Latent Profiles of Seminary Students’ Perceptions of Sense of Community Amidst the COVID-19 Pandemic

Abstract

Existing research on sense of community in educational contexts shows positive associations with well-being and negative associations with mental health symptoms. However, exploration of students’ sense of community within the seminary context is minimal. Drawing on the relational spirituality model, which posits oscillating experiences of dwelling and seeking in a dialectical growth process, we framed sense of community as horizontal dwelling. We used mixture modeling to generate subgroups using items from a measure of sense of community and then explored associations between sense of community and various demographic predictors and personal formation outcomes, including well-being and symptoms, along with virtues and religiousness/spirituality. We did so within the unique context of the COVID-19 pandemic. We used a sample of graduate students from 18 Christian seminaries across North America (N = 867; M_age = 31.95; 48.1% female; 60% White). Factor mixture results supported a 3-class solution, with one class reporting consistently greater ratings of sense of community, labeled the strongly connected, another class reporting consistent mid-level ratings, labeled the moderately connected, and a third class reporting consistently lower ratings, and labeled the disaffected. Results for the disaffected showed a pattern of associations with lower well-being and greater symptoms, and lower religiousness/spirituality, along with greater pandemic stress. The strongly connected showed greater levels of virtuousness, well-being and religiousness/spirituality, and lower symptoms, although they were also more likely to report greater illusory health. The pattern of associations for the moderately connected was a blend of similarities with the other two subgroups. The disaffected also showed a modest risk effect for lower well-being and greater symptoms over time. Findings pointed to providing greater horizontal dwelling among the disaffected and the need for greater seeking among the strongly connected.

1. Latent Profiles of Seminary Students’ Perceptions of Sense of Community Amidst the COVID-19 Pandemic

Porter et al. (2019) described an ecumenical meta-theory for the personal (i.e., religious/spiritual, character, moral) formation of students, grounded in the seminary, as an “intentional culture” of growth (p. 5). The construct of sense of community has been used to describe this potential for the seminary context to foster student growth (Chukwuorji et al. 2018). Dost and Mazzoli Smith (2023) defined sense of community within educational contexts as “feeling part of somewhere an individual can be themselves and feel confident in their personal and social identities, through secure, meaningful, and harmonious support” (p. 841). The first aspect of their definition captures an inclusive, culturally accepting environment, and the second aspect broadly taps into social support. Existing research on sense of community tends to show positive associations with subjective (e.g., positive emotion) and eudaimonic (e.g., life purpose) well-being in a variety of settings, including educational contexts (Stewart and Townley 2020). Existing research also tends to show that a greater sense of community corresponds to lower mental health symptoms (Coulombe and Krzesni 2019; Gopalan et al. 2022; Park et al. 2023).

In their meta-theory of change, Porter et al. (2019) drew on the relational spirituality model (RSM), which defines religiousness/spirituality (R/S) as “ways of relating to the sacred” (Sandage et al. 2020, p. 24). The metaphor of a crucible is used to describe a context conducive to growth, “a metaphor of intensification and containment” (Sandage et al. 2020, p. 71). Sense of community aligns with the containment aspect of the metaphor. In the RSM, containment represents one pole of the dialectic process of growth, labeled dwelling, which is grounded in “the safe haven function of the attachment system” (Porter et al. 2019, p. 20). In a seminary context, dwelling can include two dimensions: vertical and horizontal. The vertical dimension connotes perceived closeness to the divine or God (i.e., the sacred), intrinsic R/S and/or R/S commitment, or more broadly, an affective dimension, often labeled R/S salience (Pearce et al. 2013; Yonker et al. 2012). The horizontal dimension refers to aspects of interpersonal functioning in R/S relational contexts. Existing findings using samples of seminary students and practicing R/S leaders (e.g., clergy, chaplains) support direct associations between greater vertical dwelling and greater well-being and greater vertical dwelling and lower symptoms (e.g., Jankowski et al. 2019). These findings are consistent with the broader literature showing these effects for R/S salience (e.g., Hodapp and Zwingmann 2019; Yonker et al. 2012). In contrast, support for the associations between greater horizontal dwelling and greater positive outcomes is inferred from the salutary influence for positive R/S coping, which consists in part of seeking social support (e.g., Hodapp and Zwingmann 2019).

Empirical support for the positive influence of horizontal dwelling is also inferred from tests of conditional and indirect effects. For example, Sandage and Jankowski (2010) found that greater interpersonal forgivingness predicted lower emotionally dysregulated experiences with God through greater differentiation. Differentiation is an aspect of the RSM and represents a developmental construct involving healthy self- and other-relating via intra- and interpersonal regulation of negative emotion (Sandage et al. 2020). In addition, Sandage et al. (2015) found risk effects for R/S narcissism, which amplified the influence of the intervening variable of emotionally dysregulated experiences with God in the correspondence between greater insecure adult attachment and greater insecure God attachment. These findings suggest a potential for healthy relational processes to provide comfort during times of distress. Differentiation and attachment are emotion regulation constructs that operationalize the safe haven function of dwelling.

As important as providing horizontal dwelling experiences may be for student growth, the crucible metaphor also calls attention to intensification. In the RSM this is captured by the construct of seeking. Seeking connotes R/S exploration, or questing, and “maps onto the secure base function of attachment” (Sandage et al. 2020, p. 59). As such, the relational system provides a platform for venturing out or moving away from the known to the unknown. In a seminary context, this can involve students questioning prior beliefs and exploring new theological frameworks, encountering a diversity of perspectives and backgrounds, and wrestling with vocational options and a sense of calling. Seeking “can be highly anxiety-provoking or distressing” (Porter et al. 2019, p. 20), and as such, also calls forth the safe haven function. Horizontal dwelling can provide the secure base from which to launch into exploration and, if need be, provide the safe haven to return to should exploration have involved experiences of disappointment, frustration, estrangement, and/or suffering.

R/S exploration is often subsumed as an implicit aspect of negative coping or R/S struggles in the form of questioning or doubting, with greater levels typically corresponding to lower well-being and greater symptoms, although more direct assessment of R/S exploration (e.g., quest R/S) has shown a null effect (Hodapp and Zwingmann 2019). Using a direct assessment approach involving a subsample from the current study sample and the larger longitudinal investigation of seminary students’ personal formation, Jankowski et al. (2022d) found that greater R/S exploration longitudinally predicted lower joy (i.e., subjective well-being) and lower life purpose (i.e., eudaimonic well-being), when initial experiential avoidance (i.e., emotion dysregulation) was high and humility was low. Conversely, when initial experiential avoidance and humility were high, there was a positive influence for initial R/S exploration on later joy and life purpose. These findings converge with cross-sectional findings showing that greater humility predicted lower symptoms through greater differentiation when R/S exploration was high (Jankowski et al. 2019). Humility may protect against dysregulation when exploring because it involves the self-regulation of egoistic tendencies (Davis et al. 2016) grounded in secure attachment experiences (Dwiwardani et al. 2014). Taken together, dwelling can facilitate emotion regulation, which then potentiates a salutary influence for R/S exploration.

