Free Speech as a Cultural Value in the United States

Political orientation influences support for free speech, with liberals often reporting greater support for free speech than conservatives. We hypothesized that this effect should be moderated by cultural context: individualist cultures value individual self-expression and self-determination, and collectivist cultures value group harmony and conformity. These different foci should differently influence liberals and conservatives’ support for free speech within these cultures. Two studies evaluated the joint influence of political orientation and cultural context on support for free speech. Study 1, using a multilevel analysis of data from 37 U.S. states (n = 1,001), showed that conservatives report stronger support for free speech in collectivist states, whereas there were no differences between conservatives and liberals in support for free speech in individualist states. Study 2 (n = 90) confirmed this pattern by priming independent and interdependent self-construals in liberals and conservatives. Results demonstrate the importance of cultural context for free speech. Findings suggest that in the U.S. support for free speech might be embraced for different reasons: conservatives’ support for free speech appears to be motivated by a focus on collectively held values favoring free speech, while liberals’ support for free speech might be motivated by a focus on individualist self-expression.

The right to free speech and free expression is central to democracy, as governmental guarantees of freedom of speech are critical in developing a cultural context conducive to the proper functioning of democratic governments (Molnar, 2009).In the United States, the belief in the constitutionally enshrined right to freedom of expression is a cornerstone of American ideology (Patterson, 1979).Public opinion reflects the importance of this belief: a plurality of Americans (47%) highlight freedom of speech as the most important freedom citizens enjoy (First Amendment Center, 2013), and a majority of Americans (86%) believe protecting free speech is more important than protecting people from being offended (First Amendment Center, 2016).An individual's right to free speech, however, may be constrained by social or legal action based on what the individual's society believe to be acceptable boundaries.
The purpose of the present study is to investigate how cultural context (specifically, individualism-collectivism and related self-construals) and political orientation influence individuals' support for free speech within the United States.Political psychologists have long called for a culturally embedded understanding of values and public opinion, especially as they relate to support for/opposition to policies.Simplifying notions of "American public opinion" ignore the diversity of U.S. society in terms of its regional and local cultures.By focusing on cultural factors which might influence support for free speech, we follow Renshon's (2002) call for political psychologists to study political psychology in its cultural context (e.g., Almond & Verba, 1989;Inglehart, 1997).
In this article, we argue that cultural context and political orientation alone are not sufficient in understanding individuals' attitudes toward policy.Cultural context, and the cultural values and orientations of individuals within cultural contexts, are an important factor in the process of developing attitudes toward specific policies.Political orientation is similarly a potent indicator of individuals' attitudes toward various policies.Yet, frequently the relationships between political orientations and issue positions are not obvious.Often, conservatives and liberals support or oppose the same policies albeit in different contexts (e.g., Brandt et al., 2014;Cohen, 2003).We propose that cultural context serves as a critical moderator in the relationship between individuals' political orientation and attitudes toward policy.We focus specifically on freedom of speech as the policy under evaluation due to its position as one of the core values of the American democracy.

Support for Free Speech
At the core of most modern debates about freedom of speech is the need to balance individuals' right to express their views and opinions with the need to protect individuals from harm caused by hateful speech (Cowan, Resendez, Marshall, & Quist, 2002).In this debate, the United States is unique in that the law places considerably greater protections to allowing individuals' free expression, compared to protections against harm caused by the expression of hateful views (Bird, 2000;Schauer, 2005).This focus stems from both legal doctrine (i.e., specific laws regarding free speech) and legal outcomes (i.e., the manner in which courts interpret these laws).This results in an exceptionally strong right to free speech in the United States, compared to other liberal democracies (Schauer, 2005).This is not to say, however, that Americans in general agree with such broad support for free speech (although Americans generally report strong support for free speech; Chong, 2006).Rather, Americans' support for free speech varies depending on contextual and individual factors.One contextual factor that highlights the conflict between allowing free speech and protecting individuals from harmful speech is issue framing (Druckman & Nelson, 2003;Nelson, Clawson, & Oxley, 1997;Sniderman & Theriault, 2004).The effect of issue framing is relatively straightforward: when issues or laws are presented in a way that highlights the importance of free speech, individuals tend to be more supportive of free speech, compared to circumstances which highlight the importance of public order (Nelson et al., 1997), or the importance of limiting the influence of special interests in government (Druckman & Nelson, 2003).These findings highlight the constant balance between free expression and other considerations (e.g., security) at the core of balancing free speech.
Other factors beyond framing can similarly influence support for free speech.Individuals' asked to consider the consequences of speech report lower support for free speech compared to individuals who were not asked to consider these consequences (e.g., violence; Kuklinski, Riggle, Ottati, Schwarz, & Wyer, 1991).Individuals primed with the idea of free speech report greater support for free speech compared to individuals primed with the idea of equal protection under the law (Cowan et al., 2002).Perceptions of threat can similarly influence support for free speech: individuals who experience threat about the future have weaker support for civil liberties (including free speech) compared to individuals who do not experience said threat (Davis & Silver, 2004).On a broader scale, more collectivist societies tend to be less supportive of free speech, compared to individualistic societies (Van de Vliert, 2011).Support for free speech also varies based on individual characteristics of the respondent.Men are generally more supportive of free speech than women (Cowan et al., 2002;Downs & Cowan, 2012;Lambe, 2004).More religious individuals tend to be less supportive of free speech than less religious individuals (Lambe, 2004), and greater endorsement of authoritarian values (e.g., right-wing authoritarianism) is linked to lower support for free speech (Downs & Cowan, 2012).Political orientation also predicts support for free speech: generally, liberals report greater support for free speech, compared to conservatives (Davis & Silver, 2004;Downs & Cowan, 2012;Lindner & Nosek, 2009).Some researchers argue, however, that this effect is driven by speech content.That is, liberals and conservatives tend to be similar in their support for free speech when presented with speech they agree with (Crawford & Pilanski, 2014).The following section evaluates in more detail the relationship between political orientation and support for free speech.

