1. Introduction and research questions

The study of metaphor in politics spans several fields, including cognitive linguistics (Lakoff, Reference Lakoff2016), critical discourse analysis (Musolff, Reference Musolff2012), political discourse analysis (Dijk, Reference Dijk1997), critical metaphor analysis (Charteris-Black, Reference Charteris-Black and Charteris-Black2004) and cognitive linguistics critical discourse studies (Hart, Reference Hart2018). Political discourse is rich in metaphors, as political topics such as education, economy, environment, health care, and international affairs involve abstract concepts. Researchers find the use of metaphors in discourse about such recent events as Russian–Ukrainian war (Nytspol & Kobuta, Reference Nytspol and Kobuta2022), COVID-19 (Neshkovska & Trajkova, Reference Neshkovska and Trajkova2020), Brexit (Charteris-Black, Reference Charteris-Black2019; Musolff, Reference Musolff2021; Negro Alousque, Reference Negro Alousque2020), climate change (Deignan et al., Reference Deignan, Semino and Paul2019) or general, universal political problems such as corruption (Isaza & Ossewaarde, Reference Isaza and Ossewaarde2021), health care (Ervas et al., Reference Ervas, Rossi, Ojha and Indurkhya2021), elections (D’Angelo & Lombard, Reference D’Angelo and Lombard2008), terror (Lakoff, Reference Lakoff2016), power division in government (Perrez & Reuchamps, Reference Perrez and Reuchamps2015), racism (Asma, Reference Asma1995) and migration (Porto, Reference Porto2022).

These issues however, as well as being political and metaphorical, are emotionally charged. Political topics are emotional, because politicians choose to stir up public opinion with emotion in order to convince the voters to support their arguments. At the same time political topics are metaphorical, because as abstract concepts they form target domains for conceptual metaphors (Lakoff & Johnson, Reference Lakoff and Johnson1999, Reference Lakoff and Johnson2003). We find similar explanations of this interaction in studies of UK and US leaders’ speeches, where metaphor is “central to the creation of persuasive belief systems […] because it exploits the subliminal resources of language by arousing hidden associations that govern our systems of evaluation” (Charteris-Black, Reference Charteris-Black and Charteris-Black2004, p. 2). People differ in decision making if the problem is metaphorically framed in different ways, as it has been shown in an experiment with the framing of violence: as a virus or as a beast (Thibodeau & Boroditsky, Reference Thibodeau and Boroditsky2011). We believe that both frames are emotionally loaded, as they both elicit fear and anger. Possible solutions, combating violence through education reform or by strengthening penalties may also elicit mixed emotions: education reform can trigger positive emotions, whereas strengthening penalties may reassure some and frighten others.

When studying the interaction between metaphors, arguments and emotional appeals in political discourse, however, we shall not rely solely on intuitive qualitative judgements, the way we used (Thibodeau & Boroditsky, Reference Thibodeau and Boroditsky2011) as an illustration. Instead, we will use a replicable method of identifying words that appeal to emotions. We present an analysis of metaphors and emotional appeals in argumentation within political discourse based on systematic methods of annotation: Metaphor Identification Procedure (PRAGGLEJAZ Group, 2007), dictionary-based sentiment analysis (Riegel et al., Reference Riegel, Wierzba, Wypych, Żurawski, Jednoróg, Grabowska and Marchewka2015; Wierzba et al., Reference Wierzba, Riegel, Wypych, Jednorwóg, Turnau, Grabowska and Marchewka2015, Reference Wierzba, Riegel, Kocoń, Miłkowski, Janz, Klessa, Konat, Grimling, Piasecki and Marchewka2021) and argumentative discourse units identification based on Inference Anchoring Theory (IAT) (Reed & Budzynska, Reference Reed and Budzynska2011). Following IAT, we conceptualise an argument as a minimum of two propositions: premise and conclusion, in which one is used by the speaker to provide support for the other. In our previous study we analysed the relation between metaphors and arguments as dynamic discourse phenomena (Juszczyk et al., Reference Juszczyk, Konat and Fabiszak2022). In the current study, we contribute an additional layer of analysis: emotional appeals, that is, cases where speakers attempt to elicit emotional responses in hearers for rhetorical gain. We selected pre-election debates as a case material for studying interactions between metaphors, arguments, and emotional appeals as in the example below:

