Pragmatic markers and ideological positioning in EUROPARL: A corpus-based study

Full Text

1. Introduction

Language is a fundamental tool of politics, shaping power relations not only through explicit propositions but also through subtle cues. Among these, pragmatic markers have often been dismissed as minor fillers, yet research shows they can carry significant ideological weight by guiding interpretation, projecting stance, and managing interpersonal alignment (e.g., Aijmer 2013: 42, Fischer 2006: 118). In parliamentary debate, where persuasion and legitimacy are continually negotiated, markers such as well, of course, you know, or but can influence perceptions and ideological positioning. The present paper is informed by CDA-pragmatic studies that demonstrated how subtle linguistic cues such as modality and evidentiality construe ideological bias under the guise of neutrality (e.g., Larina et al. 2019).

Critical Discourse Analysis (CDA) has examined how language enacts power and ideology, but has largely focused on semantic and rhetorical strategies. Pragmatic linguistics, meanwhile, has richly described markers’ roles in coherence, politeness, and stance-taking without linking these to ideology. This study bridges these strands by analysing pragmatic markers as instruments of ideological positioning in European Parliament (EP) debates. It explores how they naturalize viewpoints as common sense, challenge opposing stances, and foster rapport with audiences.

The study aims to provide tools for the analysis of manipulation, to show how micro-level markers like of course or you know can reveal implicit persuasive strategies such as presupposing agreement or invoking shared knowledge. It also provides interdisciplinary insight by demonstrating the value of integrating CDA and Critical Discourse Theory with descriptive pragmatics to explain how markers sustain or contest power relations, while also recognizing that AI itself has been framed as an ideology reshaping social institutions through power, manipulation, and domination. It integrates AI tools by assessing the potential of large language models (e.g., ChatGPT) to detect and interpret pragmatic markers in large corpora, noting both their added value and limitations.

Following this introduction, the paper reviews the relevant literature, outlines a corpus-based methodology, presents quantitative and qualitative findings, and interprets them in light of ideological positioning and manipulative discourse strategies. The discussion also considers AI’s role in complementing the analysis, before concluding with the study’s contributions and directions for future research on language, ideology, and pragmatics in political discourse.

2. Literature review

Pragmatic markers (PMs) are linguistic items that do not alter propositional content but serve crucial textual and interpersonal functions. Classic studies (Schiffrin 1987) identified their role in organizing discourse, while Fraser (1996: 168) defined them as expressions shaping pragmatic interpretation rather than truth-conditional meaning. Subsequent approaches — from Conversation Analysis to Relevance Theory and interactional sociolinguistics — have shown that PMs manage turn-taking, mitigate face threats, and signal stance (Fischer 2006, Aijmer 2013).

More recently, PMs — traditionally viewed as non-ideological — have been re-evaluated as carriers of ideological meaning. Following Rocher’s (1969) classic sociological perspective, ideology can be understood as a structured system of ideas and judgements that both explain and justify the position of a group, drawing on shared values and orienting its future course of action. Wodak (2007: 203) urged integrating pragmatics into CDA, noting that hedges, fillers, and turn initiators can index power and stance. This aligns with Fairclough’s (1995: 136ff) concept of the “conversationalization” of public discourse, where institutional talk adopts colloquial features to appear relatable. Historical corpus research on British parliamentary debates (Hiltunen & Vartiainen 2024) confirms increased informality, with markers such as you know and well projecting solidarity and aligning with populist appeals.

In political contexts specifically, PMs support persuasion within formal norms. They can naturalize stances as self-evident (of course), downplay specifics (and so on), or manage interpersonal relations (well as a mitigator). Furkó (2019, 2020) showed that evidential markers, general extenders, and stance markers often serve manipulative ends, aligning with van Dijk’s (1993) concept of “ideological work.” Other studies highlight how boosters (indeed, clearly) reinforce authority, while adversatives (but) pivot from concession to preferred stance, foregrounding one view over another. Engagement markers such as you know can build in-group solidarity, a hallmark of populist rhetoric (Wodak 2015: section 4.1).

From a critical perspective, Critical Discourse Analysis (CDA) and Critical Discourse Studies (CDS) traditionally focus on overt rhetorical devices, but micro-level features like PMs also contribute to power dynamics. They can reinforce the “ideological square” (van Dijk 1993: 249), naturalizing in-group virtues and problematizing the out-group (We, of course… compared with They, well…). Subtle markers of attitude (frankly, honestly) may lend unwarranted credibility, while general extenders (phrases such as and so on, used at the end of lists to mark the list as incomplete) can obscure contentious details (de Saussure 2007: 152, Taubayev 2015: 254). Recent cross-cultural research (Ponton et al. 2025) expands this perspective, showing how pronouns and ‘we-strategies’ enact consensus and ideological alignment across political systems.

Methodologically, Corpus-Assisted Discourse Studies (CADS) have enabled large-scale analysis of PMs, revealing frequency spikes that correlate with ideological confrontation or political strategy shifts. This evidence reinforces qualitative descriptive (e.g., Stubbs 1996) and critical (e.g., Flowerdew & Richardson 2017) studies by demonstrating that PMs are used systematically in the construction of ideology.

In sum, PMs are polyfunctional resources operating at the intersection of cohesion, stance-taking, and ideological positioning. This study builds on Furkó (2019, 2020) and others by examining their role in European parliamentary discourse through a corpus-based CDA approach, connecting micro-level language choices to broader political strategies.

