Why Your Attachment Style Quiz Result Changed

You took it in March and it said anxious. You took it again in July and it said fearful-avoidant. One of those is presumably wrong, and the obvious conclusion is that the quiz is not worth much.

That conclusion is available and sometimes correct. But there are four ordinary reasons a result moves, and only one of them is a bad quiz. Working out which applies turns an annoying inconsistency into the most informative thing the exercise has produced.

The Four Ordinary Reasons

One: you answered about a different relationship

Attachment behaviour emerges between two specific people. It is shaped heavily by how predictable the other person is, which means the same person can produce genuinely different answers about a steady partner and about an unreliable one.

Here is the part almost no quiz mentions: the instructions usually do not say which relationship to answer about. So you supply one silently. In March you were thinking about the person you were seeing. In July you were thinking about the way it ended, or about your mother, or about nobody in particular and therefore about an average of everyone. The subject changed, so the answer changed.

If you cannot now remember who you had in mind, that is not a memory problem. It is an unrecorded variable, and it is the single largest source of movement between sittings.

Two: you answered from a different state

These questionnaires are written as trait measures and are always completed by a person in a state. Three days after an argument, an item about worrying whether someone really cares does not read the same as it does on a quiet Sunday. Nothing about you has changed. The item has.

A result taken in the week after a rupture is a state measurement wearing a trait label. So is a result taken six weeks into something that is going unusually well, which produces a flattering reading that tends not to survive the first disappointment. Illness, exhaustion and a bad stretch at work all push in the same direction.

Three: you sit near a boundary

The four names are regions of a space made by two continuous dimensions. The thing being measured is continuous; the label is discrete. Near a dividing line, a two-point shift changes the word you are given and almost nothing about where you actually sit.

There is a quick test for this. If your two results are neighbouring labels, anxious and fearful for instance, or secure and dismissive, a boundary effect is the most likely explanation and your position barely moved. If your two results are opposite corners, secure in March and fearful-avoidant in July, that is not a boundary effect and one of the other three reasons is doing the work.

Four: you actually changed

Attachment patterns do move. A year of therapy, a long stretch of a steady relationship, the end of an unsteady one, a bereavement, a new baby: all of these show up in the research on change over time, in both directions.

Note the timescale, though. This is the least likely explanation for a change over a fortnight and a very reasonable one for a change over two years. If your two results are close together in time, look at the first three reasons before crediting this one.

When the Quiz Itself Is the Problem

Sometimes it genuinely is the instrument. The signs are reasonably easy to spot: very few items, so a single answer swings the outcome; questions written so transparently that you can see which response produces which label; forced choices with no honest option for someone who does both depending on the situation; a result that arrives attached to a product.

One more, and it is the most reliable warning sign of all. Be sceptical of any quiz that gives you a confident percentage. Precision of that kind is a design choice about how the result looks, not a property of what was measured. The broader question of what these instruments can do at all is covered in how accurate attachment quizzes are.

Worth holding onto: an instrument can be perfectly stable and still wrong. A quiz with only two possible outcomes would give you the same answer every time. Consistency is not evidence of quality, which is why a moving result is weaker evidence against a quiz than most people assume.

A Protocol for a Reading That Holds Still

If you want something more stable than a single sitting, this takes about twenty minutes in total and works with any decent quiz.

Choose the relationship before you start, and write the name down. Not relationships in general. One person.
Answer when nothing is currently happening in it. Not the week of an argument, not the first month of something new.
Answer about the last six months rather than about this evening. When an item is hard to place, that longer window is the tie-breaker.
Take it three times across a month, same relationship, and treat the middle result as your reading rather than the most recent one.
Then take it once more, deliberately, about a different relationship, and treat the difference between the two as its own finding.

That last step is the one people skip and the one that pays. A large gap between two relationships tells you something about the relationships as much as about you, and it is information a single result cannot contain. Our quiz shows the result immediately and asks for nothing, which makes running it four times practical rather than tedious.

Read the Dimensions, Not the Label

Even a well-built quiz discards most of what it knows when it converts your answers into one of four words. If the result reports scores as well as a name, the scores are the part worth recording. If it does not, write two sentences instead of a word: roughly how strong the fear-of-distance side is, roughly how strong the pull-away side is.

Keep that note with the date on it. Retaking in six months against a written record is a completely different exercise from retaking against memory, because memory reliably reconstructs the past to match the present, which means it will tell you either that nothing has changed or that everything has. Neither is reporting. This is also the practical reason research instruments produce continuous scores in the first place.

A Moving Result Is Not Always a Broken Instrument

The assumption underneath the complaint is that your attachment style is a stable property of you, like a blood group, so any variation must be measurement error.

Some of it is. But a good part of what these questionnaires measure genuinely moves, because it is a description of how you are relating rather than a fact about your constitution. A thermometer that showed the same reading in every room would not be a more reliable thermometer.

So two different results a year apart are not two wrong answers. They are two data points, and the direction of travel between them is more informative than either one on its own. What deserves your attention is not that the word changed, but which dimension moved to change it, and what was happening in your life at the time.

Frequently Asked Questions

Why did my attachment style quiz give a different result the second time?

Usually one of four reasons. You had a different relationship in mind, since attachment behaviour is relationship-specific and most quizzes never say which one to answer about. You answered from a different emotional state, such as the week after an argument. You sit near a boundary between two labels, so a small shift crossed a line. Or you genuinely changed, which is normal over long periods.

How often should I retake an attachment quiz?

For a stable reading, three times across a month about the same relationship, then treat the middle result as your answer. For tracking change, roughly every six months, compared against a written record rather than against memory. Retaking repeatedly in a single week mostly measures your mood that week rather than anything more durable.

Does a changing result mean the quiz is inaccurate?

Not on its own. Some of what these questionnaires measure genuinely moves, because it describes how you are relating rather than a fixed property like a blood group. Stability is also not proof of quality: an instrument can return the same answer every time and still be wrong. Look instead at whether the two results were neighbouring labels or opposite corners.

Should I trust my first result or my most recent one?

Neither by default. A single sitting is one observation taken in one mood about one unrecorded relationship. The middle of three readings, taken across a month about the same person, is a better estimate than either the first or the latest. If you only have two and they disagree, the gap between them is the finding rather than a problem to be resolved.