1
Framing the picture: the effect of
congruence in images’ frames on attitude
towards immigrants
Corien W. Middelbeek (12538515) Master’s Thesis
Graduate School of Communication
Master’s programme Communication Science Supervisor: dr. L.C.N. Jacobs
Date: 31-01-2020
Abstract
Previous studies have established that frames in images can influence people’s attitude, and that this effect is especially strong when frames are congruent. Based on this knowledge, this study aimed to determine whether exposure to congruent frames in images depicting immigrants had a significantly different effect on attitude towards immigrants compared with exposure to incongruent stimuli. For this end, a survey embedded experiment was conducted in which participants were either exposed to congruent or incongruent frames which varied in being either threatening, empathy evoking, or both. Furthermore, the role of the socio-ideological attitude right-wing authoritarianism (RWA) in the relationship was studied in order to determine whether people’s predispositions can be used to explain images’ framing effects. Results of a one-way ANOVA and multiple regression analysis suggest that congruence in images’ frames does not generate different attitudes towards immigrants than does incongruence in frames and that RWA has no role in the relationship.
Keywords: Framing; attitude towards immigrants; frame congruence; right-wing
2 Everyone is familiar with the phrase “a picture is worth a thousand words”. This certainly holds for the photograph of the drowned Syrian toddler, Alan Kurdi, who was found washed ashore in Turkey after having tried to escape from the war in Syria. The photo, which was taken in 2015, went viral and became a top trending picture on Twitter under the hashtag ‘humanity washed ashore’ (Smith, 2015). As the humanitarian aspect of the migration crisis was made salient in the photo, empathy and concern were raised among the public as donations to charitable organisations increased steeply (Slovic et al., 2017).
Which aspects of an issue are made salient, such as in the picture of Alan Kurdi, has become a prominent topic in the field of communication science which has been referred to as ‘framing’. Frames are said to be interpretative packages in media discourse that make certain aspects of a complex issue salient to an audience in order to make the issue interpretable and understandable (Entman, 1993; Gamson & Modigliani, 1989). Relating to various subjects, some scholars have looked into framing in relation to content (Semetko & Valkenburg, 2000; Cissel, 2012; Van Gorp & Vercruysse, 2012; McMenamin et al., 2013; Curran & Gibson, 2019), whereas others have focused on the effects that framing can have on people’s attitudes, behavioural intentions and opinions (de Graaf et al., 2015; Wolsko et al., 2016; Cabrera et al., 2018).
While initially relatively underresearched (Powell et al., 2015), scientific attention has also been devoted to disentangling the framing effects that images can have. Similar to textual frames, frames in images can affect people’s attitudes, opinions and behavioural intentions (Von Sikorski et al., 2012; Powell et al., 2015). Powell et al. (2015), for example, found that exposure to images with an obligation to intervene frame brought about higher support for intervention than did exposure to images with a risk to intervene frame. Moreover, they found that these images, when presented without text, generated stronger framing effects than did the frames in the text-alone conditions (Powell et al., 2015). Furthermore, Schmuck &
3 Matthes (2017), found that when participants were exposed to both an image and a text, congruence (i.e. whether the frames of a text and an image match or not (Powell et al., 2015)), generated stronger effects on individuals’ attitudes than did incongruence in the frames.
While it appears, based on these studies, that frames in images and text can significantly affect individuals’ attitudes, behavioural intentions and opinions, these effects may actually be overestimated, as several studies showed that framing effects are dependent upon context and individuals’ prior knowledge of an issue, and people’s values and predispositions (Domke et al., 2002; Boomgaarden & Vliegenthart, 2009; Chong & Druckman, 2012; Schemer, 2012; Sohlberg et al., 2018).
Therefore, this study examines the causal relationship between frames in images and individuals’ attitudes by answering the following research question: To what extent does exposure to images concerning immigration which are congruent in their frames impact attitude towards immigrants differently than does exposure to images with incongruent frames? And is there a moderating effect of right-wing authoritarianism (RWA) on this relationship?
This study looks at the framing effects of images because images have increasingly become a prominent part of current day news reporting and since images’ frames have been found to generate stronger effects in comparison to textual frames (Powell et al., 2015). Therefore, by determining whether incongruence in images’ frames generates different effects on attitudes than does congruence in images’ frames, the extent of images’ framing effects can be established. By including the socio-ideological attitude RWA as a moderator, it furthermore becomes clear to what extent these effects are dependent upon individuals’ prior dispositions, as RWA consists of people’s beliefs about moral and traditional norms and values, coercive social control, and their belief in the importance of respect for and obedience to existing authorities (Duckitt & Sibley, 2010).
4 What’s more, this study looks at attitudes towards immigrants because the topic of immigration has become highly salient in news reporting since the onset of the European migration crisis (Eberl et al., 2018). And because studies found that this coverage is often conflict-centred and negative, and that media mostly make use of threat frames when discussing the topic, which leads to negative, more hostile attitudes towards immigrants (Eberl et al., 2018; Farris & Silber Mohamed, 2018). Therefore, by exploring the framing effects of images concerning immigrants, this study can also contribute to an understanding of the media’s role in shaping public opinion about immigrants.
Theoretical Framework
As previously mentioned, congruence of frames refers to whether the frames of a text and an image match or not (Powell et al., 2015). It is expected that congruent stimuli generates stronger framing effects than does incongruent stimuli (Chong & Druckman, 2007). This is the case because a certain type of information (i.e. a certain frame) is made increasingly salient when stimuli is congruent compared to when stimuli is incongruent. As noted by Entman (1993) and Gamson & Modigliani (1989), frames are used to understand and interpret issues. As congruence in frames enhances issue understanding, memory, and media learning (Powell et al., 2015), and because it induces cognitive argument approval (Schmuck & Matthes, 2017), it is understood why congruent stimuli has greater framing effects than incongruent stimuli.