3. Data and methodology

This study uses the English-language subset of the EUROPARL corpus (ENA, August 15, 2025)¹ with a view to avoiding translation issues and focussing on original utterances. We compiled a 1-million-word sub-corpus of parliamentary debates from the past two decades, selecting sessions with high ideological contention (e.g., immigration, sovereignty, economic policy) to capture strategic language use across parties, countries, and political alignments.

PMs were identified through a combined automated and manual process. An initial list — based on Furkó (2020: 151, 196) — included discourse markers (well, now), stance markers (I think, you know, frankly), evidential/modal items (of course, surely), conjunctive connectors (but, however), general extenders (and so on, or whatever), and fillers. We searched the corpus using Sketch Engine as well as AntConc and Python scripts, then manually excluded non-pragmatic uses (e.g., well as an adverb of manner). Each occurrence was annotated for one or more functional categories: Evidential/Certainty, Contrast/Counterargument, Interpersonal/Engagement, Hedging/Qualification, Filler/Pauser, and General Extender.

Analysis followed a corpus-based CDA framework (Wodak 2015, Hart 2018), combining quantitative measures (frequency, dispersion, collocation) with close qualitative reading. Collocational patterns (e.g., of course we, but I) and PM clusters (e.g., well, frankly) were examined for rhetorical effects and compared with the British National Corpus (BNC) as well as the Hansard Corpus (ENA, August 15, 2025)² as reference points. We aligned PM usage with five manipulative strategies from CDA: suppression, polarization, recontextualization, conversationalization, and ambiguity.

An exploratory AI-assisted component tested whether GPT-4 could reliably identify and interpret PMs taking possible hallucinations and confabulations into consideration. We provided short corpus extracts (150–200 words) and prompted the model to highlight PMs and comment on their function. Outputs were compared to manual coding for insight into the model’s utility and limitations, echoing recent computational work (Wise & El Barj 2023: 3) on machine learning detection of hedges and authority markers.

Reliability was ensured through double-coding of a subset of data by an additional analyst, resolution of disagreements via established pragmatic criteria, and statistical checks (e.g., dispersion plots) to avoid skew from single debates or speakers. Example excerpts included in the paper were selected for representativeness, while all identifying political details were anonymized.

The aim of the mixed-method design — quantitative corpus analysis, qualitative CDA interpretation, and exploratory AI-assisted review — was to enable both breadth and depth in tracing how pragmatic markers contribute to ideological positioning in European parliamentary discourse.

4. Results of the analysis

In this section, we report the findings of our corpus-based analysis, illustrating how pragmatic markers operate in parliamentary discourse to reinforce or challenge ideological positions. The analysis is structured around several functional categories of pragmatic markers, although overlaps are common since a single marker instance can serve multiple functions. For each category, we highlight quantitative trends (frequency or distributional patterns) and qualitative insights (illustrative examples and their discursive effects).

4.1. Evidential and certainty markers: Asserting common ground

Evidential or certainty markers include items that convey the speaker’s assessment of a statement’s truth, obviousness, or shared acceptance. Examples from our data include of course, indeed, obviously, clearly, in fact, surely, as well as certain uses of phrases like I believe (when used assertively rather than tentatively). These markers were found to be pervasive in parliamentary speech. Quantitatively, of course was among the most frequent multi-word pragmatic markers in the corpus, appearing on average 5.89 times per 10,000 words (henceforth tpttw). Its usage spanned speakers from different political groups, but the analysis revealed a common thread: of course often prefaced statements that the speaker wanted to present as uncontested or taken for granted.

(1)   Of course, we want a strong Europe that protects its citizens. (ENA, August 15, 2025)³

Here, of course is used to frame the proposition (we want a strong Europe that protects citizens) as something beyond doubt or debate. The effect is twofold: it posits unity and consensus (implying that everyone in the chamber, or at least the speaker’s in-group, must agree on this goal) and it implicitly marginalizes any dissent (anyone not wanting that would seem unreasonable). This aligns with the strategy of naturalization of ideology — making an ideological commitment (support for a strong protective Europe) appear as common sense. Such usage corroborates Simon-Vandenbergen et al.’s (2007: 31) point about presupposition: of course presupposes agreement. In our corpus, right-leaning and left-leaning speakers alike used of course to preemptively close debate on core values (e.g., “Of course we care about human rights,” “Of course national security is paramount”), even if they subsequently diverged on the means of policy implementation. This evidential marker thus serves as a face-saving and consensus-building device: challenging an “of course” statement risks positioning the speaker outside the presumed consensus.

Other certainty markers like indeed (2.94 tpttw) and clearly (3.58 tpttw) were also frequently used to strengthen claims. Indeed often appeared in supportive follow-ups: e.g., “Our economy is improving. Indeed, unemployment has fallen to record lows.” This marker signals reinforcement or confirmation of a point, adding emphasis that the speaker’s argument is grounded in reality or evidence. In ideological terms, indeed helps a speaker build authority, suggesting that facts are on their side (thus any opposing claims are implicitly less factual). Clearly and obviously serve a similar affirming function; however, they can carry a slightly confrontational undertone — obviously in parliamentary speech often came up in rebuttals: “Clearly, the opposition has not considered the full implications of this law,” or “Obviously, what my colleague fails to mention is…”. Here the pragmatic marker clearly casts the speaker’s subsequent correction as self-evident truth, thus describing the opponent as either ignorant or deceptive for not acknowledging it. This is a subtle form of delegitimization through pragmatic phrasing. It resonates with van Dijk’s (1993: 250) notion that elites in discourse establish their version of reality as the authoritative one, effectively dismissing others’ versions as clearly misguided.

