Racist speech in the Internet, structural racism in recruitment procedures, and racist stereotypes in advertising: racism occurs on many levels and is constantly being reproduced and updated through cultural practices. Speaking, acting and showing are closely intertwined in everyday life. Images, however, acquire a role that is different from that of texts and actions because they can exert a swift, powerfully emotional impact where texts need time to influence the feelings of a reader. Their immediacy also means that images often create an implicit imperative to act. This phenomenon can be subjected to scholarly investigation by examining visual rhetoric. Our starting point here is the observation that images can almost always be used with intentional agency. This is not a new phenomenon. Even in early history, pictorial narratives were used to create religious solidarity; portraits were used by the Greeks to serve men in powerful positions; and advertising signs were employed to lure Romans into visiting the local baths. And this phenomenon has from the very start been linked with rhetoric.

Figure 1: Cicero at his first speech against Catilina. Fresco of Cesare Maccari, 1888.

Initially, rhetoric was a theory of speaking, but already in Antiquity it was expanded into a teaching system that encompassed other media too, such as painting, architecture and music. In the early modern period, oil paintings were structured according to strict rules whose desired impact included: How can I make people cry? Or laugh? How can I get someone to think carefully about something? How can I incite someone to revolution? The art of that time constantly employed such means. When rhetoric lost its primacy in Europe in about 1800, this knowledge about the impact of images remained in hibernation, as it were, in the images themselves. Today these images can only be interpreted with specialist knowledge. Nevertheless, we feel “affects” such as awe, anger or horror when we view these emotionally powerful pictures. This is especially the case with images that depict violence – in other words, images that show something outrageous, something taboo or grisly – or with images whose impact is in some other way “violent”, such as images that slander, embarrass or mock. They can often exert their emotional impact even when the observer does not possess much contextual knowledge. If these images are then combined with a line of text that toys with horror or makes mockery explicit, then we have reached the typical hate-speech meme of our own day. The only aspect that is new here is how these images achieve large-scale, transnational dissemination through the Internet; the repercussions of this in society are not yet understood in full.

The Bernese model of rhetorical design analysis is a four-step process in which experts in visual practice analyse design objects – ranging from graphic design to typography and images – and determine their impact potential (for an extensive explanation, see Scheuermann 2017). They investigate the effect mechanisms described above in order to understand them better. This research involves experienced designers who place their practical knowledge at the service of the analysis. In an initial step, the formal and functional characteristics of the image are determined, and similarities between different media are mapped out. The results of this first step are used in a second step of the process to determine presumed impact goals – such as the emotional impact that can be triggered by a particular colour combination or a particular picture. In a third step, all those elements are systematically sought that might contradict the important intended effects of the medium used. In a fourth step, conclusions are drawn that may inform a re-design of the image or – in our case – that might provide a means of countering it. The principal difference between the applied approach and other forms of image analysis lies in our examining aspects of form and content as a single unit.

Since 2015, the research group VIRAL (“Visual Rhetoric Analysis Lab”) in the Research Unit Communication Design at the Bern University of the Arts HKB has been analysing the editorial design of the online magazine Dabiq of IS (“Islamic State”) in this manner (see Scheuermann/Beifuss 2017) and has helped the Federal Social Insurance Office (FSIO) with analysing existing counter-narratives. In what follows, we shall consider analyses of hate-speech memes, conducted in an interdisciplinary collaboration with social scientists from the Universities of Mannheim and Munich. Here, we will focus on the results of the rhetorical design analysis conducted by our group.

The hate-speech material analysed by both research groups was made available to us by the “Demokratiezentrum Baden-Württemberg” (www.demokratiezentrum-bw.de). This organisation runs the online platform respect!, where hate speech can be reported anonymously. Experts review these submissions and initiate appropriate action, which can range from requesting removal to pressing charges.

We received 312 screenshots documenting content reported between October 2017 and May 2018. These show social media posts (texts, emoticons, images, meme videos) and user profiles (profile names, image galleries). We selected a subset of 156 memes, expanding the concept to include both generic memes (featuring an image macro with customised text on the image) and all posts featuring text-image combinations.

As the origins of this corpus lie in material reported by individuals, its composition is shaped by the socio-political and legal context of the reporting platform. For someone to report a post, he or she must know the reporting platform and must recognise the post as problematic and/or illegal. The platform respect! operates under German law. As in Switzerland, the German penal code (StGB) limits freedom of speech with laws protecting individual rights and collective interests. These prohibit the coercion or denunciation of a person, depictions of extreme violence, and issuing any public call to commit criminal acts. The most significant difference between Swiss and German laws is the German prohibition of symbols of unconstitutional organisations ($86), including the National Socialist Party and its successor organisations. Using the swastika or Nazi propaganda imagery is thus explicitly forbidden under German law. More than a third of the memes reported feature forbidden symbols and/or Nazi propaganda imagery.

With regard to design aspects, the material we examined can be divided into three clusters, based on the affect techniques used:

The first cluster consists of memes that were not identified as directly aggressive or violent, but that include content prohibited by §86 of the German penal code. They seem to be self-indulgent in intent, and are situated within a context of peer-group conversations in a (neo )Nazi subculture. The design schemes and affect techniques used for these memes accordingly tend to foster a range of emotions such as familiarity, intimacy, belonging, heroism, pride and glorification. (See figure 2)

Figure 2: Here, neo-Nazism is both trivialised and “normalised” by combining it with popular culture. These memes emphasise a state of “harmlessness”, and try to establish both the legitimacy of group affiliation, and the illegitimacy of prosecuting neo-Nazism.

Another cluster consists of memes that promote “alternative information”. The affect techniques used here aim to provide credible argumentation and to sow doubt. The arguments in question are rooted in Holocaust denial or racist/anti-Semitic stereotypes. In some cases, the memes are in violation of §86 or promote violent action. (See figure 3)

Figure 3: Memes that try to be convincing often use black-and-white images. When “alternative facts” are presented, they abandon design schemes from traditional journalism, suggesting “non-mainstream” sources instead. In other instances, design schemes from popular media outlets are imitated in order to emphasise the legitimacy of a claim.

A third cluster of memes was identified as being directly aggressive and violent. These memes were designed to denounce, intimidate, shame and dehumanise, to ridicule suffering, to provoke fear, anger and hatred, and to glorify or trivialise violence. In these cases, affect techniques are used to heighten the aggressive impact. (See figure 4)

In a study that is still to be published, Rieger and Harles reach the conclusion that humour and hate speech occur frequently in connection with each other. According to Harles (2018) humour is used to disguise provocative design elements. Our analysis confirms this thesis, and reveals that authors of hate-speech memes employ emotional techniques and stylistic means in a calculated manner so as to offend and to stir up hatred. It is self-evident that memes will be understood differently according to their context, and that the boundary between humour and hatred is blurred in certain cases. We consider rhetorical image analysis as a suitable means of defining these boundaries more clearly.

Such clarification is necessary if we are to plan and realise ways of countering hate speech in a systematic fashion, and if we are to offer support to the legal authorities in their struggle against it. Even in cases where images remain within the bounds of what is legal and are thus not liable for prosecution, they might be an integral aspect of an instance of hate speech that crosses those boundaries to the point of being punishable by law. The authors of hate-speech memes have broad scope for action, and legislators and the judicature are still trying to respond to the mass, transnational dissemination of such material. Politicians today are tasked with finding a competent answer to this challenge, and we need international, interdisciplinary research into the media that spread these images in our society. It is vital that people involved in promoting anti-racism should become aware of the special affective potential of images.

Dr. Arne Scheuermann
Dr. Jachen C. Nett
Maria Mahdessian
Eliane Gerber

Depuis 2015, le groupe de recherche VIRAL (Visual Rhetoric Analysis Lab) de l’Unité de recherche en stratégie de la communication de la Haute école des arts de Berne (HKB) analyse le concept éditorial du webzine de l’État islamique Dabiq selon la méthode présentée ci-dessus (voir Scheuermann/Beifuss 2017).

Internet est une mine d’or pour quiconque veut diffuser des idées racistes et inciter les gens, de manière implicite ou explicite, à commettre ou tolérer des actes de violence. Parmi les nombreux outils offerts par le web, le recours aux images figure en très bonne place : contrairement à l’écrit, qui est un canal de transmission relativement lent, les images agissent directement et efficacement sur les émotions des internautes. De par leur immédiateté, les images constituent souvent une incitation implicite à l’action. La rhétorique visuelle étudie ces phénomènes sous un angle scientifique. Cette discipline se fonde sur la constatation que le recours aux images permet presque toujours de provoquer une réaction. Les images violentes déclenchent par exemple une forte réaction émotive, et ce généralement sans grand besoin d’informations contextuelles.

La méthode d’analyse de la rhétorique visuelle en quatre étapes développée à Berne permet de mesurer efficacement la réaction suscitée par les contenus haineux diffusés sur Internet. Par rapport à d’autres méthodes d’analyse de l’image, celle-ci a la particularité de traiter les aspects de fond et de forme comme un tout.

Le matériel soumis à l’analyse du groupe de recherche provient du Demokratiezentrum Baden-Württemberg (Centre pour la démocratie du Bade-Wurtemberg, www.demokratiezentrum-bw.de). Il s’agit de 312 captures d’écran de pages Internet qui ont été signalées à cette organisation privée entre octobre et mai 2018. Via sa plateforme en ligne respect! l’organisation reçoit des signalements anonymes de contenus web inappropriés et vérifie si ceux-ci constituent une infraction pénale. Dans le cas présent, le groupe de recherche a sélectionné 156 mèmes de cette base de données, en collaboration avec des chercheurs des Universités de Mannheim et de Munich.

La méthode d’analyse de la rhétorique visuelle a permis d’identifier des schémas récurrents et de relever plusieurs clés d’analyse qui semblaient particulièrement pertinentes. À partir de là, le groupe de recherche a classé le matériel en trois catégories, en fonction des techniques utilisées. La première catégorie rassemble les mèmes dont le caractère agressif et violent n’est pas flagrant au premier abord, mais dont le contenu tombe sous le coup de l’art. 86 du code pénal allemand (interdiction de diffuser du matériel de propagande d’organisations anticonstitutionnelles). Une deuxième catégorie regroupe les mèmes dont la stratégie consiste à présenter une argumentation crédible pour jeter le doute sur des connaissances établies. Enfin, la troisième catégorie est celle des mèmes dont le caractère agressif et violent apparaît clairement. Ici, la stratégie consiste à dénigrer, intimider, humilier ou déshumaniser certaines personnes ou catégories de personnes.

Les résultats de l’analyse suggèrent que le recours simultané au texte et à l’image fait partie intégrante des stratégies d’élaboration des contenus haineux, dont il permet d’accroître l’impact émotionnel. En conclusion, ces résultats confirment la thèse selon laquelle les techniques jouant sur l’affect et les autres outils stylistiques sont utilisés de manière tout à fait intentionnelle dans les contenus haineux diffusés en ligne.