The Role of Identification and Emotion in Narrative Perspective’s Ability to Influence Attitude, Behavioural Intention and Behaviour
By
Samuel J. Griffiths 11181249
Master’s Thesis
Master’s Programme Communication Science
Graduate School of Communication
Supervisor: Dr. Stephanie Welten Date of Completion: 02-02-2017
Abstract
Much research has recently concerned the persuasive potential of narratives. It has been shown that a first-person narrative perspective is more likely to influence the attitudes, intentions and behaviours of a reader than a third-person equivalent. This study investigates whether these effects occur because a first-person narrative perspective leads to stronger identification with the protagonist than a third-person narrative perspective, whether this stronger identification leads to stronger emotions, and whether these stronger emotions lead to stronger attitudes and behavioural intentions. The direct effect of narrative perspective on behaviour was also tested. A student sample (N = 165) recruited through social media read a text about student stress and mindfulness meditation. The text was randomised between a narrative told from the first-person perspective, a narrative told from the third-person perspective and a non-narrative control. Next, participants completed a survey measuring identification with the protagonist, how strongly they felt fear, sadness and hope while reading, their attitude towards mindfulness meditation apps and their intention to meditate through such apps over the following two weeks. To measure behaviour, participants (N = 111) reported whether they took part in mindfulness meditation two weeks later. Mediation analysis showed narrative perspective did not significantly predict attitude or behavioural intention, neither directly, nor through identification and emotion. However, the third-person narrative perspective had a significant negative indirect effect on attitude and behavioural intention through fear. Additionally, logistic regressions revealed narrative perspective did not predict behaviour during the two weeks following reading the text. These results oppose the success of the narrative perspective manipulation in prior studies, suggesting that the effectiveness of this manipulation depends on other factors. Therefore, potential moderators are proposed for future research, alongside potential advancements in the measure of
The Role of Identification and Emotion in Narrative Perspective’s Ability to Influence Attitude, Behavioural Intention and Behaviour
Narratives are an important part of human communication (Slater, Johnson, Cohen, Comello & Ewoldsen, 2014). They can be defined as “representation[s] of connected events and characters that [have] an identifiable structure, [are] bounded in space and time, and [contain] implicit or explicit messages about the topic being addressed” (Kreuter et al., 2007, p. 222). Narratives capture our attention, entertain us and help us make sense of what it is to be human, however recently, research has focussed on narrative’s persuasive potential (Braddock & Dillard, 2016; De Graaf, Sanders & Hoeken, 2016; Slater et al., 2014). This research has built upon seminal works examining how individuals engage with narratives (Cohen, 2001; Green & Brock, 2000; Oatley 1999; Slater & Rouner, 2002). A recent meta-analysis concluded that narratives can influence attitudes, intentions and behaviours (Braddock & Dillard, 2016). Their persuasiveness has been demonstrated for topics as diverse as euthanasia (De Graaf, Hoeken, Sanders & Beentjes, 2009), racism and voting (Kaufman & Libby, 2012), and safe-sex (Moyer-Gusé, Chung & Jain, 2011). Moyer-Gusé (2008) argues that people are reluctant to alter their attitudes and conduct because they wish to avoid the dissonance inherent in change, however altering something as simple as the perspective from which a narrative is told may counteract this inertia. Using a first-person perspective, where the narrator describes something that happened to them personally (“I kept thinking about the exam”), seems to be more effective in changing attitudes, intentions and behaviour than a third-person perspective, where the narrator describes events that happened to another person (“Annie kept thinking about the exam”) (De Graaf et al., 2016). This difference in persuasive potential between first- and third-person narratives is well evidenced (De Graaf et al., 2016; Nan, Dahlstrom, Richards and Rangarajan, 2015;
Winterbottom, Bekker, Conner & Mooney, 2008), however there is little research testing why this difference exists.
This study will contribute to existing literature by exploring whether identification, and subsequently emotion can explain why a first-person perspective is more persuasive than a third-person equivalent. The first-person perspective may facilitate a process called
identification, which makes it easier for readers to consider alternative viewpoints, and through their imagination, experience the thoughts, actions and emotions of another (Moyer-Gusé, 2008). The emotional aspect is of interest here because the emotions elicited in a narrative can affect its persuasive potential (Nabi & Green, 2015). Indeed, different discrete emotions can influence attitudes, intentions and behaviours in different ways (Dillard & Nabi, 2006; Dillard & Peck, 2000; Nabi, 2002; Tannenbaum, Hepler, Zimmerman, Saul, Jacobs, Wilson, & Albarracín, 2015). Message designers should elicit specific emotions to encourage a particular response, however messages intended to elicit one emotion can elicit other
emotions entirely, and many emotional appeals fail to elicit their target emotion at all (Dillard & Nabi, 2006). By encouraging identification through a first-person narrative perspective, it may be possible to elicit specific emotions reliably (Cohen, 2001; Hoeken & Sinkeldam, 2014; Moyer-Gusé, 2008).
Identification is well suited to the elicitation of specific emotional responses to a narrative (Hoeken & Sinkeldam, 2014). Through identification a reader becomes absorbed in the narrative, adopting a character’s perspective, empathising with them and internalising their goals (Cohen, 2001; Moyer-Gusé, 2008). Events in the narrative affecting these goals should elicit relevant emotions (Moors, Ellsworth, Scherer & Frijda, 2013). For example, in a text about a student, the student failing an exam will elicit sadness in a reader who identifies with that character (Nabi, 2002). The reader has adopted the character’s goals and therefore experiences the same emotions as the character when this goal is not achieved. The resulting
sadness promotes reflection and problem-solving, encouraging the reader to seek ways to alleviate the sadness (Mitchell, Brown Morris-Villagran & Villagran, 2001; Nabi, 2002). The target behaviour presented in the text is then favoured as a means of alleviating this sadness (Nabi, 2015). Therefore, identification could serve as a targeting system for message
designers, channelling the reader’s emotional experience through the perspective of a character to achieve persuasive outcomes.
Hoeken and Sinkeldam (2014) tested the relationship between identification, emotion and persuasion, finding that identification predicts the strength of a reader’s emotional response, and these emotions have a persuasive influence on attitudes. However, they manipulated identification through altering character likeability. This may be difficult to replicate in practice, as different people may like or dislike a character for different reasons. Therefore, the current experiment aims to replicate their findings with a more consistent manipulation; narrative perspective. Furthermore, this study examines whether emotion can mediate the relationship between identification and behavioural intention too (Moyer-Gusé et al., 2011). Testing this will further our understanding of two components of persuasion: identification and emotion. If identification can elicit stronger emotions, then this finding will help message designers to increase the persuasiveness of their content. Thus, the research question addressed in this paper is:
RQ: To what extent do identification and emotion mediate the relationship between the independent variable narrative perspective and the outcome variables attitude, behavioural intention and behaviour?
To test this research question, it was important to select a topic where narratives could be used as a means of persuasion. Narratives are useful tools for health initiatives trying to change attitudes or behaviours (Green, 2006). For example, stress can have serious implications for students, and several websites have suggested mindfulness meditation (MM), as a means of
coping with it (LearnPsychology.org, n.d.; Yale Health. n.d.). This suggestion has been supported by a growing body of research (Lynch, Gander, Kohls, Kudielka & Walach, 2011; Shearer, Hunt, Chowdhury & Nicol, 2016). It is important to help students manage their stress because excessive stress can lead to anxiety, heart palpitations and depression (NHS, n.d.). Mobile applications can teach students how to practice MM themselves (Chittaro & Vianello, 2016; Laurie & Blandford, 2016). However, attitudes can serve as barriers to using such apps (Laurie & Blandford, 2016), as may behavioural inertia (Moyer-Gusé, 2008). If health websites wish to increase student participation in MM, designing persuasive narratives could be an effective solution. Therefore, student stress and MM were selected as the topics of the narratives in this study.
