Whose film is it?

The measure

Every scored film here carries twenty readings — how dark, how playful, how tense, how beautiful. Two films can be placed a measurable distance apart on them, and the only comparison worth making is a matched one: as the directors piece found, most of what looks like a personal signature is the genre and the era talking.

So every number below compares two films of the same primary genre, from the same decade. Two such films picked at random sit 4.79 apart. That is the line everything is measured against.

Series are thrown out throughout. A pair of films sharing both top-billed actors is overwhelmingly a sequel — the same film again, not two films that happen to share a person — and leaving them in flatters everybody.

The order

Two films of the same genre and decade, by what they share. The dashed line is sharing nothing.

The director wins, at 19.6% tighter than the baseline. No surprise, and it is the answer the argument usually assumes.

The lead actor is the surprise: 13.7%, which is about seventy percent of what the director is worth. Across ten thousand matched pairs, with sequels removed, casting a particular person does nearly as much to how a film feels as choosing who directs it.

And the studio does almost nothing. 7.1% — a third of the director, and small enough that a house style is mostly a marketing claim. That is worth sitting with, given how much of film discourse is organised around studios having sensibilities. A studio picks projects; it does not appear to shape how they feel.

But the star’s effect does not travel

That near-parity comes with a catch, and it is the most interesting thing here.

The table above compares films of the same kind. Ask the question a different way — take everything a person has been in, across all genres, and ask how much each reading varies within their body of work against how much it varies within a genre — and the two come apart completely.

A director holds most readings together across genres. A lead actor holds none of them: every bar but ambiguity and spirituality is at or below zero, and violence is −67%.

A star’s filmography is not merely un-alike. It is more varied on almost every reading than a random genre is. Which makes sense the moment you say it out loud: a working actor takes a comedy, then a thriller, then a prestige drama, and the emotional distance between those is the point of the career.

So the two effects are different in kind, not just size. A director carries a sensibility between kinds of film — their thriller and their drama still resemble each other. A star carries nothing between kinds; what they do is make films of the same kind feel more like each other. They are a texture inside a genre rather than a hand across a body of work.

Both are real. Only one of them is what people mean by a voice.

The thing that really does repeat

One number on that chart dwarfs all three: 2.41. Two films sharing both leads and a director — a sequel, in other words — are twice as close to each other as two ordinary films of the same genre and decade.

Nothing a person brings to a film comes near it. The single most reliable way to make two films feel the same is not to hire the same director, or the same star, or to shoot them at the same studio. It is to make the same film again.

Which is the mechanism behind a finding published earlier today: the most-watched films have grown more alike over six decades while the catalogue as a whole has not. The list of what is watched most has filled up with the one thing that reliably resembles itself.

What this is not

This measures how a film feels, not how it looks, sounds or is cut. It uses the top-billed two as “the star”, which is a blunt instrument and unkind to ensemble films. It counts a director credit as authorship, which is a convention rather than a fact. And it can only see people with at least a handful of well-known films, so careers that never travelled are invisible to it.

The studio result is the one most worth arguing with, because production companies in this data include financiers and service companies alongside the labels people mean when they say a studio has a taste. A narrower definition might find more. It would have to find a great deal more to close a gap this size.

The studio whose sensibility people argue about most, as a shelf: The A24 coordinate.