The most consequential question about attachment style is also one of the most contested: is it stable, or can it change? If childhood attachment experiences produce a fixed pattern that determines adult relational life, the practical implications are different from those of a model where the pattern responds to subsequent experience and deliberate work.
The research literature gives an answer that is more useful than either extreme view. Attachment style is real, has roots in early experience, and shows meaningful stability across time. It is also genuinely responsive to relational experience and to deliberate developmental work. Change is real, slower than people often hope, and not guaranteed — but it happens.
What stability the research actually shows
The most-cited finding on attachment stability comes from Fraley's (2002) meta-analytic work. The result is more nuanced than popular summaries suggest. Stability of attachment from infancy to adulthood is real but relatively weak. Subsequent meta-analyses (Groh et al. 2014; Pinquart, Feußner, & Ahnert 2013) have confirmed this pattern. Early attachment patterns are not destiny — they correlate with later attachment, but the correlation is far from deterministic.
Within adulthood, attachment patterns show stronger but still partial stability. Test-retest correlations across short intervals (weeks or months) are high. Across longer intervals (years), correlations are moderate. Across decades, correlations are weaker but still present. This is the pattern of a real psychological characteristic that responds to ongoing experience — not a fixed trait that never changes, and not a transient state that fluctuates with mood.
Earned security: the documented phenomenon
The concept that gives the strongest empirical support to attachment change is "earned security." The term was developed in the Berkeley Adult Attachment Interview tradition by researchers including Mary Main, Erik Hesse, and others. It describes adults who report negative or insecure childhood attachment experiences but who, in adulthood, show coherent, autonomous attachment patterns characteristic of security.
The term itself implies what the research has begun to confirm: that some portion of adults whose early attachment was insecure develop secure attachment later in life. They earn it, not in the moralistic sense, but in the sense that the security develops through subsequent experience rather than carrying forward from infancy.
The Roisman et al. (2002) longitudinal study — based on 23 years of prospective data — provided important refinement to the concept. The study found that adults classified as earned-secure based on retrospective reports of negative childhood and current secure functioning did parent as effectively as adults whose security had been continuous from childhood, but they showed elevated depressive symptomatology. The study also found, importantly, that retrospectively defined earned-secures often had received supportive childhood parenting that they later remembered as more difficult. This complicates the simple narrative — earned security is real, but its operationalization matters, and not everyone who reports earning security through post-childhood experience actually has the developmental history the construct implies.
The more recent refinement involves prospective definition: identifying individuals whose attachment was measured as insecure earlier in development and as secure later. Prospectively defined earned security has weaker evidence base but represents the cleaner construct methodologically.
What changes attachment
The mechanisms underlying attachment change are increasingly well- characterized in the research literature. Several factors emerge as particularly important.
Corrective relational experience. A relationship — romantic partnership, deep friendship, therapeutic alliance, mentoring connection — with a person who is themselves securely attached and consistently responsive can, over time, revise the implicit working models that underlie attachment. The mechanism is not psychoeducational; it is experiential. Sustained contact with someone whose behavior contradicts the insecure pattern's predictions about how relationships work gradually updates those predictions.
The effect is not automatic. The insecure person must be able to tolerate the corrective experience long enough for it to register. Many insecure attachment patterns include defensive structures that protect against the kind of contact that would actually revise them — the dismissive person who maintains distance, the anxious person who tests and provokes the secure partner's withdrawal, the fearful person who oscillates in and out of contact. The corrective experience requires both a securely attached other and the insecure person's capacity to remain in the relationship long enough for the pattern to update.
Therapy specifically. Therapeutic relationships with attuned clinicians produce documented changes in attachment patterns, particularly when the work continues for sustained periods. The therapeutic alliance itself functions as a relational experience that can revise insecure patterns; specific approaches like Emotionally Focused Therapy (Sue Johnson and colleagues) explicitly target attachment dynamics in couples.
Negative life events. The research is asymmetric here. Negative events — relational loss, betrayal, sustained relationship conflict — can shift secure people toward greater insecurity. Positive events — relational success, new secure relationships — can shift insecure people toward greater security. Both directions of change are documented (Fraley, Gillath & Deboeck 2021).
Maturation and life stage. Some attachment change appears to occur through ordinary developmental processes — the entry into stable long-term partnership, the experience of parenting, the gradual accumulation of relational experience that contradicts insecure predictions. People do not have the same attachment style at 50 that they had at 20, on average. The trajectory varies significantly by person.
What attachment change is not
The popular discussion of attachment change sometimes overstates what the research supports. Several clarifications are worth making.
It is not fast. The research literature describes change over years and decades, not weeks and months. Working models of attachment are deep cognitive-emotional structures developed through repeated early experience and reinforced through years of subsequent relational patterns. They do not revise quickly. Someone who sees their attachment style on a test result and resolves to "be secure" does not become secure through resolution. The change happens through sustained contradictory experience over substantial periods.
It is not linear. Attachment change tends to occur in fits and starts. Periods of consolidation alternate with periods of regression under stress. A person who has shifted toward greater security over years may, under conditions that recapitulate early relational threats (a major loss, a damaging relationship, a serious illness), find old insecure patterns reasserting. This is not failure — it is how deeply embedded patterns work. Sustained development eventually produces more reliable security; intermittent regression is part of the process.
It is not always toward greater security. Most popular discussion treats attachment change as moving from insecure to secure. The research finds that change runs in both directions. People can lose security under conditions of sustained relational difficulty just as people can earn security under conditions of sustained relational support. Maintaining security through life's difficulties is itself a developmental achievement; it is not a default state that some lucky people happen to occupy.
It is not guaranteed by knowledge. Reading about attachment, taking an attachment assessment, understanding your style — these are useful starting points. They do not produce change. The change occurs through relational experience, sustained over time, often with deliberate support. Knowledge facilitates the change; it does not constitute it.
What this means practically
If your attachment style is insecure and you would like it to be more secure, the research suggests several things.
Sustained contact with securely attached others is the primary lever. This may be a romantic partner, a friend, a therapist, or a mentor — the role matters less than the pattern. The other person does not need to be perfectly secure (no one is) but does need to be reliable enough that, over time, your nervous system gets evidence that contradicts your insecure expectations. This is why therapeutic relationships are particularly effective: they are designed to provide consistent, attuned, responsive contact across years of work.
Patience with yourself is important. The pattern that took fifteen years of childhood to install will not revise in a year of trying. The research timeline is in decades. This is not a counsel of despair; it is a counsel of realistic expectations. The change is real, and people do experience meaningful shifts within a year or two of focused work. The full revision of the working model is a longer arc.
Working with a therapist familiar with attachment specifically is the most evidence-supported approach. Emotionally Focused Therapy (EFT) for couples, certain integrative individual therapies, and trauma-informed approaches can all address attachment patterns explicitly. The therapeutic relationship itself functions as part of the corrective experience.
Take the attachment style assessment to identify your current pattern. How attachment styles affect adult relationships covers how each style manifests in romantic relationships specifically. Your attachment style describes where you currently are. It does not describe where you must remain.