Do conservative and anxious people have stronger emotional reactions in response to threatening visual frames?

Master Thesis

Do conservative and anxious people have stronger emotional reactions in response to

threatening visual frames?

Graduate School of Communication Research Master of Communication Science

Author:

Camilla Frericks

Student number:

11797355

Supervisor:

Dr. Bert Bakker

Date of submission:

28.06.2019

Acknowledgments

I would like to thank my supervisor Bert Bakker for believing in me and supporting me from the beginning on the go into the field and measure physiological responses. A special thank you goes to Bert Molenkamp for helping me program the experimental component and to retract the physiological data. Thank you to everyone who volunteered to take part in my experiment! I did not think it was possible to get 41 unpaid participants in four days. And especially thank you Pina Kakuschke you are the best research assistant I could have imagined. Finally, I want to thank everyone who gave me mental and emotional support during this project, especially my friends and family.

Abstract

The present study examines the emotional responses to threatening visual frames, as negative visuals dominate media news and yet there has not been much research investigating how they affect individuals. Further it was tested whether the emotional responses are stronger for conservatives and individuals with an anxiety proneness. Emotional responses were assessed by asking individuals (n = 40) to evaluate their emotions in response to threatening, neutral and positive images in their valence and arousal. The study included three different threatening frames to see whether people respond differently to different kinds of threat: personal harm, outgroup and evolutionary threat. In addition, physiological responses were assessed by measuring skin conductance, which is an indicator for physiological arousal. It was predicted that threatening images would trigger stronger emotional responses than neutral and positive images especially for conservatives and individuals with trait anxiety. Results suggest that individuals in general had significantly more negative and arousing emotions when viewing threatening images compared to neutral images. Emotional reactions to positive images were not significantly evaluated as less arousing, however, as the negativity bias predicts, which posits that people tend to respond more strongly and are more attentive to negative stimuli in their environment and in turn give these stimuli more weight. Further, conservative and anxious individuals did not have stronger emotional reactions compared to either liberal or non-anxious participants. Possible explanations for the insignificant moderation effects will be discussed.

Do conservative and anxious people have stronger emotional reactions in response to threatening visual frames?

The current study focusses on the effects of visual frames as these seem to be more effective than textual frames (Powell et al., 2015; Rodriguez & Dimitrova, 2011). Framing effects, the way in which issues are formulated, are among the most commonly studied types of political communication that shape public opinion (Coe, Canelo, Vue, Hibbing, & Nicholson, 2017; Slothuus & de Vreese, 2010). Visual frames are used in media to define (frame) issues. They are frequently used in media because images attract more attention and help to emphasize specific aspects of an issue, yet not much is known about how they affect individuals (Bucy & Grabe, 2007; Powell et al., 2015).

Mass media news are central and critical components of a representative democracy (Soroka, Nir, & Fournier, 2013) and they are dominated by negative stories (Soroka & McAdams, 2015). This is probably the result of the negativity bias, which posits that people tend to respond more strongly and are more attentive to negative stimuli in their environment and in turn give these stimuli more weight (Meffert, Chung, Joiner, Waks, & Garst, 2006; Rozin & Royzman, 2001; Soroka & McAdams, 2015). Evolutionary reasons for this may be that negative experiences can be more life threatening than positive events are beneficial (Hibbing et al., 2014; Rozin & Royzman, 2001). Although the negativity bias seems to be universal, it is being suggested that there are notable individual variations, as some people seem to be more sensitive to potential threats than others (Coe et al., 2017; Hibbing et al., 2014; A. Lang, Bradley, Sparks, & Lee, 2007; Norris, Larsen, Crawford, & Cacioppo, 2011). Several studies already suggest that individuals, who are more sensitive towards threats, tend to have more politically conservative beliefs (Hatemi et al., 2013; Hibbing, Smith, & Alford, 2014, Oxley et al., 2007). Furthermore, anxious individuals also seem to be biased towards negative information since they were more attracted to and persuaded by threatening frames than non-anxious people (e.g., Arceneaux, 2012; Gadarian & Albertson, 2014).

Moreover, researchers express a demand to better understand how images affect individuals (Bucy & Grabe, 2007; P. J. Lang, 2010; Powell et al., 2015; Yegiyan & A. Lang, 2010). The current research, therefore, investigates the emotional responses of different visual frames with the focus on threatening images, as these seem to elicit the strongest reactions. Emotional reactions are usually

measured using self-reports. More researchers, however, also want to examine physiological measures as these may tap more into the deep, involuntary reactions and are less susceptible to social

desirability biases (Coe et al., 2017). Therefore, this study does not solely rely on self-reported emotions but also takes physiological measures of skin conductance (SC) into account. Further, as there seems to be personal differences in how strong individuals respond to threatening frames, it will be examined whether these effects are stronger for conservative, or anxious individuals. The

experiment of this study was conducted at the Museum for Communication in Berlin to ensure a more diverse sample and realistic simulation than a laboratory setting.

Visual framing and negativity

Investigating framing effects and the negativity bias as influential types of political

communication requires both a profound theoretical framework with defined concepts, which will be presented in the following. Furthermore, it will be demonstrated how both conservatism and the trait anxiety may function as specific moderators which influence the relationship between the type of image and its emotional responses.

Visual framing

Framing effects occur when people consume news; they are described as the ability of news media to influence attitudes and behaviors by subtle differentiations in how issues are reported

(Lecheler & de Vreese, 2018). Thereby their actual effect is determined by the type of news frame, the news story, and personal differences (de Vreese, 2012). Newspapers always show images, and videos are shown by television broadcasted news (Soroka, Loewen, Fournier, & Rubenson, 2016). Therein lies a relevance to investigate the effects of visual frames, especially as these seem to be more effective than textual frames (Powell et al., 2015). Images may be a more powerful framing device as they are more attention-grabbing and demand less cognitive processing than text, which make them more easily to be accepted without question (Rodriguez & Dimitrova, 2011). Further, they are suggested to evoke more emotional responses and consequently influence attitudes and behaviors (e.g., Powell et al., 2015; Yegiyan & Lang, 2010). Researchers express a lack of research examining emotional responses of news frames in general (e.g., Aarøe, 2011; Lecheler & de Vreese, 2018),

especially in response to visual frames (e.g., Bucy & Grabe, 2007; Lang, 2010; Powell et al., 2015; Yegiyan & Lang, 2010).

Investigating media and its communicative aspects should include moderating variables, as these “modify the direction and/or strength of the effect of media use on a given outcome.”

(Valkenburg & Peter, 2013, p. 223). Not much is known about moderating effects, however. Lecheler and de Vreese (2018), for instance, suggest that there should be a greater focus on personality traits such as trait anxiety and political beliefs, instead of relying on demographic variables.

Moreover, there is an increasing interest in examining psycho-physiological measurements, such as SC (e.g., Coe et al., 2017) as these measures could increase validity (Lecheler & de Vreese, 2018). Additionally, with these measures findings from other news framing effects research can be replicated including objective measures (de Vreese, 2012). Coe and her colleagues (2017), for instance, were among the first to show that an individual’s physiological predisposition accounts for how they respond to textual media frames. The results of their study show that participants with a higher threat sensitivity, measured by greater skin conductance level (SCL), are more persuaded by threatening frames (Coe et al., 2017), implying that these frames are more effective for individuals with a higher threat sensitivity. What if other personal differences influence how individuals respond to threats? More precisely, whether some individuals have greater SCL than others in response to threatening stimuli.

To contribute to research, this study investigates the self-reported and physiological emotional reactions of different types of images and for whom these effects may be stronger. The focus thereby, will be on threatening images. The next section describes the negativity bias and explains SC in more detail.

