Dating terms glossary: what dictionaries record

Attachment styles in dating: what the test measures

Research measures attachment as two scores, anxiety and avoidance, not four types. Where the labels come from, what the test asks and what it predicts.

The Mocha team7 min read

Attachment styles describe how people tend to feel and behave when they get close to someone: how much they worry about being left, and how comfortable they are depending on another person. In research they are measured as two scores rather than four types. One is attachment anxiety, worry about whether a partner is available and responsive. The other is attachment avoidance, discomfort with closeness and with relying on others. The familiar labels, secure, anxious, avoidant and fearful, are names for regions on that two dimensional map. The questionnaires that place someone on it, the Experiences in Close Relationships (ECR) and its revision, the ECR-R, give a score on each dimension.

That last point is the one most quizzes leave out. If a test gives you one label and no scores, it is not measuring what the research measures.

Where the labels come from

The theory began with children. The British psychoanalyst John Bowlby argued that infants are built to keep close to a caregiver, and Mary Ainsworth's laboratory studies of one year olds sorted them into three patterns: secure, anxious resistant and avoidant. Chris Fraley, a psychologist at the University of Illinois, gives the split she found in his overview of adult attachment research as roughly 60, 20 and 20 percent.

In 1987 Cindy Hazan and Phillip Shaver asked whether adults in love show the same patterns. Their questionnaire was three short paragraphs, and people simply picked the one that fit. One read, in part, "I often worry that my partner doesn't really love me or won't want to stay with me." About 60 percent chose the secure paragraph and about 20 percent each chose the avoidant and anxious ones, the same split as in infancy.

The fourth style arrived next. Kim Bartholomew argued in 1990 and 1991 that avoidance comes in two kinds: dismissing, where people keep their distance and appear untroubled by it, and fearful, where the distance comes with visible distress. Her four categories, secure, preoccupied, dismissing and fearful, are the ones most articles still use.

The four attachment styles in relationships

On the map Fraley and Shaver publish, each of Bartholomew's four styles is a corner of the two scores, and the descriptions below paraphrase her Relationships Questionnaire, which asks people to pick the paragraph closest to them.

StyleAnxietyAvoidanceWhat it sounds like in a relationship
SecurelowlowGetting close is easy; comfortable depending on a partner and being depended on
PreoccupiedhighlowWants complete emotional intimacy, feels the partner holds back, worries about being valued less
DismissinglowhighComfortable without close emotional relationships; independence matters most
FearfulhighhighWants closeness but finds it hard to trust or depend on anyone, and fears being hurt

Two findings from Fraley's overview fill in the corners. Secure adults tend to be more satisfied in their relationships, which last longer and show more trust and commitment, and they are more likely both to seek support from a partner when distressed and to give it. And dismissing people are not as unmoved as they look: in an experiment by Fraley and Shaver, they were as physiologically distressed as anyone when discussing losing a partner, but could suppress it when told to, which fearful people managed less well. Read the table as four directions on a map, not four kinds of people.

Why researchers stopped counting types

In 1998 Kelly Brennan, Catherine Clark and Phillip Shaver put every self report attachment measure they could find into one large analysis, about 60 subscales in all. Two broad factors came out: anxiety and avoidance. In the same year Fraley and Niels Waller tested whether people really fall into separate types and found no evidence for a true attachment typology. In the words of the measurement page Shaver and Fraley maintain, the types are regions in a two dimensional space, and you lose precision whenever you use a typological measure instead of the continuous scales.

In practice that means two people both called anxious can be quite different, and someone just over the line from secure has more in common with secure people than with the far end of their own label.

What the ECR-R actually asks

Fraley's overview names the ECR and the ECR-R as the most popular measures of adult attachment. The ECR-R, published by Fraley, Waller and Brennan in 2000, has 36 statements, 18 for each dimension, each rated from 1 to 7. On the anxiety scale: "I'm afraid that I will lose my partner's love." On the same scale, scored in reverse: "I rarely worry about my partner leaving me." Your anxiety score is the average of the 18 anxiety answers after reversing those items, and the same for avoidance. Fraley reports reliability of .90 or higher for both scales.

