The emotion regulation account of motivated political reasoning

Theories of motivated reasoning start with the assumption that all reasoning is motivated (Kunda, Reference Kunda1990), and distinguish between accuracy and directional goals—accuracy goals referring to motivations to reach correct conclusions, and directional goals referring to motivations to reach desired conclusions, such as conclusions consistent with preexisting political attitudes (e.g., Leeper & Slothuus, Reference Leeper and Slothuus2014). Following previous work, directionally motivated reasoning should lead voters to be skeptical toward information that is incongruent with their political attitudes (e.g.,Taber & Lodge, Reference Taber and Lodge2006). In the context of perceptions of politically relevant facts, this should lead to skepticism toward facts that are incongruent with voters’ political attitudes, which again should lead to perceptual divides where voters are more likely to assess information as true if it is congruent with their political attitudes than if it is incongruent (e.g., Flynn et al., Reference Flynn, Nyhan and Reifler2017; Hartman & Newmark, Reference Hartman and Newmark2012; Jerit & Barabas, Reference Jerit and Barabas2012; Miller et al., Reference Miller, Saunders and Farhart2016; Nyhan & Reifler, Reference Nyhan and Reifler2010; Peterson & Iyengar, Reference Peterson and Iyengar2020; Schaffner & Roche, Reference Schaffner and Roche2016).

The argument that motivated reasoning can distort perceptions of politically relevant facts has been made in a large number of domains (Baekgaard et al., Reference Baekgaard, Christensen, Dahlmann, Mathiasen and Petersen2017; Bisgaard, Reference Bisgaard2015, Reference Bisgaard2019; Christensen & Moynihan, Reference Christensen and Moynihan2020; Clayton et al., Reference Clayton, Davis, Hinckley and Horiuchi2019; Dickerson & Ondercin, Reference Dickerson and Ondercin2017; Flynn et al., Reference Flynn, Nyhan and Reifler2017; Flynn & Krupnikov, Reference Flynn and Krupnikov2019; Khanna & Sood, Reference Khanna and Sood2018; Kraft et al., Reference Kraft, Lodge and Taber2015; Robison, Reference Robison2018; Tworzecki & Markowski, Reference Tworzecki and Markowski2014; Vegetti & Mancosu, Reference Vegetti and Mancosu2020), and is largely parallel to Taber and Lodge’s (Reference Taber and Lodge2006) and Taber et al. (Reference Taber, Cann and Kucsova2009) classical finding that motivated reasoning can distort evaluations of the strength of political arguments (Kraft et al., Reference Kraft, Lodge and Taber2015).

The existing literature provides several different explanations for why motivated political reasoning occurs (see e.g., Arceneaux & Vander Wielen, Reference Arceneaux and Vander Wielen2017; Lodge & Taber, Reference Lodge and Taber2013; Mercier & Sperber, Reference Mercier and Sperber2011, Reference Mercier and Sperber2017; Stanovich, Reference Stanovich2021; Stanovich et al., Reference Stanovich, West and Toplak2013; Taber & Lodge, Reference Taber and Lodge2006, Reference Taber and Lodge2016).

Among the most prominent explanations is Taber & Lodge’s JQP model (Lodge et al., Reference Lodge, Steenbergen and Brau1995; Lodge & Taber, Reference Lodge and Taber2013; Taber et al., Reference Taber, Cann and Kucsova2009; Taber & Lodge, Reference Taber and Lodge2006, Reference Taber and Lodge2016), where affect toward sociopolitical concepts is stored in online tallies that exert a strong downstream influence on conscious deliberation (Lodge et al., Reference Lodge, Steenbergen and Brau1995; Redlawsk, Reference Redlawsk2002, Reference Redlawsk2006; Redlawsk et al., Reference Redlawsk, Civettini and Emmerson2010). In JQP, motivated political reasoning is fundamentally driven by an automatic and unconscious process called affective contagion (Erisen et al., Reference Erisen, Lodge and Taber2014; Lodge & Taber, Reference Lodge and Taber2013, p. 135; Taber & Lodge, Reference Taber and Lodge2016); a memory process that is “[…] hard wired into the very architecture of the brain […]” (Lodge & Taber, Reference Lodge and Taber2013, p. 227 emphasis in original). This implies that debiasing strategies are very unlikely to be successful (Lodge & Taber, Reference Lodge and Taber2013, p. 229).

