Religious Attendance and Social Networks
In line with previous research, religious attendance was linked to the objective characteristics of participants’ social networks, with significant effects observed for social network size [F(2, 561) = 11.98, p < 0.001]. Both FAs (20.03 ± 10.18) and IAs (16.85 ± 8.81) had larger social networks compared to NAs (15.29 ± 8.49; vs. IAs: p = 0.004; vs. FAs: p < 0.001), which is in line with H1.
When particular types of the relationships were investigated, we observed differences for relationship status [χ2(2) = 7.74, p = 0.021; in marital-like relationship: 48.3% of NAs, 56.1% of IAs, 41.2% of FAs], relatives other than partner, parents and children [F(2, 561) = 12.64, p < 0.001; more in FAs (2.25 ± 2.05, p < 0.001) and IAs (1.72 ± 1.75, p = 0.012) than in NAs (1.32 ± 1.41)], and co-students or teachers [F(2, 561) = 4.98, p = 0.007; FAs (4.58 ± 2.84) > NAs (3.60 ± 2.96) at p = 0.005]. Finally, a significant between-group difference was observed for the number of affiliates from church, temple, or other religious groups [F(2, 561) = 73.39, p < 0.001], with both FAs (1.68 ± 2.60) and IAs (0.12 ± 0.73) reporting more such affiliates than NAs at p < 0.001.
As we expected (H3), the difference in the overall size of the social network was still significant after excluding members of church or religious groups from the overall score [F(2, 561) = 5.41, p = 0.005]. Further analysis of this effect revealed that, in case of the adjusted social network size, only the difference between FAs (18.34 ± 8.99) and NAs (15.29 ± 8.49) was significant (p = 0.003), while IAs (16.73 ± 8.67) did not differ from any of the remaining groups.
No significant differences were observed between the groups in the mean level of loneliness [F(2, 561) = 2.61, p = 0.074]. However, when the R-UCLA subscales were analyzed, no between-group differences for the R-UCLA Intimate Others subscale [F(2, 561) = 1.23, p = 0.293] were observed, but FAs had lower levels of R-UCLA Social Others [F(2, 561) = 3.93, p = 0.020] and R-UCLA Belonging and Affiliation [F(2, 561) = 3.07, p = 0.047] than NAs. These findings are partially in line with our expectations of lower loneliness levels in FAs comparing to NAs (H2).
No significant differences were observed between the groups in the level of mental well-being [F(2, 561) = 1.94, p = 0.144].
Frequent religious attendance compared to non-attendance [F(2, 561) = 11.98, R2 = 0.041, p < 0.001; FAs vs. NAs: beta = 0.512, t = 4.87, p < 0.001] was a significant predictor of social network size. However, no similar effect was observed for infrequent religious attendance vs. non-attendance (IAs vs. NAs: beta = 0.169, t = 1.76, p = 0.08).
Social network size [F(3, 560) = 22.57, R2 = 0.11, p < 0.001; beta = − 0.321, t = − 7.87, p < 0.001], but neither IAs vs. NAs (beta = − 0.112, t = − 1.20, p = 0.23) nor FAs vs. NAs (beta = − 0.064, t = − 0.61, p = 0.54), predicted the loneliness level.
The indirect effect (full mediation) of the FAs vs. NAs on loneliness via social network size [beta = − 0.164, CI = (− 0.251; − 0.090)] was found to be significant.
Next, the same analysis was repeated for the adjusted SNI score. Again, FAs vs. NAs [F(2, 561) = 5.41, R2 = 0.019, p < 0.001; beta = 0.349, t = 3.28, p < 0.01], but not IAs vs. NAs (beta = 0.165, t = 1.69, p = 0.091), predicted adjusted social network size. Further, loneliness was predicted by adjusted social network size [F(3, 560) = 23.68, R2 = 0.11, p < 0.001; beta = − 0.325, t = − 8.076, p < 0.001], but not by the FAs vs. NAs (beta = − 0.115, t = − 1.12, p = 0.262) or IAs vs. NAs (beta = − 0.113, t = − 1.21, p = 0.226).
Again, the analysis revealed a significant indirect effect (full mediation) of the FAs vs. NAs on loneliness via adjusted social network size [beta = − 0.113, CI = (− 0.193; − 0.044)] (findings in line with H4).
