Relationship satisfaction during COVID-19: The role of partners' perceived support and attachment.

Objective: The main goal of this study was to examine the interplay between individuals' attachment insecurity and their perceptions of their partners' COVID-related behaviors (supportive and negative behaviors) in predicting their relationship satisfaction. Background: Stress is a well-documented risk factor for relationship satisfaction. COVID-19 related stressors thus pose a challenge to maintaining relationship satisfaction. Although partners' supportive behaviors can play a central role in mitigating these stressors, enduring individual vulnerabilities, such as attachment insecurity, are likely to moderate the effectiveness of supportive (or negative) behaviors. Method: In this two-wave study, conducted at the start of the COVID-19 pandemic in Israel, 239 participants in cohabiting Israeli couples reported their current relationship satisfaction and perceived partners' supportive and negative behaviors in response to COVID-related stress. Participants' pre-COVID reports of relationship satisfaction and attachment orientations were used to assess the extent to which partners' supportive/negative behaviors interacted with attachment orientations to predict relationship satisfaction maintenance during the first lockdown in Israel.
Results: Higher levels of support and lower levels of negative behaviors were associated with greater relationship satisfaction maintenance. Anxiously attached individuals showed greater sensitivity to their partners' support, whereas avoidantly attached individuals manifested lower reactivity to their partners' negative behaviors. Conclusions: Perceived partners' supportive and negative behaviors can predict relationship satisfaction during stressful times. However, high attachment anxiety and low attachment avoidance may render individuals more sensitive to such behaviors. Implications: The results suggest that during times of stress, it is essential to target partners with attachment insecurity to strengthen their supportive skills.

The COVID‐19 pandemic constitutes one of the most serious global health crises the world has witnessed in a century (World Health Organization, 2020). There is now ample evidence that the stress caused by the pandemic is exerting pernicious effects on many people's well‐being (e.g., Brooks et al., 2020; Torales et al., 2020). Pandemic‐related stressors have been shown to lead to psychological distress (e.g., Brown et al., 2020; Luetke et al., 2020; Pietromonaco & Overall, 2020; Prime et al., 2020). Such negative effects of the pandemic on people's well‐being have been documented in Israel as well. For example, two studies conducted on the Israeli population reported elevated levels of perceived stress and COVID‐related economic and health anxiety (Bareket‐Bojmel et al., 2021; Horesh et al., 2020).

Notably, the pandemic's adverse effects have also been found to spill over to affect people's close relationships (Brown et al., 2020; Pietromonaco & Overall, 2020; Prime et al., 2020). The measures put in place to combat the pandemic have resulted in massive disruptions to couples' daily routines. For example, stay‐at‐home policies require couples to cope with drastic changes dyadically, such as working from home and sharing responsibility for (home) childcare (Carlson et al., 2020). These challenging circumstances are likely to deplete partners' resources, which may impede them from satisfactorily addressing each other's relational needs. Thus, the external stressors related to the COVID‐19 pandemic may impact partners' ability to maintain a high level of relationship satisfaction (Pietromonaco & Overall, 2020).

The current study implemented the vulnerability‐stress‐adaptation model (Karney & Bradbury, 1995) to explore the interplay between partners' relational behaviors at the start of the pandemic in Israel and their pre‐pandemic dispositional vulnerabilities. Specifically, it focused on partners' attachment insecurity, a key vulnerability factor, in shaping partners' relationship satisfaction during the pandemic (Pietromonaco & Overall, 2020).

A satisfying romantic relationship is strongly linked to individuals' physical health (Robles et al., 2014) and mental well‐being (Braithwaite & Holt‐Lunstad, 2017; Proulx et al., 2007; Whisman & Baucom, 2012). There is consistent evidence that lower levels of relationship satisfaction and higher levels of relational distress are causal risk factors for depression (for a recent review, see Whisman et al., 2021). Relatedly, partners in highly satisfying relationships tend to live longer and have fewer health problems than unsatisfied partners or single individuals (Kiecolt‐Glaser & Newton, 2001; Robles et al., 2014). The impact of people's relationships on their long‐term health is comparable to (or greater than) that of well‐known health‐risk factors such as obesity (Holt‐Lunstad et al., 2010). This points to the need to better identify the relational and personal factors shaping partners' relationship satisfaction during times of stress.

