THEORETICAL ASSUMPTIONS ABOUT THE CONSEQUENCES OF CROSS-CUTTING EXPOSURE

Perhaps the most often cited proponent of communication across lines of difference is John Stuart Mill (1859, 21), who pointed out how a lack of contact with oppositional viewpoints diminishes the prospects for a public sphere: “If the opinion is right, they are deprived of the opportunity of exchanging error for truth; if wrong, they lose what is almost as great a benefit, the clearer perception and livelier impression of truth produced by its collision with error.” Likewise, Habermas (1989) assumes that exposure to dissimilar views will benefit the inhabitants of a public sphere by encouraging greater interpersonal deliberation and intrapersonal reflection.

Exposure to conflicting political views is also said to play an integral role in encouraging “enlarged mentality,” that is, the capacity to form an opinion “by considering a given issue from different viewpoints, by making present to my mind the standpoints of those who are absent.… The more people's standpoints I have present in my mind while I am pondering a given issue, and the better I can imagine how I would feel and think if I were in their place, the stronger will be my capacity for representative thinking and the more valid my final conclusions, my opinion” (Arendt 1968, 241). Interactions with others of differing views is assumed to be “essential for us to comprehend and to come to appreciate the perspective of others” (Benhabib 1992, 140).

Cross-cutting exposure also is assumed to promote greater awareness of oppositional views because no individual person thinking in isolation can foresee the variety of perspectives through which political issues may be perceived (Manin 1987). Thus political deliberation “teaches citizens to see things they had previously overlooked, including the views of others …” (p. 351). Awareness of rationales for oppositional views is a particularly important type of political knowledge because of its close ties to legitimacy. One purpose served by conveying rationales for oppositional views is to help render the ultimate decision or policy legitimate in the eyes of others (Manin 1987). If rationales are not made public, the losers in a given controversy will not know what reasons or arguments the winners judged to be stronger in deciding the merits of the case: “Hence discussion rather than private deliberation would be necessary to ‘put on the table’ the various reasons and arguments that different individuals had in mind, and thus to ensure that no one could see the end result as arbitrary rather than reasonable and justifiable, even if not what he or she happened to see as most justifiable” (Fearon 1998, 62).

Exposure to diverse political views is obviously tied to a wide range of outcomes that are valued in democratic systems. But it would be quite naive to suggest that only good can flow from cross-cutting interactions; conversations among those of differing views also have the capacity to result in bitter arguments, violence, and/or a hostile and uneasy silence (Scorza 1998). Thus Kingwell (1995) has stressed the importance of civility or politeness in maintaining conversations across lines of political difference. To sustain relationships that make cross-cutting discourse possible, discussants must at times refrain from saying all they could say in the interests of smooth social interaction. This type of civility via “not-saying” … “contributes to smooth social interaction, makes for tolerance of diversity and conditions a regard for the claims and interests of others” (Kingwell 1995, 219). In this view, exposure to differing views holds the potential for tremendous benefits, but only if it occurs in a context in which the collective project of getting along with one another in society is primary and the elucidation of differences secondary.

THE ROLE OF CROSS-CUTTING EXPOSURE IN EMPIRICAL RESEARCH

In addition to the assumptions of political theory, many empirical relationships have been credited to exposure to dissonant views. For example, in his classic study of tolerance, Stouffer (1955, 127) suggested that exposure to conflicting views was the main reason that education and tolerance were so closely connected:

Interpretations of tolerance levels have stressed the diversity of people's contacts, but this concept has seldom been directly measured. Nonetheless, closely related concepts support the likelihood of such an impact. For example, a personality dimension known as “openness to experience” is strongly positively related to tolerance (Marcus et al. 1995), and authoritarians have been found to live relatively sheltered lives with little exposure to alternative lifestyles and beliefs (Altemeyer 1997). In a recent study of Russian social networks, Gibson (1999) also found that support for democratic institutions was correlated with the number of “weak ties” (i.e., nonrelatives) in a person's social network.

