Skip to main content
Search
Search
    Quick Search anywhere
    Quick search in Citations
Advanced Search
Skip main navigationClose Drawer MenuOpen Drawer Menu
Previous article
Next article
Free AccessRegistered Report

Cognitive and Emotional Responses to Russian State-Sponsored Media Narratives in International Audiences

Published Online:January 19, 2023https://doi.org/10.1027/1864-1105/a000371

Abstract

Abstract: Russia utilizes state-sponsored news media outlets, such as RT or Sputnik, to project antagonistic strategic narratives into targeted societies and perturb international audiences. While psychological responses to this conduct are frequently assumed, there is a lack of causal evidence demonstrating this. Using a transdisciplinary perspective, we conducted four survey experiments that tested two path models predicting possible cognitive and emotional responses to two narrative strategies that Russian state-sponsored media employ: destruction, which portrays a state as weak and chaotic, and suppression, which portrays a state as indecent and morally deviant. The experiments had between-participant designs, where participants read either an article demonstrating a strategy or a control text, and then indicated their responses to several trust and emotional variables. Participants were either Swedish or Dutch citizens, to build on previous analyses of Russian narration about Sweden and The Netherlands. Path analyses revealed significant differences between the conditions on several response variables. However, we found no evidence that these effects were mediated by generalized realistic or symbolic threat perceptions. We contribute preliminary insights into potential causal links between Russian antagonistic narrative strategies and specific psychological responses. This study, and its overarching research agenda, should have implications for practitioners seeking to counter Russian information influence.

State-sponsored media outlets, such as RT (formerly Russia Today) and Sputnik, are considered vehicles for the Kremlin to project antagonistic strategic narration. Strategic narratives are tools that allow political actors to construct and shape perceptions of the reality of international politics in foreign or domestic audiences (Miskimmon et al., 2017), and are assembled using varying levels of factual, misleading or oversimplified, or intentionally false information. Strategic narratives imbue news about current events with meaning and storyline in a manner that benefits a political actor’s international objectives (Schmitt, 2018). Through strategic narration, political actors can antagonistically distort international audiences’ perceptions of their domestic political reality and, consequently, cultivate tension within societies of targeted states. Consequently, RT and Sputnik, which until March 2, 2022 were freely accessible to international audiences online, are thought to have bolstered the Kremlin’s ability to engage in malign information influence – the “use of information as a weapon to inflict harm upon others” (Wagnsson, 2020, p. 1). Crucially, the outlets are thought to initiate narratives that are later repackaged and dispersed through social or more mainstream media platforms (Ramsay & Robertshaw, 2019), indirectly reaching more citizens than their limited direct audiences (Crilley et al., 2020). Demonstrative of the outlets’ perceived impact internationally, the European Union (EU) controversially chose to ban RT and Sputnik in European borders as a response to Russia’s invasion of Ukraine.

Given this state of affairs, it is important to understand how international (and particularly European) audiences respond to Russian narratives describing their own state. Destabilizing societies from within is seen as central to the Kremlin’s ambitions in the West, part of its “hybrid warfare” model (Galeotti, 2017). While reductions in trust or the triggering of emotional responses are often discussed as responses to Russian malign information influence (Ingram, 2020), this psychological dimension has seldom been studied in depth. Empirical research examining effects is limited (cf. Crilley & Chatterje-Doody, 2020; Fisher, 2020), and clear causal evidence is even scarcer. Questions of whether Russian antagonistic strategic narratives elicit destabilizing responses, and what these effects specifically are, therefore warrant closer scrutiny. Evidence of causal relationships would have value for policymakers and practitioners who seek to develop countermeasures to Russian influence, providing insights that may help specify the development of counter-narratives or societal deterrence endeavors.

Experimental research has been proposed as a method of attaining such evidence (Hoyle et al., 2021a), whereby experiments can test for psychological responses to narratives identified in prior qualitative analyses. In this study, we will use survey experiments to gauge the cognitive and emotional responses to two types of narrative strategies identified in Russian state-sponsored media (Wagnsson & Barzanje, 2021). As a first study, we measure immediate responses. However, the influence of strategic narratives is a long-term process that would point to the utility of longitudinal observations, a logic also reinforced by the field (e.g., Gaines et al., 2007). Such studies are, however, resource-heavy and require intense planning. They are therefore preferable to conduct once indications of causal relationships are observed. The current study aims to deliver insights that can be explored in longitudinal formats.

