To Dog-Whistle or to Bark? Elite Communication Strategies When Invoking Conspiracy Theories

Abstract

When do politicians dog-whistle conspiracy theories (CTs), and when do they explicitly endorse – or ‘bark’ – a CT? Over time, does the use of dog-whistles shape the degree to which politicians bark? Drawing from the models of mass communication literature, we theorize that politicians who leverage CTs to garner political support have incentives to tailor their communication to their audience. When politicians speak to general audiences, they risk being punished for explicitly endorsing CTs. However, for parties that use CTs to rally their base, dog-whistling a CT may allow politicians to covertly signal support for a CT to party faithful. Conversely, amongst audiences primarily composed of party loyalists or CT believers, politicians have strong incentives to explicitly endorse CTs. We test our theory with data from Poland, where a series of CTs emerged following a 2010 plane crash in Smoleńsk, Russia that killed the Polish president and 95 other top officials. We draw on speeches and tweets discussing the crash from 2011 to 2022 by the Law and Justice (PiS) party, which sometimes endorses these CTs. We find descriptive evidence that PiS politicians both dog-whistle and ‘bark’. While they tend to dog-whistle more when the audience is more diverse, they tend to bark when the audience is more uniformly CT-supporting. We find some evidence that politicians bark more and dog-whistle less over time, which suggests that, with sustained use, dog-whistling may become understood by a wider array of audiences.

In cases as diverse as Brazil (Nicas 2023), Hungary (Plenta 2020) and India (Vaishnav 2019), populist politicians leverage conspiracy theories to rally their base (Pirro and Taggart 2022), earn electoral support (Marinov and Popova 2022), attack opponents (Pirro and Taggart 2022) and delegitimize electoral results (Nicas 2023). Invoking conspiracy theories (CTs) polarizes publics (Van Prooijen et al. 2022) and increases partisans' belief in those CTs (Enders and Smallpage 2019), which can have negative societal consequences – such as lowering vaccination rates (Juen et al. 2023) or inciting violence (Jolley et al. 2022). Though CTs can mobilize populists' supporters (Marinov and Popova 2022), invoking CTs can also alienate moderates (Przybylski 2018). Populist politicians seem aware of this dilemma. Populists sometimes explicitly push CTs, but other times only discuss the context or history that gave rise to a CT.

To what extent do populists differentiate their discursive strategies on CTs according to the potential audience exposed to their message? Do populists who spread CTs explicitly endorse CTs or use dog-whistling to spread them when speaking to different audiences? Do their strategies shift over time? Though scholars have studied which actors create or promote CTs (Bergmann and Butter 2020; Pirro and Taggart 2022) and which individuals endorse or spread them (Blackington 2021; Einstein and Glick 2015; Krasodomski-Jones 2019; Radnitz 2021), we know less about whether populists strategically leverage CTs, (de)radicalizing their positions for different audiences.

We theorize that populist politicians will use two divergent appeals when invoking a CT, depending on the audience: (1) dog-whistling or (2) barking. Populists who dog-whistle CTs use coded language that hints at their support for a CT without explicitly endorsing it (Åkerlund 2022), making them less likely to appear ‘extreme’ to general audiences. Dog-whistling enables populists to appeal simultaneously to faithful voters – who understand the dog-whistle – and to general audiences – who do not. Populists may be more willing to ‘bark’ amongst CT believers, thereby rallying their hardcore supporters by explicitly endorsing the CT. Yet, the repeated use of dog-whistles may render this rhetoric recognizable to a wider audience. We thus expect populists who dog-whistle CTs for extended periods of time to recognize when a dog-whistle loses its veiling power and switch to explicitly endorsing CTs.

We test our theory by analysing the politics of CTs propagated by leaders of a prominent Polish populist party concerning the 2010 plane crash in Russia, which killed 96 top Polish political, religious and military officials. This crash happened en route to a memorial service for the Katyń massacre – a Soviet NKVD massacre which killed over 21,000 Polish prisoners of war in 1940. The populist party Law and Justice (Prawo i Sprawiedliwość – PiS) discursively places the Katyń massacre alongside the Smoleńsk plane crash – sometimes dog-whistling similarities and sometimes endorsing CTs claiming that a rival party, Civic Platform (Platforma Obywatelska – PO), colluded with Russia to cause the crash.

Using tweets and speeches from PiS, we find descriptive evidence that populists deploy different strategies to disseminate CTs, depending on their target audience. Populists use dog-whistles to invoke historical memory in contexts where the audience is more diverse. By contrast, amongst faithful supporters, populists appear more willing to explicitly endorse crash CTs. Through our findings, we detail how political leaders infuse CTs into a constructed memory regime in systematically different ways, potentially depending on the target audience.

