The Development of the Attachment Defenses Questionnaire (ADQ-50): A Preliminary Examination of Reliability, Validity, and Factor Structure

Abstract

This paper marks the initial phase in the development of the Attachment Defenses Questionnaire (ADQ-50), a self-report tool crafted to assess defense mechanisms associated with attachment processes, catering to both clinical and research contexts. Anchored in the theoretical framework of attachment theory, the ADQ posits that an individual’s internalized attachment style plays a influential role in predicting their defense mechanisms. The paper outlines the comprehensive development and refinement process of the ADQ-50. In Study 1 a preliminary 176-item version of the ADQ was examined. Data was collected online drawing from participants sourced from Prolific and undergraduate students (N = 1994). Study 2 further refined the ADQ, evaluating its initial convergent validity with a diverse participant pool (N = 726), including undergraduates, Prolific contributors, general practice medical patients, and individuals from social media. Exploratory factor analysis revealed a robust ten-factor structure, resulting in a 50-item scale aligning with theoretical expectations and demonstrating good psychometric properties. Findings, limitations, strengths and future research directions are discussed. We posit that the ADQ holds great potential to deepen our comprehension of defense mechanisms linked to attachment, with wide-ranging implications for clinical practices.

Attachment theory (Bowlby, 1969, 1980) proposes that the interactions between caregivers and infants, along with the environmental context provided by the caregiver, lay the foundation for relational security or insecurity. Since its inception, attachment theory has found extensive application beyond child development, specifically in understanding adult emotional regulation, personality pathology, vulnerability, resilience, and the dynamics of relational and defensive processes (Darling Rasmussen et al., 2019; Hazan & Shaver, 1994; Lyddon & Sherry, 2001; Mikulincer & Shaver, 2019; Morley & Moran, 2011; Richardson et al., 2022).

Adult attachment insecurity is commonly understood within the framework of a two-dimensional model (Brennan et al., 1998). The two dimensions, attachment anxiety and attachment avoidance, entail stable patterns of thoughts, feelings, and behaviors in relationships (Fraley, 2019). Individuals need not be exclusively anxiously-attached or avoidantly-attached and can have high levels in both dimensions (Hlay et al., 2023). The dimensions are often correlated and better viewed as oblique than orthogonal (Cameron et al., 2012).

In the absence of secure attachment, insecure attachment gives rise to defense mechanisms acting as secondary strategies for dealing with distress (Bowlby, 1980; Main, 1990; Mikulincer et al., 2009; Mikulincer & Shaver, 2016). Broadly defined, psychological defenses are automatic stress-response mechanisms associated with personality functioning addressing conflict and cognitive dissonance, (Cramer, 2015; Freud, 1894, 1966; Silverman & Aafjes-van Doorn, 2023; Vaillant, 2020). Mikulincer and Shaver (2016) classify attachment-related defenses into hyperactivating mechanisms (associated with attachment anxiety) involving attention-seeking behaviors, and deactivating mechanisms (linked to attachment avoidance) involving distancing strategies and suppressed emotion. Building on this, Richardson et al. (2022) propose that the distinct defensive patterns associated with the two dimensions of insecure attachment, offers a promising framework for organizing psychological defenses in scale development.

Measuring psychological defenses

The challenges associated with measuring psychological defenses, especially through self-report tools, are widely recognized (Chabrol et al., 2005; Davidson & MacGregor, 1998; Mehlman & Slane, 1994; Mihalits & Codenotti, 2020; Ruuttu et al., 2006; Thygesen et al., 2008; Wilkinson & Ritchie, 2015). Issues include inconsistency between definitions across scales and lack of construct validity (Mehlman & Slane, 1994), along with psychometric problems (Wilkinson & Ritchie, 2015; Ruuttu et al., 2006 for DSQ-40). Criticism has been directed toward the absence of an appropriate theoretical model for guiding defense mechanisms scale development (Mihalits & Codenotti, 2020). Consequently, there is a pressing need for a more robust self-report scale to measure defense mechanisms, one that is firmly grounded in a strong theoretical foundation.

The recent DMRS-SR-30 defense scale demonstrates the potential for adequate validity (Prout et al., 2022). However, the DMRS system (Di Giuseppe et al., 2020; Perry & Henry, 2004) determines the function of a defense by its presentation rather than underlying meaning and antecedents. This top-down approach thus derives patterns from observations, rather than starting with preexisting theory. Although there is merit to basing scale development on empirical observations, there are nevertheless a number of benefits from employing theory-driven scale development (Boateng et al., 2018; Carpenter, 2018; Finch, 2020; Morgado et al., 2017). A strong theoretical and conceptual foundation is important for making specific predictions and for ensuring that the chosen solution is conceptually meaningful.

Considering this, attachment theory, with its extensive empirical foundation (Cassidy & Shaver, 2016; Holmes, 2014; Simpson & Rholes, 2015), and established association with defenses (Mikulincer & Shaver, 2016) potentially provides a robust theoretical basis for creating a comprehensive psychological defenses scale. Existing attachment assessment tools, such as the adult attachment interview (AAI; George et al., 1996; Hesse, 2016) and the adult attachment projective picture system (AAP; George & West, 2012), recognize the interplay between attachment and defensive processes, incorporating them into their coding systems. However, this integration has yet to be attempted with self-report measurement. While there are a number of self-report scales already in existence to assess general defense mechanisms, a self-report instrument dedicated to attachment-related defensive processes is yet to be explored. This paper reports the development of a new self-report measure of defense based on attachment theory, the Attachment Defenses Questionnaire (ADQ).

The theoretical framework of the Attachment Defenses Questionnaire (ADQ)

The theoretical foundation of the ADQ is based in Mikulincer and Shaver (2016) framework, attachment learning theory (Bosmans et al., 2020), and social defense theory (Ein-Dor et al., 2010; Ein-Dor & Hirschberger, 2016, 2018), and draws from the theoretical and empirical research on attachment-related defense mechanisms by Richardson and Boag (under review) and Richardson et al. (2022).

Attachment learning theory (Bosmans et al., 2020) proposes that inconsistent positive rewards for seeking support and a lack of reward for exploration contribute to attachment anxiety. The unpredictable nature of intermittent positive reinforcement may paradoxically teach children to maintain high distress levels, fostering chronic attachment system activation. Conversely, consistent negative or indifferent responses to attempts to seek support may lead individuals to avoid using support-seeking behavior when distressed. The absence of rewards following support-seeking contributes to attachment avoidance when caregivers consistently dismiss or reject the child’s needs.

Building on this understanding, the current research explores the defense mechanisms associated with the two dimensions of attachment insecurity. For more theoretical context, please consult Richardson and Boag (2024/under review). The following section consolidates the literature that forms the basis of the ADQ, offering insights into the constructs we aim to measure. Table 1 provides an overview of these constructs.

Table 1. Construct definitions for the preliminary ADQ.

