Ambiguity, responsibility and political action in the UK daily COVID-19 briefings

ABSTRACT This paper investigates how pronouns were used by UK government speakers to allocate responsibility to themselves and others in all 92 daily televised COVID-19 briefings that were held between March and June 2020. We identified the referent for every use of the first-person plural pronoun (1PL) as ‘inclusive’, ‘exclusive’, or 'ambiguous' and analysed the transitivity patterns in which these pronouns act as Participants. We argue that the UK government uses the inherent ambiguity of this pronoun to strategically mitigate their own responsibility for controlling the spread of the virus, while increasing the amount of responsibility to the general public. In doing so, we propose a transparent and replicable systematic method for identifying the referents of pronouns, which may be useful to other discourse analysts faced with the challenging task of pronoun resolution.


Introduction
On 9 March 2020 UK Prime Minister Boris Johnson gave his first televised address to the nation about COVID-19.This was followed shortly after by another on 12 March.Then, starting from 16 March, the government committed to giving the public daily updates through televised briefings, and a briefing was broadcast from Downing Street almost every day between 16 March 2020 and 23 June 2020.However, as the crisis unfolded in the UK, there was criticism of both inaction and confused messaging levelled at the government, to the extent that the briefings were described as a 'lesson in how not to do government communications' (Oliver, 2020, p. 1).
This paper examines the ways in which responsibility for political and social action in response to the virus is represented by the government in the daily briefings.In doing so, we focus on a feature of political discourse that is well-established as an effective rhetorical devicethe first-person plural (1PL) pronoun we.Pronoun resolutionthe identification of the intended referent for a pronounis a challenging task for discourse analysts.Indeed, the vagueness and slipperiness of we are fundamental to its value to political speakers.Therefore, in this paper, we outline a proposed systematic process for resolving the intended referent of we, at least insofar as inclusive and exclusive interpretations are concerned.Once this process has been described, we combine pronominal analysis with transitivity analysis as a means by which to observe how responsibility is encoded by government speakers in this context.
This study represents the first systematic discourse analysis of ambiguity and responsibility in the daily televised briefings.Adding to existing work, this study contributes to the growing body of discourse-analytic research examining the rhetorical and communicative strategies employed by the British government during a public health emergency.Therefore, it expands the base of international knowledge on the political discourse of nations around the world as the pandemic developed and continues to develop.

COVID-19 and political discourse
The COVID-19 pandemic sparked different political responses across the governments of the world, including divergent strategies for disseminating and communicating information to their citizens (e.g.Sjölander-Lindqvist et al., 2020).Research has revealed the ways in which political leaders such as former German Chancellor Angela Merkel (Jaworska, 2021) and South African president Cyril Ramaphosa (Hunt, 2021) invoked discourses of unity, togetherness and collective action as means by which to build trust and to persuade citizens to comply with COVID-19-related rules and restrictions.In contrast, the COVID-related speeches of former US President Donald Trump (Olimat, 2020) and former Australian Prime Minister Scott Morrison (Alyeksyeyeva et al., 2021) have been found to be characterised by dysphemistic metaphor, framing the virus as an invisible, foreign enemy and a dangerous threat in order to legitimise new policies and restrictions.Indeed, a large-scale comparative analysis of the first COVID-19-related statements by political leaders in 29 countries across four continents found that the construction of the virus as an 'outgroup', pandemic-as-war and pandemic-as-movement metaphors, and notions of national solidarity were rhetorical strategies drawn on by governments worldwide (Berrocal et al., 2021).
In the UK, the government's approach to public health communication and messaging during the first wave of the pandemic in 2020 was roundly criticised.This criticism is reflected by Cowper (2020, p. 1): At a time of a global respiratory pandemic for which no effective vaccine yet exists, the bare minimum that a nation, and its health system, needs from its government is two things: competence and trustworthiness.It feels surreal to be writing these words on the influences of communications and politics on the UK's response to covid-19, but the need to do so is a signal of the poverty of our government's performance that has shown so starkly the gap between their ambitious rhetoric and the appalling reality.
Several studies have collected and analysed the daily televised news briefings as part of larger datasets, combined with other COVID-19 related discourse types such as interviews, articles, statements, Parliamentary statements, press releases, websites and social media posts.Billig (2021) focuses on how the British government uses statisticsprecise, round and 'semi-magical' numbersabout the virus for rhetorical advantage and to achieve political ends.Jarvis (2021) examines the construction of temporality and narratives of time within the government's discourse, and how temporal claims situate the virus historically, project its likely development and help the government ⍰⍰communicate, justify and defend' their evolving response to the virus (Jarvis, 2021, p. 15).Meanwhile, Sanders (2020) assesses the government's messaging in terms of the communicative characteristics of so-called 'high reliability organisations' (e.g.nuclear power stations and airline companies).Using Braun and Clarke's (2006) thematic content analysis, Sanders (2020, p. 373) tentatively concludes that the Prime Minister's communication was initially complacent with regard to the threat posed by the virus and was later characterised by claims of 'following the science', with no acknowledgement or apology for any errors or failings.Finally, Andreouli and Brice (2022) also employ content analysis and found that there were five 'principal constructions of the 'good citizen'' in government communications during the pandemic: the confined, the heroic, the sacrificial, the unfree and the responsible.These constructions, they argue, are reflections of three ideological dilemmas faced by the government when developing and communicating their response to the pandemic: freedom vs control, individualism vs collectivism and active vs passive citizenship (Andreouli & Brice, 2022, p. 9).

