Attachment Style
Statistics
90+ findings from peer-reviewed meta-analyses and longitudinal studies — organized by topic across prevalence, mental health, relationships, parenting, and more.
This page compiles the most important attachment style statistics in one place, drawn exclusively from peer-reviewed meta-analyses, longitudinal cohort studies, and major primary research. Every claim links to its source. Where findings are contested or have been revised by newer data, that's flagged clearly.
The statistics are organized by topic — prevalence, mental health, relationships, intergenerational transmission, physical health, therapy outcomes, and early development — so whatever aspect of attachment you're most curious about, it's easy to find.
01
How common is each attachment style?
Prevalence & Distribution
How attachment styles are distributed across adult populations and in infancy — including the most current global meta-analytic estimates.
Adult Attachment Style Distribution — US Nationally Representative Sample (N = 8,098)
Source: Mickelson, Kessler & Shaver (1997), Journal of Personality and Social Psychology, 73(5), 1092–1106
Infant Attachment Distribution — Most Current Global Meta-Analysis (N = 20,720 dyads, 285 studies)
Note: Earlier estimates (van IJzendoorn & Kroonenberg, 1988) reported ~65% secure because the disorganized category wasn't yet included. The Madigan et al. 2023 figure is the current best estimate. Source: Madigan et al. (2023), Psychological Bulletin, 149(1–2), 99–132
Relationship Duration · Secure vs. Anxious
Secure adults had romantic relationships lasting nearly twice as long as anxious-ambivalent adults on average
10.02 vs. 4.86 yrs
Secure = 10.02 years; avoidant = 5.97 years; anxious/ambivalent = 4.86 years. F(2,568) = 15.89, p < .001. N = 620 in the original Hazan & Shaver newspaper study.
Cross-Cultural Prevalence · 62 Regions
Secure attachment was the modal style in 79% of 62 cultural regions studied worldwide
79% of regions
N = 17,804 adults. Preoccupied attachment was disproportionately prevalent in East Asian cultures. Secure attachment appears to be a near-universal default across human populations.
Cultural Context · Kibbutz vs. Home Care
Disorganized attachment reached ~48% in Israeli kibbutz communal sleeping — vs. 80% secure in home-based arrangements
80% → 48% secure
N = 48. Demonstrates how collective caregiving norms can substantially shift attachment distributions within a single culture.
Disorganized Attachment · Maltreatment Risk
Disorganized attachment climbs from ~15% in low-risk samples to ~80% in maltreated infants
15% → 80%
~25% in low-SES samples without maltreatment; up to ~48% with parental substance abuse. Maltreatment is the single most powerful predictor of disorganization.
Within-Culture vs. Cross-Cultural Variation
Variation in attachment classifications within a single country exceeds variation between countries by 1.5×
1.5× intracultural
Meaning cross-national differences in attachment distributions are smaller than the variation you'd find comparing different samples within the same country — a critical nuance for cross-cultural claims.
Adult Attachment Interview · Non-Clinical Adults
AAI distributions in non-clinical adults: 58% secure-autonomous, 23% dismissing, 19% preoccupied — based on the first 10,000 interviews
58 / 23 / 19%
An additional ~18% received a secondary unresolved classification for loss or trauma. Updated 2024 synthesis of 26,000+ interviews: 59% F, 25% Ds, 16% E.
Methodological Note
Self-report instruments (ECR, RQ) and the Adult Attachment Interview (AAI) measure different constructs and yield different distributions. A 2007 meta-analysis found the overlap between the two was "trivial to small" (r = 0.09, N = 1,221). Be cautious when comparing statistics derived from different measurement approaches. Roisman et al. (2007), Journal of Personality, 75(2).
02
Depression · Anxiety · PTSD · BPD · Eating Disorders · Suicidality
Mental Health Correlations
How insecure attachment relates to specific mental health conditions. Effect sizes are reported as published — typically small-to-moderate, not dramatic.
r = .42
Attachment anxiety correlates with negative mental health outcomes (depression, anxiety, loneliness) across 245 samples and 79,722 participants — the largest meta-analysis on this topic to date. Avoidance showed r = .28 for the same composite.
Depression
Depression · Preoccupied vs. Dismissing
Preoccupied attachment predicts depression significantly more than dismissing does — and dismissing shows essentially no association
d = 0.48 vs. d = 0.09
Preoccupied vs. secure: d = 0.48. Dismissing vs. secure: d = 0.09 (non-significant). Not all insecure attachment is equally linked to depression.
Depression · Children & Adolescents
Insecure attachment is meaningfully linked to depression in children and adolescents across 123 independent samples
r = .31
643 effect sizes from 123 samples. A significant moderate effect across development — attachment and depression are related throughout childhood, not just in adulthood.