This article's contributions are threefold. First, we answer a call to analyse how audience and context impact whether CTs are shared (Green et al. 2023) by examining populists' discursive strategies for CTs. Second, we meet a demand for increased analysis on the types of conversations occurring across online–offline divides (Tucker et al. 2018), extending fundamental questions of discourse medium to the study of CTs and politics. Third, we push forward our understanding of the micro-foundations for the relationship between populist rule and CTs (Bergmann and Butter 2020; Hawkins and Rovira Kaltwasser 2022; Pirro and Taggart 2022).

Utility of CTs for politicians

CTs blame small groups of powerful people for achieving some objectives and often include a secretive and specified out-group who allegedly undermine the interests of the broader public or political in-group (Cichocka et al. 2016; Miller et al. 2016; Uscinski and Parent 2014). Populist claims and conspiracy theorizing sometimes coincide in this way.¹ Populist politicians frequently claim that their political opponents are ‘shadowy forces that continued to hold onto illegitimate powers to undermine the voice of the people’ (Mudde et al. 2013: 503). These anti-elite assertions align with CTs, which suggest that a group of conspirators secretly harms some broader public (Uscinski et al. 2016).

Given the overlap between populists' communication strategies and conspiracy theories (Hameleers 2021), it is unsurprising that CTs can help cement populist rule. Some populists leverage CTs to rally their base (Plenta 2020; Santini et al. 2022) or to justify policy and institutional changes (Barton Hronešová 2022; Jenne 2018). Though some CTs are rooted in partisan and ideological divides, others can appeal to people who hold anti-establishment views (Uscinski et al. 2021). For supporters of populist parties, which often leverage anti-establishment appeals (Mudde 2004), the invocation of CTs by politicians may be attractive.

Populists push CTs for instrumental purposes including mobilizing voters (Plenta 2020; Vaishnav 2019), delegitimizing election results (Munn 2021; Nicas 2023), increasing support for their policies and programmes (Plenta 2020), attacking opponents, encouraging a sense of crisis and distracting from their governing failures (Pirro and Taggart 2022). Politicians' rhetoric on CTs matters because individuals' ideologies and partisanship influence their CT beliefs (Miller et al. 2016; Pasek et al. 2015; Smallpage et al. 2017; Uscinski et al. 2016). Though CT belief is not correlated with political ideology (Enders et al. 2023), CTs – particularly partisan CTs – appeal more to extremists and strong partisans than moderates (Enders and Smallpage 2019; Enders and Uscinski 2021; van Prooijen et al. 2015).

The power and salience of CTs may vary based on who spreads them and how they use CTs. Politicians who can graft CTs onto existing foreign threats or domestic polarization may gain additional engagement (Blackington and Cayton 2024). Politicians who are electoral ‘losers’ may use CTs more often, leveraging their status as outsiders to gain support (Uscinski and Parent 2014). Due to the anti-establishment nature of populist rhetoric, populists may continue to portray themselves as outsiders once in power. Populists may continue to use CTs to justify their governing decisions and explain away their failures (Pirro and Taggart 2022), which may be especially potent as populist attitudes and conspiratorial beliefs can overlap (Oana and Bojar 2023; Zulianello and Guasti 2023).

Messaging for the audience

Though populists sometimes push CTs to mobilize their base, invoking CTs risks alienating moderates (Przybylski 2018). Models of mass communication suggest that political candidates have incentives to adjust their messages to map onto the preferences of the majority of voters (Druckman et al. 2010). Politicians hoping to win elections may respond to the political preferences of their supporters (Downs 1957) by tailoring their message and emphasizing different parts of their agenda to appeal to different target audiences (Goggin 1984; Miller and Sigelman 1978; Stier et al. 2018). Some politicians present more ideologically extreme positions in primary elections and more moderate positions in general elections (Acree et al. 2020; Brady et al. 2007; Burden 2001). Though incentives to strategically radicalize and deradicalize exist, some voters perceive these strategic candidates as flip-flopping their positions (Burden 2004). As such, politicians sometimes seek other ways to signal varying levels of extremism.

Multi-vocal communication strategies like dog-whistling or double-speak enable politicians to appeal simultaneously to party faithful and general audiences (Albertson 2015; Khoo 2017; Mendelberg 2001) by projecting two messages: one to the general public and one to other activists (Feldman and Jackson 2014). Double-speak allows radical right politicians to appeal to the general public while also keeping far-right activists engaged in the movement (Korhonen 2021). In this way, ‘covert, coded, and ever-adapting language use’ enables the far right to reach broader audiences (Åkerlund 2022: 1810). Since multi-vocal communication strategies are designed to go over the heads of those who lack insider knowledge of the far right's messaging strategy, these audiences may remain unaware of their exposure to these far-right cues, which can spread these messages further (Saul 2018).

