Method and methodology are concepts that are used on a daily basis by social science students, researchers, and lecturers. Interestingly, for many students (and also researchers), both concepts have acquired a synonym-like quality. It is not exceptional to read in papers, theses, and published books a description of a method under the ‘methodology’ heading. Seen from a linguistic, anthropological, sociolinguistic, ethnographic, or in general an empirical-interpretative perspective, that is as remarkable as it is sometimes troubling. To cut to the chase: method and methodology are not synonyms. This truism is not another case of semantics either. A methodology encompasses all steps in your research. It includes theoretical perspectives, the formulation of the research question(s), the definition of what counts as data and what does not, and the methods used. A method can clearly never replace a methodology. This perspective on methodology is at the heart of our practices as empirical analysts of human behavior and has a huge impact on how we do (or should do) science, and thus on how we produce knowledge. 

This little note on methods and methodology doesn’t attempt to be exhaustive, novel, or to contribute to the development of the field of digital discourse analysis, let alone that it aspires to any groundbreaking interventions. The goal is far more modest. I want to insert the old ethnographic principle into the field of digital discourse analysis. In my understanding, methodology is ontological and epistemological perspectives and methods in one. In this merger, the perspectives are foundational and should lead us in our choice of methods and how we deploy those methods. I decided to write this note after many years of supervising student research. The idea is to explain to students what the difference is between a method and a methodology and what it says about our practices in the production of knowledge. Even though it was written for students, I can only hope that it is useful for a broader audience. 

The reality of the empirical world

Before I jump into this little exercise, it is important to make explicit that I write this as a scholar that is interested in studying, and understanding, ‘the reality of the empirical world’. This means that I position myself in a long tradition of anthropologists, ethnographers, and sociolinguists (think of people like George Herbert Mead, Herbert Blumer, Howard Becker, Dell Hymes, Clifford Geertz, Jan Blommaert, Gunther Kress, and many others) that embrace an inductive empirical-interpretative mode of producing knowledge. All these scholars share the idea that if one wants to understand the complexities of social life, one requires detailed studies of human interaction in specific situations, in context.

Taking that empirical world seriously means that ‘it is the point of departure and the point of return (…) It is the testing ground for any assertions made about the empirical world’ (Blumer, 1969: 22). The production of empirically grounded knowledge is thus always established in a dialogue with the reality ‘out there’. The knowledge that we produce of that world, should in the last instance be verified in that world. And what we call methodology, covers the principles that underlie and guide the process of our analysis of that world. From our first entry in the (digital) field until we check our conclusion with the field (see for instance Van der Aa & Blommaert, 2016): it is all informed by our methodology. Methodology applies to and covers all parts of our research of the empirical world. All research starts with our perspective on the world and on what counts as ‘knowledge’.

Methodology

The first point I want to elaborate on is that a methodology is loaded with (1) an ontology and (2) epistemology (Blommaert & Dong, 2010): it comes (1) with a specific perspective on the world, and the role of humans in that world, and (2) with a specific perspective on how knowledge is produced. In a digital ethnographic discourse analytical approach to digital data, this means that: 

1. we see language not just as a collection of words, or a thing in itself, but as the architecture of society. It is through language that human beings become social, and build groups and societies. It is also through language that humans are able to interpret actions, reflect on themselves, and give meaning to objects. The empirical world is thus deeply connected to the way people shape it in words and practices.
2. studying digital discourse is studying digitally mediatized language-in-action. It is through that action that people shape the social world. Identities, cultures, groups, and objects in that world are discursively constructed. 
3. in post-digital societies, language-in-action is increasingly ‘mediatized’. This is not a trivial finding. Mediatization is not neutral: digital media are actors that actively direct and partially shape discourse. They categorize, select, and make messages visible and invisible. Technologies are not just neutral tools, but intermediaries that transfer messages. They are mediators and thus have an active role in the production and circulation of knowledge.
4. digital discourse is the result of the interaction between humans, platforms, and increasingly also non-human actors like algorithms and datafication processes (see Maly, 2022). Digital discourse is a socio-technical assemblage (Bucher, 2018), and analyzing it as such means analyzing discourse not just in relation to the social, cultural, political, and economic context, but also the technological context (Varis, 2016). 
5. we also realize that discourse is not only produced by humans anymore but also by algorithms (see Maly, 2022 for more information)
6. as a result of the previous, digitally mediated discourse has no linear, direct and symmetrical sender-receiver structure anymore (Maly, 2018; Blommaert, 2020). Algorithms, bots, and datafication are all part of our digitally mediatized interactions and influence the input, uptake, and flow of our communication. 
7. digital platforms are not necessarily stable entities. On the contrary, they are continuously changing to meet the goals of their owners, the requirements of governments, and the wishes and practices of their users (Bucher, 2018). 

Methodology and how we look at the world

This perspective on the world, and the role of humans and platforms in it, has a substantial effect on what we study, how we study it, and what we see as data and what we do not. We study the world from the viewpoint of humankind, and how humans in interaction with (human and non-human) others and in and through (active) objects shape the world. We understand our own world and our worldviews as constructions that have a social, cultural, and political history and in which humans are seen as having agency. People construct themselves, their worldviews but also societal structures and objects that in turn influence the world. 

What does this now mean in our actual practice as researchers? Let’s make it (more) concrete. In the case of trying to understand the role and function of right-wing influencers, we – taking this perspective on board – assume that their communication on social media has an impact on the world. It is not ‘just words'’. Words – and especially words that circulate on a large scale – co-construct the world. The more reach, the more serious that impact can potentially be. Understanding that impact of language forces us to take it very seriously, and carefully analyze that language in its context, including paying close attention to the ‘natural histories of discourse’ (Silverstein & Urban, 1996). Let’s illustrate this with an example of the dating app The Sauce. When that app highlights three positive reviews on its website, we should realize that those reviews are not just reviews: they are taken out of their original context, and re-contextualized in a marketing talk. It is thus not a coincidence that all three of the reviews are positive: they are now used to sell the app as unique, safe, fun, and cool. Ergo, the perspective on mediatized language-in-action forces us to understand language in its context. That means that we look at how it is used, for which goals, at its genre, but also at the history of that (type of) discourse, and so on. Details of actual interactions matter and if we want to do justice to the empirical world, they have to be analyzed meticulously. 