Anxiety
Anxiety · Preoccupied vs. Dismissing (AAI)
Only preoccupied — not dismissing — attachment is associated with anxiety symptoms when measured via the Adult Attachment Interview
d = 0.35 (preoccupied only)
Dismissing vs. secure: d = −0.02 (NS). This challenges the assumption that all forms of insecure attachment increase anxiety.
⚠ Contested
Social Anxiety · Insecure Attachment
93% of reviewed studies found insecure attachment associated with social anxiety — with attachment anxiety as the most consistent predictor
28 of 30 studies
A systematic review of 30 studies. The near-universal direction of findings makes social anxiety one of the most reliably supported attachment-mental health links.
PTSD
| Attachment Style | Effect (ρ) | Notes |
|---|---|---|
| Fearful | ρ = .44 | Strongest PTSD association — large effect |
| Insecure (overall) | ρ = .26 | Across 46 studies, N = 9,268 |
| Secure | ρ = −.27 | Protective against PTSD severity |
| Dismissing | NS | Not significantly associated except in military samples |
Source: Woodhouse, Ayers & Field (2015), Journal of Anxiety Disorders, 35, 103–117
Borderline Personality Disorder
BPD · Universal Insecure Attachment Link
Every single reviewed empirical study found a strong link between BPD and insecure attachment — 13 out of 13
13 / 13 studies
Unresolved, preoccupied, and fearful classifications were most characteristic. No reviewed study found BPD associated with secure attachment.
BPD · Anxiety vs. Avoidance Dimension
BPD traits correlate strongly with attachment anxiety — more so than with avoidance
r = .48 (anxiety) / .30 (avoidance)
27 effect sizes. The anxiety dimension — fear of abandonment, preoccupied monitoring — shows a particularly tight relationship with BPD phenomenology.
Eating Disorders
Eating Disorders · Anxiety & Anorexia
Attachment anxiety is dramatically elevated in anorexia nervosa compared to healthy controls — a large effect by any standard
d = 1.07
95% CI: 0.87–1.27. Bulimia nervosa showed similar anxiety effects (d = 1.02) plus elevated avoidance (d = 0.86). These are among the largest effect sizes in the attachment-psychopathology literature.
Disordered Eating · General Population
Attachment anxiety predicts unhealthy eating patterns even in general, non-clinical populations
r = .27 (70 studies, N = 19,470)
Avoidance r = .12. Suggests attachment-eating links extend well beyond diagnosed eating disorders into everyday disordered eating.
Substance Use
Earlier attachment security prospectively predicts lower substance use — and the causal direction runs from attachment to substance use, not the other way around
34 longitudinal samples, N = 56,721. 95% CI: −.14 to −.08. The prospective direction was significantly stronger from attachment to substance use than the reverse — important for understanding etiology.
r = −.11
Suicidality
Suicidality · Insecure Attachment Link
Insecure attachment is associated with suicidal ideation and attempts across a large, diverse evidence base
r = .16–.26 (47 studies, N = 50,214)
A consistent small-to-moderate effect replicated across tens of thousands of participants.
Suicide Attempts · Which Styles Are Linked
Only fearful and preoccupied attachment predicted suicide attempts — dismissing and avoidant styles did not
2 of 4 styles linked
Across 58 studies. Broad claims that all insecure attachment increases suicide risk are not well supported — attachment type matters significantly.
⚠ Contested
Self-Esteem
Self-Esteem · Parental Attachment Security
Parental attachment security predicts self-esteem — and security appears to build self-esteem more than self-esteem builds security
r = .34 (202 studies, N = 81,485)
Cross-lagged: security → later self-esteem (r = .19) was stronger than self-esteem → later security (r = .08). Attachment comes first.
Self-Esteem · Anxiety vs. Avoidance
Attachment anxiety is a stronger predictor of low self-esteem than avoidance is
r = −.38 (anxiety) vs. −.23 (avoidance)
38 studies, N = 9,494. The anxiously attached person's inner experience of unworthiness is not simply insecurity — it shows up clearly in self-esteem measures.
03
Satisfaction · Conflict · Infidelity · Breakup · Sex
Romantic & Relationship Outcomes
How attachment style shapes relationship quality, conflict physiology, infidelity, and how people cope when relationships end.
46.2%
Of the variance in relationship satisfaction explained by attachment anxiety and avoidance (actor + partner effects) in a community sample of 1,014 couples. Attachment also explained 17.9% of variance in relationship instability.