Emotional Amplification	Emotional amplification, a defense mechanism linked to an overactive attachment system, involves exaggerating emotional signals. It serves as both a communication tool to express needs and a method to seek support, addressing unmet attachment needs and eliciting the caregiving response that may have been lacking at times.
Chronic Proximity Seeking	Chronic proximity seeking manifests as an obsessive urge for interpersonal contact and meaningful connection. These hyperactive behaviors simulate control and proactive steps toward intimacy. Simultaneously, these actions ease distress linked to perceived separation or emotional vulnerability from past experiences.
Defensive Reenactment	Reenactment, stemming from attachment insecurity, unconsciously prompts the compulsion to rewrite personal history for a more favorable outcome. It involves repeating past maladaptive behaviors in an attempt to gain a sense of mastery.
Hypervigilance	Hypervigilance with attachment insecurity involves heightened awareness of interpersonal cues, where individuals anticipate and respond to perceived inadequacies in others’ interest, availability, or responsiveness. This hyperactive defense mechanism helps in coping with situations that trigger attachment-related fears.
Rumination	Rumination involves repetitive thoughts on distressing emotions, often driven by perfectionism as a hyperactive defense mechanism. It serves to occupy anxiety, analyze, and prevent future mistakes. In attachment insecurity, relationship-focused rumination may create a false sense of closeness with someone actively on one’s mind.
Compulsive caregiving	Compulsive caregiving is an intense focus on others’ well-being, driven by a hyperactive attachment pattern. It serves as a defense mechanism to fulfill unmet needs for closeness, acceptance, and inclusion. This may result from past inadequate caregiving or external pressures, like caring for a sick, anxious, or hypochondriac parent..
Consumption	Consumption, an interpersonal dependency defense, drains others’ energy due to weak self-other boundaries, tied to projective identification and challenges in self-other differentiation. It conveys a longing for empathy, expressing unmet attachment needs, and manifests as cathartic hyperactive communication.
Splitting	Splitting can be understood as cognitive disconnection in the context of intolerable conflicting emotions. The process involves the emergence of polarized views of self and others. As a defense, splitting allows individuals to cope with ambivalence
Protest	Protest behaviors in the context of attachment-related defensive processes are directed toward resisting separation, eliciting a reaction, and expressing retributive anger. Such hyperactive conduct may encompass tantrums, attention-seeking strategies, or feigned reaction-seeking threats.
Compulsive Self-Reliance	Compulsive self-reliance, marked by a commitment to independent behavior and reluctance to accept help perceived as threatening or unfamiliar serves as a deactivating defense. It maintains a suppressed attachment system, rooted in pervasive distrust, enabling individuals to establish self-direction and control. This behavior emerges from the necessity to self-parent or repeatedly navigate circumstances demanding self-reliance.
Self-interest	Self-interest functions as a deactivating defense mechanism characterized by detachment and the devaluation of others within the context of attachment avoidance. Individuals with attachment avoidance are primarily driven by the motivation to protect themselves. This self-interested strategy proves advantageous during times of danger, enabling a swift response to perceived threats and the assertion of control over the surrounding environment.
Distancing	Distancing functions as a deactivating defense mechanism, entailing the control of personal space and sensitivity to physical and emotional boundaries. Rooted in early attachment experiences, it arises from adverse outcomes in seeking caregiver support. This mechanism preserves individual identity by restricting closeness with others, acting as a shield against confronting beliefs of rejection.
Emotional Suppression	Emotional suppression, as an attachment-related defense mechanism is a strategy to disconnect from feelings associated with attachment disappointment. This behavior acts as a protective measure, serving as an avoidance tactic to guard against perceived threats, rejection, or emotional pain.
Defensive exclusion	Exclusion, an attachment defense, blocks awareness of ideas activating the attachment system. It shields against vulnerability and beliefs of unworthiness and by keeping uncomfortable thoughts out of consciousness, affecting memory, perception and attention.
Dismissal of Attachment Needs	The dismissal of attachment needs functions as a denial mechanism, shielding oneself from acknowledging unfulfilled early psychological needs for connection and guarding against feelings of rejection through cognitive processes such as minimization and rationalization.
Compensatory self-inflation	Compensatory self-inflation, a psychological defense, involves self-image enhancement as a protective measure against threats to self-esteem. Personal deficiencies are denied. It is proposed to correlate with vulnerable narcissism and is hypothesized to be associated with attachment avoidance.

The defenses mechanisms associated with a hyperactive attachment system

Attachment anxiety is associated with patterns of heightened emotional reactivity, distress amplification and emotion-focused coping, evidenced by exaggerated crying, and pain catastrophizing (Drenger et al., 2017; Kratz et al., 2012; Lopez, 2001; Mikulincer et al., 1993, 2009; Wei et al., 2005). As a hyperactive defense mechanism, emotional heightening (amplification) can be understood as a strategic means to attract support and a communicative tool drawing attention to needs. Magnifying emotional signals aims to evoke a caregiving response aligning with the intensified pursuit of connection associated with attachment anxiety (Konok et al., 2016; Mikulincer et al., 2000, 2002; Patton et al., 2010; Sharpsteen & Kirkpatrick, 1997).

Consequently, we view chronic proximity seeking acting as a defense mechanism representing a hyperactive drive for interpersonal contact and connection. Intense longing for closeness coupled with unresolved attachment experiences can potentially lead to repetition compulsion (Freud, 1914). Reenactment may also act as a hyperactive defense mechanism, signifying a return to the familiar, particularly in the context of trauma (Van der Kolk, 1989). There is evidence suggesting a connection between attachment anxiety and cyclic behaviors (Faber et al., 2018; Kuipers et al., 2016; Smith & South, 2020), and research has further identified attachment anxiety as a predictor of revictimization (Bockers et al., 2014).

Compulsive caregiving also conforms to the pattern of hyperactive attachment behavior (Braun et al., 2012; Millings & Walsh, 2009; Shaver et al., 2010). This behavior entails an intense preoccupation with others’ welfare, leading individuals to invest excessive attention in nurturing them (Bowlby, 1980). Mikulincer and Shaver (2016) suggest that compulsive caregiving, as a defense mechanism, satisfies unmet needs for closeness, acceptance, inclusion and sensitivity.

Hypervigilance can also be viewed as an attachment-related defense mechanism. Ein-Dor (2015) established a connection between the “sentinel schema”—characterized by heightened threat sensitivity, high stress reactivity, and constant vigilance—and attachment anxiety. Richardson et al. (2022) similarly report a connection between attachment anxiety and defensive anticipation. Mikulincer and Shaver (2016) further note that attachment-anxiety sensitivity is especially pronounced in response to cues of rejection or abandonment.

Hypervigilance aligns with over-thinking, and rumination stands out as another hyperactivating mechanism linked to attachment anxiety in the literature (Burnette et al., 2009; Caldwell & Shaver, 2012, 2013; Chung, 2014; Garrison et al., 2014; Joel et al., 2012; Saffrey & Ehrenberg, 2007). Rumination, viewed as a defense mechanism, entails preoccupation with anxiety and can serve the pursuit of perfectionism and preventing future mistakes (Senra et al., 2018).

Building on Mikulincer and Shaver’s insights (2016), emphasizing the “fight” aspect of the hyperactivating strategy inherent in adult attachment anxiety, we suggest that “protest” can be understood as a hyperactive attachment defense. Exemplifying this is the finding that attachment anxiety is associated with behaviors such as intense crying, manipulation, defiance, and passive aggression (Drenger et al., 2017; Gruneau Brulin et al., 2022; Laan et al., 2012; Overall et al., 2014; Richardson et al., 2022).

The defense identified as consumption (Daly & Mallinckrodt, 2009) characterizes behavior that depletes or demands others’ energy. It relates to the “psychic vampire” phenomenon (Petric, 2023) which may be due to poor self-other boundaries. We posit that the construct of consumption is theoretically intertwined with projective identification, both serving as communicative mechanisms and conduits for the solicitation of empathy (Roth, 2017). In projective identification, individuals project their own feelings onto others, distorting the perception of both self and others (Richardson et al., 2022). Projective identification and difficulties in self-other differentiation, are associated with attachment anxiety (Berant & Wald, 2009; Lopez, 2001; Mikulincer & Horesh, 1999; Richardson et al., 2022).

