Relational turbulence from the COVID-19 pandemic: Within-subjects mediation by romantic partner interdependence

Method

Data analysis

Paired samples t-tests were used to compare means of repeated measures of partner interdependence, negative emotions, and relational turbulence before versus during the COVID-19 pandemic (RQ1, RQ2, H1). Pearson correlations were used to test H2. To test H3, a within-subjects mediation model (Judd et al., 2001) was estimated using MEMORE (Mediation and Moderation Analysis for Repeated Measures Designs) version 3.4.1 (Montoya & Hayes, 2017) to test the indirect effect of the pandemic on relational turbulence through partner interdependence (facilitation and interference served as parallel mediators, controlling for each other). MEMORE constructs averages and differences in repeated measures and provides bootstrapping to generate 95% confidence intervals for indirect effects to test mediation as described by Judd et al. (2001) where the same participant is measured on both the mediator and outcome in two situations (in our case, repeated measures of interdependence and turbulence recalled before and during the pandemic). Thus, MEMORE allows us to determine if changes in our mediators (pre versus during pandemic facilitation and interference) are associated with changes in relational turbulence from before to during the pandemic. We used 5,000 bias-corrected bootstrap samples for indirect effects. To complement the conceptual diagram of the within-subjects parallel mediation model displayed in Figure 1, the statistical diagram for the model is offered in Figure 2.

OPEN IN VIEWER

Figure 2.

Statistical diagram of within-subjects parallel mediation model. Note. The superscript ^c indicates mean centered.

Results

Correlations between all study variables are provided in Table 1.

OPEN IN VIEWER

Table 1.

Zero-order correlations with confidence intervals (N = 315)

Variables	1	2	3	4	5	6
1. Partner Facilitation	—	−.271[−.396, −.144]	−.269[−.386, −.151]	−.212[−.322, −.098]	−.290[−.402, −.174]	−.380[−.485, −.274]
2. Partner Interference	−.475[−.581, −.362]	—	.473[.371, .565]	.495[.378, .598]	.444[.335, .542]	.438[.345, .529]
3. Anger	−.419[−.524, −.312]	.493[.383, .594]	—	.628[.543, .703]	.716[.640, .780]	.595[.498, .681]
4. Fear	−.310[−.432, −.190]	.428[.298, .543]	.686[.602, .753]	—	.724[.644, .791]	.494[.389, .590]
5. Sadness	−.392[−.496, −.282]	.471[.353, .576]	.800[.742, .847]	.829[.771, .877]	—	.633[.549, .707]
6. Relational Turbulence	−.548[−.635, −.455]	.509[.415, .598]	.617[.519, .708]	.510[.382, .631]	.630[.532, .719]	—

Note. Confidence intervals are in brackets using 5,000 bootstrap samples (bias-corrected accelerated). Correlations before COVID-19 are below the diagonal; correlations during COVID-19 are above the diagonal.

RQ1 inquired if romantic partner interdependence with regard to everyday routines changed after the COVID-19 pandemic. Results of paired samples t-tests revealed that both partner facilitation (t = 8.335, d = .470) and interference (t = 4.213, d = .379) decreased as partners became less interdependent. H1 predicted that relational turbulence would increase during the pandemic. A paired samples t-test supported this hypothesis; on average relational turbulence was greater from before to during the pandemic (t = −6.686, d = .377). H2a and H2b predicted that during the pandemic, interference would correlate positively with anger, fear, and sadness toward interacting with the partner, and facilitation would correlate inversely with these negative emotions. H2a was confirmed as interference was correlated positively with anger (r =.473 [CI: .371, .565]), fear (r = .495 [CI: .378, .598], and sadness (r = .444 [.335, .542]). H2b was confirmed as facilitation was correlated inversely with anger (r = −.269 [CI: −.386, −.151]), fear (r = −.212 [CI: .322, −.098], and sadness (r = −.290 [−.402, −.174]).

RQ2 inquired if negative emotions that accompany interacting with a romantic partner would be heightened during the peak of the first wave of the pandemic as compared to before the pandemic. Paired samples t-tests indicated an average increase in anger (t = −3.585, d = .201), fear (t = −3.132, d = .176), and sadness (t = −4.937, d = .278) when interacting with partners. Full statistical reporting of the repeated measures before and during the pandemic are reported in Table 2, along with effect sizes for increases and decreases among variables.

OPEN IN VIEWER

Table 2.

Paired samples differences.

