The images of refugees:
A multi-methodological study about the visual representation of refugees in British and Dutch newspapers and its effects on emotions and attitudes
Aron Dekker 10279164
aron.dekker@student.uva.nl University of Amsterdam
January 29, 2016
Graduate School of Communication Master’s Thesis - Political Communication
Abstract
In media research the representation of refugees, asylum seekers and migrants has often been investigated and has demonstrated to shape citizens’ attitudes and opinions towards these outgroups. These studies, however, focus mainly on textual communication, but since visual communication is rising, research on visuals has become increasingly important. In this multi-methodological study we have therefore investigated the visual representation of refugees and its effects on attitudes, by focusing on the concepts of visual framing and visual framing effects. Furthermore, we examined to what extent visual frames evoked emotions and whether these emotions functioned as mediators for the framing effects. We first conducted a content analysis (N = 527) of British and Dutch newspapers to investigate to what extent refugees were either framed as victims or intruders, and consecutively in an experiment (N = 394) the effects of these frames were examined in a competitive and traditional experimental design. We found that refugees were predominantly portrayed as vulnerable victims rather than threatening intruders, and that there were only minor differences in this representation between newspapers. The findings of the experiment indicated that the visual frames influenced attitudes. Nevertheless, in the competitive design a direct effect was observed, whereas in the traditional design this effect was indirect through emotional responses. But either way, we showed that visuals play a crucial role in shaping citizens’ attitudes towards refugees.
Keywords: refugees, representation, visual framing, visual framing effects, emotional responses, attitudes, competitive framing
The images of refugees: A multi-methodological study about the visual representation of refugees in British and Dutch newspapers and its effects on emotions and attitudes
On the 3rd of September 2015 a picture was published of a Syrian boy who was found dead on a beach in Bodrum, Turkey. Together with his parents he tried to reach Europe by boat because they fled from war in Syria, but they never arrived as they got in trouble with their boat and drowned. Although not the first picture of its kind, this specific picture received considerable attention and put the refugee crisis high on the agenda. This Syrian boy is one of the hundreds of thousands refugees from the Middle East and Africa who have tried, and continue to make an effort, to reach the European border, as they flee from war and suppression in their home countries. According to Casella (2015), the refugee crisis that Europe is facing is the worst since the end of the Second World War. Because of its immense proportions, this crisis is discussed a lot by politicians, citizens and the media.
In media research the representation of refugees, asylum seekers and migrants has often been investigated (e.g., Van Gorp, 2005; d'Haenens & De Lange, 2001; Sulaiman-Hill, Thompson, Afsar & Hodliffe, 2011; Horsti, 2007), and these studies have demonstrated that this representation shape citizens’ attitudes and opinions towards these outgroups (e.g., Lecheler, Bos & Vliegenthart, 2015; Wood, & King, 2001). However, these studies focus mainly on textual communication. But since it was a picture that put the refugee crisis on the agenda, and overall visual information is rising (e.g., Geise & Baden, 2015), it has become increasingly important to examine the visual representation of refugees and its effects as well. Although there are some exceptions, there is still a gap in research on visual information. Because some studies have found that the visual representation is mainly negative (Bleiker, Campbell, Hutchison & Nicholson, 2013; Banks, 2012; Batziou, 2011), while other studies have found that refugees were predominantly positively portrayed (Van Gorp, 2006). However, there is not a clear explanation for this discrepancy. Therefore, in this study we
examine the visual media construction of refugees by focusing on the concept of framing. Framing is “the observation that media can portray one and the same topic in different ways, emphasizing certain evaluations or only parts of an issue at the expense of others” (Schuck & De Vreese, 2006, p. 5). In a study by Van Gorp (2005), it was found that refugees were either framed as victims or as intruders. Building on this classification, we will investigate to what extent refugees are also portrayed in terms of victims and intruders in visuals, and try to address the following research question:
RQ1: To what extent are visuals of the refugee crisis framed in terms of refugees as victims or refugees as intruders?
Furthermore, there is little knowledge about the effects of the visual representation. We will therefore also investigate to what extent the visual representation of refugees affects political attitudes, by focusing on the concept of visual framing effects. Recent studies have also shown that frames evoke emotions (Igartua, Moral-Toranzo, & Fernández, 2011), and that these emotions mediate the framing effects (e.g., Lecheler et al., 2015). Visuals can transmit emotions effectively, and these emotions are more directly and authentically by visuals than text (Brantner, Lobinger & Wetzstein, 2011), which makes it relevant to study emotional responses in the current study too. By conducting an experiment, we try to address to what extent the victim and intruder frame affect attitudes towards refugees and we examine the mediating role of emotions in this effect. The following research question will be addressed:
RQ2: To what extent does exposure to the visual victim and intruder frame affect attitudes towards refugees, and what is the role of emotions in this effect?
If we find an explanation for the discrepancy in the representation of refugees, and also conclude that this representation evokes different emotional responses, which subsequently influences participants’ attitudes, it can be argued that visuals are a crucial factor in shaping public opinion of the refugee crisis. Because we try to explain the importance of visuals in the refugee debate and also respond to the changing media environment, since visual information is rising, this study is highly societally relevant. Even though previous studies have already shown that the representation of refugees influences political attitudes, the literature on the role of visuals in shaping those attitudes is rather low. Besides we know little about the influence that visuals have on evoking emotions and to what extent these emotions are mediators for political attitudes. This makes the current study theoretically relevant, as we contribute to migration literature by focusing on the effects of the visual representation of refugees. We shall also bring more knowledge in the field of visual framing effects, as we focus on the role of emotional responses.
Theory
In this theory section, we will first outline the knowledge about the representation of refugees in the media and why this differs among different newspapers. We will then discuss studies on framing, framing effects and emotional responses to explain the role of media in shaping attitudes towards refugees, which consequently leads to the hypotheses.
Refugees in the media
The representation of refugees in the media has often been researched in the past years. In the majority of these studies, it becomes clear that refugees are commonly presented as criminals, possible terrorists, troublemakers (Horsti, 2007) or as undesirables (Wood & King, 2001). According to Horsti (2007) this is because only events in which refugees can be portrayed as such make it to news reports, whereas media only rarely focuses on more
subjected. In a study by Kaye (2001), it was found that British newspapers mainly referred to refugees with the terms “phoney” or “bogus”. However, Sulaiman-Hill et al. (2011) found that newspapers do not just cover refugees negatively. In their study Australian and New Zealand newspapers were compared over a ten-year period. New Zealand newspapers especially covered refugee-issues in a balanced tone and focused mainly on the positive contributions that refugees had made to that country. Australian newspapers, on the other hand, did focus more on the problems posed by refugee-issues and explicitly expressed their concerns. These studies all focus on the representation of refugees in texts. Nevertheless, to an increasing extent scholars focus on the visual representation of refugees too (Banks, 2012; Bleiker et al., 2013; Batziou, 2011). But as with the textual representation, refugees are mainly depicted in negative terms. Recent studies have shown that media portray refugees in visuals mainly as ‘the other’ (Batziou, 2011), as dehumanized (Bleiker et al., 2013), as a threat (Horsti, 2003) or as shadowy strangers (Banks, 2012). In two studies by Van Gorp (2005, 2006) the representation of refugees in Belgium newspapers was studied in both text and visuals and it was found that refugees were either portrayed as victims or as intruders. The focus of the majority of newspapers, however, was mainly on victims, and this contrast was even larger in visuals, since in visuals refugees were only rarely portrayed as intruders.
