5.2.1 A Corpus-Based Study of Metaphors for Cancer: The Accidental Discovery of Humorous Metaphors

As part of a project entitled ‘Metaphor in End-of-Life Care’ (MELC), a team based at Lancaster University employed a combination of corpus tools to analyse the use of metaphors for the experiences of cancer and the end of life. The 1.5-million-word corpus consisted of interviews with and online forum posts by members of three relevant stakeholder groups: people with cancer, unpaid carers looking after a person with cancer and healthcare professionals (Semino et al., Reference Semino, Demjén, Hardie, Payne and Rayson2018). The project was inspired, in part, by long-standing controversies surrounding the conventional metaphor of having cancer as a battle (e.g., ‘fighting cancer’) and the suggestion that it was more appropriate to use different metaphors, such as ‘cancer journey’ (Cancer Institute of New South Wales, 2023), or no metaphors at all (Sontag, Reference Sontag1979).

The MELC team found that what they call ‘Violence’ and ‘Journey’ metaphors are the two most frequently used types of metaphors in the data (Semino et al., Reference Semino, Demjén, Hardie, Payne and Rayson2018: 84). They also provided textual evidence of potentially harmful consequences of Violence metaphors, as when a forum contributor who has been told that her cancer has become incurable says, ‘I feel such a failure that I am not winning this battle.’ Here the feeling of failure seems to be generated by the metaphorical scenario in which the patient presents herself as ‘not winning’, in contrast with a ‘literal’ reality in which the disease cannot be cured because of the limitations of available treatments. However, the analysis of 899 instances of Violence metaphors in the online patient section of the MELC corpus also revealed that some contributors to the forum use them in ‘empowering’ ways, as in ‘my consultants recognised that I was a born fighter’. Indeed, Semino and colleagues (Reference Semino, Demjén and Hampe2017) show that both Violence and Journey metaphors can be used in empowering and disempowering ways. They therefore argue that the implications of specific uses of metaphor in terms of (dis)empowerment are more important, particularly with respect to patients’ well-being, than the type of metaphor involved (e.g., Violence versus Journey).

The team also noticed early on the humorous overtones that some of the metaphors carried, alongside presenting the person with cancer in a highly agentive, empowered position in relation to the disease. In the following examples, the humorous element arises from a combination of the metaphorical use of ‘kicking’ in relation to the cancer and the use of the informal expression ‘butt’ and the niche term ‘wahoola’.

The presence of humour in discussions of a serious illness such as cancer may be unexpected but should not be entirely surprising. Humour is not just a central part of how, as human beings, we perceive the world and communicate with one another. It has also long been associated with three main functions (Attardo, Reference Attardo1994) that are relevant to coping with adversity: disparaging someone/something else (in our case, a serious illness, the health system, etc.); releasing negative emotions (in our case, the stress and anxiety associated with being ill); and creating intimacy and social bonds with others (in our case, people on an online forum who share the same illness).

Focussing on humour was not initially part of the team’s plan for analysis. Having noticed this phenomenon while familiarising themselves with the data, however, the researchers explored the methodological issue of how to pursue the investigation of humour more systematically in the data.

5.2.2 Humorous Metaphors: From Serendipity to Corpus-Based Analyses

Humour does not have predictable linguistic realisations and is therefore not a straightforward phenomenon to investigate via corpus methods (but see Macqueen et al., Reference Macqueen, Collins, Brookes, Demjén, Semino and Slade2024). However, by following up the instances of humour that they had come across serendipitously, the MELC team identified a particular thread on the cancer online forum which was entirely dedicated to humour.

The thread is entitled ‘For those with a warped sense of humour WARNING – no punches pulled here’ (henceforth ‘Warped’). Due to the sampling approach that had been taken to create the MELC datasets, only a section of Warped was included in the MELC sub-corpus of online forum posts by patients. However, to carry out a thorough analysis of cancer-related humour, the team returned to the original data source to consider the complete Warped thread. At the time of data collection in 2012, it contained 2,544 posts contributed by 68 unique participants over 13 months, for a total of 530,055 words. This made it one of the largest of the 30,000 threads on the online forum of the UK-based cancer charity from which the team collected their data. Contributors to Warped take a humorous approach to a variety of topics, such as the side effects of treatment, irritations in the workplace, needing to buy new clothes and so on.