It is clear from Taber and Lodge’s writings (Lodge & Taber, Reference Lodge and Taber2013, p. 220, 227, 233) that they view JQP as qualitatively different from another influential explanation, which holds that people engage in motivated political reasoning to alleviate cognitive dissonance (e.g., Baekgaard et al., Reference Baekgaard, Christensen, Dahlmann, Mathiasen and Petersen2017, p. 1119; Bisgaard, Reference Bisgaard2019, p. 3; Christensen et al., Reference Christensen, Dahlmann, Mathiasen, Moynihan and Petersen2018; Festinger, Reference Festinger1962; Groenendyk, Reference Groenendyk2013, Reference Groenendyk2018; Kuklinski et al., Reference Kuklinski, Quirk, Jerit, Schwieder and Rich2000, p. 794; Kunda, Reference Kunda1990; Peterson & Iyengar, Reference Peterson and Iyengar2020, p. 135; Ploger et al., Reference Ploger, Dunaway, Fournier and Soroka2021), i.e., to control what emotions they experience (Kiil, Reference Kiil2024; Thibodeau et al., Reference Thibodeau, Peebles, Grodner and Durgin2015; Westen et al., Reference Westen, Blagov, Harenski, Kilts and Hamann2006; Westen & Blagov, Reference Westen and Blagov2007). According to this account, humans generally seek to avoid (approach) behaviors and stimuli that become associated with negative (positive) affect. This leads to directionally motivated reasoning because people “…approach and avoid judgments based on their emotional associations” (Westen et al., Reference Westen, Blagov, Harenski, Kilts and Hamann2006, p. 1947). According to the emotion regulation account, then, one central mechanism behind motivated political reasoning is a desire to minimize negative affective states and to maximize positive affective states.

This process could potentially coexist with other mechanisms—such as the automatic activation of affective contagion emphasized in the JQP model—which could operate in parallel or under different conditions, and people might even differ in the extent to which one or the other mechanism dominates, depending on contextual cues or personal dispositions, such as cognitive style (cf. Arceneaux & Vander Wielen, Reference Arceneaux and Vander Wielen2017).

Mercier and Sperber (Reference Mercier and Sperber2017, pp. 244–246) argue against the emotion regulation explanation (which they refer to as the “feel-good account”) on theoretical grounds, stressing that it would be maladaptive for humans to adopt erroneous beliefs to obtain positive emotion, and that such a psychological adaption is therefore unlikely to have developed through the evolutionary history of humans.

Direct empirical evidence for the emotion regulation account of motivated political reasoning comes from a limited number of studies. Westen et al. (Reference Westen, Blagov, Harenski, Kilts and Hamann2006), Kaplan et al. (Reference Kaplan, Gimbel and Harris2016), and Casado‐Aranda et al. (Reference Casado‐Aranda, Venkatraman, Sánchez‐Fernández and Luque‐Martínez2020) found evidence linking emotion regulation and motivated political reasoning through neuroimaging studies, showing that confrontation with attitude-incongruent information increases activity in brain regions associated with implicit and explicit emotion regulation (though some of this evidence was mixed, and some of the results contradicted each other).

Other evidence for the emotion regulation account comes from studies that successfully use self-affirmations to reduce outcomes associated with motivated political reasoning (Binning et al., Reference Binning, Sherman, Cohen and Heitland2010; Cohen et al., Reference Cohen, Aronson and Steele2000, Reference Cohen, Sherman, Bastardi, Hsu, McGoey and Ross2007). This supports the emotion regulation account because self-affirmations provide an alternative source of global self-worth (Steele, Reference Steele1988), which is thought to work through reducing the need to use motivated political reasoning to bolster self-worth (a positive emotion). However, a meta-study of more recent well-powered studies on the efficacy of self-affirmations on motivated reasoning (Lyons et al., Reference Lyons, Farhart, Hall, Kotcher, Levendusky, Miller, Nyhan, Raimi, Reifler, Saunders, Skytte and Zhao2021; see also Nyhan & Reifler, Reference Nyhan and Reifler2011) found no consistent debiasing effect, leaving the empirical status of this literature unclear.