One interesting finding was how speakers modulated certainty markers to manage epistemic stance. When a speaker wanted to avoid appearing too dogmatic, they occasionally paired certainty markers with personal attribution: e.g., “I believe, of course, that...” or “Of course, in my view,...”. By inserting “I believe” or “in my view,” the speaker adds a slight hedge acknowledging personal stance, yet still retains of course to imply that their belief aligns with common sense. This interplay of hedging and certainty reflects Hyland’s (2005: 138ff) notion of balancing boosters and hedges in academic writing, here manifesting in political discourse. It allows politicians to push a viewpoint as obvious while maintaining a veneer of humility or subjectivity (“it’s just my reasonable opinion that happens to be obvious”).

4.2. Adversative and contrastive markers: Managing counter-arguments

But (28.94 tpttw) is by far the most frequent adversative marker in our corpus — unsurprising, as argumentation thrives on contrast and rebuttal. However, its pragmatic role extends beyond that of a mere conjunction. In political speeches, but often serves to orchestrate a specific rhetoric: acknowledge something to appear fair or thorough, then pivot to the main point which often undermines what was acknowledged. We observed a pattern where speakers would use but to navigate ideological safe ground before moving to contentious claims.

(2)   The proposal has some merits, but it is ultimately unacceptable to our group because it undermines national sovereignty.” (ENA, August 15, 2025)⁴

In this example, everything before but is a strategic concession (“has some merits” is a mild positive). This concession may signal acknowledgement of a competing ideology or of a general principle (e.g., a proposal deemed socially progressive, hence “some merits”). However, the use of but indicates that the speaker’s principal stance follows—namely, rejection of the proposal on ideological grounds (c.f., national sovereignty, a typical concern of certain ideological camps). Here, but mitigates the preceding concession and steers the audience’s attention toward the subsequent argument. From a CDA perspective, this structure allows the speaker to appear reasonable and balanced (acknowledging both sides) while effectively prioritizing their partisan stance — a tactic of apparent concession that strengthens argumentative force. Such use of but is so routine in parliamentary dialectic that even listeners expect that any phrase before but might be perfunctory. It ties into the broader ideological strategy of framing: by structuring discourse as “Yes, X is true, but Y,” the speaker frames Y (their viewpoint) as the conclusion to be remembered, whereas X (the opponent’s point) is framed as subordinate or the exception to the rule.

Other contrastive markers identified include however (13.82 tpttw), yet (12.72 tpttw) and nevertheless (5.3 tpttw), which tended to appear in more formal registers (often read from prepared speeches). These function similarly to but in indicating a turn to a counterpoint. We found that however is often sentence-initial in transcripts (e.g., “However, we must consider...”), reflecting written-style influence; whereas but is more common mid-sentence in spontaneous remarks (“... merits, but it is unacceptable ...”). Pragmatically, however can carry a slightly more polite or measured tone than but. For instance, in diplomatic exchanges on the floor, an MEP might say: “I appreciate the Commission’s efforts; however, I remain skeptical about the timeline.” The difference is subtle: however separates the clauses more cleanly, allowing the speaker to delineate the positive and negative clearly, whereas but blends them into one sentence, often for punchier delivery.

One particularly political use of adversatives is in managing face and mitigating direct confrontation. Instead of directly contradicting a fellow politician, a speaker might employ a pseudo-agreement followed by but. e.g., “I understand what my colleague is saying, but I think he is overlooking...”. The phrase “I understand” here is not a PM per se but works in tandem with but as a politeness strategy. It acknowledges the colleague’s perspective (saving their face) just enough before delivering disagreement. This relates to the distinction between impersonal and personal argumentation: PMs that distance the idea from the individual (e.g., “I understand him, but the idea…”) frame the conflict as concerning ideas rather than personal competence. But can thus serve an important role in institutional politeness — it lets debate occur without overt personal attacks, adhering to parliamentary norms.

Interestingly, we noticed that some speakers, particularly those from more consensus-driven political cultures or parties, would sometimes avoid starting a sentence with but, opting for longer formulations such as “That may be so; however,...” or even breaking into a new sentence starting with However. In contrast, more combative debaters frequently used but in rapid-fire exchanges (e.g., interjecting “But that’s not true” in heated moments). This pattern suggests a stylistic divide that may correlate with ideology or debating style: a more confrontational, populist approach might use ‘but’ to directly interject and refute, whereas a more technocratic or diplomatic approach tends to employ more elaborate constructions to maintain decorum.

4.3. Interpersonal and engagement markers: Building solidarity or control

Pragmatic markers that directly engage the audience or manage the speaker-hearer relationship were also prominent. Chief among these in our corpus is you know (4.001 tpttw in EUROPARL compared with 4.61 tpttw in BNC), a classic example of an interpersonal marker. You know appeared in our data both in its canonical use (seeking confirmation or indicating shared knowledge) and as a general filler. Quantitatively, you know was less common in the formal plenary speeches (which often are prepared or read out) but more frequent in spontaneous moments such as Question and Answer sessions or (counter-)interjections. It was also more likely to appear in speeches by certain politicians known for a plainspoken style. When you know is used in the European Parliament context, it often seems intended to bring listeners onto the same page, as if appealing to common sense or shared experience.

