Nele Poldvere

• Trnavac, Radoslava & Poldvere, Nele (2024). Investigating Appraisal and the Language of Evaluation in Fake News Corpora. Corpus Pragmatics. ISSN 2509-9507. p. 1–24. doi: 10.1007/s41701-023-00162-x. Full text in Research Archive Show summary

  The present corpus study, which is grounded in Appraisal Theory, investigates evaluative language use in fake news in English. The primary aim is to find out how and why, if at all, evaluative meanings are construed differently in fake news compared to genuine news. The secondary aim is to explore potential differences between types of fake news based on contextual factors. The data are from two carefullydesigned corpora containing both fake and genuine news: a single-authored corpus and a multi-authored corpus. Both corpora contain false information that is meant to deceive, but they also differ from each other in terms of register, genre and the motivational goals of the authors. Through qualitative and quantitative analyses, we show that there are systematic differences in the occurrence of Appraisal expressions across fake and genuine news, with Appraisal being more common in the former. However, the exact nature of the affective, dialogic and modal expression of fake news is influenced by contextual factors that, so far, have largely been ignored in fake news research. Therefore, the study has important implications for the development of fake news detection systems based on data sources of different kinds, a task which is in grave need of the input of corpus linguists.

• Poldvere, Nele; Kibisova, Elizaveta & Alvestad, Silje Susanne (2023). Investigating the language of fake news across cultures. In Maci, Stefania; Demata, Massimiliano; McGlashan, Mark & Seargeant, Philip (Ed.), The Routledge Handbook of Discourse and Disinformation. Routledge. ISSN 9781003224495. p. 153–165. doi: https:/doi.org/10.4324/9781003224495-11. Full text in Research Archive Show summary

  This chapter presents the methodological considerations, the challenges and the opportunities, of investigating the language of fake news across cultures. The methodological considerations are based on the Fakespeak project, an interdisciplinary project involving linguists and computer scientists whose objectives are, firstly, to reveal the language and style of fake news in English, Norwegian, and Russian and, secondly, to see whether the findings can help improve existing fake news detection systems. The procedure of collecting data in the three languages representing three different cultures (e.g. English in the USA) presented several methodological challenges, which required innovative solutions. The challenges were quite different in each case, with access to quality data (collections of fake and genuine news individually labelled for veracity by experts) being especially limited for Norwegian and Russian, albeit for different reasons. The datasets that have emerged from the Fakespeak project so far are very diverse, including both large datasets with news written by a mix of authors (so-called general datasets) and smaller datasets with fake and genuine news written by the same author (single-authored and hybrid, multiauthored datasets). The diversity of data types is important for linguistic analyses of fake news so that meaningful comparisons with genuine news can be made, as well as for the subsequent development of automatic detection systems.

• Poldvere, Nele; Uddin, Md Zia & Thomas, Aleena (2023). The PolitiFact-Oslo Corpus: A New Dataset for Fake News Analysis and Detection. Information. ISSN 2078-2489. 14(12). doi: 10.3390/info14120627. Full text in Research Archive Show summary

  This study presents a new dataset for fake news analysis and detection, namely, the PolitiFact-Oslo Corpus. The corpus contains samples of both fake and real news in English, collected from the fact-checking website PolitiFact.com. It grew out of a need for a more controlled and effective dataset for fake news analysis and detection model development based on recent events. Three features make it uniquely placed for this: (i) the texts have been individually labelled for veracity by experts, (ii) they are complete texts that strictly correspond to the claims in question, and (iii) they are accompanied by important metadata such as text type (e.g., social media, news and blog). In relation to this, we present a pipeline for collecting quality data from major fact-checking websites, a procedure which can be replicated in future corpus building efforts. An exploratory analysis based on sentiment and part-of-speech information reveals interesting differences between fake and real news as well as between text types, thus highlighting the importance of adding contextual information to fake news corpora. Since the main application of the PolitiFact-Oslo Corpus is in automatic fake news detection, we critically examine the applicability of the corpus and another PolitiFact dataset built based on less strict criteria for various deep learning-based efficient approaches, such as Bidirectional Long Short-Term Memory (Bi-LSTM), LSTM fine-tuned transformers such as Bidirectional Encoder Representations from Transformers (BERT) and RoBERTa, and XLNet.

• Seitanidi, Eleni; Poldvere, Nele & Paradis, Carita (2023). All-cleft constructions in the London–Lund Corpora of spoken English: Empirical and methodological perspectives. Journal of Pragmatics. ISSN 0378-2166. 207, p. 78–92. doi: 10.1016/j.pragma.2023.02.003. Full text in Research Archive Show summary

  This study investigates all-cleft constructions in the London–Lund Corpora of spoken British English. It has two aims: an empirical and a methodological one. The empirical aim is to contribute to research on spoken discourse by analysing the form–meaning properties of all-clefts along with their pragmatic functions and their development since the 1950s drawing on insights from Cognitive Linguistics with special focus on Diachronic Construction Grammar. We show that all-clefts are used both as assertions and directives. They focalise an element that speakers find particularly relevant at the same time as they have the effect of blocking alternative perspectives and thereby contracting the discursive space for addressees. The methodological aim is to critically assess the comparability level of the diachronic corpora based on a detailed investigation of the distribution of all-clefts in the data. We show that there is a high degree of similarity between the designs of the corpora, which makes them suitable for diachronic investigations of recent change.

• Põldvere, Nele; Johansson, Victoria & Paradis, Carita (2021). On the London–Lund Corpus 2: Design, challenges and innovations. English Language and Linguistics. ISSN 1360-6743. 25(3), p. 459–483. doi: 10.1017/S1360674321000186. Full text in Research Archive Show summary

  This article describes and critically examines the challenging task of compiling The London–Lund Corpus 2 (LLC–2) from start to end, accounting for the methodological decisions made in each stage and highlighting the innovations. LLC–2 is a half-a-million-word corpus of contemporary spoken British English with recordings from 2014 to 2019. Its size and design are the same as those of the world's first machine-readable spoken corpus, The London–Lund Corpus of Spoken English with data from the 1950s to 1980s. In this way, LLC–2 allows not only for synchronic investigations of contemporary speech but also for principled diachronic research of spoken language across time. Each stage of the compilation of LLC–2 posed its own challenges, ranging from the design of the corpus, the recruitment of the speakers, transcription, markup and annotation procedures, to the release of the corpus to the international research community. The decisions and solutions represent state-of-the-art practices of spoken corpus compilation with important innovations that enhance the value of LLC–2 for spoken corpus research, such as the availability of both the transcriptions and the corresponding time-aligned audio files in a standard compliant format.

• Põldvere, Nele; Johansson, Victoria & Paradis, Carita (2021). Resonance in dialogue: The interplay between intersubjective motivations and cognitive facilitation. Language and Cognition. ISSN 1866-9808. 13, p. 643–669. doi: 10.1017/langcog.2021.16. Full text in Research Archive Show summary

  Dialogic resonance, when speakers reproduce constructions from prior turns, is a compelling type of coordination in everyday conversation. This study takes its starting point in resonance in stance-taking sequences with the aim to account for the interplay between intersubjective motivations and cognitive facilitation in resonance production. It analyzes stance-taking sequences in the London–Lund Corpus 2, determining (i) the type of stance alignment (agreement or disagreement), and (ii) the time lapse between the stance-taking turns. The main findings are, firstly, that resonance is more likely than non-resonance to be used by speakers who express disagreement than agreement, which we interpret as a mitigating function of resonance, and, secondly, that the turn transitions are faster in resonating sequences due to cognitive activation in the prior turn. We propose that the face-saving intersubjective motivation of resonance combines with its facilitating cognitive effect to promote appeasing communication.

