteristics can only be an antecedent of effects, whereas similarity on subjective characteristics can be an antecedent as well as a consequence. That is why this study focuses on the effects of similarity on objective characteristics. When readers are similar to characters on objective characteristics, they may become more similar on subjective characteristics like beliefs or attitudes (Hoffner & Buchanan, 2005; Hoffner & Cantor, 1991).

The idea that objective similarity may lead to beliefs more consistent with the story is in line with Social Cognitive Theory (Bandura, 1986, 2002), which posits that similar role models facilitate observational learning. Characters in narratives may serve as such role models. Prior research has shown associations between readers' perceived similarity to characters and story-consistent health beliefs (Andsager, Bemker, Choi, & Torwel, 2006; Pinkleton, Austin, & Van de Vord, 2010). In addition, Green (2004) showed that individuals who had personal experiences similar to the protagonist in a story were more transported into the story, which was in turn related to story-consistent beliefs. However, in these studies, similarity was not manipulated and thus causal conclusions could not be drawn. Beliefs that are consistent with the story prior to reading may also lead readers to perceive themselves as more similar to the protagonist of the story. Therefore, whether objective similarity of a character to a reader leads to more similar beliefs needs to be tested. The first hypothesis of this study is:

H1: Readers of a version of a story in which the protagonist is similar to them on an objective characteristic will have beliefs that are more consistent with the story than readers of a version of the same story in which the protagonist is dissimilar to them.

In addition to identifying narrative elements that increase the effectiveness of stories, establishing the mechanisms through which these elements lead to effects is also important. Previous research provides two potential underlying processes of the influence of similarity on beliefs. On the one hand, research on adapting health messages to recipients suggests that self-referencing plays a role (Kreuter, Bull, Clark, & Oswald, 1999; Kreuter & Wray, 2003). On the other hand, research on narrative experiences and effects indicates that identification with the character may be a process through which similarity leads to story-consistent beliefs (Cohen, 2001; De Graaf, Hoeken, Sanders, & Beentjes, 2012). Both of these potential mechanisms will be investigated in this study.

Self-referencing

A reason why adapting characteristics of a protagonist to the recipient may make a story with a health message more effective is that a message which is customized to a reader facilitates thinking about the message and especially thoughts that relate the message to the self (Kreuter et al., 1999). When a protagonist in a story is similar to a reader, that reader can more easily connect the experiences of the protagonist to him- or herself. Such relating of a message to the self is called self-referencing (Burnkrant & Unnava, 1989, 1995). Self-referencing includes being reminded of one's own experiences and comparing new information to existing knowledge (Burnkrant & Unnava, 1989, 1995). When reading a narrative, the description of an event may make readers remember a similar event they went through. For instance, when someone reads a story about a cancer patient, they may think about a time they, or someone who was close to them, were seriously ill. In this way, self-referencing connects the story to aspects of the self of the reader (Burnkrant & Unnava, 1989, 1995).

Self-referencing should be distinguished from personal relevance, a concept which has been studied extensively in research on the Elaboration Likelihood Model (Petty & Cacioppo, 1986). Personal relevance refers to the importance that a recipient attaches to a message topic (Petty & Cacioppo, 1986). Thus, personal relevance is a rather stable perception of a topic by a recipient, whereas self-referencing is a variable process that takes place during reading a message. The perception of personal relevance of a message topic may lead to the process of self-referencing, but the concepts are clearly distinct.

Prior research has shown that stories can trigger self-referencing in the form of memories of personal experiences and other thoughts about a reader's own life (Dunlop, Wakefield, & Kashima, 2010; Larsen & Seilman, 1988). Self-referencing can in turn increase effects of a message. The more information is related to the self, the more it will affect beliefs and attitudes. Because the self is a complex and highly organized memory structure, information that is linked to the self will be highly available and used in subsequent judgments (Burnkrant & Unnava, 1989).

Research on advertising effects has shown that self-referencing in response to ads indeed leads to attitudinal effects. Burnkrant and Unnava (1995) found that ads that reminded readers of their own prior experiences with the product by addressing them directly made participants engage in more self-referencing. Furthermore, these ads that increased self-referencing also produced more positive attitudes toward the product.

In addition, research on health narratives indicates that self-referencing in response to narratives can play a role in narrative effects. Dunlop et al. (2010) compared narrative health messages to nonnarrative messages. Although there was no difference between the two message formats on self-referencing, self-referencing significantly predicted risk perceptions of getting the disease and intentions to enact preventive behavior in both formats.

