Adult Attachment Interview (AAI) Coding Model – Mary Main, Nancy Kaplan, & Ruth Goldwyn

The assessment of attachment in adult populations underwent a paradigm shift in the mid-1980s through the pioneering work of Mary Main, Nancy Kaplan, and Ruth Goldwyn at the University of California, Berkeley. Prior to this juncture, attachment theory, originally formulated by British psychoanalyst John Bowlby and empirically operationalized by American developmental psychologist Mary Ainsworth, was predominantly conceptualized and assessed at the level of overt, non-verbal infant behavior. Ainsworth’s Strange Situation procedure had firmly established that individual differences in infant-caregiver interactions could be reliably organized into secure, insecure-avoidant, and insecure-ambivalent behavioral classifications. However, the theoretical trajectory of attachment theory necessitated an empirical leap: demonstrating how these early sensorimotor and dyadic adaptations survive developmental transitions, consolidate into stable cognitive structures, and govern socio-emotional functioning across the human lifespan.

The creation of the Adult Attachment Interview (AAI) and its accompanying, rigorous system of narrative discourse analysis resolved this empirical impasse. Conceived by Mary Main, with critical empirical and conceptual refinements by Nancy Kaplan and Ruth Goldwyn, the AAI fundamentally transformed developmental and clinical psychology by moving attachment assessment from the level of behavioral observation to the level of mental representation. The AAI does not assess an adult’s attachment status through behavioral performance, self-report personality questionnaires, or conscious self-evaluations of relationship satisfaction. Instead, it measures how an individual organizes, structures, and verbally conveys their autobiographical history of childhood experiences with attachment figures while undergoing the cognitive and emotional demands of a standardized, semi-structured clinical interview.

At the center of this methodological breakthrough is the insight that defensive strategies, cognitive processing failures, and regulatory mechanisms are encoded directly within the linguistic architecture of speech. Drawing deeply upon H. Paul Grice’s philosophical framework of conversational pragmatics and cooperative communication, Main, Kaplan, and Goldwyn developed a 9-point scale-based coding manual that evaluates narrative coherence, semantic-episodic memory integration, metacognitive monitoring, and subtle lapses in cognitive reasoning. This comprehensive analysis will explore the historical genesis, structural architecture, linguistic underpinnings, dimensional rating scales, categorical classifications, psychometric validity, and clinical and intergenerational implications of the Adult Attachment Interview coding system.

1. Historical Genesis and Theoretical Foundations of the AAI

1.1 The Epistemological Shift to the Level of Representation

The conceptual origins of the Adult Attachment Interview emerged directly out of the Berkeley Longitudinal Study, an ambitious investigation initiated by Mary Main in the 1970s that tracked a cohort of infants assessed in Ainsworth’s Strange Situation into middle childhood and adolescence. As these children grew, researchers observed that early attachment patterns were remarkably persistent; however, the overt behavioral repertoire characteristic of infancy—such as physical proximity seeking, distress vocalizations, and avoidant physical gaze aversion—naturally transformed as children developed advanced linguistic, conceptual, and executive capacities. Bowlby had hypothesized that through continuous transactional interactions with primary caregivers, infants construct internal working models of self and others. These models operate as internal cognitive, affective, and dynamic representational schema that forecast, guide, and interpret attachment-relevant interactions.

Until the development of the AAI, internal working models remained largely theoretical constructs, invisible to direct empirical observation. Main recognized that if internal working models truly govern adult socio-emotional functioning, their influence must be perceptible in the organization of thought, language, and memory retrieval when the attachment behavioral system is intentionally activated. Thus, the Berkeley group orchestrated an epistemological shift from the level of observable interactive behavior to the level of internal, symbolic mental representation. Rather than observing what adults did when distressed, Main and her colleagues evaluated how adults thought about distress, vulnerability, and caregiving relationships.

The Berkeley Longitudinal Study provided the empirical cradle for validating this representational pivot. When the parents of the children evaluated in the Strange Situation were interviewed about their own childhood histories of parental care, an extraordinary pattern emerged: the linguistic organization and structural coherence of a parent’s discourse predicted their infant’s Strange Situation behavioral classification with unprecedented statistical power. This prospective finding demonstrated that an adult’s internal state of mind regarding attachment, encoded in linguistic narrative forms, serves as the primary psychological organizer of parenting behavior and the intergenerational transmission of relational security.

1.2 The Collaborative Contributions of Main, Kaplan, and Goldwyn

The construction of the Adult Attachment Interview protocol and its exhaustive coding and classification system represents an intellectual synthesis forged by Mary Main, Nancy Kaplan, and Ruth Goldwyn. Mary Main provided the overarching theoretical architecture, identifying the structural parallels between infant behavioral adaptations and adult linguistic strategies. Main observed that the behavioral avoidance seen in Strange Situation “Group A” infants—characterized by the organized diversion of attention away from the caregiver and the suppression of distress—reappeared in adult narrative discourse as a distinct, defensive deactivation of attachment memory, manifest in extreme brevity, memory lapses, and the derogation or idealization of parents. Conversely, the infant resistant/ambivalent behavioral strategy (“Group C”) mapped precisely onto linguistic hyperactivation: an inability to mentally disengage from attachment figures, reflected in entangled, emotionally dysregulated, and unending discourse.

Nancy Kaplan played a pivotal role in operationalizing these representational concepts across developmental stages. Her early doctoral research at Berkeley explored internal working models in six-year-olds using pictorial attachment assessments, family drawings, and emotional communication tasks. Kaplan’s insights directly informed how representational models regulate the flow of information, establishing that an individual’s defensive organization determines which aspects of autobiographical memory are granted conscious access, integrated into narrative structures, or defensively segregated from cognitive awareness. Her collaborative work with Main helped bridge the gap between childhood symbolic play and adult structural discourse.

Ruth Goldwyn was indispensable in translating these complex theoretical paradigms into a standardized, psychometrically rigorous narrative scoring manual. Working closely with Main over thousands of hours of interview analysis, Goldwyn meticulously cataloged the micro-linguistic phenomena, syntactic anomalies, and pragmatic indicators that define each coding scale. Goldwyn brought an exceptional ear for linguistic nuances and conversational cadences, ensuring that the scoring manual could be systematically taught to independent researchers with high inter-rater reliability. The resulting manual, “Adult Attachment Scoring and Classification System” (Main, Goldwyn, & Hesse, 2003), stands as one of the most comprehensive qualitative-quantitative coding instruments in psychological science.

1.3 Integration with Evolutionary and Psychodynamic Paradigms

The AAI coding model is theoretically grounded at the confluence of John Bowlby’s ethological-evolutionary paradigm and psychoanalytic drive and defense theory. Bowlby conceptualized attachment as an evolved, biologically homeostatic behavioral system designed to optimize survival by maintaining physical proximity to a protective caregiver under conditions of threat, illness, or environmental stress. When the primary caregiver is accessible and responsive, the behavioral system operates with open, flexible feedback loops. However, when the caregiver is persistently rejecting, unresponsive, or frighteningly unpredictable, the child must institute defensive adaptations to preserve proximity without incurring catastrophic relational rejection or psychic fragmentation.

Main, Kaplan, and Goldwyn recognized that the psychoanalytic mechanisms of defense—specifically repression, splitting, isolation of affect, projection, and denial—could be re-conceptualized through the lens of cognitive information processing and communicative pragmatics. Under the pressure of attachment-related anxiety, defensive processes manifest as failures in cognitive monitoring and structural disruptions in conversational flow. When the attachment system is activated by interview questions that probe memories of vulnerability, rejection, and trauma, the subject’s cognitive apparatus is subjected to severe regulatory strain.

Consequently, the AAI serves as a structural bridge between early sensorimotor relational adaptations and adult metacognitive monitoring. If an individual has developed an organized, secure representational model, their executive functions remain intact during attachment activation; they can simultaneously recall episodic trauma or parental rejection while maintaining conscious, objective metacognitive oversight of their narrative. Conversely, in individuals utilizing insecure or disorganized defenses, the activation of the attachment system results in involuntary cognitive intrusions, defensive dissociations, or the wholesale deactivation of cognitive processing, leaving distinct linguistic markers within the transcript.

2. The Adult Attachment Interview Protocol: Structure and Administration

2.1 Standardized Administration and Semi-Structured Architecture

The administration of the Adult Attachment Interview is strictly governed by a semi-structured protocol comprising a sequence of twenty standardized questions, delivered in a predetermined, invariant order. The interview was explicitly designed by Main and Goldwyn to “surprise the unconscious,” deliberately nudging the participant out of rehearsed, intellectualized platitudes and forcing real-time cognitive processing of emotionally charged attachment memories. The protocol begins with benign demographic and structural queries regarding the family constellation, but rapidly shifts toward deep, emotionally demanding autobiographical terrain.

The interviewer’s stance is rigidly prescribed: it must remain neutral, supportive, non-directive, and non-therapeutic. The interviewer does not offer interpretations, emotional comfort, psychoeducation, or validation of the participant’s distress. The goal is to provide a standardized, uniform interpersonal setting that ensures comparability across different transcripts. The interviewer acts as a calm, collaborative conversational partner whose inquiries function as standardized cognitive probes. The interviewer is instructed to systematically utilize specific follow-up prompts whenever a participant responds vaguely, generalizes without supporting evidence, or attempts to deflect the query.

