A bloke I know through my son's cricket told me in winter that he had heaps of mates. He said it in passing, the way you would confirm you had enough milk in the fridge. Twenty minutes later, in the same conversation, he mentioned he had spent six weeks waiting on a scan result and had told nobody except his wife, and her only because she drove him to the appointment.

Both statements are true about the same man on the same afternoon. Neither is a lie. They are answers to two different questions, and nobody had ever put the second one to him.

I have largely stopped asking men how many friends they have. Once you see what happens when researchers ask it, the headcount looks less like a measurement and more like a wording choice.

Ask it seven ways, get seven answers

The most useful paper I have read on this is a methodology paper about a survey going wrong. Claude Fischer at Berkeley published The 2004 GSS Finding of Shrunken Social Networks: An Artifact? in the American Sociological Review in 2009, after the US General Social Survey reported that Americans had lost about a third of their confidants in twenty years. He went back through the instrument rather than the headline.

Part of what he documents is what one population says when the wording moves. Four questions, four answers on the share reporting nobody:

  • 1986, asked about close friends: 5 per cent said none.
  • 1998, asked whether they had any good friends they felt close to: 9 per cent.
  • 2004, asked to name the people with whom they discussed important matters: 25 per cent.
  • 2002, asked in a two-part sequence separating staying in contact from discussing a serious problem: 5 per cent.

Same country, same survey programme, same stretch of years. Five per cent to twenty-five, depending on how the question was built.

Current surveys behave the same way. Pew Research Center asked 5,057 US adults in July 2023 and found 8 per cent had no close friends. Pew asked 6,204 adults again and reported in January 2025 that 81 per cent had at least one close friend, putting the other side at about 19 per cent. Same organisation, same panel, different wording, more than double the figure from eighteen months earlier.

The figure I use most often, 15 per cent of American men with no close friends, comes from the Survey Center on American Life's 2021 work and needs a caveat it almost never travels with. That reading was an online probability panel. The 1990 comparison everyone pairs it with was a Gallup telephone poll with a live interviewer, and the report's own endnote concedes that men speaking to a person may have been less willing to say they had nobody. Any multiplier drawn across the two, fivefold included, carries an instrument change dressed as a social one. The men-specific figures also sit in a commentary piece rather than the main report, and that commentary carries no caveat at all. I have written elsewhere about what the 15 per cent does and does not support.

Australia's own reading is 4 per cent. The Australian Institute of Family Studies asked 10,729 adult men in the second wave of Ten to Men, in 2015 and 2016, and 4 per cent reported none, 5 per cent among men aged 45 to 60. That question counts close friends and relatives together, so it measures a different quantity from the American figures and should never be set beside them.

Nothing about the men changed between any of those readings. The instrument did.

Ten people, and nobody to tell

Here is the finding that made me put the headcount down for good.

Fischer looked at the 2004 respondents who had named zero people they discussed important matters with, the group that produced the alarming headline. Later in the same interview, 80 per cent of them said they had at least three people they kept in regular contact with. The median for that group was ten.

Ten people. Nobody to tell.

A man can produce a headcount of ten and a confidant count of zero in one sitting without contradicting himself once. It is an accurate description of a real position a lot of men are in, and it is invisible to any question beginning with "how many".

Somebody answered six hundred

The same Australian survey asked men to count their close friends and relatives. The answers ranged from 0 to 600.

I keep coming back to that. Someone sat down with a government survey and wrote six hundred, probably in good faith. He was answering a different question from the man who wrote three, and the form could not tell them apart.

The institute is careful here in a way popular writing never is. Its report calls the measure aloneness rather than loneliness, and says aloneness indicates an absence of social contacts and is not necessarily problematic or unwanted, while loneliness is a lack of desired relationships. A government research body stating in its own methodology that the count is not the thing.

Half the names would not name you back

The count is also one-sided.

Abdullah Almaatouq and colleagues published a study in PLOS ONE in March 2016 in which participants named their friends and rated how likely each was to name them back. People expected reciprocation 94 per cent of the time. Actual reciprocation ran at 53 per cent, 413 of 775 nominations.

The caveat matters. That study covered 84 undergraduates in one class, exactly the group you would expect to over-read a friendly acquaintance. Nobody has run it on 48-year-old men, which is a shame, because the direction of the error is the one that would hurt.

The gap is in use, not stock

If the headcount were the thing, the sexes would sort neatly, and they do not.

The Australian Institute of Health and Welfare's summary of HILDA data reports men more socially isolated than women, 18 per cent to 12, while loneliness runs 15 per cent for men and 16 for women. Men have fewer contacts and are no lonelier for it, at least not in any way that instrument detects.

Pew's January 2025 work lands in the same place. Men and women are about equally likely to have at least one close friend; what differs is that men contact theirs less often. For a lot of men reading this the people are already there, and the habit of using them has quietly stopped running. That is a description of the men I hear from rather than a measured share.

Better questions than "how many"

Each of these maps onto a variable somebody has actually measured, which is why I trust them more than a number.

Who did you talk to about something that mattered in the last six months? This is my own wording rather than a survey's, built from the shape the GSS uses. The time limit and the subject test are what narrow it. The time limit and the subject test are what narrow it.

Who would you discuss a serious personal problem with, as opposed to just stay in contact with? The GSS two-part sequence from 2002. That contrast is the whole instrument, and it is the distinction nobody had ever offered my cricket acquaintance.

Which of them would name you back? Almaatouq's question, undergraduates or not. Ask it of each name and watch which ones you hesitate over.

When did you last initiate? This one is mine. Pew measured how often people contact their friends, not who started it, so the initiation question is my extension of their finding rather than their finding. More than a year and you have a name on a list rather than a friendship in use.

If the tiers underneath this are the sticking point, I have written about what separates an acquaintance from a friend.

What I did with it

I worked through those four questions on paper. It took twenty minutes and was more uncomfortable than an exercise involving no other people has any right to be.

I am not going to publish what came out, because a number from one man's kitchen table is evidence of nothing. What I will say is that the four lists did not agree, and the disagreement was the useful part.

So the honest answer to "how many mates have you got" is a question about definitions. The men who say heaps are giving a straight answer to the easy version, because the easy version is the only one anyone asks. Write the names under each of the four questions and see whether the columns match. Then contact whoever turns up in the first and not the last.

Sources