Relational behaviors (whether supportive or nonsupportive) play a central role in mitigating (or exacerbating) the external effects of stressors on couples' personal and relational well‐being (Bodenmann, 2005; Conger et al., 1999; Karney & Bradbury, 1995). In one study, where couples were asked to disclose their stressful experiences, higher attentive support from the listening partner was associated with higher self‐reported satisfaction by both partners (Kuhn et al., 2018). In the context of COVID‐19, a recent daily diary study on a sample of Israeli couples found that although the disclosure of one's COVID‐related stress had no effect on daily levels of relationship satisfaction, perceived partner's constructive and destructive responses to these disclosures were significantly associated with better daily relational outcomes (Bar‐Kalifa et al., 2021). Given the stress and disruption caused by the pandemic, it is plausible that partners' needs for support have grown while, at the same time, their capacity to meet each other's needs has diminished.

Furthermore, people's personality dispositions are likely to affect relational well‐being (Kashdan et al., 2018). The vulnerability‐stress‐adaptation model (Karney & Bradbury, 1995) posits that romantic partners' relational functioning in times of stress is shaped by their dispositional vulnerabilities (e.g., difficulties in regulating distressing emotions). Support for this model comes, for example, from a longitudinal study following newlywed couples over 3 years. Couples' vulnerability traits (e.g., impulsivity) strengthened the association between external stressors (e.g., financial strain, health concerns) and physical aggression (Langer et al., 2008). In another longitudinal study on a cohort of African American romantic partners, chronic financial strain predicted lower enacted and perceived dyadic support (Clavél et al., 2017).

Pietromonaco and Overall (2020) recently adopted the vulnerability‐stress‐adaptation model as a conceptual framework to probe the possible consequences of COVID‐19 related stressors on couples' relationships. They suggested that the external stress resulting from the pandemic (e.g., economic hardship) would increase maladaptive relational behaviors (e.g., hostility), particularly for couples dealing with contextual (e.g., minority status) and personal (e.g., depression) vulnerabilities. The current study focused on attachment insecurity, an individual vulnerability factor that is likely to shape partners' relational experience during the pandemic (Pietromonaco & Overall, 2020; Steele, 2020).

Attachment is a biologically mediated inborn system that directs individuals to seek proximity with significant others (i.e., attachment figures) to receive comfort and security when feeling distressed (Bowlby,1969, 1973, 1982; Mikulincer & Shaver, 2005). Over the course of development, children develop enduring expectations (i.e., working models) as to attachment figures' responsivity in times of need. When entering adulthood, the romantic partner serves as one of the most significant attachment figures for many people (Hazan & Shaver, 1987; Mikulincer & Shaver, 2007).

Adults' attachment style can be conceptualized on two axes of attachment, termed anxiety and avoidance (Hazan & Shaver, 1994). When faced with stress, individuals high on attachment anxiety tend to adopt hyperactivating attachment strategies that are manifested in excessive proximity‐seeking behaviors and high reactivity to their partners' relational behaviors. By contrast, avoidantly attached people tend to adopt deactivating attachment strategies that are often manifested in withdrawal behaviors and insensitivity to their partners' relational behaviors (Mikulincer & Shaver, 2007; Simpson & Rholes, 2017). For example, Kordahji et al.’s (2015) study examined the moderating effect of partners' attachment anxiety on the association between emotional support and partners' reduction in cardiovascular arousal during laboratory dyadic supportive interactions. This study found that attachment anxiety was associated with greater arousal reduction when partner's emotional support was provided. In contrast, attachment avoidance was associated with lower arousal reduction in response to such support. These findings are in line with the idea that anxiously attached individuals benefit more, and are reactive to a greater extent, to their partner's supportive behaviors than avoidantly attached individuals.

The idea that pre‐pandemic attachment vulnerability could predict changes in partners' relationship functioning during the pandemic has recently been examined (Overall et al., 2020). Using a sample of 234 romantic partners who reported their attachment style before the pandemic outbreak in New Zealand, the authors found that attachment anxiety predicted greater relationship problems and lower relationship quality; attachment avoidance predicted lower family cohesion and problem‐solving efficacy.

The current study builds on these findings but goes further to examine whether attachment style renders individuals more or less sensitive to their partners' behaviors. In other words, it examines attachment style as a moderator of partners' reactivity to partners' perceived support and negative behaviors. Specifically, based on a sample of 239 Israeli participants, we examined whether pre‐COVID attachment style would interact with perceived supportive and negative behaviors in predicting relationship satisfaction during the first lockdown in April 2020.