Variable	Pre COVID-19		Post COVID-19		r	t(314)	dRM
	M	SD	M	SD
Partner Facilitation	4.592	.946	4.112	1.279	.614	8.335	.470
Partner Interference	2.401	1.101	2.159	1.145	.593	4.213	.239
Anger	2.356	1.100	2.545	1.308	.709	−3.585	.201
Fear	1.745	1.043	1.898	1.203	.710	−3.132	.176
Sadness	2.006	1.160	2.295	1.269	.638	−4.937	.278
Relational Turbulence	2.091	1.097	2.506	1.254	.569	−6.686	.377

Note. All paired samples correlations and t tests are significant at p < .001.

Hypothesis 3 was a within-subjects mediation test in line with the RTM/RTT, predicting that if the pandemic yielded changes in romantic partner interdependence, those changes (in facilitation and interference, independently) would in turn explain the hypothesized increases in relational turbulence predicted in H2. Results of a within-subjects mediation analysis with 5,000 bias-corrected bootstrap samples revealed evidence for mediation; the increase in relational turbulence was explained, in part, by a decrease in partner interdependence due to the pandemic (see Figure 2 for path coefficients). Compared to before the pandemic, partners decreased their interdependence during the pandemic, and by doing so, slightly reduced relational turbulence by interfering less with daily routines (a₁b₁ = −.036 [−.083, −.002]), but to a larger degree, also created more relational turbulence by not facilitating as much (a₂b₂ = .154 [.079, .244]). Moreover, controlling for interference and facilitation, the pandemic had a direct and positive effect on relational turbulence (c′ = .297 [.162, .432]). Figure 3 displays unstandardized path coefficients for the model.

OPEN IN VIEWER

Figure 3.

Within-subjects parallel mediation model (N = 314). Note. All path coefficients are significant at p < .05.

Discussion

In an effort to better understand how dating relationships are affected by the COVID-19 pandemic, this study provided tests consistent with the RTM/RTT and demonstrated that this pandemic affected college students’ interdependence in their romantic relationships, their experience of negative emotions in those relationships, and ultimately the stability of those relationships. Specifically, compared to how romantic relationships were recalled before the pandemic, during the pandemic, on average, interference and facilitation from a partner in everyday routines declined, negative emotions toward the partner were amplified, and relational turbulence became more prevalent. Collectively, these results confirm RTT’s claim that transitions create changes to relational environments that modify patterns of interdependence, which ultimately give rise to more chaotic relational states (Solomon et al., 2016).

The decrease in partner interdependence might be explained by college students’ restrictions for contact during the pandemic. That is, their partner’s ability to facilitate or interfere with daily routines becomes less influential if their overall contact and time spent together has decreased due to pandemic-related constraints. Put simply, romantic partners might see each other less, providing fewer opportunities to interfere or facilitate. A post-hoc analysis provided evidence for this explanation of decreased interdependence; on average, partners saw each other 4.400 times a week before the pandemic (SD = 2.199), which decreased to an average of 3.220 times a week during the pandemic (SD = 2.496); t(300) = 7.305, p < .001, d = .420. This limited contact explanation could explain the decreases in interdependence from before to during the pandemic. For instance, some students might have moved back home and now live with a parent or parents after college campuses discontinued in-person educational offerings and switched to online instruction (Sahu, 2020). Post-hoc descriptive statistics revealed additional evidence for this limited contact explanation as partners reported increases in geographical separation after the pandemic started. Prior to the pandemic, 50.2% of partners were geographically within 15 minutes of their partner (n = 158), whereas during the pandemic, only 28.3% of partners remained within a 15-minute travel proximity (n = 89). From these post-hoc contact explorations revealing decreased weekly contact with and increased geographical distance from partners, we believe that partner interdependence declined because of new living situations that were required as romantic partners moved from college campuses and were no longer enrolled in on-campus courses or experiencing an on-campus college life.

The within-subjects mediation model revealed parallel indirect effects for partner interdependence (both interference and facilitation) on relational turbulence, as well as a direct effect of the pandemic itself on relational turbulence. With regard to the former, increases in relational turbulence were explained in part by decreases in facilitation and interference. That is, slightly less relational turbulence was experienced to the extent that partners interfered less with daily routines during the pandemic, but at the same time, relational turbulence increased more to the extent that partners did not facilitate as much with daily routines either. These opposite indirect effects are in line with the wealth of research demonstrating that interference creates turbulence whereas facilitation diminishes it (e.g., McLaren et al., 2011, 2012; Solomon & Priem, 2016). However, a direct effect revealed that, controlling for changes in partner facilitation and interference, the pandemic itself was associated with a change in more relational turbulence above and beyond partners’ decreases in interdependence. Pandemic stressors, independent of partner interdependence, appear to disrupt the stability of romantic relationships. Although relational research on the pandemic is limited as of this writing, preliminary findings suggest that stressors from the pandemic, including social isolation, financial strain, and perceived stress, are associated with lower romantic relationship quality (Balzarini et al., 2020).