The study by Schmuck & Matthes (2017), supports this notion as it was found that exposure to congruence in specific types of threat frames, that is economic and symbolic threat frames, generated stronger anti-immigration attitudes than did incongruent threat frames. In other words, people who were exposed to an image and a text which had the same type of threat frame, had significantly stronger anti-immigration attitudes than did participants who were exposed to a text and an image which were incongruent in their threat frames.
5 Atwell Seate & Mastro (2016), similarly demonstrated that threatening information can negatively impact attitudes towards immigrants as their study showed that exposure to threatening intergroup messages had a direct effect on attitudes held about immigrants’ rights, in the direction that participants considered that these rights should be more restrictive. Additionally, it indirectly influenced attitudes about immigration policy via intergroup anxiety, for which the participants also preferred more restriction.
Moreover, when considering studies that look beyond the effects of congruence in frames for only image-text pairs, more evidence is found that supports the notion of congruent stimuli generating stronger framing effects. For example, when participants were exposed to incongruent news frames regarding a U.S. military incident, subjects were more critical of the incident than when they were exposed to frames that reaffirmed one another (Rowling et al., 2013).
Regarding the topic of immigration, it is shown that media messages evoking humanitarian concern for immigrants are able to increase support for policies allowing more immigration, but this effect weakens in the presence of incongruent information which evokes perceptions of cultural and economic threat (Newman et al., 2015). Similarly, negative news portrayals of immigrants can bring about negative attitudes towards immigrants, but this effect reduces when there is frequent exposure to positive news portrayals of immigrants, although these effects are dependent on people having little knowledge on the issue of immigration (Schemer, 2012).
This latter finding is in line with studies on negativity bias which have shown that negative news evokes stronger and more durable reactions than does positive news (Soroka & McAdams, 2015; Soroka et al., 2019). This can be explained by the fact that people on average tend to pay more attention to negative than to positive news, both psychologically as well as physiologically (Soroka & McAdams, 2015; Soroka et al., 2019). Research by Igartua
6 & Cheng (2009), has supported this notion as it showed that a negative frame regarding immigration (namely a crime growth frame) in comparison to a positive frame (namely an economic contribution frame) generated more negative attitudes towards immigration.
A study by Bos et al. (2016), that looked into the effects of valence of frames (i.e. whether a frame is positive or negative), however, finds that it is not so much the valence of a frame that affects people’s attitudes, but that it is rather the frame itself which impacts attitudes (Bos et al., 2016). But considering that valence of a frame can overlap to a certain extent with the frame used, as a threat frame for example is inherently negative, it is not surprising that studies found that negative news coverage of immigration negatively impacts attitudes towards immigrants (Igartua & Cheng, 2009; Eberl et al., 2018; Farris & Silber Mohamed, 2018). Moreover, other studies have demonstrated that valence is relevant in predicting anti-immigrant attitudes, as news with a negative tone was found to be able to evoke stronger negative attitudes about immigrants than positive news was able to yield positive attitudes (Boomgaarden & Vliegenthart, 2009; Jacobs & van der Linden, 2018). This might be because exposure to certain negative issues increases threat perceptions (Van der Linden & Jacobs, 2017).
Based on the finding that it is the frame that influences attitudes, this study will assess the effects of two specific frames on attitudes towards immigrants, namely that of a threat frame and an empathy frame. And these will be shown either independent of one another, or together, so that the effect of congruence in frames can be established. These two frames are chosen because the threat frame has been found to generate negative attitudes towards immigrants (Atwell Seate & Mastro; Schmuck & Matthes, 2017; Eberl et al., 2018; Farris & Silber Mohamed, 2018), while news highlighting the humanitarian aspects of immigration can generate more positive attitudes towards immigrants (Newman et al., 2015; Slovic et al., 2017; Parrott et al., 2019). As emphasizing the obligation to intervene in order to prevent
7 human rights violations is considered empathy framing (Powell et al., 2015), this study will from here on out refer to this frame as an empathy frame.
Based on results indicating that congruent stimuli generates stronger framing effects than does incongruent stimuli and that a threat frame generates a negative attitude towards immigrants, the following is hypothesized:
H1: Compared with exposure to anempathy frame, with exposure to both a threat as well as an empathy frame, and with no exposure, exposure to a threat frame is negatively associated with attitude towards immigrants.
Furthermore, based on the finding that an empathy frame can bring about a positive attitude towards immigrants, the following is hypothesized:
H2: Compared with exposure to both a threat as well as an empathy frame, and with no exposure, exposure to an empathy frame is positively associated with attitude towards immigrants.
However, the positive effect of exposure to only an empathy frame on attitude towards immigrants is expected to be smaller than that of exposure to only a threat frame, as a threat frame is likely to be perceived as more negative than an empathy frame. This is because of a negativity bias which causes the threat frame to have stronger framing effects. Therefore, the following is hypothesized about exposure to incongruence in frames, namely exposure to both a threat as well as an empathy frame, compared with when there is no exposure:
H3: Compared with no exposure, exposure to both a threat as well as an empathy frame will generate a slightly more negative attitude towards immigrants.
8 Although the previous section has elaborated on news frames’ ability to impact attitudes, which partly depends on congruence in frames, many studies have emphasized that other factors play a role as well in determining the extent of framing effects. People’s knowledge of the topic under discussion (Schemer, 2012), prior opinions (Chong & Druckman, 2012), and predispositions (Domke et al., 2002), have all been found to moderate the relationship between exposure to frames and people’s attitudes and behaviours. Strong predispositions, for example, can weaken framing effects due to people’s resistance to messages that are disconfirming (Chong & Druckman, 2007). Furthermore, audiences are not passive and can actively evaluate the strength and relevance of messages, which influences the effectiveness of frames in shaping attitudes and behaviours (Chong & Druckman, 2007).
Therefore, this research will also look into the role that RWA has in the relationship between congruence in images’ frames and attitude towards immigrants. RWA is a socio-ideological attitude (Van Assche et al., 2014), which consists of people’s beliefs in conforming to moral and traditional norms and values (i.e. conventionalism), their belief in coercive social control (i.e. authoritarian aggression), and the extent to which they consider respect for and obedience to existing authorities to be important (i.e. authoritarian submission) (Duckitt & Sibley, 2010).
