Climate Imaginaries and the Linguistic Construction of Identities on Social Media

Abstract

This article takes a linguistic approach to the construction of group identities in relation to climate imaginaries, which condition what climate futures are considered possible. It employs social identity theory along with methods from Sociolinguistics and Discourse Linguistics to analyse patterns of positioning and stance-taking in social media discourse (Twitter/X, Instagram). By identifying typical multimodal practices at a micro level, as well as introducing five functional meta categories (›awareness, advice, and agitation‹, ›ritualised appreciation‹, ›solidarity and allyship, ›voice and advocacy‹, and ›community demarcation‹), the research investigates how digital discourse reflects and shapes group identities related to climate change topics. It thereby offers insights into the role of language in co-producing climate imaginaries in digitally shared environments.

Zusammenfassung

Dieser Artikel untersucht mit sozio- und diskurslinguistischen Methoden die Konstruktion von Gruppenidentitäten hinsichtlich klimabezogener Zukunftsvorstellungen (Climate Imaginaries). Letztere bestimmen, welche klimabezogenen Zukünfte für möglich gehalten werden. Auf der Basis eines digitalen Diskursausschnitts (Twitter/X, Instagram) werden kommunikative Praktiken der Positionierung und des Stance-takings analysiert sowie typische multimodale Muster auf der Mikroebene herausgearbeitet. Übergeordnet werden fünf funktionale Metakategorien identifiziert (›Agitation‹, ›rituelles Gedenken‹, ›Allyship‹, ›Fürsprechen‹ und ›Gruppenabgrenzung‹), deren Erörterung nachzeichnet, wie digitale Diskurse Gruppenidentitäten im Zusammenhang mit Klimawandelthemen reflektieren sowie formen und wiederum von diesen geformt werden – und wie damit klimabezogene Zukunftsvorstellungen ko-produziert werden.

1 Social Media Discourse and Climate Futures

Applying Taylor’s (2004, p. 24) idea of social imaginaries, that are requisite for the collective practices making up our social life, Hulme (2022, p. 230) writes that »climate imaginaries can be understood as collectively shared sets of beliefs, narratives, technologies, discourses, and practices that condition what climate futures are thought as possible, likely or (un-)desirable« (italics original). He thus adds a futuring dimension to the purview of imaginaries, and follows Taylor’s emphasis on the social by suggesting that climate imaginaries »give meaning to events, […] shape the practices, lived experiences, and identities of social groups« (Hulme 2022, p. 230). While climate imaginaries comprise social action on various levels, the analysis of public discourse – especially social media communication – addresses some of their key elements (Pearce et al. 2019, p. 9).

The mentioned categories of ›practices‹, ›experiences‹ and ›identities‹ can, from a linguistic perspective, be traced in discourse as linguistic and multimodal practices, narratives of experience, and markers of (group) identities. Initially shaped by climate imaginaries, associated linguistic patterns can then be attributed as points of access to existing and (in social media environments) participatorily produced climate imaginaries. This coheres with approaches in Socio‑, Discourse- and Cultural Linguistics: firstly, that linguistic and communicative practices shape culture, norms, and imaginaries and, secondly, that cultural aspects manifest themselves simultaneously in linguistic and communicative practices (Bubenhofer et al. 2022, p. 6).

Participation in the rendering of climate imaginaries can occur in various shapes and forms, and can be equally abstract or concrete. To give a tangible example, research in other disciplines, such as Economics, Politics, or Organisation Studies, provides suggestions of consumer-related climate imaginaries on a high level of abstraction: for instance, Levy/Spicer (2013, p. 663) suggest key climate imaginaries such as a ›technological solution‹, ›fossil fuel regime‹, ›sustainable lifestyle‹, or an ›inevitable climate apocalypse‹, which are connected to respective world views. They identify concrete actions or ideas in the realm of these climate imaginaries as »framings of an imaginary« (Levy/Spicer 2013, p. 666, our italics). However, those specific climate-related ideas could also simply be named ›imaginaries‹ – if approached on a different scale that recognises in how far these actions or ideas are themselves embedded in specific beliefs, narratives, technologies, discourses, and practices. It is important to retain the possibility of findings (imaginaries, values, elements of cultural climate models) on various abstraction levels and, furthermore, focus on language and communicative practices that provide contributions with a modelling character in any form – displays of imagination but also of foresight or agitation (Davidson/Kemp 2023). In fact, there are numerous linguistic categories that seem promising to serve as entry points for the latter, especially in digitally shared environments. Returning to the quote from Hulme (2022) and the connection of imaginaries and climate futures with a broad collection of linguistic and multimodal practices, narratives of experiences, and markers of group identities, we delineate our approach as follows: in this article, we focus our analysis on markers of group identities and associated communicative practices as a first step toward contributing to a linguistic approach to climate futures on social media.¹

Consequently, we use a social media dataset with posts from Twitter/X and Instagram concerning the topic ›climate change‹ spanning a limited time frame (Pearce et al. 2024, see chapter 3 for details on data collection and structure). For the analytical approach, we employ aspects of social identity theory suited for public and political discourse (a.o., Oakes 2002; Mead 2013 [1934]; Abels 2017) and apply approaches from Interactional Sociolinguistics, Conversation and Discourse Linguistics (see chapter 2) to analyse both frequent and qualitatively prominent linguistic patterns. Furthermore, we focus on the particularities of this digital discourse with regard to various linguistic practices of (self-)positioning, reader addressing, narrative stance-taking, evaluation and display of (group) norms. After looking at the data in more detail, we identified several typical micro patterns at the word, text-structural, and narrative level as well as five functional meta categories connected to central displays of group identities that, moreover, organise our findings on a macro level (see chapters 4, 4.1–4.5).

2 Media Identities – Group Identities

When it comes to identity in climate discourse, we are intuitively confronted with a number of seemingly pre-existing discourse roles that are equally intuitively associated with identities of some sort: politicians, activists, scientists, entrepreneurs, tech company representatives (e.g., Schwegler 2018, pp. 471–485).² The list can easily be expanded: the marginalised, the powerful, ›ordinary‹ people, tax payers, and so on. At least two considerations appear worthy of mention: Firstly, the clear demarcation of these groups is at best an artificial conceptual support, a rhetorical strategy, or even complete fiction at worst. Secondly, these groups vary greatly in size and level of abstraction. The latter is of course connected to the first point, originating in the fuzziness of group demarcation. The element of self-attribution is therefore relevant for the definition and also the (linguistic) analysis of any kind of group. This brings us to the question of ›identity‹, both on the individual as well as on the collective level, which we need to investigate before we can delve deeper into aspects and central patterns of their linguistic realisation (see chapter 2.2).

2.1 Identities in Theory and Identity Theories

Individual identity derives from questions of the form ›Who am I?‹, ›Why do I do things the way I do them?‹, or ›How do I want to be?‹. Perceptions of the self are, however, not stable over time. Oftentimes individuals re-interpret their past actions to reorganise their identity conception – sometimes even consciously (Abels 2017, p. 196). This personal and purposeful aspect of identity accounts for differences between individuals with nearly indistinguishable biographies. The Social Sciences, however, have mostly moved away from conceptions of what can be termed an asocial conception of individual identity isolated from social interactions (Rucht 2011, p. 26). Such theories cannot account for the original emergence of the individual’s identity (Mead 2013 [1934], p. 267).

Other approaches such as social systems theory and social network theory understand identity as something that emerges in social (inter-)action, meaning that the notion of group identity originating solely from the aggregation of individual identities is not feasible (Holzer/Fuhse 2010, p. 314). These assumptions can be traced back to Mead, who differentiated ›I‹, ›me‹ and ›self‹. The first is responsible for spontaneous, non-reflexive behavioural impulses while the second emerges from the imagined view of the ›self‹ by others, which relies on role assumption (Abels 2017, pp. 212–214; Mead 2013 [1934]). The ›self‹, then, is the reflexive entity mediating between past actions of the ›I‹ and affordances observed by the ›me‹ and can therefore be understood as individual identity (Abels 2017, p. 215). This means that even what can be assumed as a personal identity is effectively socially dependent. Ultimately, this also holds true for identity aspects such as ›class‹, ›gender‹, and ›nationality‹, which are sometimes claimed to be better understood in essentialist terms (McGlashan 2020, p. 313) – a view that is rejected here.

Relational Sociology understands identity as a reaction to the need for control in a social world that confronts the individual with social interactions that are to some extent veiled and dubious (White 2008, p. 1). Resulting personal identities seem non-contradictory from the outside, while in reality they internally negotiate a considerable amount of doubt and inconsistency. This makes individual identity a fundamentally dynamic phenomenon (White 2008, p. 1; Demuth/Watzlawik 2021, p. 630). Individuals are present in multiple social ties, networks, and registers, resulting in conflicting belongings and thus a less coherent self-identity than those individuals themselves may be aware of (Liepelt 2010, p. 36). Switching between these social settings is then crucial for unique identity formation. They need to decide which aspects of prevalent norms and practices in each setting they wish to adopt. Individuals are not a certain identity, they assume aspects of individuality according to affordances of perceivable group ideals (White 2008, p. 2). Furthermore, they display, (co-)construct, and negotiate identities (linguistically) in social interaction (Page 2012, p. 17). Accordingly, individual and collective identity are closely intertwined.