The sequence begins by asking the participant to give a general overview of their childhood family environment, followed immediately by the task of selecting five adjectives that describe their early childhood relationship with their mother, and then five adjectives for their father. Once these semantic descriptors are established, the interviewer methodically returns to each individual adjective, asking the participant to provide specific, episodic memories illustrating why they selected that precise word. Subsequent questions systematically interrogate experiences of physical distress, emotional hurt, illness, experiences of rejection, discipline, parental threats, profound losses, traumatic abuses, and the overarching evolution of the participant’s relationship with their caregivers into adulthood.

2.2 Eliciting Episodic Versus Semantic Memory

The core neurocognitive engine of the AAI protocol is the intentional elicitation of dynamic tension between semantic memory and episodic memory, drawing upon Endel Tulving’s neurocognitive taxonomy of human memory systems. Semantic memory houses generalized conceptual knowledge, linguistic abstractions, and cultural scripts (e.g., “My mother was the most loving, selfless woman in the world”). In contrast, episodic memory contains autobiographical, spatiotemporally grounded representations of personally experienced events, complete with affective and sensory details (e.g., “When I fell out of the oak tree at age seven and broke my wrist, my mother locked me outside and told me to stop crying”).

By forcing the participant to initially generate high-level semantic adjectives and then immediately demanding distinct episodic narratives to substantiate those assertions, the AAI tests the structural integrity and internal consistency of the participant’s representational architecture. In individuals with secure states of mind, semantic generalizations are readily, fluently, and honestly supported by congruent episodic recollections; if an adjective like “supportive” was chosen, the subject can quickly recount an instance where support was tangibly provided, or if a negative adjective like “volatile” was chosen, illustrative episodes are provided with balanced perspective.

In contrast, insecure strategies reveal profound structural splits between these memory stores. A dismissing individual may provide superlative semantic descriptors (“loving,” “perfect,” “wonderful”), but when pressed for episodic corroboration, they produce complete recall failure (“I can’t remember anything specific”), irrelevant vignettes, or contradictory episodic memories of extreme neglect and abandonment. The AAI protocol systematically probes instances of vulnerability—asking participants where they went when emotionally upset, physically hurt, or ill—because these are the precise ecological conditions under which the evolutionary attachment system was designed to seek proximity. Assessing the cognitive retrieval of autobiographical memory under the specific emotional stress of these questions exposes how defenses manage relational affect.

2.3 Verbatim Transcription Protocols

The empirical validity of the Main, Kaplan, and Goldwyn coding model relies upon the absolute fidelity of the interview transcript. Because the primary indices of defensive organization are embedded in subtle linguistic disruptions, syntactic breaks, and communicative anomalies, the AAI cannot be scored from audio recordings alone, nor can it be scored from summarized or grammatically cleaned notes. The audio recording must be transcribed verbatim according to meticulous, standardized transcription protocols.

Every single acoustic event and conversational behavior must be explicitly preserved in the written text. This mandates the precise inclusion of:

• All disfluencies, false starts, stuttering, and self-interruptions (e.g., “I went—well, we didn’t—my mother never actually—”).
• Filler vocalizations and audible pauses, categorized by length (e.g., [pause, 4 seconds], [sighs], “uh,” “um,” “er”).
• Non-lexical acoustic phenomena, including laughter, chuckles, breathlessness, weeping, throat-clearing, and changes in vocal volume or cadence.
• Idiosyncratic grammatical errors, dialectical shifts, incomplete sentences, trailing thoughts marked by ellipses, and run-on compound structures.
• Overlapping speech, interviewer prompts, and conversational interruptions.

The structural layout of the transcript is formatted rigorously, with wide margins for coding notations and specific line numbering to allow exact textual referencing. Without this microscopic level of textual preservation, a coder cannot evaluate conversational turn-taking, pauses that indicate cognitive retrieval blockades, or subtle shifts into poetic, dissociative, or archaic syntax that mark traumatic disorganization. The transcript is treated as a clinical biopsy of the participant’s cognitive-linguistic operations.

3. Linguistic and Pragmatic Framework: Grice’s Maxims in AAI Coding

3.1 Application of H. Paul Grice’s Cooperative Principle

The theoretical breakthrough that enabled Mary Main and Ruth Goldwyn to transition from intuitive, qualitative assessments of narratives to a rigorous, standardized coding system was the operational application of philosopher H. Paul Grice’s Cooperative Principle. In his classic work “Logic and Conversation” (1975), Grice posited that human conversation is inherently a rational, collaborative enterprise governed by an overarching normative assumption: participants are expected to make their conversational contribution such as is required, at the stage at which it occurs, by the accepted purpose or direction of the talk exchange.

Grice established that successful, coherent communication requires participants to adhere to four fundamental conversational maxims: Quality, Quantity, Relation, and Manner. While Grice originally formulated these maxims within the philosophy of language to explain conversational implicatures and linguistic pragmatics, Main recognized that the unprovoked, systemic violation of these maxims during the discussion of attachment experiences reveals underlying defensive distortions of thought. In the AAI coding model, discourse coherence is operationalized directly as the sustained adherence to Gricean maxims throughout the entire interview protocol.

When an adult is asked to describe their early attachment relationships, a coherent individual remains collaborative, presenting a balanced, internally consistent, and clear account that respects the listener’s cognitive needs. Conversely, individuals whose attachment experiences are characterized by defensively organized internal working models will systematically violate Grice’s maxims. These conversational violations are not random intellectual or educational shortcomings; they represent structural psychological markers of emotional defenses, demonstrating where autobiographical memory retrieval is blocked, flooded, or structurally fragmented by attachment-related affect.

3.2 The Four Maxims in Attachment Discourse

The AAI scoring system methodically evaluates transcripts against each of Grice’s four maxims, treating specific linguistic anomalies as pathognomonic markers of distinct internal states of mind:

The Maxim of Quality (Be Truthful): This maxim requires that the speaker does not state that which they believe to be false, and does not assert claims for which they lack adequate evidence. In the AAI, violations of Quality occur when there are explicit contradictions within the transcript. The most common manifestation is the profound discrepancy between positive semantic generalizations and contradicting episodic narratives. A speaker violates Quality when they assert, “My parents never fought; we were the happiest family on earth,” and ten minutes later casually remarks, “My father would frequently smash the living room furniture during his drunken rages while my mother hid in the closet.” Quality violations also manifest as unacknowledged internal contradictions where the speaker shifts between mutually incompatible assertions without showing any awareness of their logical discordance.

The Maxim of Quantity (Be Succinct, Yet Complete): Quantity mandates that the speaker provide as much information as is required for the purpose of the exchange, while avoiding excessive verbosity. In attachment discourse, violations of Quantity manifest along two polar, defensive vectors:

• Under-elaboration (Extreme Brevity): Characteristic of dismissing (Ds) states of mind, the participant offers clipped, monosyllabic, or laconic responses. When asked about their relationship with their parents, they state, “It was fine. Normal. Nothing to report,” systematically withholding biographical information, leaving the interviewer to repeatedly probe for substance.
• Over-elaboration (Excessive Verbosity): Characteristic of preoccupied (E) states of mind, the participant produces torrential, unconstrained monologues. An inquiry regarding a childhood fever results in a twenty-minute, uninterrupted narrative detailing parental marital conflicts, current financial anxieties, and legal disputes, utterly overwhelming the conversational framework.

The Maxim of Relation (Be Relevant): Relation demands that the speaker make their conversational contributions strictly pertinent to the current inquiry. In AAI coding, violations of Relation occur when a participant wanders far afield from the topic introduced by the interviewer. When asked, “What did you do when you were physically hurt as a child?”, an individual violating Relation might reply, “Well, my mother always preferred cats to dogs, and let me tell you about the stray cat we found in 1984…” Relation violations frequently involve intrusive shifts into contemporary adult grievances, philosophical soliloquies, or unprompted commentaries about the interview process itself.

The Maxim of Manner (Be Perspicuous, Orderly, and Clear): Manner requires the speaker to avoid obscurity of expression, avoid ambiguity, and be orderly and grammatically transparent. Violations of Manner manifest as convoluted, ungrammatical, or drifting syntax. Speakers displaying high passivity of thought exhibit classic Manner violations: sentences trail off into indefiniteness (“and all that kind of stuff, you know, whatever…”), terms of vague reference proliferate (“and then they did the thing with the stuff”), and the subject shifts grammatical voice from the first person (“I felt terrified”) to distancing, second-person or third-person generalizations (“Well, you know how when one is a kid, you just naturally don’t care”). Severe Manner violations also include sudden, jarring shifts into archaic, eulogistic, or dissociative phrasing, often signaling unresolved trauma.

3.3 Metacognitive Monitoring and Reflexive Awareness

A critical dimension of discourse analysis formulated by Main and Goldwyn is the presence or absence of metacognitive monitoring. Metacognitive monitoring refers to an individual’s real-time, reflexive awareness of the fallibility, construction, and cognitive nature of their own memories, thoughts, and narrative assertions. It is the capacity to think about one’s own thinking while actively engaged in social discourse.

Within the AAI, metacognitive monitoring is evidenced when a participant spontaneously acknowledges the subjective, constructed status of their recollections, distinguishes clearly between their childhood perceptions and their contemporary adult evaluations, and actively checks their assertions for potential communicative ambiguity. For instance, a speaker demonstrating high metacognitive monitoring might state: “I remember my father as being an absolute tyrant when I was seven; but thinking back on it now, he had just lost his brother and was working two night shifts, so what I perceived as callous anger was probably just sheer exhaustion and grief. Of course, my memory of that summer might be colored by what my older sister told me later.”