In example (1) the speaker employs the argument scheme, in which the premise “Polska służba zdrowia jest w stanie przedzawałowym” (“Polish health service is in a heart-attack threatening condition…”) leads to the conclusion that there is a need for the “pakt na rzecz polskiego zdrowia” (“pact for the Polish health”). He uses the metaphor Polish health service is a patient (metaphorical expressions are underlined), which contributes to the rhetorical cohesion, and appeals to the emotions of the hearers through “zdrowie” (“health”) (marked in bold) as eliciting happiness in Polish, following the classification in (Wierzba et al., Reference Wierzba, Riegel, Kocoń, Miłkowski, Janz, Klessa, Konat, Grimling, Piasecki and Marchewka2021). If we were analyzing this example based on our intuitions rather than on automatic identification of emotional appeals, then we would also identify “zawał” (“heart attack”) and “nie przeżyje” (“won’t survive”) as expressions appealing to negative emotions, most likely fear. But the dictionary (Wierzba et al., Reference Wierzba, Riegel, Kocoń, Miłkowski, Janz, Klessa, Konat, Grimling, Piasecki and Marchewka2021), that is, a list of words that were evaluated and coded in the experiment and which is now used as a means for automatic identification of expressions appealing to emotions, does not contain “zawał” (“heart attack”) or “nie przeżyje” (“won’t survive”).Footnote ¹ In terms of argument analysis, the speaker uses the argument from fear appeal (Walton, Reference Walton2000, p. 22), where negative or threatening result is used to justify the need for an action. This example consists of negative premise (“heart attack as a problem“) and positive conclusion (“pact as a solution”).

Example (1) illustrates persuasive strategies used by politicians. In this type of discourse, the speakers will combine metaphorical expressions and emotion-eliciting words into their argumentation with the aim of greater rhetorical effect, as every utterance in pre-election debates can be considered as having persuasive intent. Based on the literature presented above, we may assume high saturation of both metaphors and emotional appeals in political discourse. Hence, the aim of this study is to analyse the frequency and interaction between three discursive phenomena present in pre-election debates: metaphors, arguments, and emotional appeals, using both concepts and methods from three theoretical traditions and combining them into one cohesive methodology.

As shown in Figure 1, there are several possible combinations of discourse phenomena:

1. 1. Texts containing only arguments (i.e., as premise or conclusion), but not metaphors or emotional appeals;

2. 2. Texts containing arguments and metaphors, but not emotional appeals;

3. 3. Texts containing arguments and emotional appeals, but not metaphors;

4. 4. Texts containing arguments, metaphors and emotional appeals;

5. 5. Texts containing only emotional appeals;

6. 6. Texts containing both metaphors and emotional appeals but not arguments;

7. 7. Texts containing only metaphors.

Figure 1. Discourse phenomena under investigation and their combinations.

The aim of this study is to analyse the frequency of the three discursive phenomena: metaphors, arguments, and emotional appeals and the frequency of their combinations. As shown in Figure 1, each of these phenomena may appear by themselves or in combinations of two or three. Our starting point is the identification of arguments, and the interaction of arguments with the remaining two phenomena, therefore we will not consider combinations of metaphors and appeals to emotion that may happen outside argumentative discourse units (this concept is explained in Section 2.2). We propose three hypotheses relating to the co-occurrences of the three phenomena in political discourse:

These hypotheses can be represented by referring to our previous Venn diagram (Figure 1):

H1 predicts that there will be more items in the sum of areas 2 and 4 than in the sum of areas 1 and 3.

H2 predicts that the sum of areas 3 and 4 will be bigger than the sum of areas 1 and 2.

H3 predicts that the number of items in area 4 will be bigger than the sum of all other types of arguments in areas 1, 2 and 3. These hypotheses can be represented on the following scale (see Figure 2).

