To save content items to your account,
please confirm that you agree to abide by our usage policies.
If this is the first time you use this feature, you will be asked to authorise Cambridge Core to connect with your account.
Find out more about saving content to .
To save content items to your Kindle, first ensure no-reply@cambridge.org
is added to your Approved Personal Document E-mail List under your Personal Document Settings
on the Manage Your Content and Devices page of your Amazon account. Then enter the ‘name’ part
of your Kindle email address below.
Find out more about saving to your Kindle.
Note you can select to save to either the @free.kindle.com or @kindle.com variations.
‘@free.kindle.com’ emails are free but can only be saved to your device when it is connected to wi-fi.
‘@kindle.com’ emails can be delivered even when you are not connected to wi-fi, but note that service fees apply.
This study explores the importance of a full(er) understanding of ‘context’ when analysing and interpreting the indirect incitement of violence, a speech act closely related to that of persuasion. Using descriptions of context from Systemic Functional Linguistics (Bowcher, 2014; Hasan, 2014, 2020), the chapter qualitatively examines a small number of selected extracts from a very particular online community – a pick-up artist (PUA) forum. The main argument here is that the contextual configuration of the discourse, in terms of the nature the community and the participants in it, is such that posts which on the face of it appear to be giving advice or expressing opinions could be potentially inciteful in this context. Incitement has typically received relatively little linguistic attention, and indirect incitement poses a specific challenge for linguistics and law alike. This chapter goes some way to providing a new perspective from which the elusive discursive action of incitement can be analysed and interpreted.
This chapter explores how members of the online incel community interact with (presumably) new members on the subreddit r/Braincels. It seeks to explore why new members might stay on a website known for hostility and poor mental health. Through the use of corpus linguistic methods, I explore ways that members of the community might interact with newcomers. I argue that users do not index their identity as new on the forum through lexemes such as new or first, and that there appears to be few repeated and explicit ways that members signal posting on the forum for the first time. As such, I analyse how incels construct their community, by using corpus linguistic methods and Martin and White’s (2005) appraisal framework. I argue that the members of the incel community constructed their group as ‘toxic’, but also as a place where they were able to receive mental health support. Ultimately, I argue that more work needs to be done to explore not only why people might engage with these kinds of forums, but also how members of the incel community convince people to stay engaged.
Traditionally, the role of language in persuasion has been mostly studied in experimental settings with a focus on how persuasive messages are understood, processed, and ultimately complied with. Recently, a new approach has emerged that focuses on the sequential properties of language-in-use mobilised in real-life persuasion-in-interaction (Humă, 2023). This body of research illuminates how aspects of sequence organisation (Humă et al., 2019, 2020), turn design (Llewellyn, 2015), and lexical choice (Sikveland & Stokoe, 2016, 2020) are implicated in persuasion. The present study contributes to this line of work by zooming in on two configurations for formatting requests in sales interactions: when-formulated and if-formulated sales requests. Using conversation analysis to examine a corpus of 159 real-life telephone calls between salespeople and prospective customers, I show that the former configuration is more effective in eliciting productive responses that advance the commercial activity. These findings can be explained in terms of the differential opportunities afforded by the two configurations to reject the sales requests. Thus, this study strengthens the claim that, in real-life social interaction, persuasion is mainly realised through the architecture of possibilities for responsive action and not through the effects of language-in-use on individual minds.
The use of the English language in many countries around the world greatly facilitates international communication. However, linguists have long pointed out that differences between established and emerging dialects of English may lead to miscommunication, especially because many users of the language may not be aware of many of these differences. Such issues may be particularly acute in high-stakes communicative endeavours such as persuading others to change their opinion or engage in a certain action. Psychologists have previously explored this construct mainly in experimental settings, whereas the present chapter uses a large database (‘corpus’) of natural language. The chapter thus provides an empirical, corpus-based analysis of the linguistic expression of persuasion across 21 international dialects (‘varieties’) of English from countries where English is widely used as a first or second language. Results indicate that there are substantial differences in the degree of overt expression of persuasion, with South Asian varieties of English indicating the lowest levels and (West) African varieties that greatest level of overt persuasion. The chapter concludes by discussing possible explanations of these patterns, implications for international communication as well as avenues for more detailed analyses of particular linguistic features involved in persuasion.
