“I don’t think education is the answer”: A corpus-assisted ecolinguistic analysis of plastics discourses in the UK

1.1 Plastics: A complex picture

Most notably hailed as a marvel of convenience in the 1950s (e.g. Life 1955), plastic is now perhaps better known for its polluting capabilities and its ubiquity as a form of ever-fragmenting waste. Its abundance and durability, which have made it a wonderfully cheap and reliable construction material for humans, have also endowed it with a lasting presence in ecosystems all around the world; some plastic has already occupied natural environments for decades beyond its original, typically fleeting, use (Lebreton et al. 2018). The long-term environmental impacts of plastic are claimed to be, at best, “unknown” and “irreversible” (Hohn et al. 2020), while other predictions are much more sobering: that plastic fragmentation is the beginning of a delayed, cumulative release of toxic chemicals, a “global plastic toxicity debt” (Rillig et al. 2021); and that plastic is “poised to dominate the 21st century as one of the yet-unchecked drivers of climate change” (Altman 2022).

However, one also cannot overstate the usefulness of plastic. Plastics enable cheap, readily available infrastructure that delivers life-sustaining water to billions of people in the form of pipelines, for example, as well as all kinds of food products in lightweight, transportable containers that can effectively preserve foods and reduce food waste; they facilitate sterile, affordable, and safe medical equipment in the form of catheters, sheets, syringes, protective packaging, and personal protective equipment; and there is an almost endless list of basic-necessity artifacts such as clothing, vehicles, technology, and affordable housing and building materials that are dependent on their plastic constituents (George 2020). Due to its complex enmeshment in global networks of production and consumption, plastic has become practically irreplaceable in many contexts. Not only is it physically enduring, but socially, culturally, and economically, too (see Evans et al. 2020).

To further complicate matters, plastic is not a straightforward material to dispose of responsibly. A range of different polymers are used and often combined, along with chemical additives, to make different kinds of plastics with specific material properties, and not all of these plastics can effectively be recycled (Bucknall 2020). Even those which theoretically could be recycled are often not, due to local logistical and economical constraints (Bucknall 2020). Citizens commonly report difficulty in knowing which items are recyclable – in the UK, this can vary even from one city to the next – and indeed cannot know for certain whether their efforts will ensure that the waste is eventually recycled (INCPEN 2021). Even when plastics are successfully recycled, the quality of the plastic is reduced with each iteration. This depends crucially on the type of plastic; for example, 100% recycled polyethylene terephthalate (PET) can be used in a closed loop if the plastic is upgraded at each stage, whereas polyolefins (polyethylene and polypropylene) need the addition of virgin polymer at every recycling event so that properties are maintained (see Eriksen et al. 2019). A life-cycle assessment, the calculation of all of the environmental impacts associated with a product, from its manufacturing to its disposal, may also rule out recycling as the most ecological option for a particular kind of plastic waste product (Walker and Rothman 2020). Beyond all of these material limitations, the story that recycling promotes – that single-use products are largely redeemable, and that disposing of them is unproblematic – does not serve to challenge our over-production and over-consumption of single-use plastics and plastic packaging. In short, recycling is not enough.

Rather than aiming to entirely avoid or simply recycle plastics, reusable plastic packaging systems are increasingly being looked to as a potential solution to plastic waste (Greenwood et al. 2021); a circular economy approach has the potential to reduce the yearly volume of plastics entering the oceans by 80% and to reduce greenhouse gas emissions by 25% (Woolven 2021). While reusable packaging may not be suitable for all products, there is a wide range of goods that have the potential to be provided on a large-scale refill or return basis. The “refill” option of such schemes normally involves taking one’s own reusable container to a store and filling it, or taking home a minimally-packaged product to refill one’s home container. The “return” option normally involves receiving a pre-packaged product and then returning the reusable container to the vendor or manufacturer, or having it collected. “Reuse” is sometimes also considered to include repurposing (i.e. buying a single-use product and later reusing it for a different purpose), but within our project, Many Happy Returns, packaging is considered to be strictly “reusable” only when it can be used more than once for the same purpose, i.e. through being refillable or returnable.