Porter et al. (2019) used the RSM language of purgation to capture a seeking process that corresponds to suffering and dysregulation. Purgation can either disrupt growth-promoting exploration and prompt a return to the familiar or motivate further seeking. The former represents complacent dwelling (Sandage et al. 2020) and, therefore, potentially a stagnant, rigid, and/or arrogant expression of R/S. In contrast, seeking that continues amidst intensification can lead to transformed dwelling, that is, the formation of qualitatively different meaning-making frames and greater differentiated and securely attached ways of relating to self, other, and the sacred. Growth is most likely to occur via the oscillation of dwelling and seeking or dialectical balancing of dwelling and seeking over time.

Jankowski et al. (2022b) identified a subgroup of seminary students that seemed to represent a balanced dwelling and seeking, which they labeled integrated. The integrated were comparable to another subgroup, labeled the closely monitored, that reported greater virtuousness and well-being. However, the integrated differed by reporting lower God monitoring, lower R/S struggles with God, lower R/S obsessions and compulsions, and lower impression management. The integrated and closely monitored reported comparable levels of R/S seeking, and the closely monitored reported higher levels of dwelling. Jankowski et al. described the relational spirituality profile of the closely monitored as internally motivated, and yet “lack[ing] autonomy and instead seem[ed] compulsory, motived by guilt or obligation … internally or externally pressured” (p. 672). In contrast, the integrated better depicted a balanced dwelling and seeking profile, consistent with differentiated and securely attached relating. They suggested that students be provided “structured opportunities to reflect on and dialogue about their identity commitments” in relationships that offered a secure base and safe haven (p. 673). Sandage et al. (2020) similarly referred to holding environments that flexibly “cultivate and repair a constructive … alliance and … resist the urges to ‘rescue or run way’ when crucible dynamics heat up” (p. 79).

2. The Current Study

Our study of seminary students’ sense of community is embedded in the unique pandemic context. Existing research on individuals’ well-being during the pandemic highlights varied experiences, with some reporting beneficial effects on their well-being and others reporting increased suffering, with the latter disproportionately reported among individuals with marginalized identities (e.g., Crabtree et al. 2023). Findings specific to a sense of community within the educational environment similarly show diverse and disproportionate adverse experiences for marginalized persons. Tice et al. (2021), for example, noted a decline in college students’ sense of belonging during the pandemic, levels which they indicated were already low prior to the pandemic. In contrast, others reported no change in college students’ sense of belonging pre- to post-pandemic declaration, except among underrepresented racial/ethnic students (Barringer et al. 2023; Gopalan et al. 2022). Gopalan et al. (2022) also found evidence that a sense of belonging acted as a protective factor, lowering mental health symptoms.

A prior study involving seminary students from the larger longitudinal investigation of seminary students’ personal formation explored the role of pandemic stress on students’ self-reported anxiety and burnout during their time in seminary. Jankowski et al. (2023) found that one subgroup displayed a resilience trajectory, reporting the highest pandemic stress, highest initial anxiety and burnout symptoms, and yet reported improved symptoms and growth in well-being. This subgroup also reported a decline in pandemic stress. In contrast, a second subgroup showed deterioration in symptoms and lower levels of well-being. This subgroup also reported no change in their moderate level of pandemic stress. Their trajectory displayed delayed onset (Schäfer et al. 2022), reporting comparable levels of anxiety to the resilient subgroup pre-pandemic declaration, which remained consistent until wave 4 when their anxiety showed an increase over the final three time points. A third subgroup displayed low pandemic stress, low symptom levels, and high well-being and was labeled flourishing. The resilient subgroup was more likely to report marginalized identification relative to the flourishing subgroup.

Accordingly, we expected to find distinct subgroups of students delineated by varying levels of their sense of community. We expected pandemic stress to predict subgroup membership. We also expected that demographic covariates, ideological commitment, and impression management would predict subgroup membership, consistent with prior research involving subgroups generated using R/S variables (Jankowski et al. 2022b). We also expected the subgroups to differ in well-being and symptoms, virtuousness, and R/S dwelling and seeking, consistent with the RSM and prior empirical findings (Jankowski et al. 2022b; Sandage et al. 2020). We also explored whether subgroups would differ by the R/S denominational tradition of the seminary and R/S formation activities.

3. Method

3.1. Participants

Longitudinal data were collected from students at 18 seminaries across North America, accredited by the Association for Theological Schools, across six time points from fall 2019 to spring 2022. The larger study (e.g., Jankowski et al. 2022b, 2023) employed an open cohort design (Mabmud 2010, para. 7), and the current study used the time 3 cohort from fall 2020 (N = 867; M_age = 31.95; SD = 10.79; range = 21–71). Among those providing demographic data, 50.5% identified as males and 48.1% as females. A majority identified as heterosexual (92%; 1.9% gay/lesbian; 2.6% bisexual, 1.2% other). Participants identified as 10.1% Black, 15.6% Asian, 60% White, 5.7% Hispanic, and 8.4% other/multiple ethnicities. A majority reported being single—never married (53.2%). Participants were primarily from evangelical Protestant seminaries (58.1%), whereas 25.5% were from mainline Protestant, 12.9% Catholic, and 3.5% Orthodox. Seventy-seven percent of students attending evangelical seminaries self-identified as evangelical, 46.4% self-identified as mainline and attended a mainline seminary, 98.2% of the students attending a Catholic seminary self-reported Catholic affiliation, and all students attending an Orthodox seminary identified as such. A majority reported they were enrolled in online classes (56.6%), 11.2% reported in-person enrollment, and 30.6% reported taking hybrid/multiple modalities.

3.2. Procedure

After obtaining IRB approval, beginning in the fall of 2019, students were invited to consent and complete an online survey in exchange for a $25 gift card. This recruitment process was repeated six times over three years. We began collecting data at wave 3 on students’ experience of their seminary context, with this sense of community as the primary construct under investigation in the current study.