In summary, similarity of a character to a reader may increase self-referencing, which in turn can lead to effects. Therefore, the second hypothesis is:

H2: The effect of character-reader similarity on story-consistent beliefs is mediated by self-referencing.

Identification

Another potential mechanism through which similarity of a protagonist to a reader may lead to effects is identification with the protagonist. A common notion is that the more similar a character is, the more readers will identify with that character (Cohen, 2001; Green, 2006; Hoffner & Cantor, 1991). Several studies have shown that readers who are similar to characters, indeed identify more with those characters. For instance, José and Brewer (1984) showed that boys identified more with male characters, and girls identified more with female characters. Similarly, Appiah (2001) showed that African American adolescents identified more with African American characters in ads.

Identification refers to the experience of a reader adopting the point of view of the character on the story and experiencing the story from that position (Cohen, 2001). Consequently, an identifying reader will have emotions that are in consonance with the events happening to the character (Oatley, 1994). As these emotions are construed by the reader as “feeling for” or “feeling with” the character, they can be referred to as empathy (Zillmann, 2006). In summary, identification with a character in a narrative refers to experiencing the story from the position of the character and feeling empathetic emotions that are consonant with the character.

Identification is related to narrative engagement (Busselle & Bilandzic, 2009; De Graaf et al., 2012). Narrative engagement refers to the intense experience people can have during reading a story (Busselle & Bilandzic, 2008, 2009). Different terms have been used for this experience, such as transportation (Green & Brock, 2000), absorption (Slater & Rouner, 2002) and involvement (Moyer-Gusé, 2008), but on a general level, these terms all refer to an experience of readers being completely engrossed in the story, losing track of time and forgetting the world around them.

Identification also entails such intense processing, but with a specific focus on characters. When readers identify, they are engrossed in imagining what it is like for a character to experience the events described in the story (Cohen, 2001). Although identification is clearly related to more general narrative experiences such as transportation, the exact nature of these relations is still unclear. These processes may operate in parallel, but they may also operate sequentially (see Busselle & Bilandzic, 2009; Green, 2006). Nonetheless, transportation and identification are distinct concepts, which can be studied separately (Tal-Or & Cohen, 2010).

Identification with characters in a narrative can lead to narrative effects (De Graaf et al., 2012; Green, 2006). As identifying readers simulate or imagine the events that happen to a character in their imagination, they may gain greater understanding of what it is like to experience the described events and beliefs may be influenced accordingly (Mar & Oatley, 2008, p. 182). Empirical research has shown that identification is indeed a mechanism of narrative impact. De Graaf et al. (2012) used stories in which two opposing characters espoused different opinions (e.g., an applicant versus an employer). They manipulated the perspective from which the story was told, such that one version of the story was told from the perspective of one character and another version was told from the perspective of the opposing character.

Results showed that participants identified more with the character whose perspective was presented in the story and that their attitudes were more consistent with the version they read. Moreover, identification mediated the effect of perspective version on attitudes, which means that identification was responsible for the influence on attitudes. When readers identify with a character, their attitudes shift in the direction of the ones implied by the character (De Graaf et al., 2012).

In summary, a character who is similar to a reader may lead to effects through identification with the character. Similarity can increase identification, which can enhance narrative effects. Therefore, the third hypothesis is:

H3: The effect of character-reader similarity on story-consistent beliefs is mediated by identification.

The two proposed mechanisms of self-referencing and identification are processes with a distinctly different focus. Self-referencing refers to a reader thinking about him- or herself in order to relate the narrative information to his or her own life. In contrast, identification refers to a reader concentrating on a character, in order to imagine what it is like for him or her. Cohen (2001) has even suggested that when readers strongly identify, they lose awareness of themselves making it impossible to relate information of the story to their self. Thus, self-referencing and identification can be seen as distinct processes through which narratives may lead to effects. By distinguishing between these processes, the exact mechanism through which similarity of the protagonist to the reader leads to narrative effects can be established.

Method

  1. Top of page
  2. Abstract
  3. Method
  4. Results
  5. Discussion
  6. References

Materials

To test the hypotheses, a story was used about a female university student who is diagnosed with intestinal cancer (De Graaf & Hustinx, 2011a; Korda & Van der Does, 1997; Robyns, 2004). Throughout the story, she experiences more and more symptoms; she has to undergo several medical tests; and ultimately she has to be operated on. Fortunately, the operation goes well and the story ends 9 months later when she is starting to pick up her normal life again.