Given that some populist parties strategically deradicalize their message to appeal to the general public (Przybylski 2018) and others use more nativist appeals when speaking to partisan outlets than in less partisan forums (Borbáth and Gessler 2023), we expect that populists will also discuss CTs differently depending on the audience. Specifically:

Hypothesis 1: Populists will dog-whistle CTs to general audiences, but ‘bark’ to faithful supporters.

If populists leverage dog-whistles repeatedly over time, mainstream audiences may learn the double-speak. Dog-whistles can be ‘steeped in the echoes of history’, which enables the message to appear coded and deniable (Moshin 2018: 37). However, repeat exposure to dog-whistles may motivate some audiences to seek out further information about their meaning (Kasimov et al. 2023) and mainstream the dog-whistle through repetition (Moshin 2018).

Can populists similarly mainstream CT dog-whistles? We expect that if populists repeatedly dog-whistle a CT, greater numbers of people will come to learn the dog-whistle's meaning. If mainstream audiences learn the dog-whistle's meaning, it will fail to operate as double-speak. From this logic, we derive our second hypothesis:

Hypothesis 2: Over time, populists may lose incentives to dog-whistle and gain incentives to explicitly endorse the CT because the dog-whistle will transform into a bark.

Empirical context

To test our theory, we analyse the Smoleńsk plane crash, which generated CTs that populists endorsed to varying degrees over time. The crash generated long-lasting politically salient CTs, allowing us to analyse changes in politicians' discursive strategies.

The roots of the Smoleńsk plane crash CTs date to 1940, when the Russian People's Commissariat of Internal Affairs (NKVD) murdered nearly 22,000 Polish intelligentsia and military officers near the town of Katyń (Drzewiecka and Hasian 2018). During communism, Soviet leaders claimed the Nazis committed this massacre. Only in 1990 did President Mikhail Gorbachev admit that the NKVD was responsible; he opened archives with corroborating documents. In 2004, President Vladimir Putin closed these archives, which harmed Polish–Russian relations. In 2010, Putin invited top Polish officials to commemorate the 70th anniversary of the Katyń massacre, which heralded a potential improvement in Polish–Russian relations. However, the plane carrying 96 Polish dignitaries to the commemoration crashed en route. All passengers died, including Polish President Lech Kaczyński as well as senior military, government and religious officials.

Official investigation reports revealed that fog and pilot error caused the crash (Zukiewicz and Zimny 2019). Nonetheless, CTs about the crash had emerged by late April 2010 (Niżyńska 2010) and were published in popular media outlets (Zukiewicz and Zimny 2019). Competing accounts of the plane crash proliferated (Myslik et al. 2021; Olechowska 2022), with the key CT motif claiming that the crash was caused by explosions.

The populist PiS party's officials sometimes dog-whistle and sometimes endorse crash CTs (Bilewicz et al. 2019; Davies 2016). For example, in 2012, Jarosław Kaczyński, PiS chairman and brother of the deceased president, accused PO Prime Minister Donald Tusk of conspiring with Russia to cover up the explosions (Davies 2016). Other PiS officials have used government resources to launch unsubstantiated crash investigations.

This case allows us to test our theory that populist CT rhetoric discursively varies by target audience because every year, on the 10 April anniversary, discussions about the crash occur offline and online. We argue that these two unique mediums – online discussion and offline commemoration marches – generate two different target audiences. Whereas the offline events create a space filled with the CT's ‘true believers’, online discourse occurs in a space with more diverse general audiences, generating incentives for populists to dog-whistle CTs rather than endorse them outright.

Offline, PiS politicians actively mobilize their electorate around Smoleńsk CTs. On the 10th of each month in Warsaw, PiS officials participate in and organize masses and marches to commemorate those who died in the crash. Attendance totals in the several thousand annually (Jaroch and Urzykowski 2022). After a mass, participants march to either the Presidential Palace or to a crash monument PiS built in central Warsaw. Kaczyński typically gives a speech surrounded by PiS politicians, clergymen and police officers. Participants in these marches are ‘actively mobilized and funded’ by PiS (Ekiert and Perry 2020: 2). Thus, marches attract strong party supporters and CT ‘true believers’ who expend substantial time and resources to participate in this collective action. Therefore, PiS politicians have few incentives to moderate their speech at these offline events.²

