Memecry: tracing the repetition-with-variation of formulas on 4chan/pol/

ABSTRACT

In this article we propose a new theoretical framework to conceptualise Internet memes and to trace their temporal variation on 4chan/pol/. We draw from literature on primary and secondary orality to conceptualise the repetition-with-variation of Internet memes as a form of memecry, which we argue is specifically pertinent to the collectivity of online subcultures. We operationalise its study through formulas: mnemonic phrases that encapsulate important elements of oral cultures, which have arguably regained prominence in ephemeral and fast-paced online environments. While Internet memes have often been studied as single images or words, formulas provide a more complex unit for tracing variation and not only circulation. We offer a quali-quantitative protocol to investigate memecry and visualise the spread and variability of 65 prominent formulas on 4chan/pol/, a far-right space known for its reliance on memes. By discussing several cases, we demonstrate how 4chan’s collective identity indeed features typical of secondary oral cultures, while revealing how the memecry of its formulas is entwined with reactionary sentiments and a subcultural struggle for distinction.

KEYWORDS:

• Memecry
• memes
• oral culture
• formulas
• 4chan
• /pol

Introduction

Since its adoption in the ‘90s, the notion of meme has become a staple in both Internet culture and Internet studies. Like their genetic counterparts, the Darwin-inspired idea of the meme as an evolutionary cultural unit owes its significance to its capacity to unite replication and imitation. Richard Dawkins (1976) defined the meme as ‘a unit of imitation’ (p. 192), but also observed that ‘meme transmission is subject to continuous mutation, and also to blending’ (p. 195). Similarly, at the turn of the nineteenth century, Gabriel Tarde already expressed a comparable idea in terms of ‘variability and heredity’:

What did Darwin’s thesis about natural selection amount to? To having proclaimed the fact of competition among living things? No, but in having for the first time combined this idea with the ideas of variability and heredity. The former idea, as it was proclaimed by Aristotle, remained sterile until it was associated with the two latter ideas. From that as a starting point, we may say that the generic term, of which invention is but a species, is the fruitful interference of repetitions (1903, p. 382).

While these insights struck a chord in the academic imagination, large-scale investigation of memes only became possible with the advent of a peculiar type – one that has transformed ‘a rather ambiguous metaphor into a concrete textual genre’ (Gal, 2018, p. 528), making it particularly amenable to quantification (Boullier, 2018; Hagen, 2022). We refer, of course, to Internet memes, the texts, images, and videos that Internet users repeat over and over, often with small modifications for purposes of humour, trolling, vibrancy, or cultural appropriation. While the evolutionary-biological roots of the meme has attracted some justified criticism (e.g., De Seta, 2016; Sampson, 2012), Internet memes undeniably display this double dynamic of imitation and mutation, being ‘imitated and/or transformed via the Internet by many users’ (Shifman, 2013, pp. 7–8) and ‘deeply embedded in popular memory through their repetition and variation’ (Jenkins et al., 2013).

In this paper, we try to reconceptualise and empirically investigate the repetition and variation of Internet memes. Tarde simply called this double dynamic ‘imitation’, but the term risks overstating the importance of repetition while understating creative transformation. A composite expression like ‘repetition-with-variation’ captures our object better, but we put forward the shorter label of ‘memecry’ for its simplicity and connection to the word ‘mimicry’ – a term which designates imitation both as a technique for parody and mockery but also as a prey/predator strategy. As we will see, this double meaning is very fitting to Internet dynamics.

Our argument proceeds in three steps.

First, we discuss memecry as a general communication technique and a feature of online cultures. While the balancing of repetition and variation can be found across cultures generally, we frame memecry as specifically pertinent to online subcultures as ‘secondary oral cultures’ (Ong, 1982). On the one hand, to survive in the ephemeral attention economy of online media, these subcultures need to generate content that is highly memorable and repeatable. On the other hand, because of the virality and connectivity of digital media, Internet memes rarely remain confined within their community of origin and, when spreading elsewhere, lose their freshness and their capacity to serve as a subcultural hallmark. To resist the forces of mainstreaming, online subcultures are then forced to constantly renew their symbols and therefore mix repetition with incessant variation. While this subcultural process of (re)appropriation is well-known (Hebdige, 1979), it is intensified by the fast pace of online attention flows.

Second, because another online novelty includes the increased traceability of memecry, here we empirically investigate its dynamic of repetition-with-variation. We specifically draw on the study of oral cultures to introduce the idea of memetic formulas – symbolically laden phrases relentlessly repeated and transformed within a subculture. While the circulation of memes rarely concerns words only, we argue textual formulas form ideal test cases to explore the characteristics of memecry within online communities.

Third, we test our operationalisation by identifying and tracing the spread and variation of prominent formulas on 4chan, the infamous Internet forum and cradle to many of the most viral memes. We focus in particular on /pol/, 4chan’s far-right and highly hostile politics subforum, notably for its distinctive subcultural nature and its political impact. We observe that, as expected in a secondary oral culture, memetic formulas encapsulate important aspects of 4chan/pol/’s collective identity and circulate in a constant loop of repetition (to assure their preservation) and variation (to maintain their vibrancy). Yet, despite the variation, we also found that memecry does not inherently mean a culture is in a constant state of flux; its form might change, but function can be consistently reactionary.

Memecry and secondary orality in online culture

While memecry is particularly pertinent to ‘secondary oral’ Internet subcultures, repetition-with-variation is a general mechanism that can be found in various cultural settings and serves different communicative purposes. We saw in the introduction how Tarde was intrigued by the Darwian idea of heredity and variability. In Tarde's thought, similarity and difference are ‘everywhere linked and contrasting like the two sides of Nature’ (1895, p. 147, our translation). The idea that ‘to exist is to differ’ (‘exister c’est différer’; ibidem, p. 148) was later elaborated by Gilles Deleuze, in particular in his seminal Différence et Répétition (1968). But the interest in replication and mutation was not limited to French theory. Discussing how repetition-with-variation played a central role in Merton’s sociological semantics, Charles Camic retraces four intellectual lineages for this notion:

In Adorno’s work at the Columbia Office of Radio Research, repetition with variation was a technique of effective musical composition; in the literatures of Western humanism, it was a device that rhetoricians, poets, and novelists employed to produce certain responses from their auditors and readers; in the literature of modern science, it was a process scientists carried out to confirm experimental results; and in the literature of contemporary social psychology, it was a practice that commercial advertisers, political propagandists, and other retailers of mass suggestions engaged in to sway the opinions of a population within the reach of their marketing campaigns. (2011, p. 177)

All these functions are alive and kicking online. Repetition-with-variation is key to the ‘vernacular creativity’ of online communities (Burgess, 2006), particularly as a way to generate jokes by creating expectation through repetition and surprise through variation (Norrick, 1993). It is standardised through ‘meme generators’ and ready-to-use templates, offering a shortcut to produce similar-but-different imagery. It is widely employed both as an experimental device and a marketing strategy, with these two functions merging in the techniques of A/B testing (Kohavi & Longbotham, 2017; Sampson, 2012; Siroker & Koomen, 2013).

There is, however, a more specific way in which repetition-with-variation is connatural to online communication, one deriving from the attention regime of the contemporary Internet. In a previous paper (Venturini, 2022), we suggested that many features of online subcultures can be explained through the notion of ‘secondary orality’ as outlined by Walter J. Ong in the 80s. Ong argued that

electronic technology has brought us into the age of “secondary orality” [and that] this new orality has striking resemblances to the old in its participatory mystique, its fostering of a communal sense, its concentration on the present moment, and even its use of formulas. (1982, pp. 133–134, italics added).

Ong was referring to telephone, radio, and television, but his idea is, we believe, even more compelling when applied to online communication. Before considering secondary orality online, let us unfold Ong’s portrayal of primary orality, i.e., the culture of pre-literate societies.

