Attachment theory posits that human beings are born with a psychobiological system -the attachment system- whose main role is to regulate stress and interpersonal relationships across the lifespan [1, 2].

In children, the experience of stress activates the attachment system, motivating the child´s proximity-seeking behaviours (e.g., crying) to significant others (attachment figures). From the caregiver´s point of view, this request activates her caregiving system: a set of behaviours to regulate the child’s distress [26].

The repetition of these interactions generates a system of social expectations in children: an “internal model” [27] of how relationships work. This model set the basis for a pattern of attachment strategies, also called attachment styles in adulthood [27]. According to longitudinal studies [28, 29], attachment has a fundamental role during the whole lifespan and can be measured throughout life with a 69%–75% concordance between attachment styles during childhood to adulthood.

Four attachment styles have been identified: secure, insecure-avoidant, insecure-anxious, and disorganised [29]. In short, people with secure attachment can show signs of their distress and look for help, expecting that others will be available to help and be actually helpful. Avoidant attachment styles are characterised by hidden distress, even when physiological signs of stress are present [30]. People with avoidant attachment style tend to be emotionally overregulated: visible stress signals are suppressed to keep the closeness with the attachment figure. This behaviour is believed to be rooted in fear of others’ rejection in front of negative emotion [2]. Attachment avoidance implies efforts to maintain distance from others, expressed in compulsive self-reliance and omission of help-seeking in front of stressful situations [2].

In contrast, anxious attachment is expressed in an intense need to be close to attachment figures, high anxiety levels upon separation, and a general emotional under-regulation. It is also characterised by an exaggerated expression of distress, excessive pursuit of other’s care, and a constant preoccupation that others will not be available when needed [2].

Lastly, people with disorganised attachment orientation show an inconsistent pattern of attachment activation-deactivation as a product of repeated experiences with threatening caregiving figures [29].

Hence, different attachment styles can potentially shape the experience of interpersonal relationships. Then, it becomes relevant to ask about the role different attachment styles have in the perception of social support as explained below.

The effect of attachment styles on perceived social support has been extensively studied [31–33]. Securely attached individuals perceive their relationships as more supportive and are generally more satisfied with the received support [32, 33]. By contrast, non-securely attached individuals report low availability of supportive figures and less satisfaction in social interactions [24, 34]. Specifically, people showing anxious attachment styles tend to initially idealise supportive figures, but they become easily disappointed, remaining with feelings of hopelessness [31, 35]. In opposition, avoidant attachment has been related to feelings of discomfort when depending on others for support, suppressing the expression of distress, and support-seeking behaviours [31]. In conclusion, attachment styles influence the way individuals experience social support.

As previously mentioned, the positive effect of social support on mental health is well documented but the characteristics of the individual differences in this relationship are still unexplored. It is relevant to understand these differences to progress in the study and treatment of mental health disorders, particularly in mood disorders like depression. Depression and depressive symptoms have been established as one of the main priorities in health research [36] due to its high and increasing prevalence and long-term consequences [37].

In the present context of the COVID-19 pandemic, depression has become of the greatest public health concern as its incidence has dramatically risen [38, 39]. The deep changes in daily social dynamics imposed by the pandemic are one of the contextual aspects of this increment [39, 40]. In this context, the role of attachment in the relationship between social support and depression becomes exceptionally relevant. Particularly, since the current COVID-19 pandemic is a known factor of stress [39, 41, 42]. As such, the pandemic has the potential to activate the attachment system [43]. In Figure 1, we offer a model for understanding the plausible effect of COVID-19 pandemic on attachment, social support, and the link with mental health outcomes such as depressive symptoms.

In insecurely attached individuals, the stress activation may result in either a hyperactivation of the attachment system (anxious attachment style) and excessive dependence on others or deactivation of the attachment system (avoidant attachment style) and compulsive self-reliance. In both cases, as discussed above, the experience of support from others would be negatively altered.

