Differential Pathways from Child Maltreatment Types to Insecure Adult Attachment Styles via Psychological and Social Resources: A Bayesian Network Analysis

Abstract

Child maltreatment has been linked to insecure adult attachment. However, it is not yet clear how different child maltreatment types are associated with attachment-related anxiety and avoidance in adulthood; and whether resilience against these insecure attachment styles is dependent on risk-specific resources. Therefore, this study explored differential pathways from child maltreatment types to attachment-related anxiety and avoidance in adulthood and examined whether psychological resources (self-esteem) and social resources (perceived social support) show risk-specific effects. An online survey retrospectively assessed experiences of child maltreatment, the level of attachment-related anxiety and avoidance in adulthood, self-esteem, and perceived social support in N = 604 former members of fundamentalist Christian faith communities (mean age = 41.27 years, SD = 12.50; 65.90% female). Cross-sectional data was analyzed using Bayesian network analysis. Only emotional child maltreatment showed direct relationships to insecure adult attachment. Specifically, emotional abuse and emotional neglect were associated with anxious and avoidant adult attachment, respectively. The effects of other child abuse types on adult attachment were mediated through emotional abuse, which indicated patterns of complex traumatization. Self-esteem mediated the effect of emotional abuse on anxious attachment, while perceived social support mediated the effect of emotional neglect on avoidant attachment. Social support was also linked to self-esteem and was therefore also important for individuals with experiences of emotional abuse. This study showed that child maltreatment types and their interactions are meaningfully linked to attachment-related anxiety and avoidance in adulthood. Interventions for survivors of child maltreatment should focus on risk-specific resources to support their resilience.

Keywords

child abuse and neglect resilience adult attachment Bayesian network analysis

Introduction

Child maltreatment is a multidimensional construct that can be differentiated into physical abuse (i.e., bodily harm and injuries) and neglect (i.e., basic physical needs are not met), emotional abuse (e.g., humiliation, hostility, rejection) and neglect (i.e., emotional needs are not met, emotionally unresponsive or unavailable caregivers), as well as sexual abuse (i.e., attempted or actual sexual contact) (Bernstein & Fink, 1994; Cicchetti & Toth, 2005; Glaser, 2002). These types are likely to be interrelated and occur together (English et al., 2005; Glaser, 2002). As child maltreatment is often perpetrated by trusted persons in a child’s life (Finkelhor et al., 2014; Hurren et al., 2018), it can have immediate and formative consequences for the child’s attachment to significant persons in childhood (i.e., parents, caregiver), as well as long-term consequences for significant relationships in adulthood (i.e., romantic partners) (Mikulincer & Shaver, 2005). With regard to long-term consequences, child maltreatment has been linked to an increased likelihood for experiencing attachment-related anxiety and avoidance in adulthood (Cyr et al., 2010; Lo et al., 2019; Reyome, 2010). These insecure adult attachment styles have been linked to a higher likelihood for experiencing intimate partner violence in adulthood (e.g., Bonache et al., 2019). Therefore, survivors of child maltreatment may become stuck in a cycle of enduring maltreatment (Widom et al., 2014), partly due to the attachment patterns formed in response to their adverse childhood experiences. Thus, to better support adults with experiences of child maltreatment, further research is needed on how child maltreatment types interact and how they may be linked to attachment-related anxiety and avoidance in adulthood.

Insecure Adult Attachment

Adult attachment theory states that the basis for an individual’s attachment style in adulthood, that is, how adults generally experience and behave in romantic relationships, begins in childhood in response to how the child is treated by its parents, caregivers, and other significant persons (Brennan et al., 1998; Fraley & Roisman, 2019). Childhood can have lasting effects on the individual’s social and emotional life, because the individual learns to read social cues, what to expect from others, and how to regulate emotions in social situations (Mikulincer & Shaver, 2005). Hence, studies have shown associations between a child’s attachment patterns with its primary caregivers and their attachment patterns in romantic relationships in adulthood (Fraley & Roisman, 2019), demonstrating their trait-like character.

The two most commonly researched attachment styles in adulthood are anxious and avoidant attachment, collectively known as insecure attachment styles (Brennan et al., 1998; Caldwell et al., 2011). The less attachment-related anxiety and avoidance an individual experiences in adult romantic relationships, the more that individual shows a secure attachment (Brennan et al., 1998). Anxious attachment is mainly characterized by a strong, unfounded fear of being abandoned and rejected by one’s partner. It has been linked to hypervigilance toward threats to the self and the romantic relationship, as well as a demand for intense physical and emotional bonding with the partner at all times (Fraley & Shaver, 2000). Adults with an anxious attachment style have been shown to have a biased perception of themselves, with a tendency to feel unworthy of being loved (Caldwell et al., 2011). Avoidant attachment is predominantly characterized by significant self-reliance due to a biased perception of others as not to be trusted or depended upon (Brennan et al., 1998; Caldwell et al., 2011). Individuals with an avoidant attachment style aim to avoid closeness, emotional bonding, and intimacy, which otherwise would lead to discomfort (Fraley & Shaver, 2000). In light of these differential characteristics of anxious and avoidant attachment styles, it may be that different types and/or combinations of child maltreatment are linked to differential insecure adult attachment styles.

Childhood Maltreatment and Insecure Adult Attachment

Only a few studies have investigated the relationship between types and/or combinations of child maltreatment and insecure adult attachment styles. For instance, a retrospective study by Caldwell et al. (2011) with at-risk mothers found a predictive effect of child emotional abuse on anxious attachment. Another retrospective study by Unger et al. (2014) investigated all three types of child abuse (separately), as well as a composite score for child neglect, in university students. Results showed a significant effect of physical abuse on avoidant attachment and a significant effect of emotional abuse on anxious attachment. No effect was found with regard to sexual abuse or the neglect composite score (Unger et al., 2014). Furthermore, a retrospective study by Espeleta et al. (2017) with college women found that emotional abuse predicted anxious and avoidant attachment, while physical abuse only predicted avoidant attachment. Sexual abuse did not show any predictive effects. In addition, a retrospective study with college students by Oshri et al. (2015) found that while emotional and sexual abuse predicted both attachment styles, emotional abuse showed a stronger effect on anxious than avoidant attachment, with the opposite effect for sexual abuse (Oshri et al., 2015). A longitudinal study found that physical abuse only predicted anxious attachment, while physical neglect predicted both anxious and avoidant attachment (Widom et al., 2018). Finally, a recent study by Maercker et al. (2021) found that emotional abuse and neglect, as well as physical neglect, were predictive of anxious attachment in individuals formerly affected by compulsory social measures and placements in childhood. Thus, the existing research provides most empirical evidence for the link between child emotional abuse and anxious attachment.