Figure 2. The scale of frequency of arguments and arguments combined with metaphors and emotional appeals from most frequent to least frequent.

These hypotheses will undergo further specification and operationalization in terms of corpus linguistics methods in Section 3. In Section 2 we present our conceptualization of key notions used in this study: metaphors, arguments and, emotional appeals. Section 4 contains quantitative summary of our results, concluded in Section 5.

2. Key concepts

Metaphors, arguments and emotional appeals: all three concepts analysed in this paper can be conceived as dynamic discourse phenomena. They are realized by the speakers using linguistic means to obtain communicative goals. They are inherently pragmatic and interactive in nature: framing an issue with metaphor, arguing or appealing to emotions makes sense only if there is an interlocutor whom we want to inform, convince, or impress. In this section, we present the theoretical background for the three concepts presented in this paper and propose an initial operationalization of them in a joint research framework.

2.1. Metaphors

Political discussions, specifically political debates, represent a unique form of verbal engagement where journalists pose questions to politicians, allowing them to highlight their perspectives, criticize adversaries, and present their party’s vision for a better world. These debates are intricate communication events because speakers not only respond to the host’s inquiries but also address in-studio attendees and television viewers. Throughout such events, politicians employ metaphors, persuasive arguments and emotional appeals to promote their vision of the current political issues. A comprehensive meta-analysis of 91 studies of metaphorical framing in political discourse has shown that “metaphors can influence individuals’ reasoning through words as well as concepts” (Brugman et al., Reference Brugman, Burgers and Vis2019).

Our perspective stems from discourse dynamics approach to metaphor, where linguistic metaphors are analysed in order to discover what people think and how they think about it. Researchers such as Cameron, Maslen, Low and many others believe that metaphor allows to “reveal something of speakers’ emotions, attitudes and values” (Cameron & Maslen, Reference Cameron and Maslen2010, p. 7). As Cameron stated: “the affective, that is, emotions and feelings that influence human activity (Damasio, Reference Damasio2003), has often been neglected in metaphor studies” (Cameron, Reference Cameron, Cameron and Maslen2010, p. 78).

Metaphors are used in political discourse to simplify political topics and frame social problems (Landau et al., Reference Landau, Robinson and Meier2014). We assume that politicians use metaphors, arguments and emotional appeals to build and present their conceptualization of the world to the audience. In political discourse, speakers use emotion-eliciting words in combination with metaphorical understanding of socio-political situations and arguments aimed at convincing the audience of their solutions to political problems identified in metaphors. Many researchers have explored the link between emotions and metaphors in corpus studies. Emotions expressed with metaphors were studied in the context of reconciliation talk (Cameron, Reference Cameron2012). Researchers (Ogarkova & Salinas, Reference Ogarkova and Salinas2014) and (Reali & Arciniegas, Reference Reali and Arciniegas2014) both highlight the role of the body in shaping emotional experiences, with Ogarkova focusing on the physiological dimension and Reali on the metaphorical, framing of emotions. Studies by (Louw & Milojkovic, Reference Louw and Milojkovic2015) and (Theodoropoulou, Reference Theodoropoulou2012) further analyze the linguistic and experiential aspects of this link, with Louw and Milojkovic emphasizing the role of shared meaning in texts and Theodoropoulou examining the embodied experience of joy and happiness. These studies collectively suggest that emotions are not only shaped by metaphors, but also have a significant impact on the way we use and understand language. A contrastive linguistic analysis of emotion concepts is performed within-corpus and cognitive linguistics (Lewandowska-Tomaszczyk, Reference Lewandowska-Tomaszczyk2019). There are also multilingual studies of emotions in metaphors, such as (Ogarkova & Soriano, Reference Ogarkova and Soriano2014), who investigate the embodied conceptualization of emotions from a cognitive linguistic perspective, focusing on the metaphorical construal of the body and its parts as containers for various types of anger in English, Russian, and Spanish. Additionally, there are reasons to assume that human annotators perceive metaphorical language as more emotional and more abstract than literal language, implying that metaphorical expressions are more emotional and abstract than literal expressions (Piccirilli & Schulte im Walde, Reference Piccirilli and Schulte im Walde2022).