(3) We’ve been negotiating this for years, and, you know, nothing has really changed on the ground. (ENA, August 15, 2025)⁵

Here you know is used as a rhetorical device to invite the audience (fellow MEPs or the public via broadcast) to agree that the situation is obvious or familiar; it functions as a softener and inviter of concurrence. In doing so, you know can create a sense of camaraderie or in-group understanding between the speaker and audience, which is powerful in ideological alignment, at the same time feeding polarization: those who disagree are implicitly cast as outsiders.

Other engagement markers include tag questions (e.g., isn’t it? or don’t we?), which were relatively infrequent in our corpus, possibly reflecting the less dialogic nature of parliamentary speech compared with everyday conversation. Nonetheless, we found instances such as “We can all agree that democracy is non-negotiable, can’t we?” uttered by a speaker trying to pressure a consensus. The tag can’t we? is a strong engagement move, turning a statement into a collective agreement check. If no one objects, it creates the record that the assembly agrees. If someone did object, they risk appearing contrary to democracy itself in this example. Hence, such markers can put argumentative pressure on the audience to align.

Well (5.18 tpttw) deserves mention here too, as it often appears at the start of responses to questions or interjections, functioning as a conversational pivot. For example, when challenged, a minister might start their reply with “Well,” — this indicates they are addressing the point but possibly disagree. In our notes we observed that well at turn-initial position frequently co-occurred with subtle shifts in footing, such as moving from defense to offense in argument, carrying the speaker over a potential moment of tension. It provides a brief pause and facilitates a transition into what could be a face-threatening act (e.g., contradicting a high-ranking official). This usage aligns with classic descriptions of well as a marker of dispreferred responses (e.g., Pomerantz 1984: 60). In parliamentary discourse, where open conflict is moderated by formal politeness, well thus shows up as a hedge, mitigator of an FTA, a function which we now turn to.

4.4. Hedges and discourse mitigators: Calibrating strength and ambiguity

In addition to expressing certainty or directness, a key function of PMs is to do the opposite — hedging or introducing ambiguity. Hedging functions of epistemic DMs such as maybe (0.9 tpttw), perhaps (9.07 tpttw), sort of (2.83 tpttw), kind of (5.81 tpttw), I think (in a tentative sense) (0.9 tpttw) were present in the corpus, though their distribution was skewed. They appeared more in deliberative contexts or when speakers were discussing complex, uncertain issues (e.g., economic forecasts, hypothetical scenarios) and less so when making ideological statements of principle. In parliamentary debates, showing uncertainty can sometimes weaken a position, so politicians often avoid too much hedging on core stances. However, we did observe a strategic use of hedges when dealing with facts or predictions that could be contested.

(4) The reforms will probably yield results in a few years, but we cannot be entirely sure at this stage. (ENA, August 15, 2025)⁶

Here probably and the phrase cannot be entirely sure serve to preempt criticism — the speaker acknowledges uncertainty proactively, which can build credibility by appearing honest and realistic. It is a way to prevent opponents from later saying “you promised X would happen.” Thus, hedging in this case is a defense against future face threat.

In manipulative political discourse, ambiguity and vagueness may at times be deliberate (Bavelas 1983: 285). Such effects can be reinforced by pragmatic markers, for instance when speakers employ general extenders or placeholders. As noted, general extenders like “and so on” (1.54 tpttw), “and things like that” (0.01 tpttw), “or whatever” (0.17 tpttw) leave statements open-ended. In our corpus, these markers were infrequent, appearing in contexts where the speaker sought to avoid specifics or gloss over details.

(5)   The opposition has obstructed, delayed, and so on, every attempt at progress. (ENA, August 15, 2025)[7]

 In this example, ‘and so on’ vaguely alludes to additional negative actions without naming them, creating a smear effect while withholding any concrete allegation the opposing side could directly rebut. Thus, this strategy is a way to imply a larger pattern of negative behavior indirectly.

Another interesting case is the use of etcetera (1.34 tpttw) in formal speech. Saying “the policy covers health, education, infrastructure, etc.” in the middle of a speech may simply reflect brevity, but it can also obscure what exactly is being referred to. When used evasively, it may conceal a lack of detailed knowledge or omit contentious sub-items subsumed under “etc.” Given the high manipulative potential of political discourse, the question arises why a speaker might trail off rather than provide a full enumeration.

We also found that some speakers employed the phrase “if you will” (0.25 tpttw), or its continental variant “so to say” (0.01 tpttw). “If you will” is a softener that suggests a formulation is not exact, allowing wiggle room.

(6)   This plan is a reset, if you will, of our economic model. (ENA, August 15, 2025)⁸

The if you will signals to the audience that reset is a metaphor or an unconventional term here, inviting a certain interpretation but not insisting on it. In terms of toning down pragmatic force, this can make a bold claim more palatable by appearing tentative or colloquial (“if you’ll allow me to use that word”). It serves as a hedge that also engages the listener’s permission. Such moves may lessen immediate pushback, as the speaker appears self-aware and receptive to nuance.

4.5. Quotation and recontextualization markers: Distancing and legitimizing

Yet another category highlighted by previous research (Furkó 2020: 41ff) is quotation markers — phrases that indicate reported speech or a shift in voice, such as “so-called” (3.38 tpttw), “quote ... unquote” (0.001 tpttw), or even tonal quotes implied in the transcript. In our largely textual analysis, detecting the latter is, naturally, challenging. As for the former, when a politician refers to an initiative as “the so-called ‘Stability Pact’”, the premodifier so-called casts doubt and distancing on the term Stability Pact. Pragmatically, so-called signals that the speaker does not endorse the legitimacy or accuracy of the quoted term, framing it as a misnomer or propaganda, delegitimizing the opponent’s framing and recontextualizing it in a skeptical light. In the example given, saying so-called Stability Pact implies it might not bring stability at all, without directly saying “the Stability and Growth Pact is misnamed.” Such markers allow an undercurrent of criticism while maintaining plausible deniability of outright attack.