During the pandemic, college students reported an increase in negative emotions toward communicating with their romantic partners including more anger, fear, and sadness. This increase in negative emotions might be indicative of partners’ daily emotional welfare being compromised as a byproduct of the pandemic. Lades et al. (2020) found when individuals stayed home during the pandemic and did not pursue outside activities, they experienced more negative affect. However, partner interference was strongly associated with these negative emotions, which is in line with previous research in other transitional contexts (Brisini & Solomon, 2019; Knobloch et al., 2007; Knobloch & Theiss, 2010; Solomon & Brisini, 2019). From a discrete emotions perspective, such negative affect may result because partner interference behaviors are perceived as threatening, uncontrollable attempts to impede one’s goals (Nabi, 2002). Axiom 2 of RTT received support as individuals felt more anger, fear, and sadness when interacting with their partner when routines were interrupted by them during the pandemic.

The findings from this investigation extend the purview of RTT to the novel context of a global health emergency and demonstrate support for the theory’s utility in the specific context of a pandemic. This research furthers a line of relatively recent scholarship which applies and tests the mechanisms of RTT when romantic partners are experiencing situations that are out of the ordinary, further illustrating RTT’s efficacy in explaining both normative and nonnormative relational experiences (e.g., Tian & Solomon, 2020). With regard to the latter, some findings emerged here which appear to be unique to the particular context under study. For example, although transitions typically evoke increased interference and attendant negative emotions from a partner (see Solomon et al., 2010), here we found that both forms of partner interdependence decreased from before to during the pandemic. It seems clear (and others have concluded) that interference is complex and unique to the transition in question (Harvey-Knowles & Faw, 2016), as some transitions naturally lend themselves to more—or in this case, less—partner interdependence, given the stringent social distancing guidelines that have characterized the pandemic in the U.S. Although partner interference is a negatively valenced construct in RTT, and thereby the reduction of interference should lead to concomitant reductions in the experience of relational turbulence, here that effect appears to have been offset by the concurrent experience of decreased partner facilitation (again, likely due to the inherent social distancing constraints of the pandemic for the college student daters in this sample, most of whom were geographically distant). In this investigation, we observed an increase in relational turbulence, which was indirectly affected by decreased partner facilitation and directly exacerbated by the nature of the pandemic itself.

Further, the findings from this study highlight the complex role of interdependence processes within RTT, and provide some evidence for the ways in which partner interference and facilitation work together (or in this case, work against each other) to impact the experience of relational turbulence in college student dating relationships. From RTT’s perspective, the least amount of turbulence should result when partner facilitation is high while partner interference is low. Here, partner interference decreased, yet so did partner facilitation. Thus, although RTT suggests that partner facilitation may at times serve a sort of buffering function against interference or the effect of transitions, this buffering effect could not be realized (despite the decrease in partner interference) because the pandemic constrained the ability to facilitate (and to interfere with) a partner’s daily routines. Perhaps because of this, the findings from this study suggested that the pandemic itself as compared to interdependence processes had the strongest impact on the experience of relational turbulence. It is important to note that the nuances of these findings regarding interdependence processes are expected to be context-dependent, as the majority of college student daters who comprised this sample were geographically distant during the pandemic. It is likely that a different pattern of results would emerge for married or cohabitating couples with regard to interdependence processes, for whom it is unlikely that both partner facilitation and interference would decrease during this extended time of social distancing, working remotely, homeschooling children, etc. For married and cohabitating couples, interdependence processes are expected to play a stronger role in the experience of relational turbulence during a pandemic (Knoster et al., 2020).