Although refugees are mainly portrayed in negative terms, the studies discussed above have also indicated that in the representation can be more positive. However, these studies are all conducted in different countries, focus on different crises, approached the representation from different theoretical backgrounds and used various methodological set-ups, which makes comparison difficult. Nevertheless, many comparative studies found that newspapers with a different political stance (conservative/ progressive) or type (tabloid/ quality) differ in the coverage of refugees (e.g., KhosraviNik, 2009; Gabrielatos & Baker, 2008). Therefore, we will elaborate more on the different characteristics of newspapers.
Differences in newspapers
When comparing tabloid and quality newspapers, it becomes clear that tabloids are overall more negative about refugees than quality newspapers (e.g., Gabrielatos & Baker, 2008). In various studies it was concluded that tabloid newspapers portray refugees more as intruders (Van Gorp, 2005) and are more hostile towards refugees (Blinder & Allen, 2015) than quality newspapers. In a critical discourse analysis by Gabrielatos and Baker (2008), it was found that tabloid newspapers focus more on the number of refugees, the manner of entry and the extent that refugees will be a threat to the country of destination. Additionally, tabloid newspapers use more words with a negative connotation regarding refugees (e.g., extremists, gangsters). Quality newspapers, on the other hand, focus on refugee-issues in a broader context by emphasizing on international, political and social affairs with overall a more balanced tone, and use more terms with a neutral or positive connotation (e.g., claimants, foreigners) when writing about refugee-issues. Although Sulaiman-Hill et al. (2011) did not distinguish between tabloid and quality newspapers, they concluded that New Zealand newspapers were more positive than Australian newspapers. But in their study, only quality newspapers from New Zealand were analyzed, whereas the analysis of Australian newspapers included both tabloid and quality newspapers. This can possibly explain the different
outcomes in their study.
When comparing progressive and conservative newspapers, research has shown that conservative newspapers are more negative than progressive newspapers. In a study by KhosraviNik (2009), it was found that British conservative newspapers hardly recognize refugees (mentioning names or someone’s qualities), unless they could portray them in
negative terms. Progressive newspapers, on the other hand, recognized refugees and portrayed them more in terms of human rights, ethics and values. A study by Batziou (2011) concluded that newspapers on the center right placed more emphasis on refugees’ otherness than
newspapers on the left, and according to Van Gorp (2005), progressive newspapers portrayed refugees more as victims than conservative newspapers.
Although comparative studies on newspaper type and political stance all conclude the same, there seems to be an interesting shift by the German newspaper Bild. Even though this newspaper is considered as one of the largest conservative tabloids in Germany, in the current refugee debate they take in a very progressive stance, as they plead for a more open and positive attitude towards refugees by the government and deprecate protests against refugees. The editor-in-chief of Bild.de explained that Bild had not shifted to a progressive stance, maintaining that a conservative newspaper should be positive about refugees, since
“conservative means conserving what is good about society (…) and one thing that is good about our society are the values it is based on” (Luyten, 2015). However, Bild seems to be an exception, since academic research found that overall conservative and tabloid newspapers were more negative about refugees than progressive and quality newspapers. However, the shift by Bild can be explained by a study by Coole (2002), in which it was found that a tragic and emotional event could change the tone of coverage by newspapers. The author found that the negative coverage of refugees changed after a tragic murder of a Turkish refugee in Glasgow to a more positive tone of coverage, in both tabloid and quality newspapers.
Interestingly, the shift by Bild started at the end of August 2015, a period in which there were many violent demonstrations against the arrival of refugees by neo-Nazi’s in Heidenau, Germany. Besides, on the 3rd of September a picture was published of a Syrian boy who was found dead on a beach in Turkey. These events were widely portrayed in the international press and could have contributed by the shift of Bild.
(Visual) framing
In the current study we examine the visual media construction of refugees by focusing on the concept of framing. Framing is the “observation that media can portray one and the
same topic in different ways, emphasizing certain evaluations or only parts of an issue at the expense of others” (Schuck & De Vreese, 2006, p. 5). According to Entman (1993), framing can be defined as selecting “some aspects of a perceived reality and make them more salient in a communicating text, in such a way as to promote a particular problem definition, causal interpretation, moral evaluation, and/or treatment recommendation for the condition
described” (p. 52). These definitions refer to emphasis framing or frames in communication (Druckman, 2001a), which means that the communicator selects or emphasizes certain aspects to lead individuals to focus on these specific aspects.1
The majority of framing research focuses on text (e.g., Sulaiman-Hill et al., 2011; Horsti, 2007; Wood & King, 2001). However, there is a desire among scholars to focus more on visual framing, since there is a rise in visual information (e.g., Geise & Baden, 2015; Coleman, 2010). Particularly online media focuses to a high extent on visuals, but there is also a rise in visual information in traditional newspapers (e.g., De Vries, 2008; Cooke, 2005). The definition by Entman (1993) originally focused on text, but can be adopted for visual information as well (Brantner et al., 2011). However, Coleman (2010) defines visual framing as “the selection of one view, scene or angle when making the image, cropping, editing or selecting it” (p. 237). In this definition framing is not only capturing the photograph, but also the selection of it, which is an important aspect since the journalist or editor who selects the visual for a certain newspaper article is not necessarily the journalist who captured the visual too. There are numerous photo archives from which journalists can select specific news images for their articles, and the selection of a visual is also an act of framing (Coleman, 2010). In most framing definitions, however, framing is related to the creation of a communicating text or visual (e.g., Entman, 1993). Therefore the definition by Coleman (2010) will be the main definition for the current study, as we are focusing on the selection rather than the creation of certain visuals by different newspapers.
In the current study we adopt the classification of Van Gorp (2005), who found that refugees were either framed as victims or as intruders. The victim frame portrays refugees as “vulnerable and innocent victims who are outside their country of nationality or habitual residence because of a well-founded fear of persecution because of his/her race, religion, nationality, membership in a particular social group or political opinion” (p. 489). These frames typically show “families with children and pictures that represent distress, fear and misery” (p. 491). The intruder frame, on the other hand, portrays refugees as “the unknown other or the dangerous stranger who is a threat to one’s own cultural and economic
achievements” (p. 489), and shows “a batch of asylum-seekers, especially single men” (p. 491). It can be argued that in the victim frame refugees are portrayed more positively and in the intruder frame more negatively. This division refers to valence framing (De Vreese and Boomgaarden, 2003; Schuck & De Vreese, 2006), which means that frames have an inherent valence of positive versus negative or good versus bad.
