To understand the appeal of far-right communication strategies we draw on social psychological theories of identity, othering, and affect. A common mechanism that many far-right nationalist political parties utilize is the construction of an exclusionary national identity group in opposition to a different threatening out-group identity. The field of social psychology has long studied how ‘we’ group identities are shaped by comparison to a ‘them’ group identity, and how this in turn influences people’s attitudes, feelings, and behaviors towards the ‘other’ (Tajfel & Turner, 1978). It is the majority and the powerful group’s norms and ideologies that define who is perceived as an opposite ‘other.’ National identity is one in-group categorization that is often emphasized in political and media discourses and in many other banal implicit everyday forms (Billig, 1995). In-group national identity and out-group minority stereotyping are often interconnected; dominant stereotypes of minorities are not arbitrary, but are shaped by the majority groups’ political interests, as well as by the attributes and behaviors of the minority group (Wagner et al., 2009). Thus identity here is not only a cognitive process of categorization and comparison, but also an affective process that is influenced by the specific cultural, social, and political contexts (Billig, 2002).

Far-right nationalist parties often construct an exclusionary in-group identity with impermeable borders, while identifying certain minority out-groups as a threat. Since the early 2000s, there has been a rise in European far-right movements that mobilize followers against a non-European out-group (Klein & Muis, 2019; Yurdakul & Korteweg, 2021). Their ideology often involves nativist worldviews, which assume that nations should be home to only the native group, and any other groups are fundamentally threatening to the homogeneity of the community (Mudde, 2007). Such ideologies are shaped by affective explanatory accounts that often rely on feelings such as hatred and anger (Cash, 1989) and rely on an affective appeal, promising security and stability through a collective in-group identity that protects against the uncertainties posed by ‘the other’ (Kinnvall, 2004).

Affect here is understood as a situated social practice and as a process of embodied dialogic meaning-making that constructs identities, social relations, as well as social orders (Wetherell, 2012). Affect plays a key role in shaping social relationships, group dynamics, and political action (Hokka & Nelimarkka, 2020; Nikunen, 2019; Wetherell, 2015). We use an integrative approach that does not separate embodied states (affect) from sense-making into familiar cultural categories (emotions), and that does not assume thinking and feeling as separate internal processes, but rather interrelated situated social acts that are influenced by cultural norms, social practices and structures, as well as our appraisal of events and the social world (Wetherell, 2012).

Affect then, is a physical, emotional and cognitive reaction mediated by and through culture and language (Benítez-Castro & Hidalgo-Tenorio, 2019; Hokka & Nelimarkka, 2020; Nikunen, 2019; Wetherell, 2012). Affective practices are those instances where people, as embodied and dynamic social actors, communicate, evaluate and negotiate based on their own experiences and the social structures and norms they are situated within (Hokka & Nelimarkka, 2020; Wetherell, 2012). Further, affect is circulated and shared, and affective practices are actively created and sustained, materializing social and psychological life, and constructing relationships between different groups (Ahmed, 2004; Wetherell, 2012).

Through this lens, we consider online visual political communication and the engagement with it as a dialogical affective practice through which identities and social differences are constructed and negotiated. Throughout we will use the term affect and affective practices as an umbrella term, but will use emotions to refer to the more specific discursive categories in the analysis of comments, where we draw on the Appraisal Framework (Bednarek, 2009; Benítez-Castro & Hidalgo-Tenorio, 2019; Martin & White, 2005), which focuses on how users utilize language to express goal-oriented emotions and evaluative meaning-making (Cavasso & Taboada, 2021).

Though originating within Systemic Functional Linguistics, it adopts a similar approach to affect, viewing it as psycho-biological, embodied, and relying on social meaning-making processes, which are mediated by language and culture (Benítez-Castro & Hidalgo-Tenorio, 2019; Martin & White, 2005). To analyze the affective practice of engagement with othering imagery, we use the framework as a practical analytical tool to capture what types of emotions and affective categories are produced in comments. Those categories are analyzed as situated, relational and integral to social practices, rather than as static categories.

Images are particularly effective as tools for constructing affective in-group and out-group identities, as they embody social representations that humanize or dehumanize different groups, and such visual representations have enduring psychological effects on how the ‘other’ is perceived (Jahoda, 1999). Images can condense abstract ideas and emotions into concrete powerful symbols (Awad, 2026) that are powerful in their emotive vividness and memorable character (Joffe, 2008), and can connect individual feelings to collective community emotions (Bleiker & Hutchison, 2008).