Theoretical Framework
The following section will explore the theory underpinning each of the relationships of interest (see Figure 1).
Figure 1. The experimental model outlining the three hypotheses.
Narrative Perspective and Identification
‘Narrative perspective’ refers to who is conveying the narrative (Hartung, Burke, Hagoort & Willems, 2016), and most common perspectives in narratives are the first- and third-person (Kaufman & Libby, 2012). These differ in persuasive potential. One meta-analysis comparing the two perspectives found that there was over double the probability of
finding persuasive effects with first-person narratives compared to third-person equivalents (Winterbottom et al., 2008). Furthermore, a later meta-analysis found first-person narratives were more likely to influence attitudes, intentions and behaviours than third-person narratives (De Graaf et al., 2016); however, the authors of these meta-analyses do not explain why this may be.
Recent experimental research suggests identification mediates the link between
narrative perspective and persuasive outcomes. Perspective (first-person vs. third-person) was manipulated by Nan et al. (2015), who found that the first-person condition was more
effective than the third- in increasing behavioural intention. They thought this manipulation worked because the readers would identify with the protagonist more in the first-person condition, however they did not measure identification.
As identification involves adopting the perspective of a character, manipulating narrative perspective should lead to significant differences in identification (Cohen, 2001; Cohen et al., 2015; De Graaf, Hoeken, Sanders & Beentjes, 2012). Indeed, in first-person narratives, readers tend to identify more with whichever character is narrating the story (Cohen et al., 2015; De Graaf et al., 2012). One way to lead a reader towards adopting either the role of ‘character’ or ‘spectator’ is through manipulating the narrative perspective
between the first- and third-person (Hartung, et al., 2016). The first-person perspective should create a more salient feeling of familiarity and closeness with the protagonist (assuming the protagonist is the narrator), which better facilitates identification than a third-person version (Kaufman and Libby, 2012). Indeed, Kaufman and Libby’s (2012) findings offer tentative support for this; when a character was a member of the same group as the reader, the level of identification experienced by participants in the first-person condition was stronger than in the third-person condition. Therefore, a first-person narrative should elicit stronger
student. However, Kaufman and Libby (2012) did not measure attitude, nor did they find an effect of narrative perspective on behavioural intention. This may be because they did not account for the mediating role of emotion in the relationship between identification and persuasive outcomes. Thus, this study will address this gap by exploring how identification may lead to persuasive outcomes through eliciting a stronger emotional response.
Identification and Emotion
Though there is strong theoretical support for the idea that identification positively predicts emotion, only one study has tested and supported this (Hoeken & Sinkeldam, 2014). The way in which readers engage with a narrative influences the strength of emotions the text is likely to elicit; however, studies addressing this relationship tend to focus on other
mechanisms of narrative persuasion, not identification (Banerjee & Greene, 2012; van Laer, Ruyter, Visconti & Wetzels, 2014; Murphy, Frank, Moran & Patnoe-Woodley, 2011; Murphy, Frank, Chatterjee & Baezconde-Garbanati. 2013). The cognitive and motivational components of identification are what differentiate it from other related constructs such as perceived similarity between reader and character, and it is these components that cause a reader to adopt the character’s goals and motivations (Cohen, 2001; Oatley, 2002). Readers should then feel emotions in response to events that affect those goals, mirroring the
character’s emotions (Hoeken & Sinkeldam, 2014). According to appraisal theories, emotion is a reaction to personally relevant events or facts (Hoeken & Sinkeldam, 2014; Moors et al., 2013). For example, if someone appraises that barriers preventing them from achieving their goals have been removed, they will feel relief; if they appraise that a situation has resulted in an irrevocable loss of something important to them, they will feel sadness (Hoeken &
Sinkeldam, 2014; Nabi, 2002). Therefore, the perspective taking component of identification is key in leading the reader to adopt the character’s goals and subsequently feeling emotions congruent with the character’s in response to events affecting these goals. Those who identify
less with the character have not adopted the character’s perspective and goals to the same degree as those who identify more strongly (Hoeken & Sinkeldam, 2014). As such, they are less likely to feel strong emotions in response to narrative events relevant to these goals. Thus, high identification should elicit stronger emotions congruent with the experiences of the character. The following section addresses why these emotions are key to the persuasion process.
Emotion, Attitude Change and Behavioural Change
In narratives, emotion can influence attitudes, intentions and behaviour. Busselle and Bilandzic (2009) found emotional engagement with a narrative was correlated with narrative-related attitudes. Similarly, De Graaf et al. (2009), demonstrated that story-consistent
attitudes were predicted by narrative-induced emotion. Bae (2008) also found that empathy was linked to attitude change. However, as these studies measure overall emotional
engagement, elicitation and empathy, they provide no information concerning how specific emotions can persuade. This is an issue, as discrete emotions are likely to have different, often opposing effects on attitudes and intentions (Nabi, 2002).
Intentions to perform a behaviour are influenced by emotional responses and the associated action tendencies (Zebregs, van den Putte, Neijens, & De Graaf, 2015). Action tendencies refer to what the emotion makes someone feel like doing or not doing in regards to a specific goal (Frijda, Kuipers & ter Schure, 1989; Chadwick, 2015). This has been shown for behaviours such as wearing a seatbelt, exercising and fruit consumption (Zebregs et al., 2015). Thus, one should elicit emotions with action tendencies that encourage the target behaviour (Zebregs et al., 2015). Three emotions are expected to be elicited in this study; fear, sadness and hope, each possessing distinct action tendencies (Nabi, 2002).
Fear. Fear is evoked by circumstances that a) pose a perceived threat to an individual
during the narrative when the student is building up to an exam they feel unprepared for. Fear’s action tendency is escape from the source of the perceived threat (Nabi, 2002). Despite the tentative support for current fear appeal models, meta-analyses indicate that fear appeals are effective in positively influencing attitudes, intention and behaviour (Nabi, 2002;
Tannenbaum et al., 2015).
Sadness. Sadness can be evoked by failing to reach a goal (Nabi, 2002). Thus,
sadness was expected to be elicited when the protagonist fails their exam and interview. There is a lack of agreement regarding the action tendency of sadness and how it elicits persuasive outcomes. Nabi (2002) argues that the action tendency of sadness is inaction and withdrawal into oneself to seek comfort and ruminate on what is lost. By contrast, Mitchell et al. (2001) and Yoo, Kreuter, Lai and Foo (2014) argue that people experiencing sadness seek to eliminate the emotion; thus, persuasive outcomes are the result of increased problem-solving and central processing of the message in search of an opportunity to alleviate the sadness. Regardless, sadness has been found to have a positive effect on attitudes (Nan, 2009). Lu (2015) found that a sadness appeal was more effective than a neutral control in changing attitudes and behavioural intentions. Furthermore, Hoeken and Sinkeldam (2014) found that stronger sadness led to more favourable target attitudes. Therefore, stronger sadness should predict stronger attitudes and intentions in this study.
