The Role of Presumed Influence and Emotions on Audience Evaluation of the Credibility of Media Content and Behavioural Tendencies

Abstract

We propose that emotional responses to a media stimulus (e.g., a news article), in addition to presumed influence, can help explain one’s evaluation of the media stimulus. Our model was evaluated using data collected from 261 college students who read a news article discussing the harmful effects of multitasking. Structural equation modelling (SEM) analysis revealed that the participants’ emotional feelings (i.e., anger, guilt and happiness) after reading the message and perceptions of how professors and administrators’ perceived the researchers interviewed in the article directly predicted the evaluation of the news article. Presumed influence of the news article on neutral Americans and professors and administrators on campus did not directly predict participants’ evaluation of the news article, the influence of which was mediated by anger. Message evaluation and select emotions explained 58 per cent and 26 per cent of the variance in supporting publication of the article and intentions to avoid multitasking, respectively. Results support the inclusion of emotional responses in news evaluation research.

Keywords

Presumed influence emotions message evaluation behavioural intentions structural equation modelling

How media consumers evaluate news or other media outlets is one of the most important and intriguing areas for media professionals and academic researchers. The model of presumed influence (Gunther & Storey, 2003) can be applied in such research. The model postulates that media consumers (a) perceive that the media may influence other media consumers and then (b) form attitudes towards (i.e., evaluate) the specific media stimulus or behaviours in response to the perceived influence. Gunther and Storey state that media consumers engage in defensive processing of media content because they perceive that those ‘less well-informed and potentially vulnerable audiences’ may be influenced by misleading information. The model holds significance in shifting the traditional direct media effects, which often may not be observed (Gunther & Storey, 2003), to analyzing the indirect effects of media content on attitudes and behaviour. However, the effect size or amount of variance explained in attitudes and behaviours is rather modest. In Gunther and Storey’s study (2003), presumed influence only explained about 4 per cent of the variance in attitudes towards health workers. Indeed in some studies, presumed influence produced inconsistent results (e.g., Gunther & Storey, 2003; Park, 2005).

We believe that the inability of the presumed influence model to account for a larger amount of variance in one’s evaluation of a media stimulus is because it does not consider the emotional responses that co-occur with cognitive judgment of the media stimulus. Because ‘emotions are not simply a response to our environment but also influence perception of situations, information preference, and desire for action’ (Arpan & Nabi, 2011, p. 5), emotional responses may help predict and explain audiences’ evaluation of news. Indeed, the media industry often employs a variety of techniques to elicit emotional responses, to command readers’ attention or to enhance viewers’ enjoyment, for example, the use of sensationalistic materials and human interest stories (Kovach & Rosenstiel, 2007; Zillmann, 2000). Therefore, attention to emotional reactions may provide a more complete picture of news evaluation and credibility research (e.g., Arpan & Nabi, 2011; Kühne & Schemer, 2013).

The goal of the present investigation is twofold. First, integrating presumed influence and emotional responses and adopting a structural equation modelling (SEM) analysis, the present analysis will examine how media consumers’ perceptions of the influence of a news article and emotions are related to their evaluation of the credibility of the article. Second, we investigate how behavioural intentions might be conceptualized and incorporated. Admittedly, constructs similar to behavioural intentions were originally incorporated in the model of presumed influence, but few studies have examined the link between news evaluation and subsequent action tendencies (Arpan & Nabi, 2011 is an exception). Accordingly, we evaluate an expanded model of presumed influence (Figure 1) in the context of news evaluation.

An Expanded Model of Presumed Influence: Evaluating the Credibility of News Articles

We use news evaluation to refer to a reader’s favourable or unfavourable evaluation of a news story’s credibility. Attitudes are conventionally defined as one’s favourable and unfavourable evaluation of a stimulus or behaviour (Fishbein & Ajzen, 2010). Thus, the audience evaluation of the credibility of a news article can be treated as one’s attitude towards media stimulus in the original model of presumed influence. Our use of evaluation of news credibility is specified at the article level and does not refer to the general trust towards the media or general attitudes towards a media programme.

According to the expanded model (Figure 1), at the first stage, participants are exposed to a news article. Research has shown that various message frames and features (e.g., gain or loss, anger- or sadness-inducing message features) may elicit both cognitive judgements and emotional responses. At the cognitive level, Gunther and Storey (2003) found that media exposure was positively related to the perceived influence of health workers. A number of scholars (Kim & Cameron, 2011; Lecheler, Schuck, & de Vreese, 2013) acknowledged the impact of message features (e.g., framing) on cognitive judgement. However, these scholars also proposed and found that various message features led to emotional responses. For example, Lecheler et al. (2013) revealed that positive and negative news frames induced positive and negative emotions, respectively.

Figure 1.

Source: Authors’ own.