Ong’s argument hinges on the temporal structure of oral communication, in particular on its inherent ephemerality.¹ The impossibility to record ideas by inscribing them in durable media (i.e., writing them down) has profound consequences on oral cultures and is a major engine of their repetition-with-variation (see also Goody, 1977; Goody & Watt, 1963). In an oral culture, everything that is not repeated risks being forgotten. This encourages its members to iterate the same sayings over and over again, expressing their creativity through minor variation rather than complete novelty. A similar mechanism, we contend, allows understanding online memecry, raising the question: is online communication (or at least part of it) a form of secondary orality?

Describing the Internet as an oral space may sound surprising. Not only does much of online communication occur in writing, but the Web was also introduced as a library technology (Berners-Lee et al., 1992). Still, with advertising becoming their main business model, online media shifted from an economy of information to an economy of attention, vindicating Herbert Simon’s prophecy that ‘a wealth of information creates a poverty of attention’ (1971, p. 40). Since Simon, many scholars have shown that the purpose of attention economies is not the conservation information but, quite the opposite, its incessant rush (Crogan & Kinsley, 2012; Lanham, 2006; Terranova, 2012). In two earlier papers, we applied this idea to digital media and described the factors that accelerate online debates and create spaces for secondary orality (Castaldo et al., 2021; Venturini, 2019).

To build on this, here we study the connection between online ephemerality and memecry, taking as our case study the infamous imageboard 4chan. In terms of raw activity, 4chan is a fairly marginal platform. Yet it is also the cradle of innumerable online memes that are disproportionately more popular than the imageboard itself. These include fairly innocuous ‘old school’ memes like Advice Animals, LOLcats, Rickrolling, and rage comics, but more recently the imageboard has become known for incendiary memes emerging from its far-right subforum /pol/, a hotbed for violent extremism and conspiracy theories (De Zeeuw et al., 2020; Tuters & Hagen, 2020). Two features of 4chan have arguably turned this platform into an example of secondary orality and memecry. The first is its anonymity. As suggested by Elizabeth Eisenstein (1980), authorship – i.e., the strong connection between a piece of content and its creator – came about with writing and print, where such connection can be enforced through records. In contrast, 4chan encourages posting anonymously under the default name ‘Anonymous’, with users referring to each other as ‘anons’. Anonymity may work as a liberating feature, facilitating the circulation of experimental and personal content (Van der Nagel & Frith, 2015), yet on 4chan it also enables hate speech, racism, misogyny, and extreme toxicity (Auerbach, 2012; Donovan et al., 2022; Knuttila, 2011).

The second feature of 4chan is its ephemerality. Unlike most platforms, 4chan has no permanent built-in archive and features a very strict policy about what counts as an ongoing discussion. At any moment, each board can contain only a few hundred threads, ranked by how recently the last comment was made. Threads that do not generate discussion are quickly down-ranked and eventually closed for commenting and pruned from the platform. Additionally, a ‘bump limit’ dictates how many posts a thread may feature before it can no longer be placed at the top of the thread index (Hagen, 2018; Tuters et al., 2018). As such, sooner or later, every thread is bound to disappear from the platform. Analysing three years of data from 4chan/pol/, Papasavva et al. (2020) found a daily average of more than 2.800 threads competing for (at the time) 150 spots on the board. During the heyday of 4chan’s /b/ ‘Random’ board, Bernstein et al. (2011) calculated that the median lifespan of its threads was below four minutes, with the longest-surviving thread lasting just above six hours. While external 4chan archives exist, they only tend to be used sporadically by 4channers.

Despite much of its content being text-based, we can hypothesise that 4chan’s anonymity, limited posting space, and absence of built-in archives endow the platform with an oral-like evanescence. Even though it first served the practical purpose of limiting expensive server storage, 4chan’s creator Christopher ‘moot’ Poole later explained how ephemerality on 4chan became a deliberate design choice aimed at promoting the creation of memorable artefacts:

The only things remembered are the things that become memes, things that are reposted, things that resonate with an audience and are re-posted over and over again, and endure the test of time. The only survivors are those things [that] swim upstream through the waterfall of content, just like salmon. (Poole, 2015)

This explains the virality with which 4chan’s memes spread to the rest of the Web, despite the marginality of their origins and extremism of their subjects. Because of its exemplary ephemerality, 4chan is the perfect case study for our investigation of memecry and the ideal Petri dish for content capable of surfing the streams of online attention economy.

This is a source of both pride and problems for 4chan’s users, especially those participating in its more active boards. Similar to how offline subcultures fight the appropriation of their style by mainstream audiences and commercial parties (Hebdige, 1979; Thornton, 1995), niche online groups often decry the mainstream circulation of ‘their’ memes (De Zeeuw & Tuters, 2020; Douglas, 2014; Phillips, 2015). The identities of 4chan boards, subreddits, Tumblr groups, and fringe communities on larger platforms are characterised by the two elements that, according to the literature, are the birthmark of subcultural groups: the reliance on style and symbols as signs of demarcation (Hebdige, 1979) and the ferocious opposition to ‘mainstream’ outgroups (Thornton, 1995). This is why subcultures are always busy reappropriating and redefining common objects (e.g., punks using safety pins as piercings and earrings) to prevent their current symbols being synchronous with those of larger ‘mainstream’ publics.

All subcultures play this attack-and-defence game, but online subcultures play it within a regime of secondary orality. As all subcultures, they need to generate cultural objects that are recognisable (inside of the subculture) and distinct (from the outside). Yet, because they operate in time rather than in space, these two needs take on a particular flavour. The need to be recognisable becomes the need to create cultural objects that are memorable and repeatable, while the need of being distinct entails changing fast enough to remain ahead of mainstreaming or ‘normiefication’ (De Zeeuw et al., 2020). This is where the metaphor of genetic evolution falls short. Internet memes mutate not to adapt to the environment, but to remain vibrant. Nissenbaum and Shifman (2017) identify this tension between repetition and variation as the engine of what we conceptualise here as memecry: ‘a largely irreconcilable inner contradiction between convention and innovation [which] leads to constant dissatisfaction, as the delicate balance required is never quite achieved’ (p. 498). The authors, however, assume that variation simply serves to avoid monotony, rather than recognising its deeper connection with (sub)cultural wars. We circle back here to the idea of memecry as a game of imitation and mockery, but also as a prey/predator strategy.

A protocol to study memecry through formulas

Memetic dynamics of fixity and novelty have been well-observed in the social sciences and humanities (e.g., Milner, 2013, 2016; Miltner, 2018; Nissenbaum & Shifman, 2017) and orality has already been related to online memetic cultures, for instance to that of the African American blogosphere (e.g., Steele, 2016). What is more novel here, however, is that our theoretical groundwork provides the handles to operationalise a tracing of memecry that is temporally sensitive – attentive to both a meme’s circulation and variation – and that builds on both qualitative and quantitative techniques. Meme studies in the social sciences and humanities, however, have so far mostly relied on qualitative methods (e.g., Milner, 2013; Shifman, 2013), where quantitative techniques are mostly used for coding of static corpora instead of following memes as fluid artefacts over time (e.g., Nissenbaum & Shifman, 2018). Quantitative computational research could address this gap, but has mostly focused on meme detection and circulation (e.g., Acker et al., 2020; Leskovec et al., 2009; Ling et al., 2021; Miliani et al., 2020) and largely ignored their transformation (see Adamic et al., 2016 for an interesting exception).

This is no critique of these valuable studies, but they do leave untapped the potential for a tracing of both sides of the memecry-coin. There are good methodological reasons for this lack of research. Images are known to compose the most archetypical memes (like image macros), but their complex and computationally intense nature makes their evolution difficult to follow at scale. For verbal memes, off-the-shelf text analysis techniques tend to be situated at a level unfit for the study of memecry; on the one hand, studies that focus on single words (Hagen & De Zeeuw, 2023; Hagen 2022; Peeters et al., 2021) work with units that are perfect to investigate diffusion, but too simple to examine transformation and change; on the other, studies relying on topic-modelling can trace the evolution of topics in well-structured texts such as scientific publications (Chavalarias & Cointet, 2013), but struggle in noisier online spaces.

