Who Talks Religion and What Are the Consequences for Social Ties? Unpacking a Sensitive Discussion Topic in Close Networks

Who Talks Religion and What Are the Consequences for Social Ties? Unpacking a Sensitive... Abstract Though religion matters greatly to many U.S. adults, it is widely considered a touchy conversational topic. Understanding how religious issues are talked about with others can elucidate a key interpersonal manifestation of Americans’ faith, yet existing research has largely overlooked the phenomena of religious discussion in social networks. This article considers which types of people talk about religion with their close ties, what relational factors underlie religious discussion, and what implications such discussion has for network turnover and stability. Applying multilevel regression methods to ego-centered networks measured in the Portraits of American Life Survey, I find that individual- and relational-level factors each predict the presence of religious discussion within close networks. Longitudinal analyses further reveal that for a large subset of Americans—namely evangelical and mainline Protestants and those involved in a congregation—religious discussion partners are especially likely to remain in the network over the course of 6 years. This association extended beyond other factors that could explain tie persistence, including relational closeness and multiple forms of homophily. Results point to several promising future directions for the study of religion and social networks. Religion is widely deemed a topic inappropriate for polite social conversation (Green 2016; Martin 1998). But salient as it is in the lives of so many believers and in the course of so many cultural and political debates (Putnam and Campbell 2010), religion seems like something Americans are inclined to talk about—with or without the etiquette experts’ blessing. And in fact, evidence suggests that just over 50% of all U.S. adults indeed discuss religion with someone outside of their family at least several times a year, with the likelihood of such conversation soaring to 83% among those who see themselves as highly religious (Cooperman 2016). From a dyadic perspective, U.S. adults report talking religion and spirituality in about half of all their (nonresidential) close tie relationships (Merino 2014). Still, existing knowledge on networks and religious discussion pushes little beyond these basic descriptive patterns. Among what types of people and in what types of relationships does the topic of religion arise? And what are the implications of religion discussions for people’s networks over time? These are questions that have yet to be clearly answered. The egocentric name-generator paradigm often employed in survey research is well-suited to identify that set of tight-knit, often homogenous ties that exerts normative influence on people’s lives and that can be called upon for support in times of need (Burt 1984; Marsden 1987).1 Still, what actually gets talked about inside these close networks—finances? current events? religion?—is glossed over in much empirical research (Bearman and Parigi 2004). To address the underlying heterogeneity of conversational topics and the possibility of topic-alter dependency within close networks, researchers have begun probing the particularities of what people discuss with their close ties (Brashears 2014; Gerber et al. 2012). Health and politics are two subject matters increasingly assessed as common discussion topics with implications for numerous important outcomes (Klofstad et al. 2009; Perry and Pescosolido 2010). Recent research, for instance, shows that health is discussed with only some members of the close network, and there is considerable between-person variability in how freely individuals talk about health issues with members of their network (Perry and Pescosolido 2010). What is more, discussing health with network members is associated with effective chronic disease detection and management (Cornwell and Waite 2012) and optimal recovery from mental health crises (Perry and Pescosolido 2015). On the politics front, recent research documents that people tend to report discussing politics between “sometimes” and “often” with friends and family (Gerber et al. 2012), yet these conversations occur most frequently and with least trepidation between people sharing political views (Cowan and Baldassarri 2017). Talking politics with members of one’s network appears to boost civic participation (Klofstad 2007; McClurg 2003), but the tendency to broach such topics mainly with like-minded partisans can magnify the impression of pervasive political polarization (Baldassarri and Bearman 2007). Despite its importance in the lives of many American adults, religion has not yet received comparable attention as an important discussion topic in people’s networks. This study represents one of the few studies to address that gap, first considering what personal and relational characteristics predict religious/spiritual discussion in U.S. adults’ close networks. The study’s second aim is to examine whether talking religion predicts the continuation of alters in a person’s network over time. Personal networks evolve amidst changing life circumstances (Morgan et al. 1996; Suitor et al. 1997), yet various relational characteristics act as relational adhesives through various transitions in the life course (Feld et al. 2007; Lubbers et al. 2010; Suitor and Keeton 1997). As an initial exploration of how religious discussion may matter for personal networks, I take up the question of network turnover and whether talking religion is associated with tie durability over the course of 6 years. Variation in religious discussion may be related to a host of other important outcomes—including the extent to which people tolerate divergent religious viewpoints and the extent to which they understand their community or society to be one marked by cultural concord or polarization. These matters cannot be empirically addressed in this study, but I hope that the basic findings presented below can help stimulate such investigations. WHO TALKS RELIGION? BACKGROUND AND THEORY By any measure, the United States is an unusually religious society among other advanced countries (Putnam and Campbell 2010). Yet there is considerable pluralism of belief and nonbelief within the American population (Machacek 2003); a classic solution to this scenario has been that set of norms relegating religion to the domain of out-of-bounds conversation topics. Still, many Americans do discuss religion, particularly with those in their close network (Merino 2014). At this point, there is no single overarching theory to explain why religious topics do or do not emerge within such networks. The following section first integrates existing theory on tie mobilization and selective confiding (Small 2013) to identify the relational processes that may underlie religious discussion. Then, I consult existing research on individual religiosity to generate empirical expectations. Relational Factors Talking religion is something that happens in a relational context, and so many of the determinants of the phenomena are likely to be dyadic in nature. I draw from three distinctive theoretical perspectives on the reasons people confide in network members (Small 2013)—topic-alter dependency, the strong ties framework, and opportune mobilization—to posit relational hypothesis about religious discussion in close networks. The premise of topic-alter dependency is that people match discussion topics with certain types of people in their personal network (Bearman and Parigi 2004). Role relations are one determining factor for this conversational calibration,2 but in the context of religious discussion, homophily may be the more crucial consideration. Specifically, I anticipate that people are more likely to talk religion with network members who share their religious background than with religiously dissimilar people. For the faithful, shared understandings and a common religious vernacular should make it easier to broach the touchy topic of religion and spirituality. By the same token, nonreligious people may find it most comfortable to discuss religion with fellow outsiders—even as the overall prevalence of such discussions may be less frequent for this group relative to religious believers. Each scenario is consistent with topic-alter dependency, but also with common understandings of homophily in social relationships. Homophily is thought to go hand-in-hand with smooth, comprehensible communication (McPherson et al. 2001), and we might expect this to be particularly true for sensitive and/or controversial conversation topics (e.g., religion).3 And indeed, prior research suggests that religious agreement predicts more frequent religious discussion among network members (Gerber et al. 2012). A second relational explanation of why networked people discuss religion comes from the strong ties perspective. This position implies that sensitive matters are discussed only with one’s most trusted confidants (Small 2013). To the extent that religion and spirituality is a delicate topic, dyadic indicators of intimacy and emotional closeness should then override other relational factors. Granovetter (1973:1361) famously defined a tie’s strength as some “combination of the amount of time, the emotional intensity, the intimacy (mutual confiding), and the reciprocal services which characterize the tie.” To investigate this second perspective, I will consider a series of predictors signaling such features of strong ties.4 Finally, the opportune mobilization argument contends that discussion topics are not primarily governed by ego–alter commonality or by a tie’s strength, but by who happens to be around for conversation. In other words, discussions—including deeps ones concerning religion or the spiritual—are ultimately practical affairs, “responsive to contextual opportunities” (Small 2013:473). Hence, we might expect people to talk religion with those who are around, due either due geographical proximity (i.e., the tie lives nearby) or the regularity of communication (in person, or enabled by technology). Individual-Level Factors Talking religion with a given network member is also likely a function of numerous individual-level factors. People who consider religion to be important in their lives and/or demonstrate it behaviorally through involvement in a congregation should tend to talk about religion with the people they deem important (Cooperman 2016). Religious affiliation may also matter. At least in the U.S. context, evangelical Protestants are known for emphasizing God’s direct, personal involvement in their everyday life and for telling others about their faith in the course of spreading the gospel message (Smith 2014). This may mean that, all else equal, Evangelicals are more likely than people with other (or no) religious affiliation to talk religion with members of their close network. It is also important to consider the impact of demographic characteristics. Across a variety of indicators, U.S. women and racial/ethnic minorities appear more religious than men and White people (Pew Research Center 2014), so we may expect network religious discussion to figure more prominently in the lives of the former groups relative to others. According to this logic, indicators of religiosity may in turn account for such patterns. Age/cohort is a third relevant characteristic. More recent cohorts of North Americans report strikingly lower levels of religious affiliation, church attendance, belief certitude, and importance of religion/spirituality than previous cohorts (Voas and Chaves 2016). Regardless of whether younger people are themselves religious, their generational milieu is one of secularization where a relatively small sliver of the cohort places much importance on religious matters. This may well mean a falling off of religious discussion across the age range of American adults. Finally, I will examine whether stressful life situations elicit religious and spiritual discussion in people’s networks. Theories of stress-buffering maintain that people facing demanding circumstances and who feel unable to cope through their own personal resources turn often to religion for consolation (Pargament 2002; Tabak and Mickelson 2009; Wei and Lui 2013). One outcome of this process may be the increased likelihood of talking about religion with close network members. Of course, the decision to talk religion with network members amidst personal adversity may hinge on what type of religiosity (if any) the individual embraces. Accordingly, my analysis will consider whether religious participation and/or affiliation dictates the impact of stressful circumstances. DOES RELIGIOUS DISCUSSION MATTER? EXPLORING IMPLICATIONS FOR TIE PERSISTENCE Describing the individual and relational predictors of religious discussion can elucidate how a highly salient—though controversial—aspect of American life unfolds in people’s close social environment. But a natural follow-up question is why this might matter for the lives of individual or for their networks. There are multiple ways one could pursue this question; as a first step, this article will take up the issue of how religious discussion intersects with network turnover. Network turnover is a ubiquitous process; life course transitions, job changes, and residential mobility effectively cycle various friends, neighbors, colleagues, and romantic partners in and out of people’s core networks (Bidart and Lavenu 2005; Morgan et al. 1996; Wellman et al. 1997). One recent analysis, for instance, reports that only 5% of older American adults retain an identical roster of close associates over a 5-year period, while nearly 60% added multiple confidants within that time frame (Cornwell and Laumann 2015). Another study using a representative U.S. sample suggests that among their four closest nonresidential associates, the typical adult swaps out and replaces two ties over the course of 6 years (Schafer and Vargas 2016). Ubiquitous network turnover is an important topic because, depending on the circumstances, it can bring about both risks and benefits. Adding new network members can boost self-esteem and enhance health (Cornwell and Laumann 2015), yet losing close ties may also disrupt stable access to resources and exacerbate inequality between income groups (Cornwell 2015; Schafer and Vargas 2016). The question for this study is whether talking religion factors into these pervasive dynamics of network change. Little research on network turnover has examined whether discussion content is associated with tie persistence, but there is reason to anticipate that for at least some segment of the population, discussing things religious will increase the likelihood that a given network member will remain in the close network over time. Social penetration theory offers a developmental account of relationships (Altman and Taylor 1973). The theory assumes that people have an “onion-layer” like set of beliefs, values, and commitments that come to be revealed to another person over time, the inner core peeled open only after conversation has moved beyond the outer, more superficial rings of the self. In this theory, people open up about sensitive matters if they believe that the rewards of deeper intimacy will outweigh the costs. And so while rapport and trust are typically a necessary starting point for increased disclosure, moving to “deeper” areas permits yet more intimacy and indicates that one forecasts the increased social penetration as a rewarding experience. In this way, social penetration shares similar expectations to the class of exchange-based theories which contend that closeness motivates exchange, but also that exchange gestures can boost positive sentiment and ultimately enhance solidarity within a dyad; for further reference, see accounts of relational cohesion theory (Thye et al. 2002) and accounts of “social-formation” in exchange theory more broadly (Lawler and Thye 1999). Continuation of the relationship is one of the main outcomes assessed by social penetration theory. Social penetration processes have an important affinity with religion and spirituality. For the faithful, religion concerns that which is sacred or of ultimate significance. Sacred matters, in turn, are imbued with powerful emotions, tend to be protected with great effort, and are intersubjective social resources—phenomena shared with other people that produce shared meaning and cohesion (Pargament and Mahoney 2005). For the faithful, then, dialogue about religion amounts to a “central, intimate facet of personality” and touches down on core issues of personal vulnerability (Altman and Taylor 1973:58). And in light of the continuous, reciprocal development of disclosure and intimacy (Altman and Taylor 1973:52; Collins and Miller 1994), religion may well represent a topical adhesive that provides a unique bond across time and space, one that stands out among other conventional markers of emotional closeness or intimacy typically assessed in studies of strong ties and network maintenance.5 In this way, religious discussion could be thought of as a potential cause and a consequence of tie closeness, but also a qualitatively different form of tie strength that predicts tie continuity. Developmental relationship models such as social penetration theory help clarify several important caveats and distinctions. First, people without a religious affiliation and/or those who do not participate in religious activity may well talk about religion from time to time, but such discussion is unlikely to touch on matters deemed sacred by the discussant. Therefore, we should expect the discussion of religious topics to be no more binding than the nondiscussion of such topics for this segment of the population. Second, the theory prompts consideration of whether talking religion with people outside of one’s own faith tradition is associated with tie durability. Intimacy is “simultaneously rewarding and risky” (Collins and Miller 1994:469), and the costs of delving into religion with a religiously dissimilar associate may outpace forecasted rewards of that disclosure. Indeed, overt conversation patterns and religious homophily could form multiplicative—not additive—associations with tie persistence. That is, religious homophily, as a form of cultural similarity, could itself prove a bonding agent in network relationships (Puetz 2015:446). And talking religion with an alter who shares one’s faith could further enhance intimacy and increase the likelihood that such a network member sticks around. Broaching such topics with a religious outsider, however, could potentially decrease chances of the tie remaining in one’s close network—particularly if the conversation was straining or awkward (Dallas 2016). One study, for instance, finds that close to 12% of Americans report having lost a friend over religious disagreement (Gerber et al. 2013). This analysis will therefore consider whether the link between religious discussion and tie persistence is contingent upon religious homophily. SUMMARY OF RESEARCH QUESTIONS In summary, two research questions guide the ensuing analyses: Research Question 1: What relational- and individual-level factors predict religious discussion in close networks? I hypothesize that religiosity (congregational involvement and affiliating as an Evangelical vs. no affiliation) is positively associated with religious discussion. In addition, I hypothesize that being older, female, and Black increases likelihood of religious discussion, as do exposures to stressors. Relational factors hypothesized to predict religious discussion include religious homophily, tie strength, and conversational opportunity. Research Question 2: Does religious discussion predict the persistence of ties in people’s close networks over time? Discussing religion is hypothesized to increase tie persistence, but only among people involved in a congregation and among those who embrace a religious affiliation. I further hypothesize that religious homophily will strengthen the association between religious discussion and tie persistence. DATA AND METHODS Sample Data for this study come from the Portraits of American Life Study (PALS), the only nationally representative panel study of which I am aware that asks about religious discussion in people’s close networks. The PALS first surveyed 2,610 noninstitutionalized American adults at least 18 years of age in 2006 (W1). The sample was obtained through a multistage process, in which zip code areas were randomly selected with probability proportionate to size, addresses were randomly selected from each zip code area, and one randomly selected adult was selected for a full interview from each selected household (after being screened for eligibility in an initial interview). The baseline survey produced an 83% contact rate, an 86% screening rate, and an 82% cooperation rate, resulting in an overall W1 response rate of 58% (0.83 × 0.86 × 0.82). With the available survey weights, the PALS sample closely mirrors basic population patterns found in the Census’ American Community Survey (Emerson et al. 2010). Further information about the PALS sampling design and recruitment is available at http://www.palsresearch.org/pals/researchers. Analyses from W1 come from the social ties among 2,435 PALS respondents who identified at least one close network tie outside of their home. Apart from the participants who reported no network members, missing data were minimal (<2% across all variables). Initial participants were re-contacted and 1,314 of them were interviewed again in 2011 (W2), yielding a 50.3% retention rate. Participants not successfully re-interviewed had larger average network size than re-interviewed participants and were disproportionately male, less-educated, lower-income, and non-White; there were not statistical differences in average age or county population size between the two groups. Those involved in a congregation at W1 were slightly more likely to be re-interviewed at W2 than noncongregants (52.9% vs. 48.3% retention), but there was no statistical difference by respondent’s religious tradition. To account for the potential biases of nonrandom attrition in the longitudinal analysis, I created inverse probability of attrition (IPA) weights.6 Longitudinal results were consistent whether using the conventional sampling weights or applying the IPA weighting scheme. Relational Measures The close networks of PALS respondents were constructed through a name-generator technique. Participants were first asked to identify up to four people “outside of your home you feel closest to.”7 They were then asked a series of questions about these network alters, the most central for this analysis being whether the person was one with whom ego “discussed religious or spiritual matters in the past 12 months.” The network name-interpreter battery also assessed alters’ demographic profile, including gender (male or female), race (whether “different race” or the same as ego), education level (4-year college degree or more vs. other), and kinship status (family members vs. other), as well as whether the tie is a coworker. Religious homophily was operationalized as whether the alter “shared [ego’s] religious faith.” Tie strength was first assessed by position in the network roster under the assumption that “the order in which [alters] were named is a reflection of [tie] strength” (Merino 2014:603). Given that Granovetter’s (1973) conception of strong ties includes the presence of confiding and supportive exchange, I also use measures of whether ego received advice (alter “provided advice that helped you make an important decision in the past 3 years”) or help (alter “volunteered their time to help you in times of need in the past 3 years”) to proxy tie strength. To capture opportunity mobilization, I use measures of whether the alter lives nearby (“within a 20 minute drive of your home”) and whether the duo interact at least once a week (in person or with technology).8 Besides assessing religious discussion, the PALS is a unique survey in that it determines tie persistence in a national sample across two waves. At W2, participants were asked to construct their ego network using the same name-generator technique as at baseline. Respondents were then asked to indicate whether present network members corresponded to any of the ties identified at W1. Respondents were shown their W1 network roster and asked whether each of the W2 names were distinct from initial members. From this information, I devised a dummy variable denoting whether the W1 alter persisted to W2. Respondent-Level Measures Individual-level predictors of religious discussion (also used as covariates in the analysis of tie persistence) include religious factors, demographic characteristics, and stressors. Congregational involvement was measured with the question “Are you currently involved in, affiliated with, or a member of a religious congregation or other place of worship?”9 Religious affiliation was categorized using the often-used scheme devised by Steensland et al. (2000); I distinguish Evangelicals, Black Protestants, Mainline Protestants, Catholics, other religion, and no religious affiliation (referred to as religious “nones”) in this study. For demographics, I account for age (years since birth at 2006), gender, race/ethnicity (White, Black, Latino non-Black, and other). A series of three dummy variables denoted whether the respondent had a variety of stress exposures in the past 3 years: “serious illness, injury, or an assault”; “a major financial crisis”; and the death of a family member or friend (“parent, child, or spouse died” or “a close family friend or relative (aunt, cousin, grandparent) died”). Additional controls include educational attainment (years of formal education), marital status (married vs. not married), working for pay (vs. not), U.S. census region (Northeast, Midwest, South, West), and county population size (the latter two variables to adjust for potential regional and/or urban/rural variation in religious discussion practices). Longitudinal analyses of network turnover further adjust for whether the respondent moved, changed working status, or changed marital status between waves. Preliminary analyses also adjusted for political affiliation, income, and parental status, but accounting for these variables had no influence on the study findings. Analysis The analysis proceeds in two main stages. After providing a descriptive account of sample statistics, I first assess the individual and relational factors, which predict religious discussion between ego (the PALS participant) and his/her network members. The outcome is a tie-level variable, so I use the multilevel modeling approach for ego-network data (van Duijn et al. 1999), where network alters (level 1 observations) are nested within the W1 PALS participants (level 2 observations). This random-intercept logistic regression model accounts for clustering of multiple observations within each respondent and incorporates predictor variables corresponding to multiple levels of analysis. The model is initially fit to the full multilevel sample, but I then examine parameter estimates separately according to egos’ religious tradition and separately for those who belong and do not belong to a congregation. This is to more fully understand whether the stress-consolation dynamics of religious discussion and/or the importance of religious homophily differs according to personal religiosity. The second stage of the analysis relies on the longitudinal sample and tests whether discussing religion is associated with tie persistence. For the PALS respondents re-interviewed in 2011, a binary variable denotes whether each W1 network alter remained in the ego network at W2 (33% of eligible W1 ties reappear in W2 participant networks). This analysis also uses a multilevel binary logistic model, incorporating baseline participant and relational characteristics to predict the odds of a given tie enduring over the course of 6 years. Consistent with the strategy mentioned above, the model is fit to both the full longitudinal sample and to subsets of the PALS sample that reflect different expressions of religious fidelity (or none whatsoever). This is primarily to assess whether talking religion is more consequential for more religious people. In addition, I will test an interaction term between religious discussion and religious homophily, both in the full sample and in the selected subsamples—this to assess whether talking religion has different implications for tie persistence depending on whether ego and alter share the same religious background. All analyses use Stata version 12 and incorporate the recommended survey weights to generalize findings to the noninstitutionalized population of U.S. adults. Robust standard errors account for multistage survey design. RESULTS Weighted descriptive statistics are shown for PALS respondents (table 1) and for the social network dyads (table 2) embedded in these survey participants. Nearly half of the baseline survey respondents were part of a religious congregation. Evangelical Protestants and Catholics were the two most common religious affiliations, both representing just over a quarter of the sample. Religious nones comprised 16% of the sample. Dyadic-level data reveal that religious discussion transpired within about half of all ego–alter relationships. There was strong evidence of religious homophily—about two-thirds of all ties were between people sharing a religious background—but this type of commonality was less pronounced than racial and gender homophily (92% and 72%, respectively). Table 1 Weighted Descriptive Statistics for Individual Level Variables (Portraits of American Life Survey, W1) Mean/prop. Standard deviation Part of cong. 0.47 Evangelical 0.27 Black Protestant 0.05 Mainline Protestant 0.14 Catholic 0.26 Other religion 0.12 No religion (nones) 0.16 Female 0.52 Black 0.11 Latino/a 0.12 Other race 0.05 Age 45.12 16.47 Serious illness/injury 0.22 Death of loved one 0.67 Financial crisis 0.23 Midwest 0.23 South 0.34 West 0.26 Northeast 0.17 County pop. Size 4.87 2.4 Education 14.46 2.84 Married 0.58 Worker 0.64 Mean/prop. Standard deviation Part of cong. 0.47 Evangelical 0.27 Black Protestant 0.05 Mainline Protestant 0.14 Catholic 0.26 Other religion 0.12 No religion (nones) 0.16 Female 0.52 Black 