On the monthly crash anniversary, there is also online discourse about Smoleńsk. Online CT discourse can circulate widely to a more diverse audience, as populists can be followed by both faithful supporters and members of the general public. According to research from the Polish Institute for Research on the Internet and Social Media, Polish social media users are heavily female (56%), young (93% aged 18–29) and live in major cities (68% live in cities with over 250,000 inhabitants). These traits correlate with support for PiS's main opposing party, PO. Indeed, 68% of Polish social media users support PO (Dabrowska-Cydzik 2023). Furthermore, on Polish Twitter, many users have politically diverse networks, such that Polish Twitter can ‘facilitate multiple-sided exposure in obtaining political news’ (Matuszewski and Szabo 2019: 10). Previous studies show that online discourse about the Smoleńsk crash is diverse, including ardent appeals to the CTs, standard media coverage and memorialization of those who died in the crash (Blackington and Cayton 2024). Altogether, Polish Twitter provides a more diverse audience than that found at offline gatherings.

Since tweets can quickly spread far beyond their direct follower network due to the retweet feature (Wu et al. 2011), politicians sometimes see Twitter as a way to convey news to, connect with, and share their positions with national and diverse audiences (Heidenreich et al. 2022; Kreiss et al. 2018; Larsson and Skogerbø 2016). Politicians may seek to create social media content that will be received well by their followers, so their followers spread the politicians' messages even further (Kelm 2020). Moderating those online posts may improve the resonance of the content. Rather than referencing the CT itself, we thus expect that online discussion of CTs will centre more around the context or histories of those events that gave rise to a CT. This quasi-memorialization can be interpreted as a dog-whistle, earning recognition by compatriots while going over the heads of non-followers (Albertson 2015). Dog-whistling a CT allows populists to appeal to their base in a more covert manner, while also seeking moderate support. Thus, on Twitter, populists may moderate their conspiratorial claims, alluding to rather than endorsing a CT outright.

As noted above, the Smoleńsk CTs have deep historical roots (Khalitova et al. 2020). Veiled references comparing the Smoleńsk plane crash to the 1940 Katyń massacre or other acts of Russian aggression against Poland may operate as dog-whistles. Over time, we expect these dog-whistles to become mainstream (Kasimov et al. 2023; Moshin 2018), such that the veiled messaging loses its ability to double-speak. As these dog-whistles become understood by a more mainstream audience, the need to dog-whistle may lessen, encouraging PiS politicians to explicitly invoke the CTs both online and offline.

Data

We create two data sources to assess our theory. First, we compile an original dataset of speeches made by PiS party leader Jarosław Kaczyński at the annual 10 April commemoration events from 2012 to 2019 and in 2022.³ The average length of a speech in Polish is 977 words, with a standard deviation of 290 words. Second, we collect a corpus of 433 tweets from the official PiS Twitter account @pisorgpl on 10 April of each year from 2010 to 2022. The average length of a tweet in Polish is 16 words, with a standard deviation of six words.⁴

Our approach builds upon other scholars' work assessing (dis)continuities in the rhetorical strategies that politicians deploy in online and offline settings (see Lacatus 2019; Lacatus and Meibauer 2023). Our article also builds upon work by scholars who compare tweets and speeches to assess communication (in)congruities across mediums (Lacatus 2021; Schaefer et al. 2023). Further, like Ipek Cinar et al. (2020) and Paolo Ricci and Gustavo Venturelli (2023), we focus on party leaders' discourse because they play a critical role in shaping the party's agenda (Fouirnaies 2018) as well as partisans' attitudes and beliefs (Broockman and Butler 2017; Zaller 1992).

Analysis

We leverage a mixed-methods approach to analyse whether populists diverge in dog-whistling or explicitly referencing CTs across mediums, deploying both structural topic models (STM) and qualitative content analysis.⁵

Structural topic models

We employ STM to parse the underlying themes in the Kaczyński speeches and PiS tweets corpora. Topic models are an unsupervised learning method in which the topics derived are ‘inferred’ rather than ‘assumed’ a priori by the researchers (Grimmer et al. 2022; Lucas et al. 2015; Roberts et al. 2014). Through STM, we can assess the extent of CT and non-CT discourse in both corpora. Further, STM allows us to model ‘topics’, or words that commonly appear alongside one another in our corpora, as a function of other metadata.⁶ Importantly, topics are derived from the corpus by the STM and not assigned a priori by researchers. We assess topic content (the words used to discuss a topic) and prevalence (how often a topic is discussed) as a function of a tweet or speech's year.⁷

We first consider what themes are prominent across both the tweets and speeches corpora. We specify the number of topics with attention to the dispersion parameter, as detailed by Matt Taddy (2012), yielding five topics for the tweet corpus. To maintain comparability, we also use five topics for the speech corpus.⁸ We present the words with the highest probability of being associated with a given topic, both for speeches (Table 1) and for tweets (Table 2). We label each topic based on these words, as well as frequent and exclusive words for each category. We include the full unlabelled topic-year results and discuss the set of words in each topic in more detail in the Supplementary Material, Appendix C.

