More Than Words?: A Test of the Effect of Emotionally Charged Photographs

On September 2, 2015, the body of a three-year old Syrian boy, Alan Kurdi, washed up on a beach near the Turkish city of Bodrum. Along with his parents and older brother, he had been on board a rubber boat trying to make the short voyage from Turkey to the Greek island of Kos. At the time of the accident, European debates about the ‘refugee’ or ‘migration’ crisis had been raging for months. Drowning fatalities were not unprecedented: In one weekend alone in April 2015, almost one thousand people drowned as they tried to cross from Libya (Yardley 2015). Yet Kurdi’s death was a game changer. The reason: a set of photographs by news photographer Nilüfer Demir, the most powerful of which show Kurdi lying face down in a sleep-like position on the beach. Within hours, the photographs were spreading through social media, and from there into the mainstream press all over the world (Rogers 2015). Countless reactions of indignation, most of them evoking Kurdi to advocate for more humane responses to the refugee crisis, followed in their wake as they made their way through the global public sphere (D’Orazio 2015). The responses were not merely symbolic: An Independent (Merrill 2015) news story the day after Kurdi’s death reported how relief and solidarity organisations experienced a surge in donations and citizens offering help.

In recent years, studies of activism and political participation have seen growing interest in the relationship between photographs and activism (e.g. Biggs 2005; Martin 2005; Arpan et al. 2006; Gray & Martin 2008; Greer & McLaughlin 2010; Halfmann & Young 2010; Corrigall-Brown & Wilkes 2012; Olesen 2013, 2017; Doerr et al. 2013; Teune 2013). Part of a broader visual turn in the humanities and social sciences since the early 1990s, this development has been a response to the near exclusive focus on text that had previously characterised the field (Doerr et al. 2013). While this lacuna is increasingly being addressed, it is routinely acknowledged that research on visuality and photography is still only nascent.

The field’s infancy is particularly evident when one searches for answers to how much and under what conditions photographs matter for activism and participation. The majority of contributions seem premised on the assumption that they do matter: It is routinely implied that they have an impact on recipients/viewers by affecting opinions and knowledge, and mobilising motivations. However, such causalities are rarely documented (but see Arpan et al. 2006), and when it comes to what is arguably one of the most central questions in the field, why some people act and participate, and others do not, there is an absence of systematic knowledge on the impact of photographs. Martin’s (2005; Gray & Martin 2008) work on the role of photographs in creating backfire, Jasper and Poulsen’s (1995) notion of moral shocks induced by photographs, and Halfmann and Young’s (2010) analysis of grotesque images are all inspiring and suggestive, but they do not provide solid data on activism and participation effects.

A reason for this gap probably resides in the obvious methodological challenges in documenting such effects. Surprisingly, activism scholars have only adopted the experimental method widely employed in psychology (e.g. Burt & Strongman 2005; Chang & Lee 2009), advertising studies (e.g. Gamliel & Kreiner 2013), and political communication and framing research (e.g. Pfau et al. 2006; Brantner et al. 2011) to a limited extent. This article seeks to advance the literature along these lines by offering a large-scale experimental study on the impact of visual framing on activism and participation in relation to the refugee crisis. The overall research question is: To what extent do photographs of suffering refugees in relation to the 2015 refugee crisis affect individual willingness to participate in pro-refugee activities?

While framing theory is now applied across a number of fields in sociology, political science, and communication, framing is predominantly understood as a textual or verbal activity. This is also the case within activism and participation studies, where framing theory has held a central position since the 1980s (e.g. Benford & Snow 2000; Snow et al. 2014). In recent years, however, scholars, mainly from outside the field (but see Corrigall-Brown & Wilkes 2012), have sought to expand the concept to encompass what is generally referred to as visual framing (e.g. Parry 2010; Brantner et al. 2011; Rodriguez & Dimitrova 2011). Visual framing works on the basis of the same logic as textual/verbal framing in that it directs the viewer’s attention to certain aspects of a situation or event and shapes the way they are understood (Entman 1993; Benford & Snow 2000). Visual framing may occur on its own, but typically, photographs appear in accompanying or supportive relationships with text. This is how photographs are used in the news, but also in activist calls for action. In such cases, text and photographs engage in a multimodal dialogue in which they mutually shape opinion processes (Kress 2010; Geise & Baden 2015).