• Põldvere, Nele; Frid, Johan; Johansson, Victoria & Paradis, Carita (2021). Challenges of releasing audio material for spoken data: The case of the London–Lund Corpus 2. Research in Corpus Linguistics. ISSN 2243-4712. 9(1), p. 35–62. doi: 10.32714/ricl.09.01.04. Full text in Research Archive Show summary

  This article aims to describe key challenges of preparing and releasing audio material for spoken data and to propose solutions to these challenges. We draw on our experience of compiling the new London-Lund Corpus 2 (LLC-2), where transcripts are released together with the audio files. However, making the audio material publicly available required careful consideration of how to, most effectively, 1) align the transcripts with the audio and 2) anonymise personal information in the recordings. First, audio-to-text alignment was solved through the insertion of timestamps in front of speaker turns in the transcription stage, which, as we show in the article, may later be used as a valuable complement to more robust automatic segmentation. Second, anonymisation was done by means of a Praat script, which replaced all personal information with a sound that made the lexical information incomprehensible but retained the prosodic characteristics. The public release of the LLC-2 audio material is a valuable feature of the corpus that allows users to extend the corpus data relative to their own research interests and, thus, broaden the scope of corpus linguistics. To illustrate this, we present three studies that have successfully used the LLC-2 audio material.

View all works in Cristin

• Poldvere, Nele; De Felice, Rachele & Paradis, Carita (2022). Advice in conversation: Corpus pragmatics meets mixed methods. Cambridge University Press. ISBN 9781009053617. 76 p. Full text in Research Archive Show summary

  This Element is a contribution to a new generation of corpus pragmatics research by taking as its starting point the multifaceted nature of speech acts in conversation, and by adopting a mixed-methods approach. Through a unique combination of theoretical, qualitative, quantitative, and statistical approaches, it provides a detailed investigation of advice-giving and advice uptake in relation to (i) the range of constructions used to give advice in different discourse contexts and at different points in time, and (ii) their interaction with dialogic and social factors of advice uptake as key components of frames of advice exchanges in natural conversation. Using data from the London-Lund Corpora of spoken British English, the Element shows, firstly, that there are systematic differences in advising between discourse contexts over the past half a century, and, secondly, that who gave the advice and how they did it are the strongest predictors of the advisee's response. This title is also available as Open Access on Cambridge Core.

View all works in Cristin

• Poldvere, Nele; Uddin, Zia & Thomas, Aleena (2024). Out of balance, out of sight: Issues with the design and accessibility of a corpus of fake and real news. Show summary

  Fake news is a topic that only recently has caught the attention of (corpus) linguists (Grieve & Woodfield, 2023; Sousa Silva, 2022; Trnavac & Põldvere, 2024). Such research has sought to identify differences in linguistic features between fake and real news based on carefully designed corpora. An example of such a corpus is the new PolitiFact-Oslo Corpus (Põldvere et al., 2023), a large dataset of fake and real news in English based on recent events (post-2019). However, in its current form the corpus has some limitations, due to the highly specific, and sensitive, nature of fake news. The present methodological study seeks solutions to these limitations with a view to facilitating future corpus building efforts around fake news, a highly promising area of study for linguists. As the name implies, the PolitiFact-Oslo Corpus relies on the fact-checking website PolitiFact.com for its data, with each news item being individually labelled for veracity by experts (from ‘True’ to ‘Pants on Fire’). In contrast to many other fake news datasets (e.g., DeClarE in Popat et al., 2018), the corpus is the result of a combination of automatic and manual procedures to have greater control over what is included. In addition to a manual approach to text selection, the corpus is accompanied by important metadata information about the texts, such as their text type (e.g., social media) and source (e.g., X). This said, the corpus currently has two major limitations. Firstly, there is a noticeable imbalance between the fake and real news samples (358,516 vs. 70,401 words, respectively), which is due to the preference of PolitiFact and other fact-checkers to debunk false information rather than to find support for true information. This limitation has serious implications for fake news analysis and detection model development based on the corpus (Põldvere et al., 2023). Secondly, due to copyright and privacy issues the corpus is currently not publicly available, a feature of the corpus which is hardly in line with current open science practices. We offer some solutions. As for the imbalance between the fake and real news samples, we have decided to extend the scope of the fact-checkers rather than to stretch out the timeline. The fact-checkers are found via Google’s Fact Check Explorer, which provides quick and easy access to more instances of (mostly or half) true news. The challenge is to ensure comparability of the ratings between the fact-checkers (what is ‘Mostly True’ according to one fact-checker may be ‘Half True’ according to another) as well as balance in terms of the metadata information (text type, source). The lack of access to the corpus is a much more complex problem to solve. Inspired by current practices in corpus linguistics, we are exploring opportunities to release the text snippets, rather than the full texts, via an online interface, which, however, is complicated by the legal challenges of distributing fake news data in our national context. We seek solutions to these challenges, too.

• Uddin, Zia; Poldvere, Nele & Thomas, Aleena (2024). The grammar of fake news: Corpus linguistics meets machine learning. Show summary

  In this study, we investigate the grammar of fake news by bringing together insights from corpus linguistics and machine learning. While the former offers a robust corpus-based register analysis of grammatical features, namely, multidimensional analysis (Biber, 1988), the latter contributes with methodological capabilities for the automatic identification of fake news based on the features. Fake news detection has made remarkable progress in natural language processing and machine learning (e.g., Rashkin et al., 2017; Põldvere et al., 2023), but it has not taken full advantage of the linguistic resources that are available. Based on the new PolitiFact-Oslo Corpus (Põldvere et al., 2023), we aim (i) to describe the grammatical differences between fake and real news across a variety of text types in a large corpus, and (ii) to develop a deep learning-based efficient approach for fake news detection based on these differences. A common distinction in multidimensional register analysis is between informational and involved styles of communication. While the former tends to contain more nouns and is common in registers with dense styles of communication such as news reportage, the latter is characterized by a more frequent use of pronouns, verbs and adjectives and is common in spontaneous conversation with lower levels of information density. Departing from the view that fake news is a register in its own right, Grieve and Woodfield (2023) analyzed 49 grammatical features in a small collection of fake and real news texts by one journalist. They found fake news to be more similar to involved styles of communication through the use of, e.g., present tense verbs, emphatics and predicative adjectives. This was different from real news which shared features with informational styles of communication. In contrast to Grieve and Woodfield (2023), in this study we make use of a large corpus of fake and real news in English: the PolitiFact-Oslo Corpus. The main strengths of the corpus are that the texts have been individually labelled for veracity by experts and are accompanied by important metadata about the text types (e.g., social media, news and blog) and sources (e.g., X, The Gateway Pundit). At present, the corpus contains 428,917 words of fake and real news, and it is growing. To extract the grammatical features, we used the Multidimensional Analysis Tagger (Nini, 2019), followed by a deep learning-based efficient approach (Attention-based Long Short-Term Memory; LSTM) to train the features incriminating fake and real news. The trained model was then used to automatically detect the fake news texts. The preliminary results based on a sample from the corpus indicate that there are systematic differences between fake and real news, which by and large are indicative of the distinction between involved and informational styles of communication, respectively. However, these differences are not the same across the text types, with social media showing lower levels of information density in fake news than news and blog. Our machine learning model based on the grammatical features also shows promising results (LSTM mean accuracy: 90%), particularly when compared to models without the grammatical features.