If we connect the discussed literature on the (visual) representation of refugees with the concept of (visual) framing, it can be expected that the visual victim frame is used more in quality and progressive newspapers, whereas the visual intruder frame is used more in tabloid and conservative newspapers. Possibly, the tragic events that started at the end of August contributed to a comparable shift in coverage in the newspapers that will be analyzed in the current study as in Bild. But since the shift by Bild is not a very common phenomenon, we will base our expectations on academic research only. Therefore, we will investigate the following hypothesis:
H1: The victim frame will be used more by quality and progressive newspapers, whereas tabloid and conservative newspapers focus more on the intruder frame.
(Visual) framing effects
In this study we will not only focus on the representation of refugees in newspapers, but also examine the effects of representation on the attitudes towards refugees. In this study we employ the concept of visual framing effects to explain the influence of different
portrayals. Previous research on framing effects has shown that frames can influence attitudes, opinions and behavior (e.g., Iyengar, 1994; Schuck & De Vreese, 2006; Nelson, Clawson & Oxley, 1997). In their study, Schuck and De Vreese (2006) showed that
participants indicated higher levels of support for the European enlargement when the issue was framed as an opportunity compared to a risk, and Nelson et al. (1997) studied whether different frames influenced the tolerance for the Ku Klux Klan (KKK), and concluded that participants had more tolerance for the KKK when the group was framed in terms of free speech rather than a disruption of public order. As in the current study, Nelson et al (1997) employed a definition based on emphasis framing, since the focus (emphasis) of the KKK was either on free speech or on the disruption of public order. By emphasizing on these aspects, a communicator leads individuals to evaluate either free speech or the disruption of public order when reaching an opinion (Druckman, 2001a). However, in contrast to what early studies on framing effects argued (e.g., Iyengar, 1994), this is not a passive process in which the frame affects the “unconscious” individual, but rather an active process in which individuals evaluate the suggested emphasis of the frames deliberately and consciously, which influences one’s beliefs and the importance that will be attached to these specific beliefs. This belief importance subsequently affects attitudes and opinions (Druckman, 2001b; Nelson & Oxley, 1999).
But as with the presence of visual frames, the majority of research on framing effects focuses on text. Nevertheless, there are some (mainly recent) studies on visual framing effects (e.g., Brantner et al., 2011; Powell, Boomgaarden, De Swert, & De Vreese, 2015; Scharrer &
Blackburn, 2015; Gibson & Zillmann, 2000). In all of these studies it is concluded that visuals, at least to some extent, affect opinions, attitudes and/or behavior too. In their study, Powell et al. (2015) found that participants indicated higher levels of support for intervention and behavioral intentions for the conflict in the Central African Republic (CAR), when they were presented to a visual in which the intervention was framed as an obligation rather than a risk. The frames were investigated in both text and visuals, and it was found that the visual frames had stronger direct effects. Interestingly, this finding is in line with earlier research on visual framing effects. According to Gibson and Zillmann (2000), this can be explained by the
picture superiority theory, which means that when a text and a visual of the same article are
incongruent, the frame of the visual is more dominant, which indicates that visuals are more powerful tools to spread frames than texts. Besides, Messaris and Abraham (2001) argued that visuals are very effective vehicles for framing, because of their analogical quality,
indexicality and the lack of an explicit propositional syntax. The analogical quality refers to
the relationship between visuals and their meanings, as this relationship is based on
similarities and analogies, rather than on social conventions (as in texts). In other words, for the recognition of certain objects in visuals, an individual does not require to have prior familiarity with the specific objects. Indexicality refers to the close connection between photographs and reality, as visuals have the implicit character of being close to the truth. This implicit character decreases the likelihood that individuals will question what they see. Finally, in a text one can clearly describe causality, comparisons or generalizations (e.g., refugees are a threat to our culture), but visuals do not have these explicit propositions, therefore in visuals, the viewer must define these propositions themselves. Because of these characteristics, frames in visuals become less obtrusive and therefore viewers of visuals are less aware of the framing process than when they are exposed to textual frames, and are less resistant to framing effects (Messaris & Abraham, 2001).
The role of emotions
In framing effects studies there is increasing attention paid to the role of emotions. Recent studies have shown that frames evoke emotional responses (Igartua et al., 2011) and that these emotions mediate the framing effects, such as political opinions, attitudes and behaviors (e.g., Gross, 2008; Lecheler et al., 2015; Powell et al., 2015). According to
Lecheler et al. (2015) these emotions are particularly interesting in political topics that cause ‘hot’ public debates. Since the current refugee crisis is an event that causes such a debate among the public, emotions might play an important role in the current study too.
In their study, Igartua et al. (2011) found that different frames evoked different emotional responses; when immigration was framed in terms of economic contribution, participants indicated more positive emotions and fewer negative emotions than when they were exposed to a frame in which immigration was linked to delinquency. Lecheler et al. (2015) went even further and concluded that different frames influenced opinions towards immigration, and that this effect was mediated by emotional responses. For instance exposure to a frame with a positive valence evoked more positive emotions that in turn resulted in more positive opinions about immigration.
Even though there is increasing evidence about the importance of emotions in framing effects research, the focus of these studies is mainly on textual frames. However, Brantner et al. (2011) argued that visuals can transmit emotions effectively and that these emotions are more directly and authentically than emotions transmitted by text. In their study they found that visual frames evoked stronger emotional responses than textual-frames, but they also claim that not every visual evoked the same emotional effects. For example, they found that visuals of Palestinian victims evoked stronger emotional responses than visuals of politicians. Iyer and Oldmeadow (2006) found that visual frames had more influence on emotional responses than textual frames, and that these emotions influenced attitudes, and Powell et al.
(2015) found that particularly visuals evoke emotions, which subsequently affect policy support and behavioral intentions. A study by Iyer, Webster, Hornsey and Vanman (2014) focused on visuals of terrorism and the extent to which these visuals evoked emotions and political attitudes. Participants were exposed to either visuals of victims or visuals of perpetrators of terrorism, and it was found that the victim condition evoked emotions of sympathy, whereas the terrorism condition evoked emotions of fear and anger. Furthermore, each of these emotions influenced support for different counterterrorism policies. Although there is still little evidence, the studies described above provide clear results that indicate that visual frames do evoke emotions that in turn affect attitudes and behavior.
Competitive framing
Although most studies find evidence of the effects of (visual) frames, in real-life environments individuals are exposed to multiple frames, as in politics and the media issues are debated in contrasting frames. In the majority of framing effects studies, however, participants are exposed to one frame only. Therefore, Chong and Druckman (2007)
introduced the concept of competitive framing (or counter framing); an experimental design in which participants will be exposed to contrasting frames, instead of just one frame.