Specifically, it was predicted that higher levels of perceived COVID‐related supportive behaviors (e.g., empathy, tangible support) and lower levels of perceived negative behaviors (e.g., criticism, withdrawal) would be associated with the maintenance of relationship satisfaction (Hypothesis 1). Based on findings showing higher dependence on partners' support in anxiously attached individuals (Campbell et al., 2005; Overall & Sibley, 2009), we further predicted that attachment anxiety would strengthen the association between perceived COVID‐19 supportive/negative behaviors and relationship satisfaction maintenance (Hypothesis 2). In contrast, based on findings showing less dependence on partners' support among avoidantly attached individuals (Collins & Feeney, 2000; Overall & Sibley, 2009), we predicted that attachment avoidance would weaken the association between perceived COVID‐19 supportive/negative behaviors and relationship satisfaction maintenance (Hypothesis 3).

METHOD

Participants

Two hundred thirty‐nine cohabiting Israeli couples who had participated in studies conducted in our lab in 2018–2019 (prior to the pandemic) were approached on April 12, 2020, when the Israeli population was under the first lockdown. Two of the original studies were daily diaries studies, and two were laboratory‐based observation studies in which couples participated in dyadic interactions. The measures used in the current study were taken from the background questionnaire administered prior to participation in the diary or the lab session. Participants in these studies were included if they were adult couples (i.e., both partners >18 years of age) who had lived together and been in a relationship for at least 6 months. A total of 240 individuals who were still living with the same partner in April 2020 completed the same set of questionnaires within a period of 2 weeks. Following Barton et al. (2020), who highlighted the importance of including individuals whose partners were unwilling to participate in dyadic studies, we included data from any individual who was still in the same relationship, even if his or her partner did not complete the follow‐up assessment. Completers and noncompleters did not differ in their attachment or relationship satisfaction scores as measured in the original studies (p > .23). One participant was excluded due to an unusual pattern in the man's attachment scores ; therefore, the final sample was composed of 239 individuals: 79 couples (80 women, 78 men) and 81 individuals (57 women, 24 men). One couple self‐identified as lesbian, one individual self‐identified as a gay man, and the rest self‐identified as heterosexual.

On average, current reports (i.e., during COVID‐19 lockdown) were filled out for 403.26 days (range: 41–734 days, SD = 221.14 days) after the initial reports were collected (i.e., in the original studies). Participants' mean age was 30.18 years for men (range: 20.3–70 years, SD = 8.33 years) and 28.85 years for women (range: 20.3–65 years, SD = 7.66 years). In relation to education, 105 participants (43.93%) had a high school education, and 133 (55.64%) reported having a higher education. Education data was missing for one participant. By the time of the study, the average duration of the relationships was 7.06 years (range: 1.17–39.17 years, SD = 6.62 years), and the average duration of cohabitation was 4.96 years (range: 10 months to 39 years, SD = 6.68 years). One hundred four participants (43.51%) were married, and 35 participants (14.64%) had at least one child.

Procedure

The data were collected between April 12 and April 27, 2020, during which Israel was under the first imposed lockdown. Home confinement regulations included a 100‐meter (328 foot) limit on venturing from home for nonessential activities, travel limitations for nonessential activities (including work), and a prohibition on meeting people (including family) from another household. Couples were approached via a personal e‐mail to participate in a follow‐up survey focusing on personal and relational coping with the current pandemic. Upon completing the study, participants were entered into a raffle, where 10 participants won and received a prize of 250 Israeli Shekels (i.e., approximately US$70). The University Institutional Review Board approved the study.

Measures

The study was administered in Hebrew, and all instruments were translated and back‐translated by native English speakers to ensure consistency with the English versions.

Attachment orientation

Partners' attachment orientation was measured in the original studies (i.e., prior to COVID‐19 outbreak) using the Experiences in Close Relationships—Relationship Structures Questionnaire (ECR‐RS; Fraley et al., 2011). On the ECR‐RS, participants rate six items assessing attachment avoidance (e.g., “I prefer not to show my partner how I feel deep down”) and three items assessing attachment anxiety (e.g., “I often worry that my partner doesn't really care for me”). Items are rated on a 7‐point Likert scale (1 = strongly disagree, 7 = strongly agree). The internal reliability of the attachment avoidance scale in our sample was .81 and .64 for men and women, respectively. The internal reliability of the attachment anxiety scale in our sample was .78 and .74 for men and women, respectively.