Finally, the popularity of specially organized deliberative forums also rests on the assumption that cross-cutting exposure is particularly beneficial. Although such events may successfully educate people on public policy issues (for a review, see Gastil and Dillard 1999; Cook et al. 1999), it is unclear from research to date whether learning is influenced by the cross-cutting interactions themselves or some other aspect of the forums such as the educational materials given to participants by organizers. Americans report that they are very unlikely to talk about politics at public meetings (Conover and Searing 1998) so the generalizability of findings from specially orchestrated forums to everyday political life is also an open question.²

The most important point to be gleaned from analyzing the role that the political diversity of social networks has played in political science research to date is that outside of work on persuasion, cross-cutting exposure typically has been an unmeasured concept, one offered in post hoc explanations for other relationships rather than as the central focus of research. Despite the important role that cross-cutting exposure plays in political theory, little empirical research has examined its consequences.

THE SOCIAL PSYCHOLOGICAL PLAUSIBILITY OF BENEFICIAL CONSEQUENCES

Embedded in these assertions are hypotheses about at least three potential beneficial effects of cross-cutting networks. Communication environments that expose people to non-like-minded political views have been assumed to promote (1) greater awareness of rationales for one's own viewpoints, (2) greater awareness of rationales for oppositional viewpoints, and (3) greater tolerance. How plausible are such benefits from the perspective of what is known about the social psychology of human interaction?

The first hypothesis rests on the assumption that confronting differences prompts people to reflect on the reasons for their own beliefs. This process is assumed to occur either in preparation for defending one's own positions or as a result of an internal need to rationalize or explain why one's own views differ from others'. Studies of cognitive response generally support the plausibility of such a reaction; exposure to counter-attitudinal advocacy enhances the production of counter-arguments, particularly for highly involving topics (e.g., Petty and Cacioppo 1977, 1979).³

The second hypothesis, that cross-cutting exposure promotes greater awareness of oppositional viewpoints, simply assumes that exposure to dissimilar views imparts new information. Psychologically this hypothesis demands nothing more than a straightforward learning process whereby rationales are transmitted from one person to another. The greatest limitation on its plausibility is the infrequency with which political conversations are likely to reach the level of depth in which rationales are articulated. But a good deal of this process may occur at the intrapersonal rather than the interpersonal level. In other words, when exposed interpersonally to political views noticeably different from their own, people may be prompted to think intrapersonally about the reasons that may have led those others to hold such views (Mutz 1998). This mental rehearsal of thoughts and search for rationales may occur even when the discussants do not explicitly articulate such reasons themselves (Burnstein and Sentis 1981; Burnstein, Vinokur, and Trope 1973).

The third hypothesis embedded in arguments about the importance of cross-cutting exposure is that it should lead to greater political tolerance. On initial consideration, this assertion sounds very similar to Allport's (1954) classic intergroup contact hypothesis, which suggests that face-to-face interaction among members of different groups can, under certain conditions, reduce prejudice. Although the contact hypothesis has been known for producing mixed evidence at best (see, e.g., Amir 1969, 1976), more recent assessments suggest that intergroup contact usually does have positive effects, even when the situation does not meet all of the conditions enumerated by Allport and subsequent researchers (see Pettigrew 1998, Pettigrew and Tropp 2000).⁴