Two versions of this story were constructed. In one version, the protagonist lives at home with her parents, whereas in the other version, the protagonist lives in student housing together with other students. The goal of these versions was to manipulate the similarity of the protagonist on a characteristic that is relevant for the target population of participants, students. The living arrangements are made clear in the beginning of the story. The parts of the story in which the symptoms, the diagnosis and the operation are described, are the same in both versions (see Table 1). Both versions consist of approximately 2,700 words.

Table 1. Example of the Story in Different Versions
Protagonist Lives With ParentsProtagonist Lives in Student Housing
  1. Note: Bold script indicates differences between versions.

[Part of beginning of story: making the living arrangements clear][Part of beginning of story: making the living arrangements clear]
That night, Anne ate with us. My mother had made spaghetti, a favorite in our home. My brother and father were also there, so that it did not fit at our 4-person kitchen table. But we had fetched an extra chair and it worked all the same.That night, all of us ate together. I had made spaghetti, easy and everyone likes it. All my housemates were there and Esther and Anne had also come, so that it did not fit at our kitchen table. But we had fetched extra chairs and it worked all the same.
“Do you have mayonnaise?” Anne asked.“Do you have mayonnaise?” Anne asked.
“You can't be serious?' said my little brother Ben, 'you can't eat mayonnaise with spaghetti.”“You can't be serious?' said Ben who is one of my housemates, 'you can't eat mayonnaise with spaghetti.”
[Part of middle of the story: description of symptoms and diagnosis][Part of middle of the story: description of symptoms and diagnosis]
The waiting room of the specialist was full of older people. I estimated the average age at 55 years. One was clearly younger, 35? No-one even came close to my own 19 years. It seemed so unfair that I had to be here.The waiting room of the specialist was full of older people. I estimated the average age at 55 years. One was clearly younger, 35? No-one even came close to my own 19 years. It seemed so unfair that I had to be here.

Participants and procedure

A total of 220 students participated in this study (76.8% female, 23.2% male). They were all enrolled in a 1st-year communication class and participated for course credit. Their mean age was 19.68 (SD = 1.55). Students participated in groups of 10–25. Within these groups, the versions were randomly distributed. Of this sample, 101 participants lived with their parents, 76 participants lived in student housing and the remaining 43 participants had different living arrangements, such as living together with their partner in a regular apartment. Because of the random distribution, the versions of the story were distributed equally among participants with different living arrangements (χ2(2) = .020, p = .92).1 All participants signed informed consent forms before taking part in the study. After everyone in a group was finished, participants were debriefed and any remaining questions were answered.

Measures

All questions except personal characteristics (gender, age, their own living arrangements), were answered on 7-point Likert scales, ranging from completely disagree to completely agree. The following questions were asked in mixed order, so as not to alert participants to underlying concepts.

Perceived similarity was measured with six questions (α = .86), based on the homophily scale by McCroskey, McCroskey, and Richmond (2006). In some questions, the fact that the protagonist had cancer was excluded specifically, for example “Before the protagonist was afflicted by cancer, her life was like mine.”

Self-referencing was measured with four questions based on Dunlop et al.'s (2010) scale of self-referencing, such as “The story reminded me of experiences in my own life” (α = .82).

Identification was measured with six questions based on Cohen (2001). However, because the original scale showed overlap with other engagement concepts, such as transportation, the scale was adapted to only capture cognitive and emotional perspective taking, which are most central to the identification concept (see Tal-Or & Cohen, 2010). For balance, there were three items for cognitive perspective taking, such as: “During reading, I imagined what it would be like to be in the position of the protagonist” and three items for emotional perspective taking, such as: “I felt tense when the character felt tense”. These six items formed a reliable scale (α = .86).2

In addition to identification, other aspects of narrative engagement were measured as well, to be able to check whether these would unexpectedly be influenced by the similarity manipulation. These measures were based on Busselle and Bilandzic (2009) and Green and Brock (2000). The focus of attention was measured with three items that formed a reliable scale (α = .85), for example “During reading, I was fully concentrated on the story.” Emotion was also measured with three items that formed a reliable scale (α = .78), for example “During reading, the story touched my emotions.” Finally, imagery was measured with three items that formed a reliable scale as well (α = .86), for example “I had a vivid image of the events in the story.”

