What Makes a Good Friend: The Research and the Honest Truth

Jeffrey Hall's 2012 study, published in the Journal of Social and Personal Relationships, asked thousands of adults what they expected from a good friend, then statistically grouped the responses into six clean categories: symmetric reciprocity (give-and-take feels balanced), agency (friends help you accomplish your goals), enjoyment (you genuinely like being around each other), instrumental aid (practical help when needed), similarity (shared values or interests), and communion (deep emotional connection and disclosure).

What the research found is that most adults rank reciprocity and enjoyment as non-negotiable, while the other four vary by individual. Some people prioritize practical help. Some prioritize intellectual similarity. Some prioritize emotional depth. There isn't a universal 'good friend' template — there's a checklist of six dimensions, and each person weights them differently. Knowing which two or three matter most to you is more useful than trying to be 'a good friend' in some abstract sense.

Most internet advice about friendship focuses on traits — loyal, honest, fun, supportive — that sound right and are almost impossible to falsify. Of course you should be loyal. Of course you should be honest. Saying so doesn't change anything.

The research is more useful because it focuses on behaviors. A loyal friend is one who shows up; a behavior, not a trait. An honest friend is one who tells you the hard thing you needed to hear at the moment you needed to hear it; a behavior, not a trait. Friendship advice that operates at the level of traits ('be a better listener') tends to be useless because it gives no specific action. Advice at the level of behaviors ('after a friend tells you about a hard thing, follow up forty-eight hours later asking how it's going') tends to actually change something. This page is going to stay at the behavior level wherever possible.

Across hundreds of conversations about adult friendship, one quality comes up again and again as the thing that separates real friendships from politely-fading ones. Good friends don't take cancellations personally. Adult life is full of weeks where someone genuinely cannot show up — a sick kid, a work crisis, a depression episode they can't name yet. The friend who responds to a canceled plan with 'okay, when's good for you instead?' is doing more for the friendship than the friend who says nothing but quietly logs it. Doing more, mathematically, because the second person is on the path to slowly disappearing.

The asymmetry here is significant. Most adults have at least one friend they used to be close to but who they slowly stopped reaching out to — not because anything bad happened, but because two or three canceled plans accumulated into 'they don't want to.' That conclusion is almost always wrong. The friend was overwhelmed, and you both quietly let the friendship lapse. Good friends interrupt that pattern, gently, until the other person is back in life enough to reciprocate.

There's a kind of cultural noise around what a good friend looks like that suggests constant availability — instant text replies, always saying yes to plans, never canceling. By that definition, almost no adult is a good friend, because almost no adult has that capacity past 30.

The more accurate picture is asymmetric. A good friend is not always available; a good friend is reliably reachable when it matters. Sometimes being a good friend means not responding to a check-in for three days because your own life is on fire, and then sending an honest update when you resurface. Sometimes it means saying no to a casual hang because you'd be drained and resentful if you went, and yes to the harder hospital-visit ask. Friendship that requires constant availability isn't friendship — it's surveillance with extra steps.

Robin Dunbar's research on layered friendship circles maps onto this: the innermost layer of 3-5 people earns the most consistent attention; the 15-person 'good friends' layer gets less; the 50-person friend layer gets less still. Being a good friend at each layer is a different behavior, and matching the layer to the appropriate level of contact is most of the skill.

If you want to actually become someone whose friends would describe as a good friend, here are five behaviors that compound across years:

1. Remember details. Write them down if you have to. Ask about the thing two weeks later. The friend who follows up on a job interview, a doctor's appointment, a hard conversation with a parent — that friend is functionally rare.

2. Send the first text. Most adults are waiting for the other person to reach out. The person who consistently sends the first 'thinking of you' message is the person who quietly holds friendships together.

3. Initiate one specific plan a month with each close friend. Not a vague 'we should hang.' A specific time, a specific place, a specific reason. Jeffrey Hall's 50-hour rule depends on showing up to the same person multiple times, and specifics are how that happens.

4. Tell the truth when the friend is about to do something you think is a mistake. Gently. Once. Then drop it and support them anyway. Friends who only validate stop being useful as friends; friends who only criticize stop being friends.

5. Show up for the unglamorous stuff. The funeral, the move, the hospital, the divorce paperwork sorting. The grand gestures don't bond people the way the unglamorous show-up does.

Most friendships that don't work out aren't dramatic — they just slowly run out of returns. But there are a few specific patterns that, when you notice them repeating, deserve a hard look:

• The friendship is consistently asymmetric — you're always the one initiating, organizing, remembering. After a year of this not changing, it's not going to change.
• The friend disappears when your life gets hard but reappears when it gets good. Some people are fair-weather without realizing it; the test is whether they show up the third time you're struggling.
• Sharing something vulnerable with them comes back to you, slightly distorted, from someone else. Once might be an accident. Twice is a pattern.
• Their value of you depends on what you do for them. If your usefulness changes — you move, you have a baby, you become less available — and the friendship suddenly cools, the friendship was about your usefulness, not you.

Noticing these patterns isn't about being judgmental. It's about being realistic about which of your relationships are likely to be load-bearing over decades and which are pleasant but circumstantial.

There's a quiet asymmetry in self-reporting: most adults rate themselves as good friends, and rate their friends slightly more critically. This is the same psychological tic that makes most people rate themselves as above-average drivers. The good news is that the gap is mostly fixable by adopting a single habit: ask your three closest friends, every year or so, what would make you a better friend to them. Not in a heavy way. As a casual 'hey, what's one thing I could do more or less of?' question on a walk.

The answers, when people give honest ones, tend to be small and specific: 'reply to my texts a little faster, even just to acknowledge,' or 'remember to ask about my mom when we see each other.' Acting on those small specific asks is what most distinguishes the friend everyone says they have from the friend everyone actually keeps.

Most adults who look honestly at their own friendships notice at least one or two relationships where they've been the slower texter, the canceler, the one who forgot the birthday. This is, in some sense, just being a human in a busy life. But it's also fixable, and the repair is usually small.

The move is rarely a grand apology. The move is usually one specific, dated message — 'I noticed I've been bad about staying in touch this year. Want to do dinner the week of the 15th? I want to hear how the new job is going.' That kind of message restarts more friendships than most adults realize. The friend on the other end has usually been quietly hoping you would write something exactly like that.