He also adds a warning about his own online version: it is a modified form, with about 16 of the original items and the rest drawn from a larger bank being tested. A quiz result is a reasonable starting point for thinking about yourself. It is not a clinical assessment, and nobody designed it to diagnose a date.

Can it change?

Less fixed than the labels suggest. Fraley summarizes the stability evidence modestly. How secure people feel with a parent and with a romantic partner overlaps at correlations of roughly .20 to .50. Attachment to parents over time is stable at about .25 to .39. The one longitudinal study he knows of that followed people from age one into adult relationships, which is unpublished, found a correlation of .17. His conclusion is that long term stability should be treated as an empirical question rather than an assumption of the theory.

What it predicts in a relationship

The largest synthesis we found is a 2024 meta-analysis by Laura Knox, Gery Karantzas and Elizabeth Ferguson at Deakin University in Australia, covering 139 studies and 38,472 people. It asked whether attachment insecurity goes with mistreating a partner, from invalidation and neglect up to abuse.

LinkAverage correlationWith stress present
Anxiety and partner mistreatment overall.19.24
Avoidance and partner mistreatment overall.14.21
Avoidance and intimate partner abuse.11.35

Most of these are small: the authors call the overall anxiety figure small to moderate, and only the .35 under stress is clearly larger. They write that not all individuals with an insecure attachment orientation will maltreat their partner, and that even those who do, do not behave that way all the time. Stress was the condition that mattered: the links were stronger in studies where people were under pressure. An anxious score tells you something about how someone handles worry, not that they will be a bad partner.

The reverse also holds: a label does not excuse control or abuse. If someone you are seeing frightens or controls you, the National Domestic Violence Hotline in the US (1-800-799-7233, or text START to 88788) and Refuge's National Domestic Abuse Helpline in England (0808 2000 247) are free and open 24 hours a day.

Using it while dating

Two findings from Fraley's overview are more useful early on than any label. Adults looking for long term relationships rate responsive qualities, attentiveness, warmth and sensitivity, as the most attractive in a partner. And some evidence suggests that people end up with partners who confirm what they already believe about relationships. Both point the same way: watch how someone responds to you, rather than deciding after three messages that they are avoidant.

The dictionaries have not caught up with the quizzes. Merriam-Webster, the American dictionary, records attachment in psychology as a strong emotional bond that an infant forms with a caregiver, and the British Cambridge Dictionary as a feeling of love and need for another person, for example for a mother by her child. Neither has an entry for attachment style. Both definitions start where the theory started, with a child.

Frequently asked questions

Where does the four style model come from?
From Kim Bartholomew, who in 1990 and 1991 added a second kind of avoidance, dismissing, to the three styles Cindy Hazan and Phillip Shaver had borrowed from infant research in 1987. Her four categories are secure, preoccupied, dismissing and fearful. The researchers who maintain the main questionnaire now recommend treating the four as regions in a space defined by two scores, anxiety and avoidance, rather than as four types.
Can your attachment style change?
It can, and the evidence for long term stability is modest. Chris Fraley reports that attachment to parents is stable at a correlation of roughly .25 to .39, that security with a parent and with a partner overlaps at roughly .20 to .50, and that the only longitudinal study he knows of, unpublished, found a correlation of .17 between security at age one and in adult relationships 20 years later.
Does an anxious style predict a bad relationship?
Weakly. A 2024 meta-analysis of 139 studies found that attachment anxiety correlated at about .19 with mistreating a partner, and avoidance at about .14. Those are small effects: not everyone with a high anxiety score mistreats a partner, as the authors themselves point out. The link was stronger under stress.
Is an online attachment style test accurate?
Only as accurate as the questionnaire behind it. The research instrument is the ECR-R, 36 statements scored on two scales, with reliability of .90 or higher. Chris Fraley notes that even his own online version has, since 2005, been a modified one, with about 16 of the original items. A quiz that names its questionnaire and gives you two scores is closer to the research than one that hands you a single label.
Do dismissive avoidants come back after ghosting?
No study we found has tested it, so a label cannot predict it. The research says something narrower: in Fraley and Shaver's experiment, dismissing people were as physiologically distressed as anyone else when talking about losing a partner, but better at suppressing it when asked to. Calm is not the same as indifferent. Whether one particular person comes back depends on that person; our ghosting page covers what studies of ghosting measured.