COVID‐related support and negative behaviors

Partners' supportive and negative behaviors in response to one's COVID‐related concerns were assessed on five items. The supportive behavior score was computed as the average of three items: “My partner is expressing understanding, empathy, and caring”; “My partner is helping me to see the stressful situation in a different light”; and “My partner is doing practical things to alleviate my stress.” The internal reliabilities in the current sample were .77 for men and .74 for women. The negative behavior score was computed as the average of two items: “My partner blames me or criticizes the way I am dealing with my stress,” and “My partner is not taking my feelings seriously or ignores them.” The internal reliabilities in the current sample were .75 for men and .79 for women. All items were rated on a 5‐point Likert scale (0 = very rarely, 4 = very often).

Relationship satisfaction

Most participants' (83%) baseline (pre‐COVID) relationship satisfaction levels were assessed on the four‐item version of the Couples Satisfaction Index (CSI‐4; Funk & Rogge, 2007) collected in the original study. For the remainder of the participants, the pre‐pandemic CSI‐4 was unavailable (i.e., it was not collected in the original study) and, therefore, could not be used as the baseline pre‐pandemic relationship satisfaction covariate. Instead, their pre‐pandemic relationship satisfaction levels were assessed using the satisfaction subscale of the Perceived Relationship Quality Components (PRQC; Fletcher et al., 2000). To construct a comparable measure of baseline pre‐pandemic relationship satisfaction, we computed the standardized relationship satisfaction scores for each participant using the distributions of the original samples employed to develop the PRQC‐satisfaction and CSI‐4 scales (i.e., Fletcher et al., 2000; Funk & Rogge, 2007). That is, the standardized z score for participants in our study was calculated using the means and standard deviations from the original validation publications. During the lockdown, relationship satisfaction was measured on the four‐item version of the CSI‐4 (Funk & Rogge, 2007). The first item of the CSI‐4 is rated on a 6‐point Likert scale, and the rest on a 5‐point Likert scale. The CSI‐4 pre‐COVID internal reliabilities in our sample were .91 for men and .90 for women. The CSI‐4 during‐COVID internal reliabilities were .90 for men and .92 for women. The PRQC‐satisfaction scale internal reliabilities were .89 for men and .94 for women.

Analytic approach

Relationship satisfaction during the lockdown was predicted by the perceived partners' behaviors (supportive/negative), their interactions with attachment anxiety and avoidance, and the participants' pre‐COVID relationship satisfaction. Note that some of the participants in the current study were coupled, therefore introducing interdependence in their relationship satisfaction scores (and hence in the model's residuals). Thus, to analyze our data, we implemented the generalized least squares regression model (implemented with the R nlme package using the GLS function; Pinheiro et al., 2021) which allowed us to model the covariation in partners' residuals. A set of additional covariates were included in the model. First, Barton et al. (2020) showed that coupled participants who participate in a study without their partner tend to report lower satisfaction levels. Therefore, we accounted for participation status by included a binary control variable indicating whether the participant's partner's data were available. In addition, the following covariates were included: gender, time elapsed since the baseline assessment, relationship duration, marital status, and the number of children. Simple slopes were computed using the emmeans R package (Length, 2020).

Power analysis

We used the Monte Carlo simulation method advocated by Lane and Hennes (2018) to calculate power in the context of models with dyads. Estimates for effect sizes were obtained from previous research (for more details, see online materials at https://osf.io/83fb6). The effect sizes were estimated to be within the medium range. The simulation indicated that the data were sufficiently powered to examine our hypotheses (.89–.99).

RESULTS

The descriptive statistics are presented in Table 1. As shown, attachment anxiety and avoidance were positively associated with each other. Attachment avoidance and anxiety were associated with perceiving less support and more negative behavior in response to COVID‐related concerns. Both attachment orientations were associated with lower relationship satisfaction. Partners' supportive and negative behaviors were associated with partners' relationship satisfaction during COVID lockdown in the expected directions. Specifically, a paired t test showed that participants' relationship satisfaction did not change significantly from the baseline (pre‐COVID) to the follow‐up (during COVID lockdown) assessments (t = 1.801, df = 136, p = .074, Cohen's d = 0.153 for women; t = 0.561, df = 101, p = .575, Cohen's d = 0.055 for men). In addition, a set of generalized least squares models indicated that female participants reported higher levels of perceived partner support behaviors and lower levels of attachment avoidance.

