Adult Attachment Scale AAS

Abstract

The Adult Attachment Scale (AAS), developed by social psychologists Nancy L. Collins and Stephen J. Read in 1990, is one of the most seminal, widely utilized psychometric instruments designed to evaluate adult attachment styles within the context of close interpersonal and romantic relationships. Grounded in the conceptual paradigm originally formulated by John Bowlby and extended into romantic partnership dynamics by Cindy Hazan and Phillip Shaver (1987), the AAS decomposed Hazan and Shaver’s initial categorical descriptions into a continuous multi-item self-report questionnaire. The original instrument consists of 18 items evaluated on a 5-point Likert scale ranging from 1 (Not at all characteristic of me) to 5 (Very characteristic of me).

Psychometrically, the AAS captures individual differences across three foundational subscales: Close (the extent to which an individual feels comfortable with emotional closeness, intimacy, and proximity), Depend (the degree to which an individual believes others are trustworthy, dependable, and available in times of need), and Anxiety (the degree to which an individual experiences pervasive fear of abandonment, romantic rejection, and insufficient love from a partner). Subsequent psychometric evaluations, including revised editions (Collins, 1996) and comparative meta-analyses (Brennan, Clark, & Shaver, 1998), have alternatively demonstrated that the Close and Depend dimensions can be combined into an overarching Attachment Avoidance axis, yielding an orthogonal two-dimensional space corresponding to Attachment Anxiety and Attachment Avoidance. The AAS exhibits robust internal consistency (Cronbach’s alpha coefficients typically ranging from .69 to .85 across subscales), satisfactory test-retest reliability across multi-month intervals, and pronounced convergent, discriminant, and predictive validity in relation to relationship satisfaction, dyadic communication, psychological well-being, and conflict resolution strategies.

Keywords

Adult Attachment Scale, AAS, attachment theory, adult attachment, internal working models, attachment anxiety, attachment avoidance, relationship satisfaction, interpersonal intimacy, psychometrics

Authors

The Adult Attachment Scale was formulated and psychometrically validated by:

• Nancy L. Collins, Ph.D. — Professor of Psychological & Brain Sciences, Department of Psychological & Brain Sciences, University of California, Santa Barbara (UCSB). Dr. Collins is an internationally recognized expert in close relationships, social support, interpersonal perception, and the cognitive-affective processes underlying attachment systems. Inquiries regarding unpublished validation datasets or scoring variations may be directed to: [email protected].
• Stephen J. Read, Ph.D. — Professor of Psychology, Department of Psychology, University of Southern California (USC). Dr. Read’s scholarship integrates social cognition, personality psychology, neural network modeling, and computational models of interpersonal behavior.

Purpose

The primary clinical and empirical purpose of the Adult Attachment Scale (AAS) is to assess the continuous dimensions of cognitive, affective, and behavioral patterns that adults manifest in their romantic and intimate attachment bonds. In the late 1980s, empirical research into adult attachment relied largely on Hazan and Shaver’s (1987) single-item categorical forced-choice measure, which classified adults into one of three rigid prototypes: Secure, Avoidant, or Anxious/Ambivalent. While groundbreaking, categorical taxonomy introduced significant psychometric constraints, including severe loss of statistical power, an inability to detect sub-threshold variance, and an untenable assumption that attachment representations were mutually exclusive and internally uniform.

Collins and Read (1990) designed the AAS to address these psychometric limitations by decomposing Hazan and Shaver’s descriptive paragraphs into discrete, continuous propositions. The purpose of this operationalization was twofold:

1. Deconstruction of Latent Attachment Structures: To empirically investigate whether adult attachment is organized along categorical boundaries or continuous underlying dimensions, and to identify the primary components that constitute an individual’s cognitive “internal working model” of self and others.
2. Investigation of Relational and Dyadic Quality: To supply researchers and clinicians with a reliable, finely calibrated diagnostic instrument capable of predicting dyadic processes such as relationship satisfaction, communication competence, emotional expression, attributional styles, and perceived partner support.

Beyond basic social psychology, the AAS has significant utility in clinical psychology, marital and family therapy, and psychiatric assessment. In therapeutic contexts, the scale enables clinicians to pinpoint specific relational vulnerabilities. For instance, high scores on the Anxiety dimension typically reveal hyperactivating strategies, such as clinging behavior, hypervigilance toward perceived threats of abandonment, and excessive reassurance seeking. Conversely, low scores on the Close and Depend dimensions highlight deactivating strategies, including compulsive self-reliance, emotional detachment, and suppression of vulnerability. Consequently, the AAS serves as an indispensable tool for formulating case conceptualizations and monitoring therapeutic interventions designed to foster an earned secure attachment.