In the remainder of this text, we subject memecry to empirical scrutiny and interrogate how it manifests in the ephemeral flow of 4chan/pol/. To do so we first look for a suitable object of study, for which we once again draw from the theory of oral cultures. A tenet of this theory is that oral culture is highly reliant on formulas: ‘more or less exactly repeated set phrases or set expressions (such as proverbs) in verse or prose, which […] have a function in oral culture more crucial and pervasive than any they may have in a writing or print’ (Ong, 1982, p. 26). Famously, the use of formulas in the Iliad and Odyssey was instrumental in Parry’s (1930, 1932) and Lord’s (1960) demonstrations that Homeric epics were composed as oral poems. According to Ong, formulas are vital to pre-literate societies as mnemonic techniques storing the most important elements of oral knowledge (1932, pp. 33–36). As memetic phrases characterised by repetition-with-variation, formulas can offer a convenient operationalisation for the idea of memecry. As a textual object, formulas are complex enough to display interesting mutations, but simple enough to be studied in large datasets and in scarcely formalised subcultures. Finally, the question of tracing formulas and their variations has a long tradition in the study of oral folklore (Russo, 2011), which this paper hopes to revive through Natural Language Processing (NLP).

The focus on formulas is justified by the fact that most communication on 4chan/pol/ is carried out through text, but does represent an important limitation of our study. In primary oral folklore, formulas are seldom composed of words only and are almost always embedded in multimodal performances that combine speech, gestures, music, and sometimes costumes. Likewise, as described among others by Nissenbaum and Shifman (2017, 2018), memes tend to be multimodal contents based on ‘templates’ that combine text with images and sometimes video or sound (Gal et al., 2016; Zulli & Zulli, 2022). It would be therefore interesting to repeat the analysis carried out below with a more multimodal approach along the lines of Zannettou et al. (2018) and Beskow et al. (2020). This could for instance be done through more advanced feature extraction using a combination of visual, textual, and/or sonic data points. Before trying such a richer but also more complex multimodal approach, however, we think a text-based study forms an ideal initial test case, both for its convenience and the significance of phrasal memes on 4chan. While multimodal change represents an important element of memecry, we believe it does not substantially change our argument about its dynamics, nor does it significantly affect the phrasal memes discussed below.

Our protocol to map memecry (Figure 1) began with collecting all 341 million posts made between 2014 and 2022 from our 4chan/pol/ dataset. This dataset was derived from the 4plebs archive (from November 2013 to mid-2018) and the data capturing tool 4CAT (from mid-2018 to the present, containing some minor data gaps; Peeters & Hagen, 2022). We then filtered and pre-processed the entire archive. This included excluding posts that were artificially or excessively repeated, including posts we marked as spam² and copy-pasted opening posts of recurring threads.³ We then tokenised all text to enable NLP methods, retaining alphanumeric words while removing stop words, URLs, numbers, and several special characters.⁴ We finally stemmed all tokens.⁵

Figure 1. Protocol diagram for mapping of memecry in 4chan/pol/’s formulas.

With the filtered and pre-processed dataset, we then had to extract salient formulas. We did so by first creating NPMI network graphs. These are co-word networks where the edges represent Normalised Pointwise Mutual Information (NPMI), where PMI is computed as $\mathrm{pmi}\left(x;y\right)=\log \left(p\right(x,y\left)/p\right(x\left)p\right(y\left)\right)$and then normalised as $\mathrm{npmi}\left(x;y\right)=\mathrm{pmi}\left(x,y\right)/\log \left(p\left(x,y\right)\right)$ (Bouma, 2009). We preferred NPMI to absolute co-occurrences because it weighs down common words (e.g., ‘fuck trump’) and favours distinct combinations (e.g., ‘praise kek’). Because calculating the NPMI scores between all words in the archive was too computationally demanding, we first filtered our dataset for posts containing words from a seed list of 250 /pol/-specific vernacular terms.⁶ We then calculated the NPMIs between all words appearing in the vicinity of each of our seed words (in a window of two).⁷ We did this per year, only keeping those word pairs that met our threshold of absolute co-occurrences and a minimum NPMI score.⁸ Having obtained these NPMI co-word pairs, we visualised their relation as yearly Gephi network graphs wherein words that were distinctly used together were grouped closely (Figure 2).

Figure 2. A zoom-in of one of the NPMI co-word graphs, visualised with Gephi (Bastian et al., 2009). Words are nodes, NPMIs are edges, and node size is the absolute amount of posts the word appears in. Spaced with ForceAtlas2. Filtered out nodes with a degree range of 1. See all networks on Zenodo: https://doi.org/10.5281/zenodo.7100864 .

We then used the yearly NPMI network graphs to qualitatively identify formulas. Here we had in mind the original stipulation by Milman Parry that formulas should be ‘made up of at least four words’ (1971, p. 275). Since we ignored stop words, triplets of three tokenised words seemed the best way to stay true to this classic definition. This thus excluded simple adjective–noun combinations (‘red pill’) while including longer phrases (‘take [the] red pill’). To identify triplets, we looked for triangular cliques in the NPMI network graphs (Figure 2), and used the 4plebs archive interface to verify whether these three words constituted a valid formula. After listing each extracted triplet and filtering out overlapping and low-occurring ones (i.e., less than 75 posts in one year), we reached a shortlist of 65 formulas.

For all the triplets, we calculated, per year, three metrics to map their spread, specificity, and variation:

1. Formula spread: the absolute number of 4chan/pol/-posts containing the formula. To account for variance, we considered a formula appearance as valid when the three triplet words appeared together within a range of five tokens (see #1 in Figure 3).

2. Formula specificity: the average NPMI between the three triplet words. This captures the specificity of the three words as a triplet (see #2 in Figure 3).

3. Formula variations: Co-words that are not in the triplet (in a window of five) with the highest average NPMI with two out of three triplet words. These other co-words thus often formed an alternative version of the formula, which allowed to map the variation of memecry (see #3 in Figure 3).⁹

Figure 3. An explanatory graph for the double flow visualisations.

We visualised the above metrics as ‘double flow graphs’, with the first two metrics as line graphs and the third as a custom flow chart. The latter consisted of an interactive RankFlow graph (Rieder, 2015) of all ‘external’ co-words above the 0.18 NPMI threshold. To get a sense of which of the triplets words these were most related to, we coloured the word-flows on the basis of their NPMI with each triplet word.¹⁰ This allowed mapping the variance of memecry: vivid/elementary colours meant a close association to one or two triplet words, possibly composing altogether different formulas. For instance, the green flow in the explanatory graph in Figure 3 points to an NPMI-connection to both word 2 and word 3. All interactive double flow graphs are available on a custom webpage (https://oilab.eu/formulas/).

With these double flow graphs, we analysed how memecry manifested on 4chan/pol/ by categorising the data using a grounded theory-inspired approach, which advocates ‘theoretical sensitivity’ to let repeating observations and topics emerge from the data (Glaser & Strauss, 1967). Considering the small sample size, we settled on only four themes we found most apparent in the data and most interesting in light of our theoretical framing: (1) far-right sentiment, referring to whether the formula displayed far-right connotations like anti-immigration rhetoric or anti-Semitism, (2) outgroup antagonism, whether the formula displayed some form of outgroup antagonism or mockery, (3) adversarial ingroup rhetoric, whether the formula displayed some form of ingroup adversity or hostility, often with a didactic function, and (4) the temporal flow of the formula, divided in the categories ‘stable’, ‘spike’, ‘decreasing’, and ‘increasing’. The first three themes could overlap, and especially outgroup antagonism and adversarial ingroup rhetoric intersected: half of the formulas labelled as adversarial ingroup rhetoric also expressed some form of outgroup antagonism by associating other 4channers to outsiders.¹¹ Using these overarching metrics as quantitative handles, we qualitatively illustrated the general findings using select double flow graphs. In this qualitative exercise we could also illustrate the memetic variation visualised in the graphs. All extracted formulas, including their annotations and a short explanation, can be found in Appendix 2.