On the other hand, the kind of people or groups one is asked to tolerate on civil libertarian grounds are seldom exactly the same as the people of opposing views that one encounters at work or in the neighborhood. Nonetheless, within the large literature on intergroup contact, a smaller group of studies of “generalized intergroup contact” confirms that contact across group lines can generalize to reduce prejudice even toward outgroups that are not part of the intergroup contact (e.g., Reich and Purbhoo 1975; Cook 1984; Pettigrew 1997; Weigert 1976). In other words, these findings support the possibility that exposure to everyday differences of political opinion may translate to an appreciation of the need to tolerate differences of political opinion among disparate groups within the larger society. People who have had to learn how to “agree to disagree” in their daily lives better understand the need to do so as a matter of public policy. For example, in support of the generalizability of contact effects, the extent of interpersonal contact across lines of religion, race, social class, culture, and nationality predicted nonprejudicial attitudes toward groups not involved in the contact, even when taking into account potential reciprocal influences (Pettigrew 1997). Moreover, the extent of contact across lines of difference also generalized to immigration policy preferences, a more policy-oriented outcome similar to tolerance measures. Although studies of intergroup contact have tended to use prejudicial attitudes as their dependent variables, their findings also appear to generalize to perspective-taking ability (see Reich and Purbhoo 1975); that is, cross-cutting contact improves people's abilities to see issues from the perspectives of others, even when they personally do not agree.⁵

As illustrated in Fig. 1, there are at least two mechanisms by which exposure to oppositional political viewpoints might lead to political tolerance.⁶

In addition to this cognitive mechanism for translating cross-cutting exposure to political tolerance, a second potential mechanism emphasizes affect over cognition. To paraphrase Stouffer (1955), one could learn from personal experience that those different from one's self are not necessarily bad people. According to this mechanism, the content and extent of people's political discussions are less important than the quality of the personal relationships that develop. It is not important that they learn about the rationales for one another's political views, but it is important that they develop close relationships with those they know to hold quite different political viewpoints. Once formed, these cross-cutting relationships make it less likely that people will support restrictions of the civil liberties of those of differing views. The bottom half of Fig. 1 illustrates how exposure to people of differing political views may lead to more intimate cross-cutting associations, and thus greater tolerance.

To summarize, interactions involving exposure to conflicting views have been assumed to benefit people largely (1) by encouraging a deeper understanding of one's own viewpoint, (2) by producing greater awareness of rationales for opposing views, and (3) by contributing to greater tolerance. With respect to the third hypothesis, if exposure to cross-cutting political views increases tolerance via its effects on awareness of rationales for oppositional points of view, then this would lend credibility to the cognitive interpretation of the benefits of cross-cutting contact. If close personal relationships across lines of political difference influence tolerance levels, then this provides support for the affective mechanism.

SURVEY DESIGN

To examine these hypotheses, I utilized data from a representative national telephone survey sponsored by the Spencer Foundation and executed by the University of Wisconsin Survey Center in the fall of 1996, immediately preceding the presidential election in November.⁷

Although the impact of discussions with like-minded others is not the central focus of this study, I include this variable in the analyses to sort out effects that may be attributed to political discussion in general, as opposed to discussions that cross lines of political difference in particular. Moreover, since political discussion of all types is likely to characterize those more politically interested, knowledgeable, and involved, controls are included for these predispositions. To the extent that the effects of exposure to dissonant views are unique and not attributable to contact that involves political agreement or to political interest and involvement more generally, then the benefits suggested by so many theorists gain support. In addition to the frequency of consonant and dissonant contact, the survey also included items that made it possible to tap the level of intimacy within dyads (see Appendix A). For each respondent, separate measures were created to represent levels of intimacy with consonant, dissonant and politically neutral discussion partners.⁹

To tap the dependent variables for the first and second hypotheses, awareness of legitimate rationales for own and for opposing views, open-ended questions were used to solicit issue-specific rationales for three separate controversies including preferences among the 1996 presidential candidates, opinions about affirmative action for women and minorities, and opinions about state versus federal control of the welfare system.¹⁰

As expected, the number of rationales that people could give for their own positions were, on average, significantly higher than those they could give for why someone might hold an opposing view (p < 0.001 in all three cases). As shown in Table 1, the measures of rationales for the two sides of a given issue were, also not surprisingly, significantly correlated with one another, thus indicating general knowledge of or interest in politics or in these specific issues. Three issues were used to get a broader sense of a given person's knowledge of dissimilar viewpoints than one issue alone would make possible and were then combined into two additive indices representing a person's overall awareness of rationales for oppositional views and overall awareness of rationales for their own viewpoints.