Reality distortion is also evident in the defense mechanism of splitting, where individuals perceive experiences or people in extremes, oversimplifying emotions and often fail to acknowledge nuanced possibilities. Fueled by conflicting emotions, splitting develops polarized views of oneself and others (Boag, 2017). Bowlby (1980) viewed splitting defenses as a result of cognitive disconnection due to attachment ambivalence. Considering the evidence indicating the association of splitting with both attachment anxiety and attachment avoidance (Lopez, 2001; Prunas et al., 2019; Richardson et al., 2022), splitting can be seen as a response to a hyperactivated attachment system because it involves the intensification of emotional reactions, exemplified by excessive idealization or devaluation.

The defenses mechanisms associated with a suppressed attachment system

Expanding upon Bowlby’s (1980) observations regarding deactivating defenses linked to attachment insecurity, compulsive self-reliance acts as an avoidant defense mechanism. Self-reliance functions as a strategy for stress management and control (Simpson & Rholes, 2017) and is evident in the negative attitude toward seeking help and the inclination to avoid support, characteristics associated with attachment avoidance (Gillath et al., 2006; Lopez et al., 1998; Vogel & Wei, 2005).

In addition to defensive self-reliance, self-interest may also act as a defense mechanism since compensatory self-interested behaviors, such as self-enhancement and the devaluation of others are associated with attachment avoidance (Berant et al., 2005; Ein-Dor, 2015; Mikulincer et al., 2009; Rom & Mikulincer, 2003). Social defense theory (Ein-Dor et al., 2010; Ein-Dor & Hirschberger, 2016, 2018) further agrees that behaviors associated with attachment avoidance, seen as “rapid responders,” may confer advantages in times of danger. These individuals prioritize self-protection, allowing them to swiftly and efficiently manage threats, potentially outperforming others in such situations. This self-serving defense strategy is potentially linked to vulnerable narcissism (Kaufman et al., 2020), and aligns with attachment avoidance beliefs that foster negative perceptions of others and heightened self-other differentiation (Bartholomew & Horowitz, 1991; Mikulincer et al., 1998).

Several studies (e.g., Akbarian et al., 2022; Fraley & Shaver, 1998; Kaitz et al., 2004) have also established a connection between distancing mechanisms and attachment avoidance. Feeney and Fitzgerald (2022) propose that defensive distance regulation entails a delicate balance between the desire for closeness and separateness, reflecting a tension between connection and autonomy. Defensive distancing can manifest in various ways, such as managing personal space, a reserved demeanor, detached/withdrawn behavior, avoiding eye contact and affection, restricting intimacy, deflective/evasive conversational tactics, and disconnecting from or denying bodily sensations (all of which are associated with attachment avoidance: Bartholomew & Horowitz, 1991; Chopik et al., 2014; Hesse, 2016; Overall et al., 2013; Richardson et al., 2022; Vieira et al., 2020).

Attachment avoidance is also notably associated with the mechanisms of emotional suppression where the range of emotional experience is actively limited (e.g. Drenger et al., 2017; Roisman et al., 2004; Wei et al., 2005; Winterheld, 2016). Furthermore research has identified attentional biases linked to attachment avoidance (Andriopoulos & Kafetsios, 2015; Edelstein & Gillath, 2008; Liu et al., 2017) indicating cognitive suppression or defensive exclusion (Bowlby, 1980)—an information processing approach to defense operating at the level of perception and attention.

While suppression is a deliberate effort to keep specific thoughts or emotions out of conscious awareness, denial is a form of defense that involves rejecting or minimizing the awareness of certain aspects (Costa, 2020). Both defenses play roles in managing psychological discomfort, and in the context of attachment-related defense, they may be employed to cope with the challenges and vulnerabilities associated with attachment needs. The dismissal of attachment needs is a denial mechanism corresponding to attachment avoidance (George et al., 1996; Hesse, 2016).

Research aims

This research aims to delineate attachment-related defensive processes by introducing a novel scale to operationalize the functioning of attachment defenses. Here we describe the development of the Attachment Defenses Questionnaire (ADQ) to address a significant gap in existing literature. Currently, there is a lack of a comprehensive self-report scale that integrates attachment theory and defense theory. Furthermore, no prior empirical study has systematically outlined attachment-related defense mechanisms specific to each attachment dimension. This study strives to fill these gaps and contribute to a deeper understanding of the intricate interplay between attachment and defense mechanisms. To achieve this, we examine the scale’s psychometric properties as a self-report instrument for assessing attachment-related defense mechanisms.

Our overarching goal is to advance the measurement of self-reported defense mechanisms by incorporating a robust theoretical foundation with strong psychometric properties and clinical relevance. By centering our defense model on attachment, we provide a distinctive perspective on perpetuating problem cycles, setting our approach apart from existing models.

We seek to derive a multifaceted structure from the preliminary ADQ data. While we have identified 16 constructs for examination and have developed corresponding items, the exploratory nature of our analysis allows for flexibility. Acknowledging potential variations or overlaps among these constructs, we aim to achieve a practical and meaningful structure through our analysis.

Additionally, we aim to establish convergent validity between the ADQ and insecure attachment by examining correlations between attachment-related defense mechanisms and insecure attachment. We anticipate that specific defenses will exhibit stronger associations with attachment anxiety, some with attachment avoidance, and others may demonstrate relevance to both dimensions (cf. Richardson et al., 2022).

Hypotheses

We hypothesize that the factors derived from the ADQ, which represent attachment-related defense mechanisms, will demonstrate significant associations with insecure attachment. This expected association serves as evidence supporting the convergent validity of the ADQ. Specifically, we hypothesize that attachment anxiety will emerge as a stronger predictor of defenses categorized as hyperactive, while attachment avoidance is expected to be a stronger predictor of defenses characterized as deactivating.

Analysis plan

Firstly, we aim to reduce and organize the items comprising the preliminary ADQ into distinct underlying factors that closely align with the constructs outlined in Table 1, reflecting various types of attachment-related defense mechanisms. In the second phase of our research, our focus is on enhancing and refining the ADQ. To achieve this, we will conduct a second study with a new sample, expecting this will reveal a factor structure closely resembling that identified in Study 1. We anticipate that a subsequent EFA and the inclusion of new items will contribute to improvements in both its reliability and overall structure. Following this, we will analyze the correlations between the ADQ and insecure attachment.

Method

Initial considerations: item and scale development

To support conceptual clarity, we developed definitions for the constructs of interest that we hypothesized would comprise the factor structure. Following best practices guidelines, we aimed to have a minimum of three items per factor (Carpenter, 2018). We also selected a 6-point Likert scale for response options (Lee & Paek, 2014; Lozano et al., 2008; Simms et al., 2019).

A manual was developed wherein 16 constructs were outlined and explicitly defined as defense mechanisms. To ensure a comprehensive understanding, we identified five features for each, creating at least ten intentional items per construct to capture diverse dimensions. We intentionally created some items with reversed wording. To ensure item quality and content validity, six clinical psychologists reviewed the item pool, which contained 200 items. They evaluated clarity and alignment using a 5-point Likert scale, referencing the ADQ manual. Items scoring below 2 were excluded, and those scoring 3–4 underwent group discussion for potential rephrasing. Qualitative feedback from psychologists during the review process offered valuable insights into the appropriateness and overall effectiveness of items in relation to the intended constructs, complementing the quantitative ratings. This expert review process supported item refinement and construct conceptualization. After this step, we retained a total of 176 items (see Table S1 in the online supplemental material). In addition, three items were included to check performance integrity (“I have been dishonest in my answers here,” “I have rushed my responses here” and “It is important to pay attention in this study. Please tick strongly agree”).