Finally, story-consistent beliefs were measured based on research by Witte (1994) and Rimal and Real (2003), which shows that perceived risk, perceived severity and perceived efficacy are important in motivating health behavior. Perceived risk was measured by two items that formed a moderately reliable scale (α = .62),3 which were “The chance that I will get intestinal cancer within 3 years is minimal” and “The chance that someone of around 20 will get intestinal cancer is negligible” (Reverse scored). Perceived severity was measured by two items that formed a reliable scale (α = .77), which were “Intestinal cancer is life-threatening” and “The chance that someone who has intestinal cancer will die is large.” Perceived efficacy was measured by two items that did not form a reliable scale (α = .18). Therefore, these items were analyzed separately, which were “If I got intestinal cancer, I would be able to manage” (efficacy to deal with diagnosis) and “It is important to go to a doctor soon if I get persistent pain in the belly” (efficacy to deal with symptoms).

Analysis

To test the hypotheses, a new variable “similarity of protagonist” was constructed. For participants who lived with their parents, this variable was set to similar if they had read the version in which the protagonist lives with her parents, and to dissimilar if they had read the version in which the protagonist lives in student housing. For participants who lived in student housing, this variable was set to similar if they had read the version in which the protagonist lives in student housing, and to dissimilar if they had read the version in which the protagonist lives with her parents. For participants who had different living arrangements, this variable was set to dissimilar in all cases. The hypotheses will be tested with univariate analyses using the new variable as a factor to uncover whether there are main effects of similarity of the protagonist. In addition, the living arrangements of the participants will be included as a factor, to rule out an (unexpected) interaction of similarity with this variable. If main effects of similarity are found, then mediation analyses using a bootstrapping procedure will be used to test the final hypothesis.

Results

  1. Top of page
  2. Abstract
  3. Method
  4. Results
  5. Discussion
  6. References

Effects of similarity

First, a manipulation check tested whether participants who had the same living arrangements as the protagonist perceived themselves as more similar to the protagonist than participants who had different living arrangements. A two-way analysis of variance (ANOVA) with similarity of the protagonist and living arrangements of the participant as factors revealed that there was a main effect of actual similarity of the living arrangements on perceived similarity (F(1, 215) = 14.97, p < .001, η2 = .065). Participants who read the story with the similar protagonist indeed perceived themselves as more similar to the protagonist (M = 4.41, SD = 1.00) than participants who read the story with the dissimilar protagonist (M = 3.79, SD = 1.02).4 There was neither an interaction effect of similarity and living arrangements of the participant, nor a main effect of these living arrangements (both p's > .50).

To test hypothesis 1, two-way ANOVA's were carried out for perceived risk, perceived severity, and perceived efficacy regarding intestinal cancer. Table 2 presents the mean scores and standard deviations of the dependent variables by the two factors. The analyses revealed main effects of similarity on perceived risk (F(1, 215) = 4.94, p < .05, η2 = .022), and on perceived efficacy to deal with symptoms (F(1, 215) = 8.06, p < .01, η2 = .036). Participants who read the story with a similar protagonist perceived more risk (M = 4.58, SD = 1.39) and more efficacy regarding symptoms of intestinal cancer (M = 5.74, SD = .88) than participants who read the story with a dissimilar protagonist (risk: M = 4.29, SD = 1.27; efficacy: M = 5.36, SD = 1.13). There was no main effect of similarity on perceived severity (F(1, 214) = 2.76, p = .098) and on perceived efficacy to deal with a diagnosis of intestinal cancer (F(1, 214) = .86, p = .35). In addition, living arrangements of the participants did not have any main effects, nor interaction effects with similarity on any of the beliefs (all p's > .05).

Table 2. Means and Standard Deviations (Between Brackets) for Beliefs by Condition
 Similar ProtagonistDissimilar Protagonist
 Participant Lives With ParentsParticipant Lives in Student HousingParticipant Lives With ParentsParticipant Lives in Student Housing
  1. Note: Bold script indicates a significant main effect of similarity.