The researcher as part of knowledge production

The key here is: we see more because we have a methodology. With a methodology, we enter the field and look at data with specific images that form lenses through which we encounter the world. This is not exceptional. We all encounter the world through an image that we have of that world. In that specific sense, scientific research is not that different from normal life. All social research is grounded in schemes and pictures to make sense of the world. The difference between our personal images and our professional vision is that the latter is produced according to a methodology, a rational discourse embedded in institutions and scientific traditions. Important here: we enter the field with our own personal biases and our professional vision. All social scientists, whether they acknowledge it or not, are part of the process of knowledge production, with all its pros and cons. We can only see, and thus also study the empirical world, through schemes, images, and ideas that we have about that world. Those pictures will shape our research questions, what we see as data and what we do not, and the relations we see in the data. It becomes visible on our premises. Ethnography is in that sense, said Hymes, (1996: 13), “continuous with ordinary life. Much of what we seek to find out in ethnography is knowledge that others already have. Our ability to learn ethnographically is an extension of what every human being must do, that is, learn the meanings, norms, patterns of a way of life. From a narrow view of science, this fact may be thought unfortunate. True objectivity may be thought to be undermined. But there is no way to avoid the fact that the ethnographer himself or herself is a factor in the inquiry. Without the general human capacity to learn culture, the inquiry would be impossible.”

Ergo, our own subjectivity cannot be surpassed, and the answer to it is not to ignore it as if it doesn’t exist, but to be aware of it and take measures to minimize unwarranted and unwanted biases. Talking about biases.  In the digital world, we encounter an algorithmically mediated world: what we see as ‘online reality’ when we open our laptops is closely related to our own algorithmic identity (Blommaert & Dong, 2019). This digital bias is a structural fact of the digital context, and should thus be taken into account. A first step in that direction is to take ourselves, our practices, and our roles seriously. If we as researchers are part of the production of knowledge, it should be clear that the more we know, the more we will notice and take with us. A social scientist that doesn’t read, just like social scientists who don’t go out in the field, will always produce superficial knowledge no matter how hyped their new method is. And social scientists who are not reflective and critical of their own biases risk reproducing them in their data and results. Identifying and assessing these biases is a key element of our methodology. The more we are outsiders of a particular field, the more we will have unnuanced images – or stereotypes in plain English – and thus the more we will need to become aware of these biases. To make the strange familiar (and if we engage in auto-ethnography to make the familiar strange) is one of the known mottos among ethnographers.  One ‘trick’ that Howard Becker introduces to counter our biases, is to focus not on categorizing people, but on their activities (Becker, 1998: 45). The focus on what people do and say in particular contexts allows us to understand behavior, and it lets reality talk back to us.

Methodology and research questions

Our ontology and epistemology not only affect how we see the world, but they also have a profound effect on what we consider as data, on our research question, and on the methods we choose. The formulation of research questions is not a trivial affair. On the contrary, it is in the interaction between the empirical world, our ontology and epistemology, and the existing theories about the world that we can formulate our research interests, define problems that deserve to be analyzed, and eventually shape our initial research questions. The research question needs to be formulated in such a way that it is answerable by studying the empirical world.

Researchability, as Verschueren (2010: 22) calls it, ‘does not only imply that it is possible to find empirical evidence supporting a finding or an answer, but also that a systematic ‘counterscreening’ of one’s data is feasible, by which identical methods can be applied to the same materials to see if there is any evidence contradicting the finding or the answer.’ ‘Counterscreening’ is of course not always possible. Even if the same data and the same methods are used, the end result can be completely different. In inductive empirical-interpretative research, the researcher has, as we just saw, an active role. Depending on the researchers, their experience with the field, and their previous research and specializations, they can see entirely different things than other researchers. Counterscreening, in our profession, is most of the time limited to checking if the existing analysis is supported by the data (even if a different analysis with a different focus is also possible). 

Taking the empirical world seriously also means that research questions should reflect problems we identified in the real world. It is thus not unusual in empirical research to first enter the field in order to explore the world. It is this exploration that allows us to detect problems that inform research (and research questions). Of course, when we enter that explorative phase, we do not enter it empty-handed or empty-minded. In our backpack are what Van den Hoonaard (1997:2) calls ‘sensitizing concepts’, that function as 'a starting point in thinking about a class of data of which the social researcher has no definite idea and provides an initial guide to her research.’ Sensitizing concepts thus suggest possible directions where to look and what to look for (Bezemer, Murtagh, Cope, Kress & Kneebone, 2011).

Research on right-wing influencers as an example

In the case of a student research project on very successful rightwing influencers, I did the explorative research and handed the students not only specific literature on Junk News, (Venturini, 2020), metapolitical influencers (Maly, 2020), quantified storytelling (Georgakopoulou, Iversen & Stage, 2020), hybrid media system (Chadwick, 2017) and algorithmic agency (Bucher, 2017). The literature and the classes were full of ‘theory’, ‘perspectives’, and ‘methodologies’ that directed the gaze on social reality. Those concepts foreshadow problems that are potentially worthy of looking at in depth. I also set an initial research question: “How do right-wing influencers use mainstream digital media to have a voice in the hybrid media system”. This research question was deeply rooted in:  

1. An exploration of the empirical world. The question is linked to the fact that right-wing influencers like Ben Shapiro, Danielle D'Souza Gill, or Dan Bongino are regularly among the most successful content producers on Facebook, even though they are accused of producing not only highly partisan viewpoints but also conspiracy theories, misinformation — or so-called fake news. This is exceptional, considering the fact that since 2016 mainstream platforms like Facebook, Instagram and Twitter have made claims to take fake news and disinformation seriously. 
2. Theory and our methodology. The assumptions of the research question are abundantly clear:

(a) there is something like a hybrid media system (Chadwick 2017), of which digital media make up a substantial part;

(b) (meta)political activists adopt influencer practices (Lewis, 2018; Maly, 2020 & 2022, Leidig & Bayarri, 2022);  

(c) and through that adoption, they succeed not only in articulating their voice, but also manage to make that voice salient in the digital ecology, and sometimes in the hybrid media system in general; 

(d) digital media are part of this voice: it is only because they use the affordances and their algorithmic knowledge (Maly, 2019a) in full, that they can realize such a reach;

(e) and last but not least, this is important because it has societal, political, and cultural effects. The moment someone succeeds not only in articulating their voice but in letting that voice be heard online (ergo realize virality), they have discursive power. The impact of that power is real. Digital media allows them to 'educate' ten to hundreds of thousands people through their posts which can have all kinds of offline effects as is visible in elections, protest marches, and even violence.

The embeddedness of the research question in our methodology allows for rigorous testing of our own assumptions, and of existing research. First of all, it forces us to understand what the influencers are actually doing: which discourse they produce. How they produce discourse, how they use the affordances of the platforms, which affordances they don’t use, how they adhere to the community guidelines, what type of content they post, and what content is completely absent. All that data is available for collection, description, and interpretation. 

Collecting data and methods

Depending on the part of reality that we want to study, our methodology, and our research questions, we will determine which type of data needs to be collected and how we want to collect data. Data selection or sampling is crucial to all scientific research. We cannot study every case that is relevant to answer our research questions. Ergo, we are always selecting data or cases and that selection is all about relevance vis à vis our research question, and thus about the potential generalizability of our research results. The goal of empirical research is, ideally, to produce outcomes that are relevant beyond the specific cases we are studying. Over time, different ways of sampling have been used, all with their pros and cons (see Becker, 1998 for a discussion). The key here is the construction of a corpus of data that is fundamentally connected to our research question. What counts as evidence and data in empirical research is not trivial, nor is it to be considered neutral in the light of different research paradigms. On the contrary, it is precisely in what is considered as ‘data’ and what not, and how to collect it, that one can see different methodologies and thus analytical practices (see Blommaert, 2004). 