Political Orientation and Free Speech
As previously noted, individuals' support for free speech can be influenced by their political orientation (Crawford & Pilanski, 2014;Lindner & Nosek, 2009).The specific effect of political orientation on support for free speech is unclear: on the one hand, some evidence suggests that conservatism is linked to lower support for free speech (e.g., Fisher et al., 1999;Lindner & Nosek, 2009).On the other hand, there is also evidence that support for free speech appears to be dependent on speech content (Crawford & Pilanski, 2014;Suedfeld, Steel, & Schmidt, 1994).Much as framing influences support for free speech by highlighting the importance of either free speech or public order (Nelson et al., 1997), political orientation influences support for free speech by establishing the "correct" position on an issue.In other words, individuals will be more supportive of speech promoting the position most congruent with their political orientation (Crawford & Pilanski, 2014).
There is both direct and indirect evidence for the "conservatism effect."By direct evidence we refer to studies in which measures of conservatism are related to lower support for free speech.Lindner and Nosek (2009) found that participants were generally supportive of an individual's right to post a controversial sign in their lawn, regardless of content (i.e., anti-Arab speech or anti-American speech).This support, however, was greater among selfidentified liberals and moderates, compared to self-identified conservatives.Another study found that social conservatism (defined as support for traditional values) was predictive of support for censorship, while economic conservatism (defined as support for limited government intervention in the economy) was not (Fisher et al., 1999).
Conservatism is similarly linked to greater support for security over civil liberties (including free speech), compared to liberals (Davis & Silver, 2004).By indirect evidence for the "conservatism effect" we refer to studies that highlight certain aspects of conservatism which can be inferred to predict lower support for free speech.Conservatism is characterized by support for traditional norms and values (Ellis & Stimson, 2012;Wilson, 1973).This implies a greater resistance to social change, as supporting traditional values implies one's preference for an existing social order (Jost, Glaser, Kruglanski, & Sulloway, 2003;Stenner, 2005Stenner, , 2009)).Consequently, conservatives can be expected to be more dogmatic (Choma, Hafer, Dywan, Segalowitz, & Busseri, 2012) and hold negative attitudes toward status quo challenges (e.g.Jost, Banaji, & Nosek, 2004).Support for free speech inherently applies to views that challenge the status quo, because views that are consistent with the status quo tend to be well received and their expression does not necessitate protections.Similarly, conservatives tend to express greater negativity bias than liberals (Hibbing, Smith, & Alford, 2014).This means that conservatives are more likely to focus on negative (as opposed to positive) outcomes, and be more influenced by negative outcomes, compared to liberals.Given that support for free speech is lowered when individuals consider the consequences of speech (Kuklinski et al., 1991), it is possible that a tendency to focus on negative consequences would result in lower support for free speech among conservatives, compared to liberals.
Evidence for the "content effect" suggests that liberals and conservatives reported similar support for censorship of topics traditionally opposed by each group: Liberals were less supportive of allowing the sale of pro-gun magazines in convenience stores, while conservatives were less supportive of allowing the sale of pornographic magazines in convenience stores (Hense & Wright, 1992;Suedfeld et al., 1994).Similarly, liberals were more tolerant of a speech critical of a conservative president (George W. Bush), while conservatives were more tolerant of an individual critical of a liberal president (Barack Obama; Crawford & Pilanski, 2014).
These contradictory findings might be a result of the multifaceted nature of conservatism (Ellis & Stimson, 2012;Stenner, 2005Stenner, , 2009)).While conservatism is characterized by a "hands off" approach to economic policy (i.e., laissez-faire conservatism; Stenner, 2005), its approach to social policy is divided: some conservatives support a similar "hands off" approach to social issues, while other conservatives support government intervention to uphold traditional values (Ellis & Stimson, 2012).The former approach would be conducive to greater support for free speech, while the latter approach would be conducive to greater support for censorship (at least, for speech considered antithetical to one's values).The issue, then, is to determine under which conditions one facet of conservatism is more salient than the other.We argue that one important condition influencing which aspect of conservatism is more salient is an individual's cultural context, specifically in terms of individualism-collectivism.
Our research will evaluate the impact of a specific cultural context (individualism-collectivism) on support for freedom of speech.The focus on individualism-collectivism stems from documented evidence, discussed further in the next section, that individualist and collectivist cultures differ in the degree to which their members accept governmental intervention on social issues (Conway, Sexton, & Tweed, 2006;Kemmelmeier, Wieczorkowska, Erb, & Burnstein, 2002;Ozawa, Crosby, & Crosby, 1996).