Presumed Influence: Presumed influence is defined as a respondent’s subjective estimate of perceived effectiveness of a media stimulus on others’ attitudes and behaviours. Related to our research,¹ when a message describes issues and problems related to one particular social group, respondents of that group may judge how others’ opinions about their group are influenced by the media stimulus (i.e., perceived identity threat), such that others may perceive or may interact negatively with that social group. In this circumstance, which is common in news credibility research (e.g., Arpan & Nabi, 2011; Arpan & Raney, 2003; Gunther & Liebhart, 2006; Vallone, Ross, & Lepper, 1985), information that threatens participants’ positive image may elicit negative evaluations among the participants (Arpan & Nabi, 2011). Vallone et al. (1985) demonstrated that Israeli and Palestinian partisans, who held strong attitudes towards their own group, perceived the same, neutral news broadcasts about the Beirut Massacre in 1982 biased against their own groups. That is, a highly involved social group (i.e., partisans) prefers positive accounts of their own social group or their actions, and thus is unlikely to perceive neutral news articles credible. In the present study, participants may estimate how an article discussing the negative consequences of multitasking and the Millennials may influence other news readers’ (e.g., neutral Americans or professor/administrators) perceptions of the Millennials and multitasking. As a result, those participants who perceive a more negative influence of the article will evaluate the article more negatively. Thus, we proposed the following hypothesis:

H1: Participants who perceive a higher influence of a news article will negatively evaluate this news article’s credibility.

Emotional Responses to Media Stimulus

Emotions refer to the state of conscious feelings that are experienced by an individual, characterized by ‘physiological changes such as arousal’ (Baumeister, Vohs, DeWall, & Zhang, 2007). As emotions may shape our responses to the environment (e.g., a news article) and guide our future behaviours, they may help further explicate the concept of media consumers’ attitudes and evaluation of a media stimulus, and predict behavioural intentions—two major constructs in the model of presumed influence. We examined how emotions can be integrated in the model of presumed influence. First, the appraisal theory indicates that emotions can be induced when participants evaluate situations that are relevant to their goals (e.g., Scherer, Schorr, & Johnstone, 2001). Recent conceptualizations state that emotions have their own appraisal patterns and own unique consequences (e.g., Ellsworth & Scherer, 2003; Lazarus, 1991). Ellsworth and Scherer (2003) stated that different personal situations (e.g., relevance or coping potential) and contexts (e.g., novelty) may introduce different emotions and different levels of such emotions. Similarly, specific message features or events in a news articles may lead to participants to cognitively appraise the events or the articles and may evoke message-specific emotions (Nabi, 2003). For example, articles discussing behaviours violating one’s own moral standards or including a responsibility frame may lead to the feeling of guilt, whereas perceived injustice or an offense from others may cause anger. Second, after an emotion is induced, the action tendency associated with that emotion guides future information processing and influences one’s judgment of media stimulus. The behavioural tendencies associated with these emotions are varied and may depend on factors such as the type of emotions, one’s own personal characteristics, coping abilities and social consequences (Ellsworth & Scherer, 2003). Similarly, message-induced emotions may lead to different patterns of information processing or different behavioural intentions. For example, anger may be negatively associated with audience’s evaluations of credibility and positively associated with the desire to retaliate and search for more information (Arpan & Nabi, 2011). Kühne and Schemer (2013) found that message-induced anger led to the accessibility of information related to punishment and preference for punitive measures, and that sadness, on the other hand, increased accessibility of information related to help and remedial measures. We believe that message-induced emotions may have different patterns related to audience’s evaluations of media credibility because different emotions may lead to different consequences.

Because each emotion has its own appraisal patterns and may only be elicited when participants evaluate relevant message features, we focus on the emotions that are most relevant to our message. More specifically, the stimulus article in our study proposed the imposition of a mandatory quiet time requiring students to unplug from all electronic devices. This policy may be perceived as restricting students’ freedom to access electronic devices that are part of their daily activities, constituting an external offense inflicted upon the students and provoking their psychological reactance (Dillard & Shen, 2005).

Anger is defined as a form of defensive reactions to an external offense (Berkowitz & Harmon-Jones, 2004) and is a common form of psychological reactance (Dillard & Shen, 2005). When being exposed to a counter-attitudinal message or a message restricting one’s freedom, participants may experience anger (Arpan & Nabi, 2011; Dillard & Shen, 2005). That is, when their identity is threatened by negative perceptions from others, participants may react defensively towards the message that may influence others’ perceptions of them. For example, Arpan and Nabi (2011) found that exposure to an identity-threatening news story (vs. exposure to a neutral or irrelevant news story) resulted in a greater anger response, which was positively related to the perceptions of bias in the news article and source derogation.

As a consequence, early research found that individuals who experienced anger often felt like ‘lashing out’, ‘kicking’, ‘yelling’ and ‘wanting to do something’ (Roseman, West, & Swartz, 1994). In the media context, the consequences of anger can manifest in a few ways. As an approach emotion, anger may lead individuals to attack, derogate and criticize a reporter or the publication and may even retaliate by demanding retraction or not supporting publication of the news article or the said media content. For example, Arpan and Nabi (2011) found that a higher level of anger, as a result of a perceived injustice or offense, was associated with more negative evaluation (i.e., attitude towards) of a news article and the reporter of the article. That is, participants with a higher level of anger (vs. lower) are more likely to blame the media for the biased information. Similarly, anger increases accessibility of and preference for punishment and retributive information and preference for punitive measures (Kühne & Schemer, 2013). Participants may also seek retaliation against the source that inflicted the offense on them. A number of other studies suggest that anger, either as part of the overall psychological reactance measure or as an individual construct, was negatively related to attitudes (e.g., evaluation of a message) and intentions (e.g., Dillard & Shen, 2005). Consistent with the action tendency associated with anger and previous research in anger and message evaluation, we hypothesized that anger mediates the relationship between presumed influence and evaluation of a news article:

H2: (a) Participants’ perceptions of presumed influence of the article on others are positively correlated with their anger responses, which are associated with (b) more negative evaluation of a news article’s credibility and (c) less support for publication (i.e., a punitive action).