Finally, both of the mediation analyses described above were extended by adding the overall well-being score as an outcome measure. Loneliness level was a significant predictor of overall mental well-being [F(2, 561) = 30.54, p < 0.001; R2 = 18, beta = 0.400, t = 9.86, p < 0.001]; however neither SNI (beta = − 0.048, t = − 1.17, p = 0.244) nor any of the religious attendance variables (IAs vs. NAs: beta = 0.034, t = 0.381, p = 0.704; FAs vs. NAs: beta = − 0.088, t = − 0.888, p = 0.375) predicted mental well-being. The same pattern of results was observed when the adjusted SNI score was included instead of the full SNI score.
In the case of both the full and adjusted SNI scores, the double mediation of the religious attendance FAs vs. NAs category via social network size and loneliness on mental well-being was observed [full SNI: beta = − 0.066, CI = (- 0.104; − 0.035); adjusted SNI: beta = − 0.045, CI = (- 0.080; − 0.017)]. In both cases no direct effects, or simple mediation (via SNI or R-UCLA) were significant. The model including FA (vs. NA) and the full SNI scores as predictors is presented in Fig. 1.
Lockdown Physical Contact and Social Resources
A main effect of religious attendance on the number of people with whom the participants were predicting to spend the next 2 weeks was observed [F (2, 561) = 8.02, p < 0.001], with FAs spending the subsequent 2 weeks with more people compared to NAs (p < 0.001). Investigation of the specific types of relationships showed that FAs were less likely to spend the next two weeks with a partner (FAs: 17.5%/ IAs: 35.6%/ NAs: 32.7%; χ2(2) = 14.75, p = 0.001) compared to other groups. At the same time, FAs were more likely to spend the next two weeks with parents (FAs: 68.2%/ IAs: 56.1%/ NAs: 49.3%; χ2(2) = 12.79, p = 0.002) and siblings (FAs: 46.6%/ IAs: 32.2%/ NAs: 28.9%; χ2(2) = 12.96, p = 0.002). No significant effects were found for the remaining categories (children, other family, roommates, coworkers, strangers).
When COVID-19-related social resources were investigated, the only significant effect was observed for less FAs expecting possible help from partners compared to other groups (FAs: 31.1%/ IAs: 47.3%/ NAs: 41.2%; χ2(2) = 9.41, p = 0.009). No between-group differences were found for the remaining categories (family, neighbors, friends, coworkers, public services, nobody).
Loneliness and Well-Being Changes Over Time
Analysis of the R-UCLA scores over one year showed a significant increase in loneliness levels [F(1, 91) = 9.62, p = 0.003; t1: 40.76 ± 10.76 vs. t2: 43.43 ± 12.30]. However, no main effect of religious attendance [F(2, 91) = 0.08, p = 0.922)] or an interaction between religious attendance and time [F(2, 91) = 0.28, p = 0.757] were found (not in line with H5).
Similarly, GHQ scores increased over the one-year period [F(1, 91) = 4.87, p = 0.030; t1: 65.88 ± 12.91 vs. t2: 70.21 ± 17.87). However, no main effects of religious attendance [F(2, 91) = 1.41, p = 0.249], nor an interaction between the factors [F(2, 91) = 1.09, p = 0.341], were found.
The current analysis aimed to examine the relationships between religious attendance, social network size, loneliness and mental well-being in young Polish adults during the COVID-19 pandemic.
Religious Attendance and Social Networks
In line with our expectations (H1), significant between-group differences in the objective characteristics of social networks were observed (i.e., larger social network size was found in both FAs and IAs, compared to NAs). This finding is congruent with previous conceptualizations of religious services as a multilayer social activity that can promote engagement in the various forms of worship-related social bonding (Dunbar, 2021; Rote et al., 2013). The notion that religious attendance may promote social contact with other attendees is reflected in the effect of religious attendance on the number of contacts associated with religious activities in the current study.
Yet, even after subtracting the number of contacts associated with religious worship from the overall social network size score, FAs still had larger social networks compared to NAs (in line with H3). This finding suggests that the effects of religious attendance on social functioning cannot be solely attributed to the additional contacts stemming from worship groups per se, which is in line with previous work emphasizing more general, structural effects that religion can have on one’s social network (Lewis et al., 2013; Ten Kate et al., 2017). More detailed investigation of the twelve different types of social relationships measured via the SNI found no between-group differences in contacts with parents and close friends. Interestingly, the lowest frequency of being in a marital-like relationship was found in the FAs group. Furthermore, in line with these results, the investigation of COVID-19 related social items revealed that FAs were less often planning to spend the subsequent weeks with their partners and they were less often expecting help from a partner in the case of COVID-19 quarantine. Given the negative association between religiosity and intimate coresidential unions (Thornton et al., 1992), these results suggest that FAs were both less likely to be in a marital-like relationship and to live with a romantic partner.