Because emotional abuse refers to a more active form of maltreatment by explicitly giving children the impression that they are worthless, not loved, etc., it aligns more closely with the biased representation of the self and the fear of not being loved that characterizes anxious attachment in adulthood (Glaser, 2002). Furthermore, although physical and sexual abuse were found to directly predict anxious attachment, particularly by fostering hypervigilance to threats (Jaffee, 2017), their psychological effects may be mediated via emotional abuse by amplifying feelings of worthlessness and not being loved (Glaser, 2002). In contrast, more passive forms of maltreatment, that is, physical and emotional neglect, may rather be linked to avoidant attachment, with emotional neglect potentially playing a more direct role than physical neglect (Glaser, 2002). Experiences of neglect can teach children that they cannot depend on attachment figures for their needs and therefore need to rely on themselves (Caldwell et al., 2011). Thus, it may be that child abuse is more likely to be linked to anxious attachment and child neglect may rather be linked to avoidant attachment.

However, there is a lack of knowledge about how different types of child maltreatment may interact together and their link to attachment-related anxiety and avoidance in adulthood. Past studies have mainly focused on their independent effects (e.g., Maercker et al., 2021; Unger et al., 2014). However, recently there has been an increase in the application of a systems or network approach in the social sciences, which stresses interdependence rather than independence to better understand real-life phenomena, including child maltreatment (Fritz et al., 2018; Thoma et al., 2020). It is of utmost importance to adequately identify these potentially differential pathways to develop effective interventions, as survivors of child maltreatment have been shown to have an increased likelihood of experiencing ongoing violence linked to their insecure adult attachment (e.g., Bonache et al., 2019). Resilience theory suggests that in order to attenuate the link between child maltreatment and insecure adult attachment, protective resources need to be identified that are meaningful within the context of a specific child maltreatment type—adult attachment relationship (Ungar, 2018).

Child Maltreatment, Insecure Adult Attachment, and Protective Resources

In the context of this study, resilience is defined as the absence of insecure adult attachment despite the experience of child maltreatment (R. Kalisch et al., 2017). An individual’s level of resilience depends on the availability and accessibility of contextually meaningful protective factors and processes, that is, resources (Ungar & Theron, 2020). According to differential susceptibility and impact theory (Ungar, 2018), the protective effect of such resources can vary with different types of risk (such as child maltreatment types). As an anxious attachment style is associated with a biased representation of the self, which can be linked to child emotional abuse (e.g., Espeleta et al., 2017; Maercker et al., 2021; Oshri et al., 2015); a psychological resource, that is, self-esteem (Rosenberg, 1979), was investigated in the current study. Self-esteem may be impacted by the experience of child maltreatment (Arslan, 2016; Thoma et al., 2020), and it has previously been found to be a meaningful mediator of the relationship between child emotional abuse and neglect and mental health (Thoma et al., 2021). As a social resource, perceived social support was investigated in the current study. Perceiving (high levels of) social support can indicate to a person that others care for them and that they are valued (Cobb, 1976). Therefore, perceived social support may be particularly affected by child neglect, because child neglect might rather be linked to a biased perception of others, and thus, related to avoidant attachment.

The Present Study

Past research suggests that emotional maltreatment during childhood may play a primary role for the presence of insecure attachment styles in adulthood, compared to physical maltreatment and sexual abuse (Glaser, 2002; Wu et al., 2021). However, clear patterns about how the specific emotional maltreatment types (i.e., abuse, neglect) may be linked to insecure adult attachment have not been identified, as well as how the other child maltreatment types are connected to adult attachment style, likely via emotional maltreatment. Based on the existing literature, the current study hypothesized that:

H1a: Emotional abuse will be associated with anxious attachment.

H1b: Emotional neglect will be associated with avoidant attachment.

H1c: Sexual and physical abuse, as well as physical neglect, will be associated with anxious and/or avoidant attachment via emotional maltreatment.

Furthermore, to shed light on how to adequately support adults with experiences of child maltreatment, this study explored if resilience against insecure attachment in adulthood is dependent on risk-specific resources. Therefore, the potentially differential role of a psychological resource (self-esteem) and a social resource (perceived social support) was investigated (Ungar, 2018). Based on the existing literature it was hypothesized that:

H2a: Adult self-esteem will significantly mediate the relationship between child emotional abuse and anxious attachment in adulthood.

H2b: Perceived social support in adulthood will significantly mediate the relationship between child emotional neglect and avoidant attachment in adulthood.

However, research has also shown that resilience-supporting resources may be interdependent, even when they belong to different resource systems (Höltge et al., 2020, 2021). In addition, differential impact theory states that resilience to more severe forms of adversity (such as child maltreatment) may be linked more to social than psychological resources (Ungar, 2018). It was therefore further hypothesized that:

H3: Social support will be positively associated with self-esteem.

In order to gain insight into these complex pathways, a Bayesian network analysis was applied. Bayesian network analysis is an explorative, data-driven statistical approach that can estimate conditional dependencies between a given set of variables based on cross-sectional data (Scutari et al., 2018). It explores the most likely unidirectional pathways from a set of predictors to outcomes of interest. In the context of this study, a Bayesian network analysis was applied to explore the most likely pathways to insecure adult attachment via different types of child maltreatment. Thus, this method indicates which child maltreatment type(s) may play a key role for the presence of insecure attachment in adulthood, as well as in the understanding of complex traumatization, that is, the experience of multiple types of child maltreatment. Bayesian network analysis was also applied to explore potentially context-specific effects of different types of protective resources and their interactions. Understanding the relationships between child maltreatment and insecure adult attachment can aid in the design of more effective and efficient intervention studies.

Method

Study Design

This secondary data analysis was based on cross-sectional data from a multicountry (i.e., Austria, Germany, and Switzerland), mixed-methods research project about the life of individuals who had left a fundamental Christian faith community “Psychological strain and resilience after leaving or exclusion by a fundamentalist Christian faith community”, which included data on child maltreatment, resilience-supporting resources, and adult attachment. Data was collected from February to June 2021 via an anonymous online survey. The study was conducted by the University of Zürich, Switzerland, in collaboration with the University of Vienna, Austria . The study protocol was approved by the Ethics Committees of the University of Vienna, Austria (ID: 00662), of the German Psychological Society [Deutsche Gesellschaft für Psychologie, DGPs], Germany (ID: 2021-01-08VA), and of the Faculty of Arts and Social Sciences in the University of Zürich, Switzerland (ID: 20.12.18).

Participants and Recruitment

Eligible participants were aged 18 years or older, German native speakers, resident in Austria, Germany, or Switzerland, and who were a former (self-identified) member of a fundamentalist Christian faith community. The recruitment was conducted in collaboration with a diverse set of organizations (e.g., information centers, support groups for ex-members of different fundamentalist Christian faith communities), various experts (e.g., specialized psychotherapists), and publicly active ex-members. These organizations and individuals advertised the study via social media (e.g., posting the study link on their respective social media channels), as well as via direct personal contacts within their social networks and communities. Snowball sampling was also used to recruit study participants.