Threat and the negativity bias

As stated in the introduction, there is evidence for an asymmetrical relationship between negative and positive information, arguing that negative events seem to have a stronger influence on

individuals’ attitudes than positive information (Soroka, 2006). This impact could be explained by the negativity bias. One definition of it is that “negative events are more salient, potent, dominant in combinations, and generally efficacious than positive events” (Rozin & Royzman, 2001, p. 297).

Dimensional theories of emotion link emotional processing and responses to activation in the appetitive and aversive motivational systems (Cacioppo & Gardner, 1999; Lang, Bradely, Sparks & Lee, 2007, Bradley & Lang, 1994; Lang, 2010; Bradley, Angelini & Sungkyoung, 2007). There is the assumption that in a natural state the appetitive system is activated. This could be due to an

evolutionary mechanism, which gets animals out to explore their environment. This system increases slowly and steadily in response to positive stimuli and is associated with approaching behavior. This is called the positivity offset (Bradley & Lang, 2007; Keene & Lang, 2016; Lang et al., 2007; Lee & Lang, 2009). In a threatening situation a quick response is needed. The aversive system thus kicks in, which begins at a lower point but increases much faster. This steeper increase in activation is called the negativity bias (Bradley & Lang, 2007; Lang, Bradley, Sparks & Lee, 2007, Cacioppo & Gardner, Keene & Lang). This quicker slope is crucial for survival as negative events can be life threatening in contrast to positive (Hibbing et al., 2014; Rozin & Royzman, 2001). Negative events are further suggested to have a stronger impact than positive events, as experiments show that negative stimuli produce decreases in heart-rate which is a sign of attention, and increases in SC, the indication for aversive activation (e.g., Soroka & McAdams, 2015). Further, people may have different emotional reactions depending on the type of threat they are exposed to. Their responses may be different depending on whether the stimuli meant a threat to survival during evolution like spiders and snakes (i.e., evolutionary threat), or a more recent threat such as a knife or gun (i.e., political threat)

(Schimmack, 2005).

In this dimensional approach, emotion is conceptualized in two dimensions: arousal (how intense the stimulus is) and valence (whether a stimulus is positive or negative), the latter also determines which motivational system will be activated (Lang et al., 2007). These emotional dimensions are used in this study. Furthermore, arousal can be assessed by physiological measures, which support the assessment of emotional responses as self-reports only reflect the surface of emotional responses to media content. Physiological measures can determine less conscious and implicit measure of emotional processes (Coe et al., 2017; Potter & Bolls, 2012). Variations in SC are considered to be good indicators of physiological arousal (e.g., Bradley, Angelini, & Lee, 2007; Potter & Bolls, 2012; Ravaja, 2004; Soroka, Nir, & Fournier, 2013; Soroka, Gidengil, Fournier, & Nir, 2016; Soroka &

McAdams, 2015). SC reflects the level of moisture released by the eccrine sweat glands, with which the skin conducts electricity better (Benedek & Kaernbach, 2010; Soroka & McAdams, 2015). In other words, this measure can show which stimuli elicit stronger reactions. In this study, SCL is analyzed which is a standard measure for studying emotional processing of media (Potter & Bolls, 2012).

According to the negativity bias, threatening stimuli should elicit negative and intense emotional reactions compared to moderate and positive stimuli, measured both by self-reported and physiologically.

The following hypotheses are thus formulated:

Hypothesis 1: Threatening images will elicit more self-reported arousal than neutral and positive

images.

Hypothesis 2: Threatening images will elicit more physiological arousal than neutral and positive

images.

Possible moderating effects of conservatism and trait anxiety

Moderating variables can explain more precisely how media really affect us, and for whom media effects may be stronger (Lecheler & de Vreese, 2018; Valkenburg & Peter, 2013). In this way they are more accurate determinants of media’s impacts in real-life.

According to Norris and her colleagues (2011), the most remarkable feature of the negativity bias is that there are strong variations between individuals, in the sense that some people react more strongly in response to negative stimuli than others (Hatemi et al., 2013; Hibbing et al., 2014; Lang et al., 2007; Norris et al., 2011). Yet, these differences may not be explained by traditional personality measures such as the Big Five (Norris et al., 2011). As personality is a very broad concept this research will only focus on two aspects of personality: political orientation and trait anxiety. The following paragraphs will justify why these aspects may function as moderators. Moreover, as stated earlier, Lecheler and de Vreese (2018) already demonstrated interest for these traits.

There already is a line of research suggesting that individuals with conservative views may be more sensitive towards negativity (e.g., Hatemi et al., 2013; Hibbing et al., 2014). There are various aspects of conservatism. Wilson (1973) who’s index of conservatism is used to assess conservatism in the current study defines this political orientation as a “resistance to change and the tendency to prefer

safe, traditional and conventional forms of institutions and behaviors” (Wilson, 1973, p. 4). Several researchers argue that the negativity bias is associated with conservatism (Hibbing, Smith & Alford, 2014; Dodd et al, 2012). An explanation could be that conservative beliefs may serve as a coping mechanism for individuals to handle their uncertainties and fears (Hatemi et al., 2013). Not only do conservative individuals report to have higher degrees of fear, they also tend to register greater physiological responses compared to more liberal individuals (Dodd et al., 2012; Hibbing et al., 2014; Oxley et al., 2008; Renshon, Lee, & Tingley, 2015). Therefore, the following moderation hypotheses are formulated:

Hypothesis 3: Conservative individuals will evaluate their emotions in response to threatening

images as more negative than liberal individuals.

Hypothesis 4: Conservative individuals will have more self-reported and physiological arousal

than liberal individuals in response to threatening images.

There could also be a connection between the negativity bias and anxiety as anxious individuals seem to have increased interest and desire to seek out more information when they are negative (Brader, 2005; Gadarian & Albertson, 2014; Marcus, College, Mackuen, & Louis, 1993; Marcus, Neuman, MacKuen, & Crigler, 2013). Studies examining the role of anxiety in response to negative political frames argue that experimentally induced anxiety moderates the pervasiveness of negative frames (e.g., Arceneaux, 2012; Gadarian & Albertson, 2014), and increases SCL (e.g., Renshon et al., 2015).

Anxiety is an ambivalent construct as it is conceptualized in different ways. The experiments presented above involved anxiety as a state, which is defined as the experience of intense dread that is internally derived and unrelated to an immediate threat (Grinker, 2013). Contrarily, trait anxiety, is conceptualized as a personal disposition, an anxiety proneness that the experience of anxiety will evolve in response to stressful situations (Spielberger, 2013). This trait therefore varies between individuals. If state anxiety produces biases towards threatening information, it seems interesting to further examine the role of trait anxiety as it is a personal disposition causing individuals to quickly feel anxious in response to threat. For instance, whether anxious citizens also show stronger emotional reactions in response to threatening stimuli than citizens who are not anxious:

Hypothesis 5: Anxious individuals will evaluate their emotions in response to threatening images

as more negative than non-anxious participants.

Hypothesis 6: Anxious individuals will have more self-reported and physiological arousal than

non-anxious individuals in response to threatening images.

Methods Sample

The experiment was conducted between the 23th _{and 26}th _{of April 2019 at the Museum for}

Communication in Berlin to have a more diverse sample than at a university in terms of political orientation and for better ecological validity. The participants were the visitors of the museum, which were recruited in the museum by the researcher and a trained research assistant. The current study had 40 valid responses (𝑀𝐴𝑔𝑒 = 38.25, SD = 13.51, 62,5% female). The age range was between 21 and 67 years. 40 percent of the sample was in their twenties, 12.5 in their thirties, 20 percent in their forties, 22.5 in their fifties and 5 percent were in their sixties. One participant was excluded because due to technical issues the sequence in which the images were shown was missing. As the experiment was conducted at a German museum the study was in German and therefore all respondents were German. Most of the participants (32.5 %) had graduated from high school (“Allgemeine Hochschulreife”). In general, the sample was well educated, 60 percent of the respondents had a university degree of which 10 percent had a master’s degree.