Psychological Construct

The Adult Attachment Scale operationalizes adult attachment not as an immutable behavioral type, but as a system of enduring cognitive-affective representations, commonly known as internal working models. According to Collins and Read (1990), these working models encompass expectations, beliefs, and defensive coping strategies that individuals construct based on their cumulative histories of relational interactions. Factor analytic investigations conducted during the instrument’s inception isolated three primary psychological dimensions:

1. The Close Dimension

The Close subscale measures the degree to which an individual feels comfortable with physical and emotional intimacy, interpersonal proximity, and vulnerability. Individuals scoring high on this dimension view emotional closeness as natural, desirable, and unthreatening. They willingly self-disclose, experience genuine comfort when their romantic partners reveal intimate emotional content, and do not perceive genuine intimacy as an infringement upon their personal autonomy. Conversely, low scores indicate interpersonal discomfort, emotional guardedness, and a marked aversion to intimacy, often manifesting as an avoidant defense mechanism to forestall potential exploitation or emotional distress.

2. The Depend Dimension

The Depend subscale evaluates the degree to which an individual believes that other people are reliable, trustworthy, and accessible when needed. This construct captures the core cognitive schema regarding the benevolence and dependability of the social world. High scorers exhibit trust that significant others will offer instrumental and emotional support during crises. Low scorers endorse pervasive cynicism regarding human reliability, operating under the assumption that “people are never there when you need them.” Such individuals often adopt compulsive self-reliance, preemptively rejecting assistance from romantic partners to avoid disappointment or betrayal.

3. The Anxiety Dimension

The Anxiety subscale assesses the extent to which an individual experiences intense apprehension regarding rejection, abandonment, emotional abandonment, and insufficient partner devotion. Rooted in an unstable or negative model of the self, high attachment anxiety triggers hyperactivating attachment strategies. Individuals with elevated Anxiety subscale scores constantly monitor their partner’s behavioral nuances for subtle cues of detachment, harbor an insatiable yearning for complete psychological merger (“enmeshment”), and routinely experience anxiety that their intense intimacy demands may overwhelm or alienate their partners. In contrast, individuals with low attachment anxiety exhibit high self-worth and confidence in their partner’s continued commitment.

The Two-Dimensional Orthogonal Representation: Avoidance and Anxiety

In subsequent psychometric scholarship (e.g., Collins, 1996; Brennan et al., 1998), researchers observed that the Close and Depend dimensions share substantial negative covariance, as both tap into fundamental avoidant tendencies. When combined, reverse-scored items from Close and Depend form an overarching Attachment Avoidance axis (discomfort with closeness and reluctance to depend on others), which stands orthogonal to Attachment Anxiety (fear of abandonment and rejection). This two-dimensional space aligns directly with the four-category attachment framework pioneered by Kim Bartholomew and Leonard M. Horowitz (1991):

• Secure: Low Anxiety, Low Avoidance (High Close, High Depend).
• Preoccupied: High Anxiety, Low Avoidance (High Close, High Depend/Ambivalent).
• Dismissing-Avoidant: Low Anxiety, High Avoidance (Low Close, Low Depend).
• Fearful-Avoidant: High Anxiety, High Avoidance (Low Close, Low Depend).

Theoretical Framework

The conceptual infrastructure of the AAS is firmly anchored in evolutionary-developmental psychology and interpersonal theory. The foundation was established by British psychoanalyst John Bowlby (1969, 1973, 1980) in his tri-volume treatise on attachment, separation, and loss. Bowlby posited that human infants possess an innate motivational control system—the attachment behavioral system—designed by natural selection to maintain proximity to a primary caregiver (the attachment figure) to ensure survival against predatory and environmental threats.

Mary Ainsworth and colleagues (1978) empirically validated Bowlby’s formulations via the laboratory-based Strange Situation protocol, identifying three organized attachment classifications in infancy: Secure (Pattern B), Anxious-Ambivalent/Resistant (Pattern C), and Anxious-Avoidant (Pattern A). Ainsworth showed that these patterns reflect adaptive behavioral strategies conditioned by the caregiver’s degree of sensitivity and responsiveness.

In 1987, Cindy Hazan and Phillip Shaver published a seminal paper extending attachment theory across the lifespan into adult romantic partnerships. They hypothesized that romantic love is an attachment process governed by the same biological and psychological systems that regulate infant-caregiver bonds. In adults, romantic partners assume the role of primary attachment figures, serving as a safe haven in moments of distress and a secure base from which the individual explores the world.

