Social isolation carries a 32 per cent increased risk of death. Friendships loosen within sixty days of silence. Nine in ten of the relationships that connect you to a different world are gone inside a year — and none of it feels like anything while it happens.
Nobody wakes up and decides to abandon a friend. It happens in the space between intentions — a reply not sent, a month that becomes three, a name you still think warmly of and have not spoken to since spring. Relationships do not end. They thin out until there is nothing holding them.
That sounds like sentiment. It is measurable, and the measurements are worse than most people expect.
That figure puts isolation in the same conversation as the risk factors medicine takes seriously and treats aggressively. It is not a metaphor about modern life. It is an effect size, drawn from ninety studies, and it survives adjustment.
The reason it lands so hard is that the thing causing it is invisible. There is no moment of injury. Nobody presents at a clinic complaining of a friendship that has quietly stopped.
Contact is what holds a relationship in place. Remove it and decay begins on a schedule far shorter than intuition suggests — sixty to ninety days before a friendship measurably loosens. Not ends. Loosens: the reply gets slower, the shared context goes stale, the reasons to reach out stop arriving on their own.
Set against that clock, two other numbers do real damage:
| Finding | Figure |
|---|---|
| Stable relationships a person can cognitively maintain (Dunbar's number) | ~150 |
| Share of social time spent on the innermost circle | 2/3 on ~15 people |
| Bridge relationships that disappear within a year | 9 in 10 |
Two-thirds of your social effort goes to fifteen people. Whatever attention remains has to cover the other hundred and thirty-five, and it does not stretch. The arithmetic is not a character flaw. It is a budget, and the budget does not balance.
Nine out of ten bridge relationships disappear within a year. A bridge relationship is the one that connects you to a world you are not already in — a different industry, country, class, generation. In network terms these are the weak ties that carry new information, and they are precisely the ties with no structural support: no shared workplace, no mutual friends, no standing reason to collide again.
So the relationships that decay fastest are the ones doing the most work. Your innermost fifteen survive on gravity. The bridge to someone unlike you survives on deliberate effort, and deliberate effort is exactly what nobody budgets.
An international student is a person whose entire relational network has been severed at once and who must rebuild it in a second language, inside a culture whose friendship conventions they are still learning. Every relationship they form is a bridge relationship. Every one of them is subject to the nine-in-ten rule.
Then they go home, and the severing happens a second time — this time to the relationships they spent years building. That is the same population, and the same mechanism, as the returnee attrition covered in the missions review. A figure the field explains with theology may be substantially explained by the ordinary physics of contact.
If decay runs on a clock, it can be anticipated. The interesting design question is not how to make people care more — they already care — but how to make the invisible visible before the sixtieth day, when recovery is still cheap.
Three things follow:
The two anchor findings — the 32% mortality figure and the two-thirds loneliness figure — are peer-reviewed and cited above. The 60–90 day decay window, the nine-in-ten bridge attrition, and the two-thirds-on-fifteen allocation are weaker: they circulate widely in the relationship-science literature but this review did not trace each to a primary study. Treat them as directional, not precise, and do not put them on a slide without sourcing them first.
Dunbar's 150 is contested in the literature. Later replications have argued the cognitive limit is softer and more variable than the original figure implies. It is useful as an order of magnitude and misleading as a precise threshold.