Political tolerance was measured using Sullivan Piereson and Marcus's (1982) content-controlled method whereby respondents first volunteer their “least-liked” group and are then asked a series of six questions about extending civil liberties to these particular groups, including the extent to which they should be banned or outlawed, be allowed to hold rallies in their city, be allowed to teach in public schools, and be subject to government phone tapping.¹³

EFFECTS ON AWARENESS OF RATIONALES FOR OWN AND OPPOSITIONAL VIEWS

The first hope of advocates of greater network diversity is that exposure to conflicting views will benefit citizens either by familiarizing them with legitimate reasons for holding opposing viewpoints or by deepening their understanding of their own views by having to defend them to others and/or to themselves. The first 2 columns of data in Table 2 show regression equations examining these two questions,¹⁴

For purposes of evaluating the first two hypotheses, what is important in Table 2 is the coefficient corresponding to exposure to dissonant political views. As shown in column 1, counter to the first hypothesis, exposure to dissonant views does not appear to have a significant impact on awareness of rationales for people's own political views. Even when examined among the most likely subgroups within the population (such as those with strongly held views or high levels of education), there is no evidence that those with cross-cutting political networks have more rationales in mind in support of their viewpoints.

However, consistent with the second hypothesis, column 2 shows that exposure to oppositional viewpoints increases awareness of legitimate rationales for opposing views. The highly significant positive coefficient supports the hypothesis that exposure to oppositional viewpoints is particularly important for purposes of familiarizing people with legitimate reasons for viewpoints that differ from their own. Nonetheless, to make a solid case for this hypothesis, it is essential to determine that it is exposure to dissonant views, and not just political discussion in general, that is driving awareness of rationales for oppositional views. Comparing the coefficients for exposure to consonant and dissonant views lends additional support to this hypothesis because the consonant coefficient is negative and significantly different from the coefficient for exposure to dissonant views.

With cross-sectional data how confident can one be that exposure to conflicting political views actually brings about greater awareness of rationales for opposing views? Because a person knows a lot about politics, he/she may be more confident of defending his/her own views, and thus be more willing to engage in cross-cutting interactions (see Conover and Searing 1998). I have attempted to rule out the most obvious spurious influences by including controls for general political knowledge, interest, and awareness specific to these issues, but reverse causation remains a possibility. The problem with this rival interpretation is that, if true, it ought to apply equally well, if not more so, to knowledge of rationales for one's own views; the more deeply committed one is to his/her position, and the more rationales in one's arsenal, the less threatened one should be by oppositional viewpoints. Table 2 shows that awareness of rationales for one's own views is not related to exposure to conflicting views in one's personal network. Although this represents a null finding with respect to the first hypothesis, ultimately this pattern strengthens the case for the idea that exposure to conflicting views contributes to greater awareness of legitimate rationales for opposing views.

How large is the effect of exposure to dissimilar views on awareness of legitimate rationales for oppositional views? To provide a general idea, I calculated the predicted values of awareness of legitimate rationales for oppositional views for those with the highest and lowest levels of exposure to dissonant political views. On average, those with high levels of exposure to dissonant views should be expected to be familiar with just over two additional rationales compared to a similar person with a homogeneous network. Given that the mean number of oppositional rationales that people in this sample could generate was 1.46, the increase due to cross-cutting networks could have quite significant consequences for the perceived legitimacy of political outcomes.