The following briefly reviews the existing work on visual framing outside the field of activism and political participation, mainly in media, communication and political communication studies, with a view to furthering our understanding of visual framing in calls for political action.

While not all the studies referred to below employ the concept of visual framing, the term is used here to encompass all studies that focus on the effects of photographs and visuals on meaning and opinion formation.

Several studies confirm strong effects of visual framing on various dependent variables. Brantner and colleagues (2011) find that news stories accompanied by photographs of Palestinian conflict victims generated stronger emotional responses than when text appeared alone or in combination with a photograph of politicians. In a study of how photographs affect perceptions of terrorism, Iyer and his colleagues (2014) show that different photographs produce different emotional responses and preferred solutions. When respondents were exposed to photographs of terrorists, rather than victims, emotional responses were fear and anger, which, in turn, led to a preference for strong counterterrorist policies. Arpan and Tüzünkan (2011) study the effect of deviance-oriented photographs of protesters and conclude that such photographs produce more negative evaluations of protestors than non-deviance oriented photographs (for other studies documenting effects, see, for example, Gibson & Zillmann 2000; Arpan et al. 2006; Small & Verrochi 2009).

There are various explanations as to why photographs affect opinion formation. First, whereas text is sometimes seen as subjective, strategic, and even manipulative, photography is considered by many as more documentary, i.e. as offering an undistorted window into the empirical world ‘as it is’ (Barthes 1977; Sontag 1979; Messaris & Abraham 2001). Second, the power of photography lies in the ability to activate immediate emotional responses and, at least momentarily, to bracket cognitive meaning formation (Iyer & Oldmeadow 2006; Brantner et al. 2011).

This is also supported at a more general level within psychological research (e.g. Zajonc 1980).

To test the hypothesis, data were collected in two steps. In the first, an online pre-test survey (November 2015) asked 252 Danes (the responsible survey institute ensured distribution on variables such as political persuasion, education, income, and gender that reasonably mirror real-world distributions) about their reactions to six photographs depicting different scenes and situations related to the 2015 refugee crisis.

Both surveys were organised and conducted by Epinion, a major Danish survey institute. Data files from the pre-test and main surveys can be acquired by contacting the author.

The photographs were identified in Google’s image search machine and selected based on two further criteria: first, that they were not already widely known, i.e. did not come with more or less pre-established meanings; and second, photographs with graphic displays of death were avoided in order not to force a reaction or propel viewers to turn away in disgust and disengage from the survey. The photograph eliciting the strongest reactions was selected for the main survey in order to maximise the potential photography effects.

In the second step, an online survey experiment with approximately 4,000 respondents (see above on representativeness) was conducted (December 2015). The experiment consisted of four conditions randomly distributed among the respondents. In condition 1 (treatment), the respondents were presented with a short ‘strong’ (using colourful language and a slightly indignant tone) text describing some of the problems suffered by refugees in the current crisis.

The strong text reads: “The civil war in Syria, now in its fifth year, has forced millions to flee from suffering and misery. In excess of 12 million Syrians – more than half the entire population – are now refugees. Around 8 million Syrians are internal refugees, while more than 4 million, according to the UNHCR (the United Nations High Commissioner for Refugees), have fled to neighboring countries such as Jordan, Lebanon, Iraq, Egypt, and Turkey – and to Europe. Since 2011, 500,000 Syrians have sought asylum in Europe, including around 13,000 in Denmark. The consequences of this flow of refugees are enormous. Families are torn apart and ruined and children traumatized. And almost every day, refugees die trying to cross the Mediterranean. The majority are now stuck in camps in neighbouring countries without any prospects for a better life somewhere else.”

In condition 3 (treatment), the respondents received a ‘weak’ (purely factual and neutral in tone) text, as well as the photograph included in condition 1.

