Pandemic and Crisis Discourse

Communicating COVID-19 and Public Health Strategy

Andreas Musolff, Ruth Breeze, Kayo Kondo, Sara Vilar-Lluch, Andreas Musolff, Ruth Breeze, Kayo Kondo, Sara Vilar-Lluch

• Editorial: Bloomsbury Academic
• Año: 2022
• ISBN:
  9781350232716
• Edición: 1
• Categoría: Languages & Linguistics
• Categoría: Linguistic Semantics

The COVID-19 pandemic has led to a host of critical reflections about discourse practises dealing with public health issues. Situating crisis communication at the centre of societal and political debates about responses to the pandemic, this volume analyses the discursive strategies used in a variety of settings. Exploring how crisis discourse has become a part of managing the public health crisis itself, this book focuses on the communicative tasks and challenges for both speakers and their public audiences in seven areas: - establishment of discursive and political authority
- official governmental and expert communication to the public
- public understanding of government communication
- legitimation of public health management as a 'war'
- judging and blaming a collective other
- cross-national comparison and rivalry
- empathy and encouragement Covering global discourses from Asia, Europe, the Middle East, North and South America, and New Zealand, chapters use corpus-based data to cast light on these issues from a variety of languages. With crisis discourse already the object of fierce national and international debates about the appropriateness of specific communicative styles, information management and 'verbal hygiene', Pandemic and Crisis Discourse offers an authoritative intervention from language experts.

PART ONE

The discourse of authority in a global crisis: Who defines (if there is) a pandemic?

CHAPTER ONE

Covid-19 press conferences across time: World Health Organization vs. Chinese Ministry of Foreign Affairs

DENNIS TAY
DEPARTMENT OF ENGLISH AND COMMUNICATION, THE HONG KONG POLYTECHNIC UNIVERSITY

INTRODUCTION

Press conferences are the standard platform for institutional representatives to communicate with the public through the media. From a critical discourse analytic perspective, the communication of factual information is only part of their nature. They also reveal ‘how ideologies are discussed and negotiated, how power relations are asserted, and how political differences on difficult issues are discussed and communicated in a positive way’ (Bhatia, 2006: 174). While linguistic aspects of press conferences still remain curiously underexamined, existing studies have focused on those with an explicit political intention, as they most clearly depict ideologies and power relations. These range from theoretical reflections on media–government relations (Fairclough, 2000) to case studies of individual politicians’ behaviour (Ryfe, 1999) and nuanced thematic analysis of politicians’ utterances (Bhatia, 2006).

The impact of Covid-19 has likewise led to multiple press conferences over a relatively short period to communicate not just facts about the pandemic, but stances towards implicated sociopolitical issues like global solidarity, responsibility and other ideological tensions. Two prominent institutions in this regard are the World Health Organization (WHO) and the Chinese government, represented internationally by its Ministry of Foreign Affairs (CMFA). As an agency of the United Nations responsible for international public health, the WHO has been in the spotlight since the onset of Covid-19. They have held regular press conferences since 22 January 2020 to provide situation updates, disseminate information on the latest measures and – perhaps most interestingly from a critical perspective – co-ordinate efforts between different countries. The Chinese government has come under considerable international scrutiny for its alleged cover-ups of key information during the onset, as well as using its financial clout to exert influence on the WHO. As it has since positioned itself as successfully containing the virus and a potential leader of the global fight, regular press conferences given by CMFA have likewise been dominated by Covid-related issues and its ideological underpinnings. These include addressing the allegations, and the role of Covid-19 in the wider strategic rivalry between China and the United States.

Given this backdrop, it becomes interesting not only to examine but also to compare the language of Covid-19 press conferences from WHO and CMFA. The former is in principle a neutral health authority with an international ambit, while the latter has not been accorded an ideologically neutral position even as it attempts to assert international influence. Nevertheless, the aforementioned allegations of their complicity lead one to wonder to what extent their stances, as linguistically manifested, might actually align. This chapter presents a study of near-daily English language press conferences by WHO and CMFA across a three-month span from 20 January, near the crisis onset, to 1 May 2020. It adopts analytic perspectives that differ from, but are complementary with, existing studies on (political) press conferences. First, rather than focusing on what gets communicated via content/thematic analyses, the present study investigates how, with computer-assisted analysis of attitudes and stances revealed through lexical choices (Pennebaker et al., 2015; Tausczik and Pennebaker, 2010). This approach moves beyond impactful, but ultimately isolated, examples of discursive strategies realized in short extracts, to make claims about more representative bodies of data. Second, the study explicitly models diachronic shifts in these lexical measures with time series analytic methods that have underexplored potential for discourse research (Tay, 2017, 2019). It addresses the following research questions and their implications for understanding socio-psychological stances towards Covid-19:

i. How does the language of WHO and CMFA Covid-19 press conferences compare in terms of socio-psychological constructs like analyticity, authenticity, clout and emotional tone?