Pronouns, responsibility and politics
The value of personal pronouns to political speakers resides in their inherent vagueness and ambiguity.This was identified by Wilson (1990, p. 76) over 30 years ago: With such manipulative possibilities provided by the pronominal system as it operates in context, it is not surprising to find that politicians make use of pronouns to good effect: to indicate, accept, deny or distance themselves from responsibility for political action; to reveal ideological bias; to encourage solidarity; to designate and identify those who are supporters (with us) as well as those who are enemies (against us); and to present specific idiosyncratic aspects of the individual politician's own personality.
The strategic purpose of personal pronouns that the current study is primarily concerned with is how they construct responsibility.Bull and Fetzer (2006) and Fetzer and Bull (2008) demonstrate the ways in which British politicians use personal pronouns strategically in televised interviews to avoid taking responsibility for their actions.They argue that political speakers use pronouns as a form of equivocation, by shifting from I to collective we, where the referent(s) of we could be understood as their political party that they represent, the government, the nation as a whole or any one of these three (Bull & Fetzer, 2006, p. 33).Such 'over-inclusion', they argue, extends the responsibility for action away from the speakers as individuals to the domain of party politics or nation more broadly, thus backgrounding their personal role and deflecting blame and culpability (Fetzer & Bull, 2008, p. 286).Similarly, Íñigo-Mora (2017) shows how British (David Cameron) and Spanish (Mariano Rajoy) leaders use we in their speeches to shield, displace and de-focalise their role(s) in political operations, sharing responsibility for these actions with the listener (in Cameron's case) or with the EU as the larger political entity (in Rajoy's case).Strategic pronominal reference is not confined only to political speeches.Teo (2004, pp. 493-494), for example, examines the way in which we is used in environmental campaign slogans in Singapore to 'co-construct the government and people as one unified, undifferentiated body' which 'creates an impression of the government and the people in partnership, sharing the responsibility of protecting and caring for environmental resources'.Mulderrig's (2012) analysis of British government policy documents or 'White Papers' also demonstrates how the ambiguous and flexible nature of person deixis can be exploited by political actors.In this context, seamless transitions between referents of inclusive and exclusive we blur the responsibility for certain (often contestable) propositions and speech acts by 'collapsing the distinction between the government and the people' which 'draws citizens into the very process of governing, thus implicating them in policy decisions' (Mulderrig, 2012, p. 707).Such deliberate and strategic uses of we allow political actors to claim consensus over political actions and as such facilitate their navigation and implementation of potentially controversial policies.
Pronominal use in relation to political talk and responsibility has been exhibited in routine political genres such as speeches and interviews.However, the COVID-19 daily briefings are anything but 'routine', given the unprecedented health crisis in which they were written and delivered.This paper, therefore, analyses a staple device from the politicians' rhetorical toolbox and investigates its use in extraordinary speech events under extraordinary conditionsconditions in which clarity, precision and unambiguous communication are a matter of public health.