Perceived risk4.40 (1.20)4.83 (1.63)4.32 (1.21)4.03 (1.44)
Perceived severity5.32 (1.09)5.51 (0.89)5.21 (1.04)5.09 (1.05)
Perceived efficacy to deal with diagnosis4.26 (1.32)4.76 (1.33)4.28 (1.57)4.35 (1.29)
Perceived efficacy to deal with symptoms5.69 (0.88)5.81 (0.89)5.34 (1.15)5.25 (1.24)

Underlying processes

To establish whether self-referencing and identification were affected by the similarity of the protagonist to the reader, two-way ANOVA's were carried out with similarity of the protagonist and living arrangements of the participant as factors. The same analyses were done for the additional aspects of narrative engagement to check whether these might unexpectedly be responsible for effects.

Table 3 presents the mean scores and standard deviations of the underlying processes by the two factors. For self-referencing, the ANOVA showed a main effect of similarity (F(1, 215) = 6.78, p = .01, η2 = .031). Participants who read the story with a similar protagonist engaged in more self-referencing (M = 4.44, SD = 1.36) than participants who read the story with a dissimilar protagonist (M = 3.96, SD = 1.52). There was no interaction effect of similarity and living arrangements of the participant, nor was there a main effect of living arrangements on self-referencing (both p's > .20).

Table 3. Means and Standard Deviations (Between Brackets) for Self-referencing, Identification, and Additional Aspects of Narrative Engagement by Condition
 Similar ProtagonistDissimilar Protagonist
 Participant Lives With ParentsParticipant Lives in Student HousingParticipant Lives With ParentsParticipant Lives in Student Housing
  1. Note: Bold script indicates a significant main effect of similarity.

Self-referencing4.21 (1.35)4.77 (1.33)3.83 (1.34)4.00 (1.81)
Identification5.54 (0.79)5.49 (0.95)5.40 (0.80)5.32 (0.93)
Attention5.25 (0.87)5.31 (1.11)5.32 (1.05)5.06 (1.21)
Emotion5.34 (0.94)5.22 (1.10)5.10 (0.90)5.02 (0.96)
Imagery5.96 (0.68)6.02 (0.94)5.88 (0.80)5.68 (0.91)

For identification, the ANOVA showed no main effect of similarity (F(1, 215) = 1.10, p = .29). Participants identified equally with the similar (M = 5.52, SD = .82) as with the dissimilar (M = 5.33, SD = .98) protagonist. There were also no interaction or main effects of living arrangements of the participant on identification (both p's > .60).

Finally, similarity did not have any unexpected effects on the additional aspects of narrative engagement: attention (F(1, 215) = .33, p = .57), emotion (F(1, 215) = 2.68, p = .11), and imagery (F(1, 215) = 2.35, p = .13), nor were there interaction effects with living arrangements of the participant (all p's > .30). These results suggest that self-referencing is the most likely candidate to be responsible for the found effects on story-consistent beliefs. Mediation analyses will establish whether self-referencing is indeed an underlying process of narrative effects.

To test hypotheses 2 and 3 regarding whether the effect of character-reader similarity on story-consistent beliefs is mediated by self-referencing and identification, mediation analysis using a bootstrapping procedure (Preacher & Hayes, 2008) were carried out. As recommended by Hayes (2009), 5,000 bootstrap samples were used to estimate the indirect effects. The analyses used Hayes' model 4 that consists of the independent variable (similarity), the dependent variable (perceived risk or perceived efficacy), and two parallel mediators (self-referencing and identification).5 These analyses show that self-referencing mediated the effects of similarity on perceived risk (b = .058, SE = .044, 95% CI = [.0022 to .1813]) and perceived efficacy to deal with symptoms (b = .052, SE = .034, 95% CI = [.0024 to .1457]).

Identification did not mediate any of the effects of similarity on perceived risk (b = .0011, SE = .030, 95% CI = [−.1038 to .0290]) and perceived efficacy (b = .007, SE = .019, 95% CI = [−.0660 to .0189]). These results show that self-referencing was responsible for the observed effects of character-reader similarity on story-consistent beliefs, whereas identification was not responsible for the effects found in this study.

Discussion

  1. Top of page
  2. Abstract
  3. Method
  4. Results
  5. Discussion
  6. References

The main goal of this experiment was to identify an “active ingredient” of narratives that increases the effectiveness of a story (see Green, 2008). The results of this study show that a protagonist who is similar to the reader is such an element that makes a story have a greater impact. Participants who had read the version of the story in which the protagonist had the same living arrangements as themselves had more story-consistent beliefs than participants who had read the version of the story in which the protagonist had different living arrangements. Hypothesis 1 can thus be accepted; readers of a version of a story in which the protagonist was similar to them on an objective characteristic had several beliefs that were more consistent with the story than readers of a version of the same story in which the protagonist was dissimilar. This means that message adaptation, which has already been found effective in other types of health communication (see Kreuter & Wray, 2003), also increases the effectiveness of narrative messages.