Finally, in the third column in Table 2, I test Kingwell's suggestion about the importance of civility in generating beneficial outcomes from cross-cutting political dialogue. Drawing on a scale widely used to classify patterns of communication within families, I operationalized the “civil” orientation to conflict as one that combines acknowledgment of the importance of expressing dissenting views with an emphasis on social harmony.¹⁶

As shown in the third column in Table 2, people with a civil orientation toward conflict are particularly likely to benefit from exposure to dissonant views. Although the inclusion of an interaction between Orientation to conflict and exposure to dissonant views slightly reduces the size of the original coefficient corresponding to exposure to dissonant views, the interaction significantly strengthens the overall model (p < 0.05), thus indicating that those who value both frank opinion expression and social harmony learn the most from their cross-cutting interactions. The size of the effect among members of this group is more than twice the size of the effect on the population as a whole.

To summarize, exposure to cross-cutting views does not appear to play a significant role in deepening people's knowledge of their own issue positions, but it does have an important impact by familiarizing them with legitimate rationales for opposing viewpoints. This impact is particularly pronounced among people who value the expression of dissenting opinions, but simultaneously care about social harmony; in other words, those who would engage in cross-cutting conversations but who would remain silent rather than risk conflict that might end a cross-cutting association altogether.

CONSEQUENCES OF POLITICAL TOLERANCE

The first column in Table 3 takes a first pass at evaluating Fig. 1's proposed mechanisms for translating exposure to dissonant views into political tolerance. If the affective ties between people are what is important for purposes of translating cross-cutting ties into political tolerance, then one would expect closeness within dissonant relationships to be particularly important to political tolerance. If the cognitive benefits are primary, then people's awareness of rationales for those views should be most predictive of political tolerance.

Consistent with the expectations in Fig. 1, exposure to dissonant views has no direct effects on political tolerance, but awareness of rationales for oppositional views and intimacy within dissonant dyads are both significantly related to tolerance. Closer relationships across lines of difference and greater knowledge of rationales for these differences both predict tolerance, even after controlling for political knowledge, political interest, extremity of issue opinions, and so forth. Notice, in contrast, that awareness of rationales for own views does not contribute significantly to political tolerance. Despite the fact that the two measures are highly correlated (see Table 1), awareness of rationales for own and opposing views represent distinctly different types of knowledge with very different consequences.

The problem with a causal interpretation of these relationships is that tolerance may have reciprocal effects on these variables. More tolerant people may be more likely to form close relationships with those of differing political views and/or may be more likely, as a result of their tolerance, to be open to learning about reasons for others' views. Although I do not carry out an analysis of the full simultaneous system, 2SLS provides a useful means of obtaining less biased estimates of the strength of the key causal relationships examined here. Fortunately, there are variables available that predict awareness of rationales for oppositional views and that predict Intimacy within dissonant dyads but that do not predict tolerance, thus making them ideal as instrumental variables in a 2SLS analysis. For example, as shown in Table 2, awareness of rationales for oppositional views is predicted by exposure to dissonant views and by awareness of rationales for own views, but they are unrelated to tolerance. Likewise, intimacy within dissonant dyads is predicted by exposure to dissonant views, exposure to consonant views, and Number of discussants in the network, but none of these variables is significantly related to tolerance judgments (see Appendix B).¹⁷

In the second column in Table 3, the endogenous variables produced by the first-stage estimations are used in a 2SLS model treating both intimacy within dissonant dyads and awareness of rationales for opposing views as endogenous. As shown in the second column, the coefficients are virtually identical when taking potential reciprocal causation into account, thus lending additional support to the proposed pathways in Fig. 1. Most interestingly, they confirm that both cognitive and affective mechanisms are at work simultaneously in translating exposure to dissonant views to greater political tolerance. If one generally perceives those opposed to one's own views to have some legitimate, if not compelling, reasons for doing so, then one also will be more likely to extend to disliked groups the rights of speech, assembly, and so forth. Likewise, close ties with those who hold differing political views can increase tolerance. It should also be noted that this effect is not a mere function of attitude extremity or general or issue-specific forms of political knowledge, as variables of this kind are already included in the model.

