There is a man in my phone I have not spoken to since 2019. The last message in the thread is his and it says "sounds good, let me know". I never let him know.
There was no reason. There is still no reason. We had no argument, nothing was said, and I would be pleased to hear from him tomorrow. The entire distance between us is that neither of us sent the next message, several hundred times in a row.
That is the version of this I find genuinely strange, and it turns out to be the common one.
The ending that never happened
Almost everybody is carrying one of these. In the first of seven studies published in Communications Psychology in April 2024, Lara Aknin and Gillian Sandstrom asked 401 people whether they had lost touch with a friend they still cared about. 90.9 per cent said yes.
Not fallen out with. Not stopped liking. Lost touch with, while still caring.
One thing to say plainly before I go further, because it matters for how much weight the rest of this carries. Those samples ran between 46 and 63 per cent women, and the paper reports no separate breakdown for men. I am not a psychologist and I have no standing to reinterpret their data by sex. So this is a finding about people, and men are inside it.
What the survey work does break out by sex is the drift itself. The Survey Center on American Life's 2021 report on American friendship, based on a survey of 2,019 adults, found that among men aged 50 to 64, 40 per cent had lost touch with a few friends in the previous twelve months and 7 per cent had lost touch with most of them. Just over half, 54 per cent, had stayed in regular contact with their friends.
Twelve months. That is the ordinary rate of quiet attrition, and none of it involves anything going wrong. I have written elsewhere about the structures that close in your forties and take the friendships resting on them with them.
What is actually in the way
Here is the number that stopped me. In Aknin and Sandstrom's third study, 453 people were asked to think of an old friend they had lost touch with, confirmed they had the contact details, said they wanted to get in touch, and were given time in the session to write the message.
27.8 per cent sent it. A separate study using the same design got 36.8 per cent.
Desire was there. Access was there. Everything that normally gets treated as the obstacle had already been removed, and roughly two-thirds of people still did not press send.
When the researchers asked why, the barriers ranked in a specific order. Top was the worry that the friend might not want to hear from them, at 5.2 out of 7. Then the expected awkwardness after so much time, at 4.9. Then, at 4.8, having no legitimate reason to make contact.
The comparison that makes the size of it clear: participants were asked how willing they would be to message an old friend, and how willing they would be to strike up a conversation with a stranger. The two came out statistically identical, both at 4.6 out of 7, with a Bayes factor of 19.52 in favour of there being no real difference. That is strong evidence for a genuine null rather than a study that failed to find something.
Twenty years of shared history buys you nothing over a stranger at a bus stop. For what it is worth, both scored above picking up litter in a park, which was also on the list.
Nobody thinks they have a good enough reason
The barrier I recognised immediately was the third one.
Asked what would count as a legitimate reason to contact an old friend, participants rated only one thing as sufficient on its own: a birthday. Everything else needed justifying. Seeing something that reminded you of them, wondering how they were, being in their city, all of it felt to people like it needed a better cover story than it has.
And the asymmetry underneath that is the part I have not been able to stop thinking about. Asked how willing they would be to hear from an old friend, people scored 4.9 out of 7. Asked how willing they would be to send the message, 3.5. That is a gap of d=0.70, which is close to large.
Which probably means the man in my phone is sitting on the other side of the same arithmetic, waiting for a birthday. It is a group mean, so it says what is likely rather than what he is thinking.
You both already know it is welcome
The second body of work here is Liu, Rim, Min and Min in the Journal of Personality and Social Psychology in 2023: thirteen preregistered studies on how much people underestimate the appreciation their message will get.
They do underestimate it, at d=.29, and the gap is widest for weak ties. A dormant friendship is precisely a weak tie. This is the case where senders are most wrong about the reception.
The honest framing of that finding is narrower than the way it usually gets quoted, though. In their data, nearly everybody in both roles rated the appreciation above the midpoint of the scale. Senders already expected to be welcome. They underrated the size of it by a modest margin. That is a smaller and more believable claim, and it is closer to how this feels from inside: I think he would be pleased to hear from me. I think it would be fine. I have still not sent it.
Isolated and lonely are two different measurements
The Australian data draws that distinction more sharply than the slogans do. The Australian Institute of Health and Welfare, working from HILDA data covering more than 17,000 Australians, reports that men are more socially isolated than women, 18 per cent against 12 per cent, while being no lonelier, 15 per cent against 16 per cent.
The gap is not in how men feel. It is in how many people they actually see.
A dormant friendship sits exactly in that gap. It is a contact that has stopped happening in a life that reports itself as fine, which is why it can run for years without anyone raising it.
Being told it will be fine does not change what you do
This is where most articles on this subject end, and where both research groups found the disappointing result.
Aknin and Sandstrom tried three attitude interventions: getting people to reflect on the value of old friendships, reassuring them the message would be welcome, and encouraging them directly. None of them significantly increased the number of messages sent. The tests came back F(2,450)=1.72, p=0.18, and F(2,601)=1.93, p=0.15.
Liu and colleagues ran the same idea from their own angle, telling people about the underestimation effect to see whether knowing about it changed behaviour. It produced no significant increase either.
Two independent labs, both preregistered, both finding that the standard closing paragraph of every article like this one does nothing. Being told it will go well is apparently something a person can agree with entirely and still not act on.
Send an easy one first
The one manipulation that moved the number was practice.
In Aknin and Sandstrom's design, some participants were first asked to send a message to a current friend, someone they were still in regular contact with, before being given the chance to message the dormant one. In that condition 53 per cent sent to the old friend, against 31 per cent in the control. The effect size was d=-0.46.
Nobody reassured those participants about anything. They had simply written one message already that day.
So the practical version, which is what I would actually do this week:
Send a message to somebody you are already in touch with. Your brother, the friend you saw a fortnight ago, someone where the message costs you nothing at all. Anything will do.
Then, in the same sitting, send the dormant one. Short, no apology for the length of the gap, no explanation of it, no manufactured reason:
Saw [the specific thing] and thought of you. How are things?
Then expect a share of them to go nowhere, because that is the ordinary rate and men who do not know it in advance tend to conclude after two dead ends that the problem is them.
That is what the practice condition shows: the message you send second is much likelier to exist than the one you try to send first. Nobody compared order against wording, so I am describing a sequence that worked in one experiment rather than a ranking of the two. If you want the fuller version of where dormant contacts sit in the sequence of things worth trying, I have set that out here.