How the friendship test was built

What we built

The four styles, the six words that can sit in front of them, and every question in the test are Kinu's own writing. The distinctions underneath them come from published friendship research and are credited below; the vocabulary is ours, and we are glad to put our name on it. It diagnoses nothing, ranks nobody, and never compares you with anyone. No validation study stands behind it, and none is claimed; what stands behind it is the research below and the testing we did ourselves.

Two sentences govern everything it says. It describes what you told us, not what your friend is like; and it is built from published friendship research, not a study of our own.

Before publishing, we tested the machinery. Thousands of constructed answer patterns went through the same arithmetic your answers do, to confirm that consistent patterns land where they should, that contradictions land in the middle, that agreeing with everything earns nobody a badge, and that the arithmetic does not hand any one style most results on its own. Constructed answers exercise the calculation rather than the wording, and patterns we built ourselves cannot say how real people will spread across the four styles; whether the questions read to you the way we intended is something only real people can answer, and we would rather know than guess.

How your result is worked out

Every statement you rate counts toward one of five readings: the two axes, who reaches first and one person or the room, and three quieter shades, memory, rhythm, and depth. Half the statements are written in reverse, so a habit of agreeing cannot pull you anywhere on its own. Then the step that shapes everything: your own typical answer is subtracted from each answer before anything is decided. The test reads you against yourself, not against a population. The only kind of claim it can make is that you answered some things more strongly than you answered most things.

Each reading then lands in one of three bands: clearly to one side, leaning, or balanced. An adjective like Steady appears only in the first case; balanced earns a plain badge, and that is a result, not a failure to have one. Nowhere is there a number. A marker on a bar shows where you sat, because a bar can be honest about the middle in a way a precise-looking figure never is.

Where the test stops

Three limits, stated plainly.

It cannot tell you what you are like in any one friendship. Who reaches first in a single pair is partly a property of the pair: when one person habitually takes the lead, the other tends to settle into responding (Kiesler, 1983). The badge is built from questions about your habits in general, so it names your overall lean. You may be an Opener in one friendship and something else entirely in another.

It cannot tell you whether a particular friendship is lopsided. Whether investment feels balanced matters to how a friendship feels and whether it lasts (reviewed in Fehr, 1996). This quiz does not measure that.

And it cannot speak for the people you name. Across six datasets, only a third to a half of named friendships were mutual (Almaatouq et al., 2016). The test never asks your friends anything, so everything it says is your side of the story. That is why its sentences keep returning to what you told us.

The research it draws on

The distinctions are older and sturdier than the vocabulary we hung on them. The activity and confidant contrast goes back to Wright's (1982) distinction between side-by-side and face-to-face friendship; Caldwell and Peplau (1982) drew the same line between friendships built on doing and friendships built on confiding. Friendships are not purely chosen; they arise from and are shaped by the contexts people share (Allan, 1989; Adams & Allan, 1998), which is where the test's question about how each friendship began comes from. The one-person-or-the-room line is not only ours. The Friendship Habits Questionnaire (Howlett et al., 2023) draws the same distinction between group and one-to-one socialising across three studies, and it predicts the size of the groups people actually enjoy. The same line also sits differently in different countries (Howlett et al., 2026), which is why a Spanish version of this test will be read on its own terms and never against the English one. Our items are our own and were not validated against either study; the distinction is theirs. And friendship networks shrink and reshuffle across adulthood as life circumstances change. How often you are in touch is a circumstance, not a kind of friendship (Wrzus et al., 2013), which is why the test treats a friendship's quiet stretches as a rhythm, never as a verdict.

Sources

1. Wright, P. H. (1982). Men's friendships, women's friendships and the alleged inferiority of the latter. Sex Roles, 8(1), 1-20.
2. Caldwell, M. A., & Peplau, L. A. (1982). Sex differences in same-sex friendship. Sex Roles, 8(7), 721-732.
3. Allan, G. (1989). Friendship: Developing a Sociological Perspective. Hemel Hempstead: Harvester Wheatsheaf / Boulder: Westview Press.
4. Adams, R. G., & Allan, G. (Eds.) (1998). Placing Friendship in Context. Cambridge University Press.
5. Kiesler, D. J. (1983). The 1982 Interpersonal Circle: A taxonomy for complementarity in human transactions. Psychological Review, 90(3), 185-214.
6. Wrzus, C., Hänel, M., Wagner, J., & Neyer, F. J. (2013). Social network changes and life events across the life span: A meta-analysis. Psychological Bulletin, 139(1), 53-80.
7. Fehr, B. (1996). Friendship Processes. Thousand Oaks, CA: SAGE.
8. Almaatouq, A., Radaelli, L., Pentland, A., & Shmueli, E. (2016). Are You Your Friends' Friend? Poor Perception of Friendship Ties Limits the Ability to Promote Behavioral Change. PLoS ONE, 11(3), e0151588.
9. Howlett, P., Baysu, G., Atkinson, A. P., Jungert, T., Rychlowska, M., & Wang, F. (2023). Friendship habits questionnaire: A measure of group- versus dyadic-oriented socializing styles. PLoS ONE, 18(6), e0285767.
10. Howlett, P., Baysu, G., Jungert, T., Atkinson, A. P., Namba, S., Sato, W., Mizuno, K., & Rychlowska, M. (2026). Friendships are more group-oriented in the United Kingdom than in Japan. British Journal of Social Psychology, 65(1), e70040.

What is different about it

Other friendship quizzes exist, and some of them are careful: they run in your browser too, and some name a cost for each type, as this one does. Two things are ours. The second half asks about the actual people in your life, by first name, and shows you your own answers about each of them, without asking those people to take anything. And every result carries the date you took it, because friendships change, and a reading from a year ago is a different reading rather than an old one.

Why names never leave the browser

The second half of the test asks about specific friendships, by first name, because the questions read differently when a real name is in them. Those names exist only on your screen. They are never sent to Kinu, never placed in a web address, never written into a share image, never included in any analytics. The whole test is worked out on your device, and your result page is assembled there too. If you share what you got, what travels is the style and its words, nothing about anyone you named. There is nothing about the people you named held on our side for anyone to see, which is not a policy so much as an absence.

The sentences about each friend follow the same principle in a second way: they are your own answers, arranged and shown back to you. The test does not measure your friend. It could not; it never met them.

Retaking it

A personality test expects you to be the same person next year. Friendship makes no such promise. People move, jobs change, an easy friendship becomes a close one, a close one goes quiet for a season, and none of that is failure; it is what the thing is. So your result carries the date you took it, and taking it again a year on is not a do-over but a genuine second reading. When you retake it, your previous positions remain visible beside the new ones, so change shows up as what it is: friendships moving, not a result flickering.