The selection of data

If we for instance want to study the ideologies embedded in a social medium like GAB, we can go in many different directions. One way is to analyze all the interviews of GAB's CEO Andrew Torba and focus on how he explains the goal of GAB. Of course, that data only allows us to analyze how the CEO wants us to see GAB. That can be important, but we should realize that the actual functioning of GAB can be very different. Collecting and analyzing all posts of GAB is, considering time limits, usually not an option. We could then decide to focus on the topics that pop up on the newsfeed for one month, using an empty account. This would at least allow us to gain insight into the type of content that is to be found on GAB and is shown to new members without a click, search or watch history. Or we could analyze the community guidelines and the actual cases of moderation. Of course, all those decisions will cause specific difficulties (in access to the data, managing the amount of data, or limiting what we can say about the data). They will also produce different types of knowledge. To highlight the paradigmatic, and thus the methodological impact of our data selection we can give another example. If – in light of our research question on right-wing influencers we talked about above – we limit our data to the actual content that influencers produce, we not only take a decision with far-reaching consequences (Thompson, 1990) would call it the fallacy of internalism), we also – consciously or not – adopt a very specific methodology that assumes that meaning-making is not produced in interaction, but is to be found in the original text itself. Such decisions have a huge impact on the end result on the status of our research results. 

The selection of our data is not trivial, and not just a matter of ‘method’, but clearly also of methodology. It is useful to highlight the difference between method and methodology here once more. In order to collect data, we need to choose a specific method to collect them. This can range from old skool observations and interviews to big-data-oriented approaches. Depending on the data we collect, and the method we use to do this, we maybe have to change our research question and our methodology, or vice versa. The methods we use, should be consistent with how we look at the object of study. That of course means that not all methods are compatible with the perspective on human and non-human behavior I outlined above. If we don’t make our methodology explicit, and just suffice by choosing an in-vogue method, we may implicitly accept a certain paradigm that contradicts our own methodology. Each method is connected to a specific ontology and epistemology, and all methods have their own biases and assumptions built into them. Methods should thus be carefully evaluated in relation to our data and our methodology.

Big data 

Take for example one type of data that in the last decade has been surrounded by an awe-factor aura that lures in funding and generates praise: so-called Big Data studies. As boyd & Crawford (2011) noticed, big data are associated with the ‘widespread belief that large data sets offer a higher form of intelligence and knowledge that can generate insights that were previously impossible, with the aura of truth, objectivity, and accuracy.’ The method and the data that are generated by the method are seen as valuable in themselves. Big data would allow us to map exactly what individuals are and what they do. Social media platforms fuel this myth for profit goals, and their critics support it (see for instance Zuboff, 2019) by supporting the idea that platforms indeed succeed in getting to know us better than ourselves. The assumption of such big data approaches is that if we have enough data, we can know people and groups. Seen from an ethnographic point of view, that might not be the case at all, as big data alone cannot account for context, the deep cultural understanding of a tweet or Instagram post (see Maly, 2019b and Burgress e.a. 2022 for a critique). What is called ‘content analysis’ in big data studies is in reality not much more than fragmented descriptions of words used by the people under study. In other words, from an ethnographic perspective, big data can never replace ethnographic immersion in the field. 

Data, methods and methodology

Such method hypes are not new. Blumer’s article on the methodology of symbolic interactionism can be read as a paradigmatic critique of such popular new methods that suggest that the empirical world can be made to fit the new models. His provocative claim that ‘the prevailing mode in the social and psychological sciences is to turn away from direct examination of the empirical social world and to give preference, instead, to theoretical schemes, to preconceive models, to arrays of vague concepts, to sophisticated techniques of research and to an almost slavish adherence to what passes as the proper protocol of research inquiry’ (Blumer, 1998: 33) reads as a critique on the contemporary turn towards ‘big data’ research. From the point of view of symbolic interaction, but also from an ethnographic, linguistic pragmatic, or sociolinguistic approach, big data are not and cannot be substitutes for explorative and thus slow research that allows us to understand human interaction in its context. In essence, technologies don’t change the basic fact that empirical social scientists research how humans interact in the world. From this perspective, a hit-and-run approach will miss the essence of empirical social research. Digital discourse analysis, seen from such an empirical-interpretative perspective, should take the following issues at heart:  

1. Just like in the 20th century, qualitative approaches and explorative research is still a conditio sine qua non. It is in this phase that we familiarize ourselves with the group of individuals we are studying. In the case of our research on right-wing influencers, it meant understanding their discourse, how they use digital media, and how they think about their practices on those media. Understanding their discursive behavior in a social, economic, political, cultural, and technological context is a step that we cannot skip. There is no substitute for context. 
2. Context is of course a very different thing in digital ethnographic discourse analysis (see Varis, 2016) than it was in traditional offline ethnography. Offline ethnography usually meant long periods of very close interaction with people. Ethnographers ‘got to know’ the people they studied, sometimes in the most intimate details. Online it is difficult to exactly know who we are interacting with. People are anonymous, use pseudonyms or lie about who they actually are. Especially in highly politicized contexts that is becoming a quite common affair, even on platforms that claim to enforce a real name policy. This poses quite some difficulties if we adopt a classic ethnographic perspective. 
3. As a result, certain methods have become far less successful in gathering data. The interview is one such example. Doing online interviews in the context of empirical research is difficult, but not impossible. Bucher’s (2017) research on the algorithmic imaginary reached out to people tweeting about ‘Algorithms and Facebook’ and ‘Algorithms and Twitter’, and then interviewed them on how they saw the impact of algorithms on their lives. This type of interviewing is very different from interviewing people you lived with for months, maybe years in a row. How people behave and communicate in email, or in a Messenger app of course hasn’t got the same status as a more intimate information session. At the same time, it can bring in new data that complement the observation already made. When one deals with ‘sensitive topics’, it can be even more difficult to actually find people willing to answer your questions face to face. Symbolic interaction can lead the way here. If we shift our attention from focusing first and foremost on ‘the identity’ of humans online, to their behavior (Becker, 1998) we can open up a new research agenda that focuses on the actual practices and discourses that create specific groups online. It is through these discursive practices that people build communities, and contribute to the political goals of these communities. 
4. Taking human behavior in digital ecologies seriously means understanding them as socio-technical assemblages. The technological context not only makes it difficult to get to know the participants and informants, but the digital platform is also an actor that should be part of the scope of research. Understanding the impact of those non-human actors in the production of discourse is usually understood as the black box dilemma (Pasquale, 2015). As the algorithms are corporate secrets, researchers have no access to the way in which those algorithms categorize discursive behavior and how that influences uptake, circulation, and eventually meaning-making. Bucher (2018) invites us to move beyond the black box metaphor and see algorithmic power as relational: the technical construct only acquires power in the sense of directing behavior and shaping our view of the world in relation to how people see and interact with them. It is in this interaction that algorithmic agency is shaped. And just like we understand human behavior as having structure, we can see technical and socio-technical behavior as non-random and structured and thus analyzable. The technical algorithms may not be visible, but we can observe human behavior in that new environment and the outcomes of that behavior. 
5. Analyzing socio-technical behavior starts from the notion that the environment in which people interact is a very ‘specific’ and layered environment. La Violette (2017) argued in relation to Reddit, that digital platforms resemble a ‘Goffmanian frame', as it “circumscribes” discourse and becomes crucial to its interpretation (Goffman 1974). But of course, we do not observe the behavior of all people on Facebook or Facebook in general but specific actors sharing specific spaces in a specific type of social media that allows certain ways of interacting and thus co-creates the culture of the page. The digital field is thus layered, stratified, and polycentric. The page of Ben Shapiro on Facebook is not the same as the one of Daniele D’Souza Gil, and as a result, not only Facebook in itself functions as a Goffmanian frame, but also the owners of that space use the affordances to shape a frame in which we should understand the discursive practices of the followers. A case study approach allows us to contextualize discursive behavior not just in a generic ‘Facebook’ context, but in the context shaped by the admin of the page, and the specific posts. 