Guilt refers to ‘an individual unpleasant emotional state associated with possible objections to his or her action’, probably due to his or her violations of moral norms (Baumeister, Stillwell, & Heatherton, 1994) or not fulfilling the individual’s responsibility or obligations. When a message discusses moral values and one’s obligations or setting standards for appropriate behaviours, it may evoke moral emotions (e.g., guilt) among the readers. The more common behavioural consequences for feeling guilty are the urge to undo the action, apologize or be forgiven (Baumeister et al., 1994; Roseman et al., 1994) such that individuals may want to make amends and change behaviours accordingly. As such, guilt was found to predict behavioural intentions. For example, Harth, Leach and Kessler (2013) found that guilt predicted intent to repair damage. Previous literature indicates that guilt may also impact message evaluation or persuasion outcomes. Hullett (2004) found that participants who felt more guilt (vs. less) formed more positive attitudes of getting tested for herpes. The probable reason is that individuals who felt guilty were those who agreed with the content in persuasive messages and hence they rated the message more credible and persuasive than those who did not feel guilty. Thus, in the case of multitasking, individuals who feel guilt may be those who understand and agree with the news article and as a consequence, may be more likely to form intentions to change their multitasking behaviour.

H3: Higher levels of guilt are positively associated with (a) evaluation of credibility and (b) intentions to avoid multitasking.

Individuals experience happiness when they achieve desired outcomes (Lazarus, 1991) or when they encounter events that are entertaining (Zillmann, 2000) or pleasant (Smith & Ellsworth, 1985). Media entertainment and news articles containing congenial information are consistent with one’s goal state to be happy. A news article discussing issues related to multitasking and call for restriction measures, on the other hand, is not pleasant or may interfere with one’s intent to remain control of their own behaviours and thus, may cause less happiness among the participants. As for the consequences, happiness may influence one’s processing of persuasive messages. Previous theorizing based on the elaboration likelihood model indicates that individuals in a happy state may be more likely to process a message in a low-effort manner (Petty & Cacioppo, 1986), perhaps because happiness may interfere with one’s cognitive capacity (Mackie & Worth, 1989) and/or happy individuals would like to maintain their happy emotional state. Similarly, according to Chaiken (1987), emotions may be used as heuristics such as ‘If I feel good, I must like it.’ More recent research (Briñol, Petty, & Barden, 2007) indicates that happiness showed a main effect on positive evaluation of a message either because they were confident in their thoughts or used happiness as a heuristic. In our study, happiness is one dimension of emotional responses to the news article. Consistent with Chaiken (1987) and Briñol et al. (2007), we suspect that individuals who have more favourable thoughts about multitasking and the content after reading the articles are more likely to be happy. Because being in a happy state will either make low-need-for-cognition individuals evaluate the message more favourably in a heuristic manner or make high-need-for-cognition individuals have more confidence in the positive thoughts they remembered, we expect an overall pattern of a positive relationship between happiness and evaluation of the news article. Furthermore, happiness has also been found to be associated with success and adoption of recommended behaviours (Lyubomirsky, King, & Diener, 2005). Thus, we forward the following hypothesis:

H4: Happiness is associated with (a) favourable evaluations of a news article’s credibility and (b) higher intentions to support circulation of the news article.

Consistent with the expanded presumed influence model, we propose Hypothesis 5.

H5: Participants’ higher levels of evaluation of credibility are positively associated with (a) stronger support for publication of the article and (b) higher intentions to avoid multitasking.

Finally, we compare the expanded model with the original model of presumed influence. More specifically related to the original model, presumed influence will predict the evaluation of a news article’s credibility, which in turn will predict supporting publication of the article and intentions to avoid multitasking. Although we suspect that the original model will also fit the data, it will explain less variance in key dependent variables (i.e., news credibility and action tendency) than will the integrative model. Formally, we propose hypothesis 6.

H6: The expanded model with emotional responses will explain more variance in the key dependent variables (e.g., credibility and behavioural intentions) than will the original model of presumed influence.

Method

Data for the present analysis were collected through an online survey. Three email invitations with the link to the survey were sent to 1,050 randomly selected participants in a large university in the United States. The survey procedure was approved by the university’s human subjects committee. Participants were informed that their participation would enter their names for a drawing of 10 cash awards of $20 each. They were instructed to read a news article related to multitasking and then complete a questionnaire. The online survey remained open for total eight days. A total of 270 students participated in the survey outside of class time. After deleting cases with incomplete data, we retained 261 cases (24.5 per cent response rate). The sample was almost evenly split between males (48.9 per cent) and females (51.1 per cent). The average age was 20.5 (SD = 1.9). The sample consisted of 8.8 per cent Asians, 5.4 per cent Blacks, 4.8 per cent Hispanics, 72.6 per cent White and 5.2 per cent other racial backgrounds. Participants were from various majors on campus.

The use of a student sample is defensible in this case because (a) the topic under investigation was relevant to students and more importantly, (b) the major purpose of the present study was to examine the theoretical relationships among various cognitive judgements, emotional responses and behavioural tendencies instead of generalizing the results to a given population. That is, internal validity is more important than external validity in our case.