Thus, while motivated political reasoning is often assumed to be driven by desires to avoid discomfort, the emotion regulation explanation has been subject to very few direct empirical tests. I argue that a testable implication of this explanation is that emotion regulation strategies should influence the extent to which people engage in motivated political reasoning, which, in turn, should influence perceptual divides over political facts. In the following, I theorize how two important emotion regulation strategies (cognitive reappraisal and emotional acceptance) should be associated with perceptual divides if the emotion regulation account of motivated political reasoning is correct. In Gross’s (Gross, Reference Gross2015) seminal emotion regulation framework, cognitive reappraisals are thoughts—or patterns of thoughts—that target the meaning of a potentially emotion-eliciting situation or its self-relevance. This, I argue, must include motivated political reasoning if—as the emotion regulation account implies—accepting attitude-incongruent political information leads to negative emotions (and accepting attitude-congruent political information leads to positive emotions) all else equal (e.g., Cohen, Reference Cohen2003; Cohen et al., Reference Cohen, Aronson and Steele2000; Groenendyk, Reference Groenendyk2013, Reference Groenendyk2018). The main thrust of the argument is that voters who strongly support a policy or a party will experience cognitive dissonance (or related types of discomfort such as self-threats) if they are led to believe something that reflects negatively on that party or policy (Cohen, Reference Cohen2003; Cohen et al., Reference Cohen, Aronson and Steele2000; Devine et al., Reference Devine, Tauer, Barron, Elliot and Vance1999; Festinger, Reference Festinger1962). This discomfort can be alleviated by altering one’s attitude toward the policy or the party in question, but this may be costly and difficult for voters with strong attitudes (e.g., Groenendyk, Reference Groenendyk2013, pp. 16–17). Instead, they can alleviate the discomfort by pursuing directional reasoning goals to avoid ending up with beliefs that contradict their political attitudes. This implies that directionally motivated political reasoning corresponds to the emotion regulation strategy of cognitive reappraisal since it involves modifying cognitive patterns to alter the emotional impact of political information (cf. Gross, Reference Gross2015).

The emotion regulation account of motivated political reasoning, therefore, implies that the more people tend to use cognitive reappraisals, the more they should tend to engage in directional motivated reasoning. Since directional motivated reasoning is generally used to explain perceptual divides (cf. e.g. Flynn et al., Reference Flynn, Nyhan and Reifler2017; Hartman & Newmark, Reference Hartman and Newmark2012; Jerit & Barabas, Reference Jerit and Barabas2012; Miller et al., Reference Miller, Saunders and Farhart2016; Nyhan & Reifler, Reference Nyhan and Reifler2010; Peterson & Iyengar, Reference Peterson and Iyengar2020; Schaffner & Roche, Reference Schaffner and Roche2016), perceptual divides should tend to be larger the more people tend to use cognitive reappraisal.

Moving on to emotional acceptance, this concerns the counterintuitive idea of calmly focusing one’s attention on uncomfortable feelings and thoughts rather than trying to alter or avoid them (Hayes et al., Reference Hayes, Pistorello and Levin2012, p. 982; Hofmann & Asmundson, Reference Hofmann and Asmundson2008, p. 5). Emotional acceptance thus represents a fundamentally different way of regulating emotions than cognitive reappraisals, as one does not have to alter one’s interpretation of the situation at hand; rather, one has to alter the way one relates to thoughts and emotions involved in and evoked by one’s appraisal of the situation.

While cognitive reappraisal and emotional acceptance can operate at both implicit and explicit levels, the present study examines these strategies in their explicit form—that is, as consciously endorsed and effortful ways of managing emotional responses. Implicit processes, on the other hand, occur automatically and without conscious intent (Bargh & Williams, Reference Bargh and Williams2007; Tamir et al., Reference Tamir, John, Srivastava and Gross2007). Indeed, emotion regulation itself can operate as an unconscious goal (Hopp et al., Reference Hopp, Troy and Mauss2011), shaping how individuals attend to and interpret emotionally salient political information. Importantly, the distinction between implicit and explicit emotion regulation is not always clear-cut (Gross, Reference Gross2015; Gyurak et al., Reference Gyurak, Gross and Etkin2011). Prior research suggests that frequent conscious use of a strategy—such as emotional acceptance—can lead to automatization, meaning that over time it becomes habitual and potentially automatic (Desbordes et al., Reference Desbordes, Negi, Pace, Wallace, Raison and Schwartz2012; Gyurak et al., Reference Gyurak, Gross and Etkin2011). Thus, although this study focuses on explicit manifestations of these strategies (via self-report and experimental priming), they may reflect broader, partly implicit regulatory tendencies.

