https://orcid.org/0000-0002-1669-2377
![Sample our Humanities journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/5939/TOC-Subject-Sample-2021-Humanities.jpg"})
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.
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).
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’ (Citation1982, 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. (Citation2021), (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.
Additional information
Funding
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]].