Data collection
The corpus for this study comprises all 92 televised daily briefings that took place between 16 March and 23 June 2020.All transcripts were sourced from the official website of the British government. 1This paper only analyses the initial prepared monologue by the governmental representative (and not any subsequent Q&A segment), as this was the point in the speech event where the government had the greatest degree of control over how they portrayed their own representation of the pandemic.Although the briefings were intended to be delivered by the Prime Minister, his ill health in late March 2020 resulted in many different cabinet ministers delivering them; this is a trend that continued even after Boris Johnson returned from illness (Table 1).

Pronominal referent coding
As discussed earlier, one reason that pronouns have received attention in political discourse analysis is because of their frequency and the potential for ambiguity their use can bring.The tendency for pronouns to be common in political discourse was also true of our dataset.The 1PL pronoun we was the fifth most frequent word in the dataset, occurring 2789 times.For reference, although other personal pronouns appeared relatively frequently, they were markedly less frequent than we; I and you occurred 1209 and 800 times respectively.
Given the importance of both the inclusive and exclusive readings of the 1PL pronoun in political discourse, the first step of the analysis involved determining, as far as was possible, whether each instance of the 1PL pronoun carried an exclusive or an inclusive reference.This is no straightforward task, as the referents of deictic expressions depend on an understanding and analysis of the context in which they are used (De Fina, 1995, p. 390;Zupnik, 1994, p. 340).Zupnik (1994, p. 340) notes how early work in this area had 'not directly addressed the complexity involved in resolution of deictic pronouns, in general, and the problems in resolution of vague deixis, in particular', and that 'the process by which the analysts determine such indexing is not discussed explicitly'.This raises important issues of reliability and replicability in the quantification and interpretation of pronouns.While this continues to be the norm, some later work (e.g.Mulderrig, 2012, p. 711) has made more transparent the processes followed in coding the forms and referents of we.For this analysis, we devised a systematic and replicable coding procedure that we followed when making determinations as to whether instances of we were inclusive or exclusive.In this procedure, different sources of evidence were used to guide our classification of each occurrence of the 1PL pronoun as being either exclusive or inclusive: 1. Is there an explicit disambiguation in the Nominal Group to either an exclusive or inclusive reading? 2. If not, the does the semantic context or world knowledge prefer one of the two possible readings (c.f.Hobbs, 1979) 3. Does lexical cohesion/possible anaphoric links in the linguistic environment favour one or the other reading? 4. What was the referent of the last use of we (current referential chain) The first point here refers to explicit disambiguation.For example, the explicit mention of as a government forces the exclusive reading of the pronoun.This is similar to Mulderrig's (2012, p. 711) 'anchor': (1) I just want to give you an update on the steps that we as a government are taking to defeat coronavirus.[MAR30DR 2 ] When no explicit disambiguation was present, the next check involved a subjective judgement about whether the verbal semantics strongly preferred one of the two readings.This mainly restricted itself to verbal processes, such as the following: (2) And, today, our armed forces are again part of that team as we announce two new deployments to the NHS Nightingale facilities in Harrogate and Bristol. [APR22DR] In this example, the inclusive reading would involve the speaker referring to the public as a co-Sayer in this verbal process.As there is no plausible way that the public could be considered an active participant in this process, cases like this were coded as exclusive.
More difficult are cases when there is no strong intuition to unambiguously code the 1PL pronoun as being either exclusive or inclusive based solely on the verbal semantics.For example: (3) As we follow our plan, our testing regime will be our guiding star [MAY21DR] In this case, there is ambiguity around who is following whose plan.After all, it is the government who devises and announces any plans.On the other hand, the plan they announced was for the country as a whole to follow.Therefore, there is an inherent tension here between the inclusive and exclusive readings.In such cases, various sources of pragmatic and discourse information were drawn upon.Specifically, the notion of lexical cohesion was invoked (Halliday & Hasan, 1976).This was because computational and psycholinguistic literature has pointed to the important role of discourse salience in the disambiguation of pronominal reference (Grosz et al., 1983;Kehler & Rohde, 2013).Lexical cohesion as defined in terms of co-occurring instances of referring expressions that associate either with the exclusive or the inclusive reading was assumed in this work to increase the salience of that potential referent.For example, occurrences in the surrounding text of lexis and phrases like nation and community would bias the analysis towards an inclusive coding; mentions of government or ministers, amongst other similar terms, would bias the analysis towards an exclusive reading.
Finally, there were cases where all the above steps still did not allow the authors to confidently decide on the appropriate code.In these cases of ambiguity, it was decided to code the example with the reading that continued the current referential chain in that grammatical position.How recent a particular potential discourse antecedent has been mentioned is argued to influence pronoun disambiguation (Ge et al., 1998), and this final choice allowed us to capture this.
The above steps were guiding principles to help make the reasoning behind our subjective judgements more transparent and reproducible.There may well be other factors that could potentially influence pronominal resolution, and we make no claim that this is the end of the story.
All instances of we in the data (n = 3156) were independently coded as being either exclusive or inclusive by both authors using the method just described.Inter-rater reliability was 85.88% and was deemed to be substantial by Cohen's Kappa (κ = 0.68).There were 477 points of disagreement between the two authors.339 of these were able to be resolved through post-analysis discussion between the authors.This left 138 instances of we unresolved in terms of their interpretation as either inclusive or exclusive.Instead of forcing one of the two readings, the decision was taken to code these 138 instances as being ambiguous, which is similar to Mulderrig's (2012, p. 709) 'ambivalent' we.