Collins and Read (1990) refined this conceptual bridge by integrating principles from contemporary social cognition. They argued that internal working models are not monolithic traits, but rather multi-layered cognitive network architectures comprising:

• Episodic Memories: Specific declarative recollections of past relational interactions.
• Beliefs and Expectations: Generalized schemas regarding whether the self is worthy of love and whether others can be relied upon to provide care.
• Goals and Needs: Motivational imperatives concerning autonomy, belongingness, and proximity.
• Defensive Strategies: Automated procedural rules that regulate emotional arousal and attentional processing when facing interpersonal threat.

The AAS was deliberately constructed to tap these underlying cognitive-affective components. By separating the evaluation of self-worth (Anxiety) from the evaluation of partner responsiveness (Depend) and preference for interpersonal distance (Close), Collins and Read provided an operational framework that translated abstract attachment constructs into measurable psychometric variables.

Validity

The Adult Attachment Scale has undergone rigorous empirical validation across thousands of participants in diverse clinical, cross-cultural, and relationship settings.

Construct and Factorial Validity

Construct validity was initially confirmed by Collins and Read (1990) across three independent college student samples ($N = 406$, $N = 289$, and $N = 101$ heterosexual dating couples). Confirmatory and exploratory evaluations confirmed that the three factors (Close, Depend, Anxiety) uniquely account for discrete variances in interpersonal functioning, distinct from general social desirability or uncalibrated neuroticism. In a comparative evaluation of self-report attachment instruments, Brennan, Clark, and Shaver (1998) analyzed 1,442 items from 60 attachment measures on a sample of 1,086 undergraduates, demonstrating that the AAS items loaded robustly onto the two higher-order canonical vectors of adult attachment: Avoidance and Anxiety.

Convergent Validity

The convergent validity of the AAS is evidenced by substantial correlations with parallel instruments designed to gauge relationship models and personality traits:

• Experiences in Close Relationships (ECR / ECR-R): The AAS Anxiety subscale correlates strongly ($r = .70$ to $.82$) with the ECR Anxiety subscale. Similarly, the composite Avoidance score from the AAS correlates robustly ($r = .65$ to $.78$) with the ECR Avoidance subscale.
• Relationship Questionnaire (RQ): Individuals classified as “Secure” on Bartholomew and Horowitz’s (1991) RQ score significantly higher on Close and Depend and lower on Anxiety. Those classified as “Preoccupied” display elevated Anxiety scores alongside moderate-to-high Close scores. “Dismissing” profiles exhibit low Anxiety coupled with very low Close and Depend scores, while “Fearful” profiles register high Anxiety alongside low Close and Depend scores.
• Self-Esteem Measures: Attachment Anxiety displays significant negative correlations with the Rosenberg Self-Esteem Scale ($r = -.45$ to $-.58$), confirming Bowlby’s hypothesis that attachment anxiety reflects an internal working model of the self as devalued or unlovable.

Discriminant Validity

Although the Anxiety subscale displays expected, moderate associations with the Big Five domain of Neuroticism ($r = .35$ to $.48$), structural equation modeling confirms that its predictive power regarding relationship outcomes remains significant after statistically controlling for broad neurotic personality variance. Furthermore, the Close and Depend subscales show weak-to-negligible correlations with social desirability scales ($r < .15$), demonstrating that respondent ratings reflect authentic attachment dynamics rather than defensive posturing or positive self-presentation.

Predictive and Criterion Validity

The AAS exhibits high criterion validity in predicting both individual psychological functioning and dyadic interaction patterns:

• Dyadic Relationship Quality: In their sample of dating couples, Collins and Read (1990) revealed that both men’s and women’s scores on the Close and Depend dimensions were positively correlated with partner satisfaction, trust, intimacy, and perceived commitment ($r$ values between $.30$ and $.55$). Attachment Anxiety was consistently linked to chronic relationship dissatisfaction, hyper-reactivity, and dysfunctional communication.
• Support-Seeking and Caregiving: Observational studies (e.g., Collins & Feeney, 2000) have demonstrated that individuals with high Depend scores actively seek emotional and instrumental support from their partners under high stress. Conversely, individuals with low Close scores actively withdraw, reject physical affection, and display defensive emotional flattening during stressful laboratory paradigms.