We also found instances where speakers explicitly say ’quote’ or use air quote around a term (this was sometimes indicated by the transcribers as “quote, unquote” around a word). For instance:

(7)   ...the quote ‘green revolution’ unquote the opposition touts.(ENA, August 15, 2025)⁹

This usage clearly shows the speaker’s disagreement with the term green revolution, insinuating that it is an empty slogan or misleading label. This constitutes metapragmatic commentary — commenting on other’s language to undermine it. By doing so, politicians engage in discursive contestation, disputing not only ideas but also the terminology and framing used to present them.

Recontextualization also occurs when speakers use pragmatic markers to insert someone else’s voice or a hypothetical voice. For example, rhetorical devices like “they say” or “some claim that … well, let me tell you” were observed. They say acts almost like a PM introducing a reported viewpoint which the speaker then often refutes. It constructs a straw man or an opposing stance for the purpose of refuting it. While they say (0.65 tpttw) is not traditionally listed as a PM, its pragmatic role in these speeches is analogous to a quotation marker — it flags an upcoming perspective as attributed to others; often unspecified others, which can be rhetorically useful because it avoids naming and potentially legitimizing a specific opponent.

(8)   They say we are spending too much, but look at the results — well, I say you can’t put a price on social stability.” (ENA, August 15, 2025)[10]

 In this example, they say introduces a criticism vaguely attributed to opponents. The speaker then uses but to counter it, and inserts the PM well in the reply “well, I say you can’t put a price…”. Here well adds a colloquial assertiveness, a tone of scoffing at the referents of “they.” It is as if the speaker takes a moment (well, I say...) to position themselves against the cited criticism, which dramatizes the contrast. This layered use of markers — they say (introduce opposition view), but (negate it), well (mark the speaker’s own retort) — exemplifies how pragmatic markers can work together to structure a dialogic narrative in a monologue, giving the impression of debate and refutation all within one speaker’s turn.

4.6. Co-occurrence highlights

Our quantitative analysis revealed that certain markers had strong co-occurrence tendencies, suggesting idiomatic political usages. For instance, of course frequently followed and (forming and of course), often when adding a point:

(9)   We need to boost innovation and of course support our small businesses. (ENA, August 15, 2025)¹¹

This phraseology might reflect a rhetorical norm of including an obvious or agreeable addition after a more controversial main point, using and of course to frame it as an afterthought that everyone agrees on. Similarly, but of course appeared when speakers conceded something but then still emphasized inevitability: “We must reduce spending, but of course, not at the expense of the most vulnerable,” blending contrast with an assurance of consensus on a value.

Markers like you know showed high collocation with personal pronouns (I and we) and cognitive verbs (think, see), which is expected as it often appears in phrases like you know I think... or you know we can’t.... This underscores its role in maintaining listener engagement in personal or collective reasoning statements.

Another pattern was the use of multiple markers in a row or in proximity, which we term pragmatic marker clustering. We saw sequences like: “Well, you know, perhaps we should…”. The layering of well + you know + perhaps at the start of a statement imbues it with interpersonal and hedging functions: well (I’m responding thoughtfully), you know (we share this understanding), perhaps (I won’t assert too strongly). The result is a highly mitigated suggestion, appropriate for a tentative proposal or broaching a delicate topic. By contrast, frankly often co-occurred with but:

(10) Frankly, I wish we could support this, but we can’t. (ENA, August 15, 2025)¹²

The frankly serves as a marker of honesty or directness, attempting to lend credibility to the unpleasant message that follows after but. These combinations show that pragmatic markers can be stacked to achieve a nuanced tone. Politicians adeptly mix them to simultaneously address multiple pragmatic needs (e.g., sounding honest while disagreeing, without alienating the audience).

Lastly, we note an interesting frequency trend: during emotionally charged debates (e.g., following a crisis or a contentious vote), there was a spike in the usage of direct appeal markers and emotive emphasis conveyed by indeed, truly and honestly. In emotionally charged moments, some speakers relied on these markers to underscore sincerity or intensity:

(11) We are truly at a crossroads indeed, and honestly, our citizens expect leadership. (ENA, August 15, 2025)¹³

In the case of truly (2.69 tpttw) and honestly (0.36 tpttw) we can observe a fuzzy boundary between PMs and propositional lexical items, i.e. their repetitive use suggests semantic bleaching they lose much of their original descriptive force (“truthful,” “sincere”) and instead function mainly as discourse devices to add emphasis or signal sincerity. (cf. Brinton 2017: 31) — with a function to persuade the audience of the speaker’s earnestness. From a critical perspective, one could argue this is a discursive strategy of ingratiation, using language to align oneself with the people’s supposed sentiment (as in authenticity claims: “honestly, I am just as frustrated as you are”).

In sum, the data show that pragmatic markers permeate parliamentary discourse in patterned ways that correlate with argumentative moves and ideological strategies. The next section will discuss what these findings mean for our understanding of political communication and the subtle mechanics of manipulation and positioning in discourse.