TABLE 1

Descriptive statistics

		Men			Women			Gender difference estimate (SE)	Zero‐order correlations
		N	M (SD)	Range	N	M (SD)	Range		1	2	3	4	5	6	7
1	Attachment avoidance	102	1.93 (0.82)	1–4.67	137	1.61 (0.57)	1–3.5	−0.31 (0.08) ***	.26 *	.50 ***	−.32 **	.27 **	−.39 ***	−.56 ***	−.71 ***
2	Attachment anxiety	102	1.70 (0.96)	1–5	137	1.92 (1.11)	1–6	0.21 (0.13)	.25 **	.05	−.25*	.17	−.52 ***	−.55 ***	−.42*
3	Perceived support behavior	102	3.73 (0.88)	1–5	137	4.01 (0.79)	1.67–5	0.28 (0.10) **	−.34 ***	−.21*	.17	−.11	.42 ***	.31 **	.29
4	Perceived negative behavior	102	1.51 (0.81)	1–5	137	1.62 (0.77)	1–4	0.10 (0.09)	.20*	.30 ***	−.30 ***	.18	−.29 **	−.08	−.67 ***
5	RS during lockdown (CSI‐4)	102	17.63 (3.21)	7–21	137	17.20 (3.43)	6–21	−0.07 (0.32)	−.47 **	−.31 **	.45 ***	−.37 ***	.58 ***	.62 ***	.61 ***
6	RS pre‐COVID (CSI‐4)	75	17.40 (3.00)	9–21	98	17.39 (2.95)	8–21	0.23 (0.33)	−.55 ***	−.34 ***	.33 ***	−.07	.70 ***	.58 ***
7	RS pre‐COVID (PRQC‐satisfaction)	27	6.36 (0.72)	5–7	39	6.21 (1.08)	2–7	−0.19 (0.14)	−.62 ***	−.21	.10	−.30	.68 ***		.73 ***

Note. CSI‐4 = Couples Satisfaction Index; PRQC = Perceived Relationship Quality Components; RS = relationship satisfaction. Women and men's differences were computed using generalized least squares modeling, where each variable was predicted by gender (0 = men, 1 = women) and partner's residuals were allowed to correlate; partners' correlations are presented on the bolded correlation matrix diagonal (for the N = 79 couples for which data were available from both partners), and men's/women's correlations are presented above/below the diagonal, respectively.

p < .05.

p < .01.

p < .001.

Hypotheses testing

Table 2 presents the standardized estimated fixed effects of this model, which were obtained by standardizing the raw variables of the entire data set and re‐estimating the model (see Baldwin et al., 2014). As predicted, supportive behaviors were positively associated with relationship satisfaction, whereas negative behaviors were negatively associated with relationship satisfaction. A significant Support × Attachment Anxiety emerged. As depicted in Figure 1, anxiously attached individuals showed greater sensitivity to their partners' supportive behaviors. Specifically, supportive behaviors were more strongly associated with relationship satisfaction for individuals high on attachment anxiety (at +1 SD β = .36, SE = .07, p < .0001), than for individuals with average attachment anxiety (β = .24, SE = .05, p < .0001(. The association was no longer significant for individuals low on attachment anxiety (at −1 SD β = .10, SE = .06, p = .08).

TABLE 2

Results of the GLS model predicting relationship satisfaction during lockdown

Predictor	Standardized estimate (SE)	p value
Support behaviors	.24 (.05)	<.001
Negative behaviors	−.20 (.05)	<.001
Attachment avoidance	.07 (.06)	.238
Attachment anxiety	−.08 (.05)	.109
Support × Attachment Avoidance	.03 (.04)	.430
Support × Attachment Anxiety	.13 (.04)	.003
Negative Behaviors × Attachment Avoidance	.10 (.05)	.033
Negative Behaviors × Attachment Anxiety	.05 (.04)	.182
Pre‐COVID RS	.52 (.06)	<.001
Partners' participation (0 = no, 1 = yes)	−.00 (.09)	.998
Time elapsed since first assessment	−.01 (.05)	.896
Relationship duration	−.16 (.07)	.034
Marriage (0 = no, 1 = yes)	.07 (.10)	.500
Kids number	−.01 (.07)	.905
Gender (−0.5 = male, 0.5 = female)	−.09 (.09)	.322

Note. GLS = generalized least squares; RS = relationship satisfaction.