Stimulus Material

This study examines the emotional responses to five different visual frames. Three of them include threatening images to see whether people respond dissimilar to different kinds of threats. The threatening frames are personal harm, outgroup, and evolutionary threat. The next frame contains neutral and the last positive images. Most images were used from either the International Affective Picture System (IAPS) or from the Geneva Affective Picture Database (GAPED). The images of these databases have been tested upon emotional reactions and have already been used by many other studies (Bakker & Arceneaux, n.d.; Bakker, Arceneaux, & Schumacher, n.d.; Coe et al., 2017; A. Lang et al., 2007; P.J. Lang, Bradley, & Cuthbert, 1997; Norris et al., 2011). Two images are from the internet because the databases did not provide an image in the way that was wanted. The threatening

images were chosen on the condition that these would elicit a strong reactions of fear, instead of pity or disgust.

There are two personal harm images, one shows a masked man holding a knife (IAPS #6510). The other image shows a gun which is directed at the person viewing the image (IAPS #6230). The outgroup image shows several men trying to climb under a barbed wire (internet1_{). This image}

indicates refugees trying to trespass a border. The images for the evolutionary threating frame consist one image displaying a biting snake (IAPS #1120), a biting dog (IAPS #1300), a big spider (GAPED #136), and one image exhibiting a forest fire (internet2_).

The neutral images present a light bulb (GAPED #Neutral061), a spoon (IAPS #7004), a basket (IAPS #7010), a glass (IAPS #7035) and a lamp (IAPS #7175). These images have already been used by other studies (Bakker & Arceneaux, n.d.; Bakker et al., n.d.; Hibbing et al., 2014). Most of the positive image have also been used by the same studies. These demonstrate a sleeping baby tugged in a blanket (GAPED #Positive007), a white beach with turquoise water (GAPED

#Positive072 ), an image of a lake, in which the blue sky and some small clouds are reflected (GAPED #Positive064), a baby seal in the snow (GAPED #Positive079), and an images of three toddlers playing together (GAPED #Positive035).

Procedure of the experiment

The current experiment consisted of two parts. The first was a short Qualtrics survey which participants filled out on a tablet. The second part was the experiment that ran with the program Presentation and for which the respondents were connected to the skin conductance (SC) measuring device. As the participants of this experiment were random visitors at a museum and their

participation was not compensated with money, the experiment was designed to be as short as possible because the duration should not discourage their participation. The first part lasted around three minutes and the second part around ten minutes.

The Qualtrics survey started with a fact sheet explaining the purpose, procedure,

confidentiality of data and the possibility to stop the experiment at any time. Participants could only

1

http://media4.s-nbcnews.com/j/newscms/2015_35/1194356/150826-150826-world-hungary-razor-wire-climb-1057a-jpg-1049_5488f76d80c387a78bdf32c3e8701353.nbcnews-ux-2880-1000.jpg)

take part in the experiment if they agreed with the informed consent and were older than 18 years old. The survey assessed respondent’s political attitudes, trait anxiety, phobia, and political ideology (left-wing to right-(left-wing). There also was an attention test. Participants were asked to select answer C out of ABC if they are paying attention. Every respondent passed this test, which indicates that they followed the experiment attentively. Further, demographics were asked, like gender, age and highest level of education (

the

https://osf.io/gurvm/?view_only=9d926b49320a4e03afead7135598e3b8).

After completion of the survey, the researcher connected the participants to the EDA

measuring device, placing two electrodes on two fingers of their non-dominant hand (see Image 1 and 2) and noise-canceling headphones were put on. This part of the study was conducted on a separate computer, which enabled another participant to already start with the first part of the study on the tablet. At the beginning the baseline SCL was measured for 30 seconds. After this, the participants saw a blanc page with a small “x” at its center. This interstimulus interval was shown for five seconds before every stimulus image, in order to get participants’ attention and as a small resting phase between the images (see Image 3). The stimuli images were shown for eight seconds and in

randomized order. After every stimulus image the participants were asked how the image made them feel according the two emotional dimensions of emotions valence and arousal using the

Image 1 Image 2

Experimental set up Hand with electrodes

Image 3

Semantic set up

Measures

Independent variables. To test the hypotheses, a repeated measures ANOVA was done

(Potter & Bolls, 2012). In this analysis the independent variable are the levels, which are the five visual frames introduced earlier.

The initially intended visual frames were four different ones: political threat, evolutionary threat and neutral and positive. The political threat frame was supposed to contain the images that now are in the personal harm (image of knife and gun) and the outgroup (image of refugees) condition. As the valence and arousal ratings of the image with the refugees does not significantly correlate with the other two images, the political threat condition was subdivided into personal harm and outgroup (see Table 4 & 5 in Appendix B).

The other three frames were kept as initially intended. In the evolutionary threat condition not all ratings of valence significantly correlate. The valence ratings of the image with the forest fire do not correlate with the image of the dog, or with the image of the snake. The valence ratings of the forest fire significantly correlate with the other images in the evolutionary threat condition. the ratings of arousal of all the images in this condition significantly correlate (r > .32, p < .05). Therefore, no changes were made to the images of evolutionary threats.

Neutral and positive images were aligned to a selection of neutral and positive images used in other studies (e.g., Bakker & Arceneaux, n.d.; e.g., Bakker et al., n.d.).

For further analyses participants were also asked whether they were afraid of concepts that were also shown in the stimulus images in order to determine whether those individuals who for instance reported to be afraid of snakes also found the image of the snake more unpleasant and

arousing. An adjusted and abbreviated version of the Symptom Checklist 90 (SCL90), which was used by Hatemi and colleagues (2013) measure (Hatemi et al., 2013; Mattsson et al. 1969) was used. Respondents were asked whether they felt discomfort over the last 30 days due to seven different issues which were also presented in the images they viewed in the second part of the study. This scale was abbreviated and adjusted version. The issues presented were: being afraid of spiders (M = 1.88,

SD = 1.20), snakes (M = 1.65, SD = 1.12), weapons (M = 2.18, SD = 1.47), natural catastrophes

(M = 2.25, SD = 1.32), refugees (M = 1.30, SD = .69), dogs (M = 1.43, SD = .90), and violence (M = 2.48, SD = 1.36) (see Table 6 in Appendix B).

Dependent variables. There are 15 dependent variables as there are five different visual

frames and three types of emotional responses that were measured for each frame: self-reported valence, self-reported arousal and physiological arousal.

The two self-reported emotional responses measured by the SAM are valence, ranging from 1 (very pleasant feelings) to 9 (very unpleasant feelings), and arousal, ranging from 1 (very aroused) to 9 (not at all aroused). The ratings of arousal were recoded to range from 1 (not at all aroused) to 9 (very aroused). This way, higher values for valence indicates more negative and for arousal more arousing. There are missing values for the SAM evaluations because the presentation continued after ten seconds without waiting for a response. For further analyses listwise deletion was implemented. The

emotional responses per visual frame was created by dividing the average of the emotional responses of each frame and dividing it by the number of images in that frame.

Following previous research (e.g., Bakker & Arceneaux, n.d.; Bakker et al., n.d.; Braithwaite, Derrick, Jones, & Rowe, 2013), the average of the natural logarithm of SCL, which was measured in microSiemens, of the five seconds stimulus interval was subtracted (i.e., the baseline SCL) by the average of the natural logarithm of SCL of the eight seconds during which the stimulus images was shown:

DV_image1 = log(SCL_exposure1) – log(SCL_InterStimulusInterval1)

The same procedure as for the SAM evaluations was applied to obtain the frames: the average of the SCL responses of each image was divided by the number of images in the frame.