Given that these findings only indirectly relate exposure to dissonant political views to political tolerance, what is the size of the net impact of cross-cutting exposure in the network on political tolerance? Using the coefficients in Tables 2 and 3 as a basis for estimating the size of the impact of the cognitive mechanism suggests that, if all else remained constant, a person at the high end of the exposure to dissonant views index would score just over 4% higher on the tolerance scale than someone with the lowest levels of exposure to dissonant views.¹⁹

From a technical standpoint, there are two ways one might view the strength of these findings. On the one hand, two of the three hypotheses have sustained extensive controls for plausible rival interpretations, the implementation of instrumental variables to take into account simultaneity bias, and a relatively small sample size. On the other hand, the effects that emerge are modest, though they are clearly statistically significant and thus lend some credibility to the many claims of democratic theorists about the benefits of cross-cutting exposure.

The relationships may be viewed in a more impressive light if one takes into consideration the crudeness of some of the operational measures relative to the concepts they represent. For example, ideally one would like a measure of awareness of legitimate rationales for oppositional views that takes into account all potential controversies. Instead this larger concept is represented in this study by only three political controversies. Likewise, the measure of the extent to which people's networks involve cross-cutting exposure has been limited by constraining respondents to only three discussants, when a more extensive network battery might generate a more valid measure including a greater number of weak ties, people with whom politics is discussed very seldom but who are politically dissimilar to the main respondent. The type of contact I examine in this study is by its very nature infrequent and often fleeting and, thus, difficult to measure. In addition, if better first-stage predictors of the endogenous variables were available, then a stronger pattern of relationships might be visible.

From a theoretical perspective, it is also worth noting that I do not make stringent assumptions about the kind of exposure to cross-cutting views that is tapped by these items. When exposed to conflicting views, I do not assume that people are truly “deliberating” according to any particular theoretical definition, nor is it assumed that when people are exposed to conflicting views the context is one in which people have equal status, reciprocity, and so forth. In this study exposure to dissonant political views requires only that people talk politics with someone who has political views that are, to some recognizable degree, different from their own (and vice versa for exposure to consonant views). Even though this is a far cry from what theorists and others envision as ultimately the most beneficial, exposure to conflicting views—even at the level defined here—appears to have the capacity to produce some beneficial effects. In short, there is undoubtedly a great deal of noise in these measures, and this needs to be taken into account in evaluating the more general theories these relationships represent. Although replication of these analyses on other data sets would be a natural next step to increase confidence in these findings, unfortunately there are few, if any, additional national surveys addressing the constitution of Americans' political networks that also measure political tolerance.²⁰

AN EXPERIMENTAL CONFIRMATION

Recognizing that statistical techniques can only go so far in strengthening causal inferences in survey data, I subjected part of this model to an experimental test. Ideally, one could test all of the relationships shown in Fig. 1 in a controlled laboratory environment, manipulating exposure to cross-cutting political views and observing the consequences. However, for the bottom half of Fig. 1, that is, the affective mechanism translating cross-cutting exposure into political tolerance, an experimental design is not feasible. At least within the context of a short-term laboratory experiment, one cannot forge cross-cutting friendships and evaluate the effects of their intimacy.

Nonetheless, the cognitive mechanism shown in the top half of Fig. 1 is amenable to experimentation. Although it is not possible to simulate the effects of ongoing, accumulated exposure to cross-cutting political viewpoints in a lab, even a large, one-time exposure to the rationales behind multiple views different from one's own could be adequate. If one learns from such an experience that those with views different from one's own have their reasons, despite the fact that one may disagree with them, such exposure should promote support for the general principle of tolerance.