1.2 Linguistic intervention

In addition to technological innovation, achieving system change requires a sensitivity to public sentiment as well as the ability to communicate complex concepts in ways that will make sense to many people. This means observing, on a large scale, the ways members of the public express their views about plastics, disposal, and reuse, and then developing communication strategies that align with (or where necessary, challenge) the most prominent and pervasive attitudes towards plastic (re)use. Encouraging everyday behavior change on a large scale involves an uncovering – and reframing – of the cultural stories (Stibbe 2021b) underpinning behavior. These are not the stories that are “read to children at bedtime, shared around a fire, or conveyed through anecdotes in formal speeches”, but “cognitive structures in the minds of individuals which influence how they think, talk and act” (Stibbe 2020: 2). For this reason, it is essential not only that analysis of the discourse used around plastics is based on the systematic examination of large, representative linguistic datasets, but that it is further underpinned by an understanding of language that aligns with the most up-to-date research in the cognitive sciences. Such an understanding is provided by cognitive linguistics, which forms a further important component of our theoretical approach.

The way we frame complex environmental issues is especially critical, according to Lakoff (2010). “Facts”, he argues, “must make sense in terms of their system of frames, or they will be ignored […] to understand something complex, a person must have a system of frames in place that can make sense of the facts” (Lakoff 2010: 73). We understand “facts” here to refer to any kind of information that is conveyed in the interest of public science communication. If we are to help laypersons, i.e. most members of the public, to properly understand the complex nature of plastic and its correct use and disposal, linguistic reframing strategies must be deployed with a sensitivity to the stories and frames that already exist in the minds of many people.

Ecolinguistics, then, which “seeks to explore linguistic phenomena found in inter-language, inter-human, and human-nature relationships from the perspective of ecological philosophy” (Chen 2016: 109), represents an ideal means of unravelling plastic’s complexities, both as a material in our society and as stories in the minds of plastic users. We maintain that only through (i) careful and critical application of ecolinguistic principles, while (ii) drawing on cognitive linguistic expertise, and (iii) basing findings on large and representative samples of linguistic data, will it be possible to craft new, empirically reliable stories that help to encourage a more mindful relationship with plastic products. Of course, linguistic research is just one piece of the plastics puzzle, but it is a crucial one.

2.1 Science communication

Faced with complex technical and environmental problems, and in a so-called “post-truth world”, effective science communication is now recognized as being more important than ever (Kopf et al. 2019). Science communication can take a diverse range of forms, from documentary films and informational flyers to stand-up comedy, even (Bowater and Yeoman 2012), and its practitioners are increasingly being referred to as “knowledge brokers” (Meyer 2010; cf. Pielke 2007) or “knowledge translators” (Bielak et al. 2008) in an effort to emphasize the co-creative and dialogic nature of good science communication. Regardless of issue or medium, it is argued that public-facing communications ought to employ listening as well as speaking, and should engage with a wide range of audiences in ways that are tailored for each audience (Bielak et al. 2008). This requires first doing empirical research to understand what laypersons already know, as well as recognizing that communication about science “does not occur in a vacuum” (Fischhoff and Scheufele 2013: 1); science communication is, in itself, a form of political communication (Scheufele 2014).

Progressive public communications in the UK have been championed perhaps most notably by the Public Interest Research Centre (PIRC),^[1] whose framing and communication toolkits have spanned topics such as nature, equality, climate justice, and the economy. PIRC work with community interest groups as well as other framing specialists such as the FrameWorks Institute^[2] to evaluate how complex issues such as the environment and the economy are understood by members of the British public, and then to produce co-created booklets of framing and narrative recommendations for a range of communicators. They also disseminate their findings in the form of posters, card decks, and postcards; their ‘The Narratives We Need’ postcard, for example, distils the findings of 15 research projects and ten years of work into five key questions for public-issue communicators: (i) “Are you talking to people’s best selves?”; (ii) “Are you showcasing a big, diverse ‘us’?”; (iii) “Are you showing we can all play a role in change?”; (iv) Are you highlighting how the system is unfairly designed?”; and (v) “Are you demonstrating that change is possible?” (PIRC 2020). Along with the FrameWorks Institute and the US-based Center for Story-Based Strategy,^[3] PIRC represents a leading example of politically sensitive, narrative-driven, public-issue communications, or knowledge brokering. Similar approaches are gradually being adopted, based on more rigorous methodological foundations, by academics working with metaphor and narrative: an output of Semino et al.’s (2017) work on cancer discourses was an illustrated “metaphor menu”^[4] for cancer patients and healthcare practitioners to use in cancer-related communications; and Lockton et al. (2019) have translated their design research into a toolkit and a deck of cards that can be used in metaphor-generating workshops, for example.

