ECR-R questionnaire: Experiences in Close Relationships, Revised

The ECR-R is a self-report questionnaire: you rate how well statements about close relationships describe you. It was introduced by R. Chris Fraley, Niels G. Waller, and Kelly A. Brennan in a 2000 paper (Fraley et al., 2000). Chris Fraley, the first author, maintains a page for the measure with the items, the scoring key, and norms.

Its predecessor, the ECR, appeared in 1998. Brennan, Clark, and Shaver pooled 323 items from the attachment questionnaires of the time and showed that two separate dimensions sit underneath them: anxiety and avoidance (Brennan et al., 1998). Two years later, Fraley and colleagues used item response theory (IRT) to reselect the items, so the scales would distinguish between people more precisely. That’s what the R stands for: revised.

The central idea behind both questionnaires is that adult attachment is a point on two axes, not a label. This was later tested directly: a taxometric analysis of two large samples concluded that adult attachment behaves like continuous dimensions rather than distinct types (Fraley et al., 2015).

The statements split evenly. 18 are about attachment anxiety: the fear that someone close will lose interest, grow colder, or leave. The other 18 are about avoidance: how uneasy it feels when closeness gets intense and you have to rely on someone. Each statement is rated from 1, “strongly disagree,” to 7, “strongly agree.”

Some statements are worded in reverse, so agreeing with them signals a low level rather than a high one. That’s a standard guard against the tendency to agree with everything, and it also helps flag careless answers. The statements are about romantic relationships in general, so people who aren’t in a relationship right now can take it too: they answer about past experience and how they usually feel.

The scoring key is simple enough to check by hand. The result is two numbers, one for each scale.

1. Answers to reverse-keyed items are flipped: 1 becomes 7, 2 becomes 6, and so on; in other words, the answer is subtracted from 8.
2. The anxiety score is the average of all answers on the first scale and falls between 1 and 7.
3. The avoidance score is computed the same way from the second scale.
4. There is no total score: the scales are scored separately and never added up.

For example, if you answered 6 to a reverse-keyed item, it counts as 2. If your anxiety answers average 3.8 after flipping, that is your anxiety score. Since June 5, 2013, Fraley’s page has carried a correction to the key: item 21 is not reversed, while items 22 and 36 are. Copies with the old key still circulate online, which is why different sites sometimes give different scores for the same answers. We use the corrected key.

A score means little on its own: it needs a point of comparison. Fraley published the averages of his online sample from the early 2000s: at least 17,000 people, 73% of them women, with an average age of 27. He advises treating these norms with some caution, since the sample isn’t representative.

A percentile tells you what share of the norm sample had a lower score than yours. For the ECR-R we compute it from a normal distribution with these means and standard deviations. If you give your age, the result adds a second line comparing you with people your age, using a formula Fraley also publishes. The avoidance distribution is probably slightly right-skewed, so percentiles at the very top are a little inflated.

The norms come mostly from English-speaking participants. In a study of 62 cultural regions, the anxious style was more common in East Asia (Schmitt et al., 2004), so comparing with a largely American sample shifts results for some languages. Once we’ve collected enough data of our own, we can compute norms for each language separately.

For research, Fraley advises against sorting people into types and recommends working with the two scores. For explaining a result to one person, though, a label helps, and if you do split people into groups, he suggests cutting each scale at its median. We cut at the norm means: in a normal distribution, the mean and the median coincide.

• Both scales below the means: secure.
• Anxiety at or above the mean, avoidance below: anxious.
• Avoidance at or above the mean, anxiety below: avoidant.
• Both scales at or above the means: fearful-avoidant, often called disorganized.

If the point lies within a quarter of a standard deviation of any boundary, the result lists the neighboring styles as well. With this split, about 32% of people come out secure, 18% anxious, 19% avoidant, and 31% fearful-avoidant. Popular articles say that more than half of people are secure. Those figures come from surveys where people picked one of several ready-made descriptions, not from splitting two scales, which is why the numbers differ. How the four styles and the borderline zone work is covered in a separate overview.

The ECR-R asks about romantic relationships in general. Its follow-up, the ECR-RS, has 9 short statements about one specific person: your mother, your father, a romantic partner, or your closest friend, and in later versions about close relationships in general as well (Fraley et al., 2011). Six statements cover avoidance, the first four of them reverse-keyed, and three cover anxiety.

Our test has 72 statements: 36 from the ECR-R and four ECR-RS blocks (close relationships in general, mother, father, friend). There is no separate partner block, because nearly all of its items match ECR-R statements. We compute the partner point from the 9 corresponding ECR-R items, so it is an approximation.

Norms for each relationship come from a study in which two online samples of 2,399 and 2,300 people answered about each of these relationships (Fraley et al., 2015). Those distributions are clearly skewed: many people give the lowest possible anxiety ratings for their mother. That’s why we compute relationship percentiles with a skewed model, and at the extremes of the scale they are approximate.

Internal consistency is high on both scales, with alpha at .90 or above. Stability holds up as well: when people retook it after three and six weeks, the latent scores shared 85% and 86% of their variance (Sibley & Liu, 2004; Sibley et al., 2005). The ECR-RS is reliable too, with alphas from .81 to .92 across relationships in the 2015 study (Fraley et al., 2015).

There are limits. Item response analysis shows that the questionnaire is most precise for people with marked anxiety or avoidance and less precise at the low end, near the secure corner. The ECR-R also captures how people themselves see their relationships. It relates only weakly to the Adult Attachment Interview (AAI), which rates how people talk about their childhood: the correlation is around .09 (Roisman et al., 2007). They are two different views of attachment, and neither is a diagnosis.

The best-known short form, the 12-item ECR-S, was built from the original 1998 ECR rather than from the ECR-R (Wei et al., 2007). Only 7 of its statements match the ECR-R, and it has its own norms. There are also short versions of the ECR-R itself, such as a 14-item Russian form (Chistopolskaya et al., 2018), and others in several languages, each with its own norms and cutoffs.

We use the full questionnaire. Short forms work well in large studies, where group averages matter, but for one person they give a rougher estimate and depend more on any single answer. We translated all 36 statements ourselves and kept the original 1 to 7 scale, so results can be compared with Fraley’s norms. Wording about the partner is gender-neutral.

1. 01Where can I find the ECR-R as a PDF?

   The statements, scoring key, and norms are on Fraley’s page for the measure, linked in the sources below. Without the authors’ permission, the scales may be used only for non-commercial research. To take the questionnaire and get your scores with percentiles right away, use our test.
2. 02Can the ECR-R be answered on a 1 to 5 scale?

   It can, but then Fraley’s norms no longer apply, because they were collected on a 1 to 7 scale. Some translations, such as the Korean one, use five points and their own cutoffs. We kept seven points so that comparing your scores with Fraley’s norms stays fair and direct.
3. 03Is the ECR-R available in other languages?

   Yes. It has been adapted into many languages, often as a shortened version with its own norms. Our test uses the original English statements and our own translations into Russian, Japanese, Italian, and Portuguese, independent of the published adaptations, always on the original seven-point scale, so every language can be compared with Fraley’s norms.
4. 04Is this the same test as the one on Chris Fraley’s site?

   The statements and the key are the same, because it’s the ECR-R. But ECRtest isn’t affiliated with the authors of the measure. Fraley runs his own research test on the university site, while we score answers with the published key and norms and add a relationship map, percentiles, and written descriptions.