Study 1: Exploratory Factor Analysis (EFA)

Participants and procedure

All study procedures adhered to ethical standards as specified by the Macquarie University Human Research Ethics Committee. Study 1 obtained a total sample of N = 2225 before screening from two sources. Source 1 participants (n = 1127) were undergraduate psychology students and were recruited via an online advertisement posted to the university research database where students self-select into a study of their choice. These participants were rewarded with course credit for their Introductory Psychology unit of study. Source 2 participants (n = 1098) were recruited via Prolific, an online recruitment platform, and were paid for 20 min of their time at an hourly rate of £7.50.

The recruited participants were invited to an online survey hosted on Qualtrics. After obtaining informed consent, participants were instructed to complete the survey, which consisted of 176 items (+ 3 attention checks) presented in random order and standard demographic questions. Overall, there were 62% female (n = 1202), 36% male (n = 709), and 2% non-binary (n = 33) participants, aged 17–82 (M = 25.46, SD = 9.31). In terms of ethnic group, participants identified as Caucasian (65.4%), Asian (19%), Middle Eastern or North African (4.6%), Black or African American (4.1%), Mixed/Multiple ethnic background (3.7%), Hispanic, Latino or Spanish origin (1.4%), Aboriginal Australian, Native Hawaiian or Other Pacific Islander (0.9%), American Indian or Alaska Native (0.3%), Some other race, ethnicity or origin (0.7%). The median completion time for the survey was 20.4 min.

Data screening and preliminary analysis

To ensure data quality, we checked for missing data and required a 100% survey completion rate for retention. Cases with incomplete surveys were excluded from the analysis. Additionally, we conducted checks to assess performance integrity (see Table S6 in the online supplemental material). Cases were excluded if their response time was less than 400 s or if they had two or more failures of the three attention-check items. Following these exclusions, the final sample size consisted of N = 1944 participants. This sample size meets the requirements for factor replicability in EFA, as it exceeds the recommended threshold of having more than 500 cases and a ratio of at least 5:1 between the number of measured variables and cases (Goretzko et al., 2021).

The distribution of the items was examined using histograms and skewness values. All items had a skewness score < 1, which was considered an acceptable level of deviation from normality in the data distribution (Kim, 2013).

As a strategy for choosing an appropriate number of factors that are reliable and interpretable, we used the parallel analysis program (O’Connor, 2000). Parallel analysis is among the most effective methods for identifying the number of factors to retain in the context of EFA (Finch, 2020). In parallel analysis, actual eigenvalues are compared with random-order eigenvalues. Factors are retained when actual eigenvalues surpass random ordered eigenvalues (Williams et al., 2010). In our study, parallel analysis indicated that 28 factors could be extracted explaining 56% of the total variance (see Table S7 in the online supplemental material), a larger number than the intended 16 factors based on theoretical considerations from the literature review.

Study 1: results

Analysis

We began with an extraction specifying 28 factors using principal axis extraction method with direct oblimin for oblique rotation and kaiser’s criteria (eigenvalue > 1 rule).

Principal axis factoring is generally considered to be appropriate for the exploration of the underlying factors for theoretical purposes (Pett et al., 2003). We used direct oblimin due to the likelihood of factors being correlated with one another, based on theory.

Decision-making processes for item exclusion and retention

Our approach aimed to decrease the number of factors to develop a scale that was more manageable in size and aligned with our intended design. To achieve this, we initially attempted to remove the factors that consisted solely of single items by eliminating the item in question. We followed an iterative process to exclude items.

At first, we identified weak factors. According to Watkins (2020), factors with two or fewer salient loadings are relatively weak, and factors with at least three salient loadings are preferred. An item was considered dominant if it was the highest-loading item on a factor.

If no dominant items were observed on a factor, meaning that all item loadings onto that factor were negligible due to higher loadings on a different factor, we specified one less factor for the subsequent round of analysis without excluding any items. If a factor had only one or two dominant items, we excluded the item that had a dominant influence on that factor, one at a time. If the weakest factor was not readily identifiable, our second strategy involved excluding the weakest item in a cluster with a loading weight of less than 0.30 (the bottom of the chain) or eliminating cross-loading items. Using this approach, a total of 90 analysis steps were taken, resulting in 94 items remaining. 11 factors were identified at this point that seemed relatively stable.

We then focused on the 11 factors and aimed to further reduce their size and increase conceptual clarity. We developed provisional labels for each factor by examining what conceptual themes their items shared. Intending to capture different dimensions of each construct while keeping the size manageable, we carefully considered any concept repetition within each factor when deciding which items to remove. To better inform this process, we also considered Cronbach’s alpha reliability analysis. If Cronbach’s alpha improved by excluding a particular item, we used this to inform decision-making to remove that item. After a thorough item reduction process involving a further 11 steps (in total from the beginning, 101 steps), we successfully reduced the original pool of 176 items to a more manageable 82 items (see Table S2 in the online supplemental material). These were organized into 11 factors (consisting of 5–10 items per factor). We applied a loading value cutoff threshold of 0.32, as suggested by Tabachnick and Fidell (2013). This table can be viewed in the appendices (Table S9 in the online supplemental material). These 11 factors covered much of the conceptual basis that this scale intended to measure, but some conceptual gaps remained, necessitating a secondary phase of item generation and refinement.

By including additional items (see Table S3 in the online supplemental material)., we aimed to enhance the comprehensiveness and precision of the ADQ measurement tool for our research. We also recognized the need to incorporate a reliable measure in the second study to ensure alignment between our constructs and attachment-insecurity. According to Clark and Watson (2019), testing for convergent validity is an important part of scale construction and involves examination of within-domain correlations. Relationships between test scores and other well-established measures intended to assess the same or similar constructs provide convergent evidence. Consequently, we included the Experiences in Close Relationships scale (ECR; Brennan et al., 1998) to investigate convergent validity with the ADQ. Specifically, our research sought to understand the associations between attachment-related defense mechanisms and insecure attachment dimensions in the context of theoretical expectations.

Study 2: EFA and convergent validity

Participants and procedure

For Study 2, the sample (n = 726 after missing data were removed) was recruited from the same sources as Study 1 (Macquarie University Introductory Psychology students, n = 520, and Prolific, n = 94), along with two new sources: First, social media, specifically academic survey exchange pages on Facebook (n = 78); Second, flyers placed in the waiting room of a GP medical clinic (n = 34). Of the total participants, 66% were female (n = 469), 32% were male (n = 229), 3% identified as non-binary (n = 20), and there were 8 missing responses. The age range of participants was between 17 and 74, (M = 24.73, SD = 9.25). In terms of ethnic group, participants identified as Caucasian (51%), Asian (21.3%), Middle Eastern or North African (7%), Black or African American (5%), Mixed/Multiple ethnic background (9.7%), Hispanic, Latino or Spanish origin (1.9%), Aboriginal Australian, Native Hawaiian or Other Pacific Islander (1.8%), Some other race, ethnicity or origin (2.2%).

Participants completed the study via an online questionnaire hosted on Qualtrics. The median completion time for the survey was 19.65 min.

Measures

The Attachment Defenses Questionnaire (ADQ)

For the factor analysis, 108 items along with 3 attention checks were utilized. Items were presented in a random order. After Study 1, we reassessed our initial ADQ design, considering the theoretical concepts we wanted to represent. We introduced 24 new items to better represent certain concepts that weren’t initially captured as intended and rephrased 2 items. After finalizing the solution, item scores within each factor were totaled to create composite scores.

