There is a particular flavour of panic I know well. A man is crossing a room towards me, arms already opening, and I know everything about him. I know he keeps koi. I know he has a bad knee and worse taste in shirts. I know we once spent forty minutes arguing about whether Bangkok traffic is actually improving or has simply learnt to look calmer.

What I do not know is his name.

For most of my life I filed this under personal defect. Bad with names, I would say, in the same tone people use for bad at maths, as though it were a settled fact about my character. The truth is duller and more forgiving. A face and a name look like two halves of one memory, but they are stored on completely different terms, and only one of them was ever set up to be found again.

A name is a word with nothing attached to it

Try this. You meet a man and someone mentions he is a baker. Then you meet another man and someone mentions that his surname is Baker. Same word. Same syllables. Same effort on your part. A week later, which one comes back?

The baker, every time, and by an embarrassing margin. Psychologists call this the Baker/baker paradox, and it was demonstrated cleanly in 1987 when McWeeny, Young, Hay and Ellis taught volunteers surnames and occupations for unfamiliar faces, using some words that could do either job. Names stayed much harder to recall than occupations, even when the word itself was identical. Nothing changed except the label it had been given.

The reason is meaning. Baker the occupation drags a whole crowd in with it: flour, ovens, four in the morning, the smell of a bakery two streets away. Every one of those is a rope back to the fact. If one snaps, you climb the next. Baker the surname turns up alone with no luggage. Gillian Cohen, the psychologist who gave the effect its name, made the point that proper names are essentially meaningless in a way ordinary words are not. A surname tells you almost nothing about the person carrying it. Most people named Baker have never made bread in their lives.

So the blank you draw has a mechanical cause. Of everything in that encounter, one item alone arrived attached to nothing, and it is the item you now need.

The name is at the back of the queue

This is the bit that made me stop apologising.

The standard model of how we recognise people, built by Vicki Bruce and Andrew Young in the mid-eighties and still doing solid work, describes a sequence rather than a single lookup. A face arrives. It gets matched against stored faces. That match opens up everything you know about the person, their job, where you met, the koi, the knee. And only after all of that does the system reach for the name. Neuroimaging work on how proper names are pulled out of the anterior temporal lobes keeps pointing the same way: the name sits downstream of everything else, on a separate and much narrower road.

Which explains the specific texture of the failure. There is nothing vague about your sense of this person. You are certain. You could describe his wedding. The knowledge is all there and fully lit, and the name is the one door at the end that will not open. Nothing has fallen out of storage. The final step has simply timed out.

Faces got given their own department

Meanwhile the face side of the operation is absurdly overqualified.

Your brain has a patch of tissue that appears to be there almost entirely for faces. Nancy Kanwisher and colleagues identified it in 1997 and called it the fusiform face area, and the evidence that it is specialised for perceiving faces rather than objects in general has held up for decades. Dedicated hardware. Faces mattered enough, for long enough, that we grew a department for them.

And that department has a long memory. In a classic study of school yearbooks, researchers found that people could still pick out former classmates’ faces with around ninety per cent accuracy up to roughly thirty-five years after leaving, sliding only to about seventy-five per cent at forty-eight years. Freely recalling the names of those same classmates fell away far faster. Thirty-five years of holding onto a face belonging to someone you merely shared a corridor with. That is a system doing its job extremely well.

The worst possible moment to hear something

There is one more thing working against you, and it is almost funny.

Think about when a name actually arrives. Someone says it in the half second before you are expected to speak, while you are arranging your own face, working out whether this is a handshake or a wai, and rehearsing your own name in case you need it. You are not listening. You are queuing.

This has a name too. The next-in-line effect, first tested by Malcolm Brenner in 1973, found that words spoken in roughly the nine seconds before your own turn to speak are recalled worse than everything else. Storage never enters into it. The information fails to get encoded at all, and goes in one ear and out of the building.

So a name means nothing, waits its turn behind everything else, and gets delivered at the exact moment you are least able to hear it. Three strikes, none of which are about how much you care.

Give the name something to hold onto

So the fix has nothing to do with trying harder. It is doing the small piece of work that nobody did for the name on its way in.

Say it back out loud. Not “nice to meet you” but “nice to meet you, Pim.” It passes as good manners, which is a bonus. The work is being done by the generation effect: information you produce yourself is remembered better than the same information you merely received. Producing the sound yourself is what files it.

Then give it a meaning it does not have. Anything will do, and the sillier the better, because you are manufacturing the rope that a surname arrived without. Rhyme it. Attach it to somebody you already know with that name. Notice that Pim wears the exact green of a mango. It will feel like a party trick for about a fortnight and then it will feel like just knowing people’s names.

And write it down within the hour. Not at the table, that is strange behaviour. But in the taxi, in your phone, with something attached: the name, plus one fact. Pim, ran the sound desk, hates coriander. It looks like admin. What you are actually doing is building the web the name should have arrived inside.

Just ask

The cheapest tool here is the one nobody wants to use.

Ask within thirty seconds and it costs nothing. Sorry, your name has already gone, say it again. Nobody has ever minded. Ask three weeks later, mid conversation, after they have mentioned your kids by name, and it costs considerably more.

We do not ask because the lapse feels like it means something. It reads as evidence that we were not paying attention, that we do not care enough. Knowing the actual mechanism helps, because it lets you ask in a level voice instead of a guilty one.

What your memory was actually doing

Years ago I ran a small chain of restaurants, and I met an implausible number of people in a decade of doing it. There was a woman who came in most Fridays, always alone, always the same table by the window. I knew her order. I knew she had a daughter studying in Melbourne, that she had put off a hip operation twice, that she tipped in cash and folded the notes in half.

I did not know her name. Not once, in three years.

I used to think that was shameful. Now I think my memory was making a reasonable call about what mattered. It kept the table, the daughter, the hip, the folded notes. It dropped a sound with nothing behind it.

Which is not an argument for giving up on names. Names are how people know you have registered them as a person rather than a Friday regular, and that is worth the small effort of manufacturing the meaning yourself. But it is an argument for dropping the self-accusation that comes attached.

You are not bad with names. You are excellent with people and clumsy with labels, which, of the two available problems, is comfortably the better one to have.