At the same time, analysis of the SNI subscales revealed that FAs had more regular contact with family members other than partners, parents and children, and with students or teachers from school, university, technical training, or adult education. Furthermore, FAs were more likely to spend the subsequent two weeks’ time with their parents and siblings. This finding suggests that FAs in the current study may be more likely to still be living with their family of origin, which in effect may have increased FAs’ propensity for contact with family members other than spouses, parents or children, as measured by the SNI. Furthermore, more participants from the FAs group reported being a student (85% of the group) and currently living in a city large enough (> 500 k inhabitants) to host an academic institution (73%). In both remaining groups, the number of students (NAs: 81%/ IAs: 79%) was more congruent with the percentage of participants living in cities large enough to host academic institutions (NAs: 88%/ IAs: 86%). This discrepancy suggests that FAs may be more likely to study out of their current residence place and to possibly commute for academic purposes. Both theoretical formulations (Tinto, 1993) and empirical data (Benson, 2007) suggest that commuter students could experience more difficulties in effectively forming new friends within the college community compared to non-commuter students. However, the methods of the current study did not distinguish between participants who study in their city of origin and those who moved to live on campus. Thus, even though the commuter students may be overrepresented in the FAs group, the difference in the size of the academic network may still stem from patterns of campus living, especially given the fact that the larger discrepancy between the fraction of the FAs currently living in (62.2%) and originating from the city (> 500 k; 25.0%). Further research will be needed to evaluate whether differences in size of the academic based network between FAs and NAs stem from different patterns of involvement in campus activities or other factors.
Religious Attendance, Loneliness and Mental Well-Being
Interestingly, despite the differences in objective social network size, religious attenders did not show lower levels of PSI than NAs (not congruent with H2). Furthermore, similar to (Rote et al., 2013), we observed a significant indirect effect of religious attendance on loneliness via social network size. However, this effect was limited to the FAs vs. NAs, and observed also for the adjusted social network size score, which did not include worship-based affiliates. In line with our hypothesis (H4) that the effects of religious attendance on mental well-being are sequentially mediated by both objective social network characteristics and subjective appraisals of one’s social bonds, this trajectory could also have been further extended towards mental well-being.
The loneliness measurement method used in the current study allows for investigation of the various aspects of PSI, ranging from loneliness in intimate relationships through loneliness in more casual social networks to the lack of a sense of belonging to a larger group or community. Interestingly, we did not observe between-group differences in the Intimate Others score, even though FAs were less often involved in marital-like relationships compared to the other groups (and particularly NAs). A single relationship status has been previously linked in young Polish adults with increased romantic and family loneliness (Adamczyk, 2016). Furthermore, it has been shown that emotional loneliness in young adults is linked only with the presence of a romantic partner, but not with the presence of a non-romantic close other (e.g., close friend), nor general social network size characteristics (Green et al., 2001).
On the other hand, the overall size of the social network (but neither the presence of a partner or close other) was shown to be significantly negatively correlated with social, but not emotional, loneliness in young adults (Green et al., 2001). Thus, the pattern of the between-group findings for the remaining loneliness subscales (lower levels of loneliness associated with more casual social relationships and a general sense of belonging to community in FAs compared to NAs) was congruent with the findings from the objective social networks measurement (increased overall network size, more contacts from further family, academic and church-based networks in FAs compared to NAs). Furthermore, a Polish study on 19–25 year old young adults also found that the effects of remaining single on romantic loneliness are mediated by perceived social support from family, with high support from the family mitigating the negative impact of prolonged single status on romantic loneliness (Adamczyk, 2016). Thus, it is plausible that the effects of an increased likelihood of not being in a marital-like relationship may have been mitigated by larger support networks in FAs. Furthermore, this effect may have suppressed the between-group differences in the overall loneliness score in the current study.
Religious Attendance, Loneliness and the COVID-19 Pandemic
Finally, additional evidence that the effects of religious attendance on loneliness and mental well-being cannot be fully explained by worship-related social groups comes from our longitudinal examination. In spite of our hypothesis (H5), we observed that PSI increased while mental well-being decreased over a one-year follow-up time in our participants, and neither of these effects were moderated by religious attendance. As religious ceremonies were among the very few exceptions for public gatherings that were not prohibited in Poland since the start of the COVID-19 restrictions, these findings suggest that being able to participate in social gatherings during worship practices did not translate into decreased loneliness in the religious attenders who completed follow-up in the current study.