How Common Is Each Attachment Style? The Numbers

You came here for a percentage, and the reason you want one is usually not statistical. You got a result, and you want to know whether you are in the ordinary group or the odd one.

So here is the number that circulates, followed by the more useful article, which is about why that number is much softer than it looks when it is printed in bold on an infographic.

The Split You Have Probably Seen

The figures repeated across most attachment content run roughly like this. Secure is the largest single group, usually given as around half. The insecure categories divide the remainder, with the anxious and dismissive-avoidant patterns each taking a meaningful share, and fearful-avoidant coming out smallest.

Treat that as approximate and commonly cited rather than established. It is a fair description of the shape most general-population samples produce. It is not a measured constant, there is no single authoritative source for it, and the precise percentages attached to it in any given article are usually inherited from another article rather than from a study.

This piece deliberately gives no exact figures. That is not evasiveness. It is that the honest interval around each of those numbers is wide enough that quoting them to the decimal point would misrepresent what is known.

Where the Familiar Numbers Came From

The four-way adult model and its associated split trace back to a small number of studies conducted mostly in North America, at a point when adult attachment research was new and samples were assembled out of whoever was available.

That is not a criticism of the original work. Early research in any field runs on convenient samples, and those studies were doing something that had not been done before. The problem is what happened next: a distribution measured in a particular population, at a particular time, with a particular instrument, got copied into general circulation as though it were a fact about human beings.

It has been repeated so often that it now sounds like something everybody knows. Frequency of repetition is not evidence.

Four Things That Move the Percentages

Prevalence estimates for attachment are not noisy in a random way. They move in predictable directions depending on four decisions the researchers made, and once you can see those decisions you can read any figure you are handed.

1. The measure

Self-report questionnaires and coded interviews produce different distributions of the same population, and they agree only modestly about individual people. Sorting adults by what they say about their relationships is not the same operation as sorting them by how coherently they can talk about their childhood, and it would be strange if the two produced identical proportions.

There is also a direction to the bias worth naming. Self-report depends on someone noticing and admitting distress, and playing down distress is close to the defining feature of the avoidant pattern. So questionnaires plausibly undercount exactly the group least inclined to fill one in honestly. How attachment is actually measured goes through what each instrument can and cannot say.

2. Who got sampled

A very large share of adult attachment research has been run on university students, because universities are where the research happens. Students are young, mostly unpartnered or newly partnered, and in the specific life phase where relationships are least settled. They are not a small version of the general public.

Clinical samples run the other way and over-represent insecurity, which is unsurprising given that distress in relationships is a common reason to seek help. Community samples of long-married adults skew secure. And online samples skew insecure for an obvious reason: people who are content in their relationships do not usually go looking for an attachment questionnaire at eleven at night.

That last one applies to this site. The population that takes an online attachment quiz is not a random sample of anything, and any distribution drawn from one would tell you about who searches, not about who exists.

3. The country

Cross-national comparisons of infant classifications found the balance of the insecure sub-types shifting between countries, in directions that tracked local childrearing norms. Some western European samples produced relatively more avoidant classifications; some Japanese and Israeli samples produced relatively more of the resistant pattern.

What that means is genuinely disputed, and the dispute is the interesting part. One reading is that cultures really do produce different distributions. Another is that a standardised laboratory separation does not mean the same thing to an infant who is routinely left with other adults as it does to one who has almost never been apart from their mother, so the procedure is not measuring identically across settings. Both readings have serious people behind them, and it is a good reason to be sceptical of any global figure.

4. Boxes or dimensions

This is the one almost nobody mentions, and it may be the largest. Most modern research measures attachment as two continuous scores, anxiety and avoidance, rather than as four types. A percentage can only be produced by drawing lines across those scores and counting what falls in each quadrant.

Move the lines slightly and the percentages move a lot, because most people sit near the middle rather than out at the corners. Nobody in the sample has changed. The categories are a summary decision applied afterwards, and any prevalence figure inherits whatever decision that study made.