Procedure

Unipark (Unipark, Germany) was used to configure the anonymous online survey. First, eligibility criteria were assessed and, if met, participants had to provide online informed consent to access the survey. The order of the scales was randomized to minimize any order or sequence effects (i.e., response bias) that may be present in survey studies. Participants were provided with a list of psychosocial support options in the case they felt distressed as a consequence of their participation. Withdrawal from the online survey was possible at any time. Participation was not financially compensated in order to have intrinsically motivated participants and to refrain from storing personal identifiers. Participants could provide their email addresses in order to receive a summary of the study results.

Measures

All data was self-reported by the participants. The respective means, standard deviations, and reliabilities of the variables, as well as their correlations, are shown in Table 1. Sex (female, male) and age (in years) were included as control variables.

Table 1.

Sample Characteristics: Reliabilities and Correlations between Study Variables.

	M (SD)	ω	1	2	3	4	5	6	7	8	9	10
Sex	65.90% female	–	–
Age	41.27 (12.50)	–	.21**
PA	8.58 (4.37)	.89	−.06	.10*
PN	8.51 (3.35)	.61	−.08	.15**	.45**
EA	12.74 (5.57)	.87	−.29**	.01	.59**	.52**
EN	14.27 (5.30)	.90	−.06	.17**	.45**	.60**	.65**
SA	7.19 (4.36)	.96	−.23**	.05	.39**	.37**	.42**	.30**
AX	57.49 (24.39)	.93	−.16**	−.15**	.15**	.16**	.33**	.28**	.25**
AV	55.21 (22.77)	.91	−.06	–.02	.15**	.18**	.28**	.34**	.24**	.47**
SS	51.95 (12.72)	.95	−.07	−.09*	−.16**	−.26**	−.25**	−.40**	−.14**	−.32**	–.44**
SE	18.99 (6.83)	.93	.14**	.19**	−.16**	−.19**	−.38**	−.30**	−.26**	−.49**	−.41**	.40**

Note. PA = physical abuse; PN = physical neglect; EA = emotional abuse; EN = emotional neglect; SA = sexual abuse; AX = anxious adult attachment; AV = avoidant adult attachment; SS = perceived social support; SE = self-esteem; M = mean; SD = standard deviation; ω = omega (reliability coefficient).

N = 604. *p < .05. **p < .01.

Childhood Trauma Questionnaire

Child maltreatment types were assessed with the Childhood Trauma Questionnaire (Bernstein & Fink, 1994), which retrospectively assessed the frequency of experienced child abuse (emotional, physical, sexual) and neglect (emotional, physical). Five items per sub-scale (i.e., child maltreatment type) were rated on a 5-point Likert scale (1 = never to 5 = very often, sum-score range per sub-scale: 5–25). Higher scores indicate a higher frequency of child maltreatment type.

Experiences in Close Relationships Inventory—Revised

The Experiences in Close Relationships Inventory—Revised (Fraley et al., 2000) assessed the extent to which individuals express attachment-related anxiety (i.e., feeling secure vs. insecure about the availability and responsiveness of romantic partners), as well as attachment-related avoidance (i.e., feeling secure vs. uncomfortable being close to or depending on others) within romantic relationships. The extent of expression of each attachment style is indicated on a separate dimension with 18-items each, rated on a seven-point Likert scale (1 = disagree strongly to 7 = agree strongly; sum-score range for each dimension: 18–126). Higher scores indicate a stronger expression of attachment-related anxiety or attachment-related avoidance.

Rosenberg Self-Esteem Scale

Self-esteem was assessed with the Rosenberg Self-Esteem Scale (Rosenberg, 1979), with 10 items rated on a four-point Likert scale (0 = strongly disagree to 3 = strongly agree, sum-score range: 0–30). Higher scores indicate higher levels of self-esteem.

Social Support Questionnaire

Perceived social support was assessed with the Social Support Questionnaire (Fydrich et al., 2009), with 14 items rated on a five-point Likert scale (1 = strongly disagree to 5 = strongly agree, sum-score range: 14–70). Higher scores indicate higher levels of perceived social support.

Statistical Analysis

First, the data was analyzed for invalid scores and response biases (e.g., no variance in responses, impossible values), duplicates, outlier/influencer analyses using Cook’s Distance, normal distribution and skewness using PP-plots, and missing values. No variables needed to be transformed due to significant skewness and no outliers/influencers were detected. List-wise deletion was used for participants with missing values on any of the variables analyzed in the present study.

Bayesian Network Analysis

A Bayesian network was estimated to infer potential pathways between types of child maltreatment, resources, anxious, and avoidant adult attachment, and demographic characteristics, using observational, cross-sectional data. This Bayesian network was represented via a directed acyclic graph (DAG; M. Kalisch & Bühlmann, 2007; McNally et al., 2017; Scutari et al., 2018). A DAG is a graphical representation of a dependency structure that is characterized by not having direct or indirect feedback loops between variables. Therefore, a variable cannot predict itself via other variables (Borsboom et al., 2013). The DAG was estimated using the Max-Min Hill-Climbing (MMHC) algorithm (Tsamardinos et al., 2006) via the bnlearn R package (version 4.5; Scutari, 2010). The MMHC algorithm is a hybrid that combines local learning, constrained-based, and search-and-score techniques to identify the best fitting model for the data (Tsamardinos et al., 2006). First, conditional independencies between variables are tested to identify the dependence structure of the variables (Scutari et al., 2018). If two variables are estimated to be marginally dependent, they may become conditionally independent if a third variable explains their direct relationship (i.e., a third variable that leads to a mediation effect). This first step provides the basis of the DAG, in which the direction of influence between two variables is not yet estimated. Second, a hill-climbing algorithm is used to identify the optimal network structure, including directional relationships. This process begins with an empty network, followed by adding, deleting, or reversing pathways between the variables, only for the relationships that were identified in the first step. The aim is to optimize a goodness-of-fit metric, in this case the Bayesian Dirichlet likelihood-equivalence uniform score, until it cannot be maximized anymore.

To ensure that the resulting network and its directed pathways are genuine and stable, 10,000 non-parametric bootstraps were performed. This procedure results in the proportion of times each relationship was identified out of the 10,000 bootstraps and how often each directed pathway occurred. A directed pathway was included in the final model if: (1) the relationship (without direction) was estimated in at least 85% of the bootstraps (which is the strength of a directed pathway); and (2) if the direction was estimated in at least 50% of the bootstraps (McNally et al., 2017; Scutari & Nagarajan, 2013). These cut-offs can be interpreted as significance in DAGs (Scutari & Nagarajan, 2013).

To aid the algorithm, several directed pathways were blacklisted (i.e., forbidden) from being included in the model: (1) All effects on age and sex, as age and sex cannot be influenced by the other variables. (2) All effects on the types of child maltreatment, except for sex, as their occurrence in childhood cannot be influenced by current age, insecure attachment style, or resources. However, females tend to experience some forms of maltreatment more than males, such as sexual abuse (Walker et al, 2004). Effects of anxious and avoidant attachment on the resources were not blacklisted in order to explore if the resource status could be a result of the present insecure attachment styles. No directed pathways were whitelisted (i.e., forced into the model), so that all estimated directed pathways in the model were discovered by the algorithm, that is, databased.

