Testing the Asymmetry Hypothesis of Tolerance: Thinking About Socially Disruptive Protest Actions

Under the asymmetry hypothesis, political tolerance and intolerance differ in their underlying psychology, making it easier to persuade the tolerant to become less tolerant than to convince the intolerant to become more tolerant. Using a representative sample of the Dutch population (N = 546), we examined this hypothesis for people’s tolerance or intolerance of socially disruptive protest actions of their least-liked group. Focusing on the relevant contrasting values of freedom of speech and public order, we found empirical evidence for the asymmetry of political tolerance: it was easier to persuade the tolerant to become less tolerant than to convince the intolerant to become more tolerant. In fact, we found a backlash effect among the intolerant participants with them showing higher intolerance as a result. These findings support the notion that tolerance is more fragile than intolerance because of the required self-restraint that involves psychological discomfort and uneasiness. However, tolerance is indispensable for our increasingly polarized liberal democratic societies making further research on the social psychology of tolerance and intolerance topical and urgent

. The psychological discomfort and uneasiness of cognitive dissonance that is involved would make tolerance more fragile than intolerance (Festinger, 1962;Harmon-Jones & Mills, 1999).
In social psychology two main models have been proposed for understanding tolerance. Simon and colleagues have developed and empirically tested their disapproval-respect model, which argues that tolerance involves disapproval based on ingroup-outgroup categorization that is restrained by respect based on shared superordinate categorization (Simon, 2020;Zitzmann et al., 2022).
The forbearance model of tolerance goes beyond categorization processes in considering tolerance as an ideological dilemma (Billig et al., 1988) that involves thinking about relevant reasons for disapproval in relation to reasons for en durance . This model identifies a general thinking process that results in tolerance or intolerance depending on the content and prioritizing of specific considerations and values relevant to the situation and case at hand. People's belief systems tend to contain competing values (Peffley et al., 2001) and the prioritization involved in weighing competing considerations can lead to higher tolerance or rather intolerance. The reasons for rejecting what one objects to can trump the reasons for being tolerant and thus decrease the psychological discomfort and the required self-restraint of toleration. For example, a commitment to freedom of speech can be overridden by values of public order and safety leading to lower tolerance of socially disruptive protest actions of adversaries (Adelman et al., 2021;Kuklinski et al., 1991).
In contrast, trying to convince the intolerant to become more tolerant by allowing that which they disapprove of is far more challenging because it is asking people to increase their feelings of psychological discomfort. The principle of cognitive consistency implies that people are motivated to hold their attitudes and actions in harmony and avoid dissonance (Festinger, 1962;Harmon-Jones & Mills, 1999). Thus, higher tolerance of disruptive protest actions by being asked to consider freedom of speech is less likely to be effective for those who are intolerant of these actions to begin with.
In fact, there is the real possibility of an "opinion backlash" (Bishin et al., 2016) in which trying to change the intoler ant in the direction of higher tolerance results in the opposite effect (e.g., Bratton, 2002;Preuhs, 2007). For example, calls for tolerance towards immigrants can lead to more positive attitudes among the already tolerant, but simultaneously to more negative attitudes among the intolerant (Djupe et al., 2015). Further, permissive policy approaches can decrease conservative peoples' support of Muslim minority members holding public rallies and demonstrations for a better recognition of their interests (Traunmüller & Helbling, 2017). Thus an emphasis on reasons for tolerance might backfire and have unanticipated consequences, similar to reactive effects in trying to reduce prejudice with invoking external social norms rather than autonomous choice (e.g., Brehm & Brehm, 2013;Legault et al., 2011).