In addition, this study set out to establish the mechanism that underlies the effect of a similar protagonist on beliefs. Results demonstrated that the extent to which readers related the story to themselves and their own lives served as a mechanism of narrative effects. Hypothesis 2 can thus be accepted; The effects of character-reader similarity on story-consistent beliefs was mediated by self-referencing. In contrast, identification did not serve as a mechanism of narrative effects in this study. Therefore, hypothesis 3 is rejected; the effect of character-reader similarity on story-consistent beliefs was not mediated by identification. In summary, similarity of the protagonist to the reader led to narrative effects by making readers think about their selves.

These results do not indicate that identification cannot be a mechanism of narrative impact. Prior research has for instance shown that identification can mediate effects of the perspective from which a story is told on story-consistent attitudes (De Graaf et al., 2012). Thus, identification is a mechanism underlying the effect of the story element perspective. This study showed that the story element of character-reader similarity can have effects through a different underlying mechanism, which is self-referencing. These results underscore the importance of looking at the role of narrative elements in narrative effects, because different elements lead to effects through different mechanisms. By establishing these mechanisms, more insight is gained into the underlying processes of narrative impact and the textual features that trigger these mechanisms.

The lack of effects of similarity on identification and the other engagement variables—attention, emotion, and imagery—provided a clean test of the mediating effect of self-referencing. Because similarity did not affect engagement with the protagonist or the story, these could not be responsible for the found effects of self-referencing. This may seem to contradict previous research that showed engagement variables play a role in narrative impact (e.g., De Graaf et al., 2012; Green & Brock, 2000). However, as explained above, the results of this study do not suggest that identification or other engagement variables cannot function as mechanisms of effects. Instead, the lack of effects of similarity on engagement variables and the full mediation effects of self-referencing on story-consistent beliefs, show that self-referencing is a mechanism of narrative effects that is independent of character and story engagement (which can be a mechanism under different circumstances). Hence, focusing on a character in identification is not the only process that can lead to story-consistent beliefs. Relating the story to oneself and one's own life in self-referencing can do so as well. Instead of contradicting previous studies, this study is thus a contribution to the extant literature, showing that self-referencing—in addition to identification—can also be a mechanism of narrative effects.

Importantly, the effects found in this study most likely reflect belief formation. Because the participants were students, who are generally young and healthy, they most likely have never thought about their risk of becoming ill with intestinal cancer, nor have they considered their efficacy of dealing with the disease before reading the story. This means that the story most likely created the risk and efficacy beliefs, rather than changed already existing beliefs about these issues. Because there are several differences between the processes of belief formation and belief change (Bohner & Dickel, 2011), the implications of this study are limited to belief formation, and future research should ascertain whether the same processes lead to belief change.

Although several beliefs were affected by the manipulation, other beliefs were not. A possible explanation for the lack of effects on the belief of perceived efficacy to deal with a diagnosis of intestinal cancer may be that this is quite a general statement (“I would be able to cope with intestinal cancer”). Previous research has shown that stories are more successful at influencing specific beliefs (e.g., De Graaf, Hoeken, Sanders, & Beentjes, 2009; Green & Brock, 2000), which is corroborated by this study that did show effects on the more specific efficacy belief of being able to deal with potential symptoms of intestinal cancer.

Perceived risk and perceived efficacy to deal with symptoms were affected by similarity. This means that several beliefs that are important in motivating health behavior (Rimal & Real, 2003; Witte, 1994) were influenced by the story. The significant effects on perceived risk and specific self-efficacy are both explained by the mediating effect of self-referencing. This mediation indicates that an increase in thoughts relating the story to the self was responsible for the effect of similarity on these beliefs. These results show that not just the narrative content regarding a health issue (e.g., description of symptoms, diagnosis, and treatment, which were the same in both versions of the story) influences readers of narrative, but also the thoughts they have about this content. When readers link the health content to their own lives, the story has a larger impact on important health beliefs.