The collective results reveal several practical implications for romantic relationships of all types, including marital relationships, during the pandemic. First, it is important to recognize that transitions are times rife with relational turbulence for couples as changes to external environments produce chaotic relational states (Solomon et al., 2010). Life, in general, has been chaotic since the pandemic began and has created stressors external to interdependent relationships that carry over into it and naturally affect its quality (Balzarini et al., 2020). It is important for partners to recognize that their romantic relationships may be strained from the pandemic, and they might expect negative affect and tumultuous sensitivities for their relationships as these are processes theorized to result from the experience of transitions (Solomon & Knobloch, 2004). Second, in order to more successfully navigate the pandemic, and to navigate extreme or nonnormative relational episodes more generally, it is important for relational partners to adapt to the “new normal” by establishing neoteric patterns of interdependence that encourage relational stability. This could include establishing new daily routines and/or maintain existing ones for psychosocial benefits (WHO, 2020). These daily routines are not all instrumental or occupational; they include hobbies as well, which have been identified as particularly important for maintaining emotional well-being during the pandemic (Lades et al., 2020).

Third, in order to more effectively manage intense relational experiences, romantic partners should put in extra efforts to facilitate their partners’ daily routines, and actively try not to interfere with them. It is possible that only so much facilitation can be reasonably and safely enacted during the pandemic (and further, some research with the RTT in the context of intense relational episodes suggests that facilitation does not always lead to positive outcomes; Tian & Solomon, 2020), so college student partners in particular—as opposed to spouses, for example—might acknowledge they have less overall influence on their partners’ daily routines since the pandemic commenced. Fourth, romantic partners should realize that their reactivity during the pandemic and other extreme relational experiences is likely to be exacerbated, and this reactivity may manifest in the form of more extreme or perhaps even volatile emotions, cognitions, and communicative behaviors (Solomon et al., 2010, 2016). Such extreme reactions have implications for a variety of relational processes, including seeking/providing social support and engaging in conflict. As such, romantic partners should be cognizant of the far-reaching impact of intense experiences such as the current global health pandemic on everyday functioning in the relationship. Being aware of the potential for such reactivity may encourage partners to pause and reassess rather than overreact and engage in communication or other behaviors that could possibly damage the relationship.

Although important insights about the pandemic’s impact on the stability of romantic relationships emerged from this investigation, the findings must be interpreted in light of the limitations that were present. The primary limitation of this research was the reliance on recollections of the romantic relationship pre-pandemic (January, 2020) with a data collection that took place during the pandemic (April, 2020). Partners reported on what their relationship was like 4 months prior and then reported on how their relationship was currently during the peak of the pandemic. This method of reporting on the relationship during a previous time and at the time of the survey has been used by previous relational turbulence scholars (Brisini et al., 2018). However, asking participants to report on “then and now” repeated measurements in the same survey presents recall limitations for modeling within-participant “change.” Yet, to design a study with two time points, we would have needed prior knowledge that a global pandemic was imminent to have collected data before it began. Thus, our repeated measurements in the same survey are a proxy for change, but cannot actually measure true changes over time, and there is a chance that recall bias was an issue (i.e., a particular fondness for “before times”). Another limitation is the college student sample which might have derived different effects due to physical separation from college campuses. As such, future researchers should examine pandemic effects in marriages and test RTT within cohabitating contexts, which might offer different conclusions from more established patterns of interdependence in shared living arrangements.

Future researchers should also continue to study major transitions as opportunities to model relational turbulence. Although transitions are not a scope condition for testing RTT because partner influence can and does occur at any point in a close relationship (Berscheid, 2002), transitions are periods of discontinuity where interdependence will change as the relational environment is affected (Solomon et al., 2016). Future researchers might also examine how the COVID-19 pandemic has created relational uncertainty, and in turn, resulted in biased cognitions as purported by RTT. To keep our survey brief with repeated measures, we only examined half of the relationship parameters in RTT. Relational uncertainty is at the core of RTT and deserves empirical attention as the pandemic continues. Finally, scholars should examine processes of relational turbulence in both dating and married samples to compare effects for generalizability (Brisini & Solomon, 2019). Although these two types of relationships have produced similar effect sizes in the relational turbulence literature (Goodboy et al., 2020), nonetheless, it remains important to continue studying both types of relationships.

This study explored changes in some of the relational processes proposed by RTT that were experienced by dating partners before the COVID-19 pandemic began to the peak of the first wave of the pandemic in the U.S. The findings revealed pandemic-related relational impacts in the form of decreased partner interdependence, increased experience of negative affect, and heightened relational turbulence (explained both by decreased partner interdependence and by the impact of the pandemic itself). These results provide continued support for RTT’s predictive and explanatory utility, and importantly, suggest practical mitigation strategies for couples who are coping with the ongoing global health crisis. This work provides support for Solomon and Brisini’s (2019) assertion that “RTT may have the greatest value when it illuminates the challenges that confront couples coping with significant life transitions, especially those that impose economic, health, or emotional burdens” (p. 2432).