Sniderman and Theriault (2004) argued that in a competitive design, frames might not be as influential as previous research suggested, because the frames may cancel each other out. This assumption was examined by Brewer and Gross (2005), who measured participants’ support for policy positions after exposure to two contrasting frames, no frames or one of the two frames. It was found that participants who were exposed to one of the two frames indicated the strongest effects, but participants in the mixed condition also indicated effects on policy positions. A more recent study by Nisbet, Hart, Myers and Ellithorpe (2013), also investigated whether competitive frames regarding climate change had more or less influence than
on participants’ support for policy positions than participants exposed to a single frame. This effect, however, was moderated by participants’ individual differences; open-minded
participants were able to weigh the costs and benefits of the presented policies, which increased their support.
Although a competitive design comes closer to a real-life environment, studies that employed this design exposed participants to an equal number of opposing frames (e.g., Brewer & Gross, 2005; Nisbet et al., 2013). In a real-life environment, however, individuals are exposed to multiple frames, and some frames are more dominant than others. Therefore, in the current study we will examine to what extent participants rely on the dominant frame in their attitude formation.
Hypotheses
To our knowledge, visual framing effects and the role of emotions have never been researched in a competitive framing design before. However, we will base our hypotheses regarding the framing effects on previous noncompetitive framing studies, since the studies by Brewer and Gross (2005) and Nisbet et al. (2013) have showed that competitive frames can have framing effect as well. But to examine whether the design of the experiment affects the outcomes of the study, we will compare the outcomes of the competitive design with a traditional design in which participants will be exposed to either the victim or the intruder frame (see method section). Therefore, all hypotheses will be examined in both designs.
Based on previous research we expect that the visual victim and intruder frame affect attitudes towards refugees. Because the victim frame has an inherent positive and the intruder frame a negative valence, we expect that this valence will result in more positive attitudes after exposure to the victim than to the intruder frame. We will therefore examine the following hypothesis:
H2: Exposure to the victim frame leads to more positive attitudes towards refugees than exposure to the intruder frame.
As recent studies on framing effects have found evidence on the role of emotions, we expect that the frames in the current study will evoke emotions too. We will examine emotions that have previously showed to be relevant in the case of immigration (e.g., Lecheler et al. 2015; Igartua et al., 2011; Verkuyten, 2004; Iyer, Leach & Crosby, 2003), which results in the following six emotions: compassion, sympathy, empathy, fear, anxiety and anger. We expect that the victim frame evokes more positive emotions (compassion, sympathy and empathy), since this frame has an inherent positive valence, and that the intruder frame evokes more negative emotions (fear, anxiety and anger), since this frame has a more negative valence. Since previous research has found that emotions mediate the framing effect, we also expect that the evoked emotions mediate the framing effect on attitude in the current study:
H3a: Exposure to the victim frame evokes more positive emotions than exposure to the intruder frame.
H3b: Exposure to the intruder frame evokes more negative emotions than exposure to the victim frame.
H4: Emotions mediate the effect of frame exposure on attitudes.
Method
For this multi-methodological study we first conducted a content analysis to investigate the representation of refugees by analyzing the presence of the victim and the intruder frame. Consecutively an experiment was conducted to examine the effects of these frames and the role of emotions in these effects.
Content analysis Sample
In the content analysis British and Dutch tabloid and quality newspapers were analyzed. The Guardian and The Daily Telegraph were selected as the British quality newspapers and The Daily Mirror as the tabloid newspaper. The Guardian and the Daily Mirror are both considered as progressive newspapers, as they have previously backed the Labour Party during elections, whereas the Daily Telegraph is considered as a conservative newspaper. For the Dutch newspapers, de Volkskrant and NRC Handelsblad were selected as the quality newspapers and De Telegraaf as the tabloid newspaper. De Volkskrant is
considered to be progressive, De Telegraaf as conservative and NRC Handelsblad as center right.
A three-month period was analyzed from the 1st of September 2015 to the 1st of December 2015. This period was selected, because the refugee crisis played a major role in the news during these months. On the 3rd of September a picture was published of a young Syrian boy who was drowned and found dead on a beach in Bodrum, Turkey. Together with his parents he tried to reach the European border by crossing the Mediterranean Sea. This picture played a major role in the refugee debate in Europe and made the crisis an everyday news event, which made this an interesting period to investigate.
Data collection and coding procedure
The newspapers were mainly collected in the public library in Amsterdam. However, not all editions of The Guardian and The Daily Mirror were available in print and were therefore collected through the online archive using Press Reader. Since every visual of the selected period was coded, every newspaper had to be scanned through to look for visuals about the refugee crisis. Visuals were only coded when at least one refugee was depicted. When it was not clear whether the visual depicted a refugee, the visual was not coded. This
resulted in a total of 527 visuals. Before the questions in the codebook had to be answered, the coder first had to look at the visual for about 10 seconds. While answering the questions, the coder was not allowed to look back at the visual. This was done to make sure that the aspects that were coded were clear and obvious visible in the visual instead of small details which readers of the newspaper would not have seen either, since most readers of a
newspaper only just look for a few seconds to a newspaper visual (e.g., Bucher & Schumacher, 2006).
Measures
A codebook was constructed with 15 binary (0-1) items to measure the victim and the intruder frame. These items were mainly based on the study by Van Gorp (2006), who measured the visual frames by the people depicted (men, women and/or children), cinematic elements (close-up/ long shot), presence of emotions and whether the visual depicts
dangerous intruders or poor victims. The textual frames of Van Gorp (2005), however, were measured on numerous items, which could also be transferred to visuals (e.g., refugees in direct danger). Besides, some items from the studies by Batziou (2011) and Bleiker et al. (2013) were also relevant for measuring the victim and the intruder frame (e.g., facial expression). Therefore, we extended and adjusted the visual frames by Van Gorp (2006) to two more detailed frames (see appendix A for the codebook). One additional item (children only) was measured for the victim frame, but was originally not included in the codebook. This item was discovered to be an important addition during the coding procedure, and based on the context variable in the codebook (a variable in which the context of the visual had to be outlined), we determined that the children only item was noticed from the start of the coding procedure and could therefore be integrated easily.
To investigate whether these questions created two reliable and distinct scales, a Mokken scale analysis was performed, as this is the most appropriate method for testing the
reliability of binary items. Based on the Loevinger H coefficients, however, the items did not created two, but five distinct scales. For the victim frame three reliable scales were created: a scale regarding identification, which measured the extent to which the depicted people are identifiable individuals and focuses on their facial expressions (H = .57), danger, which measured the extent to which the depicted people are in danger and/or are victims of violence (H = .45) and poverty, which measured the extent to which people are depicted as poor and/or injured (H = .33). Obviously, when a visual has a high level of danger, there does not
necessarily need to be a high level of identification, or when someone is in danger, he or she does not has to look poor. For the intruder frame two scales were created: no identification, which is the reversed scale of the identification scale for the victim frame, but without the facial expressions, since this is not observable when there is no identification (H = .61) and violence, which measured the extent to which the people depicted are men, perpetrators of violence and/or expressing anger (H = .79). But again the same problem occurred. For
instance, when there is a high level of violence, the people depicted do not necessarily need to be unidentifiable in the visual (See appendix B for a full overview of the Mokken scale analysis).