FIGURE 1

Perceived Support Behavior × Attachment Anxiety effect on relationship satisfaction

In addition, a significant Negative Behavior × Attachment Avoidance interaction emerged. As depicted in Figure 2, avoidantly attached individuals were less reactive to their.

FIGURE 2

Perceived Negative Behavior × Attachment Avoidance effect on relationship satisfaction

partners' negative behaviors. Specifically, negative behaviors were not associated with relationship satisfaction for individuals high on attachment avoidance (at +1 SD β = −.08, SE = .05, p = .155), whereas this association was significant for participants with average (β = −.20, SE = .04, p < .0001) and low attachment avoidance (at −1 SD β = −.32, SE = .07, p < .0001).

DISCUSSION

The COVID‐19 pandemic, the most devastating global health crisis of the current century (World Health Organization, 2020), has brought about immense disruption to individuals from all walks of life. Unsurprisingly, the pandemic and its related stressors (e.g., health concerns, economic strains, lack of childcare) have been found to exacerbate psychological distress (e.g., Brooks et al., 2020; Torales et al., 2020). This universal situation provides a unique, albeit unfortunate, opportunity to test some of the most fundamental tenets of relationship science in an ecologically valid context. The current study drew on the vulnerability‐stress‐adaptation model (Karney & Bradbury, 1995), which posits that the ultimate effect of external stress on intimate relationships is largely determined by the intersection between partners' vulnerability and the quality of their support. Based on this line of thought, this study examined the interplay between participants' perceived partner's response to COVID‐related stress and their attachment insecurity in predicting their relationship satisfaction maintenance at the start of the COVID‐19 outbreak.

Consistent with the first hypothesis, higher levels of perceived supportive behaviors (e.g., showing understanding, providing tangible help) and lower levels of perceived negative behaviors (e.g., criticism, disregarding) in response to one's COVID‐related stress were associated with better relationship satisfaction. These findings are in line with a large body of research showing that partners' reactions to each other in times of stress are critical to their satisfaction (for a review, see Falconier et al., 2015, and Randall & Bodenmann, 2017). Importantly, the reported effects of supportive and negative behaviors on partners' relationship satisfaction emerged above and beyond the couples' baseline relationship satisfaction levels measured prior to the pandemic. Thus, the findings can be interpreted as pointing to the role of responsiveness in relationship satisfaction maintenance in stressful times.

It is worth noting that the effects of perceived supportive behaviors and negative behaviors were comparable in size in the current study. This contrasts with an extensive body of work documenting the phenomenon of “bad is stronger than good”; that is, that negative stimuli have more substantial effects on various psychological outcomes (including relational) than positive stimuli (see Baumeister et al., 2001). For example, in a set of daily diary studies examining the effects of dyadic support and hindrance, the latter were found to be more robustly associated with partners' daily relationship feelings, especially negative ones (Rafaeli et al., 2008). This pattern was interpreted to reflect people's social baseline (Beckes & Coan, 2011); that is, the presence, responsiveness, and availability of close others are, in the true psychological sense, the human baseline state. Thus, a partner's behaviors that fall short of the baseline expectation may lead to a more pronounced (negative) effect (Bar‐Kalifa & Rafaeli, 2015). Nevertheless, the current results challenge this interpretation by suggesting that stressful times call baseline expectations into question, rendering people as sensitive to positive relational behaviors as to negative ones.

In partial support of the second hypothesis, attachment anxiety interacted with the partner's perceived supportive (though not negative) behaviors in predicting relationship satisfaction maintenance. In other words, anxiously attached individuals were more reactive to their partners' responsiveness with regard to COVID‐related stress. Ample research has shown that given their concerns about their attachment figures' availability, individuals with high attachment anxiety tend to adopt hyperactivating attachment strategies, mainly consisting of excessive proximity‐seeking attempts (Collins et al., 2006; Mikulincer & Shaver, 2007, 2019). These attempts are often ineffective, leaving anxiously attached individuals perpetually vigilant for cues of their partner's unavailability (Mikulincer & Shaver, 2008). Furthermore, anxiously attached individuals may be particularly susceptible to the existential fears triggered by the COVID‐19 pandemic (e.g., mortality, loss, isolation; Steele, 2020), which thus may further exacerbate their reliance on their partners' support and soothing.