To identify the distinctive linguistic characteristics of Warped among the other threads on the forum, a keyness analysis was carried out (Semino et al., Reference Semino, Demjén, Hardie, Payne and Rayson2018: 240ff.; see Chapter 1 for a brief introduction). This involved using Warped as the target corpus and a 1-million-word corpus of contributions to the forum that exclude Warped as a reference corpus. A log likelihood cut-off of 10.83 was adopted (indicating that if there was no difference between the types of language use in the two sub-corpora, there is only a 0.01 per cent chance that the item in question would be key). Semino and co-authors show how several of the key parts of speech and key semantic domains revealed by the comparison contain humorous uses of language, including metaphorical ones. Here we will focus on a particular pattern that emerged from the combination of two key items: a key part of speech and a key semantic domain.

The key part of speech category is ‘Preceding noun of title’ (with the tag NNB), which includes words such as Mr, Mrs, and Prof. In Warped, the terms are used in light-hearted personifications of cancer (e.g., Mr Crab) and in the humorous use of military titles as nicknames that some of the contributors use for one another, such as Flight Lieutenant Tom and Captain Joe.

The key semantic domain is ‘Warfare, defence and the army; weapons’ (with the tag G3). This includes words such as war, armed and bombing. Its presence among the key semantic domains in Warped was remarkable because, as we have mentioned, Violence metaphors, which frequently involve war-related vocabulary, are one of the two most-frequent types of metaphors in the whole corpus and the most frequent in the sub-corpus of online contributions by patients. Given that this vocabulary can be expected in threads across the forum, why should these terms be used significantly more often in Warped?

An initial examination of concordance lines showed that, in most cases, lexical items included under the ‘Warfare, defence and the army; weapons’ semantic domain were used metaphorically, and humorously, in relation to joint activities that forum contributors engage in:

However, a more systematic analysis of concordance lines for both key items proved problematic, as it was often difficult to understand what was being talked about. Consider the following extract about one of the contributor’s hospital test results (and note that all names/usernames in the quotes included in this chapter are replacements of the original names/usernames, following Semino and co-authors’ (Reference Semino, Demjén, Hardie, Payne and Rayson2018) approach to anonymisation):

What are the missions that the fighting force in the rescue team has successfully carried out? And what is the bin that the writers and others are supposed to travel in? Clearly the contributors shared some common ground that the researchers did not have, and this required going beyond concordance lines and reading substantial chunks from the Warped thread to understand what contributors are doing when they use the kind of vocabulary that accounted for the keyness of the ‘Warfare, defence and the army; weapons’ semantic domain.

5.2.3 Co-created Metaphorical Humour on a Forum Thread

By reading contributions to Warped sequentially, the team observed the origin, development and eventual demise of (metaphorical) in-jokes. With this additional context, references to the fighting force, as in the previously quoted extract, began to make sense. The co-creation of humour over extended stretches of interaction has been previously studied in informal oral interactions (Dynel, Reference Dynel2009) but has not yet received much attention in online discussions of illness experiences. The size of the Warped dataset made such an analysis, of this scale, possible.

Approximately a month after the inception of Warped in July 2011, some contributors started describing each other humorously as army officers on the same camp, giving one another titles such as commandant, brigadier, and colonel. These descriptions are initially used in relation to different imaginary enterprises that contributors would engage in if they met in person. In this context, perceived success is jokingly rewarded by promotion up the army ranks:

As the thread develops, the idea of Warped contributors as members of an army is used as part of an instantiation of the conventional metaphor whereby being ill with cancer is a fight against the disease:

In early 2012, however, a particular combination of posts triggered a new development of the (conventional) metaphor of people with cancer as fighters. A contributor referred to by Semino and colleagues (Reference Semino, Demjén, Hardie, Payne and Rayson2018) as ‘Pretzel’ tells the story of her attempt to dispose of an old fire alarm that was emitting a continuous sound by placing it in a wheelie bin due for collection on that day. However, the bin collectors did not turn up, leaving Pretzel with ongoing embarrassment at the sound emanating from the bin. This inspires several contributors to imagine the reaction of Pretzel’s neighbours, including a scenario in which someone fears that the sound comes from an explosive device and calls the bomb squad. While this scenario is being developed, another contributor (Suzysue) metaphorically describes a longer-than-expected hospital stay as being still imprisoned by warders. This inspires another contributor to combine the imaginary wheelie bin scenario with the metaphor of hospitalised patients as prisoners and to propose that the group should use Pretzel’s bin with the beeping smoke alarm as a distraction to get Suzysue evacuated and liberated.