Experiences in Close Relationships Scale (ECR; Brennan et al., 1998)

We measured attachment insecurity using the 36-item self-report ECR, which consists of two subscales of attachment: Avoidance (18 items) and Anxiety (18 items). The questionnaire asks participants to rate statements about their general feelings in close relationships (e.g., with romantic partners, close friends, or family members) on a 7-point Likert scale ranging from 1 (strongly disagree) to 7 (strongly agree). Statements measuring attachment anxiety include “My desire to be very close sometimes scares people away,” while statements measuring attachment avoidance include “I want to get close to others, but I keep pulling back.” Scores for each attachment dimension are obtained by calculating the mean of the items that represent the construct. In this study, Cronbach’s alphas for attachment anxiety and attachment avoidance were excellent (α = .92 and α = .91, respectively).

Study 2: results

Preliminary analysis

In Study 2, we followed a comparable approach to Study 1 to handle missing data and screen the data for quality. We removed cases with incomplete surveys to ensure the integrity of the data. Additionally, we conducted checks to exclude cases with response times below 400 s or two or more failed attempts on attention-check items. We evaluated the distribution of the data and found it to be acceptable. Furthermore, based on parallel analysis, we determined that 16 factors could be extracted from the data (see Table S8 in the online supplemental material).

Study 2, part 1—factor analysis

In Study 2, we initially started with 108 items and extracted 16 factors using the principal axis extraction method with direct oblimin rotation and Kaiser’s criteria (eigenvalue > 1 rule). We also made efforts to minimize cross-loading of items and eliminate as many as possible in the final solution. Employing the same approach as Study 1, we conducted 60 steps and ultimately arrived at a final solution with 10 factors and 50 items. The 10-factor solution explained 55.44% of the variance. Table 2 displays the results of the EFA from Study 2, showing all 10-factor labels. Detailed descriptions of these constructs can be found in the the online supplemental material. Cronbach’s alpha reliability was acceptable-to-good (α > 0.7) for all factors except Rumination and Disengagement but these were close to 0.7.

Table 2. EFA Study 2 (N = 726) results with the 10-factor solution.

	M (SD)		1	2	3	4	5	6	7	8	9	10
Factor 1. Chronic Proximity Seeking (α = .78)	2.87 (1.60)	I fall crazy-in-love so fast that I lose touch with reality.	.63	−0.02	.03	.12	.01	.03	.11	−0.11	.04	−0.20
	4.06 (1.50)	I can get caught up in fantasies about relationship scenarios.	.54	.07	−0.06	−0.25	−0.07	.12	.04	.04	.04	.17
	3.26 (1.58)	I can be obsessive during the pursuit of a relationship.	.51	−0.05	−0.02	.03	−0.03	−0.02	.05	−0.19	.16	−0.01
	3.80 (1.42)	I am prone to idealization	.49	.07	.02	−0.12	−0.05	.01	.00	.01	.12	.14
	3.02 (1.46)	I seem to return to the same type of relationship over and over again.	.32	.15	.05	−0.06	−0.07	−0.05	−0.06	−0.26	.08	−0.03
Factor 2. Distancing (α = .79)	2.74 (1.31)	I prefer to be quite separate from partners.	−0.04	.68	−0.02	.08	−0.02	−0.06	.02	−0.05	.04	−0.01
	3.37 (1.58)	I tend to be the “distancer” in my relationships in that I may be emotionally unavailable or seek to be alone.	−0.01	.60	.02	−0.05	−0.03	.00	.07	−0.04	−0.04	.18
	2.58 (1.39)	When I hear the word love, I withdraw.	.12	.51	.00	.07	−0.02	−0.17	.10	−0.06	.06	.04
	3.75 (1.51)	I seem to need more physical personal space than others seem to.	.00	.48	−0.06	−0.01	−0.16	.10	.09	.00	−0.10	.06
	3.49 (1.50)	I can feel “crowded” or “suffocated” when others try to get close.	.08	.48	.07	−0.01	−0.15	−0.02	.05	−0.06	−0.14	−0.01
Factor 3. Compulsive-caregiving (α = .74)	3.97 (1.37)	I tend to please others above taking care of myself.	−0.05	.02	.74	−0.01	−0.02	−0.06	.07	−0.03	.07	.01
	3.92 (1.37)	I have a deeply ingrained tendency to place others ahead of myself.	−0.02	.09	.63	.06	−0.02	.13	.08	.07	.06	.01
	3.99 (1.51)	I find it hard to say no to others even when things don’t suit me.	.03	.17	.56	−0.11	.02	−0.01	−0.02	−0.05	.06	.04
	4.21 (1.36)	I don’t go out of my way to please others. R	−0.03	.11	−0.53	.15	−0.05	.00	.09	.03	.01	.04
	4.01 (1.32)	I would risk my life to help others.	.02	−0.15	.43	.08	−0.08	.07	.04	.00	−0.06	.08
Factor 4. Rumination (α = .66)	3.85 (1.42)	Generally, it is fairly easy for me to put unwanted thoughts out of my mind. R	−0.03	−0.03	−0.10	.53	.03	.02	−0.06	−0.06	−0.04	.18
	4.12 (1.50)	I overthink relationship worries.	.17	−0.03	.11	−0.48	−0.06	.10	.12	−0.10	.03	.14
	4.15 (1.46)	I can get stuck in my head thinking about how relationship problems might play out.	.21	.04	.06	−0.42	−0.07	.10	.06	−0.07	.08	.21
	4.10 (1.58)	The idea that my partner could leave me unexpectedly doesn’t cross my mind. R	.00	.15	−0.06	.40	.07	.02	−0.08	.14	.00	−0.05
	3.46 (1.46)	I am pretty good at “switching off” from hard feelings. R	.00	−0.03	.00	.40	−0.11	−0.09	.06	−0.07	−0.12	.39
Factor 5. Parentification (α = .81)	3.44 (1.63)	I have had to do more self-parenting than most others have.	.02	.00	.01	.05	−0.80	.06	−0.10	.00	.04	−0.06
	3.73 (1.60)	My needs have been misread by my parent more often than not.	.02	.07	.00	−0.17	−0.62	.01	.04	.07	.12	.02
	3.47 (1.52)	I learned early in life to bury the need to be taken care of.	−0.02	.11	.12	.10	−0.58	−0.04	.09	.01	−0.07	.10
	3.79 (1.50)	I have been hardened by the way I was raised.	−0.03	.01	−0.10	.01	−0.58	.07	.02	−0.10	−0.06	.15
	2.89 (1.54)	I have negative expectations of others because of my parents.	.05	.05	.04	−0.09	−0.57	−0.11	.08	−0.05	.08	−0.08
Factor 6. Disengagement (α = .68)	2.70 (1.27)	I am quite in touch with my feelings. R	−0.05	−0.07	−0.06	−0.03	−0.03	.60	.00	−0.03	−0.01	−0.12
	2.67 (1.31)	I easily recall the emotional part of memories. R	.04	.04	−0.01	−0.04	.00	.59	.03	−0.01	.01	.00
	2.40 (1.25)	I have an abundance of empathy. R	−0.02	.03	.29	.00	−0.03	.50	−0.08	−0.02	−0.02	.01
	2.60 (1.25)	I tend to form strong emotional connections with people, places, and things. R	.15	−0.14	.17	.03	−0.05	.39	.06	.05	.13	.06
	3.24 (1.38)	I am generally quick to experience a deep sense of connection with others. R	.14	−0.12	.09	.22	.05	.35	.10	−0.10	.15	−0.02
Factor 7. Ambivalent Splitting (α = .73)	3.09 (1.52)	I see things in extremes. Things are either all-good or all-bad.	−0.05	−0.05	−0.01	−0.06	−0.02	.01	.76	−0.13	.02	−0.04
	3.21 (1.45)	My focus is totally positive or totally negative.	−0.01	.10	−0.06	.06	.01	.07	.69	.00	.05	.03
	3.68 (1.48)	My efforts are either a success or they are a failure. I find it hard to see a middle ground.	.03	.05	.08	−0.13	−0.05	−0.02	.57	.09	.02	.04
	2.92 (1.54)	I frequently switch from seeing someone/ something as entirely good to entirely bad, e.g., “evil” or “perfect.”	.21	.07	.03	.06	.01	−0.09	.44	−0.20	.01	−0.08
	3.87 (1.46)	I am constantly on guard, comparing myself and my status to others.	.10	.06	.16	−0.28	.02	−0.05	.32	.02	.09	.18
Factor 8. Defensive Protest (α = .79)	2.62 (1.59)	I have picked fights to try to stop or stall someone I care about from leaving.	.02	−0.02	.10	.02	−0.12	−0.02	−0.01	−0.68	.08	−0.07
	2.46 (1.66)	I have made threats to break up, which I did not really mean.	−0.03	.04	−0.05	.03	−0.03	.05	.08	−0.59	.08	−0.08
	2.74 (1.59)	It is quite like me to try to make someone feel jealous because I am not getting enough attention from them.	.15	.00	.00	.01	−0.03	−0.03	.01	−0.58	.07	.06
	2.88 (1.70)	I go through cycles of I love you - I hate you - don’t leave me.	.18	.01	.02	−0.02	−0.06	−0.03	.20	−0.51	−0.08	−0.05
	3.61 (1.61)	If I was mad with my partner, I might purposely ignore phone calls from them as a protest.	−0.02	.15	−0.02	−0.11	.11	.10	.03	−0.50	−0.03	.22
Factor 9. Consumption (α = .76)	3.01 (1.50)	I have a tendency to spill my emotional issues onto others.	.09	.00	.00	−0.01	−0.05	.09	.01	−0.11	.58	−0.02
	3.58 (1.56)	I have a habit of oversharing.	.06	−0.04	.05	−0.01	−0.09	.09	.06	.05	.56	.01
	3.21 (1.54)	Sometimes, I lean too heavily on others.	.09	−0.14	.03	−0.07	.09	−0.05	.08	−0.16	.53	.09
	2.62 (1.27)	Others think I am draining.	−0.01	.08	.05	−0.03	−0.17	−0.24	.18	−0.06	.38	−0.07
	3.79 (1.35)	I have a tendency to “latch onto” others I feel a connection toward.	.32	−0.10	.09	−0.05	.00	.04	.01	−0.07	.36	.15
Factor 10. Vulnerability Suppression (α = .74)	3.74 (1.48)	I tend to block thoughts that make me feel too much.	−0.04	.16	.00	.02	−0.07	−0.01	.04	−0.03	.17	.56
	3.98 (1.35)	If I remember something unpleasant, I push it away from my mind.	−0.04	.04	.02	.06	.00	.02	−0.03	−0.03	.08	.53
	3.88 (1.41)	I usually shove my emotions down when I can.	.13	.19	.15	.06	−0.03	−0.12	.10	.05	−0.10	.50
	4.02 (1.43)	I prefer to disguise vulnerability.	.08	.09	.05	−0.07	−0.08	−0.14	.05	−0.01	−0.11	.49
	4.36 (1.40)	I would hate for others to see me as fragile.	.06	−0.06	.06	−0.05	−0.09	−0.03	.07	.01	−0.09	.46