What does this now mean? It means that our methodology shapes what we see as data, and what are relevant methods to collect the data. When we collect data, we should thus be very conscience about our methods and how they influence the characteristics of our data. Text is not enough when we analyze digital discourse. From an ethnographic point of view, everything is data, and that is still a good reflex. That may seem a strange statement, but it touches upon the heart of the ethnographic methodology: it forces us to be very conscious about what we exclude from our data set as ‘not being data’. When we exclude uptake from our data, we are in reality not studying interaction or communication, but just language.

Collecting data

The attention to what counts as data and what not should be evident now: it could be that what is categorized as non-data has played a fundamental role in the meaning-making process, and is thus crucial in our understanding of what is going on. Of course, saying that ‘everything is data’ doesn’t really help students in gaining insight into what they should do. And truth be told: in practice we always make selections. Maybe the most important is to not a priori exclude something when it pops up and seems relevant. Let me give an example. In our research project on right-wing influencers, students combined different methods of data gathering: (1) observation, (2) scraping, and (3) interviews. 

Firstly, students engaged in a short-term covert ethnographic observation of the different social media accounts of Ben Shapiro and Daniele D’Souza for six weeks. Within that period they systematically collected data for one week: that means that within that week each post was screenshotted, as well as and all the comments and interactions. This method is clearly inspired by our methodology that sees language-in-action, and thus interaction as our object of research. Our initial starting point was Facebook, as those influencers had a particularly huge uptake on Facebook. But that limitation was exactly that: a serious limitation, as it became clear that those influencers were also active on Instagram and Twitter and that they used those platforms for different goals. Their message production there was important because it shed light on their Facebook communication and co-constructed the overall strategy and thus the meaning-making process. We also couldn’t ignore the hyperlinked nature of digital discourse. Shapiro and D’Souza use their social media to drive audiences to their own shows on their own platforms and to the media they work for. The posts on their socials are thus not just their comment and the link including the automatically generated abstract, picture, and title of the article, the linked article itself is also part of the communication and thus worthwhile to analyze. Of course, the article is not the post, and not all his followers will click on all the links. Point is, that the distinction between what is visible on social media and on their own platforms can become very relevant in the analysis, and maybe in understanding why they succeed in being integrated into mainstream digital media, even though they break the community guidelines on their own platforms. Following the data principle is especially important considering that data was collected with a focus on: 

1. interaction and a specific socio-technical and ethnographic understanding of interaction. Concretely this meant that students focused on the input (how and what Shapiro and D’Souza Gil communicate, including the different modes and hyperlinked character of their online communication), the uptake (how and who interacts with the posts), and the use of affordances of the platform. 
2. and on understanding how these influencers can be known for spreading ‘fake news’ and at the same time be among the most influential influencers on Facebook. This meant understanding how these influencers managed to adhere to the community guidelines, while at the same time producing ‘fake news’ or at least 'junk news' (Venturini, 2020). In order to do that it was important to understand what exactly they post on social media, but also see how this links to the article that they posted.

To collect the data, students complemented the six weeks of observation with live observations: following and mapping the interaction with a post in real-time. This type of live research brings ‘time’ into the equation: it looks at how people in real-time interact with posts, and how this influences not only meaning-making but also uptake and spreadability. This live research is embedded in a longer-term ethnographic observation (see Maly, 2019 for more information). The combination of this ‘short term’ immersion in the digital field and those live ethnographies was the foundation of the data collection, but it was further enriched by a more ‘big data’ approach and interviews. We scraped the Twitter accounts during the same week. Contrary to the multimethod approach of Shahin (2016) not the algorithmic research, but the explorative research was leading here. The algorithmic method was used in a second stage as a triangulation method and as a method that creates a new type of data. Such ‘big’ data sets do not replace the explorative research, or the detailed manual data collection, but they can be useful to shed light on the scale of things and allow us to upscale our understanding of cases that were analyzed in detail. These ‘big data’ are not used to ‘study the discourse’ of the influencers, but to get a more general idea of the uptake and understand the network around the influencer in question. In this way, big data collection injects new insights by which we can further deepen our methodological perspective. From the start, we stressed that uptake was crucial, as it is in the uptake that meaning is made. Uptake is important because (1) it changes the meaning of the post, but also because (2) it contributes to the spreadability of the post and as such (3) contributes to producing specific input discourse, as influencers adjust to the digital culture of the platform to maximize uptake. The big data approach allows us to research this on a macro scale.

But big data alone was not enough to understand uptake: likes, shares, comments, and retweets clearly matter, but the question is also: what do they take up? How do comments resonate with the input? Do the reactions create new frames to understand the input-discourse or not? Which comments are highlighted by the platform etc.? All this begs for a detailed discourse analysis. Next to our observations, we used interviews to deepen our understanding of why people interact with influencers. Students tried a similar method as Bucher (2017) to understand why people were interacting with the posts of Ben Shapiro and Danielle D’ Souza Gill and contacted people who were commenting, liking, and sharing the posts of influencers. As the topic is highly political, we expected this to be difficult. And that expectation became truth. Students working on Shapiro had to reach out to over 100 people to receive only four people/accounts willing to answer questions. The lack of intimate contact and the lack of a relationship of trust with the participants clearly had a substantial impact on our success rate. That being said, from our methodological perspective this was still potentially interesting data. Important here was to again understand those answers not as ‘knowledge’ (now we know how it works!), but as data. Meaning: we analyze them while being fully aware that respondents can lie, be mistaken, or be politically motivated in their responses. I will illustrate the difference between an empirical-interpretive methodological fueled analysis, and a more quantitative analysis below, in the Analysis section. For now, it is important to realize that the questions posed and the conditions in which we interview our subjects all influence the answer we get (or don’t get), and should thus be taken on board in our analysis. 