Table 1

Mean and standard deviations of dependent variables

Frame SAM - Valence SAM - Arousal Physiological Arousal

Personal harm 7.32 (1.35) 5.32 (1.97) .002 (.005) Out-group 6.70 (1.40) 5.14 (1.97) -.001 (.005) Evolutionary threat 6.96 (1.24) 4.86 (1.63) .001 (.003) Neutral 4.70 (.65) 1.95 (1.01) .001 (.003) Positive 2.2 (.79) 4.81 (1.77) .0001 (.003) Sample size 24 25 40

Moderating variables. Conservatism was measured with an adjusted version of the Wilson-Patterson

Conservatism-Liberalism Attitude Index (1968), which has often been used in the context of political communication (e.g., Bakker et al., n.d.; Hatemi et al., 2013; Smith, Oxley, Hibbing, Alford, & Hibbing, 2011; S Soroka et al., 2013). It consists of political statements, that reflect either more liberal or conservative views. Only six out of the 28 statements were used to ensure that the duration of the experiment stayed as short as possible. Further, instead of asking whether participants agreed with “yes” and “no”, participants’ attitudes were asked on a scale ranging from 1 (strongly oppose) to 5 (strongly support), to guarantee more variance. These statements used in this study are: homosexual marriage (M = 1.58, SD = 1.10), legalized abortion (M = 1.60, SD = .93), restrict legal immigration (M = 2.10, SD = .98) and more police cameras in public areas (M = 2.45, SD = 1.11), reduce income equality (M = 1.73, SD = .96), and raise taxes on the rich (M = 1.70, SD = .76) (see Table 2).

Response scores of items reflecting a strongly liberal view were reverse coded, so that higher values indicate a highly conservative view. For this all variables were recoded except the variables restrict

legal immigration and more police cameras in public areas.

Table 2

Mean and standard deviations of political attitudes

Political attitudes Means (Standard deviations) Homosexual marriage 1.58 (1,11) Social conservatism Legal abortion 1.60 (.93) Restrict immigration 2.10 (.98) Security conservatism Police cameras 2.45 (1.11) Reduce income inequality 1.73 (.96) Economic

conservatism _{Raise taxes on riches (.76)} 1.70

A Principal Components Analysis (PCA) using SPSS was done with varimax rotation to see whether these items measure the same construct. Barlett’s Test of Sphericity is significant

were adjusted as the PCA suggested: The items homosexual marriage and legalized abortion load on the first factor which was called social conservatism. The statements restrict legal immigration and

more police cameras in public areas, which loaded on the second factor were clustered and called

security conservatism. On the third factor the statement reduce income equality and raise taxes on the

rich loaded. These were added and called economic conservatism. For further analyses, a median split

was done for all three conservatism variables. All values above the median were coded as 1, indicating a conservative view while the rest was coded as 0 (see Table 3).

Table 3

Pearson’s correlations of the three conservatism scales and ideology. Cronbach’s alpha of the conservatism scales 1. Social conservatism 2. Security conservatism 3. Economic

conservatism 4. Ideology Cronbach's α

Mean (SD) 1. 1 .84 3.18 (1.89) 2. .18 1 .74 4.55 (1.87) 3. .40* .21 1 .61 3.53 (1.47) 4. .58** .37* .38* 1 3.95 (1.13) Note: ** p < .01, * p < .05

Another question asked for the political ideology of respondents in a more direct way. Respondents were asked to place themselves on a scale ranging from 1 (very left) to 10 (very right) according to their political views (M = 3.95, SD = 1.13). This was asked in a similar way as the European Values Survey and the European Election Social Surveys ask respondents about their political ideology. Most respondents identified themselves as more left-wing, 65 percent placed themselves on a value below the middle value (5). Most respondents placed themselves on value three (40 %). 25 percent placed themselves in the middle and 10 percent on the most right-wing value in this study which is value 6. This measure was used to assess whether there is a correlation between the direct and indirect measure of political orientation. Ideology significantly correlated with the three conservatism scales (see Table 3). There is a strong correlation between ideology and social conservatism was strong (r = .58). Whereas the strengths of the associations between ideology and security conservatism (r = .37) were of medium strength. This suggests that the political statements

either did not measure conservatism correctly or that respondents did not identify correctly on the ideology scale according to their political views.

The last moderating variable is trait anxiety. The are several forms of anxiety and multiple ways how to measure these. This study does not focus on a specific form of anxiety, rather a general anxious disposition. Sylvers and colleagues (2011) reviewed clinical research on fear and anxiety. They argue that asking whether a person experiences anxiety in response to a specific stimulus or situation is not sufficient, as this person may attribute their anxiety to the discomfort of the arousing stimulus. Instead they suggest using items that assess stimuli that are non-specific and more persistent. The brief measure for assessing generalized anxiety disorder (Spitzer, Kroenke, Williams, & Lowe, 2006) comply with these arguments and were therefore used to assess this trait. Four items out of the seven items were used. The question asked: “Over the last 2 weeks, how often have you been bothered by the following problems?”. Respondents answered on a scale ranging from 1 (not at all), 2 (several days), 3 (more than half the days) to 4 (nearly all the days). The statements were: Worrying too much about different things (M = 1.98, SD = .62), having trouble relaxing (M = 1.78, SD = .77), becoming easily annoyed or irritable (M = 1.85, SD = .48), being so restless that it is hard to sit still (M = 1.43,

SD = .59) (see Table 4).

A PCA using varimax rotation was conducted to examine whether these items measure the same construct. Barlett’s Test of Sphericity is significant χ2(6) = 36.89, p < .01. The PCA revealed that one factor explains 52.10 percent of the variance (Eigen value = 2.08) and two factors explain 78.80 percent (Eigen value = 1.07). The component and rotated component matrix both show that all items except becoming easily annoyed or irritable load on one factor with values above .5. Including all items still provides a good internal consistency (Cronbach’s α = .67), and therefore all items are kept in this scale. For further analyses a median split was done (Median = 7). All individuals with a higher score than seven were coded as 1 (40 %), while those with a score of seven or lower were coded as 0 (60 %) (see Table 7 in Appendix B).

Results Manipulation checks

To check whether the emotions in response to threatening images were indeed evaluated as more negative than the neutral and positive images, an ANOVA with repeated measures was

conducted. The independent variable is visual frames, which consist of five categories. The dependent variables are the self-assessment manikin (SAM) evaluations of valence for each frame (n = 24). Mauchly’s test indicated that the assumption of sphericity had not been violated, χ2(9) = 16.45 p = .06. The ANOVA analysis revealed a significant effect of frame on the SAM dimension of valence,

(F(4,92) = 112.11, p < .01). A post hoc test revealed that the emotions in response to the images in the threatening frames were evaluated as significantly more negative than the neutral and positive images (see Figure 1). This result indicates that the manipulation worked; threatening images were indeed experienced as more negative than neutral and positive images (see Table 8 in Appendix B for means and standard deviations of SAM for valence per frame).

Figure 1

Bar chart with means and 95% confidence intervals for self-reported valence (positive – negative) per frame

Testing of hypotheses

To test the hypotheses ANOVA’s with repeated measures were conducted, which is commonly used to analyze skin conductance level (SCL) (Potter & Bolls, 2012). The number of frames was the independent variable and self-reported or physiological arousal were the dependent variables. 0,00 2,00 4,00 6,00 8,00 10,00

Mea

n

Self-reproted valence

Main effects. With respect the first hypothesis stating that threatening images would elicit

more self-reported arousal than neutral and positive images, a significant main effect was found of frame on self-reported arousal, (F(2.99, 71.77) = 23.42, p < .01), (n = 25). Mauchly’s test indicated that the assumption of sphericity had been violated, χ2(9) = 16.83, p = .05, which means that the variances of the differences between levels are significantly different. Therefore, degrees of freedom were corrected using Greenhouse-Geisser estimates of sphericity (ε = .75). The post hoc test

demonstrated that the emotions in response to threatening images were significantly evaluated as more arousing than neutral images (see Figure 2). Threatening images however were not significantly evaluated as more arousing than positive images (see Table 8 in Appendix B for means and standard deviations of SAM for arousal per frame). This result does not fully support the negativity bias.