Testing the Asymmetry Hypothesis
Using a representative sample of the Dutch population, we examined the hypothesis of asymmetry for people's political tolerance or intolerance of socially disruptive protest actions of their least-liked ideological group. Tolerance signifies an approach towards opinions, beliefs, and related practices that one disagrees with or objects to (Cohen, 2004;Verkuyten et al., 2022). However, social groups become the proper focus of tolerance if accompanied by a defining set of group values and beliefs, such as opinion-based groups (being pro or anti an issue) or ideological adversaries. Research on political tolerance is concerned with accepting the equal civic rights of ideological groups that one is negative about, and the common term -and related measure -for this is 'the least-liked group'. The well-known least liked group technique examines political tolerance by first asking people to indicate which ideological group they like the least and subsequently whether they are willing to grant people of that group the full rights of citizenship, such as giving public speeches, seeking public office, and holding demonstrations and protest actions (Sullivan et al., 1979). Political tolerance presupposes that the perceiver dislikes the ideological target group (Gibson, 2006) and although the content-controlled least-liked group technique has its limitations, it allows for the examination of political tolerance of ideological groups that people themselves object to (Hurwitz & Mondak, 2002).
We further tested the asymmetry hypothesis by focusing on socially disruptive protest actions and considering the relevant contrasting values of freedom of speech and public order. First, political tolerance is commonly examined in terms of practices and actions that must be allowed for all citizens to express and seek support for their point-of-view, including protest actions of groups with contrasting ideological beliefs and practices. Protest actions are typically perceived in relation to a constellation of dimensions and various distinctions are proposed for different types of protest actions, such as normative and non-normative, violent and non-violent, moral and immoral, and disruptive and constructive (e.g., Feinberg et al., 2020;Piven & Cloward, 1991;Tausch et al., 2011). Protest actions that are deemed as violating a moral standard or transgressing a moral boundary (e.g., "stopping other people from speaking in public", "occupying buildings") are generally not tolerated by the public because these actions are considered unreasonable, threatening, and emotionally harmful means to reach a goal (Boch, 2020;Chong & Levy, 2018;Feinberg et al., 2020;Simpson et al., 2018;Verkuyten et al., 2023). Therefore, such actions are ill-suited for testing the asymmetry hypothesis as they would be expected to be intolerable to most. Here we instead focus on conventional protest actions (e.g. demon strations on public roads, rallies in one's neighborhood) that can be considered socially inconvenient and disruptive of everyday life and thereby the functioning of society (Shuman et al., 2021).
Second, we considered the well-known oppositional values of "freedom of speech and demonstration" vs. "public order and safety" which are relevant for evaluating socially disruptive protest actions (e.g., Adelman et al., 2021;Nelson et al., 1997;Peffley et al., 2001;Zilli Ramirez & Verkuyten, 2011). For some people, tolerance of these actions may follow from liberal freedoms of speech and expression, whereas for others, the rejection of these actions stems from concerns about public safety and order. Thus, tolerance of disruptive protest actions can be weighed, on the one hand, by the value placed on individual liberties and civic freedom that are supported by overwhelming proportions of citizens in western democracies (Wike & Simmons, 2015), and, on the other hand, by whether such practices undermine the safety, order, and wellbeing of others which are also deemed important by most citizens (Silver, 2018). These two values can conflict with each other making it important to consider the relative trade-off that people make between them for understanding their tolerance (i.e., freedom trumps order) or intolerance (i.e., order trumps freedom) of protest actions. Following cognitive dissonance theory and the asymmetry hypothesis, we expected that asking tolerant people to consider the importance of public safety and order will make them less tolerant, whereas intolerant people will not become more tolerant when asked to consider the importance of freedom of speech, and might even become more intolerant (i.e., a backlash or reactance effect).

Method Participants
Potential respondents were selected by a survey company (Kantar) which maintains a representative panel for fieldwork in the Netherlands. From this online panel, a national sample of the native Dutch population aged 18 years and older was compiled via a random stratification procedure based on the characteristics gender, age, education, household size, and region. The response rate was 54%, which is common in the Netherlands (Stoop et al., 2010). A sample of 544 Dutch respondents participated with consent in an online survey and answered the questions of interest. 1 The sample closely matches the general Dutch population in terms of demographic characteristics, but with a slight overrepresentation of older people and higher educated (Statistics Netherlands, 2019). The sample was relatively evenly divided in terms of gender (51.1% female), and age ranged from 18 to 89 years (M = 49.18, SD = 18.24). Education was measured on a 7-point ordinal scale, ranging from "no education/only lower education/integration course/Dutch language course" (1) to "Doctoral or master's degree or postgraduate education" (7). Based on the classification of low, moderate, and high education levels by Statistics Netherlands (2019), 13.2% of the sample were classified as having low levels of education, 48% as moderately educated, and 38.8% as highly educated. Participants were relatively evenly distributed in their political orientation (self-placement from 1 ('extreme left') to 7 ('extreme right'), M = 4.71, SD = 1.87; 24.8% center, 26.6% left-leaning, 34.4% right-leaning, 14.3% non-response or indicating they did not know). With weights applied to correct 1) As is common in large-scale data collections, a team was involved which resulted in various topics being examined in different randomized versions of the questionnaire, such as attitudes towards immigrants, self-confirmation, and prejudice. Here we focus on the questions that we were able to include in four of the ten versions of the questionnaire to empirically test the asymmetry hypothesis. for any deviations from the targeted quotas the findings are representative for the Dutch population in terms of gender, age, education, and political orientation. Data collection was approved by the relevant ethical board and adhered to the national legal requirements. The study was pre-registered and the data was made publicly available (see Supplementary Materials).