• Poldvere, Nele; Kibisova, Elizaveta; Trnavac, Radoslava & Alvestad, Silje Susanne (2024). Is stance in fake news expressed differently across languages? Show summary

  In this corpus study, we examine the use and distribution of stance expressions in fake news in three languages: English, Norwegian and Russian. Stance is a major communicative resource in language, and it plays an important role in how news organisations and social media users inform, persuade and entertain their audiences. Stance may also be an indicator of the ‘fakeness’ of news. Indeed, previous research (Trnavac & Põldvere, 2024) has identified significant differences in evaluative language use between fake and genuine news in English; however, it is unclear how these differences are reflected across languages and cultures. Therefore, the present study sets out to compare and contrast the occurrence of a range of lexico-grammatical features of stance across English in the US, Norwegian in Norway and Russian in Russia, based on Biber (2006). The lexico-grammatical features are divided into major structural and pragmatic categories (e.g., adverbs expressing epistemic stance: certainly). The data are from three comparable corpora, collected from fact-checking websites covering news in the respective languages. Based on previous cross-linguistic research on stance in news discourse (Marin-Arrese, 2015), we expect there to be considerable differences across the three languages due to general differences in argumentative and persuasive style, as well as specific cultural characteristics associated with each country (e.g., varying levels of freedom of expression). References Biber, D. (2006) Stance in spoken and written university registers. Journal of English for Academic Purposes, 5, 97–116. https://doi.org/10.1016/j.jeap.2006.05.001 Marin-Arrese, J. I. (2015). Epistemicity and stance: A cross-linguistic study of epistemic stance strategies in journalistic discourse in English and Spanish. Discourse Studies, 17(2), 210–225. https://doi.org/10.1177/1461445614564523 Trnavac, R., & Põldvere, N. (2024). Investigating Appraisal and the language of evaluation in fake news corpora. Corpus Pragmatics. https://doi.org/10.1007/s41701-023-00162-x

• Poldvere, Nele (2024). Manual annotation – the destroyer of fun? Not with the right tool(s).
• Poldvere, Nele (2024). Corpus approaches to stance and evaluation in fake news. Show summary

  In this talk, I report on two corpus studies on the language of fake news—on stance and evaluation in English—from the Fakespeak project, University of Oslo. Fakespeak is concerned with revealing the grammatical and stylistic features of the language of fake news with a view to adding those features to automatic fake news detection systems for better performance. Importantly, we aim to go beyond simply describing what the language of fake news is like, and what features are more common in fake news compared to genuine news: we also want to know the reasons for the observed differences, as facilitated by linguistic theory. Therefore, in the two studies we draw on two well-known theories from linguistics to study stance and evaluation in fake news, namely, Appraisal Theory (Martin & White, 2005) and Biber's (2006) framework of grammatical stance. In the latter, I also briefly touch upon our work on extending the framework to Russian and Norwegian, the two other languages of Fakespeak. The corpora for the studies are of different kinds, ranging from small, but highly controlled corpora of disinformation and deception, to large multilingual corpora containing both mis- and disinformation. References Biber, D. (2006). Stance in spoken and written university registers. Journal of English for Academic Purposes, 5(2), 97–116. https://doi.org/10.1016/j.jeap.2006.05.001 Martin, J. R., & White, P. R. R. (2005). The language of evaluation: Appraisal in English. Palgrave Macmillan.

• Poldvere, Nele; Kibisova, Elizaveta & Alvestad, Silje Susanne (2024). Fakespeak (linguistics).
• Kibisova, Elizaveta & Poldvere, Nele (2023). Stance, fake news and corpora: A contrastive study of English, Norwegian and Russian.
• Seitanidi, Eleni; Poldvere, Nele & Paradis, Carita (2023). ’That’s everything we need’: Everything-cleft constructions in spoken British English. Show summary

  This study examines the use of everything in cleft constructions in spoken British English (everything that we do is evidence-based). It investigates if everything in such constructions, in addition to having a quantificational function, can also create the smallness and the exhaustive meanings which have been considered characteristic of all- rather than everything-cleft constructions (Tellings, 2020, p. 6). The ‘smallness’ contribution of all expresses the idea that what is being discussed is ‘not much’, whereas its exhaustive reading expresses the meaning of something being ‘no more than’, e.g., all (that) John ate for lunch was a banana (Homer, 2019, p. 1; AUTHORS, forthcoming). We use the Spoken British National Corpus (BNC) 1994 (BNC Consortium, 2007), the Spoken BNC2014 (Love et al., 2017), London—Lund Corpus 1 (Greenbaum & Svartvik, 1990) and London—Lund Corpus 2 (AUTHORS, 2021) totalling 22.9 million words to address two questions: 1. How frequent are everything-cleft constructions in spoken British English? 2. What are the formal and semantic properties of everything-cleft constructions? We approach these questions from a Cognitive Linguistics perspective (e.g., Langacker, 2016) focusing on the construction’s contexts of use. We combine quantitative and qualitative analysis by analysing the frequency of occurrence of everything-clefts, their constructional variants and pragmatic functions. Fifty-five everything-cleft constructions occurred in our data. Regular everything-clefts (everything I do is right) were the most frequent (71%—39 instances), followed by demonstrative/ deictic everything-clefts (that’s everything we need) (24%—13 instances), and reversed everything-clefts (the bibliography is everything you’ve used to influence you) (5%—3 instances). In the regular everything-cleft constructions, which often contain evaluations, everything has a quantificational function meaning ‘all the things discussed in the everything clause’. However, when it occurs after the copula, everything can create the exhaustivity reading which is associated with clefts and all-clefts, but not the smallness reading. Instead, everything can at times create a sufficiency reading whereby what is discussed is ‘enough’. References AUTHORS (forthcoming) AUTHORS (2021) BNC Consortium. (2007). The British National Corpus, XML Edition. Oxford Text Archive. http://hdl.handle.net/20.500.12024/2554 Greenbaum, S., & Svartvik, J. (1990). The London–Lund Corpus of Spoken English. In J. Svartvik (Ed.), The London-Lund corpus of spoken English: Description and research (pp. 11–59). Lund University Press. Homer, V. (2019). That’s all. In R. Stockwell, M. O’Leary, Z. Xu, & Z. L. Zhou (Eds.), Proceedings of the 36th West Coast Conference on Formal Linguistics (pp. 1–21). Cascadilla Proceedings Project. Langacker, R. W. (2016). Nominal grounding and English quantifiers. 3(1), 1–31. BDSL (Bibliographie der Deutschen Sprach-und Literaturwissenschaft). Love, R., Dembry, C., Hardie, A., Brezina, V., & McEnery, T. (2017). The Spoken BNC2014: Designing and building a spoken corpus of everyday conversations. International Journal of Corpus Linguistics, 22(3), 319–344. https://doi.org/10.1075/ijcl.22.3.02lov Tellings, J. (2020). An analysis of all-clefts. Glossa: A Journal of General Linguistics, 5(1), 111. https://doi.org/10.5334/gjgl.1092

• Poldvere, Nele & Paradis, Carita (2023). Spoken English. [Internet]. CorpusCast podcast.
• Poldvere, Nele; Kibisova, Elizaveta & Alvestad, Silje Susanne (2023). Stance, fake news and corpora: A contrastive study of English, Norwegian and Russian. Show summary

  In this work-in-progress presentation, we outline the main aims and steps of implementation of a corpus-based analysis of stance in fake news in English, Norwegian and Russian. In our previous study, we found significant differences in Appraisal expressions between fake and genuine news in English (Trnavac & Põldvere, 2023), but it is unclear how these differences are reflected across languages and cultures. Based on Biber’s (2006) stance framework, we set out to compare and contrast the use and distribution of a wide range of lexico-grammatical features of stance across English in the US, Norwegian in Norway and Russian in Russia. To the best of our knowledge, the framework has not been applied to fake news research before, nor are we aware of any extensions of the framework to Norwegian and Russian to facilitate cross-linguistic comparisons. The data are from three corpora of fake and genuine news, collected from fact-checking websites covering news in the respective languages. In this presentation, we focus on and seek feedback on the first steps of the analysis: (i) the development of the stance frameworks in all three languages and (ii) the design and compilation of comparable corpora in terms of size and genre. We also present the outputs of preliminary semi-automatic searches of the stance expressions in the corpora, which give interesting first insights into how stance-taking works in fake news in English, Norwegian and Russian. References Biber, D. (2006). Stance in spoken and written university registers. Journal of English for Academic Purposes, 5, 97–116. Trnavac, R., & Põldvere, N. (2023). Investigating Appraisal and the language of evaluation in fake news corpora [Manuscript under review].