Figure 2

Bar chart with means and 95% confidence intervals for self-reported arousal (low to high) per frame

The second hypothesis predicted that threatening images would elicit more physiological arousal than neutral and positive images. Mauchly’s test indicated that the assumption of sphericity had been violated, χ2(9) = 32.39, p < .01. Therefore, degrees of freedom were corrected using Greenhouse-Geisser estimates of sphericity (ε = .74). A significant main effect of frame on

physiological arousal was found, (F(2.96,115.38) = 4.58, p < .01), (n = 40). The post hoc test indicated that the neutral images elicited significantly less physiological arousal than the threatening images in the personal harm frame, and in the evolutionary threat frame (see Figure 3). Physiological arousal in response to the threatening images in the outgroup frame was not significantly different to the neutral and positive images. The second hypothesis is partially supported firstly because only the emotions in

0 1 2 3 4 5 6 7 Mean

Self-reported arousal

response to neutral images were evaluated as significantly less arousing than the threatening images, whereas the positive images were not. Secondly, threatening images in only two frames significantly showed greater physiological arousal compared to the responses of the neutral images, but not compared to the positive images (see Table 8 in Appendix B for means and standard deviations of physiological arousal per frame).

Figure 3

Bar chart with means and 95% confidence intervals for physiological arousal (SCL) per frame

Moderation effects. To test for moderation effects a median split was done for all the

moderation variables. These were then analyzed as between-subjects factors in the repeated measures ANOVAs. Due to word limit, only the (marginally) significant moderation effects will be presented here.

The third hypothesis stated that conservative individuals will evaluate their emotions in response to threatening images as more negative than liberal participants. There was only a marginally

significant moderation effect for economic conservatism, (F(4,88) = 2.19, p = .076). Levene’s test was not significant. A post hoc test revealed that the emotions in response to threatening images were evaluated as significantly more negative compared to neutral and positive images (see Table 9 in Appendix B for means and standard deviations). Figure 4 shows that the economically conservative individuals (n = 10) evaluated emotions of the threatening images in the personal harm frame and in

-0,003 -0,002 -0,001 0 0,001 0,002 0,003 0,004 Mean

Physiological arousal

the evolutionary threat frame as more negative than economically liberals (n = 14). These differences in conservatism are not significant, however (p = .31). Hypothesis 3 is thus not supported.

Figure 4

Bar chart with means and 95% confidence intervals for valence and economic conservatism per frame

With respect the fourth hypothesis proposing that conservative individuals will have more self-reported and physiological arousal than liberal individuals in response to threatening images, two significant effects were found with security conservatism. The first is with self-reported arousal, (F(4, 92) = 5.14, p < .01). Levene’s test was not significant. The post hoc test showed that the emotions of neutral images were significantly evaluated as less arousing than threatening and positive images. Figure 5 shows that there is a significant difference in security conservatism, namely that conservative individuals significantly evaluated the emotions of the images in the personal harm frame as more arousing than the image in the outgroup frame. Figure 5 also shows secure conservative individuals (n = 9) evaluated the emotions of the images in the personal harm frame and in the evolutionary threat frame as more arousing than liberals (n = 16). Liberal individuals rated the image in the outgroup frame as more arousing than conservatives (see Table 9 in Appendix B for means and standard deviations). These differences in political orientation, however, were not significant (p = .41).

0 1 2 3 4 5 6 7 8 9 10 Mean

Self-reported valence and economic conservatism

Personal harm - Liberal Personal harm - Conservative Outgroup - Liberal

Outgroup - Conservative Evolutionary threat - Liberal Evolutionary threat - Conservative Neutral - Liberal

Neutral - Conservative Positive - Liberal

Figure 5

Bar chart with means and 95% confidence intervals for self-reported arousal and security conservatism per frame

For the physiological data there also is a significant moderation effect for security conservatism, (F(3.42, 129.81) = 3.74, p = .01). Levene’s test was not significant. Figure 6 demonstrates a significant difference in security conservatism, namely that conservatives (n = 18) significantly had less physiological arousal than liberals (n = 22) when viewing the image in the outgroup frame compared to when they saw in the images in the personal harm and the evolutionary threat frame. Further, secure conservatives had more physiological arousal when viewing threatening images from the personal harm frame and the evolutionary threat frame than liberal individuals, whereas they had less physiological arousal than liberals when viewing the threatening image in the outgroup (see Table 9 in Appendix B for means and standard deviations). These differences, however, were not significant (p = .44). Given these results the fourth hypothesis was not supported

0 1 2 3 4 5 6 7 8 9 Mean

Self-reported arousal and security conservatism

Personal harm - Liberal Personal harm - Conservative Outgroup - Liberal

Outgroup - Conservative Evolutionary threat - Liberal Evolutionary threat - Conservative Neutral - Liberal

Neutral - Conservative Positive - Liberal Positive - Conservative

Figure 6

Bar chart with means and 95% confidence intervals for physiological arousal and security conservatism per frame

In respect to the fifth hypothesis suggesting that anxious individuals will evaluate their

emotions in response to threatening images as more negative than non-anxious participants, there were no significant moderation effect, (F(4,88) = 1.08, p = .37). Hypothesis 5 is thus not supported.

A significant moderation effect was found for hypothesis 6 predicting that anxious individuals would have more self-reported and physiological arousal than non-anxious individuals in response to threatening images, (F(4, 92) = 2.48, p = .049). The effect was only significant for self-reported arousal. Figure 7 shows that anxious participants (n = 10) evaluated the emotions of the image in the outgroup frame as more negative than participants who were not anxious (n = 15) (see Figure 7). Non anxious individuals evaluated the emotions in response to the threatening images as more negative in the personal harm frame and in the evolutionary threat frame (see Table 9 in Appendix B for means and standard deviations). These differences in trait anxiety, however, were not significant (p = .76). Hence, the sixth hypothesis is not confirmed.

-0,006 -0,004 -0,002 0 0,002 0,004 0,006 Mean

Physiological arousal and security conservatism

Personal harm - Liberal Personal harm - Conservative Outgroup - Liberal

Outgroup - Conservative Evolutionary threat - Liberal Evolutionary threat - Conservative Neutral - Liberal

Neutral - Conservative Positive - Liberal Positive - Conservative

Figure 7

Bar chart with means and 95% confidence intervals for self-reported arousal and trait anxiety per frame

Further analyses. Linear regression analyses were calculated in SPSS to assess whether participants

experienced the images as negative and arousing when they reported that they are afraid of that issue. Two regression models with the issue (e.g., snake) from the phobia measure, corresponding with the image of the response as independent variable and the emotional responses (SAM and physiological) as dependent variables were significant. Being afraid of spiders predicted a more negative emotional evaluation of the image of the spider, (F(1, 36) = 9.43, p < .01). The strength of the prediction was moderate: 19 percent of the variation in the evaluation of valence can be predicted by being afraid of spiders (R 2 _{= .19). Being afraid of spiders also predicted more self-reported arousal in response to the}

image of the spider, (F(1, 37) = 4.65, p < .05). The strength of this prediction was small, with nine percent of the variation of the evaluation of arousal can be predicted by being afraid of spiders (R2 _{= .09). None of the regression analyses with physiological arousal as dependent variable was}

significant (see Table 10 in Appendix B).