ii. Does this language exhibit modellable structural regularities across time?

DATA AND METHOD

WHO and CMFA press conference transcripts were downloaded from their respective websites.¹ The time span of interest was from 20 January to 1 May 2020. This covers an approximate three-month period from when the China National Health Commission first confirmed that the virus was human-to-human transmissible,² to just after the 100th WHO situation report was released. This resulted in a corpus of 50 WHO (377,359 words) and 67 CMFA transcripts (109,647 words), as there were days without press conferences. The data was cleaned by removing irrelevant details like introductory text, but both media questions and responses were retained for analysis.

The analytical methodology comprises four distinct steps: (i) automated lexical analysis with the Linguistic Inquiry and Word Count (LIWC) program (Pennebaker et al., 2015), (ii) comparative statistical analyses of resulting variables between the two corpora, (iii) time series analysis of the variables following the Box-Jenkins method with ARIMA models (Box et al., 2015; Tay, 2019) and (iv) close contextual analysis of illustrative examples implied by the above results. Steps (i) and (iii) will be described in more detail below as they are likely to be more unfamiliar to readers.

Linguistic Inquiry and Word Count

Linguistic Inquiry and Word Count (LIWC) is a computerized text analysis program widely used in socio-psychological studies of language(s) in many contexts. It computes word frequencies in an input text under approximately ninety different lexical (e.g. cognitive, affective, social) and grammatical categories (e.g. parts of speech). Additionally, it can compute four ‘summary variables’: analytical thinking, clout, authenticity and emotional tone, each defined by some combination of the above lexical/grammatical categories. Many studies have shown that these categories co-occur and reliably differentiate input texts along the traits implied by the summary variables. These are summarized in Table 1.1. Plus signs indicate categories that are relatively frequent in texts that reflect a higher level of that summary variable, and vice versa for minus signs. The psychometric properties of LIWC are detailed in Pennebaker et al. (2015).

TABLE 1.1: Summary variables and defining lexical categories

Summary variable	Defining lexical categories
Analytical thinking	+articles, prepositions -pronouns, auxiliary verbs, conjunctions, adverbs, negations (Pennebaker et al., 2014)
Clout	+1st person plural pronouns, 2nd person pronouns -tentative words (e.g. maybe, perhaps) (Kacewicz et al., 2013)
Authenticity	+1st person singular pronouns, 3rd person pronouns, exclusive words (e.g. but, except, without) -negative emotion words (e.g. hurt, ugly, nasty), motion verbs (e.g. walk, move, go) (Newman et al., 2003)
Emotional tone	+positive emotion words (e.g. love, nice, sweet) -negative emotion words (e.g. hurt, ugly, nasty) (Cohn, Mehl and Pennebaker, 2004)

Each summary variable in an input text is scored from 0 to 100, based on standardized scores from large comparison samples. A high analytical thinking score suggests formal, logical and hierarchical thinking versus informal, personal, here-and-now and narrative thinking. This is based on a study of American college admission essays, where essays with more articles and prepositions were found to be more formal and precise in describing objects, events, goals and plans, while those with more pronouns, auxiliary verbs, etc. were more likely to involve personal stories (Pennebaker et al., 2014).

A high clout score suggests speaking/writing with high expertise and confidence versus a more tentative and humble style. This is based on studies of decision-making tasks, chats and personal correspondences. Higher-status individuals used more we/our, you/your and fewer tentative words. This was explained by an association between relative status and attentional bias. Higher-ups are more other-focused and less unsure, while lower individuals are more self-focused and tentative (Kacewicz et al., 2013).