Subcultural distinction through formulas

Purely quantitatively speaking, the 65 extracted triplets (Appendix 2) seem to confirm our hypothesis that formulas are quite prevalent in /pol/’s discourse. All formulas combined appeared in over 418.000 posts, and the most popular occurs in more than 40.000 (‘orange man bad’). This seems low compared to /pol/’s millions of monthly posts, but these metrics are quite strict, not counting variants with different words or longer formats. It also only scratches the surface of a deeper well of formulas excluded by our protocol, like those with two terms, multiple stop words, or containing words also used in many other contexts, which were weighted down by NPMI.

More importantly, the formulas we found provide a distinct glimpse into /pol/’s subcultural character and how memecry manifests on the board. In broad strokes, the majority of the phrases (45) function to uphold a subcultural distinction by antagonising an outgroup or deploying adversarial ingroup rhetoric. The boundary-work (Gieryn, 1983) that is apparent in these formulas thus underlines Ong’s observation that secondary orality serves to build a ‘strong group sense’ among disparate collectives (1982, p. 133) While some formulas rise and fall rapidly, others are remarkably consistent, thus vindicating our hypothesis that memecry can serve as a strategy to stabilise 4chan folklore. At the same time, most formulas show at least some variation, thus assuring the vibrancy and distinctiveness of /pol/’s subculture. However, this vibrancy should not be taken as a sign that memecry goes hand-in-hand with cultural variation and progressiveness. If we are right in conceptualising memecry as a form of secondary orality, the meme’s form must change so that its cultural meaning can stay the same, thus carrying out a conservative function – a function we will see is especially pertinent on the far-right /pol/.

Outgroup antagonism

The last point is clearly visible in formulas geared towards antagonising or mocking a specific outgroup, which composed a third of our formula list (22 out of 65 formulas). This is in line with literature that characterises memes as vehicles for drawing in- and outgroup boundaries (e.g., Milner, 2016; Nissenbaum & Shifman, 2017; Tuters & Hagen, 2020) as well as work on subcultures that theorises them as defined in opposition to other communities (e.g., Gelder & Thornton, 1997; Muggleton & Weinzierl, 2003). In the case of /pol/, it is no surprise that these enemies include people of colour, immigrants, Muslims, the left, and, above all, Jews. Indeed, a third of all extracted formulas contained some kind of explicit far-right sentiment. For instance, 2019 saw a rapid rise of 4chan anons calling out ‘(((their))) globalist agenda’, referring to anti-Semitic conspiracy theories where Jews are seen as pulling the strings behind worldwide societal changes, in this case combining the dog-whistle ‘globalist’ (Benkler et al. 2018) with the triple parentheses used to call-out Jewish people or groups (Tuters & Hagen, 2020). The presence of several green flows in Figure 4 also shows how ‘globalist agenda’ formed a phrasal template onto which various outgroups could be attached: from ‘anti-white globalist agenda’ to ‘globalist NWO agenda’.

Figure 4. Double flow graphs for the formula ‘(((their))) globalist agenda’.

These memetic variations point to how even if the specific phrase ‘(((their))) globalist agenda’ was fairly new, its variations already pushed an anti-Semitic message before that. The other far-right formulas further underline this consistently reactionary character of /pol/. For instance, featuring some of the highest average NPMIs and absolute appearances is a highly stable formula mockingly impersonating Jewish slang: ‘oy vey goyim’ (Figure 5). ‘Oy vey’ refers to a Yiddish expression of exasperation and ‘goyim’ to non-Jewish people, combined to imitate a Jewish person berating non-Jews on /pol/. This formula is thus ‘oral’ in the most straightforward sense, as it transliterates an oral expression.¹² On /pol/, it is commonly used in reaction to other anons presenting conspiracy theories on a Jewish cabal, e.g., in the sentence ‘oy vey, the goyim know’.¹³ Figure 5 shows how ‘oy vey’ comprises other anti-Semitic formulas as well, notably with the variations ‘oy vey, oy gevalt’ and ‘oy vey, it’s annudah shoah’, again imitating Yiddish verbal communication.

Figure 5. Double flow graphs for the formula ‘oy vey goyim’.

While these antagonising formulas and their variations do change over time, what is clear from the entire corpus is how the outgroups themselves stay relatively stable. This has implications for 4chan/pol/’s culture as well as memecry as a subcultural dynamic. As we suggested by comparing /pol/ to preliterate communities, the very function of memetic formulas is to use repetition and variation in order to preserve the most important element of Internet subcultures from the ephemerality of digital media, or to ‘shield at least part of the content of memory from the transmuting influence of the immediate pressures of the present’ (Goody & Watt, 1963, p. 308). Similarly, on /pol/ memetic phrases are often employed to pass on conservative and reactionary ideas. This conservatism is in line with Ong’s theorisation of oral cultures, describing how constant repetition of norms invites a ‘highly traditionalist or conservative set of mind that […] inhibits intellectual experimentation’ (1982, p. 41). This paradoxical politics of memecry forms a double bind that is also representative of 4chan/pol/’s culture: an extremely conservative space that is nonetheless primed towards chaos and unpredictability, inspired by a profound opposition to traditional partisanship, leadership allegiances, and trite jokes (Donovan et al., 2022; Hagen, 2022). 4chan has often been framed as ‘countercultural’ (see e.g., Beran, 2019) and, at least for a while, a commitment to irony and ‘strict non-seriousness’ (Nissenbaum & Shifman, 2017, p. 487). Yet in the case of the secondary oral culture of /pol/, memecry mostly serves to keep the subculture variable in mere form rather than ideological underpinnings.

Adversarial ingroup rhetoric

This conservative function is even clearer in extracted phrases that displayed some form of adversarial ingroup rhetoric, again appearing in a third of all formulas (21 out of 65). It is long known how 4chan’s discourse is drenched in hostile insults and ‘rhetorical one-upmanship’ between anons (Phillips, 2015, p. 50). Indeed, ‘classic’ 4chan formulas like ‘lurk moar newfag’ (7.462 posts) and ‘tits and timestamp or GTFO’ (2.736 posts) also appear in our list, in this case referring to how new users (‘newfags’) should familiarise themselves with 4chan’s culture by reading before posting (‘lurking’) and how female users should post a picture of their breasts with a written date as proof of genuineness – or otherwise ‘get the fuck out’. It is well-observed how such formulas serve to socialise newcomers into the unwritten rules of the subculture (see e.g., Fathallah, 2021; Nissenbaum & Shifman, 2017), but insults are also a classic feature of oral cultures. As Ong reminds us: ‘standard in oral societies across the world, reciprocal name-calling has been fitted with a specific name in linguistics: flyting’ (1982, p. 43). More abstractly, in these adversarial ingroup formulas we see a manifestation of the ‘extraordinarily agonistic’ tone of oral culture, with its participants commonly jousting through ‘verbal and intellectual combat’ (Ong, 1982, p. 43). In her study of 4chan/b/, Phillips argued that 4chan’s vituperation was a concretisation of broader tendencies in Western discourse, marked by androcentrism and the idea that adversity is a sign of ‘competence, superiority, [and] power’ (Moulton, 2003, p. 149; Phillips, 2015, p. 23). This once again dovetails with Ong’s observation of how the ‘agonistic dynamics of oral thought processes and expression have been central to the development of western culture’ (1982, p. 45).