Big data, interviews, and observations can all be useful methods in empirical-interpretative research, but only if we bring them in line with our methodology and our research questions, and if we fully understand the limits of the data. It is our methodology that shapes which methods we (can) use and how we understand the data they yield. What is thus different in an empirical-interpretative methodology, is that having a method in itself is not enough: the method should fit the research question, and most importantly: the data. In this type of inductive research, it is the data that is in itself guiding the research. Research question and methods can be reexamined if the data force us to do this. 

Not all data is equal

Once the data is collected, it is necessary to determine the relationship between the different data. Big data, observations, and interviews are not only collected using different methods, but they also have different statuses. Depending on one’s methodology, certain data can be worth more than others. All data are thus not equal. Interestingly, seen from the perspective of now dominant methodologies, fieldwork or observation is increasingly seen as less worthy than standardized interviews that are converted into stats or big data research. Just like big data research, statistics still connote science in many eyes of the broad public. Big data, just like standardized interviews are producing data that can be inserted into a computer that spits out ‘knowledge’. Interviews and big data are in this perspective synonyms for knowledge. They are seen as producing hard facts and thus the ideal stuff to produce knowledge. Seen from an empirical-interpretative perspective in the tradition of Blumer, Becker, or in a more ethnographic tradition, the opposite is true. Observation is foundational. Without observation it is hard, if not impossible, to overcome the outsider position and fully grasp what people are doing (people lie, people are sometimes wrong, they give answers that they think are desired, …). Big data, from this tradition, can be helpful in understanding the scale of things, gaining insight into the role of bots, mapping networks, or observing relevant patterns, but it can never be a replacement for an ethnographic exploration of social worlds. Big data are thus only useful in empirical research when they are analyzed in relation to the methods used to collect them, the limits of the database and the affordances of that method, and of course of our overall methodology, and thus our understanding of social life (and for instance the importance of context in understanding interaction). Interviews, within such an empirical tradition, are not ‘knowledge’ producing devices, but interactions between researchers and interviewees, or rather participants. They are discourse that needs to be analyzed in context. Whatever the used method, what comes out is data. Data is always a subjective representation of things that happened in the empirical world. And those data are the rough stuff that needs careful analysis before it can produce knowledge. To wrap this section up: if you choose an empirical-interpretative methodology, not all data, and not all methods are useful: special care is needed to match them up. 

Analysis

The analysis of the data, just like the data themselves, the used methods, and the research question, is influenced by our methodology. Our methodology determines how we understand our data, and how we can use our methods in relation to our research question. For outsiders of the empirical-interpretative endeavor (regardless of whether we are employing critical discourse analysis, pragmatics, conversation analysis, ….), this is the phase that is the hardest to grasp. More than having a specific list of practices, or tricks, analysis is a far more complex and less easy-to-summarize affair. Analysis is a theoretically inspired interpretation. Interpretation is also not limited to the analytical phase, it is present throughout the whole process of research. Interpretative analysis is processual and starts with our perspective on social life. It is thus also present when we ‘go into the field’ and collect our data. Whenever we gather data, we are also trying to make sense of what happens. That underlines, once more, the importance of a solid methodology and why there is no replacement for doing fieldwork and exploring the world. Interpretation is thus always there, and your methodology will guide you along the way. 

Methods and the production of knowledge

When we actually start to analyze our data, we can use different methods to make sense of our data. From statistics and big data approaches to ethnopoetics or multimodal discourse analysis, the methods limit what we can do with our data, and as already stressed our methodology limits which methods are considered useful. The actual analysis goes beyond the more descriptive phase of an exploratory set-up. Important to underline once more: data and description are related to the theory, to the professional vision. Taking this step from ‘description’ to ‘analysis’ is maybe one of the most difficult steps newbies in discourse analysis have to take. This has several reasons: (1) unfamiliarity with the methodology and method(s), (2) lack of time to become deeply integrated into a field, (3) lack of knowledge (entering the field with an open mind is good, but entering the field with an empty mind will not be helpful to create an in-depth analysis), and (4) what I would label as the fast and efficient syndrome. Researchers in general are under pressure to produce results fast. That is not that different for students. Student life is not solely life to study (and party) anymore, many combine work and study time. The stakes are high and a master year is short (too short for real empirical research). In such a context, methods that result in a quick outcome are tempting. Even going over the same data more than once, is increasingly understood as a ‘waste of time’. In such a context, ethnographic digital discourse analysis becomes counter-intuitive: it takes time, and revisiting the data several times over is the rule rather than the exception. Discourse-centered empirical-interpretative research that takes itself seriously, always starts by getting to know your data well. In the case of spoken discourse (think of Podcasts or YouTube videos), we first have to transcribe: ‘Without a transcript – a written/graphic representation – talk is impossible to analyze systematically’ (Cameron, 2001). Analyzing a YouTube clip thus starts with the question of how to ‘represent the talk’ on paper so that it does justice to what happened in the actual conversation. How do you represent dialects, register, tone of voice, and idiosyncratic discourse habits? Transcribing forces the researcher to think not just about the words uttered, but also about multimodal meaning-making in general. It forces us to interpret language as a process of actual meaning-making in which so-called details, can be very important in the overall meaning that is constructed. 

The next step is to start focusing on the actual message that is communicated in interaction, and across different interactions (see Wortham & Reyes, 2021 for an excellent introduction). What is exactly narrated? If we now work with spoken or written digital discourse (or a mixture of both), we will – if we see discourse as language-in-action – look at any interaction as an event of speaking. Digital discourse is always produced by someone addressing an audience. The poster is not posting in general to an unspecified world but always has a specific audience in mind (his friends or followers on his social media). Who one is addressing will have a fundamental effect on the message. To analyze talk, it therefore is important, not only to look at the text but also at the participants in the interaction (including intended audiences) and the general context. To make this concrete again, in our research on right-wing influencers, we encountered multimodal discourse that was produced in a very specific, layered, and polycentric context (Blommaert, 2010): the Facebook page of these influencers ànd the places where that discourse was made and taken up. These contexts are part of our analysis: it is thus important to understand the world those influencers created on different media and how this message was taken up by people all over the world. In the first instance, that means that we need a detailed analysis of the posts they make from an input perspective. Those posts are inherently multimodal, in many cases also hyperlinked and they are part of a regular stream of posts. It of course also means that we need to focus on the uptake: on how people interact with those posts. 