Stimulus Article

Before completing a questionnaire with various measures, each participant read an article that discussed the negative consequences of multitasking (e.g., not contribute to learning or solving complicated tasks in the future) and proposed imposing a quiet time when students were required to unplug from all electronic devices. Although this article discussed the negative consequences of multitasking, it also recognized Millennials’ exceptional abilities in multitasking. This article was roughly balanced and polite in pointing out the problems associated with multitasking.

Students were informed that the article was either selected from USA Today or a school newspaper in a different state. A manipulation check was conducted. Participants were asked about the perceived reach of the news article: ‘How many readers do you think this article can reach?’ Participants responded on a scale ranging from 1 (hardly any) to 11 (millions). An independent-samples t test showed that the reach manipulation was successful with M_USAToday = 7.89 (SD = 2.61) vs. M_schoolpaper = 5.19 (SD = 2.57), t(259) = 7.78, p < 0.001.

Measures

Presumed Influence on Neutral Americans and on Professors on Campus: Similar to the measure used by Vallone et al. (1985), presumed influence questions in our study focused on the consequences of presumed influence. We initially differentiated two groups of others: general neutral Americans and professors and administrators on campus who have the authority to enforce a quiet time. Four questions were asked for each group of the referents. These questions were created to reflect the context of our experimental manipulation. For example, ‘After reading this article, how will neutral Americans think about the Millennials?’, ‘…, how will neutral Americans think about multitasking?’, ‘…, how will neutral Americans think about the professors who were surveyed in the article?’ and ‘…, how will neutral Americans think about the researchers who conducted the survey?’. Responses ranged from 1 (very negative) to 7 (very positive). The same four questions were repeated after replacing ‘neutral Americans’ with ‘professors and administrators on campus’. Values ranged from 1 (very negative) to 7 (very positive).

However, factor analysis showed that these eight items did not load on their respective factors based on social distance or importance in policy decisions, that is, presumed influence on neutral Americans versus professors and administrators on campus. Two different factors showed up: neutral Americans’ and professors’ perceptions of Millennials and multitasking (Cronbach α = 0.69), and professors’ perceptions of the professors and researchers who were interviewed in the article (Cronbach α = 0.73).

Emotional Responses: Participants were asked how they felt after reading the news article. They responded to a list of eight emotional descriptors (e.g., Arpan & Nabi, 2011; Dillard & Shen, 2005) based on a scale ranging from 1 (not at all) to 7 (very much). Among these emotional descriptors, angry, irritated and annoyed formed an ‘anger’ scale (Dillard & Shen, 2005; α = 0.90). Guilty, shameful and regret formed the guilt measure (α = 0.94). Pleased and happy assessed the overall happiness dimension (α = 0.93). These items, together with items for all other measures, were submitted to an overall confirmatory factor analysis (reported in the results section) and were found to load on their respective factors.

Evaluation of the News Article’s Credibility: Participants were asked to indicate their overall assessment of the news article they read. This measure was based on five statements: ‘It was accurate/fair/meaningful/newsworthy/unbiased.’ These items were selected from Gaziano and McGrath (1986) and Meyer (1988). Values ranged from 1 (strongly disagree) to 7 (strongly agree) (Cronbach α = 0.84).

Supporting Publication and Circulation of the News Article: One important variable in the integrative model of media response and third-person effects research is whether media consumers support publication or restriction of the media content or the ideas contained in the media content. We constructed a scale based on three items: ‘Professors and administrators on campus should receive a copy of this news article’, ‘This news article should be published in Democrat and Chronicle (note: a local newspaper)’ and ‘Our university should adopt a policy of mandatory quiet times’ (Cronbach α = 0.78).

Intentions to Avoid Multitasking: The expanded model also states that actions, as a result of reading a news article, should not be confined to supporting publication or placing a restriction. Indeed, participants may be persuaded to adopt the action advocated in the news article. Thus, we included an additional measure to tackle participants’ intentions to avoid multitasking in the future: ‘I intend to avoid multitasking behaviours in the future’, ‘I plan to avoid multitasking behaviours in the future’ and ‘I will continue to multitask in the future (recoded).’ Alpha coefficient was 0.87.

Several questions related to news slant were included in the initial questionnaire. However, these questions formed a scale with low reliability (α = 0.50) and can be considered as a subcomponent of the overall news evaluation. Thus, we did not include these questions in further analysis. Demographic questions and a question related to demand characteristics were included at the end. No participant guessed the true purpose of the study. No other questions were included.

Results

Pearson correlations, means, and standard deviations of the variables, based on observed variable scores, are presented in Table 1. The present study sought to evaluate the efficacy of the expanded model of presumed influence and more specifically, six hypotheses based on the expanded model and the specific case of multitasking among college students. We used a two-step approach in SEM to evaluate the last six hypotheses. Because the reach of the newspapers was not a significant predictor of any of the cognitive and emotional responses, we decided to drop the analysis of effects of message features in the SEM analysis and collapsed data from all conditions. The two-step approach first evaluated a measurement model or the construct validity of various items used to measure the variables to be studied. Because the data did not meet the multivariate normality assumption for SEM analysis (the normalized Mardia’s coefficient = 32.1), the robust maximum likelihood estimate in EQS 6, based on the covariance matrix, was used to correct standard errors. Fit indices for the measurement model (Step 1) showed that the measurement model fit the data well, providing evidence for the construct validity of scales, Satorra-Bentler χ² (209, N = 261) = 376.2, p < 0.001, root mean square error of approximation (RMSEA) = 0.055, 90 per cent confidence interval (CI) of RMSEA (0.046, 0.064), comparative fit index (CFI) = 0.94 and standardized root mean square residue (SRMR) = 0.06. Fit statistics showed that the overall model, which included both the measurement and structural models, fit the data well. Satorra-Bentler χ² (200, N = 261) = 506.0, p < 0.001, root mean RMSEA = 0.061, 90 per cent CI of RMSEA (0.053, 0.068], CFI = 0.93 and SRMR = 0.07.