Research into RWA has demonstrated that ethnic diversity is associated with negative attitudes towards immigrants among people high in RWA, while it predicted positive attitudes for people low in RWA (Van Assche et al., 2014). Furthermore, RWA predicts opposition to immigrants and that this effect is even stronger when it interacts with perceptions of outgroup social threat (Duckitt & Sibley, 2010).
Based on this knowledge, the following is hypothesized about the interaction of exposure to a threat frame with RWA:
9 H4: Compared with exposure to an empathy frame, with exposure to both a threat as well as an empathy frame, and with no exposure, the negative effect of exposure to a threat frame on attitude towards immigrants is stronger for participants high in RWA than participants low in RWA.
As perceptions of threat are not evoked in the congruent empathy frame condition, the following is hypothesized regarding the interaction of exposure to an empathy frame with RWA:
H5: Compared with exposure to both a threat as well as an empathy frame, and with no exposure, the positive effect of exposure to an empathy frame on attitude towards immigrants is weaker for people high in RWA compared to those who are low in RWA.
However, as perceptions of threat are elicited in the incongruent threat and empathy frames condition, the following is hypothesized with regards to this condition and its interaction with RWA:
H6: Compared with no exposure, the negative effect of exposure to both a threat as well as an empathy frame on attitude towards immigrants is stronger for those who are high in RWA compared with those who are low in RWA.
Method and Measurements
Participants. By means of convenience and snowballing sampling, 174 participants
were recruited. Through social media platforms (i.e. Facebook, WhatsApp, Email and Instagram) participants were invited to fill in an online questionnaire which they could enter via a link to Qualtrics. Although a benefit of these ways of sampling is that data could be easily gathered and in a short duration of time, the disadvantages are that the sample gathered is unlikely to be representative of the population in society. This implies that there is little external validity and that the findings of this study are not generalizable to other people or
10 across situations. However, in this study, the main points of interests are changes between-subjects following exposure to different types of frames, making that population parameters are less relevant. Hence, even snowball sampling can help to gain insight into the mechanism, even though replication with a representative sample is still necessary.
A cover story was used to mask the goal of the study. Subjects were told that the study aimed to determine to which extent images with people affect memory differently than do images without people, and what role duration of exposure plays in this. Deception was necessary because this study wanted to establish whether there is a causal relationship between images’ frames and subjects’ attitude towards immigrants.
Upon entering the survey, participants were presented with the cover story and told that they would be shown some images and asked several questions, and they were informed about their rights and asked to give informed consent before being able to continue with the survey. Furthermore, as participants would be shown images of immigrants which were likely to be perceived as threatening and empathy evoking, only subjects who were 18 years or older could partake in the survey out of ethical considerations. Therefore, subjects were obliged to provide their age subsequent to having given informed consent. Moreover, after having finished the survey, participants were fully debriefed about the true purpose of the study.
Of the 174 participants recruited, only the 161 subjects who gave informed consent were included in the sample. Of these 161 participants, 14.9% had missing values on the items measuring the dependent variable and they were therefore excluded from the sample. A total of 37 participants were deleted, which resulted in a total sample size of 137 respondents (N = 137, Mage = 31.56, SD = 14.20, 22.6% = male, 74.5% = female).
Stimuli. Similar to research by Powell et al. (2015), the images that were used in this
11 pilot test was conducted in order to ensure that the stimuli used in the experiment indeed evoked perceptions of threat and empathy. In this study, a group of participants (N = 29, Mage = 28,34, SD = 12.55, 6.6% = male and 90.0% = female) was asked to rate ten individual images in which immigrants were depicted. For each image, participants were asked to which extent they found the image threatening. Answers were measured on a scale from 0 to 10 (0 = ‘not at all’, 10 = ‘extremely’). Similarly, they were asked to which extent they found each image empathy evoking, which was measured using the same scale.
The images were derived from online news sources (CBS News, BBC News, China Daily, CNN and the New York Post) and were real and non-manipulated. To ensure that it were the frames of the images that influenced participants’ attitude towards immigrants rather than the immigrants shown, images were selected on the basis of depicting a certain type of immigrant. For example, only Caucasian males were shown in the images and no images included children or elderly people. Furthermore, all images included more than one person, as outgroup size can influence perceptions of outgroup threat (Schlueter & Davidov, 2013).
Based on this pilot study, the image with the highest mean rating on the threat scale (M = 7.55, SD = 1.70) was selected and compared against the mean threat score of the other images combined (M = 4.94, SD = 1.72). A paired sample t-test showed that the difference in scores was significant, t(19) = 5.12, p <.001. These results suggest that the image with the highest mean rating on the threat scale was perceived as significantly more threatening than the other images were on average.
This test was similarly conducted for the image with the highest mean rating on the empathy scale. Results showed that participants perceived the image with the highest mean empathy rating (M = 8.69, SD = 1.59) as more empathy evoking than the other images on average (M = 5.47, SD = 1.58). This difference was found to be significant t(25) = 13.60, p
12 <.001, which suggests that the image was perceived as significantly more empathy evoking than the other images were on average.
Furthermore, in order to ensure that the image with the highest mean rating on the threat scale was not also perceived to be empathy evoking, and that the image rated on average most empathy evoking was not also perceived to be threatening, correlation tests were performed.
Results showed that no significant correlation was found between the mean threat score and the mean empathy score of the image with the highest mean threat score, r = .026, p = .900. This result indicates that the image, while bringing about perceptions of threat, did not also evoke empathy among participants. Therefore, this image was used for the exposure conditions in the experiment in which participants became exposed to an image with a threat frame. That is the congruent threat frame condition, and the incongruent threat and empathy frames condition.
For the image with the highest mean rating on the empathy scale, the correlation test showed that there was a strong significant negative correlation between how threatening participants perceived the image to be and how empathy evoking they found the image, r = -.631, p = .003. This results indicates that the more the image was seen as empathy evoking, the less it was considered to be threatening. Therefore, this image was used for the exposure conditions in the experiment in which participants were shown an image with an empathy frame. That is the congruent empathy frame condition, and the incongruent threat and empathy frames condition.