A high authenticity score suggests more honest, personal and disclosing discourse versus more guarded and distanced discourse. This is based on studies of elicited true versus false stories where the latter has fewer first- and third-person pronouns, exclusive words and more negative emotion and motion verbs. This was explained by the idea that liars tend to dissociate themselves with the lie, feel greater tension and guilt, and speak in less cognitively complex ways. These linguistic tendencies accurately distinguished truth-tellers versus liars in independent data more than 60 per cent of the time (Newman et al., 2003).

A high emotional tone score suggests a more positive and upbeat style, a low score anxiety/sadness/hostility, while a neutral score around 50 a lack of emotionality. This was based on a study of online journals prior to and after the September 11 attacks where negative emotion words increased sharply following the attack and gradually returned to pre-attack baselines after some time (Cohn, Mehl and Pennebaker, 2004). Table 1.2 summarizes the four variables and the general interpretation of high/low/middle scores.

TABLE 1.2: Summary variables and interpretation

With each transcript as an input text, LIWC thus provides multivariate profiles of the language of press conferences across the time span – an approach that has recently been applied to other discourse contexts with a socio-psychological emphasis like psychotherapy talk (Huston et al., 2019; Tay, 2020), editorials (Tay 2021a), and social media (Tay 2021b). The four summary variables are especially useful for revealing aspects such as whether Covid-19 issues are logically or narratively framed, power relations projected by speakers, extents of disclosure and emotional attitudes towards the situation. They complement semantic annotation tools like USAS (Archer et al., 2002) that focus more on content categories like ‘food and farming’ or ‘science and technology’.

Time series analysis (Box-Jenkins method)

A time series is a set of consecutive measurements of a random variable made at intervals that are usually equally spaced (e.g. minutes, days, years). Examples include stock prices, heartbeat and rainfall. A typical time series appears like erratic fluctuations with some trends and/or cyclical behaviour embedded within. Analysts have two general objectives: (i) to discover mathematically expressible regularities underlying the series, and (ii) to forecast future values on that basis. The main difference between time series analysis and standard regression models is that the latter assumes measurements to be mutually independent, and relies on how these measurements correlate with other predictor variables. The former, on the other hand, relies precisely on how the values within the series itself are correlated with one another (i.e. auto-correlated), without considering the influence of other external predictors. The Box-Jenkins method (Box et al., 2015) is one of the most widely used by time series analysts. Tay (2019) argues that many discourse contexts lead to naturally occurring time series, and demonstrates how to apply the Box-Jenkins method on news articles, classroom talk and psychotherapy sessions, the latter of which can profit from forecasting future linguistic behaviour by clients. As an exploratory modelling approach that solely considers values across time without input from other preconceived predictors, it dovetails with bottom-up discourse analytic approaches where data patterns are first established, and then interpreted or explained.

The stepwise methodology (i.e. inspecting the series, calculating autocorrelations, selecting and evaluating candidate models, performing residual diagnostics) is detailed in Tay (2019) and will not be repeated here due to space constraints. I will instead highlight the major sources of insight one can expect in a more conceptual way. At the specific level, the details of a good fitting final model for any linguistic/discursive variable can reveal much about its nature across time. Take, for example, what are known as ‘autoregressive’ (AR) models. Its generic form is y_t = C + Φ_ky_t–k + a_t where

• y_t is the present value in the series

• C is known as the constant term of the model

• Φ_k is a coefficient also known as the AR operator

• y_t–k is the past series values at time t−k

• a_t is the error term at time t

From this, we can derive information like the mean value of the series and the direction and extent to which the present value is linearly related to the value k intervals in the past. These can in turn be interpreted with careful reference to specific contextual characteristics of the discourse.

At the general level, which will be the focus of this chapter, we ...

Índice

• Cover
• Halftitle Page
• Title Page
• Contents
• List of Figures
• List of Tables
• List of Contributors
• Introduction: From declarations of war to denial to explanations: How global publics have coped with the Covid-19 pandemic
• Part I: The discourse of authority in a global crisis: Who defines (if there is) a pandemic?
• Part II: The discourse of crisis management: How is the public meant to understand the pandemic and how does it actually do so?
• Part III: The discourse of ‘War’ against the pandemic: How to ‘Fight’ Covid-19?
• Part IV: The discourse of judgement and rivalry: Blaming other/s for the pandemic and comparing national performances
• Part V: The discourse of empathy and encouragement: How to foster solidarity among doctors, patients and health experts
• Index
• Imprint