Metacognitive monitoring allows a speaker to spontaneously correct linguistic slips, notice contradictions in their own narrative, and repair violations of Gricean maxims on the fly. Main and Goldwyn establish a vital clinical and coding distinction between authentic, reflexive metacognitive monitoring and performative “pseudo-coherence.” Pseudo-coherence is an intellectualized, rigid, or rehearsed narrative posture wherein a participant utilizes psychoanalytic or psychological jargon (e.g., “My mother was narcissistic and emotionally unavailable due to her unworked-through Oedipal conflicts”) to present a facade of understanding. The skilled AAI coder recognizes that such rehearsed rhetoric lacks true reflexive flexibility, emotional spontaneity, and genuine metacognitive insight, collapsing under the specific episodic probes of the interview.

4. The AAI Coding System: Scales of Probable Childhood Experience

4.1 Constructing the Parental Experience Profile

The AAI coding system operates across a dual-tier evaluative structure: the first tier reconstructs the participant’s probable childhood experiences, while the second tier evaluates the participant’s current state of mind regarding attachment. Main and Goldwyn recognized that in order to accurately assess how an individual’s mind defends against or integrates relational reality, the coder must first infer, as objectively as possible, the actual caregiving environment to which the individual was exposed during their first sixteen years of life.

The Scales of Probable Childhood Experience evaluate maternal and paternal figures on distinct, independent 9-point inferential scales (ranging from 1 = completely absent or no evidence, to 9 = extreme, pervasive presence). It is of paramount methodological importance that these scales do not reflect what the participant claims occurred; rather, they reflect what the coder infers actually took place, based upon the rigorous, objective corroboration of specific episodic memories, historical timelines, and factual events recounted across the entire transcript.

If a participant enthusiastically claims, “My mother was the most devoted, nurturing mother imaginable,” but all cited childhood episodes reveal that she left the child home alone for days at a time to go drinking, the coder assigns a low score on Loving behavior and a high score on Neglect. The coder constructs a historical reconstruction, functioning much like an investigative forensic clinician, evaluating the objective plausibility of caregiver behavior while strictly filtering out the participant’s subjective rationalizations, emotional justifications, or defensive obfuscations.

4.2 Loving and Rejecting Behavior Scales

The two foundational pillars of the parental experience profile are the Loving and Rejecting scales, which capture the affective core of the early caregiver-child relationship.

The Loving Scale (1 to 9): The Loving scale measures the inferred extent to which the parent provided a secure emotional base, warmth, unconditional acceptance, and sensitive responsiveness to the child’s attachment signals. A score of 9 is reserved for caregivers who exhibited continuous, authentic emotional availability, sustained physical affection, validation of the child’s subjective emotional life, and prompt comfort during times of physical distress or developmental distress. A score of 1 reflects an absolute absence of loving behavior. Coders are trained to assign high loving scores only when the participant can substantiate assertions of warmth with clear, concrete, rich episodic examples demonstrating the parent acting selflessly, sensitively, and protectively in response to the child’s explicit needs.

The Rejecting Scale (1 to 9): The Rejecting scale measures the inferred degree to which the caregiver openly rebuffed, pushed away, ridiculed, or penalized the child’s proximity-seeking and attachment behaviors. Rejecting behavior is not simply a lack of warmth; it is an active, aversive response to the child’s expressions of dependency, vulnerability, sadness, or fear. A score of 9 indicates pervasive, severe emotional or physical rejection, such as a parent explicitly stating they never wanted the child, banishing the child from the home when distressed, laughing at the child’s crying, or demonstrating cold hostility. Coders must distinguish between:

• Direct, overt rejection: Active physical abandonment, hostile verbal mockery, or explicit commands to suppress tears and emotional needs.
• Covert, intellectualized rejection: Chilling emotional aloofness, severe physical stiffness when the child attempts proximity, or subtle disgust displayed when the child requires physical care during illness.

4.3 Neglecting, Involving/Role-Reversing, and Pressure to Achieve

In addition to Loving and Rejecting, the Main and Goldwyn system assesses three specific patterns of pathogenic parental care that exert profound developmental distortions upon internal working models:

The Neglecting Scale (1 to 9): Neglecting measures systematic psychological or physical unavailability, abandonment, and profound failures of parental oversight and stewardship. A high score on this scale reflects a home environment where the caregiver was physically absent, catatonically depressed, substance-dependent, or utterly disengaged from the child’s developmental reality. Unlike Rejection, which is an active, aversive pushing away of proximity, Neglect is an ambient, passive vacuum: the parent simply fails to notice, feed, supervise, protect, or interact with the child, leaving them to navigate the physical and social world in profound existential isolation.

The Involving/Role-Reversing Scale (1 to 9): Also known as the Parentification scale, this dimension operationalizes the extent to which the caregiver inverted the natural generational hierarchy, turning the child into their own attachment figure, confidant, emotional spouse, or psychic caretaker. In a role-reversing dynamic (scored 7 to 9), the parent’s emotional equilibrium depends entirely upon the child’s performance, soothing, or continuous companionship. The child is burdened with the parent’s marital complaints, suicidal ideations, or existential dread. Coders look for historical episodes where the child felt responsible for the parent’s emotional survival, felt intense guilt when attempting independent peer activities, or was forced to manage the functional operations of the household due to parental incompetence or infantilism.

The Pressure to Achieve Scale (1 to 9): This scale measures parental instrumentalization of the child, wherein love, approval, and relational safety are explicitly made contingent upon external performance, social status, academic perfection, or athletic prowess. In families scoring high on Pressure to Achieve, the child is treated not as an authentic human subject with unique emotional needs, but as a narcissistic extension of the parental ego. The parent displays intense pride and affection when the child wins accolades, but instantly withdraws affection, becomes icily critical, or subjects the child to severe punishment when performance is merely average or fails. This dynamic cultivates an internal working model wherein the self is experienced as inherently unlovable, valuable only as a functional instrument of achievement.

5. The AAI Coding System: Scales of Current State of Mind

5.1 Idealization of Parents

The evaluation of the participant’s current state of mind with respect to attachment constitutes the core of the AAI classification system. While the experience scales assess what happened in childhood, the state of mind scales evaluate how the adult’s mind currently processes, organizes, and defends against those past events. The Idealization of Parents scale is scored separately for mother and father on an anchored 1 to 9 continuum and represents one of the most reliable markers of a dismissing (Ds) organization.

Idealization is operationalized as a marked structural discrepancy between the semantic representations used to describe a caregiver and the actual episodic memories produced to corroborate those claims. An individual scoring high on Idealization (7 to 9) insists on painting the caregiver in glowing, superlative, or unblemished terms (“She was a saint,” “He was the greatest father any boy could ever dream of”), yet when pressed by the interviewer to provide an episodic memory of a time the parent was loving, the participant:

• Experiences total cognitive retrieval failure (“I can’t think of any specific times, we were just always happy”).
• Offers generalized, ritualized descriptions rather than autobiographical episodes (“Well, she cooked dinner every day”).
• Provides memories that directly contradict the superlative semantic descriptor (“He was so great; whenever I cried, he locked me in the basement until I learned to be a man”).

The coder assigns an idealization score based entirely on the presence of this semantic-episodic split, paired with the participant’s apparent lack of awareness regarding the contradiction. A score of 1 indicates an objective, balanced appraisal where semantic claims are fully substantiated by rich, congruent episodic narratives, with both strengths and flaws acknowledged naturally.

5.2 Involving Anger and Current Derogation

The AAI captures two diametrically opposed linguistic methods of managing negative affect regarding attachment figures: active, uncontained entanglement (Involving Anger) versus cool, dismissive minimization (Current Derogation).

Involving Anger (Mother and Father scored independently, 1 to 9): Involving Anger is the quintessential state of mind marker for preoccupied (E) attachment. It does not measure calm, objective anger about past abusive or negligent events; rather, it captures live, ongoing, uncontained, and psychologically unintegrated rage that actively floods the interview discourse. The participant appears caught in an ongoing psychic struggle with the parent, speaking as if the historical offense occurred mere moments ago. Linguistic markers include:

• Run-on, grammatically uncontrolled sentences filled with vitriol, sarcasm, and rhetorical questions.
• Direct, second-person accusatory shifts, where the speaker abruptly addresses the parent as if they are in the room (“And then she says to me, ‘How could you do this?’ and I say, ‘No, how could *you* do this, Mother!'”).
• Drafting the interviewer into the role of a courtroom judge, juror, or referee, soliciting agreement (“Can you believe that? Wouldn’t anyone be furious?”).

Current Derogation of Attachment (Mother, Father, and Attachment Relationships scored, 1 to 9): Derogation represents an aggressive, dismissive deactivating strategy. Rather than becoming emotionally flooded like the angry-preoccupied speaker, the derogating individual treats attachment figures, attachment vulnerability, and the interview itself with icy contempt, cynical mockery, and intellectualized belittlement. Caregivers are dismissed with biting, derogatory epithets (“She’s just a pathetic, hysterical mess,” “My father is completely irrelevant to my life, like a piece of old furniture”). The participant sneers at their own childhood distress, dismissing emotional needs as ridiculous, foolish, or disgusting. Derogation serves to defensively strip the attachment figure of all psychological significance, protecting the individual from acknowledging the profound pain of early rejection.

5.3 Passivity of Thought and Vagueness

The Passivity of Thought scale (1 to 9) evaluates an elusive yet structurally devastating discourse pathology, serving as the central identifying marker for the passive-preoccupied (E1) classification. Passivity reflects an internal working model characterized by profound cognitive drifting, lack of agency, attentional entanglement, and an inability to maintain a sustained, coherent focus on a single autobiographical line of thought.