In the threat frame image, a relatively young man is seen most prominent in the foreground who is wearing a piece of cloth to cover his mouth and nose and who appears to have thrown something, while burning piles are seen in the background. (See Appendix A). In
13 the empathy frame image, a middle-aged man wearing a lifejacket is seen on a shore. He is down on his knees with his arms out and hand palms up and he appears to be crying. A second man is beside him and appears to be comforting him (See Appendix B). The two images are similar in that they both show one man most prominently in the foreground who has another, less prominent man near him. Furthermore, both images show multiple other people in the background.
Design and Procedure. An online survey-embedded experiment was conducted in
order to test the hypotheses. Participants were randomly assigned to one of the four exposure conditions: the congruent threat frame condition (the conthreat condition) (n = 34), the congruent empathy frame condition (the conemp condition) (n = 34), the incongruent threat and empathy frames condition (the incon condition) (n = 33), and the control condition (n = 36).
Based on their assigned condition, subjects were shown either the threat frame image, the empathy frame image, both the threat and empathy frame images, or in case of the control condition, neither of these images. Similar to the research by Powell et al. (2015), each image was shown for 10 seconds after which the survey automatically progressed. Furthermore, in the incon condition, the two images were shown in subsequent order and this order varied between participants in the condition in order to ensure that possible framing effects were not attributable to the order in which the frames were shown. Subsequently, the dependent measures were presented which were followed by manipulation checks which measured the extent to which participants found the image(s) seen to be threatening and/or empathy evoking.
Furthermore, in line with the cover story, participants then viewed an image which depicted a demonstration sign about educational budget cuts. This image was shown for 3 seconds and was succeeded by 3 questions that tapped into participants’ memory regarding all
14 images depicted and prior questions asked. One question for example asked what topic the final image shown touched upon. Answers included “climate change”, “education”, “healthcare” and “don’t know”. At the end of the survey, subjects were asked to indicate their gender, country of birth, highest achieved educational level and political orientation.
Measurements
Dependent Variable. After having been exposed to the stimuli, participants were
asked about their attitude towards immigrants. Attitude towards immigrants in this study is considered an individual’s disposition to react with a level of unfavourableness or favourableness to a specific issue (Davidov & Meuleman, 2012), in this case: immigrants. Similar to previous research, this concept was measured by asking participants whether their country should allow more or less immigrants (Davidov & Meuleman, 2012; Schmuck & Matthes, 2017).
Three items derived from ESS (2019) were used to measure this concept: “To what extent do you think your country should allow people of the same race or ethnic group as most people in your country to come and live there?”, “How about people from the poorer countries outside Europe?”, and “How about people of a different race or ethnic group from most people in your country?”. Answers were given on a four point Likert scale (1 = ‘allow none’, 2 = ‘allow a few’, 3 = ‘allow some’, 4 = ‘allow many’).
A principal axis factor analysis (PAF) showed that the 3 items form a single uni-dimensional scale: only one component has an eigenvalue above 1 (eigenvalue 2.42) and there is a clear point of inflexion after this component in the scree plot. Together, these factors explain 80.71% of the variance in the original variables (items). After a direct oblimin rotation, all items correlate positively with the first factor, the variable “attitude towards immigrants from poorer countries outside Europe” has the strongest association (factor
15 loading is .95). Reliability of the scale is good, Cronbach’s alpha = .88. Therefore, it appears the scale measures “attitude towards immigrants” (M = 3.15, SD = 0.57).
Moderator. Statistical analyses by Passini (2008) into the three aspects that make up
RWA, namely conventionalism, authoritarian submission, and authoritarian aggression, showed that a model with the construct as a three-dimensional measurement was better fitting than a one-dimensional model. Additionally, the study found that, unlike conventionalism and authoritarian submission, the authoritarian aggression dimension is related to attitudes of outgroup exclusion. Therefore, it is assumed that it is this dimension that will interact with the frames shown in the exposure conditions. For that reason, only items concerning authoritarian aggression were used to measure RWA.
Specifically, participants were asked to what extent they agreed with the following four statements: “Crime rates show that we have to use extreme measures against delinquents”, “We have to be tolerant toward protesters”, “The government of my country should eliminate all opponents”, and “My country would be great if we got rid of the 'rotten apples' who are ruining everything”. Responses were given on a five point Likert-scale (1 = ‘strongly disagree’, 2 = ‘somewhat disagree’, 3 = ‘neither agree nor disagree’, 4 = ‘somewhat agree’, and 5 = ‘strongly agree’). The second item, “We have to be tolerant toward protesters”, was reverse coded.
A principal factor analysis (PAF) showed that the 4 items form a single uni-dimensional scale: only one component has an eigenvalue above 1 (eigenvalue 1.93) and there is a clear point of inflexion after this component in the scree plot. Together, these factors explain 48.15% of the variance in the original variables (items). After a direct oblimin rotation, all items correlate positively with the first factor, the variable “Crime rates show that we have to use extreme measures against delinquents” has the strongest association (factor
16 loading is .66). The scale is reasonably reliable, Cronbach’s alpha = .63. Therefore, it appears the scale measures “RWA” (M = 2.07, SD = 0.69).
Control variables. Subjects indicated what their gender was by choosing between ‘male’ and ‘female’. Country of birth was measured by asking participants to fill in in which country they were born. Furthermore, highest achieved educational level was coded into eight categories: primary school, secondary school, vocational training, bachelor of applied sciences, bachelor of (social) sciences/arts/law, master of applied sciences, master of (social) sciences/arts/law, and PhD.
Furthermore, political orientation was measured by asking participants the following question: “In politics people sometimes talk of 'left' and 'right'. Using this card, where would you place yourself on this scale, where 0 means the left and 10 means the right?”, which was derived from the European Social Survey (ESS) (2019). Correspondingly, answers were given on a 0 to 10 scale.