Procedure and Materials
Participants were asked to select from a list of pre-selected ideological groups their least-liked group. We used a list of pre-selected groups because this does not lead people to focus on extreme or extremist groups (e.g. terrorists, neo-nazis) that are beyond toleration by most. Based on previous research in the Netherlands (Verkuyten et al., 2023), participants were presented with a list of 15 mainstream ideological groups of different types and across the political spectrum which provides "everyone an opportunity to express his or her intolerance" (Gibson, 1992, p. 574; see Appendix). 2 Then, participants were asked to indicate to what extent it should be tolerated that this group engages in four types of socially disruptive protest actions: "holding protest actions in your neighborhood or city", "holding a large demonstration on the public road", "distribute pamphlets and leaflets at the entrance of public buildings", and "organiz ing meetings in busy shopping areas". Seven-point scales were used with the anchor 'never tolerated' (1) and 'always tolerated' (7) and the four items formed a reliable scale (α = .92).
Based on the sum score on these four questions, individual participants were then automatically (computerized) divided into two groups: intolerant (if sum < 16) and tolerant (if sum ≥ 16). Next the contrasting values of freedom of speech and public order were presented to the two groups. Previous research demonstrates that the Dutch public considers both values equally important in relation to specific dissenting practices (Adelman et al., 2021) and also as equally persuasive arguments for policies. 3 Specifically, the intolerant group was presented with the following text: "Safety and public order are of course very important values, but there are also other values that are important, such as freedom of speech and the right to demonstrate". By contrast, the tolerant group was told: "Freedom of speech and the right to demonstrate are of course very important values, but there are also other values that are important, such as security and public order". Subsequently, both groups of participants were asked to take some time to carefully consider and weigh up the different values in light of their answers: "Now we ask you to consider and weigh up the different values carefully. So please really take a moment to think about the importance of the different values in the light of your answers". This persuasibility intervention sought to guarantee that both the costs and the benefits of tolerance or intolerance were fully considered which allows us to test the asymmetry hypothesis. After reflecting on the different values, participants were asked to indicate the extent to which they were inclined to change their view ("to what extent are you then inclined to …") using three bipolar questions with the end-point anchors, "be less tolerant vs be more tolerant", "be more negative toward the least-liked group vs be more positive towards the least-liked group", and "listen less to arguments of the least-liked group vs listen-more to arguments of the least-liked group". Seven-point scales were used and it was explained that a score of 4 (scale midpoint) indicates that they do not change their views ("if your answer remains the same, choose the middle box of 4") and that the degree of change in one or the other direction could be indicated (α = .67).