In partial support of the third hypothesis, attachment avoidance interacted with the negative (though not supportive) behaviors in predicting relationship satisfaction maintenance. Specifically, those participants high on attachment avoidance were less reactive to their partners' negative behaviors (e.g., blame, withdrawal) with respect to COVID‐related stress. The literature has repeatedly shown that avoidantly attached individuals often find it hard to trust their attachment figures' capacity to alleviate their distress. Therefore, when distressed, they adopt deactivating attachment strategies, consisting chiefly of withdrawal, self‐reliance, and detachment from relational needs (Barry & Lawrence, 2013; Cassidy & Kobak, 1988; Mikulincer & Shaver, 2008). Thus, the lack of sensitivity to their partners' negative behaviors may reflect their emotional disengagement and distancing from their partners (Feeney & Fitzgerald, 2019; Feeney & Karantzas, 2017; Overall et al., 2013).

Caution should be exercised in interpreting these findings to imply that attachment avoidance inoculates the relationship to the potentially deleterious effects of COVID‐related stress. For example, over time, chronic engagement in deactivating strategies is likely to result in a lack of connection as well as resentment on the part of these individuals' partners (Beck et al., 2013; Overall et al., 2015). Relatedly, even though avoidantly attached individuals often do not report experiencing distress, they still react viscerally to relational transgressions. For example, in response to relational stressors, they showed heightened physiological reactivity as manifested in their autonomic nervous system and the hypothalamic–pituitary–adrenal axis (Diamond & Fagundes, 2010; Feeney & Kirkpatrick, 1996; Pietromonaco et al., 2013). Future studies examining the interplay between COVID‐related stress and attachment avoidance should pay greater attention to the long‐term effects of their deactivating strategies and collect both subjective and objective indices of distress.

In contrast to our predictions, attachment anxiety did not interact with partner's perceived negative behaviors in predicting relationship satisfaction maintenance. This finding is inconsistent with research showing that highly anxious individuals are more sensitive to relationship threats (Mikulincer & Shaver, 2003, 2012; Simpson & Rholes, 2012) and experience higher levels of distress during conflicts (Campbell et al., 2005). In addition, in contrast to our predictions, attachment avoidance did not interact with partner's supportive behaviors in predicting relationship satisfaction maintenance. Though caution should be exercised in interpreting such null results, this lack of a linear association may be interpreted in light of Girme et al.’s (2015) findings of a curvilinear effect of partner's support for avoidantly attached individuals. It is possible that social support is only effective for highly avoidant individuals when it is provided at moderate levels of intensity when the partner communicates care, but the support is not experienced as overwhelming, which would evoke a sense of dependency. Unfortunately, our samples were insufficiently powered to test the quadratic effect of perceived support, but this should be considered in future research.

These findings may also have societal implications. For example, economic strain was shown to be linked to poor dyadic support and more negative behaviors (Cutrona et al., 2003; Jackson et al., 2016; Maisel & Karney, 2012; Williamson et al., 2013; though see Ross et al., 2019). Therefore, it is likely that the association between sociodemographic risk factors (e.g., poverty, minority identity) and marital distress is related to maladaptive dyadic coping behaviors (e.g., greater negative behaviors, lower support). This may be particularly pronounced during a time of crisis that engenders various economic challenges (Bareket‐Bojmel et al., 2021). Furthermore, sociodemographically disadvantaged individuals often reported elevated levels of attachment insecurity (Li et al., 2020; Schmitt, 2008). The results of the current study suggest that such insecurity may act as a vulnerability factor that places additional demands on disadvantaged couples' relationship satisfaction. Interestingly, recent work has highlighted that couples at higher risk benefit more from relationship education programs (Stanley et al., 2020). Therefore, it is important to provide access to such programs to sociodemographically disadvantaged couples, especially in times of global health crises.

Limitations and future directions

Several limitations of the current study should be mentioned. First, relative to the general population, the couples who participated in the current study reported high levels of satisfaction and lower levels of attachment insecurity on average (see Fletcher et al., 2000; Fraley et al., 2011; Funk & Rogge, 2007). Low levels of partners' negative behaviors were reported as well. This limited variability in the focal variables threatens the generalizability of the results and could reduce the statistical power to detect some of the hypothesized effects (e.g., interaction between attachment anxiety and partners' negative behaviors). Therefore, future studies would benefit from targeting more heterogeneous samples.

Relatedly, the data were collected soon after the outbreak of the pandemic in Israel, which may explain why we were able to collect follow‐up assessments from only half of the participants we contacted. Although the data were sufficiently powered to detect effects within the medium range, the effects that were found to be significant were in the low‐medium range (β = .12 to β = .26). This hints that the interaction effects found to be insignificant maybe be underpowered to test.