Relationship Satisfaction · Long-Term Effect
The negative effect of insecure attachment on relationship satisfaction grows stronger the longer a relationship lasts — not weaker
Effect increases with time
Meta-analysis: k = 57 studies, N = 14,340. Attachment style matters more in long-term relationships, not less — an important finding for couples work.
Relationship Quality · Avoidance vs. Anxiety
Avoidant attachment does more damage to relationship satisfaction, connectedness, and support — while anxious attachment drives conflict more
Two different mechanisms
The two insecure styles don't damage relationships the same way. Avoidance erodes closeness; anxiety escalates conflict. Important for distinguishing interventions.
Infidelity · Anxious Attachment in Marriage
Newlyweds high in attachment anxiety were more than twice as likely to commit infidelity — while avoidance showed no association in married samples
2× infidelity risk
N = 207 newlywed couples. Contrasts with dating-sample findings where avoidance predicts infidelity. The context of marriage appears to shift which attachment dimension is implicated.
⚠ Contested: dating studies find avoidance predicts infidelity instead
Conflict Physiology · Cortisol & Attachment
Insecurely attached individuals show greater cortisol reactivity during relationship conflict — and their partner's attachment style independently predicts their own stress response
↑ Cortisol during conflict
Anxious men and avoidant women showed the steepest cortisol curves. Attachment shapes how conflict gets into the body — not just how it's experienced psychologically.
Infidelity · Fearful Attachment Link
Combined attachment insecurity is associated with marital infidelity — with fearful attachment showing the strongest link
r = .18 (k = 17, N = 13,666)
Fearful r = .19; dismissive r = .07; preoccupied showed no significant association. The relationship between attachment and infidelity is more nuanced than popular accounts suggest.
Breakup Rate · Avoidant Attachment
Avoidant adults were the most likely to experience actual relationship termination over a 10-week follow-up period
Highest breakup rate
N = 193 undergraduates. Attachment style predicted real relationship outcomes over real time — not just self-reported satisfaction ratings.
Post-Breakup Distress
Attachment anxiety strongly predicted preoccupation with the lost partner, exaggerated reunion attempts, anger/vengeance, and dysfunctional coping post-breakup. Avoidance predicted self-reliant withdrawal. Among divorcees, elevated distress was concentrated only in avoidant and anxious-ambivalent individuals — not secure individuals — despite identical objective stressors. Davis, Shaver & Vernon (2003); Birnbaum et al. (1997).
04
How attachment passes from parent to child
Intergenerational Transmission
How reliably parents transmit their attachment style — and the significant gap between what the data explains and what remains unknown.
| Meta-Analysis | Effect Size | Concordance | Notes |
|---|---|---|---|
| van IJzendoorn (1995) | d = 1.06 / r = .47 | 75% (3-way) | Foundational estimate; k = 18 studies |
| Fonagy, Steele & Steele (1991) | N/A | 75% | Prospective: prenatal AAI → 12-month SSP, N = 96 |
| Verhage et al. (2016) | r = .31 | 51% (4-way) | Updated; k = 95, N = 4,819. Effect has declined |
| Benoit & Parker (1994) | N/A | 65% (3-gen) | Grandmother → mother → infant concordance |
Transmission Gap · Sensitivity & AAI
Maternal sensitivity accounts for only about a quarter of the parent-to-child attachment transmission — leaving a ~75% "transmission gap" still unexplained
~25% explained
Even after correcting for attenuation, the majority of how parents pass attachment patterns to children remains unaccounted for by sensitivity alone. A central unresolved question in the field.
Maternal Sensitivity · Secure Attachment
Maternal sensitivity moves the probability of infant secure attachment from ~38% to ~62% — meaningful, but far from deterministic
r = .24 (k = 21, N = 1,099)
Sensitivity matters — but it's not the whole story. The moderate effect size helps explain why sensitivity-focused interventions produce modest changes in attachment security.
Disorganized Transmission · Unresolved AAI
Of disorganized infants, 53% have parents classified as unresolved on the AAI — meaning 47% do not, a second unexplained gap
53% concordance
The disorganized-to-unresolved transmission link exists but leaves substantial unexplained variance — a separate "transmission gap" for disorganization specifically.
Heritability · Genes vs. Environment
Infant attachment security shows near-zero heritability — it is shaped by environment, not genes, at least in infancy
0% heritability
Twin study (N = 157 pairs): shared environment = 52% of variance; nonshared + error = 48%. This contrasts sharply with most psychological traits, which show substantial heritability.
⚠ Contested: Fearon et al. (2014) found ~40% heritability in adolescence
05
Cardiovascular · Immune · Chronic Pain · Health Behaviors
Physical Health
Attachment style gets under the skin — with measurable effects on inflammation, immune function, cardiovascular risk, and health behaviors.