Where pro-environmental communication is concerned, a growing amount of research is being carried out by marketing and business scholars, in particular, on optimal ways of promoting environmentally beneficial products and practices (Taylor 2015). Despite often having a somewhat capitalistic undercurrent, industry-oriented research on “green” marketing and advertising has produced many useful, experimental findings around effective use of language, particularly with regard to promoting pro-environmental behavior change. Studies on the appropriate level of assertiveness, for example, have found that gentler and less assertive phrasing is typically found to yield more compliance on a range of issues, including environmental practices (Baek et al. 2015), and that employing a greater number of “green” messages can in fact backfire, leading to reactance (Moyer-Gusé et al. 2019) and negative brand perception (Olsen et al. 2014). There are a few instances in which more forceful messaging may be more effective, however: when the recipient already invests some effort in the promoted behavior (Baek et al. 2015); when the issue is accepted to be of extreme importance (Kronrod et al. 2012b); or when the message encourages “hedonic consumption” rather than virtuous behaviors (Kronrod et al. 2012a). Interestingly, and perhaps unsurprisingly, an experimental study on the language of returnable packaging labels found that “ease of use” messages, as opposed to messages that justify why returning packaging is important, were more effective at raising consumers’ intentions to actively participate in return schemes (Ratnichkina et al. 2021). A key finding of WRAP (2021a), however, is that consumers typically spend very little time reading such labels – around ten seconds or less – and as such they must be clear, simple, direct, and easy to find if they are to have any benefit.

2.2 Strategic use of framing

A reasonably consistent finding across messaging studies is that negatively framed messages tend to be more effective than positively framed ones for influencing attitudes and behaviors (Amatulli et al. 2019; Davis 1995; Ganzach and Karsahi 1995; Levin et al. 1998; Meyerowitz and Chaiken 1987; Olsen et al. 2014; White et al. 2011). That is to say, messages that emphasize the benefits of engaging in pro-environmental behavior are typically less persuasive than those that emphasize the risks or potential losses associated with not engaging in that behavior, e.g. the benefits of purchasing a ‘green’ product versus the disadvantages associated with not purchasing it, or positively framed product claims such as “biodegradable” and “recyclable” versus negatively framed ones such as “no pesticides” and “no phosphates” (Olsen et al. 2014). Amatulli et al. (2019) argue that this effect can be attributed specifically to the activation of the emotion of shame, or “anticipated shame”, while White et al. (2011) propose that it is due in part to fluency, i.e. ease of cognitive processing. White et al. (2011) contest, however, that negative frames are always more effective: they find that negative, or loss-framed, messages are more persuasive in contexts of low-level or concrete actions (e.g. the act of reusing or returning packaging), while positive, or gain-framed, messages are potentially more appropriate for high-level or abstract purposes (e.g. fostering beliefs about why reusing and returning packaging is important). These are also the findings of Lord (1994), whose study on recycling concludes that the most effective motivator of recycling behaviors is a negatively framed message conveyed by a personal acquaintance, while positively framed messages are more effective at promoting a favourable attitude towards recycling generally.

All of these findings align with research in cognitive linguistics, which broadly agrees on a view of linguistic negation of any kind as having a fundamentally foregrounding effect (see Hidalgo Downing [2000] for a useful overview). Givón (1993) explains that this is because negation essentially reverses the normative means through which human beings understand stasis as a conceptual background, against which descriptions of events or change stand out in our minds. Givón points out that there is a predominance of language in everyday communication which fits this standard conceptual structure: events are construed through language as more informative than non-events. This predominance also affects how we perceive departures from it made through negation, where the event, rather than inertia, is established as the ground and a non-event becomes more salient and more informative (Givón 1993: 190). When negation is used in language, therefore, it is discoursally and conceptually deviant, drawing attention both to itself as a linguistic form and to the non-events it describes.

Negation also acts to defeat our expectations in the discourse-world of a particular communication and generates implicature as a result. Nahajec (2009: 110) explains that “in order to deny a prior proposition, implicit or explicit, we have to conceptualize or create a mental representation of what is being denied”; in Olsen et al.’s (2014) examples of labels like “no phosphates” above, then, readers of these messages must first mentally represent an affirmative situation in which the product does contain phosphates, in order to then understand the reverse of that state of affairs. The label also carries a set of presuppositions and implicatures, including some assumption that the product ordinarily might be expected to, or even should contain phosphates, but does not. Cognitive linguistic research on negation thus enables us to understand the potential rhetorical power of negatively framed messaging, which places additional demands on our conceptual processing, challenges our expectations and presuppositions, and has an overall disruptive pragmatic effect in discourse.