5. Discussion

The above analysis reveals that pragmatic markers, often overlooked as mere fillers or connective devices, play a consequential role in shaping parliamentary discourse and the ideologies it conveys. In this section, we discuss the implications of these findings in light of the study’s research questions and the broader theoretical frameworks introduced earlier. We also evaluate how the integration of corpus methods and AI tools contributed to these insights, reflecting on interdisciplinary implications for linguistics and political communication research.

5.1. Pragmatic markers as vehicles of ideology and power

One of the central findings is that PMs contribute actively to ideological positioning in parliamentary discourse. They do so by operating in the background of utterances to frame statements, align speakers with audiences, and preempt resistance. This supports and extends Furkó’s (2020: 79ff) contention that discourse-pragmatic devices are integral to how authority and consensus are enacted in language. The evidence from the EUROPARL corpus shows that markers such as of course and clearly are not incidental; they help construct an ideological common ground where the speaker’s views are the norm. This is a powerful subtle tactic. By the time a parliamentarian has said “Of course, we must do X,” they have already set the terms of the debate: to disagree is to go against what is “of course.” Van Dijk’s work on ideology noted that ideologies often work through implicit assumptions and shared knowledge in discourse (van Dijk 2008: 233); here we see pragmatic markers as a linguistic means to inject those assumptions (the taken-for-granted truths) into the conversation.

Moreover, pragmatic markers facilitate the exercise of power by shaping the flow and tone of discourse. Strategic use of but, for instance, allows a speaker to steer the agenda so that attention shifts toward their counter-argument rather than the original point. This represents a linguistic manifestation of power, as it determines which aspect of the argument is foregrounded. Similarly, markers like well and you know allow a speaker to manage interruptions or challenges gracefully, maintaining a position of composure and inclusivity. These moves accumulate to maintain what Bourdieu (1991: 170) would call symbolic power: the power to impose a vision of reality (in this case, through how issues are talked about). A politician who frequently says “you know” might give the impression that the public is on their side (since “you” presumably includes the public), thus exerting a subtle claim to speak for the people — a hallmark of populist authority claims.

Our findings also illustrate how pragmatic markers tie into populist discourse strategies as described by, for example, De Cleen (2019) or Musolff (2016). Populist rhetoric often involves constructing a dichotomy between the rational, honest common folk and the corrupt, deceitful elite. Pragmatic markers facilitate this in two ways: (1) Conversationalization — making political speech sound like everyday talk — was evident in the usage spikes of colloquial markers (you know, well, just, kind of). This aligns with the trend of colloquialization observed in British parliamentary discourse in the Hansard corpus (cf. Hiltunen & Vartiainen 2024), and in EUROPARL it was often the more populist or outsider voices that embraced it. By using informal markers, these speakers discursively downplayed the institutional distance, attempting to sound like “one of the people” rather than an aloof politician. (2) Polarization and Legitimization — markers helped polarize by reinforcing in-group knowledge (e.g., “we all know”) and delegitimized out-group narratives (through quotation markers like “so-called”). When a representative says “the so-called experts”, that single marker so-called casts doubt on an entire group’s credibility, resonating with an anti-elitist ideological stance.

Additionally, pragmatic markers were found to aid in the suppression or backgrounding of counter-arguments, which is a subtle form of exercising discursive power. By using general extenders or trailing off with etc., speakers minimize what is left unsaid. This can intentionally or unintentionally suppress further detail or alternative viewpoints. If an MEP lists positives of a policy and ends with “and so on,” they might be glossing over other benefits (or costs) that they choose not to articulate — effectively removing them from the immediate debate space. This connects to the critical concept of agenda-setting in discourse: not everything gets voiced or elaborated upon. Pragmatic markers can be a linguistic tool for trimming the agenda in one’s favor. As Fairclough (2003: 55) noted, what is not said (and how it is omitted) can be as important as what is said in maintaining hegemonic narratives.

5.2. Interdisciplinary reflections: Bridging pragmatics and CDA

Our study underscores the importance of bridging pragmatic micro-analysis with critical macro-analysis in discourse. Through the literature review, we noted how CDA and CDT (Critical Discourse Theory) approach discourse at different levels — one being more empirically textual, the other more abstractly concerned with discourse in social structure. Pragmatic marker analysis provided a concrete interface between these: it is empirical and textual, yet its import is only understood via ideological concepts. For example, identifying a spike in of course usage is empirical; interpreting it as a sign of naturalizing a certain ideology requires critical theorization about hegemony and common sense. In our discussion of results, we effectively linked specific linguistic evidence to discursive strategies such as conversationalization, naturalization, polarization, and legitimation. This demonstrates a methodological payoff of combining corpus pragmatics with CDA — it allows analysts to quantify and pinpoint how exactly those strategies manifest in language patterns.

This interdisciplinary approach also necessitated clarifying terminology, as we did in adopting the term pragmatic markers over discourse markers. This choice is not merely pedantic; it reflects an intent to capture these items’ multifunctionality beyond discourse cohesion. The term pragmatic marker emphasizes their role in speaker stance and interaction, which made it easier to discuss their ideological roles. It also connects to critical pragmatics — an area that looks at how context and power relations affect pragmatic meaning. Wodak’s (2007: 210) call for cross-theoretical inquiry is essentially answered by studies like this, which treat pragmatic details as crucial evidence of broader social meaning making.