0.11 Latino/a 0.12 Other race 0.05 Age 45.12 16.47 Serious illness/injury 0.22 Death of loved one 0.67 Financial crisis 0.23 Midwest 0.23 South 0.34 West 0.26 Northeast 0.17 County pop. Size 4.87 2.4 Education 14.46 2.84 Married 0.58 Worker 0.64 Notes: n = 2,435. Standard deviations are omitted for categorical variables. View Large Table 1 Weighted Descriptive Statistics for Individual Level Variables (Portraits of American Life Survey, W1) Mean/prop. Standard deviation Part of cong. 0.47 Evangelical 0.27 Black Protestant 0.05 Mainline Protestant 0.14 Catholic 0.26 Other religion 0.12 No religion (nones) 0.16 Female 0.52 Black 0.11 Latino/a 0.12 Other race 0.05 Age 45.12 16.47 Serious illness/injury 0.22 Death of loved one 0.67 Financial crisis 0.23 Midwest 0.23 South 0.34 West 0.26 Northeast 0.17 County pop. Size 4.87 2.4 Education 14.46 2.84 Married 0.58 Worker 0.64 Mean/prop. Standard deviation Part of cong. 0.47 Evangelical 0.27 Black Protestant 0.05 Mainline Protestant 0.14 Catholic 0.26 Other religion 0.12 No religion (nones) 0.16 Female 0.52 Black 0.11 Latino/a 0.12 Other race 0.05 Age 45.12 16.47 Serious illness/injury 0.22 Death of loved one 0.67 Financial crisis 0.23 Midwest 0.23 South 0.34 West 0.26 Northeast 0.17 County pop. Size 4.87 2.4 Education 14.46 2.84 Married 0.58 Worker 0.64 Notes: n = 2,435. Standard deviations are omitted for categorical variables. View Large Table 2 Weighted Descriptive Statistics for Relational Level Variables (Portraits of American Life Survey, W1) Proportion Discussion religion 0.49 Religious homophily 0.65 Alter rank 1 0.27 Alter rank 2 0.26 Alter rank 3 0.25 Alter rank 4 0.22 Alter gave advice 0.58 Alter gave help 0.79 Lives nearby 0.52 Interact weekly 0.79 Racial homophily 0.92 Gender homophily 0.72 Alter has bachelor degree 0.34 Kin 0.46 Co-worker 0.09 Tie remain, W1–W2 0.33 Proportion Discussion religion 0.49 Religious homophily 0.65 Alter rank 1 0.27 Alter rank 2 0.26 Alter rank 3 0.25 Alter rank 4 0.22 Alter gave advice 0.58 Alter gave help 0.79 Lives nearby 0.52 Interact weekly 0.79 Racial homophily 0.92 Gender homophily 0.72 Alter has bachelor degree 0.34 Kin 0.46 Co-worker 0.09 Tie remain, W1–W2 0.33 Note: n = 8,999 (within 2,435 individual respondents). View Large Table 2 Weighted Descriptive Statistics for Relational Level Variables (Portraits of American Life Survey, W1) Proportion Discussion religion 0.49 Religious homophily 0.65 Alter rank 1 0.27 Alter rank 2 0.26 Alter rank 3 0.25 Alter rank 4 0.22 Alter gave advice 0.58 Alter gave help 0.79 Lives nearby 0.52 Interact weekly 0.79 Racial homophily 0.92 Gender homophily 0.72 Alter has bachelor degree 0.34 Kin 0.46 Co-worker 0.09 Tie remain, W1–W2 0.33 Proportion Discussion religion 0.49 Religious homophily 0.65 Alter rank 1 0.27 Alter rank 2 0.26 Alter rank 3 0.25 Alter rank 4 0.22 Alter gave advice 0.58 Alter gave help 0.79 Lives nearby 0.52 Interact weekly 0.79 Racial homophily 0.92 Gender homophily 0.72 Alter has bachelor degree 0.34 Kin 0.46 Co-worker 0.09 Tie remain, W1–W2 0.33 Note: n = 8,999 (within 2,435 individual respondents). View Large Discussing Religion Table 3 presents the full, main-sample analysis of discussing religion in a series of three models. This is done because the two core indicators of personal religiosity, congregational involvement and affiliation, overlap to a considerable degree (i.e., very few people with no affiliation are connected to a congregation). Therefore, model 1 focuses on congregational involvement, while model 2 substitutes in religious affiliation. Model 3 includes both dimensions of religiosity for the sake of comparability.10 Models 4–10 replicate the analysis for specific religious subsets of the sample.11 Table 3 Discussing Religion, Odds Ratio Estimates from Random Intercept Logistic Regression Models (Portraits of American Life Survey, W1) Full sample Part of congregation Not part of congregation Evangelicals Mainline Protestants Black Protestants Catholics Religious nones Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Model 8 Model 9 Model 10 Individual-level predictors  Part of cong. 3.79*** 4.23** 6.81*** 2.80*** 3.17** 3.00** 1.02 (2.84–5.07) (3.20–5.59) (3.51–13.21) (1.65–4.75) (1.34–7.53) (1.51–5.95) (0.32–3.24)  Evangelicala 1.84* 0.95 4.54* 0.64 (1.14–2.99) (0.60–1.49) (1.18–17.50) (0.37–1.12)  Black Protestanta 1.52 0.81 3.02 0.71 (0.82–2.83) (0.42–1.53) (0.66–13.78) (0.32–1.60)  Mainline Protestanta 0.66 0.34** 1.13 0.41** (0.42–1.03) (0.22–0.52) (0.32–4.00) (0.23–0.73)  Catholica 0.92 0.50** 1.82 0.49* - (0.56–1.51) (0.31–0.82) (0.42–7.88) (0.26–0.94)  Other 0.88 0.66 4.80* 0.50*  Religiona (0.44–1.77) (0.36–1.22) (1.26–18.32) (0.27–0.91)  Female 1.81*** 2.40*** 1.94** 1.35 2.49*** 1.91 2.56* 0.75 3.15** 1.00 (1.30–2.51) (1.60–3.58) (1.40–2.70) (0.70–2.59) (1.66–3.73) (0.95–3.83) (1.24–5.28) (0.21–2.64) (1.38–7.21) (0.39–2.58)  Blackb 0.90 0.88 0.77 0.60 1.16 0.63 0.81 0.41 0.86 (0.53–1.54) (0.51–1.51) (0.44–1.34) (0.30–1.20) (0.55–2.47) (0.23–1.73) (0.19–3.41) (0.16–1.06) (0.27–2.78)  Latino/ab 0.33*** 0.39*** 0.39** 0.25* 0.54 0.34** 0.64 0.45* 0.43 (0.21–0.52) (0.23–0.64) (0.24–0.64) (0.08–0.76) (0.28–1.06) (0.17–0.67) (0.15–2.68) (0.23–0.87) (0.10–1.80)  Other raceb 0.19*** 0.20*** 0.17** 0.13*** 0.18*** 0.14*** 0.37 0.14*** 0.55 (0.08–0.43) (0.08–0.46) (0.08–0.37) (0.05–0.35) (0.08–0.43) (0.06–0.33) (0.04–3.25) (0.05–0.40) (0.12–2.57)  Age 0.99 1.00 1.00 1.00 0.99 0.98 0.99 0.99 1.01 1.00 (0.98–1.00) (0.99–1.01) (0.99–1.01) (0.98–1.02) (0.98–1.01) (0.96–1.02) (0.96–1.02) (0.96–1.03) (0.99–1.03) (0.97–1.03)  Serious illness/injury 1.17 1.03 1.14 1.06 1.29 1.07 0.95 2.12 1.56 0.82 (0.78–1.77) (0.68–1.57) (0.78–1.68) (0.45–2.46) (0.83–1.99) (0.47–2.46) (0.40–2.27) (0.73–6.18) (0.81–3.01) (0.35–1.91)  Death of loved one 1.26 1.36 1.21 1.10 1.31 0.79 0.94 2.28 1.80* 1.04 (0.89–1.79) (0.97–1.90) (0.85–1.72) (0.61–1.99) (0.85–2.02) (0.42–1.52) (0.52–1.71) (0.82–6.33) (1.00–3.25) (0.43–2.53)  Financial crisis 1.33 1.22 1.30 1.32 1.33 1.17 2.03 1.73 1.09 2.35 (0.89–1.99) (0.79–1.90) (0.87–1.92) (0.59–2.96) (0.77–2.30) (0.65–2.11) (0.91–4.56) (0.57–5.27) (0.51–2.33) (0.57–9.69)  Midwestc 1.24 1.17 1.18 0.88 1.73 2.06 0.99 1.95 0.71 1.26 (0.73–2.10) (0.75–1.81) (0.72–1.91) (0.37–2.11) (0.92–3.25) (0.62–6.89) (0.34–2.86) (0.33–11.53) (0.33–1.51) (0.34–4.66)  Southc 2.06*** 1.88*** 1.79** 1.54 2.16** 2.34 1.31 1.68 1.00 1.95 (1.32–3.20) (1.29–2.73) (1.18–2.71) (0.72–3.32) (1.28–3.63) (0.76–7.16) (0.43–4.02) (0.33–8.65) (0.52–1.91) (0.56–6.77)  Westc 2.36*** 2.05** 2.33** 2.10 2.89*** 4.27* 2.00 1.22 1.28 1.92 (1.41–3.95) (1.24–3.39) (1.35–4.02) (0.62–7.12) (1.68–4.97) (1.09–16.72) (0.63–6.31) (0.28–5.24) (0.61–2.66) (0.49–7.61)  County pop. size 1.05 1.04 1.05 1.08 1.01 1.10 1.03 1.13 1.08 1.00 (0.96–1.15) (0.96–1.13) (0.96–1.15) (0.90–1.30) (0.90–1.14) (0.98–1.23) (0.89–1.18) (0.92–1.41) (0.98–1.19) (0.80–1.25)  Education 1.06* 1.11*** 1.08* 1.12 1.05 1.23** 1.01 1.27* 1.15** 1.13 (1.01–1.12) (1.04–1.18) (1.01–1.15) (0.96–1.31) (0.97–1.14) (1.06–1.43) (0.89–1.16) (1.02–1.59) (1.04–1.28) (0.94–1.37)  Married 0.94 1.02 0.96 1.05 0.84 1.41 0.96 1.31 1.13 0.54 (0.69–1.26) (0.76–1.38) (0.72–1.28) (0.63–1.73) (0.60–1.19) (0.83–2.39) (0.42–2.18) (0.59–2.91) (0.68–1.87) (0.23–1.26)  Worker 1.21 1.27 1.23 1.48 1.17 1.04 0.58 0.86 1.55 0.91 (0.83–1.75) (0.94–1.72) (0.90–1.69) (0.68–3.21) (0.69–1.98) (0.45–2.40) (0.26–1.28) (0.36–2.07) (0.84–2.86) (0.40–2.07) Relational-level predictors  Religious homophily 1.55*** 1.57** 1.58** 1.83*** 1.39 3.10*** 1.28 6.49*** 1.23 1.24 (1.24–1.93) (1.26–1.97) (1.27–1.96) (1.32–2.55) (0.99–1.94) (2.07–4.64) (0.71–2.32) (2.22–18.94) (0.75–2.01) (0.75–2.05)  Alter rank 1 1.88*** 1.87*** 1.88** 2.06*** 1.74** 1.61 1.33 1.94 3.81*** 1.46 (1.51–2.36) (1.50–2.34) (1.50–2.35) (1.46–2.90) (1.29–2.34) (0.86–3.00) (0.80–2.23) (0.97–3.88) (2.64–5.49) (0.76–2.82)  Alter rank 2 1.16 1.15 1.16 1.17 1.15 0.88 0.95 1.93 2.14*** 0.86 (0.92–1.45) (0.92–1.44) (0.92–1.45) (0.86–1.60) (0.84–1.59) (0.57–1.35) (0.51–1.77) (0.93–4.00) (1.47–3.12) (0.45–1.62)  Alter rank 3 0.99 0.99 0.99 1.07 0.92 0.81 1.15 1.51 1.19 0.78 (0.81–1.22) (0.80–1.21) (0.81–1.22) (0.78–1.49) (0.70–1.20) (0.47–1.39) (0.67–1.99) (0.86–2.67) (0.78–1.82) (0.46–1.30)  Alter gave advice 2.79*** 2.76*** 2.87** 3.09** 2.78*** 3.07*** 3.28*** 4.51*** 2.91*** 2.43** (2.21–3.52) (2.19–3.47) (2.26–3.65) (2.00–4.76) (2.04–3.79) (1.70–5.54) (2.00–5.37) (1.89–10.77) (1.59–5.30) (1.39–4.25)  Alter gave help 1.86*** 1.91*** 1.87** 1.90** 1.76*** 2.37*** 2.27* 1.80 1.36 2.10 (1.47–2.36) (1.48–2.47) (1.48–2.36) (1.29–2.80) (1.25–2.47) (1.51–3.73) (1.15–4.49) (0.91–3.56) (0.78–2.36) (0.89–4.94)  Lives nearby 1.40*** 1.43*** 1.41** 1.45* 1.35 1.44 1.10 1.26 1.39 1.20 (1.15–1.70) (1.18–1.74) (1.16–1.72) (1.07–1.96) (0.99–1.84) (0.90–2.28) (0.64–1.87) (0.62–2.57) (0.97–1.98) (0.65–2.23)  Interact weekly 1.73*** 1.74*** 1.71** 1.99** 1.56** 2.31** 2.01* 1.53 1.08 2.67** (1.32–2.26) (1.32–2.29) (1.31–2.24) (1.29–3.10) (1.14–2.12) (1.28–4.16) (1.00–4.05) (0.74–3.16) (0.65–1.78) (1.31–5.43)  Racial homophily 0.78 0.81 0.78 1.00 0.74 1.50 0.90 0.95 0.72 0.88 (0.50–1.20) (0.52–1.25) (0.50–1.20) (0.52–1.96) (0.45–1.23) (0.85–2.65) (0.34–2.38) (0.19–4.77) (0.22–2.33) (0.40–1.96)  Gender homophily 0.90 0.92 0.92 1.00 0.86 0.77 0.94 1.05 0.81 1.18 (0.72–1.14) (0.73–1.15) (0.73–1.16) (0.69–1.43) (0.66–1.12) (0.52–1.15) (0.58–1.54) (0.50–2.21) (0.47–1.38) (0.76–1.83)  Alter has bachelor dg. 1.32* 1.32* 1.32* 0.94 1.78*** 0.94 1.72 2.36* 1.03 1.70 (1.06–1.66) (1.05–1.65) (1.05–1.67) (0.67–1.31) (1.27–2.49) (0.59–1.52) (0.98–3.02) (1.07–5.24) (0.65–1.61) (0.85–3.40)  Kin 0.84 0.83 0.85 0.98 0.80 0.79 1.32 0.97 1.16 0.46* (0.69–1.01) (0.68–1.01) (0.71–1.02) (0.71–1.35) (0.60–1.05) (0.48–1.30) (0.75–2.32) (0.57–1.66) (0.79–1.70) (0.21–0.99)  Co-worker 0.90 0.89 0.89 1.20 0.79 0.66 1.45 2.45 0.96 0.80 (0.65–1.24) (0.64–1.26) (0.64–1.24) (0.70–2.07) (0.54–1.14) (0.30–1.45) (0.64–3.26) (0.20–29.47) (0.50–1.84) (0.33–1.93)  Level 1 5.23 5.45 5.17 6.48 4.24 7.26 3.18 5.37 5.73 4.10  Variance (0.50) (0.53) (0.51) (0.95) (0.56) (2.25) (0.71) (2.84) (1.27) (0.95)  −2LL 4841.96 4886.24 4821.47 2195.57 2591.21 1227.96 741.20 489.75 1189.65 788.78 Observations  Level 1 8,944 8,945 8,944 4,139 4,805 1,985 1,026 926 2,469 1,418  Level 2 2,419 2,420 2,419 1,102 1,317 529 270 256 683 384 Full sample Part of congregation Not part of congregation Evangelicals Mainline Protestants Black Protestants Catholics Religious nones Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Model 8 Model 9 Model 10 Individual-level predictors  Part of cong. 3.79*** 4.23** 6.81*** 2.80*** 3.17** 3.00** 1.02 (2.84–5.07) (3.20–5.59) (3.51–13.21) (1.65–4.75) (1.34–7.53) (1.51–5.95) (0.32–3.24)  Evangelicala 1.84* 0.95 4.54* 0.64 (1.14–2.99) (0.60–1.49) (1.18–17.50) (0.37–1.12)  Black Protestanta 1.52 0.81 3.02 0.71 (0.82–2.83) (0.42–1.53) (0.66–13.78) (0.32–1.60)  Mainline Protestanta 0.66 0.34** 1.13 0.41** (0.42–1.03) (0.22–0.52) (0.32–4.00) (0.23–0.73)  Catholica 0.92 0.50** 1.82 0.49* - (0.56–1.51) (0.31–0.82) (0.42–7.88) (0.26–0.94)  Other 0.88 0.66 4.80* 0.50*  Religiona (0.44–1.77) (0.36–1.22) (1.26–18.32) (0.27–0.91)  Female 1.81*** 2.40*** 1.94** 1.35 2.49*** 1.91 2.56* 0.75 3.15** 1.00 (1.30–2.51) (1.60–3.58) (1.40–2.70) (0.70–2.59) (1.66–3.73) (0.95–3.83) (1.24–5.28) (0.21–2.64) (1.38–7.21) (0.39–2.58)  Blackb 0.90 0.88 0.77 0.60 1.16 0.63 0.81 0.41 0.86 (0.53–1.54) (0.51–1.51) (0.44–1.34) (0.30–1.20) (0.55–2.47) (0.23–1.73) (0.19–3.41) (0.16–1.06) (0.27–2.78)  Latino/ab 0.33*** 0.39*** 0.39** 0.25* 0.54 0.34** 0.64 0.45* 0.43 (0.21–0.52) (0.23–0.64) (0.24–0.64) (0.08–0.76) (0.28–1.06) (0.17–0.67) (0.15–2.68) (0.23–0.87) (0.10–1.80)  Other raceb 0.19*** 0.20*** 0.17** 0.13*** 0.18*** 0.14*** 0.37 0.14*** 0.55 (0.08–0.43) (0.08–0.46) (0.08–0.37) (0.05–0.35) (0.08–0.43) (0.06–0.33) (0.04–3.25) (0.05–0.40) (0.12–2.57)  Age 0.99 1.00 1.00 1.00 0.99 0.98 0.99 0.99 1.01 1.00 (0.98–1.00) (0.99–1.01) (0.99–1.01) (0.98–1.02) (0.98–1.01) (0.96–1.02) (0.96–1.02) (0.96–1.03) (0.99–1.03) (0.97–1.03)  Serious illness/injury 1.17 1.03 1.14 1.06 1.29 1.07 0.95 2.12 1.56 0.82 (0.78–1.77) (0.68–1.57) (0.78–1.68) (0.45–2.46) (0.83–1.99) (0.47–2.46) (0.40–2.27) (0.73–6.18) (0.81–3.01) (0.35–1.91)  Death of loved one 1.26 1.36 1.21 1.10 1.31 0.79 0.94 2.28 1.80* 1.04 (0.89–1.79) (0.97–1.90) (0.85–1.72) (0.61–1.99) (0.85–2.02) (0.42–1.52) (0.52–1.71) (0.82–6.33) (1.00–3.25) (0.43–2.53)  Financial crisis 1.33 1.22 1.30 1.32 1.33 1.17 2.03 1.73 1.09 2.35 (0.89–1.99) (0.79–1.90) (0.87–1.92) (0.59–2.96) (0.77–2.30) (0.65–2.11) (0.91–4.56) (0.57–5.27) (0.51–2.33) (0.57–9.69)  Midwestc 1.24 1.17 1.18 0.88 1.73 2.06 0.99 1.95 0.71 1.26 (0.73–2.10) (0.75–1.81) (0.72–1.91) (0.37–2.11) (0.92–3.25) (0.62–6.89) (0.34–2.86) (0.33–11.53) (0.33–1.51) (0.34–4.66)  Southc 2.06*** 1.88*** 1.79** 1.54 2.16** 2.34 1.31 1.68 1.00 1.95 (1.32–3.20) (1.29–2.73) (1.18–2.71) (0.72–3.32) (1.28–3.63) (0.76–7.16) (0.43–4.02) (0.33–8.65) (0.52–1.91) (0.56–6.77)  Westc 2.36*** 2.05** 2.33** 2.10 2.89*** 4.27* 2.00 1.22 1.28 1.92 (1.41–3.95) (1.24–3.39) (1.35–4.02) (0.62–7.12) (1.68–4.97) (1.09–16.72) (0.63–6.31) (0.28–5.24) (0.61–2.66) (0.49–7.61)  County pop. size 1.05 1.04 1.05 1.08 1.01 1.10 1.03 1.13 1.08 1.00 (0.96–1.15) (0.96–1.13) (0.96–1.15) (0.90–1.30) (0.90–1.14) (0.98–1.23) (0.89–1.18) (0.92–1.41) (0.98–1.19) (0.80–1.25)  Education 1.06* 1.11*** 1.08* 1.12 1.05 1.23** 1.01 1.27* 1.15** 1.13 (1.01–1.12) (1.04–1.18) (1.01–1.15) (0.96–1.31) (0.97–1.14) (1.06–1.43) (0.89–1.16) (1.02–1.59) (1.04–1.28) (0.94–1.37)  Married 0.94 1.02 0.96 1.05 0.84 1.41 0.96 1.31 1.13 0.54 (0.69–1.26) (0.76–1.38) (0.72–1.28) (0.63–1.73) (0.60–1.19) (0.83–2.39) (0.42–2.18) (0.59–2.91) (0.68–1.87) (0.23–1.26)  Worker 1.21 1.27 1.23 1.48 1.17 1.04 0.58 0.86 1.55 0.91 (0.83–1.75) (0.94–1.72) (0.90–1.69) (0.68–3.21) (0.69–1.98) (0.45–2.40) (0.26–1.28) (0.36–2.07) (0.84–2.86) (0.40–2.07) Relational-level predictors  Religious homophily 1.55*** 1.57** 1.58** 1.83*** 1.39 3.10*** 1.28 6.49*** 1.23 1.24 (1.24–1.93) (1.26–1.97) (1.27–1.96) (1.32–2.55) (0.99–1.94) (2.07–4.64) (0.71–2.32) (2.22–18.94) (0.75–2.01) (0.75–2.05)  Alter rank 1 1.88*** 1.87*** 1.88** 2.06*** 1.74** 1.61 1.33 1.94 3.81*** 1.46 (1.51–2.36) (1.50–2.34) (1.50–2.35) (1.46–2.90) (1.29–2.34) (0.86–3.00) (0.80–2.23) (0.97–3.88) (2.64–5.49) (0.76–2.82)  Alter rank 2 1.16 1.15 1.16 1.17 1.15 0.88 0.95 1.93 2.14*** 0.86 (0.92–1.45) (0.92–1.44) (0.92–1.45) (0.86–1.60) (0.84–1.59) (0.57–1.35) (0.51–1.77) (0.93–4.00) (1.47–3.12) (0.45–1.62)  Alter rank 3 0.99 0.99 0.99 1.07 0.92 0.81 1.15 1.51 1.19 0.78 (0.81–1.22) (0.80–1.21) (0.81–1.22) (0.78–1.49) (0.70–1.20) (0.47–1.39) (0.67–1.99) (0.86–2.67) (0.78–1.82) (0.46–1.30)  Alter gave advice 2.79*** 2.76*** 2.87** 3.09** 2.78*** 3.07*** 3.28*** 4.51*** 2.91*** 2.43** (2.21–3.52) (2.19–3.47) (2.26–3.65) (2.00–4.76) (2.04–3.79) (1.70–5.54) (2.00–5.37) (1.89–10.77) (1.59–5.30) (1.39–4.25)  Alter gave help 1.86*** 1.91*** 1.87** 1.90** 1.76*** 2.37*** 2.27* 1.80 1.36 2.10 (1.47–2.36) (1.48–2.47) (1.48–2.36) (1.29–2.80) (1.25–2.47) (1.51–3.73) (1.15–4.49) (0.91–3.56) (0.78–2.36) (0.89–4.94)  Lives nearby 1.40*** 1.43*** 1.41** 1.45* 1.35 1.44 1.10 1.26 1.39 1.20 (1.15–1.70) (1.18–1.74) (1.16–1.72) (1.07–1.96) (0.99–1.84) (0.90–2.28) (0.64–1.87) (0.62–2.57) (0.97–1.98) (0.65–2.23)  Interact weekly 1.73*** 1.74*** 1.71** 1.99** 1.56** 2.31** 2.01* 1.53 1.08 2.67** (1.32–2.26) (1.32–2.29) (1.31–2.24) (1.29–3.10) (1.14–2.12) (1.28–4.16) (1.00–4.05) (0.74–3.16) (0.65–1.78) (1.31–5.43)  Racial homophily 0.78 0.81 0.78 1.00 0.74 1.50 0.90 0.95 0.72 0.88 (0.50–1.20) (0.52–1.25) (0.50–1.20) (0.52–1.96) (0.45–1.23) (0.85–2.65) (0.34–2.38) (0.19–4.77) (0.22–2.33) (0.40–1.96)  Gender homophily 0.90 0.92 0.92 1.00 0.86 0.77 0.94 1.05 0.81 1.18 (0.72–1.14) (0.73–1.15) (0.73–1.16) (0.69–1.43) (0.66–1.12) (0.52–1.15) (0.58–1.54) (0.50–2.21) (0.47–1.38) (0.76–1.83)  Alter has bachelor dg. 1.32* 1.32* 1.32* 0.94 1.78*** 0.94 1.72 2.36* 1.03 1.70 (1.06–1.66) (1.05–1.65) (1.05–1.67) (0.67–1.31) (1.27–2.49) (0.59–1.52) (0.98–3.02) (1.07–5.24) (0.65–1.61) (0.85–3.40)  Kin 0.84 0.83 0.85 0.98 0.80 0.79 1.32 0.97 1.16 0.46* (0.69–1.01) (0.68–1.01) (0.71–1.02) (0.71–1.35) (0.60–1.05) (0.48–1.30) (0.75–2.32) (0.57–1.66) (0.79–1.70) (0.21–0.99)  Co-worker 0.90 0.89 0.89 1.20 0.79 0.66 1.45 2.45 0.96 0.80 (0.65–1.24) (0.64–1.26) (0.64–1.24) (0.70–2.07) (0.54–1.14) (0.30–1.45) (0.64–3.26) (0.20–29.47) (0.50–1.84) (0.33–1.93)  Level 1 5.23 5.45 5.17 6.48 4.24 7.26 3.18 5.37 5.73 4.10  Variance (0.50) (0.53) (0.51) (0.95) (0.56) (2.25) (0.71) (2.84) (1.27) (0.95)  −2LL 4841.96 4886.24 4821.47 2195.57 2591.21 1227.96 741.20 489.75 1189.65 788.78 Observations  Level 1 8,944 8,945 8,944 4,139 4,805 1,985 1,026 926 2,469 1,418  Level 2 2,419 2,420 2,419 1,102 1,317 529 270 256 683 384 Notes: Robust standard errors are shown in parentheses. aReligions none is reference group. bWhite is reference group; Race/ethnicity not included in model 8 because the overwhelming majority of Black Protestants are Black. cNortheast is reference group. *p < .05; **p < .01; ***p < .001 (two-tailed tests). View Large Table 3 Discussing Religion, Odds Ratio Estimates from Random Intercept Logistic Regression Models (Portraits of American Life Survey, W1) Full sample Part of congregation Not part of congregation Evangelicals Mainline Protestants Black Protestants Catholics Religious nones Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Model 8 Model 9 Model 10 Individual-level predictors  Part of cong. 3.79*** 4.23** 6.81*** 2.80*** 3.17** 3.00** 1.02 (2.84–5.07) (3.20–5.59) (3.51–13.21) (1.65–4.75) (1.34–7.53) (1.51–5.95) (0.32–3.24)  Evangelicala 1.84* 0.95 4.54* 0.64 (1.14–2.99) (0.60–1.49) (1.18–17.50) (0.37–1.12)  Black Protestanta 1.52 0.81 3.02 0.71 (0.82–2.83) (0.42–1.53) (0.66–13.78) (0.32–1.60)  Mainline Protestanta 0.66 0.34** 1.13 0.41** (0.42–1.03) (0.22–0.52) (0.32–4.00) (0.23–0.73)  Catholica 0.92 0.50** 1.82 0.49* - (0.56–1.51) (0.31–0.82) (0.42–7.88) (0.26–0.94)  Other 0.88 0.66 4.80* 0.50*  Religiona (0.44–1.77) (0.36–1.22) (1.26–18.32) (0.27–0.91)  Female 1.81*** 2.40*** 1.94** 1.35 2.49*** 1.91 2.56* 0.75 3.15** 1.00 (1.30–2.51) (1.60–3.58) (1.40–2.70) (0.70–2.59) (1.66–3.73) (0.95–3.83) (1.24–5.28) (0.21–2.64) (1.38–7.21) (0.39–2.58)  Blackb 0.90 0.88 0.77 0.60 1.16 0.63 0.81 0.41 0.86 (0.53–1.54) (0.51–1.51) (0.44–1.34) (0.30–1.20) (0.55–2.47) (0.23–1.73) (0.19–3.41) (0.16–1.06) (0.27–2.78)  Latino/ab 0.33*** 0.39*** 0.39** 0.25* 0.54 0.34** 0.64 0.45* 0.43 (0.21–0.52) (0.23–0.64) (0.24–0.64) (0.08–0.76) (0.28–1.06) (0.17–0.67) (0.15–2.68) (0.23–0.87) (0.10–1.80)  Other raceb 0.19*** 0.20*** 0.17** 0.13*** 0.18*** 0.14*** 0.37 0.14*** 0.55 (0.08–0.43) (0.08–0.46) (0.08–0.37) (0.05–0.35) (0.08–0.43) (0.06–0.33) (0.04–3.25) (0.05–0.40) (0.12–2.57)  Age 0.99 1.00 1.00 1.00 0.99 0.98 0.99 0.99 1.01 1.00 (0.98–1.00) (0.99–1.01) (0.99–1.01) (0.98–1.02) (0.98–1.01) (0.96–1.02) (0.96–1.02) (0.96–1.03) (0.99–1.03) (0.97–1.03)  Serious illness/injury 1.17 1.03 1.14 1.06 1.29 1.07 0.95 2.12 1.56 0.82 (0.78–1.77) (0.68–1.57) (0.78–1.68) (0.45–2.46) (0.83–1.99) (0.47–2.46) (0.40–2.27) (0.73–6.18) (0.81–3.01) (0.35–1.91)  Death of loved one 1.26 1.36 1.21 1.10 1.31 0.79 0.94 2.28 1.80* 1.04 (0.89–1.79) (0.97–1.90) (0.85–1.72) (0.61–1.99) (0.85–2.02) (0.42–1.52) (0.52–1.71) (0.82–6.33) (1.00–3.25) (0.43–2.53)  Financial crisis 1.33 1.22 1.30 1.32 1.33 1.17 2.03 1.73 1.09 2.35 (0.89–1.99) (0.79–1.90) (0.87–1.92) (0.59–2.96) (0.77–2.30) (0.65–2.11) (0.91–4.56) (0.57–5.27) (0.51–2.33) (0.57–9.69)  Midwestc 1.24 1.17 1.18 0.88 1.73 2.06 0.99 1.95 0.71 1.26 (0.73–2.10) (0.75–1.81) (0.72–1.91) (0.37–2.11) (0.92–3.25) (0.62–6.89) (0.34–2.86) (0.33–11.53) (0.33–1.51) (0.34–4.66)  Southc 2.06*** 1.88*** 1.79** 1.54 2.16** 2.34 1.31 1.68 1.00 1.95 (1.32–3.20) (1.29–2.73) (1.18–2.71) (0.72–3.32) (1.28–3.63) (0.76–7.16) (0.43–4.02) (0.33–8.65) (0.52–1.91) (0.56–6.77)  Westc 2.36*** 2.05** 2.33** 2.10 2.89*** 4.27* 2.00 1.22 1.28 1.92 (1.41–3.95) (1.24–3.39) (1.35–4.02) (0.62–7.12) (1.68–4.97) (1.09–16.72) (0.63–6.31) (0.28–5.24) (0.61–2.66) (0.49–7.61)  County pop. size 1.05 1.04 1.05 1.08 1.01 1.10 1.03 1.13 1.08 1.00 (0.96–1.15) (0.96–1.13) (0.96–1.15) (0.90–1.30) (0.90–1.14) (0.98–1.23) (0.89–1.18) (0.92–1.41) (0.98–1.19) (0.80–1.25)  Education 1.06* 1.11*** 1.08* 1.12 1.05 1.23** 1.01 1.27* 1.15** 1.13 (1.01–1.12) (1.04–1.18) (1.01–1.15) (0.96–1.31) (0.97–1.14) (1.06–1.43) (0.89–1.16) (1.02–1.59) (1.04–1.28) (0.94–1.37)  Married 0.94 1.02 0.96 1.05 0.84 1.41 0.96 1.31 1.13 0.54 (0.69–1.26) (0.76–1.38) (0.72–1.28) (0.63–1.73) (0.60–1.19) (0.83–2.39) (0.42–2.18) (0.59–2.91) (0.68–1.87) (0.23–1.26)  Worker 1.21 1.27 1.23 1.48 1.17 1.04 0.58 0.86 1.55 0.91 (0.83–1.75) (0.94–1.72) (0.90–1.69) (0.68–3.21) (0.69–1.98) (0.45–2.40) (0.26–1.28) (0.36–2.07) (0.84–2.86) (0.40–2.07) Relational-level predictors  Religious homophily 1.55*** 1.57** 1.58** 1.83*** 1.39 3.10*** 1.28 6.49*** 1.23 1.24 (1.24–1.93) (1.26–1.97) (1.27–1.96) (1.32–2.55) (0.99–1.94) (2.07–4.64) (0.71–2.32) (2.22–18.94) (0.75–2.01) (0.75–2.05)  Alter rank 1 1.88*** 1.87*** 1.88** 2.06*** 1.74** 1.61 1.33 1.94 3.81*** 1.46 (1.51–2.36) (1.50–2.34) (1.50–2.35) (1.46–2.90) (1.29–2.34) (0.86–3.00) (0.80–2.23) (0.97–3.88) (2.64–5.49) (0.76–2.82)  Alter rank 2 1.16 1.15 1.16 1.17 1.15 0.88 0.95 1.93 2.14*** 0.86 (0.92–1.45) (0.92–1.44) (0.92–1.45) (0.86–1.60) (0.84–1.59) (0.57–1.35) (0.51–1.77) (0.93–4.00) (1.47–3.12) (0.45–1.62)  Alter rank 3 0.99 0.99 0.99 1.07 0.92 0.81 1.15 1.51 1.19 0.78 (0.81–1.22) (0.80–1.21) (0.81–1.22) (0.78–1.49) (0.70–1.20) (0.47–1.39) (0.67–1.99) (0.86–2.67) (0.78–1.82) (0.46–1.30)  Alter gave advice 2.79*** 2.76*** 2.87** 3.09** 2.78*** 3.07*** 3.28*** 4.51*** 2.91*** 2.43** (2.21–3.52) (2.19–3.47) (2.26–3.65) (2.00–4.76) (2.04–3.79) (1.70–5.54) (2.00–5.37) (1.89–10.77) (1.59–5.30) (1.39–4.25)  Alter gave help 1.86*** 1.91*** 1.87** 1.90** 1.76*** 2.37*** 2.27* 1.80 1.36 2.10 (1.47–2.36) (1.48–2.47) (1.48–2.36) (1.29–2.80) (1.25–2.47) (1.51–3.73) (1.15–4.49) (0.91–3.56) (0.78–2.36) (0.89–4.94)  Lives nearby 1.40*** 1.43*** 1.41** 1.45* 1.35 1.44 1.10 1.26 1.39 1.20 (1.15–1.70) (1.18–1.74) (1.16–1.72) (1.07–1.96) (0.99–1.84) (0.90–2.28) (0.64–1.87) (0.62–2.57) (0.97–1.98) (0.65–2.23)  Interact weekly 1.73*** 1.74*** 1.71** 1.99** 1.56** 2.31** 2.01* 1.53 1.08 2.67** (1.32–2.26) (1.32–2.29) (1.31–2.24) (1.29–3.10) (1.14–2.12) (1.28–4.16) (1.00–4.05) (0.74–3.16) (0.65–1.78) (1.31–5.43)  Racial homophily 0.78 0.81 0.78 1.00 0.74 1.50 0.90 0.95 0.72 0.88 (0.50–1.20) (0.52–1.25) (0.50–1.20) (0.52–1.96) (0.45–1.23) (0.85–2.65) (0.34–2.38) (0.19–4.77) (0.22–2.33) (0.40–1.96)  Gender homophily 0.90 0.92 0.92 1.00 0.86 0.77 0.94 1.05 0.81 1.18 (0.72–1.14) (0.73–1.15) (0.73–1.16) (0.69–1.43) (0.66–1.12) (0.52–1.15) (0.58–1.54) (0.50–2.21) (0.47–1.38) (0.76–1.83)  Alter has bachelor dg. 1.32* 1.32* 1.32* 0.94 1.78*** 0.94 1.72 2.36* 1.03 1.70 (1.06–1.66) (1.05–1.65) (1.05–1.67) (0.67–1.31) (1.27–2.49) (0.59–1.52) (0.98–3.02) (1.07–5.24) (0.65–1.61) (0.85–3.40)  Kin 0.84 0.83 0.85 0.98 0.80 0.79 1.32 0.97 1.16 0.46* (0.69–1.01) (0.68–1.01) (0.71–1.02) (0.71–1.35) (0.60–1.05) (0.48–1.30) (0.75–2.32) (0.57–1.66) (0.79–1.70) (0.21–0.99)  Co-worker 0.90 0.89 0.89 1.20 0.79 0.66 1.45 2.45 0.96 0.80 (0.65–1.24) (0.64–1.26) (0.64–1.24) (0.70–2.07) (0.54–1.14) (0.30–1.45) (0.64–3.26) (0.20–29.47) (0.50–1.84) (0.33–1.93)  Level 1 5.23 5.45 5.17 6.48 4.24 7.26 3.18 5.37 5.73 4.10  Variance (0.50) (0.53) (0.51) (0.95) (0.56) (2.25) (0.71) (2.84) (1.27) (0.95)  −2LL 4841.96 4886.24 4821.47 2195.57 2591.21 1227.96 741.20 489.75 1189.65 788.78 Observations  Level 1 8,944 8,945 8,944 4,139 4,805 1,985 1,026 926 2,469 1,418  Level 2 2,419 2,420 2,419 1,102 1,317 529 270 256 683 384 Full sample Part of congregation Not part of congregation Evangelicals Mainline Protestants Black Protestants Catholics Religious nones Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Model 8 Model 9 Model 10 Individual-level predictors  Part of cong. 3.79*** 4.23** 6.81*** 2.80*** 3.17** 3.00** 1.02 (2.84–5.07) (3.20–5.59) (3.51–13.21) (1.65–4.75) (1.34–7.53) (1.51–5.95) (0.32–3.24)  Evangelicala 1.84* 0.95 4.54* 0.64 (1.14–2.99) (0.60–1.49) (1.18–17.50) (0.37–1.12)  Black Protestanta 1.52 0.81 3.02 0.71 (0.82–2.83) (0.42–1.53) (0.66–13.78) (0.32–1.60)  Mainline Protestanta 0.66 0.34** 1.13 0.41** (0.42–1.03) (0.22–0.52) (0.32–4.00) (0.23–0.73)  Catholica 0.92 0.50** 1.82 0.49* - (0.56–1.51) (0.31–0.82) (0.42–7.88) (0.26–0.94)  Other 0.88 0.66 4.80* 0.50*  Religiona (0.44–1.77) (0.36–1.22) (1.26–18.32) (0.27–0.91)  Female 1.81*** 2.40*** 1.94** 1.35 2.49*** 1.91 2.56* 0.75 3.15** 1.00 (1.30–2.51) (1.60–3.58) (1.40–2.70) (0.70–2.59) (1.66–3.73) (0.95–3.83) (1.24–5.28) (0.21–2.64) (1.38–7.21) (0.39–2.58)  Blackb 0.90 0.88 0.77 0.60 1.16 0.63 0.81 0.41 0.86 (0.53–1.54) (0.51–1.51) (0.44–1.34) (0.30–1.20) (0.55–2.47) (0.23–1.73) (0.19–3.41) (0.16–1.06) (0.27–2.78)  Latino/ab 0.33*** 0.39*** 0.39** 0.25* 0.54 0.34** 0.64 0.45* 0.43 (0.21–0.52) (0.23–0.64) (0.24–0.64) (0.08–0.76) (0.28–1.06) (0.17–0.67) (0.15–2.68) (0.23–0.87) (0.10–1.80)  Other raceb 0.19*** 0.20*** 0.17** 0.13*** 0.18*** 0.14*** 0.37 0.14*** 0.55 (0.08–0.43) (0.08–0.46) (0.08–0.37) (0.05–0.35) (0.08–0.43) (0.06–0.33) (0.04–3.25) (0.05–0.40) (0.12–2.57)  Age 0.99 1.00 1.00 1.00 0.99 0.98 0.99 0.99 1.01 1.00 (0.98–1.00) (0.99–1.01) (0.99–1.01) (0.98–1.02) (0.98–1.01) (0.96–1.02) (0.96–1.02) (0.96–1.03) (0.99–1.03) (0.97–1.03)  Serious illness/injury 1.17 1.03 1.14 1.06 1.29 1.07 0.95 2.12 1.56 0.82 (0.78–1.77) (0.68–1.57) (0.78–1.68) (0.45–2.46) (0.83–1.99) (0.47–2.46) (0.40–2.27) (0.73–6.18) (0.81–3.01) (0.35–1.91)  Death of loved one 1.26 1.36 1.21 1.10 1.31 0.79 0.94 2.28 1.80* 1.04 (0.89–1.79) (0.97–1.90) (0.85–1.72) (0.61–1.99) (0.85–2.02) (0.42–1.52) (0.52–1.71) (0.82–6.33) (1.00–3.25) (0.43–2.53)  Financial crisis 1.33 1.22 1.30 1.32 1.33 1.17 2.03 1.73 1.09 2.35 (0.89–1.99) (0.79–1.90) (0.87–1.92) (0.59–2.96) (0.77–2.30) (0.65–2.11) (0.91–4.56) (0.57–5.27) (0.51–2.33) (0.57–9.69)  Midwestc 1.24 1.17 1.18 0.88 1.73 2.06 0.99 1.95 0.71 1.26 (0.73–2.10) (0.75–1.81) (0.72–1.91) (0.37–2.11) (0.92–3.25) (0.62–6.89) (0.34–2.86) (0.33–11.53) (0.33–1.51) (0.34–4.66)  Southc 2.06*** 1.88*** 1.79** 1.54 2.16** 2.34 1.31 1.68 1.00 1.95 (1.32–3.20) (1.29–2.73) (1.18–2.71) (0.72–3.32) (1.28–3.63) (0.76–7.16) (0.43–4.02) (0.33–8.65) (0.52–1.91) (0.56–6.77)  Westc 2.36*** 2.05** 2.33** 2.10 2.89*** 4.27* 2.00 1.22 1.28 1.92 (1.41–3.95) (1.24–3.39) (1.35–4.02) (0.62–7.12) (1.68–4.97) (1.09–16.72) (0.63–6.31) (0.28–5.24) (0.61–2.66) (0.49–7.61)  County pop. size 1.05 1.04 1.05 1.08 1.01 1.10 1.03 1.13 1.08 1.00 (0.96–1.15) (0.96–1.13) (0.96–1.15) (0.90–1.30) (0.90–1.14) (0.98–1.23) (0.89–1.18) (0.92–1.41) (0.98–1.19) (0.80–1.25)  Education 1.06* 1.11*** 1.08* 1.12 1.05 1.23** 1.01 1.27* 1.15** 1.13 (1.01–1.12) (1.04–1.18) (1.01–1.15) (0.96–1.31) (0.97–1.14) (1.06–1.43) (0.89–1.16) (1.02–1.59) (1.04–1.28) (0.94–1.37)  Married 0.94 1.02 0.96 1.05 0.84 1.41 0.96 1.31 1.13 0.54 (0.69–1.26) (0.76–1.38) (0.72–1.28) (0.63–1.73) (0.60–1.19) (0.83–2.39) (0.42–2.18) (0.59–2.91) (0.68–1.87) (0.23–1.26)  Worker 1.21 1.27 1.23 1.48 1.17 1.04 0.58 0.86 1.55 0.91 (0.83–1.75) (0.94–1.72) (0.90–1.69) (0.68–3.21) (0.69–1.98) (0.45–2.40) (0.26–1.28) (0.36–2.07) (0.84–2.86) (0.40–2.07) Relational-level predictors  Religious homophily 1.55*** 1.57** 1.58** 1.83*** 1.39 3.10*** 1.28 6.49*** 1.23 1.24 (1.24–1.93) (1.26–1.97) (1.27–1.96) (1.32–2.55) (0.99–1.94) (2.07–4.64) (0.71–2.32) (2.22–18.94) (0.75–2.01) (0.75–2.05)  Alter rank 1 1.88*** 1.87*** 1.88** 2.06*** 1.74** 1.61 1.33 1.94 3.81*** 1.46 (1.51–2.36) (1.50–2.34) (1.50–2.35) (1.46–2.90) (1.29–2.34) (0.86–3.00) (0.80–2.23) (0.97–3.88) (2.64–5.49) (0.76–2.82)  Alter rank 2 1.16 1.15 1.16 1.17 1.15 0.88 0.95 1.93 2.14*** 0.86 (0.92–1.45) (0.92–1.44) (0.92–1.45) (0.86–1.60) (0.84–1.59) (0.57–1.35) (0.51–1.77) (0.93–4.00) (1.47–3.12) (0.45–1.62)  Alter rank 3 0.99 0.99 0.99 1.07 0.92 0.81 1.15 1.51 1.19 0.78 (0.81–1.22) (0.80–1.21) (0.81–1.22) (0.78–1.49) (0.70–1.20) (0.47–1.39) (0.67–1.99) (0.86–2.67) (0.78–1.82) (0.46–1.30)  Alter gave advice 2.79*** 2.76*** 2.87** 3.09** 2.78*** 3.07*** 3.28*** 4.51*** 2.91*** 2.43** (2.21–3.52) (2.19–3.47) (2.26–3.65) (2.00–4.76) (2.04–3.79) (1.70–5.54) (2.00–5.37) (1.89–10.77) (1.59–5.30) (1.39–4.25)  Alter gave help 1.86*** 1.91*** 1.87** 1.90** 1.76*** 2.37*** 2.27* 1.80 1.36 2.10 (1.47–2.36) (1.48–2.47) (1.48–2.36) (1.29–2.80) (1.25–2.47) (1.51–3.73) (1.15–4.49) (0.91–3.56) (0.78–2.36) (0.89–4.94)  Lives nearby 1.40*** 1.43*** 1.41** 1.45* 1.35 1.44 1.10 1.26 1.39 1.20 (1.15–1.70) (1.18–1.74) (1.16–1.72) (1.07–1.96) (0.99–1.84) (0.90–2.28) (0.64–1.87) (0.62–2.57) (0.97–1.98) (0.65–2.23)  Interact weekly 1.73*** 1.74*** 1.71** 1.99** 1.56** 2.31** 2.01* 1.53 1.08 2.67** (1.32–2.26) (1.32–2.29) (1.31–2.24) (1.29–3.10) (1.14–2.12) (1.28–4.16) (1.00–4.05) (0.74–3.16) (0.65–1.78) (1.31–5.43)  Racial homophily 0.78 0.81 0.78 1.00 0.74 1.50 0.90 0.95 0.72 0.88 (0.50–1.20) (0.52–1.25) (0.50–1.20) (0.52–1.96) (0.45–1.23) (0.85–2.65) (0.34–2.38) (0.19–4.77) (0.22–2.33) (0.40–1.96)  Gender homophily 0.90 0.92 0.92 1.00 0.86 0.77 0.94 1.05 0.81 1.18 (0.72–1.14) (0.73–1.15) (0.73–1.16) (0.69–1.43) (0.66–1.12) (0.52–1.15) (0.58–1.54) (0.50–2.21) (0.47–1.38) (0.76–1.83)  Alter has bachelor dg. 1.32* 1.32* 1.32* 0.94 1.78*** 0.94 1.72 2.36* 1.03 1.70 (1.06–1.66) (1.05–1.65) (1.05–1.67) (0.67–1.31) (1.27–2.49) (0.59–1.52) (0.98–3.02) (1.07–5.24) (0.65–1.61) (0.85–3.40)  Kin 0.84 0.83 0.85 0.98 0.80 0.79 1.32 0.97 1.16 0.46* (0.69–1.01) (0.68–1.01) (0.71–1.02) (0.71–1.35) (0.60–1.05) (0.48–1.30) (0.75–2.32) (0.57–1.66) (0.79–1.70) (0.21–0.99)  Co-worker 0.90 0.89 0.89 1.20 0.79 0.66 1.45 2.45 0.96 0.80 (0.65–1.24) (0.64–1.26) (0.64–1.24) (0.70–2.07) (0.54–1.14) (0.30–1.45) (0.64–3.26) (0.20–29.47) (0.50–1.84) (0.33–1.93)  Level 1 5.23 5.45 5.17 6.48 4.24 7.26 3.18 5.37 5.73 4.10  Variance (0.50) (0.53) (0.51) (0.95) (0.56) (2.25) (0.71) (2.84) (1.27) (0.95)  −2LL 4841.96 4886.24 4821.47 2195.57 2591.21 1227.96 741.20 489.75 1189.65 788.78 Observations  Level 1 8,944 8,945 8,944 4,139 4,805 1,985 1,026 926 2,469 1,418  Level 2 2,419 2,420 2,419 1,102 1,317 529 270 256 683 384 Notes: Robust standard errors are shown in parentheses. aReligions none is reference group. bWhite is reference group; Race/ethnicity not included in model 8 because the overwhelming majority of Black Protestants are Black. cNortheast is reference group. *p < .05; **p < .01; ***p < .001 (two-tailed tests). View Large Across the various models, being