2.3 Linguistic approaches

Cognitive linguistic studies, drawing on both psychology and narrative theory, provide supporting evidence for the benefits of a narrative-based approach, both to the communication of scientific research and to the creation and nurturing of movements for social change. Perhaps most notably in recent years, we have seen the rise of emotive, storytelling documentaries that have succeeded in motivating public and even policy discussions around environmental concerns (Nolan et al. 2022), a phenomenon commonly dubbed the “Blue Planet effect” (Gell 2019). The powerful role such storytelling can play in creating or encouraging mass social movements and systemic change is well attested in cognitive accounts of narrative persuasion (see Green et al. 2002). Jacobs (2002), for instance, argues that narrative structures are instrumental in the forging of collective social identities and also allow members of those collectives to be furnished with a set of stories about a shared past, present, and future. Narratives, in essence, allow people to rationalise and order events in a conceptually manageable chronological sequence and to position themselves and others within that sequence in roles such as “hero” or “villain”. As Jacobs puts it, the creation of plot “encourages the public concentration of attention onto specific events, encouraging discussion about the meaning of those events”, while narrative features such as opposing characters “serve to dramatize a society’s deep cultural codes, increasing the likelihood of a continued emotional investment in public life” (Jacobs 2002: 218).

Ecolinguists, not to be confused with language ecologists (Alexander and Stibbe 2014), contribute to our understanding of environmental discourses by critically analyzing instances of language for its pro- and anti-ecological sentiment. Stibbe, especially, has furthered the field of ecological discourse analysis by drawing on cognitive linguistic concepts of story, framing, and metaphor, as well as critical discourse analysis methods (e.g. Fairclough 2001; see Stibbe 2001), to develop a coherent approach to ecolinguistics: a positive, critical, ecological discourse analysis (see Stibbe 2021a, 2021b). Doing ecolinguistics in this sense involves taking an explicit ecological and philosophical stance and then using a range of discourse analytic approaches to determine whether or not a discourse is aligned with that philosophy (see Section 3.1 for more on ecosophy). Ecolinguistics has been used to analyze the language of economics textbooks (Stibbe 2020, 2021b), haiku poetry (Stibbe 2007, 2021b), computer games (Poole and Spangler 2020), road users (Caimotto 2020), environmental debates (Poole 2018), climate change news stories (Norton and Hulme 2019), news reports (Cheng and He 2021), and tourism texts (Ponton 2022), to name a handful of recent examples. To our knowledge, it has not yet been used to analyze the language of plastics and plastic packaging, nor has it been used in industry-facing research, e.g. on consumer messaging strategies.

Increasingly, ecolinguistics is also being employed in conjunction with corpus linguistic methods, a survey of which can be found in Poole (2022), the first book-length treatment of corpus-assisted ecolinguistics. Doing linguistics with a corpus – literally, a “body” of text – simply means conducting linguistic analysis on a very large, computer-readable text or set of texts and, consequently, employing computational methods in processing and analyzing said text(s). The fact that a corpus can be constructed from any of a wide range of texts (or indeed other modalities), and that there is no prescribed method for doing corpus analysis, means that corpus methods can be applied to a wide range of language data and research questions. Corpus-assisted discourse analysis, for instance, is now commonplace and has been used to explore a host of societal and environmental issues, from media representations of mental health (Hunt and Brookes 2020), Islam (Baker et al. 2013), and feminism (Jaworska and Krishnamurthy 2012), to discourses of nonhuman animals (Sealey and Charles 2013), the environment (Lischinsky 2015), and climate change (Grundmann and Krishnamurthy 2010; Koteyko et al. 2013; Liu and Huang 2022), amongst many others. When it comes to corpus studies on plastic packaging and related themes, we are aware of just two: Napolitano (2018), a corpus-assisted analysis of media discourse around the plastic bag ban in Australia; and Niceforo (2021), a corpus-assisted ecolinguistic discourse analysis of two public-facing reports on plastic pollution.