Russian Antagonistic Narrative Strategies

Wagnsson and Barzanje (2021) advanced a framework accounting for the narrative strategies Russian state-sponsored media uses to narrate, and consequently, harm targeted states. Two key strategies defined were destruction and suppression. Destruction characterizes narratives that seek to denigrate a state’s power, authority, and control. These narratives focus on amplifying crime, disorder, and internal divisions; weakening domestic support for the military; and harming the target’s image as an attractive partner for alliances (p. 12). Conversely, suppression narratives aim to tarnish a state’s cultural image and status through narratives that pathologize its “way of life, leadership or population,” presenting it as “morally deviant,” and emphasizing liberal or nontraditional values within the populace (p. 12). Several studies have demonstrated the use of destruction and suppression in their analyses examining Russian narration of different states (e.g., Deverell et al., 2020; Hellman, 2021). Some of the richest examples can be found in qualitative analyses of Sputnik’s narration of Swedish society between 2014 and 2018 (Wagnsson & Barzanje, 2021) – which yielded the strategies’ conceptualization – and of RT’s narration of the Dutch society between 2018 and 2020 (Hoyle et al., 2021b). These analyses demonstrate remarkable convergence in how the strategies are constructed across different states, with several narratives appearing to be almost exact replicas of each other.

For destruction, narratives such as “the unsafe space” in Swedish coverage and “a dangerous society” in Dutch coverage portrayed the states as overrun with crime, frequently sensationalizing one-off, random instances of violence or crime. “The conflict-torn space” narrative in Swedish narration shared aspects of the “divided society” narrative in Dutch narration, both depicting the societies as acutely polarized across an array of contentious societal issues and on the brink of internal conflict. “Foolish institutions,” a narrative identified in Dutch coverage, further converged with Swedish narration emphasizing the culpability of various state institutions – such as the government, police, or military – for causing, or at least for failing to address, the dire state of affairs in either society. These articles often mocked Dutch or Swedish state institutions for their incompetence, portraying them as disorganized and clueless.

Suppression, meanwhile, was built through the emphasis of untraditional values festering in both states. This frequently involved blaming “the other” – progressive groups such as pro-feminist, anti-racism, or pro-LGBTQ+ activists, or migrants – for threatening the status quo and imposing their nontraditional views on others. Narratives such as “the unsexy space” or “the ultra-modern space” in Swedish coverage, or “a weird society” in Dutch coverage, derided the states’ liberal tendencies and accentuated the detriment of their progressive policymaking for the general population. For The Netherlands, this “hyperliberalization” was positioned more as inherent to Dutch society, whereas it was positioned more as an effect of Swedish “moral decay,” leading to Sweden’s downfall from moral superiority. Sometimes, articles contained references to social media reactions, as if to evidence the scorn of the international community. Indeed, the suppression narratives identified appeared intent on making both Sweden and The Netherlands “look like ridiculous space[s] that [are] not to be taken seriously” (Wagnsson & Barzanje, 2021, p. 8).

The identification of these strategies connects to broader understandings of how the Kremlin uses international media to carve an image aspiring to be seen as a prestigious and authoritative actor in the international community (Miskimmon & O’Loughlin, 2017) and as a global defender of traditional values (Feklyunina, 2016), and how it uses antagonistic narration of other states as a way of fortifying its own image. In reflecting these fundamental drivers of Russian foreign policy, destruction and suppression account for the strategic intentions behind the projection of certain narratives by Kremlin-affiliated media. They, therefore, provide a solid foundation upon which to begin building models of responses that Russian antagonistic narration may elicit in international audiences. However, to properly account for these responses, transdisciplinarity is essential: This international relations perspective must be integrated with psychological theory pertinent to how foreign audiences exposed to these narratives might react. A vital step, then, is to assimilate these disciplines into testable transdisciplinary models.

Toward a Transdisciplinary Model

Extant literature on psychological responses to news media suggests that exposure to such narratives should indeed elicit an array of cognitive and emotional responses. News media coverage of topics fitting the destruction strategy has been predominantly shown to trigger effects on institutional trust, anger, and fear. For example, news media emphasizing the threat of crime or terrorism in society decreased institutional trust (Liebertz & Bunch, 2019) and anger and fear (e.g., Nellis & Savage, 2012; Shoshani & Slone, 2008). Similarly, coverage of institutional misconduct or incompetence by, for example, the national government or police, has been shown to trigger similar effects (e.g., Kepplinger et al., 2012; van Elsas et al., 2020).

Anger responses have also been evidenced after exposure to news media about suppression topics, such as groups who challenge the status quo and whose lifestyles deviate from traditional societal values (e.g., Shaver et al., 2017). Similarly, media representation of such groups can trigger disgust responses (e.g., Casey, 2016; Dalsklev & Kunst, 2015), and affect elements of outgroup trust, such as reducing feelings of warmth (Shaver et al., 2017) and positive attitudes toward the outgroup (Pinsof & Haselton, 2017). News media portrayals of one’s nation as deviant have also been shown to invoke shame feelings in citizens (Chekroun & Nugier, 2011). There is evidence to suggest, then, that destruction and suppression narratives should trigger direct effects on such response outcomes.