In politics, I argue, emotional acceptance implies noticing and embracing emotions and thoughts elicited by processing political information without trying to alter or escape them. This includes noticing and embracing discomfort associated with believing attitude-incongruent information. An anti-immigration voter, for instance, would notice and embrace the feelings of anxiety and self-threat elicited by earnestly listening to a factual claim that challenges an anti-immigration stance without taking steps to avoid or alter those feelings. Therefore, using emotional acceptance as an emotion regulation strategy when processing attitude-incongruent political information should enable the voter to believe attitude-incongruent political information and simultaneously downregulate negative emotions. Voters prone to using emotional acceptance should therefore engage less in motivated political reasoning when assessing political facts and therefore display smaller perceptual divides. This leads to the following hypothesis:

People who are prone to using cognitive reappraisal, on the other hand, should engage more in motivated political reasoning when assessing political facts and should therefore display larger perceptual divides. This leads to the second hypothesis:

While hypothesis 1 and hypothesis 2 concern individual differences in people’s general tendency to use emotion regulation strategies, people’s use of emotion regulation strategies can change from situation to situation. Many psychotherapeutic interventions aim at altering clients’ emotion regulation habits by promoting the use of acceptance or associated emotion regulation strategies (e.g., Acceptance and Commitment Therapy (Hayes et al., Reference Hayes, Pistorello and Levin2012), Mindfulness-Based Stress Reduction (Kabat‐Zinn, Reference Kabat‐Zinn2003; Kabat-Zinn, Reference Kabat-Zinn1982), and Metacognitive Therapy (Wells, Reference Wells2011). We should therefore expect situational features, which influence use of emotion regulation strategy, to influence motivated political reasoning and thus also to influence perceptions of politically relevant facts. This leads to hypothesis 3:

For H1 to find support, we should expect individuals high in emotional acceptance to be more likely to believe attitude-incongruent facts and to display a lower difference between belief in attitude-congruent facts and attitude-incongruent facts (i.e., a smaller perceptual divide). For H2 to find support, individuals high in cognitive reappraisal should be less likely to believe attitude-incongruent facts and display a larger difference between belief in attitude-congruent facts and attitude-incongruent facts (i.e., a larger perceptual divide).

Finally, for H3 to find support, performing an emotional acceptance-inducing task should lead to greater belief in attitude-incongruent facts and a smaller difference between belief in attitude-congruent facts and attitude-incongruent facts.

Overview of the studies

I test the hypotheses in the context of perceptions of facts related to immigration politics in Denmark. Though Denmark is a relatively consensus-based multiparty system, immigration politics is a divisive and highly salient issue, which has dominated party politics over the last decade (cf. e.g., Hansen & Stubager, Reference Hansen and Stubager2017). I contend that the immigration debate in Denmark tends to elicit strong emotions and that directionally motivated reasoning and perceptual divides, therefore, are very likely to be triggered in this context.

In all three studies, I recruited respondents from Dynata’s (formerly SSI) panel. Participants were invited to answer an online survey for economic compensation (see Table 1 for an overview of the studies). In study 1, I collected 1254 subjects in the fall of 2019. In studies 2 and 3, I collected 1314 and 1618 subjects in early and mid-2020, respectively. Sample sizes were decided based on power analyses. All samples are approximately representative of the Danish voting age population in terms of age, gender, education, and income (see Supplementary SI 1 for sample characteristics). In all studies, I use an instructional attention check (Oppenheimer et al., Reference Oppenheimer, Meyvis and Davidenko2009) to weed out respondents who do not pay sufficient attention to the question texts (see Supplementary SI 2 for details). Only respondents who pass the attention check are counted as part of the full sample and included in the analyses. Studies 2 and 3 were preregistered (see Supplementary SI 3 for minor deviations from preregistrations not described in the main text).

Table 1. Overview of studies

In study 1, I investigate hypothesis 1 through a cross-sectional analysis with perceptual divides as the dependent variable. The evaluated facts are attributed to a non-expert, partisan source (leftwing or rightwing politicians). In study 2, I use a mix of cross-sectional and experimental analyses to investigate hypotheses 1, 2, and 3. This time, the facts are attributed to an expert, non-partisan source (Statistics Denmark). In this study, I focus entirely on attitude-incongruent facts. In study 3, I investigate hypotheses 1 and 2 with a mix of cross-sectional and experimental analyses. I use both attitude-congruent and attitude-incongruent facts, and whether the facts are attributed to an expert or a non-expert source is determined through random assignment.

