Questions for couples: the 36, and four things everyone gets wrong

All thirty-six below, free, with nothing to sign up for. Then the more interesting part: almost everything written about them is describing a magazine essay rather than the study — and the study says something more useful than the essay does.

They come from a 1997 paper by Arthur Aron and colleagues. Pairs of students who had never met sat together in a large psychology class, worked through three sets of twelve slips, and were stopped by an experimenter at fifteen minutes whether they had finished a set or not. Forty-five minutes in total.

The thirty-six

Four things everyone gets wrong

1. The four minutes of eye contact was never in it

There is no gaze task anywhere in that paper — not in either condition, not in the appendix, not in any of the three studies. The staring comes from a different and earlier piece of work, a 1989 study on mutual gaze and feelings of romantic love.

The version everybody knows — thirty-six questions, then four minutes of looking at each other — comes from Mandy Len Catron’s 2015 Modern Love essay about trying it with an acquaintance in a bar. It is a lovely piece of writing and it is not the study, which she has said herself more than once.

2. The control group was not small talk

This is the most mangled part, and correcting it makes the finding better rather than worse. The comparison condition was a matched list of thirty-six shallow questions, in the same three sets, on the same clock, read aloud and answered by both people in exactly the same way. From the paper’s own appendix:

Do you prefer digital watches and clocks or the kind with hands? Why? · What are the advantages and disadvantages of artificial Christmas trees? · Describe your mother’s best friend.

the control condition, Aron et al. (1997), appendix

So the result is not a list of questions beats no list. It is that the content of the list changes the outcome when the structure, the clock and the format are held identical. That is a narrower claim and a much more interesting one.

3. “Closer than their closest friend” is backwards

You have probably read that the strangers ended up closer than the participants’ closest relationships. The paper says the opposite. It compares the result against a separate sample of students rating their own closest, deepest, most intimate relationship, and reports that after forty-five minutes the strangers landed at about the thirtieth percentile of those ratings.

Forty-five minutes with a stranger beat the closest relationship of roughly a third of students. Which is a remarkable thing for forty-five minutes to do, and is not the thing it gets quoted as.

4. The couple who married were not in this study

That detail appears in the 1997 paper as a parenthesis about an earlier, differently designed pilot — a ninety-minute version built around something else entirely. One pair, out of fifty-three. It is the single most repeated fact about this research and it is an aside about a different experiment.

What actually made it work

The authors did not set out to write a love potion. Their stated aim was to make “being in a relationship” something a laboratory could switch on, in the way that mood induction procedures let you switch on a mood. It is a methods paper.

And they tested four ingredients to see which mattered. Only one of them did.

• The content of the tasks — escalating self-disclosure rather than shallow questions. This was the whole effect.
• Matching people by attitudes so they would not disagree — no effect.
• Telling them to expect that they would like each other — no effect.
• Naming closeness as the goal — no overall effect, and it mattered only for introverts, who caught up with extraverts when it was named.

There is a fourth, unnumbered demonstration in the same paper that makes the point bluntly: 222 students at a different university, no questionnaire, no matching, paired on the spot — and the closeness score came out the same. The elaborate apparatus contributed nothing. The questions did all of it.

The detail that matters most if you are already a couple

Here is a small piece of study design that nobody mentions and that changes how you should use the list.

Aron did not have people rate their closeness in front of each other. When the forty-five minutes were up, participants were told to move to another spot in the room, as far from their partner as possible, and then fill in the questionnaire.

That is a methodologist’s judgement rather than a tested hypothesis. But it is the most on-point fact in the paper for anyone doing this with a partner rather than a stranger, and it is the problem with the list as it is usually used: you go first, your answer sets the terms, and the second person is no longer answering the question. They are responding to you.

And the honest complication

We believe strongly in answering separately, so it is worth saying that the research on this does not all point our way.

Studies of survey answers given with a spouse in the room do find that presence changes the answers. But the effect is convergence rather than straightforward impression management — couples’ answers move towards each other, and in one large German panel study the partners’ accounts of who did which housework became more consistent and less likely to add up to an impossible total. Those authors conclude that partner presence may improve data quality rather than damage it.

The work we would point to on the other side — that social influence narrows the range of answers and raises confidence without improving accuracy — was done on factual estimates, not on people describing their own relationship, so it is an analogy rather than evidence. Nobody, as far as we can find, has run the actual experiment: two partners answering the same relationship question blind, then revealing.

What we would say is narrower than a finding. Answering separately gives you two pictures instead of one negotiated one, and the distance between them is usually the interesting part of the evening.

If you have been together years, do something else

This is the recommendation the research supports and no page about the 36 questions makes. The list is a stranger protocol. It was built to manufacture closeness between people with no shared history, and a great deal of its power is that everything being disclosed is new.

Aron’s own later work on couples who are already together points somewhere different. In three experiments, seven minutes of a novel shared activity raised how couples rated their relationship, compared with a pleasant but mundane one and with doing nothing.

It is worth knowing what the active ingredient in that turned out to be, because it is usually reported wrongly. The original tasks were both new and physically arousing, so the two could not be told apart. A later set of four studies separated them and found that the newness predicted the benefit and the arousal did not. So the prescription is not “do something exciting”. It is: do something that actually teaches you something.

Which puts the list in its proper place rather than dismissing it. A question whose answer you could already have written is not doing the work, however deep it sounds. A question you genuinely cannot predict is — and after four years together those are rarer, and less likely to be on a famous list, than they were on the day you met.

A second study found something odder and rather charming: sixty couples did a high-disclosure session with another couple, and came away feeling closer to their own partner.

So use the list, by all means — but if you have been together four years, the honest advice is that a new thing done together on a Saturday has better evidence behind it than question nineteen.

Where inbetween fits

Our version of these is built on the one thing the list as published cannot do: you each answer where the other cannot see, and you both see everything in the same second. No one goes first, so nobody is responding rather than answering — and it works across a time difference, because you do not have to be awake at the same time.

The games run on the same mechanic for an evening that does not need to be heavy, and the exercises page has the material with the best evidence behind it if things are harder than that.