It Doesn’t Hurt to Ask: Why Follow-Up Questions Make People Like You

By Noam Shemla · The Studies

TL;DR Across two experiments and a speed-dating study, people who asked more questions were liked more by the person they were talking to, and follow-up questions, the ones that build on what the other person just said, did most of the work. The askers never saw it coming: they did not expect to be liked any more for asking.

• Paper: Huang, K., Yeomans, M., Brooks, A. W., Minson, J., & Gino, F. (2017). It doesn’t hurt to ask: Question-asking increases liking. Journal of Personality and Social Psychology, 113(3), 430–452.
• Design: Two randomised experiments (Studies 1 and 2), an observer study reading their transcripts, and an observational study of real speed dates (Study 3).
• Sample: 398 lab participants in 199 pairs; 338 online participants in 169 pairs; 612 observers; 110 speed daters in 1,961 date observations.
• Manipulation: One or both people in a 15-minute text chat were told to ask either “at least nine” or “at most four” questions.
• Outcome: How much the partner liked the asker (four items, 1–7); in the speed dates, whether the partner wanted a second date.
• Method: Mean comparisons and multilevel models; mediation through rated responsiveness; a machine-learning classifier to detect follow-up questions in the speed-date transcripts.
• Headline: Asking more questions, and especially more follow-up questions, raised liking, and people consistently failed to predict it.

There is an old piece of advice about being good company: be interested rather than interesting. It sounds like manners. This paper, from a team at Harvard, set out to test whether it is also true, and if so, why.1

The design is unusually clean for conversation research. Instead of watching people talk and noticing that the likeable ones ask a lot of questions, which could simply mean that likeable people do everything well, the researchers told some people to ask more questions and others to ask fewer, and measured what happened to how much they were liked.

Study 1: nine questions or four

The first experiment took place in a university behavioural lab. Participants sat in separate cubicles and chatted by instant message with a stranger for 15 minutes, with one instruction: get to know each other. In each pair, one person was secretly given an extra rule. Half of them were told to ask at least nine questions, the other half at most four. Those two numbers were the 25th and 75th percentiles of question-asking in a pilot run in the same lab, so both were well within the range of a normal conversation.

The instruction worked. The "many questions" group asked 10.2 questions on average and the "few questions" group 4.3. Then the partners, who knew nothing about the instruction, rated how much they liked the person they had chatted with.

The difference, a third of a standard deviation, is a real but moderate effect. What made the study interesting was the other side of it. The askers were also asked how much they thought their partner liked them, and there the two groups were identical (5.27 and 5.19). People who asked more questions were liked more, and had no idea.

Study 2: when both people are told what to do

Study 1 left a loophole: the partner who received the questions was free to ask as many as they liked in return. Maybe what people enjoy is a conversation that is balanced, or one that is lopsided in their favour. So in Study 2 the team ran the chat online with 169 new pairs and gave both people an instruction, at random and independently. Some pairs were two heavy askers, some two light ones, and some one of each.

The result replicated, and the loophole closed. People liked partners who asked many questions (6.02 against 5.79), and that held regardless of how many questions they themselves had been told to ask. It is not balance people respond to. It is being asked.

The team then showed the transcripts of those conversations to 612 outside readers and asked them to rate both people. The readers did not like the heavy askers any more. If anything they preferred the people who had been asked a lot of questions, the ones who had done most of the talking. Question-asking, in other words, is not a trait that makes you look likeable from the outside. It works on the person being asked.

Why it works: responsiveness

The authors' explanation is a construct from relationship research called responsiveness: the sense that the other person has listened to you, understood you, and cares about what you said. Coders who read the chats blind to condition rated the heavy askers as more responsive, and a mediation analysis found that responsiveness carried part of the effect of question-asking on liking in both experiments (a standardised indirect effect of .12 in Study 1 and .07 in Study 2).

Not all questions are equal

If responsiveness is the mechanism, some questions should do more than others. So the team had three coders sort every question in 368 of the chats into types. Their scheme had six, of which four were common enough to analyse:

What kind of questions people asked, and which kind was linked to being liked

The pattern is sharp. Follow-up questions were the most common kind and the only kind that predicted liking. Switch questions, the ones that abandon what the other person said for a new topic, went with lower rated responsiveness. And when people were told to ask more questions, they did not just ask more of everything: the share that were follow-ups rose from 28% to 42%, and that shift itself carried part of the effect.

A follow-up question is hard to fake. You cannot ask one without having listened. That is exactly why it works.

Study 3: speed dates

The first two studies were text chats between strangers with no stake in the outcome. For the third, the team went to the same recorded speed-dating evenings that McFarland, Jurafsky and Rawlings used for their study of what makes two people click: 110 graduate students, each on 15 to 19 four-minute dates.2

There were far too many questions to code by hand, so the team trained a classifier on the 4,209 question turns their coders had already labelled in Studies 1 and 2, and let it label the speed dates. It identified follow-up questions with 87% accuracy on held-out examples.