7.5×
Anxious-resistant infants were approximately 7.5 times more likely to report inflammation-based illness at age 32 — in a 30-year prospective longitudinal study (N = 163). Anxious-avoidant infants were approximately 3 times more likely.
Chronic Illness · Diabetes & Colitis
70–72% of patients with diabetes or ulcerative colitis were classified as insecurely attached — vs. ~40–45% in healthy comparisons
70–72% insecure
A striking overrepresentation. Insecure attachment was strongly elevated in both of these chronic conditions compared to general population samples.
Inflammation · Avoidance & IL-6
During marital conflict, high attachment avoidance produced an 11% increase in the inflammatory marker IL-6 — vs. a 6% decrease for those low in avoidance
+11% IL-6 (avoidance)
Avoidant individuals appear emotionally disengaged but show a paradoxically greater inflammatory response to relational stress. The body registers what the mind dismisses.
Immune Function · Anxiety & T Cells
Higher attachment anxiety predicted significantly fewer T cells in married couples — mediated by elevated cortisol dysregulation
↓ CD3+, CD4+, CD8+ T cells
N = 85 couples. Jaremka et al. (2013) show attachment anxiety dysregulates the HPA axis in ways that tangibly compromise immune function — a direct biological pathway.
Chronic Pain · Insecure Attachment
Insecure attachment is associated with approximately twice the experience of chronic widespread pain compared to secure attachment
~2× chronic pain
The pain-attachment link persists even after controlling for psychological distress — suggesting attachment may have independent effects on pain processing and tolerance.
Research Gap
Direct attachment-to-mortality evidence remains scarce. Pietromonaco & Beck (2019) characterize this as a gap in the literature — most evidence linking attachment to mortality is indirect, flowing through broader social connection research. More longitudinal work is needed before causal claims can be made. Current Opinion in Psychology, 25, 115–120.
06
Alliance · Outcome · Attachment-Based Interventions
Therapy & Treatment Outcomes
How attachment style shapes engagement in therapy, response to treatment, and the effectiveness of attachment-targeted interventions.
Therapeutic Alliance · Attachment Style
Secure attachment predicts a stronger therapeutic alliance — and both anxious and avoidant styles undermine it
r = .17 (alliance); avoidance r = −.14
k = 17, N = 886. Alliance is a key mediator of therapy outcomes — and attachment shapes it from the very first session.
Therapy Outcomes · Anxiety vs. Security
Pretreatment attachment anxiety predicts significantly worse therapy outcomes — while pretreatment security predicts better ones
d = −.46 (anxiety) / d = +.37 (security)
k = 14, N = 1,467. Improvements in security during therapy coincided with better outcomes in the updated 36-study synthesis. Attachment changes and symptom changes move together.
Modality Matching · CBT vs. IPT
High avoidance predicted better outcomes in CBT for depression; low avoidance predicted better outcomes in IPT — a genuine aptitude-treatment interaction
CBT vs. IPT depends on avoidance
Attachment style should inform modality selection, not just treatment content. This is one of the stronger arguments for using attachment assessment at the start of treatment.
Parenting Interventions · Security & Sensitivity
Attachment-based parenting interventions improve infant attachment security — but shorter, focused programs outperform long, multi-component ones
d = 0.20 (security); d = 0.33 (sensitivity)
70 studies, N = 1,503. More intensive intervention is not always better — a key finding for program design.
EFT Outcomes · Couples Therapy
About 70% of couples completing Emotionally Focused Therapy were symptom-free by the end of treatment — with a large effect size
~70% symptom-free; d = 0.93
20-study meta-analysis. EFT vs. alternatives: d = 0.44. Note: earlier estimates of d = 1.3 have been revised downward in this more comprehensive synthesis.
MBT vs. TAU · BPD Long-Term
Mentalization-Based Treatment for BPD produced dramatically better 8-year outcomes than treatment as usual — including an 87% vs. 13% rate of continued diagnosis
13% vs. 87% continuing BPD
N = 41. Suicidality: 23% MBT vs. 74% TAU; global functioning >60: 45% vs. 10%. The most striking long-term attachment-informed outcome data available in the literature.
Contested Finding
Despite wide adoption, the largest Circle of Security RCT (N = 141 Head Start mothers) found no significant main effects on child attachment or behavior. Cassidy et al. (2017), Development and Psychopathology, 29(2), 651–673. Popularity of an intervention and measured effectiveness are not the same thing.