The weak text reads: “The civil war in Syria is now in its fifth year and in excess of 12 million Syrians, i.e. more than half of the entire population, are now refugees. Around 8 million Syrians are internal refugees, while more than 4 million, according to the UNHCR (the United Nations High Commissioner for Refugees), have fled to neighboring countries such as Jordan, Lebanon, Iraq, Egypt, and Turkey – and to Europe. Since 2011, 500,000 Syrians have sought asylum in Europe, including around 13,000 in Denmark.”

The rationale behind using both a strong and weak text was to test whether photographs may have stronger effects where the text does not actively attempt to shape opinion and garner indignation. The texts attempt to mimic parts of a call from an NGO attempting to persuade citizens to act in favour of refugees in one way or another. However, none of the texts mention specific actors in order to avoid a sender bias in which the respondents engage or disengage with participation calls based on sympathy for the sender of the message (e.g. Slothuus 2010).

Following the exposure, the respondents were asked to what extent they would be willing (on a five-level Likert-scale) to (1) take part in a demonstration to pressure Danish politicians to receive more refugees in Denmark; (2) sign a petition to the same end; (3) donate DKK 150 (approx. EUR 20) to support refugees. The dependent variable was split into three options to make it possible to test, for example, if photographs are more likely to motivate ‘weak’ and less demanding forms of participation (i.e. donating money) as opposed to ‘strong’ and more political forms such as demonstrations. These questions were followed by a battery of questions on variables such as political persuasion (voting at the most recent election), education, political interest and gender.

As mentioned earlier, the first step in the data collection process was to select a photograph for the experimental survey. The respondents were shown six photographs and then asked to select the one that emotionally impacted them the most. The procedure was repeated with the remaining photographs and so on until all the photographs had been selected. The results in Figure 1 show which photographs were selected by the respondents in the first round.

Figure 1

Photograph 2 was chosen by 30 per cent, with photograph 6 (28 per cent) coming in second. On this basis, photograph 2 was selected to accompany the texts in the main survey. Following the categories represented in Table 1, this photograph combines high suffering/conflict and a focus on children.

Figure 1 shows how the respondents prioritised between the photographs, but it does not say anything about the degree of impact. To probe this, the respondents were asked to assess (on a scale from 1 to 10) how much the photograph chosen in the first round impacted them emotionally. Collapsing 8–10 on the scale, Figure 2 shows that more than two thirds of the respondents (69 per cent) reported a very high or high degree of impact. This information is important in order to contextualise the findings from the main survey. It tells us that individuals do in fact feel the power of photographs. Whether this is also sufficient to affect their willingness to act is another question.

Figure 2

Figure 3 presents the overall results of the main survey for each of the four conditions and the three participation forms. The diagrams show the collapsed percentages of individuals expressing ‘a high degree’ or ‘some degree’ of willingness to participate in each of the activities (the respondents were asked to rate their willingness on all three participation forms, and percentages can therefore sum up to more than 100 per cent). The differences between participation forms confirm the theoretical expectation that there is a participation hierarchy determined by degrees of time investment and politisisation.

However, as is immediately evident in Figure 3, the response patterns in the four conditions are almost identical, and an ANOVA test shows that all differences are insignificant. Thus, adding a photograph to the text did not turn out to have a positive effect on individuals’ willingness to participate.

Figure 3

Furthermore, the small differences between the conditions that are in fact detected are not systematically in the expected direction of photography effects. Comparing conditions 1 and 2, there is, in fact, slightly more support in condition 2 (i.e. text without photography). Comparing conditions 3 and 4, there are differences in the direction of photography effects, but they are weak and insignificant. In sum, the inclusion of a photograph with a text makes no significant difference for individual willingness to participate in any of the three activities (but see the qualifications in the discussion section). This is true for both the strong and weak text conditions.