However, we believe there is scope to nuance – and thus increase the explanatory power behind – the hypothesized relationships between the narrative strategies and described responses. Specifically, we propose that the strategies can be reconceptualized as psychological mechanisms of intergroup threat theory (Stephan & Stephan, 2000). This theory asserts that the perception of threat dictates how people respond to certain stimuli (e.g., a media narrative), defining two types of perceived threats: realistic threats and symbolic threats. Realistic threats are perceived threats to an individual or their ingroup’s power, physical wellbeing, or economic prospects, whereas symbolic threats are perceived threats to an individual or their ingroup’s values, image, identity, or way of life. Given the destruction strategy’s focus on portraying rising criminality, societal division, and the failure and incompetency of state institutions, we propose that destruction narratives act as mechanisms that increase generalized realistic threat perceptions. Conversely, we conceptualize suppression narratives, which focus on marring a state’s national image as moral and decent, as mechanisms that increase generalized symbolic threats perceptions. One can therefore conceptualize perceptions of realistic and symbolic threats as mediators between the narrative strategies and any consequent responses they might trigger. In defining these mediation roles, one can draw on a breadth of research testing intergroup threat theory to support the predictions of specific psychological responses to the two narrative strategies.

Studies investigating realistic threat perceptions reinforce predictions of reductions in institutional trust and increases in anger and fear responses in international audiences. Research has shown that realistic threat perceptions, such as perceived danger and low security provisions, economic precarity, or disingenuousness and incompetence in institutional actors, can lead to significant reductions in institutional trust levels in citizens (e.g., Ervasti et al., 2019; Maier, 2011). Studies have also associated such realistic threat perceptions with increased fear (e.g., Jackson, 2009; Jetten et al., 2017) and anger responses (e.g., Ditton et al., 2017; Sadler et al., 2005), with the degree of anger, as opposed to fear, shown to increase as attributions of blame on state institutions were made more explicit (Wagner, 2014).

Extant literature examining symbolic threat perceptions also supports predictions of reduced outgroup trust and increased anger, shame, and disgust responses in international audiences. The effects of suppression, in part, will be linked to reputation, with the Russian media’s focus on pathologizing foreign states’ liberal proclivities leading to a perceived degradation of this reputation for citizens of those states – a symbolic threat (Chen et al., 2020). This seems especially likely considering emerging research revealing that RT and Sputnik’s messaging is predominantly sought by more conservative citizens who are already critical of their state’s liberal values (Wagnsson, 2022). This notion of reputation degradation converges with previous assertions by Wagnsson and Barzanje that suppression should trigger “status loss for the readers” (2021, p. 12). Perceived general degradation to an ingroup’s image, such as national identity, has been linked to anger and shame responses (Allpress et al., 2014; Iyer et al., 2007). Research has also shown that the symbolic threat of an outgroup imposing on the ingroup, particularly if they seem to be attempting to promote unconventional or untraditional views, can lead to reductions in positive attitudes or trust (Schmuck & Matthes, 2015, 2017) and heightened disgust responses toward the outgroup (e.g., Tapias et al., 2007; Wirtz et al., 2016).

The Present Study

With this, we can create two independent transdisciplinary path models of predicted responses in international audiences, depicted in Figure 1. As a first attempt at defining possible psychological responses to Russian antagonistic narration, the models are purposefully preliminary, and responses are conceptualized as independent. This does, however, omit possible complexity between the variables. One might propose dependencies between observed trust and emotion responses that would suggest repositioning them in a temporal sequence where trust responses are predicated on emotional responses. It also discounts the effects of possible moderators, such as personality traits or audience characteristics, which might influence responses. With audiences commonly viewed as active interpreters, processing media messaging against the backdrop of their own personal and social settings, understanding who is most reactive to types of messaging is key to understanding reception of Russian antagonistic strategic narration. Augmenting the models by including such aspects are future goals of a research agenda that the current study initiates.

Figure 1 Path models depicting the expected cognitive and emotional responses to destruction (A) and suppression (B) narratives.

The current study aims to test the assertions of these proposed path models, with an overarching research question of: What cognitive and emotional responses are triggered in international audiences when they are exposed to Russian state-sponsored media narratives, employing the destruction or suppression strategies, about their own state? We will conduct four separate studies, testing both strategies in two different audiences: Sweden and The Netherlands. Both are states that have been recently targeted by Russian malign information influence (AIVD, 2021; SVT Nyheter, 2018), and whose targeting has been the focus of intense domestic discourse. Further, both states have been identified as two examples of institutionally strong European states whom Russia seeks to destabilize (Galeotti, 2017), and have international images as beacons of European progressive policy-making, providing germane material for Russia to malign. Lastly, the aforementioned qualitative analyses of Russian media narration about these states provide further motivation as they offer germane narrative material to test experimentally. We do not expect major differences between these two countries, and results will be used to generalize our conclusions.