In passive discourse, the speaker repeatedly loses the conceptual thread of the conversation. The linguistic indicators of passivity are subtle, pervasive, and unmistakable:

• Sentence abandonment and trailing thoughts: The participant regularly begins a sentence, falters, and allows the thought to trail off into complete silence or meaningless filler phrases (“And then my dad came home and he was… well, you know, just, sort of, whatever…”).
• Indefinite words and empty speech: Excessive reliance on non-specific terms such as “things,” “stuff,” “all that jazz,” “whatnot,” and “you know the drill.”
• Childlike vocal cadences and linguistic regressions: The adult speaker suddenly begins utilizing childish terminology (“My mommy was an icky person,” “I was an owie-prone kid”), revealing an unmediated emotional collapse back into the childhood developmental state.
• Unrepaired oscillations: The speaker articulates two mutually exclusive perspectives within the exact same breath without noticing or attempting to resolve the contradiction (“My father was always there for me, well, he was never home and lived in another country, but he was completely supportive”).

When passivity reaches severe levels (scores of 7 to 9), the entire transcript acquires a foggy, ethereal, ungrounded quality. The interviewer’s questions fail to gain traction, and the narrative floats through associational tangents, leaving the coder with an overwhelming sense that the participant lacks an integrated, autonomous ego center capable of reflecting upon their own life history.

5.4 Transcript Coherence and Mind Coherence Scales

The structural apex of the Main, Kaplan, and Goldwyn coding model resides in the two overarching macro-scales: Coherence of Transcript and Coherence of Mind. These 9-point anchor scales integrate all qualitative, linguistic, and structural observations into a definitive measure of narrative organization.

Coherence of Transcript (1 to 9): Coherence of Transcript evaluates the degree to which the interview adheres cleanly, elegantly, and collaboratively to Grice’s maxims of Quality, Quantity, Relation, and Manner. A highly coherent transcript (scored 7 to 9) presents a seamless, internally consistent, and well-contextualized autobiographical account. The speaker is open, emotionally honest, concise yet adequately descriptive, strictly relevant to the inquiries, and linguistically clear. Crucially, transcript coherence does not require a happy, benign childhood; an individual who experienced horrifying parental abuse, severe abandonment, and chronic trauma can receive an 8 or 9 on transcript coherence if they can describe those catastrophic events with psychological truthfulness, emotional balance, syntactic order, and clear perspective.

Coherence of Mind (1 to 9): While Transcript Coherence focuses heavily on conversational pragmatics and communicative collaboration, Coherence of Mind assesses the underlying, internal representational architecture of the self. It evaluates whether the participant possesses a unified, non-fragmented, reality-oriented psychological structure. A high score on Coherence of Mind indicates that the individual has successfully processed and integrated their developmental history, whatever that history was. They do not maintain incompatible representations of their parents, they do not defensively segregate memories from affect, and they do not succumb to dissociative lapses when confronting painful material.

The scoring anchors for the Coherence scales are calibrated with exquisite psychometric precision:

• Score 1 (Profound Fragmentation): Complete incoherence. The transcript is riddled with irreconcilable contradictions, massive lapses in reasoning, severe passivity, uncontained vitriolic rages, or pervasive idealization that utterly collapses under questioning. Grice’s maxims are violated continuously on nearly every page.
• Score 3 (Markedly Incoherent): Systematic, organized defensive distortions. The narrative is heavily dominated by either rigid dismissing deactivation (severe idealization, under-elaboration) or entangled preoccupied hyperactivation (excessive verbosity, passivity, involving anger).
• Score 5 (Borderline/Mixed Coherence): The narrative exhibits moderate coherence punctuated by intermittent defensive breakdowns. The participant may narrate one parental relationship with balance and clarity, but become defensive, angry, or laconic when addressing the other parent.
• Score 7 (Clearly Coherent): The discourse is substantially collaborative, objective, and well-organized. Minor, isolated violations of Gricean maxims may occur under intense emotional pressure, but the participant quickly demonstrates metacognitive monitoring, repairing the lapse spontaneously.
• Score 9 (Exceptional Integration): Rare, profound narrative integration. The participant demonstrates crystalline clarity, deep emotional authenticity, exquisite collaborative balance, and advanced metacognitive awareness. The narrative flows with emotional grace, acknowledging complex ambivalences without defensive distortions.

6. Classification Group F: Autonomous / Free with Regard to Attachment

6.1 Core Discourse Markers of Group F

Participants classified into Group F (Autonomous / Free with Regard to Attachment) possess an internal working model characterized by open information processing, emotional security, and an autonomous, integrated state of mind regarding attachment. The definitive psychometric signature of Group F is elevated scores on both Coherence of Transcript and Coherence of Mind, typically falling solidly within the 6 to 9 range, coupled with low scores on Idealization, Derogation, Involving Anger, and Passivity of Thought.

The discourse of an Autonomous individual is marked by a fundamental stance of non-defensive acceptance toward their personal history. When asked to evaluate their childhood relationships, they do not attempt to protect their parents through idealizing platitudes, nor do they seek revenge through vindictive, angry monologues. Instead, they present an objective, balanced appraisal, acknowledging both the positive strengths and the inevitable limitations or failures of their caregivers. They discuss painful memories, losses, and rejections with appropriate, modulated affect, openly feeling the sadness or hurt of those experiences without becoming overwhelmed, flooded, or structurally disorganized.

Linguistically, Group F discourse embodies Grice’s Cooperative Principle in its purest form. The speaker provides succinct yet vivid episodic illustrations to support their semantic assertions (adhering to Quality and Quantity), responds directly to the interviewer’s questions without rambling or drifting into irrelevant tangents (adhering to Relation), and expresses their thoughts with syntactic clarity, precision, and order (adhering to Manner). Furthermore, autonomous speakers demonstrate consistent metacognitive monitoring, viewing their parents historically, within the context of the parents’ own generational backgrounds, marital difficulties, and socioeconomic realities.

6.2 Continuous-Secure Versus Earned-Secure Attachment

One of the most theoretically profound discoveries yielded by the AAI coding system is the empirical distinction between two distinct developmental pathways leading to an Autonomous state of mind: Continuous-Secure and Earned-Secure attachment.

Continuous-Secure Individuals: These individuals enjoyed an early childhood environment characterized by high inferred parental Loving behavior, sensitive caregiving, and low rejection, neglect, or trauma. Their high narrative coherence reflects a direct, uninterrupted continuation of an early secure infant-caregiver attachment. Their autobiographical recall is fluid, uninhibited, and naturally congruent; the warm, supportive representations established in their sensorimotor infancy have simply matured into an articulate, balanced, and confident representational state of mind in adulthood. Their discourse flows with an effortless, unforced emotional openness.

Earned-Secure Individuals: In sharp, dramatic contrast, Earned-Secure individuals experienced childhood caregiving environments that were objectively adverse, abusive, deeply rejecting, or neglectful, yielding low scores on parental Loving scales and high scores on Rejecting, Neglecting, or Role-Reversing scales. Despite this severe developmental adversity, the Earned-Secure participant narrates their painful history with extraordinary coherence, objective perspective, and emotional integration, scoring 6 or higher on Transcript Coherence.

The linguistic discourse of the Earned-Secure individual bears clear structural evidence of arduous emotional transformation, psychic separation, and therapeutic resolution. They do not minimize the trauma, nor are they consumed by current rage toward the abusive parent. Instead, they articulate how they survived the adversity, frequently noting the healing influence of alternative attachment figures (such as a supportive grandparent, mentor, or romantic partner) or long-term psychotherapeutic intervention. The identification of Earned Security provides definitive empirical proof that internal working models are not rigid, deterministic scripts; they are dynamic, open representational systems capable of substantial reorganization through corrective emotional experiences across the lifespan.

6.3 Sub-Classifications within Group F

To capture the subtle stylistic nuances within the Autonomous organization, Main and Goldwyn established five distinct sub-classifications (F1 through F5):

F1: Somewhat Set Aside / Moderately Dismissing Tendencies: While achieving an overall coherent narrative that meets the threshold for Group F, the F1 transcript exhibits minor, trace elements of dismissing deactivation. The speaker may show a slight preference for independence, provide slightly brief answers to questions regarding vulnerability, or briefly express a desire to “leave the past in the past.” However, unlike a truly dismissing individual, when pressed, the F1 speaker readily engages, substantiates their claims with clear episodic memories, and demonstrates authentic narrative integration.

F2: Continuous/Moderately Contextualized: A balanced, secure profile where the individual exhibits slight deactivating or intellectualizing tendencies that are seamlessly integrated into an objective, overarching perspective. The individual may discuss painful experiences with a slightly intellectualized tone, but their discourse remains collaborative, deeply thoughtful, and grounded in rich, uncontradicted autobiographical evidence.

F3: The Prototypical Autonomous Narrative: The absolute gold standard of psychological security. The F3 transcript represents the quintessential balance between semantic and episodic memory, pristine adherence to Gricean maxims, effortless emotional fluency, and advanced metacognitive awareness. Whether narrating joyful or agonizing life events, the F3 speaker maintains total collaborative rapport, emotional honesty, and reflective insight. There is no trace of defensive idealization, derogatory contempt, angry entanglement, or passivity.