While Figure 3 demonstrates a lack of effects at the aggregate level, there may potentially be effects within intervening variables such as political persuasion, political interest, education, and gender. It might seem paradoxical to expect positive effects within these variables, as that would presuppose a reverse effect for other respondent categories within the relevant variable. If, for example, individuals on the Left side of the political spectrum turn out to be affected by photographs, the non-effect at the aggregate level would necessarily imply that individuals on the Right are negatively affected (i.e. inclined not to act because of the photograph). However, to firmly conclude a non-effect, this possibility was investigated.

Relating willingness to different background variables, the results show considerable variation in who is willing to participate. Centre-Left political orientation, high to some political interest, high educational level, and gender (female) are strong predictors of participation willingness. However, when comparing across the conditions, there are no visible signs of photography effects within the variables (the results are not reported here but are available from the author). Furthermore, the results indicate a lack of systematism in the direction of photography effects, with the photograph-included conditions sometimes having higher degrees of support inclination, and in other cases, less. Even if we see small differences in the direction of effects in some categories within the intervening variables, the lack of systematism serves to underline the overall findings, i.e. that photographs do not affect participation inclination in a consistent manner.

The findings stand in some contrast to Figure 2, where the majority of respondents indicate that the photographs they selected did, in fact, have a strong emotional impact on them. Yet it appears that this impact is not significant enough to increase the willingness to participate in any of the participation forms. The unequivocal nature of the results is interesting when considering the widespread assumption in activism and participation studies that photographs do matter, and the rather strong support for effects discussed in the theoretical section on visual framing. The following offers some potential theoretical explanations for this outcome. They focus on the importance of timing, resistance, and saturation.

Even studies that confirm photography effects often point to their conditionality. Furthermore, some scholars highlight no, limited or even reverse effects of photography. Starting with the conditionality argument, Small and Verrochi (2009), for example, analysed the impact of the facial expressions of victims on viewer sympathy and concluded that certain emotional expressions, i.e. sadness (compared with neutral or smiling faces), was the strongest predictor of sympathy. Further, Arpan and colleagues (2006), while demonstrating significant effects of protester photographs on audience perceptions, found that the effects were strongest on those viewers/readers who had a relatively high interest in the relevant protest issue (see also Geise & Baden 2015 for a discussion of conditionality). Effects and their intensity thus vary considerably across both photograph (Small & Verrochi 2009) and recipient characteristics (Arpan et al. 2006).

Another set of works go even further and document no, limited, or even reverse effects of photographs. In an early intervention, Isen and Noonberg (1979), for example, found that the willingness to donate to handicapped children was negatively affected by the inclusion of photographs of handicapped children. Analysing Third World issue donation, Dyck and Coldevin (1992) concluded that the inclusion of photographs in appeals did not positively affect donation levels. Further, Domke et al.’s (2002) analysis of the impact of a well-known photograph of a protestor killed during anti-Vietnam protests at Kent State University in 1970 demonstrated that, although the photograph unequivocally documented death and suffering on the part of the protestors, it did not steer sympathy towards the protestors, but rather towards the government and the authorities.

These findings suggest that the effect explanations discussed earlier, the one-to-one relationship between photographs and reality, emotional impact, and picture superiority, do not automatically produce positive effects. Photographs, and perhaps especially those depicting suffering (Halfmann & Young 2010), may even work in reverse and cause disengagement or resistance. When a photograph is shown in the context of a question about, for example, the willingness to act, it may be viewed as obtrusive and aggressive, i.e. as an attempt to force a certain emotionally driven evaluation on the viewer (Isen & Noonberg 1979; Domke et al. 2002). While not necessarily questioning the empirical reality of the photograph, it is considered to portray the issue manipulatively and one-sidedly.

Such resistance and scepticism is likely to be strongest when citizens have crystallised opinions. Framing research (e.g. Zaller 1992; Chong & Druckman 2010; Druckman & Leeper 2012), for example, suggests that framing effects diminish when citizens have developed relatively strong opinions on an issue. In other words: Timing matters. There are reasons to expect similar dynamics in visual framing. While broader in scope, the research on the CNN effect (i.e. the impact of television on foreign policy) supports this assumption. It argues that television effects are most notable in situations of uncertainty, i.e. when citizens and policy makers are unsure about the character of a critical event/situation and how it should be addressed (Robinson 2002). As opinions mature and crystallise, a set of competing frames for making sense of the issue develop and become available.