Culture and Free Speech
Culture plays an important role in the process of negotiating the boundaries of civil rights (e.g., Turiel & Wainryb, 1998).The present research examines the cultural context of individualism-collectivism (Hofstede, 2001;Oyserman, Coon, & Kemmelmeier, 2002;Triandis, 1995).Broadly speaking, individualism and collectivism reflect the manner in which those living within a particular culture construe the relationship between individuals and society, and whether the individual or the group is the focal point of society (Oyserman & Lee, 2008).With individualist cultures focusing on individuals, and collectivist cultures centered on groups, societies differ with regard to societal values and goals (Oyserman et al., 2002).Individualist societies, such as the United States and Western Europe, champion individual self-determination, the primacy of rights over duties, and a focus on individual self-fulfillment (Triandis, 1995).In this cultural context, behavior is typically viewed as a reflection of internal attributes of the self, with social rules being viewed as constraining authentic expression of the individual (Markus & Kitayama, 1991).Collectivist societies, such as many East Asian or African countries, highlight the interconnectedness of individuals with their groups and often stress individuals' obligations to their families and communities (Hofstede, 2001;Oyserman et al., 2002).Such cultural contexts tend to emphasize interpersonal relationships and increase the motivation to fit in with others, such that behavior is shaped by what is appropriate or acceptable within the social context (Hui & Triandis, 1986).
Societal differences in individualism-collectivism tend to be associated with different types of self-construal, i.e., the set of characteristics which allows individuals to understand themselves as either independent from or interdependent with others (Cross, Hardin, & Gercek-Swing, 2011;Markus & Kitayama, 1991).All individuals think of themselves as independent or interdependent at different times and in different contexts.Cultural practices and messages in individualist societies tend to highlight independent self-construals, such that individuals are more likely to think of themselves as autonomous agents, unique and different from others.Conversely, collectivist societies emphasize interdependent self-construals such that individuals are more likely to conceive themselves as embedded in their groups and defined by their social relationships (Kitayama, Park, Sevincer, Karasawa, & Üskül, 2009).Because self-construals are highly sensitive to contextual factors, researchers have used experimental manipulations of self-construals to investigate psychological processes related to individualism/collectivism, thus simulating some aspects of cross-cultural differences (e.g., Gardner, Gabriel, & Lee, 1999).
Importantly, variations in individualism-collectivism have been related to greater governmental regulation and greater support for governmental regulation.Conway and colleagues (2006) demonstrated that, whether looking at cross-national or cross-regional differences, collectivist societies are more likely to impose governmental restrictions on individual behavior, including freedom of speech.Other studies indirectly suggest that collectivism is linked to greater government intervention.Ozawa and colleagues (1996) demonstrated that affirmative action is more acceptable in a collectivist society as it reflects a group-based approach to justice (see also Kemmelmeier, 2003).Similarly, Kemmelmeier et al. (2002) demonstrated that people endorsing individualist values or those living in highly individualist U.S. states found physician-assisted suicide more acceptable as an individual choice than those not endorsing such values and those living in collectivist U.S. states.In both cases, collectivism is related to greater government involvement in citizen's lives, whether it is by enacting "affirmative action" policies (a type of government policy aimed at promoting greater representation of different groups in educational settings), or by promoting greater governmental oversight of doctor-patient relationship.
We argue that these relationships between collectivism and greater government involvement might be a reflection of collectivist contexts highlighting the facet of conservatism that supports government intervention to uphold traditional values (Ellis & Stimson, 2012).This is supported by the fact that collectivism is linked to greater endorsement of tradition and conformity (Cukur, De Guzman, & Carlo, 2004) and to greater endorsement of conventionalism (i.e., support for conventional morality and support for traditional values; e.g., Kemmelmeier et al., 2003).Collectivism, however, does not necessarily imply a greater endorsement of conservatism (cf.Kemmelmeier et al., 2003).
Conservatism is characterized by a tendency to support the status quo (Jost et al., 2003).What individuals define as the status quo, however, is likely to be influenced by their culture.For example, conservatives within individualist cultures should be more likely to endorse individual rights and freedom of choice, because a focus on individuals (rather than communities) is a defining aspect of individualist cultures.Conservatives within collectivist cultures, however, should be more likely to endorse tradition and prioritizing familial and communal needs over individual needs, because a focus on community (rather than individuals) is a defining aspect of collectivist cultures.In this sense, conservatism should "amplify" the focus of the dominant culture.This amplification should be strongest within collectivist cultures, because individuals in these cultures are more strongly motivated to "fit in," compared to individuals in individualist cultures.Because liberals are less likely than conservatives to endorse the status quo, this amplification effect is not expected to appear among liberals.The dynamics might be reversed in individualist cultures, such that the focus on individuals (rather than communities) in these cultural contexts may seem more compatible with liberals' emphasis on personal freedom and self-determination.Yet, compared to collectivist contexts, individualist contexts may exert less pressure toward conformity (e.g.Bond & Smith, 1996;Fincher, Thornhill, Murray, & Schaller, 2008).Thus, in individualistic contexts conservatives and liberals may be less likely to differ in their social views than in collectivistic contexts.
In the context of support for free speech, culture should moderate the effect of political orientation on support for free speech.Given that collectivist cultures are characterized by greater support for governmental regulation of individuals' lives (e.g., Conway et al., 2006;Ozawa et al., 1996), we would expect conservatives in collectivist cultures to "amplify" this by expressing greater support toward governmental regulation of freedom of speech, relative to liberal individuals.In other words, conservatives should be less supportive of freedom of speech than liberals in collectivist contexts.Conservatives and liberals may be less likely to differ in their social views in individualistic contexts, however, since individualist contexts exert less pressure toward conformity.