Although the Mokken scale analysis did not create two distinct scales, we decided to compute the three victim and two intruder scales to two distinct scales and proceeded with the analyses, for the following reasons: (1) the items were all based on literature that measured the items before (either in text or visuals), (2) the five scales individually were all reliable, (3) a second order factor solution showed that the three victim scales and the two intruder scales created two clear scales, which indicated that the underlying scales measured the same.2 These scales, however, were not very reliable α = 0.28 (victim scale) and α = 0.33 (intruder scale), and (4) we measured, besides the actual frames, to what extent the coder perceived each visuals as either an intruder or victim frame. A Pearson correlation indicated that there
was a positive correlation between perceived victim frame and the actual victim frame r = 0.72, p < 0.001, and between the perceived intruder frame and the actual intruder frame r = 0.71, p < 0.001. Therefore, we argued that there is enough evidence for the distinction between the victim and intruder frame. However, some items (e.g., wealth) did not correlate with a single item and were therefore deleted. Because there were 9 items measuring the victim frame and 6 items measuring the intruder frame, the items were computed and divided by the number of items of the frame, which created two equal 0-1 scales.
Intercoder reliability
Even though the coding procedure was done by just one coder, an intercoder reliability test was performed to investigate whether the items in the codebook were clear and with no room for interpretation. Therefore an additional coder was instructed about the codebook and 10% of the total sample was analyzed by both coders. The results of the intercoder reliability test indicated that the agreement between the coders was reliable. The average Krippendorff’s Alpha for the victim frame is α = 0.89 and for the intruder frame α = 0.93 (see appendix C). Data analysis
To measure to what extent the six newspapers differed in the use of the victim and the intruder frame, a one-way analysis of variance (ANOVA) with Bonferroni post-hoc test was performed twice; once for the three Dutch and once for the three British newspapers.
Experiment
Participants
Participants were recruited by a convenience sample, as the URL of the survey-experiment was shared on social media. In total 394 Dutch individuals participated in the experiment. The majority of the participants were female (63.7%) and the ages ranged from 18 to 100 years old (M = 32.02, SD = 13.51), however the vast majority was younger than 30 years old (66.0%), was highly educated (78.4%) and lived in an urban area (63.7%).
Design and procedure
Because in the current study we examined to what extent the design (traditional versus competitive) influenced the results, we set-up three experimental conditions: a victim
condition (n=132) and an intruder condition (n=140) for the traditional design and a
competitive condition (n=122) for the competitive design.3 Before participants were exposed to the experimental condition, they first had to answer questions regarding their demographic factors, political knowledge and issue importance. After, they were randomly assigned to one of the three experimental conditions. The five visuals were presented one by one, and to make sure that participants have been exposed to the experimental condition, they were forced to look for at least six seconds at the visuals. After the exposure, participants had to indicate their emotions and had to answer questions regarding their attitudes towards refugees.
A randomization check was performed between the victim, intruder and competitive condition, and within the competitive condition on age, gender, educational level,
employment status, political stance, political knowledge and issue importance. The results indicated that there were no significant differences between the experimental conditions and within the competitive condition.4
Stimulus material
The stimulus material consisted of 10 newspaper visuals. Five of these visuals had a clear victim frame and five a clear intruder frame. The codebook of the content analysis of the current study was used to indicate to what extent the visuals contained either a victim or an intruder frame (see appendix D for an overview of the visuals). Participants in the competitive condition were not exposed to one frame only, but to five random visuals from the set of 10 victim or intruder framed visuals. However, to make sure that participants in this condition were not randomly assigned to five of the same frames, we programmed the questionnaire so that the maximum of a specific frame was four. By using this design, we could determine
whether an increasing number of exposures to one of the frames lead to stronger emotional and attitudinal changes. In this design, increasing exposure to the victim frame was the same as a decrease in the intruder frame and vice versa (e.g., when a the victim frame increased from three to four, the intruder frame decreased from two to one). In other words, the scales of the frames were reversed from each other. This design increased the external validity, as it came closer to a real-life environment. In the traditional design participants were exposed to the same visuals, but either the five visuals with the intruder frame or the five visuals with the victim frames.
Manipulation check
The manipulation check was measured on two levels. The first level was a factual check in which participants had to indicate whether they were exposed to certain
characteristics of the visuals (four characteristics, e.g., have you been exposed to children?), as the victim and intruder frame both had different distinctive aspects. This was done to check whether participants paid attention to the visuals while they were exposed to the experimental condition. These were the first questions that had to be answered after exposure to the visuals. In the traditional design, it was found that participants in the victim condition (M = 3.70, SD = 0.48) indicated that they had been exposed more to the characteristics of the victim frame than participants in the intruder condition (M = 0.69, SD = 0.82) t(258) = -35.65, p <0.001, and participants in the intruder condition (M = 3.31, SD = 0.82) indicated that they had been exposed to more characteristics of the intruder frame than participants in the victim condition (M = 0.30, SD = 0.48) t(258) = -35.65, p <0.001. In the competitive design, it was found that an increase of exposure to the victim frame was positively related to more correct answers on the question regarding the characteristics of the victim frame r = 0.91, p <.001.
An additional check was performed to indicate whether participants understood the manipulation of the different conditions. To make sure that this check did not influence the
results of the experiment, these questions were the last of the experiment. Participants had to indicate three questions on a 5-point scale (1= victims, 5= intruders) to what extent they perceived the people on the intruder framed visuals as intruders and the people on the victim framed visuals as victims. In the traditional design, it was found that participants in the intruder condition (M=2.35, SD=1.00) perceived the people depicted on the visuals more as intruders than participants in the victim condition (M=1.53, SD=0.72) t (270) = 7.74, p< 0.001. However, no correlation was found in the competitive design between increasing exposure to the intruder frame and the extent to which participants indicated the people depicted on the visual as intruders r= -0.02, p = 0.822. But since the participants in the competitive design were exposed to a mix of frames, exposure to one victim frame might have been enough for perceiving the depicted people as victims rather than intruders, which possibly influenced the results. The significant results for the other manipulation checks allowed us to proceed with this design and the study.
Measures
Emotional responses were measured by six emotions that were relevant in the case of immigration and refugee crises (e.g., Lecheler et al. 2015; Igartua et al., 2011; Verkuyten, 2004; Iyer, Leach & Crosby, 2003). Although in the hypotheses we make a distinction between positive and negative emotions, we will test each emotion individually. Participants had to indicate the following emotional responses on a 7-point scale (1= not at all, 7 = very much): compassion (M = 5.33, SD = 1.53), empathy (M = 5.21, SD = 1.52) and sympathy (M = 4.76, SD = 1.62) as the positive emotions, and fear (M = 2.85, SD = 1.77), anxiety (M = 3.11, SD = 1.91) and anger (M = 4.00, SD = 1.93) as the negative emotions.