Study 1

Study 1 focuses on H1 and uses a cross-sectional design to test whether individual differences in emotional acceptance predict belief in attitude-incongruent facts and perceptual divides. Following such studies as Peterson and Iyengar (Reference Peterson and Iyengar2020), I use a mix of correct and incorrect factual claims that have received significant public attention. I use two pro-immigration factual claims and two anti-immigration factual claims. The correct pro-immigration claim reads: “The right-wing’s ‘Start help” [a renowned Danish social program], which halved unemployment benefits for immigrants, has increased poverty-related crimes and led many immigrant women to exit the labor force” whereas the incorrect pro-immigration claim reads: “Support for equality between men and women is just as high among non-western immigrants and descendants as among ethnic Danes.” The correct anti-immigration claim reads”: Non-western immigration costs the welfare state approximately 33 billion kroner a year. This is approximately three times as much as is used on police every year.” The incorrect anti-immigration claim reads: “Non-western immigrants are on average six times as likely to commit punishable crimes as ethnic Danes.”

Pro-immigration statements were introduced with the text”: Politicians on the political left have made the following claim.” Anti-immigration statements were introduced with the text”: Politicians on the political right have made the following claim.” Respondents were asked to evaluate to what extent they agree that the statement is correct on a five-point Likert scale (cf. e.g., Nyhan & Reifler, Reference Nyhan and Reifler2010).

I create an index for belief in attitude-incongruent facts (M = .36, SD = .25) and for attitude-congruent facts (M = .65, SD = .19), both ranging from 0 to 1. Attitude-congruency of each statement is determined by dividing respondents into anti- and pro-immigration respondents (dependent on whether their mean score on the immigration attitudes index was higher than .5). I create a perceptual divide measure by subtracting belief in attitude-incongruent facts from belief in attitude-congruent facts (M = .65, SD = .19), ranging from 0 to 1 with high values indicating a large perceptual divide.

I create an immigration attitudes index (α = .81, M = .62, SD = .27) ranging from 0 to 1, with high values indicating anti-immigration attitudes. I used a 10-point immigration attitude self-placement scale and two five-point Likert scale items concerning punishments for criminal immigrants and whether immigration constitutes a threat.

To measure individual differences in trait emotional acceptance, I use the acceptance subcomponent of the Cognitive Affective Mindfulness Scale—Revised (CAMS-R) (Feldman et al., Reference Feldman, Hayes, Kumar, Greeson and Laurenceau2007). The CAMS-R is a validated measure that correlates with mindfulness and well-being and negatively predicts anxiety, distress, and depression (Feldman et al., Reference Feldman, Hayes, Kumar, Greeson and Laurenceau2007). The acceptance subscale includes three items: “I am able to accept the thoughts and feelings I have,” “I can accept things I cannot change,” and “I can tolerate emotional pain.” Scores are recorded on a scale from 1 (almost never) to 5 (almost always). I create an index ranging from 0 to 1 (α = .80, M = .66, SD = .24) with high values indicating high emotional acceptance.

I control for standard sociodemographic factors (education, age, gender, income, and ethnicity) and a series of psychological covariates, which have been shown in previous studies to be important predictors of motivated political reasoning: political sophistication, attitude strength (Taber & Lodge, Reference Taber and Lodge2006), need for cognition, and need for affect (Arceneaux & Vander Wielen, Reference Arceneaux and Vander Wielen2013, Reference Arceneaux and Vander Wielen2017). I measure political sophistication with four knowledge questions from the Danish national election study conducted in 2019 (Hansen, Reference Hansen2021), e.g., “Which parties were in government up until the election in 2019?” I measure attitude strength with two questions about how strong participants’ attitudes on immigration politics are, such as “How much do you personally care about the issue of immigration?” I measure need for cognition with three items from Cacioppo et al. (Reference Cacioppo, Petty and Feng Kao1984), such as” I like to have the responsibility of handling a situation that requires a lot of thinking.” To measure need for affect, I use three questions from each of the two dimensions in Need For Affect (approach and avoidance) in Maio & Esses (Reference Maio and Esses2001), such as “I am a very emotional person.” See Supplementary SI 4 for full question wordings and Supplementary SI 5 for descriptive statistics for measures where these are not provided in the main text. Supplementary see SI 6 for pairwise correlations between continuous predictors.

Results. As expected, respondents believe attitude-congruent facts much more than attitude-incongruent facts. On average, respondents report 31 percentage points more belief in attitude-congruent than attitude-incongruent facts (p < .001). This shows that perceptual divides identified by previous studies in other contexts (Bisgaard, Reference Bisgaard2015; Jerit & Barabas, Reference Jerit and Barabas2012; Miller et al., Reference Miller, Saunders and Farhart2016; Nyhan & Reifler, Reference Nyhan and Reifler2010; Peterson & Iyengar, Reference Peterson and Iyengar2020; Robison, Reference Robison2018; Tworzecki & Markowski, Reference Tworzecki and Markowski2014; Vegetti & Mancosu, Reference Vegetti and Mancosu2020) replicate in Denmark.