Table 1.

Pearson Correlations, Mean, and Standard Deviations of the Variables Used in the Study based on Observed Variable Scores (N = 261)

Variable	1	2	3	4	5	6	7	8
1. Presumed influence on Others	–
2. Presumed influence on Professors’ Perceptions of Information Source	0.10	–
3. Anger	–0.17	0.05	–
4. Happiness	0.28	0.12	0.02	–
5. Guilt	–0.04	0.12	0.32	0.44	–
6. News Credibility	0.22	0.18	–0.28	0.52	0.25	–
7. Support for Publication	0.28	0.28	–0.24	0.50	0.26	0.61	–
8. Intentions to Avoid Multitasking	–0.06	0.03	–0.04	0.22	0.25	0.34	0.43	–
Mean	3.72	4.41	3.16	3.06	2.36	4.08	3.22	2.60
SD	0.85	0.87	1.58	1.43	1.49	1.17	1.43	1.47

Source: Authors’ own.

Note: All scales were based on a 1–7 point scale, with 1 indicating less or negative and 7 indicating more or positive.

* Pearson correlations > 0.12 were significant (p < 0.05).

As noted in the method section, factor analysis showed that two types of presumed influence emerged: (a) perceptions of multitasking among both neutral Americans at large and professors on campus and (b) professors’ perceptions and evaluations of researchers and professors that were interviewed in the article. Regarding Hypothesis 1, presumed influence on neutral Americans at large and professors on campus did not directly predict participants’ evaluation of the news article’s credibility (β = 0.06, ns). However, when participants assumed that professors on campus would evaluate the professors and researchers positively were more likely to perceive the news to be credible (β = 0.15, p < 0.05). Thus, Hypothesis 1 was partially supported.

Hypothesis 2 examined the mediating role of anger. Figure 2 shows that presumed influence on neutral Americans and professors on campus positively predicted participants’ anger responses, which in turn was negatively related to readers’ evaluation of news credibility (β = –0.30, p < 0.001), lending support for Hypothesis 2(a) and 2(b). It should be acknowledged that anger did not directly predict support for publication (i.e., Hypothesis 2(c)), indicating that anger was only directly related to evaluation of the article’s credibility but not to action tendency in this case. The indirect effect was 0.19, p < 0.001.

For Hypothesis 3(a), guilt positively predicted one’s evaluation of news credibility, indicating that those who experienced more moral emotions, perhaps due to agreeing the article and recognizing one’s own failures, perceived a higher level of a news article’s credibility (β = 0.16, p < 0.001). For Hypothesis 3(b), guilt also directly predicted intentions to avoid multitasking (β = 0.22, p < 0.05). For Hypothesis 4(a) and 4(b), happiness was positively associated with the news article’s credibility (β = 0.46, p < 0.001) and was also a direct predictor of support for publication (β = 0.19, p < 0.05).

Figure 2.

All paths shown above were standardized coefficients: Solid lines (vs. dashed) indicate significant paths (p < 0.05, two-tailed). Satorra-Bentler χ²(200, N = 261) = 506.0, p < 0.001, root mean square error of approximation (RMSEA) = 0.061, 90 per cent CI of RMSEA (0.053, 0.068), comparative fit index = 0.93, and standardized root mean square residual (RMR) = 0.07. Standardized factor loadings for measurement items ranged from 0.52 to 0.98 and are not shown above. We specified the correlations among all exogenous variables. The relationships among the three emotional responses: r = 0.06 between happiness and anger, r = 0.42 between happiness and guilt, r = 0.27 between anger and guilt. * Presumed influence: Neutral Americans’ and Professors on campus’ perceptions of the Millennials and multitasking. ** Presumed influence: Professors on campus’ perceptions of professors and researchers interviewed in the news article.

For Hypothesis 5(a/b), the evaluation of the credibility of a news article directly and positively predicted intentions to avoid multitasking (β = 0.39, p < 0.001) and support for publication (β = 0.61, p < 0.001).

Overall, readers’ evaluation of an article’s credibility and select emotions explained 58 per cent and 26 per cent of the variance in supporting publication and intent to avoid multitasking, respectively.