Using a simple, three-group design in which 82 student subjects²¹

After the pretest, each subject was exposed to rationales for dissonant or consonant views or to nothing at all. Because the cognitive mechanism in Fig. 1 hypothesizes a purely informational effect from cross-cutting exposure on tolerance, the manipulation was limited to simply conveying information about arguments behind oppositional positions, without any face to face contact with another human being. Further, because this hypothesis is about the effects of generalized exposure to contradictory views on tolerance, and not about effects from exposure to any one topic or area of controversy, each subject was exposed either to multiple rationales for multiple political viewpoints that matched their own, rationales for views they were known to oppose, or no new information. To strengthen the manipulation, exposure to consonant or dissonant views was carried out by exposing respondents to consistently agreeable or disagreeable arguments for three separate issues. A randomized schedule dictated for which of eight pretest issue controversies each subject received stimuli and in which order.²⁴

In the dissonant and consonant exposure conditions, three brief “assignments” provided a context for exposing people to rationales in support of, or in opposition to, their own views. For each assignment, subjects were given a stack of five cards, each bearing a rationale in support of an issue position.²⁵

The raw mean comparisons resulting from this experiment were in the expected direction with lower tolerance in the control condition (

) relative to the dissonant views condition (

), but these differences were not statistically significant among any of the three groups. However, when the efficiency of the model was improved by including covariates and taking into account subjects' perspective-taking ability and dogmatism, significant effects on tolerance were evident among those respondents with high perspective-taking ability. As predicted, perspective-taking ability was directly related to political tolerance, but it also served as an important contingent condition for the effects of cross-cutting exposure. In contrast, analyses revealed no significant differences or interactions in analyses comparing the control condition to subjects exposed to rationales for consonant views.²⁸

As Fig. 2 demonstrates, among those high in perspective-taking ability, mean levels of tolerance were higher when subjects were exposed to rationales for dissonant views. However, among those low in perspective-taking ability, tolerance levels were lower when subjects were exposed to dissonant views, although the higher variance among this group makes this a suggestive, though not significant difference.²⁹

To assess the size of this effect while taking other variables into account, Table 4 presents these results as regression equations. Column 1 in Table 4 shows that among those with high perspective-taking ability, those receiving exposure to rationales for dissonant views on three issues scored about 14% higher on the tolerance scale. Although the small, relatively homogeneous student sample used in this experiment is by no means a representative one, it is impressive that an effect of this size was generated by one single disembodied exposure to cross-cutting political views. As the second column in Table 4 shows, exposure to viewpoints consonant with the subject's own views produced no such effects, thus confirming that it is exposure to dissonant views that is encouraging tolerance rather than simply exposure to political viewpoints more generally.

DISCUSSION

Exposure to dissimilar views has been deemed a central element—if not the sine qua non—of the kind of political dialogue that is needed to maintain a democratic citizenry: “Democratic public discourse does not depend on preexisting harmony or similarity among citizens … but rather on the ability to create meaningful discourses across lines of difference” (Calhoun 1988, 220). The extent to which people are exposed to cross-cutting political viewpoints has become of increasing concern to observers of American politics as a result of trends toward increasing residential balkanization (e.g., Harrison and Bennett 1995; Frey 1995). If people self select into lifestyle enclaves with similar-minded others, their exposure to dissimilar political views should suffer.

To the extent that residential balkanization and other trends translate to a decline in communication across lines of political difference, one of its adverse effects may include fewer opportunities for people to learn about legitimate rationales for oppositional viewpoints. Particularly when policies or candidates other than one's own top preferences carry the day, the findings of this study suggest that the perceived legitimacy of the winning people and policies may be hindered by a lack of awareness of legitimate reasons for opposing views. If people are surrounded by people who think much like they do, they will be less aware of the legitimate arguments on the other side of contemporary political controversies.

Beyond legitimacy, the extent of exposure to dissonant political views may also be important for its indirect contributions to political tolerance. The capacity to see that there is more than one side to an issue, that a political conflict is, in fact, a legitimate controversy with rationales on both sides, translates to greater willingness to extend civil liberties to even those groups whose political views one dislikes a great deal.