Corpus linguistics and cognitive linguistics have also been successfully combined, often with corpus data providing an additional testbed for existing cognitive linguistic theories (Grondelaers et al. 2007; see also Arppe et al. [2010] and Gries and Stefanowitsch [2006] for full discussion of the benefits and challenges of combining corpus-linguistic methodologies with cognitive theory). Cognitive linguistics also offers valuable insight into metaphor, instances of which can reliably be observed with the use of a corpus (Semino 2017). The relationship between corpus frequency and specific cognitive processes such as entrenchment (Schmid 2007), however, is not entirely certain, and has been complicated by methodological variation in how researchers measure both corpus frequency and processes including entrenchment (cf. Mehl 2018). Nonetheless, text frequency – one of the main affordances of corpus linguistics – can be interpreted as representing an exposure rate (Wallis 2012), and the more a language user is exposed to a linguistic feature, the more the cognitive activation of that feature ought to become routinized or entrenched (Schmid 2007: 119–120). This claim would apply to the entrenchment of linguistic frames or stories such as those around plastics and reuse.

2.4 Language and Many Happy Returns

Many Happy Returns is a large, multidisciplinary project funded by UK Research and Innovation (UKRI), specifically as part of the UKRI’s Enabling Research competition in its Smart Sustainable Plastic Packaging Challenge. The aims of this funding initiative are “to find solutions to existing issues with plastic packaging, reduce plastic pollution and unlock barriers to create fundamental changes in the industry” (UKRI 2020), and MHR’s proposed contribution, as mentioned in Section 1, is to innovate and encourage reusable plastic packaging systems as a means of reducing single-use plastic waste. To this end, it has five research teams, each covering one of the work packages (Language, Change, Willingness, Technology, and Life Cycle), and represents seven departments (English, Geography, Psychology, Mechanical Engineering, Chemical and Biological Engineering, the Management School, and Chemistry) across four faculties at the University of Sheffield. Beyond the University, the project has external partners ranging from non-profit organizations such as OPRL (On-Pack Recycling Label),^[5] City to Sea,^[6] and WRAP,^[7] to major UK supermarkets such as Morrisons, Co-op, and Marks & Spencer, as well as larger, multinational corporations including Nestlé, Unilever, and Berry Global.

While our Language work package is certainly distinct, it is also designed to feed into and simultaneously be informed by the other areas of research on the MHR project. Our geographer colleagues in the Change team, for instance, are documenting the everyday lives of members of the public by observing and photographing their engagement with plastics and reuse within the home. Meanwhile, psychologists on the Willingness team are running controlled experiments to determine the limits of perceived aging and staining that British consumers are likely to accept in reusable plastic food containers (Baird et al. 2022). Elsewhere in the project, mechanical engineers are submitting candidate container materials to a range of stress tests, such as repeated scratching and industrial washing cycles, while chemical engineers are calculating the most ecologically viable packaging solutions by means of detailed life-cycle assessments. Our ecolinguistic work has closer ties with some of these work packages than with others, but ultimately learns from, and can contribute to, all areas of this multidisciplinary research.

The cross-disciplinary relevance of (eco)linguistics is not always obvious to all stakeholders, however. Of the ten university-led projects participating in this UKRI challenge, MHR is the only one to include a work package dedicated specifically to language. Reasonably, one might ask where linguists, with little to no background in plastics research, fit when it comes to research on packaging technology and polymer science. We would argue that our limited background in plastics is not only acceptable, for linguistic analysis, but in fact preferable. To be new to a subject can be an advantage for linguists, according to Stibbe, as “the stories that a discipline are based on may be so ingrained within the discipline that they go unnoticed” (Stibbe 2021a: 80). Drawing on Meyer and Land’s (2005) work on “threshold concepts” – the concepts that, once understood, irreversibly change one’s outlook and can even lead to shifts in disciplinary identity – Stibbe proposes that it may be better for discourses to be analyzed by those who have not yet crossed such thresholds (Stibbe 2021a: 80). Another central argument in his discussion is that ecolinguistics is not a subject or a methodology but can be considered a transdisciplinary movement that is not to be confined to the linguistic; to do ecolinguistic work is to be engaged in the “real world” and to have an “ethical and practical dimension” (Stibbe 2021a: 84). It makes sense, as we see it, to try to use (eco)linguistic research to transcend the disciplinary boundaries in a project as broad and complex as this one, and we recognize the general lack of linguistics in plastic research as a significant disciplinary gap.