Hope. Hope is a yearning for a better state of affairs than those that exist at present
(Nabi, 2002). It encourages one to capitalise on opportunities and to steer clear of potential punishments (Chadwick, 2015). Hope’s action tendency involves mobilisation and goal-commitment (Prestin, 2013). However, Chadwick (2015) found that hope did not predict behavioural intention; she argues that a hope appeal without a recommended course of action does not allow the reader to channel their emotion productively. Prestin (2013) demonstrated that the inclusion of such a component increased participants’ motivation regarding their
goals. Thus, with a recommended course of action it is expected that participants who feel hope will be motivated to pursue the suggested behaviour (MM) a) to help them reach their academic goals and b) to avoid stress. There is limited research concerning the relationship between hope and attitude change, however Chadwick (2015) found hope positively affected perceived effectiveness, which is an important precursor to attitude change. As such, it seems logical that hope will elicit more positive attitudes towards behaviours that readers believe will facilitate reaching their goals.
Though this study focuses on discrete emotions, separate hypotheses were not crafted for each emotion. This was because, as shown above, the action tendencies of fear, sadness and hope are all expected have a positive effect on attitude and behavioural intention. Fear is expected to make the participant seek out a means of reducing their fear (Nabi, 2002), sadness is expected to increase issue-relevant thinking and problem-solving (Yoo et al., 2014), and hope is expected to make the participant seek a means of achieving their goals (succeeding in their studies) while avoiding suffering (stress) (Chadwick, 2015). These should all lead to more positive attitudes and behavioural intentions regarding the target behaviour, MM with an app or website. Though the same effect was expected from the three emotions, each was measured and analysed independently to examine this relationship in more detail.
Based upon the above literature the following hypotheses are proposed: H1: Narrative perspective will have an effect on attitude that is mediated first by
identification, and subsequently emotion (fear, sadness, hope), so that in comparison to the third-person and control conditions, the first-person perspective leads to stronger
identification, which leads to stronger emotion, which leads to more positive attitudes. H2: Narrative perspective will have an effect on message-consistent behavioural intention that is mediated first by identification, and subsequently emotion (fear, sadness, hope), so that
in comparison to a third-person and control conditions, the first-person perspective leads to stronger identification, which leads to stronger emotion, which leads to stronger message-consistent behavioural intention.
H3: Narrative perspective will predict behaviour, so that participants in the first-person narrative perspective condition will be more likely to perform the target behaviour than those in the third-person perspective condition1.
Method
Participants, Design and Procedure
Three hundred and thirty-two participants were randomly allocated to one of three narrative perspective conditions (first-person narrative vs. third-person narrative vs. control group), creating a between-subjects online experiment. Participants in the third-person
condition were further randomly assigned to either the male or female version of the narrative in case participants identified more with a character that was the same gender as themselves. Snowball sampling was used to recruit students through social media. Participants were excluded if they were under 18 (N = 1), were no longer a student (N = 20), did not complete the survey (N = 143) or listed their gender as ‘other’ (N = 3) (this facilitated the process of testing gender differences between the two third-person conditions). After this, 165
participants who completed the first survey (T1) remained (Mage = 22.76, SDage = 2.88; 121 women, 44 men). Out of the 149 willing to be contacted for the second part of the study two weeks later (T2), 111 completed the measures (Mage = 22.79, SDage = 2.82; 84 women, 44 men).
After giving informed consent, participants were presented with their text followed by the T1 questionnaire. Finally, participants were asked if they wanted to take part in the
second part of the study two weeks later. Those who agreed were asked to provide an email address and a unique identifier. Participants who did not agree were sent to a debrief page.
All participants were presented with links to webpages about coping with student stress and websites providing MM courses and apps. All participants were provided with a text-box in which they could provide feedback.
Those who agreed to take part in the second part were emailed a link two weeks later. This led them to the T2 questionnaire. After completing the study, participants were
debriefed and thanked for taking part.
Materials
Student, medical and psychological websites (LearnPsychology.org, n.d.; NHS, n.d.; WebMD, n.d.) were consulted to ensure a realistic representation of the physical and
psychological manifestations of stress. Thus, the narrative focused on a student and was structured to show fear prior to an exam, sadness after they failed the exam and reflected upon their unreached goals, and hope as they are introduced to a MM app which facilitates their studying and opens the possibility of future success (Chadwick, 2015; Nabi, 2002). The pronouns and matching word-endings were altered (Kaufman & Libby, 2012; Nan et al., 2015) to manipulate the narrative perspective between first- and third-person. The control text contrasted with the narratives by focusing on arguments, reasoning and claims, as opposed to the narratives which exhibited a temporally-bound progression of events involving, a protagonist, scenes, conflict and resolution (Lu, 2013) (see Appendix A for examples of the texts). All texts were pretested by having students proof-read them and note necessary changes or improvements. The first-person text was 1,953 words long (Flesch reading ease: 78.2), the third-person 1,976 words long (Flesch reading ease: 77.0) and the control text 1,328 words long (Flesch reading ease: 58.6). This discrepancy in length was necessary as much of the length in the narratives came from character and plot development and contextual details. No additional material was incorporated to increase the length of the control text to ensure all texts provided the same information.
Measures
T1 (see Appendix 2 for more details on the items). Identification was measured
using ten 7-point Likert items (1 = Strongly disagree; 7 = Strongly agree) adapted from Cohen (2001). Principal components analysis revealed 2 factors with an eigenvalue above 1, however, based upon analysis of the scree plot’s point of inflexion, a one factor solution was selected as the best fit (E.V. = 5.38; Cronbach’s α = .90, Field, 2009).
All items measuring emotion followed the phrase “While reading the text, to what extent did you experience these emotions?”. Items were measured on Likert scales ranging from 1 (not at all) to 7 (an extreme amount). Sadness (lonely, grief, sad, empty) (E.V. = 2.84; Cronbach’s α = .861) and fear (terror, scared, fear, panic) (E.V. = 3.33; Cronbach’s α = .93) were measured with 4 items each, all taken from the discrete emotion questionnaire
developed by Harmon-Jones, Bastian and Harmon-Jones (2016). Hope (hopeful, inspired) was measured with two items taken from Prestin (2013) ( r = .62, p < .001).
Attitude towards MM apps and websites was measured using seven 7-point semantic differentials adapted from Kim and Niederdeppe (2014) (e.g. foolish-wise,
harmful-beneficial, useless-useful) which followed the clause “For me, taking part in a mindfulness meditation app or website would be” (E.V. = 4.81; Cronbach’s α = .92).
Behavioural intention was measured with three items adapted from Conner and Sparks (2005), adding the openings “I intend to…” (1 = definitely do; 7 = definitely do not), “I will try to…” (1 = definitely will not; 7 = definitely will) and “I will make an effort to…” (1 = definitely false; 7 = definitely true) to the phrase “…take part in mindfulness meditation with a mindfulness meditation app or website within the next two weeks.” (E.V. = 2.46; Cronbach’s α = .89).
T2. Behaviour was measured with the item; “I took part in mindfulness meditation
from Conner & Sparks, 2005). One other item was included as a measure of behaviour; “I took part in mindfulness meditation with a mindfulness meditation app or website within the last two weeks” (yes; no). This was included because some participants indicated in their comments that they wanted to try MM but not through apps.