Hypothesis 6 stated that the expanded model was a better model than the original model of presumed influence. Because the expanded model involved more variables than did the original model, model comparison cannot be conducted by examining the significance of the χ² difference between the models and the degrees of freedom or relying on the differences between Akaike information criterion. Thus, in this case, the amount of variance explained in the immediate dependent variable (i.e., evaluation of news credibility) is used for comparison. A two-step SEM analysis, including variables related to the model of presumed influence, was conducted. The final model (Figure 3) fit the data well. Satorra-Bentler χ² (114, N = 261) = 232.7, p < 0.001, RMSEA = 0.064, 90 per cent CI of RMSEA (0.053, 0.076), CFI = 0.92 and SRMR = 0.08. The amounts of explained variance in perceived credibility (15 per cent), support for publication (55 per cent) and intentions to avoid multitasking (21 per cent) were lower compared to those in the expanded model. Particularly, the original model explained 15 per cent of the variance in perceived credibility, compared to 51 per cent in the expanded model. Thus, evidence indicates that the expanded model can explain the evaluation of media content better than the original model of presumed influence.²

Figure 3.

Source: Authors’ own.

Discussion

The present study proposed that the inclusion of emotional responses may help to explain better the audience evaluations of news articles and situated emotional reactions within the model of influence of presumed influence. We will first discuss the importance of emotional reactions in the model of presumed influence and will then discuss the specific variables in Figure 1.

First, our comparison showed that the amount of explained variance in the immediate dependent variable (i.e., news credibility) was improved from 15 per cent in the original model to 51 per cent in the expanded model. The original model was problematic because traditionally, presumed influence only explains a small amount of variance (e.g., Gunther & Storey, 2003), meaning a large amount of variance is left unexplained. Furthermore, scholars in SEM or regression analysis caution the possibility of under-specifying regression models (e.g., Shadish, Cook, & Campbell, 2003; Tomarken & Waller, 2003). Under-specifying regression models put the statistical validity of the results in question. That is, the significant relationship between presumed influence on multitasking and perceived credibility could be spurious when other important predictors are not controlled for. As our study shows, the presumed influence in the expanded model became nonsignificant when emotional responses were included. Admittedly, all models are under-specified and may omit some unknown, yet important variables. That is, the expanded model with emotional responses is probably under-specified as well. Future research should continue to explore and consider additional, theoretically relevant variables on a parsimonious basis. However, as of now, it offers a better account of news evaluation compared to the original model of presumed influence.

Second, it is interesting to note that although the presumed influence on others did not directly predict millennial students’ evaluation of news credibility, it indirectly predicted students’ evaluation through anger. That is, presumed influence led to anger responses, which then predicted students’ evaluation of news credibility. Consistent with the literature (Arpan & Nabi, 2011; Dillard & Shen, 2005), our result showed an important pathway for the role of presumed influence on message evaluation, which should also be included in the expanded model of presumed influence. These results also showed that news evaluation was predicted by the perceptions of how professors on campus evaluated the researchers and professors—those who were interviewed in the article. That is, in our investigation, how the professors and researchers in the news article (e.g., source of the information) were evaluated is directly related to perceived credibility. Participants might project their own assessment of the source of the information onto how professors and administrators on their campus judge the source of information. Thus, it is possible that the perceptions of whether the news article may influence professors on campus’ evaluation of the interviewees in the article positively predicted favourable evaluation of the news. It is possible that participants make assessment of a news article’s credibility based on projected others’ opinions. Overall, we might need to reconsider the construct of presumed influence related to an identity threat.

Third, the addition of emotional variables in media effects research is an important undertaking not only because they explain more variance one’s evaluation of a media stimulus, but also perhaps because most news articles may evoke some sort of emotional responses among news consumers (Kovach & Rosenstiel, 2007). These emotional responses may be associated with perceived credibility and future news selection or behaviours (Arpan & Nabi, 2011). Thus, it is of both theoretical and practical importance. However, evidence related to the association between emotions and intentions is somewhat mixed. Arpan and Nabi found that anger, but not other discrete emotions (e.g., guilt and sadness), predicted bias perception of news articles related to cheating on campus. In our investigation, anger, guilt and happiness were all associated with the evaluation of news credibility in a multitasking article. Dillard, Shen, and Vail (2007) found that anger and guilt were associated with perceived effectiveness of an alcohol-related public service announcement, and that anger and fear predicted perceived effectiveness for participants who read a public service announcement that recommended flossing. The evidence seems to indicate that the topic areas or issues being addressed in news articles or persuasive messages may determine which emotions are significant predictors. That is, consistent with the appraisal theories of emotions (e.g., Ellsworth & Scherer, 2002) and more recent media research (Kühne & Schemer, 2015), different message features may evoke different emotions, which may lead to different consequences. Furthermore, we found that guilt and regret-based moral emotions, after recognizing one’s failure to perform a good deed, may directly predict their future behavioural intentions to avoid multitasking, whereas happiness directly predicted intentions to support circulation of the article. These relationships indicate that perceived credibility does not fully mediate the relationship between these two classes of emotions and intentions. Taken together, results indicate that emotions may influence one’s evaluation of media content and may also directly influence one’s behavioural tendency. Granted, these relationships are largely consistent with theorizing in emotions and behavioural intentions (Baumeister et al., 2007) where emotions can be related to behavioural choices and yet each class of emotions may have its own behavioural consequences.

Similar to the propositions in the original presumed influence model, we postulate that perceived credibility may predict intentions to perform more than one behaviour because media consumers may consider the disposition of the article (e.g., restriction of circulation or future media use) and the actions that the article might advocate. Because previous theorizing indicates that evaluations (i.e., attitudes) (Fishbein & Ajzen, 2010) and emotions do not directly predict behaviours and are mediated by intentions, it is acceptable to use intentions as the outcome variables in media response. Future research may consider including behavioural measures as well.