part of a congregation is a strong predictor of talking religion with a given network tie. This is true without accounting for affiliation (model 1), after adjusting for affiliation (model 3), and among Evangelicals, Black Protestants, Mainline Protestants, and Catholics. Evangelical affiliation also seems to dictate whether people talk religion (model 2), but this can be explained by the fact that Evangelicals have high levels of congregational involvement relative to other groups, particularly religious nones (Evangelical odds ratio is no longer significant in model 3). Additional analysis reveals that over 60% of Evangelicals are connected to a congregation; this figure is higher than all other religious groups. Notably, once accounting for congregational involvement, Catholics and Mainline Protestants are actually less likely to talk religion with a given network tie than are religious nones (model 3). This suppression pattern reflects the fact that both groups are relatively unlikely to report religious discussions, yet both are far more likely than the religious nones to be part of a congregation (fewer than 10% of nones have a congregational involvement)—the factor which itself is the strongest predictor of talking religion. Indeed, Catholics and Mainline Protestants who do belong to a congregation are indeed far more likely to talk religion with a given tie than are their noncongregational counterparts (models 7 and 9). Supplementary analyses considered one other dimension of individual religiosity, self-assessed religious salience (from “not at all important” to “by far the most important part of your life”). This indicator of religiosity tended to perform much like congregational involvement (strongly associated with religious discussion in the main sample and across all affiliation groups), but it was not included in final models because it overlapped to a great degree with the other two aspects of personal religiosity. Among the other individual-level factors, being female is associated with religious discussion (models 1–3), but this pattern is driven mainly by noncongregants, by Catholics, and by Mainline Protestants (see models 5, 7, and 9). The association between gender and talking religion was consistent whether or not individual religiosity was included in the model; this suggests that women’s proclivity to discuss religion is not merely a function of their being more religious than men. In contrast to my expectations, there is no evidence that Black people—or those identifying as Black Protestants—talk religion any more or less than White adults. This was true in models (not shown) that excluded individual religiosity. However, Latino/as were less likely than Whites to report talking religion, particularly among Evangelical and Catholic subgroups. Expectations about age/cohort and stressful life events were not substantiated. None of these factors emerged as significant predictors of religious discussion in the main sample or in any of the subgroups.12 The sole exception was the case of a loved one dying; this experience increased the odds of talking religion only among Catholics. Finally, there is some evidence that Southerners and those living in the West are more likely than Northeasterners to talk religion, and that having higher education is associated with greater odds of discussing religion with a network tie. The latter pattern is most pronounced among Evangelicals, Black Protestants, and Catholics. Each relational theory receives some element of support in table 3. Religious homophily is a strong predictor of religious discussion in the full sample (OR = 1.58; 95% CI = 1.27–1.96). However, subgroup analysis reveals that the pattern is driven largely by people involved in a congregation (model 4) and by Evangelicals and Black Protestants (models 6 and 8). Indicators of tie strength also emerged as strong predictors. Alters named first in the roster had odds of religious discussion about 88% higher than did alters nominated last (models 1–3). Being top of alter list appears to be particularly important among Catholics, increasing the odds of religious discussion by a factor of 3.81 (model 9). Being top-ranked was less consistently influential among the other religious affiliation subgroups. Also of note, the main distinction by roster ordering—where observed—seemed to be between those named first and everyone else; there was little evidence that ordering predicted religious discussion beyond the top-ranked alter for any of the models. Religious discussion also coincided with social support provision, especially advice-giving. Indeed, the link between advice given and religious discussion was remarkably consistent across all respondent subgroups. Alters who provided practical help also tended to be people with whom religion was discussed, though this pattern was somewhat less robust across all subgroups than was advice exchange. Frequent interaction—one aspect of tie strength, but also a reflection of conversational opportunity—boosted the odds of talking religion by about 70% in the main sample. Living within a 20-minute drive of alter also increased odds of religious discussion (about 40% in the main sample), but this estimate was not consistently different from nil across the various subgroup models.13 Though not the focus of the analysis, there was some evidence that people were more likely to talk about religion with highly educated network members, though this pattern was most apparent among those not affiliated with a Congregation and among Black Protestants.14 Tie Persistence To conclude the analysis, I consider one potential network outcome of religious discussion, tie persistence. Descriptive results from table 2 reveal that just over 33% of all initially identified alters persist when PALS respondents were re-interviewed in 2011. Results from the full sample in table 4 indicate that talking religion with an alter in W1 increases the odds by 43% that the alter will re-appear in ego’s W2 close network. This association is net of religious homophily, race and gender homophily, initial tie closeness, and the full set of other relational and individual-level covariates. Religious homophily, in fact, was not significantly associated with tie persistence, even in supplementary models where religious discussion was removed. Further, religious homophily did not significantly interact with religious discussion (results not shown for sake of space). These findings cast doubt on the possibility that mere religious homophily is the reason why talking religion is associated with tie durability. Table 4 Tie Persistence, Odds Ratio Estimates from Random Intercept Logistic Regression Models (Portraits of American Life Survey, W1/W2) Full sample Part of Congregation Not Part of Congregation Evangelicals Mainline Protestants Black Protestants Catholics Religious nones Model 1 Model 2 Model 3 Model 4 Model 6 Model 5 Model 7 Model 8 Discuss religion with alter 1.44* 1.80** 1.24 2.45** 3.11** 1.33 1.24 0.91 (1.05–1.98) (1.22–2.67) (0.83–1.85) (1.39–4.30) (1.43–6.74) (0.40–4.38) (0.79–1.97) (0.47–1.78) Individual level covariates  Part of cong. 0.62** 0.76 0.80 0.62 0.94 0.11*** (0.46–0.84) (0.46–1.25) (0.33–1.94) (0.21–1.85) (0.55–1.60) (0.03–0.41)  Evangelicala 0.97 4.50** 0.71 (0.62–1.51) (1.50–13.50) (0.43–1.19)  Black Protestanta 0.71 3.41 0.60 (0.44–1.15) (0.82–14.15) (0.23–1.58)  Mainline Protestanta 1.16 5.92*** 0.82 (0.65–2.06) (2.06–17.03) (0.31–2.15)  Catholica 0.97 5.43** 0.63* (0.63–1.49) (1.83–16.11) (0.40–0.99)  Other religiona 0.53** 1.58 0.68 (0.33–0.84) (0.48–5.23) (0.43–1.04)  Female 1.31* 1.56* 1.04 1.60* 0.95 1.43 1.18 1.03 (1.01–1.70) (1.06–2.29) (0.79–1.37) (1.05–2.45) (0.50–1.81) (0.48–4.31) (0.68–2.05) (0.64–1.66)  Blackb 0.96 0.88 0.93 1.12 0.38 0.49 0.81 (0.67–1.39) (0.42–1.87) (0.53–1.63) (0.54–2.34) (0.05–3.10) (0.16–1.53) (0.38–1.74)  Latino/ab 1.03 1.12 1.00 1.95 0.17* 0.82 0.67 (0.69–1.54) (0.59–2.15) (0.51–1.97) (0.76–4.97) (0.04–0.73) (0.41–1.66) (0.22–2.04)  Other raceb 0.67 0.38 0.96 1.95 0.22* 0.34 0.69 (0.33–1.33) (0.11–1.34) (0.42–2.22) (0.22–17.52) (0.05–0.86) (0.09–1.19) (0.14–3.33)  Age 1.01 0.99 1.02** 1.00 0.96* 1.00 1.01 1.03* (1.00–1.02) (0.98–1.01) (1.01–1.04) (0.99–1.02) (0.93–0.99) (0.97–1.04) (1.00–1.03) (1.00–1.06)  Serious illness/injury 0.78 0.89 0.62 1.27 0.35 0.61 0.79 0.44** (0.55–1.12) (0.54–1.45) (0.39–1.00) (0.74–2.19) (0.11–1.07) (0.16–2.28) (0.50–1.26) (0.25–0.77)  Death of loved one 0.80 1.17 0.62* 0.61 0.50* 0.51 0.88 0.74 (0.63–1.02) (0.82–1.66) (0.42–0.90) (0.37–1.00) (0.25–0.99) (0.20–1.34) (0.55–1.39) (0.41–1.35)  Financial crisis 1.26 1.49 1.08 1.20 0.56 1.16 1.43 1.02 (0.96–1.65) (0.98–2.27) (0.71–1.63) (0.74–1.92) (0.24–1.32) (0.50–2.70) (0.71–2.85) (0.52–2.01)  Midwestc 0.61** 0.61* 0.64 0.47* 1.09 0.55 0.59 0.83 (0.42–0.89) (0.38–0.96) (0.36–1.14) (0.23–0.98) (0.31–3.79) (0.20–1.53) (0.29–1.19) (0.31–2.23)  Southc 0.65** 0.52** 0.84 0.56 1.01 1.07 0.72 0.85 (0.47–0.89) (0.33–0.80) (0.59–1.19) (0.28–1.11) (0.29–3.52) (0.34–3.33) (0.36–1.47) (0.41–1.80)  Westc 0.84 0.81 0.85 0.76 1.20 0.57 0.79 0.93 (0.58–1.20) (0.41–1.60) (0.58–1.14) (0.26–2.23) (0.35–4.12) (0.19–1.74) (0.36–1.75) (0.39–2.20)  County pop. size 0.99 1.04 0.98 0.97 1.04 1.08 0.97 1.05 (0.93–1.07) (0.96–1.14) (0.89–1.07) (0.85–1.10) (0.89–1.22) (0.89–1.33) (0.87–1.08) (0.92–1.18)  Education 1.08** 1.11*** 1.00 0.97 0.99 1.06 1.09 1.03 (1.02–1.13) (1.04–1.18) (0.93–1.07) (0.86–1.09) (0.88–1.12) (0.94–1.19) (0.98–1.21) (0.92–1.15)  Married 0.95 1.22 0.76 1.10 0.49 0.56 0.90 1.10 (0.70–1.28) (0.77–1.94) (0.54–1.09) (0.64–1.90) (0.18–1.32) (0.30–1.07) (0.55–1.45) (0.64–1.88)  Worker 0.91 0.77 0.96 0.57* 0.80 1.18 1.28 1.35 (0.71–1.16) (0.53–1.12) (0.68–1.37) (0.33–1.00) (0.45–1.42) (0.39–3.63) (0.80–2.04) (0.81–2.26)  Moved (W1–W2) 0.88 0.86 0.84 0.86 0.67 0.61 0.77 0.85 (0.63–1.21) (0.59–1.25) (0.55–1.27) (0.54–1.35) (0.23–1.94) (0.22–1.68) (0.46–1.26) (0.44–1.64)  Change in marital status (W1–W2) 0.83 0.90 0.76 0.34** 0.73 1.13 0.99 0.84 (0.62–1.10) (0.60–1.36) (0.53–1.07) (0.17–0.66) (0.30–1.78) (0.41–3.14) (0.64–1.54) (0.41–1.70)  Change in work status (W1–W2) 0.84 0.63* 1.05 0.68 0.22*** 1.65 1.08 1.73* (0.58–1.20) (0.41–0.96) (0.68–1.61) (0.38–1.24) (0.10–0.50) (0.56–4.88) (0.47–2.46) (1.05–2.85) Relational level covariates  Religious homophily 1.08 1.16 1.01 1.01 1.33 0.66 1.40 1.27 (0.83–1.39) (0.84–1.61) (0.69–1.48) (0.56–1.81) (0.66–2.70) (0.33–1.31) (0.92–2.15) (0.83–1.94)  Alter rank 1 3.68*** 3.25*** 3.94*** 4.34*** 1.89 2.71 4.14*** 4.80*** (2.61–5.18) (2.19–4.83) (2.34–6.63) (2.32–8.11) (0.66–5.39) (0.93–7.85) (2.34–7.31) (2.37–9.71)  Alter rank 2 2.05*** 1.84*** 2.20*** 1.91 1.65 2.37* 2.40** 1.70 (1.56–2.69) (1.29–2.63) (1.42–3.41) (0.97–3.74) (0.70–3.89) (1.01–5.56) (1.35–4.24) (0.96–3.00)  Alter rank 3 1.36* 1.29 1.42 1.18 0.81 1.52 1.93** 1.50 (1.04–1.78) (0.90–1.83) (0.91–2.21) (0.64–2.17) (0.35–1.91) (0.44–5.19) (1.17–3.18) (0.73–3.07)  Alter gave advice 1.62** 1.33 1.97*** 1.44 0.67 1.18 1.09 2.73*** (1.20–2.18) (0.85–2.08) (1.39–2.79) (0.81–2.56) (0.26–1.72) (0.67–2.05) (0.60–1.99) (1.47–5.07)  Alter gave help 1.27 1.47 1.19 0.99 0.66 0.86 2.12** 1.21 (0.92–1.74) (0.92–2.36) (0.74–1.92) (0.58–1.69) (0.28–1.57) (0.41–1.78) (1.23–3.67) (0.56–2.63)  Lives nearby 1.09 1.07 1.14 0.89 1.01 1.93* 1.19 1.14 (0.83–1.43) (0.73–1.57) (0.82–1.59) (0.57–1.38) (0.49–2.10) (1.16–3.23) (0.76–1.86) (0.67–1.94)  Interact weekly 1.32 1.24 1.39 1.46 3.64** 1.58 1.51 0.67 (0.96–1.81) (0.86–1.77) (0.91–2.15) (0.81–2.62) (1.45–9.13) (0.67–3.72) (0.84–2.71) (0.34–1.31)  Racial homophily 1.95** 2.20* 1.58 0.88 3.56 8.40 1.05 2.28* (1.24–3.06) (1.10–4.40) (0.90–2.77) (0.27–2.86) (0.44–28.68) (0.36–197.71) (0.48–2.28) (1.07–4.88)  Gender homophily 1.26* 1.48* 1.14 4.33 2.31* 0.77 0.94 1.21 (1.01–1.56) (1.03–2.11) (0.88–1.47) (1.61– 11.67)** (1.08–4.92) (0.30–1.97) (0.61–1.45) (0.78–1.87)  Alter has bachelor dg. 0.93 0.90 1.04 1.14 0.90 1.77 0.63* 1.10 (0.76–1.13) (0.68–1.20) (0.79–1.37) (0.67–1.94) (0.45–1.80) (0.73–4.31) (0.39–1.00) (0.61–1.99)  Kin 2.45*** 1.91** 2.88*** 0.83 3.58*** 0.90 1.84*** 2.32* (1.90–3.17) (1.27–2.87) (1.99–4.15) (0.47–1.48) (1.68–7.64) (0.40–2.02) (1.27–2.68) (1.22–4.44)  Co-worker 0.80 0.62 0.97 2.45 0.14** 0.96 0.93 1.70 (0.48–1.35) (0.33–1.18) (0.52–1.80) (1.52– 3.95)** (0.04–0.57) (0.27–3.34) (0.42–2.06) (0.68–4.24)  Level 1 variance 1.25 1.24 1.02 0.62 1.21 1.15 0.97 0.56 (0.17) (0.24) (0.21) (0.31) (0.42) (0.67) (0.31) (0.34)  −2LL 859,928,640 390,913,472 452,567,008 222,713,520 88,179,943 19,448,486 230,494,992 114,654,792 Observations  Level 1 4619 2224 2395 1031 518 385 1223 752  Level 2 1233 583 650 273 134 106 331 202 Full sample Part of Congregation Not Part of Congregation Evangelicals Mainline Protestants Black Protestants Catholics Religious nones Model 1 Model 2 Model 3 Model 4 Model 6 Model 5 Model 7 Model 8 Discuss religion with alter 1.44* 1.80** 1.24 2.45** 3.11** 1.33 1.24 0.91 (1.05–1.98) (1.22–2.67) (0.83–1.85) (1.39–4.30) (1.43–6.74) (0.40–4.38) (0.79–1.97) (0.47–1.78) Individual level covariates  Part of cong. 0.62** 0.76 0.80 0.62 0.94 0.11*** (0.46–0.84) (0.46–1.25) (0.33–1.94) (0.21–1.85) (0.55–1.60) (0.03–0.41)  Evangelicala 0.97 4.50** 0.71 (0.62–1.51) (1.50–13.50) (0.43–1.19)  Black Protestanta 0.71 3.41 0.60 (0.44–1.15) (0.82–14.15) (0.23–1.58)  Mainline Protestanta 1.16 5.92*** 0.82 (0.65–2.06) (2.06–17.03) (0.31–2.15)  Catholica 0.97 5.43** 0.63* (0.63–1.49) (1.83–16.11) (0.40–0.99)  Other religiona 0.53** 1.58 0.68 (0.33–0.84) (0.48–5.23) (0.43–1.04)  Female 1.31* 1.56* 1.04 1.60* 0.95 1.43 1.18 1.03 (1.01–1.70) (1.06–2.29) (0.79–1.37) (1.05–2.45) (0.50–1.81) (0.48–4.31) (0.68–2.05) (0.64–1.66)  Blackb 0.96 0.88 0.93 1.12 0.38 0.49 0.81 (0.67–1.39) (0.42–1.87) (0.53–1.63) (0.54–2.34) (0.05–3.10) (0.16–1.53) (0.38–1.74)  Latino/ab 1.03 1.12 1.00 1.95 0.17* 0.82 0.67 (0.69–1.54) (0.59–2.15) (0.51–1.97) (0.76–4.97) (0.04–0.73) (0.41–1.66) (0.22–2.04)  Other raceb 0.67 0.38 0.96 1.95 0.22* 0.34 0.69 (0.33–1.33) (0.11–1.34) (0.42–2.22) (0.22–17.52) (0.05–0.86) (0.09–1.19) (0.14–3.33)  Age 1.01 0.99 1.02** 1.00 0.96* 1.00 1.01 1.03* (1.00–1.02) (0.98–1.01) (1.01–1.04) (0.99–1.02) (0.93–0.99) (0.97–1.04) (1.00–1.03) (1.00–1.06)  Serious illness/injury 0.78 0.89 0.62 1.27 0.35 0.61 0.79 0.44** (0.55–1.12) (0.54–1.45) (0.39–1.00) (0.74–2.19) (0.11–1.07) (0.16–2.28) (0.50–1.26) (0.25–0.77)  Death of loved one 0.80 1.17 0.62* 0.61 0.50* 0.51 0.88 0.74 (0.63–1.02) (0.82–1.66) (0.42–0.90) (0.37–1.00) (0.25–0.99) (0.20–1.34) (0.55–1.39) (0.41–1.35)  Financial crisis 1.26 1.49 1.08 1.20 0.56 1.16 1.43 1.02 (0.96–1.65) (0.98–2.27) (0.71–1.63) (0.74–1.92) (0.24–1.32) (0.50–2.70) (0.71–2.85) (0.52–2.01)  Midwestc 0.61** 0.61* 0.64 0.47* 1.09 0.55 0.59 0.83 (0.42–0.89) (0.38–0.96) (0.36–1.14) (0.23–0.98) (0.31–3.79) (0.20–1.53) (0.29–1.19) (0.31–2.23)  Southc 0.65** 0.52** 0.84 0.56 1.01 1.07 0.72 0.85 (0.47–0.89) (0.33–0.80) (0.59–1.19) (0.28–1.11) (0.29–3.52) (0.34–3.33) (0.36–1.47) (0.41–1.80)  Westc 0.84 0.81 0.85 0.76 1.20 0.57 0.79 0.93 (0.58–1.20) (0.41–1.60) (0.58–1.14) (0.26–2.23) (0.35–4.12) (0.19–1.74) (0.36–1.75) (0.39–2.20)  County pop. size 0.99 1.04 0.98 0.97 1.04 1.08 0.97 1.05 (0.93–1.07) (0.96–1.14) (0.89–1.07) (0.85–1.10) (0.89–1.22) (0.89–1.33) (0.87–1.08) (0.92–1.18)  Education 1.08** 1.11*** 1.00 0.97 0.99 1.06 1.09 1.03 (1.02–1.13) (1.04–1.18) (0.93–1.07) (0.86–1.09) (0.88–1.12) (0.94–1.19) (0.98–1.21) (0.92–1.15)  Married 0.95 1.22 0.76 1.10 0.49 0.56 0.90 1.10 (0.70–1.28) (0.77–1.94) (0.54–1.09) (0.64–1.90) (0.18–1.32) (0.30–1.07) (0.55–1.45) (0.64–1.88)  Worker 0.91 0.77 0.96 0.57* 0.80 1.18 1.28 1.35 (0.71–1.16) (0.53–1.12) (0.68–1.37) (0.33–1.00) (0.45–1.42) (0.39–3.63) (0.80–2.04) (0.81–2.26)  Moved (W1–W2) 0.88 0.86 0.84 0.86 0.67 0.61 0.77 0.85 (0.63–1.21) (0.59–1.25) (0.55–1.27) (0.54–1.35) (0.23–1.94) (0.22–1.68) (0.46–1.26) (0.44–1.64)  Change in marital status (W1–W2) 0.83 0.90 0.76 0.34** 0.73 1.13 0.99 0.84 (0.62–1.10) (0.60–1.36) (0.53–1.07) (0.17–0.66) (0.30–1.78) (0.41–3.14) (0.64–1.54) (0.41–1.70)  Change in work status (W1–W2) 0.84 0.63* 1.05 0.68 0.22*** 1.65 1.08 1.73* (0.58–1.20) (0.41–0.96) (0.68–1.61) (0.38–1.24) (0.10–0.50) (0.56–4.88) (0.47–2.46) (1.05–2.85) Relational level covariates  Religious homophily 1.08 1.16 1.01 1.01 1.33 0.66 1.40 1.27 (0.83–1.39) (0.84–1.61) (0.69–1.48) (0.56–1.81) (0.66–2.70) (0.33–1.31) (0.92–2.15) (0.83–1.94)  Alter rank 1 3.68*** 3.25*** 3.94*** 4.34*** 1.89 2.71 4.14*** 4.80*** (2.61–5.18) (2.19–4.83) (2.34–6.63) (2.32–8.11) (0.66–5.39) (0.93–7.85) (2.34–7.31) (2.37–9.71)  Alter rank 2 2.05*** 1.84*** 2.20*** 1.91 1.65 2.37* 2.40** 1.70 (1.56–2.69) (1.29–2.63) (1.42–3.41) (0.97–3.74) (0.70–3.89) (1.01–5.56) (1.35–4.24) (0.96–3.00)  Alter rank 3 1.36* 1.29 1.42 1.18 0.81 1.52 1.93** 1.50 (1.04–1.78) (0.90–1.83) (0.91–2.21) (0.64–2.17) (0.35–1.91) (0.44–5.19) (1.17–3.18) (0.73–3.07)  Alter gave advice 1.62** 1.33 1.97*** 1.44 0.67 1.18 1.09 2.73*** (1.20–2.18) (0.85–2.08) (1.39–2.79) (0.81–2.56) (0.26–1.72) (0.67–2.05) (0.60–1.99) (1.47–5.07)  Alter gave help 1.27 1.47 1.19 0.99 0.66 0.86 2.12** 1.21 (0.92–1.74) (0.92–2.36) (0.74–1.92) (0.58–1.69) (0.28–1.57) (0.41–1.78) (1.23–3.67) (0.56–2.63)  Lives nearby 1.09 1.07 1.14 0.89 1.01 1.93* 1.19 1.14 (0.83–1.43) (0.73–1.57) (0.82–1.59) (0.57–1.38) (0.49–2.10) (1.16–3.23) (0.76–1.86) (0.67–1.94)  Interact weekly 1.32 1.24 1.39 1.46 3.64** 1.58 1.51 0.67 (0.96–1.81) (0.86–1.77) (0.91–2.15) (0.81–2.62) (1.45–9.13) (0.67–3.72) (0.84–2.71) (0.34–1.31)  Racial homophily 1.95** 2.20* 1.58 0.88 3.56 8.40 1.05 2.28* (1.24–3.06) (1.10–4.40) (0.90–2.77) (0.27–2.86) (0.44–28.68) (0.36–197.71) (0.48–2.28) (1.07–4.88)  Gender homophily 1.26* 1.48* 1.14 4.33 2.31* 0.77 0.94 1.21 (1.01–1.56) (1.03–2.11) (0.88–1.47) (1.61– 11.67)** (1.08–4.92) (0.30–1.97) (0.61–1.45) (0.78–1.87)  Alter has bachelor dg. 0.93 0.90 1.04 1.14 0.90 1.77 0.63* 1.10 (0.76–1.13) (0.68–1.20) (0.79–1.37) (0.67–1.94) (0.45–1.80) (0.73–4.31) (0.39–1.00) (0.61–1.99)  Kin 2.45*** 1.91** 2.88*** 0.83 3.58*** 0.90 1.84*** 2.32* (1.90–3.17) (1.27–2.87) (1.99–4.15) (0.47–1.48) (1.68–7.64) (0.40–2.02) (1.27–2.68) (1.22–4.44)  Co-worker 0.80 0.62 0.97 2.45 0.14** 0.96 0.93 1.70 (0.48–1.35) (0.33–1.18) (0.52–1.80) (1.52– 3.95)** (0.04–0.57) (0.27–3.34) (0.42–2.06) (0.68–4.24)  Level 1 variance 1.25 1.24 1.02 0.62 1.21 1.15 0.97 0.56 (0.17) (0.24) (0.21) (0.31) (0.42) (0.67) (0.31) (0.34)  −2LL 859,928,640 390,913,472 452,567,008 222,713,520 88,179,943 19,448,486 230,494,992 114,654,792 Observations  Level 1 4619 2224 2395 1031 518 385 1223 752  Level 2 1233 583 650 273 134 106 331 202 Notes: Robust standard errors are shown in parentheses. aReligions none is reference group. bWhite is reference group; Race/ethnicity not included in model 5 because the overwhelming majority of Black Protestants are Black. cNortheast is reference group. ***p < .001; **p < .01; *p < .05 (two-tailed tests). View Large Table 4 Tie Persistence, Odds Ratio Estimates from Random Intercept Logistic Regression Models (Portraits of American Life Survey, W1/W2) Full sample Part of Congregation Not Part of Congregation Evangelicals Mainline Protestants Black Protestants Catholics Religious nones Model 1 Model 2 Model 3 Model 4 Model 6 Model 5 Model 7 Model 8 Discuss religion with alter 1.44* 1.80** 1.24 2.45** 3.11** 1.33 1.24 0.91 (1.05–1.98) (1.22–2.67) (0.83–1.85) (1.39–4.30) (1.43–6.74) (0.40–4.38) (0.79–1.97) (0.47–1.78) Individual level covariates  Part of cong. 0.62** 0.76 0.80 0.62 0.94 0.11*** (0.46–0.84) (0.46–1.25) (0.33–1.94) (0.21–1.85) (0.55–1.60) (0.03–0.41)  Evangelicala 0.97 4.50** 0.71 (0.62–1.51) (1.50–13.50) (0.43–1.19)  Black Protestanta 0.71 3.41 0.60 (0.44–1.15) (0.82–14.15) (0.23–1.58)  Mainline Protestanta 1.16 5.92*** 0.82 (0.65–2.06) (2.06–17.03) (0.31–2.15)  Catholica 0.97 5.43** 0.63* (0.63–1.49) (1.83–16.11) (0.40–0.99)  Other religiona 0.53** 1.58 0.68 (0.33–0.84) (0.48–5.23) (0.43–1.04)  Female 1.31* 1.56* 1.04 1.60* 0.95 1.43 1.18 1.03 (1.01–1.70) (1.06–2.29) (0.79–1.37) (1.05–2.45) (0.50–1.81) (0.48–4.31) (0.68–2.05) (0.64–1.66)  Blackb 0.96 0.88 0.93 1.12 0.38 0.49 0.81 (0.67–1.39) (0.42–1.87) (0.53–1.63) (0.54–2.34) (0.05–3.10) (0.16–1.53) (0.38–1.74)  Latino/ab 1.03 1.12 1.00 1.95 0.17* 0.82 0.67 (0.69–1.54) (0.59–2.15) (0.51–1.97) (0.76–4.97) (0.04–0.73) (0.41–1.66) (0.22–2.04)  Other raceb 0.67 0.38 0.96 1.95 0.22* 0.34 0.69 (0.33–1.33) (0.11–1.34) (0.42–2.22) (0.22–17.52) (0.05–0.86) (0.09–1.19) (0.14–3.33)  Age 1.01 0.99 1.02** 1.00 0.96* 1.00 1.01 1.03* (1.00–1.02) (0.98–1.01) (1.01–1.04) (0.99–1.02) (0.93–0.99) (0.97–1.04) (1.00–1.03) (1.00–1.06)  Serious illness/injury 0.78 0.89 0.62 1.27 0.35 0.61 0.79 0.44** (0.55–1.12) (0.54–1.45) (0.39–1.00) (0.74–2.19) (0.11–1.07) (0.16–2.28) (0.50–1.26) (0.25–0.77)  Death of loved one 0.80 1.17 0.62* 0.61 0.50* 0.51 0.88 0.74 (0.63–1.02) (0.82–1.66) (0.42–0.90) (0.37–1.00) (0.25–0.99) (0.20–1.34) (0.55–1.39) (0.41–1.35)  Financial crisis 1.26 1.49 1.08 1.20 0.56 1.16 1.43 1.02 (0.96–1.65) (0.98–2.27) (0.71–1.63) (0.74–1.92) (0.24–1.32) (0.50–2.70) (0.71–2.85) (0.52–2.01)  Midwestc 0.61** 0.61* 0.64 0.47* 1.09 0.55 0.59 0.83 (0.42–0.89) (0.38–0.96) (0.36–1.14) (0.23–0.98) (0.31–3.79) (0.20–1.53) (0.29–1.19) (0.31–2.23)  Southc 0.65** 0.52** 0.84 0.56 1.01 1.07 0.72 0.85 (0.47–0.89) (0.33–0.80) (0.59–1.19) (0.28–1.11) (0.29–3.52) (0.34–3.33) (0.36–1.47) (0.41–1.80)  Westc 0.84 0.81 0.85 0.76 1.20 0.57 0.79 0.93 (0.58–1.20) (0.41–1.60) (0.58–1.14) (0.26–2.23) (0.35–4.12) (0.19–1.74) (0.36–1.75) (0.39–2.20)  County pop. size 0.99 1.04 0.98 0.97 1.04 1.08 0.97 1.05 (0.93–1.07) (0.96–1.14) (0.89–1.07) (0.85–1.10) (0.89–1.22) (0.89–1.33) (0.87–1.08) (0.92–1.18)  Education 1.08** 1.11*** 1.00 0.97 0.99 1.06 1.09 1.03 (1.02–1.13) (1.04–1.18) (0.93–1.07) (0.86–1.09) (0.88–1.12) (0.94–1.19) (0.98–1.21) (0.92–1.15)  Married 0.95 1.22 0.76 1.10 0.49 0.56 0.90 1.10 (0.70–1.28) (0.77–1.94) (0.54–1.09) (0.64–1.90) (0.18–1.32) (0.30–1.07) (0.55–1.45) (0.64–1.88)  Worker 0.91 0.77 0.96 0.57* 0.80 1.18 1.28 1.35 (0.71–1.16) (0.53–1.12) (0.68–1.37) (0.33–1.00) (0.45–1.42) (0.39–3.63) (0.80–2.04) (0.81–2.26)  Moved (W1–W2) 0.88 0.86 0.84 0.86 0.67 0.61 0.77 0.85 (0.63–1.21) (0.59–1.25) (0.55–1.27) (0.54–1.35) (0.23–1.94) (0.22–1.68) (0.46–1.26) (0.44–1.64)  Change in marital status (W1–W2) 0.83 0.90 0.76 0.34** 0.73 1.13 0.99 0.84 (0.62–1.10) (0.60–1.36) (0.53–1.07) (0.17–0.66) (0.30–1.78) (0.41–3.14) (0.64–1.54) (0.41–1.70)  Change in work status (W1–W2) 0.84 0.63* 1.05 0.68 0.22*** 1.65 1.08 1.73* (0.58–1.20) (0.41–0.96) (0.68–1.61) (0.38–1.24) (0.10–0.50) (0.56–4.88) (0.47–2.46) (1.05–2.85) Relational level covariates  Religious homophily 1.08 1.16 1.01 1.01 1.33 0.66 1.40 1.27 (0.83–1.39) (0.84–1.61) (0.69–1.48) (0.56–1.81) (0.66–2.70) (0.33–1.31) (0.92–2.15) (0.83–1.94)  Alter rank 1 3.68*** 3.25*** 3.94*** 4.34*** 1.89 2.71 4.14*** 4.80*** (2.61–5.18) (2.19–4.83) (2.34–6.63) (2.32–8.11) (0.66–5.39) (0.93–7.85) (2.34–7.31) (2.37–9.71)  Alter rank 2 2.05*** 1.84*** 2.20*** 1.91 1.65 2.37* 2.40** 1.70 (1.56–2.69) (1.29–2.63) (1.42–3.41) (0.97–3.74) (0.70–3.89) (1.01–5.56) (1.35–4.24) (0.96–3.00)  Alter rank 3 1.36* 1.29 1.42 1.18 0.81 1.52 1.93** 1.50 (1.04–1.78) (0.90–1.83) (0.91–2.21) (0.64–2.17) (0.35–1.91) (0.44–5.19) (1.17–3.18) (0.73–3.07)  Alter gave advice 1.62** 1.33 1.97*** 1.44 0.67 1.18 1.09 2.73*** (1.20–2.18) (0.85–2.08) (1.39–2.79) (0.81–2.56) (0.26–1.72) (0.67–2.05) (0.60–1.99) (1.47–5.07)  Alter gave help 1.27 1.47 1.19 0.99 0.66 0.86 2.12** 1.21 (0.92–1.74) (0.92–2.36) (0.74–1.92) (0.58–1.69) (0.28–1.57) (0.41–1.78) (1.23–3.67) (0.56–2.63)  Lives nearby 1.09 1.07 1.14 0.89 1.01 1.93* 1.19 1.14 (0.83–1.43) (0.73–1.57) (0.82–1.59) (0.57–1.38) (0.49–2.10) (1.16–3.23) (0.76–1.86) (0.67–1.94)  Interact weekly 1.32 1.24 1.39 1.46 3.64** 1.58 1.51 0.67 (0.96–1.81) (0.86–1.77) (0.91–2.15) (0.81–2.62) (1.45–9.13) (0.67–3.72) (0.84–2.71) (0.34–1.31)  Racial homophily 1.95** 2.20* 1.58 0.88 3.56 8.40 1.05 2.28* (1.24–3.06) (1.10–4.40) (0.90–2.77) (0.27–2.86) (0.44–28.68) (0.36–197.71) (0.48–2.28) (1.07–4.88)  Gender homophily 1.26* 1.48* 1.14 4.33 2.31* 0.77 0.94 1.21 (1.01–1.56) (1.03–2.11) (0.88–1.47) (1.61– 11.67)** (1.08–4.92) (0.30–1.97) (0.61–1.45) (0.78–1.87)  Alter has bachelor dg. 0.93 0.90 1.04 1.14 0.90 1.77 0.63* 1.10 (0.76–1.13) (0.68–1.20) (0.79–1.37) (0.67–1.94) (0.45–1.80) (0.73–4.31) (0.39–1.00) (0.61–1.99)  Kin 2.45*** 1.91** 2.88*** 0.83 3.58*** 0.90 1.84*** 2.32* (1.90–3.17) (1.27–2.87) (1.99–4.15) (0.47–1.48) (1.68–7.64) (0.40–2.02) (1.27–2.68) (1.22–4.44)  Co-worker 0.80 0.62 0.97 2.45 0.14** 0.96 0.93 1.70 (0.48–1.35) (0.33–1.18) (0.52–1.80) (1.52– 3.95)** (0.04–0.57) (0.27–3.34) (0.42–2.06) (0.68–4.24)  Level 1 variance 1.25 1.24 1.02 0.62 1.21 1.15 0.97 0.56 (0.17) (0.24) (0.21) (0.31) (0.42) (0.67) (0.31) (0.34)  −2LL 859,928,640 390,913,472 452,567,008 222,713,520 88,179,943 19,448,486 230,494,992 114,654,792 Observations  Level 1 4619 2224 2395 1031 518 385 1223 752  Level 2 1233 583 650 273 134 106 331 202 Full sample Part of Congregation Not Part of Congregation Evangelicals Mainline Protestants Black Protestants Catholics Religious nones Model 1 Model 2 Model 3 Model 4 Model 6 Model 5 Model 7 Model 8 Discuss religion with alter 1.44* 1.80** 1.24 2.45** 3.11** 1.33 1.24 0.91 (1.05–1.98) (1.22–2.67) (0.83–1.85) (1.39–4.30) (1.43–6.74) (0.40–4.38) (0.79–1.97) (0.47–1.78) Individual level covariates  Part of cong. 0.62** 0.76 0.80 0.62 0.94 0.11*** (0.46–0.84) (0.46–1.25) (0.33–1.94) (0.21–1.85) (0.55–1.60) (0.03–0.41)  Evangelicala 0.97 4.50** 0.71 (0.62–1.51) (1.50–13.50) (0.43–1.19)  Black Protestanta 0.71 3.41 0.60 (0.44–1.15) (0.82–14.15) (0.23–1.58)  Mainline Protestanta 1.16 5.92*** 0.82 (0.65–2.06) (2.06–17.03) (0.31–2.15)  Catholica 0.97 5.43** 0.63* (0.63–1.49) (1.83–16.11) (0.40–0.99)  Other religiona 0.53** 1.58 0.68 (0.33–0.84) (0.48–5.23) (0.43–1.04)  Female 1.31* 1.56* 1.04 1.60* 0.95 1.43 1.18 1.03 (1.01–1.70) (1.06–2.29) (0.79–1.37) (1.05–2.45) (0.50–1.81) (0.48–4.31) (0.68–2.05) (0.64–1.66)  Blackb 0.96 0.88 0.93 1.12 0.38 0.49 0.81 (0.67–1.39) (0.42–1.87) (0.53–1.63) (0.54–2.34) (0.05–3.10) (0.16–1.53) (0.38–1.74)  Latino/ab 1.03 1.12 1.00 1.95 0.17* 0.82 0.67 (0.69–1.54) (0.59–2.15) (0.51–1.97) (0.76–4.97) (0.04–0.73) (0.41–1.66) (0.22–2.04)  Other raceb 0.67 0.38 0.96 1.95 0.22* 0.34 0.69 (0.33–1.33) (0.11–1.34) (0.42–2.22) (0.22–17.52) (0.05–0.86) (0.09–1.19) (0.14–3.33)  Age 1.01 0.99 1.02** 1.00 0.96* 1.00 1.01 1.03* (1.00–1.02) (0.98–1.01) (1.01–1.04) (0.99–1.02) (0.93–0.99) (0.97–1.04) (1.00–1.03) (1.00–1.06)  Serious illness/injury 0.78 0.89 0.62 1.27 0.35 0.61 0.79 0.44** (0.55–1.12) (0.54–1.45) (0.39–1.00) (0.74–2.19) (0.11–1.07) (0.16–2.28) (0.50–1.26) (0.25–0.77)  Death of loved one 0.80 1.17 0.62* 0.61 0.50* 0.51 0.88 0.74 (0.63–1.02) (0.82–1.66) (0.42–0.90) (0.37–1.00) (0.25–0.99) (0.20–1.34) (0.55–1.39) (0.41–1.35)  Financial crisis 1.26 1.49 1.08 1.20 0.56 1.16 1.43 1.02 (0.96–1.65) (0.98–2.27) (0.71–1.63) (0.74–1.92) (0.24–1.32) (0.50–2.70) (0.71–2.85) (0.52–2.01)  Midwestc 0.61** 0.61* 0.64 0.47* 1.09 0.55 0.59 0.83 (0.42–0.89) (0.38–0.96) (0.36–1.14) (0.23–0.98) (0.31–3.79) (0.20–1.53) (0.29–1.19) (0.31–2.23)  Southc 0.65** 0.52** 0.84 0.56 1.01 1.07 0.72 0.85 (0.47–0.89) (0.33–0.80) (0.59–1.19) (0.28–1.11) (0.29–3.52) (0.34–3.33) (0.36–1.47) (0.41–1.80)  Westc 0.84 0.81 0.85 0.76 1.20 0.57 0.79 0.93 (0.58–1.20) (0.41–1.60) (0.58–1.14) (0.26–2.23) (0.35–4.12) (0.19–1.74) (0.36–1.75) (0.39–2.20)  County pop. size 0.99 1.04 0.98 0.97 1.04 1.08 0.97 1.05 (0.93–1.07) (0.96–1.14) (0.89–1.07) (0.85–1.10) (0.89–1.22) (0.89–1.33) (0.87–1.08) (0.92–1.18)  Education 1.08** 1.11*** 1.00 0.97 0.99 1.06 1.09 1.03 (1.02–1.13) (1.04–1.18) (0.93–1.07) (0.86–1.09) (0.88–1.12) (0.94–1.19) (0.98–1.21) (0.92–1.15)  Married 0.95 1.22 0.76 1.10 0.49 0.56 0.90 1.10 (0.70–1.28) (0.77–1.94) (0.54–1.09) (0.64–1.90) (0.18–1.32) (0.30–1.07) (0.55–1.45) (0.64–1.88)  Worker 0.91 0.77 0.96 0.57* 0.80 1.18 1.28 1.35 (0.71–1.16) (0.53–1.12) (0.68–1.37) (0.33–1.00) (0.45–1.42) (0.39–3.63) (0.80–2.04) (0.81–2.26)  Moved (W1–W2) 0.88 0.86 0.84 0.86 0.67 0.61 0.77 0.85 (0.63–1.21) (0.59–1.25) (0.55–1.27) (0.54–1.35) (0.23–1.94) (0.22–1.68) (0.46–1.26) (0.44–1.64)  Change in marital status (W1–W2) 0.83 0.90 0.76 0.34** 0.73 1.13 0.99 0.84 (0.62–1.10) (0.60–1.36) (0.53–1.07) (0.17–0.66) (0.30–1.78) (0.41–3.14) (0.64–1.54) (0.41–1.70)  Change in work status (W1–W2) 0.84 0.63* 1.05 0.68 0.22*** 1.65 1.08 1.73* (0.58–1.20) (0.41–0.96) (0.68–1.61) (0.38–1.24) (0.10–0.50) (0.56–4.88) (0.47–2.46) (1.05–2.85) Relational level covariates  Religious homophily 1.08 1.16 1.01 1.01 1.33 0.66 1.40 1.27 (0.83–1.39) (0.84–1.61) (0.69–1.48) (0.56–1.81) (0.66–2.70) (0.33–1.31) (0.92–2.15) (0.83–1.94)  Alter rank 1 3.68*** 3.25*** 3.94*** 4.34*** 1.89 2.71 4.14*** 4.80*** (2.61–5.18) (2.19–4.83) (2.34–6.63) (2.32–8.11) (0.66–5.39) (0.93–7.85) (2.34–7.31) (2.37–9.71)  Alter rank 2 2.05*** 1.84*** 2.20*** 1.91 1.65 2.37* 2.40** 1.70 (1.56–2.69) (1.29–2.63) (1.42–3.41) (0.97–3.74) (0.70–3.89) (1.01–5.56) (1.35–4.24) (0.96–3.00)  Alter rank 3 1.36* 1.29 1.42 1.18 0.81 1.52 1.93** 1.50 (1.04–1.78) (0.90–1.83) (0.91–2.21) (0.64–2.17) (0.35–1.91) (0.44–5.19) (1.17–3.18) (0.73–3.07)  Alter gave advice 1.62** 1.33 1.97*** 1.44 0.67 1.18 1.09 2.73*** (1.20–2.18) (0.85–2.08) (1.39–2.79) (0.81–2.56) (0.26–1.72) (0.67–2.05) (0.60–1.99) (1.47–5.07)  Alter gave help 1.27 1.47 1.19 0.99 0.66 0.86 2.12** 1.21 (0.92–1.74) (0.92–2.36) (0.74–1.92) (0.58–1.69) (0.28–1.57) (0.41–1.78) (1.23–3.67) (0.56–2.63)  Lives nearby 1.09 1.07 1.14 0.89 1.01 1.93* 1.19 1.14 (0.83–1.43) (0.73–1.57) (0.82–1.59) (0.57–1.38) (0.49–2.10) (1.16–3.23) (0.76–1.86) (0.67–1.94)  Interact weekly 1.32 1.24 1.39 1.46 3.64** 1.58 1.51 0.67 (0.96–1.81) (0.86–1.77) (0.91–2.15) (0.81–2.62) (1.45–9.13) (0.67–3.72) (0.84–2.71) (0.34–1.31)  Racial homophily 1.95** 2.20* 1.58 0.88 3.56 8.40 1.05 2.28* (1.24–3.06) (1.10–4.40) (0.90–2.77) (0.27–2.86) (0.44–28.68) (0.36–197.71) (0.48–2.28) (1.07–4.88)  Gender homophily 1.26* 1.48* 1.14 4.33 2.31* 0.77 0.94 1.21 (1.01–1.56) (1.03–2.11) (0.88–1.47) (1.61– 11.67)** (1.08–4.92) (0.30–1.97) (0.61–1.45) (0.78–1.87)  Alter has bachelor dg. 0.93 0.90 1.04 1.14 0.90 1.77 0.63* 1.10 (0.76–1.13) (0.68–1.20) (0.79–1.37) (0.67–1.94) (0.45–1.80) (0.73–4.31) (0.39–1.00) (0.61–1.99)  Kin 2.45*** 1.91** 2.88*** 0.83 3.58*** 0.90 1.84*** 2.32* (1.90–3.17) (1.27–2.87) (1.99–4.15) (0.47–1.48) (1.68–7.64) (0.40–2.02) (1.27–2.68) (1.22–4.44)  Co-worker 0.80 0.62 0.97 2.45 0.14** 0.96 0.93 1.70 (0.48–1.35) (0.33–1.18) (0.52–1.80) (1.52– 