Method
An Analysis of Variance was conducted (one-way ANOVA) to establish whether there is a significant difference in attitude towards immigrants between the different exposure conditions. This was done because a One way ANOVA is the appropriate statistical test when the dependent variable is continuous and the independent variable consists of two or more discrete groups (Statistics Solutions, 2013), which consists in this study of the four exposure conditions conthreat, conemp, incon and control.
Additionally, a multiple regression analysis was run to determine the relationship between the dependent variable attitude towards immigrants and the exposure conditions, RWA, and the interactions of RWA with the exposure conditions. As RWA is a continuous variable while the exposure condition variable is categorical, dummy variables were
17 computed for the latter variable so that differences between the interactions of RWA with each exposure condition could be explored. A multiple regression analysis was used because this test allows the modelling of the dependent variable as a function of multiple independent variables. In this case, RWA, the exposure conditions, and the interaction variables of RWA with the exposure conditions.
In order to ensure that randomization of the participants in terms of age, gender, educational level, political orientation, country of birth and RWA level across the conditions was successful, the means of each of these variables across the exposure conditions were compared. As shown in Table 1, no significant differences were found between any of the conditions. Therefore, it can be concluded that randomization of the participants across all exposure conditions was successful and the items were for that reason not included in analysis of variance nor the multiple regression analysis.
Table 1
Results of a one factor analysis of variance
SS df MS F p Eta-squared Age Education Political orientation Country of birth Gender = Male RWA 319.90 3.72 11.48 49.03 0.17 2.89 3 3 3 3 3 3 106.63 1.24 3.83 16.34 0.06 0.96 0.52 0.48 0.91 1.01 0.30 2.10 .667 .697 .439 .390 .826 .104 0.01 0.01 0.02 0.02 0.01 0.05 Note. N varied between variables.
Analyses and Results
Descriptives. In Table 2, the descriptive results of the variables are provided.
Regarding the dependent variable attitude towards immigrants, it is observed that participants overall held a rather favourable attitude towards immigrants. And when looking at the
18 moderator, it is seen that participants on average were low in RWA. Additionally, the table shows that only 23% of the participants was male and that participants overall were more left in their political orientation. Moreover, the results indicate that the average of highest achieved educational level is quite high.
Table 2
Descriptive statistics of the variables
n Min. Max. M SD
Dependent variable
Attitude towards immigrants Moderator RWA Participants’ characteristics Age Education Political orientation Country of birth Gender = Male 137 137 137 134 134 132 133 1.00 1.00 19 1 0 1 0 4.00 5.00 82 8 10 18 1 3.15 2.07 31.56 5.22 4.05 2.61 0.23 0.57 0.69 14.20 1.60 2.05 4.02
Further descriptives on the percentages of the highest achieved educational level among the participants shows the following: primary school (0%), secondary school (8%), vocational training (5.8%), bachelor of applied sciences (16.1%), bachelor of (social) sciences/arts/law (27.7%), master of applied sciences (6.6%), master of (social) sciences/arts/law (33.6%), and
19 PhD (0%). From these percentages we can draw the conclusion that, on average, subjects had indeed quite a high level of educational attainment.
Manipulation checks. To check whether the threat frame image brought about similar
perceptions of threat among participants in the conthreat condition as among participants in the incon condition, subjects in these conditions were asked to which extent they found the image(s) they saw to be threatening. Answers were recorded on a 1 to 7 scale, in which 1 represented ‘not at all’ and 7 meant ‘to a great extent’. This question followed the items measuring the dependent variable.
When comparing the means of these two groups on this measurement for the threat frame image, results showed that subjects who were in the conthreat condition found the image slightly more threatening (M = 4.68, SD = 1.42) than did subjects in the incon condition (M = 4.15, SD = 1.38). A one-way analysis of variance was carried out to assess whether the perceived threat was significantly different between the two conditions. No significant difference was found, F (1, 62) = 2.27, p = .137, Eta-squared = .035.
Similarly, participants who were in the conemp condition and participants who were in the incon condition were asked to which extent they found the image(s) they saw to be empathy evoking. Answers were also recorded on a 1 to 7 scale, in which 1 represented ‘not at all’ and 7 meant ‘to a great extent’ and this question was also posed following the items measuring the dependent variable.
When comparing the means of these two groups on this measurement for the empathy frame image, results indicated that participants in the conemp condition found the image less empathy evoking (M = 5.06, SD = 1.60) than did participants in the incon condition (M = 5.79, SD = 1.02). A one-way analysis of variance showed that the perceived empathy was
20 significantly different between the two conditions, F (1, 64) = 4.84, p = .031, Eta-squared = .070.
While these findings are interesting in that they indicate that the threat frame image was perceived as similarly threatening across the conthreat and the incon conditions, whereas the level of empathy evoked by the empathy frame image was significantly lower in the conemp condition than it was in the incon condition, these results do not indicate whether the manipulations of the threat frame and the empathy frame were successful. In order to test this, it is necessary to have measurements of the perceived level of threat for all four conditions, so also for the conemp condition as well as the control condition. Similarly, measurements of the level of empathy evoked after exposure to the threat frame image and no exposure are necessary to establish whether the empathy frame image indeed evoked empathy.
However, these manipulation check measurements are missing from the dataset because the checks were not included in the survey that was used for this study. Therefore no proper manipulation checks could be conducted. However, as the mean average scores on the manipulation checks for the threat frame image (M = 4.41, SD = 1.41) as well as the average score on the manipulation checks for the empathy frame image (M = 5.42, SD = 1.38), are quite high, this might be an indication that the manipulations operated as intended.
Main effects. Subjects who were in the conemp condition were found to have the
most positive attitude towards immigrants (M = 3.21, SD = 0.49), followed by the subjects in the conthreat condition (M = 3.15, SD = 0.51), the control condition (M = 3.13, SD = 0.67), and lastly the incon condition (M = 3.12, SD = 0.59).
A one factor analysis of variance was conducted to determine whether the discrete exposure conditions significantly influenced attitude towards immigrants. Because
21 randomization of the control variables over the exposure conditions was successful, the control variables were not included in the analysis.