Figure 1 displays the association between emotional acceptance and belief in attitude-incongruent facts and perceptual divides (Supplementary see SI 7 for full regression tables for all three studies). Emotional acceptance is not associated with either belief in attitude-incongruent facts (b = −.03, p = .49) or perceptual divides (b = .03, p = .24) at levels approaching statistical significance. These results contradict the theoretical expectation expressed in H1, as emotional acceptance neither predicts belief in attitude-incongruent facts nor predicts perceptual divides. This suggests that emotional acceptance does not reduce partisan perceptions of political facts, at least in the context of facts related to immigration politics in Denmark.

Figure 1. Study 1 results. Effect of emotional acceptance on belief in attitude-incongruent facts and perceptual divides in study 1. Estimates are unstandardized OLS regression coefficients, and vertical lines indicate 95% confidence intervals. Immigration attitude, political sophistication, attitude strength, age, gender, income, ethnicity, and education are included as covariates.

The factual claims used in study 1 are accompanied by a non-expert source cue. Given the relatively low credibility of the source cues (a partisan source), one could argue that it is relatively rational (or reasonable) to be skeptical of the attitude-incongruent factual claims. While perceptual divides, such as those studied here, have previously been attributed to motivated political reasoning (Flynn et al., Reference Flynn, Nyhan and Reifler2017; Hartman & Newmark, Reference Hartman and Newmark2012; Jerit & Barabas, Reference Jerit and Barabas2012; Miller et al., Reference Miller, Saunders and Farhart2016; Nyhan & Reifler, Reference Nyhan and Reifler2010; Peterson & Iyengar, Reference Peterson and Iyengar2020; Schaffner & Roche, Reference Schaffner and Roche2016), others have argued that they may be attributable to more rational processes (e.g,. Druckman & McGrath, Reference Druckman and McGrath2019; Gerber & Green, Reference Gerber and Green1999). This implies a risk that the ineffectiveness of emotional acceptance may be due to the perceptual divides being driven by rational processes—and thus not by motivated reasoning. One way to accommodate this issue is to use a highly credible expert source and only use true factual claims. Perceptual divides, which persist even under such conditions, are more likely to be caused by motivated political reasoning, as there are fewer “reasonable” grounds for disbelieving the facts.

Study 2

In study 2, I augment the research design used in study 1 by attributing all factual statements to an expert source and by only using true factual claims. Furthermore, to ensure that the factual claims arouse emotions, I accompany each factual statement with a graph (Supplementary see SI 4 for details) illustrating the claim to make it more tangible (Nyhan & Reifler, Reference Nyhan and Reifler2018). Following Khanna and Sood (Reference Khanna and Sood2018), in this study, I only expose respondents to attitude-incongruent factual claims. I do so to focus narrowly on the specific context where emotional acceptance is theorized to influence reasoning.

In this study, I also test H2 by investigating whether individual differences in cognitive reappraisal are associated with low belief in attitude-incongruent facts. Moreover, I test H3 by investigating whether situational cues prompting the use of emotional acceptance increase belief in attitude-incongruent facts. The test of H3 occurs through an experiment where participants in the treatment group are prompted to use emotional acceptance to regulate emotions while evaluating a set of attitude-incongruent political facts (see Supplementary SI 6 for a balance test).

The procedure is as follows: After reporting immigration attitudes, emotion regulation strategies, and covariates, participants read and evaluate two attitude-incongruent factual statements. Participants in the treatment group then listen to an audio clip that guides them through a short exercise in accepting emotions, and they are invited to use emotional acceptance when subsequently evaluating two other attitude-incongruent facts. Participants in the control group engage in a guided mind-wandering task before evaluating the two other attitude-incongruent facts. All participants then answer a manipulation check before reading a thorough debriefing.

Respondents only evaluate attitude-incongruent facts as determined by their immigration attitude position. The facts selected for study 2 are reported in Table 2. They are designed to elicit strong emotional reactions, to be very congenial to one side of the immigration debate, and to be correct. In all cases, the facts are attributed to Statistics Denmark (a highly authoritative expert source), which is also their true source, and they are accompanied by a graph illustrating the claim (see Supplementary SI). I create an index for belief in the attitude-incongruent facts used before the intervention (M = .45, SD = .29), one for the facts used after the intervention (M = .46, SD = .3), and one for all items combined (M = .45, SD = .25). All indices range from 0 to 1.