3.95)** (0.04–0.57) (0.27–3.34) (0.42–2.06) (0.68–4.24)  Level 1 variance 1.25 1.24 1.02 0.62 1.21 1.15 0.97 0.56 (0.17) (0.24) (0.21) (0.31) (0.42) (0.67) (0.31) (0.34)  −2LL 859,928,640 390,913,472 452,567,008 222,713,520 88,179,943 19,448,486 230,494,992 114,654,792 Observations  Level 1 4619 2224 2395 1031 518 385 1223 752  Level 2 1233 583 650 273 134 106 331 202 Notes: Robust standard errors are shown in parentheses. aReligions none is reference group. bWhite is reference group; Race/ethnicity not included in model 5 because the overwhelming majority of Black Protestants are Black. cNortheast is reference group. ***p < .001; **p < .01; *p < .05 (two-tailed tests). View Large Though religious discussion seems a relational adhesive in the full sample model (model 1), further analysis reveals that this phenomenon is contingent upon ego’s own religiosity. Among those who are connected to a religious congregation, discussing religion is associated with an 80% increase in the odds (95% CI = 1.22–2.67) that the tie will remain at W2. The relationship is nonsignificant among that half of the sample who has no congregational involvement (OR = 1.24; 95% CI = 0.83–1.85). Protestants are the other subgroups that stand out. Alters identified by Evangelicals, the conservative branch of Protestantism, have 225% higher odds of remaining in the W2 network if they and ego shared religious conversation than if they did not. For Mainline Protestants, talking religion increases the odds of tie persistence by a factor of 2.88. Among Catholics, Black Protestants, or religious nones, however, there was no evidence that religious discussion with a given alter increased tie durability between waves; each of these groups had odds ratios statistically indistinguishable from 1.0. Supplementary analyses further examined whether congregational affiliates, Evangelicals, and Mainline Protestants held on to their religious discussion partner ties merely because these alters tended to belong to their churches. I incorporated information about whether the tie was a co-congregant from the name-interpreter section of the PALS survey (respondents were asked “Which of the people you mentioned [as network members] are involved in your congregation”) and re-estimated models 2, 4, and 6. Results indicate that co-affiliation in a congregation—a classic instance of focused activity as a basis of network connectedness (Feld 1981)—was unable to explain why religious discussion partners tended to have staying power in these ego networks. Finally, supplementary analyses considered whether the longitudinal findings reported in table 4 could be observed beyond the strongest of people’s strong ties. Taking the largest subgroup of religious participants, those connected to a congregation, I limited network ties to those with whom ego did not receive helpful advice and who fell in the bottom half of the roster (i.e., named 3rd and 4th rather than given 1st or 2nd priority). Results from this “less-core” subset of the close network—those lacking the strongest signals of tie strength at W1—reveal that the association between religious discussion and tie persistence remained quite strong (OR = 2.21, p < .001). DISCUSSION This study addressed two main research questions. First, what individual and relational factors predict the discussion of religion in close networks? Analyses reveal that American adults report discussing religion with about half of their closest nonresidential ties in the past year. Among the strongest individual-level predictors of discussing religion was ego’s involvement in a congregation. This dimension of religiosity surpassed the importance of denominational affiliation, suggesting that for discussion of religion in networks, the behavioral expression of religious fidelity is more important than the religious group with whom one identifies. Contrary to hypotheses, there was no evidence that religious discussion differed by age, or (with the exception of Catholics) that people were more likely to have talked religion with network members amidst stressful life conditions such as financial strain or a loved one’s death. Findings did, however, provide support for each of the three theoretical frameworks offered as relational explanations for religious discussion. Namely, sharing a religious background, being strongly tied, and having ample opportunity for conversation each predicted religious discussion between PALS respondents and their alters. Relational explanations tended to differ somewhat, however, according to religious tradition. Most notably, religious homophily was influential among people connected to a religious congregation and among Mainline and Evangelical Protestants, while sharing a nonreligious outlook with one’s ties did not appear to matter among religious nones. The one relational predictor that was observed across all stripes of religiosity was whether the alter had provided useful advice—ties that offer counsel tend to be ties with whom ego is most likely to discuss religion, and this whether or not ego him/herself is religious. Building from these initial findings, the second research question considered whether discussion of things religious predicts the presence of a given tie over the course of six years. This inquiry addresses the central question of network change (“why do some ties persist more than others?”; Suitor et al. 1997:4), recognizing that discussion topics themselves could plausibly solidify or destabilize social bonds. Results indicate that talking about religion was associated with a greater likelihood of the network member remaining in ego’s close network. It may be that delving into matters of ultimate significance (at least from the perspective of ego) imbues a particular relationship with heightened significance and cements its presence in a network over time. Alternatively, the types of ties that venture into religious topics may be the ones with the strongest staying power. Theoretical accounts of relationship progression and intimacy-disclosure dynamics suggest that both processes are likely in operation, though this study is unable to decisively arbitrate causal directionality. Other studies have considered how factors such as kinship status, closeness, and contact frequency are associated with tie persistence (Lubbers et al. 2010; Morgan et al. 1996; Suitor and Keeton 1997), but this study breaks new ground by examining the role of conversational substance. There may be additional discussion topics that too serve as relational adhesives—talking about personal health or politics, for instance, could plausibly enhance tie durability. At the same time, it is uncertain whether these types of topics carry the existential or sacred weight that characterizes religion among so many U.S. adults. Nevertheless, the topical exceptionalism of religion relative to other discussion matters is a hypothesis that could be addressed in future research. It is worth noting that the close networks measured in the PALS include only up to four members, so it is quite possible that initial alters identified at baseline but left unnamed at survey follow-up were still an active presence in the respondent’s life. Still, the fact that religious discussion is such a strong predictor of remaining among one’s top four associates—above and beyond the power of receiving social support, sharing the same gender and racial background, talking frequently, or being prioritized as a “top” tie at W1—tells us something of the significance of talking religion, at least when it comes to a person’s inner circle. If not working as a “relational adhesive” in any strictly causal sense, religious discussion would then seem to capture an unacknowledged and undertheorized dimension of tie strength. As expected, discussing religion seems to matter most among those who are involved in a religious congregation (particularly among Mainline and Evangelical Protestants). And, for the religious, talking religion is more important than merely sharing the same faith background or even sharing a congregation. This study contributes to research on religious networks and on the discussion of sensitive matters in close networks. Existing research on touchy topics is dominated by analyses of health-focused or politically focused conversation (Klofstad et al. 2009; Perry and Pescosolido 2010). Religion, surprisingly, has been largely left out of consideration. Indeed, research more generally addressing the overlap between social networks and religion remains at an early stage, mainly because few large-scale representative surveys with ego-network methodology have seriously considered religion and spirituality (Everton 2016). A recent string series of studies has begun to address the role of religion in the social ties in lives of American adults, demonstrating how network ties’ religiosity aids volunteer recruitment efforts (Merino 2013) and shapes social support provision (Merino 2014; Schafer 2015). This study complements these efforts by shedding light on the basic matter of religion as a discussion topic in close relationships. Still, more research is needed to advance the themes presented in this study. First, what do people actually talk about when they discuss religion? It would be helpful, for instance, to distinguish talk of personal religious beliefs and spirituality from conversations about the intersection of religion and public life. This may be a place to more carefully consider gender differences. The current findings revealed that women are more likely than men to talk religion with their network members, but it is unclear whether men and women talk about religion in gender-specific ways (Ozorak 1996; Thompson 1991). The finding of regional variation also raises questions about whether people in the American South versus the American West—both regions where people outtalk those in the distinctively “post-Christian” Northeast (Barna Group 2017)—discuss religion in varied ways. Chronicling precisely how people talk religion within various regional contexts will deepen our understanding of the microlevel, interactional dimensions of society’s secular turn (Voas and Chaves 2016). Understanding the disclosure process would be another fruitful direction for future research. People tend to broach political topics when they anticipate that their network member shares their beliefs (Cowan and Baldassarri 2017). Do similar processes apply when it comes to disclosing religious beliefs? Further, does this disclosure differ for religious and nonreligious Americans? As a distrusted minority group in the United States (Edgell et al. 2016), the nonreligious may be more guarded than are people in more sizeable religious groups. Third, future research should consider other results of religious conversation within social networks. This study used network turnover as an initial outcome to understand why religious discussion may be consequential, but there are other processes that could be affected by talking religion or by religious disclosure. For instance, does religious discussion between religiously heterogeneous dyads spark tension or increase argumentation? Does frequently talking religion lead to more tolerance or instead to ideological polarization within networks? Such questions are timely, given the widespread perception of a growing religious divide in the United States (Stetzer 2015). Whole-network analyses may usefully supplement the current study’s ego-centered network approach and demonstrate the bottom-up structural dynamics of religious division in particular populations. Though helping stimulate these promising avenues for future work, this study is not without its limitations. The name-generator technique used in the PALS survey asks about people’s four closest nonresidential ties, so I am unable to document how much people discussion religion with weaker ties. This may be a reason why talking religion with religiously dissimilar alters did not predict tie loss (or negatively moderate the association between talking religion and tie durability); limiting the network to only one’s four closest associates could mean that the ties are close enough to withstand disagreement over a touchy topic or the occasional awkward conversation. Another measurement limitation is that respondents report only if they discussed religion with their tie, not how often (or in what ways) they had such discussions. Likewise, in their efforts to limit the time demands of the survey, the research team failed to ask participants how long they had known the alter or anything about the person’s political views. In spite of these drawbacks, the PALS data offer several clear advantages. It is, to my knowledge, the only publically available, nationally representative survey study that addresses religious discussion in close networks. Further, its panel design and careful documentation of W1 and W2 network members’ identity enable the longitudinal analysis of tie persistence. Still, much could be gained from collecting data with more nuanced attention to the particularities of religious discussion and in relation to other important features of the relationship. Finally, as insinuated above, the analysis of relational factors as predictors of religious discussion can provide little leverage on questions of causality. The majority of PALS respondents’ alters named at W1 did not reappear in W2 network rosters, making it untenable to adequately test competing longitudinal processes in PALS respondents’ networks (i.e., religious homophily → religious discussion vs. religious discussion → religious homophily). In conclusion, sociologists have long contended that “religion is an eminently social thing” (Durkheim 1912 [1995]:9), a system of beliefs and practices sustained by groups and capable of fostering social solidarity. Though many of the social forms of religious behavior occur mainly within organizational settings such as churches or mosques, social networks are other important—if poorly understood—contexts for religious expression. This study finds that both individual and relational factors shape the discussion of religion within close networks, and it also reveals that for many Americans, talking religion has significant implications for network continuity. In so doing, it invites future investigation into what specific religious topics emerge in people’s networks and what are the other implications of discussing religion for social network processes. ACKNOWLEDGMENTS This article has benefited from constructive feedback by participants at the University of Western Ontario Sociology Colloquium series and from helpful comments from the Editor and reviewers of Sociology of Religion. Data for this study come from the Portraits of American Life Study, a publicly available dataset available at http://www.palsresearch.org/pals/researchers. Footnotes 1 In the terminology of social network analysis, an ego is the focal individual (e.g., the sampled survey respondent), while an alter is the person identified by ego as a member of his/her social network. 2 For instance, money and household finances are “important matters” that people often discuss with a spouse, whereas community issues or politics tend to be the matters about which people talk with friends (Bearman and Parigi 2004). 3 For a similar argument in the context of politics, see Cowan and Baldassarri (2017). 4 Though this study is concerned primarily with people’s close networks—all members of which could be considered “strong ties”—there is dyadic variability on such traits even within this inner circle (Merino 2014). 5 Closeness and kinship status, for instance, increase the durability of a social support tie over 10 years (Suitor and Keeton 1997). 6 Specifically, I used a variety of W1 survey measures to predict the conditional probability of remaining in the PALS sample. I then computed case-specific predicted probabilities from the logistic regression model and took the inverse of this value. Finally, I multiplied each propensity score by the standard survey weights. In effect, this procedure up-weights those cases least likely to have contributed data at W2. 7 This name-generator tool departs from its close ego-network counterpart in the General Social Survey (GSS), but in ways advantageous to the present research questions. Specifically, the PALS network is not defined by whether people discuss important matters, an issue of inevitable interpretive difference across various respondents (e.g., is religion an “important matter”?). Rather, the network starts with people who are close, and in the name-interpreter, follow-up questions ascertain whether or they happen to discuss religion. That being said, it is important to note that the PALS close network is wholly nonresidential, whereas the GSS allows housemates as members of the core discussion network. 8 Frequency of communication is also considered an element of tie strength (Granovetter 1973), so a significant association with religious discussion would offer support for both the opportunity mobilization and the strength of tie framework. 9 The interviewer further specified that congregation refers to a “church, temple, synagogue, mosque, or other place of worship.” 10 Fitting a model that excludes affiliation also has the advantage of generating a more sensible estimate of whether Black people are more/less likely to talk religion than White adults. Participants who identify as Black Protestant are overwhelming Black, generating considerable multicollinearity in a model that adjusts for both race/ethnicity and denominational affiliation. 11 “Other religion” is not analyzed as its own subgroup in these models because it is a heterogeneous mix of religious affiliations. 12 Coefficients for quadratic and cubic specifications of age failed to achieve statistical significance in supplementary analyses. 13 Supplementary analyses further examined the potential distinctiveness of the South with respect to relational characteristics. Most associations were consistent between Northeast and South when the sample was divided into regional subgroups though several differences did emerge. First, religious homophily was not significantly associated with religious discussion in the South, whereas it was in the Northeast. Second, being highly ranked in the network and receiving help is a strong predictor of talking religion in the South, whereas these factors are noticeably weaker and nonsignificant in the Northeast. Frequency of contact, however, was a stronger predictor in the Northeast. 14 Supplementary analyses reveal that the positive association between college-educated alter and religious discussion was driven primarily by cases where the alter’s education exceeded that of the respondent. REFERENCES Altman , Irwin , and Dalmas A. Taylor . 1973 . Social Penetration: The Development of Interpersonal Relationships . New York, NY : Holt, Rinehart & Winston . Baldassarri , Delia , and Peter Bearman . 2007 . “ Dynamics of Political Polarization .” American Sociological Review 72 : 784 – 811 . Google Scholar CrossRef Search ADS Barna Group . “The Most Post-Christian Cities in America: 2017.” Barna: Cities. https://www.barna.com/research/post-christian-cities-america-2017/. Bearman , Peter S. , and Paolo Parigi . 2004 . “ Cloning Headless Frogs and Other Important Matters: Conversation Topics and Network Structure .” Social Forces 83 : 535 – 57 . Google Scholar CrossRef Search ADS Bidart , Claire , and Daniel Lavenu . 2005 . “ Evolutions of Personal Networks and Life Events .” Social Networks 27 : 359 – 76 . Google Scholar CrossRef Search ADS Brashears , Matthew E . 2014 . “ ‘Trivial’ Topics and Rich Ties: The Relationship between Discussion Topic, Alter Role, and Resource Availability Using the ‘Important Matters’ Name Generator .” Sociological Science 1 : 493 – 511 . Google Scholar CrossRef Search ADS Burt , Ronald S . 1984 . “ Network Items and the General Social Survey .” Social Networks 6 : 293 – 339 . Google Scholar CrossRef Search ADS Collins , Nancy L. , and Lynn Carol Miller . 1994 . “ Self-Disclosure and Liking: A Meta-Analytic Review .” Psychological Bulletin 116 : 457 – 75 . Google Scholar CrossRef Search ADS PubMed Cooperman , Alan . 2016 . Many Americans Don’t Argue about Religion—or Even Talk about It . Washington, DC : Pew Research Center . http://www.pewresearch.org/fact-tank/2016/04/15/many-americans-dont-argue-about-religion-or-even-talk-about-it/. Cornwell , Benjamin . 2015 . “ Social Disadvantage and Network Turnover .” Journals of Gerontology, Series B: Psychological Sciences and Social Sciences 70 : 132 – 42 . Google Scholar CrossRef Search ADS Cornwell , Benjamin , and Edward O. Laumann . 2015 . “ The Health Benefits of Network Growth: New Evidence from a National Survey of Older Adults .” Social Science & Medicine 125 : 94 – 106 . Google Scholar CrossRef Search ADS Cornwell , Erin Y. , and Linda J. Waite . 2012 . “ Social Network Resources and Management of Hypertension .” Journal of Health and Social Behavior 53 : 215 – 231 . Google Scholar CrossRef Search ADS PubMed Cowan , Sarah K. , and Delia Baldassarri . 2017 . “ ‘It Could Turn Ugly’: Selective Disclosure of Attitudes in Political Discussion Networks .” Social Networks 52 : 1 – 17 . Google Scholar CrossRef Search ADS Dallas , Kelsey . 2016 . “What Keeps Believers from Talking about Faith in Public?” DeseretNews.Com , May 19 . http://www.deseretnews.com/article/865654517/Whats-lost-when-people-stop-talking-about-faith-in-public.html. Durkheim , Emile . 1912 . Elementary Forms of the Religious Life . New York, NY : Free Press . Edgell , Penny , Douglas Hartmann , Evan Stewart , and Joseph Gerteis . 2016 . “ Atheists and Other Cultural Outsiders: Moral Boundaries and the Non-Religious in the United States .” Social Forces 95 : 607 – 38 . Google Scholar CrossRef Search ADS Emerson , Michael O. , David Sikkink , and Adele D. James . 2010 . “ The Panel Study on American Religion and Ethnicity: Background, Methods, and Selected Results .” Journal for the Scientific Study of Religion 49 : 162 – 71 . Google Scholar CrossRef Search ADS Everton , Sean F . 2016 . “ Networks and Religion: Ties that Bind, Loose, Build Up, and Tear Down .” Connections 16 : 1 – 33 . Feld , Scott L . 1981 . “ The Focused Organization of Social Ties .” American Journal of Sociology 86 : 1015 – 35 . Google Scholar CrossRef Search ADS Feld , Scott L. , J. Jill Suitor , and Jordana Gartner Hoegh . 2007 . “ Describing Changes in Personal Networks over Time .” Field Methods 19 : 218 – 36 . Google Scholar CrossRef Search ADS Gerber , Alan S. , Gregory A. Huber , David Doherty , and Conor M. Dowling . 2012 . “ Disagreement and the Avoidance of Political Discussion: Aggregate Relationships and Differences across Personality Traits .” American Journal of Political Science 56 : 849 – 74 . Google Scholar CrossRef Search ADS Gerber , Alan S. , Gregory A. Huber , David Doherty , Conor M. Dowling , and Seth J. Hill . 2013 . “ Who Wants to Discuss Vote Choices with Others? Polarization in Preferences for Deliberation .” Public Opinion Quarterly 77 : 474 – 96 . Google Scholar CrossRef Search ADS Granovetter , Mark S . 1973 . “ The Strength of Weak Ties .” American Journal of Sociology 78 : 1360 – 80 . Google Scholar CrossRef Search ADS Green , Emma . 2016 . “The Impoliteness of Talking about Religion.” The Atlantic . April 12 . https://www.theatlantic.com/politics/archive/2016/04/the-impoliteness-of-talking- about-religion/477834/. Klofstad , Casey A . 2007 . “ Talk Leads to Recruitment: How Discussions about Politics and Current Events Increase Civic Participation .” Political Research Quarterly 60 : 180 – 91 . Google Scholar CrossRef Search ADS Klofstad , Casey A. , Scott D. McClurg , and Meredith Rolfe . 2009 . “ Measurement of Political Discussion Networks: A Comparison of Two ‘Name Generator’ Procedures .” Public Opinion Quarterly 73 : 462 – 83 . Google Scholar CrossRef Search ADS Lawler , Edward J. , and Shane R. Thye . 1999 . “ Bringing Emotions into Social Exchange Theory .” Annual Review of Sociology 25 : 217 – 44 . Google Scholar CrossRef Search ADS Lubbers , Miranda J. , Jose Luis Molina , Jurgen Lerner , Ulrik Brandes , Javier Avila , and Christopher McCarty . 2010 . “ Longitudinal Analysis of Personal Networks. The Case of Argentinean Migrants in Spain .” Social Networks 32 : 91 – 104 . Google Scholar CrossRef Search ADS Machacek , David W . 2003 . “ The Problem of Pluralism .” Sociology of Religion 64 : 145 – 61 . Google Scholar CrossRef Search ADS Marsden , Peter V . 1987 . “ Core Discussion Networks of Americans .” American Sociological Review 52 : 122 – 31 . Google Scholar CrossRef Search ADS Martin , Judith . 1998 . “ A Few Topics to Avoid in Social Conversation .” Chicago Tribune , January 1 . McClurg , Scott D . 2003 . “ Social Networks and Political Participation: The Role of Social Interaction in Explaining Political Participation .” Political Research Quarterly 56 : 449 – 64 . Google Scholar CrossRef Search ADS McPherson , Miller , Lynn Smith-Lovin , and James M. Cook . 2001 . “ Birds of a Feather: Homophily in Social Networks .” Annual Review of Sociology 27 : 415 – 44 . Google Scholar CrossRef Search ADS ———. 2013 . “ Religious Social Networks and Volunteering: Examining Recruitment via Close Ties .” Review of Religious Research 55 : 509 – 527 . CrossRef Search ADS Merino , Stephen M . 2014 . “ Social Support and the Religious Dimensions of Close Ties .” Journal for the Scientific Study of Religion 53 : 595 – 612 . Google Scholar CrossRef Search ADS Morgan , David L. , Margaret B. Neal , and Paula Carder . 1996 . “ The Stability of Core and Peripheral Networks Over Time .” Social Networks 19 : 9 – 25 . Google Scholar CrossRef Search ADS Ozorak , Elizabeth Weiss . 1996 . “ The Power, but Not the Glory: How Women Empower Themselves Through Religion .” Journal for the Scientific Study of Religion 35 : 17 – 29 . Google Scholar CrossRef Search ADS Pargament , Kenneth I . 2002 . “ The Bitter and the Sweet: An Evaluation of the Costs and Benefits of Religiousness .” Psychological Inquiry 13 : 168 – 81 . Google Scholar CrossRef Search ADS Pargament , Kenneth I. , and Annette Mahoney . 2005 . “ Sacred Matters: Sanctification as a Vital Topic for the Psychology of Religion .” The International Journal for the Psychology of Religion 15 : 179 – 98 . Google Scholar CrossRef Search ADS Perry , Brea L. , and Bernice A. Pescosolido . 2010 . “ Functional Specificity in Discussion Networks: The Influence of General and Problem-Specific Networks on Health Outcomes .” Social Networks 32 : 345 – 57 . Google Scholar CrossRef Search ADS ———. 2015 . “ Social Network Activation: The Role of Health Discussion Partners in Recovery from Mental Illness .” Social Science & Medicine 125 : 116 – 128 . CrossRef Search ADS Pew Research Center . 2014 . “ Religious Landscape Study .” Pew Research Center’s Religion & Public Life Project . http://www.pewforum.org/religious-landscape-study/. Puetz , Kyle . 2015 . “ Consumer Culture, Taste Preferences, and Social Network Formation .” Sociology Compass 9 : 438 – 49 . Google Scholar CrossRef Search ADS Putnam , Robert D. , and David E. Campbell . 2010 . American Grace: How Religion Divides and Unites Us . New York, NY : Simon and Schuster . Schafer , Markus H . 2015 . “ Religiously Traditional, Unusually Supportive? Examining Who Gives, Helps, and Advises in Americans’ Close Networks .” Social Currents 2 : 81 – 104 . Google Scholar CrossRef Search ADS Schafer , Markus H. , and Nicholas Vargas . 2016 . “ The Dynamics of Social Support Inequality: Maintenance Gaps by Socioeconomic Status and Race ?” Social Forces 94 : 1795 – 822 . Google Scholar CrossRef Search ADS Small , Mario Luis . 2013 . “ Weak Ties and the Core Discussion Network: Why People Regularly Discuss Important Matters with Unimportant Alters .” Social Networks 35 : 470 – 83 . Google Scholar CrossRef Search ADS Smith , Christian . 2014 . American Evangelicalism: Embattled and Thriving . Chicago, IL : University of Chicago Press . Steensland , Brian , Lynn D. Robinson , W. Bradford Wilcox , Jerry Z. Park , Mark D. Regnerus , and Robert D. Woodberry . 2000 . “ The Measure of American Religion: Toward Improving the State of the Art .” Social Forces 79 : 291 – 318 .[AU: ] Google Scholar CrossRef Search ADS Stetzer , Ed . 2015 . “ Religious Polarization on the Way .” USA Today , November 8 . Suitor , J. Jill , Barry Wellman , and David L. Morgan . 1997 . “ It’s about Time: How, Why, and When Networks Change .” Social Networks 1 : 1 – 7 . Google Scholar CrossRef Search ADS Suitor , Jill , and Shirley Keeton . 1997 . “ Once a Friend, Always a Friend? Effects of Homophily on Women’s Support Networks across a Decade .” Social Networks 19 : 51 – 62 . Google Scholar CrossRef Search ADS Tabak , Melanie A. , and Kristin D. Mickelson . 2009 . “ Religious Service Attendance and Distress: The Moderating Role of Stressful Life Events and Race/Ethnicity .” Sociology of Religion 70 : 49 – 64 . Google Scholar CrossRef Search ADS Thompson , Edward H . 1991 . “ Beneath the Status Characteristic: Gender Variations in Religiousness .” Journal for the Scientific Study of Religion 30 : 381 – 94 . Google Scholar CrossRef Search ADS Thye , Shane R. , Jeongkoo Yoon , and Edward J. Lawler . 2002 . “ The Theory of Relational Cohesion: Review of a Research Program .” Advances in Group Processes 19 : 139 – 66 . Google Scholar CrossRef Search ADS van Duijn , Marijtje A. J. , Jooske T. van Busschbach , and Tom A. B. Snijders . 1999 . “ Multilevel Analysis of Personal Networks as Dependent Variables .” Social Networks 21 : 187 – 210 . Google Scholar CrossRef Search ADS Voas , David , and Mark Chaves . 2016 . “ Is the United States a Counterexample to the Secularization Thesis ?” American Journal of Sociology 121 : 1517 – 56 . Wei , Dedong , and Eric Y. Liu . 2013 . “ Religious Involvement and Depression: Evidence for Curvilinear and Stress-Moderating Effects among Young Women in Rural China .” Journal for the Scientific Study of Religion 522 : 349 – 67 . Google Scholar CrossRef Search ADS © The Author(s) 2018. Published by Oxford University Press on behalf of the Association for the Sociology of Religion. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Sociology of Religion Oxford University Press

Who Talks Religion and What Are the Consequences for Social Ties? Unpacking a Sensitive Discussion Topic in Close Networks

Sociology of Religion , Volume Advance Article – Feb 17, 2018

Loading next page...