• Poldvere, Nele; Kibisova, Elizaveta; Alvestad, Silje Susanne & Trnavac, Radoslava (2023). Fake news around the world: A corpus-based analysis of stance in fake news in English, Norwegian and Russian. Show summary

  This talk presents a corpus-based analysis of stance in fake news in English, Norwegian and Russian. In our previous research, we found significant differences in Appraisal expressions between fake and genuine news in English (Trnavac & Põldvere, 2023), but it is unclear how these differences are reflected across languages and cultures. Based on Biber’s (2006) stance framework, we set out to compare and contrast the use and distribution of a wide range of lexico-grammatical features of stance across English in the US, Norwegian in Norway and Russian in Russia. The lexico-grammatical features are divided into major structural categories with meanings related to epistemic stance, evidentiality, attitude, and more (e.g., I believe, evidently, it is surprising that, and equivalent constructions in other languages). To the best of our knowledge, the stance framework has not been used in fake news research before, nor are we aware of any extensions of the framework to Norwegian and Russian to facilitate cross-linguistic comparisons. The data are from three comparable corpora of fake and genuine news, collected from fact-checking websites covering news in the respective languages. For Russian, for example, we relied on websites operating in countries neighbouring Russia, taking into account the recent implementation of laws in Russia which have considerably limited freedom of expression in the country. The corpora are comparable in terms of size and the fact that the fake news stories are mainly from news websites, blogs and social media posts. Based on previous cross-linguistic research on stance in news discourse (Marin-Arrese, 2015), we expect there to be considerable differences in the way in which stance is expressed in fake news in the three languages due to general differences in argumentative and persuasive style, as well as specific cultural characteristics associated with each country (e.g., high levels of state propaganda in Russia). References Biber, D. (2006). Stance in spoken and written university registers. Journal of English for Academic Purposes, 5, 97–116. Marin-Arrese, J. I. (2015). Epistemicity and stance: A cross-linguistic study of epistemic stance strategies in journalistic discourse in English and Spanish. Discourse Studies, 17(2), 210–225. Trnavac, R., & Põldvere, N. (2023). Investigating Appraisal and the language of evaluation in fake news corpora [Manuscript in preparation].

• Trnavac, Radoslava & Poldvere, Nele (2023). Is fake news more evaluative? Comparing Appraisal expressions across fake and genuine news in English. Show summary

  Fake news has become an important topic of research in a variety of disciplines including corpus linguistics. The main strength of corpus approaches to fake news is that we have access to carefully designed datasets of both fake and genuine news, which can be used to identify salient linguistic features of fake news (Asr and Taboada, 2019). Moreover, the choice of the linguistic features is grounded in existing linguistic theory, which can inform our analyses in important ways (Grieve and Woodfield, forthcoming). The goal of this paper is to compare the use and distribution of evaluative features across fake and genuine news in English using qualitative and quantitative corpus linguistic techniques. The qualitative analysis, grounded in Appraisal Theory—ATTITUDE, ENGAGEMENT and GRADUATION (Martin and White, 2005)—sheds new light on how and why evaluative meanings are construed differently in fake news compared to genuine news. The Appraisal expressions were identified manually in the corpus texts, followed by classification into the Appraisal categories and their subcategories (e.g., CONTRACTIVE vs. EXPANSIVE functions of ENGAGEMENT). The quantitative analysis involved the implementation of a chi-square test to measure the association between the evaluative features, on the one hand, and fake and genuine news, on the other. The corpora were extracted from two sources. The first corpus is based on the writings of The New York Times journalist Jayson Blair, who was fired from the newspaper for fabricating news stories in the early 2000s (Grieve and Woodfield, forthcoming). The corpus contains approximately 56,000 words, with samples of both fake and genuine news. The second corpus is based on the writings of seven other journalists who have been identified by various fact-checking services to have written both fake and genuine news. This corpus contains approximately 13,000 words from online news websites. On the one hand, the corpora are similar in the sense that both are controlled for well-known confounding variables such as authorship. On the other hand, they differ in terms of genre (mainstream newspaper vs. alternative news websites) and the journalists’ motivation to lie. While Blair’s motivation was personal, the other journalists’ motivation was more ideological. In this way, the corpora contain different types of fake news, which may or may not lead to differences in terms of evaluation. The preliminary results showed interesting differences between the two corpora, and between fake and genuine news. So, is fake news more evaluative? This is certainly the case in the Jayson Blair corpus. Moreover, Blair’s fake news is characterized by a greater use of ATTITUDE, which might be due to his lack of access to basic information about the news events (the who, what, when and where of journalism), since he never attended them. These results are slightly different in the other corpus, where fake news is also more evaluative, but where the key Appraisal category is ENGAGEMENT instead. In particular, the journalists seem to draw heavily on the CONTRACTIVE function of ENGAGEMENT to fend of alternative viewpoints and to present their ideology as ‘true’ and ‘valid’. Thus, evaluation provides important cues for fake news detection, but more work is required to understand the effect of genre and motivation on the linguistic outcomes. References Fatima Asr and Maite Taboada. 2019. Big Data and quality data for fake news and misinformation detection. Big Data & Society. https://doi.org/10.1177/2053951719843310 Jack Grieve and Helena Woodfield. forthcoming. The language of fake news. Cambridge Elements in Forensic Linguistics. Cambridge University Press. Jim Martin and Peter White. 2005. The language of evaluation: Appraisal in English. Palgrave Macmillan.

• Poldvere, Nele; Johansson, Victoria & Paradis, Carita (2023). Why do we use the same words when we disagree with each other? An experimental approach to interactive alignment. Show summary

  Interactive alignment, or the reuse of ideas and constructions from prior speaker turns, is a compelling type of meaning negotiation, dialogue management and stance coordination in everyday face-to-face conversation. Du Bois (2014) argues that engaging in this way with prior speaker contributions is a socially motivated phenomenon that speakers use for various communicative purposes. Garrod and Pickering (2004) consider alignment to be a largely automatic cognitive process that makes conversation smooth and easy. In a previous corpus study (Põldvere, Johansson & Paradis, 2021), we brought these two approaches together to investigate why and when speakers make use of each other’s contributions in conversation, and what the social and cognitive underpinnings of this behaviour might be. The study showed that (i) speakers tend to align with each other when they disagree, and (ii) the disagreements are produced very quickly (with no-gap-no-overlap). Both of these results point to the mitigating function of alignment in disagreement situations. Using the overhearer paradigm (Bögels & Levinson, 2017), we investigate how participants perceive interactive alignment in disagreement situations in a controlled laboratory setting. This experimental study concerns the social processes underpinning the interpretation of alignment by drawing on insights from Du Bois’ notion of dialogic engagement and Garrod’s and Pickering’s observations of turn transitions. From the perspective of an overhearer of a conversation, does alignment influence the overhearer’s judgments of social closeness among speakers in disagreement situations? Are there differences between the different types of alignment? What role does timing play in the overhearer’s reactions? In the experiment, the participants listen to pre-recorded conversations between two people who disagree with each other. The stimuli were recorded with two actors who spoke with a British English accent. The manipulations involved the presence or absence of alignment, the extent of the alignment (formal vs. semantic) and the timing of the disagreement (no-gap-no-overlap vs. a gap of one second). The participants in the experiment are native speakers of British English. After listening to the recordings, the participants rate their agreement with two statements on a seven-point Likert scale, designed to elicit their intuitions about social closeness among speakers in the recordings (I feel that these people share similar views, these people seem to be comfortable with each other). Contrary to expectation, though, alignment, and particularly formal alignment, received lower ratings of social closeness in the pilot study than non-alignment. This result was modulated by time whereby alignment expressed after a no-gap-no-overlap was rated higher than non-alignment that was expressed after a gap. References Põldvere, N., Johansson, V., & Paradis, C. (2021). Resonance in dialogue: The interplay between intersubjective motivations and cognitive facilitation. Language and Cognition, 13, 643–669. Bögels, S., & Levinson, S. C. (2017). The brain behind the response: Insights into turn-taking in conversation from neuroimaging. Research on Language and Social Interaction, 50(1), 71–89. Du Bois, J. W. (2014). Towards a dialogic syntax. Cognitive Linguistics, 25(3), 359–410. Garrod, S., & Pickering, M. J. (2004). Why is conversation so easy? TRENDS in Cognitive Sciences, 8(1), 8–11.