Table 2. Facts evaluated in studies 2 and 3

Designing an experimental manipulation strong enough to impact emotion regulation strategy in an online survey is a challenge, but has been accomplished successfully before (Ford et al., Reference Ford, Feinberg, Lam, Mauss and John2018; Hafenbrack et al., Reference Hafenbrack, Kinias and Barsade2014; see also Halperin et al., Reference Halperin, Porat, Tamir and Gross2013). Participants in both the treatment and the control group listen to a 3 minutes 30 seconds long audio file. In the treatment group, the audio file guides respondents to scan their body for an unpleasant sensation, to focus their attention on that sensation, and to let go of the struggle with the sensation (to accept it). The manipulation is based on a transcript of a guided emotional acceptance exercise from Harris (Reference Harris2011). In the control condition, the audio file guides participants through a short mind-wandering task, designed to mimic a “waking, baseline mental state” (Hafenbrack et al., Reference Hafenbrack, Kinias and Barsade2014, p. 371). For transcripts of both audio files, see Supplementary SI 8.

As a manipulation check, I use three items from the Toronto Mindfulness Scale (TMS), which measure “decentering”—a concept closely associated with emotional acceptance (Lau et al., Reference Lau, Bishop, Segal, Buis, Anderson, Carlson, Shapiro, Carmody, Abbey and Devins2006). The TMS is a state measure rather than a dispositional measure, which makes it useful as a subjective manipulation check in this context. An example of an item is “I approached each feeling by trying to accept it, no matter whether it was pleasant or unpleasant.” The index ranges from 0 to 1 (α = .7, M = .63, SD = .21). I pilot-tested the acceptance manipulation on a sample of 213 respondents on Amazon M-Turk, and I found a 10 percentage point increase in TMS of the treatment (p < .01).

To measure cognitive reappraisal, I use the emotion regulation questionnaire (ERQ) developed by Gross & John (Reference Gross and John2003, p. 403). The ERQ is a validated measure (Aldao et al., Reference Aldao, Nolen-Hoeksema and Schweizer2010) with two subscales; one measuring the use of cognitive reappraisal, and one measuring suppression. I only use the four items measuring cognitive reappraisals: (1) “When I want to feel more positive emotion (such as joy or amusement), I change what I’m thinking about”; (2) “when I want to feel less negative emotion (such as sadness or anger), I change what I’m thinking about”; (3) “when I want to feel more positive emotion, I change the way I’m thinking about the situation”; and (4) “when I want to feel less negative emotion, I change the way I’m thinking about the situation.” Scores are reported on a Likert-type five-point scale. I create an index ranging from 0 to 1 (α = .85, M = .54, SD = .25).

Individual differences in emotional acceptance are operationalized the same way as in study 1 (α = .70, M = .67, SD = .21). I again control for political sophistication (measured the same way as in study 1) and attitude strength as well as income, education, gender, and age (see Supplementary SI 4 for details).

Results. As expected, participants with moderate immigration attitudes believe attitude-incongruent political facts much more than participants with strong and extreme attitudes. Attitude extremity decreases belief in attitude-incongruent facts by 27 percentage points (p < .001). This is similar to the results in study 1.

Figure 2 shows the associations between emotion regulation strategy and belief in attitude-incongruent facts. Individual differences in emotional acceptance predict approximately 12 percentage points (p < .01) increased belief in the attitude-incongruent facts evaluated before the treatment. This supports H1 as emotional acceptance is associated with high belief in attitude-incongruent facts, and, thus contradicts the results found in study 1. Individual differences in cognitive reappraisal, in contrast, are unrelated to belief in attitude-incongruent facts (p = .42). This contradicts the expectation in H2, according to which the two should be negatively associated.

Figure 2. Study 2 results. Effect of emotional acceptance and cognitive reappraisal on belief in attitude-incongruent facts in study 2. Estimates are unstandardized OLS regression coefficients, and vertical lines indicate 95% confidence intervals. Immigration attitude, political sophistication, attitude strength, age, gender, income, and education are included as covariates. All models were preregistered.

Moving on to the effect of situational emotional acceptance, the results show that the experimental treatment has no effect on belief in attitude-incongruent facts (p = .40), although it leads to a 6 percentage points increase in the manipulation check (p < .001). This indicates that situational cues increasing emotional acceptance do not increase belief in attitude-incongruent facts, which contradicts H3.