In a secure attachment, the activation of the attachment system and the consequent help-seeking behaviours can help look for social contact and positively use those social interactions. However, the range of measures to protect public health includes social distancing and social isolation as key prevention of COVID-19 incidence. In this context, even securely attached people who could benefit from social support in normal conditions cannot rely on their usual stress regulation strategies. Thus, the usual strategies to deal with stress are unavailable, leading the stress systems to continue to be activated.

In this scenario, secure attachment strategies are not necessarily helpful, and therefore insecure attachment strategies may get activated, which, in turn, impede the perception of social support. This process has the potential to affect mental health negatively.

Although there is growing evidence for the role of social support and attachment in mental health [25, 44–46], and particularly in depression [46, 47], no study to date has investigated the moderating role of attachment in the relationship between social support and depression in the context of COVID-19 pandemic. Reports show that the pandemic sanitary procedures (e.g., social distancing, repeated lockdowns, school closures) and the fear of contracting the virus among other precautionary measures have heavily impacted mental health [39, 48, 49].

Chilean circumstances have been particularly challenging. First, the pandemic started above a disrupted social context due to the 2019 social outbreak that has had important social, political and economic consequences [50], which are known risk factors for mental health [51]. Second, since mid-March 2020 Chile adopted severe restrictions on movement and social contact [52] with already known negative consequences on people’s mental health [53]. Third, because of both sanitary and social crisis, a large percentage of the population has faced serious economic difficulties [52].

Therefore, the present study aims to explore the predictive role of social support on depressive symptoms and the moderating role of attachment styles in the association between social support and depressive symptomatology.

Based on the previous literature [3, 25, 44, 47, 54], we hypothesised that securely attached individuals are more likely to display support seeking behaviours and that these strategies will positively influence their mental health. In contrast, insecurely attached individuals (either anxious or avoidant) under stress tend to isolate and behave in a compulsively self-reliant way; therefore, they will not look for help from others, impeding the benefit from social support.

Based on the evidence mentioned above, we expect to find 1) higher perceived social support to be associated with lower depressive symptoms; 2) higher attachment security to be associated with lower depressive symptoms, and 3) a moderating role of attachment in the association between social support and depressive symptomatology.

One hundred and eighty-eight participants answered the set of online questionnaires and were included in the study analyses (N women = 148) during December 2020 and January 2021 (i.e., after 9 months of severe restrictions to social contact and during one of the strict lockdown periods). Five participants left an item blank on one of the self-reported scales. These missing values were substituted by the mean series method, using the mean of the values answered by the participant on the same scale. The sample’s mean age was 39.78 (SD = 14.72), ranging between 18 and 80 years old. 87.7% of participants reported having completed undergraduate or postgraduate education. Analysis related to sociodemographic information and the study variables is available in the Supplementary File S1.

Participants were invited to voluntarily join the study through different online platforms. All participants agreed through informed consent, which was approved on 12/29/2020 by the Ethics Committee of Universidad del Desarrollo. As retribution for participating, three gift cards valued at $100.000 CLP (around $120 US) were raffled at the end of the data collection.

The sample size was previously calculated through power analysis using the G*Power program. Anticipating an effect size of 0.15, p-value > 0.05 and B = 0.95, the minimum sample size required was N = 129. Questionnaires were displayed on the Survey monkey online platform (https://es.surveymonkey.com/). The whole survey took approximately 10–15 min for participants to complete. Only participants with completed surveys were considered as part of this study.

Participants were asked to complete the following three scales. Further details of these instruments and data analysis plan are available in the Supplementary File S1.

The Medical Outcome Study Social Support Survey (MOSS; [55]) measures perceived social support and it is validated for the Chilean population [56]. This 18 items instrument uses a 5-point Likert scale (from 1 = Never; to 5 = Always) and an additional item assessing social network size. MOSS includes five dimensions: positive social interactions, affectionate support, instrumental support, emotional/informational support, and the social network size.