Transitivity coding
To analyse how exclusive and inclusive pronominal reference was used by the speakers as part of their representation of the unfolding pandemic and their reaction to it, a transitivity analysis was conducted on all clauses in which 1PL pronouns appeared as participants.
It is important at this stage that we are fully transparent in how this transitivity analysis was approached because, as has been often noted in the systemic functional literature, there is considerable variability and debate about how best to identify the different process types (Gwilliams & Fontaine, 2015;O'Donnell et al., 2008).
Firstly, we followed the major classificational system of the 'Sydney' model of Systemic Functional Grammar (Halliday & Matthiessen, 2014) with its major division of six process types: material, mental, relational, verbal, behavioural or existential.The major methodological decision that arises in a transitivity analysis is the relative importance of different sources of information in determining a clause's classification.As these terms come from a functional grammar, both lexicogrammatical and semantic criteria are potentially relevant in the identification of process types (for instance, see the criteria for process types given in Eggins, 2004;Fontaine, 2013; as well as the detailed descriptions given in Halliday & Matthiessen, 2014).This links to there being distinct 'models' of process type classification, where one of these different sources of information is preferred (O'Donnell et al., 2008).In this work, we prioritised lexicogrammatical diagnostics for transitivity coding over purely semantic ones.The reason for this is that foregrounding the meaning over the grammar runs the risk of confusing two different levels of analysis, and in so doing potentially omits linguistic observation that might be relevant and important in this context.Specifically, it risks missing cases where, for instance, a figure of doing, saying or thinking, is markedly realised by its non-typical corresponding process type (Halliday & Matthiessen, 1999).Transitivity coding was conducted by the first author on all clauses which contained the 1PL pronoun (n = 3156).
Following the resolution of pronominal referents and process types, the analysis then turned to identifying whether there were any qualitative patterns and/or differences in the clausal contexts in which the 1PL pronoun was used.Since the analysis is focused on the notion of responsibility, the patterning of 1PL pronouns appearing in material processes was analysed.Because material processes typically realise clausal level meanings that relate to overt actions that could potentially have an effect upon the world, these clausal meanings are the ones most likely to contain the representation of actions taken in response to the developing pandemic.