108 items reduced to 50. The bold values highlight significant loadings indicating factor patterns.

Study 2, part 2 – convergent validity test

Hypotheses

We predict a significant positive association between attachment anxiety and the defense mechanisms chronic proximity seeking, compulsive caregiving, rumination, defensive protest, consumption, and splitting. Furthermore, we predict a significant positive association between attachment avoidance and the defense mechanisms of distancing, disengagement, suppression, and parentification.

Preliminary analysis: independent t-tests

ADQ factor scores were analyzed by gender to assess differences. Table 3 presents descriptive statistics and significant gender differences. Apart from male and female participants, a small number identified as non-binary gender (n = 20). Due to the limited size of this group, they were excluded from the comparative analysis.

Table 3. Descriptive statistics and gender differences in attachment and defenses.

	Mean (SD)
	Overall (n = 726)	Male (n = 229)	Female (n = 469)	t	d
ECR: Attachment
Avoidance	3.65 (1.06)	3.63 (0.98)	3.64 (1.10)	−0.74	−0.01
Anxiety	4.15 (1.19)	3.87 (1.08)	4.27 (1.21)	−4.40**	−0.34
ADQ: Defense Mechanism
Chronic Proximity-Seeking	17.01 (5.50)	16.30 (4.86)	17.25 (5.65)	−2.31*	−0.18
Distancing	15.94 (5.37)	15.67 (5.06)	16.00 (5.51)	−0.77	−0.06
Compulsive-caregiving	20.10 (4.85)	18.68 (4.08)	20.65 (5.03)	−5.56**	−0.42
Rumination	19.67 (4.82)	17.79 (4.24)	20.47 (4.76)	−7.52**	−0.58
Parentification	17.32 (5.86)	16.46 (5.01)	17.48 (6.22)	−2.32*	−0.17
Ambivalent-Splitting	16.77 (5.50)	16.05 (4.82)	16.95 (5.71)	−2.18*	−0.17
Defensive-Protest	14.31 (6.00)	13.42 (5.10)	14.76 (6.26)	−3.02**	−0.23
Consumption	16.21 (5.26)	15.40 (4.63)	16.51(5.43)	−2.81**	−0.21
Vulnerable-Suppression	19.97 (4.96)	19.90 (4.50)	19.91(5.10)	−0.05	.00
Disengagement	13.61 (4.27)	14.66 (4.33)	13.07 (4.09)	4.75**	.38

d is Cohen’s d for effect size.

*p < .05; **p < .01.

df = 724.

The results showed significant gender differences in attachment anxiety with a small effect size. Specifically, women reported higher levels of attachment anxiety than men, but no differences were found in attachment avoidance. Our findings showed significant gender differences in 8 of the 10 defense mechanisms. Specifically, there was a moderate gender difference in Rumination and Compulsive Caregiving, with women reporting higher levels. Additionally, there were small differences in Chronic-proximity seeking, Parentification, Ambivalent-splitting, Defensive-protest, and Consumption, with women reporting slightly higher levels. Conversely, men scored higher on Disengagement with a moderate-sized difference. Given the evidence of small to moderate but significant gender differences, it was necessary to control for gender in subsequent analyses.

Correlational analysis

Correlational analyses were performed to assess the relationship between ADQ factors and attachment styles, shown in Table 4. Regarding the individual subscales, attachment anxiety exhibited significant moderate-strong correlations with the defense mechanisms: Chronic-proximity-seeking, Consumption, Defensive protest, Ambivalent-splitting, Rumination, and Compulsive-caregiving. These findings are in line with our predictions. Additionally, attachment anxiety demonstrated a less strong but still significant correlation with Vulnerability-suppression and a weak correlation with Distancing.

Table 4. Correlations between attachment styles and the factors derived from Study 2.