An explanation of the effects of similarity found in this study may lie in the increased relevance a reader perceives when a story has a similar protagonist. As explained in the introduction, personal relevance, and self-referencing are distinct concepts, but relevance is likely to influence self-referencing (Kreuter & Wray, 2003). Therefore, whether effects observed for personal relevance also hold for self-referencing becomes an interesting question. Petty and Cacioppo (1986) found that when personal relevance is high, strong arguments lead to message effects, but weak arguments lead to boomerang effects. Self-referencing may similarly be effective only when strong arguments are provided in a message. For narratives, however, what would constitute a strong argument remains unclear, because narrative texts are very different from rhetorical texts (see Green & Brock, 2002). Future research should take up on this issue by defining what would be strong and weak “story arguments” and subsequently testing effects under conditions of high self-referencing and low self-referencing.

This study has several limitations that should also be taken into account in future research. First, this study used a single story to test the hypotheses. Therefore, specific features of this story may have influenced the results. For instance, the story was written from a first-person perspective, which plausibly made readers feel close to the protagonist regardless of whether they were similar to her (Sanders & Redeker, 1996). When reading stories written from a third-person perspective, there may be an influence of similarity on identification.

Also, the protagonist of the story was female. Even though results showed that women did not perceive themselves more similar to the protagonist than men, the fact is that women shared more characteristics with the protagonist than men. Accordingly, the role of gender in relating to characters merits further exploration. As another specific feature, the fact that it was a written story may have limited the amount of engagement of university students. Perhaps a similar protagonist in a filmed narrative may have made this group more engaged in the story. Therefore, future research should test the effects of similarity with different types of stories.

In addition, the sample of participants from the specific group of university students may show different effects than the general population. Future research could look at effects in different groups. Finally, some of the measures were not optimal. For instance, identification was measured with a scale that has not formally been validated. Because this is a central construct in studies on narrative impact, a validated scale would have been desirable (Igartua, 2010; Igartua & Barrios, 2012).

The identification scale used in this study did not include items explicitly capturing the aspect of losing awareness of the self and temporarily “merging” with the character (Cohen, 2001). Future research should explore whether this aspect would be influenced by similarity and mediate effects on story-consistent beliefs. In addition, when measuring the story-consistent beliefs, the two items for efficacy did not form a reliable scale. Future research should preferably use validated scales.

In addition to addressing the limitations of the present study, future research should also extend this study. This study focused on one narrative element that was found to make a narrative more effective, but there are many more potential “active ingredients” of narratives. For instance, the structure of a story seems to influence readers' engagement (Brewer & Lichtenstein, 1982; De Graaf & Hustinx, 2011b). Whether this also leads to more effects should be tested. Also, other features of characters, like the morality of their actions, can influence the level of readers' identification or empathy (Tal-Or & Cohen, 2010; Zillmann, 2006). By establishing if and through which mechanisms such narrative elements lead to more story-consistent beliefs, this line of research expands our insight on the causes and underlying processes of narrative effects. As a step forward in this effort, this study showed that a similar protagonist leads to effects through self-referencing.

Notes
  1. 1

    Whether age differed between participants with different living arrangements was checked, but the association did not reach significance (χ2(2) = 18.08, p = .054). When analyses are run with age as a covariate, results do not change.

  2. 2

    To ensure the scale was unidimensional, a factor analysis was carried out. This analysis showed that the items for cognitive and emotional perspective taking loaded on a single dimension, with factor loadings ranging from .70 to .88 (EV = 3.72).

  3. 3

    Although this value is lower than the normative .70 for the alpha coefficient, alpha underestimates the reliability of short scales (Cortina, 1993). Because this was only a two-item scale, the alpha of .62 was taken as an indication of sufficient consistency.

  4. 4

    Whether the gender of the participants affected similarity was examined, because the protagonist was female in both versions. An ANOVA that included gender showed that there was no significant difference on perceived similarity between men (M = 3.75; SD = 1.10) and women (M = 4.12; SD = 1.03; F(1, 210) = 2.68, ns). Also, there were no interactions with similarity of the protagonist or living arrangements of the participant (all p's > .35). When analyses are run with gender as a control variable, results do not change.

  5. 5

    To rule out alternative explanations, a mediational analysis with self-referencing and identification as sequential mediators was also performed, using model 6. However, the mediation of these two variables in sequence was not significant for percieved risk (b = .007, SE = .018, 95% CI = [−.0535 to .0212]) nor for perceived efficacy (b = .005, SE = .013, 95% CI = [−.0383 to .0151]).

References

  1. Top of page
  2. Abstract
  3. Method
  4. Results
  5. Discussion
  6. References