Control variables. All control variables were measured at T1 only. Perceived stress
was measured on 7-point Likert scales (1 = never; 7 = very often) with the ten-item version of Cohen, Kamarck and Mermelstein’s (1983) perceived stress scale (PSS-10) (Cole, 1999). All items followed the phrase “in the last month, how often have you…”. Principal components analysis revealed that one reverse-coded item (“…dealt successfully with irritating life hassles”), did not load on to the same factor as the other nine when coefficients smaller than .40 were suppressed, thus this item was not included in the scale. Furthermore, there were 2 factors that had an eigenvalue above 1, however based upon analysis of the scree plot’s point of inflexion a one factor solution was deemed to be the best fit (E.V. = 4.44; Cronbach’s α = .87, Field, 2009).
Four items were created to measure T1 MM behaviour. Each was analysed separately as it was reasoned that they measure separate constructs. These were “Up until now how regularly have you practiced mindfulness meditation?” (1 = never; 7 = regularly), “Do you currently have a mindfulness meditation app installed on your phone?” (yes; no), “Up until now how regularly have you used an app or website to practice mindfulness meditation?” (1 = never; 7 = regularly), and “In regards to mindfulness, how knowledgeable do you consider yourself to be?” (1 = not at all knowledgeable; 7 = extremely knowledgeable).
Participants were asked to provide their gender, age and nationality. Proficiency in English was also measured on a 7 point Likert scale (1 = very basic; 7 = at a native level), as it was reasoned that this could affect their comprehension of the narrative. For descriptive statistics of both samples, see Table 1.
Table 1
Means, Standard Deviations and Ranges of Measures Describing both the T1 and T2 Samples
T1 (N = 165) T2 (N = 111)
Variable M SD Range M SD Range
Age 22.76 2.88 18 – 36 22.79 2.82 18 – 36 Perceived stress 4.23 1.15 1.56 – 7.00 4.23 1.14 1.56 – 6.56 T1 MM 2.21 1.62 1 – 7 2.22 1.61 1 – 7 T1 MM app use 2.76 1.38 1 – 7 1.80 1.37 1 – 7 T1 app installation 1.86 .35 1 – 2 1.84 .37 1 – 2 T1 mindfulness knowledge 3.05 1.75 1 – 7 3.10 1.78 1 – 7 English Proficiency 6.43 .87 2 – 7 6.47 .85 3 – 7 Results
To check if the two third-person conditions (male and female protagonists) could be combined, two independent t-tests were conducted with gender of the third-person condition as the IV and identification as the DV. The first included only the female participants in the two third-person conditions (N = 46). No significant difference was found between the female protagonist (M = 4.98, SE = .27) and the male protagonist condition (M = 5.18, SE = .28) t(44) = -.519, p = .606. This analysis was repeated with only the males (N = 15). A bootstrap of 10,000 samples was used, as the sample for this analysis included less than 30 participants. No significant difference was found between the female (M = 4.98, SE = .27) and the male conditions (M = 5.18, SE = .28), t(13) = 1.20, p = .253. Thus, gender had no significant effect on identification, so the two third-person conditions were combined.
To confirm the manipulation worked, two ANOVAs were conducted on items adapted from Batson, Early and Salvarani (1997). The first used the Likert item “...to what extent did you concentrate on the feelings of the student portrayed in the text?” (1 = Not at all; 7 = Very much) as the DV, and narrative perspective (first-person vs third-person vs control) as the IV. Participants in the third-person condition were expected to focus more on the character’s feelings than those in the first-person or control conditions. There was no significant difference, F(2, 162) = .835, p = .436, with participants in the third-person narrative perspective condition not scoring any higher on focusing on the character’s feelings (M = 4.62, SD = 1.56) than those in the first-person narrative perspective condition (M = 4.93, SD = 1.31) or the control condition (M = 4.66, SD = 1.35). The second ANOVA used the Likert item “...to what extent did you concentrate on how you yourself were feeling?” (1 = Not at all; 7 = Very much) as the DV, and the narrative perspectives (first-person vs third-person vs control) as the IV. If the reader adopted the character’s perspective as they were expected to in the first-person narrative condition, then their experience of the narrative’s events and the character’s experience should be indistinguishable. There was no significant effect, F(2, 162) = 1.708, p = .184, with participants in the first-person narrative perspective condition not scoring any higher on focusing on their own emotions (M = 4.13, SD = 1.73) than those in the third-person narrative perspective condition (M = 4.20, SD = 1.88), or the control condition (M = 4.75, SD = 1.81). According to this check, the manipulation failed.
Randomisation check
A randomisation check was performed via a MANOVA. Narrative perspective was used as the IV and age, T1 MM, T1 app use, T1 mindfulness knowledge, perceived stress and English proficiency were used as DVs to assess whether the conditions differed on these measures. In the T1 sample there was no significant multivariate effect, Wilks’ Lambda = .955, F(12, 312) = .600, p = .842. Additionally, there were no significant univariate effects of
narrative perspective on age, perceived stress, T1 MM, T1 MM app use, T1 mindfulness knowledge or English proficiency, all Fs(2, 161) <= .672, all ps >= .512. This was also the case for the T2 sample, with no significant MANOVA effect, Wilks’ Lambda = .927 F(12, 208) = .662, p = .787, or individual effects of narrative perspective on age, perceived stress, T1 MM, T1 MM app use, T1 mindfulness knowledge or English proficiency, all Fs(2, 108) <= .861, all ps >= .426.
Two chi-square analyses were performed for the T1 sample. The first included narrative perspective and gender. In the first-person condition (N = 60), 42 people (70.0%) were female. This was not significantly different from the third-person condition (N = 61) of whom 46 (75.4%) were female, or the control condition (N = 44), of whom 33 (75.0%) were female, χ2 (2) = .538, p = .764. The next chi-square analysis included narrative perspective and T1 app installation. In the first-person condition, 10 people (16.7%) had a MM app on their phone. This was not significantly different from the third-person condition of whom 7 (11.5%) had an app, or the control condition, of whom 6 (13.1%) had an app, χ2 (2) = .684, p = .710.
These analyses were repeated for the T2 sample. In the first-person condition (N = 41), 31 (75.6%) were female. This was not significantly different from the third-person condition (N = 42) of whom 32 (76.2%) were female, or the control condition (N = 28), of whom 21 (75.0%) were female, χ2 (2) = .013, p = .993. The next chi-square analysis included narrative perspective and T1 app installation. In the first-person condition, 8 (19.5%) Had a MM app on their phone. This was not significantly different the third-person condition of whom 6 (14.3%) had an app, or the control condition, of whom 4 (14.3%) had an app, χ2 (2) = .520, p = .771.
Therefore, participants were correctly randomly distributed between conditions, so no covariates were included beyond those identified in the correlation analysis below.
Correlation analysis
A correlation analysis was run on the DVs plus the additional variables gender, age, T1 MM, T1 MM app use, T1 app installation, T1 mindfulness knowledge, perceived stress and English proficiency to assess if any of them could be covariates for the DVs. Correlations between DVs and control variables found to be significant at the .05 level led to the control variable being included as a covariate in analyses containing the relevant DV (see Tables 2 and 3).