Table 1. Speech Topics

Note: Bolded words are conspiratorial. Italicized words memorialize the crash.

Table 2. Twitter Topics

Note: Bolded words are conspiratorial. Italicized words memorialize the crash.

Assessing the offline speeches, we find that explicit references to the Smoleńsk CTs comprise over 33% of the speech corpus. This topic explicitly references activists who promote monthly events that regularly invoke crash CTs, such as the editor-in-chief of the right-wing weekly Gazeta Polska, Tomasz Sakiewicz. Indeed, both Sakiewicz's first name and Gazeta Polska are regularly mentioned in Kaczyński's annual speeches. Similarly, the second most common topic – making up 22% of the speech content – asserts that the events in Smoleńsk were a ‘crime’ for which the ‘truth’ should be identified. We title this group ‘crash investigation and CT’, as these phrases are suggestive of a need to ‘discuss’ and re-evaluate the ‘response’ to the crash. The offline crash investigation discourse is unsurprising, given PiS's reopening of an investigation into the origins of the crash when it returned to power in 2015. The commission tasked with investigating the crash deliberately ignored evidence that would falsify Smoleńsk CTs (Ptak 2022). Finally, the smallest topic of the speech corpus explicitly characterizes the crash as an ‘evil attack’. Together, over 69% of offline speeches advance crash CTs.

By contrast, the largest topic in the Twitter corpus is memorialization – appearing in almost 33% of the corpus. This category explicitly characterizes the crash as a ‘disaster’, a unique word choice that does not attribute blame for what happened at Smoleńsk but acknowledges the tragedy. This contrasts with the language terming the crash a ‘crime’ or ‘attack’ in Kaczyński's speeches. This category also acknowledges the ‘anniversary’ of the death of ‘President Kaczyński’, focusing more on his loss than the political origins and/or ramifications surrounding his death.

Similarly, the second most common topic for the tweet corpus mnemonically ties those who died in the crash to the Katyń massacre, the event that Polish officials flew to Smoleńsk to commemorate. Linking Smoleńsk to Katyń is a dog-whistle for the Smoleńsk CTs, as it portrays the plane crash as ‘the latest narrative arc in a calamitous book of Polish memory’ (Soroka 2022: 335) and suggests support for Smoleńsk CTs without endorsing them. The next two categories focus on logistics of the offline speech, directing followers to where and how they may find information. Finally, the smallest category – at only 12% of the corpus – appeals to discovering the ‘truth’. While we label this category as CTs, the words identified by the STM are less explicit CT references than those in the speeches.

In both corpora, we find support for our first and second hypotheses. PiS invokes more explicit appeals to CTs amongst its base. When speaking to a broader audience, CT references are less explicit but rather are buried amongst other appeals to historical memory.

To gain a sense of whether and how these appeals may mainstream, we leverage the STM's ability to account for how a topic's content and prevalence varies as a function of time. Figures 1 and 2 provide the expected proportion of the corpus comprising each topic over time. In both figures we colour code topics by those that make CT appeals (red), those appealing to politically relevant topics that are not specific CT endorsements (blue) and those topics focusing on programmatic or logistics related information (black).

Figure 1. Expected Topic Proportion Over Time – Speeches

Figure 2. Expected Topic Proportions Over Time – Tweets

Figure 1 shows that Kaczyński's rhetoric at marches oscillates between explicit CT appeals, agitating to spread CTs offline, or detailing the CT investigation. All three topics remain common in his discourse over time. Though we must remain conservative in our interpretation given the limited number of speeches in our corpus, the ‘crash CT investigation’ category rises from a predicted 0% of the corpus in 2012 to roughly 50% by 2022.

PiS also leverages memorialization rhetoric to greater degrees over time in both corpora. Between 2019 and 2022, Kaczyński's speeches invoke more memorialization rhetoric. This topic jumps from about 10% of the speeches to roughly 40%. Similarly, on Twitter, while CTs remain a small proportion of PiS tweets, explicit CT endorsements increase over time. In this manner, the speech and tweet time trends resemble each other.