Results of the analysis showed that no significant effect among the subjects of the exposure conditions on respondents’ attitude towards immigrants was found, F (3, 133) = 0.15, p = .929. A post-hoc test furthermore, confirmed that there were no significant differences in attitudes towards immigrants between the participants of the four experimental conditions. Those who were in the conthreat condition did not differ significantly from those in the conemp condition (M difference = -0.06, p = 1.000), nor from those in the incon condition (M difference = 0.03, p = 1.000), nor from those in the control condition (M difference = 0.02, p = 1.000). Therefore, hypothesis 1 is rejected.
Furthermore, the post-hoc test also showed that subjects in the conemp condition did not differ significantly in their attitude towards immigrants from those in the incon condition (M difference = 0.08, p = 1.000), nor from those in the control condition (M difference = 0.08, p = 1.000), thereby rejecting hypothesis 2.
Also attitudes towards immigrants did not differ significantly between participants who were exposed to the incongruent threat and empathy frames compared with participants who were not exposed to a frame (M difference = -0.01, p = 1.000), which rejects hypothesis 3.
Interaction effects. In order to check if the assumption of homoscedasticity was met,
a scatterplot was plotted. From the scatterplot it appears that the residuals are normally distributed. Therefore, the assumption of homoscedasticity is met. Furthermore, the multiple regression model with attitude towards immigrants as the dependent variable, and RWA, the exposure conditions and the interaction variables of RWA with the exposure conditions as independent variables, is insignificant, F (7, 129) = 2.03, p = .056, R2 = .10. The model can therefore not be used to predict attitude towards immigrants. As model 1 in Table 3 shows,
22 when the conthreat condition and its interaction with RWA were used as the reference categories, no independent variables were found to be significantly associated with attitude towards immigrants.
Furthermore, as can be observed from model 2 in the table in which conemp and its interaction with RWA were the reference categories, it was similarly found that none of the variables was significantly correlated with attitude towards immigrants. Except for the interaction variable of RWA with the control condition, b* = -0.44, t = -2.24, p = .027, 95% CI [-0.83, -0.05]. This result indicates that no exposure negatively predicts attitude towards immigrants compared with exposure to incongruent threat and empathy frames, but only for participants who are high in RWA.
Lastly, as can be seen in model 3 of the table, when the incon condition and its interaction with RWA were used as the reference categories, neither RWA, b* = 0.27, t = -1.61, p = .109, 95% CI [-0.59, 0.06], nor the control condition, b* = -0.45, t = 0.99, p = .324, 95% CI [-0.45, 1.34], nor the interaction between these two variables, b* = -0.22, t = -1.00, p = .320, 95% CI [-0.67, 0.22], were found to be significantly associated with attitude towards immigrants.
As the sample size in this study was small, which could account for the lack of interaction effects found, four additional multiple regression analyses were run in which only one of the interaction variables was included per analysis. The model with the interaction variable of RWA with the conthreat condition showed to be insignificant, F (5, 131) = 1.78, p = .120, R2 = .25. Similarly, the model with the interaction of RWA with the conemp condition, F (5, 131) = 2.16, p = .062, R2 = .08, and the model with the interaction between RWA with the incon condition, F (5, 131) = 1.72, p = .135, R2 = .06, were found not to be significantly associated with attitude towards immigrants.
23 However, the model with the interaction of RWA with the control condition was found to be significant, F (5, 131) = 2.64, p = .026, R2 = .09. This indicates that the model can be used to predict attitude towards immigrants. The strength of the prediction is weak: 9% of the variation in attitude towards immigrants can be predicted on the basis of the incon condition and the interaction of the control condition with RWA (R2 = .09). The incon condition, b* = -0.72, t = -2.01, p = <.05, 95% CI [-1.43, -0.01], has a moderately strong association with attitude towards immigrants, while the interaction of RWA with the control condition has a relatively moderate association, , b* = -0.37, t = -2.12, p = <.05, 95% CI [-0.71, -0.02].
These results indicate that being in the incon condition has a significantly more negative effect on attitude towards immigrants compared with no exposure to frames, (b = -0.72, p < .05). Furthermore, no exposure has a negative effect on attitude towards immigrants compared with exposure, but only for participants who are high in RWA (b = -0.37, p < .05).
However, as hypotheses 4, 5 and 6, are concerned with the differences in the interaction effects of RWA with the exposure conditions, the hypotheses have to be rejected as the model including all interaction variables, as shown in Table 3, proved to be insignificant.
Discussion
The aim of this study was to determine whether exposure to congruent frames in images depicting immigrants would bring about a significantly different attitude towards immigrants than would incongruence in images’ frames. An image with a threat frame and an image with an empathy frame were used for this purpose and shown to participants either independently or in an incongruent pair. Contrary to expectations, analyses showed that exposure to congruent stimuli did not generate a significantly different attitude towards immigrants than did exposure to incongruent stimuli. Furthermore, neither did the threat
24 frame image bring about a significantly different attitude towards immigrants than did the empathy frame image. Furthermore, when exploring whether RWA had an influence on the extent to which the frames in the exposure conditions were able to affect attitudes towards immigrants, no significant interaction effects were found.
A few explanations could account for the lack of significant results. First of all, the sample size in this study is small for the amount of experimental conditions included in the design. Ideally, each of the exposure conditions would contain 50 participants in order to ensure statistical power. Therefore, the lack of insignificant findings could be a consequence of this small sample size. Regarding the interactions, the items measuring RWA, and more specifically authoritarian aggression, proved not to be to reliable. Therefore, it is possible that RWA was not properly measured which could explain why no interaction effects were found.