Metaphors as unit of analysis, following MIP (Metaphor Identification Procedure) (PRAGGLEJAZ Group, 2007), are understood here as words whose basic dictionary meaning and contextual meaning differ. Such words are metaphorically used expressions which, according to Conceptual Metaphor Theory (Lakoff, Reference Lakoff1993, Reference Lakoff2016; Lakoff & Johnson, Reference Lakoff and Johnson1999, Reference Lakoff and Johnson2003), are realizations of conceptual metaphors. Moving from the word level to the conceptual level allows researchers to identify systematic mappings between concrete and abstract concepts, where the latter can be expressed, described and understood through the former.

2.2. Arguments

Inference Anchoring Theory (IAT) (Reed & Budzynska, Reference Reed and Budzynska2011) provides the framework for understanding arguments as dynamic discourse phenomena occurring when the speaker attempts to persuade the hearer during dialogical interaction. Based on the concepts of dialogical turns and illocutionary forces, IAT postulates that each inference (i.e., the relation between premise and conclusion of an argument) is “anchored” in dialogical exchange. According to IAT, arguments are constructed between dialogical turns and are related to the speech act of arguing, making them highly dynamic and pragmatic in nature. This approach is suitable for the material proposed in the current paper – pre-election debates – which consists of spoken arguments, uttered in highly persuasive contexts.

In terms of argument structure, IAT postulates that an utterance containing just one proposition is not yet an argument – it is simply an assertion, where the speaker is stating their opinion. It is only in the context of dialogical challenges, when the speaker is providing a justification for their proposition, that an argument is constructed. An argument then, is a pair of (at least) two propositions: a conclusion (a claim) and a premise justifying it. In Example 1, presented in Section 1 of this paper, the proposition “Polish health service is in a heart-attack threatening condition” serves as a premise, justifying “we propose the pact for Polish health service…”, which then becomes a conclusion. Following IAT, we adopt a descriptive rather than prescriptive approach in argument analysis, refraining from judging the quality of arguments proposed by the speaker. The aim of our annotation is to identify argument structures (i.e., premise-conclusion pairs), based on the context indicating argumentative intention (realized in the speech act of arguing). Whether a given premise really justifies a given conclusion is not assessed in our analysis. This method allows us to separate argumentative fragments of texts from non-argumentative ones, allowing us to identify only those metaphors and emotional appeals that appear within argument structures. Figure 3 presents the premise and conclusion pair in OVA+ software, used to annotateFootnote ² them in text.

Figure 3. Example of premise and conclusion pair (English translation of Example 2, analyzed in detail in Section 3.3).

2.4. Operationalization of hypotheses

In sum, the conceptualization of three dynamic discourse phenomena proposed in this paper allows us to investigate the interaction between them in one analytical frame presented in Table 1.

Table 1. Conceptualization of metaphors, arguments and emotion-eliciting words as used in this paper

Based on such conceptualizations we propose the following operationalization of hypotheses presented in Section 1:

3.1. Political debates

Our study is based on the corpora of two Polish 2019 pre-election debates (18,783 words in total), annotated with metaphors using Metaphor Identification Procedure (PRAGGLEJAZ Group, 2007) and arguments using Inference Anchoring Theory (Reed & Budzynska, Reference Reed and Budzynska2011). The third layer consists of automatic identification of emotion-eliciting words based on EMEAN and NAWL databases (Wierzba et al., Reference Wierzba, Riegel, Kocoń, Miłkowski, Janz, Klessa, Konat, Grimling, Piasecki and Marchewka2021). This allows us to capture the elements of the debate where speakers not only use metaphors in their argumentation (Juszczyk et al., Reference Juszczyk, Konat and Fabiszak2022), but they are also adding an emotional appeal to increase the persuasive power of their words. The information about the debates is summarised in Table 2.

Table 2. Description of the TVP and TVN corpus of 2019 pre-election debates

3.2. Procedure

In our procedure, three layers of analysis are conducted independently, by separate groups of analysts.