The Present Studies
Two studies sought to test this amplification hypothesis concerning the moderating role of cultural context on the relationship between political orientation and individuals' support for free speech.We argue that U.S. respondents' support for free speech will be influenced by cultural context and political orientation.Specifically, greater support for free speech will vary among liberals and conservatives depending on whether their immediate cultural context or self-construal highlights an individual or collective focus.Study 1 evaluated the relationship between political orientation and state-level cultural context, and support for free speech using a national sample of U.S. respondents.
Study 2 tested the variation in the relationship between political orientation and support for free speech through an experimental manipulation of individuals' cultural self-construal.Notably, across our two studies, we investigated the same conceptual hypothesis at the culture (aggregate) level as well as at the individual level of analysis.This was critical because we cannot assume that findings obtained at one level of analysis will transfer or be applicable to a different level of analysis (e.g., Na et al., 2010;Shweder, 1973); rather, this has to be empirically demonstrated.

Study 1
The purpose of Study 1 was to investigate the relationship between individualism-collectivism, political orientation, and support for free speech throughout the United States.To this end, we used survey data to investigate how support for free speech among U.S. residents varies based on participants' state of residence.Relying on documented differences in collectivism between different states in the United States (Vandello & Cohen, 1999) the present research examined to what extent the association between individuals' political orientation and support for free speech varies as a function of state levels of collectivism.

Method Participants
The survey data include responses from 1,001 individuals from 36 U.S. states, with state-level sample sizes ranging from 9 (Oklahoma) to 105 (California), very similar to previous research (e.g., Kemmelmeier et al., 2002).The 14 states not included were disproportionately small in population and low in collectivism (e.g., South Dakota, Alaska).Participants were recruited using a classic cluster design, aimed at creating a representative sample of the adult U.S. population living in households.The survey was conducted in person, and for each household, one member over the age of 18 was chosen to participate using the most recent birthday method.The sample was 56.3% female, with a median age of 44 years (age range: 18-90).The ethnic breakdown was 72.5% White, 15.8% African-American, 9.7% Hispanic, 1.7% Asian and 0.3% "Other."

Procedure
Data were drawn from the United States Citizenship, Involvement, Democracy Survey (CID; Howard, Gibson, & Stolle, 2006).The CID is a survey of attitudes about civic involvement, social capital, and democracy, including participants' attitudes toward free speech regulation.CID data were collected through in-person interviews conducted in 2005.

Measures Support for Free Speech
We identified six variables in the CID data relevant to the study goals, which were combined into a measure of support for free speech (α = .89).Respondents were asked to select from a list the two social groups which they dislike the most (see Control variables), and indicate their agreement or disagreement with three items for each group on a 5-point Likert scale: whether the group a) should be allowed to make a speech in the respondent's community, b) should be banned from running for public office, and c) should be allowed to hold public demonstrations in the respondents' community.Variables were recoded so that higher scores indicate greater support for free speech.

Political Orientation
Participants' political orientation was measured using a 10-point Likert item (1 = Very Liberal, 10 = Very Conservative).Such a single-item assessment of ideology has been demonstrated to be highly effective in predicting political attitudes across a broad range of domains (see Jost, 2006 for a review).

Collectivism Index
We included data on individualism/collectivism for each U.S. state in the sample using the U.S. state collectivism index developed by Vandello and Cohen's (1999), and which has been widely used in the cultural psychology literature (e.g., Allik & Realo, 2004;Conway, Ryder, Tweed, & Sokol, 2001;Conway et al., 2006;Crowder & Kemmelmeier, 2017;Harrington & Gelfand, 2014;Kemmelmeier et al., 2002;Kemmelmeier, Jambor, & Letner, 2006;Thornhill & Fincher, 2011;Varnum, 2012).This index was developed by combining eight variables indicative of individualist or collectivist tendencies among the population of each state.These variables are a. the percentage of people living alone: people in individualist societies are more likely to perceive themselves as autonomous, which in turn should promote greater rates of people living alone, compared to collectivist societies (Triandis, 1989).
b. the percentage of elderly people living alone: Related to the first variable, people in individualist societies are less likely to allow their parents to live with them, compared to people in collectivist societies (Triandis, 1989).c. the ratio of divorce rate to marriage rate: Individualist cultures around the world have higher rates of divorce than collectivist cultures (Hamamura, 2012).d. the percentage of people with no religious affiliation: religion promotes integration and regulates moral conduct, as a result it is expected to promote collectivism (Farias & Lalljee, 2008).Similarly, individualism should be related to lower interest in religion (Triandis, 1995).

e. the average percentage of votes for the Libertarian party during the past four presidential elections:
Libertarianism is focused on individual freedoms and self-governance, values which are also central to individualism (Vandello & Cohen, 1999).
f. the percentage of self-employed people: self-employment reflects individualist values (e.g., self-governance) and provides fewer opportunities for socialization with coworkers.These factors relate self-employment with individualism (Vandello & Cohen 1999).
g. the ratio of people carpooling to work compared to those driving alone: carpooling is perceived as a collectivist activity, and individualist people are therefore expected to be more likely to drive alone (Vandello & Cohen, 1999).
h. the percentage of households with grandchildren in them: households in collectivist societies are expected to include extended family members more than households in individualist societies (Triandis, 1989).
With the exception of the last two variables, lower values indicate higher levels of collectivism.For the combined index, higher values indicate higher collectivism.Each respondent was assigned the index score corresponding to their reported state of residence.

