Two dimensions instead of three categories
Attachment research started with babies. Mary Ainsworth's strange-situation experiments sorted infants into secure, anxious-ambivalent and avoidant, and for a while adult research copied the habit of sorting people into types. The problem is that a type label throws away information. Two people both labelled "avoidant" can be very different, and plenty of people do not fit any of the boxes cleanly.
The Experiences in Close Relationships scales take the other approach. They measure two continuous dimensions and let your position on those dimensions do the work. Attachment anxiety covers how much you worry about being rejected, abandoned or not valued enough. Attachment avoidance covers how uncomfortable you are with closeness and depending on someone.
Keep those two dimensions separate in your head and the four familiar style names become combinations rather than identities.
The 36 items
The original ECR came from Kim Brennan, Catherine Clark and Phillip Shaver in 1998. It has 36 items, 18 for each dimension, rated from 1 ("strongly disagree") to 7 ("strongly agree"), with 4 as a neutral midpoint. Some items are reverse-scored, meaning agreement counts as low anxiety or low avoidance. The Chinese version used on this site keys 14 of the 36 items in reverse.
Item wording stays concrete. Anxiety items ask about worrying that a partner will lose interest, needing reassurance, or feeling upset when a partner is not available. Avoidance items ask about preferring not to show feelings, finding it hard to depend on someone, or pulling away just as things get close.
Why there is a revised version
By the late 1990s there were more than a dozen self-report attachment questionnaires in circulation, and no good way to compare them. Fraley, Waller and Brennan (2000) took 323 items drawn from 14 of those inventories and administered them to 1,085 undergraduates at the University of Texas at Austin, 682 women and 403 men with a median age of 18. They then applied item response theory, a family of models that describes how much information each item carries at each level of the underlying trait.
The comparison was unflattering to the field. The ECR's test information functions were clearly higher than those of the other instruments in the study, but its precision was uneven across the range: it measured the anxious and avoidant ends reasonably well and the secure end poorly. That unevenness matters because it can make a continuous distribution look lumpy, as though people cluster into groups when they do not.
So the authors built a replacement from the same item pool. They ran a principal-axis factor analysis on 30 clusters of similar items, then picked the 18 items with the highest discrimination for each dimension. Thirteen of the 18 anxiety items, 72%, came from the original ECR; for avoidance only 7 of 18, or 39%, did. The result was ECR-R, which increased measurement precision by 50% to 100% without adding a single item. The information curve for the revised anxiety scale is roughly double that of the original.
One limit stayed. Both versions still measure the low-anxiety, low-avoidance region less precisely than the rest, so a very secure-looking score is the least trustworthy part of the profile.
Scoring
Each subscale is the average of its 18 items after reverse-scoring, so both scores run from 1 to 7. There is no total score, and there should not be one: someone can be high on both dimensions, low on both, or high on one and low on the other, and averaging that away destroys the useful part.
The four patterns
| Anxiety | Avoidance | Pattern |
|---|---|---|
| Low | Low | Secure. Comfortable with closeness and with independence |
| High | Low | Preoccupied. Wants closeness badly and worries about losing it |
| Low | High | Dismissing. Values independence over intimacy and keeps distance |
| High | High | Fearful. Wants closeness and expects it to hurt, often at the same time |
The split between "low" and "high" has to be drawn somewhere, and different studies draw it in different places. The version on this site divides each dimension at 2.5 on the 1-to-7 scale. That is a reporting convenience, not a clinical boundary, and a score just above the split is not meaningfully different from one just below it.
Reliability
The original ECR recorded alpha of .94 for avoidance and .91 for anxiety in an undergraduate sample. Across studies and populations, alpha for both dimensions stays above 0.80 and often reaches 0.90 or higher, with retest intervals of weeks to months holding up.
Fraley's simulations, run a hundred times over the observed item parameters, put retest correlations for ECR-R at .93 and .94 for anxiety and .95 and .95 for avoidance, against .91 and .94, .90 and .91 for the original. The practical difference is small at the level of an individual score, which is another way of saying the original was not broken. The revised version is better at the extremes and better distributed in between.
What it predicts
Both dimensions show up in relationship outcomes. Avoidance predicts less self-disclosure and lower relationship satisfaction, and it shows up in how conflicts get handled, with more withdrawal and less repair. Anxiety predicts more reassurance-seeking and more vigilance for signs of rejection, which can create the very distance the person fears. The dimensions also correlate with emotion-regulation style and remain distinguishable from neuroticism, which matters because otherwise the whole thing could be dismissed as a repackaging of that broader trait.
Attachment style is reasonably stable but not fixed. Scores shift with relationship experiences and with therapy, and the two dimensions can move independently.
The Chinese version
Li Tonggui and Kato Kazuo published a Chinese revision in Acta Psychologica Sinica in 2006. They tested 371 Chinese university students, analysed the 231 who were currently in a relationship, and retested 59 of them four weeks later, checking validity against the Rosenberg Self-Esteem Scale and other measures. Both internal consistency and retest reliability came out acceptable, along with construct and criterion validity.
Take the test
The CheckPsych version has all 36 items and takes about 10 minutes. You get separate anxiety and avoidance scores between 1 and 7, plus the pattern those two scores form.
References
Brennan, K. A., Clark, C. L., & Shaver, P. R. (1998). Self-report measurement of adult attachment: An integrative overview. In J. A. Simpson & W. S. Rholes (Eds.), Attachment theory and close relationships (pp. 46-76). Guilford Press.
Fraley, R. C., Waller, N. G., & Brennan, K. A. (2000). An item response theory analysis of self-report measures of adult attachment. Journal of Personality and Social Psychology, 78(2), 350-365.
李同归, 加藤和生. (2006). 成人依恋的测量：亲密关系经历量表（ECR）中文版. 心理学报, 38(3), 399-406.
Disclaimer: this article and the ECR-R are for general education and self-reflection. They are not a diagnosis and not a treatment recommendation. Attachment scores describe tendencies in close relationships, they shift with circumstances, and the 2.5 split used here is a reporting convention rather than a clinical threshold. If relationship patterns are causing you distress, a qualified therapist can help you work with them.