	Attachment anxiety		Attachment avoidance
	r	p	r	p
Attachment anxiety	1		.08	.03
Attachment avoidance	.08	.03	1
Defense Mechanisms
Chronic Proximity Seeking	.68	.00	.12	.00
Consumption	.68	.01	−0.10	.00
Compulsive-caregiving	.40	.00	.03	.48
Rumination	.56	.00	.08	.04
Distancing	.09	.01	.67	.00
Disengagement	−0.27	.00	.38	.00
Vulnerability Suppression	.26	.00	.48	.00
Parentification	.32	.00	.41	.00
Ambivalent Splitting	.57	.00	.34	.00
Defensive Protest	.58	.00	.18	.00

Significance is at level p < .01.

df = 724.

On the other hand, attachment avoidance was significantly correlated with Distancing Suppression, Parentification, Disengagement, and Ambivalent-splitting. The strengths of these associations were moderate-strong. These findings support our predictions. Additionally, attachment avoidance displayed weak positive correlations with Defensive-protest and Chronic-proximity-seeking.

Regression analysis

To address the third research question of whether attachment can predict defenses, and to determine which attachment dimension was an independent predictor (controlling for gender differences), we employed multiple linear regression analyses to ascertain the distinct relationships between each defensive mechanism and the attachment dimensions. Please refer to Table 5. In this analysis, defensive mechanisms acted as the dependent variables, while the two attachment dimensions were incorporated as independent variables. Gender was included as a covariate in all models to account for the disparities observed in the t-tests. Participants who identified as non-binary (n = 20) were excluded from this analysis because of their comparative small number in the sample which did not allow for meaningful comparisons. The significance level was set to a conservative p < .01 given multiple regression models were being fit.

Table 5. Linear regression analyses with attachment dimensions as predictors (IVS) of attachment defenses controlling for gender.

	Attachment avoidance					Attachment anxiety					Gender
Dependent Variable	B	SE B	β	t	p	B	SE B	β	t	p	B	SE B	β	t	p	R Square
ADQ 10 factors
Chronic Proximity   Seeking	.34	.14	.07	2.4	.02	3.09	.13	.68	23.93	.00	−0.28	.32	−0.02	−0.86	.39	.46
Distancing	3.33	.14	.66	23.25	.00	.20	.13	.04	1.51	.13	.23	.33	.02	.72	.47	.44
Compulsive-caregiving	−0.06	.16	−0.01	−0.36	.72	1.56	.14	.38	11.01	.00	1.35	.36	.13	3.78	.00	.18
Rumination	.08	.14	.02	.61	.54	2.17	.12	.54	17.41	.00	1.81	.31	.18	5.80	.00	.36
Parentification	2.12	.18	.38	11.65	.00	1.37	.17	.28	8.29	.00	.46	.42	.04	1.10	.27	.25
Disengagement	1.59	.13	.40	12.13	.00	−1.04	.12	−0.29	−0.8.75	.00	−1.19	.30	−0.13	−3.98	.00	.25
Defensive Protest	.76	.17	.14	4.49	.00	2.86	.15	.57	18.57	.00	.19	.39	.02	.49	.62	.36
Consumption	−0.81	.13	−0.17	−6.03	.00	3.02	.12	.69	24.73	.00	−0.09	.31	−0.01	−0.29	.77	.48
Vulnerable Suppression	2.21	.15	.48	14.88	.00	.88	.13	.21	6.51	.00	−0.35	.34	−0.03	−1.02	.31	.29
Ambivalent Splitting	1.50	.15	.29	10.12	.00	2.55	.13	.56	18.95	.00	−0.13	.34	−0.01	−0.38	.71	.42

Note. The Beta values represent how strongly each of the attachment styles is related to the corresponding defense.

Significance is at level p < .01.

Regarding the 10-factor defense structure, attachment anxiety was found to be the more robust and sole positive predictor of Chronic-proximity-seeking, Rumination, Compulsive-caregiving, and Consumption. On the other hand, attachment avoidance was the stronger sole positive predictor of Distancing and Disengagement.

When defense mechanisms were significantly predicted by both attachment dimensions, effect sizes in the form of standardized betas demonstrate which is the more dominant predictor. For instance, Parentification was significantly predicted by both attachment avoidance (β = .38) and attachment anxiety (β = .28). Defensive-protest was more substantially influenced by attachment anxiety (β = .57) compared to attachment avoidance (β = .14). Similarly, Ambivalent-splitting was significantly predicted by both attachment dimensions, with attachment anxiety (β = .55) exhibiting a stronger association compared to attachment avoidance (β = .30). On the other hand, Vulnerability-suppression was more strongly predicted attachment avoidance (β = .48) compared to attachment anxiety (β = .21).

Taken together, these findings are largely consistent with our hypotheses and this provides initial support for the external validity of the ADQ.

Discussion

We have introduced the Attachment Defenses Questionnaire (ADQ-50), a newly developed scale that intends to measure attachment-related defense mechanisms. Unlike other defense mechanism scales, this scale was grounded in attachment theory. Two studies were conducted, both incorporating factor analysis. The second study additionally explored the associations between the ADQ and attachment dimensions measured through the ECR. Our results affirm our hypotheses, as we successfully identified a factor structure for the ADQ that is both clinically relevant and theory-driven while maintaining a manageable size. Furthermore, our study provides preliminary evidence supporting the convergent validity of the ADQ.

We hypothesized that attachment anxiety would be the stronger predictor of defenses categorized as hyperactive (chronic proximity seeking, compulsive caregiving, rumination, defensive protest, consumption, and splitting) while attachment avoidance would be the stronger predictor of defenses characterized as deactivating (distancing, disengagement, suppression, and self-reliance/parentification). In terms of the relationship between our scale and attachment dimensions, our results support our hypotheses and were predominately in line with theoretical expectations with some minor exceptions. The unexpected positive correlation between suppression and attachment anxiety, albeit weak, contrasts with previous studies indicating challenges in cognitive detachment for individuals with attachment anxiety (Mikulincer et al., 2004; Richardson et al., 2022). However, Clear and Zimmer-Gembeck (2017) revealed a noteworthy connection between attachment anxiety and the suppression of anger, but no correlation to the suppression of sadness or worry. This suggests that attachment anxiety may be associated with specific facets of emotional suppression, rather than a global emotional suppression, highlighting potential avenues for future research. Another unexpected finding was the positive association between what we initially termed “learned self-reliance” and attachment anxiety. In theory, attachment anxiety is predicted to negatively correlate with self-reliance, given that it aligns with dependence-incompetence schemas and a need for approval (Cantazaro & Wei, 2010; Jakubiak et al., 2023; Prunas et al., 2019; Simard et al., 2011). Consequently, we reevaluated the labeling of this factor and, after scrutinizing the five items, determined that “parentification” more precisely captures the phenomenon we are measuring in the ADQ-50.

Limitations and future research

The current work has some limitations. The use of self-report methods to assess psychological defense mechanisms may seem paradoxical (Davidson & MacGregor, 1998). However, individual differences on self-report attachment measures relate to measurable unconscious processes (Shaver & Mikulincer, 2002). This approach allows for the exploration of attitudes and behaviors believed to signify an underlying defense, even when not explicitly addressing the defense itself.

Another limitation is our sample. Given the requirement for a large sample size, we opted for a convenience sample, which can pose challenges to the generalization of findings. While our results provide valuable insights for the specific group we studied, caution must be exercised when attempting to generalize these findings to a wider or different population. Moreover, despite our best attempts at obtaining a balanced sample of ethnicity and gender, the number of participants per group is not even.