It was further assessed whether the self-reported evaluation of arousal corresponded with the physiological measures of arousal for each image. Pearson’s correlation coefficient does not show any significant correlations between the self-reported arousal of an image and the physiological arousal (see Table 11 in Appendix B).

0 1 2 3 4 5 6 7 8 Mean

Self-reported arousal and trait anxiety

Personal harm - Non-anxious Personal harm - Anxious Outgroup - Non-anxious Outgroup - Anxious

Evolutionary threat - Non-anxious Evolutionary threat - Anxious Neutral - Non-anxious Neutral - Anxious Positive - Non-anxious Positive - Anxious

Discussion and conclusion

The aim of this experiment was to examine the emotional responses to threatening visual frames, as negative visuals dominate media news and yet there has not been much research

investigating how they affect individuals (Bucy & Grabe, 2007; Powell et al., 2015). Further it was tested whether the emotional responses are stronger for conservatives and individuals with an anxiety proneness. Emotional responses were assessed by asking individuals to evaluate their emotions in response to threatening, neutral and positive images in their valence and arousal. The study included three different threatening frames to see whether people respond differently to different kinds of threat. These are: personal harm, outgroup and evolutionary threat. In addition, physiological responses were assessed by measuring skin conductance (SC), an indicator for physiological arousal (e.g., Bradley et al., 2007; Potter & Bolls, 2012; Ravaja, 2004). It was predicted that threatening images would trigger stronger emotional responses than neutral and positive images especially for conservatives and individuals with trait anxiety.

In general, the results of the current study indicate that threatening images are experienced as negative and arousing. More in detail, compared to neutral images, threatening images did

significantly elicit more self-reported and physiological arousal like the negativity bias predicts. Yet the negativity bias was not fully supported in this experiment as participants did not experience more arousal when viewing threatening images compared to when they saw positive images. A reason for this could be that the negative images were not experienced as negative as the positive images were experienced as positive. Therefore, a pilot study testing the negativity of the negative images and the positivity of the positive images would have been a good contribution to the validity of this study. Moreover, the threatening images may have not been as threatening, as for instance seeing the image of a gun pointed at a person in an experiment setting is not the same as someone actually holding a gun to their face.

There were several significant moderation effects. For instance, there was a significant difference for security conservatism in participant’s self-reported and physiological arousal, namely that conservatives significantly experienced more arousal when viewing the images from the personal harm frame compared to the image in the outgroup frame. Respondents also had significantly more

physiological arousal when they viewed the images from the evolutionary threat frame compared to the image in the outgroup frame. The results of this study did not support the hypothesis that conservative individuals have stronger emotional reactions compared to liberals. One significant difference in conservatism was found in the other direction, namely that security conservatives had significantly less physiological arousal than liberals when they viewed the image in the outgroup frame. These findings contradict with results from other studies suggesting that conservative individuals see outgroups as a threat (Hatemi et al., 2013; Hibbing et al., 2014; Oxley et al., 2008). Bakker and his colleagues (Bakker, Arceneaux, et al., n.d.; Bakker, Schumacher, Gothreau, &

Arceneaux, n.d.) tried to replicate the findings by Oxley et al., (2008) and by Hibbing et al., (2014) in three different countries and also did not find that conservatives had stronger physiological responses to threatening stimuli compared to liberals. The association between conservatism and threat,

therefore, needs further investigation. A possible explanation for the effect in the other direction as predicted is that firstly the sample was not very conservative as most respondents agreed with liberal statements and self-identified as more left-wing (see Table 2 and Table 3). This indicates that those participants that were coded as conservative (for median split) may not hold strong conservative views. Instead of seeing the image of the outgroup as a threat, they may not have cared much about them and therefore experienced less arousal than liberals, who may have felt pity towards immigrants. The challenge with measuring SC is that it is only an indicator of emotional arousal and not an

indicator of specific emotions (Potter & Bolls, 2012; Ravaja, 2004; Renshon et al., 2015) and therefore the true nature of the experienced emotions can only be speculated. For future research it is therefore suggested to have a more diverse sample in levels of conservatism.

For trait anxiety the results suggest that the emotions in response to the threatening images in the personal harm frame and in the evolutionary threat frame were evaluated as less arousing by anxious individuals compared to non-anxious individuals which is also contrary to the findings of other studies examining anxiety and threatening frames (e.g., Arceneaux, 2012; Gadarian & Albertson, 2014; Renshon et al., 2015). The differences in emotional responses for anxious and non-anxious participants in this study, however, were not significant. A possible explanation for these insignificant effects may be that the participants in this sample did not have strong dispositions of anxiety (see

Table 7 in Appendix B). The significant increase in SC for anxious individuals as found by Renshon and his colleagues (2015) may be because they induced anxiety by having participants watch a movie clip. This difference in experimentation design may be the reason for the dissimilar findings in this study as the current study was more interested in the proneness of anxiety and not anxiety as a state. It would still be interesting to investigate the role of trait anxiety in response to threatening frames in future research.

Strangely almost none of the regression analyses were significant testing whether individuals reporting that they are afraid of an issue also predicted that they would have stronger emotional reactions to an image corresponding to that fear. Only being afraid of spiders predicted more self-reported negative and arousing emotions. A possible explanation may be that respondents were asked whether they had been afraid of the issues during the last 30 days. It could be that participants generally were afraid of the issues but had not been exposed to that issue for a longer time than 30 days are therefore did not report this. This explanation is supported by the significant regression analysis for being afraid of spiders, as spiders are more common in Germany than for instance snakes, natural catastrophes, and weapons. Further the item being afraid of refugees may have been better formulated as being afraid of the refugee crisis.

Another strange result is that there were no significant correlations between self-reported and physiological arousal. The other studies relying on physiological measures did not report whether these measures correlated with self-reported emotional arousal (Bradley et al., 2007; Soroka et al., 2013; Soroka & McAdams, 2015). It would be interesting to discover whether SC correlated with self-reported arousal in these studies. SC is a highly sensitive measure and participants may have been distracted by noises in the museum while viewing the images.

Several limitations of this study have already been mentioned while discussing the results. In general, the results of this study are restricted to its small sample size, especially for the self-reported emotional responses. Multiple evaluations of the Self-Assessment Manikins were lost because the presentation of the experiment was not programmed to wait for the participants to evaluate their emotions, which is a major limitation. Another limitation may be that the scales that were used in the study were abbreviated to shorten the duration of the experiment because shortening the scales could

have affected their validity (Bakker & Lelkes, 2018). It should be considered that this experiment was in the course of a master thesis that did not have funding to compensate the participants and not as much time to conduct this research. Although the sample was small, there were still significant effects even for physiological arousal, which is not that common.

References

Aarøe, L. (2011). Investigating frame strength: The case of episodic and thematic frames. Political

Communication, 28(2), 207–226. https://doi.org/10.1080/10584609.2011.568041

Bakker, B., & Arceneaux, K. (n.d.). Concordance in Emotional Processing and Political Attitudes

Pre-analysis Plan.

Bakker, B. N., Arceneaux, K., & Schumacher, G. (n.d.). Ideological differences in threat sensitivity?

Three replications of the association between conservatism and threat sensitivity.

Bakker, B. N., & Lelkes, Y. (2018). Selling Ourselves Short? How Abbreviated Measures of

Personality Change the Way We Think about Personality and Politics. Journal of Politics, 80(4), 1311–1325.

Benedek, M., & Kaernbach, C. (2010). A continuous measure of phasic electrodermal activity.