Analysis
The frequency and proportions of the three different types of 1PL pronominal referenceexclusive, inclusive and ambiguousare provided in Table 2.
As can be seen, using the 1PL with its exclusive reading was the most frequent use by the government speakers.In terms of transitivity patterns, the exclusive reading occurs in verbal processes more frequently than inclusive readings, and the reverse is true of relational processes.However, perhaps unsurprisingly, material processes account for the overwhelming majority of instances of both inclusive and exclusive 1PL (Table 3).Although material processes are the most common process type across all readings of the 1PL pronoun, this does not mean that the same clausal meanings are being encoded in each case.In the analysis below we outline the major patterns that emerge firstly when comparing the exclusive and inclusive readings, followed by the types of environments in which the ambiguous cases appear.

Inclusive vs exclusive: helping, controlling and 'grammatical distancing'
Examination of the specific types of material processes in which the exclusive and inclusive 1PL pronouns appear reveals a distinction in the actions and situations that are represented.To give a sense of this, Table 4 lists the 10 most common verbs (plus ties) appearing in these two readings.
Starting with the exclusive readings, the theme of government as provider emerges, as they provide, support and help: (4) We E must be confident that we E are able to provide sufficient critical care and specialist treatment right across the UK [APR16DR] 3   (5) We E must support millions of businesses and tens of millions of families and individuals over the coming months [MAY17BJ] (6) By making use of the people and the resources of the Department of Transport, its agencies and arm's lengths bodies, we E will be helping NHS trusts and local resilience groups to fight the pandemic [APR24GS] In contrast, we see a different theme emerging with inclusive readings; instead of portraying the wider public as co-Actors in processes of 'managing actions' (Mulderrig, 2012, p. 713)  This inclusive meaning appears to be a rhetorical attempt to discursively cultivate national unity, togetherness and solidarity that research on COVID-19-related speeches in other countries has identified (Hunt, 2021;Jaworska, 2021).In ( 9) and ( 10) this combines with an attempt at compelling and persuading citizens to adhere to the strict rules being enforced.
It is important to note that there is nothing necessarily inaccurate in the government's representation of the public's role here.After all, actions taken by the public, including social distancing and mask wearing, have been promoted by various national and international health bodies (e.g.NHS, CDC and WHO) to be crucial in lowering rates of transmission.Nor is it the case that these same verbs do not appear in sentential contexts with the exclusive reading.But when they do, they consistently appear in different constructions that have consequences for how responsibility is being linguistically coded.To illustrate this point, consider (11): (11) As a Government, we E continue to take steps necessary to slow the spread of the virus [APR22DR] On the surface, this example seems very similar to the material processes that predominate clauses where the inclusive pronoun acts as Actor.However, a closer examination of the transitivity structure shows an important difference (Table 5). 4  Instead of the exclusive pronoun maintaining its position as Actor in the embedded clause, this position is filled by the elided steps.It is this embedded clause that contains the linguistic representation of the direct action to counter the spread of COVID-19.This contrasts with the examples seen with the inclusive pronoun, where the inclusive pronoun was encoded as a direct Actor in the relevant processes.
A similar effect is also created through a separate but closely related structure, where the exclusive pronoun does not appear in the main transitivity pattern which represents the counter-measures to COVID-19, but instead appears embedded within a defining relative clause modifying lexemes such as steps or decisions.Again, it is these steps or decisions Table 5. Transitivity analysis of As a Government, we E continue to take steps necessary to slow the spread of the virus.that are linguistically codified as the direct Actor of pertinent material processes.Below is an example of such a construction, taken from Alok Sharma's briefing on 17 April 2020 (Table 6).These two extracts are not isolated examples; these constructions jointly form a consistent pattern throughout the government briefings in clauses which contain the exclusive pronoun, further exemplified in ( 12)-( 15 We propose that such constructions have a direct consequence for the linguistic representation of responsibility during the pandemic.On one hand, we have the inclusive pronoun directly appearing in the Actor position in relevant material processes such as saving lives, protecting the NHS and controlling the virus.On the other hand, when the exclusive pronoun appears in clauses with similar meanings, the most frequent pattern is that they are what we term 'grammatically distanced' from the syntactic position that typically encodes meanings of direct agency.The claim here is not that in such examples the government is encoding themselves as having no responsibility for the processes included in clauses like those above.Instead, we argue that there is an attempt at linguistically encoding reduced responsibility, at least when compared to similar clauses containing the inclusive pronoun.In so doing, the government gives the impression of taking decisive political action by taking up a subject position, but not the subject position of the verb encoding the most direct actions against the virus.Comparing the exclusive and inclusive uses of the 1PL pronoun in material processes in the briefings allows a paradigmatic pattern to emerge.This pattern is shown in ( 16)-( 18).These examples are adapted from similar cases discussed above, but modified for the purpose of clear exemplification: (16) We're protecting the NHS (17) We're taking steps to protect the NHS (18) The steps that we are taking to protect the NHS Functionally, there are different roles for the 1PL pronoun here.In ( 16), it is acting as the direct actor in the material process of protect.In (17), the pronoun is acting as an Actor but is represented as enacting measures which then in turn aim to protect the NHS.In (18), the pronoun functions as a Qualifier within the nominal group to characterise the head noun steps (Halliday & Matthiessen, 2014, p. 382).Given standard ideas about agency and their links to clausal representation, this then creates what we term a 'hierarchy of responsibility'.In ( 16) the pronoun is represented as having the highest degree of responsibility for the action of protecting, while (18) contains the least.