From this point on, Warped contributors take turns to produce increasingly bizarre versions of metaphorical scenarios in which they are an army conducting rescue missions to liberate cancer patients from hospitals by using, among other things, Pretzel’s wheelie bin. This leads to the kind of humorous metaphorical narrative that had initially flummoxed the MELC team when encountered in concordance lines:

Semino and colleagues (Reference Semino, Demjén, Hardie, Payne and Rayson2018) discuss the characteristics and functions of these chains of descriptions of metaphorical rescue missions. From the perspective of metaphor use, they involve the collaborative creative extension of the conventional metaphor of having cancer as a battle and cancer patients as fighters. In the rescue mission scenarios, the patients fight the health system rather than the illness and do so via rather unorthodox means. From a humour perspective, contributors co-create and co-develop an in-joke that makes light of a stressful situation (hospital stays) and that strengthens the bonds of solidarity and complicity between them. This can make a difficult situation easier to bear. Moreover, as Semino and colleagues (Reference Semino, Demjén, Hardie, Payne and Rayson2018: 255) put it:

Two further studies of metaphorical in-jokes on Warped have employed Dynamic Systems Theory (Larsen-Freeman and Cameron, Reference Larsen-Freeman and Cameron2008) to explain the way in which these humorous metaphors emerge, are collaboratively developed and eventually decline (Semino and Demjén, Reference Semino, Demjén and Hampe2017; Demjén, Reference Demjén2018). From this perspective, the interactions on the Warped thread ‘can be seen as a self-organizing system which evolves over multiple time-scales as a result of complex interactions between internal and external forces’ (Semino and Demjén, Reference Semino, Demjén and Hampe2017: 186). Humorous metaphors such as those we have discussed are seen as emerging from the dynamic interaction between external factors, such as the conventional metaphors of having cancer as a battle, and internal factors, such as the nature of the Warped thread and previous posts by contributors (Gibbs and Cameron, Reference Gibbs and Cameron2008). The development of humorous metaphors over time involves patterns of both stability and variation. For example, Pretzel’s story about the wheelie bin causes a perturbation in the system that results in a new metaphorical scenario, where rescue missions from hospitals involve the beeping wheelie bin. This scenario then becomes established as a pattern of stability in the system, until it eventually stops being used around July 2012 (six months after the first reference to the wheelie bin), as the community focuses on other patterns of shared humour.

5.3.1 Response Structure in Online Forums

While online forums function to host a potentially endless variety of conversation topics, there is a reasonably consistent structure across platforms that reflects their asynchronous, multiparty affordances (Morzy, Reference Morzy, Özyer, Rokne, Wagner and Reuser2013). Most online forums involve some degree of threading, visually indicating continuity between posts in the way that demonstrates their topical contiguity, even when a message is posted sometime after the post to which it responds. In light of this asynchronicity, respondents can use a range of features to maintain coherence when their messages are often not the next contribution to the discussion, chronologically speaking, such as lexical repetition and lexical substitution (Meredith, Reference Meredith2019: 247), or even explicitly targeting the recipient of their post through @-tagging.

In its naturally occurring form, the thread of forum posts is typically represented visually on the forum interface, providing a kind of flow diagram of the conversation. When these online interactions are encoded in plain text for the purposes of (corpus) analysis, these thread relations can be documented as metadata. HealthUnlocked provided the forum data to the researchers as a spreadsheet with metadata in columns alongside the message content. Table 5.1 shows a simplified version of the spreadsheet information and demonstrates how the structure of discussion threads across the forum is captured according to thread and message identifiers (Thread ID, Parent ID, Post ID). The examples shown in Table 5.1 represent the complete set of contributions to the discussion thread 1241, with each post given a unique numerical identifier (Post ID) and a timestamp indicating when it was posted.

Table 5.1Forum post metadata for discussion thread 1241

Each post is also given a Parent ID, which corresponds with the Post ID of the message ‘under’ which the contribution was posted as a response; for example, the second row in Table 5.1 was posted as a direct reply to the Opening post (Post ID: 1241) and the third contribution was posted as a reply to the second message. By referring to the User ID, we can see that the discussion involved three different contributors and that it was the original poster (#001) who returned to reply, first to participant #002 and then user #003.

While Table 5.1 presents the contributions chronologically, showing that the entire discussion thread took place between 10:22 a.m. on 8 November 2016 and 6:37 p.m. on 9 November 2016, Figure 5.1 demonstrates the response structure between the three contributors to the discussion thread. From this, we can see that user #003 responded not to the Opening post (Post ID: 1241) but rather to post 3394, which came from the initial exchange between users #001 and #002. While there is no direct interaction between users #002 and #003, it is important to remember that all posts are visible to those who visit the discussion thread (and may not post).

Figure 5.1

Response structure for discussion thread 1241.