We divide the overarching research question into two main subquestions. (1) Do destruction and suppression narratives about a particular state elicit direct psychological responses in the audiences from that state? We expect that:

Hypothesis 1 (H1):

Exposure to destruction narratives should lead to lower levels of institutional trust (H1a) and higher levels of anger (H1b) and fear (H1c) when compared to a control.

Hypothesis 2 (H2):

Exposure to suppression narratives should lead to lower levels of outgroup trust (H2a) and higher levels of anger (H2b), shame (H2c), and disgust (H2d) when compared to a control.

Our second subquestion concerns the proposed mediation mechanism by realistic and symbolic threat perceptions: Do (types of) threat perceptions mediate the relationships between the destruction or suppression narratives and the psychological responses they elicit? We expect that:

Hypothesis 3 (H3):

Realistic threat perceptions will mediate the effects of destruction narratives on institutional trust (H3a), anger (H3b), and fear (H3c) levels.

Hypothesis 4 (H4):

Symbolic threat perceptions will mediate the effect of suppression narratives on outgroup trust (H4a), anger (H4b), shame (H4c), and disgust (H4d) levels.

All hypotheses are depicted in Figure 1.

Method

Participants

We collected representative samples via Novus, a Swedish market research company. Sample weights were applied to ensure representativeness. For more information, see S2 in the Electronic Supplemental Material, ESM 1. Novus recruit participants randomly through telephone interviews and targeted personal invitations to underrepresented target groups. They are in accordance with GDPR legislation, and are a full member of ESOMAR. Participants were Dutch or Swedish citizens and had to have command of the Swedish or Dutch language. Only participants older than 18 years were included. Outliers in completion time (+3.00 SD) and participants who failed the attention check were excluded. Our final sample consisted of 340 participants for Study 1 (63 removed), 353 participants for Study 2 (63 removed), 366 participants for Study 3 (73 removed), and 375 participants for Study 4 (63 removed). More specific demographic insights can be found in ESM 1 (S1 – Table E1).

Design

The research was divided into four survey experiments. Studies 1 and 2 used Swedish participants, whereas Studies 3 and 4 used Dutch participants. Conversely, Studies 1 and 3 investigated responses to destruction narratives, and Studies 2 and 4 investigated responses to suppression narratives. We treated these as distinct studies as opposed to employing a factorial design since the dependent variables and encompassing path models are different for either strategy. Each study employed a single-factor, between-participants experimental design. The design scheme is also illustrated in Figure 2.

Figure 2 Diagram presenting the configuration of the studies.

Materials

Stimulus Materials

All described stimulus materials can be viewed on the OSF repository (https://osf.io/yb6kr/).

Experimental Conditions

Each experimental condition consisted of exposure to two news articles pulled from Sputnik or RT’s coverage of Swedish or Dutch society, based on the aforementioned qualitative analyses (Hoyle et al., 2021b; Wagnsson & Barzanje, 2021). The articles were chosen based on their representativeness of the antagonistic strategy that they represent, and therefore each strategy focuses on different topics. While some themes are possible to broach from both strategies, most are distinct to one strategy. For example, while crime could be narrated from a suppression perspective (e.g., because of societal deviance), it is primarily a destruction topic. This distinction, therefore, precluded the use of manipulations presenting a specific topic from both strategies’ perspectives as cross-contamination could occur. Nevertheless, all stimuli centered on domestic life and society in the particular state.

In Study 1, the destruction strategy was operationalized by two articles highlighting, first, crime in Sweden and, second, the ineptitude of the police and government in tackling this. These articles broach themes identified in analyses of Swedish narration, namely that an “unsafe space” has emerged due to passive and naïve Swedish leadership. For Study 2, the suppression strategy was operationalized by two articles describing, first, how the Church of Sweden is composing a book of hymns that will be more inclusive and gender-neutral, and second, how the Swedish national museum has added “insane” racism warnings to various artworks. The articles relay these stories in a caustic and derogatory tone, seemingly to undermine these actions. By deriding inclusivity in this way, the articles touch upon strategic narratives unearthed by the aforementioned analyses of Sputnik’s Swedish coverage.