Control Variables
The individual-level control variables included participants' political interest, education, religiosity, and participants' specific choice of least-liked groups.
Political interest was measured using a 4-point Likert item ("How interested would you say you are in politics?" 1 = Not at all interested; 4 = Very interested).Education was measured using a categorical variable with eight levels: no high school education (dropped out of school before grade 9), some high school education (dropped out of school between grades 9-11), high school diploma (the reference category), GED (a high school equivalent degree for individuals who did not finish high school), technical/vocational degree, some college education without a degree, college degree, and postgraduate degree (medicine, law, PhD).Religiosity was measured using an 11-point Likert item ("Regardless of whether you belong to a particular religion, how religious would you say you are?" 0 = Not at all religious, 10 = Very religious).
Another set of control variables sought to control for the possibility that the relationship between participants' political orientation and support for free speech was influenced by their specific choice of least liked group or second least liked group.For this purpose, we focused on the six groups that were most commonly chosen by participants as their two least-liked groups (the American Nazi Party, the American Communist Party, the Ku Klux Klan, people who support the military takeover of the government, radical Muslims, and gay rights activists).For each of these groups we created a variable, which was coded as 0 if the participant did not choose the specific group as their least liked or second least liked group, and 1 if the participant did choose the specific group as their least liked or second least liked group.See Table 1 for descriptive statistics on all variables used in the analysis.
Other common control variables (age, gender, and race) had non-significant effects; thus, they were not included in the final set of models.

Results and Discussion
The relationship between cultural context and support for free speech was analyzed using multi-level linear regression, with respondents' score on the measure of support for free speech (our dependent variable) being treated as nested within the state (modeled as a random effect).The level-1 (individual-level) independent variables were respondents' education, political interest, political orientation, and religiosity, with all continuous predictors being grand-mean centered.The level-2 (state-level) independent variable was collectivism scores for U.S. states (Vandello & Cohen, 1999) entered as a grand-mean centered variable.
In addition, our model included a cross-level interaction between political orientation and state-level collectivism to test our prediction that the relationship between political orientation and support for free speech depends on the cultural context.See Table 2 for a summary of the final model, and see Appendix A for details on all models briefly described next.
The null model (-2 Log Likelihood = 2,492.576)included only the intercept and the nesting variable (U.S. states).
The intraclass correlation for our dependent variable was .157,indicating that almost 16% of the variance in support for free speech occurred between U.S. states.
Free Speech as a Cultural Value 716 Model 1 contained the level-1 control variables.Level-1 control variables were participants' political interest, education, religiosity, and participants' six most commonly chosen least liked groups (the American Nazi Party, the American Communist Party, the Ku Klux Klan, people who support the military takeover of the government, radical Muslims, and gay rights activists).This model improved the model fit over the null model (-2 LL = 2,389.897vs. 2,492.576;ΔΧ 2 (15) = 102.679,p < .001).In this model, participants' political interest was significantly related to support for free speech, such that greater interest in politics was related to lower support for free speech.Participants' education was also related to support for free speech, such that participants with some college education, participants with college degrees, and participants with advanced college degrees, all reported greater support for free speech, compared to participants with only a high school education.More religious participants were less supportive of free speech.Participants' choice of least liked or second least liked group was also related to support for free speech: participants who choose the American Communist Party, the Ku Klux Klan, the American Nazi Party, gay rights activists, and radical Muslims, were all less supportive of free speech, compared to individuals who did not choose these groups.Conversely, participants who chose "people who support the military takeover of the government" as their least liked or second least liked group were more supportive of free speech, compared to participants who did not.This model accounted for 14.3% of the level-1 variance compared to the null model.
Model 2 added the level-1 main effect: participants' political orientation.(1) = 5.726, p = .017).All significant control variables from the second model remained significant.In addition, the interaction effect of political orientation and state-level collectivism on support for free speech was significant.Region of significance analyses indicate that the relationship between political orientation and state collectivism on support for free speech is significant in states with collectivism scores lower than 29.853 or higher than 65.983.Specifically, within our sample of states, these values roughly pertain to Nebraska and Wyoming, and South Carolina at the lower and upper end of the collectivism distribution (see Figure 1).However, we should note that there are actually no states in our data with collectivism scores lower than 29.853.As seen in Figure 1, the relationship between political orientation and state collectivism is significant in collectivist states but not in individualist states.Overall, this model accounted for 21.6% of the level-1 variance.
In addition, this model accounted for 26.5% of the variance of the random effect of political orientation, compared to the previous model without the critical cross-level interaction between political orientation and state-level collectivism.As we anticipated, Study 1 indicated conservatives in individualist U.S. states were not significantly different in their support for free speech than liberals.Yet, contrary to our expectations, conservatives in collectivist U.S.
states were more supportive of free speech than liberals.In other words, American conservatives were more supportive of free speech than American liberals when their environment highlighted collectivist values.This result, though not strong in terms of magnitude, was unexpected and contrasted with theory.Moreover, neither the inclusion of the state-level measure of collectivism nor the inclusion of the theoretically predicted interaction term improved the overall model fit.This suggests that there might be other state-level factors which might also moderate the relationship between political orientation and support for free speech.We believed that the results of the study, although unexpected, merited additional attention.Furthermore, while small effects might be unimpressive, they are critical for theory testing and development (Prentice & Miller, 1992).
Therefore, we conducted a follow-up study to replicate the pattern and examine the causal influence of self-construal and political orientation on support for free speech.