Table 2
Key Descriptive Statistics of all Significant Covariates, Listed with their Corresponding Variables in the T1 analyses (N = 165) Dependent variable Covariate M SD Pearson correlation p
Identification Perceived stress 4.27 1.21 .453 .000
Gender 1.25 0.43 -.171 .028
Fear Perceived stress 4.27 1.21 .333 .000
T1 mindfulness knowledge 2.99 1.81 -.172 .027
Sadness Perceived stress 4.27 1.21 .349 .000
English proficiency 6.43 .873 -.175 .025
Hope Perceived stress 4.27 1.21 .211 .007
Attitude Perceived stress 4.27 1.21 .238 .002
Gender 1.25 .43 -.203 .009
T1 MM 2.22 1.69 .332 .000
T1 MM app use 1.83 1.51 .345 .000
T1 app installation 1.88 0.33 -.237 .002
Behavioural intention
Perceived stress 4.27 1.21 .281 .000
Table 3
Key Descriptive Statistics of all Significant Covariates, Listed with their Corresponding Variables in the T2 analyses (N = 111)
Dependent variable
Covariate name Mean SD Pearson
correlation p.
Identification Perceived stress 4.23 1.14 .523 .000
MM T1 MM 2.22 1.61 -.408 .000 T1 MM app use 1.80 1.37 -.282 .003 T1 app installation 1.84 0.37 .263 .005 T1 mindfulness knowledge 3.20 1.78 -.217 .022 MM with app or website T1 MM 2.18 1.61 -.281 .003 T1 app installation 1.83 0.37 .237 .012 Main Analyses
Hypothesis 1: Attitude. The first hypothesis, that the first-person narrative
perspective would have a stronger positive effect on attitude than the third-person narrative perspective and control condition, and that this effect would be positively mediated by identification and emotion (fear, sadness, hope), was tested using PROCESS (model 6; 10,000 bootstraps, 95% confidence interval) (Hayes, 2013). Conditions were dummy-coded to compare the first- to the person, the first-person to the control group, and the third-person to the control group. See Table 2 for covariates. To provide context for all following
analyses, means and standard deviations for all experimental variables across the three conditions are provided in Table 4.
Table 4
Mean Scores and Standard Deviations of Identification, Fear, Sadness, Hope, Attitude and Behavioural Intention Between the Three Experimental Conditions
DV IV M SD Identification First-person 4.97 0.92 Third-person 4.89 1.31 Control 4.85 0.91 Fear First-person 3.58 1.50 Third-person 3.32 1.73 Control 2.70 1.59 Sadness First-person 3.00 1.33 Third-person 3.03 1.62 Control 2.31 1.17 Hope First-person 3.74 1.41 Third-person 3.72 1.60 Control 3.35 1.27 Attitude First-person 4.77 1.25 Third-person 4.66 1.29 Control 4.98 0.94
Behavioural intention First-person 3.69 1.62
Third-person 3.67 1.87
Control 3.97 1.71
Fear. The first-person vs third-person analysis showed no significant indirect effect of narrative perspective on attitude (see Figure 2). Narrative perspective did not significantly predict identification however participants who experienced more identification felt more fear. Those who experienced stronger fear did not report significantly more positive attitudes. This pattern was also present in the first-person vs control analysis (see Figure 3). In the third-person vs control analysis those who experienced more identification felt more fear and those who felt more fear reported significantly less positive attitudes (see Figure 4). In the first-person vs third-person analysis a confidence interval for the indirect effect (a1d21b2 =
.004) included zero (95%CI = -.0718 to .0161) meaning that the effect was not significant. This was also true for the first-person vs control analysis (a1d21b2 = -.003, 95%CI = -.0475 to
.0108), and the third-person vs control analysis (a1d21b2 = -.018, 95%CI = -.1019 to .0430).
Therefore, narrative perspective did not significantly predict attitude, nor did it have a significant indirect effect through identification and fear. However, in the third-person vs control analysis narrative perspective did have a significant indirect effect on attitude through fear when identification was not included. A third-person perspective elicited more fear than the control condition which in turn predicted less positive attitudes in the third-person condition than in the control group. A confidence interval for this indirect effect (a2b2 =
-.147) was entirely below zero (95%CI = -.3757 to -.0174) meaning that the effect was significant.
Figure 2. A model displaying coefficients and significance levels of paths in a mediation testing whether narrative perspective (first-person vs. third-person) has an indirect effect on attitude through identification and fear. *p < .05. ** p < .01.
Figure 3. A model displaying coefficients and significance levels of paths in a mediation testing whether narrative perspective (first-person vs. control) has an indirect effect on attitude through identification and fear.
*p < .05. ** p < .01.
Figure 4. A model displaying coefficients and significance levels of paths in a mediation testing whether narrative perspective (third-person vs. control) has an effect on attitude through identification and fear.
*p < .05. ** p < .01.
Sadness. The first-person vs third-person analysis showed no significant indirect effect of narrative perspective on attitude (see Figure 5). Narrative perspective did not significantly predict identification however participants who experienced more identification felt more sadness. Those who experienced stronger sadness did not report significantly more positive attitudes. This pattern was also present in the first-person vs control and third-person
vs control analyses (Figure 6 and Figure 7). In the first-person vs third-person analysis a confidence interval for the indirect effect (a1d21b2 = .000) included zero (95%CI = -.0168 to
.0169). This was also the case for both the first-person vs control (a1d21b2 = .005, 95%CI =
-.0430 to .0071) and the third-person vs control analyses (a1d21b2 = -.006, 95%CI = -.0519 to
.0058). Therefore, narrative perspective did not significantly predict attitude, nor did it have a significant indirect effect through identification and sadness.
Figure 5. A model displaying coefficients and significance levels of paths in a mediation testing whether narrative perspective (first-person vs. third-person) has an effect on attitude through identification and sadness.
*p < .05. ** p < .01.
Figure 6. A model displaying coefficients and significance levels of paths in a mediation testing whether narrative perspective (first-person vs. control) has an effect on attitude through identification and sadness.
*p < .05. ** p < .01.
Figure 7. A model displaying coefficients and significance levels of paths in a mediation testing whether narrative perspective (third-person vs. control) has an effect on attitude through identification and sadness.
Hope. The first-person vs third-person analysis showed no significant indirect effect of narrative perspective on attitude (see Figure 8). Narrative perspective did not significantly predict identification, however participants who experienced more identification felt more hope. Those who experienced stronger hope did not report significantly more positive attitudes. This pattern was also present in the third-person vs control analysis (see Figure 9) In the first-person vs control analysis those who experienced more identification felt more hope and those who felt more hope reported significantly more positive attitudes (see Figure 10). In the first-person vs third-person analysis a confidence interval for the indirect effect (a1d21b2 = .000) included zero (95%CI = -.0152 to .0237). This was also the case in both the
first-person vs control (a1d21b2 = .017, 95%CI = -.0055 to .0782) and the third-person vs
control analyses (a1d21b2 = .003, 95%CI = -.0096 to .0504). Therefore, narrative perspective
did not significantly predict attitude, nor did it have a significant indirect effect through identification and hope, so H1 was rejected.
Figure 8. A model displaying coefficients and significance levels of paths in a mediation testing whether narrative perspective (first-person vs. third-person) has an effect on attitude through identification and hope.
*p < .05. ** p < .01.
Figure 9. A model displaying coefficients and significance levels of paths in a mediation testing whether narrative perspective (third-person vs. control) has an effect on attitude through identification and hope
*p < .05. ** p < .01.