This procedure was used to test the following models: (1) if there were direct differential pathways from certain child maltreatment types to anxious and/or avoidant attachment; and (2) if self-esteem and perceived social support showed risk-specific effects. While DAGs are usually described based on the strength and direction of the identified pathways (McNally et al., 2017), an additional analysis was performed, conditional probability queries (CPQ; Nagarajan et al., 2013), to provide estimates for the prediction of anxious and avoidant attachment by their related variables in model 1 and model 2. In CPQ, the averaged Bayesian network of the 10,000 bootstraps is viewed as an expert system. This model can be used to estimate the posterior probability of an outcome based on the available evidence on the values of certain variables/predictors (Nagarajan et al., 2013). 10⁴ samples were generated for each conditional probability query for precise probability estimates. The resulting estimates indicate how much the outcome changes if the predictor increases by one point.

Results

Sample Characteristics

In total, 622 individuals participated in the study. However, n = 18 individuals were excluded from the present analysis due to missing values in sex (n = 4), age (n = 1), or adult attachment style (n = 13, n = 10 had more than 80% missing items and n = 3 had less than 50% missing items). The excluded and included participants significantly differed in their self-esteem (t[615] = 2.77, p < .01) and perceived social support (t[615] = 4.15, p < .01) in that the excluded group reported significantly less levels in these resources.

The final sample consisted of N = 604 participants (with n = 508 living in Germany, n = 73 in Switzerland, n = 23 in Austria), and the majority being female (65.90%). The most frequently reported child maltreatment type was emotional neglect and emotional abuse, followed by physical abuse and physical neglect. Sexual abuse showed the lowest prevalence. All marginal correlations between the different child maltreatment types were significant, ranging from .30 (emotional neglect-sexual abuse) to .65 (emotional abuse-emotional neglect). All child maltreatment types were also significantly and positively related to both insecure adult attachment styles. Anxious and avoidant adult attachment showed a correlation of .47. Both resources (i.e., self-esteem and perceived social support) correlated positively with each other and showed significant negative correlations with all child maltreatment types, as well as both insecure attachment styles. With regard to sex differences, female participants reported significantly more emotional and sexual abuse, as well as higher scores of anxious attachment. Male and female participants did not differ with respect to their levels of avoidant attachment. Regarding the resources, male participants showed higher scores in self-esteem. See Table 1 for an overview of the sample characteristics.

Model 1: Child Maltreatment Types and Insecure Adult Attachment

Model 1 tested if emotional abuse is associated with anxious attachment (H1a), if emotional neglect is associated with avoidant attachment (H1b), and if sexual and physical abuse as well as physical neglect are indirectly related to insecure adult attachment via emotional maltreatment (H1c). As shown in Figure 1a, emotional neglect was only associated with avoidant attachment (b = 1.58, SE = .04, t = 39.22, p < .01, R² = .13), and emotional abuse was only associated with anxious attachment (b = 1.35, SE = .04, t = 32.61, p < .01, R² = .10). However, emotional neglect had an indirect effect on anxious attachment, as avoidant attachment was shown to predict anxious attachment (in 64% of the bootstraps) rather than the other way round. Furthermore, physical abuse showed an indirect effect on anxious attachment via emotional abuse, while sexual abuse showed an indirect effect on anxious attachment via physical and emotional abuse. No significant effects on insecure adult attachment were found for physical neglect.

Figure 1.

(a) Estimated Bayesian Network of Model 1 and (b) Estimated Bayesian Network of Model 2.

With regards to the covariates, sex was found to act as a significant predictor of sexual abuse and emotional abuse, with female participants reporting more experiences in both abuse types, as was expected based on the marginal correlations (see Table 1). Therefore, sex also showed an indirect effect on anxious attachment via emotional and sexual abuse. Physical neglect was only predicted by physical abuse and emotional neglect and was therefore not indirectly related to either insecure attachment style. Age was not found to be a significant predictor of any of the other model variables.

Model 2: The Differential Roles of Self-Esteem and Perceived Social Support

Model 2 tested if adult self-esteem is a significant mediator in the relationship between child emotional abuse and anxious attachment in adulthood (H2a), if perceived social support in adulthood is a significant mediator in the relationship between child emotional neglect and avoidant attachment in adulthood (H2b), and if social support is positively associated with self-esteem (H3). Model 2 replicated the directed pathways between the child maltreatment types from model 1 and the effects of sex (see Figure 1b). Furthermore, it found that self-esteem and perceived social support show risk-specific effects: Self-esteem significantly explained the relationship between emotional abuse and anxious attachment, while the relationship between emotional abuse and anxious attachment remained significant, indicating a partial mediation (effect of emotional abuse on anxious attachment when controlling for self-esteem: b = 0.13, SE = 0.04, t = 3.47, p < .01, R² = .00). Social support significantly and fully explained the relationship between emotional neglect and avoidant attachment (effect of emotional neglect on avoidant attachment when controlling for social support: b = 0.01, SE = 0.04, t = 0.28, p > .05, R² = .00), indicating a full mediation. Furthermore, social support had an indirect effect on anxious attachment via a positive effect on self-esteem. Age also had a significant and positive effect on self-esteem.

Discussion

The first aim of this study was to investigate which types of child maltreatment (i.e., physical and emotional abuse and neglect, sexual abuse) were primarily related to which insecure adult attachment styles (i.e., anxious and avoidant) by controlling for their respective mutual dependence structure. The second aim was to investigate whether psychological (i.e., self-esteem) and social (i.e., perceived social support) resources have risk-specific effects on these potentially differential relationships. The results showed that only child emotional maltreatment had a direct relationship with insecure adult attachment: Emotional abuse was linked to anxious attachment, and emotional neglect was linked to avoidant attachment. The effects of other child abuse types on adult attachment were mediated primarily through emotional abuse. In addition, self-esteem directly mediated the effect of emotional abuse on anxious attachment, whereas perceived social support directly mediated the effect of emotional neglect on avoidant attachment. These findings are discussed in the following sections.

Differential Pathways from Child Maltreatment to Insecure Adult Attachment

Interactions in childhood with significant others, such as parents or caregivers, can have a formative influence on adult romantic relationships, that is, adult attachment (Fraley & Roisman, 2019). Past research provides a basis for the expectation that emotional maltreatment during childhood will play a primary role in the presence of insecure attachment in adulthood, compared to physical maltreatment and sexual abuse (Glaser, 2002; Wu et al., 2021). Therefore, it was hypothesized that only child emotional neglect and abuse are directly linked to adult attachment styles, and each is related to a different attachment style. The current study confirms this expectation using an exploratory, data-driven approach: Emotional abuse was linked to anxious attachment and emotional neglect was linked to avoidant attachment. These differential pathways may be due to the way in which emotional abuse and neglect affect the individual. While emotional abuse may negatively affect the child’s integrity and sense of worth, child neglect may negatively affect the child’s perspective of others as being unsupportive and unaccountable (Caldwell et al., 2011). These two biases toward the self versus others are reflected in the respective insecure adult attachment styles. Individuals with an anxious attachment are often anxious about losing their partner due to feelings of being unworthy of love (Brennan et al., 1998; Fraley & Shaver, 2000). Whereas individuals with an avoidant attachment may struggle to trust in or rely on other people and often keep their distance (Brennan et al., 1998; Fraley & Shaver, 2000).