First, the recording of both debates is transcribed manually.

Second, metaphors are manually annotated. They are identified in transcripts as metaphorical lexical units using Metaphor Identification Procedure (PRAGGLEJAZ Group, 2007) and eMargin software (Kehoe & Gee, Reference Kehoe and Gee2013). Metaphorical expressions were annotated in accordance with (Juszczyk & Kamasa, Reference Juszczyk, Kamasa, Knapik, Odelski, Chruszczewski and Chłopicki2016), which adapted the Metaphor Identification Procedure (MIP) (PRAGGLEJAZ, 2007). In order to facilitate online group discussions about metaphorical units the team of annotatorsFootnote ⁴ used the e-Margin: A Collaborative Textual Annotation Tool (Kehoe & Gee, Reference Kehoe and Gee2013), and the resulting data is publicly available.Footnote ⁵ After reading the text excerpt (step 1 in MIP) and identification of lexical units (step 2 in MIP), raters established contextual meaning of each unit (step 3a in MIP) and determined if it has a more basic contemporary meaning in other contexts (step 3b in MIP). The lexical unit was marked as metaphorical if its contextual meaning contrasted with the basic meaning but could be understood in comparison with it (steps 3c and 4 in MIP). Contextual meanings were identified using Wielki Słownik Języka Polskiego (Great Dictionary of the Polish Language) (Żmigrodzki, Reference Żmigrodzki2019). There are 814 metaphors identified in the entirety of the transcripts and 313 metaphors in our corpus of arguments.

Third, argument structures are manually annotated. The debate transcript is segmented into Argumentative Discourse Units (ADUs). In this way, we identify premise–conclusion pairs. In our corpus, there are N = 615 such pairs. Argumentative Discourse Units are marked using Online Visualization of Arguments (OVA+) tool (Janier et al., Reference Janier, Lawrence and Reed2014), which allows for the diagramming of the structures described in Inference Anchoring Theory and storing the resulting annotations in the form of Argument Interchange Format – AIF ontology (Chesnevar et al., Reference Chesnevar, McGinnis, Modgil, Rahwan, Reed and Simari2006) as part of AIFdb publicly available database.

Fourth, emotion-eliciting words are automatically extracted using lexical lists: EMEAN (Wierzba et al., Reference Wierzba, Riegel, Kocoń, Miłkowski, Janz, Klessa, Konat, Grimling, Piasecki and Marchewka2021) and NAWL (Riegel et al., Reference Riegel, Wierzba, Wypych, Żurawski, Jednoróg, Grabowska and Marchewka2015; Wierzba et al., Reference Wierzba, Riegel, Wypych, Jednorwóg, Turnau, Grabowska and Marchewka2015). Intersection of NAWL and EMEAN gave us a list of 7674 unique items. In doing so, we follow the psychological tradition in researching emotional appeal. We do not use wordlists produced for clarin or WordNet.

Fifth, text in MLU and ADUs is lemmatized and sixth, overlap between metaphors, arguments, and emotional appeals is identified using Excel formulas. Our analytic procedure can be summarized in steps shown on Figure 4:

Figure 4. The procedure of data analysis. MLU: Metaphorical Lexical Units; ADU: Argumentative Discourse Units, EEW: Emotion Eliciting Words. Numbers refer to sets of discourse phenomena and their combinations presented in Figure 1.

3.3. Data overview (types and tokens)

Table 3 shows descriptive statistics of types and tokens of lemmatized pairs of premises and conclusions. This data was obtained using AntConc (Anthony, Reference Anthony2022) and Excel formulas.

Table 3. Type-token analysis of the data

As we can see on the chart in Figure 5, happiness is the emotion most intensely appealed to, followed by fear, anger and sadness, which have equal intensity, while disgust is the least intense emotion in our study.

Figure 5. Average intensity of five basic emotion-eliciting words in premise-conclusion pairs.

Example (2) (presented below and in Figure 3) presents three layers of annotation integrated into one data set. It is a premise and conclusion pair with emotion-eliciting words and metaphorical lexical unit: