Political dispositions

The first of these categories follows naturally from deliberative democrats’ preoccupation with everyday political talk as an activity whose essence consists in the expression, exchange, and mutual consideration of views about politics. In this perspective, its meaning for those performing it is believed to emanate exclusively from its topical content. Accordingly, dispositions acquired during individuals’ political socialization can be expected to affect their views concerning the prospect of discussing politics (Beck and Jennings, Reference Beck and Jennings1982). Most prominent among potentially relevant political dispositions is interest in politics which can be understood as a stable ‘expectation that engaging with political content […] in the future will turn out to be rewarding’ (Prior, Reference Prior2019: 4). It should give rise to the belief that political conversations are gratifying experiences because they offer the opportunity to occupy oneself intensely with a subject matter deemed intrinsically attractive. Accordingly, it should lead to favorable attitudes toward everyday political talk.

Passionate stances with regard to certain political worldviews, ideological camps or parties could also allow political discussions to appear as gratifying. Persons with strong attitudes concerning directional alignments like partisanship and ideology may see them as welcome opportunities to affirm their political identity, express it to others, and persuade them of the superiority of their views. As a generalized self-confidence in political matters, internal political efficacy (Craig and Maggiotto, Reference Craig and Maggiotto1982) might also nurture the expectation that political conversations offer rewards. We accordingly hypothesize:

The social dimension of everyday political talk

The other two categories acknowledge that political conversations are a specific mode of interpersonal communication and presuppose the participation of fellow citizens as communication partners. For a more complete understanding of people’s views about everyday political talk it is crucial to take into account that it implies interactions with other people. When discussing politics, individuals encounter each other not only in the political role of citizens but also as social beings. Alongside the cognitive dimension of the themes, issues, and questions that constitute conversations’ substance, they are inevitably also concerned with the affective dimension of interlocutors’ linkages to each other (Watzlawick et al., Reference Watzlawick, Beavin and Jackson2011: 29–52). Scholarship of interpersonal communication has long recognized that its meaning cannot be fully understood by focusing exclusively on its topical content. Drawing on seminal work by anthropologist Bronisław Malinowski (Reference Malinowski, Ogden and Richards1923) and linguist Roman Jakobson (Reference Jakobson and Sebeok1960) it has adopted the term ‘phatic’ to denote conversations’ distinctly social function of managing the bonds between their participants (Thornbury and Slade, Reference Thornbury and Slade2006: 19–20; Senft, Reference Senft, Senft, Östman and Verschueren2009).

When evaluating the prospect of engaging in everyday political talk, individuals can accordingly be expected to envisage this kind of behavior not only, and perhaps not even primarily, as a political activity, but also in terms of the social interactions that it necessarily entails (Rosenberg, Reference Rosenberg1954: 351). Studies of everyday political talk have not paid much attention to this dimension. What little research there is, however, suggests that certain social features of political talk might indeed be at odds with the instrumental political function assigned to it by deliberative democratic theory. In his seminal study of political apathy, Rosenberg already argued that many people may ‘be blocked from talking […] in behalf of their political beliefs out of fear of losing friends, alienating neighbors, endangering marriages, jeopardizing their positions in groups, losing business, jeopardizing their jobs, endangering production in their plants, facing community pressures, or exposing their feelings of self-esteem to threat’ (Rosenberg, Reference Rosenberg1954: 354).

Over the years, other (mostly qualitative) studies voiced similar concerns. From varying vantage points they converged on the diagnosis that discussing politics with one’s peers can be a quite stressful and adverse experience that many people find uncomfortable and prefer to avoid (Scheuch, Reference Scheuch, Scheuch and Wildenmann1965; Schudson, Reference Schudson1997; Eliasoph, Reference Eliasoph1998; Conover et al., Reference Conover, Searing and Crewe2002; Cramer, Reference Cramer2004; Mutz, Reference Mutz2006). Two aspects, in particular, have been highlighted by this literature: that political views may be considered too personal and private to be shared with others, and that discussing such matters may lead to unpleasant or even painful controversies. This suggests that attitudes toward everyday political talk can be expected to be affected by psychological dispositions of relevance for individuals’ interactions with other people. The fact that this behavior is an inherently social activity also implies that it inevitably takes place in specific social settings. The circumstances of these interpersonal contexts might likewise impinge on people’s attitudes toward this activity.

Interpersonally relevant psychological dispositions

Political psychology has in recent years made significant progress in revealing how political behavior is affected by features of citizens’ intra-individually highly stable, but inter-individually variable personalities (Caprara and Vecchione, Reference Caprara, Vecchione, Huddy, Sears and Levy2013). These personalities are conceived as multi-dimensional internal structures of an enduring nature that are to a large extent genetically rooted and thus heritable. They influence individuals’ behavior in all walks of life and therefore can be expected to leave an imprint also in the realm of politics (Mondak, Reference Mondak2010; Cawvey et al., Reference Cawvey, Hayes, Canache and Mondak2017). Research on political talk has thus far mainly referred to the comprehensive ‘Big Five’ framework of personality traits (Hibbing et al., Reference Hibbing, Ritchie and Anderson2011; Gerber et al., Reference Gerber, Huber, Doherty and Dowling2012). For the purpose of our study, we need a more fine-grained approach that specifically refers to personality facets of direct relevance for individuals’ relationships to other persons. Without claiming to be exhaustive, such interpersonally relevant psychological dispositions entail individuals’ need to belong, conflict orientations, subjective discussion competence, and social trust.

Need to belong is ‘an inherent, primary human need to form and maintain stable interpersonal bonds’ (Morey and Yamamoto, Reference Morey and Yamamoto2020: 83). It is a powerful human desire to experience enduring interactions of mutual concern with other persons (Baumeister, Reference Baumeister, Van Lange, Kruglanski and Higgins2011). For individuals with a high need to belong, the potentially controversial character of discussions about politics might appear threatening. A strong need to belong can therefore be expected to be associated with rather negative attitudes toward everyday political talk.

Individuals’ conflict orientation is ‘a specific type of approach/avoidance motivation; it is a stable personality trait centered around how people experience and react to conflict – whether they are excited by arguments, uncomfortable when others fight in public, or happy to handle a disagreement face to face’ (Sydnor, Reference Sydnor2019: 29). It is defined by the opposing poles of conflict aversion (dislike of political confrontation and argument) and conflict seeking (feelings of excitement and enjoyment about dispute and contention). Since everyday political talk may lead to conflictive experiences, conflict aversion can be expected to lead to unfavorable attitudes toward this behavior, whereas conflict seeking should stimulate positive views (Mutz, Reference Mutz2006; Neblo, Reference Neblo2015: 129–144).

Topically focused discussions in general, and political discussions in particular, demand certain skills, among them a basic understanding of the thematized subject matters and some measure of articulateness and eloquence. People differ in these skills, and in their confidence to command them. People endowed with high subjective competence to discuss politics (Rubin et al., Reference Rubin, Martin, Bruning and Powers1993) should consider everyday political talk an activity in which they can excel and that is accordingly rewarding. Those with little self-confidence, by contrast, face the ‘[t]hreat of ego-deflation’, associated with ‘the prospect of revealing factual ignorance or committing gross logical errors’, and consequently a desire ‘to avoid the feeling of defeat, abashment, humiliation, or other discomfiture’ (Rosenberg, Reference Rosenberg1954: 353). Accordingly, they should view political discussions rather negatively.

Trusting an actor involves one’s confidence that this actor will deliberately refrain from behaving in ways that violate one’s interests, even if the actor could do so without fearing sanctions. In the case of social trust, this actor is the anonymous aggregate of people outside one’s intimate core networks. This orientation has often been described as a lubricant that enables smooth cooperation between the members of societies (Nannestad, Reference Nannestad2008) – a beneficial function that can be assumed to extend to everyday political talk as well. If people expect fellow citizens to exploit opportunities to cause them harm, they will associate a high risk of humiliating or otherwise painful experiences with interactions that involve revealing personal opinions, such as one’s views about politics. Trusting persons, by contrast, should feel unhampered by such worries and accordingly develop more positive attitudes (Wyatt et al., Reference Wyatt, Katz and Kim2000).

The following formal hypothesis summarizes these expectations:

Interpersonal contexts

Ordinary citizens’ talk about public affairs typically takes place between the members of social networks (Huckfeldt and Sprague, Reference Huckfeldt and Sprague1995; Tanasoca, Reference Tanasoca2020). Since these networks define contextual circumstances for political conversations, their attributes should play an important role for attitudes toward this behavior. Two characteristics appear particularly important: the relationships between network members (Morey et al., Reference Morey, Eveland and Hutchens2012; Schmitt-Beck and Grill, Reference Schmitt-Beck and Grill2020), and the amount of disagreement individuals encounter when discussing politics with these persons (Nir, Reference Nir, Kenski and Jamieson2017). With regard to both, our expectations run counter to normative claims emphasized by theorists of deliberative democracy.

Concerning relationships between interlocutors, the distinction between strong ties and weak ties appears crucial. Strong ties are characterized by intimacy, mutual regard, and positive affect (Straits, Reference Straits1991). They circumscribe political talk of a private nature, involving close associates in protected spaces, such as people’s homes, and encompass two types of relationships: kinship bonds with spouses and other family members (Zuckerman et al., Reference Zuckerman, Dasovic and Fitzgerald2007), and friendships (Fischer, Reference Fischer1982). Weak ties connect people who know each other, although not necessarily well, and are outside of each other’s circles of intimates, such as co-workers, neighbors, and other acquaintances (Huckfeldt and Sprague, Reference Huckfeldt and Sprague1995). Since they involve interactions in more public settings like workplaces, bars or over the proverbial garden fence, they are often considered more valuable from a deliberative democratic point of view (Huckfeldt et al., Reference Huckfeldt, Johnson and Sprague2004; Tanasoca, Reference Tanasoca2020). However, in view of the challenging character of discussions about politics, strong ties can be expected to offer a more comfortable context (Schudson, Reference Schudson1997; Morey et al., Reference Morey, Eveland and Hutchens2012) which, consequently, should be associated with more favorable attitudes toward this activity.

Advocates of deliberative democracy praise discussions between those holding opposing views as the most constructive mode to address the pluralism of societal interests and value orientations (Thompson, Reference Thompson2008: 502; Sunstein, Reference Sunstein2003). Yet, deliberative democrats’ appreciation of disagreeable encounters seems to go against the grain of how ordinary people want to structure their social life. Many find disagreement a source of tension, with a potential to disrupt highly valued social relationships, and accordingly uncomfortable (Mutz, Reference Mutz2006). Studies indicate that citizens try to avoid political conversations when they anticipate controversies (Gerber et al., Reference Gerber, Huber, Doherty and Dowling2012; Settle and Carlson, Reference Settle and Carlson2019). This suggests that attitudes toward everyday political talk should be more negative when these conversations are expected to entail experiences of political disagreement. Concern about the ‘tyranny’ of majority opinion voiced by classics like de Tocqueville (Reference De Tocqueville1990: 254–286) and Mill (Reference Mill2015: 5–54), but also modern-day spiral-of-silence theory, suggests that individuals find encounters with disagreeable views particularly unpleasant when they are held by a majority (Noelle-Neumann, Reference Noelle-Neumann1974; Hayes et al., Reference Hayes, Shanahan and Glynn2001). Envisaging such experiences should, accordingly, lead to even more negative attitudes toward this behavior.

We accordingly hypothesize:

A web of conditioning relationships

Everyday political talk needs to be ‘safely integrated into the fabric of everyday lives and relationships’ (Mutz, Reference Mutz2006: 137). It therefore requires balancing political motives with the presumably more profound desire to maintain harmonious relationships with one’s associates (Pennington and Winfrey, Reference Pennington and Winfrey2021: 100). Interpersonal motives are closely associated with individuals’ fundamental longing for affiliation, social identification, self-esteem and well-being as accepted members of their groups and communities (Zayas and Sakman, Reference Zayas and Sakman2020). They are therefore constantly activated as people lead their daily lives. Political dispositions, by contrast, pertain only to one specific domain of life that is, moreover, fairly remote from people’s lifeworld, accordingly more abstract, and often not very salient (van Deth, Reference Van Deth2000). Individuals can therefore be assumed to function as social beings first, and as zooi politikoi only second. As noted by Mutz, ‘[p]eople tend to care more about social harmony in their immediate face-to-face personal relationships than about the larger political world’ (Mutz, Reference Mutz2006: 106). This general perspective bears two implications for our research. First of all, it suggests that social factors exert an overall stronger impact on persons’ attitudes toward everyday political talk than their political dispositions. We accordingly hypothesize (not visualized in Figure 1):

In addition, over and above this merely additive specification in terms of competing direct effects, the general assumption that interpersonal factors are more powerful than political ones leads us to expect that they affect attitudes toward everyday political talk also indirectly by conditioning the influence of political dispositions. Mondak et al. (Reference Mondak, Hibbing, Canache, Seligson and Anderson2010), among others, have argued that the impact of distal personality features on political orientations should indeed be seen as primarily interactive. Supporting this claim, a small body of research on various modes of political communication has demonstrated such moderating effects of conflict orientations (Mutz, Reference Mutz2006; Dalisay, Reference Dalisay2012; Testa et al., Reference Testa, Hibbing and Ritchie2014; Sydnor, Reference Sydnor2019; Wolak, Reference Wolak2022), and the need to belong (Eck et al., Reference Eck, Schoel and Greifeneder2017). We accordingly hypothesize that individuals’ psychological dispositions condition the effects of political dispositions on attitudes toward everyday political talk. If persons value social rewards higher than political gratifications, then strong political interest, intense directional passions, and marked internal efficacy should most clearly nurture the expectation that everyday political talk is associated with rewarding experiences when their need to belong is weak, when they enjoy rather than abhor conflict, when they feel competent to stand their ground in discussions, and when they trust their fellow citizens strongly. By contrast, under conditions of a strong need to belong, marked conflict avoidance, little subjective discussion competence, and weak social trust, the dreaded social costs of such conversations should suppress the beneficial effects of political dispositions. Put formally:

The same line of thought leads us to expect that the impact of dispositions is moderated by the contextual circumstances of everyday political talk as ultimate exogeneous factors. Mondak et al. (Reference Mondak, Hibbing, Canache, Seligson and Anderson2010; see also Mondak, Reference Mondak2010; Cawvey et al., Reference Cawvey, Hayes, Canache and Mondak2017) have made a strong case that effects of psychological dispositions should be conceived as conditional on the situational settings of behaviors. To the extent that features of contexts resonate with personality facets, individuals differently endowed with these features can be expected to respond in dissimilar ways to identical social situations. In other words, variable contextual conditions can be expected to moderate and thereby cause heterogeneity in the effects of psychological dispositions. They ‘contribute to patterns of political behavior, but the expression of these effects will often be contingent on the situation’ (Mondak et al., Reference Mondak, Hibbing, Canache, Seligson and Anderson2010: 85). It seems highly plausible to expect similar situational contingencies also for political dispositions.

Importantly, in bearing with our previous hypotheses, contextually induced heterogeneity should work in opposite directions for psychological and political dispositions. We assume that political dispositions affect attitudes toward everyday political talk more strongly when interpersonal settings are favorable (analogous to H2). Conversely, the impact of political dispositions should be muted under more adverse contextual circumstances. The contrary pattern should emerge for psychological dispositions. Their impact should be boosted, especially when they resonate with these contextual features. Conflict orientations, for instance, should become more influential when conversations are envisaged as disagreeable (Mutz, Reference Mutz2006; Sydnor, Reference Sydnor2019). Thinking further along these lines also suggests an even more complex three-way interaction. It implies that the contextual circumstances affect not only the direct effects of psychological and political dispositions, but also the conditioning impact of the former on the effects of the latter.

Dependent variable and methodological approach

As indicated above, respondents’ assessments of everyday political talk were recorded for nine different interpersonal contextual circumstances, resulting from a cross-classification of three types of relationships between conversation partners and three kinds of disagreement experiences. To turn these contextual conditions into variables whose impact on attitudes toward everyday political talk can be estimated in conjunction with respondents’ political and psychological dispositions, the survey dataset was reshaped into a long format. It was organized in such a way that the information contained in the original questions was shifted into a generic variable indicating respondents’ like or dislike of political talk and two variables that specify relationships between interlocutors, respectively, the amount of disagreement. The dataset is stacked by disagreement nested within relationships nested within respondents, resulting in nine observations per respondent. In the linear ordinary least squares regression modelsFootnote ² presented below, the generic like-dislike variable serves as the dependent variable, whereas the two indicators of contextual circumstances are used as independent variables.

Independent and control variables

Political dispositions: Political interest is measured by means of self-reports on a five-point scale. The indicator of partisanship takes the form of a five-point scale ranging from non-partisans to very strong party identifiers. Ideological extremity is measured by means of an 11-point left-right scale folded at the midpoint. Internal political efficacy is measured by an additive scale based on two items.