We employed parallel analysis as a preliminary step to ascertain the optimal number of factors for initial extraction. Despite being widely recognized as the gold standard for evaluating factor retention, as asserted by Goretzko (2022), our specific application of parallel analysis suggested an impractically large number of factors for a feasible solution. It is important to acknowledge that parallel analysis, while valuable, may exhibit limitations, particularly in cases where there is a risk of over-extraction of factors, especially in large sample sizes.

Additionally, we intended to include an equal number of reversed items, but we could not fulfill this aim since some of the items that were intended to be reversed did not meet the necessary criteria, e.g. failing to load above the predefined threshold. At the same time, the scale we developed was modeled after the ECR, and we note that in the ECR, attachment avoidance has more assigned reverse items than attachment anxiety. It is possible that this uneven distribution of reverse items may reflect an inherent characteristic of attachment avoidance that makes negative wording more suitable for such items.

Another potential limitation lies in the fact that our expert review group, which contributed to formulating the ADQ items, did not possess extensive experience with the attachment literature and defense theory literature or clinical practice associated with psychic defenses. Nevertheless, the review group did have extensive training in standard practices and therapies, including schema therapy.

One additional limitation in our study is that the reliability analysis revealed that the alpha values for 2 out of 10 subfactors are slightly below the optimal range. There are different reports about the acceptable values of Cronbach’s alpha, ranging from 0.70 to 0.95 (Tavakol & Dennick, 2011). Values near 0.7 can be deemed minimally acceptable and although not ideal, are nevertheless satisfactory. Future research can potentially address this limitation, and potential avenues include refining the existing items to enhance the reliability values or considering the inclusion of new items that better align with the intended constructs.

Other potential avenues for future research involve investigating the test-retest reliability of the ADQ, scrutinizing its structure and validity within a psychiatric clinical population, and exploring the developmental foundations of these defensive mechanisms. Exploring variations related to age and ethnic groups could be intriguing for future studies, especially considering that attachment theory is considered cross-cultural (Mesman et al., 2016) and defensive functioning may be influenced by age (Diehl et al., 1996). Future research could also attempt to establish a connection between the ADQ and real-life defensive maneuvers evaluated within psychotherapy sessions. Given the association of attachment and defenses with borderline personality disorder (Smith & South, 2020; Zanarini et al., 2009), we anticipate that the ADQ will demonstrate valuable clinical utility within this particular client population.

There are also some unanswered questions that require addressing. Our goal was to encapsulate constructs representing the sentinel schema of attachment anxiety through hypervigilance and the rapid responder schema of attachment avoidance via the construct defensive self-interest, as per social defense theory (Ein-Dor & Hirschberger, 2018). However, we encountered challenges in maintaining distinct factors. Future revisions of the ADQ may find ways to incorporate these constructs more distinctly, or perhaps, these schemas are better understood as overarching umbrella constructs.

Further examination is needed to explore the connection between hyperactivating and deactivating processes, and to determine whether this aligns with the nature of the association between attachment dimensions. The current findings suggest that any defense mechanism may feasibly incorporate elements of both hyperactivation and deactivation. For example, the defense splitting suggests a tension between approach and avoidance, wherein an individual is simultaneously attracted to and repelled by the same goal or situation (see Mikulincer et al., 2010 on pushes and pulls of close relationships). This is consistent with Cameron et al. (2012) report that the attachment dimensions are not orthogonal in practice. Furthermore, a contradictory or seemingly disorganized defense strategy may be explained by high attachment anxiety and high attachment avoidance occurring together also known as “fearful” attachment (Bartholomew, 1990; Hlay et al., 2023). Given the potential for diverse perspectives on this topic, it presents a compelling area for future research exploration.

In psychological terms, the correlation between attachment-anxiety related defenses and attachment-avoidance related defenses indicates a nuanced dynamic. One compelling question revolves around understanding how two distinct patterns became internalized. This serves as a rich topic, prompting exploration into the roles of primary and secondary caregivers, and the evolution of internal working models shaped by the influence of romantic partners and life events. Another area of application is examining how the combination effects of hyperactivating defenses and deactivating defenses play out in various contexts.

Future research could explore an interaction term between attachment-avoidance and attachment-anxiety and between deactivating defenses and hyperactivating defenses. The interaction term allows for an exploration of whether the combined influence or effect of both dimensions has a unique impact on a particular outcome, compared to the effects of each dimension separately. Further exploration of how these defenses interact within the context of specific psychological processes or disorders could provide valuable insights for clinical practice and intervention strategies.

Conclusion

The culmination of this research is the development of the ADQ-50. The ADQ underwent multiple revisions. Initially, 200 items were created based on a thorough literature review. An expert review reduced the count to 176, refining phrasing. In Study 1, EFA narrowed this to 82 items, identifying 11 factors. In Study 2, 26 new items were added. A second EFA analyzed 108 items, resulting in the final ADQ-50 with 10 factors and 5 items per factor.

The limitations in this research are reasonably balanced by its strengths. One notable strength of the first study was the utilization of a large sample size, which bolstered the reliability of the analysis. Another strength of the study is its breadth; not only does it provide insights into inter-reliability, but it also offers evidence of external validity. A further strength was our attention to detail. We made every effort to be thorough and to meticulously document the decision-making steps throughout the scale development process. Given the numerous instances in the literature where EFA has been misapplied, as highlighted by Carpenter (2018), we aimed to adhere to best practices and ensure no essential details were overlooked. Furthermore, careful construct conceptualization is important for the development of a scale and many papers on scale development often fall short in providing detailed information on this critical aspect (Kyriazos & Stalikas, 2018) To address this gap and offer clarity, we have presented a definition table in the online supplemental material (Table S10). An additional positive aspect is that each factor in our model consists of five items, exceeding the recommended minimum of three items per factor (Carpenter, 2018), enhancing the reliability and stability of the identified factors.

The ADQ also goes beyond the ECR by emphasizing the interplay between defense mechanisms and attachment orientation, providing detailed insights for analyzing coping patterns. In clinical settings, it helps identify intervention targets and self-perpetuating cycles from maladaptive coping behaviors. This focus is particularly beneficial for individuals with vulnerable personality traits, given the established connections between borderline personality and both attachment insecurity (Smith & South, 2020) and defense mechanisms (Zanarini et al., 2009).

Researchers employing the ADQ should be mindful of the potential for simultaneous high levels of hyperactivating and deactivating defenses within an individual. These mechanisms only appear as if they should be opposites reflecting the complexity of attachment dynamics. Recognizing these nuances provides clinicians with valuable insights for addressing conflict and the use of defense mechanisms in therapy. However, when analyzed at a large group level, this has the potential to confound hypothesized associations. Therefore, the interpretation of ADQ data should take this into account.

In conclusion, based on the findings presented, the ADQ-50 shows considerable promise and demonstrates potential as a reliable and valid scale for assessing attachment defenses. However, further research is needed to validate the ADQ-50 and to explore the potential applications of the attachment defenses model in diverse contexts. Future studies could investigate the predictive validity of the ADQ-50, the stability of attachment defenses over time, and the potential links between attachment defenses and clinical outcomes. We aspire to have provided clinicians and researchers with a valuable tool for future psychological assessment and intervention.

Informed consent

Informed consent was obtained from all individual participants included in the study. The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.

Disclosure statement

No funding or other support was received. The authors have no competing interests to declare that are relevant to the content of this article. This study was approved by the Humanities and Social Sciences Human Research Ethics Committee, Macquarie University and was performed in accordance with the National Statement on Ethical Conduct in Human Research (2018/2007).