Attitude towards refugees was measured by four questions (e.g., refugees are victims of war and suppression) on a 5-point scale (1= strongly disagree, 5= strongly agree) (M= 3.85, SD= 0.61), in which a high score indicated stronger positive attitudes. A factor analysis
indicated that the four attitude questions could be computed, since the questions created one component (eigenvalue = 2.07) with factor loadings ranging from 0.63 to 0.79. A reliability test showed that the four questions created a reliable scale α = 0.68 (see appendix D for an overview of all questions). A normality test indicated that the emotional responses and attitudes were all normally distributed.5
Data analysis
Since the experiment was divided into two designs, all analyses to answer the
hypotheses had to be performed twice; once of the traditional condition (exposure to only the victim versus intruder frame) and once in the competitive condition (exposure to a mix of the victim and intruder frames). For the second hypothesis (H2), an independent-samples t-test was performed to test the differences in attitude between the participants in the traditional design. For the competitive design, an ANOVA with Bonferroni post-hoc test was performed, to indicate whether increasing exposure the victim frame increased positive emotions. For the third hypothesis (H3), for both the traditional as the competitive design, six regression
analyses were performed (one for each emotion), to test whether exposure to only the victim or intruder frame evoked emotions (traditional design), and whether increasing exposure to either the victim or intruder frame evoked emotions (competitive design). For the final hypothesis (H4), it was tested whether the evoked emotions (of H3) functioned as mediators for the framing effect (of H2). To test these mediators, we used the Process macro (Hayes, 2013), which is a tool that makes it possible to perform a multi-mediation analysis in SPSS. This tool is based on an ordinary least squares (OLS) regression analysis for estimating the direct and indirect effects.
Results
Victim and intruder frame
in the newspapers than the intruder frame (M = 0.17, SD = 0.20, 95% CI [0.16, 0.19]) (binary 0-1 scale). However, the confidence intervals (CI) indicated that there was no difference between the two frames within the Dutch newspaper De Telegraaf (see table 1). For all other newspapers the victim frame was significantly more pronounced than the intruder frame.
An ANOVA between the Dutch newspapers indicated that there was no difference in the use of the intruder frame F (2, 285) = 2.88, p = 0.058, 𝜂2 = 0.02. However, for the victim frame there were significant differences between the Dutch newspapers F (2, 285) = 8.05, p < 0.001, 𝜂2 = 0.04. A Bonferroni post-hoc test indicated that NRC Handelsblad used the victim frame significantly more than De Telegraaf Mdifferenc= 0.10, p = 0.001, and de Volkskrant Mdifferenc= 0.07, p = 0.008. A second ANOVA was performed between the British newspapers, but there was no difference observable between de newspapers in the use of the victim frame F (2, 236) = 0.67, p = 0.513, 𝜂2 = 0.01, and the intruder frame F (2, 236) = 1.31, p = 0.272, 𝜂2 = 0.01.
Table 1
Presence of the victim and intruder frame in the six newspapers
Victim frame Intruder frame
Newspapers M SD 95% CI M SD 95% CI De Volkskrant (n= 130) 0.29 0.15 [0.27, 0.32] 0.18 0.19 [0.14, 0.21] NRC Handelsblad (n= 99) 0.36 0.18 [0.33, 0.39] 0.15 0.16 [0.11, 0.18] De Telegraaf (n= 59) 0.26 0.18 [0.21, 0.30] 0.22 0.22 [0.16, 0.28] The Guardian (n= 96) 0.33 0.19 [0.29, 0.37] 0.14 0.19 [0.11, 0.18] Daily Telegraph (n= 65) 0.35 0.23 [0.29, 0.41] 0.20 0.24 [0.14, 0.25] Daily Mirror (n= 78) 0.37 0.22 [0.32, 0.42] 0.18 0.22 [0.13, 0.23] Note: means range from 0 to 1. CI = confidence intervals
In the first hypothesis (H1) it was predicted that the victim frame was used more by quality and progressive newspapers and that tabloid and conservative newspapers focused
more on the intruder frame. However, the results above have indicated that there are just small differences between the newspapers in the use of both frames, and thus we did not found support for the first hypothesis. Despite the fact that we only found minor significant differences, an interesting pattern is observable. The only conservative tabloid (De Telegraaf) used the intruder frame the most and the victim frame the least of all newspapers. Besides, within both countries, the findings indicated that newspapers that focused to the highest extent on the victim frame used the intruder frame to a lower level than the other newspapers,
whereas newspapers that used the intruder frame to the highest extent (although not higher than the victim frame) used the victim frame to a lower level.
Influence of frames on attitude
In an experiment the framing effects were investigated. It was hypothesized that exposure to the victim frame lead to more positive attitudes towards refugees than exposure to the intruder frame (H2). An independent-samples T test showed that, in the traditional design, there was no difference in attitude between participants who were exposed to the victim frame (M = 3.88, SD = 0.63) or the intruder frame (M = 3.82, SD = 0.62), t (268,896) = -0.739, p = 0.461, which indicates that there was no direct effect of the frames on attitudes towards refugees. However, within the competitive condition we found differences between the victim and intruder frame, as we observed a significant ANOVA, F (3, 118) = 3.68, p = 0.014, 𝜂2 = 0.09. A Bonferroni post-hoc test specified that participants who were exposed to one victim frame (and four intruder frames) indicated to have a less positive attitude towards refugees than participants who were exposed to three victim frames (and two intruder frames) Mdifferenc= -0.45, p=0.038, or four victim frames (and one intruder frame) Mdifferenc= -0.68, p=0.019 (one tailed), which indicated that in the competitive condition participants had more positive attitudes when the victim frame is the dominant frame, and less positive attitudes
when the intruder frame is the dominant frame (see table 2). Therefore, we found partly support for this hypothesis.
Tabel 2
The influence of exposure to the victim and intruder frame on attitudes Exposure to the
victim and intruder frame
Attitude towards refugees
N M SD 95% CI 0 victim/ 5 intruder 140 3.83 0.62 [3.72, 3.93] 1 victim/ 4 intruder 13 3.50 0.91 [2.95, 4.05] 2 victim / 3 intruder 44 3.75 0.57 [3.58, 3.92] 3 victim / 2 intruder 55 3.95 0.50 [3.81, 4.10] 4 victim / 1 intruder 10 4.18 0.46 [3.85, 4.50] 5 victim / 0 intruder 132 3.88 0.63 [3.78, 3.99] Total 394 3.85 0.62 [3.62, 4.08]
Note: In this table, both the results of the traditional design (0 and 5) as for the competitive design (1 to 4) are presented.