To the extent that the respondents have already ‘bought’ one of these frames, photographs may have a limited effect on opinion formation. For those in some/full agreement with the normative line in a multimodal text-photograph constellation, the photograph will be confirmative. For those in disagreement, it will not be strong enough to alter opinion.

The timing aspect may not only explain resistance and scepticism but also neutral and/or disengagement reactions. This may be the case, for example, when individuals experience prolonged exposure to photographs of suffering. In a study of Swedes’ reaction to images from the Balkan Wars in the 1990s, Höijer (2004) saw a certain compassion fatigue (e.g. Moeller 1999) and saturation as news reporting on suffering continued over several weeks. This dynamic is likely to work in interaction with the resistance dynamic. Simply put: It is easier to resist a photograph when it is not novel or shocking.

As noted earlier, the survey experiment was conducted in December 2015. At that time, the debate on refugees had been active for several months. The most intensive period was September (see Figure 4). The September debate, moreover, was characterised by a relatively strong humanitarian perspective emphasising solidarity with the refugees and a high level of policy uncertainty, with both Danish and international political actors debating responses to the unprecedented refugee wave.

Figure 4

By December, the debate intensity, as well as the humanitarian perspective, had receded. In terms of interpreting the survey results, we may expect individual opinions to have crystallised and stabilised at this time. During the fall 2015, Danish political parties passed from a period of policy uncertainty to developing positions and profiles on the matter. Similarly, individuals had had time to debate the issue with friends and family and, with the parties finding their feet, they had also found clearer navigation posts in the political landscape. It is plausible to suggest that by the time of the survey, opinions had reached a stable state where emotional stimuli in the form of photographs were unable to make a significant difference in the willingness to participate on behalf of refugees.

The crystallisation-over-time dynamic was potentially amplified by saturation dynamics. As suggested by Höijer (2004) above, prolonged periods of exposure to suffering may create a certain degree of compassion fatigue. During the intense debate month of September, readers of Danish newspapers and social media users were exposed to countless photographs documenting the plight, suffering, and distress of refugees. The photographs of Alan Kurdi, briefly discussed in the introduction, were only the most widely publicised in a steady stream of photographic documentation. By the time of the survey in December 2015, it is thus probable that the respondents had reached a certain saturation point. This may appear to contradict Figure 2 to some degree, which reports a strong emotional impact. However, stating that photographs have an emotional impact does not necessarily translate into an opinion affecting impact. In fact, we might speculate that opinion altering effects mainly derive from novel/unexpected photographs.

Figure 5 condenses the explanatory account for the observed non-effect. Of course, this account cannot be verified within the parameters of the present study. What it does is only suggest the possibility that if the survey had been conducted in, for example, October 2015, we might have observed at least some effect. We do not, however, have any way of knowing whether the results might, in fact, have been the same even in a different temporal setting. Thus, the credibility of the explanatory account suggested here resides primarily in its theoretical grounding (see below for a discussion of methodological strategies for a stronger test of the hypotheses indicated in and by the account).

Figure 5

Some qualifications are in place to contextualise the strong non-finding reported here and to avoid a complete rejection of the power of photography. First, the results do not rule out the possibility that the respondents willing to participate on behalf of refugees had been motivated in this direction by photographs at an earlier stage of their opinion formation. As noted in the introduction, several organisations reported an increase in interest and donations after the publication of the Alan Kurdi photographs in early September. Put differently: Perhaps the power of photography had already done its job by December. This interpretation is consistent with the saturation argument. Second, the data do not allow us to see if some respondents who were perhaps reluctantly willing to participate were strengthened in this conviction by the photograph, i.e. the data do not enable us to distinguish between degrees within the willingness ratings. Third, we cannot disregard the possibility that some respondents were actually affected by the photograph, but that this effect was cancelled out by respondents who were affected in a reverse way, i.e. expressed unwillingness to participate because of the photograph. As discussed above, the latter reaction finds some support in the literature. Isen and Noonberg (1979), for example, suggest that photographs may be experienced as manipulative and attempting to force a certain response. The results from the pre-test lend some support to this line. In it, the respondents were asked to briefly describe how they experienced the six photographs. While the large majority used this space to describe emotions of sorrow, despair, anger, etc., some also expressed that they felt a certain manipulative drive in the exposure to the photographs. Fourth, a more method-internal explanation might suggest that the photographic stimulus was not strong enough. In other words, we cannot be sure that a stronger and more graphic photograph would not have generated an effect on participation willingness.