 
/lp/ou_press/who-talks-religion-and-what-are-the-consequences-for-social-ties-dxoTgTsFn0
Publisher
Oxford University Press
Copyright
© The Author(s) 2018. Published by Oxford University Press on behalf of the Association for the Sociology of Religion. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com.
ISSN
1069-4404
eISSN
1759-8818
D.O.I.
10.1093/socrel/srx069
Publisher site
See Article on Publisher Site

Abstract

Abstract Though religion matters greatly to many U.S. adults, it is widely considered a touchy conversational topic. Understanding how religious issues are talked about with others can elucidate a key interpersonal manifestation of Americans’ faith, yet existing research has largely overlooked the phenomena of religious discussion in social networks. This article considers which types of people talk about religion with their close ties, what relational factors underlie religious discussion, and what implications such discussion has for network turnover and stability. Applying multilevel regression methods to ego-centered networks measured in the Portraits of American Life Survey, I find that individual- and relational-level factors each predict the presence of religious discussion within close networks. Longitudinal analyses further reveal that for a large subset of Americans—namely evangelical and mainline Protestants and those involved in a congregation—religious discussion partners are especially likely to remain in the network over the course of 6 years. This association extended beyond other factors that could explain tie persistence, including relational closeness and multiple forms of homophily. Results point to several promising future directions for the study of religion and social networks. Religion is widely deemed a topic inappropriate for polite social conversation (Green 2016; Martin 1998). But salient as it is in the lives of so many believers and in the course of so many cultural and political debates (Putnam and Campbell 2010), religion seems like something Americans are inclined to talk about—with or without the etiquette experts’ blessing. And in fact, evidence suggests that just over 50% of all U.S. adults indeed discuss religion with someone outside of their family at least several times a year, with the likelihood of such conversation soaring to 83% among those who see themselves as highly religious (Cooperman 2016). From a dyadic perspective, U.S. adults report talking religion and spirituality in about half of all their (nonresidential) close tie relationships (Merino 2014). Still, existing knowledge on networks and religious discussion pushes little beyond these basic descriptive patterns. Among what types of people and in what types of relationships does the topic of religion arise? And what are the implications of religion discussions for people’s networks over time? These are questions that have yet to be clearly answered. The egocentric name-generator paradigm often employed in survey research is well-suited to identify that set of tight-knit, often homogenous ties that exerts normative influence on people’s lives and that can be called upon for support in times of need (Burt 1984; Marsden 1987).1 Still, what actually gets talked about inside these close networks—finances? current events? religion?—is glossed over in much empirical research (Bearman and Parigi 2004). To address the underlying heterogeneity of conversational topics and the possibility of topic-alter dependency within close networks, researchers have begun probing the particularities of what people discuss with their close ties (Brashears 2014; Gerber et al. 2012). Health and politics are two subject matters increasingly assessed as common discussion topics with implications for numerous important outcomes (Klofstad et al. 2009; Perry and Pescosolido 2010). Recent research, for instance, shows that health is discussed with only some members of the close network, and there is considerable between-person variability in how freely individuals talk about health issues with members of their network (Perry and Pescosolido 2010). What is more, discussing health with network members is associated with effective chronic disease detection and management (Cornwell and Waite 2012) and optimal recovery from mental health crises (Perry and Pescosolido 2015). On the politics front, recent research documents that people tend to report discussing politics between “sometimes” and “often” with friends and family (Gerber et al. 2012), yet these conversations occur most frequently and with least trepidation between people sharing political views (Cowan and Baldassarri 2017). Talking politics with members of one’s network appears to boost civic participation (Klofstad 2007; McClurg 2003), but the tendency to broach such topics mainly with like-minded partisans can magnify the impression of pervasive political polarization (Baldassarri and Bearman 2007). Despite its importance in the lives of many American adults, religion has not yet received comparable attention as an important discussion topic in people’s networks. This study represents one of the few studies to address that gap, first considering what personal and relational characteristics predict religious/spiritual discussion in U.S. adults’ close networks. The study’s second aim is to examine whether talking religion predicts the continuation of alters in a person’s network over time. Personal networks evolve amidst changing life circumstances (Morgan et al. 1996; Suitor et al. 1997), yet various relational characteristics act as relational adhesives through various transitions in the life course (Feld et al. 2007; Lubbers et al. 2010; Suitor and Keeton 1997). As an initial exploration of how religious discussion may matter for personal networks, I take up the question of network turnover and whether talking religion is associated with tie durability over the course of 6 years. Variation in religious discussion may be related to a host of other important outcomes—including the extent to which people tolerate divergent religious viewpoints and the extent to which they understand their community or society to be one marked by cultural concord or polarization. These matters cannot be empirically addressed in this study, but I hope that the basic findings presented below can help stimulate such investigations. WHO TALKS RELIGION? BACKGROUND AND THEORY By any measure, the United States is an unusually religious society among other advanced countries (Putnam and Campbell 2010). Yet there is considerable pluralism of belief and nonbelief within the American population (Machacek 2003); a classic solution to this scenario has been that set of norms relegating religion to the domain of out-of-bounds conversation topics. Still, many Americans do discuss religion, particularly with those in their close network (Merino 2014). At this point, there is no single overarching theory to explain why religious topics do or do not emerge within such networks. The following section first integrates existing theory on tie mobilization and selective confiding (Small 2013) to identify the relational processes that may underlie religious discussion. Then, I consult existing research on individual religiosity to generate empirical expectations. Relational Factors Talking religion is something that happens in a relational context, and so many of the determinants of the phenomena are likely to be dyadic in nature. I draw from three distinctive theoretical perspectives on the reasons people confide in network members (Small 2013)—topic-alter dependency, the strong ties framework, and opportune mobilization—to posit relational hypothesis about religious discussion in close networks. The premise of topic-alter dependency is that people match discussion topics with certain types of people in their personal network (Bearman and Parigi 2004). Role relations are one determining factor for this conversational calibration,2 but in the context of religious discussion, homophily may be the more crucial consideration. Specifically, I anticipate that people are more likely to talk religion with network members who share their religious background than with religiously dissimilar people. For the faithful, shared understandings and a common religious vernacular should make it easier to broach the touchy topic of religion and spirituality. By the same token, nonreligious people may find it most comfortable to discuss religion with fellow outsiders—even as the overall prevalence of such discussions may be less frequent for this group relative to religious believers. Each scenario is consistent with topic-alter dependency, but also with common understandings of homophily in social relationships. Homophily is thought to go hand-in-hand with smooth, comprehensible communication (McPherson et al. 2001), and we might expect this to be particularly true for sensitive and/or controversial conversation topics (e.g., religion).3 And indeed, prior research suggests that religious agreement predicts more frequent religious discussion among network members (Gerber et al. 2012). A second relational explanation of why networked people discuss religion comes from the strong ties perspective. This position implies that sensitive matters are discussed only with one’s most trusted confidants (Small 2013). To the extent that religion and spirituality is a delicate topic, dyadic indicators of intimacy and emotional closeness should then override other relational factors. Granovetter (1973:1361) famously defined a tie’s strength as some “combination of the amount of time, the emotional intensity, the intimacy (mutual confiding), and the reciprocal services which characterize the tie.” To investigate this second perspective, I will consider a series of predictors signaling such features of strong ties.4 Finally, the opportune mobilization argument contends that discussion topics are not primarily governed by ego–alter commonality or by a tie’s strength, but by who happens to be around for conversation. In other words, discussions—including deeps ones concerning religion or the spiritual—are ultimately practical affairs, “responsive to contextual opportunities” (Small 2013:473). Hence, we might expect people to talk religion with those who are around, due either due geographical proximity (i.e., the tie lives nearby) or the regularity of communication (in person, or enabled by technology). Individual-Level Factors Talking religion with a given network member is also likely a function of numerous individual-level factors. People who consider religion to be important in their lives and/or demonstrate it behaviorally through involvement in a congregation should tend to talk about religion with the people they deem important (Cooperman 2016). Religious affiliation may also matter. At least in the U.S. context, evangelical Protestants are known for emphasizing God’s direct, personal involvement in their everyday life and for telling others about their faith in the course of spreading the gospel message (Smith 2014). This may mean that, all else equal, Evangelicals are more likely than people with other (or no) religious affiliation to talk religion with members of their close network. It is also important to consider the impact of demographic characteristics. Across a variety of indicators, U.S. women and racial/ethnic minorities appear more religious than men and White people (Pew Research Center 2014), so we may expect network religious discussion to figure more prominently in the lives of the former groups relative to others. According to this logic, indicators of religiosity may in turn account for such patterns. Age/cohort is a third relevant characteristic. More recent cohorts of North Americans report strikingly lower levels of religious affiliation, church attendance, belief certitude, and importance of religion/spirituality than previous cohorts (Voas and Chaves 2016). Regardless of whether younger people are themselves religious, their generational milieu is one of secularization where a relatively small sliver of the cohort places much importance on religious matters. This may well mean a falling off of religious discussion across the age range of American adults. Finally, I will examine whether stressful life situations elicit religious and spiritual discussion in people’s networks. Theories of stress-buffering maintain that people facing demanding circumstances and who feel unable to cope through their own personal resources turn often to religion for consolation (Pargament 2002; Tabak and Mickelson 2009; Wei and Lui 2013). One outcome of this process may be the increased likelihood of talking about religion with close network members. Of course, the decision to talk religion with network members amidst personal adversity may hinge on what type of religiosity (if any) the individual embraces. Accordingly, my analysis will consider whether religious participation and/or affiliation dictates the impact of stressful circumstances. DOES RELIGIOUS DISCUSSION MATTER? EXPLORING IMPLICATIONS FOR TIE PERSISTENCE Describing the individual and relational predictors of religious discussion can elucidate how a highly salient—though controversial—aspect of American life unfolds in people’s close social environment. But a natural follow-up question is why this might matter for the lives of individual or for their networks. There are multiple ways one could pursue this question; as a first step, this article will take up the issue of how religious discussion intersects with network turnover. Network turnover is a ubiquitous process; life course transitions, job changes, and residential mobility effectively cycle various friends, neighbors, colleagues, and romantic partners in and out of people’s core networks (Bidart and Lavenu 2005; Morgan et al. 1996; Wellman et al. 1997). One recent analysis, for instance, reports that only 5% of older American adults retain an identical roster of close associates over a 5-year period, while nearly 60% added multiple confidants within that time frame (Cornwell and Laumann 2015). Another study using a representative U.S. sample suggests that among their four closest nonresidential associates, the typical adult swaps out and replaces two ties over the course of 6 years (Schafer and Vargas 2016). Ubiquitous network turnover is an important topic because, depending on the circumstances, it can bring about both risks and benefits. Adding new network members can boost self-esteem and enhance health (Cornwell and Laumann 2015), yet losing close ties may also disrupt stable access to resources and exacerbate inequality between income groups (Cornwell 2015; Schafer and Vargas 2016). The question for this study is whether talking religion factors into these pervasive dynamics of network change. Little research on network turnover has examined whether discussion content is associated with tie persistence, but there is reason to anticipate that for at least some segment of the population, discussing things religious will increase the likelihood that a given network member will remain in the close network over time. Social penetration theory offers a developmental account of relationships (Altman and Taylor 1973). The theory assumes that people have an “onion-layer” like set of beliefs, values, and commitments that come to be revealed to another person over time, the inner core peeled open only after conversation has moved beyond the outer, more superficial rings of the self. In this theory, people open up about sensitive matters if they believe that the rewards of deeper intimacy will outweigh the costs. And so while rapport and trust are typically a necessary starting point for increased disclosure, moving to “deeper” areas permits yet more intimacy and indicates that one forecasts the increased social penetration as a rewarding experience. In this way, social penetration shares similar expectations to the class of exchange-based theories which contend that closeness motivates exchange, but also that exchange gestures can boost positive sentiment and ultimately enhance solidarity within a dyad; for further reference, see accounts of relational cohesion theory (Thye et al. 2002) and accounts of “social-formation” in exchange theory more broadly (Lawler and Thye 1999). Continuation of the relationship is one of the main outcomes assessed by social penetration theory. Social penetration processes have an important affinity with religion and spirituality. For the faithful, religion concerns that which is sacred or of ultimate significance. Sacred matters, in turn, are imbued with powerful emotions, tend to be protected with great effort, and are intersubjective social resources—phenomena shared with other people that produce shared meaning and cohesion (Pargament and Mahoney 2005). For the faithful, then, dialogue about religion amounts to a “central, intimate facet of personality” and touches down on core issues of personal vulnerability (Altman and Taylor 1973:58). And in light of the continuous, reciprocal development of disclosure and intimacy (Altman and Taylor 1973:52; Collins and Miller 1994), religion may well represent a topical adhesive that provides a unique bond across time and space, one that stands out among other conventional markers of emotional closeness or intimacy typically assessed in studies of strong ties and network maintenance.5 In this way, religious discussion could be thought of as a potential cause and a consequence of tie closeness, but also a qualitatively different form of tie strength that predicts tie continuity. Developmental relationship models such as social penetration theory help clarify several important caveats and distinctions. First, people without a religious affiliation and/or those who do not participate in religious activity may well talk about religion from time to time, but such discussion is unlikely to touch on matters deemed sacred by the discussant. Therefore, we should expect the discussion of religious topics to be no more binding than the nondiscussion of such topics for this segment of the population. Second, the theory prompts consideration of whether talking religion with people outside of one’s own faith tradition is associated with tie durability. Intimacy is “simultaneously rewarding and risky” (Collins and Miller 1994:469), and the costs of delving into religion with a religiously dissimilar associate may outpace forecasted rewards of that disclosure. Indeed, overt conversation patterns and religious homophily could form multiplicative—not additive—associations with tie persistence. That is, religious homophily, as a form of cultural similarity, could itself prove a bonding agent in network relationships (Puetz 2015:446). And talking religion with an alter who shares one’s faith could further enhance intimacy and increase the likelihood that such a network member sticks around. Broaching such topics with a religious outsider, however, could potentially decrease chances of the tie remaining in one’s close network—particularly if the conversation was straining or awkward (Dallas 2016). One study, for instance, finds that close to 12% of Americans report having lost a friend over religious disagreement (Gerber et al. 2013). This analysis will therefore consider whether the link between religious discussion and tie persistence is contingent upon religious homophily. SUMMARY OF RESEARCH QUESTIONS In summary, two research questions guide the ensuing analyses: Research Question 1: What relational- and individual-level factors predict religious discussion in close networks? I hypothesize that religiosity (congregational involvement and affiliating as an Evangelical vs. no affiliation) is positively associated with religious discussion. In addition, I hypothesize that being older, female, and Black increases likelihood of religious discussion, as do exposures to stressors. Relational factors hypothesized to predict religious discussion include religious homophily, tie strength, and conversational opportunity. Research Question 2: Does religious discussion predict the persistence of ties in people’s close networks over time? Discussing religion is hypothesized to increase tie persistence, but only among people involved in a congregation and among those who embrace a religious affiliation. I further hypothesize that religious homophily will strengthen the association between religious discussion and tie persistence. DATA AND METHODS Sample Data for this study come from the Portraits of American Life Study (PALS), the only nationally representative panel study of which I am aware that asks about religious discussion in people’s close networks. The PALS first surveyed 2,610 noninstitutionalized American adults at least 18 years of age in 2006 (W1). The sample was obtained through a multistage process, in which zip code areas were randomly selected with probability proportionate to size, addresses were randomly selected from each zip code area, and one randomly selected adult was selected for a full interview from each selected household (after being screened for eligibility in an initial interview). The baseline survey produced an 83% contact rate, an 86% screening rate, and an 82% cooperation rate, resulting in an overall W1 response rate of 58% (0.83 × 0.86 × 0.82). With the available survey weights, the PALS sample closely mirrors basic population patterns found in the Census’ American Community Survey (Emerson et al. 2010). Further information about the PALS sampling design and recruitment is available at http://www.palsresearch.org/pals/researchers. Analyses from W1 come from the social ties among 2,435 PALS respondents who identified at least one close network tie outside of their home. Apart from the participants who reported no network members, missing data were minimal (<2% across all variables). Initial participants were re-contacted and 1,314 of them were interviewed again in 2011 (W2), yielding a 50.3% retention rate. Participants not successfully re-interviewed had larger average network size than re-interviewed participants and were disproportionately male, less-educated, lower-income, and non-White; there were not statistical differences in average age or county population size between the two groups. Those involved in a congregation at W1 were slightly more likely to be re-interviewed at W2 than noncongregants (52.9% vs. 48.3% retention), but there was no statistical difference by respondent’s religious tradition. To account for the potential biases of nonrandom attrition in the longitudinal analysis, I created inverse probability of attrition (IPA) weights.6 Longitudinal results were consistent whether using the conventional sampling weights or applying the IPA weighting scheme. Relational Measures The close networks of PALS respondents were constructed through a name-generator technique. Participants were first asked to identify up to four people “outside of your home you feel closest to.”7 They were then asked a series of questions about these network alters, the most central for this analysis being whether the person was one with whom ego “discussed religious or spiritual matters in the past 12 months.” The network name-interpreter battery also assessed alters’ demographic profile, including gender (male or female), race (whether “different race” or the same as ego), education level (4-year college degree or more vs. other), and kinship status (family members vs. other), as well as whether the tie is a coworker. Religious homophily was operationalized as whether the alter “shared [ego’s] religious faith.” Tie strength was first assessed by position in the network roster under the assumption that “the order in which [alters] were named is a reflection of [tie] strength” (Merino 2014:603). Given that Granovetter’s (1973) conception of strong ties includes the presence of confiding and supportive exchange, I also use measures of whether ego received advice (alter “provided advice that helped you make an important decision in the past 3 years”) or help (alter “volunteered their time to help you in times of need in the past 3 years”) to proxy tie strength. To capture opportunity mobilization, I use measures of whether the alter lives nearby (“within a 20 minute drive of your home”) and whether the duo interact at least once a week (in person or with technology).8 Besides assessing religious discussion, the PALS is a unique survey in that it determines tie persistence in a national sample across two waves. At W2, participants were asked to construct their ego network using the same name-generator technique as at baseline. Respondents were then asked to indicate whether present network members corresponded to any of the ties identified at W1. Respondents were shown their W1 network roster and asked whether each of the W2 names were distinct from initial members. From this information, I devised a dummy variable denoting whether the W1 alter persisted to W2. Respondent-Level Measures Individual-level predictors of religious discussion (also used as covariates in the analysis of tie persistence) include religious factors, demographic characteristics, and stressors. Congregational involvement was measured with the question “Are you currently involved in, affiliated with, or a member of a religious congregation or other place of worship?”9 Religious affiliation was categorized using the often-used scheme devised by Steensland et al. (2000); I distinguish Evangelicals, Black Protestants, Mainline Protestants, Catholics, other religion, and no religious affiliation (referred to as religious “nones”) in this study. For demographics, I account for age (years since birth at 2006), gender, race/ethnicity (White, Black, Latino non-Black, and other). A series of three dummy variables denoted whether the respondent had a variety of stress exposures in the past 3 years: “serious illness, injury, or an assault”; “a major financial crisis”; and the death of a family member or friend (“parent, child, or spouse died” or “a close family friend or relative (aunt, cousin, grandparent) died”). Additional controls include educational attainment (years of formal education), marital status (married vs. not married), working for pay (vs. not), U.S. census region (Northeast, Midwest, South, West), and county population size (the latter two variables to adjust for potential regional and/or urban/rural variation in religious discussion practices). Longitudinal analyses of network turnover further adjust for whether the respondent moved, changed working status, or changed marital status between waves. Preliminary analyses also adjusted for political affiliation, income, and parental status, but accounting for these variables had no influence on the study findings. Analysis The analysis proceeds in two main stages. After providing a descriptive account of sample statistics, I first assess the individual and relational factors, which predict religious discussion between ego (the PALS participant) and his/her network members. The outcome is a tie-level variable, so I use the multilevel modeling approach for ego-network data (van Duijn et al. 1999), where network alters (level 1 observations) are nested within the W1 PALS participants (level 2 observations). This random-intercept logistic regression model accounts for clustering of multiple observations within each respondent and incorporates predictor variables corresponding to multiple levels of analysis. The model is initially fit to the full multilevel sample, but I then examine parameter estimates separately according to egos’ religious tradition and separately for those who belong and do not belong to a congregation. This is to more fully understand whether the stress-consolation dynamics of religious discussion and/or the importance of religious homophily differs according to personal religiosity. The second stage of the analysis relies on the longitudinal sample and tests whether discussing religion is associated with tie persistence. For the PALS respondents re-interviewed in 2011, a binary variable denotes whether each W1 network alter remained in the ego network at W2 (33% of eligible W1 ties reappear in W2 participant networks). This analysis also uses a multilevel binary logistic model, incorporating baseline participant and relational characteristics to predict the odds of a given tie enduring over the course of 6 years. Consistent with the strategy mentioned above, the model is fit to both the full longitudinal sample and to subsets of the PALS sample that reflect different expressions of religious fidelity (or none whatsoever). This is primarily to assess whether talking religion is more consequential for more religious people. In addition, I will test an interaction term between religious discussion and religious homophily, both in the full sample and in the selected subsamples—this to assess whether talking religion has different implications for tie persistence depending on whether ego and alter share the same religious background. All analyses use Stata version 12 and incorporate the recommended survey weights to generalize findings to the noninstitutionalized population of U.S. adults. Robust standard errors account for multistage survey design. RESULTS Weighted descriptive statistics are shown for PALS respondents (table 1) and for the social network dyads (table 2) embedded in these survey participants. Nearly half of the baseline survey respondents were part of a religious congregation. Evangelical Protestants and Catholics were the two most common religious affiliations, both representing just over a quarter of the sample. Religious nones comprised 16% of the sample. Dyadic-level data reveal that religious discussion transpired within about half of all ego–alter relationships. There was strong evidence of religious homophily—about two-thirds of all ties were between people sharing a religious background—but this type of commonality was less pronounced than racial and gender homophily (92% and 72%, respectively). Table 1 Weighted Descriptive Statistics for Individual Level Variables (Portraits of American Life Survey, W1) Mean/prop. Standard deviation Part of cong. 0.47 Evangelical 0.27 Black Protestant 0.05 Mainline Protestant 0.14 Catholic 0.26 Other religion 0.12 No religion (nones) 0.16 Female 0.52 Black 0.11 Latino/a 0.12 Other race 0.05 Age 45.12 16.47 Serious illness/injury 0.22 Death of loved one 0.67 Financial crisis 0.23 Midwest 0.23 South 0.34 West 0.26 Northeast 0.17 County pop. Size 4.87 2.4 Education 14.46 2.84 Married 0.58 Worker 0.64 