Influence of frames on emotional responses
In the third hypothesis it was predicted that the victim frame evoked more positive emotions than the intruder frame (H3a). A regression analysis showed that exposure to victim frame increased the positive emotions compassion, empathy and sympathy (see table 3). However, this effect was only observable for participants in the traditional design. In other words, these emotions were only evoked when participants were exposed to only the victim or intruder frame. When participants were exposed to a mix of frames, these emotions were possibly neutralized, since we did not observed any differences in the competitive design. Therefore, we found partly support for this hypothesis.
In the second part of third hypothesis it was predicted that the intruder frame evoked more negative emotions than the victim frame (H3b). However, a regression analysis
same in both experimental designs (see table 4). Therefore, we did not found any support for this hypothesis.
Table 3
Positive emotions after exposure to the victim frame.
Compassion Sympathy Empathy
Frames B SE β B SE β B SE β Traditional design Victim frame .860*** .190 .267 .591** .197 .180 .625** .189 .198 R2 .071 .032 .039 Competitive design Victim frame .061 .152 .037 .110 .170 .059 -.021 .151 -.012 R2 .001 .004 .000
Note: This table shows the results of a regression analysis. ***p<.001, **p<.01, *p<.05. Table 4
Negative emotions after exposure to the intruder frame
Fear Anxiety Anger
Frames B SE β B SE β B SE β Traditional design Intruder frame .015 .214 .004 .065 .227 .017 -.271 .241 -.068 R2 .000 .000 .005 Competitive design Intruder frame -.161 .207 -.072 -.072 .229 -.029 -.011 .207 -.005 R2 .005 .001 .000
Note: This table shows the results of a regression analysis. ***p<.001, **p<.01, *p<.05.
Emotions as a mediator for framing effects
In the fourth hypothesis (H4) it was predicted that the evoked emotions functioned as mediators for the main framing effects. Mediation can be tested through multiple statistical
methods, but the most common technique is still the causal step approach by Baron and Kenny (1986). However, in the first step of this approach, the authors argued that a mediation model requires a direct effect from the independent (X) on the dependent variable (Y).
Following Baron and Kenny (1986), we had to conclude that emotions were not a mediator (M) for attitudes, because (1) in the traditional design we did not found a direct effect of the frames on attitudes, and (2) although in the competitive design a direct effect was found, in this design the frames did not evoke any emotions, and these emotions were perceived as the mediators. However, in two recent studies the mediation model by Baron and Kenny (1986) was revisited, and it was argued that for a mediation model there is no significant direct effect needed from X to Y but only significant effects from X to M, and from X to Y when
controlling for M (Hayes, 2009; Zhao, Lynch & Chen, 2010). According to Zhao et al. (2010), a mediation model without a direct effect is an indirect-only mediation. Nevertheless, Mathieu and Taylor (2006) suggested making a distinction between mediation and indirect effects, where mediation refers to a model in which there is a significant direct effect of X to Y, and an indirect effect refers to a model in which there is no direct effect. Following Mathieu and Taylor (2006), we thus cannot have a mediation model, as we did not found a direct effect of X to Y. But as can be seen in figure 1, we found an effect of the frame on the three positive emotions (X to M), and from the frame on the positive attitudes through compassion and sympathy (X to Y, controlling for M), which indicates an indirect effect. To examine whether the indirect effects were significant, we used the bootstrapping method with five thousand bootstrapping samples. When a 95% BCa CI did not contain 0, we determined the effect as significant. We thus found a significant total indirect effect through compassion, sympathy and empathy (b = 0.136, SE = 0.048, 95% BCa CI [0.06, 0.25], and indirect effects through compassion (b = 0.060, SE = 0.035, 95% BCa CI [0.01, 0.15]) and sympathy (b = 0.044, SE = 0.023, 95% BCa CI [0.01, 0.11]). However, no significant indirect effect was
THE IMAGES OF REFUGEES 30 Frame Attitude Compassion Sympathy Empathy c’ (c) path = -.081 (.055) b path1 = .071* b path2 = .077** b path3 = .051 a path1 = .839*** a path2 = .574** a path3 = .625** a1 a2 a3 0 = Intruder frame 1 = Victim frame b3 R2 = .20 F (4, 265) = 16.51 b1 b2
observed for empathy (b = 0.032, SE = 0.024, 95% BCa CI [-0.004, 0.091]). Because the effect size was the largest for compassion, this emotion had the strongest influence on attitude. Since the victim and intruder frames were binary variables, this figure shows that exposure to the victim frame evoked compassion, sympathy and empathy, and the former two emotions affected attitudes towards refugees positively. Although the findings indicated only an indirect effect for the traditional condition, we found support for the fourth hypothesis.
Figure 1: This figure shows the results of a multi-mediation model. The three paths show unstandardized (OLS)
regression coefficients. ***p<.001, **p<.01, *p<.05. The solid lines are significant and the dashed lines nonsignificant.
Discussion
The starting point for this study was the current refugee crisis in which hundreds of thousands citizens from the Middle East and Africa have fled from war and suppression. Because this crisis reached immense proportions, it has become an everyday news event. Previous research showed that the representation of refugees in the media is mainly negative and affects political attitudes. However, the media landscape is changing, as there is a rise in visual communication. In the current study we have therefore investigated the visual representation of refugees and the effects of these visuals on emotional responses and attitudes.
By conducting a content analysis, we examined to what extent visuals of the refugee crisis were framed in terms of refugees as victims or refugees as intruders, and expected that the victim frame was used more in quality and progressive newspapers and the intruder frame more in tabloid and conservative newspapers (H1). The findings showed that newspapers portrayed refugees in visuals mainly as victims by focusing particularly on identifiable families who expressed distress, fear and misery, instead of portraying them as threatening intruders by focusing on unidentifiable and anonymous groups of violent and angry men. These results are in line with the studies by Van Gorp (2005, 2006) who also found that refugees were portrayed as victims rather than intruders. However, since the victim frame has an inherent positive valence and the intruder frame a negative valence, the results of the current study are contrasting to the majority of research on the visual representation of refugees, as these studies concluded that refugees were mainly negatively portrayed. Interestingly, the (mainly) positive representation of refugees that was found in the current study is in line with the shift by the German newspaper Bild. Coole (2002) found that newspapers could shift their tone of coverage after a tragic or emotional event. We studied a time period just after the tragic dead of the young Syrian boy, which thus possibly made the newspapers focus mainly on victims. Although we cannot conclude whether newspapers shifted, since we did not compared the findings with data from before the tragic event, the theory by Coole (2002) can be the explanation for the predominantly focus on the victim rather than the intruder frame. When comparing newspapers on type and political stance, the findings showed that there was no newspaper that focused more on the intruder than on the victim frame. Within the Dutch newspapers, however, NRC Handelsblad used the victim frame more than de Volkskrant and De Telegraaf, but no differences were observed for the intruder frame. Within the British newspapers there were no differences between the newspapers in the use of the victim and intruder frame. Although only small significant differences were observed, we found an interesting pattern in the use of the victim and intruder frame between British and Dutch
newspapers. Newspapers that focused to the highest extent on the victim frame used the intruder frame to a lower level than the other newspapers, and vice versa. Interestingly, the progressive quality newspapers (de Volkskrant and The Guardian) use neither of the frames to the highest extent, and The Guardian uses both frames to the lowest extent of the British newspapers, which indicates that these newspapers are possibly less ‘outspoken’ in the visual portrayal of refugees The Dutch conservative tabloid De Telegraaf used the intruder frame to the highest extent of all newspapers, and did not used the victim frame more than the intruder frame. This pattern
indicated that there are differences between the newspapers in the expected direction. But as these differences are not significant, we did not found support for the first hypothesis. Nevertheless, to answer the first research question, we conclude that (1) the victim and intruder frames exists next to each other, but that (2) overall the victim frame is more pronounced in British and Dutch newspapers than the intruder frame.
