The Jump Scare: A Defence and a Prosecution

There is a website whose entire function is to tell you, by timestamp, where the jump scares are in a film so you can brace for them. It has been running for years and it has a large audience. Consider what that implies: a substantial number of people treat the horror genre’s most famous device as a hazard to be routed around, like roadworks. No other art form has a service for this. Nobody publishes a timecode list of the sad bits in a drama.

That website is the prosecution’s exhibit A, and it is devastating. It also, on inspection, convicts the wrong defendant.

What a startle actually is

Start with the physiology, because the argument is unwinnable without it.

The acoustic startle response is a reflex. A sufficiently loud, sufficiently sudden sound produces a whole-body flinch within a few tens of milliseconds — faster than conscious thought, mediated in the brainstem, and completely uninterested in whether you are frightened. You will flinch at a dropped saucepan. You will flinch at a car backfiring outside the cinema. You will flinch at a violin stab in a film you are bored by, and you will flinch at the same one on the second viewing, when you know it is coming.

So a jump scare measures nothing about a film’s ability to frighten. It is an assault on a reflex arc and it works on everybody, always, regardless of merit. That is the honest core of the prosecution: the technique is a cheat because it cannot fail, and a device that cannot fail cannot be a craft.

Here is why I still defend it. The reflex leaves a residue. For some seconds after a startle your heart rate is up, your attention is scattered, your breathing has changed, and your capacity to hold a sceptical distance from the film is measurably degraded. You have been softened. Whether that softening becomes fear or annoyance is decided entirely by what the film does next, and that is where the craft lives. The startle is a door being kicked open. The question is whether anything walks through it.

The directors people sneer at are frequently very good at this and get no credit, because the flinch is loud and the ten seconds afterwards are quiet. James Wan’s real skill in the Conjuring films is the recovery — the way he lets a room settle wrongly after a bang, so the audience’s still-firing nervous system has to sit in a silence it does not trust. That silence is the film. The bang was just admission.

Where it comes from

The device has a birth certificate and it is an honourable one.

In Cat People (1942), Jacques Tourneur has a woman walking alone at night, sure she is being followed, the tension racked to breaking — and then a bus pulls in and its air brakes let go with a hiss that is pitched, deliberately, like a big cat. The audience jumps. The threat was a bus. The technique has been called the “Lewton bus” ever since, after producer Val Lewton, and it has been in continuous use for eighty-four years.

What is critical is why Lewton’s unit invented it. They had no money. RKO gave Lewton a title, a budget ceiling and no monster, so terror had to be manufactured out of a woman, a pavement and a sound. The jump scare was a solution to poverty, and the poverty made it an aesthetic. It was invented by the most restrained horror unit in the history of the medium, which ought to give the technique’s despisers pause.

Notice too what Tourneur does with his false alarm, because this is the whole distinction. The bus is a lie, and then the film immediately shows you sheep torn apart in a field. The false alarm is paid for, within thirty seconds, with evidence that the real thing exists. The audience has been made to jump at nothing and then told: your instinct was correct, there is something out there, you simply mis-identified it. That is a proof.

Spielberg’s Ben Gardner scare in Jaws (1975) is the same architecture. It was famously an afterthought: Spielberg wanted a bigger jolt after previews and reshot it in Verna Fields’s swimming pool with milk poured in for murk, reportedly paying for it himself. It is engineered, artificial, added late for effect, and it is one of the great moments in the genre, because the film has spent an hour teaching you that the danger is what you cannot see and the head arrives to confirm it. The jolt collects on a debt already run up.

The prosecution’s real case

So what went wrong? One habit.

It is the unpaid false alarm. The cat that jumps out of the cupboard, the friend who grabs a shoulder, the hand on the arm in the dark — the shock that is revealed to be nothing and is then never redeemed. Tourneur’s bus is followed by the sheep. The modern cat is followed by another cat. Do that four times in ninety minutes and you have taught your audience the single most destructive lesson a horror film can teach: that the film’s alarms carry no information. Once alarms are noise, the audience stops scanning the frame, and an audience that has stopped scanning cannot be dreaded at.

The second charge is the loudness arms race, and this is a modern crime, because the tools got better. Everything that lets a film build a silence you can feel in your chest also gives a lazy director twelve extra decibels to spend on a cupboard. The device migrated from a suspense mechanism to a mix decision, and mix decisions can be made by someone who has not thought about the story.

The third is the trailer, which spends every scare the film has before you buy a ticket, and which is why the timestamp website exists at all. The audience arrives pre-braced. A braced reflex is a smaller reflex.

Underneath all three sits an economic fact. The Blumhouse model works by making horror films for a few million dollars against a possible nine-figure return, and at that budget the startle is the most reliable unit of value per pound in the medium: it needs no star, no effect, no location, and it survives translation into every market on earth. A studio that can buy a guaranteed physical response for the price of a sound cue is going to buy several. That is arithmetic, and arithmetic is a harder opponent than fashion.

The defence rests, partially

I have to concede the strongest counter-evidence, because it is overwhelming: the finest horror of the last fifteen years mostly does not need this at all.

Hereditary has one or two and they are barely the point; its worst moment is a sound in a car and a held frame. It Follows puts its threat in the middle distance in daylight, walking, and asks you to find it. Robert Wise’s The Haunting (1963) frightens people with a door bulging and a sound, and never shows a thing. Kiyoshi Kurosawa has built forty years of dread with essentially no startles at all, and the held frame is a real instrument in the right hands. If the best work does not need the tool, the tool is at best optional.

But “optional” is not “illegitimate”, and the current fashion for treating any startle as a mark of vulgarity has produced its own problem — a strand of the genre so committed to slow dread that it has forgotten films are supposed to do something to you. The Exorcist is a machine for assaulting an audience, and Friedkin was entirely unembarrassed about that, and it is better than almost everything the tasteful end has produced since.

The rule I would offer is simple enough to test. A jump scare is legitimate when it pays a debt the film has already incurred. It is a cheat when it borrows one. Everything else — volume, timing, the cat — is detail.

Spoilers below

Two of the best jump scares ever cut give away the endings of their films, so they belong down here.

Carrie (1976). Brian De Palma ends on a dream sequence — sunlit, slow, scored as a benediction — in which a survivor walks to lay flowers at the ruined White house, and a hand comes out of the ground and takes her arm. The audience screams and then the film simply stops. What makes it work is that De Palma runs the whole shot in reverse and slow motion, so the sequence has a dream’s floating wrongness before anything happens; the viewer’s unease has nowhere to go, and the hand is a release. De Palma has been candid that he took the idea from Deliverance (1972) and its hand rising from the water. Twenty-plus years of horror films ended with the monster’s eye opening because of that shot, and almost none of them earned it, because almost none of them had spent two hours establishing that the world was already a nightmare their heroine could not wake from. De Palma’s whole cinema is about the audience’s complicity in looking, and that hand is the audience being grabbed back.

Mulholland Drive (2001). A man in a diner tells a friend about a dream in which there is a figure behind the building, and says he has come to the diner to prove the figure is not real. They walk round the back. The figure is there. This is the greatest jump scare in cinema and it is the perfect illustration of my rule, because Lynch announces it in dialogue. He tells you what is going to happen, where, and to whom. The debt is stated aloud and then paid in full, and the scene is still unbearable — because the announcement is precisely what makes it unbearable. There is only a promise kept, which is the last thing anyone expects from a jump scare, and the reason Lynch remains the great diagnostician of American unease.

Do that, and nobody puts your film on a timestamp website. They put it in a sentence that begins “there’s a bit where”.