5.3.2 Coding Discourse Units

With an understanding that online forums are characterised as providing information and emotional support (Yip, Reference Yip2020), Collins and Baker referred to the framework developed by Biber and colleagues (Reference Biber, Egbert, Keller and Wizner2021) to capture the functional units of interactions. In other words, we can refer to ‘discourse units’ (DUs) to document how participants in the forum achieve communicative goals such as ‘giving advice and instructions’ or ‘describing or explaining the past’ in this digital, multi-party asynchronous mode.

The coding of discourse units is also described in Egbert and colleagues (Reference Egbert, Wizner, Keller, Biber, McEnery and Baker2021: 725), wherein the authors provide their operational definition for a discourse unit as follows:

1. 1. Coherent for its overarching communicative goal, which is both the primary objective of the DU and the task that the interlocutors are doing with language in the DU

2. 2. Characterised by one or more communicative purposes, where a communicative purpose is a finite set of actions that serve to help accomplish the communicative goal of a DU

3. 3. Recognisably self-contained: a DU has an identifiable beginning and end

Egbert and colleagues (Reference Egbert, Wizner, Keller, Biber, McEnery and Baker2021) explain that the range of possibilities for communicative goals (which might be complaining about annoying co-workers or making plans for buying Christmas presents) is boundless and so not coded. However, they developed a taxonomy for nine types of communicative purpose, which are summarised in Table 5.2. In their application of this taxonomy to examine informal spoken conversations, the authors also found that imposing a length requirement optimised the coding between raters and so established the additional criterion:

1. 4. A discourse unit has a minimum of five utterances or 100 words.

Similarly motivated by reaching optimal agreement between coders, Collins and Baker (Reference Collins and Baker2023) forewent a length requirement when coding discourse units in the Anxiety Support forum data, finding that the communicative purpose ‘giving advice and instructions’, for example, could be observed in a short forum contribution such as ‘You should go see your GP’.

Table 5.2Biber et al.’s (2021) taxonomy for nine communicative purposes

The nine communicative purposes, as described in Biber and colleagues (Reference Biber, Egbert, Keller and Wizner2021: 25), are presented in Table 5.2.

Through their application of the coding framework, Biber and colleagues (Reference Biber, Egbert, Keller and Wizner2021) recognised that an individual DU can serve multiple communicative purposes and so implemented the tags according to a 0–3 quantitative scale in which 0 = purpose not present (not recorded), 1 = minor purpose, 2 = major purpose and 3 = dominant purpose. To demonstrate this, we have produced the full message of the Opening post of discussion thread 1241 in Table 5.3, showing how the post can be divided into two DUs and how each DU was coded according to its dominant and additional functions. This coding reflects our reading of the post as beginning with situation-dependent commentary, through references to this page and the forum more generally, before the contributor expresses the difficult feelings they are experiencing (the communicative purpose fel) while also offering a description of their experiences of anxiety (the communicative purpose des).

Table 5.3Discourse unit coding for an Opening forum post (Post ID: 1241)

Table 5.3 Long description

The table consists of two columns: message content and communicative purpose. The communicative purpose column is subdivided into 3, 2, and 1. The message content is divided into two rows: Row 1. Message content: Title: Relentless Hi folks, im new to this page, i just needed somewhere to go to get my anxiety down (spelled as d-w-n) into words; communicative purpose 3: s d c; communicative purpose 2: blank; communicative purpose 1: blank. Row 2: Message content: i feel im at my wits end for months now I have (spelled as i-v) lived with my anxiety from the minute i open my eyes until i fall asleep then im anxious in my dreams its never ending, i constantly feel sick my heart beats so hard i can hear it, I have (spelled as i-v) no patience im extremely emotional i get so wound up i become breathless, i constantly think bad is going to happen, how do i cope with this, its taking (spelled as t-a-k-i-n) over my life and im at a loss as to what (spelled as w-o-t) to do; communicative purpose 3: f e l; communicative purpose 2: d e s; communicative purpose 1: blank.

Collins and Baker (Reference Collins and Baker2023) report the occurrence of each communicative purpose across a randomised sample of one discussion thread from each calendar month of the data, which amounted to 103 discussion threads, comprising 822 individual posts. They found that the most common communicative purpose observed was the expression of opinions and personal perspectives (fel), followed by explanations (des) and advice-giving (adv), with minimal occurrences of conflict communication (con) and joking (jok) in the forum interactions.