Furthermore, the applied analysis method facilitated an examination of more complex pathways in the data. Results showed that emotional abuse may affect avoidant attachment via emotional neglect, with avoidant attachment also found to be linked to anxious attachment. This could indicate that individuals with an avoidant attachment may also possess a biased representation of the self, which may enhance their vulnerability for anxious attachment. Additionally, the result that avoidant attachment predicted anxious attachment, rather than the other way round, could indicate that a biased self-perception and being dependent on others is more pervasive than a biased perspective of others and strong self-reliance. One mechanism that may explain why avoidant attachment was linked to anxious attachment could be that a strong self-reliance has limits in the demands of adult daily life, which could contribute to the emergence of internalized feelings of worthlessness originating from the child abuse (English et al., 2005; Glaser, 2002).

Finally, it was hypothesized that sexual and physical abuse as well as physical neglect are indirectly related to insecure adult attachment via emotional maltreatment. The results confirmed that child physical and sexual abuse were not directly related to insecure adult attachment, and a pathway was estimated via physical and emotional abuse. This pathway may indicate a bottom-up process: Sexual abuse may be viewed as a specific type of physical abuse, thereby contributing to reports of physical abuse. Subsequently, physical abuse may contribute to an individual’s overall feeling of being emotionally abused (Glaser, 2002). Hence, each child maltreatment type might have direct detrimental effects in adulthood, and in the case of adult attachment, emotional abuse appears to play a primary role, through which other child abuse types may exert an effect. It is worth noting that physical neglect was not linked to insecure adult attachment (neither directly, nor indirectly). While its general relationship with emotional neglect was sufficiently stable to be included in the final model, the direction of this pathway was among the lowest (estimated in 63% out of all bootstrapped samples). This might indicate that physical and emotional neglect share a bidirectional rather than unidirectional relationship. In summary, the results provide further evidence for the crucial role of child emotional maltreatment in the relationship between child maltreatment and insecure attachment in adulthood, and that the effects of other types of maltreatment on adult attachment seem to be mediated via emotional maltreatment. Additionally, the study further supports the differential impact of emotional abuse and neglect on insecure adult attachment.

Differential Impact of Psychological and Social Resources

Due to the expected differential pathways from child maltreatment types to insecure adult attachment styles, it was hypothesized that risk-specific resources are needed for affected adults to show resilience. Hence, the role of a psychological (i.e., self-esteem) and social (perceived social support) resource was investigated. As expected, self-esteem was only significant in the relationship between emotional abuse and anxious attachment (i.e., self-esteem mediated the effect of emotional abuse on anxious attachment: Emotional abuse was linked to lower self-esteem, which in turn was linked to anxious attachment). Whereas perceived social support was only significant in the relationship between emotional neglect and avoidant attachment (i.e., self-esteem mediated the effect of emotional neglect on avoidant attachment: Emotional neglect was linked to lower perceived social support, which in turn was linked to avoidant attachment). This finding suggests that a biased self-perception may be linked to emotional abuse, while a biased perception of others may be linked to emotional neglect. The current explorative, data-driven analysis did not find any other explaining effects of the resources on the relationship between child maltreatment types and adult attachment. This suggests that resilience against insecure adult attachment in the aftermath of specific child maltreatment types may be based on risk-specific resources, as predicted by differential impact and susceptibility theory (Ungar, 2018).

Furthermore, it was hypothesized that social support and self-esteem would be related to each other (Höltge et al., 2020, 2021). Examining more complex patterns in the data showed that social support had an indirect, resilience-supporting effect on anxious attachment via self-esteem. This is consistent with previous research that found that perceived social support can increase self-esteem by providing individuals with the feeling of being loved and of worth to others (Goodwin et al., 2004). The estimated direction of this effect is in line with observations that resilience against more severe forms of adversity, such as child maltreatment, can be linked to the availability of meaningful external resources (Ungar, 2018). Thus, in the context of child maltreatment, the relationship between self-esteem and perceived social support may differ from that found in studies without this adverse context (e.g., self-esteem affecting perceived social support; Marshall et al., 2014).

Taken together, the current findings suggest that while self-esteem was only significant in the relationship between emotional abuse and anxious attachment; perceived social support was important for individuals with experiences of emotional neglect and avoidant attachment, as well as indirectly for emotional abuse and anxious attachment because it was positively linked to self-esteem. Hence, in line with resilience theory (Ungar & Theron, 2020), the results show risk-specific effects of the studied resources as well as the interaction between psychological and social resources. Also, these results give further evidence that the relationship between emotional abuse and anxious attachment is primarily characterized by a biased representation of the self, and the relationship between emotional neglect and avoidant attachment by a biased perception of others. It is worth noting that the stability of the directed pathways between perceived social support and avoidant attachment was the lowest observed, as it was estimated in only 56% of the bootstrapped samples. In comparison to the rather stable pathway between self-esteem and anxious attachment (estimated in 83% of the bootstrapped samples), this finding might indicate that avoidant attachment could also reinforce low perceived social support, while anxious attachment may not further decrease self-esteem. However, given that this is just speculative based on cross-sectional observations, these relationships should be further examined in future longitudinal studies.

Practical Implications

The study provides empirical evidence that experiences of child emotional abuse and neglect are particularly important in relation to insecure adult attachment. Interventions that aim to support adults with insecure attachment should be particularly aware of the potential primary roles played by child emotional abuse for anxious attachment and child emotional neglect for avoidant attachment. Furthermore, as results suggest that avoidant attachment may be linked to higher expression of anxious attachment, rather than the other way round, it may be important for practitioners to consider interventions that stabilize and increase the patient’s self-worth, particularly for those who primarily experienced child neglect. With regard to resources, individuals who experienced emotionally neglect may need to address the way in which they perceive the available support from others, particularly from their partner, in the context of avoidant adult attachment. Although this study did not assess social network size, avoidance has been found to lead to a limited social network (Fiori et al., 2011; Gillath et al., 2017). Therefore, increasing the social network may also be a useful target for intervention, as increases in perceived social support may also increase the self-esteem of individuals who experienced emotional abuse. It may also be beneficial for practitioners who work with adults with insecure attachment styles to initially screen for different experiences of child maltreatment, as well as the status of self-esteem and perceived social support.