Journal of Neuroscience Methods, 190(1), 80–91.

https://doi.org/10.1016/j.jneumeth.2010.04.028

Brader, T. (2005). Striking a Responsive Chord: How Political Ads Motivate and Persuade Voters by Appealing to Emotions. American Journal of Political Science, 49(2), 388–405.

https://doi.org/10.1111/j.0092-5853.2005.00130.x

Bradley, M. M., & Lang, P. J. (2007). Emotion and Motivation. In J. T. Cacioppo, L. G. Tassinary, & G. G. Berntson (Eds.), Handbook of Psychophysiology (3rd ed., pp. 581–608). Retrieved from http://brain-mind.med.uoc.gr/sites/default/files/aaaaEmotion_0.pdf

Bradley, S. D., Angelini, J. R., & Lee, S. (2007). Psychophysiological and Memory Effects of Negative Political ADS: Aversive, Arousing, and Well Remembered. Journal of Advertising,

36(4), 115–127. https://doi.org/10.2753/JOA0091-3367360409

Braithwaite, J. J., Derrick, W. G., Jones, R., & Rowe, M. (2013). A Guide for Analysing

Electrodermal Activity (EDA) and Skin Conductance Responses (SCRs) for Psychological Experiments. Psychophysiology, 49(1), 1017–1034. Retrieved from

http://www.bhamlive.bham.ac.uk/Documents/college-les/psych/saal/guide-electrodermal- activity.pdf%5Cnhttp://www.birmingham.ac.uk/documents/college-les/psych/saal/guide-electrodermal-activity.pdf%0Ahttps://www.birmingham.ac.uk/Documents/college-les/psych/sa

Bucy, E. P., & Grabe, M. E. (2007). Taking television seriously: A sound and image bite analysis of presidential campaign coverage, 1992-2004. Journal of Communication, 57(4), 652–675. https://doi.org/10.1111/j.1460-2466.2007.00362.x

Coe, C. M., Canelo, K. S., Vue, K., Hibbing, M. V., & Nicholson, S. P. (2017). The Physiology of Framing Effects: Threat Sensitivity and the Persuasiveness of Political Arguments. The Journal

of Politics, 79(4), 1465–1468. https://doi.org/10.1086/692737

de Vreese, C. H. (2012). New avenues for framing research. American Behavioral Scientist, 56(3), 365–375. https://doi.org/10.1177/0002764211426331

Dodd, M. D., Balzer, A., Jacobs, C. M., Gruszczynski, M. W., Smith, K. B., & Hibbing, J. R. (2012). The political left rolls with the good and the political right confronts the bad: Connecting

physiologiy and cognition to preferences.pdf. Philosophical Transactions of the Royal Society B:

Biological Sciences, 367(1589), 640–649.

Gadarian, S. K., & Albertson, B. (2014). Anxiety, immigration, and the search for information.

Political Psychology, 35(2), 133–164. https://doi.org/10.1111/pops.12034

Hatemi, P. K., Mcdermott, R., Eaves, L. J., Kendler, K. S., & Neale, M. C. (2013). Fear as a Disposition and an Emotional State: A Genetic and Environmental Approach to Out-Group Political Preferences. American Journal of Political Science, 57(2), 279–293.

https://doi.org/10.1111/ajps.12016

Hibbing, J. R., Smith, K. B., & Alford, J. R. (2014). Differences in negativity bias underlie variations in political ideology. Behavioral and Brain Sciences, 37(3), 297–307.

https://doi.org/10.1017/S0140525X13001192

Keene, J. R., & Lang, A. (2016). Dynamic motivated processing of emotional trajectories in public service announcements. Communication Monographs, 83(4), 468–485.

https://doi.org/10.1080/03637751.2016.1198040

Lang, A., Bradley, S. D., Sparks, J. V., & Lee, S. (2007). The Motivation Activation Measure (MAM): How Well Does MAM Predict Individual Differences in Physiological Indicators of Appetitive and Aversive Activation? Communication Methods and Measures, 1(2), 113–136.

Lang, P.J., Bradley, M. M., & Cuthbert, B. N. (1997). International Affective Picture System (IAPS): Technical Manual and Affective Ratings. NIMH Center for the Study of Emotion and Attention, 39–58. https://doi.org/10.1027/0269-8803/a000147

Lang, Peter J. (2010). Emotion and motivation: Toward consensus definitions and a common research purpose. Emotion Review, 2(3), 229–233. https://doi.org/10.1177/1754073910361984

Lecheler, S., & de Vreese, C. H. (2018). News framing effects: Theory and practice. In News Framing

Effects. https://doi.org/10.4324/9781315208077

Lee, S., & Lang, A. (2009). Discrete emotion and motivation: Relative activation in the appetitive and aversive motivational systems as a function of anger, sadness, fear, and joy during televised information campaigns. Media Psychology, 12(2), 148–170.

https://doi.org/10.1080/15213260902849927

Marcus, G. E., College, W., Mackuen, M. B., & Louis, S. (1993). Anxiety , Enthusiasm , and the Vote : The Emotional Underpinnings of Learning and Involvement During Presidential Campaigns Author ( s ): George E . Marcus and Michael B . Mackuen Source : The American Political Science Review , Vol . 87 , No . 3 ( Sep .,. The American Political Science Review,

87(3), 672–685.

Marcus, G. E., Neuman, W. R., MacKuen, M., & Crigler, A. N. (2013). The Affect Effect. The Affect

Effect. https://doi.org/10.7208/chicago/9780226574431.001.0001

Meffert, M. F., Chung, S., Joiner, A. J., Waks, L., & Garst, J. (2006). The effects of negativity and motivated information processing during a political campaign. Journal of Communication, 56(1), 27–51. https://doi.org/10.1111/j.1460-2466.2006.00003.x

Norris, C. J., Larsen, J. T., Crawford, L. E., & Cacioppo, J. T. (2011). Better (or worse) for some than others: Individual differences in the positivity offset and negativity bias. Journal of Research in

Personality, 45(1), 100–111. https://doi.org/10.1016/j.jrp.2010.12.001

Oxley, D. R., Smith, K. B., Alford, J. R., Hibbing, M. V, Miller, J. L., Scalora, M., … Hibbing, J. R. (2008). Political attitudes vary with Physiological Traits. 430(2001), 1667–1670.

https://doi.org/10.1126/science.1157627

Psychophysiological measurement and meaning: Cognitive and emotional processing of media

(pp. 101–139). Routledge.

Powell, T. E., Boomgaarden, H. G., de Swert, K., & de Vreese, C. H. (2015). A Clearer Picture: The Contribution of Visuals and Text to Framing Effects. Journal of Communication, 65(6), 997– 1017. https://doi.org/10.1111/jcom.12184

Ravaja, N. (2004). Contributions of Psychophysiology to Media Research : Review and Recommendations. Media Psychology, 6(2), 193–235.

https://doi.org/10.1207/s1532785xmep0602

Renshon, J., Lee, J. J., & Tingley, D. (2015). Physiological arousal and political beliefs. Political

Psychology, 36(5), 569–585.

Rodriguez, L., & Dimitrova, D. V. (2011). The levels of visual framing. Journal of Visual Literacy,

30(1), 48–65. https://doi.org/10.1080/23796529.2011.11674684

Rozin, P., & Royzman, E. B. (2001). Negativity Bias, Negativity Dominance, and Contagion.

Personality and Social Psychology Review, 5(4), 296–320.

Schimmack, U. (2005). Attentional interference effects of emotional pictures: Threat, negativity, or arousal? Emotion, 5(1), 55–66. https://doi.org/10.1037/1528-3542.5.1.55

Slothuus, R., & de Vreese, C. H. (2010). Political parties, motivated reasoning, and issue framing effects. Journal of Politics, 72(3), 630–645. https://doi.org/10.1017/S002238161000006X Smith, K. B., Oxley, D., Hibbing, M. V., Alford, J. R., & Hibbing, J. R. (2011). Disgust sensitivity and

the neurophysiology of left-right political orientations. PLoS ONE, 6(10). https://doi.org/10.1371/journal.pone.0025552

Soroka, S. N. (2006). Good news and bad news: Asymmetric responses to economic information.