F4: Somewhat Preoccupied / Emotionally Resonant: A secure narrative profile displaying minor, trace elements of preoccupied hyperactivation. The F4 speaker may exhibit slight emotional volatility, highly detailed and vivid recollections of past relational conflicts, or mild conversational verbosity. However, unlike a truly preoccupied individual, the F4 speaker never loses their metacognitive anchor; they successfully pull back from emotional intensity, recognize their own affective charge, and maintain narrative coherence and collaborative control.

F5: Unresolved Attachment-Related Trauma / Loss (with High Coherence Base): Assigned to individuals whose overarching state of mind regarding general attachment relationships is robustly autonomous and secure, but who exhibit localized, contained, or minor indicators of unresolved loss or trauma that do not cross the quantitative threshold for a primary Unresolved/Disorganized (U/d) classification. The structural core of the mind remains resilient and integrated.

7. Classification Group Ds: Dismissing of Attachment

7.1 Core Defensive Architecture of Group Ds

Classification Group Ds (Dismissing of Attachment) represents the representational manifestation of the avoidant behavioral adaptations observed in Ainsworth’s Strange Situation. The overarching defensive architecture of the Dismissing state of mind is organized around cognitive deactivation of the attachment behavioral system. Deactivating strategies are designed to systematically divert attention away from attachment-related thoughts, memories, affects, and wishes, neutralizing any perceived need for relational proximity, intimacy, or emotional reliance upon others.

The internal working model of the Dismissing individual is built upon an insistence on complete personal self-reliance, emotional invulnerability, and autonomy. Relational vulnerability is perceived as intrinsically hazardous; having historically experienced an attachment figure who was rejecting, cold, or punitive when proximity was sought, the individual learned that the safest survival strategy is to defensively suppress the conscious activation of attachment distress. Consequently, in the AAI, the Dismissing participant attempts to convince the interviewer—and themselves—that their childhood was unblemished, that parental rejection was completely inconsequential, and that emotional relationships play no significant role in their current adult life.

Psychologically, the Dismissing strategy requires the active, continuous cognitive compartmentalization of painful autobiographical experiences. Affect is isolated from cognition. When confronted with interview probes designed to activate attachment vulnerability, the Dismissing individual mobilizes a suite of defensive maneuvers: claiming complete amnesia for early childhood, normalizing neglectful or abusive parental practices as “good discipline that made me strong,” and vigorously dismissing the emotional significance of childhood injuries, illnesses, or separations.

7.2 Linguistic and Structural Hallmarks

The linguistic manifestation of the Dismissing state of mind is characterized by pervasive, organized violations of Grice’s maxims of Quality and Quantity:

Severe Violations of Quality (Through Idealization): As detailed previously, the hallmark of the Ds transcript is the unresolvable split between semantic generalizations and episodic recall. The participant opens with glowing, superlative semantic claims regarding the caregivers, but completely fails to corroborate these claims with congruent autobiographical episodes. When the interviewer notes that the cited memories depict coldness or rejection rather than love, the participant dismisses the contradiction or rapidly changes the subject.

Severe Violations of Quantity (Extreme Laconic Brevity): Dismissing transcripts are frequently among the shortest in absolute word count. The speaker routinely provides clipped, laconic, monosyllabic answers, actively refusing to expand upon their narratives. Probes regarding emotional distress are met with conversational dead ends:

• Interviewer: “What did you do when you were upset as a child?”
• Participant: “Nothing.”
• Interviewer: “Do you remember any specific time you were upset and went to someone?”
• Participant: “Nope. I just dealt with it myself. Always have.”

This persistent withholding of biographical data violates the Maxim of Quantity, placing an immense burden upon the interviewer, who must repeatedly pry for the barest autobiographical details.

Pervasive Normalization and Amnesia: When forced to acknowledge objectively abusive or rejecting caregiver behavior (e.g., being struck with belts or locked outside in winter), the Ds speaker engages in instant, defensive normalization: “It was completely normal, all parents did that back then, it was good for me, it taught me respect.” Alternatively, the speaker deploys defensive amnesia as a primary shield: “I don’t remember my childhood at all. My memory begins at age eighteen. I just don’t think about the past; it’s useless to dwell on things.” This amnesia is not organic; it is a selective, functional cognitive blockade specifically targeting attachment-relevant autobiographical nodes.

7.3 The Sub-Classifications: Ds1 through Ds4

Main and Goldwyn delineated four distinct sub-classifications within the Dismissing group, reflecting the diverse structural strategies utilized to achieve cognitive deactivation:

Ds1: Idealizing: The prototypically dismissing profile. The Ds1 transcript is organized entirely around the defense of parental idealization. The speaker exhibits elevated scores on Idealization of Mother, Father, or both (typically 6 to 9 on the 9-point scale). The narrative is marked by continuous, unyielding positive semantic portrayals of the family, completely unbacked or directly contradicted by episodic recollections. The speaker maintains a polished, cheerful, and unruffled facade, aggressively deflecting any attempts by the interviewer to probe beneath the superficial surface of the family mythology.

Ds2: Devaluing of Attachment: The Ds2 profile achieves deactivation not through polite idealization, but through active, contemptuous derogation. The speaker scores high on the Current Derogation of Attachment scale. Rather than claiming their childhood was wonderful, the Ds2 speaker openly describes their parents as useless, contemptible, or utterly irrelevant. Attachment relationships, tenderness, love, and psychotherapy are dismissed with biting sarcasm, cynical laughter, and intellectualized hostility. By actively devaluing and sneering at the entire realm of human attachment, the Ds2 individual immunizes themselves against the conscious awareness of relational deprivation.

Ds3: Restricted in Feeling: Characterized by profound narrative constriction, flat emotional cadence, and severe emotional blunting. The Ds3 speaker does not exhibit the aggressive derogation of Ds2 nor the overt idealization of Ds1; instead, their transcript is marked by a pervasive, robotic neutrality. Affect is completely evacuated from the text. Childhood experiences of trauma, abandonment, or illness are recounted with the emotional detachment of an actuary reading a dry balance sheet. The speaker is completely alienated from their own affective life, presenting an internal working model wherein feelings are simply non-existent entities.

Ds4: Fear of Death / Unconscious Vulnerability: A specialized dismissing sub-classification marked by the continuous, defensive suppression of deep existential terror regarding mortality, physical vulnerability, or catastrophic illness. While the speaker maintains a standard dismissing, self-reliant posture regarding interpersonal relationships, the narrative reveals involuntary, defensive leaks regarding extreme fear of physical dissolution, death of the self, or medical trauma. The individual’s deactivating defenses are strained to the limit by the inescapable reality of biological vulnerability.

8. Classification Group E: Preoccupied with Attachment

8.1 Hyperactivation and Entanglement

Classification Group E (Preoccupied with Attachment) represents the representational counterpart to the insecure-resistant/ambivalent (Group C) infant behavioral pattern. The primary regulatory strategy organizing the Preoccupied state of mind is hyperactivation of the attachment behavioral system. Hyperactivating strategies involve the continuous, heightened vigilance to attachment-related cues, the persistent maximization of vulnerability and distress, and an absolute inability to mentally disengage from historical and contemporary attachment conflicts.

The internal working model of the Preoccupied individual is fundamentally entangled with their primary caregivers. Unlike the Dismissing individual who attempts to push the past away, the Preoccupied individual is hopelessly drowning in the past. There is no psychic boundary between historical childhood events and current emotional life; past parental failures, cruelties, and rejections are experienced as live, active wounds that continue to bleed into the immediate present. The individual’s attentional apparatus is permanently hijacked by attachment dynamics, preventing the consolidation of an autonomous, stable ego identity.

Psychologically, hyperactivation maintains proximity to inconsistent, unpredictable, or role-reversing caregivers by remaining perpetually distressed, angry, or infantilized. The Preoccupied individual has learned that emotional calm, self-reliance, or independent functioning results in parental abandonment or severe guilt; therefore, the attachment system is kept in a state of permanent, high-voltage arousal. In the AAI, this internal dynamic produces discourse that is chaotic, emotionally uncontained, syntactically fragmented, and overwhelming to the conversational partner.

8.2 Discourse Pathology: Violations of Quantity and Manner

The Preoccupied transcript is an extraordinary linguistic showcase of massive, unconstrained violations of Grice’s maxims, concentrated primarily within the maxims of Quantity and Manner:

Catastrophic Violations of Quantity (Excessive Verbosity): Preoccupied transcripts are often massive documents, sometimes spanning hundreds of pages. The speaker suffers from a profound inability to inhibit narrative output. A simple inquiry regarding where the family lived when the participant was five triggers an unconstrained deluge of biographical data, familial gossip, psychological speculations, and ongoing domestic disputes. The speaker floods the interview with irrelevant narrative tributaries, utterly losing track of the original question and overwhelming the interviewer’s capacity to process the information.

Pervasive Violations of Manner (Convoluted, Entangled, and Drifting Syntax): The internal chaotic flooding of affect wrecks the syntactic structure of the speaker’s language. Sentences stretch across an entire page, bound together by endless conjunctions, parenthetical insertions, abrupt changes of tense, and grammatical breakdowns:

• Syntactic Run-ons: “And my mother—who never understood my brother anyway because he was always running around with that terrible crowd down by the docks which my father warned her about back in 1978 before they even bought the blue station wagon—she comes in and screams at me for not cleaning my room, which wasn’t even dirty, but you know how she always gets that look in her eye, like she’s about to explode, and so I said…”
• Rhetorical Invocations and Involving Jargon: The speaker continuously attempts to collapse conversational distance, demanding the interviewer participate in their psychic reality through linguistic tags like “Right?”, “You know what I mean?”, “Wouldn’t you?”, or utilizing psychobabble to articulate ongoing parental combat.