In Study 3, the destruction strategy was operationalized similarly to Study 1, by two articles describing crime in The Netherlands and detailing how the police have voiced concerns of being overstretched. Again, these articles focus on portraying The Netherlands as a dangerous place, with criminals running riot and victims largely unprotected. They also chastise the Dutch state for leaving the police insufficiently trained and under-financed. These articles combine several narratives observed in RT’s Dutch coverage. For Study 4, the suppression strategy was operationalized through articles describing, first, the unveiling of a transgender pride zebra-crossing in the Dutch city of Almere, and second, an inclusivity measure to diversify street names in Rotterdam. These stories are, again, typical of suppression, with quotes selected that seek to deride these Dutch initiatives and their underlying progressive motivations.

Control Conditions

The control conditions consisted of a neutral text containing factual information on the topic of the corresponding experimental condition. Using texts that present simple facts for the control conditions is effective because participants are thereby exposed to similar themes and information as the participants in the corresponding experimental condition, but with the absence of the narratives that we seek to test. For Studies 1 and 3, the two control texts consisted of information about the crime or the national police force in Sweden or The Netherlands. In Studies 2 and 4, the control conditions presented factual information about LGBTQ+ progress or cultural diversity initiatives created in Sweden or The Netherlands.

Stimulus Presentation

We ensured similarity in the presentation and length of the stimulus materials. The articles are of similar lengths and will be presented solely as text so that participants are exposed to the narrative content without the influence of images/videos. Participants were informed that the articles are from international media before reading, but not that they are from RT/Sputnik. As discussed, audiences are more likely to be exposed to RT/Sputnik media content unknowingly as it is often repackaged by other, more widely read, social or local/mainstream media outlets. Therefore, we omitted this information to more realistically capture how citizens may respond to the narrative content. These presentation decisions echo those of previous research on responses to Sputnik/RT media (e.g., Edenborg, 2021).

Variables and Instruments

Full scales can be found on the repository.

Perceived Generalized Realistic and Symbolic Threat

To our knowledge, no preexisting scales measure the experience of realistic and symbolic threats on a macro-level. (Brambilla and Butz [2013] examine perceived macro-level symbolic and realistic threats. However, the two items they use are more objective, asking the extent that participants felt a situation described in a vignette could represent a threat to the countries’ values and traditions/economic system.) We therefore developed two reliable scales (realistic threat: ω for the Swedish sample = .74, ω for the Dutch sample = .78; symbolic threat: ω for the Swedish sample = .82, ω for the Dutch sample = .85). Details about their construction and development can be found on the repository. Participants were asked how far they agree with statements (e.g., “The moral values of [Swedish/Dutch] society are deteriorating” or “The [Swedish/Dutch] state provides well for me”; reverse-coded) on a 7-point Likert scale (1 = completely disagree, 7 = completely agree).

Emotional Responses

We assessed emotional responses with a 10-item scale measuring different emotions, which was used in previous research measuring emotional responses to news stories (Igartua et al., 2011). Participants were asked to indicate how much the articles made them feel a certain emotion on a 7-point Likert scale (1 = not at all, 7 = very much).

Institutional Trust

We assessed institutional trust by adapting preexisting items in the institutional trust scale (ω for the Swedish sample = .88; ω for the Dutch sample = .86) developed by Spadaro et al. (2020). Participants were asked to think about state institutions such as the government or the police, and then asked to indicate how far they agree with statements (e.g., “I trust state institutions in [Sweden/the Netherlands] because they are fulfilling their tasks well”) on a 7-point Likert scale (1 = completely disagree, 7 = completely agree).

Outgroup Trust

We assessed trust in societal groups othered by Russian state-sponsored media by adapting a 6-item scale (ω for the Swedish sample = .76; ω for the Dutch sample = .69) developed by Noor et al. (2008). Participants were asked to think about groups in society who do not think or look like them, and then to indicate how far they agree with statements (e.g., “Most members of other groups in [Sweden/the Netherlands] try to be fair”) on a 7-point Likert scale (1 = completely disagree, 7 = completely agree).

Manipulation Check

Participants indicated the extent to which they felt the stimulus they read focused on the failure of Swedish/Dutch state institutions or on criticism of Swedish/Dutch progressive policies. This was measured on a scale from 1 (= not at all) to 7 (= very much).

Attention Check

Participants were posed an instructed-response item embedded in one of the existing scales.

Exploratory Control Variables

Participants self-reported their political orientation on both social issues and economic issues on two scales from 1 (= very liberal) to 7 (= very conservative). Self-reported trust in news media was assessed on a scale from 1 (= not at all) to 7 (= very much). Finally, we administered adapted versions of the aforementioned perceived realistic and symbolic threat items used by Brambilla and Butz (2013).

Procedure

Studies were administered using an online survey in either Swedish or Dutch. After providing informed consent and providing demographic data, participants were randomly assigned to either an experimental condition or a corresponding control condition. Participants were then asked to read the stimulus material, with an imposed minimum time of 2 min for this stage. Lastly, they completed the counterbalanced posttest items and final check items, before being thanked and debriefed.