Psychological dispositions: Subjective discussion competence and need to belong are indicated by additive scales based on two items each. The latter’s coding is reversed because a low need to belong should be associated with positive attitudes toward everyday political talk. Six items were used to construct an additive scale of conflict orientations, with low values indicating conflict aversion and high values conflict seeking. For measuring social trust, an additive scale was used that combined items pertaining to respondents’ trust in acquaintances and strangers.

Interpersonal contexts: To indicate the strength of social ties a dummy variable is used that distinguishes weak ties (1 = acquaintances) from strong ties (0 = family members and friends). The amount of disagreement is indicated by two dummy variables. The first pertains to disagreement as such, the second to majority disagreement. They are coded in such a way that the estimate for the former expresses the contrast with the base condition without disagreement prime, whereas majority disagreement is contrasted with simple disagreement.

All models include the following socio-demographic controls: gender, age, immigration background, education, occupational status and economic well-being. In addition, the models control for the psychological traits need to evaluate, need for cognition, and need for cognitive closure (see Online Appendix for details of instrumentation).

Findings

The findings for model (1) in Table 1 are mostly in line with the basic hypotheses H1a, H1b, and H1c. H1a receives only partial support since political interest is the only political disposition that makes a difference in attitudes toward everyday political talk. This suggests that highly interested individuals find political talk much more rewarding than those with little or no interest, whereas partisan or ideological passions and internal efficacy have no additional impact. Therefore, these orientations will no longer be considered in the following analyses.Footnote ³ The other findings are unequivocally in line with H1b and H1c. Individuals’ conflict orientations are the overall strongest predictor in the model. Conflict seeking is associated with much more positive attitudes toward everyday political talk than conflict avoidance. The effects of social trust, discussion competence and need to belong are substantial, but weaker. The impact of interpersonal contexts is overall somewhat less pronounced, but all effects are likewise in line with our expectations. Political conversations within strong ties are viewed more positively than those within weak ties. The prospect of disagreement renders attitudes toward everyday political talk less favorable, and the most challenging scenario of having to face a disagreeable majority gives an additional push toward more negative attitudes.

Table 1. Direct effects of political dispositions, psychological dispositions, and interpersonal contexts on attitudes toward everyday political talk (unstandardized regression coefficients)

Robust standard errors (clustered by respondents) in parentheses. All variables are normalized to range 0 to 1.

^# p < 0.10, *p < 0.05, **p < 0.01, ***p < 0.001.

Comparing the incremental model fit statistics of models (3) to (6) (relative to the baseline model (2)) allows assessment of H1d. Compared to the purely political model (3), the incremental explanatory power of the complete social model (6), which entails both interpersonally relevant psychological dispositions and interpersonal contexts, is more than twice as large. The interpersonal predictors thus outweigh the political ones by a large margin. Most of this model’s superior fit is due to the psychological dispositions, however (cf. model (4)), whereas the overall impact of contextual circumstances alone (model (5)) is smaller than the explanatory power of the block of political dispositions. H1d thus receives qualified support. The analysis suggests that interpersonally relevant psychological dispositions are more important for attitudes toward everyday political talk than political dispositions, whereas contextual circumstances play a more limited role – at least in a purely additive perspective.

In addition to these direct effects, we expect social factors to affect attitudes toward everyday political talk also conditionally in a complex interplay of moderating relationships with political dispositions. To test these hypotheses for political interest (as the only political disposition of relevance) we add multiplicative interaction terms to the full model (1) displayed in Table 1. We pursue a two-track strategy that combines a more permissive with an extremely conservative approach to modeling. The first approach consists of a series of models in which each interaction effect implied in the generic conditional hypotheses is estimated separately. In the restrictive approach, all proposed interactions are examined simultaneously in a combined model. Since even simple two-way interaction models include the predictors of interest twice in the same equation they are by definition affected with considerable multicollinearity (Brambor et al., Reference Brambor, Clark and Golder2006). This erects high hurdles for achieving conventional levels of statistical significance already in the permissive models. Including key predictors multiple times, as is necessarily the case in the combined models, raises the bar even higher.