3.3. Measures

Sense of community. We used 13 items adapted from measures of organizational support and commitment (Allen and Meyer 1990; Worley et al. 2009). We used the recommended robust diagonally weighted least squares estimation and the polychoric correlation matrix (Sellbom and Tellegen 2019) to conduct exploratory factor analysis (EFA) using FACTOR (version 12.04.05; Lorenzo-Seva and Ferrando (2023)). We used the default oblique Promin rotation. The Hull method based on the common part accounted for fit index (Lorenzo-Seva et al. 2011), the minimum average partial test (Lorenzo-Seva and Ferrando 2023), and classical parallel analysis (Timmerman and Lorenzo-Seva 2011) each suggested retaining two factors. The 2-factor first-order solution fit the data (Comparative Fit Index = 0.976, Root Mean Square Error of Approximation = 0.08, 90% confidence interval [CI] [0.07, 0.09]). The inter-factor correlation was significant (r = 0.63, bias-corrected 95% CI [0.58, 0.71]) but demonstrated acceptable discriminant validity since the CI did not cover 0.80 (Rönkkö and Cho 2022).

The 3 items on the first factor (ρ = 0.94; λs > 0.80; e.g., “My seminary is sensitive to cultural differences among students”) tapped into the dimension of cultural sensitivity. The 10 items on the second factor (ρ = 0.91; λs > |0.40|; e.g., “My seminary is willing to be flexible according to my unique needs”) were consistent with student perceptions of the institution as supportive of their psychological needs, academic preparation, vocational goals, and R/S formation, and labeled supportive.

An oblique exploratory bi-factor model showed evidence of a strong general factor, with two specific factors (λs > 0.30 for a majority of the items on the general and specific factors; Lorenzo-Seva and Ferrando (2019)). The bi-factor results suggested that the 13 items can be used to assess global sense of community (ω = 0.90). The explained common variance value was below the threshold of 0.85, supporting multidimensionality (Lorenzo-Seva and Ferrando 2023) and the use of the two specific factors.

COVID-19 stress. We used a single item (i.e., “All things considered, how would you assess the impact of COVID-19 on your life?”; Crabtree et al. 2023), rated from 0 (negatively) to 100 (positively), with higher scores indicating less felt stress.

R/S impression management. We used five items (ω = 0.79; e.g., “I am always in the mood to pray”; Hall and Edwards 2002), rated from 1 (not at all true) to 5 (very), with higher scores indicating socially desirable responding in the form of greater exaggerated sense of relating to the divine. The interpretation of exaggerated R/S is consistent with the strong correlation in our current study between impression management and perceived closeness to God (r = 0.58, p < 0.001).

Ideological commitment. We used a single item (Perry 2015) to assess R/S ideological commitment ranging from 1 (very conservative) to 7 (very liberal).

R/S formation activities. We used four items that were rated as most influential in response to a prompt asking participants to rank the “seminary-related activities based on their positive influence on your spiritual life over the last 6 months,” using a scale from 1 (most influential) to 10 (least influential). We created dichotomous variables using the median response category, coded 0 = less positive, 1 = more positive, for each of the following activities: “In-class experiences (lectures, discussions, in-class activities)”, “Informal friendships/community experiences at my school”, “Outside of class assignments and readings”, and “Small group experience required by my school”.

Joy. We used three items (ω = 0.86; e.g., “I consistently feel a subtle but enduring feeling of joy”; Watkins et al. (2018)), rated from 1 (disagree strongly) to 7 (agree strongly), with higher scores indicating greater positive emotion.

Humility. We used the 9-item Expressed Humility Scale (ω = 0.82; e.g., “I admit it when I don’t know how to do something”; Owens et al. (2013)), with items rated from 1 (disagree strongly) to 5 (agree strongly), with higher scores indicating greater humility.

Social justice activism. We used five items (ω = 0.84; e.g., “I speak out for equality for immigrant communities”; Jankowski et al. 2022e), rated from 1 (never true) to 5 (always true), with higher scores indicating greater activism.

Calling. We used the 2-item presence subscale from the Brief Calling Scale (Spearman-Brown coefficient = 0.82; e.g., “I have a good understanding of my calling as it applies to my career”; Dik et al. 2012), rated from 1 (not at all true of me) to 5 (totally true of me). Higher scores indicated a greater sense of “a transcendent summons” to a specific job (Dik et al. 2012, p. 244), consistent with greater life purpose (e.g., r = 0.57, p < 0.001; for the correlation between calling–presence and life purpose–presence (Bronk et al. 2018) in the current study sample).

Differentiation. We used the 3-item emotion cutoff subscale from the Differentiation of Self Inventory—Short Form (ω = 0.89; e.g., “I tend to distance myself when people get too close to me”; Drake et al. (2015)), rated from 1 (not at all) to 6 (very characteristic of me), with higher scores indicating lower emotional cutoff (i.e., greater interpersonal emotion regulation).

Disconnection. We used three social connectedness items from a measure of belongingness (ω = 0.84; e.g., “I have little sense of togetherness with my peers”; Lee and Robbins (1995)), rated from 1 (strongly disagree) to 6 (strongly agree), with higher scores indicating greater disconnection.

Perceived closeness to God. We used five divine items from the Daily Spiritual Experience Scale (ω = 0.86; e.g., “I feel God’s presence”; Underwood and Teresi (2002)), rated from 1 (never or almost never) to 6 (many times a day), with higher scores indicating greater perceived closeness.

R/S exploration. We used six items (ω = 0.84; e.g., “I have been placing a lot of effort in exploring spiritual and/or religious questions”; Beck and Jessup (2004)), rated from 1 (disagree strongly) to 7 (agree strongly), with higher scores indicating greater exploration.

R/S struggles. We used the 3-item divine subscale from the 14-item Religious and Spiritual Struggles Scale (RSSS; ω = 0.78; e.g., “Felt as though God had abandoned me”; Exline et al. (2023)), rated from 1 (not at all/does not apply) to 5 (a great deal), with higher scores indicating greater struggles most consistent with greater anxious attachment to God (e.g., r = 0.44, p < 0.001; for the correlation between divine struggles and anxious attachment to God [Beck and McDonald 2004] in the current study sample). We also used the 3-item interpersonal subscale from the 14-item RSSS (ω = 0.77; e.g., “felt hurt, mistreated, or offended by religious/spiritual people”; Exline et al. (2023)), with higher scores indicating greater interpersonal R/S conflict. The divine and interpersonal subscales were moderately correlated (r = 0.42, p < 0.001, in the current study sample).