Future Research Directions and Limitations

One limitation of our research was the use of a student sample. Although the use of a student sample is defensible for model testing purposes where internal validity of the results is more important and the experimental stimulus is closely related to them, future research may consider using other samples to provide evidence for external validity—whether the relationships specified in this model hold for other samples. It should be acknowledged that sample characteristics are just one of the several factors (e.g., settings, outcomes and treatments/operationalization) that may threaten external validity (Shadish et al. 2002). More studies need to be completed for the theory-building purposes and for generalization. Another limitation was that we could not ascertain the cause–effect relationship based on the survey data that were collected at one time shot because we cannot establish time order that is required for causal inference. Reliability coefficients for the presumed influence scales were close to the lower bound of satisfactory reliability (0.69 and 0.73). However, the use of SEM analysis corrects measure errors when estimating path coefficients and thus, a limited number of scales of approximately 0.70 coefficients should not pose problems in data analysis. The last limitation was related to the use of an experimental setting where the assessment of participants’ behavioural tendencies was limited to questions we included in the questionnaire. In a more natural setting, participants may respond with additional behavioural tendencies. For example, angry participants may yell or may search for more information related to multitasking.

It should also be noted that there are a few variations in using the expanded model (Figure 1). First, exposure to media can be explicated as exposure to different forms of media or a similar message with different content or news framing, or in the present investigation, exposure to one message. If it is a one-message scenario, the model will start from presumed influence and emotions to credibility and behavioural intentions. If it is the former scenario, there might be variations in participants’ responses in presumed influence and emotions, and thus, models start from media exposure (e.g., different variations or amount of exposure) to presumed influence and emotions. The former scenario will help us understand message variables may affect responses to news articles or media fare. Our experimental manipulation aimed to bring out some variations in cognitive judgments and emotional responses and did not result in any differences in any of the variables included in the structural equation models. Future research might want to consider stronger manipulations to examine whether message factors (e.g., emotion vs. no emotion, or various news frames) may introduce differences in cognitive judgments and emotional responses (e.g., Gross & Brewer, 2007). Lecheler et al. (2013) conceptualized that emotions are not just the result of message features (e.g., framing), but also the mediators of message effects and related to other important variables. More specifically, Lecheler et al. found that anger and enthusiasm, but not fear and contentment, mediated the relationship between news features and policy support. The combination of emotional responses and cognitive judgments as multiple mediators can be a venue for future research. Furthermore, the topic of an issue may also delineate the types of emotions and presumed influence to be included. For example, news on terrorism or protests may invoke strong anger responses or news of one’s favourite team winning an important match may induce happiness. Also, as shown in Figure 1, the general version of the expanded model did not specify the exact presumed influence, emotions and behaviours to be investigated, whereas Figure 2 was more specific. That is, the general model can be used as an overall theoretical framework, yet more detailed analysis of the topic under investigation should delineate the exact nature or types of the presumed influence, emotions and behaviours.

In conclusion, our study proposes an expanded model of presumed influence with both cognitive and emotional variables. The expanded model is consistent with a recent stream of news credibility and persuasion research that incorporates emotions (e.g., Arpan & Nabi, 2011; Dillard et al., 2007) and can account for a greater portion of variance explained in readers’ evaluation of news credibility in the original model of presumed influence. Our results also reveal an intricate relationship between two types of presumed influence and behavioural intentions. Certainly, more research is needed to further evaluate and explicate the concepts in the model and should consider employing different topics and samples. Finally, as previous research shows, the presumed influence model has been applied to a number of domains, including health and social marketing issues (e.g., Gunther & Storey, 2003), such that the expanded model may be applied to these domains as well and should not be limited to news credibility.

Footnotes

Notes

References

Arpan

L. M.

, & Nabi

R. L.

(2011). Exploring anger in the hostile media process: Effects on news preferences and source evaluation. Journalism & Mass Communication Quarterly, 88(1), 5–22.

Arpan

L. M.

, & Raney

A. A.

(2003). An experimental investigation of news source and the hostile media effect. Journalism & Mass Communication Quarterly, 80(2), 265–281.

Baumeister

R. F.

, Stillwell

A. M.

, & Heatherton

T. F.

(1994). Guilt: An interpersonal approach. Psychological Bulletin, 115(2), 243–267.

Baumeister

R. F.

, Vohs

K. D.

, DeWall

C. N.

, & Zhang

(2007). How emotion shapes behavior: Feedback, anticipation, and reflection, rather than direct causation. Personality and Social Psychology Review, 11(2), 167–203.

Berkowitz

, & Harmon-Jones

(2004). Toward an understanding of the determinants of anger. Emotion, 4(2), 107–130.

Briñol

, Petty

R. E.

, & Barden

(2007). Happiness versus sadness as a determinant of thought confidence in persuasion: A self-validation analysis. Journal of Personality and Social Psychology, 93(5), 711–727.

Chaiken

(1987). The heuristic model of persuasion. In Zanna

M. P.

, Olson

J. M.

, & Herman

C. P.

(Eds.), Social influence: The Ontario symposium (Vol. 5, pp. 3–39). Hillsdale, NJ: Erlbaum.

Dillard

J. P.

, & Shen

(2005). On the nature of reactance and its role in persuasive health communication. Communication Monographs, 72(2), 144–168.