In this paper, we investigate whether appeals to expertise make robots persuasive and provide evidence on the influence of single persuasive messages in human-robot interactions. We explore the effects of two different kinds of persuasive strategies on people’s behavior and subjective evaluation of the robot: appeals to participants’ own expertise on the one hand and reference to research on the other. We present a controlled lab study in a healthcare scenario with professional elderly care workers as our participants, where the aim is to address dehydration. We study attitudinal and behavioral effects of these strategies of influence; specifically, we measure participants’ water intake after the interaction, as well as their subjective ratings of the robot. Our results show that both strategies have influence on participants’ water intake while the reference to one’s own expertise yields significant behavioral effects.
The chapter explores viewpoint across various topics and genres of political discourse. Viewpoint is defined as a pervasive property of language and conceptualisation which is exhibited across a broad range of linguistic and conceptual phenomena. The chapter starts by looking at deixis and deictic shifts in media discourses of immigration and political protests. The ideological role of viewpoints evoked by transitive versus reciprocal verbs is also considered in the context of media coverage of political protests. Subjective versus objective construal is further analysed as a viewpoint phenomenon and the role of objective construals in official communication around Covid-19 is highlighted. Viewpoint as an inherent feature in the mental spaces networks configured in response to modal and conditional constructions are considered in the context of Brexit discourse. Finally, conducted within the framework of discourse space theory, an analysis is given of distance and proximity (relative to a deictically specified viewpoint) in the discourse of the far-right organisation Britain First.
The chapter is concerned with metaphor and focusses specifically on war metaphors in political discourses. The cognitive mechanisms at work in metaphor are described with an emphasis on frames as the unit of conceptual organisation that gets mapped in political metaphors. Recent experimental studies demonstrating the framing effects of metaphor are discussed. The war frame is described to include discussion of intertextuality as a means of accessing it. Three case studies are then presented exploring war metaphors in discourses of Covid-19, Brexit and immigration. Analogies with the first and second world wars in particular are highlighted and critiqued. The chapter defines and discusses extreme metaphors illustrated through examples in which immigrants are compared to animals and closes with a discussion of how readers may resist extreme metaphors.
The chapter explores co-speech gestures in spoken political discourse. It defines co-speech gesture as a fundamental feature of communication which is implicated in the discursive performance of prejudice. Gesture-speech relations are discussed and a classification of gestures is provided. It is shown how speech and gesture may interact with respect to schematisation, viewpoint, attention and metaphor. Two case studies focussed on the gestural style of right-wing populism are presented. The first considers the co-speech gestures executed by Donald Trump during a campaign rally. The analysis highlights his comedic use of gestures, the use of iconic and enactment gestures in connection with his border wall policy, and his use of points and shrugs to engage with his audience in different ways. The second focusses on co-speech gestures in the anti-immigration discourse of Nigel Farage. The analysis shows that legitimating moves characteristic of immigration discourse, including focussing, denial, authorisation, deixis, proximisation, metaphor, quantification and aspectising, when performed in spoken discourse are multimodal and involve a gestural component.
The chapter outlines key principles in Cognitive CDA, which inherits its social theory from CDA and from cognitive linguistics inherits a particular view of language and a framework for analysing language (as well as other semiotic modes). In connection with CDA, the chapter describes the dialectical relationship conceived between discourse and society. Key concepts relating to the dialogicality of discourse are also introduced, namely intertextuality and interdiscursivity. The central role of discourse in maintaining power and inequality is described with a focus on the ideological and legitimating functions of language and conceptualisation. In connection with cognitive linguistics, the chapter describes the non-autonomous nature of language, the continuity between grammar and the lexicon and the experiential grounding of language. The key concept of construal and its implications for ideology in language and conceptualisation are discussed. A framework in which construal operations are related to discursive strategies and domain-general cognitive systems and processes is set out. The chapter closes by briefly introducing the main models and methods of Cognitive CDA.
The chapter addresses attentional distribution in conceptualisations of events. It argues that language directs attention over particular portions of an event-structure selecting certain elements for focal attention while conceptually backgrounding other elements. The ideological implications of attentional distribution are discussed with reference to mystification whereby either human agency in or the human impact of harmful social actions is obscured. Two case studies are presented. The first considers action-chain profiling in media coverage of fatalities on the Gaza border. It shows how attentional distributions evoked by intransitive, passive and agentless passive constructions as well as nominalisations conceptually background those responsible for the fatalities. It further shows the conceptual means by which the impact of violent actions may be mitigated. The second considers path-profiling in immigration discourse. It shows how different verb choices serve to highlight humanitarian motivations for migration versus the impact of migration on host countries and considers the role of metonymy in legitimating hostile immigration policies.