Journal of Politics, 68(2), 372–385. https://doi.org/10.1111/j.1468-2508.2006.00413.x

Soroka, S, Nir, L., & Fournier, P. (2013). Negativity Biases in Reactions to Network News: A Cross-National, Psychophysiological Pilot Study. Paper Prepared for Delivery at the Toronto Political

Behavior Workshop, (November), 1–35.

Soroka, Stuart, Gidengil, E., Fournier, P., & Nir, L. (2016). Do Women and Men Respond Differently to Negative News? Politics and Gender, 12(2), 344–368.

https://doi.org/10.1017/S1743923X16000131

Soroka, Stuart, Loewen, P., Fournier, P., & Rubenson, D. (2016). The Impact of News Photos on Support for Military Action. Political Communication, 33(4), 563–582.

https://doi.org/10.1080/10584609.2015.1133745

Soroka, Stuart, & McAdams, S. (2015). News, Politics, and Negativity. Political Communication,

32(1), 1–22. https://doi.org/10.1080/10584609.2014.881942

Spitzer, R. L., Kroenke, K., Williams, J. B. W., & Lowe, B. (2006). A Brief Measure for Assessing Generalized Anxiety Disorder. Archives of Internal Medicine, 166(10), 1092–1097.

Sylvers, P., Lilienfeld, S. O., & LaPrairie, J. L. (2011). Differences between trait fear and trait anxiety: Implications for psychopathology. Clinical Psychology Review, 31(1), 122–137.

https://doi.org/10.1016/j.cpr.2010.08.004

Valkenburg, P. M., & Peter, J. (2013). The differential susceptibility to media effects model. Journal

of Communication, 63(2), 221–243. https://doi.org/10.1111/jcom.12024

Yegiyan, N. S., & Lang, A. (2010). Processing central and peripheral detail: How content arousal and emotional tone influence encoding. Media Psychology, 13(1), 77–99.

Appendix A - Self-Assessment Manikins (SAM)

Dimension valence

Appendix B - Tables

Table 4

Correlation matrix for SAM evaluations of valence

Political

threat Knife Gun Refugee

Knife 1 .69** .23

Gun .69** 1 .21

Refugee .23 .21 1

Evolutionary

threat Dog Fire Snake Spider

Dog 1 .41** .68** .78**

Fire .41** 1 .26 .52**

Snake .68** .26 1 .63**

Spider .78** .52** .63** 1

Neutral Lightbulb Spoon Glass Basket Lamp

Lightbulb 1 -.06 .04 -.28 .06

Spoon -.06 1 0.01 .18 .46**

Glass .04 .01 1 .58** _.61**

Basket -.28 .18 .58** 1 .47**

Lamp .06 .46** _.61** _.47** ₁

Positive Lake Baby Beach Seal Babies

Lake 1 .24 .47** .17 .32 Baby .24 1 .59** .32 .61** Beach .47** .59** 1 .42* .38* Seal .17 .32 .42* 1 .34* Babies .32 .61** .38* .34* 1 Note: ** p < .01, * p < .05

Table 5

Correlation matrix for SAM evaluation of arousal

Political threat Knife Gun Refugee

Knife 1 .47** .36*

Gun .47** 1 .01

Refugee .36* .01 1

Evolutionary

threat Dog Fire Snake Spider

Dog 1 .36** .64** .53**

Fire .36** 1 .26 .48**

Snake .64** .26 1 .63**

Spider .53** .48** .63** 1

Neutral Lightbulb Spoon Glass Basket Lamp

Lightbulb 1 .30 .23 .26 .25

Spoon .30 1 .43** .43** _.27

Glass .23 .47** 1 .50** .28

Basket .26 .43** .50** 1 .42**

Lamp 0.25 .27 .28 .42** ₁

Positive Lake Baby Beach Seal Babies

Lake 1 .62** .80** .64** .62** Baby .62** 1 .62** .40* .78** Beach .80** .62** 1 .48** .51** Seal .64** .40** .48** 1 .37* Babies .62** .77** .51** .37* 1 Note: ** p < .01, * p < .05

Table 6

Means and standard deviations (SD) of phobia scale

Phobia Mean (SD) Spiders 1.88 (1.20) Snakes 1.65 (1.12) Weapons 2.18 (1.47) Natural catastrophes 2.25 (1.32) Refugees 1.30 (69) Violence 2.48 (1.36) Table 7

Means and SDs of trait anxiety

Trait anxiety

Mean (Standard deviation) Worrying too much about different things 1.98

(.62)

Having trouble relaxing 1.78

(77) Becoming easily annoyed or irritable 1.85 (.48) Being so restless that it is hard to sit still 1.43

Table 8

Means, SDs and 95 % Confidence Intervals (CIs) of main effects of repeated measures ANOVAs per frame

Dependent variable

Personal

harm Outgroup

Evolutionary

threat Neutral Positive

SAM - valence Mean (SD) 7.48 (1.38) 6.92 (1.14) 6.96 (1.25) 4.75 (0.54) 2.01 (0.82) 95% CI .56 .47 .51 .22 .33 SAM -arousal Mean (SD) 5.46 (1.84) 5.00 (2.00) 4.84 (1.56) 2.13 (1.05) 4.86 (1.79) 95% CI .74 .80 .62 .42 .71 Physiological arousal Mean (SD) .002 (.005) -.001 (.005) .001 (.003) -.001 (.003) .0001 (.003) 95% CI .002 .001 .001 .001 .001

Table 9

Means, SDs and 95 % CIs of moderation effects of repeated measures ANOVAs per frame

Dependent and moderation variable Personal harm - Liberal Personal harm - Conservative Outgroup - Liberal Outgroup - Conservative Evolutionary threat - Liberal Evolutionary threat - Conservative Neutral - Liberal Neutral - Conservative Positive - Liberal Positive - Conservative SAM - valence & economic conservatism Mean (SD) 7.11 (1.27) 8.00 (1.41) 6.93 (1.07) 6.90 (1.29) 6.59 (1.31) 7.48 (1.02) 4.83 (.26) 4.64 (.79) 2.20 (.98) 1.74 (.42) 95% CI .68 .89 .57 .81 .70 .64 .14 .50 .52 .27 SAM - arousal & security conservatism Mean (SD) 4.81 (1.46) 6.61 (1.96) 5.50 (2.03) 4.11 (1.69) 4.44 (1.23) 5.56 (1.88) 2.05 (.91) 2.27 (1.32) 4.75 (1.57) 5.04 (2.21) 95% CI .73 1.31 1.02 1.13 .62 1.25 .45 .88 .79 1.47 Physiological arousal & security conservatism Mean (SD) .001 (.005) .003 (.005) .002 (.004) -.003 (.004) .001 (.004) .001 (.004) -.001 (.002) -.001 (.004) .0003 (.003) -.0001 (.002) 95% CI 1.03 .87 1.20 .86 .84 .97 .43 .81 .90 1.17 SAM - arousal & trait

anxiety Personal harm - Non-anxious Personal harm - Anxious Outgroup - Non-anxious Outgroup - Anxious Evolutionary threat - Non-anxious Evolutionary threat - Anxious Neutral - Non-anxious Neutral - Anxious Positive - Non-anxious Positive - Anxious Mean (SD) 5.93 (2.00) 4.75 (1.38) 4.67 (2.32) 5.50 (1.35) 4.93 (1.62) 4.70 (1.53) 1.88 (.82) 2.50 (1.28) 4.57 (1.74) 5.28 (1.86) 95% CI 1.03 .87 1.20 .86 .84 .97 .43 .81 .90 1.17