The Long Take as an Instrument of Dread

Most writing about the long take treats it as a feat — a thing to be timed, counted and admired, like a free climb. That is the least interesting property it has. What the long take actually does is remove the cut, and the cut is doing much more work in a horror film than anyone gives it credit for.

Think about what a cut tells an audience. It tells them somebody chose this. A cut is an editorial statement: here is where you should be looking now, because it matters. Over a century the medium has trained us so thoroughly in that grammar that we read it below consciousness — we know a scene is about to turn because the coverage tightens, we know we are safe because the film cut away, we know the thing behind the door is important because the film bothered to show us the door. The cut is a hand on the shoulder. It says: I am here, I am curating this, you are being looked after.

Take it away and something structural changes. A shot that will not cut has stopped being a statement about what matters and has become, simply, a place. The frame acquires corners. It acquires an off-screen — a genuine one, because the film cannot reassure you by cutting there and back. And in a place with corners and no curator, anything can arrive.

The removal is the mechanism

This is why the long take and dread are natural partners, and why the long take and spectacle are natural partners too, which is the source of most of the confusion about it.

Michael Haneke understands the removal better than anyone alive and uses it as a weapon. Funny Games (1997) holds, after the worst thing in the film has happened, on a single unbroken shot of a wrecked living room for something close to ten minutes: a woman trying to free herself, a television burbling, nobody arriving. The camera does not move to help her. It does not cut away out of tact. The scene’s horror is entirely the absence of an editor’s mercy, and Haneke’s point — a hostile, deliberate, slightly insufferable point — is that a cut would have been a kindness the audience has not earned. Caché (2005) opens on a static shot of a street so long that the audience begins to suspect the shot itself, which turns out to be exactly what the film wants.

Kiyoshi Kurosawa built a whole career out of the same insight and is gentler about it. In Cure (1997) he shoots long, wide and static, frequently from a position that includes an exit — a doorway, a corridor mouth, a window — and then simply lets the shot run. Nothing needs to come through the doorway. The doorway is the event. Kurosawa’s genius is that he refuses to promote the frightening thing with a cut, so the audience has to do its own promoting, and an audience that is scanning a frame is an audience that has been conscripted.

The ten-minute wall

The history of the form is largely a history of a physical limit, and knowing it changes how you watch the older examples.

A 35mm magazine held about a thousand feet of film, which at sound speed runs a little under eleven minutes. That was the wall. Hitchcock hit it head-on with Rope (1948) and dealt with it by choreographing the camera into the back of a jacket or a piece of furniture at each reel change, so the darkness could swallow the join. The result is a film with roughly ten cuts, several of them hidden and several of them not, and it is the medium’s first sustained argument that continuity itself could be the subject.

Welles took the same limit and used it for a bomb. The opening of Touch of Evil (1958) — a crane shot that plants an explosive in a boot and then wanders off to follow a couple across a border town for over three minutes — works because the audience knows something the frame keeps failing to prioritise. The camera will not cut back to the car. It has other business. That refusal is precisely what makes Welles’s grubby border picture feel like a trap closing, and it is a dread mechanism executed in 1958 with a crane and a stopwatch.

Digital removed the wall entirely, and it is worth noticing what happened next. Once a take could run for an hour, the technique’s centre of gravity shifted from suspense to endurance. The limit had been doing something useful: it forced a director to spend the take on one idea. When Haneke holds his living room he is spending an entire magazine on the absence of help, and you can feel the cost of it. A digital director holding for the same duration has spent nothing, and audiences can tell.

The horror canon of the unbroken shot

The genre’s foundational example is the opening of Halloween (1978), and it is instructive because it does the opposite of Haneke and arrives at the same result.

Carpenter’s near-continuous Panaglide take walks us up to a house, in through a kitchen, up a staircase, through a murder and back out onto a lawn, all inside a mask. The camera is the killer. The lack of cutting means there is no third party — no editor standing between the audience and the act, no one to blame, no cut to establish that a film is being made. And Carpenter’s real trick is the reverse at the end: when the mask comes off, the film cuts, hard, to a wide. The withheld cut has been storing energy for four minutes and he spends it in one frame. The whole blueprint is in that sequence and most of the imitators copied the mask.