Intervention and Assignment Check
To check whether participants followed the instruction to "really take a moment to think about the importance of the different values in the light of your answers", we used a covert timestamp that was inserted after participants read the instructions and again after they had answered the three questions about whether they were intended to change their 2) We did not consider the participants (N = 37) who did not choose one of the pre-given groups. The least-liked group of these participants is not known and therefore these participants could not be asked questions about their intentions to change their views towards this group.
3) With a different national representative sample (N = 912) and using 7-point scales, we found that 'freedom of speech' (M = 4.95, SD = 1.35) and 'public peace and order' (M = 4.91, SD = 1.30) were considered equally convincing policy arguments ("How convincing do you find [specific value] as a decisive argument for policy?"). views. On average, completion time was 60.40 seconds (SD = 70.11; median = 48; mode = 44; range 7 -946) which indicates that in general participants deliberated during the task.
To validate the assignment of the participants to the tolerant or the intolerant group, we checked whether the tolerant group generally valued 'freedom of speech' more than 'public order' and the intolerant group indeed prioritized 'public order'. Participants were asked at the end of the questionnaire to evaluate both values. Four filler items with oth er values were used (e.g., 'equal opportunities', 'personal responsibility') and the values were introduced and presented together so that participants could evaluate them in relation to each other. For each value, three questions with 7-point scales were presented: "How important is this value for you in comparison to other values?", "How strongly is this value for you a moral principle?", and "How important do you find this value for making policy decisions". The three items for freedom of speech formed a reliable scale (α = .80; M = 5.22, SD = 1.00) and this was also the case for the three items for public order (α = .85; M = 5.05, SD = 1.04). Following our theoretical reasoning and previous research, we examined the process of value prioritization by using a difference measure in which an individual's value for social order is subtracted from the importance attached to freedom of speech (Adelman et al., 2021;Gibson et al., 2020;Peffley et al., 2001). Thus, a higher or more positive score means that in general participants consider the value of freedom of speech relatively more important as a basis for policy than public order, and vice versa for the lower or more negative score. Overall, there was a somewhat stronger emphasis on freedom of speech over social order (M = .18, SD = 1.12). Importantly and in support of the computerized assignment of the participants, the tolerant group considered freedom of speech relatively more important than public order (M = .52, SD = 1.07), compared to the intolerant group (M = -.05, SD = 1.09). The difference between the two groups is significant and substantial, t(545) = 6.14, p < .001, d = .54. Moreover, the tolerant group valued freedom of speech more than the intolerant group, t(545) = 3.44, p < .001, d = .30; M = 5.41, SD = .95, and M = 5.11, SD = 1.01, respectively, and valued social order less, t(545) = -3.15, p = .002, d = .28; M = 4.88, SD = 1.07, and M = 5.16, SD = 1.00. Yet, on average, both groups considered freedom of speech and public order to be important values with the average scores significantly above the midpoint of the scales, p s < .001. This suggests that appeals to both values can resonate in both groups, making it meaningful to ask participants in the two groups to seriously consider these values in deciding whether to tolerate the protest actions of their least-liked group (Nelson & Garst, 2005;Peffley & Hurwitz, 2007).

Tolerance
We first examined how tolerant the national sample of participants was of the socially disruptive actions of their least-liked group. Following previous research on tolerance (Adelman et al., 2021;Gibson et al., 2020), and the labelling of the scale endpoints together with the subsequent automatic (computerized) division of the participants into two groups, we found that a majority of participants were generally more intolerant (61.1%; score < 16) than tolerant (38.9%, score ≥ 16), which supports the notion that the protest actions were considered socially disruptive (Verkuyten et al., 2023). The average tolerance score for the four protest actions was 3.34 (SD = 1.57) which is significantly below the midpoint, t(545) = -9.82, p < .001. Compared to tolerant participants, the intolerant participants were more strongly right-wing oriented politically, t(467) = 3.18, p = .002 (M = 3.91, SD = 1.32, and M = 4.33, SD = 1.45, respectively) and had a lower formal educated, t(545) = 4.76, p < .001, but the two groups did not differ significantly in gender and age.

Change of Views
After the tolerant group was asked to carefully consider the importance of public order and safety and the intolerant group the importance of freedom of speech and the right to demonstrate, both groups were asked whether they were inclined to change their views on accepting the protest actions. On average, there were no significant differences in the amount of time both groups of participants took to reflect on the importance of the additional value and to respond to the three items, t(545) = .48, p = .64, suggesting that both tolerant and intolerant people engaged in deliberative thinking to a similar extent.
Participants were instructed to choose the midpoint (4) of the bipolar scales if they were not inclined to change their views. As expected, compared to this midpoint, the tolerant group became significantly less accepting of their least-liked group, t(212) = 3.31, p = 001, d = .23; M = 3.80, SD = .85. Thus, in line with the asymmetry hypothesis, carefully considering the importance of public order and safety made tolerant people somewhat less accepting. However, the intolerant group also became less accepting towards their least-liked group after being asked to consider a value to accept their least-liked group indicating a contrast effect, t(333) = 11.39, p < .001, d = .62; M = 3.44, SD = .90. Thus, in line with the asymmetry hypothesis, being asked to carefully consider the importance of freedom of speech and the right to demonstrate did not make the intolerant participants more tolerant. Rather, a contrast effect emerged and they became even more intolerant, and more strongly so compared to the tolerant group: the difference in average change between both groups was significant, t(545) = 4.81, p < .001, d = .42.