However, what is different on /pol/ compared to how authors have previously theorised ingroup hostility on 4chan’s /b/-board (Nissenbaum & Shifman, 2017; Phillips, 2015) is the density of conspiratorial suspicion. For example, the formula ‘gr8 b8 m8’, short for ‘great bait mate, I rate eight out of eight’, is used in reply to anons suspected to be maliciously ‘baiting’ for responses. While already used on /b/ in 2013, the phrase became one of the most consistently used formulas on /pol/. The same suspicion is also featured in other formulas like ‘low quality bait’, used to judge posts as obvious provocations. These phrases might be used in a tongue-in-cheek manner, but on /pol/, formulaic suspicion can also take on a more serious paranoid character. For instance, multiple extracted formulas were entangled with accusations of others being a ‘shill’, i.e., someone being involved in a conspiracy without disclosing it. This conspiratorial ingroup suspicion seems consistent over time even if the specific formulas change. For instance, we can see how ‘JIDF shill detected’ was taken over by variants ‘Sharieblue shill detected’ (Figure 6; in reference to Shareblue Media, a pro-Clinton media organisations). Coming full circle, this paranoia is so common on /pol/ as to evoke an opposite adversarial formula: ‘take your meds, schizo’, used to dismiss other 4channers as schizophrenic conspiracists.

Figure 6. Double flow graph of the formula ‘JIDF shill detected’.

The rise and fall of formulas

Using the temporal trends of our 65 formulas, we lastly want to discuss the ‘life cycles’ of formulas to illustrate the prey/predator game of memecry. Firstly, six of our formulas decreased in volume, slowly falling out of fashion on /pol/, while 23 displayed a short spike of usage. This evanescence is for instance the case with ‘check your privilege, shitlord’, a formula impersonating liberals calling-out those in privileged positions, where the pejorative ‘shitlord’ generally refers to a bigot on the Internet. On /pol/ the formula was tied to the ‘online culture wars’ that saw far-right agitators embroiled with feminist progressives (Massanari, 2017). As ‘check your privilege’ only became popular on platforms like Twitter in 2020, /pol/-anons had already exhausted variations on the formula, like ‘check your privilege, heliphobe’ (Figure 7),¹⁴ dropping the phrase altogether by 2018 (see Appendix 3). As such, the death of this formula illustrates both the time-bound character of some phrases, as well as the failure of memetic variation to ensure novelty and avoid mainstreaming. Yet, as we argued above, newer formulas were quick to fulfil the same antagonistic function, for instance through the most-used formula that became popular from 2017 onwards: ‘orange man bad’, which similarly antagonises liberals as engaging in mindless critique, in this case of Donald Trump.

Figure 7. The double flow graph of the formula ‘check your privilege, shitlord’.

Next to simply ceasing to be used, some formulas morph in and out of distinct shapes. In our corpus, this was most visible with ‘praise lord Kek’, a phrase referencing the Egyptian god Kek that became entwined with a semi-ironic religion harnessing the power of ‘meme magic’ (Asprem, 2020). Figure 8 clearly indicates how ‘praise lord’ indeed became associated with ‘kek’ in 2016 in the context of the U.S. Presidential election. With the so-called ‘cult of Kek’ quickly spilling over to YouTube and Twitter, the green flows in Figure 8 show how on /pol/, ‘praise lord’ came back to its earlier association with a generic religious discourse (e.g ‘praise our lord and saviour’ or ‘praise lord Jesus’). Such tracings form a vindication of the inherent duality of memes where ‘the demand to follow the rules and codes of meme use exists alongside a contrasting demand for innovation and creativity’ (Nissenbaum & Shifman, 2017, p. 493). In other words, here we see in action how formulas constitute ‘unstable equilibriums’ serving to anchor and move forward the subculture (Nissenbaum & Shifman, 2017).

Figure 8. The double flow graph for the formula ‘praise lord Kek’.

Conclusion

This paper introduced the notion of ‘memecry’ to define and investigate the dynamics of repetition-with-variation characteristic to Internet (sub)cultures. Memecry, we argue, plays a crucial role as a strategy to deal with the growing communicational evanescence of digital media. In the advertising-driven economy of digital platforms, contents are streamed at an increasingly rapid pace and with an ever-shorter memory buffer. In line with Ong’s theory of secondary orality, this acceleration generates cultural dynamics similar to those of pre-literate societies, which heavily relied on repetition-with-variation as a way to preserve their lore through the evanescent medium of sound. Likewise, Internet memes and memecry represent a way for online communities to uphold their culture in the increasingly ephemeral attention regime of digital media.

This paper provides an empirical investigation of memecry on the imageboard 4chan. We chose this space for its ephemerality and lack of memory, and we concentrated on the far-right /pol/ ‘Politically Incorrect’ board for its notoriety and cultural influence. We used a quali-quantitative approach, starting from 341 million posts circulating on the board since 2014 and devised a computational operationalisation of the classic notion of oral formulas: short phrases repeated-with-variation. We used text and network analyses to single out 65 prevalent formulas, operationalised as triplets that occur often together and with high specificity. We also extracted other words associated with the triplets, thus providing variants of the expressions. To demonstrate the temporal flux of these formulas, we published an interactive visualisation online.

As we hypothesised, formulas are common on 4chan/pol/, providing some empirical validation that memes form 4chan’s ‘locus of memory’ (Coleman & Brunton, 2010), just like formulas anchor oral cultures (Ong, 1982). More importantly, our extracted formulas encapsulated crucial features of /pol/’s collective identity. Distant and close reading revealed strong outgroup antagonisms and in-group frictions, both playing a crucial role in preserving the consistency of the community and socialising new members. Once again, like oral communities, the members of 4chan/pol/ use formulas to name and shame their imagined enemies. All these cultural traits are known to scholars of 4chan, but we empirically traced how they get compressed into a number of memorable and memeable expressions – exactly as one should expect in a (secondary) oral community. Moreover, while a dominant far-right sentiment on /pol/ was to be expected, our case study offers further support to studies that point out to the conservative social function of memes (e.g., Nissenbaum & Shifman, 2018), clashing with earlier optimism about their disruptive and democratising power (Tuters & Hagen, 2020). This also teaches that the ever-evolving character of memecry does not have to be accompanied by progressivism. At the same time, secondary oral online subcultures cannot remain discursively stable for too long, as our analysis showed how, without variation, repetition is not enough to preserve subcultural distinctiveness. While we made these findings on the basis of a fairly specific case – both concerning the platform and the textual focus – these conservative-yet-dynamic characteristics of memecry are likely common outside of our study as well, having been identified through the use of visual memes on 4chan (see Nissenbaum & Shifman, 2017) and on different platforms like Reddit (Massanari, 2015).

We may now also distinguish the memecry of secondary oral online subcultures from that of primary oral cultures. While the two share the same mnemonic role, they fulfil this function in very different settings. As observed by Ong, secondary orality is ‘not antecedent to writing and print, as primary orality is, but consequent upon and dependent upon writing and print’ (1982, p. 167). This entails, among other things, that secondary orality is more amenable to tracing and monitoring than primary orality. This is true for scholars (and indeed this traceability underpins this very paper), but it is also true for social actors. Where members of pre-literate societies have little means of tracking how their cultures change and spread, the networked character of online sociality means cultural diffusion is easy to spot. 4chan posters can use external archives of their forum and other platforms’ search engines to get a quick but fairly accurate idea of the spread of their ideas. And not only are they well aware of this possibility, they also strategically play with it. They proudly flaunt the virality of their memes, but are also quick to variate or abandon them when they become too popular.