White (2008) postulates four layers of identity: The primary dimension are individual preferences (which are ultimately also learned in social experience). The secondary dimension can be understood as the cultural consensus of groups where members of the group are discernible both from the inside as well as from the outside. Such groups establish ideas of their self in the form of coherent self-perceptions. The tertiary layer consists of conflict and dissent that take the form of internal and external negotiation of divergence and adaption. Lastly, groups organise another layer of reflexion about the third layer, where conflicts are assessed and rationalised (White 2008, pp. 10–11). In the posts we analysed for this contribution, we primarily find examples of the second and third dimension of identity construction. While individual preferences might also be deciphered from social media posts, the actors we look at generally speak as members of certain groups, which will become clear later. The analysis of the fourth layer would require a detailed inquiry into the comment sections and group discussions of the posts examined, which is considered only marginally in this article. In our conclusion (chapter 5), we will come back to these four dimensions of identity and provide first ideas about how our findings give clues to approach this fourth layer.

›Group identities‹ are thus sets of norms, rules, and conventions of a number of people who interact with one another for a certain time (Dolata/Schrape 2018, p. 22). Still, this should not be interpreted to mean that all members of a group constitute a homogenous body of like-minded individuals. »Identities are, thus, recognised as multilayered, dynamic, and often ambiguous, as social actors can project a number of different, sometimes conflicting positions and can have their positions interpreted by others differently than they had intended« (Giaxoglou/Georgakopoulou 2021, p. 245). Group identities, however, do have a feedback effect on group members’ choices, preferences, and actions (Dolata/Schrape 2018, p. 24).

›Collective identities‹ rely heavily on constant demarcation and distinction of other identities (Mützel/Fuhse 2010, p. 14) to form an understanding of the in-group (Kleinke/Bös 2018, p. 156). Mere allocation of individuals to certain groups is not enough, though, as social meaning of and for those distinctions has to be formed and provided in collective action (Oakes 2002, p. 816). This can be viewed as a co-constructive endeavour mediated by self-display as well as reactions of the other to that self-display, as well as external perception of third parties, i.e. the out-group (Tagg 2015, p. 146).

It appears that when we speak of ›identity‹ we might in fact be dealing with two separate terms. One denotation defines it as being what makes a person unique, signifying that person’s utmost own characteristics, ideals, values, and personality. The other understands ›identity‹ as a certain set of shared characteristics, ideals, values, and beliefs that a person can identify with. This is then understood as the identity of a group. Phrases like ›My identity as an activist‹ or ›My identity as a football fan‹ are no rare utterances in contemporary discourse. We could describe this reading of the term as identity as a resource, because individuals can choose from a variety of different group identities to make up their own personal identity. Individual identity is thus always ›patchwork identity‹ (Keupp 2008 [1999]). Foucault (2021 [2005], p. 244) highlights this paradoxical nature of identity, emphasising its simultaneous deeply personal and profoundly communal aspects – a relationship marked by the tension between the pursuit of individuality and the longing to transcend all barriers that divide humans from each other.

2.2 Linguistic Identity Analysis

While the co-constructional nature of identities and the shaping character of the above-mentioned imaginaries are fairly undisputable, we now want to emphasise the essential feature and the role of language therein. Specifically, we focus on the concrete displays of identities on social media, particularly the constituting communicative practices, which span all linguistic dimensions. The associated multimodal and linguistic patterns are continually evolving, and appear in multifaceted ways. Accordingly, Mazur points out that »identity is something that is done, not something that simply is. Identity is a verb, not a noun« (2021, p. 451). Likewise, the linguistic category of identity is well known in both Sociolinguistics and Conversational Analysis as a »process rather than fixed aspects of cognitive structure or personality« (Oakes 2002, p. 817, our italics). An adequate and brief overview is provided by Hinnenkamp (2018, p. 149–163). As linguists, we approach identities in practice (Steen 2015, p. 53); we focus our analysis on how actors are ›doing identity‹ or how they are ›generating identities through communication‹ – this applies to the individual, the group, and the community perspective. This is very much in line with the theoretical approaches presented above that also understand identity as a fluid, dynamic result of continuous co-construction, even when individual identity is concerned.

Understanding identity as a process implies feedback effects of group identities on the group members’ choices, preferences, and actions as well as the complexity of the identity-related development of actors in discourse.³ While an ethnographic tracing of identity development is an interesting endeavour, our aim is situated within the linguistic pragmatics of public online discourse (appropriate to our dataset): similar to Page (2012, p. 22), we are thus not so much interested in tracing and interpreting the different facets of a single group’s or a specific set of participants’ identities constructed and displayed in online discourse. Rather, we seek to reveal linguistic factors, forms, and functions that contribute to group-related identity work on social media. We begin the analysis by systematically applying knowledge from previous studies on salient word forms (e.g., Müller 2009; Kleinke/Bös 2018; Torres Cajo 2022). Based on the structure of the dataset, this means employing both quantitative and qualitative steps of analysing, inductively extracting, and describing linguistic patterns (see chapter 3 to follow our corpus approach).

The central threads of our final identity analysis are inductively identified categories on a mid-level of abstraction, in form of the following contextual meta functions: 1) ›awareness, advice, and agitation‹, 2) ›ritualised appreciation‹, 3) ›solidarity and allyship‹, 4) ›voice and advocacy‹, and 5) ›community demarcation‹. On the margin, this overall functional perspective can support the attempt to explore the occasions of engagement in specific online discourses. In combination with the respectively arranged and exemplarily analysed posts, these functional meta-categories can additionally help to tangibly illustrate the online climate change discourse community. Therefore, we use these categories to structure the result section of this article, although functional aspects regularly overlap, occur jointly or simultaneously. We want to emphasise that these functional categories are never mutually exclusive. However, looking at linguistic markers of usefulness in various linguistic dimensions (see Hausendorf et al. 2017, pp. 241–244), we can sort posts according to their most prominent function, as primary function, while acknowledging other possible functions (Heinemann/Viehweger 1991).

In chapter 4, we consider each contextual meta function in turn by briefly introducing the respective applied framework for our analysis and by demonstrating the fine-grained identity analysis of prominent examples. For the latter, we draw on several linguistic patterns of identity construction which are well known and well described in the literature – both in the context of individual interactional identity analysis and reflections on collective identity in Text and Discourse Analysis. They are (self-)positioning (Bamberg/Georgakopoulou 2008; Deppermann 2013; Giaxoglou/Georgakopoulou 2021; Torres Cajo 2022; Spieß 2024), reader addressing (Hausendorf et al. 2017, p. 135), narrative stance-taking (Niemelä 2011; De Fina/Georgakopoulou 2020, pp. 105–106; Merten 2023), evaluation and display of (group) norms (e.g., Schwegler 2018; Dang-Anh 2023). In our data, these linguistic patterns of identity construction explicitly or implicitly serve as references to climate imaginaries.

Deppermann (2013, p. 63) argues that the analytical »positioning perspective seems to be well equipped to attend to the fine-grained work of invoking, ascribing and negotiating identities«. This also extends to online environments (Spieß 2024). The category of ›positioning‹ is a vital basis to the analysis of narrative identity (Lucius-Hoene/Deppermann 2004) and equally important to the analysis of interactional storytelling, supporting the sociolinguistic concept of ›small stories‹: Bamberg/Georgakopoulou (2008, p. 382) prominently demonstrate the analytical value of »how people use small stories in their interactive engagements to construct a sense of who they are« by focusing on the associated (self-)positioning. In social media contexts, similar patterns occur and, furthermore, storytelling is »often equated with the presentation of an ›authentic‹ self and experience« (Giaxoglou/Georgakopoulou 2021, p. 241, see chapter 4.1). Thus, story content – however small the story is – can be well analysed from a functional perspective to display the users’ engagement, which opens up new insights into identity construction (see Frick in this issue). As Giaxoglou/Georgakopoulou (2021, p. 243) sum up, small stories »are parts of larger interactive activities«; they can also be part of discursive activities in the sense of social media community building and, therefore, support group or community identity developments.

Moreover, joint storytelling – that is, posts from more than one account in shared authorship – is a relatively new platform feature (e.g., on Instagram) which adds an interesting twist to the question about the above-mentioned authenticity. The entanglement of identities is sometimes visible in the »plot« (Giaxoglou/Georgakopoulou 2021, p. 243)⁴ of co-authored posts. Here, an initial text-structural analysis can reveal shares of identities, sometimes in line with, sometimes more or less additional to the main story of the post (see chapter 4.1). Not only co-authors but also other entities can be linked to a post – either through actual links or textual and visual mentions. Looking at displays of those entities such as places, events, and interconnected actors, we can find out more about themes and the aboutness of the posts (Giaxoglou/Georgakopoulou 2021, p. 243). In a social media environment, this implies that we can learn about the function of a post or »emplotment« (White 2000 [1973]; Giaxoglou/Georgakopoulou 2021, p. 224)⁵ and thereby the imagining of climate futures within the discursive context of joint storytelling.