Mean/prop. Standard deviation Part of cong. 0.47 Evangelical 0.27 Black Protestant 0.05 Mainline Protestant 0.14 Catholic 0.26 Other religion 0.12 No religion (nones) 0.16 Female 0.52 Black 0.11 Latino/a 0.12 Other race 0.05 Age 45.12 16.47 Serious illness/injury 0.22 Death of loved one 0.67 Financial crisis 0.23 Midwest 0.23 South 0.34 West 0.26 Northeast 0.17 County pop. Size 4.87 2.4 Education 14.46 2.84 Married 0.58 Worker 0.64 Notes: n = 2,435. Standard deviations are omitted for categorical variables. View Large Table 1 Weighted Descriptive Statistics for Individual Level Variables (Portraits of American Life Survey, W1) Mean/prop. Standard deviation Part of cong. 0.47 Evangelical 0.27 Black Protestant 0.05 Mainline Protestant 0.14 Catholic 0.26 Other religion 0.12 No religion (nones) 0.16 Female 0.52 Black 0.11 Latino/a 0.12 Other race 0.05 Age 45.12 16.47 Serious illness/injury 0.22 Death of loved one 0.67 Financial crisis 0.23 Midwest 0.23 South 0.34 West 0.26 Northeast 0.17 County pop. Size 4.87 2.4 Education 14.46 2.84 Married 0.58 Worker 0.64 Mean/prop. Standard deviation Part of cong. 0.47 Evangelical 0.27 Black Protestant 0.05 Mainline Protestant 0.14 Catholic 0.26 Other religion 0.12 No religion (nones) 0.16 Female 0.52 Black 0.11 Latino/a 0.12 Other race 0.05 Age 45.12 16.47 Serious illness/injury 0.22 Death of loved one 0.67 Financial crisis 0.23 Midwest 0.23 South 0.34 West 0.26 Northeast 0.17 County pop. Size 4.87 2.4 Education 14.46 2.84 Married 0.58 Worker 0.64 Notes: n = 2,435. Standard deviations are omitted for categorical variables. View Large Table 2 Weighted Descriptive Statistics for Relational Level Variables (Portraits of American Life Survey, W1) Proportion Discussion religion 0.49 Religious homophily 0.65 Alter rank 1 0.27 Alter rank 2 0.26 Alter rank 3 0.25 Alter rank 4 0.22 Alter gave advice 0.58 Alter gave help 0.79 Lives nearby 0.52 Interact weekly 0.79 Racial homophily 0.92 Gender homophily 0.72 Alter has bachelor degree 0.34 Kin 0.46 Co-worker 0.09 Tie remain, W1–W2 0.33 Proportion Discussion religion 0.49 Religious homophily 0.65 Alter rank 1 0.27 Alter rank 2 0.26 Alter rank 3 0.25 Alter rank 4 0.22 Alter gave advice 0.58 Alter gave help 0.79 Lives nearby 0.52 Interact weekly 0.79 Racial homophily 0.92 Gender homophily 0.72 Alter has bachelor degree 0.34 Kin 0.46 Co-worker 0.09 Tie remain, W1–W2 0.33 Note: n = 8,999 (within 2,435 individual respondents). View Large Table 2 Weighted Descriptive Statistics for Relational Level Variables (Portraits of American Life Survey, W1) Proportion Discussion religion 0.49 Religious homophily 0.65 Alter rank 1 0.27 Alter rank 2 0.26 Alter rank 3 0.25 Alter rank 4 0.22 Alter gave advice 0.58 Alter gave help 0.79 Lives nearby 0.52 Interact weekly 0.79 Racial homophily 0.92 Gender homophily 0.72 Alter has bachelor degree 0.34 Kin 0.46 Co-worker 0.09 Tie remain, W1–W2 0.33 Proportion Discussion religion 0.49 Religious homophily 0.65 Alter rank 1 0.27 Alter rank 2 0.26 Alter rank 3 0.25 Alter rank 4 0.22 Alter gave advice 0.58 Alter gave help 0.79 Lives nearby 0.52 Interact weekly 0.79 Racial homophily 0.92 Gender homophily 0.72 Alter has bachelor degree 0.34 Kin 0.46 Co-worker 0.09 Tie remain, W1–W2 0.33 Note: n = 8,999 (within 2,435 individual respondents). View Large Discussing Religion Table 3 presents the full, main-sample analysis of discussing religion in a series of three models. This is done because the two core indicators of personal religiosity, congregational involvement and affiliation, overlap to a considerable degree (i.e., very few people with no affiliation are connected to a congregation). Therefore, model 1 focuses on congregational involvement, while model 2 substitutes in religious affiliation. Model 3 includes both dimensions of religiosity for the sake of comparability.10 Models 4–10 replicate the analysis for specific religious subsets of the sample.11 Table 3 Discussing Religion, Odds Ratio Estimates from Random Intercept Logistic Regression Models (Portraits of American Life Survey, W1) Full sample Part of congregation Not part of congregation Evangelicals Mainline Protestants Black Protestants Catholics Religious nones Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Model 8 Model 9 Model 10 Individual-level predictors  Part of cong. 3.79*** 4.23** 6.81*** 2.80*** 3.17** 3.00** 1.02 (2.84–5.07) (3.20–5.59) (3.51–13.21) (1.65–4.75) (1.34–7.53) (1.51–5.95) (0.32–3.24)  Evangelicala 1.84* 0.95 4.54* 0.64 (1.14–2.99) (0.60–1.49) (1.18–17.50) (0.37–1.12)  Black Protestanta 1.52 0.81 3.02 0.71 (0.82–2.83) (0.42–1.53) (0.66–13.78) (0.32–1.60)  Mainline Protestanta 0.66 0.34** 1.13 0.41** (0.42–1.03) (0.22–0.52) (0.32–4.00) (0.23–0.73)  Catholica 0.92 0.50** 1.82 0.49* - (0.56–1.51) (0.31–0.82) (0.42–7.88) (0.26–0.94)  Other 0.88 0.66 4.80* 0.50*  Religiona (0.44–1.77) (0.36–1.22) (1.26–18.32) (0.27–0.91)  Female 1.81*** 2.40*** 1.94** 1.35 2.49*** 1.91 2.56* 0.75 3.15** 1.00 (1.30–2.51) (1.60–3.58) (1.40–2.70) (0.70–2.59) (1.66–3.73) (0.95–3.83) (1.24–5.28) (0.21–2.64) (1.38–7.21) (0.39–2.58)  Blackb 0.90 0.88 0.77 0.60 1.16 0.63 0.81 0.41 0.86 (0.53–1.54) (0.51–1.51) (0.44–1.34) (0.30–1.20) (0.55–2.47) (0.23–1.73) (0.19–3.41) (0.16–1.06) (0.27–2.78)  Latino/ab 0.33*** 0.39*** 0.39** 0.25* 0.54 0.34** 0.64 0.45* 0.43 (0.21–0.52) (0.23–0.64) (0.24–0.64) (0.08–0.76) (0.28–1.06) (0.17–0.67) (0.15–2.68) (0.23–0.87) (0.10–1.80)  Other raceb 0.19*** 0.20*** 0.17** 0.13*** 0.18*** 0.14*** 0.37 0.14*** 0.55 (0.08–0.43) (0.08–0.46) (0.08–0.37) (0.05–0.35) (0.08–0.43) (0.06–0.33) (0.04–3.25) (0.05–0.40) (0.12–2.57)  Age 0.99 1.00 1.00 1.00 0.99 0.98 0.99 0.99 1.01 1.00 (0.98–1.00) (0.99–1.01) (0.99–1.01) (0.98–1.02) (0.98–1.01) (0.96–1.02) (0.96–1.02) (0.96–1.03) (0.99–1.03) (0.97–1.03)  Serious illness/injury 1.17 1.03 1.14 1.06 1.29 1.07 0.95 2.12 1.56 0.82 (0.78–1.77) (0.68–1.57) (0.78–1.68) (0.45–2.46) (0.83–1.99) (0.47–2.46) (0.40–2.27) (0.73–6.18) (0.81–3.01) (0.35–1.91)  Death of loved one 1.26 1.36 1.21 1.10 1.31 0.79 0.94 2.28 1.80* 1.04 (0.89–1.79) (0.97–1.90) (0.85–1.72) (0.61–1.99) (0.85–2.02) (0.42–1.52) (0.52–1.71) (0.82–6.33) (1.00–3.25) (0.43–2.53)  Financial crisis 1.33 1.22 1.30 1.32 1.33 1.17 2.03 1.73 1.09 2.35 (0.89–1.99) (0.79–1.90) (0.87–1.92) (0.59–2.96) (0.77–2.30) (0.65–2.11) (0.91–4.56) (0.57–5.27) (0.51–2.33) (0.57–9.69)  Midwestc 1.24 1.17 1.18 0.88 1.73 2.06 0.99 1.95 0.71 1.26 (0.73–2.10) (0.75–1.81) (0.72–1.91) (0.37–2.11) (0.92–3.25) (0.62–6.89) (0.34–2.86) (0.33–11.53) (0.33–1.51) (0.34–4.66)  Southc 2.06*** 1.88*** 1.79** 1.54 2.16** 2.34 1.31 1.68 1.00 1.95 (1.32–3.20) (1.29–2.73) (1.18–2.71) (0.72–3.32) (1.28–3.63) (0.76–7.16) (0.43–4.02) (0.33–8.65) (0.52–1.91) (0.56–6.77)  Westc 2.36*** 2.05** 2.33** 2.10 2.89*** 4.27* 2.00 1.22 1.28 1.92 (1.41–3.95) (1.24–3.39) (1.35–4.02) (0.62–7.12) (1.68–4.97) (1.09–16.72) (0.63–6.31) (0.28–5.24) (0.61–2.66) (0.49–7.61)  County pop. size 1.05 1.04 1.05 1.08 1.01 1.10 1.03 1.13 1.08 1.00 (0.96–1.15) (0.96–1.13) (0.96–1.15) (0.90–1.30) (0.90–1.14) (0.98–1.23) (0.89–1.18) (0.92–1.41) (0.98–1.19) (0.80–1.25)  Education 1.06* 1.11*** 1.08* 1.12 1.05 1.23** 1.01 1.27* 1.15** 1.13 (1.01–1.12) (1.04–1.18) (1.01–1.15) (0.96–1.31) (0.97–1.14) (1.06–1.43) (0.89–1.16) (1.02–1.59) (1.04–1.28) (0.94–1.37)  Married 0.94 1.02 0.96 1.05 0.84 1.41 0.96 1.31 1.13 0.54 (0.69–1.26) (0.76–1.38) (0.72–1.28) (0.63–1.73) (0.60–1.19) (0.83–2.39) (0.42–2.18) (0.59–2.91) (0.68–1.87) (0.23–1.26)  Worker 1.21 1.27 1.23 1.48 1.17 1.04 0.58 0.86 1.55 0.91 (0.83–1.75) (0.94–1.72) (0.90–1.69) (0.68–3.21) (0.69–1.98) (0.45–2.40) (0.26–1.28) (0.36–2.07) (0.84–2.86) (0.40–2.07) Relational-level predictors  Religious homophily 1.55*** 1.57** 1.58** 1.83*** 1.39 3.10*** 1.28 6.49*** 1.23 1.24 (1.24–1.93) (1.26–1.97) (1.27–1.96) (1.32–2.55) (0.99–1.94) (2.07–4.64) (0.71–2.32) (2.22–18.94) (0.75–2.01) (0.75–2.05)  Alter rank 1 1.88*** 1.87*** 1.88** 2.06*** 1.74** 1.61 1.33 1.94 3.81*** 1.46 (1.51–2.36) (1.50–2.34) (1.50–2.35) (1.46–2.90) (1.29–2.34) (0.86–3.00) (0.80–2.23) (0.97–3.88) (2.64–5.49) (0.76–2.82)  Alter rank 2 1.16 1.15 1.16 1.17 1.15 0.88 0.95 1.93 2.14*** 0.86 (0.92–1.45) (0.92–1.44) (0.92–1.45) (0.86–1.60) (0.84–1.59) (0.57–1.35) (0.51–1.77) (0.93–4.00) (1.47–3.12) (0.45–1.62)  Alter rank 3 0.99 0.99 0.99 1.07 0.92 0.81 1.15 1.51 1.19 0.78 (0.81–1.22) (0.80–1.21) (0.81–1.22) (0.78–1.49) (0.70–1.20) (0.47–1.39) (0.67–1.99) (0.86–2.67) (0.78–1.82) (0.46–1.30)  Alter gave advice 2.79*** 2.76*** 2.87** 3.09** 2.78*** 3.07*** 3.28*** 4.51*** 2.91*** 2.43** (2.21–3.52) (2.19–3.47) (2.26–3.65) (2.00–4.76) (2.04–3.79) (1.70–5.54) (2.00–5.37) (1.89–10.77) (1.59–5.30) (1.39–4.25)  Alter gave help 1.86*** 1.91*** 1.87** 1.90** 1.76*** 2.37*** 2.27* 1.80 1.36 2.10 (1.47–2.36) (1.48–2.47) (1.48–2.36) (1.29–2.80) (1.25–2.47) (1.51–3.73) (1.15–4.49) (0.91–3.56) (0.78–2.36) (0.89–4.94)  Lives nearby 1.40*** 1.43*** 1.41** 1.45* 1.35 1.44 1.10 1.26 1.39 1.20 (1.15–1.70) (1.18–1.74) (1.16–1.72) (1.07–1.96) (0.99–1.84) (0.90–2.28) (0.64–1.87) (0.62–2.57) (0.97–1.98) (0.65–2.23)  Interact weekly 1.73*** 1.74*** 1.71** 1.99** 1.56** 2.31** 2.01* 1.53 1.08 2.67** (1.32–2.26) (1.32–2.29) (1.31–2.24) (1.29–3.10) (1.14–2.12) (1.28–4.16) (1.00–4.05) (0.74–3.16) (0.65–1.78) (1.31–5.43)  Racial homophily 0.78 0.81 0.78 1.00 0.74 1.50 0.90 0.95 0.72 0.88 (0.50–1.20) (0.52–1.25) (0.50–1.20) (0.52–1.96) (0.45–1.23) (0.85–2.65) (0.34–2.38) (0.19–4.77) (0.22–2.33) (0.40–1.96)  Gender homophily 0.90 0.92 0.92 1.00 0.86 0.77 0.94 1.05 0.81 1.18 (0.72–1.14) (0.73–1.15) (0.73–1.16) (0.69–1.43) (0.66–1.12) (0.52–1.15) (0.58–1.54) (0.50–2.21) (0.47–1.38) (0.76–1.83)  Alter has bachelor dg. 1.32* 1.32* 1.32* 0.94 1.78*** 0.94 1.72 2.36* 1.03 1.70 (1.06–1.66) (1.05–1.65) (1.05–1.67) (0.67–1.31) (1.27–2.49) (0.59–1.52) (0.98–3.02) (1.07–5.24) (0.65–1.61) (0.85–3.40)  Kin 0.84 0.83 0.85 0.98 0.80 0.79 1.32 0.97 1.16 0.46* (0.69–1.01) (0.68–1.01) (0.71–1.02) (0.71–1.35) (0.60–1.05) (0.48–1.30) (0.75–2.32) (0.57–1.66) (0.79–1.70) (0.21–0.99)  Co-worker 0.90 0.89 0.89 1.20 0.79 0.66 1.45 2.45 0.96 0.80 (0.65–1.24) (0.64–1.26) (0.64–1.24) (0.70–2.07) (0.54–1.14) (0.30–1.45) (0.64–3.26) (0.20–29.47) (0.50–1.84) (0.33–1.93)  Level 1 5.23 5.45 5.17 6.48 4.24 7.26 3.18 5.37 5.73 4.10  Variance (0.50) (0.53) (0.51) (0.95) (0.56) (2.25) (0.71) (2.84) (1.27) (0.95)  −2LL 4841.96 4886.24 4821.47 2195.57 2591.21 1227.96 741.20 489.75 1189.65 788.78 Observations  Level 1 8,944 8,945 8,944 4,139 4,805 1,985 1,026 926 2,469 1,418  Level 2 2,419 2,420 2,419 1,102 1,317 529 270 256 683 384 Full sample Part of congregation Not part of congregation Evangelicals Mainline Protestants Black Protestants Catholics Religious nones Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Model 8 Model 9 Model 10 Individual-level predictors  Part of cong. 3.79*** 4.23** 6.81*** 2.80*** 3.17** 3.00** 1.02 (2.84–5.07) (3.20–5.59) (3.51–13.21) (1.65–4.75) (1.34–7.53) (1.51–5.95) (0.32–3.24)  Evangelicala 1.84* 0.95 4.54* 0.64 (1.14–2.99) (0.60–1.49) (1.18–17.50) (0.37–1.12)  Black Protestanta 1.52 0.81 3.02 0.71 (0.82–2.83) (0.42–1.53) (0.66–13.78) (0.32–1.60)  Mainline Protestanta 0.66 0.34** 1.13 0.41** (0.42–1.03) (0.22–0.52) (0.32–4.00) (0.23–0.73)  Catholica 0.92 0.50** 1.82 0.49* - (0.56–1.51) (0.31–0.82) (0.42–7.88) (0.26–0.94)  Other 0.88 0.66 4.80* 0.50*  Religiona (0.44–1.77) (0.36–1.22) (1.26–18.32) (0.27–0.91)  Female 1.81*** 2.40*** 1.94** 1.35 2.49*** 1.91 2.56* 0.75 3.15** 1.00 (1.30–2.51) (1.60–3.58) (1.40–2.70) (0.70–2.59) (1.66–3.73) (0.95–3.83) (1.24–5.28) (0.21–2.64) (1.38–7.21) (0.39–2.58)  Blackb 0.90 0.88 0.77 0.60 1.16 0.63 0.81 0.41 0.86 (0.53–1.54) (0.51–1.51) (0.44–1.34) (0.30–1.20) (0.55–2.47) (0.23–1.73) (0.19–3.41) (0.16–1.06) (0.27–2.78)  Latino/ab 0.33*** 0.39*** 0.39** 0.25* 0.54 0.34** 0.64 0.45* 0.43 (0.21–0.52) (0.23–0.64) (0.24–0.64) (0.08–0.76) (0.28–1.06) (0.17–0.67) (0.15–2.68) (0.23–0.87) (0.10–1.80)  Other raceb 0.19*** 0.20*** 0.17** 0.13*** 0.18*** 0.14*** 0.37 0.14*** 0.55 (0.08–0.43) (0.08–0.46) (0.08–0.37) (0.05–0.35) (0.08–0.43) (0.06–0.33) (0.04–3.25) (0.05–0.40) (0.12–2.57)  Age 0.99 1.00 1.00 1.00 0.99 0.98 0.99 0.99 1.01 1.00 (0.98–1.00) (0.99–1.01) (0.99–1.01) (0.98–1.02) (0.98–1.01) (0.96–1.02) (0.96–1.02) (0.96–1.03) (0.99–1.03) (0.97–1.03)  Serious illness/injury 1.17 1.03 1.14 1.06 1.29 1.07 0.95 2.12 1.56 0.82 (0.78–1.77) (0.68–1.57) (0.78–1.68) (0.45–2.46) (0.83–1.99) (0.47–2.46) (0.40–2.27) (0.73–6.18) (0.81–3.01) (0.35–1.91)  Death of loved one 1.26 1.36 1.21 1.10 1.31 0.79 0.94 2.28 1.80* 1.04 (0.89–1.79) (0.97–1.90) (0.85–1.72) (0.61–1.99) (0.85–2.02) (0.42–1.52) (0.52–1.71) (0.82–6.33) (1.00–3.25) (0.43–2.53)  Financial crisis 1.33 1.22 1.30 1.32 1.33 1.17 2.03 1.73 1.09 2.35 (0.89–1.99) (0.79–1.90) (0.87–1.92) (0.59–2.96) (0.77–2.30) (0.65–2.11) (0.91–4.56) (0.57–5.27) (0.51–2.33) (0.57–9.69)  Midwestc 1.24 1.17 1.18 0.88 1.73 2.06 0.99 1.95 0.71 1.26 (0.73–2.10) (0.75–1.81) (0.72–1.91) (0.37–2.11) (0.92–3.25) (0.62–6.89) (0.34–2.86) (0.33–11.53) (0.33–1.51) (0.34–4.66)  Southc 2.06*** 1.88*** 1.79** 1.54 2.16** 2.34 1.31 1.68 1.00 1.95 (1.32–3.20) (1.29–2.73) (1.18–2.71) (0.72–3.32) (1.28–3.63) (0.76–7.16) (0.43–4.02) (0.33–8.65) (0.52–1.91) (0.56–6.77)  Westc 2.36*** 2.05** 2.33** 2.10 2.89*** 4.27* 2.00 1.22 1.28 1.92 (1.41–3.95) (1.24–3.39) (1.35–4.02) (0.62–7.12) (1.68–4.97) (1.09–16.72) (0.63–6.31) (0.28–5.24) (0.61–2.66) (0.49–7.61)  County pop. size 1.05 1.04 1.05 1.08 1.01 1.10 1.03 1.13 1.08 1.00 (0.96–1.15) (0.96–1.13) (0.96–1.15) (0.90–1.30) (0.90–1.14) (0.98–1.23) (0.89–1.18) (0.92–1.41) (0.98–1.19) (0.80–1.25)  Education 1.06* 1.11*** 1.08* 1.12 1.05 1.23** 1.01 1.27* 1.15** 1.13 (1.01–1.12) (1.04–1.18) (1.01–1.15) (0.96–1.31) (0.97–1.14) (1.06–1.43) (0.89–1.16) (1.02–1.59) (1.04–1.28) (0.94–1.37)  Married 0.94 1.02 0.96 1.05 0.84 1.41 0.96 1.31 1.13 0.54 (0.69–1.26) (0.76–1.38) (0.72–1.28) (0.63–1.73) (0.60–1.19) (0.83–2.39) (0.42–2.18) (0.59–2.91) (0.68–1.87) (0.23–1.26)  Worker 1.21 1.27 1.23 1.48 1.17 1.04 0.58 0.86 1.55 0.91 (0.83–1.75) (0.94–1.72) (0.90–1.69) (0.68–3.21) (0.69–1.98) (0.45–2.40) (0.26–1.28) (0.36–2.07) (0.84–2.86) (0.40–2.07) Relational-level predictors  Religious homophily 1.55*** 1.57** 1.58** 1.83*** 1.39 3.10*** 1.28 6.49*** 1.23 1.24 (1.24–1.93) (1.26–1.97) (1.27–1.96) (1.32–2.55) (0.99–1.94) (2.07–4.64) (0.71–2.32) (2.22–18.94) (0.75–2.01) (0.75–2.05)  Alter rank 1 1.88*** 1.87*** 1.88** 2.06*** 1.74** 1.61 1.33 1.94 3.81*** 1.46 (1.51–2.36) (1.50–2.34) (1.50–2.35) (1.46–2.90) (1.29–2.34) (0.86–3.00) (0.80–2.23) (0.97–3.88) (2.64–5.49) (0.76–2.82)  Alter rank 2 1.16 1.15 1.16 1.17 1.15 0.88 0.95 1.93 2.14*** 0.86 (0.92–1.45) (0.92–1.44) (0.92–1.45) (0.86–1.60) (0.84–1.59) (0.57–1.35) (0.51–1.77) (0.93–4.00) (1.47–3.12) (0.45–1.62)  Alter rank 3 0.99 0.99 0.99 1.07 0.92 0.81 1.15 1.51 1.19 0.78 (0.81–1.22) (0.80–1.21) (0.81–1.22) (0.78–1.49) (0.70–1.20) (0.47–1.39) (0.67–1.99) (0.86–2.67) (0.78–1.82) (0.46–1.30)  Alter gave advice 2.79*** 2.76*** 2.87** 3.09** 2.78*** 3.07*** 3.28*** 4.51*** 2.91*** 2.43** (2.21–3.52) (2.19–3.47) (2.26–3.65) (2.00–4.76) (2.04–3.79) (1.70–5.54) (2.00–5.37) (1.89–10.77) (1.59–5.30) (1.39–4.25)  Alter gave help 1.86*** 1.91*** 1.87** 1.90** 1.76*** 2.37*** 2.27* 1.80 1.36 2.10 (1.47–2.36) (1.48–2.47) (1.48–2.36) (1.29–2.80) (1.25–2.47) (1.51–3.73) (1.15–4.49) (0.91–3.56) (0.78–2.36) (0.89–4.94)  Lives nearby 1.40*** 1.43*** 1.41** 1.45* 1.35 1.44 1.10 1.26 1.39 1.20 (1.15–1.70) (1.18–1.74) (1.16–1.72) (1.07–1.96) (0.99–1.84) (0.90–2.28) (0.64–1.87) (0.62–2.57) (0.97–1.98) (0.65–2.23)  Interact weekly 1.73*** 1.74*** 1.71** 1.99** 1.56** 2.31** 2.01* 1.53 1.08 2.67** (1.32–2.26) (1.32–2.29) (1.31–2.24) (1.29–3.10) (1.14–2.12) (1.28–4.16) (1.00–4.05) (0.74–3.16) (0.65–1.78) (1.31–5.43)  Racial homophily 0.78 0.81 0.78 1.00 0.74 1.50 0.90 0.95 0.72 0.88 (0.50–1.20) (0.52–1.25) (0.50–1.20) (0.52–1.96) (0.45–1.23) (0.85–2.65) (0.34–2.38) (0.19–4.77) (0.22–2.33) (0.40–1.96)  Gender homophily 0.90 0.92 0.92 1.00 0.86 0.77 0.94 1.05 0.81 1.18 (0.72–1.14) (0.73–1.15) (0.73–1.16) (0.69–1.43) (0.66–1.12) (0.52–1.15) (0.58–1.54) (0.50–2.21) (0.47–1.38) (0.76–1.83)  Alter has bachelor dg. 1.32* 1.32* 1.32* 0.94 1.78*** 0.94 1.72 2.36* 1.03 1.70 (1.06–1.66) (1.05–1.65) (1.05–1.67) (0.67–1.31) (1.27–2.49) (0.59–1.52) (0.98–3.02) (1.07–5.24) (0.65–1.61) (0.85–3.40)  Kin 0.84 0.83 0.85 0.98 0.80 0.79 1.32 0.97 1.16 0.46* (0.69–1.01) (0.68–1.01) (0.71–1.02) (0.71–1.35) (0.60–1.05) (0.48–1.30) (0.75–2.32) (0.57–1.66) (0.79–1.70) (0.21–0.99)  Co-worker 0.90 0.89 0.89 1.20 0.79 0.66 1.45 2.45 0.96 0.80 (0.65–1.24) (0.64–1.26) (0.64–1.24) (0.70–2.07) (0.54–1.14) (0.30–1.45) (0.64–3.26) (0.20–29.47) (0.50–1.84) (0.33–1.93)  Level 1 5.23 5.45 5.17 6.48 4.24 7.26 3.18 5.37 5.73 4.10  Variance (0.50) (0.53) (0.51) (0.95) (0.56) (2.25) (0.71) (2.84) (1.27) (0.95)  −2LL 4841.96 4886.24 4821.47 2195.57 2591.21 1227.96 741.20 489.75 1189.65 788.78 Observations  Level 1 8,944 8,945 8,944 4,139 4,805 1,985 1,026 926 2,469 1,418  Level 2 2,419 2,420 2,419 1,102 1,317 529 270 256 683 384 Notes: Robust standard errors are shown in parentheses. aReligions none is reference group. bWhite is reference group; Race/ethnicity not included in model 8 because the overwhelming majority of Black Protestants are Black. cNortheast is reference group. *p < .05; **p < .01; ***p < .001 (two-tailed tests). View Large Table 3 Discussing Religion, Odds Ratio Estimates from Random Intercept Logistic Regression Models (Portraits of American Life Survey, W1) Full sample Part of congregation Not part of congregation Evangelicals Mainline Protestants Black Protestants Catholics Religious nones Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Model 8 Model 9 Model 10 Individual-level predictors  Part of cong. 3.79*** 4.23** 6.81*** 2.80*** 3.17** 3.00** 1.02 (2.84–5.07) (3.20–5.59) (3.51–13.21) (1.65–4.75) (1.34–7.53) (1.51–5.95) (0.32–3.24)  Evangelicala 1.84* 0.95 4.54* 0.64 (1.14–2.99) (0.60–1.49) (1.18–17.50) (0.37–1.12)  Black Protestanta 1.52 0.81 3.02 0.71 (0.82–2.83) (0.42–1.53) (0.66–13.78) (0.32–1.60)  Mainline Protestanta 0.66 0.34** 1.13 0.41** (0.42–1.03) (0.22–0.52) (0.32–4.00) (0.23–0.73)  Catholica 0.92 0.50** 1.82 0.49* - (0.56–1.51) (0.31–0.82) (0.42–7.88) (0.26–0.94)  Other 0.88 0.66 4.80* 0.50*  Religiona (0.44–1.77) (0.36–1.22) (1.26–18.32) (0.27–0.91)  Female 1.81*** 2.40*** 1.94** 1.35 2.49*** 1.91 2.56* 0.75 3.15** 1.00 (1.30–2.51) (1.60–3.58) (1.40–2.70) (0.70–2.59) (1.66–3.73) (0.95–3.83) (1.24–5.28) (0.21–2.64) (1.38–7.21) (0.39–2.58)  Blackb 0.90 0.88 0.77 0.60 1.16 0.63 0.81 0.41 0.86 (0.53–1.54) (0.51–1.51) (0.44–1.34) (0.30–1.20) (0.55–2.47) (0.23–1.73) (0.19–3.41) (0.16–1.06) (0.27–2.78)  Latino/ab 0.33*** 0.39*** 0.39** 0.25* 0.54 0.34** 0.64 0.45* 0.43 (0.21–0.52) (0.23–0.64) (0.24–0.64) (0.08–0.76) (0.28–1.06) (0.17–0.67) (0.15–2.68) (0.23–0.87) (0.10–1.80)  Other raceb 0.19*** 0.20*** 0.17** 0.13*** 0.18*** 0.14*** 0.37 0.14*** 0.55 (0.08–0.43) (0.08–0.46) (0.08–0.37) (0.05–0.35) (0.08–0.43) (0.06–0.33) (0.04–3.25) (0.05–0.40) (0.12–2.57)  Age 0.99 1.00 1.00 1.00 0.99 0.98 0.99 0.99 1.01 1.00 (0.98–1.00) (0.99–1.01) (0.99–1.01) (0.98–1.02) (0.98–1.01) (0.96–1.02) (0.96–1.02) (0.96–1.03) (0.99–1.03) (0.97–1.03)  Serious illness/injury 1.17 1.03 1.14 1.06 1.29 1.07 0.95 2.12 1.56 0.82 (0.78–1.77) (0.68–1.57) (0.78–1.68) (0.45–2.46) (0.83–1.99) (0.47–2.46) (0.40–2.27) (0.73–6.18) (0.81–3.01) (0.35–1.91)  Death of loved one 1.26 1.36 1.21 1.10 1.31 0.79 0.94 2.28 1.80* 1.04 (0.89–1.79) (0.97–1.90) (0.85–1.72) (0.61–1.99) (0.85–2.02) (0.42–1.52) (0.52–1.71) (0.82–6.33) (1.00–3.25) (0.43–2.53)  Financial crisis 1.33 1.22 1.30 1.32 1.33 1.17 2.03 1.73 1.09 2.35 (0.89–1.99) (0.79–1.90) (0.87–1.92) (0.59–2.96) (0.77–2.30) (0.65–2.11) (0.91–4.56) (0.57–5.27) (0.51–2.33) (0.57–9.69)  Midwestc 1.24 1.17 1.18 0.88 1.73 2.06 0.99 1.95 0.71 1.26 (0.73–2.10) (0.75–1.81) (0.72–1.91) (0.37–2.11) (0.92–3.25) (0.62–6.89) (0.34–2.86) (0.33–11.53) (0.33–1.51) (0.34–4.66)  Southc 2.06*** 1.88*** 1.79** 1.54 2.16** 2.34 1.31 1.68 1.00 1.95 (1.32–3.20) (1.29–2.73) (1.18–2.71) (0.72–3.32) (1.28–3.63) (0.76–7.16) (0.43–4.02) (0.33–8.65) (0.52–1.91) (0.56–6.77)  Westc 2.36*** 2.05** 2.33** 2.10 2.89*** 4.27* 2.00 1.22 1.28 1.92 (1.41–3.95) (1.24–3.39) (1.35–4.02) (0.62–7.12) (1.68–4.97) (1.09–16.72) (0.63–6.31) (0.28–5.24) (0.61–2.66) (0.49–7.61)  County pop. size 1.05 1.04 1.05 1.08 1.01 1.10 1.03 1.13 1.08 1.00 (0.96–1.15) (0.96–1.13) (0.96–1.15) (0.90–1.30) (0.90–1.14) (0.98–1.23) (0.89–1.18) (0.92–1.41) (0.98–1.19) (0.80–1.25)  Education 1.06* 1.11*** 1.08* 1.12 1.05 1.23** 1.01 1.27* 1.15** 1.13 (1.01–1.12) (1.04–1.18) (1.01–1.15) (0.96–1.31) (0.97–1.14) (1.06–1.43) (0.89–1.16) (1.02–1.59) (1.04–1.28) (0.94–1.37)  Married 0.94 1.02 0.96 1.05 0.84 1.41 0.96 1.31 1.13 0.54 (0.69–1.26) (0.76–1.38) (0.72–1.28) (0.63–1.73) (0.60–1.19) (0.83–2.39) (0.42–2.18) (0.59–2.91) (0.68–1.87) (0.23–1.26)  Worker 1.21 1.27 1.23 1.48 1.17 1.04 0.58 0.86 1.55 0.91 (0.83–1.75) (0.94–1.72) (0.90–1.69) (0.68–3.21) (0.69–1.98) (0.45–2.40) (0.26–1.28) (0.36–2.07) (0.84–2.86) (0.40–2.07) Relational-level predictors  Religious homophily 1.55*** 1.57** 1.58** 1.83*** 1.39 3.10*** 1.28 6.49*** 1.23 1.24 (1.24–1.93) (1.26–1.97) (1.27–1.96) (1.32–2.55) (0.99–1.94) (2.07–4.64) (0.71–2.32) (2.22–18.94) (0.75–2.01) (0.75–2.05)  Alter rank 1 1.88*** 1.87*** 1.88** 2.06*** 1.74** 1.61 1.33 1.94 3.81*** 1.46 (1.51–2.36) (1.50–2.34) (1.50–2.35) (1.46–2.90) (1.29–2.34) (0.86–3.00) (0.80–2.23) (0.97–3.88) (2.64–5.49) (0.76–2.82)  Alter rank 2 1.16 1.15 1.16 1.17 1.15 0.88 0.95 1.93 2.14*** 0.86 (0.92–1.45) (0.92–1.44) (0.92–1.45) (0.86–1.60) (0.84–1.59) (0.57–1.35) (0.51–1.77) (0.93–4.00) (1.47–3.12) (0.45–1.62)  Alter rank 3 0.99 0.99 0.99 1.07 0.92 0.81 1.15 1.51 1.19 0.78 (0.81–1.22) (0.80–1.21) (0.81–1.22) (0.78–1.49) (0.70–1.20) (0.47–1.39) (0.67–1.99) (0.86–2.67) (0.78–1.82) (0.46–1.30)  Alter gave advice 2.79*** 2.76*** 2.87** 3.09** 2.78*** 3.07*** 3.28*** 4.51*** 2.91*** 2.43** (2.21–3.52) (2.19–3.47) (2.26–3.65) (2.00–4.76) (2.04–3.79) (1.70–5.54) (2.00–5.37) (1.89–10.77) (1.59–5.30) (1.39–4.25)  Alter gave help 1.86*** 1.91*** 1.87** 1.90** 1.76*** 2.37*** 2.27* 1.80 1.36 2.10 (1.47–2.36) (1.48–2.47) (1.48–2.36) (1.29–2.80) (1.25–2.47) (1.51–3.73) (1.15–4.49) (0.91–3.56) (0.78–2.36) (0.89–4.94)  Lives nearby 1.40*** 1.43*** 1.41** 1.45* 1.35 1.44 1.10 1.26 1.39 1.20 (1.15–1.70) (1.18–1.74) (1.16–1.72) (1.07–1.96) (0.99–1.84) (0.90–2.28) (0.64–1.87) (0.62–2.57) (0.97–1.98) (0.65–2.23)  Interact weekly 1.73*** 1.74*** 1.71** 1.99** 1.56** 2.31** 2.01* 1.53 1.08 2.67** (1.32–2.26) (1.32–2.29) (1.31–2.24) (1.29–3.10) (1.14–2.12) (1.28–4.16) (1.00–4.05) (0.74–3.16) (0.65–1.78) (1.31–5.43)  Racial homophily 0.78 0.81 0.78 1.00 0.74 1.50 0.90 0.95 0.72 0.88 (0.50–1.20) (0.52–1.25) (0.50–1.20) (0.52–1.96) (0.45–1.23) (0.85–2.65) (0.34–2.38) (0.19–4.77) (0.22–2.33) (0.40–1.96)  Gender homophily 0.90 0.92 0.92 1.00 0.86 0.77 0.94 1.05 0.81 1.18 (0.72–1.14) (0.73–1.15) (0.73–1.16) (0.69–1.43) (0.66–1.12) (0.52–1.15) (0.58–1.54) (0.50–2.21) (0.47–1.38) (0.76–1.83)  Alter has bachelor dg. 1.32* 1.32* 1.32* 0.94 1.78*** 0.94 1.72 2.36* 1.03 1.70 (1.06–1.66) (1.05–1.65) (1.05–1.67) (0.67–1.31) (1.27–2.49) (0.59–1.52) (0.98–3.02) (1.07–5.24) (0.65–1.61) (0.85–3.40)  Kin 0.84 0.83 0.85 0.98 0.80 0.79 1.32 0.97 1.16 0.46* (0.69–1.01) (0.68–1.01) (0.71–1.02) (0.71–1.35) (0.60–1.05) (0.48–1.30) (0.75–2.32) (0.57–1.66) (0.79–1.70) (0.21–0.99)  Co-worker 0.90 0.89 0.89 1.20 0.79 0.66 1.45 2.45 0.96 0.80 (0.65–1.24) (0.64–1.26) (0.64–1.24) (0.70–2.07) (0.54–1.14) (0.30–1.45) (0.64–3.26) (0.20–29.47) (0.50–1.84) (0.33–1.93)  Level 1 5.23 5.45 5.17 6.48 4.24 7.26 3.18 5.37 5.73 4.10  Variance (0.50) (0.53) (0.51) (0.95) (0.56) (2.25) (0.71) (2.84) (1.27) (0.95)  −2LL 4841.96 4886.24 4821.47 2195.57 2591.21 1227.96 741.20 489.75 1189.65 788.78 Observations  Level 1 8,944 8,945 8,944 4,139 4,805 1,985 1,026 926 2,469 1,418  Level 2 2,419 2,420 2,419 1,102 1,317 529 270 256 683 384 Full sample Part of congregation Not part of congregation Evangelicals Mainline Protestants Black Protestants Catholics Religious nones Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Model 8 Model 9 Model 10 Individual-level predictors  Part of cong. 3.79*** 4.23** 6.81*** 2.80*** 3.17** 3.00** 1.02 (2.84–5.07) (3.20–5.59) (3.51–13.21) (1.65–4.75) (1.34–7.53) (1.51–5.95) (0.32–3.24)  Evangelicala 1.84* 0.95 4.54* 0.64 (1.14–2.99) (0.60–1.49) (1.18–17.50) (0.37–1.12)  Black Protestanta 1.52 0.81 3.02 0.71 (0.82–2.83) (0.42–1.53) (0.66–13.78) (0.32–1.60)  Mainline Protestanta 0.66 0.34** 1.13 0.41** (0.42–1.03) (0.22–0.52) (0.32–4.00) (0.23–0.73)  Catholica 0.92 0.50** 1.82 0.49* - (0.56–1.51) (0.31–0.82) (0.42–7.88) (0.26–0.94)  Other 0.88 0.66 4.80* 0.50*  Religiona (0.44–1.77) (0.36–1.22) (1.26–18.32) (0.27–0.91)  Female 1.81*** 2.40*** 1.94** 1.35 2.49*** 1.91 2.56* 0.75 3.15** 1.00 (1.30–2.51) (1.60–3.58) (1.40–2.70) (0.70–2.59) (1.66–3.73) (0.95–3.83) (1.24–5.28) (0.21–2.64) (1.38–7.21) (0.39–2.58)  Blackb 0.90 0.88 0.77 0.60 1.16 0.63 0.81 0.41 0.86 (0.53–1.54) (0.51–1.51) (0.44–1.34) (0.30–1.20) (0.55–2.47) (0.23–1.73) (0.19–3.41) (0.16–1.06) (0.27–2.78)  Latino/ab 0.33*** 0.39*** 0.39** 0.25* 0.54 0.34** 0.64 0.45* 0.43 (0.21–0.52) (0.23–0.64) (0.24–0.64) (0.08–0.76) (0.28–1.06) (0.17–0.67) (0.15–2.68) (0.23–0.87) (0.10–1.80)  Other raceb 0.19*** 0.20*** 0.17** 0.13*** 0.18*** 0.14*** 0.37 0.14*** 0.55 (0.08–0.43) (0.08–0.46) (0.08–0.37) (0.05–0.35) (0.08–0.43) (0.06–0.33) (0.04–3.25) (0.05–0.40) (0.12–2.57)  Age 0.99 1.00 1.00 1.00 0.99 0.98 0.99 0.99 1.01 1.00 (0.98–1.00) (0.99–1.01) (0.99–1.01) (0.98–1.02) (0.98–1.01) (0.96–1.02) (0.96–1.02) (0.96–1.03) (0.99–1.03) (0.97–1.03)  Serious illness/injury 1.17 1.03 1.14 1.06 1.29 1.07 0.95 2.12 1.56 0.82 (0.78–1.77) (0.68–1.57) (0.78–1.68) (0.45–2.46) (0.83–1.99) (0.47–2.46) (0.40–2.27) (0.73–6.18) (0.81–3.01) (0.35–1.91)  Death of loved one 1.26 1.36 1.21 1.10 1.31 0.79 0.94 2.28 1.80* 1.04 (0.89–1.79) (0.97–1.90) (0.85–1.72) (0.61–1.99) (0.85–2.02) (0.42–1.52) (0.52–1.71) (0.82–6.33) (1.00–3.25) (0.43–2.53)  Financial crisis 1.33 1.22 1.30 1.32 1.33 1.17 2.03 1.73 1.09 2.35 (0.89–1.99) (0.79–1.90) (0.87–1.92) (0.59–2.96) (0.77–2.30) (0.65–2.11) (0.91–4.56) (0.57–5.27) (0.51–2.33) (0.57–9.69)  Midwestc 1.24 1.17 1.18 0.88 1.73 2.06 0.99 1.95 0.71 1.26 (0.73–2.10) (0.75–1.81) (0.72–1.91) (0.37–2.11) (0.92–3.25) (0.62–6.89) (0.34–2.86) (0.33–11.53) (0.33–1.51) (0.34–4.66)  Southc 2.06*** 1.88*** 1.79** 1.54 2.16** 2.34 1.31 1.68 1.00 1.95 (1.32–3.20) (1.29–2.73) (1.18–2.71) (0.72–3.32) (1.28–3.63) (0.76–7.16) (0.43–4.02) (0.33–8.65) (0.52–1.91) (0.56–6.77)  Westc 2.36*** 2.05** 2.33** 2.10 2.89*** 4.27* 2.00 1.22 1.28 1.92 (1.41–3.95) (1.24–3.39) (1.35–4.02) (0.62–7.12) (1.68–4.97) (1.09–16.72) (0.63–6.31) (0.28–5.24) (0.61–2.66) (0.49–7.61)  County pop. size 1.05 1.04 1.05 1.08 1.01 1.10 1.03 1.13 1.08 1.00 (0.96–1.15) (0.96–1.13) (0.96–1.15) (0.90–1.30) (0.90–1.14) (0.98–1.23) (0.89–1.18) (0.92–1.41) (0.98–1.19) (0.80–1.25)  Education 1.06* 1.11*** 1.08* 1.12 1.05 1.23** 1.01 1.27* 1.15** 1.13 (1.01–1.12) (1.04–1.18) (1.01–1.15) (0.96–1.31) (0.97–1.14) (1.06–1.43) (0.89–1.16) (1.02–1.59) (1.04–1.28) (0.94–1.37)  Married 0.94 1.02 0.96 1.05 0.84 1.41 0.96 1.31 1.13 0.54 (0.69–1.26) (0.76–1.38) (0.72–1.28) (0.63–1.73) (0.60–1.19) (0.83–2.39) (0.42–2.18) (0.59–2.91) (0.68–1.87) (0.23–1.26)  Worker 1.21 1.27 1.23 1.48 1.17 1.04 0.58 0.86 1.55 0.91 (0.83–1.75) (0.94–1.72) (0.90–1.69) (0.68–3.21) (0.69–1.98) (0.45–2.40) (0.26–1.28) (0.36–2.07) (0.84–2.86) (0.40–2.07) Relational-level predictors  Religious homophily 1.55*** 1.57** 1.58** 1.83*** 1.39 3.10*** 1.28 6.49*** 1.23 1.24 (1.24–1.93) (1.26–1.97) (1.27–1.96) (1.32–2.55) (0.99–1.94) (2.07–4.64) (0.71–2.32) (2.22–18.94) (0.75–2.01) (0.75–2.05)  Alter rank 1 1.88*** 1.87*** 1.88** 2.06*** 1.74** 1.61 1.33 1.94 3.81*** 1.46 (1.51–2.36) (1.50–2.34) (1.50–2.35) (1.46–2.90) (1.29–2.34) (0.86–3.00) (0.80–2.23) (0.97–3.88) (2.64–5.49) (0.76–2.82)  Alter rank 2 1.16 1.15 1.16 1.17 1.15 0.88 0.95 1.93 2.14*** 0.86 (0.92–1.45) (0.92–1.44) (0.92–1.45) (0.86–1.60) (0.84–1.59) (0.57–1.35) (0.51–1.77) (0.93–4.00) (1.47–3.12) (0.45–1.62)  Alter rank 3 0.99 0.99 0.99 1.07 0.92 0.81 1.15 1.51 1.19 0.78 (0.81–1.22) (0.80–1.21) (0.81–1.22) (0.78–1.49) (0.70–1.20) (0.47–1.39) (0.67–1.99) (0.86–2.67) (0.78–1.82) (0.46–1.30)  Alter gave advice 2.79*** 2.76*** 2.87** 3.09** 2.78*** 3.07*** 3.28*** 4.51*** 2.91*** 2.43** (2.21–3.52) (2.19–3.47) (2.26–3.65) (2.00–4.76) (2.04–3.79) (1.70–5.54) (2.00–5.37) (1.89–10.77) (1.59–5.30) (1.39–4.25)  Alter gave help 1.86*** 1.91*** 1.87** 1.90** 1.76*** 2.37*** 2.27* 1.80 1.36 2.10 (1.47–2.36) (1.48–2.47) (1.48–2.36) (1.29–2.80) (1.25–2.47) (1.51–3.73) (1.15–4.49) (0.91–3.56) (0.78–2.36) (0.89–4.94)  Lives nearby 1.40*** 1.43*** 1.41** 1.45* 1.35 1.44 1.10 1.26 1.39 1.20 (1.15–1.70) (1.18–1.74) (1.16–1.72) (1.07–1.96) (0.99–1.84) (0.90–2.28) (0.64–1.87) (0.62–2.57) (0.97–1.98) (0.65–2.23)  Interact weekly 1.73*** 1.74*** 1.71** 1.99** 1.56** 2.31** 2.01* 1.53 1.08 2.67** (1.32–2.26) (1.32–2.29) (1.31–2.24) (1.29–3.10) (1.14–2.12) (1.28–4.16) (1.00–4.05) (0.74–3.16) (0.65–1.78) (1.31–5.43)  Racial homophily 0.78 0.81 0.78 1.00 0.74 1.50 0.90 0.95 0.72 0.88 (0.50–1.20) (0.52–1.25) (0.50–1.20) (0.52–1.96) (0.45–1.23) (0.85–2.65) (0.34–2.38) (0.19–4.77) (0.22–2.33) (0.40–1.96)  Gender homophily 0.90 0.92 0.92 1.00 0.86 0.77 0.94 1.05 0.81 1.18 (0.72–1.14) (0.73–1.15) (0.73–1.16) (0.69–1.43) (0.66–1.12) (0.52–1.15) (0.58–1.54) (0.50–2.21) (0.47–1.38) (0.76–1.83)  Alter has bachelor dg. 1.32* 1.32* 1.32* 0.94 1.78*** 0.94 1.72 2.36* 1.03 1.70 (1.06–1.66) (1.05–1.65) (1.05–1.67) (0.67–1.31) (1.27–2.49) (0.59–1.52) (0.98–3.02) (1.07–5.24) (0.65–1.61) (0.85–3.40)  Kin 0.84 0.83 0.85 0.98 0.80 0.79 1.32 0.97 1.16 0.46* (0.69–1.01) (0.68–1.01) (0.71–1.02) (0.71–1.35) (0.60–1.05) (0.48–1.30) (0.75–2.32) (0.57–1.66) (0.79–1.70) (0.21–0.99)  Co-worker 0.90 0.89 0.89 1.20 0.79 0.66 1.45 2.45 0.96 0.80 (0.65–1.24) (0.64–1.26) (0.64–1.24) (0.70–2.07) (0.54–1.14) (0.30–1.45) (0.64–3.26) (0.20–29.47) (0.50–1.84) (0.33–1.93)  Level 1 5.23 5.45 5.17 6.48 4.24 7.26 3.18 5.37 5.73 4.10  Variance (0.50) (0.53) (0.51) (0.95) (0.56) (2.25) (0.71) (2.84) (1.27) (0.95)  −2LL 4841.96 4886.24 4821.47 2195.57 2591.21 1227.96 741.20 489.75 1189.65 788.78 Observations  Level 1 8,944 8,945 8,944 4,139 4,805 1,985 1,026 926 2,469 1,418  Level 2 2,419 2,420 2,419 1,102 1,317 529 270 256 683 384 Notes: Robust standard errors are shown in parentheses. aReligions none is reference group. bWhite is reference group; Race/ethnicity not included in model 8 because the overwhelming majority