While the topic of meme culture is widely studied, one of its subtopics has only rarely been explored despite its widespread popularity: the use of GIFs in political debates occurring on text-oriented social media platforms such as Twitter. This article argues that Twitter users who use GIFs in political statements intended to persuade others are driven by an urge to cathartically alleviate frustrations stemming from the absence of non-verbal communication affordances. Two significant propositions follow from this. First, through a revisitation of Stuart Hall’s reception theory, it is argued that the cathartic practice relies on the assumption that GIFs can convey at least some of the posters’ emotional states. Second, it is proposed that posters assume that GIFs facilitate emotional transfer by depicting basic, common displays of emotion that may influence and persuade other users through affect. It follows that a foundational set of these basic, common displays of emotions can be identified and categorized using previously established applications of affect theory. To begin the process of decomplexifying of these political exchanges of affect, a preliminary typology is provided. The typology is based on the theoretical assertions of the article, including a dual-track view of affect theory that combines the more recent Spinozist-Deleuzean view with the theory’s practical-empirical origins.
As social media platforms have become venues of political debates between users, it seems natural that the affordances provided by these platforms would become instruments of argumentation and persuasion in such debates. As exemplified by political exchanges in the short videos on TikTok, visual performance can be a highly effective way of demonstrating a personal viewpoint.1 However, while political discussions are popular in visually-oriented social media such as TikTok and Twitch, they appear even more popular on text-oriented platforms like Reddit and Twitter. Of course, engaging in political debates on a primarily text-based social platform does not turn users into calm, rational debaters, and emotions that may be difficult to express in words still run high. This relates to a problem users have tried to remedy since the birth of computer-mediated communication: the absence of non-verbal cues. Perhaps the most resilient surrogate for non-verbal communication in text-based chats are emoticons, the juxtaposed characters that are meant to resemble different types of facial expressions and which can be traced back to the ARPANET and PLATO systems (and non-computer-based sources before that.)2 Their graphical successors, the emoji, came along in the 1990s, but both symbol types are still in use today.3 Also still in use and serving similar purposes are animated GIFs, which were not initially used as such non-verbal surrogates. Today, however, animated GIFs, also known as reaction GIFs, are frequently used to express that which cannot be expressed in text and would likely be expressed as a non-verbal cue in other interpersonal communication settings. In particular, GIFs remain popular among users who engage with politics through online discussions, as witnessed by the collection of tweets studied here. And yet, despite being part of the oft-studied meme culture, GIFs are generally underexplored, especially in the context of online political debates.
Jiang et al. demonstrate a disconnect in which a GIF may evoke sensations in the person viewing it that are different from those experienced and/or intended by the poster.4 This conceptual-definitional article elucidates how, despite the risk of such a disconnect, Twitter users still use GIFs to express non-verbal sentiments when trying to persuade opponents in political discussions. It is argued here that this occurs because GIF posters believe that GIFs have the capacity to facilitate emotional transfers (thereby also potentially facilitating persuasion) within the boundaries of a basic, emotional “common ground” that the posters share with potential receivers. Hence, building upon previous theoretical and empirical work that has already established the use of animated GIFs as proxies for emotion on social media, the article presents arguments for two propositions:
When posting a GIF, posters assume that they share some common perceptions of emotional GIF depictions with receivers.
Some of these non-verbal displays can be identified by constructing a preliminary typology based on previously established typologies of emotional displays in political TV debates.