Violations of Relation (Current Attentional Capture): Preoccupied individuals continuously violate the Maxim of Relation by hijacking the interviewer’s focus. Asked about their childhood, they launch into a detailed diatribe regarding a telephone argument they had with their mother that very morning, completely incapable of maintaining historical focus.

8.3 The Sub-Classifications: E1 through E3

Main and Goldwyn delineated three distinct sub-classifications within Group E, differentiating the narrative posture through which hyperactivation and entanglement are expressed:

E1: Passive / Drifting / Vague: The E1 profile is defined by elevated scores on the Passivity of Thought scale (6 to 9). The speaker’s strategy is not active, angry confrontation, but rather helpless, childlike drifting. The E1 narrative is characterized by linguistic vagueness, sentence abandonment, childlike terminology, and ethereal non-sequiturs. The speaker presents as a fragile, helpless child trapped in an adult body, incapable of making definitive assertions or holding a sustained, clear line of thought regarding their own life. The parental figures are experienced as massive, enveloping clouds that completely obscure the individual’s subjective reality.

E2: Angry / Conflicted / Argumentative: The E2 profile is the classic embodiment of Involving Anger (scoring 6 to 9 on Involving Anger for one or both parents). The transcript is an unconstrained, vitriolic battleground. The speaker is actively, furiously fighting with their parents across the pages of the interview. The language is sharp, sarcastic, accusatory, and emotionally dysregulated. The speaker utilizes direct quotation, reenacts historical arguments with dramatic vocal shifts, and continually appeals to the interviewer to validate their profound, unhealed, and burning sense of relational grievance.

E3: Fearfully Preoccupied / Traumatized Entanglement: The E3 profile captures individuals who are hyperactivated by pervasive, ambient fear and trauma. The speaker is paralyzed by dread regarding past or ongoing attachment catastrophes—such as parental madness, terrifying domestic violence, or overwhelming environmental chaos. The narrative is haunted by persistent feelings of catastrophic vulnerability, terror, and helplessness. The speaker has not developed an organized defense of anger (E2) or passivity (E1); they remain permanently in the behavioral grip of raw attachment alarm, terrified of parental abandonment, insanity, or death.

9. The Unresolved / Disorganized (U/d) Classification

9.1 Etiology and Operational Criteria for Unresolved Status

The formulation of the Unresolved / Disorganized (U/d) classification represents one of Mary Main’s most groundbreaking contributions to developmental psychopathology, emerging directly out of her prior discovery (with Judith Solomon) of the Disorganized/Disoriented (Group D) infant attachment pattern in the Strange Situation. In infancy, disorganized attachment is triggered when the caregiver serves simultaneously as the biological source of fear and the biological haven of safety—a catastrophic relational paradox that Bowlby termed “fright without solution.” The infant’s behavioral system collapses into temporary paralysis, freezing, stereotypies, and asymmetrical movements because the evolutionary drive to flee from danger conflicts directly with the drive to flee to the attachment figure.

In the Adult Attachment Interview, Main and Goldwyn discovered that this underlying disorganization reemerges in adulthood as a localized, catastrophic breakdown of cognitive and metacognitive monitoring specifically when the participant is asked to discuss experiences of loss through death or experiences of physical, sexual, or severe emotional abuse. The Unresolved (U/d) classification is not assigned based on whether a person suffered trauma; it is assigned based entirely on whether the cognitive processing and linguistic reporting of that trauma displays acute structural disorientation.

Methodologically, the U/d classification is assigned on a 9-point scale evaluating Unresolved Loss and Unresolved Trauma. A score of 5 or higher warrants a primary classification of U/d. Crucially, U/d is never assigned in isolation; it is always assigned alongside an underlying, secondary organized classification (F, Ds, or E) that reflects the individual’s overarching baseline state of mind when not discussing the unresolved event (e.g., U/d-Ds1, or U/d-F3). During discussion of trauma, however, the central executive architecture of the mind undergoes temporary micro-dissociation.

9.2 Lapses in the Monitoring of Reasoning

Lapses in the Monitoring of Reasoning are extraordinary linguistic events wherein the speaker, in the course of discussing a death or trauma, articulates causal beliefs, temporal sequences, or physical assertions that violate basic physical laws, biological reality, or normative logic. The speaker makes these assertions spontaneously, without metaphoric framing and without realizing that their statements are irrational or mutually impossible.

Main and Goldwyn cataloged several distinct manifestations of reasoning lapses:

• Attribution of Physical Reality to Deceased Figures (Magical Thinking): The speaker asserts that a deceased individual is simultaneously dead and physically alive in the spatial world. For example, a participant speaking of their mother who died ten years prior casually states: “She died in 2012, but she still comes into my kitchen on Tuesdays to help with the laundry, physically turning the dials on the washer.” The coder must verify that this is not poetic metaphor or cultural/spiritual convention, but a concrete cognitive belief.
• Temporal Disorientation / Spatial Collapse: The speaker becomes deeply confused regarding the temporal boundaries of life and death, stating: “I was born three years after my father died, so he raised me strictly during my elementary school years.”
• Bizarre Causal Beliefs Regarding Self-Responsibility: The speaker attributes the death of an attachment figure entirely to their own childhood childhood thoughts, minor misbehaviors, or magical agency: “My grandfather had a fatal heart attack in another city because I didn’t finish my oatmeal that morning. I killed him.”
• Confusion Between the Self and the Abuser: During the narration of abuse, the speaker’s boundaries of psychic identity dissolve: “When my father was beating me, I was looking down from the ceiling and I was holding the stick, hitting myself to purge the devil.”

9.3 Lapses in the Monitoring of Discourse

While lapses in reasoning reflect failures of logic and reality testing, Lapses in the Monitoring of Discourse represent acute disruptions in conversational pragmatics, executive focus, and communicative awareness. During the discussion of loss or trauma, the participant’s language suddenly detaches from the interpersonal reality of the interview setting.

The primary markers of discourse lapses include:

• Sudden Intrusion of Eulogistic, Archaic, or Poetic Language: In the middle of an otherwise mundane conversational cadence, the speaker abruptly shifts into an elevated, funereal, scriptural, or archaic register when mentioning a deceased individual: “And so she was borne away on the wings of the celestial cherubim into the eternal ether of the boundless firmament, where the sacred dew droppeth forevermore.” This is spoken with solemn, trance-like gravity, utterly ignoring the conversational context.
• Protracted, Unprompted Silences: The participant falls completely silent in the middle of a sentence regarding trauma, freezing for an extended duration (often 10, 20, or even 40 seconds) without acknowledging the freeze. When they resume speaking, they do not complete the sentence, frequently resuming on an entirely different topic, indicating an episode of micro-dissociation or an internal cognitive void.
• Inappropriate Sensory Detail: The speaker becomes flooded with hyper-specific, macabre sensory recollections that have no communicative relevance to the question, describing the precise smell of rotting flesh, the visual splattering of blood, or the microscopic texture of an abuser’s skin, speaking in a robotic, detached whisper.
• Loss of Conversational Anchor: The participant suddenly forgets who the interviewer is, where they are, or asks bizarre rhetorical questions that reveal a temporary loss of situational orientation.

10. The Cannot Classify (CC) Category and Atypical Profiles

10.1 Structural Criteria for the Cannot Classify Designation

The Cannot Classify (CC) designation, originally conceptualized by Mary Main and extensively systematized by Erik Hesse, represents the ultimate structural breakdown of organized attachment strategies. In standard AAI coding, an interview is placed into one of the three primary organized categories (Group F, Group Ds, or Group E), even if it also receives a secondary Unresolved (U/d) status. The fundamental assumption of the tripartite model is that human beings possess an organized, overarching strategy—either deactivating (Ds), hyperactivating (E), or integrated (F)—to manage attachment information.

In a Cannot Classify transcript, this overarching strategic organization has completely disintegrated. The participant does not possess a single, coherent or insecure strategy; instead, they exhibit mutually exclusive, diametrically contradictory states of mind within the same interview text. The criteria for assigning a Cannot Classify designation are met under several explicit structural conditions:

• Co-existence of Opposing Extreme States of Mind: The transcript displays high Idealization of one parent (e.g., Mother Idealization scored 7 or 8, indicating a Dismissing strategy) coupled simultaneously with severe Involving Anger toward the other parent (e.g., Father Involving Anger scored 7 or 8, indicating a Preoccupied strategy).
• Global Narrative Strategy Shifts: The participant begins the interview utilizing an intensely Dismissing strategy (extreme brevity, cold normalization, denial of memory), but midway through the protocol—typically following the questions on trauma or loss—abruptly collapses into unconstrained Preoccupied flooding (torrential verbosity, catastrophic syntactic breakdown, deep passivity).
• Pervasive Unresolved Trauma Across Multiple Domains: An interview that scores 7 or higher on multiple Unresolved Loss and Trauma scales, such that the disorganization is no longer a localized lapse, but an ambient, structural dissolution of the entire cognitive apparatus.