What Stays Stable When the Percentages Do Not

It would be easy to read all of that as the numbers are meaningless, and that goes too far. Some things survive the variation.

• In general-population samples, secure is reliably the largest single group. That ordering holds across measures and countries even where the percentages disagree substantially.
• The insecure categories together are a minority, and a large one. Insecure attachment is not rare or unusual in any sample anyone has looked at.
• Fearful-avoidant is consistently the smallest of the four, however it is measured.
• In samples drawn from people already in distress, the balance shifts toward insecurity, which is what you would expect and not a finding about the general public.

That is a real result and it is worth holding onto. The ordering is far more robust than the numbers, and the ordering is what most people are actually asking about when they ask how common their result is.

Why the Fearful-Avoidant Figure Is the Least Trustworthy

If you have seen wildly different numbers quoted for this one, there is a reason, and it is structural rather than careless.

The fearful-avoidant pattern is defined by scoring high on both dimensions at once. That places it in a corner of the space where relatively few people sit, and where a small change in a cut-off flips a noticeable number of people in or out. Small groups defined by a corner are exactly the groups whose estimates jump around between samples.

There is also a naming problem. Some sources use fearful-avoidant and disorganised interchangeably, and they are not the same construct or measured the same way. When two studies count different things and both print a percentage, the percentages are not comparable, whatever the labels say.

A Better Question Than Whether You Are Rare

Underneath the interest in prevalence is usually a different question: am I unusual, and is that bad.

The category answer is not much help with either. Belonging to a large group tells you nothing about how you experience relationships, and belonging to a small one is not a severity rating. What does carry information is the underlying scores, because anxiety and avoidance are two separate dimensions and where you sit on each is more informative than which quadrant those two positions place you in.

Someone marginally over the line into anxious and someone at the far end of the scale receive the same word from a quiz. Their weeks look nothing alike. That difference is invisible in every prevalence table ever published, and it is the difference you actually live in.

How to Read Any Prevalence Number You Are Given

Five questions, which take about ten seconds and will disqualify most of what you encounter.

1. Was it measured by questionnaire or by interview? The two give different distributions and only modest agreement.
2. Who was in the sample? Students, patients, married adults and website visitors are four different populations.
3. Which country, and was the measure designed there?
4. Were the categories cut from continuous scores, and if so, where were the lines drawn?
5. Is there a source at all, or does the trail end at another article that also had no source?

That fifth one eliminates a startling amount of it. Follow the citations on attachment prevalence figures far enough and they frequently loop back into other blog posts, which is how a rough approximation from a specific sample became a number people quote as though it were the height of Everest.

Frequently Asked Questions

What percentage of people are securely attached?

The figure usually circulated is around half, and it should be treated as approximate and commonly cited rather than established. Secure is reliably the largest single group in general-population samples, and that ordering is much more robust than the percentage attached to it. The exact number moves with the measure used, the population sampled and where the study drew its category boundaries.

Which attachment style is the rarest?

Fearful-avoidant is consistently the smallest of the four. It is also the estimate that varies most between studies, for two structural reasons: it is defined by scoring high on both anxiety and avoidance at once, so small changes in a cut-off move a lot of people in or out, and some sources conflate it with disorganised attachment, which is a different construct measured a different way.

Why do different sources give different attachment style percentages?

Because four decisions change the answer: whether attachment was measured by questionnaire or by coded interview, who was sampled, which country the sample came from, and where the boundaries were drawn when continuous scores were converted into categories. Two honest studies can produce noticeably different distributions from the same underlying reality without either being wrong.

Does being in a rare attachment category mean my situation is worse?

No. Prevalence is not severity. How common a category is says nothing about how much difficulty any individual in it experiences, and the categories themselves are summaries of two continuous scores. Where you sit on attachment anxiety and on attachment avoidance carries far more information about your actual experience than which of the four boxes those positions place you in.