Results

Means and standard deviations are presented in Table 1. Exploratory analyses can also be found in ESM 1 (S3). Code for our analyses can be found on the repository.

Table 1 Means and standard deviations by study and condition

Confirmatory Analysis

Our originally specified models were just-identified and fit indices were therefore unavailable. Consequently, we decided to fit our models while controlling for age and gender to increase the degrees of freedom. For transparency, we report these models here and the coefficients for the originally specified models in ESM 1 (S4 – Table E2). There were no significant differences between the originally specified models and the newly fitted models (for more information, see S3 in ESM 1). Unless stated otherwise, we report weighted fit indices and estimates (for more information, see S2 in ESM 1).

Study 1 – Testing Destruction Narratives in Swedish Participants

Manipulation Check

A Welch two-sample t test showed that perceptions that the text focused on the failure of Swedish institutions were significantly higher in the destruction condition (M = 4.90, SD = 1.48) than in the corresponding control condition (M = 3.48, SD = 1.71): t(316.69) = 7.96, p < .001, d = 0.88.

Model Estimation

A path analysis model was estimated with lavaan, using the MLMV estimator due to (multivariate) nonnormality and the presence of heteroskedasticity in (some of) the emotional variables. This resulted in a well-fitting model based on our prespecified cut-off values, χ2 = 0.79, df = 2, p = .674, CFI = 1.000, RMSEA = .000, 90% CI [.000, .082], SRMR = .011.

Hypothesis Testing

We tested our hypotheses using this estimated model (Figure 3). We present the model’s path coefficients and statistics in Table 2. H1b, testing the direct effect on anger, was supported. However, H1a and H1c, which tested the direct effects on institutional trust and fear, were not supported. Moreover, H3a, H3b and H3c, testing the indirect effects on the response variables, were also not supported.

Figure 3 Path model testing responses of Swedish participants to destruction narratives. Paths are depicted using weighted standardized estimates. Dotted lines indicate nonsignificant pathways.
Table 2 Weighted path coefficients and statistics for confirmatory analyses

Study 2 – Testing Suppression Narratives in Swedish Participants

Manipulation Check

A Welch two-sample t test showed that perceptions that the text focused on the criticizing the progressiveness of Sweden were significantly higher in the suppression condition (M = 4.16, SD = 1.81) than in the corresponding control condition (M = 2.72, SD = 1.44): t(305.27) = 7.77, p < .001, d = 0.87.

Model Estimation

Fitting a path model using the procedure stated in Study 1 resulted in a reasonable model fit, χ2 = 25.15, df = 2, p < .001, CFI = 0.914, RMSEA = .187, 90% CI [.126, .255], SRMR = 0.048.

Hypothesis Testing

We tested our hypotheses using this estimated model (Figure 4). We present the model’s path coefficients and statistics in Table 2. H2b and H2d, testing the direct effects on anger and disgust, were supported. However, H2a and H2c, testing the direct effect on outgroup trust and shame, were not supported. Moreover, H4a, H4b, H4c, and H4d, testing the indirect effects on the response variables, were also not supported.

Figure 4 Path model testing responses of Swedish participants to suppression narratives. Paths are depicted using weighted standardized estimates. Dotted lines indicate nonsignificant pathways.

Study 3 – Testing Destruction Strategies in Dutch Participants

Manipulation Check

A Welch two-sample t test showed that perceptions that the manipulation focused on the failure of Dutch institutions were significantly higher in the destruction condition (M = 4.96, SD = 1.29) than in the corresponding control condition (M = 3.30, SD = 1.54): t(317.67) = 10.81, p < .001, d = 1.18.

Model Estimation

Fitting a path model using the procedure stated in Study 1 resulted in a well-fitting model, χ2 = 18.73, df = 2, p < .001, CFI = 0.949, RMSEA = .152, 90% CI [.094, .218], SRMR = .045.

Hypothesis Testing

We tested our hypotheses using this estimated model (Figure 5). We present the model’s path coefficients and statistics in Table 2. H1a, H1b, and H1c, testing the direct effects on institutional trust, anger and disgust, were all supported. However, H3a, H3b, and H3c, testing the indirect effects on these variables, were not supported.

Figure 5 Path model testing responses of Dutch participants to destruction narratives. Paths are depicted using weighted standardized estimates. Dotted lines indicate nonsignificant pathways.

Study 4 – Testing Suppression Strategies in Dutch Participants

Manipulation Check

A Welch two-sample t test showed that perceptions that the text focused on the criticizing the progressiveness of the Netherlands were significantly higher in the suppression condition (M = 4.18, SD = 1.45) than in the corresponding control condition (M = 3.26, SD = 1.53): t(336) = 5.65, p < .001, d = 0.61.