• Poldvere, Nele (2023). Is there a language of fake news? Some answers from the Fakespeak project. Show summary

  In this talk, I will present some findings from a large project on the language of fake news, called Fakespeak. The goal of the project is to find out whether or not fake news is characterised by a distinct set of linguistic features that would allow us to speak of a language of fake news. Research on this topic has gained much ground in other disciplines, such as computer science, but in linguistics, it is still quite new. Grieve and Woodfield (forthcoming) have taken important steps towards establishing a linguistically driven approach to fake news, one that is grounded in existing linguistic theory and where the data are of high quality, i.e., controlled for a number of variables. In the Fakespeak project, we add to their research by drawing on the linguistic theories of Conceptual Metaphor Theory (Lakoff & Johnson, 1980) and Appraisal Theory (Martin & White, 2005) to determine the defining metaphorical and evaluative features of fake news, respectively. The data are of different kinds. We make use of the Jayson Blair corpus used in Grieve and Woodfield, but we have also built new corpora, both large and small, to be able to generalise our results to different types of fake news. References Grieve, J., & Woodfield, H. (forthcoming). The language of fake news. Cambridge Elements in Forensic Linguistics. Cambridge University Press. Lakoff, G., & Johnson, M. (1980). Metaphors we live by. Chicago University Press. Martin, J. R., & White, P. R. R. (2005). The language of evaluation: Appraisal in English. Palgrave Macmillan.

• Poldvere, Nele (2022). Fake news: Linguistic, legal and computational perspectives.
• Poldvere, Nele (2022). Metaphor and evaluation in fake news in English: Some data, some results.
• Poldvere, Nele (2022). Metaphor in fake and genuine news: Same or different?
• Seitanidi, Eleni; Poldvere, Nele & Paradis, Carita (2022). All-cleft constructions in the London–Lund Corpus 2 (LLC–2) spoken British English. Show summary

  This study contributes to the hitherto limited research on all-cleft constructions (all you have to do is flip it). All-cleft constructions comprise two clauses connected with the copula be, where one clause is introduced by all, which is synonymous with ‘only’ (Traugott, 2008). The construction often conveys an assertion of the scarcity or inadequacy of what is discussed (Homer, 2019; Tellings, 2020). Given the construction’s evaluative character (Traugott, 2008: 9), we argue that speakers use the ‘not much’ meaning of the construction to construe the action modified by the modal (when there is one) and the main verb of the all-clause as either (i) positive, e.g. a request by the speaker as causing minimal inconvenience (all I need to know is do you want), or (ii) negative, e.g. the action taken by the addressee or someone else as being inadequate (all he has to do is forward an email to the solicitor and just hasn't). We use data from the new London-Lund Corpus (LLC–2) (Põldvere et al., 2021) of spoken British English, and we adapt Biber at al.’s (1999) classification of wh-cleft constructions and apply it to all-clefts, categorizing all-cleft variants into (1) regular (all I did is I wrote to them), (2) reversed (hydrogen is all I have available), and (3) demonstrative (be an adult that's all I'm asking you to do) all-cleft constructions. Adopting a Construction Grammar approach, we account for the (1) formal features, i.e., collocational patterns involving modal and main verbs, and intonation placement and contours, (2) meanings and (3) discourse functions of the all-cleft construction addressing the following research questions: 1. What are the form-meaning characteristics of the construction? 2. What are the communicative functions of the construction? 3. How do the communicative functions of the construction vary by discourse context, e.g. spontaneous commentary? We combine quantitative and qualitative methods using concordances and word lists to study the formal features of the construction. We conduct close analyses of the texts to study the construction’s communicative functions in each discourse context. Furthermore, the availability of the corpus audio files allows us to study the intonational patterns of the all-cleft constructions using specialised software. Some preliminary findings suggest that proper all-clefts are the most frequent, followed by demonstrative all-clefts, and reversed all-clefts. With regard to the main verbs found in the construction, the majority express ACTION, i.e. do, followed by MENTAL STATES, i.e. need, want, take, then SPEAKING, i.e. say, and MENTAL ACTIVITY, e.g. think, know. In terms of the distribution of the construction across the discourse contexts of the corpus, preliminary findings suggest that the construction most frequently occurs in spontaneous commentary, followed by prepared speech and distanced conversations, face-to-face conversation, and, finally, legal proceedings. The distribution may be due to the fact that spontaneous commentary comprises monologues from cooking and science demonstrations where speakers often use the construction to explain the procedures and present them as straightforward.

• Seitanidi, Eleni; Poldvere, Nele & Paradis, Carita (2022). The focalizing all-cleft construction in the London–Lund Corpora of spoken British English. Show summary

  Adopting a Construction Grammar approach, we examine the focalizing all-cleft construction in spoken British English in London–Lund Corpus 1 (LLC–1) (Greenbaum & Svartvik, 1990) and London–Lund Corpus 2 (LLC–2) (Põldvere et al., 2021). All-cleft constructions comprise two clauses connected by the copula be with one clause introduced by all, which means ‘only’ (Traugott, 2008) and which contributes to the construction’s ‘below expectation’ reading (Kay, 2013: 6). We aim to contribute to the scant research on all-cleft constructions (Traugott, 2008: 9) by determining their formal and semantic features and their discourse functions thereby highlighting their development over the past half a century. Following Biber at al.’s (1999) classification of cleft constructions, all-cleft constructions are grouped into (1) proper (all I was going to say was I don`t know), (2) reversed (hydrogen is all I have available), and (3) demonstrative (that`s all I know) all-cleft constructions. In order to identify the discourse functions served by the construction, we consider the discourse context, e.g., spontaneous commentary, in which the construction is used and the relationship between the discourse participants, e.g., equals/disparates, combining quantitative and qualitative analyses, e.g., frequency word lists and collocations. 80 Preliminary findings suggest that all-cleft constructions are used more in LLC–2 monologues and dialogues. In both corpora, proper all-clefts are the most frequent, followed by demonstrative all-clefts. Regardless of power relations, LLC–2 speakers seem to use all-cleft constructions more for explicitly criticising the addressee or another person, e.g., all he has to do is forward an email to the solicitor and just hasn’t, whereas LLC–1 speakers express their criticism less directly, e.g., if that`s all we can offer yes. Given the construction’s evaluative character (Traugott, 2008: 9), the process underlying the development of the construction seems to be subjectification. References Biber, D., Johansson, S., Leech, G., Conrad, S. and Finegan, E. (1999) Longman Grammar of Spoken and Written English, Harlow: Pearson Education. Greenbaum, S. and Svartvik, J. (1990) The London–Lund Corpus of Spoken English, in J. Svartvik (ed.), The London–Lund Corpus of Spoken English: Description and Research, pp:11–59, Lund: Lund University Press. Kay, P. (2013) The Limits of Construction Grammar, in G. Trousdale and T. Hoffmann (eds.) The Oxford Handbook of Construction Grammar, New York: Oxford University Press. Põldvere, N., Johansson, V., and Paradis, C. (2021) On the London–Lund Corpus 2: Design, challenges and innovations, English Language and Linguistics, 25(3), 459–483. https://doi.org/10.1017/S1360674321000186 Traugott, E., C. (2008) “All that he endeavored to prove was…”: On the emergence of grammatical constructions in dialogual and dialogic contexts, in R. Cooper and R. Kempson (eds.) Language in Flux: Dialogue Coordination, Language Variation, Change and Evolution, pp: 143–177, London: Kings College Publications.