This is a crucial element to keep this in mind when doing research on online subcultures, and particularly on the most toxic ones such as 4chan/pol/. Investigating these radical communities always comes with the risk of offering them increased visibility and legitimacy (Topinka et al., 2021). Yet what really provides online trolls the ‘oxygen of amplification’ (Phillips, 2018) is not so much their scholarly description, but the moral panic that it sometimes comes with. This is where the temporally-aware approach of this paper can help: instead of emphasising the unbridled spread of toxic ideas, temporal analysis can map how these ideas wither quickly as they drift away from their original subcultural enunciation. The acceleration of online attention cycles is the source of much toxic memecry, but it is also the reason for its generally short lifespan. Thus, even more than the oxygen of amplification, what should be denied to online trolls is the fuel of duration.

Acknowledgements

Thanks to Stijn Peeters, Mathieu Jacomy, and Erik Borra for technical assistance.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

The first author has received a PhD in the Humanities grant from the Dutch Research Council (NWO) under grant number PGW.19.030. The second author works has been supported by the European Union – Horizon 2020 Framework Program under the scheme ‘INFRAIA-01-2018-2019 – Integrating Activities for Advanced Communities’, Grant Agreement n.871042, ‘SoBigData++: European Integrated Infrastructure for Social Mining and Big Data Analytics’ (http://www.sobigdata.eu).

Notes on contributors

Sal Hagen

Sal Hagen is a PhD candidate at the University of Amsterdam, Department of Media Studies. During his MA in Media Studies, he co-founded OILab, a research group studying online political subcultures. His PhD concerns the collectivity of online subcultures. Methodologically, his work combines media theory with computational methods [email: [email protected]].

Tommaso Venturini

Tommaso Venturini is researcher at the CNRS Centre for Internet and Society, associate professor at the Medi@lab of the University of Geneva, and founder of the Public Data Lab. In 2017 and 2018, Tommaso has been a researcher at the École Normale Supérieure of Lyon and recipient of the ‘Advanced Research’ fellowship of the French Institute for Research in Computer Science and Automation. In 2016, he was ‘digital methods lecturer’ at the Department of Digital Humanities of King’s College London. From 2009 to 2015, he coordinated the research activities of the médialab of Sciences Po Paris [email: [email protected]].

Notes

1 As Ong notes, ‘sound exists only when it is going out of existence. It is not simply perishable but essentially evanescent, and it is sensed as evanescent. When I pronounce the word “permanence”, by the time I get to the “- nence”, the “perma-” is gone, and has to be gone’ (1982, pp. 31–32).

2 We identified spammy posts as having a post body of over 100 characters, and where the text sequence from the 15th to the 100th character appeared at least 25 times in other posts. We set this window to ignore reply-texts at the start of the post (e.g. ‘>>123456789’) and slight variations that are appended to spam texts to prevent deletion. Posts with ten or more newline characters were also skipped since they often involved copy-pasted lists.

3 We skipped the opening posts of these so-called ‘general threads’ by checking if the word ‘general’ appeared in the post’s subject (e.g. ‘/ptg/ President Trump General’). This filter did not catch all general thread opening posts, but was sufficient for the purposes of this paper.

4 We allowed for the following special characters: $ @ # & - — () * / since they sometimes have vernacular meaning in 4chan’s culture (e.g. ‘b&’ as a bastardisation of ‘banned’).

5 Stemming was carried out through nltk’s SnowballStemmer.

6 This list was compiled by combining (1) neologisms and contextually changing words extracted by Peeters et al. (2021), (2) vernacular imageboard terms collected by OILab, and (3) additional terms added by the authors (Appendix 1)

7 If one of the neighbouring words was also in the seedlist, we extended the window (e.g. ‘green, frog, kek, mighty, thank, pepe, frog, meme’), and considered all words in the window as co-words.

8 We filtered out all word pairs appearing in less than 0.0002% of the total posts per year (thresholds ranging from 30 to 117 posts) and with NPMI values below 0.18.

9 To filter out accidentally high-NPMI co-words (e.g. through misspellings), the absolute co-word count of each combination had to be fifty or more.

10 For the colouring, we converted the three NPMI scores between the root word and each of the three triplet words to a value in a RYB (red, yellow, blue) code, where an NPMI of zero (or a failure to meet the absolute co-word threshold of 0.0002% posts per year) meant a colour value of 0 and an NPMI of 1 a colour value of 255. We then normalised the values so that they added up to 1 and rounded them out so they equalled the colours of the graph’s legend.

11 For instance, the formula ‘remove [yourself] from the gene pool’ could simultaneously display adversity to other anons as well as carrying out broader boundary-work by alluding to genetic purity.

12 These written translations of oral accents are present in several other /pol/-formulas as well, notably with ‘dey dindu nuffin’ (stereotyping the pronunciation of ‘they didn’t do anything’ in an African American accent) and ‘hurr durr muh’ (used to slam a thoughtless argument, e.g. ‘hurr durr muh elections were stolen’).

13 The formula ‘Oy vey, the goyim know’ is designated as a ‘hate slogan’ by the Anti-Defamation League (n.d.).

14 ‘Heliphobe’ is in reference to the then-common copy-pasted sentence starting with ‘I sexually identify as an Attack Helicopter’, intended to mock discourse on gender identification and political correctness.