10.2 Clinical Correlates and Severe Psychopathology

The Cannot Classify profile is exceptionally rare in normative, non-clinical cohorts (typically appearing in less than 2-3% of general population samples). However, in psychiatric, forensic, and severe clinical populations, CC prevalence rates skyrocket, frequently appearing in 30% to 50% of evaluated individuals. Empirical research has demonstrated a robust, direct correlation between the CC classification and the most severe forms of psychological disturbance:

Borderline Personality Disorder (BPD): The rapid, unintegrated oscillations between idealization and devaluation that define BPD are the precise psychological equivalent of the linguistic contradictions seen in CC transcripts. The individual cannot maintain a continuous, stable internal working model of relational figures; caregivers are split into all-good, idealized saviors or all-bad, terrifying abusers, with the narrative violently careening between these poles.

Complex Post-Traumatic Stress Disorder (CPTSD) and Dissociative Disorders: CC is strongly linked to chronic, developmental relational trauma—such as enduring childhood incest, sadistic physical torture, or severe parental psychosis. In these individuals, the developmental architecture of the ego was systematically shattered before any organized attachment strategy could consolidate. The resulting narrative is characterized by profound dissociative fragmentation, amnesic barriers, and identity alteration.

Severe Suicidality and Inpatient Psychopathology: Transcripts of individuals with histories of lethal suicide attempts, intractable treatment-resistant depression, and severe substance dependence disproportionately receive the Cannot Classify designation. The structural inability to organize thought and affect around attachment crises leaves the individual utterly defenseless when confronted with interpersonal stress in adult life.

10.3 Coding Dilemmas and Differential Diagnostics

Coding a potential Cannot Classify transcript represents the most formidable challenge a certified AAI rater can encounter. It requires meticulous differential diagnosis to avoid two dangerous coding errors: prematurely defaulting to CC when an organized strategy is actually present, or forcing an organized classification onto a narrative that is fundamentally fragmented.

Differentiating Severe Passivity (E1) from Global Collapse (CC): Highly passive speakers (E1) can appear superficially disorganized; their sentences trail off, their thoughts are vague, and their language is childlike. However, the skilled coder recognizes that severe passivity is itself an *organized* strategy of hyperactivating helplessness. In E1, the passivity is continuous, uniform, and homogenous throughout the text. In CC, by contrast, passivity is violently interrupted by sudden bursts of rigid idealization or cold derogation, shattering the stylistic unity of the narrative.

Managing Conflicting Sub-scale Signals: Coders frequently encounter transcripts that contain conflicting micro-markers—such as a slightly elevated Idealization score alongside a moderate Passivity score. Main and Goldwyn establish explicit hierarchical scoring rules: a transcript must not be classified as CC unless the conflicting scales cross the critical clinical threshold of 6 or higher on the 9-point scale. If the contradictory scores are low (e.g., Idealization 4, Passivity 4), the coder must determine which strategy organizes the majority of the interview, utilizing the sub-classifications (e.g., F1, F4, Ds1, E1) to capture the minor counter-thematic elements without invoking the severe CC designation.

11. Psychometric Rigor, Training, and Methodological Considerations

11.1 Rater Training and Certification Standards

The Adult Attachment Interview coding system is internationally renowned for its uncompromising psychometric rigor and the demanding standards required for coder certification. Conceived by Mary Main and Erik Hesse, the training process was intentionally constructed to prevent the subjective, impressionistic drift that often plagues qualitative discourse analysis. Becoming a reliable, certified AAI coder is an intellectual endeavor comparable to mastering a complex neuropsychological test battery.

The standard certification pipeline requires researchers to attend an intensive, two-week full-time residential training institute led by an authorized master trainer. During these institutes, candidates are immersed in the microscopic, sentence-by-sentence analysis of dozens of training transcripts, learning to operationalize Gricean maxims, distinguish between semantic and episodic memory, and calibrate the 9-point scales. Following the institute, candidates must undergo a rigorous, blind reliability testing process lasting several months. Candidates are sent an un-coded test set of thirty to forty highly complex, standardized benchmark transcripts drawn from diverse clinical and non-clinical cohorts.

To achieve formal certification, the candidate’s ratings must achieve an exceptionally high statistical concordance with the master coding consensus established by Main and Hesse:

• At least 80% to 85% agreement on the major 3-way categorical classifications (F, Ds, E).
• At least 80% agreement on the assignment of the Unresolved (U/d) and Cannot Classify (CC) status.
• High inter-rater correlation coefficients (typically Cronbach’s alpha or intraclass correlations exceeding .80) across all 9-point dimensional scales.

In empirical research contexts, the methodological standard mandates dual-independent, blind coding. Two certified raters code transcripts independently, without any access to the participant’s demographic background, psychopathology, or Strange Situation infant data. Discrepancies are resolved through consensus conferences or by submitting the transcript to a third master coder, ensuring that the resulting classifications represent objective, psychometrically immaculate data.

11.2 Reliability and Discriminant Validity

Over four decades of empirical investigation, the AAI coding model has demonstrated extraordinary psychometric stability, construct validity, and discriminant validity across thousands of empirical studies published in major psychological journals.

Test-Retest Stability: Longitudinal studies assessing individuals across intervals of one year, five years, and even fifteen years have revealed remarkable test-retest stability for AAI classifications in non-clinical cohorts. In the absence of major intervening life events—such as catastrophic relational trauma, profound bereavement, or intensive, reconstructive psychoanalysis—an individual’s state of mind regarding attachment remains exceptionally stable across the adult lifespan. The cognitive-linguistic architecture tapped by the AAI behaves not as a fleeting emotional state, but as a deeply entrenched structural trait.

Discriminant Validity from Verbal IQ and General Cognitive Intelligence: A critical empirical question historically raised by critics was whether AAI coherence simply reflects general verbal intelligence, higher education, narrative eloquence, or cognitive processing speed. Decades of research have thoroughly demolished this critique. Studies measuring verbal IQ, working memory, and general cognitive ability alongside the AAI consistently demonstrate that narrative coherence is statistically independent of IQ. Highly articulate, intellectually brilliant academics often produce intensely dismissing, idealizing, or preoccupied transcripts that score near the bottom of the coherence scale, while individuals with limited formal education and modest vocabularies frequently achieve exceptional scores of 8 or 9 on transcript coherence, narrating their lives with humble, crystalline truthfulness and balanced emotional integration.

Independence from Autobiographical Memory Capacity and Narrative Talent: Research has further confirmed that AAI coherence does not measure general storytelling talent or autobiographical memory capacity. The AAI coding model does not evaluate how well someone constructs an entertaining, dramatic story; it evaluates whether their discourse adheres to Gricean truthfulness, succintness, relevance, and clarity. A participant can have severe memory deficits regarding factual dates and historical trivia, yet achieve an Autonomous classification if they manage those limitations with honest, transparent metacognitive monitoring (e.g., “I simply cannot recall my early childhood home, and it’s important I don’t make things up to fill the silence”).

11.3 Cross-Cultural Applicability and Sociolinguistic Critiques

As the AAI coding system expanded across the globe, developmental researchers confronted fundamental questions regarding its cross-cultural validity. Can a coding instrument rooted in Western conversational philosophy (Grice) and developed within a Californian university sample be applied validly to individuals in Japan, sub-Saharan Africa, South America, or indigenous cultures? Sociolinguistic critics argued that communication norms vary widely across human societies: what appears as dismissing laconic brevity in a middle-class American sample might reflect culturally normative modesty, reticence, or filial piety in a traditional Confucian culture; conversely, what appears as angry preoccupation might reflect normative conversational expressivity in Mediterranean or Latin American contexts.

In response, massive cross-cultural meta-analyses—such as those conducted by Marinus van IJzendoorn and Abraham Sagi-Schwartz—demonstrated that the foundational distribution of AAI classifications is remarkably consistent across diverse global cultures. The human infant’s need for an accessible, sensitive attachment figure, and the adult’s capacity to speak about relational vulnerability with coherence and truthfulness, appears to be an evolutionary universal that transcends local linguistic variations.

However, cross-cultural application requires strict translation and sociolinguistic adaptation protocols. Master coders must understand local linguistic idioms, conversational cadences, and cultural norms regarding parental authority. Certified translators and native-speaking coders are trained to preserve the pragmatic micro-markers—such as hesitation pauses, polite conversational hedging, and culturally specific terms of emotional reference—ensuring that the underlying cognitive-defensive structure is accurately captured without penalizing culturally sanctioned communication styles.

12. Clinical Applications and Intergenerational Transmission

12.1 The Intergenerational Transmission of Attachment

The most profound empirical triumph of the Adult Attachment Interview coding model is its unrivaled capacity to demonstrate the intergenerational transmission of attachment. In the late 1980s and 1990s, seminal longitudinal studies conducted by Peter Fonagy, Miriam Steele, and Howard Steele in London, alongside independent replications worldwide, established a breathtaking predictive finding: an expectant mother’s AAI classification assessed prior to the birth of her first child predicts that infant’s Strange Situation classification with that mother at twelve months of age with an accuracy rate exceeding 75% to 80%.

The predictive correspondences map with mathematical elegance:

• An Autonomous (Group F) mother overwhelmingly raises a Secure (Group B) infant.
• A Dismissing (Group Ds) mother overwhelmingly raises an Insecure-Avoidant (Group A) infant.
• A Preoccupied (Group E) mother overwhelmingly raises an Insecure-Resistant/Ambivalent (Group C) infant.
• An Unresolved (U/d) mother overwhelmingly raises a Disorganized/Disoriented (Group D) infant.