• Nykiel, Joanna & Poldvere, Nele (2022). Ellipsis in a less well known environment: The reactive what-x construction in English. Show summary

  In this talk, we present, and suggest a theoretical analysis of, what is known as the reactive what-X construction, depicted in (1)-(2). This construction has interrogative and clarificational, as well as anaphoric properties (i.e., it requires an antecedent in the preceding context), as has been shown by recent corpus work on spoken British English (Põldvere & Paradis 2019, 2020). We begin by summarizing the syntactic properties and the main functions of the reactive what-Xconstruction as requests for information and requests for verification, based on data drawn from the London-Lund Corpora. In the second part of this talk, we address patterns that haven’t been noticed before, i.e., that this construction also has properties aligning it with certain kinds of elliptical constructions, in particular, with fragments (see (2)). We next pursue a theoretical analysis that explicitly captures the elliptical properties of the reactive what-X construction. We do so by taking as our starting point the Construction Grammar analysis of this construction offered in Põldvere & Paradis (2020) and relating the construction to reprise fragments, shown in (3). This allows us to analyze the reactive what-X construction as a type of in-situ interrogative clause whose elliptical properties are licensed by a constraint that permits a fragment to be the sole daughter of an S node (see Ginzburg & Sag 2000, Nykiel & Kim 2021). (1) A: She was like, it would've been gross to put it in my car and leave it there for a whole day, and I was like, yeah it's gross to just pick it up anyway. It's got its head coming off, Jesus. B: What she would eat it? (2) A: Have you ever been out there before? B: What to Russia? (3) A: This is it for Harvey. B: Harvey?

• Poldvere, Nele & Saric, Ljiljana (2022). A study of metaphorical cues of deception in fake news. Show summary

  Fake news is an age-old problem, but it is only with the growth of digital communication in recent years that is has become a serious threat to national security. However, current language-based approaches to fake news detection have focused mainly on their topical trends rather than on the linguistic explanations for how and why the language of fake news differs from that of real news. This study seeks to fill this gap by drawing on Conceptual Metaphor Theory (Lakoff & Johnson, 1980) to (i) identify differences in metaphorical language use between fake and real news, and to (ii) provide tentative explanations for the occurrence of metaphorical cues of deception in fake news. Previous research has found metaphor to be a useful resource for describing personal experiences (Gibbs & Franks, 2002) and so it is reasonable to assume that this applies to lived experiences, too. Therefore, we expect fake news to be characterised by a more limited use of metaphor. The data are from the Jayson Blair corpus (Grieve & Woodfield, forthcoming), a collection of 36 fake news and 28 real news articles by the former New York Times journalist Jayson Blair. In the fake news articles, Blair lied about being on site and fabricated comments, among other things, which required him to write about events and places that he did not experience. The methodological approach is a combination of Cameron’s (2003) and Pragglejaz Group’s (2007) procedures. The texts were read in full to identify the full range of metaphorical expressions, which were then classified into source domains. The findings provide support for the above expectation. Moreover, the fake news articles contain more conventional domains of experience such as events are objects, as opposed to real news, which favours more creative instantiations of the war and sports domains. References Cameron, L. (2003). Metaphor in educational discourse. Continuum. Gibbs, R. W. Jr., & Franks, H. (2002). Embodied metaphor in women’s narratives about their experiences with cancer. Health Communication, 14(2), 139–165. Grieve, J., & Woodfield, H. (forthcoming). The language of fake news. Cambridge Elements in Applied Linguistics. Cambridge University Press. Lakoff, G., & Johnson, M. (1980). Metaphors we live by. The University of Chicago Press. Pragglejaz Group. (2007). MIP: A method for identifying metaphorically used words in discourse. Metaphor and Symbol 22(1), 1–39.

• Poldvere, Nele (2022). On the importance of dialogue and dialogicity in language change. Show summary

  This talk draws on my work on dialogic aspects of language change as observed in large collections of contemporary English dialogues. Dialogicity is an important aspect of human sense-making, with each utterance being linked to prior or successive discourse in one way or another; however, some constructions have become so tied up with the surrounding discourse that they have incorporated these dialogic properties into their constructional representation. One such construction is the reactive what-x construction, illustrated in (1). (1) A: have you ever been out there before B: what to Russia A: mm B: no The reactive what-x construction has interesting form–meaning properties: it comprises what and a subsequent phrasal or clausal complement x, which, importantly, always form one and the same tone unit. The dialogic properties manifest themselves in several ways such as the sequential placement of the construction in the surrounding discourse (the construction is responsive and always follows an immediately preceding turn by another speaker) and its meaning potential (i.e., to react to the preceding turn in order to negotiate and call it into question). In addition, the dialogic embedding of the construction in different discourse contexts has given rise to several dialogic functions, of which requests for verification is only one (the example above). In this talk, I explore the diachronic processes by which the construction has acquired these properties. In doing so, I propose an extension to Diachronic Construction Grammar to go beyond form–meaning properties in the strict lexical–semantic sense and into the domains of dialogicity, prosody and social interaction. In addition, I present a corpus of contemporary English dialogues, namely, the London–Lund Corpus 2 of spoken British English, which together with the first London–Lund Corpus provide an excellent resource for the study of short-term language change.

• Poldvere, Nele & Saric, Ljiljana (2022). How (not) to get away with deception: Metaphorical cues of deception in fake news. Show summary

  Jayson Blair was fired from The New York Times in the early 2000s for writing news articles that contained intentionally false information, such as not being in the location where the news was reported. Grieve and Woodfield (forthcoming) have even noted that, in many ways, Blair “ushered in the modern era of fake news” (p. 31), a serious threat to today’s democracies. Moreover, the authors found Blair’s fake news articles to be grammatically distinct from his real news articles, a finding that contributes to the growing field of the language of fake news. This study explores further features of this genre through the study of metaphorical language in Blair’s articles. Metaphor (e.g., COVID-19 comes in waves) has been found to vary considerably depending on the discourse community to which the speaker/writer belongs, and conceptual metaphor, in particular (e.g., disease is a natural disaster), reveals a fundamental part of the way people think, reason and imagine (Gibbs, 2017). In this study, we explore the hypothesis that metaphor is more limited in Blair’s fake news articles compared to his real news articles due to the lack of perceptual information in such contexts. In other words, it might be difficult to recount real-life stories without being able to rely on one’s embodied experiences, such as having been in a place that one is writing about. The results of this study can have important implications, not only for the study of the language and psychology of fake news, but also for deception more generally. References Gibbs, R. W. Jr. (2017). Metaphor wars: Conceptual metaphors in human life. Cambridge University Press. Grieve, J., & Woodfield, H. (forthcoming). The language of fake news. Cambridge Elements in Applied Linguistics. Cambridge University Press.

• Paradis, Carita; Johansson, Victoria & Põldvere, Nele (2021). Special issue on spoken language in time and across time: Introduction. English Language and Linguistics. ISSN 1360-6743. 25(3), p. 449–457. doi: 10.1017/S1360674321000174. Show summary

  The idea of this special issue on Spoken language in time and across time emerged at an international symposium on this topic that we organised at Lund University on 20 September 2019. The purpose of the symposium was to celebrate important past and present achievements of spoken language research as well as past and present corpora available for such research. Some speakers reported on academic and technical advances from the past, while others offered information about state-of-the-art research on spoken language and spoken corpus compilation. Our idea with the symposium was also to bring together early career scholars, somewhat more senior scholars as well as senior scholars – the latter actually active when interest in spoken language and spoken corpus compilation was in its infancy. The type of spoken corpora in focus extended from the world's first publicly available, machine-readable spoken corpus, The London–Lund Corpus of Spoken English (Svartvik 1990), nowadays referred to as LLC–1, through to the spoken parts of The British National Corpora (BNC) from 1994 (BNC Consortium 2007) and 2014 (Love et al. 2017), The Diachronic Corpus of Present-Day Spoken English (DCPSE) consisting of LLC–1 and the British component of The International Corpus of English (ICE–GB), Santa Barbara Corpus of Spoken American English (SBCSAE) (Du Bois et al. 2000–5), The Corpus of Contemporary American English (COCA) (Davies 2008–) and finally the most recent one, The London–Lund Corpus 2 (LLC–2) (Põldvere, Johansson & Paradis 2021a). The symposium thus covered approximately half a century of data from publicly available corpora compiled for multipurpose use by the academic community for research on spoken English in different contexts.