However, on a pragmatic micro level and concerning the above-mentioned in-group-oriented aspect (see chapter 2.1), identities are constructed to a large degree by indexicality in interaction; all linguistic dimensions – semantics, pragmatics, but also grammar and lexicon – can be utilised as indexical markers of group affiliation (Tagg 2015, p. 224). Concerning the out-group, demarcations of group identities relative to other group identities are vital. They can take the form of either self-description, where the ›other‹ is only implied, or of an ex negativo definition, where norms and conventions that are undesirable or to be rejected are explicated (Tagg 2015, p. 226 as well as chapter 4, in particular 4.5). Both forms are performative distinctions. Practices that display stance, direct or indirect commenting on groups, or addressing of others, namely the way ›how characters are positioned within a post‹, provide insights into the actors’ own positioning and identities as well as their understanding of themselves with regard to the discursive environment. On account of this multifaceted situation, Giaxoglou/Georgakopoulou (2021, p. 246, with reference to Bamberg 2004 and Bamberg/Georgakopoulou 2008) emphasise a theoretical and analytical distinction between three levels or dimensions of positioning, starting with (self-)positioning in interaction on level one and two. As Dang-Anh (2023, p. 16) points out for digital, political, and activist discourses, level three positioning is the relevant dimension for our purposes: here, the focus lies on a positioning act’s indication of »alignment or disalignment with related dominant discourses or master narratives about related social and moral identities« (Giaxoglou/Georgakopoulou 2021, p. 246). Thus, in terms of findings, we concentrate on group-related identity patterns that express aspects of social meaning and collective action (Oakes 2002, p. 816) in language. In order to display group affiliation and demarcation by means of concrete (self-)positioning, reader addressing, narrative stance-taking, evaluation and display of (group) norms – directly or ex negativo – the following linguistic saliencies appear prominently in our dataset:

• Firstly, at the word level: distinctive pronoun usage especially in certain collocations with specific verbs; explicit group and role designations (concrete, unspecific, by relationship or function); (metonymic) category designations; emotional adjectives and illustrations as emotional framing; direct prompts.

• Secondly, at a text-structural level: chronological timeline or milestone order for historical narratives; list-making⁶ to prioritise group-related facts, stances, and actions.

• Thirdly, at a narrative level: generalisations; calls for or displays of support; (implicit) arguments against popular or stereotypical discourse positions, e.g., in shared experiences.

All of these linguistic patterns and communicative practices were inductively extracted in the process of our analysis. In this article, they are repeatedly addressed in several exemplary contexts within the five functional meta categories (chapters 4.1–4.5). Before that, the following paragraphs (chapter 3) describe in detail the overall approach to the social media dataset.

3 Data Rationale and Approach

Our analysis draws on a subset of a social media dataset that was initially collected for a joint and interdisciplinary multimodal cross-platform analysis (Pearce et al. 2024). This dataset will be continuously updated for later research and is situated in a larger interdisciplinary project (see footnote 1).⁷ In order to build the dataset, posts concerning the topic ›climate change‹ were collected from TikTok, Instagram, and Twitter/X during the time of the COP28 in November and December 2023, the 28^th Conference of the Parties. Therefore, the event plays a central role in the dataset. In addition, ›netzero‹ and ›veganism‹ are relevant subtopics, since the initial search queries were adapted to these solution-oriented aspects as project-related interests in the range of climate change and climate futures. For this article, we only investigated the platforms Twitter/X and Instagram. When speaking of ›data‹ and ›dataset‹, from this point on we are referring to this subset of the larger data. We have split up this subset into an Instagram and a Twitter/X corpus to analyse them separately using corpus linguistic methods. We will refer to each ›corpus‹ when a finding is only applicable to one of the platforms, and use ›data‹, ›dataset‹ or ›corpora‹ when referring to both.

Even though our research contribution is methodologically focused on a detailed, exemplary close reading analysis of prominent social media posts, we nonetheless are confronted with a rather large amount of data to choose such posts from. We therefore approached the corpora from a quantitative perspective first (Bubenhofer 2015, 2013, p. 110) in order to detect linguistic data that is relevant to our interest in group and community identities. However, the aim was not to rely on statistically significant frequencies of certain words only. Rather, the quantitative pre-selection was guided by questions of how identity construction, climate imaginaries, and group demarcations within the discourse about climate change are linguistically characterised. As an initial approach, we therefore did not look up general frequencies of words or n‑grams, but decided to build corpus queries based on collocation expectancies in regard to said questions (see for an elaboration of this kind of approach Bubenhofer 2022, especially pp. 198–200). A number of such expectations could be discarded very quickly with this approach.

Collocations pointing toward positioning and self-identifications as part of a specific group such as ›as climate activists‹ or ›as a vegan‹, for instance described by Torres Cajo (2022, p. 71) or Merten (2023, p. 106), are practically non-existent in both corpora. Verbs and nouns co-occurring with the lemmas ›we‹ and ›our‹ respectively proved to yield more promising results. However, the detailed analysis of the significance of ›we‹ and the ›we‹-set in the respective context (Günthner 2021) required subsequent steps of close reading. The table in figure 1 shows the collocations with their absolute frequency.⁸ Collocations promising to be good indicators for posts relevant for our research question are highlighted in bold letters.

Fig. 1

Collocations with their absolute frequency

After this initial approach, we started looking into the now more manageable numbers of posts, firstly, in hope of encountering speech indicating identity construction. Our process is described below. Secondly, we included the images of the posts in order to identify further relevant patterns, especially in text-in-image combinations.

Continuing with the captions, grouping ›we want‹ and ›we need‹ together, and searching for subsequent collocations of the form to verb_fin, a noticeable collocation was ›we want to AMPLIFY‹, where ›amplify‹ could be realised by either of the following: ›amplify‹, ›unite‹, ›emphasise‹, ›spread‹, ›help‹, ›raise‹, and ›enhance‹. Even though these verbs can hardly all be seen as good synonyms, their realisation in the posts where they occur illustrate similar expressions of group desires and/or aims.⁹ Likewise, collocations of ›we are‹ and ›we must/should‹ appeared to be useful candidates for further investigation. The tables in figure 2 show which verbs proved to indicate posts relevant for our academic interest. It is easily recognisable that by this method we quickly reach a point where statistical relevance is not (and also should not be) the relevant factor for data selection (Bubenhofer 2013, p. 109). This is especially true when the methodology is ultimately committed to qualitative approaches. To make sense of the large corpora, however, it seemed and proved prudent to start the search for relevant data with an informed corpus inquiry based on both pre-assumptions and statistical salience.

Fig. 2

Tables of relevant collocations of ›we are‹ (left side), ›we must/should‹ (middle), and ›we want/need to‹ (right side) in Instagram corpus

Adding to these reflections, it is part of the nature of social media data and the ways in which we can gain access to it that a large portion of textual material is not actually found in the ›body‹ of the post. This is especially true for Instagram, where the caption of a post may even contain less text than the image(s) of the post themselves. But the same is true for tweets, as they too may contain images or screenshots of other tweets or even postings from other platforms that do not find their way into the text data in our corpora via text mining procedures. This fact renders a purely statistical approach to find relevant data all the more inadequate and shows the importance of analysing social media posts holistically as multimodal ensembles (Stöckl 2019) encompassing both textual and pictorial elements. In the next step, we derived categories from the posts (see method section in chapter 2.2) to describe various strategies of identity construction, which are presented in the following chapter.

4 Traces and Constructions of Identities in Five Functional Meta Categories

The analysis has not only identified the previously mentioned five contextual meta functions (›awareness, advice, and agitation‹, ›ritualised appreciation‹, ›solidarity and allyship‹, ›voice and advocacy‹, ›community demarcation‹, see chapters 4.1–4.5), identity construction, and saliences on a pragmatic micro level (for a summary see chapter 2.2), but rather provided further observations that categorise our dataset beyond that. These findings are probably typical for a social media discourse concerning ›climate change‹, although some of these aspects might be even generic to social media (or at least to certain platforms).

The initial observation concerns the occasions of the posts. We found two distinctive occasions for climate change-related posts that involve identity construction: Firstly, we discovered content moving »between online and offline« (Giaxoglou/Georgakopoulou 2021, p. 243) with an explicit link to an event. Such events can be public and international but locally bound, e.g., the COP28 conference, which took place in Dubai. However, they can also be lacking a concrete location and be global and predominantly digital instead, such as Earth Day or Indigenous People’s Day (see chapter 4.2). Those events are usually temporally fixed (Earth Day is always on April 22^nd; Indigenous People’s Day is on the first Monday in October in many countries). Rather individual events were also found as posting occasions in our data. This was, for example, the launch of a publication or a specific individual instance of activism. Secondly, we identified specific accounts showing a (serial) routine of social media climate activism, which is a typical emerging pattern (Pearce et al. 2019). Their posts do not have to be connected to certain special events. Rather, they post regularly concerning various topic-related aspects and produce content in order to maintain and increase their audience and influence (see Vu et al. 2021 and the example of NGOs).

A further, very general observation concerns the different practices of positioning and stance-taking on a meta level, in the case of communication about someone/some entity. While it is fairly common in some arenas of digital activism, in particular in »emancipatory discourses« (Zottola 2024, p. 45), to contribute to discussions by way of digital narrative sharing (Tienken 2013; Androutsopoulos 2014) in a first-person singular mode, our dataset shows only a few explicit examples of that practice (see chapter 4.1, ex. 3). Overall, (recurrent) digital narrative sharing, as it is regularly observed in other thematical discourses (e.g., Tienken 2013; Androutsopoulos 2014; Marx 2020), might be rather atypical for climate change or climate activist discourse, at least according to our dataset. However, a typical speaker’s position is a group- or community-oriented positioning, emphasised by the pronoun ›we‹ or the ›we‹-set (for collocations see fig. 1), which can have several meanings. For instance, it can include or exclude individuals addressed and/or individuals spoken of (Günthner 2021, p. 292). It can also be utilised as generic ›we‹. Other forms of positioning involve writing about third parties. This can include a ›tribute to someone/some entity‹ or to act as the ›voice of someone/some entity‹. We will come back to these patterns in subchapters 4.3 and 4.4. In the following, we will describe and discuss examples of our qualitative analysis by presenting the functional key categories and related linguistic patterns as well as the construction of identities and communicative practices in turn.