This predictive power led researchers to explore the precise mediating mechanism bridging the mother’s internal mind and the infant’s behavioral development. Peter Fonagy and Mary Target demonstrated that the primary psychological engine mediating this transmission is Parental Reflective Functioning—the parent’s capacity to mentalize, envisioning the infant as a psychological agent with their own distinct intentional states, desires, thoughts, and emotional feelings. An Autonomous parent, possessing an integrated, non-defensive state of mind, can accurately read the infant’s subtle non-verbal distress signals, mirroring and regulating those affects without becoming defensively cold (Dismissing) or emotionally overwhelmed (Preoccupied). This sensitive responsiveness provides the relational soil within which the infant constructs a secure working model of self and other.

Crucially, the identification of Earned-Secure attachment demonstrated that the cycle of intergenerational transmission is not an unalterable genetic destiny. When an adult processes their childhood adversity, achieving high narrative coherence through corrective relational experiences or psychotherapy, they break the transgenerational cycle of trauma, raising secure children despite their own painful history.

12.2 Application in Psychotherapy and Mentalization-Based Interventions

In contemporary clinical practice, the AAI has transcended its origins as a developmental research tool, evolving into a sophisticated diagnostic and treatment-planning instrument within psychoanalysis, relational psychotherapy, and mentalization-based interventions. Administering an AAI at the intake phase of long-term psychotherapy provides the clinician with a high-resolution, structural map of the patient’s defensive architecture, cognitive blind spots, and focal attachment vulnerabilities.

Knowledge of a patient’s AAI classification fundamentally dictates clinical technique:

• Working with Dismissing (Ds) Patients: The clinician knows that the patient will deploy cognitive deactivation, intellectualization, laconic minimization, and defensive amnesia. Directly demanding emotional vulnerability will only trigger deeper withdrawal or cynical derogation. The therapist must patiently help the patient slowly mentalize somatic sensations, gradually puncture idealizing defenses through gentle reality testing, and validate the deep, unacknowledged sorrow beneath the self-reliant facade.
• Working with Preoccupied (E) Patients: The clinician must actively serve as a cognitive and emotional container, establishing rigid therapeutic boundaries to prevent the patient from flooding the sessions with chaotic, torrential verbosity. The technical focus is on deceleration, dampening hyperactivating affect, analyzing Involving Anger, and assisting the patient to step out of the familial battlefield to claim their own autonomous, differentiated identity.
• Working with Unresolved (U/d) Patients: The therapist must anticipate acute dissociative ruptures, magical causal thinking, and paralyzing terror whenever traumatic material is approached. The work requires grounding techniques, trauma processing, and the meticulous restoration of reality testing and metacognitive monitoring.

Furthermore, serial AAI administrations have become the gold standard for tracking structural, characterological change in psychotherapy outcome research. True structural cure is evidenced not merely by the reduction of symptomatic distress, but by the measurable transformation of the patient’s narrative coherence, showing an empirical movement from Insecure (Ds, E) or Unresolved (U/d) classifications toward an Earned-Autonomous (F) state of mind.

12.3 Contemporary Research Trajectories and Neurobiological Correlates

In the twenty-first century, the Adult Attachment Interview has forged profound interdisciplinary alliances with cognitive neuroscience, functional neuroimaging, epigenetics, and computational linguistics, validating Main, Kaplan, and Goldwyn’s model at the biological level.

Functional Neuroimaging (fMRI) Correlates: Contemporary neuroimaging studies have demonstrated that the distinct linguistic strategies observed in the AAI correspond to distinct patterns of neural circuit activation. When exposed to attachment-relevant emotional stimuli (such as facial expressions of distress or maternal cues):

• Dismissing individuals exhibit profound hypo-activation of the limbic system (specifically the amygdala and insula), paired with hyper-activation of the dorsolateral prefrontal cortex (dlPFC), reflecting the conscious, effortful neural suppression of emotional processing.
• Preoccupied individuals exhibit sustained, runaway hyper-activation of the amygdala, anterior cingulate cortex (ACC), and temporal poles, with marked hypo-activation of prefrontal executive inhibitory regions, visually confirming the failure of top-down affective regulation.
• Autonomous individuals demonstrate balanced, synchronized functional connectivity between prefrontal executive networks and subcortical emotional centers, reflecting seamless emotional-cognitive integration.

Physiological and Epigenetic Markers: Research examining autonomic nervous system functioning reveals that during the AAI, Dismissing individuals display massive surges in electrodermal skin conductance, heart rate, and cortisol secretion, despite maintaining a cool, detached, and unruffled behavioral posture. Their physiological bodies are in a state of high autonomic panic, even as their linguistic apparatus insists that “everything was completely fine.” Epigenetic research has identified distinct methylation patterns in the oxytocin receptor gene (OXTR) and glucocorticoid receptor genes associated with Unresolved and Insecure AAI classifications, illustrating how relational adversity literally engraves itself into the biological expression of the human genome.

Computational Linguistics and Machine Learning: The latest frontier in AAI methodology involves the application of natural language processing (NLP), machine learning algorithms, and automated linguistic analysis (such as LIWC – Linguistic Inquiry and Word Count) to interview transcripts. Researchers are training deep learning neural networks on certified AAI databases to automatically detect syntactic anomalies, Gricean violations, semantic-episodic discrepancies, and micro-pauses. While computational tools cannot yet replace the clinical intuition and holistic evaluation of a certified human master coder, they are opening unprecedented avenues for large-scale epidemiological screening, bringing the visionary insights of Mary Main, Nancy Kaplan, and Ruth Goldwyn into the digital era of psychological science.

Conclusion: The Enduring Legacy of the AAI Coding Model

The Adult Attachment Interview coding model formulated by Mary Main, Nancy Kaplan, and Ruth Goldwyn represents a monumental achievement in the history of developmental psychology and psychological assessment. By orchestrating a profound epistemological shift from observable behavior to internal representation, and by operationalizing the assessment of internal working models through the rigorous lens of linguistic pragmatics, they provided clinical science with an objective, replicable window into the human mind. The insight that an individual’s emotional security is encoded not in the idyllic perfection of their childhood history, but in their capacity to speak about that history with truthfulness, clarity, balance, and coherence, remains one of the most liberating and clinically transformative discoveries in psychological science.

Through its rigorous dimensional scales, its robust categorical taxonomies, and its profound predictive power regarding the intergenerational transmission of attachment, the AAI has forever altered our understanding of human development, psychopathology, and the therapeutic process. It stands as a testament to the fact that while we cannot alter the objective historical events of our early childhood, we possess the neuroplastic and psychological capacity to reorganize how those events are held within our minds, achieving an integrated, autonomous state of mind that breaks the cycle of historical trauma and fosters psychological freedom across generations.

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.
• 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.
• Fonagy, P., Steele, H., & Steele, M. (1991). Maternal representations of attachment during pregnancy predict the organization of infant-mother attachment at one year of age. Child Development, 62(5), 891–905. https://doi.org/10.2307/1131141
• Fonagy, P., & Target, M. (1997). Attachment and reflective function: Their role in self-organization. Development and Psychopathology, 9(4), 679–700. https://doi.org/10.1017/s0954579497001399
• Grice, H. P. (1975). Logic and conversation. In P. Cole & J. L. Morgan (Eds.), Syntax and semantics: Vol. 3. Speech acts (pp. 41–58). Academic Press.
• Hesse, E. (2008). The Adult Attachment Interview: Protocol, method of analysis, and empirical studies. In J. Cassidy & P. R. Shaver (Eds.), Handbook of attachment: Theory, research, and clinical applications (2nd ed., pp. 552–598). The Guilford Press.
• Main, M., & Goldwyn, R. (1984). Adult attachment scoring and classification system. Unpublished scoring manual, Department of Psychology, University of California, Berkeley.
• Main, M., Goldwyn, R., & Hesse, E. (2003). Adult attachment scoring and classification system (Version 7.2). Unpublished scoring manual, Department of Psychology, University of California, Berkeley.
• Main, M., & Kaplan, N. (1985). Security in infancy, childhood, and adulthood: A move to the level of representation. Monographs of the Society for Research in Child Development, 50(1/2), 66–104. https://doi.org/10.2307/3333827
• Main, M., Kaplan, N., & Cassidy, J. (1985). Security in infancy, childhood, and adulthood: A move to the level of representation. In I. Bretherton & E. Waters (Eds.), Growing points of attachment theory and research. Monographs of the Society for Research in Child Development, 50(1-2, Serial No. 209), 66–104.
• Main, M., & Solomon, J. (1986). Discovery of an insecure-disorganized/disoriented attachment pattern. In T. B. Brazelton & M. W. Yogman (Eds.), Affective development in infancy (pp. 95–124). Ablex Publishing.
• Main, M., & Solomon, J. (1990). Procedures for identifying infants as disorganized/disoriented during the Ainsworth Strange Situation. In M. T. Greenberg, D. Cicchetti, & E. M. Cummings (Eds.), Attachment in the preschool years: Theory, research, and intervention (pp. 121–160). University of Chicago Press.
• Tulving, E. (1985). Memory and consciousness. Canadian Psychology/Psychologie canadienne, 26(1), 1–12. https://doi.org/10.1037/h0080017
• van IJzendoorn, M. H. (1995). Adult attachment representations, parental responsiveness, and infant attachment: A meta-analysis on the predictive validity of the Adult Attachment Interview. Psychological Bulletin, 117(3), 387–403. https://doi.org/10.1037/0033-2909.117.3.387
• van IJzendoorn, M. H., & Bakermans-Kranenburg, M. J. (1996). Attachment representations in mothers, fathers, adolescents, and clinical groups: A meta-analytic search for normative data. Journal of Consulting and Clinical Psychology, 64(1), 8–21. https://doi.org/10.1037/0022-006X.64.1.8