3.1 Ecosophy

Over the course of our research on this project, we developed and subsequently formalized our ecosophy, or ecological philosophy. Now commonplace in ecological and ecolinguistic research, the ecosophy is an ethical framework to be set out at the start of the work, inspired principally by Naess (1990). It provides a philosophical grounding for the research and an ethical standard against which we can compare instances of language to determine whether or not they align with our values and overall mission (the overarching mission of our wider project being a reduction in single-use plastics). Drawing on Guattari’s (2000) concept of ecosophy, we emphasize the need for an integrated understanding of ecology, i.e. one that considers the interconnectedness of social, mental, and environmental ecologies. Following Stibbe (2021b) and others working in ecolinguistics, we present the main tenets of our ecosophy (E) as a list of values: (E1) Protection (of the planet and its inhabitants over profits); (E2) Recognizing the rights of all (to health, to safety, to wellbeing, to a future); (E3) Systems thinking (i.e. recognizing a network of causality, rather than placing full blame or onus on individual actors); and (E4) Fairness (towards all actors in the network). These values do not preclude the conventional ecosophy tenets (e.g. wellbeing, social justice, care, and compassion for others); rather, we have chosen to foreground these priorities to meet the needs of our research, which is especially social, political, and industrial in focus. In order to achieve ecological wellbeing for our planet as a whole, we must first, in this instance, envisage ways of addressing the material problems of logistics and infrastructure while reimagining the rights and responsibilities of each party involved.

It should be clarified here, for those unfamiliar with ecosophies, that while they are expressions of one’s values, they are also based in evidence and are subject to change as new evidence comes to light (Stibbe 2021b). As briefly described at the outset of this article, there is currently significant evidence to suggest that the production, consumption, and disposal of plastics is a major source of harm to our planet and its inhabitants (Rochman et al. 2013), whether in the form of carbon emissions from petrochemical industries (Zheng and Suh 2019) and the social injustices this pollution creates (UNEP 2021); plastic’s implications for the climate (Shen et al. 2020; Stoett and Vince 2019); plastic waste overwhelming communities around the world (Letcher 2020), constituting, some have argued, colonialism (Liboiron 2021); or harmful chemicals and microplastics entering ecosystems (Rochman et al. 2016; Verma et al. 2016), nonhuman bodies (Haave et al. 2021) and our bodies (Ragusa et al. 2021), the consequences of which are still relatively poorly understood (Nava and Leoni 2021; Vethaak and Legler 2021). These harms and hazards are to be weighed against plastic’s benefits, such as its convenience and affordability; its light weight for transport; its vast contributions to food provision, safety, hygiene, and medical care; and its central role as a flexible construction material in everyday artifacts such as technology and clothing (Andrady and Neal 2009; George 2020). With all of these factors in mind, we have designed our ecosophy around the values that we deem to be logically and morally relevant for the wellbeing of our planet as a whole when considering the supply and use of plastics, particularly single-use plastics.

3.2 Data

Given that one of our main objectives on the MHR project is to compare the language of plastics producers and plastics professionals with the language of members of the public, i.e. non-specialists, we set about building a corpus with two parts, each of which can also be considered a corpus in its own right. Corpus 1 is comprised of plastic- and reuse-oriented language of governmental and public-facing bodies, while Corpus 2 features language of members of the public around those same themes. Together, they form the MHR Corpus.

For Corpus 1, we identified two main sources of online, publicly available text reflecting governmental and public-facing plastics-oriented discourse: webpages from UK campaigns on recycling and plastic information (namely ‘Recycle Now’ and ‘Clear on Plastics’); and guidance from the UK government on refuse disposal across different local authority areas (‘Gov.uk’ council webpages). For Recycle Now and Clear on Plastics, the websites were small enough that it was possible to manually identify the sections of the websites deemed relevant to this work. The URLs for these pages were then fed into a freely available corpus-building program, BootCat,^[8] which scraped the text from the webpages and exported it to files that could then be cleaned, tagged, and added to the Corpus 1 collection. BootCat is typically used to build a corpus from scratch, i.e. by specifying key terms of interest (‘seeds’) which it then combines randomly into search queries (‘tuples’) to be used in Google searches to generate the relevant URLs, and this is how we used BootCat to collect the data from the ‘Gov.uk’ webpages (see Appendix A for the list of tuples used).

Another major source of public-facing discourse around plastics, especially packaging, is that of UK supermarket websites. Using BootCat again, we identified and scraped relevant webpages from the websites of eight of the UK’s largest supermarkets, with varying degrees of success: Co-op, Ocado, M&S, Asda, Sainsbury’s, Morrisons, Tesco, and Waitrose. The tuples used to gather this data are listed in Appendix A.