Appendices

Appendix 1. Seed list words

Table

word	stemmed_word	source
(((their)))	(((their)))	custom
(((them)))	(((them)))	custom
(((they)))	(((they)))	custom
(you)	(you)	custom
/ourguy/	/ourguy/	custom
/thread	/thread	custom
afc	afc	expert
ahmed	ahm	expert
aladin	aladin	expert
amerilard	amerilard	expert
anon	anon	expert
anti-white	anti-whit	neologism
autist	autist	neologism
awakening	awaken	expert
b-but	b-but	neologism
b&	b&	expert
b8	b8	expert
background	background	changing context
bait	bait	expert
baldcel	baldcel	expert
based	base	expert
bbc	bbc	expert
bc	bc	changing context
becky	becki	expert
beta	beta	expert
betabuxx	betabuxx	expert
bgc	bgc	expert
blackpill	blackpil	expert
block	block	changing context
bluepill	bluepil	expert
bonus	bonus	changing context
boomer	boomer	expert
brad	brad	expert
brap	brap	expert
breadpilled	breadpil	expert
btfo	btfo	expert
btw	btw	neologism
buck	buck	expert
buckbreaking	buckbreak	expert
bulgarbros	bulgarbro	expert
bull	bull	changing context
bump	bump	expert
burger	burger	changing context
butthurt	butthurt	neologism
cap	cap	changing context
captain	captain	expert
ceo	ceo	changing context
chad	chad	expert
chamber	chamber	changing context
chaos	chao	changing context
checked	check	expert
cheeto	cheeto	expert
chinz	chinz	expert
cnn	cnn	changing context
coalburner	coalburn	expert
commie	commi	neologism
coomer	coomer	expert
cope	cope	expert
copium	copium	expert
cracker	cracker	expert
cringe	cring	expert
crumbs	crumb	expert
cuck	cuck	expert
cucklefuck	cucklefuck	expert
cunnypilled	cunnypil	expert
cunts	cunt	neologism
desu	desu	expert
dindu	dindu	expert
dipshit	dipshit	neologism
district	district	changing context
doomer	doomer	expert
dumbass	dumbass	neologism
ebic	ebic	expert
entry	entri	changing context
error	error	changing context
ethnicel	ethnicel	expert
euro	euro	changing context
faggotry	faggotri	neologism
fakecel	fakecel	expert
femanon	femanon	custom
feminazi	feminazi	expert
flags	flag	changing context
foid	foid	expert
foundation	foundat	changing context
fpbp	fpbp	expert
framecel	framecel	expert
fren	fren	expert
ftfy	ftfi	neologism
gandy	gandi	expert
gf	gf	custom
gigachad	gigachad	expert
globohomo	globohomo	expert
glowie	glowi	expert
glownigger	glownigg	expert
golem	golem	expert
gook	gook	expert
gop	gop	changing context
goy	goy	neologism
goyim	goyim	neologism
grillpilled	grillpil	expert
gtfo	gtfo	changing context
heightcel	heightcel	expert
hopium	hopium	expert
host	host	changing context
hotwheels	hotwheel	expert
hugbox	hugbox	expert
hurr	hurr	neologism
hypergamy	hypergami	expert
idk	idk	neologism
incel	incel	expert
investigation	investig	changing context
irl	irl	neologism
jidf	jidf	neologism
jogger	jogger	expert
joot	joot	expert
judge	judg	changing context
kebab	kebab	neologism
kek	kek	neologism
khhv	khhv	expert
kikes	kike	neologism
kino	kino	expert
kys	kys	expert
landwhale	landwhal	expert
larp	larp	expert
leaf	leaf	expert
lel	lel	neologism
libtards	libtard	neologism
lmao	lmao	neologism
lookism	lookism	expert
looksmaxx	looksmaxx	expert
lulz	lulz	neologism
magic	magic	changing context
manlet	manlet	neologism
meatspace	meatspac	expert
meh	meh	neologism
mentalcel	mentalcel	expert
merchant	merchant	changing context
mexicucks	mexicuck	expert
mfw	mfw	neologism
miga	miga	expert
moar	moar	neologism
mods	mod	neologism
moor	moor	expert
moot	moot	expert
mossad	mossad	changing context
mudskin	mudskin	expert
mudslimes	mudslim	neologism
mujahiddin	mujahiddin	expert
murrishit	murrishit	expert
nasa	nasa	changing context
natsoc	natsoc	neologism
neet	neet	neologism
newfag	newfag	neologism
niggas	nigga	neologism
nigs	nig	neologism
nippon	nippon	expert
non-white	non-whit	neologism
normalfag	normalfag	expert
normie	normi	expert
nothingburger	nothingburg	expert
npc	npc	custom
npcs	npcs	expert
ntmaxx	ntmaxx	expert
nypa	nypa	expert
object	object	changing context
opposition	opposit	expert
orange	orang	changing context
orbiter	orbit	expert
ourgal	ourgal	custom
ourguy	ourguy	custom
oy	oy	changing context
panic	panic	changing context
patrician	patrician	expert
pattern	pattern	changing context
paypig	paypig	expert
pc	pc	changing context
pedo	pedo	neologism
picrel	picrel	expert
pics	pic	neologism
pill	pill	changing context
plz	plz	neologism
pool	pool	changing context
pussies	pussi	neologism
qanon	qanon	expert
qt	qt	custom
quadroon	quadroon	expert
radfem	radfem	expert
rapefugee	rapefuge	expert
rarted	rart	expert
reddit	reddit	neologism
redpill	redpil	neologism
rekt	rekt	neologism
right-wing	right-w	neologism
ring	ring	changing context
roastie	roasti	expert
root	root	changing context
rush	rush	changing context
sa	sa	changing context
sage	sage	expert
sandniggers	sandnigg	neologism
schizo	schizo	expert
shill	shill	neologism
shithole	shithol	neologism
shitpost	shitpost	neologism
shitskin	shitskin	neologism
shock	shock	changing context
sides	side	expert
silent	silent	changing context
sjw	sjw	neologism
sjws	sjws	neologism
skyking	skyke	expert
sneed	sneed	expert
soy	soy	custom
soyboy	soyboy	custom
spic	spic	neologism
stacy	staci	expert
storm	storm	changing context
strawman	strawman	neologism
stream	stream	changing context
suifuel	suifuel	expert
summerfag	summerfag	expert
tendie	tendi	expert
terf	terf	expert
tfw	tfw	neologism
thot	thot	expert
today	today	changing context
toothpaste	toothpast	expert
toxic	toxic	changing context
track	track	changing context
trad	trad	expert
tradthot	tradthot	expert
trannies	tranni	neologism
trap	trap	expert
trigger	trigger	changing context
trips	trip	changing context
unfathomably	unfathom	custom
unironically	uniron	custom
v&	v&	expert
vice	vice	changing context
virgin	virgin	expert
waifu	waifu	neologism
warning	warn	changing context
wizard	wizard	expert
wut	wut	neologism
wwg1wga	wwg1wga	expert
xd	xd	neologism
ylyl	ylyl	expert
you(s)	you(s)	custom
zygo	zygo	expert