5.3. The role of AI tools in analyzing pragmatic markers

One of the innovative perspectives of this study was exploring AI assistance in discourse analysis. The experiment with ChatGPT, though limited, provided revealing observations. On the positive side, the AI was quite adept at identifying common pragmatic markers (it reliably highlighted items like well, but, of course, you know in the input segments). It also generated plausible explanations for their functions, often consistent with established descriptions in pragmatics. For instance, for a sentence in our test excerpt, “Well, we should consider the alternatives,” the AI noted that “well” introduces a suggestion, softening a potential disagreement. This is essentially correct and matched our analysis. Such capability suggests that AI could be useful as a first-pass tool in scanning large volumes of text for potential pragmatic phenomena. It could flag sentences with PM clusters for in-depth human analysis, thereby expediting some of the labor-intensive aspects of corpus analysis.

However, the AI also displayed notable limitations. It struggled with more nuanced or context-dependent aspects. For example, when given a passage where “of course” was used ironically (the speaker was actually being sarcastic saying “Of course, the minister has answered everything — not!”), the AI did not catch the sarcasm; it interpreted of course straightforwardly as indicating obviousness. This suggests that AI, at present, lacks true pragmatic competence — it does not grasp tone, irony, or the extra-linguistic knowledge needed to see when a PM is used sincerely versus sarcastically. A human analyst immediately sensed the sarcasm from context (and perhaps tone, if audio were available), understanding that “of course” in that context meant the opposite. The AI’s misinterpretation in such cases is a cautionary tale: context is quintessential in discourse analysis, and large language models, while context-sensitive in a textual sense, do not have the real-world awareness or discourse situation awareness to fully emulate human interpretation.

Additionally, ChatGPT occasionally over-generalized its explanations. At times it ascribed a manipulative intent to a marker usage where a human would see it as routine. For example, it suggested that a particular use of “well” was to “stall for time and deceive the audience” — an overreach not supported by evidence (it was simply a typical conversational well). This hints at another limitation: AI might introduce bias or make assumptions that are not textually grounded, especially since it has been trained on myriad texts including possibly some with conspiracy or overinterpretation. It underscores the need for a critical human perspective: it is necessary for the analyst to confirm whether a purported function is fulfilled in a particular context.

Nonetheless, the integration of AI is promising for scalability. A tool such as the PragMaBERT model (Wise & Houda 2023) could process entire corpora and statistically highlight anomalies or patterns (e.g., a model could flag that “frankly” is unusually frequent in a certain politician’s speeches relative to others, which might correlate with a certain persona or strategy). AI can also help in performing tasks such as clustering contexts of a pragmatic marker to see the common threads. In our case, manual analysis found patterns (e.g., of course often in initial position signaling assumed agreement); an AI might cluster all instances of of course and help quickly surface that pattern.

Importantly, the use of AI in critical analysis raises a meta-issue of whether algorithms can detect manipulation. As our study shows, markers contribute to manipulation in often subtle ways. Teaching an AI model what counts as manipulative use of language requires not just linguistic input but a theory of manipulation. Some progress is being made — for example, labeling instances of clear populist rhetoric or known propaganda techniques. Wise & El Barj (2023: 3) claim their fine-tuned BERT can identify context-dependent manipulative PM use to a degree. However, such models largely recognize patterns they have seen; they might not detect novel or highly context-specific manipulations. We therefore view AI as a tool for human analysts rather than a replacement. It can handle the “what” (finding markers, counting, basic function labeling) quite well; but the “so what” — the ideological significance — still requires human critical reasoning.

Our own brief trial suggests that a productive workflow could be AI-assisted coding followed by human critical interpretation. This resonates with the notion in digital humanities of “distant reading” (getting the big patterns via computational means) combined with “close reading” (interpreting specific instances in depth). In critical discourse studies, where the stakes include understanding propaganda and ideology, maintaining this human-in-the-loop approach is crucial to avoid missing cultural nuance or ethical implications. The commentary on AI usage in this paper also contributes to the emerging discussion on critical algorithmic studies: just as we critically examine political language, we must also critically evaluate the AI tools used for that analysis—questioning their underlying motives, norms, and assumptions, much like Blodgett et al. (2020) recommend in their call for interrogating bias measurement methods in NLP.

5.4. Implications for political communication and public discourse

Understanding pragmatic markers in parliamentary discourse has practical implications beyond linguistics. It sheds light on how politicians achieve persuasive impact not only through grand rhetoric but through the minutiae of language. Media training for politicians often focuses on messaging and staying on point; our findings suggest that training could also usefully focus on pragmatic markers — for example, advising a speaker to use of course to project confidence, or cautioning that overuse of hedges like maybe can undermine perceived decisiveness. Conversely, from a media literacy or public awareness perspective, teaching citizens to spot these markers and reflect on their effect can be empowering. If voters recognise that of course, indeed, truly, etc. can function as prompts to accept a claim as true, they may respond with greater scrutiny to statements presented as self-evident.

Furthermore, this study’s approach can inform analysis of parliamentary transcripts by journalists, fact-checkers, or analysts. For instance, identifying that a leader’s speech relies heavily on “we all know” and “of course” might prompt an examination of what unspoken assumptions are being pushed. At the same time, it is important to acknowledge that frequent reliance on such markers may also reflect individual style or rhetorical habit rather than deliberate ideological bias. However, our approach also helps differentiate political styles: one politician’s discourse full of well, I think, perhaps paints a different ethos (maybe more cautious or scholarly) than another’s filled with clearly, of course, in fact (more assertive, authoritative). These stylistic differences influence how audiences receive messages. For example, an assertive style may convince some, whereas others prefer a more subtle approach. Moreover, each style can be portrayed by media in various lights (e.g., as confident as opposed to overconfident or cautious as opposed to weak), and each framing carries political consequences.