Online media and digital technologies have facilitated the use of images by far-right parties to mobilize engagement through the creation of political identities and fostering social connections (Hokka & Nelimarkka, 2020; Nikunen, 2015; Nikunen et al., 2021). These practices involve the constant sharing and circulation of certain images to attract and engage people in an embodied online experience and to motivate participants to act in particular ways (Nikunen, 2015). These images then become floating symbols that gain affective value as they are reproduced and shared (Ahmed, 2004; Hokka & Nelimarkka, 2020; Nikunen, 2015). The emotions most often expressed in far-right images online include fear, anger, and resentment (Hokka & Nelimarkka, 2020; Salmela & Scheve, 2017). Those emotions are connected to higher degrees of affective polarization, which negatively influences people’s moral evaluations (Nikunen, 2019; Nikunen et al., 2021; Renström et al., 2023) and makes them more likely to rely on stereotypes in their opinions (Renström et al., 2023). In a post-modern society that is preoccupied with insecurity and fear, those emotions are easily triggered and channeled towards perceived ‘enemies’ (e.g. immigrants and refugees) (Hidalgo-Tenorio & Benítez-Castro, 2022; Salmela & Scheve, 2017).

Social media platforms have contributed to the spread of far-right visual affective practices. The participatory and engagement features online have facilitated the connection between far-right groups and their sympathizers (Baele et al., 2023), the dissemination of far-right views outside of their networks (Bhat & Klein, 2020), and the circulation of their content across different groups and contexts helping to create international affective mobilization and unity among different right-wing movements (Hokka & Nelimarkka, 2020). The algorithmic structure of search engines and social media also makes it possible for the far-right to utilize the logic of non-restriction on the flow of information (Daniels, 2018; Nikunen et al., 2021) and to diffuse far-right ideas into mainstream politics (Heikkilä, 2017). The more followers are exposed to these practices, the more they are desensitized to online hate, and the more they normalize racism and discrimination (Bilewicz & Soral, 2020).

The data for this study consist of all images (N = 968) posted by the Danish People’s Party from January 2017 to October 2022 on their official Facebook page. The party created its Facebook page in 2012 but only the images posted after 2017 were collected for several reasons. Firstly, the number of images posted per year and the number of followers were significantly lower before 2017 than after. Secondly, in 2016, Facebook changed its interaction infrastructure to include the reactions “Love, Haha, Wow, Sad and Angry,” in addition to “Like”, which broadened the options and enticement to engage with posts. And finally, this period falls after the party became the second largest party in the 2015 national elections, and extends up to the 2022 election day (1 November) where it fell in popularity.

In the first stage of analysis, all images were coded manually by the first and third author for their visual and textual content and the engagement they received on Facebook. The coding process utilized methods from image content analysis (Bell, 2012) in combination with thematic analysis to account for both pre-decided codes as well as emerging codes and to identify main patterns in the data (Braun & Clarke, 2022; Gleeson, 2011). The codes for the visual content included the different persons, objects, landscapes, and institutions portrayed in the images. The textual content was coded for the topic covered by the image and the main attitude towards that topic. Images were assigned multiple codes based on their content, so reported frequencies reflect total occurrences rather than mutually exclusive categories. We utilized an iterative inter-coder approach, where two coders independently coded a shared subset of 100 images sampled across the dataset timeframe. The results were then compared and discrepancies were discussed to refine code definitions and develop a preliminary codebook. This process was repeated with a second subset of 100 images, allowing for further additions and clarification of coding categories. The finalized codebook was then completed and the entire dataset was coded. Throughout this phase, any uncertainties were discussed to maintain consistency in coding decisions.

For engagement, the total number of reactions, comments, and shares for each image was recorded. Engagement is understood here as the number of users pressing a ‘reaction,’ commenting on an image, or sharing it. These are interpreted as indicators of attention, engagement, or resonance (Rieder et al., 2015). To account for variations in the number of page followers over time, these figures were also converted into a percentage of the total number of followers in the year of posting. Historical follower counts were retrieved from archived versions of the Facebook page using the Internet Archive’s Wayback Machine. These snapshots provided approximate follower figures for the relevant time periods. Across the period studied, the number of followers varied from approximately 82,000 in 2017 to 106,000 in 2022. While this represents some growth over time, the relative difference is modest, suggesting that variation in audience size is unlikely to be the primary driver of observed engagement patterns. Additionally, users can react and comment on posts even if they are not followers of the party’s Facebook page. Due to these limitations, we report in the results section raw engagement figures rather than percentages.