Consecutively, we investigated to what extent exposure to the victim and intruder frame affected attitudes towards refugees, and examined the role of emotions in this effect. We hypothesized that exposure to the victim frame led to more positive attitudes towards refugees (H2) and evoked more positive emotions (H3a) than exposure to the intruder frame, whereas exposure to the intruder frame evoked more negative emotions than exposure to the victim frame (H3b), and that these emotions mediated the effect of frame exposure on attitudes (H4). The findings indicated that the representation influenced people’s attitude. As hypothesized it was found that exposure to the victim frame led to more positive attitudes towards refugees than exposure to the intruder frame. This direct effect, however, was only found for participants in the competitive experimental design, as we did not found a direct framing effect between participants in the traditional design. This means that an increase of the victim frame (and a decrease of the intruder frame) led to more positive attitudes, but only when this frame was presented next to the intruder frame. When participants were exposed to the victim frame only, their attitude towards
refugees did not differ from participants who were exposed to only the intruder frame.
Interestingly, these findings were in line with a study by Nisbet et al. (2013), who also found that participants in the competitive condition indicated stronger framing effects. In their study this finding was opposed to their hypothesis and they thus did not provide a clear explanation. Nevertheless, they argued that exposure to contrasting frames possibly changes the framing process, as participants have a “wider range of considerations in a deliberative and systematic manner” (p. 778). In other words, when someone is exposed to two frames, there seems to be a wider range to weigh the overall ‘costs and benefits’ of the refugee crisis. The dominant frame subsequently has more influence on someone’s attitude, than when someone is exposed to one frame only. Moreover, Messaris and Abraham (2001) argued that frames in visuals are less obtrusive than in text, which makes visuals effective vehicles for framing. However, when participants are exposed to five visuals with the same frame, the manipulation possibly does become obtrusive and makes participants aware of the framing process. Thus, when participants are exposed to opposing frames, but with one dominant frame, the manipulation becomes less obtrusive which makes participants less resistant for the framing process, and therefore the dominant frame becomes more effective.
As expected, we found that the victim frame evoked more positive emotions than the intruder frame, but did not found support for the hypothesis regarding the negative emotions. The victim frame, however, only evoked more positive emotions than the intruder frame when
participants were exposed to the victim frame only (traditional condition). When participants were exposed to a mix of frames (competitive condition), the emotions were possibly neutralized. In other words, the positive emotions evoked by the victim frame canceled the negative emotions evoked by the intruder frame and vice versa. Moreover, we found support for the fourth hypothesis, as the findings indicated that the evoked positive emotions (compassion and
previous literature (e.g., Gross, 2008; Lecheler et al., 2015; Powell et al., 2015). But since we only found that the victim frame evoked more positive emotions within the traditional design, the indirect effects were only found in this condition. To answer the second research question, we thus found that in both experimental designs the frames affected attitudes towards refugees. In the competitive design, however, we found a direct effect in which emotions did not play a role at all, whereas in the traditional condition there was no direct effect observable, but an indirect effect through compassion and sympathy. But since the content analysis indicated that citizens will only rarely (or even never) be exposed to only one frame, in a real-life environment these emotions will probably not be as strong as in a traditional experimental design, and therefore we question the generalizability of the emotions. But either way, the visual frames have shown to influence attitudes towards refugees, and we thus found evidence that visuals are effective vehicles for framing.
Usually, experiments are judged for their low external validity. But since in this multi-methodological study a content analysis first indicated that there was a distinction in the visual representation between refugees portrayed as victims and intruders, we could use this distinction in the experiment as well. Because we showed that these frames in reality were present in the media, the external validity of the experiment increased. Furthermore, the content analysis indicated that both the victim and the intruder frames existed next to each other. Because we employed an experimental design in which participants were exposed to both frames as well, the findings of the experiment came closer to a real-life environment, which also increased the external validity.
Although we found interesting results, this study has also some limitations. In this study we computed a victim frame and intruder frame variable based on 15 binary items that we measured in a content analysis. Although we had theoretical considerations to compute these items, a Mokken scale analysis indicated that, statistically, the items did not created two distinct scales.
Besides, we compared the presence of the victim and intruder frame and concluded that the victim frame was more dominant in the media than the intruder frame. But since these frames were measured by different items and focused on different characteristics of the visuals, these frames might not be comparable. Therefore, the findings give a clear indication of the presence of these frames, but they have to be taken with some caution. Moreover, we had an appropriate sample size for the experiment, but within the competitive design participants were divided over four
conditions, which resulted in (mainly two) smaller conditions compared to the conditions in the traditional condition. Therefore, in the comparison between the victim and intruder frame, we compared a large versus a small sample, which could have affected the results.
We have shown that newspapers were predominantly positive towards refugees and argued that the newspapers might be influenced by the tragic events that happened at the beginning of September. But since we did not compared two time frames, we cannot determine whether newspapers shifted to a more positive tone, or that this positivity was their nature. Therefore, we recommend for a future study to focus on a wider time frame to examine the influence of
emotional and tragic events on the representation of refugees. Moreover, we only found small significant differences in the use of the victim and intruder frame between the newspapers, but did observe an interesting pattern, which indicated that newspapers did use different visuals.
Therefore, we suggest for a future study to examine the visual representation of refugees more in-depth and between a wider array of newspapers.
We finally want to emphasize the importance of the experimental design in which framing effects are investigated, as we did found different framing effects between the competitive and traditional experimental design. Because the former design comes much closer to a real-life environment, we suggest to focus more on competitive designs in future experimental studies. More specifically, it would be essential to find an explanation for why a competitive design showed to have direct framing effects, whereas these effects were not observable in a traditional
design. Because if we find more empirical evidence for the influence of competitive frames, in a real-life environment, frames might have even more influence than previous studies suggested.
In the introduction we claimed that this study was societally and theoretically relevant because there was not much knowledge about the visual representation of refugees and the effect of different portrayals on attitudes. Because the findings of this study have shown that visuals play a crucial role in shaping citizens’ attitudes about refugees, and that the choice of a newspaper thus affects someone’s image of refugee-issues, this study brought more knowledge in the field of migration and framing research, and makes an important contribution to the literature.