Appendix 2. Formulas and short explanations

Table

formula	query terms	explanation	total posts	average npmi	temporal type	far-right sentiment	outgroup antagonism	ingroup (debate) antagonism
(((their))) globalist agenda	(((their)))__agenda__globalist	Used to refer that certain things are part of the "globalist" plans of a Jewish elite, where the three parentheses are used to refer to Jewish people.	1194	0.35	spike	x	x
(30) yr old boomer	boomer__old__yr	Used to refer to certain non-babyboom generations yet featuring their characteristics, e.g., by stating "millenials are just 30 yr old boomers".	1050	0.23	spike		x
(big) tiddy goth gf	gf__goth__tiddi	Phrase referring to the fantasy of having a girlfriend from the goth subculture with big breasts.	738	0.4	stable		x
AIDS Skrillex and Carl (the Cuck)	aid__carl__skrillex	Referring to two memes depicting stereotypical U.S. liberals that were attributed the names "Carl the Cuck" and "AIDS Skrillex".	1150	0.49	spike	x	x
alpha (or) beta male	alpha__beta__male	Used to refer to dominant "alpha males" and subordinate "beta males".	3508	0.39	stable
anti-racist is code for anti-white	anti-racist__anti-whit__code	Claim that (liberal) people use the term "anti-racist" as code language for "anti-white".	3804	0.39	stable	x
autistic histrionic screeching	autist__histrion__screech	Used to mock someone else's exasperation, especially that of other 4channers or concerned liberals.	513	0.38	spike
boomers and gen Xers	boomer__gen__xer	Used to refer to the generations of baby boomers and gen X.	1972	0.42	increasing
bump limit reached	bump__limit__reach	Used to indicate that a thread has reached the threshold of posts that prevents it from being sorted to the front of the board's index.	7701	0.19	stable
Cerification & Designation: BGC	bgc__certif__design	Used to state things are designated for a "Big Greek Cock", a phrasal meme usually associated with images of the Greek actor Tasos Nousias.	557	0.35	increasing
chad vs. virgin	chad__virgin__vs	Memetic format to juxtapose the stereotypes of attractive and sexually active "Chads" versus sexually unsuccessful virgins.	14563	0.29	increasing
check your privilege, shitlord	check__privileg__shitlord	Used to mockingly impersonate liberals telling other to be wary of their privileged position, where 'shitlord' generally refers to a bigoted person online.	2384	0.27	decreasing	x	x	x
child trafficking ring	child__ring__traffick	Referring to a network of paedophilic trafficking of kids, usually in a conspiracist fashion.	2986	0.3	increasing
cool clock Ahmed	ahm__clock__cool	Mocking phrase in reference to a 2015 incident where a digital clock of a student (Ahmed Mohamed) was mistaken for a bomb.	1279	0.17	spike	x	x	x
cope, seeth, and dilate	cope__dilat__seeth	Used to mock angry 4chan users, specifically telling them to "cope" with their loss, seeth, and letting their pupils dilate.	12737	0.32	increasing			x
cosmic background radiation	background__cosmic__radiat	Used to refer to cosmic radiation stemming from the early stage of the universe.	498	0.32	stable
CTR/Shareblue shill	ctr__shareblu__shill	Referring to alleged infiltrants (shills) of liberal- and Clinton-related organisations "Correct the Record" and "Shareblue".	578	0.35	spike	x	x	x
cut-rate memers (over at) reddit	cut-rat__memer__reddit	Part of the "Heh … not bad, kid" copy-pasta.	262	0.45	spike		x
desu baka senpai	baka__desu__senpai	A listing of Japanese terms commonly used on 4chan, where "desu" roughly means "it is", "baka" refers to "fool", and "senpai" meaning a master or teacher.	6410	0.44	spike
dey dindu nuffin	dey__dindu__nuffin	Racist saying mocking African-American accents, in this case "they didn't do anything".	3524	0.57	stable	x	x	x
dude weed lmao	dude__lmao__weed	Phrase mocking the enthusiasm of weed enthusiasts.	17941	0.21	spike		x
EUPHORIC PELICAN LMAO	euphor__lmao__pelican	Part of a copy-pasta depicting a pelican through ASCII art.	500	0.55	decreasing
full panic mode	full__mode__panic	Referring to people or groups displaying hysteria.	3292	0.28	stable
gas the kikes, racewar (now)	gas__kike__racewar	Anti-Semitic phrase calling to annihalate Jews and start a war between ethnic groups.	1013	0.26	stable	x	x
gr8 b8 m8	b8__gr8__m8	Bastardisation of "great bait mate", used to claim another post is maliciously provocative "bait" to evoke replies.	11264	0.61	stable			x
guac bowl merchant	bowl__guac__merchant	Nickname for Jeb Bush after he claimed to know how to make "a really good guacamole" and selling a guacamole bowl.	600	0.51	spike		x
hello, uh, (…) department?	depart__hello__uh	Used to act out if someone is calling the "based department" to signal someone's acts are approved and thus considered "based".	983	0.27	spike
homicidal gas chamber	chamber__gas__homicid	Anti-Semitic phrase referring to the gas chambers in WWII as "homocidal".	2650	0.39	stable	x
hugbox and echo chamber	chamber__echo__hugbox	Used to refer to certain (online) spaces that lack in contradicting or genuine opinions. Often used in reference to Reddit.	626	0.48	spike		x
hugbox cuntsucking reddit (friends)	cuntsuck__hugbox__reddit	Part of the "Fresh off the boat, from reddit, kid?" copy-pasta.	508	0.48	increasing		x
hurr durr muh	durr__hurr__muh	Used to mock an allegedly thoughtless argument, e.g., "hurr durr muh elections were stolen".	6308	0.42	stable		x	x
JIDF shill detected	detect__jidf__shill	Referring to alleged infiltrants (shills) of the Jewish Internet Defense Force.	331	0.19	decreasing	x		x
kissless handholdless virgin	handholdless__kissless__virgin	Used reference to a KHHV or "kissless, handholdless, hugless virgin".	724	0.6	increasing			x
latest shitpost copypasta	copypasta__latest__shitpost	Part of the "Heh … not bad, kid" copy-pasta.	272	0.16	spike
le epic (…) XD	epic__le__xd	Used to mock hyper-enthusiastic discourse related to Reddit, e.g., "le epic meme XD".	1757	0.35	stable		x	x
le happy merchant	happi__le__merchant	Phrase referring to the anti-Semitic meme depicting a grossly stereotyped Jewish salesman, "the happy merchant".	2740	0.21	decreasing	x	x
lel, top kek	kek__lel__top	Used to express joy, where "lel" is a play on "lol" (laughing out loud) and "top kek" being roughly synonymous to "lol".	391	0.18	decreasing
low quality bait	bait__low__qualiti	Used to indicate a 4chan post is easy to recognise as malicious provocation.	9220	0.23	stable			x
lurk moar newfag	lurk__moar__newfag	Used to command inexperiences 4channers ("newfags") to observe the culture of the board without posting ("lurking").	7462	0.46	increasing			x
magic sky wizard	magic__sky__wizard	Used to mock the idea of a god by ridiculing it as a wizard in the sky.	837	0.3	decreasing
mayn meteor stream	mayn__meteor__stream	Used to mock British pronunciation of "mainstream media", specifically by YouTuber Paul Joseph Watson.	168	0.39	spike
mossad false flag	fals__flag__mossad	Used to argue something is a staged "false flag" operation by the Israeli secret services.	6357	0.26	spike
NEET basement dweller	neet__basement__dwell	Referring to "beta" men living in their parents' basement and who are "Not in Education, Employment, or Training".	2020	0.38	stable			x
nice strawman argument	argument__nice__strawman	Used to ironally praise someone's argument when it features a strawman fallacy.	553	0.17	stable			x
orange man bad	bad__man__orang	Used to mock criticism of Donald Trump as mindless imitation.	44940	0.24	spike	x	x	x
oy vey goyim	goyim__oy__vey	Anti-Semitic phrase mimicking Jewish people, where "oy vey" is a Yiddish saying to express exasperation and "goyim" refers to non-Jewish people.	24864	0.58	stable	x	x
praise lord Kek	kek__lord__prais	Used to praise the deity "Kek", who became entangled with the semi-ironic religion of the "cult of Kek".	2932	0.23	spike
prep the bull, sven	bull__prep__sven	Racist meme calling Swedes ("Svens") to orally please a black person in preparation of them having sex with "Sven's" partner.	943	0.38	spike	x	x	x
prepare for the storm maggot	maggot__prepar__storm	Part of the infamous "navy Seal" copypasta.	4600	0.38	stable
qt asian waifu	qt__asian__waifu	Used to refer to a "cute asian wife".	524	0.23	stable
read Mein Kampf	kampf__mein__read	Used to state one has read Hitler's Mein Kampf, or as a commandment that others should do so.	16609	0.48	stable	x
remove from the gene pool	gene__pool__remov	Used to argue certain people or groups should be annihalated, i.e., "removed from the gene pool".	11111	0.33	stable		x	x
Seed's Feed & Sneed	feed__seed__sneed	In reference to an episode of The Simpson's where a shop is called "Seed's Feed & Sneed (formerly Chuck)" and often used in an intentionally pointless fashion.	31004	0.47	spike
slide thread, sage	sage__slide__thread	Referring to a thread that "slid" from the original topic of conversation, where "saging" means replying to a thread without bumping it, thus contributing to quicker archival.	26660	0.33	increasing
sloppy job mossad	sloppi__job__mossad	Used to point out an alleged bad insider job by the Israeli secret services.	12445	0.4	spike	x		x
take the red pill	pill__red__take	Command to become "redpilled", meaning awaking from a world of lies (in reference to The Matrix).	15318	0.25	stable	x
take your meds schizo	med__schizo__take	Used to tell another 4chan user they are paranoid or obsessive.	28321	0.24	spike			x
tfw no qt gf	gf__qt__tfw	Short for "that face when no cutie girlfriend". Often used in combination with a meme depicting a face with a recognisable emotional expression.	6800	0.34	stable
thanks based poseidon	base__poseidon__thank	Referring to thanking "based Poseidon" for letting immigrants drown, where "based" roughly means "cool".	5659	0.22	spike	x
third world shithole	shithol__third__world	Derogatory name for countries labelled as third-world.	34262	0.28	stable	x
thirsty beta orbiter	beta__orbit__thirsti	Phrase used to mock men who linger around women (i.e., "orbiting" around them).	287	0.39	spike			x
tippity top kek	kek__tippiti__top	Conjunction of things being "tippity top" and "top kek", with "kek" roughly meaning "lol".	576	0.39	spike
tits and timestamp or GTFO	gtfo__timestamp__tit	Used to tell off female users of 4chan, commanding them to show their breasts with a written-down timestamp (as proof of genuinity) or otherwise to "get the fuck out".	2736	0.45	increasing			x
wave the confederate flag	confeder__flag__wave	Used to say one will wave the flag of the U.S. Confederate States.	1412	0.29	stable
yo, (…) finna boutta	boutta__finna__yo	Racist phrase imitating African-American accents, in this case the phrase "yo, (…) finally about to".	197	0.54	spike	x	x

Appendix 3. Counts of “check your privilege” on 4chan/pol/ and Twitter

Figure A9. Tweets and posts on 4chan/pol/ containing ‘check your privilege’ (tweets and retweets).