Sources

  1. R. Chris Fraley, A Brief Overview of Adult Attachment Theory and Research, University of Illinois Read October 7, 2026, rechecked October 9, 2026. Supports Bowlby and Ainsworth's origins, the Hazan and Shaver 1987 paragraphs and the roughly 60, 20 and 20 percent split, Brennan's two dimensions, the shift to dimensional measurement, the ECR and ECR-R as the most popular measures, the parent to partner overlap of about .20 to .50, stability of about .25 to .39, the unpublished infancy to adulthood correlation of .17, Zeifman and Hazan on responsiveness as attractive, and stability as an empirical question. Reread October 16, 2026: also supports secure adults being more satisfied in their relationships, with greater longevity, trust, commitment and interdependence, and more likely to seek and give support; and Fraley and Shaver's 1997 finding that dismissing individuals (high avoidance, low anxiety) were as physiologically distressed as others when discussing losing a partner but could suppress it when instructed, while fearfully avoidant individuals were less successful.
  2. Phillip R. Shaver and R. Chris Fraley, Self-Report Measures of Adult Attachment Read October 7, 2026. Supports Bartholomew's four category scheme and the Relationships Questionnaire, Brennan, Clark and Shaver's 1998 analysis of 60 subscales, the recommendation to think in dimensions because Fraley and Waller found no evidence for a true typology, and the description of styles as regions in a two dimensional space. Reread October 16, 2026, with Figure 1 (secure at low anxiety and low avoidance, preoccupied at high anxiety and low avoidance, dismissing-avoidant at low anxiety and high avoidance, fearful-avoidant at high on both) and the Relationships Questionnaire page (labs.psychology.illinois.edu/~rcfraley/measures/rq.html), whose four descriptions are paraphrased in the table; the measures page maps styles A to D to secure, fearful-avoidant, preoccupied and dismissing-avoidant.
  3. R. Chris Fraley, Information on the Experiences in Close Relationships-Revised (ECR-R) questionnaire Read October 7, 2026, rechecked October 9, 2026. Supports the 36 items, 18 per scale, the quoted items, reverse keying and averaging, reliability of .90 or higher, and the note that the online questionnaire is a modified version with about 16 original items.
  4. Knox, Karantzas and Ferguson (2024), The Role of Attachment, Insecurity, and Stress in Partner Maltreatment: A Meta-Analysis, Trauma, Violence and Abuse Read October 7, 2026, rechecked October 9, 2026. Supports the 139 studies and 38,472 participants, the overall correlations of .19 for anxiety and .14 for avoidance, the higher figures where stress was present, the .11 and .35 figures for avoidance and intimate partner abuse, and the authors' statement that not all insecurely attached individuals maltreat their partners.
  5. Merriam-Webster, entry attachment Read October 9, 2026 in the Internet Archive copy of the page dated September 11, 2026. Supports the psychology sense, a strong emotional bond that an infant forms with a caregiver, especially when viewed as a basis for normal emotional and social development, and the list of phrases containing the word (attachment parenting, on attachment), which does not include attachment style.
  6. Cambridge Dictionary, entry attachment Read October 7, 2026. Supports the sense a feeling of love and need for another person, for example for a mother by her child. Cambridge has no entry for attachment style.