The argument for Proposition 1 takes a two-pronged approach by revisiting and adapting the reception theory of Stuart Hall to memetic culture and by extending the work of Rachel Kuo on the cathartic effect of GIF posting.5 The arguments for Proposition 2 are based on the two predominant theoretical tracks that have historically emerged in affect theory, the Spinozist-Deleuzean track and the Ekman-derived track. A typology is constructed by applying these theories to the real-life use of GIFs in political discussions on Twitter. The typology identifies and illustrates a “common ground” (as Clark and Brennan define the term) of emotional displays in GIFs.6 Though the construction of the typology is preliminary, its successful construction demonstrates the basic viability of identifying such emotional displays in GIFs.7
GIFs came into existence in 1987 as a way for the early online service CompuServe to show compressed images on their pages with a minimum of bandwidth usage.8 This also made GIFs a popular image format for the growing World Wide Web in the early 1990s. The format allowed for displaying several images in an animated sequence, similar to a flip-book. Animated GIFs adorned many early websites but fell temporarily out of fashion around the millennium until social media re-popularized them in the mid-2000s. Since then, GIF use has increased, and while the format mostly adheres to a majority of meme genre conventions, it is also unique in a number of ways. A distinctive characteristic of all memes is the potential for recontextualization.9 But in contrast to still-image memes such as image macros, where an image from pop culture is remixed when a user adds text to the actual image itself, letting the GIF appear in a new context is the remix.10 To this point, Grădinaru and others argue that a GIF is free of its original context when a user combines the GIF with new text in a social media post, even if the original context might have been similar to the new one.11
GIFs have long been viewed by scholars as having the capacity for conveying emotion.12 GIFs can be used to performatively represent the user’s situation as one of several social media affordances intended to correct imbalances emerging from the lack of non-verbal communication.13 In this sense, GIFs fulfill the same purpose as emoji, and just like the latter, GIFs can be accessed quickly and easily through keyboards on smartphones – the platform most people use to access social media now.14 In a 2017 survey, more than two-thirds of U.S. respondents in the age group 18-34 claimed that they could express emotional states better through GIFs than through words.15 In a study by Bakshi et al., animated GIFs received more shares and likes on social media than still-image memes.16 The authors note that many of the study participants “pointed out the value of animated GIFs in effectively communicating emotions and gestures,” and they conclude that animated GIFs “are significantly more engaging and viral than other types of content.”17 Adding to Stark and Crawford’s argument, Kuo identifies another similarity between GIFs and emoji, stating that just as they do with emoji, “users deploy GIFs to take up conventional poses” of “embodied affect.”18
There appears to be a general consensus among those scholars who have studied GIFs that they stand out from other meme categories because of their heightened affective capabilities. But since emotions are experienced subjectively, how can anyone who uses a GIF know that its emotional content will be received as intended or as eliciting the same emotional response as the poster experienced when the GIF was chosen? This disconnect has been explored in non-political contexts by Jiang et al., who demonstrate the importance of a shared interpretation of context if GIFs are to be interpreted even remotely as they were intended by the poster. 19 Jiang et al. also successfully show that intent is a highly influential factor when users choose GIFs to represent non-verbal dimensions of their communication. In their study, users expressed frustration over not being able to find a GIF that precisely corresponded to the emotion they wanted to convey or which they trusted would be interpreted by the receiver as intended. This “mismatch” of meaning calls to mind Stuart Hall’s reception model, in which he famously described how the receiver of a mediated message would either read it through dominant code used by the message “producer,” fit the message to their own circumstances using negotiated code, or reject the message’s perceived intention, operating within oppositional code instead. 20
A user who posts a GIF intended to convey an emotional display in political social media discourse is clearly also trying to convey a position. By choosing and posting a particular GIF that the poster perceives as illustrative of their political position and as matching the emotion they wish to convey, the poster engages in a performance of affect intended to be decoded in alignment with the poster’s perception. Although perceptions of them vary wildly, GIFs are unlikely to be posted with the express purpose of being misunderstood. Following Hall, the poster tries to establish a form of "micro-hegemony" extending only to the post in which the GIF appears and where their perception of the emotion conveyed in the GIF is dominant. If there is congruence between the poster's "micro-hegemony" and how the viewer of GIF receives it, then the GIF is decoded using a dominant code. In Hall’s terms, the poster re-presents the GIF, having separated it from its original context. This separation may be so severe, as Grădinaru has convincingly argued, that GIFs may even become Barthesian floating chains of signifieds with no specific referent. 21 However, as the study by Jiang et al. shows, dominant code readings are rare. Negotiated code readings occur much more frequently, in which factors like the receiver's own background, circumstances, and affective state, impact the reading of the GIF. Such negotiated code readings can happen with reference to specific or unspecific referents per Grădinaru.