Avoidant attachment to God. We used five items from the 14-item avoidance subscale from the Attachment to God Inventory (ω = 0.76; e.g., “I prefer not to depend too much on God”; Beck and McDonald 2004), rated from 1 (disagree strongly) to 7 (agree strongly), with higher scores representing greater perceived insecurity in relating to God.

Mental health symptoms. We used the Generalized Anxiety Disorder 2-item scale (Spearman –Brown coefficient = 0.83; e.g., “Not being able to stop or control worrying”; Kroenke et al. (2010)) and the Patient Health Questionnaire 2-item scale (Spearman-Brown coefficient = 0.80; e.g., “Feeling down, depressed, or hopeless”; Kroenke et al. (2010)), with items rated 0 (not at all) to 3 (nearly every day), and higher scores indicating greater anxiety and depression symptoms, respectively.

3.4. Data Analytic Plan

Data were analyzed using mixture modeling procedures in Mplus (version 8.4; Muthén and Muthén 1998–2017, 2019; i.e., type = mixture). Data were multivariate non-normal (Mardia’s multivariate kurtosis statistic = 1761.29, p < 0.001; Cain et al. 2017). Data were assumed missing at random based on the low normed χ² (χ²/df) of 1.32 for Little’s MCAR test. We used maximum likelihood estimation with robust standard errors to account for non-normality and to handle missing data by using all available data to estimate parameters. Class enumeration was based on the Bayesian Information Criterion (BIC) and the Lo–Mendell–Rubin adjusted likelihood ratio test (LMR-LRT; Nylund-Gibson and Choi 2018). We also considered the quality of class separation and accuracy using entropy values > 0.80 considered good (Nylund-Gibson and Choi 2018). We used the manual two-step procedure to examine the influence of covariates on sense of community class membership and to examine differences by class on the distal outcomes (Bakk and Kuha 2018, 2021). We opted for the two-step procedure because it permitted us to build a more complex model involving multiple latent variables (Bakk and Kuha 2021). Specifically, we linked two latent categorical variables in a multinomial logistic regression, creating a type of hybrid mixture model (Sterba 2013). For the second latent variable, we explored change in flourishing using longitudinal mixture modeling (van der Nest et al. 2020) with data from waves 2 (n = 540) through 4 (n = 749), during the months preceding wave 3 (M = 4.50, SD = 2.78; wave 2–wave 3) and the following months (M = 4.84, SD = 2.96; wave 3–wave 4).

4. Results

Initial exploration using LPA revealed three- (BIC = 27,797.68) and four-class solutions (BIC = 27,491.65) as candidate models. In addition to the lower BIC, the four-class solution was also supported by a plateauing of the likelihood increment percentage per parameter index after the fourth class (Grimm et al. 2021), whereas the three-class solution was supported by an elbow in the plot of BIC values and the Lo–Mendell–Rubin adjusted likelihood ratio test in which the p-value became > 0.05 with the four-class solution (Nylund-Gibson and Choi 2018). However, there were significant correlations among the sense of community indicator variables, suggesting a violation of the assumption of local independence (Muthén 2018). Unaccounted-for correlations among indicators can confound decisions about the number of classes that best fit the data (Muthén 2004). One option is to relax the assumption by correlating the indicators and conducting mixture modeling (Muthén 2020; Muthén and Muthén 1998–2017). A second option is to conduct factor mixture modeling, in which a continuous factor model is used to capture the correlations among indicators within classes (Muthén 2009, 2015; Sterba 2013). We proceeded with a model-building and comparison process to determine a final solution (Clark et al. 2013).

We decided that a two-factor, three-class factor mixture model (FMM) best fit the data (BIC = 26,436.05; entropy = 0.996). In this model, factor loadings were invariant across classes, whereas indicator means, covariances, and variances varied across classes, and factor means were fixed to zero. Clark et al. (2013) labeled this variation FMM-3, in contrast to FFM-1 and FMM-2 variations (see also Kim et al. 2023). The three-class model fit better relative to two-factor, two- (BIC = 26,594.86) and four-class (BIC = 26,465.67) FMM-3 solutions. This model also fit better than the best-fitting three-class FMM-1 model (BIC = 26,681.67), in which factor loadings and indicator means were invariant across classes, the inter-factor covariance was set to zero, and factor means were allowed to vary. The three-class FMM-3 model also fit better than the best fitting two-class FMM-2 model (BIC = 26,654.26), which differed from the FMM-1 by allowing the inter-factor covariance to vary across classes. The FMM-3 model is determined by participants’ item responses rather than differences in the factor means as in the FMM-1 and FMM-2 variations. The two-factor FMM-3 model also fit better relative to the best-fitting one-factor solutions, and specifically, the three-class FMM-1 (BIC = 27,434.08), two-class FMM-2 (BIC = 27,439.53) and two-class FMM-3 (BIC = 27,022.75) models. We encountered convergence issues when fitting the FMM-4 variation, which differed from the FMM-3 model by also allowing factor loadings to vary within each class (Clark et al. 2013). Last, the two-factor, three-class FMM-3 model fit better relative to the best-fitting four-class mixture model with correlated indicators (BIC = 26,599.75).

Figure 1 depicts the plot of estimated means on the sense of community indicators by class. Only three pairwise comparisons fell below the threshold of Cohen’s d < 0.50 as an indication of poor separation between classes (Diallo et al. 2017), with the difference between classes 1 and 2 on item 10 the only nonsignificant comparison (see Figure 1). A majority of the comparisons were consistent with the commonly reported between-class differences of d > 0.80, showing acceptable class separation (Tein et al. 2013). By comparison, each of the differences between classes for items 12 and 13 represented good class separation of d > 1.25 (Tein et al. 2013; See Figure 1). We labeled class 1 the disaffected (Bulbulia et al. 2020), which connotes lower levels of R/S (Bulbulia et al. 2020), and in an educational context, can connote “a low sense of belonging” and “less engagement in school activities” (Kuang et al. 2020, p. 141), or disconnected (Patte et al. 2021) and disengaged (Lingo and Chen 2023). The construct of disaffected also connotes a lower sense of community or less felt need fulfillment, which includes needs for belonging and mattering (Stewart and Townley 2020). As such, our use of disaffected is consistent with an R/S orientation (Bulbulia et al. 2020) or way of relating to the sacred (Sandage et al. 2020), which is distinct from disaffiliation (i.e., “‘dropping out’ and … disidentification”) and deconversion (i.e., “no longer believ[ing]”), although social disconnection is often an aspect of disaffiliation and deconversion (Fisher 2017, p. 362). We opted to label class 3, our reference class, the strongly connected, and class 2, moderately connected, given more positive ratings on the vast majority of sense of coherence items relative to the disaffected.