Another limitation to interpreting engagement numbers in relation to the visual and textual content is that each image was assigned multiple codes, limiting the ability to isolate the effect of specific visual or textual content on engagement. Therefore, the analysis of engagement numbers is intended to give a descriptive overview of engagement patterns and associations rather than a predictive or inferential model of what drives engagement.

While the first stage of the study provided an overview of engagement patterns, it offered limited insight into the nature and affect of those engagements. For instance, reaction numbers include ‘love’ as well as ‘hate’ and sharing numbers could mean the user is sharing the image to support it, yet they might also share it to refute or mock it. To better understand engagement, the second stage focused on the comments section of a subset of the data. Guided by the findings from stage one, we selected the visual code associated with the highest engagement numbers, and we qualitatively analyzed the first 500 comments for each of the top 5 images within this visual code (2,500 comments = 59,470 words). Although this methodological choice captures the early affective framing of the discussion—which can shape subsequent thread dynamics—we acknowledge that the first 500 comments may differ from later responses and that these posts represent highly engaged cases rather than typical images. Nevertheless, focusing on these highly visible and often sensational posts remains relevant, as such material is central to how far-right actors generate attention, and it allows us to analyze the kinds of affective responses that accompany and sustain unusually high levels of engagement.

We draw on the qualitative and discourse-based Appraisal Framework (Bednarek, 2009; Benítez-Castro & Hidalgo-Tenorio, 2019; Martin & White, 2005), which focuses on the subjectivity within communication, and on how commentators use language to express their (in)direct emotions and their evaluative meaning-making (Cavasso & Taboada, 2021). This approach helps capture the dialogical and evaluative nature of communication, which is characteristic of online commenting, as comments always consist of some form of a reaction or a response to certain content, usually a post, an image, or other comments (Cavasso & Taboada, 2021).

The original model developed by Martin and White (2005) consists of three domains: attitude, engagement, and graduation. The attitude system deals with emotional, moral and aesthetic evaluation, while the latter two subsidiary items are concerned with the dialogical stance of the commentator and the rhetorical scalability of evaluative responses, respectively (Qiao & Jiang, 2022). Because the primary aim of the study is to examine how opinions and emotions are negotiated and expressed in online comments, the analysis will focus on the attitude system of the model.

In line with previous adaptations of the attitude domain presented by Benítez-Castro and Hidalgo-Tenorio (2019) and Bednarek (2009), we substitute the attitude label by promoting affect to the superordinate system to account for how affect is always already part of language, attitudes and opinions. The new affect is subsequently divided into emotion and opinion categories, a practical rather than theoretical separation between evaluative responses and expressions ascribed to emotion categories (emotion) and the assessment of normative, inherent properties of what is being evaluated (opinion) (Bednarek, 2009; Benítez-Castro & Hidalgo-Tenorio, 2019). We also adopt the changes to the new emotion sub-system presented by Benítez-Castro and colleagues (2019, 2024) which offer a clear, psychologically informed classification system that emphasizes the evaluative, goal-related nature and social directedness of affect (see Table 1). In addition to the adjustments above, we also included an overt/covert distinction, following Bednarek (2009), to allow for the coding of more subtle emotional responses and invocations of affect, and to make the model more suitable for online comments, where the prevalence of indirect emotional responses tends to be high (Cavasso & Taboada, 2021).

Two further codes were included in addition to the affect-framework to account for the overall tone and nature of the engagement with far-right visuals. First, we included a marker for polarity in order to gauge the distribution and intensity of positive and negative evaluations, as determined by coders on the basis of the general tone of the comment, with the underlying assumption that comments may vary in intensity and force even within emotion types (e.g. hate vs. indifference) and depending on the manner of writing. Secondly, we were interested in capturing the directionality of these evaluations and thus included emerging codes for subject/object of affect. A total of 7 subject-groups were identified within the analyzed comments: DF, Immigrants/Refugees, IS-Fighters, Politicians/Government, Other commentators, The legal system/Police, and Denmark.

All 2,500 comments were manually coded by the second and fourth author into the adapted model above. A decision was made to code each comment only once within each of the categories of emotion, opinion, polarity and subject of affect. A principle of salience was therefore employed, in which only the most prominent feature of each category was coded. While this allows for an overall impression of the valence of comments befitting the scope of the current article, it does limit the display of nuance observed within individual comments.

Given the subjective nature of qualitatively interpreting the comments, inter-coder reliability and agreement were prioritized. An initial test of the coding scheme was performed on the first 150 comments on Image 2, to give the coders a chance to familiarize themselves with the coding process and subject matter. The coders then met up to discuss ambiguities in interpreting the coding scheme and developed a set of guidelines for coding and resolving disagreements. Further collective training was performed, after which the guidelines were adjusted, and a decision was made to have regular meetings throughout. A further 30 comments or more for each image were also coded in collaboration, followed by a discussion to allow particularities relating to the specific subject matter of that image to emerge and be considered. Additionally, we adopted a consensus approach, in which any comments deemed ambiguous by the coders were collectively coded and discussed, which is a useful strategy when examining qualitative data that requires sensitivity towards subtle affective meanings, and where multiple codes might apply to the same unit of analysis (Campbell et al., 2013). The remaining comments were divided between the coders and individually coded. For these comments, an initial intercoder reliability score was calculated based on simple proportion agreement by comparing 5 sets of 10 comments (50 comments) for all five images (86%, 84%, 80%, 88% and 90% for Image 1, 2, 3, 4 and 5 respectively) before any remaining disagreements were negotiated and resolved for the final dataset.

The data used in the study are from a public page and posts which are public on Facebook, which have been manually obtained in compliance with the platform’s terms of service. No information about private accounts is provided here, and all quotations have been translated, with identifying information redacted or anonymized.

Even when pursued in the service of critical assessment, collecting, analyzing and reproducing potentially harmful discourses for scientific research purposes requires careful considerations, due to the processing of personal data, the risk of amplifying such discourses and violent rhetoric, and the risk of contributing to the continued normalization of different types of othering (Askanius, 2021). To mitigate these risks while also balancing transparency, only the necessary data were collected and stored safely, and illustrative quotes have been provided only to the extent that they support research goals and facilitate the critical evaluation of the affective practices under study.

The visual content coding resulted in 41 codes and the textual content coding resulted in 20 codes. Table 2 shows the codes that had a frequency of 40 or higher (i.e., over about 4% of the sample). The table shows that the party’s communication relies on certain representations of the ‘in-group.’ These visual representations include the Danish people (P3-P6, n = 192), the Danish nation represented through the flag (S3, n = 65), and the Danish nature represented in green landscape (N3, n = 40). In addition to these in-group defining images, another recurrent representation is that of constructing the ‘other,’ often illustrated through depictions of Muslims and criminals (P8-P10, n = 202). When coding for all people represented in the data, politicians’ images were highest (P1-P2, n = 297), followed by the ‘other,’ followed by images of generic ethnic Danes. Aside from images of politicians, it is the image of Muslim women that is most frequent (n = 85). When looking at the topics most frequently expressed textually in the images, the “anti-immigration and refugees” topic is the most frequent of all the topics (n = 309). This aligns with studies presented earlier on far-right nationalist identity construction and how it involves a high focus on distinguishing an out-group.

Average engagement numbers show that across codes of people represented in images, it is the images that include Muslim women (P10) that have the highest average engagement per image (5,216). It is more specifically those images that include Muslim women in a facecover (O5) that have the highest engagement numbers across all visual codes (5,806). Across codes of textual content and attitude, it is the anti-Muslim statements that have the highest engagement numbers (4,535). Note that those codes are not exclusive, as any image coded for facecover is simultaneously coded for Muslim women, and many also have anti-Muslim textual content. Understanding how these images can be associated with engagement requires an intertextual and contextual analysis that considers the interplay of visual symbols and text, as well as the influence of political debates at the time of posting.

To explore this intertextuality further, we selected the visual code with the highest engagement (facecover) and analyzed it in relation to the accompanying textual content and other accompanying visual codes. The most frequent textual content accompanying the 49 images of facecover was: anti-immigration and refugees (n = 27), anti-Muslim (n = 20), pro-law enforcement (n = 7), and anti-Red block parties (n = 6). In terms of other visual codes, aside from Muslim women and facecover, the most frequent accompanying visuals were: Muslim men in traditional Muslim clothing (n = 12), migrant child (n = 8), criminal (n = 6), Danish women (n = 4), and Muslim women in headscarf (n = 4).

This preliminary overview indicates that the facecover has been repeatedly used by the party to associate and elaborate its meanings and affect with other issues. The facecover imagery is consistently used to represent the broader immigrant and refugee out-groups, using it to advocate for firmer law enforcement and to criticize other political parties. This makes the facecover a visual floating symbol, filled over time with different associations and emotional registers that gain affective value as it is reproduced and circulated (Ahmed, 2004; Hokka & Nelimarkka, 2020; Nikunen, 2015). Visually the symbol is also associated with other ‘threat’ out-group representations such as Muslim men and associations with criminality. It is also used in juxtapositions that contrast Muslim women wearing a facecover with Danish women. This latter juxtaposition is a positioning strategy often used in images to make an argument through contrasting two identity groups (Awad, 2026). Here the Muslim woman becomes an embodiment of the irreconcilable difference between what she represents and the Danish cultural values represented through the figure of the Danish woman.

Based on the findings above, we looked further into the visual code associated with the highest engagement numbers; that of the facecover. We collected the first 500 comments for the top five images with the highest engagement numbers within the facecover visual code. The five images (see Figure 1) were posted by DF between May 2018 and July 2019. In every image at least one woman is wearing a face-covering burka, but the context and purpose of posting the images differ. Understanding the context of each image was relevant to coding (e.g. subject of affect) and to the dialogical and contextual interpretation of the comments.

The context of the first image relates to DF’s stance against taking in refugees, in particular those who are able to visit their home countries. The image portrays a woman in a burqa walking with children in public and the text reads: ‘If refugees are able to vacation in their homeland, they can also stay there permanently! Agree? Please share’ (transl.). The second image tackles a highly active debate at the time on whether or not to bring Islamic State (IS) fighters with Danish citizenship back to Denmark after having fought in Syria on the side of IS. The image portrays a prohibition symbol placed over a graphic illustration of three women wearing facecovers depicted without eyes and carrying weapons. The text reads: ‘IS-fighters should be deprived of their Danish citizenship! They have turned their back on Denmark! Please share’ (transl.). The third image informs viewers of the newly announced decision to accept the UN mandated quota refugees. The image portrays three women wearing facecovers walking in a public space, and the text reads: ‘The government has now notified the UN that Denmark will receive the mandated quota of refugees in 2019. We, DF, say no to mandated quota refugees. What do you say? (transl.). The fourth image relates to DF’s political victory in enforcing a ban on facecovering, colloquially referred to as ‘the Burka Ban’. The image was posted on the same day the new law was passed in parliament. The image portrays a prohibition symbol placed over an illustration of a woman wearing a facecover, and the text reads: ‘No more burkas in Denmark as of today!’ (transl.). DF stresses in the post how face covering is oppressive and they encourage people to like and share the news. The fifth image relates to DF’s stance on whether the borders should be closed to asylum seekers in the wake of the Covid-19 pandemic. The image shows an illustration of a man wearing traditional ‘galabia’ and a woman wearing a facecover, both without any facial features, and the text reads: ‘Close the borders to asylum seekers! Agree? Please share!’ (transl.).

It is important to note that among the textual content presented in the five images, only the fourth image addresses the facecovering and the Danish ban. The other images use the imagery of women in facecover as a prototype –a generalized stereotype image– for refugees and asylum seekers (Image 1, 3, and 5) and terrorist fighters (Image 2).

The findings show that the polarity of expressed affect in the comments was overwhelmingly negative, with the polarity of –1 making up more than 75% of all comments. Positive comments made up only 14%, with +1 being the most frequent polarity measured. In terms of the prevalence of different emotions, it is unsurprising, then, that negative emotions made up 85.5% of affective comments. These findings align with earlier findings by Cavasso and Taboada (2021) who found predominantly negative evaluations in online news comments, as well as Qiao and Jiang (2022), who found X (Twitter) users to be primarily attentive to the negative aspects of polarizing topics.

Interpreting the polarity of online comments required several considerations. We started with the assumption that online comments are often exaggerated expressions that come out of people strongly feeling with or against something and therefore expected there to be more comments with high polarity: -2 or +2. We also considered that dissatisfaction is milder than repulsion and satisfaction is milder than attraction, and thus expected this to be reflected in the polarity. The results, however, show that most of the comments fall under -1 and +1, and that those polarities included all four most common emotions. To illustrate with a few examples, the comment “It's certainly not because I don't like other people, but it's their behavior that I detest” (transl., Image 2, Comment 839) is written in a sober yet negative way, but the content reflects hate and intolerance, which led us to code –1 and Repulsion. The comment “The question is basically how they got a Danish passport.” (transl., Image 2, Comment 905) reflects sober negative polarity with confusion and distrust, which was coded –1 and Dissatisfaction. This means that a comment may contain, for example, much hate or intolerance towards immigrants or DF, yet still be written in a concise non-intense manner.