• Põldvere, Nele (2021). Advice in conversation: A mixed-methods approach to corpus pragmatics.
• Põldvere, Nele (2021). Spoken English in time and across time: Constructions, context, corpora. Show summary

  Spoken dialogue is the most common use of language, but it is also incredibly complex and dynamic. It puts on full display the intricate ways in which speakers coordinate their contributions to make sense of the world and negotiate social relations with each other. A fruitful method for studying spoken dialogue is to consult language corpora based on spoken, conversational data. However, the shortage of such corpora has long been an obstacle. This talk brings together the main findings of my recent work on spoken English both from a synchronic and a diachronic perspective. It provides a novel and empirically grounded account of the dynamic negotiation of meaning in spoken dialogue including the constructional properties and socio-cognitive processes that play a role. It also reports on the compilation of a new corpus of spoken English, the London–Lund Corpus 2, designed according to the same principles as the first London–Lund Corpus with data from the 1950s to 1980s. In this way, the corpus allows not only for synchronic investigations of contemporary speech but also for principled diachronic research of spoken language across time.

• Põldvere, Nele (2021). A mixed-methods and mixed-models approach to spoken English. Show summary

  In this talk, I will mainly present work related to my PhD thesis, which I defended at Lund University in September 2019. The thesis was concerned with developing a mixed-methods and mixed-models approach to spoken English. More precisely, it aimed to further our understanding of the use and development of constructions in spoken dialogue, and to propose a dynamic and socio-cognitive description and explanation of dialogic meaning-making. The work fell within the broad framework of usage-based Cognitive–Functional Linguistics, which adopts a contextualised conception of meaning in language. However, not even cognitive and functional linguistic approaches to language have been immune to the written language bias evident in the language sciences more generally, and even if they have generated investigations that embrace the idiosyncrasies of spoken dialogue, these investigations have emphasised either the cognitive or interactive dimensions of speech, largely ignoring the interaction between them. Also, we have limited knowledge of the kinds of properties that influence constructional meaning and the underlying processes that govern the emergence, interpretation and development of constructions in spoken dialogue. These are exactly the kinds of issues that I investigated in my thesis, by drawing on a range of constructions, cognitive–functional approaches and methodological techniques. One of the methodological innovations of the thesis was the compilation and use of the London–Lund Corpus 2, a new half-a-million-word corpus of spoken British English. In the talk, I will explain how and why the corpus was compiled as well as how one can gain access to it. I will end the talk by giving a brief overview of my current work on the language of fake news within the Fakespeak project in the Department of Literature, Area Studies and European Languages.

• Põldvere, Nele (2021). Tracking change in epistemic stance constructions expressing advice in the London–Lund Corpora. Show summary

  Epistemic stance constructions such as modal verbs are commonly used to give advice in everyday conversation (e.g., you could do it). They express uncertainty about the validity of the future event, and they are also weakly deontic, making them suitable for countering the negative social consequences associated with advice-giving. This paper investigates recent change of epistemic modal constructions expressing advice in English conversation based on data from the London–Lund Corpora from the 1950s up until the present day. It seeks to determine how the use of the constructions has changed over the last half-century, and the socio-communicative factors that may have played a role. Pervasive in human encounters, advice-giving is at the heart of how we organize social relations with each other. However, telling another person what to do is a notoriously sensitive undertaking, meaning that the way we express ourselves is of crucial importance. A common way to give advice in English is to use epistemic modal constructions such as could (Couper-Kuhlen & Thompson, 2017). These constructions are less direct than some of the other constructions (e.g., imperatives; do it), and they also express a lower degree of epistemic control and certainty about the prospective event (cf. Paradis, in press). Leech (2003) reports that modal verbs such as can and could are on the rise in Present-Day English, possibly due to the growing tendency for speakers to avoid constructions that assert power and authority over the addressee. However, more research is needed to determine if this explanation applies to different socio-communicative functions of modal verbs such as advice-giving. The data for the analysis come from everyday conversation in two comparable corpora of Present-Day spoken British English, London¬–Lund Corpus 1 (LLC–1) with data from the 1950s–1980s and the brand new London–Lund Corpus 2 (LLC–2) from 2014–2019. In order to determine the relative frequency of epistemic modal constructions (e.g., can/could, may/might, will/would, and their negative counterparts) in the corpora, I extract the full range of constructions expressing advice, using text-processing tools such as part-of-speech taggers, parsers and speech-act taggers. I also annotate the constructions for several formal and functional factors (e.g., modal verb, agentivity, speaker commitment). A preliminary comparison of epistemic modal constructions expressing advice in the two corpora suggests that the use of the constructions has increased from LLC–1 to LLC–2. This increase seems to mirror the general trend where speakers in the later corpus are considerably more likely to share their thoughts about what other people should (or should not) do than speakers in the earlier corpus. A closer look at the individual occurrences of the epistemic modal constructions reveals a shift to advice expressed in more indirect ways, involving the modal verb could (rather than can), the inclusive we (e.g., we could do it) even if the act is to be carried out by the addressee only, and the lack of intensifiers. This suggests that, in English conversation, there is an increased preference for epistemic modal constructions that make the act of giving advice less explicit and that warrant a high degree of caution and tact. References Couper-Kuhlen, E., & Thompson, S. A. (2017, July). Linguistic formats for suggesting and advising in everyday talk. Paper presented at IPrA2017, Belfast, UK. Leech, G. (2003). Modality on the move: The English modal auxiliaries 1961–1992. In R. Facchinetti, M. Krug & F. Palmer (Eds.), Modality in Contemporary English (pp. 223–240). Berlin: Mouton de Gruyter. Paradis, C. (in press). Two layers of modal grounding in recommendations.

• Põldvere, Nele (2021). Using the London–Lund Corpora to investigate recent change in advice-giving in spoken English. Show summary

  Several new spoken corpora have seen the light of day in recent years. Few, however, are comparable to a corpus from an earlier period to facilitate diachronic investigations of spoken language. The new London–Lund Corpus 2 (LLC–2) of spoken British English, collected 2014–2019, fills this gap. The size and design of LLC–2 are comparable to those of the world’s first spoken corpus, the London–Lund Corpus (LLC–1) with data mainly from the 1960s–1970s, which means that, together, the two corpora allow us to make principled diachronic comparisons of speech over the past 50 years. This paper has two aims: (i) to critically examine the methodological aspects of the London–Lund Corpora as a diachronic resource, and (ii) to demonstrate the usefulness of the resource based on a case study of the development of constructions expressing advice in spontaneous conversation. One of the most important challenges of the compilation of LLC–2 was to capture recordings of conversations that are comparable to those in LLC–1 in terms of the situational contexts in which they occur. In order to overcome this challenge, a large number of the conversations in LLC–2 were recorded at the University College London, which was also the main site of recording for LLC–1. This gave us speech situations that revolve around similar subject matters (e.g., common-room conversations, tutorials), and it also ensured that the speakers in LLC–2 share many of the same characteristics with those in LLC–1. The usefulness of the London–Lund Corpora as a diachronic resource is demonstrated in a case study of the development of constructions expressing advice in spontaneous conversation. Advice is a pervasive feature of human communication, but it is also a notoriously sensitive undertaking where language plays a crucial role. Advice can be expressed in a variety of ways, ranging from more deontic (e.g., imperatives; do it!) to less deontic constructions (e.g., various types of modals; I would do it). Based on a quantitative analysis of a large number of constructions expressing advice in the London–Lund Corpora, this case study aims to find out whether speakers today prefer or avoid constructions that assert power and authority compared to 50 years ago (Leech, 2003). The preliminary results suggest an increased preference for advice to be expressed in more direct ways. A closer inspection of the distribution of the advice constructions shows that there are conversations in both corpora that slightly skew the results; i.e., they occur in advice-heavy contexts that do not exist in the other corpus (e.g., study abroad sessions in LLC–2). However, the overall distribution of the constructions is relatively even, thus suggesting that the London–Lund Corpora are a suitable resource for studying recent change in spoken English. References Leech, G. (2003). Modality on the move: The English modal auxiliaries 1961–1992. In R. Facchinetti, M. Krug & F. Palmer (Eds.), Modality in contemporary English (pp. 223–240). Mouton de Gruyter.