4.1. Covariate Effects

Table 1. Probabilities, means, and standard deviations for the covariates and distal outcomes by class.

	Disaffected (16.7%)	Moderately Connected (36.3%)	Strongly Connected (47%)
		P		M (SD) for Entire Sample	Range for Entire Sample
Covariate
Age	NS	NS	--	31.95 (10.79)	21.00–71.00
Gender	NS	−0.37	--
Marital Status	NS	0.42	--
Sexual Orientation	NS	NS	--
Race	NS	NS	--
COVID-19 Stress	−0.01	NS	--	46.01 (21.97)	0.00–100.00
Impression Management	−0.26	−0.22	--	2.96 (0.85)	1.00–5.00
Ideological Commitment	0.19	NS	--	3.70 (1.73)	1.00–7.00
Categorical Outcomes
Seminary Ecclesial Family
Catholic	0.07 a	0.11 ab	0.16 b
Evangelical	0.44 a	0.57 b	0.64 b
Mainline	0.47 a	0.27 b	0.17 c
Orthodox	0.02 a	0.05 a	0.03 a
Spiritual Formation Activity
In-class experiences	0.49 a	0.54 a	0.49 a
Informal friendship	0.72 a	0.56 b	0.44 c
Out-of-class assignments	0.67 a	0.52 b	0.47 b
Required small group	0.38 a	0.52 a	0.63 b
		M
Continuous Outcomes
Joy	4.80 a	5.04 a	5.45 b	5.19 (1.22)	1.00–7.00
Humility	4.10 a	4.13 a	4.29 b	4.20 (0.47)	1.00–5.00
Social Justice Activism	3.74 a	3.71 a	3.93 b	3.82 (0.86)	1.00–5.00
Calling–Presence	3.36 a	3.62 b	3.80 c	3.66 (1.01)	1.00–5.00
Interpersonal Differentiation	4.30 a	4.45 ab	4.58 b	4.49 (1.26)	1.00–6.00
Disconnection	2.60 a	2.38 b	2.21 c	2.34 (1.15)	1.00–6.00
Perceived Closeness to God	3.94 a	4.16 b	4.48 c	4.27 (0.96)	1.00–6.00
Attachment to God—Avoidance	2.53 a	2.24 b	1.99 c	2.17 (1.06)	1.00–7.00
R/S Exploration	4.84 a	4.89 a	5.14 b	5.00 (1.18)	1.00–7.00
R/S Struggles—God	2.12 a	2.02 ab	1.91 b	1.98 (0.92)	1.00–5.00
R/S Struggles—Interpersonal	2.79 a	2.62 a	2.37 b	2.53 (1.07)	1.00–5.00
Anxiety Symptoms	1.91 a	1.56 b	1.47 b	1.58 (1.57)	0.00–6.00
Depression Symptoms	1.56 a	1.48 a	1.18 b	1.35 (1.45)	0.00–6.00

Note: -- = Strongly connected as the reference subgroup. Gender coded 0 = male, 1 = female; marital status coded 0 = non-single, never married, 1 = single, never married; sexual orientation coded 0 = non-heterosexual, 1 = heterosexual; race coded 0 = non-White, 1 = White. P = probability/likelihood, M = mean, SD = standard deviation. Probabilities or means with different superscripts are significantly different between subgroups, with differences derived from pairwise comparisons on the outcome variables using the Model Constraint command, z-test, using the 2-step method. NS = nonsignificant. R/S = religious/spiritual. Ecclesial family = seminary denominational affiliation.

presents the results from the tests of covariate influence on class membership. Our inclusion of covariates followed the guidelines that they are theory-driven and have potential to clarify the meaning of class profiles (Masyn 2013). Prior research has shown that demographic variables delineating marginalized and non-marginalized identities, along with ideological commitment and socially desirable responding, influenced subgroups generated using R/S-related variables (e.g., Jankowski et al. 2022b). We also included COVID-19 stress as a covariate, given the larger social context within which data collection for the current study took place.

In the current study, females were less likely to belong to the moderately connected relative to the strongly connected, whereas the likelihood of belonging to the moderately connected was greater for those identifying as single—never married. As reports of COVID-19 stress decreased (i.e., higher ratings of a positive impact from the pandemic), participants were less likely to belong to the disaffected relative to the strongly connected. As ratings of an exaggerated sense of R/S increased (i.e., greater impression management), participants were less likely to belong to the disaffected and moderately connected relative to the strongly connected. As participants identified as more ideologically liberal, their likelihood of membership in the disaffected increased relative to the strongly connected, and while not reported in Table 1, this was also the case for the disaffected relative to the moderately connected (B = 0.12, SE = 0.06, p = 0.04).

4.2. Distal Outcomes

Table 1 also presents the results from the tests of differences by class on the distal outcome variables. Attending an evangelical seminary corresponded to a greater likelihood of belonging to the moderately or strongly connected, whereas attending a mainline seminary corresponded to a greater likelihood of belonging to the disaffected. Students at Catholic seminaries were also more likely to belong to the strongly connected. The disaffected were more likely to rate the R/S formation activities of informal friendships/community experiences and out-of-class assignments as more positive to their spiritual life relative to the moderately connected and strongly connected. Conversely, the strongly connected were more likely to rate the formation activity of required small group experience as more positive to their spiritual life, relative to the disaffected and the moderately connected. The strongly connected differed from both the disaffected and moderately connected by reporting higher levels of virtues, well-being, and healthy R/S outcomes. The strongly and moderately connected were comparable in divine struggles and anxiety symptoms, reporting lower levels relative to the disaffected. Relative to both the moderately and strongly connected, the disaffected reported lowest levels of calling–presence and perceived closeness to God, and highest levels of social disconnectedness, the latter of which offered construct validity evidence for the sense of community measure. The disaffected were comparable to the moderately connected on the virtues, interpersonal differentiation, R/S exploration, R/S struggles, and depression symptoms. Learning modality (i.e., online, in-person, hybrid) was not associated with class membership, as subgroups did not differ.