Furthermore, the moderate sample size prevented us from examining a more complex set of effects. Specifically, given the inherently dyadic nature of stress coping (Bodenmann, 2005; Bodenmann et al., 2006), it is imperative to consider the effect of one's partner's attachment insecurity as it relates to one's own attachment insecurity. For example, in one lab study observing couples discussing a relational conflict, anxiously attached women had difficulty recognizing their avoidant partners' distress; avoidantly attached men, for their part, had difficulties approaching and supporting their anxious partner (Beck et al., 2013). Therefore, as has recently been pointed out by Pietromonaco and Overall (2020), the pairing of attachment anxiety and avoidance is likely to result in poorer adaption to the current crisis.

It is also critical to consider the specific juncture at which the current study took place; that is, shortly after the outbreak of the crisis, when the Israeli population was under the first lockdown. In other words, even though the current study provides an ecologically valid glimpse into how partners cope with stress, when the data were collected, participants were likely to have experienced the pandemic as acute (vs. chronic) stress. Dyadic support can play a pivotal role in coping with both acute and chronic stress (Randall & Bodenmann, 2009), but enduring stress often depletes people of their resources and thus places them in greater need of their partner's support (Karney et al., 2005). By following romantic partners over a more extended period of time, future studies could achieve a more in‐depth understanding of how acute versus chronic stress affects partners. For example, we failed to find evidence that anxiously attached individuals were more reactive to their partners' negative behaviors. However, over time, when their personal resources are exhausted, anxiously attached individuals may react more strongly to their partners' destructive behaviors toward their COVID‐related stress.

Another issue worth mentioning is that partners' supportive and negative behaviors were measured using participants' self‐reports. Though it could be argued that the perceptions themselves are what matters in assessing one's own satisfaction, discrepancies between perceptions and behaviors effects on relational outcome could also be meaningful to explore. Specifically, it is likely that participants' attachment anxiety and avoidance biased their perceptions of their partners' relational behaviors (Barry et al., 2007; Overall et al., 2015). Therefore, to better understand the interplay between partners' behaviors and attachment style in the context of a measured stressor, future studies could incorporate self‐reports with observation methods. This would allow a more direct examination of how attachment style shapes both partners' perceptions and actual behaviors.

CONCLUSION AND IMPLICATIONS

The current pandemic constitutes a unique, although tragic, opportunity to explore the interplay between partners' dispositional vulnerabilities and dyadic coping in the face of a major ecological stressor. In this study, partners' relationship satisfaction served as the focal outcome, which is known to be robustly associated with partners' mental and physical health (e.g., Robles et al., 2014; Slatcher & Schoebi, 2017; Whisman & Baucom, 2012). The results indicated that the perceived quality of the partner's response to one's own COVID‐related stress predicted relationship satisfaction maintenance pre‐ to post‐COVID outbreak. Therefore, interventions aimed at strengthening partners' responsiveness and dyadic coping may be useful for partners when traversing these difficult times, especially approaches that can be delivered through a self‐directed, online platform (Spencer & Anderson, 2021), which is better suited to social‐distancing regulations. For example, Bodenmann and Shantinath (2004) developed an empirically based dyadic skill‐learning intervention targeting partners' dyadic coping. There is evidence to suggest that this program can be effectively provided online (Bodenmann et al., 2014). Therefore, policy makers may want to consider funding this type of intervention to make it more accessible to the larger community during stressful periods.

The results suggested that partners' preexisting dispositional vulnerability, namely, attachment insecurity, interacts with the perceived partner's responsiveness when predicting changes in relationship satisfaction. These results may help family counselors better identify couples with enduring vulnerability (e.g., those experiencing elevated attachment anxiety) who may be at greater risk during these stressogenic circumstances. In addition, the results highlight the need to implement interventions that directly address attachment insecurity when working with couples in distressing times. For example, emotionally focused therapy for couples (Johnson et al., 1999) has been shown to be effective in changing partners' attachment orientations (Burgess Moser et al., 2016; Wiebe et al., 2017). It is essential to integrate some of the interventions developed by this approach such as withdrawal re‐engagement, blamer softening (i.e., the therapist helps the more withdrawn partner and the more critical partner to express their vulnerable emotions and attachment‐related needs) in stressful global times.

Appendix S1 Supporting Information

Click here for additional data file.