Are Attachment Styles Permanent or Can They Change? What 40 Years of Research Shows

Attachment styles are not permanent. Longitudinal research shows roughly 60 to 70 percent of adults keep the same attachment style over years, while about 30 to 40 percent shift at least once, usually toward security. Change is gradual, driven by repeated relationship experiences that retrain expectations, not by willpower or a single insight. Your earliest patterns act like a default setting, not a locked one: they can be rewritten across months to years through consistent, corrective interactions and, in some cases, structured therapy.

The mechanism: why attachment is stable but not fixed

Attachment style is a learned prediction system, not a personality trait. In the first 18 months of life, an infant's brain builds internal working models: expectations about whether others will respond when needed. These models live in procedural memory, the same system that stores how to ride a bike. That is why they run automatically and resist change in a way verbal beliefs do not.

The key mechanism is what researchers call a feedback loop between expectation and behavior. A person with anxious attachment expects abandonment, so they monitor for signs of distance, and that hypervigilance can produce the very withdrawal they fear. An avoidant person expects their needs to be dismissed, so they suppress them, which keeps partners from ever learning what support would look like. Each cycle confirms the original model.

Change happens when the loop is interrupted. A partner, therapist, or close friend responds in a way that disconfirms the prediction often enough that the brain updates its forecast. This is why change is slow: the model was built from thousands of repetitions, and it takes a meaningful number of new ones to overwrite it. In plain terms, attachment change is less like flipping a switch and more like learning a new habit under stress.

Neurobiologically, the relevant systems are the ones governing threat detection, including the amygdala, and the ones governing social reward and caregiving, including the ventral striatum and oxytocin-linked circuits. Secure relationships appear to dampen threat reactivity over time. This is established at the level of general mechanism; the precise timing and size of neural change in adults is still an active research question, not settled fact.

How stable, exactly? The numbers behind the claim

Stability estimates come from longitudinal cohorts that measured attachment in infancy or early childhood and again decades later. The most cited findings, from studies following children into adulthood, converge on a moderate figure: generally 60 to 75 percent continuity, depending on the sample and the measurement method.

A frequently referenced meta-analytic synthesis of stability studies found that attachment security shows moderate stability across the lifespan, with correlation coefficients typically in the 0.3 to 0.5 range over long intervals. That is substantial but far from destiny. In practical terms, a correlation of 0.4 means early experience explains roughly 16 percent of the variance in adult attachment, leaving most of the outcome governed by later experience, relationships, and context.

Adult attachment measured at two points a few years apart is more stable than infancy-to-adulthood comparisons, often 70 percent or higher. The longer the interval and the more major life events in between, the more room for revision.

In plain terms: your style is likely to stay put from year to year, but over a decade or across a big life event, roughly one in three people moves.

The four styles and how much each one tends to move

Attachment is usually scored on two dimensions, anxiety and avoidance, and the four styles are regions on that grid. Change rates differ by style, and the direction of movement is not random.

Secure attachment is the most stable category, with continuity often above 70 percent. Securely attached people already have a working model that handles stress well, so there is less pressure to revise it.

Anxious attachment shows moderate change, and the most common shift is toward security, especially when a person enters a stable, responsive relationship. The anxious style is also the one most likely to respond to therapy, because the underlying fear can be directly targeted.

Avoidant attachment changes somewhat less often, and when it does, it often moves toward security slowly, sometimes only after a relationship that persists despite the person's distancing. Dismissive-avoidant patterns are among the hardest to shift because the person may not experience the pattern as a problem.

Fearful-avoidant attachment, a mix of high anxiety and high avoidance, shows the most fluctuation over time. This instability reflects internal conflict rather than a stable strategy, and it also carries the highest rates of shift in either direction, including temporary worsening under stress.

Earned security: how change actually happens

The term earned security describes a person who did not start secure but developed security through later relationships. Research on this group is one of the most encouraging findings in the field: adults who report earned security look, on many measures, much like people who were secure from childhood, including in parenting sensitivity and relationship satisfaction.

There are three documented routes to change. The first is a corrective relationship, typically a romantic partner or close friendship that stays responsive over time. The second is psychotherapy, particularly approaches that target the working model directly. The third is reflective capacity, the ability to notice your own patterns and question them, which can be built through journaling, structured reflection, or therapy.

Therapy effects are modest but real. Meta-analytic reviews of attachment-focused interventions report small-to-moderate improvements in attachment security, with effect sizes around 0.2 to 0.4. That means change is measurable but not dramatic, and it accumulates over the course of treatment rather than appearing suddenly. This is established enough to support therapy as a route, but the exact mechanism of change is still debated.

A non-obvious caveat: change often shows up first in one relationship and generalizes slowly. Someone may become secure with a specific partner while still reacting anxiously with a parent or in the workplace. Attachment is partly relationship-specific, not a single global switch.

A realistic timeline for change

Change does not follow a fixed schedule, but research and clinical observation suggest a rough time course. The table below reflects typical patterns, not guarantees.

In plain terms: expect to feel the first real shift in about three to six months, and expect the change to hold across situations only after one to two years of consistent experience. Anyone promising overnight transformation is overselling.

Common misconceptions, corrected

The biggest misconception is that attachment style is a fixed label. It is better understood as a probability distribution: you have a dominant style and a range of responses, and context shifts which one shows up.

A second misconception is that you must "fix" an insecure style to be healthy. Research does not support that framing. Insecure styles are adaptations that made sense in a person's history, and many people function well with them. The goal is flexibility, not erasure.

A third is that change requires a perfect partner. The evidence points instead to consistency, not perfection. A partner who repairs after conflict is more influential than one who never errs.

A fourth, corrected by recent work: anxious and avoidant people are not opposites who cannot understand each other. Their underlying fear is often the same, rejection or abandonment, expressed through opposite strategies. This shared root is why couples therapy can work for both at once.

What to actually do if you want to change

Change is more likely when you take specific, repeated actions rather than waiting for insight.

• Notice the moment of activation. Learn to spot the physical and mental signature of your pattern, such as the urge to withdraw or the spike of reassurance-seeking, before you act on it.
• Practice one disconfirming behavior per week. If you are avoidant, share one need; if anxious, tolerate one unanswered message without escalating.
• Build repair into conflict. A single repair after a rupture does more for security than avoiding conflict entirely.
• Choose relationships that are consistent, not intense. Stability is the active ingredient, and intensity without reliability does not retrain the model.
• Consider structured therapy if the pattern causes distress or blocks relationships. Attachment-focused and emotion-focused approaches have the strongest evidence base.
• Track change over months, not days. Expect a visible shift within 3 to 6 months and consolidation over 1 to 2 years.

Limitations and what science does not yet know

Several limits are worth stating plainly. Most longitudinal samples are Western and drawn from middle-class communities, so stability and change rates may not generalize across cultures with different caregiving norms. Measurement matters too: self-report questionnaires capture conscious models, while interview-based methods capture more implicit ones, and the two can disagree.

Change rates also depend heavily on life events, which are hard to control in research. Someone who moves countries, changes jobs, or loses a partner may shift in ways no study can predict. Finally, the field distinguishes clearly between what is established, the direction and rough size of change, and what is hypothesis, the precise neural mechanisms and optimal timing of interventions.

The honest summary: attachment styles can and do change, for roughly one in three people over a decade, and for more when therapy or a stable relationship is involved. Early patterns set a default. Later experience rewrites it.