Additional Analyses
To examine the robustness of the findings, we conducted two additional analyses. First, we considered completion time for examining whether outliers impacted on the findings. Participants (N = 28) who took less time than one-third of the median were considered speeders (Miller et al., 2020). Excluding them from the analysis did not change the findings: there was again a similar significant difference in the intention to change one's views between the tolerant (M = 3.77, SD = .86) and intolerant group (M = 3.40, SD = .88), t(517) = 4.57, p < .001, d = .42. Additionally, we excluded participants who were slower than one-third of the median and this also did not change the pattern of results.
Second, using the general linear model (GLM) univariate procedure, we compared the difference in average change intention of the tolerant and intolerant groups while controlling statistically for age, gender, level of education 4 , and political orientation (as covariates). The general linear model is a flexible generalization of analysis of variance and regression analysis and yields similar results (Rutherford, 2001). This analysis showed a similar significant difference between the tolerant group and the intolerant group in the inclination to change views, F(1, 492) = 18.66, p < .001, η p 2 = .04, and there were no significant main effects for any of the four demographic variables (p s > .18). Additionally, there were no significant interaction effects between any of these variables and the two groups (p s > .24). Thus, the asymmetry hypothesis was supported regardless of whether we considered the age, gender, education level, or political orientation of the participants, and was similar for older and younger participants, females and males, higher and lower educated, and right-wing and left-wing oriented people.

Discussion
Using a national representative sample, we found empirical evidence for the asymmetry of political tolerance: it was easier to persuade the tolerant to become less tolerant than to convince the intolerant to become more tolerant (e.g., Gibson, 2006;Peffley et al., 2001). In fact, we found among the intolerant participants a backlash effect where consideration of alternate reasons to tolerate their least-liked ideological group resulted in higher intolerance, as found in other studies (e.g., Djupe et al., 2015;Traunmüller & Helbling, 2017). This supports the notion that tolerance is more fragile than intolerance because of the psychological discomfort and uneasiness of cognitive dissonance (Festinger, 1962;Harmon-Jones & Mills, 1999) that is involved . The empirical role of psychological discomfort as the underlying mechanism might be studied in future research. For example, recent research using brain activity (EEG) has indeed found higher cognitive conflict following an outgroup toleration task than a control task . Such brain imaging techniques can help shed light on the underlying mechanisms involved.
Theoretically, the asymmetry of tolerance and the related psychological processes further suggests that tolerance and intolerance are indeed not "merely polar opposites on an unidimensional continuum" (see quote above, Gibson, 4) Education was measured on a 7-point ordinal scale, but similar to other research in the Netherlands (e.g. De Graaf et al., 2000;Van de Werfhorst & Van Tubergen, 2007), it was treated as a continuous variable in the analysis. An additional analysis with the categories of 'low, middle, and high' level of education in which 'low education' was the refence category, yielded similar results. 2006, p. 29). This is also indicated in research using person-centred approaches such as latent class and profile analysis. For example, in a US study on political tolerance, the best model for the data in a latent class analysis required four categories of individuals rather than a continuum of tolerance-intolerance (McCutcheon, 1985). These four categories could not be readily placed on a unidimensional positive-negative continuum because there was no monotonic change across the four groups of individuals. Rather, they formed four latent classes of political tolerance. Similarly, in examining tolerance among a nationwide sample in the US and also in the Netherlands, four qualitative different latent profiles were found including subgroups of individuals that were generally intolerant or generally tolerant (Adelman & Verkuyten, 2020;Mather & Tranby, 2014).
The practical importance of the findings is that it is easier to create intolerance out of tolerance, than vice versa. This suggests, for example, that populist appeals to intolerance towards the full rights of citizenship of political adversaries or minority groups might be more persuasive to the public than appeals to tolerance. Furthermore, the psychological consistency of intolerance (you reject what you object to) makes it more likely that people negatively act on their negative attitude: intolerance tends to have stronger behavioural consequences than does tolerance with its psychological dissonance and self-restraint (Gibson, 2006). These possible implications make it important to more fully examine the psychological differences between tolerance and intolerance and when and why intolerance increases, or rather how tolerance can be stimulated.