The Experiences in Close Relationship Scale Short Form [57] was adapted to measure adult attachment in general relationships (Adaptation process in Supplementary File S2). The modified version of the ECR-S used in this study has 12 items with seven response levels (from 1 = strongly disagreer; to 7 = strongly agree). The instrument allows to obtain an index of anxious and avoidant attachment for respondents.

The Beck Depression Inventory-Short Form (BDI-SF; [58]) measures depressive symptoms. We used the version validated for the Chilean population [59], and shows good reliability and validity in various populations [58, 59].

The linear regression analysis evidenced a significant negative association between social support and reported depressive symptoms (F(1,186) = 33.971, p < 0.001) with a R 2 = 0.154. Specifically, for each increasing point of perceived social support, reported depressive symptoms were reduced by −0.178 points (see Figure 2).

To explore the differential effect of each social support dimension on reported depressive symptoms a multiple linear regression was run. This model (F(5,176) = 7.469, p < 0.01, R 2 = 0.175), revealed that positive social interactions (b = −0.341, t(181) = −2.177, p = 0.031) was the only significant predictor of lower depressive symptoms among the other social support dimensions (all other p-values > 0.05, See Table 1 for details).

A multiple linear regression was run to test the relationship between attachment and reported depressive symptoms. Importantly, these two dimensions of attachment insecurity evidenced no correlation (r = 0.103, p = 0.16). The regression showed a significant and positive relationship between depression and ECR-S attachment styles (F(1,185) =19.786, p < 0.01) with a R 2 = 0.176. Importantly, both attachment styles resulted as significant coefficients in this model, being avoidance (b = 0.336, t(187) = 5.041, p < 0.01) a slightly stronger predictor than anxiety (b = 0.192, t(187) = 3.218, p < 0.01). Therefore, higher levels of these attachment styles (and therefore, higher levels of attachment insecurity) are related with higher levels of reported depressive symptoms.

A moderated moderation was run to test the role of the two attachment dimensions on the relationship between perceived social support and depression (See a representation in Supplementary File S3). Under this aim, social support scores, insecure attachment dimensions and all the interactions among these predictors were included to explain depression score variance.

As shown in Table 2, this analysis evidenced a significant simple effect of avoidance (t(180) = 2.671, p = 0.01). The model also showed significant interactions between avoidance and social support (t(180) = −2.854, p < 0.01), and between the two insecure attachment dimensions (t(180) = −2.583, p = 0.01). Importantly, the third order interaction with the relevant independent variable and the two moderators was also significant (t(180) = 2.914, p < 0.01). This interaction explained a rather small but significant increase in depression variance, Δ R 2 = 0.03, F(1,180) = 8.49, p < 0.01. Thus, anxiety and avoidance attachment styles are significant moderators of the relationship between social support and depression.

As reported above, educational level and age resulted in significant correlations with self-reported depression symptoms. Therefore, a new moderated moderation model was run including these two variables as covariates to control for their possible effect. This model evidenced the same significant effects; the simple effect of avoidance and the interaction between this attachment style with social support, and the two insecure attachment dimensions. Importantly, the three order interaction of social support, anxiety and avoidant attachment style was also significant. Regarding the covariates, educational level did not reach the level of significance while age did (t(178) = −3.176, p = 0.01). As expected, this model still explained a small and significant portion of depression variance, Δ R 2 = 0.02, F(1,178) = 6.48, p = 0.01.

Through the exploration of the three-way interaction of the model, it is observed that insecure attachment levels (i.e., attachment avoidance and anxiety) impacted the relationship between social support and depression in five different combinations (see Figure 3). Therefore, as detailed in Table 2, insecure attachment is a relevant moderator for low and medium levels of avoidance together with low and medium levels of anxiety. Also, for medium levels of both insecure styles, and for high levels of avoidance together with low and medium levels of anxious attachment. These conditional effects evidenced that the negative relationship between social support and depressive symptoms is effective only for certain profiles of insecure attachment. Interestingly, participants scoring high on both avoidant and anxious style have a slight increase of depressive symptoms related to higher levels of social support. Although this result did not reach a significant level, it is important to consider in further studies.