The 36 Questions, Rewritten for Friendship

The famous 36 questions come from a 1997 psychology experiment that generated closeness between strangers in a laboratory, not from a study about friendship or love. They still work as a method, but the list needs rewriting before it fits a real friendship. Below is what the original actually was, why it does not transfer cleanly, and a set of 36 friendship questions in three escalating sets.

What the 36 questions actually are

The source is a paper by Arthur Aron and colleagues, published in 1997 in Personality and Social Psychology Bulletin under the title The Experimental Generation of Interpersonal Closeness. It describes a procedure, not a relationship recipe. Pairs of strangers worked through 36 questions in three sets of twelve across roughly 45 minutes, about 15 minutes per set, with each set more probing than the one before. The measured outcome was reported closeness immediately afterwards.

Three details get lost in almost every version circulating online:

• The paper is not called "the questions that lead to love". That framing comes from a 2015 New York Times Modern Love essay by Mandy Len Catron, who tried the questions with an acquaintance and later married him. It is a good essay and a single personal story. It is not a finding.
• The eye-contact step is poorly documented. The four minutes of silent staring that usually travel with the list appear with different durations in different retellings, anywhere from two to four minutes. Treat it as folklore attached to the list rather than as an established part of it.
• Closeness measured minutes later is not friendship.Feeling close to somebody at the end of a 45-minute session is a real effect and a genuinely useful one. It says nothing about whether the two of you are still talking in six months.

None of this makes the method worthless. It makes the method interesting for the right reason: reciprocal, gradually escalating self-disclosure reliably produces a feeling of knowing somebody. That is a mechanism you can borrow.

Why the original list does not fit friendship

People hand each other the original 36 and it goes flat. Usually for one of four reasons.

• The framing is romantic. Several questions assume a partner and a romantic register. Asked between two people exploring a friendship, they land as either a proposition or a joke.
• It was built for one synchronous sitting. Three sets in 45 minutes, in a room, with a timer. Friendships are built in fragments across weeks, often in writing. A procedure designed for one continuous session does not survive being spread over a fortnight without changes.
• The lab condition is the opposite of the friendship condition. Participants disclosed freely partly because they expected never to see the other person again. Consequence-free disclosure is easy. With somebody who might actually become part of your life, every answer costs something.
• It front-loads intensity. Set three arrives fast. In a lab with a researcher and a timetable, that works. Between two strangers on their own, question 31 arriving on day one reads as a stunt.

36 questions for friendship, in three sets

These are not Aron's questions reworded. They are a friendship set built on the same escalation logic, drawn from the question bank we actually use, and they keep the three devices that make the original method work rather than just its topics: a stated commonality, a complete-the-sentence prompt, and a request for advice on something real.

If you would rather use the original, the Greater Good in Action version is free and faithful to the paper.

Set 1: easy ground

Low cost, still revealing. Nobody has to be brave yet, and the point of this set is to establish that you will both actually answer.

1. A free weekend with nothing you have to do. What actually gives you energy back?
2. What is something small that reliably improves your day?
3. When are you genuinely at your best?
4. What did you want to be when you were ten, and what happened to that?
5. What is your usual role in a group of friends?
6. Be honest: what actually makes you laugh?
7. What is something you enjoy that you would never put on a profile?
8. What does a normal evening at your place look like?
9. How do you like to make plans, and how far ahead?
10. What is the last thing you got genuinely absorbed in?
11. Name three things you think we already have in common.
12. What do your favourite conversations tend to run on?

Set 2: how you actually live

This is where compatibility shows up. Not interests, but how somebody handles time, attention, conflict and other people.

13. A friend comes to you with a problem. Do you solve it, listen, lighten it, or say the uncomfortable thing?
14. Underneath everything, what is a real friendship mostly about?
15. What gives your life meaning, most of the time?
16. How much of your free time do you actually want to share with somebody close?
17. Who usually starts things between you and the people you are close to?
18. When your head and your heart disagree, which one usually wins?
19. What is something you have changed your mind about in the last few years?
20. Tell me about a time somebody showed up for you when they did not have to.
21. What are you working on in yourself right now, if anything?
22. When the world around you changes fast, what is your first honest reaction?
23. What are you good at that people rarely notice?
24. Complete this sentence: what I want more of in my life is.

Set 3: the part that costs something

Only worth reaching once sets 1 and 2 went well. If either of you has been giving one-line answers, stop here rather than pushing.