The vast majority (97.2%) of emotions coded were covert (ex. “And let it be closed to asylum seekers for all eternity”, transl., Image 5, Comment 2035), rather than overt (ex. “Another meaningless piece of virtuesignaling to solve a non-existent problem. What will they think of next… I am appalled”, transl., Image 4, Comment 1735) expressions of emotion. Similarly, only a very small number of comments were not double coded for both emotion and opinion (1.4%). These findings support the inclusion of the covert label in our analysis model, as well as the decision to allow for double coding of these categories, without which we would have missed crucial and nuanced insights into emotions and affect expressed by the commentators and the influence of emotion on judgement. As for the opinion category, the subdivision of social sanction was the most frequent (ex., “No, no, no and no again. Their culture fits very poorly with Danish culture. And they are siphoning money from the state. Red block must be stopped”, transl., Image 3, Comment 1381), making up 78.7% of double coded comments. Social esteem (ex. “The whole lot at Christiansborg are sleeping, they get paid really well for their non-work:( ”, transl., Image 1, Comment 96) made up only the remaining 19.8%, which suggests that commentators more often engaged in moral evaluation according to social norms and laws, rather than the judgement of abilities and personality. Further, no appreciation was found in our data. One explanation pertains to the nature and subject matter of the images being studied. Othering imagery may primarily encourage engagement involving the judgement of people rather than objects. Even when the image context concerned an object-discourse like the Burka Ban, the judgement was directed at either the legislators proposing a ban or judgement of the people involved in face-covering practices, not the burka itself.

The most prevalent subject of affect overall was DF (35.2%), with Immigrants following shortly after (30%). When looking at only the comments pertaining to positive emotions (Attraction and Satisfaction) however, the most prevalent subject was Immigrants (51.8%), with DF coming in second with 28.1%. As for the negative emotions (Repulsion and Dissatisfaction), DF was the most prevalent subject of affect (36.9%), with Immigrants coming in second with 27.4%. An overview of the prevalence of the different subjects of affect is displayed in Table 3.

The model also included codes for both Inclination and Grabbing emotion types, but neither was prevalent in the data. Only one comment was coded for inclination: “[Tagged individual]. I’m more inclined towards the New Right party” (transl., Image 5, Comment 2295). The lack of Inclination is in line with Glăveanu and colleagues’ (2018) earlier study showing that the nature of online commenting is more about stating one’s stance, and less about being open to dialogue and perspective-taking. The manner of coding and the principle of salience could also have influenced the frequency with which certain emotions were observed. Other articles have opted for a further breakdown of the text into smaller components (e.g. Cavasso & Taboada, 2021), which may have allowed for the observation of more instances of emotions such as surprise or inclination.

Our results regarding the subject of affect bring in a nuanced perspective that contradicts the idea that hate and othering images attract online engagement mostly from supporters of those ideas. While the comments overall are very negative, this negativity is often directed towards the party itself (31.3% of all comments). As one commentator states: “DF has always mastered the art of spreading hatred and discord to perfection” (transl., Image 1, Comment 1). Many other commentators share the same sentiments, with most emotions pertaining to Dissatisfaction (the most prevalent emotion type) being directed towards DF. Similarly, when looking at the arguably limited number of comments coded for positive emotions, the subject of Immigrants has the highest frequency (51% of positive comments, 7.2% of overall comments), indicating an opposition to the values and opinions usually expressed by DF politicians and supporters. Looking at these patterns together, upwards of 40% of all comments do not agree with or only partially agree with the opinions expressed by the party. One reason could be that, while these images are posted on the official DF Facebook page, this is not a closed group and images circulate to a broader audience than those who have liked or followed the page. As Cavasso and Taboada (2021) point out, the reaction options available on many social media platforms also make it possible to show agreement without having to go through the trouble of commenting, leaving the commenting to those who most strongly disagree or have a desire to express their opinion.

The results support previous studies arguing that the prevalence of negativity in comment-engagement is characteristic of the evaluative nature of online commenting. User engagements with images are driven by negative affect, regardless of whether these are directed towards the content of the post itself, the sender, or other commentators. This is further supported by the results of Qiao and Jiang (2022) and Jing-Schmidt (2007), which confirm the negative asymmetry phenomenon and negativity bias, denoting how people are more likely to pay attention to and engage with negative events and topics that generate negative affect.