Importantly, over time, the confidence intervals widen in both the tweet and the speech corpora for both the memorialization (blue) and CT (red) topics. These widened confidence intervals occur alongside the uptick in both CT and memorialization rhetoric. The ambiguity around topic usage rates and increased usage of both rhetorical strategies suggest that PiS officials increasingly couple dog-whistles with explicit CT endorsements online and offline.

Though descriptive, this trend could reflect a blurring of the distinction between dog-whistling and CT endorsements. Dog-whistling and barking may not be stable strategies that elites can continuously peddle to different audiences over time. Instead, as elites use dog-whistles repeatedly, this rhetoric loses its ability to double-speak. We interpret this evidence as supporting our third hypothesis that the distinction between ‘dog-whistles’ and ‘barks’ may become less explicit as elite CT appeals are recognized by mainstream audiences.

Qualitative content analysis

Our unsupervised STMs suggest a substantive distinction between online and offline PiS CT rhetoric. To gain a better sense of the tenor of these conversations, we turn to the rhetoric itself. We find that while Kaczyński sometimes invokes dog-whistles, he almost always explicitly endorses CTs about the crash in his speeches. Likewise, online, even language appearing within the ‘conspiracy theory’ category often reflects dog-whistling rhetoric.

Offline

In seven of the nine years in our sample of speeches, Kaczyński ‘barked’ crash CTs. In 2022, he stated that the crash was ‘an assassination’. He subsequently noted a need to identify ‘how to treat those responsible for the crime … those that made the decision and those who carried it out, here in Poland, but above all, there in Russia’. Kaczyński's rhetoric endorsed crash CTs at this 2022 event. Similarly, in 2016, Kaczyński stated of Tusk and PO:

They wanted to kill this memory [of the crash]; they are afraid of being banned … that previous government was responsible for it. Not this one of Kopacz, of course, only the one of Donald Tusk … The Smoleńsk tragedy was no accident.

Alongside explicit statements blaming PO for instigating the crash, Kaczyński interwove various dog-whistles. He implied that foreign or domestic actors were responsible for hiding ‘the truth’, which Poles needed to identify. These statements imply support for a CT without explicitly invoking one. For example, in 2019, Kaczyński stated that

The trip to Smoleńsk for Katyń [commemorations] was not an ordinary trip. They were on a mission which promoted a broader cause than just the memory of those who were murdered 70 years ago, now almost 80. It was about Poland … about what is extremely important and what was taken from us not only in the terrible years of World War II, but also later after World War II and after 1989: It is the right to truth. In the case of our nation, this right to truth is also the right to dignity … We must fight for this truth.

To believers of Smoleńsk CTs, this quote dog-whistles by suggesting that the truth surrounding the plane crash remains unknown. However, to a general audience, this quote appears more mild – focused on dignity, truth and national identity rather than a CT.

Similarly, in 2018, Kaczyński used double-speak by claiming that the marches and investigations ‘have contributed to the fact that we can say once again that we have won … we served our homeland, we served the truth and dignity of Poland’. In this quote, ‘the truth’ appears neutral. However, for CT believers, ‘the truth’ suggests that Kaczyński is uncovering the allegedly nefarious causes of the plane crash. Indeed, dog-whistling rhetoric focusing on truth appears in all of Kaczyński's anniversary speeches in our sample.

Online

The large number of tweets within our corpus allow STM to identify the ‘most’ representative tweets in each topic identified within our corpus.⁹ These tweets provide a qualitative depiction of the party's online rhetoric.

We start with the 12% of the Twitter corpus identified by the STM analysis as CT references. All five examples of the most representative tweets in this topic quote excerpts of Kaczyński's offline speeches, meaning online barks are recycled from the offline events. For example, in 2015 PiS tweeted, ‘Without the truth about Smoleńsk, we cannot build a strong country – says J. Kaczyński in front of the Presidential Palace.’ This vague appeal to the ‘truth’ of Smoleńsk invokes the crash but is less explicit than the suggestions of Russian explosives or invocation of Tusk's culpability found in the speeches.

More commonly, tweets emphasize standard mourning practices like attending mass or laying flowers on the graves of those who died. For example, in 2021, PiS tweeted on the crash anniversary that ‘On behalf of the PiS Parliamentary Club … PiS MP @TeresaWargocka laid flowers at the memorial to the Smoleńsk catastrophe.’ Similar tweets comprise all five of the ‘memorialization’ categories of tweets, which are roughly 33% of the corpus. These exemplar tweets are largely culled from earlier years (2013, 2014, 2016, 2019 and 2021), reflecting the emphasis on pure memorialization online in the early years after the crash.