Figure 10. A model displaying coefficients and significance levels of paths in a mediation testing whether narrative perspective (first-person vs. control) has an effect on attitude through identification and hope.
*p < .05. ** p < .01.
Hypothesis 2: Behavioural intention. The second hypothesis, that the first-person
narrative perspective would have a stronger positive effect on behavioural intention than the third-person narrative perspective and control conditions, and that this effect would be
positively mediated by identification and emotion, was also tested using PROCESS (model 6; 10,000 bootstraps, 95% confidence interval). Conditions were dummy-coded and compared as before. See Table 2 for covariates.
Fear. The first-person vs third-person analysis showed no significant indirect effect of narrative perspective on behavioural intention (see Figure 11). Narrative perspective did not significantly predict identification, however participants who experienced more identification felt more fear. Those who experienced stronger fear did not report significantly stronger behavioural intention. This pattern was also present in both the first-person vs control and person vs control analyses (Figure 12 and Figure 13). In the first-person vs third-person condition a confidence interval for the indirect effect (a1d21b2 = .006) included zero
(95%CI = -.0162 to .0928). This was also the case for the first-person vs control (a1d21b2 =
-.008, 95%CI = -.0736 to .0084) and third-person vs control analyses (a1d21b2 = -.0181,
95%CI = -.1147 to .0371). Therefore, narrative perspective did not significantly predict behavioural intention, nor did it have a significant indirect effect through identification and
fear together. However, in the third-person vs control analysis narrative perspective did have a significant effect on behavioural intention through fear when identification was not
included. A third-person perspective elicited more fear than the control condition which in turn predicted less behavioural intention in the third-person condition than in the control group. This mirrors the mediating role of fear in the relationship between narrative
perspective and attitude shown in the third-person vs control analysis above. A confidence interval for this indirect effect (a2b2 = .156) was entirely below zero (95%CI = .4315 to
-.0149).
Figure 11. A model displaying coefficients and significance levels of paths in a mediation testing whether narrative perspective (first-person vs. third-person) has an effect on behavioural intention through identification and fear. *p < .05. ** p < .01.
Figure 12. A model displaying coefficients and significance levels of paths in a mediation testing whether narrative perspective (first-person vs. control) has an effect on behavioural intention through identification and fear. *p < .05. ** p < .01.
Figure 13. A model displaying coefficients and significance levels of paths in a mediation testing whether narrative perspective (third-person vs. control) has an effect on behavioural intention through identification and fear. *p < .05. ** p < .01.
Sadness. The first-person vs third-person analysis showed no significant indirect effect of narrative perspective on behavioural intention (see Figure 14). Narrative perspective did not significantly predict identification, however participants who experienced more identification felt more sadness. Those who experienced stronger sadness did not report significantly stronger behavioural intention. This pattern was also present in the first-person vs control and third-person vs control analyses (Figure 15 and Figure 16). In the first-person vs third-person analysis, a confidence interval for the indirect effect (a1d21b2 = -.0014)
included zero (95%CI = -.0151 to .0440). This was also the case for both the first-person vs control (a1d21b2 = -.008, 95%CI = -.0823 to .0067) and the third-person vs control analyses
(a1d21b2 = -.008, 95%CI = -.0823 to .0067). Therefore, narrative perspective did not
significantly predict behavioural intention, nor did it have a significant indirect effect through identification and sadness.
Figure 14. A model displaying coefficients and significance levels of paths in a mediation testing whether narrative perspective (first-person vs. third-person) has an effect on behavioural intention through identification and sadness. *p < .05. ** p < .01.
Figure 15. A model displaying coefficients and significance levels of paths in a mediation testing whether narrative perspective (first-person vs. control) has an effect on behavioural intention through identification and sadness. *p < .05. ** p < .01.
Figure 16. A model displaying coefficients and significance levels of paths in a mediation testing whether narrative perspective (third-person vs. control) has an effect on behavioural intention through identification and sadness. *p < .05. ** p < .01.
Hope. The first-person vs third-person analysis showed no significant indirect effect of narrative perspective on behavioural intention (see Figure 17). Narrative perspective did not significantly predict identification, however participants who experienced more
identification felt more hope. Those who experienced stronger hope did not report
significantly stronger behavioural intention. This pattern was also present in the third-person vs control analysis (see Figure 18). In the first-person vs control analysis those who
experienced more identification felt more hope and those who felt more hope reported significantly stronger behavioural intention (see Figure 19). In the first-person vs third-person condition a confidence interval for the indirect effect (a1d21b2 = .0136) included zero
(95%CI = -.0240 to .1122). This was also the case for both the first-person vs control (a1d21b2 = .0345, 95%CI = -.0105 to .1426) and third-person vs control analyses (a1d21b2 =
.010, 95%CI = -.0197 to .1109). Therefore, narrative perspective did not significantly predict behavioural intention, nor did it have a significant indirect effect through identification and hope, so H2 was rejected.
Figure 17. A model displaying coefficients and significance levels of paths in a mediation testing whether narrative perspective (first-person vs. third-person) has an effect on behavioural intention through identification and hope. *p < .05. ** p < .01.
Figure 18. A model displaying coefficients and significance levels of paths in a mediation testing whether narrative perspective (third-person vs. control) has an effect on behavioural intention through identification and hope. *p < .05. ** p < .01.
Figure 19. A model displaying coefficients and significance levels of paths in a mediation testing whether narrative perspective (first-person vs. control) has an effect on behavioural intention through identification and hope. *p < .05. ** p < .01.
Hypothesis 3: Behaviour. MM with a MM app or website. To test hypothesis 3, that
participants exposed to the first-person text would be more likely to take part in MM with a MM app or website than those exposed to the third-person or control texts, three logistic regressions were performed. See table 3 for covariates and table 5 for a summary of
coefficients. These logistic regressions compared the first-person condition to the third- (-2 Log Likelihood = 64.38, Nagelkerke R2 = .15, χ2 (8) = 4.88, p = .770), the first-person to the control condition (-2 Log Likelihood = 48.58, Nagelkerke R2 = .13, χ2 (8) = 6.29, p = .614) and the third-person to the control condition (-2 Log Likelihood = 49.86, Nagelkerke R2 = .18, χ2 (8) = 5.59, p = .693).
Contrary to H3, participants in the first-person condition were not significantly more likely to use a MM app or website than participants exposed to the third-person condition (b = -.10, OR = .90, p = .875) or the control condition (b = .36, OR = 1.43, p = .663). Neither were participants in the third-person condition more likely to use an app or website than those in the control condition (b = .43, OR = 1.53, p = .592).
General MM. These analyses was repeated using general MM as the outcome variable, comparing the first-person condition to the third- (-2 Log Likelihood = 73.63, Nagelkerke R2 = .29, χ2 (8) = 6.36, p = .607), the first-person to the control condition (-2 Log Likelihood = 69.44, Nagelkerke R2 = .19, χ2 (8) = 10.83, p = .212) and the third-person to the control condition (-2 Log Likelihood = 54.61, Nagelkerke R2 = .39, χ2 (8) = 9.13, p = .331).
Again, contrary to H3, participants in the first-person condition were not significantly more likely to take part in MM generally than either the participants exposed to the third-person condition (b = .53, OR = 1.70, p = .391) or the control condition (b = .06, OR = 1.06, p = .920). Neither were participants in the third-person condition more likely to use an app or website than those in the control condition (b = -.66, OR = .52, p = .376). Narrative
perspective did not significantly predict MM with a MM app or website, or general MM so H3 was rejected.