Sam Raimi’s answer is the other pole. The Evil Dead (1981) has its shakycam force — a camera on a plank, run through the woods by two men — and the take is long because the pursuit has to be unbroken to be a pursuit. If the camera cut, it would become a series of shots of woods. Unbroken, it becomes a thing that is coming, and the audience’s inability to escape the shot is the audience’s inability to escape it.

Kubrick’s use in The Shining (1980) is the most quietly radical of the lot. Garrett Brown had invented the Steadicam in the mid-70s — it debuted in Bound for Glory and Rocky in 1976 — and Kubrick immediately grasped that its real gift was continuous space. Following the trike through corridor after corridor without a cut is what allows the Overlook to be geographically wrong: you cannot cheat a room’s position past an audience that has never blinked. The hotel’s impossibility is a product of the take. Cut it up and the film is a haunted house. Leave it whole and the building is lying to you in real time.

Found footage took the whole apparatus and made it a premise. REC (2007) is a stairwell shot as a single ascending nightmare, and the reason the form’s best entries work is that they inherit the long take’s contract — the camera is a fact rather than a narrator, so it cannot promise you anything.

The line between dread and brag

Now the counter-argument, which is substantial.

The long take is also the single loudest way for a crew to announce itself, and the moment an audience starts counting, dread is over. Children of Men (2006) is the honest hard case: Emmanuel Lubezki’s car ambush and the Bexhill assault are astonishing, and they are astonishing partly as stunts. I have watched the Bexhill sequence with people who spent it wondering how the blood got on the lens and whether it stayed there. The film survives the tension between despair and showmanship because Cuarón has an argument to make with the continuity — the refusal to cut is a refusal to look away from a war — but the seam is visible.

Push further and you get Birdman (2014), where Lubezki stitched a film into an apparent single shot, and 1917 (2019), which did the same for a war. Both are magnificent and neither is frightening, because a shot that cannot end is a shot that cannot kill you. The audience knows the film has committed to a technique, and a committed technique is a curator by another name. Russian Ark (2002) — Sokurov’s genuine unbroken ninety-six minutes through the Hermitage — and Victoria (2015), Sebastian Schipper’s real 138-minute take, are logistical miracles and neither is doing what Haneke’s living room does with a locked-off camera and no budget.

The distinction I would draw is this: the long take produces dread when it is deprived, and spectacle when it is equipped. Haneke’s camera cannot help. Lubezki’s camera can go anywhere. Park Chan-wook’s celebrated corridor fight in Oldboy (2003) — a lateral, side-scrolling brawl shot over roughly three days — is a masterpiece of exhaustion rather than fear, and that is the correct outcome for it, because the film wants you to feel a man’s stamina. The technique served the aim. That is all any technique can do.

Where it leaves us

The long take is a subtraction, and subtractions only work if the thing removed was load-bearing. Remove the cut from a film with nothing in its corners and you have a boring shot of a room. That is why so much of the slow-cinema end of the genre is unwatchable, and why the handful of directors who can do it are the ones with something in the frame you have not noticed yet.

Fernando Meirelles shot City of God (2002) like a bullet and cut it like one, and it is terrifying. So the enemy is reassurance, and the long take is simply the cheapest way to stop offering any. The slow zoom does a version of the same job with a fraction of the equipment.

There is a reason directors reach for the shock instead, and I have some sympathy: the startle is a reliable instrument and the held frame is not. A jump scare works on anybody. A four-minute hold only works on an audience that has been given a reason to search the frame, and giving them that reason is the entire job of the preceding hour.

Point a locked-off camera at a doorway. Do not cut. Do not put anything through it. If your film has been working, that is unbearable — and if it has not, you have just proved it in public.