Attachment styles and repeating patterns: a psychodynamic view

Where attachment theory comes from

John Bowlby and Mary Ainsworth built attachment theory from the 1950s onward; Bretherton's 1992 history traces how their separate and joint work fitted together. The core idea is that an infant forms a bond with a caregiver as a system for staying safe, and that the pattern of that bond becomes a template for what to expect from other people. Ainsworth's observational work distinguished secure infants from two insecure patterns, avoidant and anxious/ambivalent.

In 1987 Cindy Hazan and Phillip Shaver carried the idea into adult romance. Across two questionnaire studies, they reported that the three patterns appeared in adult couples in roughly the same proportions as in infants. Fraley's overview of the research describes the measure they used, in which people chose the paragraph that best described them: about 60% chose the secure paragraph, and about 20% each chose the avoidant and anxious ones.

That is the version that travelled onto the internet. It is also where the research moved on.

From types to two dimensions

Later work asked whether people fall into distinct boxes. Brennan, Clark and Shaver (1998) found that responses to attachment statements clustered into two dimensions: attachment anxiety, a worry about whether a partner will be available, and attachment avoidance, a preference for not relying on or opening up to others. A secure adult scores low on both. Together with taxometric research by Fraley and Waller (1998), this led most researchers to treat attachment as varying in degree, not in kind. All of this is as summarised in Fraley's overview.

So "I am anxious-avoidant" is a shorthand for a position on two continuous scales, not membership of a category with edges.

How stable is it?

Two honest answers sit side by side: moderately, and not completely.

Fraley's 2002 meta-analysis found that attachment security is moderately stable across the first 19 years of life, and that the prototype model, in which early representations keep shaping later patterns, best accounted for the data. In a 20-year follow-up by Waters and colleagues (2000), 60 infants were assessed at 12 months and 50 were reassessed in early adulthood; 72% kept the same secure or insecure classification. Negative life events, such as parental loss, divorce, serious illness or family abuse, were linked to change: 44% of infants whose mothers reported such events changed classification (8 of 18), against 22% of those without (7 of 32). The sample was small and homogeneous, so treat the figures as illustrative.

Fraley and Roisman (2019) draw four lessons from longitudinal research. Two matter here: the links between early care and adult attachment are modest, and early experiences do not determine adult outcomes. Attachment is more malleable in childhood and adolescence than in adulthood, but it is not frozen.

How it shows up in a real life

Consider a type of situation familiar to many clinicians: a person who, in several different relationships, finds that the moment someone becomes reliably available, they begin to feel crowded and look for a reason to leave. In another life, the pattern runs the other way: the more a partner steps back, the harder the person pursues. Both are recognisable positions on the two dimensions. Neither is a diagnosis.

A psychodynamic reading adds what the dimensions leave out. It asks what the closeness means to the person, whose face they expect behind a partner's, and what the repetition protects them from. This is the territory of repetition compulsion and transference: the expectation about what others will do is carried into new rooms, including the consulting room.

Two people, two positions

One feature of attachment patterns that quizzes cannot show is that a relationship has two people in it. A person who scores high on anxiety and a person who scores high on avoidance can fall into a loop together: one pursues, the other withdraws, and each reads the other's move as proof of what they already expected. The pattern lives in the pair as much as in either person. This is part of why a pattern can appear to change when the partner changes, and why one person's "style" in a first relationship may look different in the next.

It is also why an outside view can add something. How a friend, a sibling or a former partner describes you in relationships may differ from how you describe yourself. Neither version is the truth, but the gap between them is often where the work starts.

What the pattern is not

• It is not a verdict on your character. The two dimensions describe expectations about others, built from experience. Expectations can be revised.
• It is not a full explanation of why you repeat a relationship. Fraley and Roisman (2019) describe the links from early care to adult attachment as modest, which leaves room for other influences: family roles, the person's own wishes, and what they are working through now.
• It is not a reason to diagnose a partner. Labelling someone else's behaviour from a quiz tends to close a conversation.

How psychodynamic work uses attachment

Attachment is one lens, not the whole picture. Psychodynamic therapists tend to attend to how the pattern appears in the relationship with the therapist, as well as in the stories told about outside relationships. A method named for exactly that, transference-focused psychotherapy, was tested in a randomised trial by Levy and colleagues (2006): 90 patients with borderline personality disorder were assigned to transference-focused psychotherapy, dialectical behavior therapy or a modified psychodynamic supportive therapy for a year. Only the first group showed a significant rise in secure classifications on the Adult Attachment Interview, along with gains in reflective function.

Read that carefully. It is one trial, in one clinical group, using an interview measure. It supports the idea that attachment organisation can shift in treatment. It does not show that any given therapy will shift yours.

What online quizzes get wrong

• They return a type. Researchers largely moved to two dimensions; a four-box answer is a simplification of how the work is now scored.
• They measure one day. A self-report reflects your mood, your current relationship and how you read the question.
• They treat a pattern as a trait. The research above shows attachment can respond to experience.
• They stop at the label. The label does not say what the pattern is for, which is the question therapy tends to ask.

A quiz can still be a fair starting point. Fraley's own page links one that estimates anxiety and avoidance; use it as a question to take elsewhere.

Where TwentyThird fits

TwentyThird is a psychodynamic AI platform for self-understanding. Its agent, Oneli, holds written conversations and tracks what repeats in what a member says, including how the same relational pattern appears in different people and settings. Members can also invite the people who know them to answer questions about them, which adds an outside view. Supervision by a licensed psychologist is in early access. Read how the method works and what the evidence behind it is. It does not assign you an attachment style or diagnose you.

What this does not replace

A pattern named by an app or a quiz is a hypothesis. A psychologist who knows you can test it against your history, the way you behave in the room, and what you are ready to hear. If a relationship pattern is hurting you or someone else, bring it to a clinician. If you are in danger, contact local emergency services now. In the EU, call 112.