Summary of logistic regression analyses for narrative perspective predicting mindfulness meditation (MM) with an app or website and general MM.
MM with app or website General MM Analysis B SE B OR B SE B OR 1st person vs 3rd person Narrative perspective -.10 0.65 0.90 .53 0.62 1.70 Perceived stress .02 0.30 1.02 .36 0.30 1.43 T1 app installation -.74 0.80 0.48 -.27 0.80 0.76 T1 MM .38 0.21 1.46 .82* 0.37 2.26 T1 MM with app - - - -.11 0.33 0.90 T1 MM knowledge - - - -.14 0.25 0.87 Constant -1.29 1.81 0.28 -3.93 2.66 0.02 1st person vs control Narrative perspective .37 0.82 1.43 .06 0.62 1.06 Perceived stress -.04 0.36 0.96 .19 0.30 1.21 T1 app installation -.1.26 0.94 0.28 -.63 0.86 0.53 T1 MM .18 0.26 1.20 .67* 0.34 1.96 T1 MM with app - - - -.30 0.35 0.74 T1 MM knowledge - - - -.15 0.24 0.87 Constant -.23 2.65 0.80 -1.33 2.53 0.26 3rd person vs control Narrative perspective .43 0.79 1.53 -.66 0.75 0.52
Perceived stress .38 0.37 1.46 .96* 0.42 2.62 T1 app installation .06 1.02 .1.07 .34 1.14 1.40 T1 MM .51* 0.24 1.66 .93 0.38 2.52 T1 MM with app - - - -.01 0.34 0.99 T1 MM knowledge - - - -.08 0.29 0.92 Constant -5.19 3.35 2.40 -7.86 3.92 0.00
B = unstandardized regression coefficient, SE B = standard error associated with the b-coefficient and OR = odds ratio *p < .05
Conclusion and Discussion
The purpose of this study was to further explore the mechanisms through which narrative perspective (first- vs third-person) can alter a narrative’s ability to elicit persuasive outcomes. It was hypothesised that narrative perspective would influence attitude and behavioural intention concerning MM through identification and emotion, and that it would have a direct effect on behaviour(see Footnote 1). There was no significant difference in the ability of the three narrative perspective conditions to predict identification, attitude,
behavioural intention or behaviour. In every analysis testing whether identification or emotion mediates the relationship between narrative perspective and attitude or behavioural intention, identification predicted emotion. Both the first- and third-person conditions predicted stronger fear and sadness than the control condition. This suggests narratives predict stronger emotion than non-narratives. Furthermore, in the third-person narrative perspective vs control analyses, the results found a significant indirect effect of narrative perspective on both attitude and behavioural intention through fear. The third-person narrative perspective predicted more fear than the control text, and this increased fear predicted weaker attitudes and behavioural intention. Analysis of the T2 data revealed that narrative perspective did not significantly predict behaviour either.
Before discussing these findings, we must address one limitation of this study; the non-significance of the manipulation check. On reflection, this may have been due to the items used rather than the effectiveness of the manipulation. The manipulation of narrative perspective has been effective in eliciting identification, attitudes, behavioural intention and behaviour previously (De Graaf et al., 2016; Kaufman & Libby, 2012; Nan et al., 2015). The items measured the extent to which the participant focused on their own feelings and the character’s feelings, overlapping with the empathetic component of emotion, and serving more as a dependent variable than a manipulation check. Thus, even though the manipulation did not significantly predict the DVs, it may not have failed. Due to this limitation, it is impossible to determine from the present data whether the manipulation of narrative perspective succeeded or not. Furthermore, the extent to which the participants focused on their feelings, or the characters, likely varied as they read. This extends to the identification measure. Cohen (2001) acknowledges that identification fluctuates as the text progresses and develops, thus its frequency should be measured as well as its intensity. If we measure identification with a single scale after reading, insight into how this mechanism works is limited. For example, readers may have identified with the student during the section about the exam and job-seeking, but not when they began to meditate. In future research, this subtlety could be observed using fMRI. A meta-analysis by Fan, Duncan, de Greck and Northoff (2011) has shown there is a core network in the brain that is consistently active during empathy. They argue the brain regions involved could be the root of the simulation of the thoughts and feelings of others inherent in empathy. This overlaps with Cohen’s
conception of identification (2001). Thus, future studies could gain insight into patterns of identification during exposure to a narrative with an fMRI scanner. By understanding how identification varies during reading, we could examine whether identification’s persuasive potential depends on when or how often the reader identifies with the protagonist. Perhaps a
certain frequency or duration of identification is as important as the strength of identification, or perhaps identification only has an effect when the reader experiences strong identification during the recommended action component of the narrative, with mean identification over the course of reading having little effect. Thus, fMRI measures could further our insight in to the role of identification in the persuasion process, allowing message designers to craft more effective messages.
Why Did Narrative Perspective Not Predict Identification?
The first-person narrative perspective did not predict stronger identification than either the third-person or control conditions as in previous literature (Kaufman & Libby, 2012; Nan et al., 2015). One possible explanation is that the effect of the manipulation was suppressed by strong perceived similarity; something provided across all conditions by the student character and university setting. The texts focused on a student struggling with exams and job-hunting, drawing from real-life examples of student stress and relevant medical information. Thus, participants likely recognised the experience of the student protagonist as being familiar and representative of their lives, making it easier to identify with them. This could have caused participants across all conditions to identify with the student in the text and show no significant difference in persuasive outcomes. Hoeken, Kolthoff & Sanders (2016) examined the effect of both narrative perspective and perceived similarity on attitude through the mediator identification. For some characters but not others, perceived similarity neutralised the effect of identification. Hoeken et al. (2016) acknowledge that scores of perceived similarity were quite low, and suggest examining whether tailoring a character to the target audience elicits enough perceived similarity to overpower narrative perspective. This makes sense, since the first-person narrative perspective supposedly facilitates
the character (Kaufman and Libby, 2012). perceived similarity only seems to affect
persuasive outcomes when the character is in a setting familiar to the reader (De Graaf et al., 2016). Furthermore, participants are more likely to identify with a character in a situation familiar to them, regardless of narrative perspective (De Graaf et al., 2011). This may explain why previous studies using a student protagonist and a student sample found a significant effect of narrative perspective; they focused on portraying a topic (voting in a US primary election) likely to be unfamiliar to the participants (Kaufman & Libby, 2012). Thus, the perceived similarity between the reader and the character may have suppressed the effect of the manipulation on identification, and the familiarity of the university situation may have facilitated this.
Why did the Narratives Predict Stronger Emotions than the Control Text?
Despite the narrative perspective conditions not predicting differences in
identification, the narratives still predicted stronger emotions than the control condition. This could be due to transportation, an immersive process similar to identification without the focus on a specific character. It entails a complete immersion in the narrative that includes heightened emotional reactions (Murphy et al., 2011). Though identification and
transportation are conceptually linked, they have been empirically differentiated and independently manipulated, thus the narratives could have elicited stronger transportation than the control text, but not stronger identification (Tal-Or & Cohen, 2010). Transportation is facilitated by the imagery of a story plot (van Laer et al., 2014), so the added contextual detail and plot development in the two narrative conditions may have emotionally engaged participants more than the control text. Future narrative perspective research should isolate the effects of identification by measuring and controlling for other dimensions of narrative engagement such as transportation and perceived similarity (De Graaf, 2014).