Over time, dog-whistling rhetoric wanes whereas explicit CT endorsements increase. This is visible by looking at the words specific to each topic for each year. For a full list of unique words for each year, see Supplementary Material, Appendix C. By 2022, the ‘Katyń memory’ category uniquely included the word ‘crime’ and ‘Ukraine’, as visible in Table 3. Whereas early years of this category focused on attending masses or laying flowers on the graves of those who died in the plane crash, by 2022 even this more neutral category included explicit conspiratorial references. PiS elites suggest that the memorialization of Katyń, the 2010 crash, and Russia's full-fledged invasion of Ukraine constitute examples of ‘crime[s]’.

Table 3. Katyń Memory Tweet STM Topic: Unique Words to Topic by Year

Dog-whistling and barking rhetoric merges with time. On the one hand, this fusion of dog-whistles and barks may emerge from an event rendering some CT claims as more ‘believable’. In this case, Russia's willingness to seek violent regime change in Ukraine may lend credibility to the idea that Russia similarly may have caused the Smoleńsk crash. On the other hand, after more than a decade of PiS elites dog-whistling Smoleńsk CTs, the average person may understand the meaning of the dog-whistle. In this way, dog-whistles may lose their veiling power, which could create incentives for politicians to simply bark.

The recent blurring between the CT and memorialization rhetoric suggested in our STM is also visible in the raw tweets. For example, in a tweet quoting Kaczyński's 2022 speech, PiS tweets, ‘What happened at Smoleńsk? We knew that it was a catastrophe, but was it an ordinary one, an accident, or something entirely different?’ This appeal – much more explicit than the vague references to ‘truth’ or ‘justice’ referenced in earlier years' tweets – suggests a newfound willingness to bark both on- and offline.

Qualitative analysis of the rhetoric deployed in both Kaczyński's speeches and PiS's tweets on 10 April indicates that while PiS politicians regularly dog-whistle online, PiS politicians regularly endorse CTs in-person. Where true believers gather, politicians may lack incentives to moderate their claims. Further, since dog-whistling often occurs alongside explicit endorsements, these offline outlets may ‘train’ potential CT believers to spot dog-whistles. Finally, as dog-whistles become more mainstreamed, politicians may distinguish less between target audiences when tailoring CT rhetoric.

Alternative explanations

Other inherent differences between offline speeches and tweets may drive our findings, such as text length. Politicians have more chances to endorse CTs in speeches than in character-limited tweets. Since existing research finds that politicians often use similar communication strategies when communicating in speeches and tweets (Lacatus 2021; Schaefer et al. 2023), we think the differences we find are due to strategy, not length. While politicians often sync their communication strategies across mediums (Lacatus 2021; Schaefer et al. 2023), they might not for CT endorsements.

Differences between the PiS party's tweets and Kaczyński's speeches could also emerge from variation in speaker. We doubt this for three reasons. First, Kaczyński plays ‘a leading role’ in the official PiS Twitter account (Rawski et al. 2021: 14), which bolsters the comparability of his speeches and PiS tweets. Second, 164 out of 433 tweets in our corpus directly quote from Jarosław Kaczyński's speeches, which indicates a connection between Kaczyński's speeches and PiS tweets. Since the account quotes less conspiratorial parts of Kaczyński's speeches on Twitter, PiS may strategically minimize its conspiratorial messages online.¹⁰ Third, PiS is described as ‘ruled single-handedly by Jarosław Kaczyński’ as he ‘makes all critical decisions, whether personnel- or policy-related’ (Pludowski 2022: 157). He often operates as a ‘backstage string-puller’ (Stanley 2020: 184) and has an ‘absolutist leadership style’ over ‘strongly centralized, hierarchical’ party structures (Pytlas 2021: 341). Given Kaczyński's extensive formal powers, tweets posted by the party likely could not substantially diverge from his own messaging without his knowledge or consent.

If differences existed between the PiS tweets and Kaczyński speeches due to who controls the messaging alone, we would expect to see differences in how other PiS officials discuss Smoleńsk CTs. To offer further evidence that other PiS officials similarly invoke CTs at offline events, we transcribe, translate and analyse speeches given by President Andrzej Duda from 2015 to 2023. Results shown in Supplementary Material, Appendix D show that Duda also invokes and endorses CTs in his commemoration speeches over time. Our results suggest a broader PiS trend in endorsing CTs more strongly at in-person events. We also evaluate whether our findings hold for a broader subset of PiS officials online, besides just the party's official account. We present these results in Supplementary Material, Appendix B; they largely mirror our main text results. PiS officials tend to invoke dog-whistles more and explicitly endorse CTs less than non-PiS officials on Twitter.