The results in studies 1 and 2 are inconsistent, as individual differences in emotional acceptance were associated with belief in attitude-incongruent facts in study 2 but not in study 1. This raises the question of whether this inconsistency is due to sampling error or to differences in the design of the two studies, such as the use of a non-expert source cue in study 1 and an expert source cue in Study 2. To resolve this issue, I conducted a third study, which provides a final test of the influence of individual differences in emotional acceptance on perceptual divides. Study 3 also provides an additional test of H2 (the effect of individual differences in cognitive reappraisal), which did not find support in Study 2.

Study 3

Study 3 provides two core methodological advantages over Study 2. First, I experimentally manipulate whether the facts are attributed to an expert source or not. This enables testing whether individual differences in emotional acceptance are only associated with belief in attitude-incongruent facts when they are attributed to an expert source. Second, participants evaluate both congruent and incongruent facts (i.e., all eight facts displayed in Table 2), thus enabling a comparison of belief in attitude-congruent and attitude-incongruent facts.

Participants are randomly assigned to one of two conditions: one where all facts are attributed to a non-expert partisan source, as in study 1, or one where all facts are attributed to the same expert source as in study 2. I use the same factual statements as used in study 2 and the same graphical illustrations. Immigration attitudes and attitude strength are measured as in study 1, and remaining individual differences variables are measured as in study 2 (see Supplementary SI 4, 5, and 6 for details, descriptive statistics, and balance test).

Results. As in studies 1 and 2, participants report 24 percentage points (p < .001) higher belief in attitude-congruent than attitude-incongruent facts and thus display clear perceptual divides. Figure 3 shows the effect of individual differences in emotional acceptance and cognitive reappraisal on belief in attitude-incongruent facts and perceptual divides, dependent on whether the facts were attributed to an expert source or a non-expert source. Individual differences in emotional acceptance have no significant interaction effect with expert source on belief in attitude-incongruent facts (p = .15), on belief in attitude-congruent facts (p = .45), or on perceptual divides (p = .64). Furthermore, emotional acceptance has no significant effect on attitude-incongruent facts, regardless of whether the source is expert or non-expert (p > .05 in both cases). Emotional acceptance does have a significant positive effect on belief in attitude-congruent facts (b = .85, p = .029) when the source is expert but not when the source is non-expert (p = .19). This adds up to a non-significant association between emotional acceptance and perceptual divides (belief in attitude-congruent facts minus belief in attitude-incongruent facts), regardless of whether the source is expert or non-expert (p > .05 in both cases). These results contradict the preregistered expectations and H1. The results indicate that the divergence between the findings in studies 1 and 2 did not stem from the difference in source attribution but from something else, possibly sampling error. The results indicate that emotional acceptance does not reduce perceptual divides over politically relevant facts as theorized.

Figure 3. Study 3 results. Effect of emotional acceptance and cognitive reappraisal on belief in attitude-incongruent facts and perceptual divides in study 3 (N = 1407). Estimates are unstandardized OLS regression coefficients, and vertical lines indicate 95% confidence intervals. Immigration attitude, political sophistication, attitude strength, age, gender, income, and education are included as covariates. Analyses of effects of emotional acceptance were preregistered (N = 1477); analyses of effects of cognitive reappraisal were not (N = 1407).

Cognitive reappraisal, in contrast, is significantly positively associated with belief in attitude-incongruent facts and negatively associated with perceptual divides. While I find no sign of an interaction effect with expert source, the effect-sizes at different source conditions range between 4 and 9 percentage points (p < .05 in all cases) and between 8 and 7 percentage points (p < .001 in all cases) when pooling observations with expert and non-exert source cues (these analyses were not preregistered). These results are exactly the opposite of the expectation in H2 as I find cognitive reappraisal to be negatively associated with perceptual divides and positively associated with belief in attitude-incongruent political facts. The results are inconsistent with the findings in study 2 where cognitive reappraisal was not significantly associated with belief in attitude-incongruent political facts.

Across all three studies, I thus find that neither individual differences in emotional acceptance nor in cognitive reappraisal have a consistent association with belief in attitude-incongruent facts or with perceptual divides. To the extent that cognitive reappraisals do predict perceptual divides, the association is in the exact opposite direction of the theoretical expectation. Furthermore, an experimental manipulation, which successfully increases emotional acceptance, does not increase belief in attitude-incongruent facts. I thus do not find support for any of the theoretical expectations derived from the emotion regulation account of motivated political reasoning.