Reliability

The psychometric stability and internal consistency of the AAS have been demonstrated across numerous normative and clinical cohorts:

Internal Consistency

In the seminal psychometric validation study by Collins and Read (1990), internal consistency estimates (Cronbach’s alpha) for the 6-item subscales established strong scale reliability:

• Close Subscale: $\alpha = .69$ to $.77$
• Depend Subscale: $\alpha = .72$ to $.78$
• Anxiety Subscale: $\alpha = .72$ to $.85$

In Collins’ (1996) revised operationalization (the Revised Adult Attachment Scale, RAAS), which modified item wording to enhance internal consistency, alpha coefficients improved further, yielding $\alpha = .81$ for Close, $\alpha = .78$ for Depend, and $\alpha = .85$ for Anxiety. When scored using the two-dimensional framework (Avoidance vs. Anxiety), the expanded 12-item Avoidance composite routinely achieves Cronbach’s alpha values exceeding $.84$.

Test-Retest Stability

Attachment styles reflect relatively stable trait-like schemas that nonetheless remain susceptible to significant life transitions and relationship changes. Empirical evaluations of the AAS over various time horizons demonstrate solid stability:

• Two-Month Interval: Collins and Read (1990) observed test-retest correlations of $r = .68$ for Close, $r = .71$ for Depend, and $r = .52$ for Anxiety over a 2-month span. The slightly lower temporal stability for Anxiety reflects its heightened sensitivity to acute relational stressors and fluctuations in partner behavior.
• Long-Term Stability: Longitudinal evaluations spanning 8 months to 2 years have reported test-retest coefficients hovering between $.55$ and $.65$, indicating substantial longitudinal stability alongside sufficient plasticity to capture evolving relationship dynamics.

Factor Analysis

The structural composition of the Adult Attachment Scale has been thoroughly evaluated using both Exploratory Factor Analysis (EFA) and Confirmatory Factor Analysis (CFA).

Original Exploratory Factor Analysis (Collins & Read, 1990)

In developing the AAS, Collins and Read subjected the 18 decomposed items to principal-components analysis followed by orthogonal (Varimax) and oblique (Promax) rotations across multiple validation cohorts. The mathematical analysis cleanly yielded a three-factor solution explaining approximately 41.5% to 46.2% of total item variance. The factor structure loaded cleanly:

• Factor 1 (Depend): Items reflecting confidence in the availability and trustworthiness of others (e.g., “I know that others will be there when I need them”) exhibited primary loadings ranging from $.58$ to $.78$, with minimal secondary cross-loadings.
• Factor 2 (Anxiety): Items tapping profound fears of rejection, separation, and unreciprocated love (e.g., “I often worry that my partner does not really love me”) demonstrated high, distinct factor loadings ranging from $.61$ to $.81$.
• Factor 3 (Close): Items indexing comfort with interpersonal intimacy and mutual emotional contact (e.g., “I find it relatively easy to get close to others”) yielded primary loadings ranging from $.52$ to $.76$.

Confirmatory Factor Analysis and Structural Modeling

Subsequent psychometric investigations have tested the comparative fit of Collins and Read’s original three-factor model versus alternative structural formulations (such as a single-factor general attachment model or a two-factor orthogonal model consisting of Attachment Anxiety and Attachment Avoidance). While the three-factor model demonstrates adequate empirical fit:

• Comparative Fit Index ($ ext{CFI}$) =$.90 – .93$
• Tucker-Lewis Index ($ ext{TLI}$) =$.88 – .91$
• Root Mean Square Error of Approximation ($ ext{RMSEA}$) =$.055 – .068$
• Standardized Root Mean Square Residual ($ ext{SRMR}$) =$.051 – .062$

Modern psychometricians frequently collapse the Close and Depend dimensions into a single latent Avoidance construct due to high inter-factor correlation ($r = -.50$ to $-.68$). When reconfigured into a hierarchical model or a two-factor structure (Avoidance vs. Anxiety), CFA fit indices typically improve ($ ext{CFI} > .94$,$ ext{RMSEA} < .05$), fully corroborating the two-dimensional conceptual model that dominates contemporary adult attachment theory.

Instrument / Measurement Tool

The operational specifications of the Adult Attachment Scale (AAS) are structured as follows:

• Format: 18-item self-report questionnaire.
• Administration Modality: Paper-and-pencil, digital survey engine, or computer-assisted clinical evaluation.
• Administration Time: Approximately 3 to 6 minutes.
• Target Population: Adults and adolescents aged 16 years and older involved in, or reflecting upon, romantic and close interpersonal relationships.
• Response Metric: 5-point Likert rating format:
• 1 = Not at all characteristic of me
• 2 = Slightly characteristic of me
• 3 = Moderately characteristic of me
• 4 = Substantially characteristic of me
• 5 = Very characteristic of me
• Original Three-Subscale Scoring Key (Collins & Read, 1990):
• CLOSE Subscale: Assesses comfort with intimacy and closeness.
  Items: 1, 7, 9*, 13, 15*, 17*
• DEPEND Subscale: Assesses trust in the dependability of others.
  Items: 3*, 6, 8*, 14, 16*, 18*
• ANXIETY Subscale: Assesses fear of rejection and abandonment.
  Items: 2*, 4, 5, 10, 11, 12
• Note on original scoring: Items marked with an asterisk (*) must be reverse-scored prior to calculating composite scores ($1
  ightarrow 5, 2
  ightarrow 4, 3
  ightarrow 3, 4
  ightarrow 2, 5
  ightarrow 1$). Subscale scores are obtained by computing the mean of the designated items.
• Alternative Two-Dimensional Scoring Key (Brennan et al., 1998; Collins, 1996):
• ANXIETY Dimension: Items 2*, 4, 5, 10, 11, 12
• AVOIDANCE Dimension: Items 1*, 3, 6*, 7*, 8, 9, 13*, 14*, 15, 16, 17, 18
• Note on alternative scoring: Asterisked (*) items are reversed before calculating the unweighted mean score for each dimension.

Permissions & Fee and Test Year

The Adult Attachment Scale was officially published in 1990 in the Journal of Personality and Social Psychology. As an academic psychometric tool developed under institutional and federal research support, the instrument is in the public domain for non-commercial educational, scientific, and empirical research purposes. No royalty fees, formal registration charges, or proprietary test licensing fees are required to utilize the scale in scholarly investigations or therapeutic assessments.

Investigators and clinicians intending to employ the AAS in clinical trials, dissertation research, or cross-cultural adaptations are encouraged to cite the original seminal validation paper (Collins & Read, 1990). Inquiries regarding the instrument, permission for specialized adaptations, or requests for unpublished technical reports should be directed to Dr. Nancy L. Collins at [email protected].

References

• Ainsworth, M. D. S., Blehar, M. C., Waters, E., & Wall, S. (1978). Patterns of attachment: A psychological study of the strange situation. Lawrence Erlbaum Associates.
• Bartholomew, K., & Horowitz, L. M. (1991). Attachment styles among young adults: A test of a four-category model. Journal of Personality and Social Psychology, 61(2), 226–244. https://doi.org/10.1037/0022-3514.61.2.226
• Bartholomew, K., & Shaver, P. R. (1998). Methods of assessing adult attachment: Do they converge? In J. A. Simpson & W. S. Rholes (Eds.), Attachment theory and close relationships (pp. 25–45). Guilford Press.
• Bowlby, J. (1969). Attachment and loss: Vol. 1. Attachment. Basic Books.
• Bowlby, J. (1973). Attachment and loss: Vol. 2. Separation: Anxiety and anger. Basic Books.
• Bowlby, J. (1980). Attachment and loss: Vol. 3. Loss: Sadness and depression. Basic Books.
• Brennan, K. A., Clark, C. L., & Shaver, P. R. (1998). Self-report measurement of adult attachment: An integrative overview. In J. A. Simpson & W. S. Rholes (Eds.), Attachment theory and close relationships (pp. 46–76). Guilford Press.
• Collins, N. L. (1996). Working models of attachment: Implications for explanation, emotion, and behavior. Journal of Personality and Social Psychology, 71(4), 810–832. https://doi.org/10.1037/0022-3514.71.4.810
• Collins, N. L., & Feeney, B. C. (2000). A safe haven: An attachment-theoretical analysis of support-seeking and support-providing in relationships. Journal of Personality and Social Psychology, 78(6), 1053–1073. https://doi.org/10.1037/0022-3514.78.6.1053
• Collins, N. L., & Read, S. J. (1990). Adult attachment, working models, and relationship quality in dating couples. Journal of Personality and Social Psychology, 58(4), 644–663. https://doi.org/10.1037/0022-3514.58.4.644
• Collins, N. L., & Read, S. J. (1994). Cognitive representations of attachment: The structure and function of working models. In K. Bartholomew & D. Perlman (Eds.), Advances in personal relationships: Attachment processes in adulthood (Vol. 5, pp. 53–90). Jessica Kingsley Publishers.
• Hazan, C., & Shaver, P. (1987). Romantic love conceptualized as an attachment process. Journal of Personality and Social Psychology, 52(3), 511–524. https://doi.org/10.1037/0022-3514.52.3.511
• Mikulincer, M., & Shaver, P. R. (2007). Attachment in adulthood: Structure, dynamics, and change. Guilford Press.

Items of the Scale