We might also be worried that the increase in joint use of dog-whistles alongside CTs together is due to something besides the habituation of the audiences to the dog-whistles. Namely, major political events like elections could make CTs more salient. If elections drove our results, we would expect to see a marked uptick or downtick in CT endorsements in Figures 1 and 2 in election years. We do not observe such changes.¹¹ We also do not see a spike or fall in CTs that blame Tusk or PO for the crash online or offline during election years. Instead, a smooth rise in CT endorsement occurs, which supports our interpretation of a mainstreaming of dog-whistles and a merging of dog-whistling and barking strategies.

Another major event that could increase the use of CTs relative to dog-whistles in this case is the full-fledged Russian invasion of Ukraine in February 2022. Elsewhere, we find that PiS politicians earn more social media engagement when invoking CTs after the invasion than before it (Blackington and Cayton 2024). In Appendix B, we test whether PiS officials also use CTs more often. We find that the majority of PiS officials' rhetoric includes dog-whistles though we observe an uptick in conspiratorial discourse after the invasion. Thus, dog-whistles and barks continue to merge over time – in line with our theory.

Alternatively, the rise of right-wing populism during this period might shift the Overton window for politicians, enabling the mainstreaming of CTs. If this was true, we would expect that CT use would be on the rise, while dog-whistles would decrease. This explanation is not necessarily at odds with our posited mechanism. Indeed, if prominent populist leaders endorse CTs abroad, then PiS politicians may also feel more emboldened to do so. This aligns with our observed increase in CT use to audiences comprised of party faithful. However, we also find that, amongst general audiences, dog-whistles and CTs increase together. Thus, while the Overton window may reflect the normalization of CT barks to party faithful, it does not account for the strategic, combined use of both dog-whistles and barks to general audiences.

Finally, over time, PiS may become increasingly less concerned with what the EU thinks of it peddling CTs. If PiS were indifferent to the EU, this would decrease the need to dog-whistle. In that instance, PiS may continue radicalizing its Smoleńsk rhetoric. Given that 89% of Poles have a positive view of the EU, including 83% of PiS partisans (Poushter et al. 2022), PiS officials likely do not make decisions about CT rhetoric with indifference towards the EU.

Conclusion

Bringing together textual data across online and offline discussions of the Smoleńsk CT, we offer descriptive evidence that politicians engage with CT discourse differently depending on the audience. Amongst general audiences, populists dog-whistle CTs, while explicitly endorsing these CTs to supporters. At events attended by party faithful, CT endorsements by populists are common. Over time, dog-whistles may become mainstreamed. As time passes, politicians may no longer distinguish between target audiences and may instead explicitly invoke CTs, perhaps knowing that the dog-whistle is no longer covert.

Future research should investigate whether and how CT dog-whistles are created and how audiences learn the underlying meanings of these dog-whistles. The process of intertwining dog-whistles with explicit CT references may train the meaning of these dog-whistles such that CT believers can recognize this double-speak in other contexts. While we identify that populists may leverage different rhetorical strategies when invoking CTs to appeal to different audiences, several questions remain open. Do people with different dispositions come to different conclusions about the meaning of the dog-whistles? Are conspiracy theorists simply motivated to hear what they want when exposed to dog-whistles, like other sources of partisan motivation? Alternatively, are conspiratorial dog-whistles carefully trained over time on audiences of party loyalists who, in turn, mainstream the conspiratorial message?

Our results suggest that the literature's broad findings about how online political communication changes the behaviour of ordinary people may not apply to elite political communication. Whereas ordinary people are less civil and more anti-social online than offline (Baek et al. 2012; Cheng et al. 2017) and whereas status-seeking individuals engage in more bombastic conversations online (Bor and Petersen 2022), politicians behave differently. Even those politicians from parties that use CTs to build mass support, such as PiS (Przybylski 2018), exhibit different behavioural patterns online and offline. While populists bark at offline events, they more frequently dog-whistle online.

These findings suggest a need to better understand differences in how elites target different audiences when propagating CTs. Indeed, work by Maria Snegovaya and Tsveta Petrova (2020) shows that governing populists generate a feeling amongst their supporters that they are being threatened or under a state of emergency – and that the populists can help address these perceptions. Our article suggests that barking and dog-whistling CTs may be one way in which populists can maintain a sense of threat.

There is a growing need to understand elite opinion and behaviour (Druckman 2022), which is particularly acute for the populism and CT literature. Elites remain central to populist success (Tworzecki 2019), which makes understanding their online and offline communication strategies important. We provide descriptive evidence, which suggests that there is variation in elite CT communication, depending on the target audience, which is especially important given that CTs often spread via elite endorsement.