The Question the Research Actually Answers

Whether attachment styles can change is a question that generates a lot of confident assertions in psychology-adjacent spaces, mostly in the direction of "not really" — the implication being that anxious people will always be anxious in relationships, avoidant people will always withdraw, and the best anyone can hope for is awareness of their pattern and enough self-management to keep it from destroying their relationships. The evidence for this view is real but partial. Attachment styles do show meaningful stability across time and across relationships. The internal working models developed in early childhood — the templates for how attachment figures behave and how worthy of care the self is — are not easily updated. But stability is not immutability, and the research on attachment change is more encouraging than the popular framing suggests.

What the Longitudinal Studies Show

The most rigorous longitudinal work on attachment in adults has found that attachment classifications change for a meaningful proportion of people over the course of years. Research from the University of California, Berkeley's Institute of Human Development, following participants across several decades, found that roughly 20 to 25 percent of participants showed different attachment classifications at follow-up than at baseline, with changes in both directions — some becoming more secure over time, some becoming less so. What predicted the changes was specific: positive relationship experiences with a responsive partner were the strongest predictor of movement toward security. Significant loss or trauma — particularly loss of close attachment figures — was the strongest predictor of movement toward insecurity. The internal working model, it turns out, is a working model. It updates in response to experience, though slowly and with considerable inertia.

The Mechanism of Change

Understanding why change is slow clarifies what change actually requires. Internal working models are not beliefs in the propositional sense — they are not simply thoughts about relationships that can be corrected by new information. They are procedural templates: expectations about how attachment figures will respond that organize attention, emotion, and behavior before conscious thought enters the picture. When an anxiously attached person perceives a small ambiguity in a partner's behavior — a slightly distracted response, a delayed text — the internal working model generates an interpretation and an emotional response before the conscious mind has the chance to apply more calibrated reasoning. The response is automatic, fast, and based on templates laid down in early experience. The fact that the partner is actually reliable does not automatically update the template. The template updates only through repeated experience of violations of its expectations that are specific enough and consistent enough to constitute evidence against it. This is why "just trust them more" is not adequate advice. The template cannot simply be overridden by a decision.

The Tangent About Therapeutic Relationships

One of the more interesting findings in attachment research is that the therapeutic relationship itself appears to function as an attachment relationship capable of updating internal working models — slowly, over the course of sustained work. The therapist who is consistently available, responsive, and non-retaliatory in the face of the client's testing behaviors provides a relational experience that the internal working model has to reckon with. A 2021 meta-analysis from researchers at the University of Amsterdam reviewing 47 studies on attachment-oriented therapy found significant effects on attachment security outcomes, with larger effects associated with longer treatment duration and stronger therapeutic alliance — consistent with the hypothesis that the mechanism is genuinely relational rather than primarily cognitive.

What Earnest Effort Can and Cannot Do

Self-awareness of your attachment pattern, while genuinely useful, is not the same as change. It is possible to accurately describe your anxious or avoidant tendencies with great precision while still being fully governed by them under relationship stress. The description and the behavior operate in different psychological systems. What actually produces change, based on the evidence, is sustained experience in relationships where the expected responses don't occur — where the template's predictions are repeatedly violated and the disconfirming evidence has time to accumulate. This happens primarily through real relationships, not through insight alone. A 2018 study from Stony Brook University following adults in new romantic relationships over 18 months found that relationship quality predicted movement in attachment security more reliably than personal history, pre-existing therapy, or self-reported intention to change. The relationship did the work. Insight about the relationship was secondary. Your attachment style is not your destiny. But the path toward change runs through actual experience, not just understanding.