Why did These Emotional Responses not Predict Persuasive Outcomes?
Possibly, the stronger fear and sadness predicted by the first- and third-person narrative perspective conditions did not consistently predict stronger attitude or behavioural intention than the control condition because the students already have their own coping methods. Some participants mentioned alternatives in their feedback. For example;
“Maybe doing sports is an interesting perspective to take in regards to coping methods, in my opinion a savior during my studies and exam periods.”
“I feel like there are more coping [methods] than meditation, such as sports, taking a break, talking to people in all sorts of life situations.”
Thus, stronger emotions may have motivated the participants to engage in stress reducing behaviour – just not MM. In line with the extended parallel processing model, response-efficacy determines whether a reader adopts a behaviour proposed in a fear appeal; theory and empirical research suggests the same is true for other types of emotional appeal (Dillard & Nabi, 2006; Lewis, Watson & White, 2013; Witte & Allen, 2000). Narratives are interpreted in line with a reader’s existing knowledge (van Laer et al., 2014). Therefore, if participants already have coping methods with a high response-efficacy then heightened emotions may direct them towards these behaviours rather than a new, untested alternative. This makes sense considering other narrative studies that have succeeded in eliciting persuasive outcomes; when presented with the threat of HPV, the target behaviour of
vaccination may seem to be the only efficacious response possible, or when presented with a text about esoteric environmental issues participants are likely to be unaware of any way to help beyond the methods suggested (Lu, 2015; Nan et al., 2015). These target responses have
no competition, whereas mindfulness meditation had to compete with other responses such as exercise and talking to friends.
One limitation of this study is the focus on discrete emotions when emotion was not manipulated. The three discrete emotions expected to be elicited by the texts all have unique action tendencies; however, these action tendencies were all expected to have the same positive effect on attitude and behavioural intention, and so it was deemed suitable to not have separate hypotheses for each emotion. This proved useful as one exception emerged that defied the trend of stronger emotions not predicting persuasive outcomes. The third-person narrative perspective condition did predict stronger fear than the control condition, and this fear predicted weaker attitudes and behavioural intention. Because this pattern was present in both the attitude and behavioural intention analyses, this may not be merely a chance
occurrence. This was unexpected, as no known prior research found an effect in the third-person condition, but not the first-third-person. Both the first-third-person and third-third-person narrative perspective conditions predicted stronger fear than the control condition, but only the third-person narrative perspective showed this fear mediating an indirect effect between narrative perspective and attitude and behavioural intention.
Therefore, the fact that this study’s results contradict previous findings suggests that the success of the manipulation is dependent on other factors; until we explore these factors further, it is uncertain how generalizable these findings are. Because of this contradiction, future research should investigate which factors cause the manipulation to succeed in some studies but not in others, testing moderators likely to either supress or facilitate the efficacy of perspective manipulation. Specifically, perceived similarity of the character to the reader (e.g. a student participant reading about either a student or a non-student character) and setting familiarity (e.g. set on a university campus or in a medical centre) provide promising avenues
of enquiry. Furthermore, different kinds of perceived similarity may have different effects on the manipulation of narrative perspective. Though Murphy et al. (2013) argue that physical similarity (i.e. race, age, sex) may be vital for grabbing the audience’s attention, Cohen (2006) argues that demographic similarity is far less potent a predictor of identification than psychological similarity. However, demographic similarities may promote identification when they pertain to the main character’s goals (Hoeken et al., 2016); so, in this context, the protagonist’s student status may have guided identification, as this demographic feature was integral to the goal of passing an exam. Alternatively, the demographic information may have been less important than characters similarity in attitude or psychology (Hoeken et al., 2016). Clearly there is a complex interplay between identification, transportation, perceived
similarity and setting familiarity, and future research should examine them together to explore these interactions.
The results of this study serve as a warning to message designers about the
effectiveness of manipulating narrative perspective. Though past evidence strongly suggests that a first-person perspective is more effective in eliciting persuasive outcomes than a third-person perspective, this study suggests that such an effect may be dependent on other factors. Thus, use of the first-person perspective should not be relied upon by message designers looking to improve their narratives until we understand the conditions under which it is effective.
In conclusion, this study provided further insight into the complexities of narrative persuasion. The perspective from which a narrative about student stress was told did not have a direct effect on attitudes, intentions or behaviours, nor did it have an indirect effect on attitudes or intentions through identification and emotion. Thus, this study runs counter to its predecessors, extending our understanding of narrative persuasion by showing that a
first-person perspective does not always increase the persuasiveness of a text. This insight led to the proposal of promising lines of enquiry that, if explored, could help message designers to craft more persuasive narratives in the future.
Footnotes
1. Based upon extant theory, my initial proposal was that the relationship between narrative perspective and behaviour would also be mediated by identification and emotion; however, because of the dichotomous outcome variable measurement, the scope of such an analysis was deemed to exceed both the level and parameters of a Master’s thesis. Therefore, individual logistic regressions were used to explore the relationship between narrative perspective and behaviour, excluding the mediators identification and emotion. This permitted exploration of the total effect, even if a more detailed mediation analysis was not possible.
References
Bae, H. S. (2008). Entertainment-education and recruitment of cornea donors: The role of emotion and issue involvement. Journal of Health Communication, 13(1), 20-36. Banerjee, S. C., & Greene, K. (2012). Role of transportation in the persuasion process:
Cognitive and affective responses to antidrug narratives. Journal of Health Communication, 17(5), 564-581.
Braddock, K., & Dillard, J. P. (2016). Meta-analytic evidence for the persuasive effect of narratives on beliefs, attitudes, intentions, and behaviors. Communication
Monographs, 1-24.
Busselle, R., & Bilandzic, H. (2009). Measuring narrative engagement. Media Psychology, 12(4), 321-347.
Chadwick, A. E. (2015). Toward a theory of persuasive hope: Effects of cognitive appraisals, hope appeals, and hope in the context of climate change. Health Communication, 30(6), 598-611.
Chittaro, L., & Vianello, A. (2016). Evaluation of a mobile mindfulness app distributed through on-line stores: A 4-week study. International Journal of Human-Computer Studies, 86, 63-80.
Cohen, J. (2001). Defining identification: A theoretical look at the identification of audiences with media characters. Mass Communication & Society, 4(3), 245-264.
Cohen, J. (2006). Audience identification with media characters. Psychology of Entertainment, 13, 183-197.
Cohen, S., Kamarck, T., & Mermelstein, R. (1983). A global measure of perceived stress. Journal of Health and Social Behavior, 385-396.
Cohen, J., Tal‐Or, N., & Mazor‐Tregerman, M. (2015). The tempering effect of transportation: Exploring the effects of transportation and identification during exposure to controversial two‐sided narratives. Journal of Communication, 65(2), 237-258.
Cole, S. R. (1999). Assessment of differential item functioning in the Perceived Stress Scale-10. Journal of Epidemiology and Community Health, 53(5), 319-320.
Conner, M., & Sparks, P. (2005). Theory of planned behaviour and health behaviour. In M. Conner & P. Norman (Eds.), Predicting Health Behaviour: Research and Practice with Social Cognition Models (2nd ed., pp. 170-222). Maidenhead, UK: Open University Press