This cognitive mechanism is augmented by influence that flows through the affective ties that people maintain across lines of political difference. Close personal ties with those of differing political views contribute to greater political tolerance. It is interesting to note that, from this perspective, the fact that Americans so seldom discuss politics in any depth is probably a feature, not a liability. Because politics is such a small part of people's day-to-day lives, when they come into contact with people of opposing views, it is relatively easy for them to ignore this dimension of difference or to discover it late enough that a friendship of some kind has already been initiated or established. Political views need not be at the forefront of daily life or daily conversation to produce beneficial consequences.

Moreover, the positive role played by affective ties to politically dissimilar others suggests a need to reconsider the role of emotion in democratic judgment. Particularly in research on political tolerance, there is a tendency to think of emotion as something that promotes intolerance and prejudicial reactions to others (cf. Kuklinski et al. 1991; Theiss-Morse, Marcus and Sullivan 1993). Although evidence on this point remains unresolved, the emotional versus deliberative citizen dichotomy often fails to acknowledge that through social interaction people form relationships with affective components as well as judgments based on the information that is conveyed.

Although in this study I have attempted to separate network characteristics such as intimacy, frequency, and agreement for analytic purposes, it should be acknowledged that in the real world they are inextricably intertwined. People generally feel closer to those who share their values, political and otherwise, and they talk more frequently with those to whom they are close. Thus it is important to note that efforts to increase exposure to disagreement may necessitate trade-offs in other network characteristics that are also generally valued. For example, to increase levels of exposure to oppositional views in the population, people will need to have a greater number of weak ties and probably less intimacy on average within their networks. And although trust is not directly examined in this study, it also goes hand in hand with homogeneity of views (e.g., Gibson 1999; Baldassare 1985); thus dense networks of tight-knit social relationships and their characteristic high levels of trust may come only at the expense of exposure to cross-cutting views. Close relationships obviously have their virtues, but large pluralistic societies such as the United States undoubtedly need citizens with a good number of weak ties in their social networks to sustain support for democratic freedoms in the midst of great heterogeneity (e.g., Simmel 1955; Karatnycky 1999).

Ultimately, political tolerance is about formalized ways in which people agree to disagree. It is primarily about restraint and not doing, rather than political action. Thus people's capacity to carry on conversations across lines of political difference, conversations in which one must agree to disagree at a micro level, may teach important lessons about the necessity of political tolerance, the public policy rendition of agreeing to disagree at the macro level.

These findings also have implications for the burgeoning empirical literature on deliberative democracy. A spate of recent studies, primarily experimental or quasi-experimental in nature, has attempted to manipulate deliberation by bringing people together to talk in small groups (see, e.g., Morrell 2000; Muhlberger and Butts 1998; Price and Cappella 2001; Simon and Sulkin 2000; Weber 1998). While such studies have provided many new insights on what happens when people are compelled to talk to one another about controversial issues, the broad and variable nature of their interactions also makes it difficult to determine which aspects of the experience are producing the observed effects. Moreover, because every small group exchange is somewhat different from the next, causal arguments have been more difficult to make than in most experimental studies. For example, it is difficult to know if effects are due to information gains through social interaction, the camaraderie of social interaction, group dynamics, and so forth. In reality, deliberation is a conglomeration of many variables, and it is often impossible to disentangle their effects when they are all varied simultaneously. From a social–psychological perspective, the advantage of this study is that it isolates one particular aspect of the deliberative encounter, the extent of cross-cutting exposure, and examines its consequences using both survey and experimental evidence. While advantageous for methodological reasons, and for purposes of understanding underlying processes of influence, this narrowness also limits the scope of the conclusions that should be drawn from it.

Does the composition of people's social networks have meaningful consequences for political tolerance and democratic legitimacy? My tentative answer to this question is yes, though they are relatively modest ones based on this survey and experimental evidence. Although these findings do not support the argument that more deliberation per se is what contemporary American politics need most, they do lend supporting evidence to arguments about the benefits of cross-cutting networks of political communication.