Finally, to hear directly from those who work with plastic packaging in a professional capacity – whether in terms of labeling, sales, design, or campaigning – we disseminated an industry survey posing questions specifically around language and labeling practices for plastic packaging (see Appendix A for the list of questions). We received 28 responses, with respondents ranging from retailer and manufacturer representatives to business owners, packaging designers, and consultants, among others. The responses to the survey, though relatively small in size, were included in Corpus 1 along with the public guidance and the supermarket website texts. Table 1 shows the final composition of Corpus 1.

Next, in order to build up a broad picture of how members of the general British public speak and write about plastics, packaging, and reuse, we gathered linguistic data from a range of sources for inclusion in Corpus 2. Twitter was identified as an abundant source of geolocated statements from members of the public, and using the Twitter Developer Academic Research product track, we downloaded 66,393 tweets via the Twitter API.^[9] The criteria for tweets to be downloaded were: (i) they must contain one or more of the relevant terms (see Appendix B); (ii) they must be in English and be geotagged to the UK, to try and capture British English language patterns; (iii) they must fall within the timeframe January 2016 to January 2021; and (iv) they must not be a retweet, so as to keep the level of corpus noise to a minimum.

We then selected two general-discussion internet forums which have a large UK userbase and identified relevant subforums and threads within them: the social media platform Reddit, and the UK-oriented parenting forum Mumsnet. Within Reddit, we located three suitable subreddits (‘AskUK’, ‘CasualUK’, and ‘BritishProblems’) and then used the Reddit API^[10] to download threads from within those subreddits that met our search term criteria (see Appendix B for these terms), posted between January 2016 and March 2021. For Mumsnet, given that there is no API available and the website is more difficult to scrape, we used a combination of BootCat and Octoparse,^[11] another freely available webpage scraper, to locate and download relevant threads that were posted within the same timeframe as those of Reddit. The BootCat tuples used to determine the Mumsnet threads are given in Appendix B. In all cases of social media and forum post scraping, only publicly available posts and comments were gathered for inclusion in the corpus, and all usernames were pseudonymized.

In addition to the social media and forum data that we downloaded, we also held online focus group discussions with members of the British public to explore views on plastics and packaging and to generate spoken language data to be included in the corpus. We ran five focus groups in total, involving 28 participants: three of the focus groups were recruited using participants from a previous experiment run by our psychologist colleagues on the MHR project, who had consented to being contacted about future experiments; and the other two were recruited using a combination of online calls for participation via environmentally oriented, UK-based Facebook groups (e.g. ‘Zero-Waste London’, ‘Sustainable Lifestyle UK’) and a call via the mailing list of one of our environmental charity project partners, City to Sea. As such, the first three focus groups involved members of the public who do not explicitly identify as being pro-environmental, and the final two involved those who do. The transcripts of the five focus group recordings were added to the social media and forum data to complete Corpus 2. Table 2 shows the final Corpus 2 composition.

We note here that there can be some contention around the inclusion of elicited language data in a corpus. We bear in mind for our analysis that this kind of data (focus groups, interviews, surveys, etc.) is of a slightly different nature to naturally occurring language data, but we are also not laboring under any illusions that an “authentic language” corpus is somehow insulated from bias, sampling issues, and so on. We also highlight here that there is a distinction to be drawn between written and spoken language data; the focus group transcripts are an example of the latter.

Both Corpus 1 and 2 – which, put together, form the MHR Corpus – will be made available to download, open-access, along with the full corpus documentation.^[12]

3.3 Analysis

Both corpora, Corpus 1 and 2, were loaded into AntConc,^[13] a freely available corpus analysis program. Standard initial corpus analyses were carried out for each corpus: the generation of a wordlist (a list of all of the words, or ‘types’, in the corpus, in descending order of frequency); a keywords analysis (a comparison of the wordlists of each of the corpora to show which terms are unusually (in)frequent, or statistically ‘key’); and a clusters analysis, which is similar to a wordlist but it lists the most frequent phrases, or ‘n-grams’, rather than single words. Clusters can be requested with certain parameters, e.g. by asking the program to only return n-grams between two and three words long, for instance, or to retrieve clusters that contain a term of interest.