4.3. Predicting Flourishing

We then examined “a generalized form of LTA [latent transition analysis]” in which latent categorical variables representing different constructs are modeled (Muthén and Muthén 2000, p. 890). Specifically, we fixed the parameters of the separate mixture solutions and used the latent categorical variable for sense of community to predict the latent categorical variable for flourishing and conducted multinomial logistic regression (Asparouhov and Muthén 2021). The latent categorical variable for flourishing consisted of a single shared categorical variable for the trajectories (StatModel n.d.) involving subjective and eudaimonic well-being, comprised of joy (i.e., positive emotion, as an aspect of subjective well-being), calling–presence (i.e., life purpose specific to a sense of divine appointing to a particular vocation, as an aspect of eudaimonic well-being), and perceived closeness to God (i.e., R/S well-being, as an aspect of eudaimonic well-being). The shared categorical variable also involved trajectories for anxiety symptoms. Levels of flourishing can be delineated by variations on the distinct yet related dimensions of well-being and symptoms (e.g., Jankowski et al. 2022b).

A three-class model (BIC = 23,886.57; entropy = 0.78), in which the nonsignificant slope variances for calling, perceived closeness, and symptoms were set to zero to aid convergence, was determined as best fitting the data based on the plot of BIC values which showed a plateau after the 3rd class and the LMR-LRT which became nonsignificant with the 4th class (p = 0.20). We labeled class 1 in Figure 2 the symptomatic subgroup because of the increased anxiety symptoms reported over times 2–4, along with declines in joy and perceived closeness with God. We labeled class 2 in Figure 2 the growth subgroup because of a self-reported gain in calling–presence and a decline in anxiety symptoms. We labeled Class 3 in Figure 2 stable flourishing because of consistent levels of greater calling–presence, perceived closeness with God, and joy, and consistently low levels of anxiety.

We then examined the influence of the demographic covariates at time 2 on class membership. Females were more likely to belong to the symptomatic (B = 1.44, SE = 0.52, p = 0.01) relative to the stable flourishing subgroup. Identifying as White corresponded to a lower likelihood of belonging to the symptomatic (B = −0.98, SE = 0.42, p = 0.02) relative to the stable flourishing. Females were also more likely to belong to the growth subgroup (B = 0.68, SE = 0.34, p = 0.047) relative to the stable flourishing. Consistent with other findings (Crabtree et al. 2023; Jankowski et al. 2023), marginalized identification was a risk factor for greater symptoms and lower well-being.

The multinomial logistic regression predicting the latent categorical variable for flourishing from the latent categorical variable for sense of community showed that those belonging to the disaffected subgroup were more likely to belong to the symptomatic subgroup (B = 0.58, SE = 0.25, p = 0.02), relative to the stable flourishing subgroup. The probability of belonging to the symptomatic among the disaffected was 0.20. By comparison, the likelihood of belonging to the stable flourishing subgroup for the three sense of community subgroups was p = 0.71 for the disaffected, p = 0.79 for the moderately connected, and p = 0.78 for the strongly connected. These results converge with those of Coulombe and Krzesni (2019), who generated latent profiles using the indicators of sense of community, subjective and eudaimonic well-being, and mental health symptoms. They found two subgroups with a low sense of community, one labeled lower sense of community, languishing wellbeing, which displayed lower well-being and higher symptoms relative to the low sense of community, consistently moderate wellbeing subgroup. In addition, the subgroup reporting a mid-level sense of community reported symptom levels comparable to the lower sense of community, languishing wellbeing subgroup. They concluded that sense of community and well-being, and sense of community and symptoms “appear not to be linked for everyone” (p. 17). Taken together, disaffection exerted a modest risk effect in our sample for declines in subjective (i.e., positive emotion of joy) and eudaimonic well-being (i.e., perceived connection with the sacred) and increased anxiety symptoms.

5. Discussion

Within our sample of seminary students, we identified a small subgroup that reported a lower sense of community. This particular subgroup, which we labeled the disaffected, was more likely to report more liberal ideological commitment and more likely to attend mainline seminaries relative to both of the connected subgroups. These findings converge with those of Bulbulia et al. (2020), who found their disaffected subgroup to be more likely to belong to mainline denominations, and Holleman and Eagle (2023), who found that mainline clergy reported greater disconnection with their local R/S community. Bulbulia et al.’s (2020) subgroup of disaffected uniquely scored low on group narcissism, which refers to an exaggerated sense of the importance of one’s R/S group in making the world a better place, for example. Bulbulia et al. also characterized the disaffected as “inclusive religious types” (p. 11), with low scores on fundamentalism, and juxtaposed against “exclusive and highly committed religious people” (p. 2). In our sample, we found exclusive R/S types among the strongly connected, as this group was more likely to score high on impression management (i.e., an exaggerated sense of their relationship with the sacred) and scored highest on vertical dwelling in the form of greater perceived closeness to God (i.e., experiential aspect of R/S commitment/intrinsic religiousness; Pearce et al. 2013; Yonker et al. 2012), lower avoidant attachment relating to God, and a greater sense of “a transcendent summons” to a specific job (i.e., calling–presence; Dik et al. 2012, p. 244).

In contrast, the disaffected were unique, relative to both connected subgroups, in scoring highest on social disconnectedness (i.e., low belongingness) and avoidant attachment to God, lowest on perceived closeness to God and calling–presence, and they reported the highest level of anxiety symptoms. Also unique to the disaffected subgroup was their identification of informal friendships/community experiences and out-of-class learning opportunities as most influential to their personal formation. They were also less likely to show socially desirable responding. Disaffected subgroup membership was also a modest risk factor for declines in perceived closeness to God and positive affect and increased anxiety symptoms over time. These differences provide a context for understanding the disaffected, particularly when juxtaposed against the strongly connected.

To make sense of our findings, we took a lived religion approach (Ammerman 2013; Packard and Ferguson 2019) by drawing on prior scholarship about experiences of disaffection, which converge on unconventionality and individuality in their relating to the sacred. Shogren (2005) summarized the emergence of the ultra-charismatic, who were marginalized and turned inward and away from the establishment or church proper, feeling alienated from the establishment. Martí (2017), writing about the emerging church movement, noted that individuals “take ownership of their own religious convictions” (p. 7). Martí further described disaffected-driven R/S as a “challenge and confrontation against conventional Christianity … the continual practice of deconstruction” (p. 7). The disaffected also tend to value inclusiveness and plurality (Martí 2017). Packard and Ferguson (2019) described those who maintain an R/S identity and yet are “dechurched” and/or “living lives that are ‘decentered’ from congregations” (p. 501). These religious dones tend to relate to the sacred “through activism … and non-bureaucratized community” (p. 501). They value inclusion and pluralism, find conventional congregational life stifling for “authentic spiritual expression” (p. 511), and experience a sense of community in non-traditional, non-institutionalized spaces.