4.1 Awareness, Advice, and Agitation

The findings in this category represent several facets of encouraging others and thereby give insights into identity construction through positioning, addressing, stance-taking, and display of group norms or imaginaries. To get a better grasp of the functional spectrum, we summarised the facets of encouragement in three closely connected meta functions of linguistic acts: actors positioning themselves as individuals raising awareness for a cause relevant to climate change, giving advice about something climate-related, or even issuing a call for climate action. Concerning the latter, we want to suggest the term ›agitation‹ according to Davidson/Kemp, who describe it as one mode of envisioning climate catastrophe that leads to the demonstration of the stakes of the crisis or the necessity of action, with the primary objective to encourage climate action (Davidson/Kemp 2023, p. 2). However, it is not our aim to separate posts into three groups connected to the functional aspects ›awareness‹, ›advice‹, and ›agitation‹. Those are merely different accentuations of one category of encouragement which we differentiated from other functions, such as the display of solidarity, allyship, or advocacy, and the other categories presented as subchapters.

›Raising awareness‹ is probably the most basic function of climate activist social media activity (Karsgaard 2023, p. 216); it is almost always a co-function of both acting modes ›advice‹ and ›agitation‹ – and of all facets of encouragement in between. Moreover, the difference between ›advice‹ and ›agitation‹ relates to the associated urgency or emphasis of the encouragement, gradually visible in communicative details such as exclamation marks, repetition of direct addressing, strong visual emplotment (images, emojis), and many others. For instance, example (1) shows direct addressing: the expression is part of the first slide of post (A)¹⁰ and is followed by slides depicting an advising list of information and possible actions. In contrast, example (2), which is part of the image of post (B), illustrates a clear call for action.

• (1) »let’s talk about earth day…« (post A)

• (2) »we want to hear from you!« (post B).

However, ›let us‹ or ›let’s‹ in combination with a verb such as ›amplify‹, ›unite‹, ›emphasise‹, ›spread‹, ›help‹, ›raise‹, or ›enhance‹ is a prominent pattern in our dataset. Advice or calls for action that begin with these word combinations demonstrate a broad and inclusive group identity (both individuals addressed and individuals spoken of are in the in-group). In the following, we present further selected examples of posts of this functional category in order to illustrate typical associated constructions of group identities.

As previously stated, digital narrative sharing in a first-person singular mode is very rare in our dataset. In cases where this practice occurs it usually involves activist posts concerning the subtopic ›veganism‹. However, our examples support Giaxoglou/Georgakopoulou (2021, p. 241) in their notion that this type of storytelling can be equated with authenticity. Accordingly, the caption of post (C) demonstrates that such positioning serves the specific rhetorical purpose of raising awareness and giving advice from an authentic and experienced position (see ex. 3).

• (3) »I am nowhere near perfect, even when it comes to Veganism. I have been Vegan for 7+ years and I am still learning that certain things are not Vegan. For example, a lot of toilet paper isn’t Vegan. So weird right? However, I am grateful I learned that information about 8 months ago – and also VERY grateful that there is in fact Vegan and cruelty free toilet paper! It’s just a matter of bringing it into awareness and making compassionate necessary changes that are within our control. It’s as simple as ordering online or moving two inches to the left or right at the store and grabbing the Vegan TP instead of the TP that is made of skin, bones, cartilage, and tendons of animals. Now I shop for Vegan and cruelty free toilet paper! We cannot do better until we know better. Now I know better and can choose the most compassionate choice possible.« (post C, our italics)

In this caption, the ›authentic self‹ shares an experience by narrating a story, while self-evaluating and giving advice simultaneously. On a text-structural level, the alternation between »I« and »we/our« shows the combination of the narrated experience and the stance-taking or lesson learned. The latter is shared as advice, but at the same time, it is indirectly normatively mandated. This latent appeal becomes explicit in the highlighted line »We cannot do better until we know better«, since the reader now knows, too, and is indirectly addressed through the inclusive »we«. Furthermore, the post uses ten slides; each slide adding another generic statement such as example (4).

• (4) »Veganism isn’t about being perfect.« (post C, slide 1)

Here, veganism as a noun or a state of being stands totum pro parte for the associated group members (vegans). It attests them that they do not have to be perfect to share the group identity. The author’s demarcation between ›veganism‹ and ›perfectionism‹ reflects or reacts to a pragmatic presupposition, an invisible but well-known discursive or public prejudice towards vegans: the expectation of perfectionism and the exposure of ›imperfect‹ actions of vegans is a typical pattern of counter rhetoric (by the out-group) in public and moral negotiations about ethical and climate-related dietary decisions. Thus, the advice of the post concerns not only the providing of vegan knowledge but equally addresses the externally imposed fallacy of faultlessness.

In our next example (fig. 3, post D), the informative aspect of providing knowledge about climate change is combined with an agitative dimension. Possible consequences of unrestrained climate change are depicted on the slides, ranging from floods to forest fires. The temporal perspective is thus set to the future, constructing catastrophic climate imaginaries. Linguistically, the use of the ›we‹-set is prevalent: »[t]he longer we delay« (fig. 3, slide 1), »our planet« (fig. 3, slide 2), »brings us closer« (fig. 3, slide 3), and »the more we lose« (fig. 3, slide 4). In the caption of the post, this is continued.

Fig. 3

Post (D), Instagram, 09.08.2021, slides 1–4 and caption

The post as a whole can therefore be understood as a call for action directed at the group evoked by »we«. This is either done implicitly and rather vaguely as on slides 1, 2 and 4 (fig. 3), where the visual scenarios of catastrophe and loss are accompanied by text that portrays these visuals as the consequence of inaction. For example, »[e]very bit of warming makes our planet less safe« (fig. 3, slide 2) builds on the tacit knowledge that the ›warming‹ mentioned is causally connected to possible concrete action against climate change. What exactly this action could be remains unclear at first. Nonetheless, the respective slides exhibit the same functional structure: an undesirable scenario is shown, while the co-text (text-in-image) construes this scenario as the outcome of idleness, implying that something can be done to prevent these undesired situations from happening. As a result of this implication, alleged agency is attributed to the addressees, which, however, has less of a character of ›control‹ but rather resembles ›responsibility‹.¹¹

This is made explicit on slide 3 (fig. 3), where it reads that »[e]very action taken to limit emissions brings us closer to a safer future«. This is echoed in the caption with »we can make a difference« (fig. 3, line 13) and »we must act now« (fig. 3, line 16–17). Additionally, the imagery used on slide 3 (fig. 3) draws from collective knowledge within the climate movement (and possibly climate policy making). The text is written in colourful letters assembled in four rows, while a colour gradient of blue and red hues is visible from left to right. Users familiar with these stripes will be able to recognise this as a citation of the »warming stripes« Ed Hawkins, Professor of Climate Science, created, which have become a symbol of climate activism globally (Rosch 2023). They can themselves be interpreted as a call for action, as they display the change in mean global temperature over the last 150 years, while the red stripes indicate an increase that is cause to alarm. The use of this symbolism suggests that, in this post (D), ›we‹ denotes an in-group of climate activists. Looking at the caption, this impression is reinforced by the fact that »Wildlife Trusts across the UK« (fig. 3, line 21) and »global leadership« (fig. 3, line 29) are both syntactically differentiated from »we«. The in-group, then, appears to be a vaguely defined set of people who care deeply about stopping climate change but who are not themselves policy makers in any way. The actions called for remain on a rather abstract level as stated in slide 3 (fig. 3): »action taken to limit emissions« or in the caption: »[r]estoring nature« (fig. 3, line 18), and »improving resilience to inevitable climate change« (fig. 3, line 20–21).

Overall, post (D) is a typical example for climate activism on social media, where several linguistic and multimodal micro practices are utilised to raise awareness and move people to take action. Viewing the slides one after the other, the post can be read as a list of consequences climate change could have. This kind of list-making can be encountered frequently, like in post (E), from which we have extracted example (5). This post, too, uses the technique of displaying catastrophic scenarios of the future that are to be prevented, while the measures that could be taken remain implicit or vague but are nonetheless stated to be vital. The last slide reads (text-in-image):

• (5) »we cannot afford climate inaction« (post E)

The practice of ›list-making‹ as a call for action can also be used more directly, as in post (F). Here, three actors are co-authoring, which is a newer but fairly common Instagram feature. The platform affordances allow users to speak as a group which makes an even better argument for the analysis of group identities and adds another meaning of ›we‹ (only individuals who post) to the collection. Additionally, we found some inconsistent text-structural aspects in a few co-authored posts, where further research is needed. In post (F), the caption entails a more generic and short call for action; the focus lies on the visible part of the multimodal ensemble. With continuous use of the practice of ›list-making‹, the post assembles possible responses to the challenge of climate change that could build a better future. Rather than showing the negative outcomes of climate change as a trigger for action, positive consequences of actions are used here. Notably, both varieties of this strategy do not provide any hands-on suggestions for action, but rather operate with desired outcomes of actions (e.g., ex. 6 and 7).

• (6) »tripling of global renewable energy« (post F, slide 4)

• (7) »halt and reversal of deforestation« (post F, slide 5)

While the ›list-making‹ of the posts (E) and (F) is spread over multiple slides with one slide for each (catastrophic) aspect or bullet point, post (A) (fig. 4) illustrates a different but even more compact form of a list. Slide 4 (fig. 4) includes a collection of six circles raising awareness for scientific information concerning »the State of the Planet« in a comprehensible and rather non-agitative manner. In contrast to the above-mentioned typical lists, slides 6 and 7 (fig. 4) entail a list of very concrete projective practices (Ayaß 2021). These are formulated as advice for individuals’ everyday life, probably at individuals from a high-income group, since (free) time and (garden) space are required to realise the suggestions.