Limitations and Future Research

As this study applied a Bayesian network analysis based on observational, cross-sectional data, it does not allow for the analysis of feedback loops and limits the validity of the estimated directed pathways. Therefore, some of the estimated directed associations between model variables (such as emotional—physical neglect, sexual—physical abuse, and perceived social support—avoidant adult attachment), may in fact be of a bidirectional nature that can only be estimated using longitudinal data (Borsboom et al., 2013; Epskamp et al., 2018). Also, future studies may benefit from more contextualized measures of resources. While the measure for adult attachment focused on romantic relationships, a general measure of perceived social support was applied, which did not clearly focus on any specific relationship. Therefore, future studies should apply a measure of social support that focuses on the attachment relationship in question. This measure should also be accompanied by an objective measure of social support from the partner in order to investigate if the insecurely attached individual has a bias toward others or if they are actually receiving less social support. Additionally, differential impact theory (Ungar, 2018) and the multisystemic theory of resilience (Ungar & Theron, 2020) propose that resilience in the face of higher levels of child maltreatment arises from a diversified and functional resource network. Therefore, future studies should investigate the influence of present or absent resources from different systems (e.g., physiological, psychological, social, ecological) in the relationships between child maltreatment and insecure adult attachment. Furthermore, the data for this study was from a research project with a distinct population, that is, ex-members of fundamentalist Christian faith communities, which limits its generalizability to other populations. Hence, the results may indicate that child emotional abuse and neglect are linked to insecure adult attachment specifically in this population. Nevertheless, this study replicated effects that were previously found in other populations (e.g., Caldwell et al., 2011; Unger et al., 2014). Future studies are needed with other defined subpopulations to compare these processes across contexts in order to identify potentially generalizable associations (Molenaar & Campbell, 2009). Also, this study did not include attention checks in the survey, which could have helped to indicate the quality of individual responses. To improve such survey research, future studies should include attention checks. Additionally, although the studied sample consisted of participants from three German-speaking countries, the majority of the sample were residents in Germany. Hence, future studies should strive for a more balanced sample when assessing multicountry data. As this was also a secondary data analysis, studies are needed that are specifically designed to research the effects of child maltreatment types on adult attachment in different populations to investigate the generalizability of the identified pathways. Also, the applied child maltreatment measure only assessed the frequency of five types of maltreatment. Future studies should include a larger variety of indicators of child maltreatment, such as additional types of child maltreatment (e.g., domestic violence, parental substance abuse, or mental illness), the relationship to and characteristics of the perpetrator, the duration and severity of the maltreatment, and age at occurrence (see e.g., Mathews et al., 2020). Additionally, future studies should also take into account other types of assault in adulthood. For instance, insecure adult attachment styles have been linked to intimate partner violence in adulthood, which may reinforce insecure adult attachment (Bonache et al., 2019). Finally, to prevent missing values in future studies, several measures could be taken: (1) Online questionnaires can be programmed so that participants must answer all questions in order to proceed; (2) a pilot study can be conducted to identify items that may not be missing at random; or (3) the study team could consult with representatives of the target population to evaluate the comprehensibility and appropriateness of the survey items. Related to this, the randomization of the survey scales may have led to a certain order in some cases, which may have increased the likelihood of missing values or response patterns that could influence the results. Pilot studies could be used in future studies to identify a contextually appropriate order of scales for the target population (Brookes et al., 2018).

Conclusion

Different child maltreatment types can be linked to different styles of insecure adult attachment. Of the child maltreatment types that were investigated in this study, emotional abuse and neglect showed the only direct relationships to anxious and avoidant adult attachment, respectively. Sexual abuse and physical abuse appeared to exert their effect on anxious attachment via emotional abuse. In line with these differential pathways, self-esteem and perceived social support showed risk-specific effects, with self-esteem being important for resilience to emotional abuse, and social support being important for resilience to emotional neglect. Nevertheless, social support was also linked to self-esteem and is therefore crucial for individuals with experiences of emotional abuse. This study suggests that it is important to study the specific effects of child maltreatment types on insecure adult attachment. The findings also have implications for future intervention research, indicating that a focus should be placed on particular protective resources that correspond to these risk-specific patterns.

Footnotes

Acknowledgements

The authors wish to thank all participants who took part in this study. The authors also extend their gratitude to all project members who were involved in various stages of the project (listed in alphabetical order by last name): Keegan W. M. Evangelista, Andreas Goreis, Rahel M. Hermann, Nina Keller, Emily Kulowatz, Andreas Maercker, Melanie Roos, Valentina Schmid, and Vanja Sumi. We thank Linda Theron for her valuable feedback on our manuscript.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interests with respect to the authorship and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research and/or authorship of this article: JH position was funded by the Swiss National Science Foundation (grant number P400PS_194538).

ORCID iDs

Jan Höltge

Shauna Rohner

Eva Maria Heim

Myriam Thoma

Author Biographies

Jan Höltge, PhD, is a visiting Postdoctoral Researcher at the University of Hawai’i at Mānoa funded by the Swiss National Science Foundation, and further affiliated with the Resilience Research Centre at Dalhousie University. His research focuses on multisystemic resilience and contextualized resilience practices of different cultural groups. He uses mixed-methods designs and statistical methods that can investigate complex patterns of resilience.

Shauna L. Rohner, PhD, is a Health Psychologist (FSP) and Postdoctoral Researcher at the Division of Psychopathology and Clinical Intervention, in the Psychological Institute, University of Zurich. Her current research focuses on experiences of trauma, stress, resilience, and adaptive development in later life. She is further interested in research on the psychological, biological, social, and environmental aspects of health and aging.

Eva M. Heim, PhD, is an Associate Professor at the Institute of Psychology, University of Lausanne, Switzerland. Her research focuses on cultural aspects in the phenomenology and etiology of mental disorders, and the cultural adaptation of psychological interventions for diverse cultural groups.

Urs Nater is a Full Professor of Clinical Psychology at the University of Vienna. His research focuses on identifying the underlying mechanisms that translate stress into illness. He is the Director of the Stress Lab and of the Music & Health Lab at the Faculty of Psychology and currently serves as Head of the Institute of Clinical and Health Psychology.

Myriam V. Thoma, PhD, is a senior research and teaching associate at the University of Zurich (UZH), Switzerland. Additionally, she works as a coordinator and psychotherapist at the psychotherapeutic outpatient center of the psychological institute of the UZH. Her research focuses on psychopathology, mental health, stress, and resilience across the lifespan, as well as clinical ecopsychology.