In our sample, marginalized identifications were not directly associated with sense of community subgroup membership, but like other experiences of disaffection, our disaffected subgroup reported valuing autonomy, as indicated by greater avoidant attachment to God (e.g., “I am uncomfortable allowing God to control every aspect of my life”) and greater disconnectedness (e.g., “In general, I don’t feel I participate with anyone or any group”). Our disaffected subgroup also identified as more ideologically liberal, which can connote inclusiveness and pluralism (Krull 2020). They also appeared more likely to push back on required small group experiences as personally transformative, which could be perceived as a challenge and confrontation against conventional Christianity. Davies (2020) described different small group types aimed at generating experiences of community, delineated a conventional type characterized by the “top down” formation of the small group (p. 139) and noted that non-conventional types involve “offering relationship and help on the other person’s terms” (p. 50). Davies suggested that organically formed small group communities may be more transformative.

Last, it is possible that disaffection represents a “Mainline Protestant distinctiveness” (Holleman and Eagle 2023, p. 599), given the greater likelihood for those attending mainline seminaries to be counted among the disaffected. However, students attending evangelical seminaries were comparably represented among the disaffected, even though they were more likely to belong to the moderately and strongly connected subgroups. So, while more students attending evangelical seminaries comprised the moderately and strongly connected, relative to the disaffected, their presence among the disaffected suggests that disaffection may not be limited to a mainline phenomenon. In addition, in the current study, disaffection was tied to COVID-19 stress, liberal ideological commitment, and autonomous relating to the sacred. While the latter two factors tend to be associated with mainline Protestantism, social justice activism also tends to characterize mainline Protestants (Jankowski et al. 2022b; Krull 2020) and this was not distinctive of the disaffected. There was also greater diversity of R/S identifications among the students attending mainline and evangelical seminaries relative to Orthodox and Catholic seminaries. In fact, beyond the affiliation percentages reported earlier, students at mainline seminaries were also more likely to self-identify as unaffiliated/none (7.3%) or historically Black protestant (8.6%), whereas at evangelical seminaries, only 1.8% identified as unaffiliated/none and 2.0% identified as historically Black protestant. Our use of seminary denominational affiliation as an external variable for comparison across classes rather than students’ self-reported affiliation limits being able to discern a mainline distinctiveness. Nevertheless, it made sense to use an institution-level variable for denominational affiliation, given that our measure of sense of community assessed the students’ experience of their seminary at a larger systems level, with the survey prompt “thinking about your seminary and seminary environment …”

5.1. Practical Implications

At the larger systems level, the disaffected serve an institutional change function (Bulbulia et al. 2020; Martí 2017). For Bulbulia et al. (2020), this social change function is a natural oscillating dynamic. As an institution becomes more rigid and conventional over time, there will be some who desire something different, an alternative R/S experience. For Martí (2017), social change is more a function of intentional individuals who “provoke change among the norm-constrained religious majority” (p. 9). Our findings suggest that the disaffected may represent a subgroup of seminary students especially engaged in the “move between institutionalized and interstitial” R/S (Manglos-Weber et al. 2024, p. 4), with interstitial referring to hybridity or pluralism that draws from a variety of R/S and non-R/S formation practices (e.g., physical exercise, yoga, nature hikes). Interstitial forms of R/S emerge out of the provocation directed at conventional R/S, in what Manglos-Weber et al. (2024) referred to as a “spirituality of deconstruction” (p. 1).

Our findings, therefore, suggest ways that seminaries can respond to the needs of the disaffected and promote personal transformation among the strongly connected. We focus on these two subgroups because the moderately connected appear to be similar to Keyes’ (2002) moderately healthy, who scored mid-range on well-being and psychosocial functioning, relative to symptomatic and low well-being subgroups, and a non-symptomatic high well-being subgroup (i.e., flourishing). In addition, the vast majority (82%) of Keyes’ moderately healthy described their subjective well-being as good or very good. The moderately connected in our study were mid-range on well-being indicators and mid-range on perceived closeness to God. They were less likely to engage in socially desirable responding, relative to the strongly connected, and were comparable to the strongly connected on low R/S struggles with God and low anxiety symptoms. At the same time, the moderately connected were comparable to the disaffected on their level of virtues, R/S exploration, interpersonal R/S struggles, and depression symptoms. The moderately connected also found informal friendships to be more personally transformative than the strongly connected and were more like the disaffected in reporting required small group experiences as less transformative. Taken together, the moderately connected appear to be “doing okay” (Keyes 2009, p. 1679), whereas the distinctives of the disaffected and strongly connected call for particular responses from the academic institutions, which we highlight below.

First though, our findings also draw attention to pre-pandemic calls for educational institutions to attend more closely to the mental health of their graduate students (e.g., Lipson et al. 2016). Relative to a national sample of religious leaders (Holleman and Eagle 2023), our sample reported greater depression symptoms. Holleman and Eagle (2023) noted that data collection for the 2019–2020 National Survey of Religious Leaders was interrupted due to the pandemic, so it seems likely that their data were predominantly pre-pandemic. Thus, the difference in depression symptoms between the two studies could be due to the fact that our data were collected post-pandemic declaration, and numerous studies have highlighted the risk effect of the pandemic on individuals’ mental health and well-being, particularly for those who report marginalized identifications (Crabtree et al. 2023; Jankowski et al. 2023). This risk effect showed up in the current study indirectly as greater COVID-19 stress predicted a greater likelihood of membership in the disaffected subgroup, and the disaffected were more likely to belong to the symptomatic subgroup (i.e., increased anxiety symptoms and declines in joy and perceived closeness with God over time). Female students and non-White students were more likely to belong to the symptomatic subgroup. However, it is also likely that the difference in depression symptoms between the two studies is due to the general prevalence of depression symptoms among graduate students, including seminary students (Smith et al. 2022). Financial stress may be a particular risk factor for graduate students’ poorer mental health (e.g., Blea et al. 2021), along with academic and interpersonal stressors and less healthy lifestyle beliefs and behaviors (e.g., Hoying et al. 2020; Jones-White et al. 2022).

The disaffected may benefit from environments structured to provide opportunities for informal friendships to evolve organically (e.g., welcoming spaces for informal congregating, non-R/S focused social activities, greater inclusive curricula, and school policies), thereby potentially promoting a sense of community. This seems particularly important among the disaffected, given that a greater sense of community has been shown to exert a protective influence against mental health symptoms (Gopalan et al. 2022). In addition, as part of the unique context for the current study, greater COVID-19 stress was a predictor of subgroup membership in the disaffected. This finding draws attention to the ways in which seminaries could tangibly offer support to students experiencing stressors of various kinds. By tangible, we mean directly assisting or referring students to internal and external community resources that offer help in specific life situations, such as assisting with transportation for someone diagnosed with a medical condition, providing child care or community meals (to help offset food insecurity) for single-parent students, or connecting students to employment opportunities amidst job loss/financial strain.