It is worth pointing out, however, that there are still many comments expressing Satisfaction and Attraction towards DF, just as there are many comments expressing Dissatisfaction and Repulsion type emotions towards Immigrants, which highlights the diversity of expressed emotions and opinions found in the data. When looking at Repulsion in isolation from the combined negative, for instance, the most prevalent subject of affect is Immigrants, making up 14.5% of all comments (ex. “The Mediterranean should be overflowing with mines so none of them had come up here in the first place < 3”, transl., Image 2, Comment 692). While there could be many possible explanations, one reason why Immigrants are the most common subject of Repulsion-type emotions relates to the political strategy often used by far-right parties, relying heavily on othering imagery to mobilize strong feelings of contempt and disrespect (Awad et al., 2022; Klein & Muis, 2019; Nissen et al., 2021; Richardson & Wodak, 2009; Yurdakul & Korteweg, 2021). As intended by such imagery, the posts may be more likely to provoke these emotions, especially intolerance towards out-group members, at least from those who share the party’s views. Importantly, the Repulsion category includes intolerance, which was the most common indicator for a Repulsion-type coding when Immigrants were the subject of affect, as noted by the coders.

There are several arguments that could explain why the visual symbol of the facecover was associated with the highest engagement numbers in our data. To elaborate on its psychological influence as an affective symbol, we discuss below the social life of that image (Awad, 2026), which includes its production and circulation by DF, its viewers’ engagement, its specific context within Danish culture and politics, and its broader historical context.

The frequency and engagement with the facecover image could be interpreted in the context of the public debate in Denmark around the Burka Ban that was presented by the center-right government coalition in 2018 and resulted in a law banning garments that cover the face. However, that would only explain its prominence during these years, where the party could have been using this visual to mobilize the passing of the law. Looking at the data longitudinally, controlling for dates, Table 4 shows that it is mostly in 2020 and 2021 that the facecover image was used.

The use of the image is also situated in a broader context of the development of Danish political debate since the 1990s, where the focus became less centered on economic policies and more on new issues that polarized people in different ways than during the Cold War (Borre, 2011). This included discussions on refugees and immigration, which gave rise to a ‘new’ nationalism, represented mainly by DF at that time (Borre, 2011). This development also arose in response to the increasing numbers of non-western immigrants in the 1960s who came from Ex-Yugoslavia, Morocco, Turkey, and Pakistan (Jacobsen, 2015). This has led to an increased political and media attention to the topic of immigration, with a focus on Muslim minorities as posing a cultural threat to Danish values.

There are no clear statistics on the number of Muslims in Denmark, but according to one study, in 2020 there were around 256,000 Muslims in Denmark, which corresponds to around 4.4% of the population (Kristensen, 2020). It is even harder to estimate the number of Muslim women wearing headscarf or face covering, since this group is a rare and elusive religious sub-culture group, but according to a study by Warburg et al. (2013), they estimated that there were around 150 women in Denmark who wore face covering in 2013, which corresponds to 0.1–0.2% of Muslim women in Denmark. One could imagine that the number might have decreased from that estimate due to the ban in 2018.

Despite this small number, images of women in facecover occurred 49 times in images posted by DF. The continuous use of such images promotes the idea that the number and magnitude of the ‘problem’ is much greater than it actually is. It also emphasizes, by repetition and mere exposure, an association between women wearing facecover and Muslim women, immigrants, and refugees. This visual strategy is contested and pointed out in around 5% of the comments analyzed, where the commentator makes an explicit reference to the misleading use of the facecover in the image. For example: “What a tasteless and misleading image – typical DF scaremongering. OF COURSE we should take in quota refugees.” (transl., Image 3, Comment 2128), “Funny how you illustrate quota refugees with burkas, which you have banned anyway. Why is that? Manipulation?” (transl., Image 3, Comment 2056), “Interesting choice of image, which even before the ban wasn’t representative of quota refugees but rather representative of DF’s warped views of humanity” (transl., Image 3, Comment 2072), and “This image is so distasteful.. It is not a coincidence that it depicts muslim women in burkas. DF are really trying to brainwash – You are sending the message that we should be afraid of everyone who looks muslim.. disgusting, you’re so far beyond reproach, and people are eating it up..” (transl., Image 2, Comment 590).