Negotiated code readings also happen in more indirect ways. One of the defining characteristics of meme culture is the appropriation and recontextualization of pop culture content fragments along with the ability to distribute or redistribute, almost regardless of the sender’s place in any power structure.22 The same Hollywood studios which used to enforce their film and TV copyrights rigorously have now embraced the GIF as a marketing instrument, even though these GIFs often distribute rights-protected material without consent. Such GIFs, sourced from other pop culture products, may invite a response beyond the one elicited solely by the emotional display in the GIF. Recognition of the person or fictional character portrayed in the GIF may bring up feelings related to the original context, pushing the recipient back towards preferred readings and the dominant code used in the original pop culture source, which in turn inhibits the negotiation of codes in the re-presentation. But a more astute recipient may initiate a meta-process in which the GIF’s original context is recalled along with a decoding using the new context, and this is clearly also use of negotiated code per Hall. 23
Of course, Hall was writing in a time when content producers were more likely to let their encoding be influenced by hegemonic ideologies and the cumulative preferred readings dominating society, mostly because of the risk-aversion caused by the higher costs of TV production. Thus, at the time, it may have been easier for receivers to decode a re-presentation encoded in alignment with hegemonic ideologies using Hall’s third way of reading, which employs oppositional code. In the current, much more complex mediasphere, where we find the case of GIFs in political discussions on Twitter, an oppositional code reading requires more analysis on the part of the recipient. In this environment, it is not unreasonable to assume that social media users are aware that GIFs can be used in contexts other than the one to which they are responding. Hence, an oppositional code reading would require that the receiver draws a reason-driven conclusion about what the poster intends to convey by posting the GIF. To reject the message, no matter the reason, the receiver must have some clarity about what they assume the poster means and engage in a deliberate, interpretative process. Hence, oppositional code readings, just like negotiated code and dominant code readings, require that poster and receiver share at least some common ground in terms of how to interpret non-verbal expressions and language. For the exchange to reach even the most basic level of communication, Hall argues, there must be some overlaps between the poster’s and the receiver's individual "frameworks of knowledge."24 Hall specifically includes “visual language” in the discourse that produces these "frameworks of knowledge."25 The aforementioned overlaps must, therefore, also cover non-verbal and visual communication. In the present case, effective, non-“mismatched” communication (per Hall) must thus entail some overlap between how the poster and the receiver perceive the non-verbal displays of emotion in the GIFs. 26
Consequently, though Jiang et al. has presented empirical evidence for disconnects, there must conversely be a minimum of common ground in place, or the communication would not happen at all. It appears, then, that readings can fall on a range in between a total lack of congruence and full, dominant code congruence between the poster’s and the receiver’s perceptions of a GIF. As shown above, negotiated and oppositional code readings must reside somewhere in this range, past the threshold of minimum common ground, for communication to happen.
When readings occur with some level of congruence, GIFs can “cathartically” add the missing jigsaw puzzle piece of non-verbal expression to written rhetoric. As Kuo describes this catharsis, GIFs have the power to “substitute the embodiment of the speaker” and to redistribute and circulate emotions, creating affective economies similar to those described by Ahmed. 27 Extending the catharsis described by Kuo, I argue that the cathartic effect is, in fact, a dual process happening simultaneously at the dialectic and the metacommunicative levels. Dialectic catharsis is related to the message itself. It is cathartic satisfaction with the completeness of the content of a message and the perception that it, with the GIF included, may sufficiently address a discursive element in a debate. Metacommunicative catharsis, on the other hand, is relief from the frustration of not being capable of expressing yourself fully because of restrictions on the media platform. When the main expression affordance is text, as is the case with Twitter, not being fully capable of expressing the entirety of a message as you would in a face-to-face situation can feel seriously debilitating. Metacommunicative catharsis, then, sets in when these debilitations are overcome, for example, by adding a GIF.