25
Table 3
Summary of the separate multiple regression models predicting positive attitude towards immigrants
Model 1 Model 2 Model 3
b s.e. p b s.e. p b s.e. p
RWA (right-wing authoritarianism) Exposure conditions
Congruent threat frame Congruent empathy frame
Incongruent threat and empathy frames Control
Interactions
Congruent threat X RWA Congruent empathy X RWA Incongruent X RWA Control X RWA -0.12 Ref. -0.09 0.22 0.67 Ref. 0.07 -0.15 -0.37 0.14 Ref. 0.43 0.45 0.44 Ref. 0.18 0.21 0.21 -0.05 Ref. 0.31 0.76 Ref. -0.22 -0.44 0.12 Ref. 0.44 0.43 Ref. 0.21 0.20 * -0.27 Ref. 0.45 Ref. -0.22 0.16 Ref. 0.45 Ref. 0.22 Note. N = 137. * p < .05. ** p < .01. *** p <.001.
26 Furthermore, regarding the stimuli used in this study, it is possible that the image which was used as the threat frame image did not actually bring about perceptions of threat. And similarly, it is possible that the image used as the empathy frame image did not evoke empathy. While it is assumed, based on the pilot study and the mean scores on the manipulation checks, that the manipulations in the exposure conditions operated as intended, no statistical analyses could be performed to confirm this assumption. Due to a lack of items measuring perceptions of threat in the conemp and control condition, and a lack of items measuring levels of evoked empathy in the conthreat and control condition, it was not possible to conduct proper manipulation checks. This is a great limitation of this study as this makes it impossible to establish whether the frames were perceived as intended but did not have an effect on attitude towards immigrants, or whether the lack of findings should be attributed to the images’ frames not having been perceived as threatening or empathy evoking.
Another, more theoretical explanation could lie in the lack of information provided to the subjects. Unlike previous studies, no references were made to the topic being researched prior to exposure to the frame(s). Schmuck & Matthes (2007), and Parrott (2019), for example, had a text on the issue being studied which was presented with the image. And Powell et al. (2015), informed participants that they would be seeing an image regarding the conflict in the Central African Republic prior to exposure. This study, on the contrary, used a cover story to mask the true purpose and no references to immigrants or immigration were made prior to exposure.
Therefore, it is possible that the absence of information regarding the issue of immigration might account for the lack of significant effects found, as participants could only deduct that the images concerned immigrants from the images themselves. As previous studies have shown that both issue importance affects the extent of framing effects (Lecheler
27 et al., 2009), as well as real world context (Boomgaarden & Vliegenthart, 2009), future studies might benefit from investigating the extent of images’ framing effects when images are presented both with and without information on what the topic being studied is.
A final explanation for the lack of significant framing effects is that the images’ frames simply had no effect on participants’ attitude towards immigrants even though the frames did succeed in evoking empathy or threat perceptions. With the issue of immigration having been a highly prominent topic in recent years (Eberl et al., 2018), it is possible that most participants in the study already had a strong attitude towards immigrants and were knowledgeable about the issue. As both strong prior dispositions, as well as issue-specific knowledge can weaken framing effects (Chong & Druckman, 2007; Schemer, 2012), it could be that the participants’ in the study were no longer affected by the frames in images depicting immigrants. What is also possible is that with a general increase in images’ prominence in the news, people become more used to seeing a variety of frames in images and therefore are no longer affected by every image they see. In a media world in which soft news reporting, infotainment and sensationalism are on the rise (Otto et al., 2017), a lack of framing effects might then turn out not to be such a bad thing.
References
Atwell Seate, A., & Mastro, D. (2016). Media's influence on immigration attitudes: An intergroup threat theory approach. Communication Monographs, 83(2), 194-213. Berry, M., Garcia-Blank, I., and Moore, K. (2015) Press Coverage of the Refugee and
Migrant Crisis in the EU: A Content Analysis of Five European Countries. Cardiff: Cardiff School of Journalism.
Boomgaarden, H. G., & Vliegenthart, R. (2009). How news content influences anti‐ immigration attitudes: Germany, 1993–2005. European Journal of Political Research, 48(4), 516-542.
28 Boomgaarden, H., Boukes, M., & Iorgoveanu, A. (2016). Image versus text: How newspaper reports affect evaluations of political candidates. International Journal of Communication, 10, 27.
Bos, L., Lecheler, S., Mewafi, M., & Vliegenthart, R. (2016). It's the frame that matters: Immigrant integration and media framing effects in the Netherlands. International Journal of Intercultural Relations, 55, 97-108.
Cabrera, A. A., Hidalgo, F. C., & Durán, C. E. (2018). Use of Framing in Communication about Health Issues and its effect on Attitudes, Impact and Decision Making. CUICIID 2018, 377.
Chong, D., & Druckman, J. N. (2007). Framing theory. Annual Review Political Science, 10, 103-126.
Chong, D., & Druckman, J. N. (2012). Counterframing effects. The Journal of Politics, 75(1), 1-16.
Cissel, M. (2012). Media Framing: a comparative content analysis on mainstream and alternative news coverage of Occupy Wall Street. The Elon Journal of Undergraduate Research in Communications, 3(1), 67-77.
Curran, N. M., & Gibson, J. (2019). Conflict and responsibility: Content analysis of American news media organizations’ framing of North Korea. Media, War & Conflict, 1750635219839203.
Davidov, E., & Meuleman, B. (2012). Explaining attitudes towards immigration policies in European countries: The role of human values. Journal of Ethnic and Migration Studies, 38(5), 757-775.
De Graaf, A., van den Putte, B., & de Bruijn, G. J. (2015). Effects of issue involvement and framing of a responsible drinking message on attitudes, intentions, and behavior. Journal of health communication, 20(8), 989-994.
Domke, D., Perlmutter, D., & Spratt, M. (2002). The primes of our times? An examination of the ‘power’of visual images. Journalism, 3(2), 131-159.
Duckitt, J., & Sibley, C. G. (2010). Right‐wing authoritarianism and social dominance orientation differentially moderate intergroup effects on prejudice. European Journal of Personality, 24(7), 583-601.
Eberl, J. M., Meltzer, C. E., Heidenreich, T., Herrero, B., Theorin, N., Lind, F., ... & Strömbäck, J. (2018). The European media discourse on immigration and its effects: a literature review. Annals of the International Communication Association, 42(3), 207-223.