Dillard

J. P.

, Shen

, & Vail

R. G.

(2007). Does perceived message effectiveness cause persuasion or vice versa? 17 consistent answers. Human Communication Research, 33(4), 467–488.

10.

Ellsworth

P. C.

, & Scherer

K. R.

(2003). Appraisal processes in emotion. In Davidson

R. J.

, Scherer

K. R.

, & Goldsmith

H. H.

(Eds.), Handbook of affective sciences (pp. 572–595). New York: Oxford University Press.

11.

Fishbein

, & Ajzen

(2010). Predicting and changing behavior: The reasoned action approach. New York: Psychology Press.

12.

Gaziano

, & McGrath

(1986). Measuring the concept of credibility. Journalism Quarterly, 63(3), 451–462.

13.

Gross

, & Brewer

P. R.

(2007). Sore losers: News frames, policy debates, and emotions. The Harvard International Journal of Press/Politics, 12(1), 122–133.

14.

Gunther

A. C.

, Bolt

, Borzekowski

D. L.

, Liebhart

J. L.

, & Dillard

J. P.

(2006). Presumed influence on peer norms: How mass media indirectly affect adolescent smoking. Journal of Communication, 56(1), 52–68.

15.

Gunther

A. C.

, & Storey

J. D.

(2003). The influence of presumed influence. Journal of Communication, 53(2), 199–215.

16.

Gunther

A. C.

, & Liebhart

J. L.

(2006). Broad reach or biased source? Decomposing the hostile media effect. Journal of Communication, 56(3), 449–466.

17.

Harth

N. S.

, Leach

C. W.

, & Kessler

(2013). Guilt, anger, and pride about in-group environmental behaviour: Different emotions predict distinct intentions. Journal of Environmental Psychology, 34, 18–26.

18.

Hullett

C. R.

(2004). Using functional theory to promote sexually transmitted disease (STD) testing: The impact of value-expressive messages and guilt. Communication Research, 31(4), 363–396.

19.

Kim

H. J.

, & Cameron

G. T.

(2011). Emotions matter in crisis: The role of anger and sadness in the publics’ response to crisis news framing and corporate crisis response. Communication Research, 38, 826–855.

20.

Kovach

, & Rosenstiel

(2007). The elements of journalism: What newspeople should know and the public should expect. New York: Three Rivers Press.

21.

Kühne

, & Schemer

(2013). The emotional effects of news frames on information processing and opinion formation. Communication Research, 42(3), 387–407.

22.

Lazarus

R. S.

(1991). Emotion and adaptation. New York: Oxford University Press.

23.

Lecheler

, Schuck

A. R. T.

, & de Vreese

C. H.

(2013). Dealing with feelings: Positive and negative discrete emotions as mediators of news framing effects. Communications: The European Journal of Communication Research, 38(2), 189–209.

24.

Lyubomirsky

, King

, & Diener

(2005). The benefits of frequent positive affect: Does happiness lead to success? Psychological Bulletin, 131(6), 803–855.

25.

Mackie

D. M.

, & Worth

L. T.

(1989). Processing deficits and the mediation of positive affect in persuasion. Journal of Personality and Social Psychology, 57(1), 27–40.

26.

Meyer

(1988). Defining and measuring credibility of newspapers: Developing an index. Journalism & Mass Communication Quarterly, 65(3), 567–574.

27.

Nabi

R. L.

(2003). Exploring the framing effects of emotion do discrete emotions differentially influence information accessibility, information seeking, and policy preference? Communication Research, 30(2), 224–247.

28.

Park

(2005). The influence of presumed media influence on women’s desire to be thin. Communication Research, 32(5), 594–614.

29.

Petty

R. E.

, & Cacioppo

J. T.

(1986). The elaboration likelihood model of persuasion. In Berkowitz

(Ed.), Advances in experimental social psychology (Vol. 19, pp. 123–205). New York: Academic Press.

30.

Roseman

I. J.

, Wiest

, & Swartz

T. S.

(1994). Phenomenology, behaviors, and goals differentiate discrete emotions. Journal of Personality and Social Psychology, 67(2), 206–221.

31.

Scherer

K. R.

, Schorr

, & Johnstone

(Eds.). (2001). Appraisal processes in emotion: Theory, methods, research. USA: Oxford University Press.

32.

Shadish

W. R.

, Cook

T. D.

, & Campbell

D. T.

(2002). Experimental and quasi-experimental designs for generalized causal inference. Boston, MA: Houghton Mifflin.

33.

Smith

C. A.

, & Ellsworth

P. C.

(1985). Patterns of cognitive appraisal in emotion. Journal of Personality and Social Psychology, 48(4), 813–838.

34.

Tomarken

A. J.

, & Waller

N. G.

(2003). Potential problems with ‘well fitting’ models. Journal of Abnormal Psychology, 112(4), 578–598.

35.

Vallone

R. P.

, Ross

, & Lepper

M. R.

(1985). The hostile media phenomenon: Biased perception and perceptions of media bias in coverage of the Beirut massacre. Journal of Personality and Social Psychology, 49(3), 577–585.

36.

Zillmann

(2000). The coming of media entertainment. In Zillmann

& Vorderer

(Eds.), Media entertainment: The psychology of its appeals (pp. 1–20). Mahwah, NJ: Lawrence Erlbaum.