In addition, seminary contexts with higher levels of students fitting the disaffected subgroup face the challenge of trying to foster personal formation among students who want alternative R/S experiences rather than those consistent with more traditional R/S beliefs and practices. This could represent an opportunity for seminary leadership to draw upon inclusive psychosocially-informed approaches to personal formation, such as the RSM (Sandage et al. 2020), or what Ammerman (2013) and Wildman et al. (2024) described as a cultural packages approach to R/S. Cultural packages refer to diverse and complex socially constructed meanings for R/S. Some packages encourage varied creative practices that appeal to those who are more inclined toward broader experiences of transcendence and/or connection and/or promoting self- and other well-being through doing good rather than experiences and practices tied to conventional, institutionalized R/S. The latter refers to cultural packages that may be more theistic and/or about “belonging to and belief in the doctrines of their particular religious tradition” (Wildman et al. 2024, p. 50).

In the current study, the strongly connected seemed to benefit from conventional R/S strategies (i.e., required small group experiences), and in many ways appeared to be flourishing, and yet, their higher scores on impression management, relative to the other two subgroups, present a challenge to making sense of their overall profile. Greater impression management can connote “illusory spiritual health” (Hall and Edwards 2002, p. 342), and in the context of R/S—flourishing associations, greater “illusory mental health” (de Oliveira Maraldi 2020, p. 774). Hall and Edwards (2002) found that R/S-specific impression management correlated strongly with the vertical dwelling dimension of God-awareness. God awareness connotes perceived closeness to God. In our sample, perceived closeness to God was strongly correlated with impression management. As Sedikides and Gebauer (2021) noted, R/S individuals appear more likely to “self-enhance on domains central to their self-concept,” particularly so in seminary contexts that privilege R/S as a good (p. 29). The notion of impression management as self-enhancement intersects with the construct of group narcissism (Sedikides and Gebauer 2021), which, as we intimated above, appears to characterize the strongly connected more so than the disaffected and the strongly connected more so than the moderately connected. Prior research has shown protective effects for humility and differentiation against the negative influence of narcissism on seminary students’ well-being (Jankowski et al. 2022a) and evidence that a humility cultivation psychoeducational program promoted gains in humility and differentiation (Jankowski et al. 2022c). Despite their self-reported high level of humility, the strongly connected may benefit from intentional humility cultivation practices.

It seems likely that the higher scores on sense of community among the strongly connected represent a combination of person–environment fit (i.e., conventional R/S students in conventional R/S contexts) and illusory R/S health. The latter raises the question about how to promote a more authentic, integrated dwelling and seeking R/S orientation to protect against risks that might come later as a result of the stressors that come with professional R/S leadership roles. Based on the RSM, growth toward greater authenticity involves intensification (Sandage et al. 2020). While the disaffected may benefit from tangible expressions of containment, the strongly connected may benefit from having the “heat” turned up via “personal challenge, self-confrontation, and spiritual and existential seeking” (Sandage et al. 2020, p. 150), which may involve intentional reflection on R/S struggles. For the disaffected, engagement with contemplative R/S practices may be facilitative of growth, whereas the strongly connected could be encouraged to engage with difference by interacting with diverse others and/or engaging in practices outside of their conventional R/S. For example, a more conservative evangelical identifying student who typically engages in small group bible study might be encouraged to engage in social justice activism, meditative prayer, or liturgical forms of worship.

It is likely that the incorporation of interstitial R/S practices into the seminary curricular structure or programming may risk institutionalization (Manglos-Weber et al. 2024). Or, it may be inevitable for alternative R/S practices generated by the deconstructive process to not become conventional over time (Bulbulia et al. 2020). Nevertheless, it may be possible to live in a liminal space, in between institutional and interstitial R/S contexts and practices (Manglos-Weber et al. 2024; Packard and Ferguson 2019). A space where the strongly connected and disaffected could converge and maybe where the moderately connected already reside or could be encouraged to reside more fully.

5.2. Limitations and Future Directions

We used a diverse sample of adults attending Christian theological seminaries in North America. However, our sample does not represent all R/S identifications and may not generalize to other R/S and cultural contexts. Future research should examine sense of community in other cultural contexts and use a greater range of R/S identifications, including those who identify as atheist, agnostic, and non-religious (i.e., nones) and those who identify as spiritual but not religious. In the current study, only 3% identified as none, and a smaller percentage yet identified as affiliated with another religion. More broadly, there is a need to develop measures of R/S that apply more to seminary students fitting the disaffected, and other non-traditional ways of relating to the sacred. Such measures might show different associations with a sense of community and formation outcomes.

We also used a measure of sense of community that was constructed uniquely for our longitudinal survey design. We engaged in “validating a set of items” for our particular research context (VanderWeele et al. 2020, p. 4) and found evidence that the items were psychometrically sound. Nevertheless, additional construct validation of these items in other R/S and cultural contexts is needed. In general, we used a select set of items from a measure rather than the full-scale set of items. As VanderWeele et al. (2020) suggested, longitudinal survey designs often require careful consideration of the “constraints on the number of items that can be used … in specific contexts” (p. 2).

Last, the results from mixture modeling procedures should be interpreted in light of the indicators selected, psychometric evidence offered, and the particular analytic specifications used, and subgroups should not be reified (Masyn 2013). As Ram and Grimm (2009) stated, mixture modeling is an “exploratory … post-hoc analysis technique that seeks out the story the data are trying to tell—a story that is limited by the specific bounds imposed during model specification” (p. 572). As such, results are largely descriptive, and the meaning of a class is relative to the others in a specific research context.

6. Conclusions

We identified three subgroups of seminary students’ sense of community, which operationalized the RSM construct of horizontal dwelling. Consistent with the RSM, those who reported greater horizontal dwelling, labeled the moderately and strongly connected, reported lower anxiety symptoms and greater eudaimonic well-being. In contrast, the disaffected subgroup scored lowest on horizontal dwelling and reported greater COVID-19 stress, greater disconnectedness (i.e., lower belongingness), greater anxiety symptoms, lower calling, and lower R/S well-being. Belonging to the disaffected was also associated with a modest risk effect for greater symptoms and lower well-being over time.