of Black Protestants are Black. cNortheast is reference group. *p < .05; **p < .01; ***p < .001 (two-tailed tests). View Large Across the various models, being part of a congregation is a strong predictor of talking religion with a given network tie. This is true without accounting for affiliation (model 1), after adjusting for affiliation (model 3), and among Evangelicals, Black Protestants, Mainline Protestants, and Catholics. Evangelical affiliation also seems to dictate whether people talk religion (model 2), but this can be explained by the fact that Evangelicals have high levels of congregational involvement relative to other groups, particularly religious nones (Evangelical odds ratio is no longer significant in model 3). Additional analysis reveals that over 60% of Evangelicals are connected to a congregation; this figure is higher than all other religious groups. Notably, once accounting for congregational involvement, Catholics and Mainline Protestants are actually less likely to talk religion with a given network tie than are religious nones (model 3). This suppression pattern reflects the fact that both groups are relatively unlikely to report religious discussions, yet both are far more likely than the religious nones to be part of a congregation (fewer than 10% of nones have a congregational involvement)—the factor which itself is the strongest predictor of talking religion. Indeed, Catholics and Mainline Protestants who do belong to a congregation are indeed far more likely to talk religion with a given tie than are their noncongregational counterparts (models 7 and 9). Supplementary analyses considered one other dimension of individual religiosity, self-assessed religious salience (from “not at all important” to “by far the most important part of your life”). This indicator of religiosity tended to perform much like congregational involvement (strongly associated with religious discussion in the main sample and across all affiliation groups), but it was not included in final models because it overlapped to a great degree with the other two aspects of personal religiosity. Among the other individual-level factors, being female is associated with religious discussion (models 1–3), but this pattern is driven mainly by noncongregants, by Catholics, and by Mainline Protestants (see models 5, 7, and 9). The association between gender and talking religion was consistent whether or not individual religiosity was included in the model; this suggests that women’s proclivity to discuss religion is not merely a function of their being more religious than men. In contrast to my expectations, there is no evidence that Black people—or those identifying as Black Protestants—talk religion any more or less than White adults. This was true in models (not shown) that excluded individual religiosity. However, Latino/as were less likely than Whites to report talking religion, particularly among Evangelical and Catholic subgroups. Expectations about age/cohort and stressful life events were not substantiated. None of these factors emerged as significant predictors of religious discussion in the main sample or in any of the subgroups.12 The sole exception was the case of a loved one dying; this experience increased the odds of talking religion only among Catholics. Finally, there is some evidence that Southerners and those living in the West are more likely than Northeasterners to talk religion, and that having higher education is associated with greater odds of discussing religion with a network tie. The latter pattern is most pronounced among Evangelicals, Black Protestants, and Catholics. Each relational theory receives some element of support in table 3. Religious homophily is a strong predictor of religious discussion in the full sample (OR = 1.58; 95% CI = 1.27–1.96). However, subgroup analysis reveals that the pattern is driven largely by people involved in a congregation (model 4) and by Evangelicals and Black Protestants (models 6 and 8). Indicators of tie strength also emerged as strong predictors. Alters named first in the roster had odds of religious discussion about 88% higher than did alters nominated last (models 1–3). Being top of alter list appears to be particularly important among Catholics, increasing the odds of religious discussion by a factor of 3.81 (model 9). Being top-ranked was less consistently influential among the other religious affiliation subgroups. Also of note, the main distinction by roster ordering—where observed—seemed to be between those named first and everyone else; there was little evidence that ordering predicted religious discussion beyond the top-ranked alter for any of the models. Religious discussion also coincided with social support provision, especially advice-giving. Indeed, the link between advice given and religious discussion was remarkably consistent across all respondent subgroups. Alters who provided practical help also tended to be people with whom religion was discussed, though this pattern was somewhat less robust across all subgroups than was advice exchange. Frequent interaction—one aspect of tie strength, but also a reflection of conversational opportunity—boosted the odds of talking religion by about 70% in the main sample. Living within a 20-minute drive of alter also increased odds of religious discussion (about 40% in the main sample), but this estimate was not consistently different from nil across the various subgroup models.13 Though not the focus of the analysis, there was some evidence that people were more likely to talk about religion with highly educated network members, though this pattern was most apparent among those not affiliated with a Congregation and among Black Protestants.14 Tie Persistence To conclude the analysis, I consider one potential network outcome of religious discussion, tie persistence. Descriptive results from table 2 reveal that just over 33% of all initially identified alters persist when PALS respondents were re-interviewed in 2011. Results from the full sample in table 4 indicate that talking religion with an alter in W1 increases the odds by 43% that the alter will re-appear in ego’s W2 close network. This association is net of religious homophily, race and gender homophily, initial tie closeness, and the full set of other relational and individual-level covariates. Religious homophily, in fact, was not significantly associated with tie persistence, even in supplementary models where religious discussion was removed. Further, religious homophily did not significantly interact with religious discussion (results not shown for sake of space). These findings cast doubt on the possibility that mere religious homophily is the reason why talking religion is associated with tie durability. Table 4 Tie Persistence, Odds Ratio Estimates from Random Intercept Logistic Regression Models (Portraits of American Life Survey, W1/W2) Full sample Part of Congregation Not Part of Congregation Evangelicals Mainline Protestants Black Protestants Catholics Religious nones Model 1 Model 2 Model 3 Model 4 Model 6 Model 5 Model 7 Model 8 Discuss religion with alter 1.44* 1.80** 1.24 2.45** 3.11** 1.33 1.24 0.91 (1.05–1.98) (1.22–2.67) (0.83–1.85) (1.39–4.30) (1.43–6.74) (0.40–4.38) (0.79–1.97) (0.47–1.78) Individual level covariates  Part of cong. 0.62** 0.76 0.80 0.62 0.94 0.11*** (0.46–0.84) (0.46–1.25) (0.33–1.94) (0.21–1.85) (0.55–1.60) (0.03–0.41)  Evangelicala 0.97 4.50** 0.71 (0.62–1.51) (1.50–13.50) (0.43–1.19)  Black Protestanta 0.71 3.41 0.60 (0.44–1.15) (0.82–14.15) (0.23–1.58)  Mainline Protestanta 1.16 5.92*** 0.82 (0.65–2.06) (2.06–17.03) (0.31–2.15)  Catholica 0.97 5.43** 0.63* (0.63–1.49) (1.83–16.11) (0.40–0.99)  Other religiona 0.53** 1.58 0.68 (0.33–0.84) (0.48–5.23) (0.43–1.04)  Female 1.31* 1.56* 1.04 1.60* 0.95 1.43 1.18 1.03 (1.01–1.70) (1.06–2.29) (0.79–1.37) (1.05–2.45) (0.50–1.81) (0.48–4.31) (0.68–2.05) (0.64–1.66)  Blackb 0.96 0.88 0.93 1.12 0.38 0.49 0.81 (0.67–1.39) (0.42–1.87) (0.53–1.63) (0.54–2.34) (0.05–3.10) (0.16–1.53) (0.38–1.74)  Latino/ab 1.03 1.12 1.00 1.95 0.17* 0.82 0.67 (0.69–1.54) (0.59–2.15) (0.51–1.97) (0.76–4.97) (0.04–0.73) (0.41–1.66) (0.22–2.04)  Other raceb 0.67 0.38 0.96 1.95 0.22* 0.34 0.69 (0.33–1.33) (0.11–1.34) (0.42–2.22) (0.22–17.52) (0.05–0.86) (0.09–1.19) (0.14–3.33)  Age 1.01 0.99 1.02** 1.00 0.96* 1.00 1.01 1.03* (1.00–1.02) (0.98–1.01) (1.01–1.04) (0.99–1.02) (0.93–0.99) (0.97–1.04) (1.00–1.03) (1.00–1.06)  Serious illness/injury 0.78 0.89 0.62 1.27 0.35 0.61 0.79 0.44** (0.55–1.12) (0.54–1.45) (0.39–1.00) (0.74–2.19) (0.11–1.07) (0.16–2.28) (0.50–1.26) (0.25–0.77)  Death of loved one 0.80 1.17 0.62* 0.61 0.50* 0.51 0.88 0.74 (0.63–1.02) (0.82–1.66) (0.42–0.90) (0.37–1.00) (0.25–0.99) (0.20–1.34) (0.55–1.39) (0.41–1.35)  Financial crisis 1.26 1.49 1.08 1.20 0.56 1.16 1.43 1.02 (0.96–1.65) (0.98–2.27) (0.71–1.63) (0.74–1.92) (0.24–1.32) (0.50–2.70) (0.71–2.85) (0.52–2.01)  Midwestc 0.61** 0.61* 0.64 0.47* 1.09 0.55 0.59 0.83 (0.42–0.89) (0.38–0.96) (0.36–1.14) (0.23–0.98) (0.31–3.79) (0.20–1.53) (0.29–1.19) (0.31–2.23)  Southc 0.65** 0.52** 0.84 0.56 1.01 1.07 0.72 0.85 (0.47–0.89) (0.33–0.80) (0.59–1.19) (0.28–1.11) (0.29–3.52) (0.34–3.33) (0.36–1.47) (0.41–1.80)  Westc 0.84 0.81 0.85 0.76 1.20 0.57 0.79 0.93 (0.58–1.20) (0.41–1.60) (0.58–1.14) (0.26–2.23) (0.35–4.12) (0.19–1.74) (0.36–1.75) (0.39–2.20)  County pop. size 0.99 1.04 0.98 0.97 1.04 1.08 0.97 1.05 (0.93–1.07) (0.96–1.14) (0.89–1.07) (0.85–1.10) (0.89–1.22) (0.89–1.33) (0.87–1.08) (0.92–1.18)  Education 1.08** 1.11*** 1.00 0.97 0.99 1.06 1.09 1.03 (1.02–1.13) (1.04–1.18) (0.93–1.07) (0.86–1.09) (0.88–1.12) (0.94–1.19) (0.98–1.21) (0.92–1.15)  Married 0.95 1.22 0.76 1.10 0.49 0.56 0.90 1.10 (0.70–1.28) (0.77–1.94) (0.54–1.09) (0.64–1.90) (0.18–1.32) (0.30–1.07) (0.55–1.45) (0.64–1.88)  Worker 0.91 0.77 0.96 0.57* 0.80 1.18 1.28 1.35 (0.71–1.16) (0.53–1.12) (0.68–1.37) (0.33–1.00) (0.45–1.42) (0.39–3.63) (0.80–2.04) (0.81–2.26)  Moved (W1–W2) 0.88 0.86 0.84 0.86 0.67 0.61 0.77 0.85 (0.63–1.21) (0.59–1.25) (0.55–1.27) (0.54–1.35) (0.23–1.94) (0.22–1.68) (0.46–1.26) (0.44–1.64)  Change in marital status (W1–W2) 0.83 0.90 0.76 0.34** 0.73 1.13 0.99 0.84 (0.62–1.10) (0.60–1.36) (0.53–1.07) (0.17–0.66) (0.30–1.78) (0.41–3.14) (0.64–1.54) (0.41–1.70)  Change in work status (W1–W2) 0.84 0.63* 1.05 0.68 0.22*** 1.65 1.08 1.73* (0.58–1.20) (0.41–0.96) (0.68–1.61) (0.38–1.24) (0.10–0.50) (0.56–4.88) (0.47–2.46) (1.05–2.85) Relational level covariates  Religious homophily 1.08 1.16 1.01 1.01 1.33 0.66 1.40 1.27 (0.83–1.39) (0.84–1.61) (0.69–1.48) (0.56–1.81) (0.66–2.70) (0.33–1.31) (0.92–2.15) (0.83–1.94)  Alter rank 1 3.68*** 3.25*** 3.94*** 4.34*** 1.89 2.71 4.14*** 4.80*** (2.61–5.18) (2.19–4.83) (2.34–6.63) (2.32–8.11) (0.66–5.39) (0.93–7.85) (2.34–7.31) (2.37–9.71)  Alter rank 2 2.05*** 1.84*** 2.20*** 1.91 1.65 2.37* 2.40** 1.70 (1.56–2.69) (1.29–2.63) (1.42–3.41) (0.97–3.74) (0.70–3.89) (1.01–5.56) (1.35–4.24) (0.96–3.00)  Alter rank 3 1.36* 1.29 1.42 1.18 0.81 1.52 1.93** 1.50 (1.04–1.78) (0.90–1.83) (0.91–2.21) (0.64–2.17) (0.35–1.91) (0.44–5.19) (1.17–3.18) (0.73–3.07)  Alter gave advice 1.62** 1.33 1.97*** 1.44 0.67 1.18 1.09 2.73*** (1.20–2.18) (0.85–2.08) (1.39–2.79) (0.81–2.56) (0.26–1.72) (0.67–2.05) (0.60–1.99) (1.47–5.07)  Alter gave help 1.27 1.47 1.19 0.99 0.66 0.86 2.12** 1.21 (0.92–1.74) (0.92–2.36) (0.74–1.92) (0.58–1.69) (0.28–1.57) (0.41–1.78) (1.23–3.67) (0.56–2.63)  Lives nearby 1.09 1.07 1.14 0.89 1.01 1.93* 1.19 1.14 (0.83–1.43) (0.73–1.57) (0.82–1.59) (0.57–1.38) (0.49–2.10) (1.16–3.23) (0.76–1.86) (0.67–1.94)  Interact weekly 1.32 1.24 1.39 1.46 3.64** 1.58 1.51 0.67 (0.96–1.81) (0.86–1.77) (0.91–2.15) (0.81–2.62) (1.45–9.13) (0.67–3.72) (0.84–2.71) (0.34–1.31)  Racial homophily 1.95** 2.20* 1.58 0.88 3.56 8.40 1.05 2.28* (1.24–3.06) (1.10–4.40) (0.90–2.77) (0.27–2.86) (0.44–28.68) (0.36–197.71) (0.48–2.28) (1.07–4.88)  Gender homophily 1.26* 1.48* 1.14 4.33 2.31* 0.77 0.94 1.21 (1.01–1.56) (1.03–2.11) (0.88–1.47) (1.61– 11.67)** (1.08–4.92) (0.30–1.97) (0.61–1.45) (0.78–1.87)  Alter has bachelor dg. 0.93 0.90 1.04 1.14 0.90 1.77 0.63* 1.10 (0.76–1.13) (0.68–1.20) (0.79–1.37) (0.67–1.94) (0.45–1.80) (0.73–4.31) (0.39–1.00) (0.61–1.99)  Kin 2.45*** 1.91** 2.88*** 0.83 3.58*** 0.90 1.84*** 2.32* (1.90–3.17) (1.27–2.87) (1.99–4.15) (0.47–1.48) (1.68–7.64) (0.40–2.02) (1.27–2.68) (1.22–4.44)  Co-worker 0.80 0.62 0.97 2.45 0.14** 0.96 0.93 1.70 (0.48–1.35) (0.33–1.18) (0.52–1.80) (1.52– 3.95)** (0.04–0.57) (0.27–3.34) (0.42–2.06) (0.68–4.24)  Level 1 variance 1.25 1.24 1.02 0.62 1.21 1.15 0.97 0.56 (0.17) (0.24) (0.21) (0.31) (0.42) (0.67) (0.31) (0.34)  −2LL 859,928,640 390,913,472 452,567,008 222,713,520 88,179,943 19,448,486 230,494,992 114,654,792 Observations  Level 1 4619 2224 2395 1031 518 385 1223 752  Level 2 1233 583 650 273 134 106 331 202 Full sample Part of Congregation Not Part of Congregation Evangelicals Mainline Protestants Black Protestants Catholics Religious nones Model 1 Model 2 Model 3 Model 4 Model 6 Model 5 Model 7 Model 8 Discuss religion with alter 1.44* 1.80** 1.24 2.45** 3.11** 1.33 1.24 0.91 (1.05–1.98) (1.22–2.67) (0.83–1.85) (1.39–4.30) (1.43–6.74) (0.40–4.38) (0.79–1.97) (0.47–1.78) Individual level covariates  Part of cong. 0.62** 0.76 0.80 0.62 0.94 0.11*** (0.46–0.84) (0.46–1.25) (0.33–1.94) (0.21–1.85) (0.55–1.60) (0.03–0.41)  Evangelicala 0.97 4.50** 0.71 (0.62–1.51) (1.50–13.50) (0.43–1.19)  Black Protestanta 0.71 3.41 0.60 (0.44–1.15) (0.82–14.15) (0.23–1.58)  Mainline Protestanta 1.16 5.92*** 0.82 (0.65–2.06) (2.06–17.03) (0.31–2.15)  Catholica 0.97 5.43** 0.63* (0.63–1.49) (1.83–16.11) (0.40–0.99)  Other religiona 0.53** 1.58 0.68 (0.33–0.84) (0.48–5.23) (0.43–1.04)  Female 1.31* 1.56* 1.04 1.60* 0.95 1.43 1.18 1.03 (1.01–1.70) (1.06–2.29) (0.79–1.37) (1.05–2.45) (0.50–1.81) (0.48–4.31) (0.68–2.05) (0.64–1.66)  Blackb 0.96 0.88 0.93 1.12 0.38 0.49 0.81 (0.67–1.39) (0.42–1.87) (0.53–1.63) (0.54–2.34) (0.05–3.10) (0.16–1.53) (0.38–1.74)  Latino/ab 1.03 1.12 1.00 1.95 0.17* 0.82 0.67 (0.69–1.54) (0.59–2.15) (0.51–1.97) (0.76–4.97) (0.04–0.73) (0.41–1.66) (0.22–2.04)  Other raceb 0.67 0.38 0.96 1.95 0.22* 0.34 0.69 (0.33–1.33) (0.11–1.34) (0.42–2.22) (0.22–17.52) (0.05–0.86) (0.09–1.19) (0.14–3.33)  Age 1.01 0.99 1.02** 1.00 0.96* 1.00 1.01 1.03* (1.00–1.02) (0.98–1.01) (1.01–1.04) (0.99–1.02) (0.93–0.99) (0.97–1.04) (1.00–1.03) (1.00–1.06)  Serious illness/injury 0.78 0.89 0.62 1.27 0.35 0.61 0.79 0.44** (0.55–1.12) (0.54–1.45) (0.39–1.00) (0.74–2.19) (0.11–1.07) (0.16–2.28) (0.50–1.26) (0.25–0.77)  Death of loved one 0.80 1.17 0.62* 0.61 0.50* 0.51 0.88 0.74 (0.63–1.02) (0.82–1.66) (0.42–0.90) (0.37–1.00) (0.25–0.99) (0.20–1.34) (0.55–1.39) (0.41–1.35)  Financial crisis 1.26 1.49 1.08 1.20 0.56 1.16 1.43 1.02 (0.96–1.65) (0.98–2.27) (0.71–1.63) (0.74–1.92) (0.24–1.32) (0.50–2.70) (0.71–2.85) (0.52–2.01)  Midwestc 0.61** 0.61* 0.64 0.47* 1.09 0.55 0.59 0.83 (0.42–0.89) (0.38–0.96) (0.36–1.14) (0.23–0.98) (0.31–3.79) (0.20–1.53) (0.29–1.19) (0.31–2.23)  Southc 0.65** 0.52** 0.84 0.56 1.01 1.07 0.72 0.85 (0.47–0.89) (0.33–0.80) (0.59–1.19) (0.28–1.11) (0.29–3.52) (0.34–3.33) (0.36–1.47) (0.41–1.80)  Westc 0.84 0.81 0.85 0.76 1.20 0.57 0.79 0.93 (0.58–1.20) (0.41–1.60) (0.58–1.14) (0.26–2.23) (0.35–4.12) (0.19–1.74) (0.36–1.75) (0.39–2.20)  County pop. size 0.99 1.04 0.98 0.97 1.04 1.08 0.97 1.05 (0.93–1.07) (0.96–1.14) (0.89–1.07) (0.85–1.10) (0.89–1.22) (0.89–1.33) (0.87–1.08) (0.92–1.18)  Education 1.08** 1.11*** 1.00 0.97 0.99 1.06 1.09 1.03 (1.02–1.13) (1.04–1.18) (0.93–1.07) (0.86–1.09) (0.88–1.12) (0.94–1.19) (0.98–1.21) (0.92–1.15)  Married 0.95 1.22 0.76 1.10 0.49 0.56 0.90 1.10 (0.70–1.28) (0.77–1.94) (0.54–1.09) (0.64–1.90) (0.18–1.32) (0.30–1.07) (0.55–1.45) (0.64–1.88)  Worker 0.91 0.77 0.96 0.57* 0.80 1.18 1.28 1.35 (0.71–1.16) (0.53–1.12) (0.68–1.37) (0.33–1.00) (0.45–1.42) (0.39–3.63) (0.80–2.04) (0.81–2.26)  Moved (W1–W2) 0.88 0.86 0.84 0.86 0.67 0.61 0.77 0.85 (0.63–1.21) (0.59–1.25) (0.55–1.27) (0.54–1.35) (0.23–1.94) (0.22–1.68) (0.46–1.26) (0.44–1.64)  Change in marital status (W1–W2) 0.83 0.90 0.76 0.34** 0.73 1.13 0.99 0.84 (0.62–1.10) (0.60–1.36) (0.53–1.07) (0.17–0.66) (0.30–1.78) (0.41–3.14) (0.64–1.54) (0.41–1.70)  Change in work status (W1–W2) 0.84 0.63* 1.05 0.68 0.22*** 1.65 1.08 1.73* (0.58–1.20) (0.41–0.96) (0.68–1.61) (0.38–1.24) (0.10–0.50) (0.56–4.88) (0.47–2.46) (1.05–2.85) Relational level covariates  Religious homophily 1.08 1.16 1.01 1.01 1.33 0.66 1.40 1.27 (0.83–1.39) (0.84–1.61) (0.69–1.48) (0.56–1.81) (0.66–2.70) (0.33–1.31) (0.92–2.15) (0.83–1.94)  Alter rank 1 3.68*** 3.25*** 3.94*** 4.34*** 1.89 2.71 4.14*** 4.80*** (2.61–5.18) (2.19–4.83) (2.34–6.63) (2.32–8.11) (0.66–5.39) (0.93–7.85) (2.34–7.31) (2.37–9.71)  Alter rank 2 2.05*** 1.84*** 2.20*** 1.91 1.65 2.37* 2.40** 1.70 (1.56–2.69) (1.29–2.63) (1.42–3.41) (0.97–3.74) (0.70–3.89) (1.01–5.56) (1.35–4.24) (0.96–3.00)  Alter rank 3 1.36* 1.29 1.42 1.18 0.81 1.52 1.93** 1.50 (1.04–1.78) (0.90–1.83) (0.91–2.21) (0.64–2.17) (0.35–1.91) (0.44–5.19) (1.17–3.18) (0.73–3.07)  Alter gave advice 1.62** 1.33 1.97*** 1.44 0.67 1.18 1.09 2.73*** (1.20–2.18) (0.85–2.08) (1.39–2.79) (0.81–2.56) (0.26–1.72) (0.67–2.05) (0.60–1.99) (1.47–5.07)  Alter gave help 1.27 1.47 1.19 0.99 0.66 0.86 2.12** 1.21 (0.92–1.74) (0.92–2.36) (0.74–1.92) (0.58–1.69) (0.28–1.57) (0.41–1.78) (1.23–3.67) (0.56–2.63)  Lives nearby 1.09 1.07 1.14 0.89 1.01 1.93* 1.19 1.14 (0.83–1.43) (0.73–1.57) (0.82–1.59) (0.57–1.38) (0.49–2.10) (1.16–3.23) (0.76–1.86) (0.67–1.94)  Interact weekly 1.32 1.24 1.39 1.46 3.64** 1.58 1.51 0.67 (0.96–1.81) (0.86–1.77) (0.91–2.15) (0.81–2.62) (1.45–9.13) (0.67–3.72) (0.84–2.71) (0.34–1.31)  Racial homophily 1.95** 2.20* 1.58 0.88 3.56 8.40 1.05 2.28* (1.24–3.06) (1.10–4.40) (0.90–2.77) (0.27–2.86) (0.44–28.68) (0.36–197.71) (0.48–2.28) (1.07–4.88)  Gender homophily 1.26* 1.48* 1.14 4.33 2.31* 0.77 0.94 1.21 (1.01–1.56) (1.03–2.11) (0.88–1.47) (1.61– 11.67)** (1.08–4.92) (0.30–1.97) (0.61–1.45) (0.78–1.87)  Alter has bachelor dg. 0.93 0.90 1.04 1.14 0.90 1.77 0.63* 1.10 (0.76–1.13) (0.68–1.20) (0.79–1.37) (0.67–1.94) (0.45–1.80) (0.73–4.31) (0.39–1.00) (0.61–1.99)  Kin 2.45*** 1.91** 2.88*** 0.83 3.58*** 0.90 1.84*** 2.32* (1.90–3.17) (1.27–2.87) (1.99–4.15) (0.47–1.48) (1.68–7.64) (0.40–2.02) (1.27–2.68) (1.22–4.44)  Co-worker 0.80 0.62 0.97 2.45 0.14** 0.96 0.93 1.70 (0.48–1.35) (0.33–1.18) (0.52–1.80) (1.52– 3.95)** (0.04–0.57) (0.27–3.34) (0.42–2.06) (0.68–4.24)  Level 1 variance 1.25 1.24 1.02 0.62 1.21 1.15 0.97 0.56 (0.17) (0.24) (0.21) (0.31) (0.42) (0.67) (0.31) (0.34)  −2LL 859,928,640 390,913,472 452,567,008 222,713,520 88,179,943 19,448,486 230,494,992 114,654,792 Observations  Level 1 4619 2224 2395 1031 518 385 1223 752  Level 2 1233 583 650 273 134 106 331 202 Notes: Robust standard errors are shown in parentheses. aReligions none is reference group. bWhite is reference group; Race/ethnicity not included in model 5 because the overwhelming majority of Black Protestants are Black. cNortheast is reference group. ***p < .001; **p < .01; *p < .05 (two-tailed tests). View Large Table 4 Tie Persistence, Odds Ratio Estimates from Random Intercept Logistic Regression Models (Portraits of American Life Survey, W1/W2) Full sample Part of Congregation Not Part of Congregation Evangelicals Mainline Protestants Black Protestants Catholics Religious nones Model 1 Model 2 Model 3 Model 4 Model 6 Model 5 Model 7 Model 8 Discuss religion with alter 1.44* 1.80** 1.24 2.45** 3.11** 1.33 1.24 0.91 (1.05–1.98) (1.22–2.67) (0.83–1.85) (1.39–4.30) (1.43–6.74) (0.40–4.38) (0.79–1.97) (0.47–1.78) Individual level covariates  Part of cong. 0.62** 0.76 0.80 0.62 0.94 0.11*** (0.46–0.84) (0.46–1.25) (0.33–1.94) (0.21–1.85) (0.55–1.60) (0.03–0.41)  Evangelicala 0.97 4.50** 0.71 (0.62–1.51) (1.50–13.50) (0.43–1.19)  Black Protestanta 0.71 3.41 0.60 (0.44–1.15) (0.82–14.15) (0.23–1.58)  Mainline Protestanta 1.16 5.92*** 0.82 (0.65–2.06) (2.06–17.03) (0.31–2.15)  Catholica 0.97 5.43** 0.63* (0.63–1.49) (1.83–16.11) (0.40–0.99)  Other religiona 0.53** 1.58 0.68 (0.33–0.84) (0.48–5.23) (0.43–1.04)  Female 1.31* 1.56* 1.04 1.60* 0.95 1.43 1.18 1.03 (1.01–1.70) (1.06–2.29) (0.79–1.37) (1.05–2.45) (0.50–1.81) (0.48–4.31) (0.68–2.05) (0.64–1.66)  Blackb 0.96 0.88 0.93 1.12 0.38 0.49 0.81 (0.67–1.39) (0.42–1.87) (0.53–1.63) (0.54–2.34) (0.05–3.10) (0.16–1.53) (0.38–1.74)  Latino/ab 1.03 1.12 1.00 1.95 0.17* 0.82 0.67 (0.69–1.54) (0.59–2.15) (0.51–1.97) (0.76–4.97) (0.04–0.73) (0.41–1.66) (0.22–2.04)  Other raceb 0.67 0.38 0.96 1.95 0.22* 0.34 0.69 (0.33–1.33) (0.11–1.34) (0.42–2.22) (0.22–17.52) (0.05–0.86) (0.09–1.19) (0.14–3.33)  Age 1.01 0.99 1.02** 1.00 0.96* 1.00 1.01 1.03* (1.00–1.02) (0.98–1.01) (1.01–1.04) (0.99–1.02) (0.93–0.99) (0.97–1.04) (1.00–1.03) (1.00–1.06)  Serious illness/injury 0.78 0.89 0.62 1.27 0.35 0.61 0.79 0.44** (0.55–1.12) (0.54–1.45) (0.39–1.00) (0.74–2.19) (0.11–1.07) (0.16–2.28) (0.50–1.26) (0.25–0.77)  Death of loved one 0.80 1.17 0.62* 0.61 0.50* 0.51 0.88 0.74 (0.63–1.02) (0.82–1.66) (0.42–0.90) (0.37–1.00) (0.25–0.99) (0.20–1.34) (0.55–1.39) (0.41–1.35)  Financial crisis 1.26 1.49 1.08 1.20 0.56 1.16 1.43 1.02 (0.96–1.65) (0.98–2.27) (0.71–1.63) (0.74–1.92) (0.24–1.32) (0.50–2.70) (0.71–2.85) (0.52–2.01)  Midwestc 0.61** 0.61* 0.64 0.47* 1.09 0.55 0.59 0.83 (0.42–0.89) (0.38–0.96) (0.36–1.14) (0.23–0.98) (0.31–3.79) (0.20–1.53) (0.29–1.19) (0.31–2.23)  Southc 0.65** 0.52** 0.84 0.56 1.01 1.07 0.72 0.85 (0.47–0.89) (0.33–0.80) (0.59–1.19) (0.28–1.11) (0.29–3.52) (0.34–3.33) (0.36–1.47) (0.41–1.80)  Westc 0.84 0.81 0.85 0.76 1.20 0.57 0.79 0.93 (0.58–1.20) (0.41–1.60) (0.58–1.14) (0.26–2.23) (0.35–4.12) (0.19–1.74) (0.36–1.75) (0.39–2.20)  County pop. size 0.99 1.04 0.98 0.97 1.04 1.08 0.97 1.05 (0.93–1.07) (0.96–1.14) (0.89–1.07) (0.85–1.10) (0.89–1.22) (0.89–1.33) (0.87–1.08) (0.92–1.18)  Education 1.08** 1.11*** 1.00 0.97 0.99 1.06 1.09 1.03 (1.02–1.13) (1.04–1.18) (0.93–1.07) (0.86–1.09) (0.88–1.12) (0.94–1.19) (0.98–1.21) (0.92–1.15)  Married 0.95 1.22 0.76 1.10 0.49 0.56 0.90 1.10 (0.70–1.28) (0.77–1.94) (0.54–1.09) (0.64–1.90) (0.18–1.32) (0.30–1.07) (0.55–1.45) (0.64–1.88)  Worker 0.91 0.77 0.96 0.57* 0.80 1.18 1.28 1.35 (0.71–1.16) (0.53–1.12) (0.68–1.37) (0.33–1.00) (0.45–1.42) (0.39–3.63) (0.80–2.04) (0.81–2.26)  Moved (W1–W2) 0.88 0.86 0.84 0.86 0.67 0.61 0.77 0.85 (0.63–1.21) (0.59–1.25) (0.55–1.27) (0.54–1.35) (0.23–1.94) (0.22–1.68) (0.46–1.26) (0.44–1.64)  Change in marital status (W1–W2) 0.83 0.90 0.76 0.34** 0.73 1.13 0.99 0.84 (0.62–1.10) (0.60–1.36) (0.53–1.07) (0.17–0.66) (0.30–1.78) (0.41–3.14) (0.64–1.54) (0.41–1.70)  Change in work status (W1–W2) 0.84 0.63* 1.05 0.68 0.22*** 1.65 1.08 1.73* (0.58–1.20) (0.41–0.96) (0.68–1.61) (0.38–1.24) (0.10–0.50) (0.56–4.88) (0.47–2.46) (1.05–2.85) Relational level covariates  Religious homophily 1.08 1.16 1.01 1.01 1.33 0.66 1.40 1.27 (0.83–1.39) (0.84–1.61) (0.69–1.48) (0.56–1.81) (0.66–2.70) (0.33–1.31) (0.92–2.15) (0.83–1.94)  Alter rank 1 3.68*** 3.25*** 3.94*** 4.34*** 1.89 2.71 4.14*** 4.80*** (2.61–5.18) (2.19–4.83) (2.34–6.63) (2.32–8.11) (0.66–5.39) (0.93–7.85) (2.34–7.31) (2.37–9.71)  Alter rank 2 2.05*** 1.84*** 2.20*** 1.91 1.65 2.37* 2.40** 1.70 (1.56–2.69) (1.29–2.63) (1.42–3.41) (0.97–3.74) (0.70–3.89) (1.01–5.56) (1.35–4.24) (0.96–3.00)  Alter rank 3 1.36* 1.29 1.42 1.18 0.81 1.52 1.93** 1.50 (1.04–1.78) (0.90–1.83) (0.91–2.21) (0.64–2.17) (0.35–1.91) (0.44–5.19) (1.17–3.18) (0.73–3.07)  Alter gave advice 1.62** 1.33 1.97*** 1.44 0.67 1.18 1.09 2.73*** (1.20–2.18) (0.85–2.08) (1.39–2.79) (0.81–2.56) (0.26–1.72) (0.67–2.05) (0.60–1.99) (1.47–5.07)  Alter gave help 1.27 1.47 1.19 0.99 0.66 0.86 2.12** 1.21 (0.92–1.74) (0.92–2.36) (0.74–1.92) (0.58–1.69) (0.28–1.57) (0.41–1.78) (1.23–3.67) (0.56–2.63)  Lives nearby 1.09 1.07 1.14 0.89 1.01 1.93* 1.19 1.14 (0.83–1.43) (0.73–1.57) (0.82–1.59) (0.57–1.38) (0.49–2.10) (1.16–3.23) (0.76–1.86) (0.67–1.94)  Interact weekly 1.32 1.24 1.39 1.46 3.64** 1.58 1.51 0.67 (0.96–1.81) (0.86–1.77) (0.91–2.15) (0.81–2.62) (1.45–9.13) (0.67–3.72) (0.84–2.71) (0.34–1.31)  Racial homophily 1.95** 2.20* 1.58 0.88 3.56 8.40 1.05 2.28* (1.24–3.06) (1.10–4.40) (0.90–2.77) (0.27–2.86) (0.44–28.68) (0.36–197.71) (0.48–2.28) (1.07–4.88)  Gender homophily 1.26* 1.48* 1.14 4.33 2.31* 0.77 0.94 1.21 (1.01–1.56) (1.03–2.11) (0.88–1.47) (1.61– 11.67)** (1.08–4.92) (0.30–1.97) (0.61–1.45) (0.78–1.87)  Alter has bachelor dg. 0.93 0.90 1.04 1.14 0.90 1.77 0.63* 1.10 (0.76–1.13) (0.68–1.20) (0.79–1.37) (0.67–1.94) (0.45–1.80) (0.73–4.31) (0.39–1.00) (0.61–1.99)  Kin 2.45*** 1.91** 2.88*** 0.83 3.58*** 0.90 1.84*** 2.32* (1.90–3.17) (1.27–2.87) (1.99–4.15) (0.47–1.48) (1.68–7.64) (0.40–2.02) (1.27–2.68) (1.22–4.44)  Co-worker 0.80 0.62 0.97 2.45 0.14** 0.96 0.93 1.70 (0.48–1.35) (0.33–1.18) (0.52–1.80) (1.52– 3.95)** (0.04–0.57) (0.27–3.34) (0.42–2.06) (0.68–4.24)  Level 1 variance 1.25 1.24 1.02 0.62 1.21 1.15 0.97 0.56 (0.17) (0.24) (0.21) (0.31) (0.42) (0.67) (0.31) (0.34)  −2LL 859,928,640 390,913,472 452,567,008 222,713,520 88,179,943 19,448,486 230,494,992 114,654,792 Observations  Level 1 4619 2224 2395 1031 518 385 1223 752  Level 2 1233 583 650 273 134 106 331 202 Full sample Part of Congregation Not Part of Congregation Evangelicals Mainline Protestants Black Protestants Catholics Religious nones Model 1 Model 2 Model 3 Model 4 Model 6 Model 5 Model 7 Model 8 Discuss religion with alter 1.44* 1.80** 1.24 2.45** 3.11** 1.33 1.24 0.91 (1.05–1.98) (1.22–2.67) (0.83–1.85) (1.39–4.30) (1.43–6.74) (0.40–4.38) (0.79–1.97) (0.47–1.78) Individual level covariates  Part of cong. 0.62** 0.76 0.80 0.62 0.94 0.11*** (0.46–0.84) (0.46–1.25) (0.33–1.94) (0.21–1.85) (0.55–1.60) (0.03–0.41)  Evangelicala 0.97 4.50** 0.71 (0.62–1.51) (1.50–13.50) (0.43–1.19)  Black Protestanta 0.71 3.41 0.60 (0.44–1.15) (0.82–14.15) (0.23–1.58)  Mainline Protestanta 1.16 5.92*** 0.82 (0.65–2.06) (2.06–17.03) (0.31–2.15)  Catholica 0.97 5.43** 0.63* (0.63–1.49) (1.83–16.11) (0.40–0.99)  Other religiona 0.53** 1.58 0.68 (0.33–0.84) (0.48–5.23) (0.43–1.04)  Female 1.31* 1.56* 1.04 1.60* 0.95 1.43 1.18 1.03 (1.01–1.70) (1.06–2.29) (0.79–1.37) (1.05–2.45) (0.50–1.81) (0.48–4.31) (0.68–2.05) (0.64–1.66)  Blackb 0.96 0.88 0.93 1.12 0.38 0.49 0.81 (0.67–1.39) (0.42–1.87) (0.53–1.63) (0.54–2.34) (0.05–3.10) (0.16–1.53) (0.38–1.74)  Latino/ab 1.03 1.12 1.00 1.95 0.17* 0.82 0.67 (0.69–1.54) (0.59–2.15) (0.51–1.97) (0.76–4.97) (0.04–0.73) (0.41–1.66) (0.22–2.04)  Other raceb 0.67 0.38 0.96 1.95 0.22* 0.34 0.69 (0.33–1.33) (0.11–1.34) (0.42–2.22) (0.22–17.52) (0.05–0.86) (0.09–1.19) (0.14–3.33)  Age 1.01 0.99 1.02** 1.00 0.96* 1.00 1.01 1.03* (1.00–1.02) (0.98–1.01) (1.01–1.04) (0.99–1.02) (0.93–0.99) (0.97–1.04) (1.00–1.03) (1.00–1.06)  Serious illness/injury 0.78 0.89 0.62 1.27 0.35 0.61 0.79 0.44** (0.55–1.12) (0.54–1.45) (0.39–1.00) (0.74–2.19) (0.11–1.07) (0.16–2.28) (0.50–1.26) (0.25–0.77)  Death of loved one 0.80 1.17 0.62* 0.61 0.50* 0.51 0.88 0.74 (0.63–1.02) (0.82–1.66) (0.42–0.90) (0.37–1.00) (0.25–0.99) (0.20–1.34) (0.55–1.39) (0.41–1.35)  Financial crisis 1.26 1.49 1.08 1.20 0.56 1.16 1.43 1.02 (0.96–1.65) (0.98–2.27) (0.71–1.63) (0.74–1.92) (0.24–1.32) (0.50–2.70) (0.71–2.85) (0.52–2.01)  Midwestc 0.61** 0.61* 0.64 0.47* 1.09 0.55 0.59 0.83 (0.42–0.89) (0.38–0.96) (0.36–1.14) (0.23–0.98) (0.31–3.79) (0.20–1.53) (0.29–1.19) (0.31–2.23)  Southc 0.65** 0.52** 0.84 0.56 1.01 1.07 0.72 0.85 (0.47–0.89) (0.33–0.80) (0.59–1.19) (0.28–1.11) (0.29–3.52) (0.34–3.33) (0.36–1.47) (0.41–1.80)  Westc 0.84 0.81 0.85 0.76 1.20 0.57 0.79 0.93 (0.58–1.20) (0.41–1.60) (0.58–1.14) (0.26–2.23) (0.35–4.12) (0.19–1.74) (0.36–1.75) (0.39–2.20)  County pop. size 0.99 1.04 0.98 0.97 1.04 1.08 0.97 1.05 (0.93–1.07) (0.96–1.14) (0.89–1.07) (0.85–1.10) (0.89–1.22) (0.89–1.33) (0.87–1.08) (0.92–1.18)  Education 1.08** 1.11*** 1.00 0.97 0.99 1.06 1.09 1.03 (1.02–1.13) (1.04–1.18) (0.93–1.07) (0.86–1.09) (0.88–1.12) (0.94–1.19) (0.98–1.21) (0.92–1.15)  Married 0.95 1.22 0.76 1.10 0.49 0.56 0.90 1.10 (0.70–1.28) (0.77–1.94) (0.54–1.09) (0.64–1.90) (0.18–1.32) (0.30–1.07) (0.55–1.45) (0.64–1.88)  Worker 0.91 0.77 0.96 0.57* 0.80 1.18 1.28 1.35 (0.71–1.16) (0.53–1.12) (0.68–1.37) (0.33–1.00) (0.45–1.42) (0.39–3.63) (0.80–2.04) (0.81–2.26)  Moved (W1–W2) 0.88 0.86 0.84 0.86 0.67 0.61 0.77 0.85 (0.63–1.21) (0.59–1.25) (0.55–1.27) (0.54–1.35) (0.23–1.94) (0.22–1.68) (0.46–1.26) (0.44–1.64)  Change in marital status (W1–W2) 0.83 0.90 0.76 0.34** 0.73 1.13 0.99 0.84 (0.62–1.10) (0.60–1.36) (0.53–1.07) (0.17–0.66) (0.30–1.78) (0.41–3.14) (0.64–1.54) (0.41–1.70)  Change in work status (W1–W2) 0.84 0.63* 1.05 0.68 0.22*** 1.65 1.08 1.73* (0.58–1.20) (0.41–0.96) (0.68–1.61) (0.38–1.24) (0.10–0.50) (0.56–4.88) (0.47–2.46) (1.05–2.85) Relational level covariates  Religious homophily 1.08 1.16 1.01 1.01 1.33 0.66 1.40 1.27 (0.83–1.39) (0.84–1.61) (0.69–1.48) (0.56–1.81) (0.66–2.70) (0.33–1.31) (0.92–2.15) (0.83–1.94)  Alter rank 1 3.68*** 3.25*** 3.94*** 4.34*** 1.89 2.71 4.14*** 4.80*** (2.61–5.18) (2.19–4.83) (2.34–6.63) (2.32–8.11) (0.66–5.39) (0.93–7.85) (2.34–7.31) (2.37–9.71)  Alter rank 2 2.05*** 1.84*** 2.20*** 1.91 1.65 2.37* 2.40** 1.70 (1.56–2.69) (1.29–2.63) (1.42–3.41) (0.97–3.74) (0.70–3.89) (1.01–5.56) (1.35–4.24) (0.96–3.00)  Alter rank 3 1.36* 1.29 1.42 1.18 0.81 1.52 1.93** 1.50 (1.04–1.78) (0.90–1.83) (0.91–2.21) (0.64–2.17) (0.35–1.91) (0.44–5.19) (1.17–3.18) (0.73–3.07)  Alter gave advice 1.62** 1.33 1.97*** 1.44 0.67 1.18 1.09 2.73*** (1.20–2.18) (0.85–2.08) (1.39–2.79) (0.81–2.56) (0.26–1.72) (0.67–2.05) (0.60–1.99) (1.47–5.07)  Alter gave help 1.27 1.47 1.19 0.99 0.66 0.86 2.12** 1.21 (0.92–1.74) (0.92–2.36) (0.74–1.92) (0.58–1.69) (0.28–1.57) (0.41–1.78) (1.23–3.67) (0.56–2.63)  Lives nearby 1.09 1.07 1.14 0.89 1.01 1.93* 1.19 1.14 (0.83–1.43) (0.73–1.57) (0.82–1.59) (0.57–1.38) (0.49–2.10) (1.16–3.23) (0.76–1.86) (0.67–1.94)  Interact weekly 1.32 1.24 1.39 1.46 3.64** 1.58 1.51 0.67 (0.96–1.81) (0.86–1.77) (0.91–2.15) (0.81–2.62) (1.45–9.13) (0.67–3.72) (0.84–2.71) (0.34–1.31)  Racial homophily 1.95** 2.20* 1.58 0.88 3.56 8.40 1.05 2.28* (1.24–3.06) (1.10–4.40) (0.90–2.77) (0.27–2.86) (0.44–28.68) (0.36–197.71) (0.48–2.28) (1.07–4.88)  Gender homophily 1.26* 1.48* 1.14 4.33 2.31* 0.77 0.94 1.21 (1.01–1.56) (1.03–2.11) (0.88–1.47) (1.61– 11.67)** (1.08–4.92) (0.30–1.97) (0.61–1.45) (0.78–1.87)  Alter has bachelor dg. 0.93 0.90 1.04 1.14 0.90 1.77 0.63* 1.10 (0.76–1.13) (0.68–1.20) (0.79–1.37) (0.67–1.94) (0.45–1.80) (0.73–4.31) (0.39–1.00) (0.61–1.99)  Kin 2.45*** 1.91** 2.88*** 0.83 3.58*** 0.90 1.84*** 2.32* (1.90–3.17) (1.27–2.87) (1.99–4.15) (0.47–1.48) (1.68–7.64) (0.40–2.02) (1.27–2.68) (1.22–4.44)  