Study 2
The purpose of the second study was to investigate the causal impact of self-construal and its interplay with political orientation on participants' support for free speech, this time focused on the individual level of analysis.Our study relied on the documented fact that independent self-construals are dominant in individualist societies, whereas interdependent self-construals are more prevalent in collectivist societies (e.g., Oyserman et al., 2002;Trafimow, Triandis, & Goto, 1991).Indeed, some authors have argued that self-construal priming in many instances can effectively model cultural differences between individualist and collectivist cultural contexts (e.g., Oyserman & Lee, 2007, 2008).To simulate this cultural difference, participants in our experiment were randomly assigned to one of two possible priming conditions: independent self-construal prime or interdependent self-construal prime.
The second factor (Eigenvalue = 2.107, 14.04% of variance explained) had good reliability (α = .86)and included eight items broadly focused on opposition to censorship (e.g., "If it causes severe distress on others, public speech should be heavily restricted," reverse coded).All responses were recorded on a 7-point Likert scale (1 = Strongly Disagree, 7 = Strongly Agree).See Appendix B for a full list of items.

Self-Construal
Supplementary to the experimental manipulation, participants completed the 24-item Self-Construal Scale by Singelis (1994).This scale consists of two 12-item subscales measuring participants' endorsement of independent ("I act the same way no matter who I am with;" α = .73)and interdependent self-construal ("My happiness depends on the happiness of those around me" α = .65).Participants also completed the 11-item relational interdependence scale (RISC; Cross, Bacon, & Morris, 2000;α = .88),which measures participants' endorsement of interdependent self-construal in terms of one's relation to close others.All responses were recorded on a 7-point Likert scale (1 = Strongly Disagree, 7 = Strongly Agree).

Political Orientation
Participants' political orientation was measured using a 10-point Likert item (1 = Very Liberal, 10 = Very Conservative).

Control Variables
Gender was included as a dummy-coded control variable (1 = female; 0 = male).Age and race/ethnicity were dropped from the analysis, as they were not significant.

Results and Discussion
Participants' mean score in the support for free speech subscale was above the mid-point of 4, M(SD) = 5.057(0.939),t(89) = 10.685,p < .001.Participants' mean score in the opposition to censorship subscale was similarly above the mid-point, M(SD) = 4.264(1.129),t(89) = 2.221, p = .029.This indicates that participants generally supported free speech and opposed restrictions on free speech.Participants' mean score for political orientation was not significantly different from the mid-point of 4, M(SD) = 3.933(1.259),t(89) = -0.505,p = .615.Similarly, there were no differences in mean scores for political orientation among participants assigned to the independent self-construal condition compared to participants assigned to the interdependent self-construal condition, M(SD) = 3.978(1.256)vs. 3.884(1.276),t(89) = 0.352, p = .726.These results indicate that the study sample was not biased toward liberalism or conservatism.
When looking at correlations, individual differences in independent self-construal were positively related to support for free speech, consistent with the idea that free speech is reflective of individual self-determination and self-expression.Conversely, the relationship between opposition to censorship and interdependence was negative.
Notably, participants' political orientation was unrelated to their opposition to censorship or their support for free speech (see Table 3 for means, standard deviations, and correlations).
ii  The two free speech scales were moderately correlated (r = .33,p = .002).As a result, both factors were analyzed simultaneously in a multivariate linear regression model.In this model, we used priming condition as a categorical predictor as well as political orientation (mean-centered) as a continuous predictor, while also including the interaction term and controlling for gender.The model accounted for 12% of the variance in participants' support for censorship, R 2 = .117,and 9% of the variance in participants' support for free speech, R 2 = .087.Univariate analyses revealed the interaction term to be significant for both the censorship scale, and the freedom scale (see  A closer evaluation of the interaction effects shows that among participants primed with independent self-construal, political orientation was not significantly related to support for free speech, b(SE) = -0.350(0.396),t( 89 Free Speech as a Cultural Value with interdependent self-construal, greater conservatism is related to stronger support for free speech and stronger opposition to censorship (see Figures 2 and 3).Study 2 demonstrated that support for free speech (e.g., allowing hate speech or opposing government censorship of it) varies depending on individuals' salient self-construal and political orientation.Specifically, when interdependent self-construals are more salient, conservative individuals express greater support for free speech and prefer lower regulation of free speech compared to liberal individuals.A caveat to these findings is that there were only four cases in our sample within the regions of significance outlined above.Despite this issue, these results nevertheless replicated Study 1.The combined findings suggest that the moderating effect of collectivism/interdependent self-construal on the relationship between political orientation and support for free speech is small but robust.
Moreover, the experimental design used in Study 2 provides evidence of a causal effect of self-construals on support for free speech among conservatives and liberals.That is, self-construals that are considered critical to differences between individualist and collectivist cultures served as a causal agent (e.g., Kemmelmeier, 2003).