Analysing a snippet of data

Let’s illustrate this with a concrete example. Students researching Ben Shapiro, as already mentioned, asked several people who interacted with Ben Shapiro’s posts on his social media page some standardized questions using the Messenger app. Those questions were of course not randomly chosen but emerged out of the explorative phase of our research project. One of the persons who responded was contacted because he commented on one of Shapiro’s posts on Facebook. In that post, Shapiro shared an article from the Daily Wire on the Rainbow flag. In his comment that accompanied that link, Shapiro was targeting the ‘dogmatic obeisance’ of the rainbow flag ‘that is forced through our throats’, and the ‘filified American Flag’. The respondent is a so-called ‘Top Fan’. This label is part of the ‘community-building’-affordances of Facebook pages. In a workshop session the students showed the following answer to one of their questions on screen: 

‘The ‘Top Fans’ thing I think is not so much of a big deal. Anyone who posts or likes every now and then is offered this Top Fan badge sooner or later. But I accepted it because I like Ben Shapiro. I think he represents a sane kind of conservativism we could also need here in Europe. As a Noahide I also appreciate that the kind of thought he stands for is based on Torah values. Therefore his conservatism is not contaminated with ethnic nationalism, racism or Islamophobia, like you find it on the Right in Europe or the Alt-Right in the U.S.’

Depending on one’s methodology and the used method to analyze this answer, we could go in different directions. I can imagine that social science students who work with statistics would see this answer as insignificant as only four people responded. But if they had to work with it, they could depending on their research question, categorize this answer in many different categories: ‘think top fans is not important’, or see it as an illustration of the poster's anti-racist commitments, or as somebody who is not far-right, not alt-right and so on. All those categorizations take the information provided by the respondent at face value. In such categorizations, the poster is assumed to be sincere about his convictions, and thus actively taking a stance against the radical and extreme right by following and interacting with the post on Shapiro’s page. As a consequence of such categorizations, Shapiro is indirectly also seen and framed as a ‘non-radical right’.

More discourse-inspired researchers would argue that his answer is far more complex, that there is or at least can be a fundamental difference between what people say, and what people do. But also between what people think and what is actually the case. In a more ethnographic-inspired discourse analysis, we should look at the context to make sense of this answer. That context is transnational, layered, and complex.

First of all, note that our respondent is answering and visiting that profile explicitly from a European perspective. Secondly, the fact that this answer was posted in Facebook Messenger after students contacted him because he was active on Ben Shapiro’s Facebook page, is relevant to make sense of the answer. Facebook in general, and that specific page in particular, is thus part of the context, just as the conversation in Messenger is part of the context. Let’s unpack this and start with the page of Shapiro. If we understand and see Ben Shapiro not just as a podcaster and influencer in it solely for the money, but as a (meta)political influencer, then we understand Shapiro’s posts in the context of a larger culture war to normalize an ideology (Maly, 2022b). His communication is then fueled by a (meta-)political purpose. In other words, his communication is then seen as trying to normalize an ideology by presenting his discourse as ‘normal’, ‘rational’, and ‘factual’. And that is exactly what students find if they look a Shapiro’s communication. Shapiro is clearly engaged in topics that are typical for the contemporary culture wars: abortion, LGBTQ+, culture Marxism, and of course a consistent negative framing of Biden and the ‘radical’ ‘left’ liberals. Interestingly, these topics are also consistently framed in an ‘intellectual discourse’ indexed by all the ‘difficult’ words he uses. Shapiro is presenting himself as an intellectual, and his discourse as intellectual common sense. On his Twitter account, this position is indexed by his pinned tweet that states ‘Facts don’t care about your feelings’. This pinned tweet frames his message in general, and thus all his posts on that medium as factual. Even more, it also frames the ones who are shocked by his posts as irrational, being completely absorbed by their ‘feelings’.

Looking from this piece of content to the answer of our respondent again, it becomes clear that the fact that our respondent frames Shapiro’s brand of conservativism as ‘sane’ not only refers to this particular brand of conservatism as not extreme right, 'not contaminated with ethnic nationalism, racism or Islamphobia’ but also as ‘sane’ in contrast with ‘LGBTQ activists’. A topic that is not only quite present on Shapiro’s wall but was also commented on by our respondent. Remember, it was this comment, and his Top Fan badge that was one of the reasons the students contacted him for an interview: it is thus part of the context of this answer. And the fact that his comment frames all of Shapiro‘s social media activity as not radical, not racist, and not extreme but as sane, now becomes a performative deed.

Thirdly, the answer of this respondent doesn’t come out of thin air. His answer is data because students contacted him in the context of a research project on Shapiro. From an empirical-interpretative tradition, this context needs to be taken into account when we analyze the comment. Remember, out of the blue, students contact him and asked him about his affiliation with Shapiro. This 'interaction order' as Goffman would say, clearly frames this interaction between the respondent and the students and gives it meaning. It is thus key to take the question itself into consideration. The opening question for the student was: 

"Hello (name respondent) I'm doing a research on Ben Shapiro and his media interactions. Statistics show that you are a "Top Fan" Of Shapiro, which means that you engage a lot with his posts and his media presence. How do you relate to his ideas and messages? What's your reasoning behind interacting with his posts? Can you please explain a little bit?"

We now understand why the respondent starts his answer with his opinion about the ‘Top Fan’ badge. What also becomes clear is that the question clearly presupposes that he has been surveilled by the students and that he is questioned about his behavior on Shapiro’s page. Even though students frame it as research on Shapiro, it is clear that he himself and his behavior is also part of that research. As a result of this question, he is not just one of the many posters on Shapiro’s page but is now turned into a subject in research that has an impact on the outcome of that research. All of a sudden he also does not communicate on the page of Shapiro anymore, but with ‘undefined researchers’. Note also that the students do not present themselves as students of a university, but as an ‘individual researcher’ using a more informal chat style communication. All these elements potentially influence communication and need to be taken into account. 

If we now look at his answer from within this context, we understand that he ‘normalizes’ or even ‘banalizes’ his behavior on the page. First, he claims that ‘the 'Top Fans’ thing’ is not a big deal. Ergo, it is not something that makes him an activist or a loyal fan: he is just a normal man that has accepted this banal label because he thinks Shapiro is a sane voice. This is nothing special, it is something that everyone gets. Of course, the reality is a tad more complex. According to Facebook, ‘You can become eligible to turn on a top fan badge by being one of the most active people on a Facebook Page or profile, which can include watching the Page's videos, liking or reacting to its content, and commenting on or sharing its posts.’  Ergo, not all people who sometimes like, comment, or share content get that label, only ‘the most active people on the page’ can acquire it. His behavior on the page is thus exceptional, but he sees or at least presents it as non-exceptional, normal even. In other words, he actively frames his behavior to the researcher as not exceptional. We can now interpret this in many ways. On the one hand, we can see it as part of his algorithmic imagination (Bucher, 2017): he doesn’t consider himself as someone who interacts a lot, so it is not abnormal to get it. It says something about how he sees the functioning of algorithms in awarding the ‘Top Fan’ badge: could be that he indeed doesn’t like and comment much, but that he views the videos a lot and is not aware that this is also one of the criteria to be awarded that badge. 

There is of course another interpretation possible. Seen from the side of the influencer, the ‘Top Fan’ badge is important. The Top Fan badge gamifies the interactions on the page. This gamification generates uptake for the influencer and thus contributes to the spread of his content to a broader audience (Maly, 2020b). From an influencer’s, but also from a metapolitical perspective, the acceptance of a Top Fan badge is important. Top Fans will not only interact more, but the Facebook algorithms will also push the content of that page in their newsfeed. Interestingly, that is also how our respondent experiences the Facebook algorithms. When asked if he is subscribed to the Daily Wire (the online medium of Ben Shapiro), our respondent answers that he doesn’t have a subscription but that ‘obviously the facebook (sic) algorithm cares enough to regularly show me the latest posts of Ben Shapiro so I don’t miss anything important’. 

Emblematic cases or the complexity of data

In the analysis of our respondent's short answer, we see a complex socio-technical assemblage that contributes to the metapolitics of Shapiro’s page.

1. First, we see how our respondent normalizes Ben Shapiro’s discourse. This allows us to understand his answers as metapolitical as well: he is actively framing Shapiro’s intervention as ‘non-radical’, sane, and not to be compared with the racist extreme right. His answer is thus performative, it provides researchers with a lens to understand Shapiro’s ideological position. Something that, if we would take those answers at face value, would contribute to a scientific legitimation of Shapiro’s discourse as ‘sane’, ‘non-radical’, and not extreme.
2. Secondly, we see how the banalization of the Top Fan badge contributes to the metapolitical effect: we shouldn’t see him as political or special.
3. Thirdly, we see how his discourse is a socio-technical assemblage constructed in relation to Facebook’s algorithms that now serve him all the articles Shapiro is posting. His interaction with the page thus creates an echo chamber in which the interaction should be understood: it clearly normalizes Shapiro’s points of view further and what he says to our interviewer, is clearly an effect of socialization within the bubble.
4. Concretely, we have to realize that our data are discursive data produced in very specific contexts and thus need to be analyzed as such. This becomes even clearer if we look at the comment our respondent left underneath the Daily Wired article Shapiro shared. Our respondent labeled the rainbow flag the ‘Hammer and Sickle of Cultural Marxism’. This type of discourse shows a deep integration into the radical right: Cultural Marxism has for many years now become the go-to explanation of the ‘radical left in power’. This post, together with his Top Fan badge, and his normalization of Shapiro’s discourse positions him right in the echo chamber. And in that echo chamber, he is not just a follower, but he performs audience labor that reinforces the metapolitical ambitions of Shapiro. 

This is of course only one piece of data that is analyzed, a piece of data that in any other research paradigm wouldn’t even be considered data. In our methodology, that snippet of data, that is only partially analyzed here, and only from one specific perspective and interest, becomes quite interesting. In an ethnographic approach, such a snippet can be relevant, especially when it is analyzed in relation to the collection and analysis of all the other data we collected. When we do that, we see that the discourse of our respondent is not exceptional at all, but complements the observations that were made. Whenever Shapiro posts we can watch different ‘Top Fans’ posting very similar messages. For instance, when Shapiro posts a link to a Daily Wire article ‘New Disney CEO Says Company Will ‘Quiet Things Down,’ Regrets Predecessor Battled DeSantis’ and comments ‘Not sure I’m buying the sudden change of heart’, we can witness several Top Fan comments highlighted by Facebook as ‘a most relevant comment' articulating very similar discourses. There seems to be a consensus that Disney is pushing woke politics that ‘shove their ideology down America's throat regardless.’ Ergo, the type of socio-technical interaction within the context of Shapiro’s page as displayed by our respondent is clearly not unique, on the contrary. The intertextual connection between Shapiro’s discourse and his ‘Top Fans’ is unmistakable. In this sense, we can see it as emblematic of a type of metapolitical interaction that reinforces the reach ànd depth of Shapiro’s discourse. The depth here refers to the fact that such comments radicalize and legitimize Shapiro’s discourse. 

Such comments are emblematic because they summarize a common practice, a very specific type of metapolitical interaction. As such they appear to us as key incidents from which much more can be inferred. Key incidents have a long history within ethnographic research (see for instance Kroon & Sturm, 2007 for an overview). Green and Bloome (1997) describe how ethnographers use ‘theoretical frameworks to guide analysis of the patterns and practices of members and how these bits of everyday life frame a particular view of social organization.’ This theoretical lens then allows the ethnographer to identify so-called key events. A key event says Erickson (1985, quoted in Kroon & Sturm, 2007:103) is ‘key in the sense that the researcher assumes intuitively that the event chosen has the potential to make explicit a theoretical ‘loading’. A key event is key in that it brings about to awareness latent, intuitive, judgements the analyst has already made about salient patterns in the data. Once brought to awareness these judgements can be reflected upon critically.’ A key event is in that respect similar to what Blommaert and Dong see as emblematic cases. Cases in their ethnographic perspective are not just descriptions of events or incidents but are cases of something. Something becomes a case when we apply our theoretical models to our case. The ethnographer describes the case ‘in functional and relational terms; explores links to other incidents, events, phenomena, or theoretical constructs; places the events in relation to other events or to wider social contexts; and then constructs a description so that others may see what members of a social group, need to know, produce, understand, interpret and produce to participate in appropriate ways.’ (Green & Bloome, 1997:186 ). To call something a case, or a key incident thus equals making a theoretical claim.

Methodology, digital discourse analysis, and methods

And thus we end where we started: our methodology allows us to analyze our data in very specific ways. This in turn allows us to go beyond mere description (summary) of the discourse. Theory enables us to see more in the data. In class, I usually state that concepts are our tools. What a hammer is for a carpenter, can be the notion of algorithmic agency, or entextualization for us. This metaphor immediately triggers the idea that if there is only a hammer in your toolkit, all problems look like nails. The more concepts we have, the more we will recognize, and the more detailed and refined our concepts are, the more precise our analysis can be. Interpretation is thus always subjective in the sense that it is not only about what is in the data, but also what a researcher can recognize and what not. Claiming that concepts are our tools should of course not be read as the slavishly use of accepted concepts. Doing empirical research is always in the last instance checked with reality. Concepts should follow reality, instead of reality being molded to fit the concepts. That means that we at least, in theory, will have a need for new concepts to make sense of a fast-changing world.

If you want more details about such discourse analytical methods, there are many excellent guides out there (see for instance the works of Cameron, 2001; Cameron & Panovic, 2014, Verschueren, 2012, Machin & Myr, 2012, Jewitt, Bezemer & O’Halloran, 2016). Analyzing those posts using a brand of discourse analysis is quite a lot of work, and even though it allows us to go in-depth and take context seriously, it also presents a limitation to what we can do within our methodology, especially in the context of digital discourse analysis. Digital contexts force us to adapt our analytical tools (Jones, Chik, Hafner, 2015), but we should do this from a well-developed methodological point of view. Analyzing digital discourse from an empirical-interpretative methodology, and thus observing the influencers and understanding their social world is quite different than scraping the posts of the last week. Scraping always alters the data, one loses context or one or more modes of communication. Seen from an empirical-interpretative methodology, ‘big data’ are quite different from ethnographic data. Scraping, if we like it or not, decontextualizes data. At the same time, adopting a mixed method approach that combines discourse analysis and quantitative methods can create knowledge on ‘both micro and macro levels by mixing in-depth explorations of textual examples with the analysis of larger discursive patterns’ (Georgakopoulou, Iversen & Stage, 2020: 20). The quantitative data can show us patterns that are invisible on the detailed level. Network analysis and other computer-assisted scrapings can help us generalize or find and place them in their networked context. What is key though, is that these methods can’t replace empirical-interpretative research, and even more, methods cannot replace a methodology. 

 

I thank all 2022-2023 students from my 'Discourse and digital media' class, and my class on 'Interaction in the hybrid media system' for their work, and to let me use their data as examples here. Sjaak Kroon has delivered substantial feedback on the first version of this working paper, the text would have looked quite different without him. 

References

Becker, H. (1998). Tricks of the trade. How to think about your research while you’re doing it. Chicago. 

Bezemer, J. Murtagh, G., Cope, A., Kress, G.  & Kneebone, R.  (2011). “Scissors, please’: The practical accomplishment of surgical work in the operating theater. Symbolic interaction Vol. 34., pg 398-414. 

Blumer, H. (1969). Symbolic interaction. California: University Of California Press

boyd, d. & Crawford, K. (2012) Critical questions for big data. Information, Communication & Society, 15:5, 662-679, DOI: 10.1080/1369118X.2012.678878

Bucher, T. (2017). The algorithmic imaginary: exploring the ordinary affects of Facebook algorithms. Information, Communication & Society, 20:1, 30-44, DOI: 10.1080/1369118X.2016.1154086

Bucher, T. (2018). If…Then. Algorithmic power and politics. New York: Oxford University Press. 

Blommaert, J. (2004). Workshopping. Professional vision, practices and critique in discourse analysis. Ghent: Academia Press. 

Blommaert, J. (2010). A sociolinguistics of globalization. Cambridge: Cambridge University Press. 

Blommaert, J. & Dong, J. (2010). Ethnographic fieldwork: a beginner’s guide.  Bristol, Buffalo, Toronto: Multilingual matters. 

Blommaert, J. & Dong, J. (2019). When your field goes online. Tilburg Papers in Culture Studies, paper 227. 

Burgress, J., Albury, K. McKCosker & Wilken R. (2022) Everyday Data Cultures. Polity.

Cameron, D. (2001). Working with spoken discourse. Los Angeles, London, New Dehli, Singapore, Washington DC: Sage.  

Cameron, D. & Panovic, I. (2014). Working with written discourse. Los Angeles, London, New Dehli, Singapore, Washington DC: Sage.  

Chadwick, (2017). The hybrid media system. Politics and power. Oxford: Oxford University Press. 

Georgakopoulou, A. &  Spilioti, T. (2020). The Routledge Handbook of Language and Digital Communication. London & New York: Routledge. 

Georgakopoulou, A., Iversen, S. & Stage, C. (2020). Quantified Storytelling. A narrative analysis of metrics on social media. Cham: Palgrave Macmillan. 

Jenkins, H., Ford, S., & Green, J. (2013). Spreadable media. Creating value and meaning in a networked culture. New York: New York University Press.

Jewitt, C., Bezemer, J. & O’Halloran, K. (2016). Introducing multimodality. London & New York: Routledge. 

Jones, R., Chik, A. & Hafner, C. (2015). Discourse analysis and digital practices. Doing discourse analysis in the digital age. London & New York: Routledge. 

Kroon, S., & Sturm, J. F. (2007). Key Incident Analysis and International Triangulation. In W. Herrlitz, S. Ongstad & P.-H. van de Ven (Eds.), Research on mother tongue education in a comparative international perspective. Theoretical and methodological issues. (pp. 99-118). Amsterdam, New York: Rodopi.

Leidig, E. & Bayarri, G. (2022). Not Your Grandma’s Fascism: Fame, Femininity, and Race in Far-Right Postcolonial India and Brazil. Social Politics: International Studies in Gender, State & Society, DOI: 10.1093/sp/jxac013

Machin, D. & Myr, A. (2012). How to do critical discourse analysis. Los Angeles, London, New Dehli, Singapore, Washington DC: Sage.  

Maly, I. (2019). New Right Metapolitics and the Algorithmic Activism of Schild & Vrienden. Social Media + Society, 5(2). DOI: 10.1177/2056305119856700 

Maly, I. (2020a). Metapolitical New Right Influencers: The Case of Brittany Pettibone. Social Sciences, 9(7), 113. DOI: 10.3390/socsci9070113 

Maly, I. (2020b). Algorithmic populism and the datafication and gamification of the people by Flemish Interest in Belgium. Dossiê • Trab. linguist. apl. 59 (1) • Jan-Apr 2020 • DOI: 10.1590/01031813685881620200409   

Maly, I. (2022). Algorithms, interaction and power.  A research agenda for digital discourse analysis. Working Papers in Urban Language & Literacies nr 205. 

Maly, I. (2020). Review essay: The age of surveillance capitalism ( Shoshana Zuboff). Diggit Magazine. 

Pasquale, F. (2015). The black box society. The secret algorithms that control money and information. Cambridge: Harvard University Press. 

Shahin, S. (2016). When Scale Meets Depth: Integrating Natural Language Processing and Textual Analysis for Studying Digital Corpora. Communication Methods and Measures, 10:1, 28-50, DOI: 10.1080/19312458.2015.1118447

Silverstein, M. & Urban, G. (1996). Natural histories of discourse. Chicago: Chicago University Press. 

Thompson, J.P. (1990). Ideology and modern culture. Cambridge: Polity. 

Van den Hoonaard, W.C. (1996). Working with Sensitizing Concepts. Analytical Field Research. Sage. 

Van der Aa, J. & Blommaert, J. (2016). Ethnographic monitoring and the study of complexity. In Martin-Jones, M. & Martin, D. (Eds.), Researching Multilingualism. Critical and ethnographic perspectives. London: Routledge.

Varis, P. (2016). Digital ethnography. In Georgakopoulou, A. & Spilioti, T. (Eds.), The Routledge Handbook of Language and Digital Communication (pp. 55-68). Routledge.

Venturini, T. (2020). From fake news to junk news: the data politics of online virality. In Bgo, D., Isin, E. & Ruppert, E. (Eds.), Data politics. Worlds, subjects, rights. London & New York: Routledge.

Verschueren, J. (2010). Ideology in language use. Pragmatic guidelines for empirical research. Cambridge University Press.

Wortham, S. & Reyes, A. (2021). Discourse Analysis Beyond the Speech Event. New York: Routledge.

Zuboff, S. (2019). The age of surveillance capitalism. The fight for a human future at the new frontier of power. Profile Books