References

Arslan

(2016). Psychological maltreatment, emotional and behavioral problems in adolescents: The mediating role of resilience and self-esteem. Child Abuse & Neglect, 52, 200–209. https://doi.org/10.1016/j.chiabu.2015.09.010

Bernstein

D. P.

Fink

Handelsman

Foote

Lovejoy

Wenzel

Sapareto

Ruggiero

(1994). Initial reliability and validity of a new retrospective measure of child abuse and neglect. The American Journal of Psychiatry, 151(8), 1132–1136. https://doi.org/10.1176/ajp.151.8.1132

Bonache

Gonzalez-Mendez

Krahé

(2019). Adult attachment styles, destructive conflict resolution, and the experience of intimate partner violence. Journal of Interpersonal Violence, 34(2), 287–309. https://doi.org/10.1177/0886260516640776

Borsboom

Cramer

A. O

. (2013). Network analysis: An integrative approach to the structure of psychopathology. Annual Review of Clinical Psychology, 9, 91–121. https://doi.org/10.1146/annurev-clinpsy-050212-185608

Brennan

K. A.

Clark

C. L.

Shaver

P. R.

(1998). Self-report measurement of adult attachment. In Simpson

J. A.

Rholes

W. S.

(Eds.), Attachment theory and close relationships (pp. 46–76). The Guilford Press.

Brookes

S. T.

Chalmers

K. A.

Avery

K. N.

Coulman

Blazeby

J. M.

(2018). Impact of question order on prioritisation of outcomes in the development of a core outcome set: A randomised controlled trial. Trials, 19(1), 1–11. https://doi.org/10.1186/s13063-017-2405-6

Caldwell

J. G.

Shaver

P. R.

C. S.

Minzenberg

M. J.

(2011). Childhood maltreatment, adult attachment, and depression as predictors of parental self-efficacy in at-risk mothers. Journal of Aggression, Maltreatment & Trauma, 20(6), 595–616. https://doi.org/10.1080/10926771.2011.595763

Cicchetti

Toth

S. L.

(2005). Child maltreatment. Annual Review of Clinical Psychology, 1, 409–438. https://doi.org/10.1146/annurev.clinpsy.1.102803.144029

Cobb

(1976). Social support as a moderator of life stress. Psychosomatic Medicine, 38, 300–314. https://doi.org/10.1097/00006842-197609000-00003

10.

Cyr

Euser

E. M.

Bakermans-Kranenburg

M. J.

Van Ijzendoorn

M. H.

(2010). Attachment security and disorganization in maltreating and high-risk families: A series of meta-analyses. Development and Psychopathology, 22(1), 87–108. https://doi.org/10.1017/S0954579409990289

11.

English

D. J.

Bangdiwala

S. I.

Runyan

D. K.

(2005). The dimensions of maltreatment: Introduction. Child Abuse & Neglect, 29(5), 441–460. https://doi.org/10.1016/j.chiabu.2003.09.023

12.

Epskamp

van Borkulo

C. D.

van der Veen

D. C.

Servaas

M. N.

Isvoranu

A. M.

Riese

Cramer

A. O. J.

(2018). Personalized network modeling in psychopathology: The importance of contemporaneous and temporal connections. Clinical Psychological Science, 6, 416–427. https://doi.org/10.1177/2167702617744325

13.

Espeleta

H. C.

Palasciano-Barton

Messman-Moore

T. L.

(2017). The impact of child abuse severity on adult attachment anxiety and avoidance in college women: The role of emotion dysregulation. Journal of Family Violence, 32(4), 399–407. https://doi.org/10.1007/s10896-016-9816-0

14.

Finkelhor

Vanderminden

Turner

Hamby

Shattuck

(2014). Child maltreatment rates assessed in a national household survey of caregivers and youth. Child Abuse & Neglect, 38(9), 1421–1435. https://doi.org/10.1016/j.chiabu.2014.05.005

15.

Fiori

K. L.

Consedine

N. S.

Merz

E. M.

(2011). Attachment, social network size, and patterns of social exchange in later life. Research on Aging, 33(4), 465–493. https://doi.org/10.1177/0164027511401038

16.

Fraley

R. C.

Roisman

G. I.

(2019). The development of adult attachment styles: Four lessons. Current Opinion in Psychology, 25, 26–30. https://doi.org/10.1016/j.copsyc.2018.02.008

17.

Fraley

R. C.

Shaver

P. R.

(2000). Adult romantic attachment: Theoretical developments, emerging controversies, and unanswered questions. Review of General Psychology, 4(2), 132–154. https://doi.org/10.1037/1089-2680.4.2.132

18.

Fritz

Fried

E. I.

Goodyer

I. M.

Wilkinson

P. O.

van Harmelen

A. L.

(2018). A network model of resilience factors for adolescents with and without exposure to childhood adversity. Scientific Reports, 8, 15774. https://doi.org/10.1038/s41598-018-34130-2

19.

Fydrich

Sommer

Tydecks

Brähler

(2009). Fragebogen zur sozialen Unterstützung (F-SozU): Normierung der Kurzform (K-14) [Social Support Questionnaire (F-SozU): Standardization of short form (K-14)]. Zeitschrift für Medizinische Psychologie, 18(1), 43–48.

20.

Gillath

Karantzas

G. C.

Selcuk

(2017). A net of friends: Investigating friendship by integrating attachment theory and social network analysis. Personality and Social Psychology Bulletin, 43(11), 1546–1565. https://doi.org/10.1177/0146167217719731

21.

Glaser

(2002). Emotional abuse and neglect (psychological maltreatment): A conceptual framework. Child Abuse & Neglect, 26(6–7), 697–714. https://doi.org/10.1016/S0145-2134(02)00342-3

22.

Goodwin

Costa

Adonu

(2004). Social support and its consequences: ‘Positive’ and ‘deficiency’ values and their implications for support and self-esteem. British Journal of Social Psychology, 43(3), 465–474. https://doi.org/10.1348/0144666042038006

23.

Höltge

Theron

Cowden

R.G.

Govender

Maximo

S.I.

Carranza

J.S.

Kapoor

Tomar

van Rensburg

Cavioni

Agliati

Grazzani

Smedema

Kaur

Hurlington

Sanders

Munford

Ungar

(2020) A cross-country network analysis of adolescent resilience. Journal of Adolescent Health, 68(3), 580–588. https://doi.org/10.1016/j.jadohealth.2020.07.010

24.

Höltge

Theron

Ungar

(2021). A multisystemic perspective on the temporal interplay between adolescent depression and resilience-supporting individual and social resources. Journal of Affective Disorders, 297, 225–232. https://doi.org/10.1016/j.jad.2021.10.030

25.

Hurren

Thompson

Jenkins

Chrzanowski

Allard

Stewart

(2018). Who are the perpetrators of child maltreatment? Criminology Research Advisory Council. https://www.aic.gov.au/sites/default/files/2020-05/18-1314-FinalReport.pdf

26.

Jaffee

S. R.

(2017). Child maltreatment and risk for psychopathology in childhood and adulthood. Annual Review of Clinical Psychology, 13, 525–551. https://doi.org/10.1146/annurev-clinpsy-032816-045005

27.

Kalisch

Bühlman

(2007). Estimating high-dimensional directed acyclic graphs with the PC-algorithm. Journal of Machine Learning Research, 8(3), 613–636.

28.