General Discussion
Based on the available literature on individualism-collectivism (e.g., Conway et al., 2006;Van de Vliert, 2011), we expected individualist contexts and salient independent self-construals to be related to greater support and collectivist contexts and salient interdependent self-construals to be related to lower support for free speech.We also expected the combined effect of interdependent self-construals/collectivism and political orientation to be amplified such that conservatives would report lower support for free speech compared to liberals in collectivist contexts.However, we expected no differences between conservatives and liberals in individualist contexts.
Our results, however, show a starkly different pattern, both in a large-scale national sample as well as a traditional social-psychological lab experiment.First, there was no main effect of cultural context or self-construal priming.
Rather, we obtained an interaction such that the relationship between political orientation and support for free speech in the United States was moderated by cultural context.These results were consistent regardless of whether differences in cultural context were a function of geographical location (Study 1) or a function of the experimental manipulation of self-construals (Study 2).Specifically, within collectivist cultural contexts, conservative participants reported greater support for free speech compared to liberal participants, contrary to our expectations.
Within individualist cultural contexts, however, there was no effect of political orientation on support for free speech.
What might explain the unexpected results?
At first glance, these findings appear to run counter to our initial theoretical assumptions that conservative would be negatively related to support for free speech, and that this effect would be amplified by collectivist contexts.
However, recall that free speech has a special place in American culture.Free speech is one of the cornerstones of American political discourse, and though issues regarding free speech can be very contentious, it is a central value in American society (Patterson, 1979).According to the 1993 General Social Survey, both the majority of self-described conservatives and liberals indicated support for free speech even if it is harmful or offensive to members of other religious or racial groups (Smith, Hout, & Marsden, 2013).Yet, the underlying dynamics of why Americans from different ends of the political spectrum support free speech are less straightforward.Rather, the strength of conservatives' endorsement of free speech values is highly contingent on the cultural context.
Collectivist contexts make tradition and ingroup values salient.To Americans, freedom represents a value that is central to their national history and collective consciousness.Indeed, Americans' understanding of themselves and their culture is predicated on the value of freedom (Cowan et al., 2002).Therefore, embracing one's community and one's traditions is not seen as antithetical to free speech, nor individual freedom in general.To the contrary, American conservatives are particularly committed to freedom (Lipset, 1996).Greater support for free speech would be a natural extension of this commitment.Free speech might seem particularly important to conservatives in collectivist contexts or when interdependent self-construals are salient, both of which highlight traditional values-such as free speech.In this context, conservatives are the "amplifiers" of dominant and traditional values.
By comparison, these traditional values might be less prominent for liberals in interdependent contexts, as liberals are generally less likely to endorse traditional values, compared to conservatives (Jost et al., 2003).
Within individualist contexts or when independent self-construals are salient, traditional cultural values are less influential, as individuals' experience less pressure to conform to perceived social norms relative to individuals' in collectivist contexts or when interdependent self-construals are salient (Cukur et al., 2004).This reduced pressure to conform would explain the lack of a moderating effect of individualist contexts and independent selfconstrual on the relationship between political orientation and free speech: when there is no pressure to conform to an established standard, individuals' views toward free speech will vary based on their individual beliefs on the subject.
The present findings suggest that liberals and conservatives' support for free speech is a function of whether their immediate cultural context (as opposed to the broader American cultural context) is compatible with supporting free speech.There is considerable regional variation in individualism-collectivism across the U.S. (Vandello & Cohen, 1999).The present findings are consistent with the idea that among conservatives, support for free speech stems from an endorsement of traditional American values, which include individuals' right to free expression, whereas support for restricted free speech occurs because conservatives are generally less open to dissent (Jost, Nosek, & Gosling, 2008).
The variation found in support for free speech among individuals of similar political orientation suggests that contextual factors are a critical part of the relationship between these two variables.Specifically, differences in individualism/collectivism and the related salience of independent or interdependent self-construals influence whether decisions are made with a focus on the individual or the group (Markus & Kitayama, 1991; see also Trafimow, Triandis, & Goto, 1991).These different foci, in turn, impact individuals' support for free speech depending on their political orientation.
The present findings have implications for our understanding of the values that underlie the support for free speech.
At least in the United States, greater support for free speech is not inherently a liberal or conservative issue.Individuals' support for free speech was dependent on whether contextual factors highlighted the importance of tradition and community values.However, which context did so varied for individuals of different political persuasion.This might occur due to different cultural contexts leading individuals to emphasizing different moral foundations (Graham, Haidt, & Nosek, 2009;Haidt, 2007).Individualist cultural contexts might emphasize the importance of fairness in allowing all individuals the right to freely express their views, whereas collectivist cultural contexts might emphasize the importance of self-expression as a sacred value of one's ingroup.Furthermore, depending on contextual factors opposing political orientations can result in similar levels of support for free speech.This suggests that individuals of different political orientations might arrive at a consensus on political issues, albeit by relying on different rationales for doing so.
In the United States, support for free speech varies as a function of political orientation and cultural context.
Conservatives in cultural contexts that highlight interdependence or collectivism report greater support for free speech (compared to liberals).These findings have two important implications: first, an analysis of the values or beliefs which underlie support for free speech must take into account individuals' political orientation and cultural context.In other words, individuals within collectivist cultural contexts might rely on different values than individuals within individualist cultural contexts when justifying their support for free speech.Second, the findings suggest a novel approach to increasing individuals' support for more expansive free speech.Specifically, we reason that messages that promote free speech while highlighting the interdependent nature of the individual in society should be effective at increasing conservatives' support for free speech.