Ambiguous cases: pronouns and slogans
This section turns to a discussion of patterns that emerge from examining the clausal environments of usages of the 1PL pronoun which were deemed to be ambiguous according to the resolution method described above.To recap, there were 138 instances of ambiguous uses of the 1PL pronoun.Ninety of these were when the pronoun is being used as a Participant in material processes.
One notable trend was for ambiguity to arise in clauses which contained language that was also present in some of the set phrases and slogans that the UK government used in their messaging to the public both within these briefings, but also in wider public service announcements in adverts and informational posters.In the UK, one prominent slogan during this time was 'Stay at home.Protect the NHS.Save Lives'.A common clausal environment in which the ambiguous cases arise is those instances which discuss the notions of protecting and saving the NHS and people's lives, in symmetry to the sloganeering campaign which was underway at the time.The ambiguity around these particular meanings arises both intertextually (across different briefings), and intratextually (within the same briefing).Intertextual ambiguity arises because there are instances of these meanings being used with a we that is biased towards its exclusive reading through lexical choices in the immediately preceding context.For example, consider the following: (19) At each point we E have been following scientific and medical advice and we E have been deliberate in our actions, taking the right steps at the right time.We E are also taking unprecedented action to increase NHS capacity by dramatically expanding the number of beds, key staff, and life-saving equipment on the front-line to give people the care they need when they need it most.This is why we E are instructing people to stay at home, so we E can protect our NHS and save lives.[APR17AS] In the above, much of the context preceding the slogan contains representation of what the government itself is doing, and as such the pronouns all have an exclusive reading.For instance, there are unambiguous references to the government following scientific and medical advice, taking unprecedented action and instructing people to stay at home.This textual information then makes salient the exclusive reading in the final two clauses of protecting and saving.
In different briefings, however, we see similar slogan-like constructions, but used in contexts which raise the salience of the inclusive reading: (20) And I use 'we' very deliberately.Because we I all have our part to play.This is a national effort and we I all have a role.[…] If we I can thwart that purpose, we I can control the virus and ultimately defeat it.We I must all follow the NHS Test and Trace instructions, because this is how we I control the virus and protect the NHS and save lives.[MAY27MH] In this case, there is a direct recognition by the then-Health Secretary of the intended referent of the pronoun; it is to be interpreted as inclusive.This inclusive-biasing of the pronoun then continues through instructions to the country as a whole.At the end of this extract, the phrase protect the NHS and save lives again appears, but this time the reference is clearly inclusive in nature.
This variation in reference then leads to intratextual ambiguity in cases when the same slogan can again be seen, but the preceding context does not strongly make salient one reading or the other: (21) At every point, we E have followed expert advice to be controlled on our actions, taking the right steps at the right time.We E are taking unprecedented action to increase NHS capacity by increasing the numbers of beds, key staff, and life saving equipment on the front line to give people the care they need.That is why it is absolutely critical people follow our instructions to stay at home, so we A can protect our NHS and save lives.[MAR26RS] In this example, conflicting pieces of intratextual information can be seen.In the preceding context, there are two instances of we in subject position with an exclusive reading.This would increase the salience of the ambiguous pronoun in the last clause towards the exclusive reading, given that pronominal resolution is affected by both the previous reference of the same term, and the same structural position (Grosz et al., 1995).On the other hand, immediately preceding the ambiguous pronoun there is an explicit mention of people making reference to the wider publica potentially explicit anchor for the pronoun.Such textual sequencing of we in this way amounts to what Mulderrig (2012, p. 719) calls 'strategic vagueness'.The conflicting pieces of information in (21) contributed to the ambiguous coding of the pronoun appearing in the use of the slogan message.
The Independent Scientific Advisory Group for Emergencies (SAGE) has advised the UK government on various medical and social aspects of COVID-19 and the appropriate response.In their recommendations in relation to government messaging about COVID-19 mitigations, SAGE (2020, p. 1) explicitly advised that 'branding or sloganeering should not come at the expense of clarity and precision'.What we have found in our analysis is that it is in these slogans that the reference of the pronoun, and therefore who is responsible for taking the related action, is in fact the least clear and precise.This goes against the very purpose of using slogans in messaging.This ambiguity arises because the meaning of deictic terms is, by their very definition, not consistent across contexts, and the embedding of these messages in different speeches results in the various inter-and intratextual factors preferring one reading or the other, or neither, in different contexts.