Model Estimation

Fitting a path model using the procedure stated in Study 1 resulted in a well-fitting model, χ2 = 6.40, df = 2, p = .041, CFI = 0.987, RMSEA = .079, 90% CI [.014, .151], SRMR = .023.

Hypothesis Testing

We tested our hypotheses using this estimated model (Figure 6). We present the model’s path coefficients and statistics in Table 2. H2c, testing the direct effect on shame, and H2d, testing the direct effect on disgust, were both supported. However, H2a and H2b, which tested a direct effect on outgroup trust and anger, were not supported. Moreover, H4a, H4b, H4c, and H4d, testing the indirect effects on these variables, were also not supported.

Figure 6 Path model testing responses of Dutch participants to suppression narratives. Paths are depicted using weighted standardized estimates. Dotted lines indicate nonsignificant pathways.

Discussion

This study aimed to investigate the psychological responses elicited in international audiences who consume Russian state-sponsored media narratives. We predicted that two antagonistic narrative strategies, destruction and suppression, would act as mechanisms that increase the perception of (different types of) threat, which in turn would predict responses on trust and emotional variables in Dutch and Swedish audiences. These predictions were not supported: realistic or symbolic threat perceptions did not mediate the effects of exposure to destruction or suppression narratives on our hypothesized response variables. However, predictions that the narrative strategies would have direct effects on our response variables were partially supported.

Given the seemingly strong theoretical convergence between Wagnsson and Barzanje’s (2021) destruction and suppression strategies and Stephan and Stephan’s (2000) realistic and symbolic threat perception model, it is curious that the threat perceptions did not mediate the strategies’ effects on the response variables, and specifically, that the manipulation did not trigger significant differences in these threat perceptions. Studies have evidenced the malleability of threat perceptions through more specific measurements (e.g., Schmuck & Matthes, 2017). A possible explanation, then, could be that our measurements were too broad and assessed more stable, trait-like perceptions rather than perceptions susceptible to change. Although we pretested to confirm sensitivity (see Study 3 in the file “Generalized realistic and symbolic threat scales development” for more details), it could nevertheless be difficult for news articles to elicit differences on such scales in the short term. In this context, more specific measurements might be more appropriate.

Contrary to predictions, there were differences in the direct effects that narratives elicited in Swedish and Dutch participants, with analyses indicating that the models were statistically noninvariant between groups. This was unexpected, as the countries are thought to be broadly culturally comparable (e.g., Shulgin et al., 2017). Inspecting the specific differences per strategy, one can speculate why this is. One option may be that specific nuances within each country’s stimulus materials may have prompted different responses. However, exploratory analyses seem to point to the effects on our response variables being driven by our intended manipulations (see S5f for more details). A second explanation may be that the differences reflect sensitivities in the populations caused by recent political events. While both countries have seen recent political controversies, the controversies in The Netherlands might be considered more impactful on the public (Ipsos, 2021) than those in Sweden (Martinsson & Andersson, 2021). Therefore, the responses of Dutch participants to destruction narratives, for example, may stem from frustrations being more easily triggered.

In general, there were more significant differences in emotional responses than in trust variables. Trust only differed significantly when comparing institutional trust in Dutch participants exposed to destruction narratives and the corresponding control condition. Interestingly, there were no significant differences in outgroup trust responses despite the stimulus materials for both expressly portraying liberal groups as unreasonable. One possible explanation is that the measurement of trust was lacking specificity: Being asked broadly about different social groups or several public institutions may elicit weaker responses than being asked about specific groups (e.g., LGBTQIA+ groups) or state institutions (e.g., the police). Another explanation could be that, like perceptions of general realistic or symbolic threats, changes to feelings of trust are also difficult to elicit through short-term exposure and that more long-term, repeated exposure is required (e.g., Shaver et al., 2017). Indeed, trust is typically considered a more stable attribute that changes gradually, meaning the capturing effect on trust is perhaps preferable in more longitudinal designs (e.g., Mutz & Reeves, 2005). Alternatively, in more short-term designs, specific attitudinal differences might be more appropriate to measure. This might also stimulate, as mentioned before, rearranging emotions and trust variables into a temporal sequence, an idea that has empirical support (Dunn & Schweitzer, 2005).

While we find several significant differences on emotional variables, both hypothesized and exploratory, we have a limited impression of the reasons behind these differences. Take anger, for example: Participants could be angrily agreeing that introducing “gender-neutral hymns” is ridiculous, or they could be angrily opposing this message. This could even differ across participants. One might argue that the direction of the anger is not relevant; that any cultivation of tension within a society is considered favorable by the Kremlin. Yet given that a goal of this study’s overarching research agenda is to provide practical insights for countermeasures such as counternarratives, we could extend the utility of these findings by nuancing our understanding of the reasons behind these emotional responses. To this end, future research might consider alternative measures, or qualitative research methods such as interviews or focus groups, which would these nuances to be elaborated.