Limitations and Future Research
In evaluating our findings, some limitations should be considered that provide directions for future research. First, we focused on political tolerance and it is unclear whether the asymmetry also exists in social domains in which people have to deal with questions of whether to endure certain beliefs and practices that they object to, such as cultural and religious differences as well as views and behaviors of friends, disagreements at work, or trouble in the neighborhood.
Second, we focused on socially disruptive but conventional protest actions and it is likely that there is no asymmetry in (in)tolerance for protest actions that violate moral standards and therefore are generally considered unreasonable, threatening, and emotionally harmful (e.g., Feinberg et al., 2020;Simpson et al., 2018;Verkuyten et al., 2023). Further more, we focused on people's tolerance of disruptive protest actions of mainstream ideological groups that are typically democratic and differ from radical extremist groups (e.g. terrorists, neo-nazis). However, average levels of tolerance can vary across mainstream disliked groups (Gibson et al., 2020) and this might matter for the pliability of tolerance and intolerance. For example, some of these groups might be perceived as less democratic making it easier to persuade people to limit the civil liberties of these groups (Petersen et al., 2011). Hence, future research should examine the asymmetry hypothesis of tolerance in relation to the nature of protest actions and the nature of the least-liked ideological group.
Third, in examining the asymmetry hypothesis, we considered the contrasting values of liberty and order. These two values are found to be central in people's thinking about disruptive protest actions and dissenting practices. For example, the Dutch public considers both values as equally decisive arguments for policies (Footnote 2) and equally important in relation to specific dissenting minority practices (Adelman et al., 2021). Furthermore, valuing freedom of speech over public order (or vice versa) distinguished tolerant and intolerant participants in the current research, and has been found to exert strong effects on political tolerance (e.g., Gibson et al., 2020;Nelson et al., 1997;Peffley et al., 2001). However, future research could investigate the pliability of tolerance and intolerance by focusing on additional values (e.g., responsibility, equality, reciprocity) and the ways in which people weigh and balance these in deciding to tolerate or not to tolerate disruptive protest actions (Verkuyten et al., 2023).
In addition to the type of values and considerations it is possible to examine the effectiveness of different ways in which persuasive messages and counter arguments are presented. We asked people to reflect on the contrasting values and the completion times suggest that they did engage in reflective thinking. However, it could also be examined whether, for example, a custom animated video or a writing task is more effective in stimulating tolerance of disliked groups and create less reactance.
And finally, we measured change by explicitly asking participants to what degree they were inclined to change their views after being asked to carefully consider an alternative value. However, change could also be measured more indirectly or implicitly which would go beyond self-reported change and rule out the possibility that participants report an inclination to change because of potential response tendencies such as social desirable responding. Additionally, for measuring tolerance and the inclination to change views, we used familiar bipolar scales with end-point anchors. However, these anchors were not explained to the participants and therefore we do not know which understanding of tolerance the participants had. Yet, research using open-ended questions in a survey on national sample of Dutch adults showed that the large majority understood tolerance as having to do with endurance and putting up with something one objects to (Verkuyten & Kollar, 2021).
Despite these limitations and venues for future work, the present findings demonstrate that tolerance is more pliable than intolerance. Tolerance is more fragile because it involves "the suppression of the tendency to suppress" (Schuyt, 1997, p. 169) and the required self-restraint implies dissonance and that has been described as 'painful' and 'to suffer' (Tønder, 2013;Williams, 1996). Yet, tolerance is indispensable for our increasingly polarized liberal democratic societies making further research on the social psychology of tolerance and intolerance topical and urgent.

Acknowledgments:
The authors have no additional (i.e., non-financial) support to report.

Competing Interests:
The authors have declared that no competing interests exist.

Ethics Statement:
The research was ethically approved by the Ethics Committee of the Faculty of Social and Behavioral Sciences, Utrecht University. File number: 21-0151.
Data Availability: All the data for this study is publicly available (see Verkuyten, Yogeeswaran, & Adelman, 2023). They are also stored at the special storage facility of Utrecht University and are openly accessible at the Data Archive and Networking Services of the Royal Dutch Academy of Sciences.

Supplementary Materials
The Supplementary Materials contain the following items (for access see Index of Supplementary Materials below): • The pre-registration protocol for the study • All research data to replicate the findings