• De Felice, Rachele; Põldvere, Nele & Paradis, Carita (2021). Advice-giving and advice uptake in conversation: A corpus pragmatic study. Show summary

  This talk presents the main findings of a multi-year project on advice in British English (BrE) conversation. Giving and receiving advice are sensitive undertakings since both the formulation of the advice and its uptake can affect the relationship between interlocutors. Despite a number of studies on advice-giving in institutional (e.g. Butler et al., 2009; Heritage & Sefi, 1992) and everyday conversation (e.g. Adolphs, 2008; Shaw et al., 2015), there has been to date no large-scale, systematic corpus study of the pragmatic and dialogic aspects of advice-giving and uptake across different contexts. Moreover, no studies have explored the motivations and mechanisms related to the development of advice constructions over time. Our study used the London–Lund Corpora of spoken BrE, comprising the first London–Lund Corpus (1950s–1980s) and the new London–Lund Corpus 2 (2014-2019), to a) provide a comprehensive overview of the linguistic resources used for advice-giving and uptake in different settings over 50 years; and b) explain how these interact with temporal, social and dialogic factors. We extracted over 2,000 constructions and manually annotated them for formal, dialogic, and pragmatic features including the wording of the advice, its uptake (e.g. acceptance, rejection), the interlocutors’ relationship, etc. Our analysis combined quantitative methods (conditional inference trees and random forests, suitable for this kind of data with a high degree of correlation) and close qualitative reading of example-rich extracts. The preliminary results of this project are promising. Systematic corpus searches revealed that BrE speakers use a much wider range of advice constructions than reported to date, including very indirect constructions (e.g., sentence fragments). This said, there is also an increased preference in recent history for more explicit advice-giving constructions (e.g., imperatives). However, these lead to lower levels of acceptance, particularly with high-status advisers, reinforcing the complex nature of advising in conversation. References Adolphs, S. (2008). Corpus and Context: Investigating Pragmatic Functions in Spoken Discourse. Amsterdam: John Benjamins. Butler, C. W., Danby, S., Emmison, M., & Thorpe, K. (2009). Managing medical advice seeking in calls to Child Health Line. Sociology of Health & Illness, 31(6), 817–834. Heritage, J., & Sefi, S. (1992). Dilemmas of advice: Aspects of the delivery and reception of advice in interactions between health visitors and first-time mothers. In P. Drew & J. Heritage (Eds.), Talk at Work: Interaction in Institutional Settings (pp. 359–417). Cambridge: Cambridge University Press. Shaw, C., Potter, J., & Hepburn, A. (2015). Advice-implicative actions: Using interrogatives and assessments to deliver advice in mundane conversation. Discourse Studies, 17(3), 317–342.

• Nykiel, Joanna & Põldvere, Nele (2021). Ellipsis in a less well-known environment: The reactive what-x construction in English. Show summary

  In this talk, we present, and suggest a theoretical analysis of, what is known as the reactive what-x construction, depicted in (1B). This construction has interrogative and anaphoric properties (i.e., it requires an antecedent in the preceding context), as has been shown by recent corpus work on English (Põldvere & Paradis 2019, 2020). We begin by summarizing this work and data drawn from the London-Lund Corpora, as well as a Construction Grammar analysis offered in Põldvere & Paradis (2020). In the second part of this talk, we address patterns that haven’t been noticed before, i.e., that this construction has properties aligning it with certain kinds of elliptical constructions (e.g., sluicing and fragments). This suggests that we could pursue a theoretical analysis that explicitly captures the elliptical properties of the reactive what-x construction. We outline ways of doing so in a constraint-based framework, and this includes suggesting a multiple inheritance network to license the entire set of the construction’s properties. (1) A: Have you ever been out there before? B: What to Russia? References Põldvere, N., & Paradis, C. (2019). Motivations and mechanisms for the development of the reactive what-x construction in spoken dialogue. Journal of Pragmatics, 143, 65-84. Põldvere, N., & Paradis, C. (2020). ‘What and then a little robot brings it to you?’ The reactive what-x construction in spoken dialogue. English Language and Linguistics, 24(2), 307-332.

• Põldvere, Nele & Alvestad, Silje Susanne (2021). Fakespeak – A pilot study. Show summary

  In this talk, I present the results of a pilot study within the research project “Fakespeak: The language of fake news”. The project involves linguists and computer scientists whose primary objective is to reveal the grammatical and stylistic features of the language of fake news, Fakespeak. The secondary objective is to improve existing fake news detection systems through the automation of the defining linguistic features of Fakespeak. Fake news is defined here as news items that are meant to be deceptive and in which case the author knows that they are false. The pilot study is inspired by a study conducted by Woodfield and Grieve (2019) on Jayson Blair, who was found fabricating news stories for The New York Times. The newspaper tagged the fabricated stories, resulting in two datasets, one of genuine and one of fabricated stories. The linguists subjected the two datasets to Register Analysis, hypothesising that, given the distinct communicative purposes of the stories—to deceive versus to inform—the genuine and the fabricated stories should be grammatically distinct. Indeed, they found that the features of Blair’s genuine stories were similar to those of informationally dense writing, whereas the features of his false stories were similar to those of interactive discourse. I will use the Jayson Blair corpus to extend the analysis to include socio-cognitive features of language, namely metaphor. Based on previous findings that metaphor is limited in interactive spoken discourse (Cameron, 2003), I expect fake news to make use of metaphor to a lesser extent than genuine news, thus providing support for Woodfield and Grieve’s (2019) study. References Cameron, L. (2003). Metaphor in educational discourse. Continuum. Woodfield, H., & Grieve, J. (2019, August 29). Individual differences in fake news: The case of Jayson Blair [Conference session]. British Association of Applied Linguistics Annual Conference, Manchester, UK.

• Põldvere, Nele & Alvestad, Silje Susanne (2021). Is metaphor the key to detecting fake news? First results from the Fakespeak project. Show summary

  This study presents the first results from Fakespeak, a project which defines the linguistic features of fake news in order to improve currently existing detection systems. We explore the possibility that the key to distinguishing between fake and genuine news is metaphor, a resource for talking about one thing (usually complex and abstract) in terms of another (simple and concrete). For instance, talking about viruses as coming in waves reflects the conceptual metaphor disease is a natural disaster, where disease is the ‘target’ domain and natural disaster the ‘source’ domain (Lakoff & Johnson, 1980). Previous research has suggested that metaphor is often used to describe personal experiences, which leads us to believe that genuine news makes use of metaphor to a greater extent than fake news. Moreover, we expect there to be differences in the source domains used. In order to evade issues of comparability, the data are from news written by a single journalist, Jayson Blair, who was caught fabricating stories for The New York Times in the early 2000s (Woodfield & Grieve, 2019). The dataset consists of 38 redacted and 47 non-redacted articles, which were read in full in order to identify the full range of metaphorical expressions. The preliminary results provide support for the expectation that metaphor is more limited in fake news. Moreover, the source domains used in fake news reflect very basic bodily experiences of the world, compared to the diverse range of complex and vivid metaphors used in genuine news (e.g., nature, war, sports). References Lakoff, G., & Johnson, M. (1980). Metaphors we live by. The University of Chicago Press. Woodfield, H., & Grieve, J. (2019, August 29). Individual differences in fake news: The case of Jayson Blair [Conference session]. British Association of Applied Linguistics Annual Conference, Manchester, UK.

View all works in Cristin