07
Age · Gender · Culture · Socioeconomic Status
Demographics
How attachment style distributions vary by age, gender, culture, and socioeconomic background.
| Age Group | Dismissing/Avoidant Rate | Notes |
|---|---|---|
| Young Adults (~18–30) | ~18% | Based on AAI distributions; Diehl et al. (1998) |
| Middle-Aged (~40–60) | ~22% | Dismissing increases through midlife |
| Older Adults (~70+) | ~40–78% | Strikingly high; likely reflects cohort effects |
Gender Differences · Avoidance & Anxiety
Men show higher avoidance and lower anxiety than women in self-report studies — but this difference disappears when measured with the Adult Attachment Interview
D = .28 (community samples)
113 samples, N = 66,132. Effect sizes vary substantially by sample type. The measurement tool you use changes what gender differences you find.
⚠ AAI studies find no reliable sex differences
Race & Attachment · SES Explains All
Apparent racial differences in infant attachment security disappear entirely when socioeconomic status is accounted for
SES, not ethnicity
NICHD sample: African-American mean AQS = .20 vs. White = .30. After SES adjustment: no significant difference. Poverty and adversity drive attachment insecurity, not race.
Adversity vs. Demographics · NCS Data
Childhood adversity — not demographics — is the strongest predictor of adult insecure attachment in a nationally representative US sample
Adversity > demographics
N = 8,098 (National Comorbidity Survey). Insecure attachment was more common among lower-income and non-White respondents, but adversity explained those associations.
08
Stability · Development · Maltreatment · Institutional Care
Childhood & Early Development
How early attachment shapes developmental trajectories — and how stable those early patterns actually are across time.
r = .39
Mean stability of attachment from infancy to adulthood across 127 papers and 21,072 attachments. Stability is significant for intervals up to 15 years, but not beyond. Life events — especially adversity — can and do shift attachment classification.
Attachment Stability · 20-Year Follow-Up
72% of children maintained their secure vs. insecure classification from 12 months to age 20 — but negative life events doubled the rate of change
72% stable over 20 years
N = 50. 44% changed classification if mothers reported negative life events vs. 22% without. Life events are the key moderator of attachment stability.
Maltreatment Impact · Effect Size
The difference in attachment security between maltreated and low-risk children is so large it represents near-complete non-overlap of distributions
d = 2.10 (security); d = 2.19 (disorganization)
Children with 5+ socioeconomic risks (d = 1.20) approached maltreated levels of disorganization — poverty alone can approach the effects of direct maltreatment.
Disorganization · Maltreatment Rate
Disorganized attachment reaches 86–90% in maltreated infant samples — vs. ~15% in low-risk middle-class samples
86–90% in maltreatment
Maltreatment is the single strongest predictor of disorganized attachment. The contrast with normative samples underscores how extreme the effect of abuse and neglect is on early attachment organization.
Carlson, Cicchetti et al. (1989); Cicchetti, Rogosch & Toth (2006)
Internalizing Behavior · Grows With Age
Attachment effects on behavior problems appear to grow as children age — not shrink — with internalizing effects roughly doubling between infancy and adolescence
Infant d = 0.15 → Adolescent d = 0.58
Infant-era internalizing: d = 0.15 (k = 42, N = 4,614). Older children/adolescents: d = 0.58 (k = 165, N = 48,224). Externalizing follows a similar pattern.
Disorganization · Adolescent Dissociation
Disorganized attachment in infancy predicted dissociation in adolescence — a direct line from early disorganization to later psychological fragmentation
r = .36 (12–18 months → age 17½)
Minnesota Longitudinal Study of Risk and Adaptation, N = 267 high-risk children. Early disorganization also predicted overall psychopathology in adolescence and young adulthood.
Foster Care · Bucharest RCT
Moving maltreated children from institutions to foster care nearly tripled the rate of secure attachment at 42 months — especially for children placed before age 2
49% (foster) vs. 18% (institution)
Community comparison: 65%. Bucharest Early Intervention RCT. Demonstrates that early relational intervention can meaningfully shift attachment even in high-adversity contexts.
Externalizing Behavior · Disorganized Strongest
Insecure attachment predicts externalizing behavior in children — with disorganized attachment showing the strongest effect, especially in boys
d = 0.31 overall; d = 0.34 (disorganized)
k = 69, N = 5,947. Effect larger for boys (d = 0.35) and in higher-risk samples (d = 0.49). Avoidant attachment alone showed d = 0.12.
Social Competence · Early Attachment
Early attachment combined with broader early-care quality predicted approximately 40% of variance in social competence in middle childhood
~40% of variance explained
Minnesota Longitudinal Study. When attachment is examined alongside other early relationship quality factors, its predictive power for later social functioning increases substantially.