Earlier discussions of the reasons behind between-group differences in responses raise important methodological reflections regarding the stimulus materials. First, we created our stimulus materials by pulling directly from articles highlighted in previous analyses of RT and Sputnik’s narration of Sweden and The Netherlands. This decision was made to preserve as much ecological validity as possible. However, it can be criticized for compromising consistency across countries, an issue we now potentially run into as we observe these between-country differences. As discussed, alternative options were considered, such as focusing on one topic – as described in the Materials section – or artificially crafting our own “replicas,” but it was decided that the potential loss of ecological validity and risk of manipulation contamination outweighed the possible consistency drawbacks. Future research seeking to specifically investigate responses between groups may seek to replicate our paradigm using one of the alternative options.

Relatedly, we can reflect on the decision not to inform participants that they were reading articles taken from Russian media. Despite justifying this in our Materials section, it is worth considering the possible differences in results should the participants have had this knowledge. While extant literature has shown that awareness of the source makes little difference to shifts in opinion following exposure to Russian media narratives (e.g., Fisher, 2020; Carter & Carter, 2021), a possible methodological revision could be to replicate the study with the addition of a third condition where participants are made aware of the Russian source, to observe any potential impact. It is important to note that, in the case of Russian state-sponsored media, awareness of the source might have more of an impact after Russia’s invasion of Ukraine.

Third, we might reflect on the design of the control condition. Exposing participants to an assembly of factual information on the corresponding topic allowed us to isolate the effect of the strategic narrative. However, a possible revision could be to include other media articles, so that both conditions consume media articles. A difficulty arises here, however, when selecting the media outlet whose reporting should act as an acceptably neutral control condition. Nevertheless, future studies might consider this methodological modification.

A final reflection centers on the timeliness of choosing to study Dutch and Swedish audiences. As described, both are European states that Russia seeks to malign and destabilize and have previously been identified as targets of Russian information influence. However, geopolitical events occurring during the execution of this research – namely, Russia’s invasion of Ukraine and the subsequent ban of RT and Sputnik in the European media space – have brought their relevancy, and the significance of this research more broadly, into sharper focus. Studies investigating psychological responses to narratives from these outlets in EU audiences are therefore particularly pertinent and can inform the debate regarding the necessity and appropriateness of this ban despite legitimate concerns about its implementation (Baarde, 2022). Yet, with Putin’s recent invasion of Ukraine underscoring the Kremlin’s ambition to adjust the international order and reassert its influence over its “near abroad” (Stent, 2022), one may argue that the psychological responses of audiences in EU states such as the Baltic states, where large Russophonic communities reside, are perhaps more pertinent to investigate.

Conclusion

The current study marks a transition in a broader research agenda from qualitatively analyzing Russian antagonistic narratives about foreign states to experimentally testing audience responses to said narratives. To our knowledge, the results are the first to experimentally evidence such responses to Russian antagonistic narration in international audiences. Our results are, therefore, preliminary and we have outlined several potential improvements and recommended directions for future research. With recent global events emphasizing the relevance that understanding the possible destabilizing effects of Russian antagonistic narration can have in international audiences, the authors hope that these results will motivate further research on this topic. Furthermore, thorough knowledge of the effects of Russian narration can serve as a foundation for scrutiny of antagonistic messaging across national borders by other actors, including terrorist organizations and other authoritarian states.

Author Biographies

Aiden Hoyle is a PhD candidate at the University of Amsterdam, TNO Defense, Safety and Security, and the Netherlands Defence Academy. His research focuses on building transdisciplinary models of cognitive and emotional responses that Russian state-sponsored media can elicit in international audiences.

Charlotte Wagnsson is Professor of Political Science at the Swedish Defence University. Her research interests include European security and political communication in the security sphere. She has recently published articles on the use of strategic narratives in malign information influence projected across borders.

Helma van den Berg is Senior Human Behavior Specialist at TNO Defence, Security, and Safety, The Netherlands. Helma works on analyzing and changing behavior in the safety and security domain.

Bertjan Doosje is Associate Professor of Psychology at the University of Amsterdam, The Netherlands. He is interested in intergroup relations from a social and cultural psychological perspective.

Martijn Kitzen is Professor of Irregular Warfare & Special Operations at the Netherlands Defence Academy. Martijn focuses on irregular warfare and special operations and holds a particular interest in embedding population-centric approaches in military operations.

References