25. What would damage a friendship beyond repair for you?
26. What is your secret fear in friendships?
27. When you are really struggling, what do you actually do?
28. What do you need from people that you find hard to ask for?
29. Tell me about the last time you felt genuinely known by somebody.
30. What is something you are still a little embarrassed about?
31. When did you last cry, and what was it about?
32. Is there anything you treat as too serious to joke about?
33. What would somebody have to understand about you to actually get you?
34. Tell me one thing you have noticed and appreciated about me so far.
35. Describe a problem you are actually dealing with, and ask me what I would do.
36. If we stopped talking after today, what would you want me to remember about you?

The rule that carries all 36

One mechanic matters more than the list: both people answer before either one reads the other's answer. It removes the interviewer role, stops you rounding your answer toward theirs, and turns a disagreement into information rather than an awkward correction. We explain why that works in the guide to deep questions to ask a new friend.

Aron's procedure got this for free, because both participants answered every question in the same room in turn. Over text you have to arrange it deliberately or it collapses back into an interview.

What we saw when people did this for real

ValuePair runs a shorter version of the same idea: two people answer the same 14 questions, and neither answer unlocks until both have answered. Counts read from our production database on 27 July 2026, when the beta had roughly 450 people, so read them as history rather than as a rate:

• 447 structured conversations had received a first answer, and 389 of them reached the final question. Close to nine in ten.
• Almost all of the loss happened before the first answer existed, not in the personal questions near the end.

The practical lesson for a 36-question attempt: the risk is not that set three is too intense. The risk is that you never get past question one. Everything you do to make the first answer easy is worth more than anything you do to make the later questions deeper.

Two limits worth stating plainly. A few hundred beta participants over a few weeks is not a study, and completion is not friendship. Nothing here shows that a question sequence produces a lasting relationship, and no matching system can know that in advance.

How to actually run this

1. Do not attempt 45 minutes cold. The original had a researcher, a room and a timetable. You have neither. One set per sitting is plenty.
2. Agree the answer-first rule out loud. Both write, then both read. Say it before you start, or the first person to answer becomes the subject.
3. Keep answers to a few sentences, and add the reason. The reason behind an answer is what the other person actually learns from.
4. Stop where the energy stops. Finishing set 2 well beats limping through set 3. The list is not a test you pass.
5. Follow the differences. Where you match you have common ground. Where you differ you have the only genuinely interesting question available: why do you see it that way?

Mistakes that kill it

• Using the romantic original for a friendship. Either adapt the framing or use a friendship set. Do not ask somebody what they like about their partner and hope it lands.
• Asking without answering. The fastest way to make somebody close up, and the failure mode the original procedure was specifically designed to avoid.
• Treating it as a compatibility test. These are not a scoring instrument. A difference is a topic, not a verdict.
• Expecting the effect to be the relationship. Feeling close after a long honest conversation is real, and it is a beginning rather than a result.

If you want the structure without arranging it

Running this with somebody you already know works well. The harder case is not having that person yet.

ValuePair introduces one compatible person at a time, keeps photos out of the matching decision, and enforces the answer-first rule for both of you across 14 questions so nobody has to police it. It is a free, open beta with no ads and no payments. You can create a profile and join the matching pool, or read how the format works in our guide to a friendship app without swiping.

Common questions

Do the 36 questions actually work?

They reliably produce a feeling of closeness at the end of the conversation, which is what the 1997 procedure was built to do and what it measured. Whether that becomes a friendship depends on everything that happens afterwards, and no study shows otherwise.

Do we have to do all 36?

No. The escalation matters, the total does not. Twelve questions answered properly by both people beat 36 answered by one.

Can I use the original list instead?

Yes, and the Greater Good in Action version is faithful and free. Expect to reword the questions that assume a romantic partner.

Does this work over text?

Yes, and written answers are often more considered because nobody is filling a silence. The cost is pace, so you need patience and a way to keep both people moving. Our FAQ explains how ValuePair paces it.

Is this for dating or for friendship?

The set above is written for friendship. Sets 1 and 2 transfer to dating with no changes. ValuePair itself is friendship-first, with dating optional and only where both people's stated intentions overlap.

Sources and editorial note

ValuePair publishes this guide and is the product described near the end of it. The engagement counts above were read from our own production database on 27 July 2026 and are reported in aggregate.

The original procedure is Aron, Melinat, Aron, Vallone and Bator, 1997, which measured closeness generated between strangers in a controlled setting. The popular "questions that lead to love" framing comes from Mandy Len Catron's 2015 New York Times Modern Love essay, a personal account rather than a research result. We make no claim that any question sequence predicts friendship or relationship outcomes.