Finally, from a democratic discourse standpoint, awareness of these subtle linguistic strategies could lead to calls for more clarity and sincerity in politics. If overused, pragmatic markers can also be double-edged — for example, too many honestly or frankly can start to ring hollow, potentially eroding trust. At the same time, it can be risky to rely on such cues as straightforward indicators of sincerity or credibility, since their pragmatic force depends heavily on context, speaker style, and audience perception. This once again underlines the need to emphasize the complementary nature of different analytical tools and methodologies.

6. Conclusion

This study investigated how pragmatic markers contribute to ideological positioning and manipulative strategies in parliamentary discourse, employing a corpus-based CDA approach on European Parliament debates. The analysis confirms that pragmatic markers — words and phrases such as of course, well, but, you know, among others — function as linguistic hinges on which the framing of arguments and the management of interpersonal relations turn. Through these small pivots of language, speakers suppress dissent, invite agreement, construct in-groups and out-groups, and steer the interpretation of their statements in ways that align with their ideological objectives.

Several key insights emerge from the research. First, pragmatic markers are instrumental in making certain ideologies appear as common sense. By embedding presuppositions and shared assumptions into debate (e.g., “of course we all agree on X”), politicians can naturalize their viewpoints and subtly delegitimize opposing perspectives without overt confrontation. While it is true that similar strategies also occur in everyday dialogue and interaction, their use in parliamentary discourse is particularly consequential because of the heightened stakes of political communication and the potential to shape collective decision-making. Second, pragmatic markers facilitate strategic maneuvering in argumentation: they help speakers balance politeness with assertiveness (through hedges and boosters), manage counter-arguments (through adversatives like but and concessive structures), and maintain a persona of credibility or relatability (through engagement markers like you know and conversational tone). These micro-level tactics accumulate into macro-level persuasive and manipulative effects that are central to critical discourse concerns. In essence, the competition for public support in parliamentary debates is waged not by content alone but also through the cadence and pragmatic cues of language.

Methodologically, the study demonstrates the value of combining corpus linguistics with critical discourse analysis to study political language. The corpus approach provided empirical evidence of patterns (such as frequency trends and co-occurrences) that lend weight to our interpretations, moving the analysis beyond anecdotal observations to more robust generalizations. At the same time, the CDA perspective ensured that we kept sight of power relations and ideology when interpreting those patterns. The exploratory integration of AI (via a large language model) highlighted a frontier for future research — one where human expertise and machine assistance could jointly handle the analysis of ever-growing political text archives. While current AI tools have limitations in grasping nuance and context, they hold promise for preprocessing and highlighting potential areas of interest, thus freeing analysts to focus on deeper interpretative work. Future advancements might see more sophisticated models capable of detecting pragmatic and rhetorical strategies, but our findings suggest that human critical judgment will remain indispensable to correctly interpret and contextualize what the machines find.

For the fields of linguistic pragmatics and discourse analysis, this study reinforces that semantically bleached, extremely context-dependent elements such as pragmatic markers deserve a central place in analyses of ideology and power. It encourages further corpus-based pragmatic studies across different languages and settings — for instance, comparing how pragmatic marker usage in parliaments varies between cultures or political systems, or how it evolves over time with changing political norms. The fact that our data was from a multilingual institution (the EU Parliament) also invites cross-linguistic questions: the question arises if equivalent markers in other languages carry the same ideological functions, or if there are culturally specific pragmatic devices that play similar roles. Given the scope of EUROPARL, future research could expand into those directions, enhancing our understanding of pragmatics in a global political context.

In conclusion, by zooming into the “small” words of parliamentary debates, we gain insight into the dynamics of persuasion, consensus, and dissent in democratic processes. Pragmatic markers serve both cohesive and mitigating functions in parliamentary dialogue: they bind arguments into a coherent, seemingly commonsense narrative and smooth over disagreements and transitions. Recognizing their role enriches our comprehension of political rhetoric, reminding us that every well, of course, or you know in a political speech may be doing more covert work than meets the ear. Awareness of these cues can help both analysts and citizens engage more critically with political language. While it is often observed that citizens are increasingly positioned as passive consumers of political discourse, fostering such awareness can contribute to more active, reflective forms of participation. From a research perspective, continued examination of these subtle mechanisms contributes to a more nuanced understanding of how language, power, and ideology intertwine in discourse. The present study has aimed to contribute to this ongoing inquiry by offering an empirically grounded account of pragmatic markers in the service of political persuasion and ideological positioning.

_{¹ https://www.statmt.org/europarl/}

_{² http://hansard-corpus.org}

_{³ https://www.statmt.org/europarl/}

_{⁴ https://www.statmt.org/europarl/}

_{⁵ https://www.statmt.org/europarl/}

_{⁶ https://www.statmt.org/europarl/}

_{⁷ https://www.statmt.org/europarl/}

_{⁸ https://www.statmt.org/europarl/}

_{⁹ https://www.statmt.org/europarl/}

_{¹⁰ https://www.statmt.org/europarl/}

_{¹¹ https://www.statmt.org/europarl/}

_{¹² https://www.statmt.org/europarl/}

_{¹³ https://www.statmt.org/europarl/}

About the authors

Péter Bálint Furkó

Author for correspondence.
Email: furko.peter@kre.hu
ORCID iD: 0000-0002-9650-4785

DSc (Doctor of the Hungarian Academy of Sciences), is Full Professor of Linguistics at the Department of English Linguistics

References

Supplementary files