This dual catharsis is solipsistic in nature, however. As argued above, it is impossible for the sender to know, at the time of transmission, to what extent the message will be received as intended. Therefore, the sender can only attempt to speculate on and predict the reception effect of the message on the majority of viewers, in a variation of what Suler dubbed solipsistic introjection.28
For this effort to be meaningful at all, the poster must heuristically assume that there is at least some likelihood that parts of the GIF’s emotional content will be understood similarly by the receiver and that the communication will be effective either through dominant, negotiated, or oppositional code readings. Otherwise, the poster would not feel that the restrictions of text-only messaging had been lifted and hence would feel no catharsis.
Establishing that there must be at least a limited amount of commonality between how posters and receivers perceive emotions depicted in GIFs - and that GIF posters may heuristically assume this commonality when posting - is moderately significant. Identifying some characteristics of this commonality would be even more useful. For this reason, I will now proceed to develop a preliminary typology of common emotional expressions in GIFs. To begin identifying some of these emotional expressions, it is helpful first to unpack theories related to mediated emotion, and I will do so through the examination of the two main tracks of affect theory, what could be called the Spinozist-Deleuzean track and what could be called the Ekman-derived track, both of which gave rise to their own, respective research subfields.
It was Silvan Tomkins who first used the term "affect theory" in any meaningful way. He also provided the first typology of emotions and how they may be displayed physiologically. 29 It is important to note, however, that Tomkins’ use of the term “affect,” although still distinct from it, was not as far separated from “emotion” as later affect theorists on the Spinozist-Deleuzean track would make it. Like many of his successors, Tomkins described the emotional states of interest with word pairs. He described six negative affects: anger-rage, fear-terror, distress-anguish, shame-humiliation, disgust, and dissmell.30 That last “affect” was a term coined by Tomkins to describe a form of arrogant rejection. On the positive side, Tomkins offered interest-excitement and enjoyment-joy. He also added a neutral affect, surprise-startle. As mentioned, Tomkins’ work served as the impetus for one of the two dominant vectors in affect theory and the affective turn of recent decades, as identified by Seigworth and Gregg: “Tomkins’ psychobiology of differential affects…and Gilles Deleuze’s Spinozist ethology of bodily capacities.” 31 Seigworth and Gregg go on to note how the Deleuze-Spinoza track came to prominence in the English-speaking world through Massumi, while Tomkins became part of the affect theory movement by way of Sedgwick and Frank.32 Both of these breakthroughs happened in papers published by these authors in 1995, which Seigworth and Gregg describe as a watershed moment for the affective turn.
More than three decades prior, Tomkins’ work had additionally inspired a separate approach, rooted more firmly in empirical studies of psychology, which would eventually intersect with communication and media studies. Building on Tomkins’ work, psychologist Paul Ekman described a typology that categorized human gestures and facial displays into at least six categories: anger, sadness, happiness, fear, disgust, and surprise. 33 Ekman worked directly with Tomkins, and he writes that Tomkins’ influence on him began as far back as 1964.34 This is important, as the Ekman connection reveals a link between Tomkins’ affect theory and its critical, philosophical style of inquiry and the later, highly empirical, psychological studies of mediated expressions of emotion. The typology that will be presented later in this paper is an indirect extension of an Ekman-derived framework constructed by what could be called the “Dartmouth group.” Beginning in the 1970s and led by scholars such as Roger Masters, Dennis Sullivan, and John Lanzetta, the group performed studies of political expression occurring in mass media based on Ekman’s and (indirectly) Tomkins’ work. 35 Extending both the theoretical grounding and the typological structure of the concept, the Dartmouth group presented a framework that categorized facial displays and gestures performed by political leaders in, at first, newspaper photos, but later also video in the form of news segments and debates on TV. 36 Inspired by the growth of visual media outlets available to voters, the work of the Dartmouth group has been revitalized and greatly expanded by Erik Bucy with numerous colleagues in recent years, and it is this currently most developed framework that predominantly informs the GIF typology in the present study. 37
Further extending the Ekman-derived track, Bucy and Grabe present an alternative to rhetoric-oriented studies of political messaging. This comes in the form of visuals that, similarly to GIFs, do not require sound to be understood as part of political discourse: “nonverbal analysis of political communication recognizes a distinction between sound bites, defined as video segments in which candidates are shown speaking, and image bites, defined as video segments in which candidates are shown but not heard.” 38 Although the similarities between image bites and GIFs are plentiful, there are also some important differences. Bucy and Grabe’s concept of image bites is settled firmly within the analysis of political communication and discourse, while GIFs are memetic, shareable, and can be used in a very wide range of contexts. In contrast to image bites, the person posting a GIF is rarely featured in that GIF, meaning that the transmission of emotional cues happens by proxy. Furthermore, image bites are usually centered on facial expressions and gestures, while only a subset of all GIFs in existence even contains human beings. GIFs can also be accompanied by text, either in the form of the social media post in which they are embedded or with words appearing directly inside GIF – this is rarely the case for image bites. Still, within these boundaries, the image bite approach appears valid as a method for identifying common ground emotional displays in GIFs.
Bucy and Grabe demonstrate how voter attitudes may be more influenced by image bites than by sound bites due to, among other things, a bias towards the visual in the brains of those who have eyesight and the prototypicality and familiarity of facial expressions and gestures. 39 From this core principle, Bucy and colleagues have developed typologies that help elucidate aspects of political debate in visual, electronic mass media. The categorization used in these typologies can be summarized as in Fig. 1 below.
Lowered and furrowed
Inner corners raised
Upper raised/lower tightened
Wide, normal, widened
Focused then cut off
Forward or lowered
Retracted or normal
Retracted and/or raised
Lower or none
Upper or both
Variable or none
Away from the source
Forward from Trunk
Turned down from vertical
Tilted from vertical
Angle to vertical
Fig.1: Criteria for classifying facial expressions. Adapted from the summarizing typology provided by Stewart et al.
It should be recognized that any typology derived from Ekman is limited in its scope. Ekman’s work has been criticized for perpetuating a eurocentricity introduced by Darwin, thereby reinforcing racial stereotypes.40 Ekman and others have disputed this, pointing to the global nature of his empirical work.41 Still, it should be self-evident that typologies such as these are broad in scope and do not represent absolute facts. They are only frameworks, basic guides intended to enable deeper investigations. Unfortunately, the typologies have often been extended beyond their intended use. Over Ekman’s objections, for example, typologies derived from his work have been used to add validity to the type of profiling used in facial recognition systems for security purposes and automated, algorithmic emotion recognition in public spaces. 42 Such systems have been shown to be highly error-prone and racially biased. 43 It is important to understand that the observations of facial displays used in the typology presented here were made by humans. Any transfer of the typologies to a nonhuman system will lose the complexity added to the observation by human consciousness and therefore be incomplete and prone to error.
Given their original origin story, the distance between the empirical, quantitative, TV-focused work from Bucy et al. and the work of Spinozist-Deleuzean affect theorists is shorter than it first appears. It seems natural, therefore, to also construct the GIF typology below using theoretical elements from contemporary affect theory. Consequently, I submit that there are three dimensions of affect to consider when GIFs appear in political exchanges on social media. The first dimension is the original emotion invoked by whatever event the poster feels the need to respond to. The content of a tweet, or an event happening elsewhere, may immediately and intuitively invoke disgust, empathy with any victims involved, or moral indignation, for example. The second dimension is the following urge to post. A person’s drive to react instead of just moving on is in itself affective. It may be driven by a need for validation, to feel part of a community, or be a reaction to a threat. The final and third affect dimension is one dominated by performativity. You perform this emotional display when you post a GIF. This is the only dimension visible to the receivers, though all three dimensions influence each other. GIF posting can never be fully spontaneous (unlike the facial displays of political leaders in TV debates) and must therefore be performative because there is a selection process involved, during which the poster has time to rethink or moderate their response.
As previously mentioned, Rachel Kuo has powerfully unpacked how GIFs are vehicles for the performance of affect.44 James Ash goes further and links GIFs to Deleuzean notions of affect and sensation more intricately, thereby building a bridge to the Massumi-led track in affect theory.45 Ash also contributes a stronger distinction between Deleuze’s concepts of sensation and affect: “sensation can be understood as the rhythmic organization of organic and inorganic forces and the transmission of these forces. Affects can be understood as the encounter of those organized forces with other bodies, which in turn shapes what these bodies are and the sensations they can generate.” 46 Ash goes on to argue that different characteristics of GIFs, including short duration, color, and their repetitive nature, enable sensations to “travel” as one of several “human, technical, and nonhuman means” that facilitate this. “Affects do not simply germinate at a point of encounter between two bodies,” Ash points out, noting instead that Deleuze’s notion of force plays a vital role, and only by following the forces and the traveling sensations is it possible to truly understand the emergence of affect. 47 Ash also proposes an allotropical view of GIFs and affect, which considers the networked environment, including the social and historical circumstances of the environment itself and the individuals in it. Applying Ash’s concepts to GIFs in political debate on social media thus tells us that the nature of the discourse and the parameters of political deliberation is likely to shape the sensations experienced by both poster and viewer of the GIF. It also emphasizes the force of sensation transmission inherent in the affective communication that GIFs facilitate.
Furthermore, Ash’s expansion of the distinction between sensation and affect supports the disconnect between the emotional experience of the receiver and the poster as described by Jiang et al. and discussed above. When the theoretical contributions of Ash, Kuo, and the overall Spinozist-Deleuzean track are taken into account, it becomes clear that the Ekman-derived track cannot stand alone, even with all its empirical explanatory power. While there is great utility to be found in the typologies built by the Dartmouth group, Bucy, etc., upon the foundation provided by Ekman and Tomlins, the Spinozist-Deleuzean track affirms the complexities of affect and emotional transfer that makes the whole effort worthwhile. If sensation could not travel and if affect was not force-ful per Deleuze, typologies such as the Ekman-derived ones would have no reason to exist, as they rely on the assumption that receivers may in some way be impacted by emotional displays. This is precisely what the Spinozist-Deleuzean track of affect theory both affirms and explores. In other words, the Ekman-derived track in affect theory is contingent on the Spinozist-Deleuzean track, while conversely, the Ekman-derived track provides dimensions of material praxis and empirical presence to the otherwise theoretically abstract Spinozist-Deleuzean track. Viewed as a dyad, these two tracks may wield highly convincing explanatory power.
Having established the above view of affect theory as a dyad made viable by Spinozist-Deleuzean theorists and made practical by Ekman-derived theory, I will now continue to show how this applies to real-life GIF examples. The typology below is extended from Ekman-derived work by Bucy and colleagues but stands on Spinozist-Deleuzean pillars. The latter establishes that sensations travel and affect can be mediated. But affect theory also tells us that without some form of interaction with the receiver, for example, through measurement or self-reporting, we cannot know to what extent there is emotional congruence between poster and receiver. Revisiting Hall’s reception theory demonstrated that there must be some emotional “common ground” (per Clark and Brennan) between poster and receiver, but this article cannot go beyond Hall’s insights about reception due to limitations of scope and resources. The typology below is thus not constructed to obtain a deeper understanding of affect generally but simply as an assistive tool for analyses of posts containing emotional displays. By focusing on encoding, the typology may be used to make more informed inferences about posters in political debates on social media and to provide some structure to similar prospective studies. But any assumptions made about reception are contingent on how much “common ground” is shared by poster and receiver. This caveat should be considered if the typology is used in any form of reception analysis.
Another caveat to consider is the data from which the typology is constructed. It is based on a data set collected and coded as part of a separate research project. 48 The GIFs were collected during the two most recent national elections in the U.S. at the time of writing, the 2018 midterm elections and the 2020 presidential election, as political debates (and the related use of GIFs) on social media debate tend to be more frequent during elections. Coding and category construction occurred using a Grounded Theory-inspired process involving two coders working initially from a codebook corresponding to the typology in Fig. 1. 49 As coding progressed, the complexity of the emotional displays made it clear that it was necessary to create subcategories, of which three were then established. Sarcasm and Angry Incredulity were added to the category Anger/Threat, and the subcategory Performative Shock was added to Fear/Evasion. Fig. 2 below provides an overview of the final categories and subcategories used.
Fig. 2: Main categories and subcategories used in this study.
The typology itself is shown in Fig. 3 (main categories) and Fig. 4 (subcategories). The figures also explain the categories through the above theoretical lenses and provide examples selected according to a criterion of how well they represent the rest of the GIFs in their category.
Above: Happiness/Reassurance: Raised eyebrows, wide eyes, upward-turned mouth corners, and a little “shimmy dance” add positive emotional cues to a motivational tweet text. The poster’s encoding connects excitement and joy with the democratic process.
Below: Fear/Evasion: Lowered eyebrows, a furrowed brow, upper eyelids lowered, lower eyelids tight. Mouth corners retracted, eyes averted, head moves laterally. The tweet text explicates that this is an encoding of the emotional state of the poster, further elaborated upon in a subsequent reply.
Above: Anger/Threat: Mouth corners face downward, eyes staring, upper and lower eyelids are wide open, brow is lowered. Both upper and lower teeth are showing, normally indicative of an expression of happiness, but this is likely attributable to the pronunciation of the word “lies.” Given that the poster (at least according to the profile image) identifies as white, this is a case of what Lauren Michele Jackson calls "digital blackface ."The poster encodes the message using non-realistic stereotypes of blackness from pop culture for effect.50
Below: Sadness/Appeasement. Eyes are closed, but eyebrows are not lowered, and their inner corners are slightly raised. Lips are shut tightly, but mouth corners are lowered rather than retracted. Teeth become hidden, and head sinks down as the GIF plays. It is a common expression of disappointment used here to encode a reaction to a post about a presidential candidate.
Fig. 3: Example tweets coded into the four main categories. Adopted from Bucy et al. (Happiness/Reassurance, Anger/Threat, Fear/Evasion) and Stewart et al. (Sadness/Appeasement).
Above: Performative Shock. Upper eyelids are first raised, then closed as eyes are averted, mouth corners are retracted, the brow is lowered as the head moves slowly down and slightly to one side. Used in a non-sarcastic context, this display of fear and evasion is combined with a figurative “pearl-clutching” gesture. Because GIFs are never spontaneous reactions, this GIF can be seen as a performative encoding rather portraying actual shock.
Below right: Sarcasm. The widened eyes, retracted mouth edges, furrowed brow, and slow, horizontal movement of the head are all indications that should code the GIF as Fear/Evasion. The context, however, is the conservative radio host Grace Vasquez amplifying a previous poster’s reaction to statements made on CNN by political pundits, whom the previous poster accuses of exhibiting dramatic, moral hypocrisy. The GIF is encoded facetiously, sarcastically portraying the moral outrage attributed to the CNN pundits.
Above: Angry Incredulity. In this subcategory of the main Anger/Threat category, bewilderment is angrily feigned through the juxtaposition of two different facial expressions. The man on the left (whose words are captioned) is staring at someone off-screen, corners of his mouth turned down (considering its natural shape), lower teeth are showing. This would normally be coded as Anger/Threat. Yet, eyes and eyebrows remain neutral, which is not typical for the category. This facial juxtaposition may indicate suspended anger, temporarily withheld until the incredulity is resolved either peacefully or angrily. Encoding suspended anger in written text is difficult, and this, therefore, serves as an example of how GIFs are helpful in this regard.
Fig. 4. The three new subcategories with examples.
Other than exemplifying the categories, the typology demonstrates several notions. First and foremost, it shows that Twitter users engaged in political debates do not tweet GIFs randomly. On the contrary, both the categorizability and the affective force of the tweets show that GIF posting happens quite deliberately, with intent and direction. Apropos categorizability: while Hall’s arguments for a minimal “common ground” in encoding and decoding is applicable and strong enough in itself, the fact that an encoding-focused analysis yields results whose characteristics are categorizable indicates that posters assume some “common ground” when posting. Why else would so many posters use the same emotional displays in GIFs (or even the same GIFs) as others instead of creating their own? And as previously mentioned, why would anyone be interested in posting a GIF with the express purpose of being misunderstood? Thus, the mere viability of even a basic typology such as the one presented here supports Proposition 1 by identifying at least some “common ground” emotional characteristics.
Furthermore, Proposition 1 is supported by the dual-cathartic process argued for above. The extension of Kuo’s catharsis concept requires the poster to solipsistically imagine how the message might be received. No catharsis occurs if the poster thinks communication has failed. In other words, if catharsis is one of the main drivers for the act of posting GIFs, posters must assume a minimal transfer of meaning enabled by a common framework of knowledge, per Hall. The risk of not obtaining catharsis would be too high otherwise.
Proposition 2 is also supported by the applicability of the typology and the categorizability of the GIFs and tweets. The TV-based typology described in Fig. 1., for example, was shown to be highly effective as a codebook, and only some more precise specification through subcategories was necessary. This is perhaps not so surprising given that image bites are TV clips, and many GIFs are based on clips from TV and film.
It does not appear to be controversial among meme scholars that GIFs are used to convey non-verbal and emotional content. Building on this foundation, I have argued that this act of posting a GIF is in itself affective, in that it can be cathartic by releasing emotional tension arising from not being able to use the full range of means of verbal and non-verbal expression, in addition to the catharsis of voicing, e.g., frustration or joy. Furthermore, by revisiting Hall’s reception theory and combining it with a full, dual-track view of affect theory, I have also argued that GIFs are vessels for sensations as they travel between bodies and take part in the affective process, even when the sensations experienced by poster and viewer only have little in common.51 I have argued that such basic, common ground of non-verbal expression via GIFs can be identified, at least as a preliminary measure, through the typology provided above. The typology, while extending the work of scholars following the Ekman-derived track in affect theory, is contingent on the validation of the possibility of emotional transfer provided by the Spinozist-Deleuzean affect theory track. 52 Though the typology does not describe the complete affective process, it does provide a non-exhaustive basis from which debate analysis and interpretation of social media posts may be performed.
The implications of these conclusions for the study of political discourse on social media are not unsubstantial. As persuasion scholars have long maintained, being brought into a different emotional state is one of the most effective preconditions for attitude adjustment.53 As this study shows, GIFs not only have the affective capacity for such emotional conditioning, they are also used with this intention by participants in political debates on social media. Hence, future studies of persuasion in mediated political communication must consider GIFs meaning-making statements on par with views expressed in writing and rhetorical persuasion.
Further studies based on conclusions made here should focus on reception and the exponential degree to which, with Hall’s terms, different encoding and decoding of GIFs can happen. Though this paper has argued that GIF exchanges provide sufficient emotional “common ground” for communication and some degree of affect to occur, the complexity of the many possible combinations of poster intent and receiver perception is an alluring invitation to further studies. The allure only grows when considering other short-snippet video formats used in political debate and as persuasive tactics in political communication that may rely on similar (or even the same) categories of expression as studied here. Examples include politically-themed reaction videos on TikTok and Instagram. While it is outside the scope of this paper to propose methods that may be used in analyses of such complex multidimensional debates, its conclusions clearly demonstrate that future work towards developing such methods would be of benefit to the field.
Morten Bay, Ph.D. is a lecturer at the Annenberg School for Communication and Journalism at University of Southern California. His research focuses on the societal and democratic implications of emergent digital media technologies with a particular focus on ethics, justice, and new forms of political communication. He has worked as a media strategist, a political advisor, and as a journalist. In the latter role, Bay authored four popular/trade books on technology and society, primarily published in his native Denmark. One of them, Homo Conexus, has been published in English. He still writes for media outlets in the U.S. and Europe.