5.3.3 Sequences of Discourse Units

The coding of discourse units enabled Collins and Baker (Reference Collins and Baker2023) to consider not only how communicative purposes were combined in one discourse unit but also how discourse units were combined within a single forum post. For example, the message in Table 5.3 carries both the communicative purpose of discussing something in the shared situational context of the visitors to the forum (i.e., the forum itself) and conveying their personal perspective. As was the case with Biber and colleagues (Reference Biber, Egbert, Keller and Wizner2021), Collins and Baker found that it was common for messages to serve multiple communicative purposes; for example, the pas code was used alongside explanations or advice-giving to provide a context drawn from lived experience. The researchers also discuss resultant communicative purposes in replies according to Opening posts that are characterised according to a particular communicative purpose (i.e., discussion threads initiated by Sharing feelings and opinions (fel)). However, they report that even when the communicative purpose of an Opening post appears to invite advice or explanations (such as an Opening post characterised in terms of ‘figuring things out’), respondents seemed to prefer to present their recommendations indirectly, as personal perspectives (i.e., fel).

Here, we discuss the contributions to discussion thread 1241 to consider the extent to which the communicative purpose of the posts is informed by the purposes and structure of preceding messages. Ultimately, the aim of this application of the framework for coding discourse units is to focus on the sequencing of discussion threads to consider the interactional dynamics of how different communicative purposes are realised in the forum and how members create a dialogue that pursues particular communicative goals.

Table 5.4 demonstrates the communicative purpose codes applied to the discourse units we identified in comments appearing as part of discussion thread 1241. The shaded rows indicate the messages posted by the author of the Opening post. From this, we can see that there are 22 DUs, that 6 of the 13 posts included more than 1 DU and that the number of DUs in a post ranged from 1 to 4. Five of the nine communicative purposes are represented in this discussion and 13 (59.09 per cent) of the DUs were coded as having more than one communicative purpose.

Table 5.4Coding for communicative purposes of discourse units in the thread 1241

If we refer back to the Opening post, as shown in full in Table 5.3, there is an apparent prompt for advice in the contributor’s question, ‘How do I cope with this?’, and they assert that they are ‘at a loss as to wot to do’. Collins and Baker (Reference Collins and Baker2023) report that many participants seem to post to the forum when they have exhausted their options or coping strategies and seek recommendations from other members. It is curious, then, to see that in response to this Opening post, User #002 does offer a contribution that functions as advice (which is for User #001 to check out User #002’s profile page for a list of recommended responses that have helped them), but here User #002’s dominant communicative purpose is to express their own personal perspective (fel), including the empathic statement I feel your pain. Thus, User #002 has foregrounded this display of empathy and emotional support over giving advice, which was directly requested.

User #001 and User #002 then continue to trade messages (Post IDs 3394, 3415) that are characterised by the expression of feelings (fel) about the difficulties of their respective anxiety experiences, which is combined with explanations (des) as to how the effects of anxiety manifest in their day-to-day activities and how they try to cope. User #003 also demonstrates this reciprocal sharing of personal perspectives, picking up on the description from User #001 in post 3394 and going on to explain their own ways of coping in posts 7904, 8404 and 8548. We are reminded of the significance of finding a community of individuals who can appreciate such struggles in an exchange where User #002 states, ‘No one understands what you’re going through unless they have been through it too’ (Post ID: 3289). Other users respond by reassuring ‘it’s all in your head’ (Post ID: 3415) and User #001 similarly asserts that ‘its so hard to get ppl to understand who havnt experienced it’ (Post ID: 8562).

While there appears to be an impetus to demonstrate mutuality, as shown in User #003’s contribution, ‘I do that too!’, there is also the recognition that anxiety experiences and coping strategies may differ for each individual: ‘If you have something that works I say take it if you can!’ Nevertheless, having established mutuality, the conversation between User #001 and User #002 appears to shift towards advice-giving (adv) and thinking ahead to the next steps (fut), as observed in the progression of post 3415 and post 3466. Thus, it would appear that both User #002 and User #003 work to establish reciprocity and a shared understanding of the difficult feelings associated with anxiety before proffering advice and looking ahead to what User #001, in particular, will do next. Establishing this understanding arguably puts them in a more legitimate position to offer advice and equally, for their recommendations to be taken on board by the original poster.

In this brief examination of a short sequence of posts within a discussion thread, we have shown how the coding of communicative purposes has helped document how social actions such as establishing mutuality, showing empathy, offering advice and planning for the future are achieved over the course of an interaction. With the appropriate annotation for these distinct code categories in place, analysts can use corpus tools to investigate the frequency, distribution, combination and sequence of such communicative purposes to investigate patterns in the dynamics of how users attend both to those communicative purposes elicited by other members, such as a request for advice, and what they perceive to be the communicative functions of platforms such as online support forums.