Who is building and testing?
The final set of examples we want to highlight concerns situations where an ambiguous pronoun appears in contexts where the material processes refer to actions meant to counter the pandemic.For example, consider the following (the referent of the pronouns is left out for the moment): (22) Over the past few months, we have built a critical national infrastructure for testing on a massive scale.We have already put in place the building blocks.We have developed the test, we've built the test centres, and the lab capacity.We've created home testing kits.
Looking at the verbal semantics of the clauses in which we appears in this example, they refer to actions which the general public has little (if any) active role in.This would bias these examples to an exclusive reading.However, the immediately preceding context provides textual information that would bias an inclusive reading, hence creating ambiguity.Below is the longer extract to illustrate this point: (23) We I must all renew our efforts.Over the course of this pandemic people, all of across the UK have been making difficult but vital sacrifices for the greater good.So let's not go back to square one.
We I can all play our part in the national effort, getting R down and keeping R down, and controlling the virus so we I can restore more of the things that make life worth living.As we I follow our plan, our testing regime will be our guiding star.It is the information that helps us I search out and defeat the virus.
Over the past few months, we A have built a critical national infrastructure for testing on a massive scale.We A have already put in place the building blocks.We A have developed the test, we A 've built the test centres, and the lab capacity.We A 've created home testing kits.[MAY21DR] The last referents for the 1PL before the ambiguous cases are clearly inclusive, since they are explicitly disambiguated twice at the beginning of this extract.This then creates the conflict between the verbal semantics, which preferences the exclusive reading, and the preceding referential and textual information, which biases the inclusive reading.The ambiguity in the pronominal reference here has clear ramifications for the ascription of responsibility, since if it is not precisely clear who is being represented here as an actor in these processes, then the question of who has responsibility over the success or failure of these measures is equally unclear.
Similar tensions between textual and referential information in the preceding contexts affected other examples in the sample.Below are some further examples of these situations (24) The more we A restrict contact, the more we A slow the spread of the infection, the more we A can help the NHS build capacity needed to care for those most in need [MAR31MG] (25) We A 've more than doubled the capacity of NHS and Public Health England Labs, and created 3 brand new mega-labs to analyse the results [MAY01MH] (26) We A have developed the test, we A 've built the test centres and the lab capacity, we A 've created the home testing kits.[MAY21DR] These examples share something in common with the slogan examples discussed above.Firstly, there are cases of pronominal ambiguity in the representation of actions directly intended to counter the pandemic, and secondly, this ambiguity arises because of the conflicting sources of information that speakers may use to achieve pronominal ambiguity, and hearers may rely on when interpreting pronominal reference, as has been observed as commonplace in the literature in routine, non-emergency political discourse.