To run a keywords analysis, it is necessary to specify a reference corpus against which the target corpus (the corpus of interest) can be compared. As one of our research aims is to assess in what ways the language of public-facing bodies (Corpus 1) and members of the public (Corpus 2) differ, our two MHR corpora were each used as a reference corpus for the other. The rationale for this is that they are both built around the same themes (plastic, packaging, reuse, and recycling) and by comparing their wordlists, we are able to establish which words and phrases are statistically more salient than expected, considering that both corpora are already based on the same themes (if we were to take, for example, a general English corpus as a reference corpus, as it would most likely simply bring up the predicted themes of plastic, packaging, reuse, and recycling). For Corpus 2, we decided to also run a second, heuristic keywords analysis against a different reference corpus to see if we could account for genre effects on pronoun use. This second reference corpus was an ad hoc collection of similar texts (social media posts, website comments, focus group transcripts) from previous corpus-building projects, unfortunately not shareable for public use. Appendix C gives more information about the texts in this ad hoc corpus.

Running the above analyses produced simple and typical corpus software output, i.e. lists of terms and frequencies (examples of which are given in Section 4), but this output alone is not sufficient for a full, critical, (eco)linguistic analysis. These raw outputs were therefore submitted to further, manual analyses, including annotation of samples of concordance lines (keywords shown in their contexts), further corpus querying of terms of interest as and when they arose, and finally a critical evaluation of these findings from ecolinguistic and cognitive linguistic perspectives, also taking into account the findings in the literature.

While corpus methods go some way towards reducing researcher bias in linguistic research, and this is one of its strengths – it draws on very large amounts of data which enables more reliable generalizations, and relies on machines to identify salient terms for closer investigation – it is not entirely insulated from bias. Baker (2015), for instance, demonstrated that when several corpus linguists are given the same corpus data and the same research question, there is no guarantee that they will reach the same conclusions in a discourse analysis task. In this respect, the application of our ecosophy was of central relevance to our study and to our overall findings, as illustrated in the following section.

4.1 Corpus outputs

The wordlists for both corpora are given, truncated to the top 50, in Appendix D. To summarize some of the key findings from the basic corpus analyses described in the section above, we can report that:

1. There is a far higher preoccupation with plastic, generally, in Corpus 2 (social media texts and focus groups) than there is in Corpus 1 (public-facing texts and the industry survey); plastic* occurred 16,030 times per million words (pmw) in Corpus 2, compared with 5,597 times pmw in Corpus 1.

2. According to the lists of keywords (shown in Table 3), our is the most statistically salient^[14] term in Corpus 1, while in Corpus 2 it’s I (or plastic, if comparing against the ad hoc reference corpus).

3. The keyword lists reflect the more corporate nature of Corpus 1 (e.g. financial, customers, report, business, colleagues, executive, cash, value, assets), as well as the interpersonal (I, my, you, me, they) and conflicted (but, just, don[’t], think) sentiments of Corpus 2. When compared against the ad hoc reference corpus, which is of a similar (mainly social media) genre to Corpus 2, we can also see confirmation of the plastic- and packaging-oriented nature of Corpus 2, along with some statistically salient pronouns (we, they, them).

4. At a glance, we can see that there are marked lexical differences in the ways retailers (Corpus 1) and members of the British public (Corpus 2) communicate about plastics and reuse.

Results such as these are quantitatively substantiated, and – given the same corpus – reproducible by other researchers. They also provide us with a defensible order in which to prioritize terms of interest for closer investigation; higher frequencies of terms indicate a greater preoccupation with those terms and their associated concepts (Schmid 2007: 119–120). However, meaningful interpretation of corpus data always requires careful and critical analysis through chosen analytical frameworks; in our case, we chose to consider the corpus outputs through the lens of cognitive linguistics and with our predefined ecosophy in mind. Our findings are reported in more detail elsewhere, particularly in relation to metaphor and narrative (forthcoming), and for the purposes of this article we will discuss just a few short examples as a demonstration of a corpus-ecolinguistic analysis.

4.2 Applying the ecosophy: Key pronouns

Looking at Table 3, we see that a few personal pronouns ranked especially high in our keywords analysis results, notably our in Corpus 1, and I in Corpus 2. Here, we take these two terms as starting points for a discussion around the role of an ecosophy in corpus linguistic analysis. For the purposes of this article, which we dedicate to the demonstration of an ecosophy in practice for the development of communications recommendations, we do not go into extensive depth on each result but rather seek to relate these results to our ecosophy.