Given-name fashions

Why do given names rise and collapse in popularity like fashions when nobody coordinates the choice?

Plot the popularity of a first name over a century and you get a shape you have seen before: a slow rise, a sharp climb, a peak, a long collapse. It is the curve of a hemline. But hold the comparison up to the light and it should not work. Clothes have an industry that retires last year's line, and they wear out, and they cost money. A name has none of that: nobody manufactures names, nobody advertises them, they are free, they never fray, and no parent is pressed to replace one. Strip out the producer and strip out the cost, and the churn should stop. It does not. So what is left doing the work?

Start by asking what a parent is actually choosing. Not a sound, exactly. Along with the sound comes everyone the parent has met who wore it, and everyone they have heard of. A name is a small package of associations, and the parent is choosing the package.

Now watch what happens to that package as the name spreads, because this is the whole engine. Suppose a name is uncommon and pleasant. Parents who hear it like it, and some use it; their children are heard by other parents; the name climbs. At what point does it stop being appealing? Not when it becomes unpleasant — the sound has not changed. It stops when the package changes: when the name now says three girls in the class, or the parents were doing what everyone else did. What made it attractive was that it was not everywhere, and being chosen puts it everywhere.

That much is the ordinary fashion loop, and it needs no industry to turn it. But names have a second decay channel clothes do not, and it is the more powerful. Because a name is given at birth and kept for life, everyone who has one was born in a narrow window, so a name becomes an age marker. To a parent in 2026, some names simply mean a woman of seventy — not a judgement about the name but an accurate inference, and nothing but time undoes it.

Which predicts something specific: recovery must be slow, roughly as long as a human life, because a name cannot return while its bearers are still visible as an age group. That is what people mean by the folk rule that names revive after about a century — a rule of thumb rather than a measured constant, but the returns of Hazel, Ada and Arthur are the pattern it describes.

Now for evidence that parents react to the movement and not merely the level. Jonah Berger and Gaël Le Mens, working with United States and French naming records, found that names which rose faster subsequently fell faster — and, in experiments, that people shown a name's adoption curve inferred faddishness from its steepness and liked the steep ones less.

The story deserves a serious rival, and there is one. Suppose parents have no preference for distinctiveness at all — suppose each simply copies a name in proportion to how common it currently is, with a trickle of new names entering. Models of that kind, applied to naming data by Alex Bentley, Matthew Hahn and colleagues, reproduce the observed frequency distribution and the steady turnover in the top ranks with no status motive whatever. The raw shape of the curves is therefore weak evidence for the differentiation story. What the neutral model does not naturally predict is the asymmetry — that rising fast kills a name — or the age-marker effect, and that is where the discriminating evidence lies. The weighting is genuinely argued rather than settled.

Finally, the sharpest test, a natural experiment rather than a model. If the loop runs on parents expressing something, it should vanish wherever naming is not an expressive act. And it does. Where names are assigned by rule — patronymics, the saint's day, the grandparent's name — distributions are stable across generations, and in early modern England a handful of names covered a very large share of the male population for centuries. The churn is not a property of names but of naming being a choice.

Three refinements.

The first is the reconciliation this piece exists for. The general account of fashion churn leans on two things names lack: an industry that manufactures obsolescence, and objects that cost money and wear out. Names strip both away and the churn survives, which makes those two accelerants rather than causes. What remains is the irreducible core — a signal whose value is its scarcity, and adoption that spends the scarcity. Names are the control condition for the fashion argument.

The second is a caution against a strong reading. Almost no parent experiences this as strategy; the felt experience is that a name is "lovely" or "a bit much now". The mechanism requires only that people take cues from what they hear.

The third is where I would be shown wrong. The account predicts that as naming became more expressive and less rule-bound, concentration should fall — and in United States records the share of babies given the single most popular name is far lower now than in the late nineteenth century. But my picture would be in trouble if a population with free, visible, expressive naming held stable name frequencies across many generations, or if the rise-fast-fall-fast asymmetry failed to replicate in larger or non-Western datasets.

• Why surnames, which are inherited rather than chosen, show no such cycle at all — and what that isolates.

Every term the collection defines is gathered in the glossary.