• 9.8 Questions per date. per person, about 22% of their turns
• 4.5 Of them follow-ups. the most common type here too
• 57% Men who said yes. to a second date
• 37% Women who said yes. to a second date

Questions were everywhere in four-minute dates, and women were much more selective than men, which is why every model controls for gender.

The outcome was blunt and behavioural: did the partner tick "yes, I'd like to see them again"? The authors ran logistic regressions of that yes on each person's question-asking, with standard errors corrected for the fact that the same people appear in many dates.

In the authors' own translation, a difference of about 8% in someone's follow-up question rate was associated with roughly one extra second-date yes over an evening of 20 dates.

That fits with the McFarland study rather than contradicting it. McFarland and colleagues found that questions went with less clicking in the same corpus, and when they read the transcripts, the questions in the flat dates were the "where are you from?" kind, used to keep a stalling conversation alive. Those are switch questions. The questions this paper finds helpful are the other kind.

Has it held up?

Better than most, with two things worth knowing.

The speed-dating analysis was challenged. In 2019 Avraham Kluger and Thomas Malloy reanalysed Study 3 with a social relations model, a method built for data where everyone meets everyone. They found that only about a third of the variation in question-asking was a stable trait of the asker, a similar share belonged to the particular pair, and that the latent scores for follow-up and switch questions were hard to tell apart. They also argued that a second-date yes is not the same thing as liking.3 The original authors replied that the reanalysis tested the wrong variables, made modelling choices that do not suit this data set, and left aside results in all three studies, including the two experiments, that answer its central points.4 The paper's own supplementary models point the same careful way: once each asker's general desirability is controlled for, the follow-up effect in the speed dates shrinks and is no longer significant.

The paper was audited. In 2025 the journal published a correction after an independent audit of the data and analysis code. The audit found minor errors in how some results were reported (the kind where a reported 5.79 reproduces as 5.78), confirmed every conclusion, and found no reason to doubt the integrity of the data.5 The numbers in this article are the published ones.

So the solid part is the experimental one: in two randomised studies, asking more questions caused people to be liked more. The speed-dating result is consistent with it but, like any observational result, cannot carry that claim on its own.

What this study cannot tell you

• The experiments were text chats. Fifteen minutes of instant messaging between strangers is not a dinner, an interview or a first date, and tone of voice played no part in it.
• More is not always better. The instruction was nine questions in fifteen minutes. The paper does not test what happens at twenty, and the observers' preference for the people who got to talk is a reminder that a conversation needs answers too.
• Liking is a rating, not a relationship. Nobody was followed up to see whether the heavy askers made more friends.
• The speed-date link is correlational and contested, as above.

What a speaker can take from it

The practical rule is short. Ask about what the other person just said. A follow-up question is proof of listening, and the study suggests people feel that proof and credit you for it, while you underestimate how much it matters.

It also explains a quiet mistake many of us make in conversations that matter: we spend the other person's turn preparing our next point. That is the opposite of the behaviour that made people liked here. Asking one more question about their answer costs nothing and, on this evidence, is what gets remembered.

This article summarises a published paper. Every number above is the authors’ own, from the paper’s text, Figures 1 and 2, and Tables 3 and 4, except the confidence intervals in the forest plot, which we computed from the standard errors the paper reports.

Measured

• Numbers of questions, counted from chat and speed-date transcripts
• Question types, coded by three people per transcript, then by a classifier in the speed dates
• Liking, rated by the partner on four items after each conversation

Inferred, not measured

• Responsiveness as the mechanism: supported by mediation analyses, which are correlational steps inside an experiment
• The speed-date link between follow-up questions and second dates is observational

References

1. Huang, K., Yeomans, M., Brooks, A. W., Minson, J., & Gino, F. (2017). It doesn’t hurt to ask: question-asking increases liking. Journal of Personality and Social Psychology, 113(3), 430–452. https://doi.org/10.1037/pspi0000097
2. McFarland, D. A., Jurafsky, D., & Rawlings, C. (2013). Making the connection: social bonding in courtship situations. American Journal of Sociology, 118(6), 1596–1649. https://doi.org/10.1086/670240
3. Kluger, A. N., & Malloy, T. E. (2019). Question asking as a dyadic behavior. Journal of Personality and Social Psychology, 117(6), 1127–1138. https://doi.org/10.1037/pspi0000156
4. Yeomans, M., Brooks, A. W., Huang, K., Minson, J., & Gino, F. (2019). It helps to ask: the cumulative benefits of asking follow-up questions. Journal of Personality and Social Psychology, 117(6), 1139–1144. https://doi.org/10.1037/pspi0000220
5. American Psychological Association (2025). Correction to “It doesn’t hurt to ask: question-asking increases liking” by Huang et al. (2017). Journal of Personality and Social Psychology, 128(3), 669. https://doi.org/10.1037/pspi0000491