Co-worker 0.80 0.62 0.97 2.45 0.14** 0.96 0.93 1.70 (0.48–1.35) (0.33–1.18) (0.52–1.80) (1.52– 3.95)** (0.04–0.57) (0.27–3.34) (0.42–2.06) (0.68–4.24)  Level 1 variance 1.25 1.24 1.02 0.62 1.21 1.15 0.97 0.56 (0.17) (0.24) (0.21) (0.31) (0.42) (0.67) (0.31) (0.34)  −2LL 859,928,640 390,913,472 452,567,008 222,713,520 88,179,943 19,448,486 230,494,992 114,654,792 Observations  Level 1 4619 2224 2395 1031 518 385 1223 752  Level 2 1233 583 650 273 134 106 331 202 Notes: Robust standard errors are shown in parentheses. aReligions none is reference group. bWhite is reference group; Race/ethnicity not included in model 5 because the overwhelming majority of Black Protestants are Black. cNortheast is reference group. ***p < .001; **p < .01; *p < .05 (two-tailed tests). View Large Though religious discussion seems a relational adhesive in the full sample model (model 1), further analysis reveals that this phenomenon is contingent upon ego’s own religiosity. Among those who are connected to a religious congregation, discussing religion is associated with an 80% increase in the odds (95% CI = 1.22–2.67) that the tie will remain at W2. The relationship is nonsignificant among that half of the sample who has no congregational involvement (OR = 1.24; 95% CI = 0.83–1.85). Protestants are the other subgroups that stand out. Alters identified by Evangelicals, the conservative branch of Protestantism, have 225% higher odds of remaining in the W2 network if they and ego shared religious conversation than if they did not. For Mainline Protestants, talking religion increases the odds of tie persistence by a factor of 2.88. Among Catholics, Black Protestants, or religious nones, however, there was no evidence that religious discussion with a given alter increased tie durability between waves; each of these groups had odds ratios statistically indistinguishable from 1.0. Supplementary analyses further examined whether congregational affiliates, Evangelicals, and Mainline Protestants held on to their religious discussion partner ties merely because these alters tended to belong to their churches. I incorporated information about whether the tie was a co-congregant from the name-interpreter section of the PALS survey (respondents were asked “Which of the people you mentioned [as network members] are involved in your congregation”) and re-estimated models 2, 4, and 6. Results indicate that co-affiliation in a congregation—a classic instance of focused activity as a basis of network connectedness (Feld 1981)—was unable to explain why religious discussion partners tended to have staying power in these ego networks. Finally, supplementary analyses considered whether the longitudinal findings reported in table 4 could be observed beyond the strongest of people’s strong ties. Taking the largest subgroup of religious participants, those connected to a congregation, I limited network ties to those with whom ego did not receive helpful advice and who fell in the bottom half of the roster (i.e., named 3rd and 4th rather than given 1st or 2nd priority). Results from this “less-core” subset of the close network—those lacking the strongest signals of tie strength at W1—reveal that the association between religious discussion and tie persistence remained quite strong (OR = 2.21, p < .001). DISCUSSION This study addressed two main research questions. First, what individual and relational factors predict the discussion of religion in close networks? Analyses reveal that American adults report discussing religion with about half of their closest nonresidential ties in the past year. Among the strongest individual-level predictors of discussing religion was ego’s involvement in a congregation. This dimension of religiosity surpassed the importance of denominational affiliation, suggesting that for discussion of religion in networks, the behavioral expression of religious fidelity is more important than the religious group with whom one identifies. Contrary to hypotheses, there was no evidence that religious discussion differed by age, or (with the exception of Catholics) that people were more likely to have talked religion with network members amidst stressful life conditions such as financial strain or a loved one’s death. Findings did, however, provide support for each of the three theoretical frameworks offered as relational explanations for religious discussion. Namely, sharing a religious background, being strongly tied, and having ample opportunity for conversation each predicted religious discussion between PALS respondents and their alters. Relational explanations tended to differ somewhat, however, according to religious tradition. Most notably, religious homophily was influential among people connected to a religious congregation and among Mainline and Evangelical Protestants, while sharing a nonreligious outlook with one’s ties did not appear to matter among religious nones. The one relational predictor that was observed across all stripes of religiosity was whether the alter had provided useful advice—ties that offer counsel tend to be ties with whom ego is most likely to discuss religion, and this whether or not ego him/herself is religious. Building from these initial findings, the second research question considered whether discussion of things religious predicts the presence of a given tie over the course of six years. This inquiry addresses the central question of network change (“why do some ties persist more than others?”; Suitor et al. 1997:4), recognizing that discussion topics themselves could plausibly solidify or destabilize social bonds. Results indicate that talking about religion was associated with a greater likelihood of the network member remaining in ego’s close network. It may be that delving into matters of ultimate significance (at least from the perspective of ego) imbues a particular relationship with heightened significance and cements its presence in a network over time. Alternatively, the types of ties that venture into religious topics may be the ones with the strongest staying power. Theoretical accounts of relationship progression and intimacy-disclosure dynamics suggest that both processes are likely in operation, though this study is unable to decisively arbitrate causal directionality. Other studies have considered how factors such as kinship status, closeness, and contact frequency are associated with tie persistence (Lubbers et al. 2010; Morgan et al. 1996; Suitor and Keeton 1997), but this study breaks new ground by examining the role of conversational substance. There may be additional discussion topics that too serve as relational adhesives—talking about personal health or politics, for instance, could plausibly enhance tie durability. At the same time, it is uncertain whether these types of topics carry the existential or sacred weight that characterizes religion among so many U.S. adults. Nevertheless, the topical exceptionalism of religion relative to other discussion matters is a hypothesis that could be addressed in future research. It is worth noting that the close networks measured in the PALS include only up to four members, so it is quite possible that initial alters identified at baseline but left unnamed at survey follow-up were still an active presence in the respondent’s life. Still, the fact that religious discussion is such a strong predictor of remaining among one’s top four associates—above and beyond the power of receiving social support, sharing the same gender and racial background, talking frequently, or being prioritized as a “top” tie at W1—tells us something of the significance of talking religion, at least when it comes to a person’s inner circle. If not working as a “relational adhesive” in any strictly causal sense, religious discussion would then seem to capture an unacknowledged and undertheorized dimension of tie strength. As expected, discussing religion seems to matter most among those who are involved in a religious congregation (particularly among Mainline and Evangelical Protestants). And, for the religious, talking religion is more important than merely sharing the same faith background or even sharing a congregation. This study contributes to research on religious networks and on the discussion of sensitive matters in close networks. Existing research on touchy topics is dominated by analyses of health-focused or politically focused conversation (Klofstad et al. 2009; Perry and Pescosolido 2010). Religion, surprisingly, has been largely left out of consideration. Indeed, research more generally addressing the overlap between social networks and religion remains at an early stage, mainly because few large-scale representative surveys with ego-network methodology have seriously considered religion and spirituality (Everton 2016). A recent string series of studies has begun to address the role of religion in the social ties in lives of American adults, demonstrating how network ties’ religiosity aids volunteer recruitment efforts (Merino 2013) and shapes social support provision (Merino 2014; Schafer 2015). This study complements these efforts by shedding light on the basic matter of religion as a discussion topic in close relationships. Still, more research is needed to advance the themes presented in this study. First, what do people actually talk about when they discuss religion? It would be helpful, for instance, to distinguish talk of personal religious beliefs and spirituality from conversations about the intersection of religion and public life. This may be a place to more carefully consider gender differences. The current findings revealed that women are more likely than men to talk religion with their network members, but it is unclear whether men and women talk about religion in gender-specific ways (Ozorak 1996; Thompson 1991). The finding of regional variation also raises questions about whether people in the American South versus the American West—both regions where people outtalk those in the distinctively “post-Christian” Northeast (Barna Group 2017)—discuss religion in varied ways. Chronicling precisely how people talk religion within various regional contexts will deepen our understanding of the microlevel, interactional dimensions of society’s secular turn (Voas and Chaves 2016). Understanding the disclosure process would be another fruitful direction for future research. People tend to broach political topics when they anticipate that their network member shares their beliefs (Cowan and Baldassarri 2017). Do similar processes apply when it comes to disclosing religious beliefs? Further, does this disclosure differ for religious and nonreligious Americans? As a distrusted minority group in the United States (Edgell et al. 2016), the nonreligious may be more guarded than are people in more sizeable religious groups. Third, future research should consider other results of religious conversation within social networks. This study used network turnover as an initial outcome to understand why religious discussion may be consequential, but there are other processes that could be affected by talking religion or by religious disclosure. For instance, does religious discussion between religiously heterogeneous dyads spark tension or increase argumentation? Does frequently talking religion lead to more tolerance or instead to ideological polarization within networks? Such questions are timely, given the widespread perception of a growing religious divide in the United States (Stetzer 2015). Whole-network analyses may usefully supplement the current study’s ego-centered network approach and demonstrate the bottom-up structural dynamics of religious division in particular populations. Though helping stimulate these promising avenues for future work, this study is not without its limitations. The name-generator technique used in the PALS survey asks about people’s four closest nonresidential ties, so I am unable to document how much people discussion religion with weaker ties. This may be a reason why talking religion with religiously dissimilar alters did not predict tie loss (or negatively moderate the association between talking religion and tie durability); limiting the network to only one’s four closest associates could mean that the ties are close enough to withstand disagreement over a touchy topic or the occasional awkward conversation. Another measurement limitation is that respondents report only if they discussed religion with their tie, not how often (or in what ways) they had such discussions. Likewise, in their efforts to limit the time demands of the survey, the research team failed to ask participants how long they had known the alter or anything about the person’s political views. In spite of these drawbacks, the PALS data offer several clear advantages. It is, to my knowledge, the only publically available, nationally representative survey study that addresses religious discussion in close networks. Further, its panel design and careful documentation of W1 and W2 network members’ identity enable the longitudinal analysis of tie persistence. Still, much could be gained from collecting data with more nuanced attention to the particularities of religious discussion and in relation to other important features of the relationship. Finally, as insinuated above, the analysis of relational factors as predictors of religious discussion can provide little leverage on questions of causality. The majority of PALS respondents’ alters named at W1 did not reappear in W2 network rosters, making it untenable to adequately test competing longitudinal processes in PALS respondents’ networks (i.e., religious homophily → religious discussion vs. religious discussion → religious homophily). In conclusion, sociologists have long contended that “religion is an eminently social thing” (Durkheim 1912 [1995]:9), a system of beliefs and practices sustained by groups and capable of fostering social solidarity. Though many of the social forms of religious behavior occur mainly within organizational settings such as churches or mosques, social networks are other important—if poorly understood—contexts for religious expression. This study finds that both individual and relational factors shape the discussion of religion within close networks, and it also reveals that for many Americans, talking religion has significant implications for network continuity. In so doing, it invites future investigation into what specific religious topics emerge in people’s networks and what are the other implications of discussing religion for social network processes. ACKNOWLEDGMENTS This article has benefited from constructive feedback by participants at the University of Western Ontario Sociology Colloquium series and from helpful comments from the Editor and reviewers of Sociology of Religion. Data for this study come from the Portraits of American Life Study, a publicly available dataset available at http://www.palsresearch.org/pals/researchers. Footnotes 1 In the terminology of social network analysis, an ego is the focal individual (e.g., the sampled survey respondent), while an alter is the person identified by ego as a member of his/her social network. 2 For instance, money and household finances are “important matters” that people often discuss with a spouse, whereas community issues or politics tend to be the matters about which people talk with friends (Bearman and Parigi 2004). 3 For a similar argument in the context of politics, see Cowan and Baldassarri (2017). 4 Though this study is concerned primarily with people’s close networks—all members of which could be considered “strong ties”—there is dyadic variability on such traits even within this inner circle (Merino 2014). 5 Closeness and kinship status, for instance, increase the durability of a social support tie over 10 years (Suitor and Keeton 1997). 6 Specifically, I used a variety of W1 survey measures to predict the conditional probability of remaining in the PALS sample. I then computed case-specific predicted probabilities from the logistic regression model and took the inverse of this value. Finally, I multiplied each propensity score by the standard survey weights. In effect, this procedure up-weights those cases least likely to have contributed data at W2. 7 This name-generator tool departs from its close ego-network counterpart in the General Social Survey (GSS), but in ways advantageous to the present research questions. Specifically, the PALS network is not defined by whether people discuss important matters, an issue of inevitable interpretive difference across various respondents (e.g., is religion an “important matter”?). Rather, the network starts with people who are close, and in the name-interpreter, follow-up questions ascertain whether or they happen to discuss religion. That being said, it is important to note that the PALS close network is wholly nonresidential, whereas the GSS allows housemates as members of the core discussion network. 8 Frequency of communication is also considered an element of tie strength (Granovetter 1973), so a significant association with religious discussion would offer support for both the opportunity mobilization and the strength of tie framework. 9 The interviewer further specified that congregation refers to a “church, temple, synagogue, mosque, or other place of worship.” 10 Fitting a model that excludes affiliation also has the advantage of generating a more sensible estimate of whether Black people are more/less likely to talk religion than White adults. Participants who identify as Black Protestant are overwhelming Black, generating considerable multicollinearity in a model that adjusts for both race/ethnicity and denominational affiliation. 11 “Other religion” is not analyzed as its own subgroup in these models because it is a heterogeneous mix of religious affiliations. 12 Coefficients for quadratic and cubic specifications of age failed to achieve statistical significance in supplementary analyses. 13 Supplementary analyses further examined the potential distinctiveness of the South with respect to relational characteristics. Most associations were consistent between Northeast and South when the sample was divided into regional subgroups though several differences did emerge. First, religious homophily was not significantly associated with religious discussion in the South, whereas it was in the Northeast. Second, being highly ranked in the network and receiving help is a strong predictor of talking religion in the South, whereas these factors are noticeably weaker and nonsignificant in the Northeast. Frequency of contact, however, was a stronger predictor in the Northeast. 14 Supplementary analyses reveal that the positive association between college-educated alter and religious discussion was driven primarily by cases where the alter’s education exceeded that of the respondent. REFERENCES Altman , Irwin , and Dalmas A. Taylor . 1973 . Social Penetration: The Development of Interpersonal Relationships . New York, NY : Holt, Rinehart & Winston . Baldassarri , Delia , and Peter Bearman . 2007 . “ Dynamics of Political Polarization .” American Sociological Review 72 : 784 – 811 . Google Scholar CrossRef Search ADS Barna Group . “The Most Post-Christian Cities in America: 2017.” Barna: Cities. https://www.barna.com/research/post-christian-cities-america-2017/. Bearman , Peter S. , and Paolo Parigi . 2004 . “ Cloning Headless Frogs and Other Important Matters: Conversation Topics and Network Structure .” Social Forces 83 : 535 – 57 . Google Scholar CrossRef Search ADS Bidart , Claire , and Daniel Lavenu . 2005 . “ Evolutions of Personal Networks and Life Events .” Social Networks 27 : 359 – 76 . Google Scholar CrossRef Search ADS Brashears , Matthew E . 2014 . “ ‘Trivial’ Topics and Rich Ties: The Relationship between Discussion Topic, Alter Role, and Resource Availability Using the ‘Important Matters’ Name Generator .” Sociological Science 1 : 493 – 511 . Google Scholar CrossRef Search ADS Burt , Ronald S . 1984 . “ Network Items and the General Social Survey .” Social Networks 6 : 293 – 339 . Google Scholar CrossRef Search ADS Collins , Nancy L. , and Lynn Carol Miller . 1994 . “ Self-Disclosure and Liking: A Meta-Analytic Review .” Psychological Bulletin 116 : 457 – 75 . Google Scholar CrossRef Search ADS PubMed Cooperman , Alan . 2016 . Many Americans Don’t Argue about Religion—or Even Talk about It . Washington, DC : Pew Research Center . http://www.pewresearch.org/fact-tank/2016/04/15/many-americans-dont-argue-about-religion-or-even-talk-about-it/. Cornwell , Benjamin . 2015 . “ Social Disadvantage and Network Turnover .” Journals of Gerontology, Series B: Psychological Sciences and Social Sciences 70 : 132 – 42 . Google Scholar CrossRef Search ADS Cornwell , Benjamin , and Edward O. Laumann . 2015 . “ The Health Benefits of Network Growth: New Evidence from a National Survey of Older Adults .” Social Science & Medicine 125 : 94 – 106 . Google Scholar CrossRef Search ADS Cornwell , Erin Y. , and Linda J. Waite . 2012 . “ Social Network Resources and Management of Hypertension .” Journal of Health and Social Behavior 53 : 215 – 231 . Google Scholar CrossRef Search ADS PubMed Cowan , Sarah K. , and Delia Baldassarri . 2017 . “ ‘It Could Turn Ugly’: Selective Disclosure of Attitudes in Political Discussion Networks .” Social Networks 52 : 1 – 17 . Google Scholar CrossRef Search ADS Dallas , Kelsey . 2016 . “What Keeps Believers from Talking about Faith in Public?” DeseretNews.Com , May 19 . http://www.deseretnews.com/article/865654517/Whats-lost-when-people-stop-talking-about-faith-in-public.html. Durkheim , Emile . 1912 . Elementary Forms of the Religious Life . New York, NY : Free Press . Edgell , Penny , Douglas Hartmann , Evan Stewart , and Joseph Gerteis . 2016 . “ Atheists and Other Cultural Outsiders: Moral Boundaries and the Non-Religious in the United States .” Social Forces 95 : 607 – 38 . Google Scholar CrossRef Search ADS Emerson , Michael O. , David Sikkink , and Adele D. James . 2010 . “ The Panel Study on American Religion and Ethnicity: Background, Methods, and Selected Results .” Journal for the Scientific Study of Religion 49 : 162 – 71 . Google Scholar CrossRef Search ADS Everton , Sean F . 2016 . “ Networks and Religion: Ties that Bind, Loose, Build Up, and Tear Down .” Connections 16 : 1 – 33 . Feld , Scott L . 1981 . “ The Focused Organization of Social Ties .” American Journal of Sociology 86 : 1015 – 35 . Google Scholar CrossRef Search ADS Feld , Scott L. , J. Jill Suitor , and Jordana Gartner Hoegh . 2007 . “ Describing Changes in Personal Networks over Time .” Field Methods 19 : 218 – 36 . Google Scholar CrossRef Search ADS Gerber , Alan S. , Gregory A. Huber , David Doherty , and Conor M. Dowling . 2012 . “ Disagreement and the Avoidance of Political Discussion: Aggregate Relationships and Differences across Personality Traits .” American Journal of Political Science 56 : 849 – 74 . Google Scholar CrossRef Search ADS Gerber , Alan S. , Gregory A. Huber , David Doherty , Conor M. Dowling , and Seth J. Hill . 2013 . “ Who Wants to Discuss Vote Choices with Others? Polarization in Preferences for Deliberation .” Public Opinion Quarterly 77 : 474 – 96 . Google Scholar CrossRef Search ADS Granovetter , Mark S . 1973 . “ The Strength of Weak Ties .” American Journal of Sociology 78 : 1360 – 80 . Google Scholar CrossRef Search ADS Green , Emma . 2016 . “The Impoliteness of Talking about Religion.” The Atlantic . April 12 . https://www.theatlantic.com/politics/archive/2016/04/the-impoliteness-of-talking- about-religion/477834/. Klofstad , Casey A . 2007 . “ Talk Leads to Recruitment: How Discussions about Politics and Current Events Increase Civic Participation .” Political Research Quarterly 60 : 180 – 91 . Google Scholar CrossRef Search ADS Klofstad , Casey A. , Scott D. McClurg , and Meredith Rolfe . 2009 . “ Measurement of Political Discussion Networks: A Comparison of Two ‘Name Generator’ Procedures .” Public Opinion Quarterly 73 : 462 – 83 . Google Scholar CrossRef Search ADS Lawler , Edward J. , and Shane R. Thye . 1999 . “ Bringing Emotions into Social Exchange Theory .” Annual Review of Sociology 25 : 217 – 44 . Google Scholar CrossRef Search ADS Lubbers , Miranda J. , Jose Luis Molina , Jurgen Lerner , Ulrik Brandes , Javier Avila , and Christopher McCarty . 2010 . “ Longitudinal Analysis of Personal Networks. The Case of Argentinean Migrants in Spain .” Social Networks 32 : 91 – 104 . Google Scholar CrossRef Search ADS Machacek , David W . 2003 . “ The Problem of Pluralism .” Sociology of Religion 64 : 145 – 61 . Google Scholar CrossRef Search ADS Marsden , Peter V . 1987 . “ Core Discussion Networks of Americans .” American Sociological Review 52 : 122 – 31 . Google Scholar CrossRef Search ADS Martin , Judith . 1998 . “ A Few Topics to Avoid in Social Conversation .” Chicago Tribune , January 1 . McClurg , Scott D . 2003 . “ Social Networks and Political Participation: The Role of Social Interaction in Explaining Political Participation .” Political Research Quarterly 56 : 449 – 64 . Google Scholar CrossRef Search ADS McPherson , Miller , Lynn Smith-Lovin , and James M. Cook . 2001 . “ Birds of a Feather: Homophily in Social Networks .” Annual Review of Sociology 27 : 415 – 44 . Google Scholar CrossRef Search ADS ———. 2013 . “ Religious Social Networks and Volunteering: Examining Recruitment via Close Ties .” Review of Religious Research 55 : 509 – 527 . CrossRef Search ADS Merino , Stephen M . 2014 . “ Social Support and the Religious Dimensions of Close Ties .” Journal for the Scientific Study of Religion 53 : 595 – 612 . Google Scholar CrossRef Search ADS Morgan , David L. , Margaret B. Neal , and Paula Carder . 1996 . “ The Stability of Core and Peripheral Networks Over Time .” Social Networks 19 : 9 – 25 . Google Scholar CrossRef Search ADS Ozorak , Elizabeth Weiss . 1996 . “ The Power, but Not the Glory: How Women Empower Themselves Through Religion .” Journal for the Scientific Study of Religion 35 : 17 – 29 . Google Scholar CrossRef Search ADS Pargament , Kenneth I . 2002 . “ The Bitter and the Sweet: An Evaluation of the Costs and Benefits of Religiousness .” Psychological Inquiry 13 : 168 – 81 . Google Scholar CrossRef Search ADS Pargament , Kenneth I. , and Annette Mahoney . 2005 . “ Sacred Matters: Sanctification as a Vital Topic for the Psychology of Religion .” The International Journal for the Psychology of Religion 15 : 179 – 98 . Google Scholar CrossRef Search ADS Perry , Brea L. , and Bernice A. Pescosolido . 2010 . “ Functional Specificity in Discussion Networks: The Influence of General and Problem-Specific Networks on Health Outcomes .” Social Networks 32 : 345 – 57 . Google Scholar CrossRef Search ADS ———. 2015 . “ Social Network Activation: The Role of Health Discussion Partners in Recovery from Mental Illness .” Social Science & Medicine 125 : 116 – 128 . CrossRef Search ADS Pew Research Center . 2014 . “ Religious Landscape Study .” Pew Research Center’s Religion & Public Life Project . http://www.pewforum.org/religious-landscape-study/. Puetz , Kyle . 2015 . “ Consumer Culture, Taste Preferences, and Social Network Formation .” Sociology Compass 9 : 438 – 49 . Google Scholar CrossRef Search ADS Putnam , Robert D. , and David E. Campbell . 2010 . American Grace: How Religion Divides and Unites Us . New York, NY : Simon and Schuster . Schafer , Markus H . 2015 . “ Religiously Traditional, Unusually Supportive? Examining Who Gives, Helps, and Advises in Americans’ Close Networks .” Social Currents 2 : 81 – 104 . Google Scholar CrossRef Search ADS Schafer , Markus H. , and Nicholas Vargas . 2016 . “ The Dynamics of Social Support Inequality: Maintenance Gaps by Socioeconomic Status and Race ?” Social Forces 94 : 1795 – 822 . Google Scholar CrossRef Search ADS Small , Mario Luis . 2013 . “ Weak Ties and the Core Discussion Network: Why People Regularly Discuss Important Matters with Unimportant Alters .” Social Networks 35 : 470 – 83 . Google Scholar CrossRef Search ADS Smith , Christian . 2014 . American Evangelicalism: Embattled and Thriving . Chicago, IL : University of Chicago Press . Steensland , Brian , Lynn D. Robinson , W. Bradford Wilcox , Jerry Z. Park , Mark D. Regnerus , and Robert D. Woodberry . 2000 . “ The Measure of American Religion: Toward Improving the State of the Art .” Social Forces 79 : 291 – 318 .[AU: ] Google Scholar CrossRef Search ADS Stetzer , Ed . 2015 . “ Religious Polarization on the Way .” USA Today , November 8 . Suitor , J. Jill , Barry Wellman , and David L. Morgan . 1997 . “ It’s about Time: How, Why, and When Networks Change .” Social Networks 1 : 1 – 7 . Google Scholar CrossRef Search ADS Suitor , Jill , and Shirley Keeton . 1997 . “ Once a Friend, Always a Friend? Effects of Homophily on Women’s Support Networks across a Decade .” Social Networks 19 : 51 – 62 . Google Scholar CrossRef Search ADS Tabak , Melanie A. , and Kristin D. Mickelson . 2009 . “ Religious Service Attendance and Distress: The Moderating Role of Stressful Life Events and Race/Ethnicity .” Sociology of Religion 70 : 49 – 64 . Google Scholar CrossRef Search ADS Thompson , Edward H . 1991 . “ Beneath the Status Characteristic: Gender Variations in Religiousness .” Journal for the Scientific Study of Religion 30 : 381 – 94 . Google Scholar CrossRef Search ADS Thye , Shane R. , Jeongkoo Yoon , and Edward J. Lawler . 2002 . “ The Theory of Relational Cohesion: Review of a Research Program .” Advances in Group Processes 19 : 139 – 66 . Google Scholar CrossRef Search ADS van Duijn , Marijtje A. J. , Jooske T. van Busschbach , and Tom A. B. Snijders . 1999 . “ Multilevel Analysis of Personal Networks as Dependent Variables .” Social Networks 21 : 187 – 210 . Google Scholar CrossRef Search ADS Voas , David , and Mark Chaves . 2016 . “ Is the United States a Counterexample to the Secularization Thesis ?” American Journal of Sociology 121 : 1517 – 56 . Wei , Dedong , and Eric Y. Liu . 2013 . “ Religious Involvement and Depression: Evidence for Curvilinear and Stress-Moderating Effects among Young Women in Rural China .” Journal for the Scientific Study of Religion 522 : 349 – 67 . Google Scholar CrossRef Search ADS © The Author(s) 2018. Published by Oxford University Press on behalf of the Association for the Sociology of Religion. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com.

Journal

Sociology of ReligionOxford University Press

Published: Feb 17, 2018

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off