How to be a better friend, according to the research

A 2023 systematic review gathered 38 studies of adult friendship and wellbeing published between 2000 and 2019 and sorted their findings by what they measured. Most of its variables are states: how many friends, how close, how much time together. But four of the things correlated with wellbeing are behaviours, the only part of a friendship one person can change on their own. They were how friends react when you share good news, whether they support your autonomy, whether either of you makes an effort to maintain the friendship, and perceived mattering, the sense that you count to a specific other person.

Two of those did more than correlate. The effort to maintain the friendship and perceived mattering were among six variables that sat in the chain between friendship and wellbeing, which is as close as this literature gets to saying that the upkeep is where the value is made. The variables that reached across years in the nine longitudinal studies were friendship quality and time actually spent together, and quality is what these behaviours add up to.

The limits are real. It is a review without a meta-analysis, so there is no pooled effect size; the studies measured wellbeing in inconsistent ways; most are snapshots; 28 of the 38 used young adults, mainly university students, so it says little about friendship at fifty; and a large share came from one research group. Nothing in it shows that behaving well as a friend causes anyone’s wellbeing. Happier people may simply make and keep better friends.

The second finding is the condition on the first. In the relationship literature, intimacy is defined operationally as feeling understood, validated and cared for, and the construct that predicts the behaviours which keep a relationship going is perceived partner responsiveness. The word doing the work is perceived. What moves a relationship is not what you did but what the other person registered.

A meta-analysis of relationship maintenance found the same thing from the other side: the correlation with relationship quality was stronger for perceptions of a partner’s maintenance than for the partner’s own reports of doing it. It is a framework resting on correlational work rather than a single experiment, and responsiveness and relationship quality almost certainly feed each other, so no direction is established. The practical implication survives the caveat. Effort that stays invisible, the thinking about someone, the meaning to call, does very little, because the other person never receives it.

One study of conversation points at how effort becomes visible. People who asked more questions, and follow-up questions in particular, were liked more, and the effect ran through perceived responsiveness. That study carries two caveats: one co-author has faced well-publicised data-integrity allegations in other work, and the effect has been contested and defended in print, so nothing here rests on it alone. The point it corroborates is simpler than the study. A follow-up question can only be produced by someone who took in what was just said, and the same is true of a detail remembered from last time.

Of the four behaviours, the reaction to good news is the most specific and the easiest to get wrong. The review found that how friends respond when you share something good was correlated with wellbeing. It did not test particular responses, so what follows is our reading of it alongside the conversation research, not a finding of the review itself.

The ordinary failure is not indifference. It is the thumbs-up, the "amazing!" and the change of subject, which registers the news without engaging with it. The conversation research suggests the engaged version is a second question: not "congratulations on the job" but "when do you start, and what did they say in the end?". The second question is the part that says the news mattered to you, because it cannot be asked without having listened to the first answer.

The same goes for saying something warm outright. In experiments where people wrote gratitude letters and predicted the reaction, writers underestimated how positive recipients would feel and overestimated how awkward it would be. Writers were grading their own wording; recipients were registering warmth. Those were one-off exchanges measured immediately, so there is no long-run data. The direction is consistent, though: the awkwardness you expect is mostly in the forecast.

Maintenance effort appears in the review as both a correlate and a mediator, and a separate 18-month study explains why it is the behaviour that cannot be skipped. Tracking closeness and contact across people’s personal networks, it found that emotional closeness to friends declined steadily once regular contact stopped, while closeness to family held up over the same gaps. It is observational, and closeness and contact almost certainly drive each other. But it means friendship is the tie that has to be maintained, and the bill comes due whether or not anyone intended to stop paying.

The trap is to turn maintenance into accounting. The research on communal and exchange relationships draws the line: keeping track of individual inputs is how exchange relationships work, and where closeness is wanted, prompt repayment of a benefit reduces attraction rather than increasing it. The studies were mostly laboratory and short-lived, and real relationships mix the two norms. The line still holds. Tracking what someone needs and what they told you is communal. Tracking who owes whom a message is exchange.

So the effort that counts is not a contact quota. It is the kind that carries content: knowing that her mother’s surgery is in March, that he is waiting to hear about the flat, that the dog is not allowed to run for six weeks. That is also the form of effort the other person can see, which, from the responsiveness research, is the only kind that does anything.

After you see or speak to a friend, write down one thing that is still open: the result they are waiting for, the decision they have not made, the thing they are dreading. Next time, ask about it first, and ask a second question about the answer. That one habit covers three of the four behaviours at once: it is maintenance effort, it is visible, and it tells them they count to you.

When they tell you good news, stay on it for two more questions before you move on. When you think something warm about a friend, say it in the plain words you would use, and treat the hesitation as the forecast bias it is. And when they are deciding something, support the choice they are making rather than steering it; support for autonomy was in the review alongside the rest.

None of this needs an app. A short note under each friend’s name with the last thing they told you, read before you see them, is the whole method. What does not work is a running tally of who messaged last, because that is the one kind of tracking the research says reduces attraction in close relationships.