Fig. 4

Post (A), »Happy Earth Day!«, Instagram, 22.04.2023, slides 1–4, 6–7

Besides the list, post (A) (fig. 4) depicts drawings of the anime character Totoro on mostly greenish and pastel coloured ground. In the Japanese animated movie My Neighbour Totoro (1988) by Hayao Miyazaki, the eponymous Totoro represents a nature spirit. Here, this medialised and anthropomorphised entity is hugging the planet (fig. 4, slide 1 and 2) and holding a sign (fig. 4, slide 3) that says »protect our home« to raise awareness. However, here, too, it is unclear who the designated in-group (›our‹) is. Addressees could be individuals of an in-group that recognise Tororo as a nature spirit as well as individuals who feel aesthetically or content-wise appealed to by the design or the information on the slides. From a metaphorical point of view, Totoro can stand for both a representative of the natural world/nature addressing humans as well as an inclusive caller, who speaks for and addresses all planetary entities at the same time. An indication towards the latter is that Totoro also appears as part of the illustrated list (fig. 4, slide 6 and 7). Here, the post is communicating trough Totoro to raise awareness and give advice (concerning the aspect of ›voice‹, see also chapter 4.4). This list elaborates eleven answers in the manner of projective practices concerning the question »What can I do to help our planet?«. In contrast to the variations of ›list-making‹ mentioned above, this one most closely resembles the category of ›special list-making‹ (see von Contzen 2018, p. 319 and her example of a New Year’s resolution).

Furthermore, as the slides 1–3 (fig. 4) as well as the used hashtags in the caption show, the post is related to the annual event ›Earth Day‹. Correspondingly, post (F) is connected to the annual event ›COP‹. This broadens the audience immediately and, therefore, the potential in-group. Additionally, we can observe the intertwined manner of our functional categories. Although a central function of those two posts (A and F) is to raise awareness, give concrete advice, and call for action, the posts are similarly relevant for another functional category, which is named ›ritualised appreciation‹ and will be elaborated in the following by returning to both of those examples.

4.2 Ritualised Appreciation

Celebrating rituals, holidays and commemoration days evoke well-known offline practices of remembrance, acknowledgement, and appreciation – for instance, national or religious holidays.¹² Additionally, we can count an increasing number of soft or commercial holidays (e.g., the current form of Valentine’s Day or Black Friday) as well as international commemoration days concerned with important social or political topics, such as International Women’s Day and Earth Day (Weaver 2005). Many of those holidays and commemoration days are celebrated offline and online with the respective (communicative) practices at hand. Jones et al. point out that »digital technologies, because of the different configurations of modes and materialities they make available, both make possible new kinds of social practices and alter the way people engage in old ones« (Jones et al. 2015, p. 3); this also applies to the ritualised appreciation of holidays and commemoration days. Hence, digital celebrations of such days also evolved into approaches of social media marketing.¹³ While using those days in connection with individual marketing goals has become a conventional digital practice, social media activist communities utilise the more or less genuine celebration of such days in a different way. Nonetheless, the lines are fluid or at least blurred; activists are sometimes equally interested in improving their social media engagement or increasing their audience and influence (Vu et al. 2021). From a linguistic perspective, we can merely recognise the marketing potential of international commemoration days but cannot investigate or verify it. We rather concentrate on the associated practices, especially on a text-structural level, in an activist discourse.

Returning to the above-mentioned post (A) (fig. 4), slide 2 and 3 show multimodal practices on a micro level which we identified as typical for this category. Slide 2 resembles slide 1 in terms of the main visual (a drawing of the globe hugged by the Anime character Totoro) but mentions below – in a dictionary style – that Earth Day is a »noun« (fig. 4). The slide includes the phonetic spelling of ›Earth Day‹ as well as further information, again, in form of a small list: firstly, date and recurrence frequency, secondly, the purpose of the day. Slide 3 (fig. 4), however, shows another typical structural element: a timeline of connected historical aspects and events in two columns covering the time between the first Earth Day to the year of the post. Interestingly, the points on the timeline refer neither to all and nor exclusively to Earth Days; they consist partially of selected Earth Days and also include milestones such as the publication of Rachel Carson’s Silent Spring (1962). The prioritisation of positive news seems to significantly influence the decision on which historical aspect is worth mentioning, creating a specific positively connotated narrative. The caption’s beginning (8) emphasises the positive and appreciative notion as well as the associated ritual.

• (8) »Happy Earth Day!

  

  

  an update to last year’s post« (post A, caption)

Not only international soft holidays such as Earth Day, but also other events such as COP, for instance, can involve ritualised appreciative practices. The above-mentioned post (F) uses the annual ritual of the COP conferences to lift the group spirits (see ex. 9 and 10).

• (9) »As COP28 enters its final days, we are at a tipping point« (post F, slide 1)

• (10) »To reach this positive tipping point, we need the following →« (post F, slide 1)

Here (ex. 9 and 10), the term ›tipping point‹, usually used to describe the cataclysmic point of no return in climate change is recoined to designate the point where climate policies turn for the better. This positive recoining of a concept or the one-sided selection of news or other content is a typical framing practice of the category ›ritualised appreciation‹. Typical posts of this category use well-known hashtags, display positive narratives and introduce cheerful verbs such as the following examples (11 and 12) demonstrate.

• (11) »Happy World Vegan Day« (post G, slide 1, our italics)

• (12) »

  

  Embracing Vegan Day: Debunking Misconceptions

  

  On November 1st, we celebrate Vegan Day, a perfect time to shed light on some common misconceptions about veganism.

  

  « (post G, beginning of caption, our italics)

Expressions such as ›embracing‹, ›celebrate‹, and ›perfect day‹ set a positive emotional tone on the word level. Other posts of this category take on the task of providing group-related facts, stances, and actions in form of positive narratives or news in a monthly or weekly rhythm or ritual. For instance, post (H) uses sentences such as (ex. 13), frequently reminding readers of positive developments.

• (13) »positive climate news of the week« (post H, slide 1)

Although this type of climate ›activism‹ and social media interaction certainly attracts a lot of attention, such posts can pave the way to ›slacktivism‹ (Greijdanus et al. 2020), which is critically discussed regarding its effects for genuine online and offline activism. Engaging in some cheerful posts alone or changing one’s picture every now and then applies not only to the concept of ›slacktivism‹ but also to ›performative allyship‹ (Kutlaca/Radke 2023). This aspect will be discussed in the following category, the meta-function of showing solidarity and being an ally.

4.3 Solidarity and Allyship

Taking a closer look at our data and the relationships between actors, others, and groups, a spectrum of in-group and out-group solidarity¹⁴ becomes evident. This spectrum ranges from a clear support of individuals ›being in the same boat‹ (which can be described as social or group solidarity) to forms of political solidarity, which can be a matter of commitment to a movement, e.g., on behalf another group (Laitinen 2022, p. 8). Following up on the latter, a further specification is analytically beneficial: support and solidarity in the form of allyship.

Before delving into the functional category of ›allyship‹ within the context of group identity construction and against the backdrop of the concept of solidarity, it is important to emphasise that the communicative acts of ›demonstrating solidarity‹ or ›appealing to solidarity‹ serve as exemplary illustrations of the fluid and dynamic nature of identities as a result of (communicative) co-construction. Our data presents a resurfaced post (I) from a previous COP conference, clarifying the interrelation: in 2021, Steven Guilbeault was appointed as Canada’s Minister of Environment and Climate Change. The caption of the post highlights that this Minister is a former Greenpeace activist. The text in the image underscores this by combining the expression in example (14) with an arrow pointing to two activists attaching a protest banner to a large building.

• (14) »This is Canada’s new Minister« (post I, our italics)

In the beginning of the caption, the post positions Guilbeault as an in-group activist, primarily to remind him of his activist roots and affiliation with the in-group, while subsequently leveraging the advantages of his political identity to achieve climate-related objectives for Canada at COP. Both parties, the Minister and Greenpeace, can speak for themselves, are heard, and can position themselves as well as each other co-constructively.