The use of the facecover as an affective symbol extends beyond Danish context and the Burka Ban debate, reflecting a broader historical and cultural significance. Analyses of the historical social life of images of Muslim women show that such imagery has been central to Western constructions of the Muslim “other” (Razack, 2004; Yegenoglu, 1998). The veiled woman is usually constructed as someone to be simultaneously feared, pitied, desired, and in need of saving, and the veil itself is portrayed as ahistorical and static; a symbol of gender oppression and civilizational deficit (Hoodfar, 1991). In Edward Said’s (1978) terms, this group is the ‘ideal other’ that ‘we’ have nothing in common with and are justified in needing protection from. It is the ‘exoticized other’ that threatens a cultural invasion. This cultural invasion aspect makes it a ‘safer’ out-group signifier than a less ‘safe’ out-group signifier that would include skin color, for example, which could risk being sanctioned for racism. The gendered dimension of this imagery also reflects how women’s bodies, more broadly, have historically been mobilized to signify a pathological form of otherness (Hall, 1997). As such, women’s bodies frequently become sites of moral, cultural, and political debate and contestation (Hafez, 2014).

Given this context and building on broader theories of the psychological power of images (Awad, 2026), we argue that the significance of the facecover image lies in key psychological processes of boundary construction, social positioning, and de-individualization.

The image creates in-group positive distinctiveness through a clear boundary against the out-group. The in-group becomes the one valuing liberal values and women’s emancipation. Meanwhile, the facecover becomes a symbol that brings the crucial irreconcilable difference, the difference that matters more than any similarity. This aligns with research showing how far-right groups in Europe consistently use images of Muslim women as an oppressed group who pose a threat to the freedom of Western women and liberal cultural values (Berg, 2018; Sayan-Cengiz & Tekin, 2022).

The image thus becomes a powerful tool for social positioning. By representing a unified image of the ‘oppressed’ woman, as opposed to the ‘emancipated’ European woman, the viewer is invited to position themselves in relation to the perceived arguments of the imagery. Through this image, the party positions itself as defender of national culture and values, as well as ethnocentric ideas of gender equality, a framing common in Danish politics (Yilmaz, 2015). The facecover image is thus effective in serving as a communicative camouflage (Adami, 2020), where far-right positions are expressed and moderated through the image, such as presenting and advancing anti-immigrant agendas as a feminist issue (Farris, 2017). However, as the comment analysis shows, many viewers who engaged with the images rejected such framing. This highlights how viewers are not passive receivers of these images, but actively engage and dialogue with content they encounter, and position themselves in relation to the arguments they perceive the image to be posing.

Finally, the image of the facecover provides an easy tool to de-individualize the ‘other.’ The facecover as a religious practice is ‘de-individuating’ in the sense that it obscures most facial features. Research shows that images de-individualizing a certain out-group are often used to portray them as a threat, rather than as people to empathize with (Falk, 2010). Further, obscured or blurred faces in news images of refugees can reinforce perceptions of them as deviant and threatening, contributing to processes of visual stigmatization (Banks, 2012). This is closely linked to the formation and reinforcement of stereotypes, generalizing this image to the whole group of immigrants and refugees, as the party does through the textual content on the images. De-individualization not only sustains stereotyping but may also contribute to forms of dehumanization, attributing the ‘other’ fewer human qualities and justifying harmful or discriminatory behavior.

The normalization of such visual affective practices by far-right groups can have broader societal implications. By de-individualizing and visually stigmatizing the ‘other,’ these images can amplify feelings of moral outrage, fear, and hostility, facilitating both the justification of discriminatory behavior and support for exclusionary policies. Such representations also have consequences for minority groups’ self-perception, as ingroup identity and outgroup stereotyping are interconnected, and minority groups develop a strong sense of themselves not only as a group, but also in relation to the expectations and views of the wider society (Howarth, 2002).

This study contributes to the body of literature on far-right visual practices online by connecting social psychological identity processes with affective practices and digital platform engagement dynamics in a culturally situated case. There are two main findings regarding the affective engagement with far-right visuals. First, that images of othering were associated with more engagement than images of the ‘in-group.’ Second, that engagement was not coming primarily from supporters of the party, but also from people contesting the image and its message. These findings imply that the engagement comes from a more nuanced audience than initially assumed. Further investigations are needed, however, to better understand who the viewers are and their complex social positions. Digital media analysis has its limitations in understanding the demographic characteristics of the viewers, how the content circulates, and how algorithms play a part in which images are seen the most by followers vs. non-followers of far-right groups. It could also be relevant to look beyond viewer engagement online to explore how different visual content mobilizes different forms of action, such as voting behavior or discrimination.