PHILIA · FRIENDSHIP
On what a relationship actually rests on — and on how narrow Aristotle thought the best kind was available.
You leave a job, or a team, or a city, or a role. The last day is warm. People say the things people say, and several of them mean it.
For about three weeks the messages continue at something like the old rate. Then they thin. Somebody suggests a drink and it does not get arranged, and then it does not get re-suggested. By month four the contact is effectively zero with a number of people you would have described, without hesitating, as friends.
Nothing happened. There was no argument, nobody did anything, and if you met any of them tomorrow it would be warm and easy and you would both mean that too. It simply stopped, and both of you are faintly puzzled by it in a way neither will mention.
Almost everybody takes this personally, at least a little. It is almost always structural. Something was holding those relationships that stopped existing on the day you left, and it was not anybody’s affection.
Aristotle gave friendship two of the ten books of the Ethics — more space than courage, justice or wisdom, which tells you what he thought it was worth — and he explains this in the first few pages of it.
The sort is by basis. Some friendships are held by usefulness, some by pleasure, and some — the third kind — by something about who the other person is.
The first two dissolve when the thing holding them does. Not because anybody was insincere, and not as a failure: the relationship was built on something, and the something stopped being there.
The mistake that produces the puzzlement in month four is a misclassification rather than an ending. Daily contact and shared difficulty feel like the deepest bond available, and quite often they are a utility friendship that is genuinely warm. Nothing is wrong with that. Aristotle is explicit that the first two kinds are good, are most of a life, and are not failed versions of the third — and treating them as counterfeit is a reliable way to end up lonely among people who like you.
The third kind, Aristotle says, requires time, and he uses an expression about having eaten a great deal of salt together: many meals, over years.
He is explicit that it cannot be rushed, and that people who declare it quickly have not got it however strongly they feel. What they have is the wish for it, which arises fast and is a different thing.
He also says you cannot have many. Not as a lament — as a structural consequence of a relationship that requires that much time from a person who has a finite amount.
Research on personal networks finds that people maintain a limited number of close ties, that the layers are consistent in size across populations, and that ties decay without contact at a fairly predictable rate.4
The decay finding is the one that bears on this essay directly: relationships maintained by proximity rather than by deliberate contact fall off quickly once the proximity goes. That is the month-four phenomenon, measured.
It is worth saying what this does not show. It does not vindicate the three-kind taxonomy, which no study has tested and which is cleaner than reality in any case. It supports one narrower claim — that contact sustained by circumstance ends with the circumstance — and that claim is enough for everything practical here.
The general evidence on relationships and wellbeing is separately among the sturdiest in psychology: social connection predicts health and longevity at effect sizes comparable to well-established risk factors.5 That is a reason to treat the classification as worth an hour rather than as idle sociology.
The account is narrower than it looks, in a way that matters.
The case against
The third kind requires both parties to be good, and “good” in his sense requires leisure, education and citizenship. Women, enslaved people and labourers are excluded structurally rather than incidentally — he says as much about friendship between unequals, and about what an enslaved person can be a friend as.
That is not a detail to be tactfully omitted. It is the frame the whole taxonomy sits in, and a reader who takes the three kinds without it inherits an aristocratic picture of who is capable of the best thing in life.
The second objection is smaller: the three-way sort is cleaner than reality. Most relationships are mixtures, the mixture changes, and a friendship that began in utility can acquire the other thing. The sort is a diagnostic question, not a set of boxes.
For each significant relationship, ask what it rests on. Shared circumstance, shared enjoyment, or something about who the person is.
Do it privately and without ranking anybody. The point is prediction, not a hierarchy of your friends: relationships built on a circumstance end with the circumstance, and knowing that in advance is the entire value of the exercise.
Then decide which ones you want to survive a change, and give those a basis that is not the situation — contact that is deliberate rather than incidental. The decay research is why this has to be arranged rather than assumed.
And when somebody leaves and the contact starts thinning, say the structural thing out loud. It lands far better than the silence does, and it has the advantage of being true.
Nobody sees this, which is what makes it honest.
Month four is not a verdict on anybody’s character, including yours. It is what happens to a relationship whose basis has been removed, and it would have happened whatever either of you felt.
That is the whole use of the sort: it tells you in advance which ones will do that, which turns a small recurring grief into a piece of information.
And the instruction is short. Decide which ones you want to survive, and give them something to stand on that is not the situation. Nothing else in this collection has better evidence behind it.
Relationships built on shared circumstance end when the circumstance does — not because anyone was insincere, but because the thing holding them stopped being there. Knowing that in advance removes most of the sting.
Aristotle thought the best kind was available only to leisured male citizens. That exclusion is the frame the taxonomy sits in, not a detail to omit.
Quotations are from the Nicomachean Ethics unless noted, cited by book, chapter and Bekker number; renderings follow W. D. Ross (rev. Urmson), Terence Irwin or Roger Crisp, and a condensed line is marked “after Aristotle”. Modern research is cited by author and publication.