Entman, R. M. (1993). Framing: Toward clarification of a fractured paradigm. Journal of Communication, 43(4), 51–58.
ESS Round 9: European Social Survey (2019): ESS-9 2018 Documentation Report. Edition 1.2. Bergen, European Social Survey Data Archive, NSD - Norwegian Centre for Research Data for ESS ERIC. doi:10.21338/NSD-ESS9-2018.
Farris, E. M., & Silber Mohamed, H. (2018). Picturing immigration: how the media criminalizes immigrants. Politics, Groups, and Identities, 6(4), 814-824.
29 Gamson,W. A., & Modigliani, A. (1989). Media discourse and public opinion on nuclear
power: A constructionist approach. American Journal of Sociology, 95(1), 1–37. Igartua, J. J., & Cheng, L. (2009). Moderating effect of group cue while processing news on
immigration: Is the framing effect a heuristic process?. Journal of Communication, 59(4), 726-749.
Jacobs, L., & van der Linden, M. (2018). Tone matters: Effects of exposure to positive and negative tone of television news stories on anti-immigrant attitudes and carry-over effects to uninvolved immigrant groups. International Journal of Public Opinion Research, 30(2), 211-232.
Lecheler, S., de Vreese, C., & Slothuus, R. (2009). Issue importance as a moderator of framing effects. Communication research, 36(3), 400-425.
McMenamin, I., Flynn, R., O’Malley, E., & Rafter, K. (2013). Commercialism and election framing: A content analysis of twelve newspapers in the 2011 Irish general election. The International Journal of Press/Politics, 18(2), 167-187.
Nenov, S. (Photographer). (2015, September 15). Clashes at Hungarian border [digital image]. CBS News. Retrieved from https://www.cbsnews.com/pictures/europe-struggles-with-migrant-refugee-crisis/21/.
Newman, B. J., Hartman, T. K., Lown, P. L., & Feldman, S. (2015). Easing the heavy hand: Humanitarian concern, empathy, and opinion on immigration. British Journal of Political Science, 45(3), 583-607.
Otto, L., Glogger, I., & Boukes, M. (2017). The softening of journalistic political communication: A comprehensive framework model of sensationalism, soft news, infotainment, and tabloidization. Communication Theory, 27(2), 136-155.
Papanikos, G. (Photographer). (2015, September 7). Europe Migrant Crisis [digital image]. CBS News. Retrieved from https://www.cbsnews.com/pictures/europe-struggles-with-migrant-refugee-crisis/49/
Parrott, S., Hoewe, J., Fan, M., & Huffman, K. (2019). Portrayals of Immigrants and Refugees in US News Media: Visual Framing and its Effect on Emotions and Attitudes. Journal of Broadcasting & Electronic Media, 1-22.
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.
Rowling, C. M., Sheets, P., & Jones, T. M. (2013). Frame contestation in the news: National identity, cultural resonance, and US drone policy. International Journal of Communication, 7, 23.
Schemer, C. (2012). The influence of news media on stereotypic attitudes toward immigrants in a political campaign. Journal of Communication, 62(5), 739-757.
Schlueter, E., & Davidov, E. (2013). Contextual sources of perceived group threat: Negative immigration-related news reports, immigrant group size and their interaction, Spain 1996–2007. European Sociological Review, 29(2), 179-191.
30 Schmuck, D., & Matthes, J. (2017). Effects of economic and symbolic threat appeals in
right-wing populist advertising on anti-immigrant attitudes: The impact of textual and visual appeals. Political Communication, 34(4), 607-626.
Semetko, H. A., & Valkenburg, P. M. (2000). Framing European politics: A content analysis of press and television news. Journal of communication, 50(2), 93-109.
Slovic, P., Västfjäll, D., Erlandsson, A., & Gregory, R. (2017). Iconic photographs and the ebb and flow of empathic response to humanitarian disasters. Proceedings of the National Academy of Sciences, 114(4), 640-644.
Smith, H. (2015, September 2). Shocking images of drowned Syrian boy show tragic plight of refugees. The Guardian. Retrieved (2019, November 19) from: https://www.theguardian.com/world/2015/sep/02/shocking-image-of-drowned-syrian-boy-shows-tragic-plight-of-refugees.
Sohlberg, J., Esaiasson, P., & Martinsson, J. (2018). The changing political impact of compassion-evoking pictures: the case of the drowned toddler Alan Kurdi. Journal of Ethnic and Migration Studies, 1-14.
Soroka, S., & McAdams, S. (2015). News, politics, and negativity. Political Communication, 32(1), 1-22.
Soroka, S., Fournier, P., & Nir, L. (2019). Cross-national evidence of a negativity bias in psychophysiological reactions to news. Proceedings of the National Academy of Sciences, 116(38), 18888-18892.
Statistics Solutions. (2013). Data analysis plan: One Way ANOVA [WWW Document]. Retrieved from http://www.statisticssolutions.com/academic-solutions/member- resources/member-profile/data-analysis-plan-templates/data-analysis-plan-one-way-anova/
Van Gorp, B., & Vercruysse, T. (2012). Frames and counter-frames giving meaning to dementia: A framing analysis of media content. Social Science & Medicine, 74(8), 1274-1281.
Van der Linden, M., & Jacobs, L. (2017). The impact of cultural, economic, and safety issues in Flemish television news coverage (2003–13) of North African immigrants on perceptions of intergroup threat. Ethnic and Racial Studies, 40(15), 2823-2841.
Von Sikorski, C., Schierl, T., Möller, C., & Oberhäuser, K. P. (2012). Visual news framing and effects on recipients’ attitudes toward athletes with physical disabilities. International Journal of Sport Communication, 5(1), 69-86.
Wolsko, C., Ariceaga, H., & Seiden, J. (2016). Red, white, and blue enough to be green: Effects of moral framing on climate change attitudes and conservation behaviors. Journal of Experimental Social Psychology, 65, 7-19.
31 Appendix
A.
(Source: Nenov, 2015).
32 (Source: Papanikos, 2015).