›Allyship‹, however, draws on solidarity but differs significantly as a concept. It is a well-studied phenomenon in the psychological literature on intergroup relations; arising in social justice discourses about racism, LGBTQIA+ rights, or gender. Today, the term is known, discussed, and also sometimes criticised in several public or academic online and offline arenas which relate to various marginalised groups (Brown/Ostrove 2013). Generally, an ›ally‹ can be defined as a member of a dominant group who actively supports marginalised communities with the aim of promoting social justice and communicating support for social change (Washington/Evans, 1991, p. 195; Brown/Ostrove 2013, p. 2212; Collier-Spruel/Ryan 2024, p. 84). In the discursive arena of social media, ›allyship‹ means enhancing and amplifying voices of groups that are normally not heard in certain discourses but are nevertheless present. An example would be (offline) protesting garment workers, who have clear demands (Amnesty International 2024). In social media climate change discourses, their voices can be enhanced by allies such as the actor fash_rev (fig. 5).¹⁵

Fig. 5

Post (J), Instagram, 05.12.2023, slides 2, 8, and caption

Academically, psychological discussions on allyship include, for instance, the reasons for and the meaning of ›active support‹, the different manifestations of (actual) allyship, and questions about the effects of it (Brown/Ostrove 2013; Droogendyk et al. 2016) or its quality as a helping action (Collier-Spruel/Ryan 2024). From a linguistic perspective, which focuses on the display and sedimentation of communicative patterns, social media solidarity is an interesting variation of allyship behaviour. Here, adapted concepts such as ›performative allyship‹ (Kutlaca/Radke 2023) or ›slacktivism‹ (Greijdanus et al. 2020) arise. Furthermore, recent scholarship has shifted towards conceptualising ›allyship‹ as a facet of ›identity‹ rather than solely an action¹⁶ (Carlson et al. 2020 with reference to McKinnon 2017 and McKenzie 2014). Both those developments serve our interest to regard ›solidarity and allyship‹ as one of five key functional meta categories within our linguistic analysis of the construction of (group) identities on social media. With the following example, we want to show how social media actors display support and allyship in their interactive engagements and simultaneously construct a sense of who they are and in which constellation (opposing) group affiliations are traceable. Thus, in terms of linguistic markers of identity, ›allyship‹ is a fascinating category which displays the entanglement of individual and group identities as well as the multilayered, dynamic, and ambiguous facets of (self-)positioning.

We have decided to discuss post (J) (fig. 5) concerning the call »to end fossil fashion« (slide 2), here because it stems from a somewhat peripheral discourse within the climate debate, namely that of the climate impact of fashion products. However, the occasion of the posting is both the social and climate-related situation in the fashion industry in general as well as the COP28 event. The post is largely an informative contribution raising ›awareness‹ for its cause, but we want to look at the communicative act of ›allyship‹ being performed in more detail. On slide 8 (fig. 5), the exploitation of »garment workers« and their unacceptably low wages are pointed out. This is a typical scenario and an exemplary narrative pattern for social media allyship that enhances the voice of groups that cannot be heard in the current discourse (here: COP28) and, on the other hand, constructs an identity for the author that is based on the empathetic adoption of the plight of other marginalised groups. The associated narrative practices have a strong ethical foundation, which is why we understand them as displays of ethical imaginaries.

For the act of ›allyship‹, the ultimate goal is the amplification of the marginalised group in question. This becomes evident in the caption, where it is first stated that »the resourcefulness and creativity to tackle climate change already exists within the fashion community« (fig. 5, line 20–22). This is notable, because for allyship to make any sense, the group that is marginalised needs to exist. If there were no one in the fashion industry fighting for improvement, there would be no option for allying. Allyship identity, then, is defined greatly by the support for another group. Hence, for its construction, the other necessarily needs to be co-produced in language and/or image. Structurally, the final step of ›doing allyship‹ is to lead attention away from the own group and highlight the other group. Here, this is done by stating that they want to »ensure that workers’ voices are heard and the people who make our clothes aren’t left behind« (fig. 5, line 36–39), and by mentioning further supporting groups. Such narrative practices of calling for support are simultaneously displays of support contributing to the post’s function of ›allying‹. The central identity trait of ally communities is paradoxical in its performativity because its main strategy is to background itself and foregrounding someone else. Rather than providing rich, positive self-descriptions, allies’ identities’ fundamental self-description is negative in the sense that it constantly points to the fact that they are not the group they feel strong compassion for. One could thus argue that their identity relies on the continual rejection of the notion that they are identical with the group they identify with.

4.4 Voice and Advocacy

Two meta functions which we want to combine here and that are very similar to ›allyship‹ is what we call ›voice and advocacy‹. In fact, it proves difficult to always tell the three apart. This shows that functional language use is not restricted to one singular aspect, but rather unifies various functional means that are closely intertwined (Hausendorf et al. 2017, pp. 241–244). Still, we find the differentiation into two categories and chapters (4.3 and 4.4) useful because it points to nuanced variations of identity construction.

Both ›voice‹ and ›advocacy‹ are used to describe a practice where a group’s grievances are spoken for in a surrogate way. We encounter this primarily whenever individuals or subgroups which are deemed representative of a larger group speak on behalf of that larger group. In this sense, the speaker acts as the voice and as the advocate of said group. We are able to distinguish two variants of this, which we call ›voice‹ and ›advocacy‹ respectively: the first describes instances where non-human entities who literally cannot speak for themselves are given voice. The other involves cases where marginalised groups who do not possess societal power to speak are given voice (Spivak 1988). The first we find prototypically when animal rights are discussed by vegan activist groups (e.g., fig. 8, post O, see analysis below), the second appears in our data, for instance, when a smaller group of Indigenous Peoples speaks on behalf of Indigenous Peoples globally concerning climate justice (e.g., fig. 6, post M, see analysis below). Both seem akin to the concept of ›allyship‹, and in everyday language, these are probably not differentiated. Nevertheless, both are similar to ›allyship‹ in different aspects.

Fig. 6

Post (M), Instagram, 04.12.2023, slides 1, 3–7

›Voice‹ resembles ›allyship‹ in terms of identity structure because the people speaking on behalf of animals are not themselves animals. Prerequisite for ›voice‹ therefore means not being part of the group you want to speak for. In the case of ›advocacy‹, however, the speakers are indeed part of the group they are advocating for. The following example should highlight the difference. In a post by the NGO culturalsurvival about the problems of transitioning towards a more sustainable economy, it reads:

• (15) »A ›green‹ economy based on corporate profit and environmental destruction is no solution. A Just Transition that centers on Indigenous Peoples’ rights and is led by Indigenous People’s knowledge and stewardship is the only viable solution.« (post K, slide 5)

The call for recognition of Indigenous People’s perspectives is evident. What is less visible is that the organisation who posted this is itself run by Indigenous Peoples (at least they claim to be). Thus, instead of pointing towards another group whose voice is to be heard, as is done with allyship, the speaker constructs themselves as part of that group and as an eligible representative. In the post mentioned above (fig. 8, post O), where ›voice‹ is given to animals, this is precisely not the case. It is very clear that the speakers themselves are not part of the group they are providing ›voice‹ for. Why, then, did we choose to place these into the same functional category? Admittedly, there would be good reasons to understand ›voice for animals‹ as a form of ›allyship‹ because the identity structure seems to be the same: a speaker constructs themselves as identical with a different group (here even of entities) than the one they speak for. Likewise, what we termed ›advocacy‹ can be seen as ›allyship‹ because structurally, marginalised people are given a voice. Another alternative would have been to view all three as very distinct categories.

For us, the important argument is this: the common element of ›voice‹ and ›advocacy‹ is that instead of enhancing the voice that normally cannot be heard in discourse, the speakers here are that voice. Conversely, a vital aspect of allyship in identity discourses is that those who are allies do not speak for the marginalised but point to their discursive contributions and thus try to render them visible (see chapter 4.3). With ›voice‹ and ›advocacy‹, the same applies in the sense that the unheard are made audible, but in a way where the speaker becomes the voice authorised to speak themselves on behalf of the marginalised.

Identifying and distinguishing these functions, then, can prove to be difficult, as further co(n)textual information, e.g., about the author, has to be considered in order to decide which meta function is most prominent in a respective post (Hausendorf et al. 2017, p. 244). Another example (16) shows how meta functions can occur simultaneously and entangled; this tweet reads:

• (16) »Climate change is not equally felt across populations with the most vulnerable being disproportionately affected, Women are often depicted as victims of climate induced water insecurity, yet they are also proactive adaptation actors« (post L)

The author of the tweet states their location being in Kenya, and judging from the profile picture and the first name, it can be assumed that he is male. If this is true, then the tweet can be read as a case of ›allyship‹, because the grievances of women as a marginalised group are pointed out as well as their proactive and adaptive actions. The focus is thus shifted to unseen actions of the other group. Applying a holistic view and including other communicative dimensions as Hausendorf et al. (2017, p. 244) suggest, alters the assessment: when the image that is posted alongside the text is taken into account, a case for the post being a form of advocacy can also be made. The picture shows a number of Black people in a flooded area, apparently attempting to cross through the water. Power line masts in the background indicate that the flooded area could once have been a road. There is no way to tell if the majority of the people depicted are women, but it appears that a large number of them are male. The visual element of the post thus references the textual mentioning of ›populations‹, rather than ›women‹ and their proactive measures. Presumably being from the Global South himself, the author therefore also speaks globally for the populations most affected by climate change as a pars pro toto.

Let us then discuss a more clear-cut example of the ›advocacy‹ function in the sense of a pars pro toto. In post (M) (fig. 6), it is stated on slide 4 that »[o]nly a small number of Indigenous Peoples attending COP28 will have badges that allow access to negotiations«. This is indicated as a citation by Jennifer Tauli Corpuz, Managing Director of Policy of the organisation Nia Tero who runs the account the post stems from. She is depicted on the slide and her jewellery, the mentioning of the Indigenous People’s name she belongs to (»Kankana-ey Igorot«), and the claim that most of the leaders of the organisation are themselves Indigenous Peoples all suggest that she herself identifies as a member of that group. She also states this as a self-identification on her profile on the website Climate One (Climateone 2024). The whole post (M) reads as an attempt to raise awareness for Indigenous Peoples’ role in climate policies. The argument in example (17) explains why this is important.

• (17) »Indigenous Peoples hold the keys to address the #climatecrisis, utilizing traditional knowledge cultivated over millenia« (post M, caption)

The caption continues that these goals can be achieved by ensuring better representation of this group at conferences such as COP28. Visually, it draws from well-known iconography prototypically associated with Indigenous Peoples; the elderly woman with dark hair and facial complexion associated with exposure to nature and the outside wearing what appears to be a fur coat is a realisation of a well-known topos not only of Indigenous Peoples, but also of wisdom and traditional knowledge. The background that shows both sunlight and rain enhance this impression. The use of such stereotypical visual elements reinforces the assumption that this is an advocacy post both by and for Indigenous Peoples. Posting such an image as an ally could well bring about allegations of tokenism or even racism.

Here, then, it is the Indigenous themselves who address their own plight (Karsgaard 2023). On slide 5 (fig. 6), the post informs about the efforts and contributions Indigenous Peoples make despite them making up a rather small percentage of the world population. On slides 6 and 7 (fig. 6), the rights as given by the »United Nations Declaration on the Rights of Indigenous Peoples« are spelled out. This can be seen as communication toward the in-group, reminding them of the rights they can claim for themselves. At the same time, it is a form of out-group communication calling for adherence to the laws by the out-group. This can be interpreted as either a call for support, addressing possible allies, or as a reminder for people who are not allies, possibly cautioning them that they can and will be held accountable. In the complementary logic of ›allyship‹ and ›advocacy‹, this post (M) can easily be quoted, distributed and hyperlinked by allies. Since it is created by the group itself and deemed appropriate by them, allies don’t need to be concerned about producing content that might be viewed as stereotypical, offensive, or racist.

On a micro level, another notable aspect of this post (M) is the linguistic practice of generalisation. We have already pointed to the rhetorical element of pars pro toto in the aforementioned example. Though the construction of the identity ›Indigenous Peoples‹ becomes self-evident in the juxtaposition of image and text where the identity is designated, it is not at all obvious to make this association from the image alone. Even though we have ourselves described the iconography as indicative of (stereo-)typical features of Indigenous Peoples, the image on slide 1 (fig. 6) shows, on a less abstract level, recognisable elements of a very particular group of Indigenous Peoples. It can confidently be assumed that many would be able to identify visual features of the Inuit. The Indigenous People of the Kankana-ey Igorot of whom Jennifer Tauli Corpuz is a member, however, live in the Philippines. Given the very disparate geographical, historical, climatic and cultural circumstances of these peoples, it is quite remarkable how easily the common identity of ›Indigenous Peoples‹ is readily available in today’s discourse. By means of depicting both of these women as parts of a common whole that is designated by »Indigenous Peoples« (e.g., fig. 6, slide 4), the textually stated »traditional knowledge« (ex. 17) is similarly construed as a structural whole, as if all Indigenous Peoples globally had access to the same knowledge. Consequently, this shared knowledge is described on an abstract level as »knowledge […] about how to best live in reciprocity with our planet’s lands, waters and skies« (fig. 6, slide 3). This is evidence for the notion that group identities are discursively constructed by linguistically claiming that a common set of norms, ideals, or other cultural elements is existent and shared by a certain group. In this way, irrefutable striking differences within the group can be shifted to the background and rendered insignificant, resulting in identities such as ›Indigenous Peoples‹ that are in reality quite surprising, but seem self-evident because of discursive mechanisms and their results. ›Advocacy‹ as a meta function plays a crucial role in this process as visual and textual identification is used to designate individuals as members of a specific group, thereby constructing these individuals as legitimate representatives of the group as a whole.

As outlined in the beginning of this chapter, the functional category ›voice‹ is related to ›advocacy‹ concerning the surrogative way of speaking for an entity. This comes into play in instances where non-human entities who literally cannot speak for themselves are given voice in the context of climate change topics. The following examples illustrate three distinct forms of ›voice‹, all respectively speaking for animals.¹⁷ The first (fig. 7, post N) pertains to wildlife and their representation in international political negotiations, such as those at COP28. The other (fig. 8, post O) relates to dietary and consumer habits within the climate-related subtopic of ›veganism‹.

Fig. 7

Post (N), Instagram, 29.11.2023, slides 1, 3, and caption

Fig. 8

Post (O), Instagram, 05.12.2023, slides 1, 4, and caption

Post (N) (fig. 7) shows a cartoon with six slides plus one photographic slide, posted on Instagram in shared authorship. Here, the collective statement is short and the relations between the authors is made visible via citation in the caption. It becomes clear that each of the five actors fulfils a specific role in this group. The caption sums up the communicative function of post (N): it calls for like-minded individuals to »[a]dd [their] voice to ensure that animals aren’t forgotten at COP28« (fig. 7, line 2–4). As Steen (2022) demonstrates in various chapters of her book on human animal linguistics, the communicative appearance of medialised animals can be investigated with respect to several facets of the human-animal-relationship. For our purposes, it is interesting how activists communicate for animals (Deininger/Steimer 2019), giving them a voice, and thereby positioning themselves.¹⁸ This can be done by integrating animals as actors in various formats, for instance, as an image or even a cartoon. The text-in-image of slide 2 (fig. 7) is a speech bubble giving ›voice‹ to the anthropomorphised giraffe saying »Hey! What about our seat?«, referring to the seats at the COP28 conference. In the image, the giraffe – a wildlife animal – positions the group of five (a lion, a rhino, a snake, and presumably a mouse or another rodent) as interested in participating in COP28. From a different perspective, the giraffe – as a medialised animal – is positioned by the human authors of the post. On the surface, as the caption of the post reads, the goal of the illustrated animals and the social media actors alike is to »secure a seat for wildlife at COP28« (fig. 7, line 7 and 8) and the function of the post is giving ›voice‹. Morally this is unquestionably an honourable act. However, the (cartoon) animals are only linguistically conceptualised as beings with agency (Steen 2022, pp. 397–432); they still lack agency or voice here, in spite of the speech bubble. As long as there is no proof of the respective animals’ aims or desires, it is rather a proclamation of animal-related messages through the animals by using these anthropomorphised cartoons as a medium. However, the authors not only position themselves as supporters of animals by giving them a voice, but slide 3 (fig. 7) also adds another aspect of identity construction to the situation. With this slide, the authors show that their stance in the caption serves the function of ›demarcation‹ (see chapter 4.5). They distance themselves from their own group representatives (humans) at the COP table – or at least from their cartoon representations, who do not want to make an effort for the wildlife and who respond to the animals’ question with: »Oh sorry. We would but there’s simply no room« (fig. 7, slide 3). Serving as pars pro toto for the COP discussion round, they illustrate that, according to these authors, the conference seems to reject animal perspectives.

The climate-related discourse about animals concerns not only wildlife but also other animals and pets. This includes the topic of ›veganism‹ and, therefore, the animals enmeshed in respective human dietary and consumer habits. Figure 8 shows two slides from a larger post (O) that includes slide 1 and eight slides similar to slide 4 (fig. 8), each illustrating a different animal that the author is speaking for and giving ›voice‹. Let us begin with the introductory text, though, since this post is multifunctional and has a long caption (over 2,000 letters, not completely printed in fig. 8) that formulates statements about vegans and non-vegans. It differentiates those groups according to their »philosophical belief system« (fig. 8, line 3) and the associated food and consumer decisions (ex. 18).

• (18) »Every time you make a choice regarding the lives of other animals, you are making a decision that is either vegan or is carnist in nature.« (post O, caption)

The caption classifies the entire post (O) as a »reminder« (fig. 8, line 1), which means it is informative, raises awareness, and – at least implicitly or through the creation of moral pressure – calls for veganism. With the rhetoric practice totum pro parte in the first line (fig. 8), the vegan in-group is described ex negativo by what ›veganism isn’t‹.

The post uses several other linguistic and multimodal practices, for instance, the above-mentioned list-making that is extensively explained with various examples in the chapters 4.1. and 4.2. The ›list-making‹ practice used here in slides 2–8 (e.g., fig. 8, slide 4) resembles in form the one in post (E) and post (F) in chapter 4.1. However, slide 1 (fig. 8) shows another kind of list, a demarcation list, comparable to a pros and cons list. On it, we can find expressions that we can sort into several categories such as ›characteristics‹, ›beliefs‹, ›stances‹, ›mindsets‹, ›features‹, ›connected concepts‹, ›acts‹, and even ›associated locations‹. Here, we can observe how list-making implicitly contributes to evoking semantic connections between concepts that are not overt elsewhere. It is the co- and context (in this case also the parallel visual structure) that enables recipients to merge neighbouring concepts semantically (Scharloth 2022, p. 223). Each line compares the side »veganism«, written in green font, to the side »carnism«, written in red font (fig. 8, slide 1). The same colour code is used in slides 2–8 (text-in-image, see slide 4 as an example, fig. 8) to provide a sentence completion task for the recipient: two possible endings (»My choice« and »Your choice«) for the sentence starting with the phrase »MY BODY« are given in a first-person singular mode. This means that, similar to post (N) (fig. 7), the author of post (O) (fig. 8) communicates for the animal through the animal. Giving a voice is the underlying action and function of the slides.

We picked this particular example (fig. 8, slide 4) for a specific reason, though. The presumably preferred choice of sentence endings results in the phrase »my body, my choice«, which is well-known in another public discourse, namely feminist and abortion debates (Weiss 2015). This immediately frames the slide with the respective connotation, cross-referencing the two discourses, and, therefore, adds the question about self-determination and agency to the post.