The qualifications gesture towards some weaknesses in the experimental method. Experiments are snapshots and poorly equipped to handle context and process. A possible solution would be to conduct several experiments over an extended period to map out how photographs may matter differently in different temporal contexts. While such a strategy involves serious challenges in regard to cross-time comparability, there is really no way around it if we wish to take the time dimension seriously. Another route would supplement experimental data with other forms of data collection. As indicated above, some types of information that might actually suggest a photography effect are simply not visible in the experimental data. This blindness could be balanced by following up on the experiment with focus group interviews in which the participants are invited to collectively reflect on the role of photographs in their opinion formation processes.

These qualifying remarks caution us not to generalise and draw overly negative conclusions about the power of visual framing based on the study’s findings. The results do, however, motivate a more conditional, contextual, and, not least, sceptical-realist understanding of the power of photography in relation to activism and participation. As suggested earlier, there are several relevant conditionalities to take into account. This study primarily indicates the need to consider the time dimension. In different ways, all the explanatory factors in Figure 5 call for incorporating time. Students of political participation and activism need to take such sensitivities and conditionalities on board. As discussed in the introduction, the literature on activism and participation seems to assume quite significant effects, but without providing firm evidence and detailing the conditions under which photography matters (and does not matter).

The explanatory account in Figure 5 emphasises the role of time and context in explaining the lack of effect. An alternative explanation, with special relevance for students of participation and activism, would point to the character of the dependent variable: participation. It was noted earlier that the strong non-finding of the article is surprising given the rather consistent documentation of visual framing effects in the literature; yet it is worth pointing out that none of these works explicitly test the power of visual framing on political participation. It might be considered whether participation is, in fact, a very hard test of visual framing in the sense that it is a much more moral-political commitment and investment heavy decision/response than, for example, expressing sympathy, trust, indignation, etc. (commonly employed dependent variables in the study of visual framing). If this is indeed the case, we might entertain the idea that even in a different temporal and political context, the impact of photographs on participation would have been negligible. This ‘suspicion’, which is of utmost importance for advancing our knowledge, could be addressed through some of the methodological strategies suggested above, but also by including alternative – less commitment and investment heavy – reaction variables in future experiments.

The article presents the results of a large-scale experimental study of the effects of photographs on individuals’ willingness to engage in political participation. In general, the results demonstrate that there is no effect. The respondents do state that photographs have a significant emotional impact on them, but the data suggest that this impact is not sufficient to significantly alter the willingness to engage in pro-refugee political participation. The article points to the timing of the survey as a potential explanation. At the time of the survey in December 2015, opinions had crystallised, policy uncertainty had diminished, and a certain compassion fatigue had set in that together limit the effect of photographs on opinion formation. The findings are highly relevant for participation and activism studies, which tend to assume rather strong effects. These insights should not lead us to dismiss the power of photography as a whole. Rather, the article calls for more theoretical and analytical sensitivity to the conditional and contextual nature of photography effects. We would do well, therefore, to heed the concern of Druckman and his colleagues (2011) about what they refer to as a publication bias in experimental studies within political science and sociology, i.e. a preference for studies that demonstrate and confirm the effects of the manipulated variables. It is absolutely crucial that researchers and journal editors are willing to publish not only confirmative results, but also those, like this one, that offer no or limited support for the tested relationships. This is the only way towards a more refined and, not least, realistic and systematic understanding of the intricate relationship between political activism and photography.