← Journal

Questions to get to know someone (that are not small talk)

By Kenneth Kintsugi, on friendship · Aug 5, 2026 · 8 min read

Most lists of questions to get to know someone are 100 questions long and unusable. Nobody has ever said “what is your biggest regret” to a person they met twenty minutes ago and had it go well. The useful thing is not more questions. It is knowing which register you are in, and having three that open something.

There is a real piece of research underneath all of this, and it is more interesting than the version that gets passed around. In 1997 Arthur Aron and four colleagues put pairs of strangers in a room for 45 minutes and gave them a list of 36 questions that escalated from ordinary to genuinely personal. A control group got 36 superficial questions instead. The escalating list produced more closeness. That is the finding.

It went around the internet in 2015 as “the 36 questions to fall in love” after a New York Times Modern Love essay used them on a first date. The paper itself never says love. It says closeness, and it was measuring how close two people felt after three quarters of an hour, not whether they ended up together.

The part of the study nobody quotes

Aron and his colleagues ran three studies, not one. The first is the famous one. The second and third tested whether you could improve on it, and the answer was no, in three separate ways.

Matching the pairs so they agreed on things that mattered to them: no significant effect. Telling the pairs in advance that they would probably like each other: no significant effect. Telling them explicitly that the goal was to get close: no significant effect. The authors note there was adequate statistical power to detect these, so it is not a case of the study being too small to see anything.

Read that as a set and it is a fairly blunt message. Compatibility on paper did not help. Optimism did not help. Trying did not help. What worked was two people taking turns answering questions that got gradually harder to answer.

Source: Aron, A., Melinat, E., Aron, E. N., Vallone, R. D., & Bator, R. J. (1997). “The Experimental Generation of Interpersonal Closeness: A Procedure and Some Preliminary Findings.” Personality and Social Psychology Bulletin. The “fall in love” framing comes from Mandy Len Catron’s 2015 essay in the New York Times, not from the paper.

Three registers, and why mixing them up is what makes it awkward

The 1997 list works because of the order, not the content. Escalation is the mechanism. Which means the practical skill is not memorising questions, it is noticing which of three registers you are currently in and not jumping two ahead.

Almost every failed attempt at this is a deepening question asked while both people are still in opening.

Opening questions that are not weather

The job here is low. You want them talking, and you want to hand them something easier to answer than “so what do you do”, which makes half the population brace.

• What have you been doing too much of lately?
• What is the last thing you got properly absorbed in?
• Are you a morning person, or do you just pretend at work?
• What did you think you would be doing at this age?
• Where did you grow up, and do you still sound like it?

Notice these all have an easy version and a real version, and the other person picks. That is the whole trick. A question that can be answered in four words without rudeness is a question people relax into.

Turning questions, where you find out if there is anything here

This is the register most conversations never leave, and it is where most of the useful information actually lives. You are not after confession. You are after opinion, because opinions are where people become specific.

• What is something everyone around you seems to enjoy that you quietly do not?
• What would you do with a completely free Saturday, if nobody would judge you for it?
• What is a thing you have changed your mind about?
• Who do you know who has the life you would want, and which part of it?
• What are you better at than you let on?
• What is the most useless thing you know a lot about?

The third one is the strongest question on this page. People almost never get asked it, everyone has an answer, and the answer is always about something they care enough to have thought twice about.

Deepening questions, and when you have earned them

These are the ones the 1997 study leaned on, and they are the ones people misuse. The signal that you can go here is not how long you have known someone. It is whether the last two answers you got were longer than the questions.

• What is something you are working on that nobody has noticed?
• When did you last surprise yourself?
• What do you miss that you would not want back?
• Who did you used to be close to, and what happened?
• What do you think people get wrong about you?

Answer these yourself first, every time. The escalation in the original study was mutual and turn-based, not one person interviewing another. If you ask a deepening question and then give a shallow answer to the same question, you have not built closeness, you have run an audit.

When it stalls, stop asking questions

The common failure is not running out of questions. It is that a stack of unrelated questions makes the other person feel processed. Three things work better than a new question.

Follow the last answer instead of moving on. Almost every answer contains a smaller thing you could have asked about, and asking about it signals you were listening rather than waiting. Someone mentions they moved here for a job they have since left, and there are four questions inside that sentence.

Say something instead of asking something. Conversation stalls when it becomes unidirectional. Offering a piece of yourself unprompted gives the other person somewhere to go that is not answering.

Let the pause happen. Roughly two seconds of quiet reads as a disaster to the person who asked and as thinking time to the person answering. Filling it usually costs you the better answer.

What to leave off the list

Anything where the honest answer is a diagnosis. Anything about money, weight or their family’s health, until they raise it. Rapid-fire either/or questions, which feel like a game show and produce nothing you can follow up. And the big abstract ones, the biggest regret and the deathbed question, which sound profound on a list and land as theatrical in a room.

One more: do not ask a question you would not answer. It is the cleanest test available and it removes about a third of most published lists.

Escalation is the mechanism. Matching people up, expecting to like each other and actively trying to get close all failed to move the needle in the original research. Taking turns answering slightly harder questions was the thing that worked.

Common questions

How many questions do I actually need?

Three you would genuinely answer yourself, one from each register. The 1997 study used 36 because it needed to fill 45 minutes under lab conditions, not because 36 is a magic number.

Do the 36 questions work on a first date?

They work at generating closeness, which is what they were built and tested for. Whether closeness after 45 minutes turns into anything is a separate question the paper does not answer.

Is it strange to use questions on purpose?

Only if you hide it. “I read about this thing, want to try it” is a normal sentence and it makes the whole exercise mutual, which is the condition the research actually tested.

What if the other person gives short answers to everything?

Give a long answer to your own question and see if it changes. If it does not after two attempts, they are not in the mood, and that is information rather than a failure.

Keep reading