Conclusion
This paper has analysed the ways in which responsibility was allocated in the UK government's daily televised COVID-19 briefings in 2020.This was examined by combining the analysis of the first-personal plural pronoun we and its position within transitivity structures.While the government represented themselves as provider, helping and supporting, the inclusive properties of we were used to construct collective responsibility with the people for the political and social actions that were designed to directly tackle the virus.We found that, generally, the government speakers employed different lexicogrammatical strategies to reduce and mitigate the amount of responsibility to themselves solely, while increasing the amount assigned to the general public, using strategies that we have termed 'grammatical distancing' and the construction of a 'hierarchy of responsibility'.At the same time, we argue that the government strategically invoked the inherent ambiguity of we over the course of briefings to render opaque precisely who is responsible for key, contextually pertinent processes and actions in fighting the virus.Moreover, we found that this also occurred in the expression of slogans, which turned the very pieces of language designed for clarity and precision into ambiguous messages.
In pursuing this analysis, we set out a transparent and reproducible process for the resolution of referents of we.The first-person plural pronoun is renowned for its utility for political speakers, precisely because of its potential to be vague, slippery and ambiguous.It is for these reasons that the identification of who is being referred to by we can be a difficult task for discourse analysts wishing to explore its uses in their data.The process outlined and employed here may offer a step towards developing a more explicit means of pronominal disambiguation in research.This, combined with a principled approach to transitivity analysis, has revealed the links between grammar, representation and responsibility.The key finding here is that the government resorted to ordinary rhetorical strategies during extra-ordinary times.As O' Malley et al. (2009, p. 617) note: 'there are ethical, strategic and public health imperatives that point to the need for transparency in communication of information during a public health emergency'.On the basis of the analysis presented here, some may argue that these imperatives were not fully met by the UK government.

Notes
David Wright is an Associate Professor in Linguistics at Nottingham Trent University.His research applies methods of corpus linguistics and discourse analysis in legal contexts.His research spans across a range of intersections between language and the law and justice, language in crime and evidence, and discourses of power, harassment and discrimination.
): (12) But first, let me update you on the steps we E are taking to defeat coronavirus, and the decisions we have taken today [APR16DR] (13) I would like to update you on the steps we E are taking to defeat this pandemic [APR01AS] (14) And today I want to set out further that we E are taking further steps to protect the NHS and especially around face coverings and face masks.[JUN05MH](15) We've E taken unprecedented action to increase NHS capacity, with more beds, more staff and more equipment on the front line.[APR08RS]

Table 1 .
Speakers, positions within government at the time and their frequency in the dataset.
Secretary of State for Digital, Culture, Media and Sport 3 George Eustice Secretary of State for Agriculture, Fisheries and Food 2

Table 2 .
Frequency of different we referents.

Table 3 .
Frequency of process types per pronoun reference.

Table 4 .
The most common lexical verbs used in material processes contain exclusive and inclusive 1PL as Participant., we see them being represented as Actors in processes that are more directly relevant to combatting the spread of the virus.This includes processes of controlling, saving, beating and protecting; examples (7-10) are indicative of these kinds of processes: (7) We I must continue to control the virus and [we I must continue to] save lives [MAY10BJ] We I must all follow the NHS Test and Trace instructions because this is how we I control the virus and [how we I ] protect the NHS and [how we I ] save lives.[MAY27MH]

Table 6 .
Transitivity analysis of I would like to set out the steps we are taking to defeat coronavirus.