Kalisch

Baker

D. G.

Basten

Boks

M. P.

Bonanno

G. A.

Brummelman

Chmitorz

Fernàndez

Fiebach

C. J.

Galatzer-Levy

Geuze

Groppa

Helmreich

Hendler

Hermans

E. J.

Jovanovic

Kubiak

Lieb

Lutz

Kleim

(2017). The resilience framework as a strategy to combat stress-related disorders. Nature Human Behaviour, 1(11), 784–790. https://doi.org/10.1038/s41562-017-0200-8

29.

C. K.

Chan

K. L.

(2019). Insecure adult attachment and child maltreatment: A meta-analysis. Trauma, Violence, & Abuse, 20(5), 706–719. https://doi.org/10.1177/1524838017730579

30.

Maercker

Bernays

Rohner

S. L.

Thoma

M. V.

(2022). A cascade model of complex posttraumatic stress disorder centered on childhood trauma and maltreatment, attachment, and socio-interpersonal factors. Journal of Traumatic Stress, 35(2), 446–460. https://doi.org/10.1002/jts.22756

31.

Marshall

S. L.

Parker

P. D.

Ciarrochi

Heaven

P. C.

(2014). Is self-esteem a cause or consequence of social support? A 4-year longitudinal study. Child Development, 85(3), 1275–1291. https://doi.org/10.1111/cdev.12176

32.

Mathews

Pacella

Dunne

M. P.

Simunovic

Marston

(2020). Improving measurement of child abuse and neglect: A systematic review and analysis of national prevalence studies. PLoS One, 15(1), e0227884. https://doi.org/10.1371/journal.pone.0227884

33.

McNally

R. J.

Mair

Mugno

B. L.

Riemann

B. C.

(2017). Co-morbid obsessive–compulsive disorder and depression: A Bayesian network approach. Psychological Medicine, 47(7), 1204–1214. https://doi.org/10.1017/S0033291716003287

34.

Mikulincer

Shaver

P. R.

(2005). Attachment theory and emotions in close relationships: Exploring the attachment-related dynamics of emotional reactions to relational events. Personal Relationships, 12(2), 149–168. https://doi.org/10.1111/j.1350-4126.2005.00108.x

35.

Molenaar

P. C.

Campbell

C. G.

(2009). The new person-specific paradigm in psychology. Current Directions in Psychological Science, 18(2), 112–117. https://doi.org/10.1111/j.1467-8721.2009.01619.x

36.

Nagarajan

Scutari

Lèbre

(2013). Bayesian networks in R with applications in systems biology. Springer.

37.

Oshri

Sutton

T. E.

Clay-Warner

Miller

J. D.

(2015). Child maltreatment types and risk behaviors: Associations with attachment style and emotion regulation dimensions. Personality and Individual Differences, 73, 127–133. https://doi.org/10.1016/j.paid.2014.09.015

38.

Reyome

N. D.

(2010). The effect of childhood emotional maltreatment on the emerging attachment system and later intimate relationships. Journal of Aggression, Maltreatment & Trauma, 19(1), 1–4. https://doi.org/10.1080/10926770903486007

39.

Rosenberg

(1979). Conceiving the Self. Basic Books.

40.

Scutari

(2010). Learning Bayesian Networks with the bnlearn R Package. Journal of Statistical Software, 35(3), 1–22. https://doi.org/10.18637/jss.v035.i03

41.

Scutari

Nagarajan

(2013). Identifying significant edges in graphical models of molecular networks. Artificial Intelligence in Medicine, 57(3), 207–217. https://doi.org/10.1016/j.artmed.2012.12.006

42.

Scutari

Graafland

C. E.

Gutiérrez

J. M.

(2018). Who learns better Bayesian network structures: Constraint-based, score-based or hybrid algorithms? Proceedings of Machine Learning Research, 72, 416–427.

43.

Seeds

P. M.

Harkness

K. L.

Quilty

L. C.

(2010). Parental maltreatment, bullying, and adolescent depression: Evidence for the mediating role of perceived social support. Journal of Clinical Child & Adolescent Psychology, 39(5), 681–692. https://doi.org/10.1080/15374416.2010.501289

44.

Thoma

M. V.

Bernays

Eising

C. M.

Maercker

Rohner

S. L.

(2021). Child maltreatment, lifetime trauma, and mental health in Swiss older survivors of enforced child welfare practices: Investigating the mediating role of self-esteem and self-compassion. Child Abuse & Neglect, 113, 104925. https://doi.org/10.1016/j.chiabu.2020.104925

45.

Thoma

M. V.

Höltge

Eising

C. M.

Pfluger

Rohner

S. L.

(2020). Resilience and stress in later life: A network analysis approach depicting complex interactions of resilience resources and stress-related risk factors in older adults. Frontiers in Behavioral Neuroscience, 14, 580969. https://doi.org/10.3389/fnbeh.2020.580969

46.

Tsamardinos

Brown

L. E.

Aliferis

C. F.

(2006). The max-min hill-climbing Bayesian network structure learning algorithm. Machine Learning, 65(1), 31–78. https://doi.org/10.1007/s10994-006-6889-7

47.

Ungar

(2018). The differential impact of social services on young people’s resilience. Child Abuse & Neglect, 78, 4–12. https://doi.org/10.1016/j.chiabu.2017.09.024

48.

Ungar

Theron

(2020). Resilience and mental health: How multisystemic processes contribute to positive outcomes. The Lancet Psychiatry, 7(5), 441–448. https://doi.org/10.1016/S2215-0366(19)30434-1

49.

Unger

J. A. M.

De Luca

R. V.

(2014). The relationship between childhood physical abuse and adult attachment styles. Journal of Family Violence, 29(3), 223–234. https://doi.org/10.1007/s10896-014-9588-3

50.

Walker

J. L.

Carey

P. D.

Mohr

Stein

D. J.

Seedat

(2004). Gender differences in the prevalence of childhood sexual abuse and in the development of pediatric PTSD. Archives of Women’s Mental Health, 7(2), 111-121. https://doi.org/10.1007/s00737-003-0039-z

51.

Widom

C. S.

Czaja

Dutton

M. A.

(2014). Child abuse and neglect and intimate partner violence victimization and perpetration: A prospective investigation. Child Abuse & Neglect, 38(4), 650–663. https://doi.org/10.1016/j.chiabu.2013.11.004

52.

Widom

C. S.

Czaja

S. J.

Kozakowski

S. S.

Chauhan

(2018). Does adult attachment style mediate the relationship between childhood maltreatment and mental and physical health outcomes? Child Abuse & Neglect, 76, 533–545. https://doi.org/10.1016/j.chiabu.2017.05.002

53.

Cao

Lin

Zhou

Chi

(2021). Child maltreatment and subjective well-being in Chinese emerging adults: A process model involving self-esteem and self-compassion. Journal of Interpersonal Violence, 37(15–16), NP13685–NP13706 0886260521993924. https://doi.org/10.1177/0886260521993924