According to our first conditional hypothesis (H2), the positive effect of political interest should to some extent depend on favorable psychological dispositions. All interaction effects displayed in Figure 3 are in line with this expectation, but most of them fail to attain statistical significance. In the separate models, conflict seeking and high confidence in one’s ability to discuss politics are associated with a stronger impact of political interest, as expected. High political interest thus appears to induce more favorable attitudes toward everyday political talk primarily among individuals that are confident in their ability to stand their ground in such discussions and do not shy away from political conflicts. But these effects are not robust when subjected to the tough test of the restrictive combined model.

Figure 3. Moderation of effect of political interest by psychological dispositions.

Note: Entries are two-way interaction effects from multiple linear regression models with robust standard errors (tails indicate 95-% and 90-% confidence intervals). Empty symbols are based on permissive models, filled symbols on restrictive models. See Online Appendix for complete models.

The second set of conditional hypotheses claims that the impact of political and psychological dispositions on citizens’ views of everyday political talk is in opposite ways conditioned by the interpersonal contexts within which this activity takes place. The estimates of the interaction terms displayed in Figure 4 show for political interest and the four psychological dispositions how these associations are amplified or mitigated by weak ties in comparison to strong ties, as well as by disagreement in contrast to no disagreement, and by majority disagreement compared to simple disagreement. H3a predicts that weak ties, as well as disagreement of growing intensity, depress the positive effect of political interest. This expectation is largely supported by our data. Attitudes toward everyday political talk are considerably less responsive to differences in citizens’ political interest when the envisaged discussions take place within weak ties or entail disagreement. These patterns emerge not only in the separate models but very clearly also in the combined models. However, it makes no further difference whether disagreeable views are expressed by a majority of one’s interlocutors.

Figure 4. Moderation of effects of political interest and psychological dispositions by interpersonal contexts.

Note: Entries are two-way interaction effects from multiple linear regression models with robust standard errors (tails indicate 95-% and 90-% confidence intervals). Empty symbols are based on permissive models, filled symbols on restrictive models. See Online Appendix for complete models.

H3b posits the opposite pattern for psychological dispositions. Across the board, they should be more relevant under the more challenging conditions of weak ties and disagreeable interlocutors. With some qualifications, this expectation is also sustained. Social trust, for instance, is much more strongly associated with attitudes toward everyday political talk within weak ties. Disagreement, by contrast, does not appear to condition the role of social trust. For the need to belong, we also see interaction effects in the expected direction – less pronounced with regard to the strength of ties, but unlike social trust also pertaining to disagreement (but not majority disagreement). Concerning conflict orientations and discussion competence, the picture is also not uniform. The separate models suggest statistically meaningful conditioning effects of the strength of ties in the wrong direction. But they are not robust to the more restrictive model specification. Moderating effects in the hypothesized direction emerge for conversations that entail disagreement. Majority disagreement appears to boost these effects further, but only under permissive model specifications. Figure 3 thus suggests that the role of social trust is only conditioned by the strength of ties, whereas conflict orientations and discussion competence are primarily important when disagreement is envisaged to occur in conversations. Need to belong has weaker conditioning effects under both contextual conditions. What does not appear to matter much is the difference between the simple presence of disagreement and disagreement expressed by a majority.

The final analysis combines the perspectives of the previous ones in three-way interaction models. This erects even higher hurdles for interaction terms to achieve statistical significance. H4 posits that the deflating role of unfavorable psychological dispositions on the effect of political interest is intensified by adverse contextual circumstances, that is, for conversations within weak ties or in the presence of disagreement, especially majority disagreement. When all hypothesized three-way interactions are modeled simultaneously, none of them is statistically significant (Figure 5). But this finding should not be overrated, given the enormous complexity of these models. The separate models, by contrast, find the effect of political interest on attitudes toward everyday political talk indeed being significantly more sensitive to conflict orientations within weak ties compared to strong ties. The same pattern is detected for discussion competence. Among the conflict averse and those with little confidence in their ability to persist in discussions, strong political interest translates to a lesser extent into a positive view of everyday political talk when it is envisaged to take place outside strong ties. Regarding the moderating role of discussion competence, we also see effects in the expected direction conditional on disagreement as well as majority disagreement, but they do not attain sufficient levels of statistical significance.

Figure 5. Moderation of interactions between political interest and psychological dispositions by interpersonal contexts.

Note: Entries are three-way interaction effects from multiple linear   regression models with robust standard errors (tails indicate 95-% and 90-% confidence intervals). Empty symbols are based on permissive models, filled symbols on restrictive models. See Online Appendix for complete models.