4.5 Community Demarcation

As stated in chapter 2.1, delaminating one’s own community from others is a fundamental aspect of collective or group identity construction. It is therefore self-evident that, functionally, all posts discussed here entail some element of what we call ›community demarcation‹; the own group is to some extent defined, described, stated to have certain norms and values, etc., while other groups are stated or displayed as being different. Both types of those evaluations include – direct or ex negativo – linguistic displays of positioning and stance-taking and, therefore, markers of group identities. Despite ›community demarcation‹ being an elementary category, we would argue that there are posts in our data where demarcation is the primary meta function. This is the case when the other meta functions described above are not clearly visible or hard to construe from textual and visual evidence, while differentiating between groups is explicit.

A few examples may clarify our argument. In figure 9 (post P) we find very little information about how to become vegan or what a vegan lifestyle looks like, which would be the characteristics of an informative post. Neither are there any tips, vegan recipes or other forms of advice given (e.g., in contrast to post C, ex. 3 and 4, chapter 4.1). It might be argued that there is an implicit form of ›advice‹ here, since one could construe »thinking about the victims« (fig. 9, text-in-image line 5) as a recommendation for those who want to be vegan but are struggling with discipline. One could also make an argument for an affirmative meta function, since people who already find it easy to live off a plant-based diet are at least implicitly congratulated for their attitude. However, the most striking aspect of the post is the explicit indication of »[t]he difference between those who find veganism easy and those who find veganism hard« (fig. 9, text-in-image lines 1–3). This is remarkable because even though it is very clear that a marked difference is indicated, it is not entirely unambiguous, as this difference could be understood as being between vegans and non-vegans, or between vegans and other vegans. In the text in the image (fig. 9), it would seem that the latter is true. The caption, however, can be read as a definition of true veganism, which is characterised precisely by the rejection of »products of evil« (fig. 9, line 3) while putting »the victim at the forefront of the thought process« (fig. 9, lines 6 and 7). In this sense, this post (P) is really about the difference between vegans and non-vegans, because people who refrain from consuming animal products for the wrong reasons are conceptualised as non-vegans. For instance, individuals who see their vegan consumer choices as an aspect of a sustainable, climate-friendly lifestyle – and who have corresponding ideals and values not only with regard to animal welfare, but even more in terms of planetary welfare – might be those conceptualised as ›non-vegan‹ vegans by the author of this post.

Fig. 9

Post P, Instagram, 10.09.2023

Here it becomes evident why we view ›community demarcation‹ as the primary meta function in this post (P). The focus lies on defining what the group’s core characteristics actually are in opposition to features of other groups. It draws the line between those who are truly vegan and do not find it difficult because their values are in line with the in-group’s morals and those who abstain from animal products for selfish reasons. It is noteworthy that a vocabulary of morality is used for this. The very term ›sacrifice‹ is renounced to describe the rejection of animal products. Whoever sees living vegan as a burden is already constructed as part of the out-group. The attribution of the term ›evil‹ to animal products, then, is in line with the moral, almost religious argument. In this post (P), we do not only find the construction of identity by self-description and delineation. There is also an argument to view this as identity construction of the fourth dimension described by White (2008, p. 11), for it seems that a fine-tuning of group identity is performed here, negotiating already established differences (consuming animal products vs. not consuming them) by introducing further levels of differentiation (rejecting animal products for selfish reasons vs. rejecting them for the good of the animals). Post (C) shows that this can take quite an opposite shape as it provides a more forgiving definition of veganism, stating that it is not important to practice the lifestyle perfectly. Rather, it is about another juxtaposition (ex. 19):

• (19) »the power of LOVE vs the love of power« (post C, slide 8)

We have already discussed this example in chapter 4.1, and indeed the aspect of giving advice is more dominant than that of demarcation. Our last example is one that does not align perfectly with our categories but is nonetheless (or maybe specifically for this reason) noteworthy.

• (20) »Let’s celebrate […]. Not in the slightest« (post Q)

The post (Q) (ex. 20) looks like an ›appreciation‹ post at first glance as it uses the distinctive phrase ›Let’s celebrate‹ to appreciate the successes of previous climate conferences (see chapter 4.2). However, this is meant sarcastically, as it turns out that there is nothing to celebrate; as the last sentence in example (20) shows. Given the accompanied graph that shows rising emissions despite climate conferences being held over the last 60 years, this can be seen as an information post relying on humorous rhetoric to raise awareness. There is also an interesting function of community demarcation here that is indicated by the refusal to use the pronoun ›we‹. Linguistically, this post is about ›the others‹ who are responsible for the conferences being »complete failures«. A clear difference is drawn between ›them‹ and the author, even though it is not clear what kind of group (activists, possibly) he belongs to.

5 Conclusion: Doing Identity on Social Media

Our multimodal analysis as well as the elaboration and differentiation of several social actions reveal that cultural aspects manifest themselves in various communicative practices and therefore can be traced with linguistic methods. Likewise, we demonstrate that those patterns play an important role for the public negotiation of societal norms and (group) identities, which become meaningful in the context of future imaginaries. As remarked upon at the outset, the distinction between macroscopic climate imaginaries such as a technological solution, fossil fuel regime, sustainable lifestyle, or an inevitable climate apocalypse (Levy/Spicer 2013, p. 663) are, in our eyes, too broad. We believe we would underappreciate the promising concept of ›climate imaginaries‹ for a Humanities approach if we did not utilise it to examine more concrete ideas, practices, and communicative patterns within public social media discourse to gain deeper insights into societies’ perceptions of climate futures.

For this, we wanted to understand the functional use of linguistic and pictorial material: ›giving advice‹, ›raising awareness‹, ›agitating‹, and ›displaying appreciation‹ as well as building group identities through ›community demarcation‹, ›doing allyship‹, ›advocating for others‹ or ›giving them a voice‹ are thus all indicative of certain social imaginaries oriented toward a future dependent on climate policy, environmental developments, and individual or group choices. In chapter 4, we have delineated diverse forms these meta functions can assume at the level of micro practices, providing numerous examples from both Instagram and Twitter/X. The benefit of such detailed linguistic analyses is that we can address different levels on which we can identify and elaborate communicative patterns within positioning and stance-taking. In chapter 2, we mentioned several aspects: pronouns and designations at the word level and various modes of action that can be found at a narrative level. Furthermore, interesting visual and structural aspects are promising entry points to describe how social actions are realised on social media. In this regard, we want to emphasise the described practice of ›list-making‹, which can involve different visual and textual patterns, indicate importance, and demonstrate preoccupation with existential (climate-related) issues.

Summing up the essence of the detailed analysis above, we noticed that further categorisations are possible which we want to outline as a conclusion of this article: Without seeking to confine climate imaginaries to specific topics, we aim to find a useful way to distinguish their nature: firstly, with respect to the temporal dimension, classifying them as retrospective, hopeful/utopian, or dystopian; secondly, with respect to agency and accountability, classifying them as individually agentive, communally agentive, or passive/mediated. All of these should, however, be understood not as clear-cut categories but rather as scalar concepts that the practices we outlined above can be allocated to and associated with as indices. Especially the agency categories can oftentimes not be determined definitively when we are looking at functional aspects. Allyship, for instance, is by definition ambiguous in this regard. As shown in post (J) (fig. 5), it is indicative of an agentive climate imaginary, as calling for support of marginalised groups is evidently something that has to be done.

However, this example also reveals that the responsibility for the real action needed to better the future ›should be taken by another group‹. In this sense, ›allyship‹ is evidence of a more mediated agency in the realm of climate imaginaries. The paradoxical identity construction is therefore mirrored in the allocation of responsibility. Generally, ›allyship‹ attests for a hopeful imaginary, for it presupposes that improvement is possible. The same applies to cases like post (O) (fig. 8), where ›voice‹ is given to animals. It can be concluded that all communicative practices that rely on ethical or moral stance-taking entail a hopeful perspective of the future. Similarly, social media posts providing ›advice‹ for others can hardly be said to think of the future as already lost. There also seems to be an implicit notion that only communal efforts exhibit prospects of success in preventing climate change. Examples such as post (C), which allows its addressed group members imperfection, can therefore take a much more forgiving approach to encouragement than agitative content like post (D) (fig. 3), where catastrophic outcomes of climate change are foregrounded. These seem to go hand in hand with communal agency constructions, where the accountability for climate action is mystified in the form of an indistinct ›we‹. This is in stark contrast to the attribution of individual agency that we find in ›advice giving‹ posts like post (A), where the question stated is »What can I do?« (fig. 4, slides 6 and 7). While it could be argued that the shared positive beliefs are more implicit in ›advice‹, examples like post (F) are more utopian than just hopeful, as they attempt to portray positive scenarios rather than apocalyptic imaginaries, while at the same time construing communal agency of possible change.

We are thus able to observe how identity construction on all levels described by White (2008) have a lot to do with conceptions about who is to do something about the future of the climate. When advocating as a legitimate representative of marginalised groups as in post (M) (fig. 6), this naturally tends to correlate with the focus on that groups’ agency for a better future. Allies of the same groups tend to echo this notion while simultaneously defining their own, less direct role of support – thus constructing their identity as allies. Appreciative posts about accomplishments of climate activism (post H) presumably strengthen activists’ identity and their conception of self-efficacy in contrast to the out-group that is implicitly constructed as inactive. Posts working with catastrophic and negative content like floods and animal cruelty tend to call upon a wider, less defined out-group, while the own role remains ›raising awareness‹, thus defining the in-group.

Whether or not these correlations between identity construction and climate imaginaries really are generalisable on the scale we suggest here we cannot say from our data; further research could investigate this on a quantitative level. However, it is apparent that the aspect of any given climate imaginary that construes the actors of change, who either build a better future or prevent an apocalyptic one, is connected to the ways in which community identities are functionally constructed.