Limitations and Future Directions
This research presented no examination of non-U.S.societies.The lack of non-U.S.data is relevant for two reasons: First, Americans have experienced relatively little suppression of speech by the government.In societies which have experienced government-imposed restrictions to free speech, support for free speech is a higher priority because it cannot be taken for granted (see Peng, Nisbett, & Wong, 1997, for the idea of deprivation-based preferences).It is easy to imagine that research in such societies might obtain different patterns concerning the interactive implications of political orientation and cultural context for free speech.Second, Americans tend to be less supportive of state intervention, compared to individuals in other developed nations (e.g., Lipset, 1996).As a result, Americans who identify as conservatives, especially those who also identify as libertarians, might be more focused on small government/free market beliefs, compared to conservatives in other nations.This, in turn, might influence free speech support, as small government/free market beliefs are often considered to be conducive to expansive civic liberties (cf.Ellis & Stimson, 2012).In brief, future research should investigate whether the present findings replicate across other nations.
Although Study 1 clearly had a sufficient sample size, the size of our Study 2 sample might have been insufficient to detect very small effects, a limitation which might also explain the lack of significant results for the independent self-construal prime in Study 2. In addition, Study 2 was likely underpowered.Post-hoc power analyses showed that the interaction effect for both the censorship scale (β = .696)and the freedom scale (β = .768)fell short of the recommended standard of β = .80(Cohen, 1988).Therefore, additional studies might seek to replicate these findings with larger samples to better determine the influence (if any) of individualism/independent self-construal and political orientation on support for free speech.At the same time, the amount of variance explained by our respective models in Study 1 (21%) and Study 2 (12% and 9%) was modest.This implies that there are very likely other relevant variables that were not considered in the present work.Future research might wish to examine the robustness of our findings in light of other, potentially overlapping predictors The absence of a control condition in Study 2 (i.e., one which did not prime either independent or interdependent self-construal) could be a limitation to its findings.This absence is justified in part by the fact that while individuals might switch between independent and interdependent self-construal based on the situation, there is no instance in which neither self-construal would be salient (Markus & Kitayama, 1991).As a result, any differences would be difficult to interpret, as individuals would likely act based on the self-construal most salient at the time of the study.
Evidence from research that has included control conditions suggests that interdependent self-construals are more salient among individuals in control conditions, but only under specific circumstances (MacDonald, Sandry, & Rice, 2012;van Prooijen & Zwenk, 2009).Similarly, some critics might be concerned about the absence of a manipulation check in Study 2. Arguably, one could have expected the experimental manipulation to produce variation on the self-reported self-construals.However, because such priming effects are typically rather shortlived, much of the relevant studies in the literature do not employ any kind of manipulation check as the expected findings typically materialize on the dependent variables (see Oyserman & Lee, 2008 for a review).
Finally, as pointed out repeatedly in the literature, speech content can impact the relationship between individuals' political orientation and support for censorship (e.g., Crawford & Pilanski, 2014;Hense & Wright, 1992;Suedfeld et al., 1994).With especially religion and religiosity emerging as a critical cultural difference (Saucier et al., 2015), it would not be surprising to observe that free speech critical of religion is subject to a different dynamic than speech concerning non-religious controversial issues.Conversely, although the present investigation focused on political orientation in terms of conservatism-liberalism, it is conceivable that endorsing specific values is critical in shaping individuals' responses to the content of speech.Again, we must leave it up to future work to elucidate these issues.

Conclusion
The results presented here demonstrate a variable relationship between political ideology and support for free speech.Our studies suggest that, in the United States, when collectivists values are salient, conservatives express greater support for free speech compared to liberals.Though the underlying mechanism is not yet well-documented, this might occur because collectivist cultural contexts emphasize traditional values, one of which, in the United States, is the right to free speech.These findings highlight the complexity of political ideology by demonstrating that individuals' attitudes about political issues (e.g., free speech) are in part influenced by contextual factors, such as regional culture.Furthermore, our results highlight the importance of incorporating cultural differences when studying the relationship between political orientation and attitudes toward political issues (Hudson & Sampson, 1999).

Notes
i) The other groups which significantly predicted support for free speech (the American Nazi Party, people who would ban all abortions, people who support the military takeover of the government, gay rights activists, and people who wish to ban all religions) were excluded as random effects because their inclusion resulted in a Hessian matrix error.

Figure 1 .
Figure 1.Predicted values for participants' free speech scale scores based on state collectivism (grand mean-centered) and self-ratings of political orientation (state mean-centered) (Study 1, United States Citizenship, Involvement, Democracy Survey).Note.Estimates for collectivism used values for Nebraska (Collectivism Index = 35) and South Carolina (Collectivism Index = 70) as exemplars of states which lie beyond or close to the region of significance for the interaction.*p < .05.

Figure 2 .
Figure 2. Mean values for participants' free speech subscale scores based on experimental condition and mean-centered self-ratings of political orientation (Study 2).*p < .05.

Figure 3 .
Figure 3. Mean values for participants' opposition to censorship subscale scores based on experimental condition and mean-centered self-ratings of political orientation (Study 2).*p < .05.

Table 1
Means and Standard Deviations for Continuous Variables in Multilevel Models Measuring Impact of Individualism/Collectivism and PoliticalOrientation on Support for Free Speech (Study 1) GED stands for General Education Diploma, a U.S. degree awarded to individuals who meet requirements of a high school diploma without having completed high school.
a Reference category = High School graduate.b

Table 2
Estimated Coefficients and Significance for Variables in Multilevel Models Measuring Impact of Individualism/Collectivism and PoliticalOrientation on Support for Free Speech (Study 1) GED stands for General Education Diploma, a U.S. degree awarded to individuals who meet requirements of a high school diploma without having completed high school.
a Reference category = High School graduate.b

Table 3
Means, Standard Deviations, and Correlations for all Scales (Study 2)

Table 4
for all model coefficients).

Table 4
Regression Coefficients for a Multivariate Regression ModelWith Support for Censorship and Support for Free Speech as Dependent Variables, and Priming Condition, Political Orientation (Mean Centered), Gender, and Experimental Prime by Political Orientation Interaction as the Independent Variables = .117,p < .001.b R 2 = .087,p < .001.c Reference category = Male.d Reference category = Independent self-construal prime.
a R 2