Ambiguous Endings and the Trust They Demand

An ambiguous ending is a loan. The film borrows against your attention for ninety minutes, promises to settle up in the last reel, and then hands you an IOU instead. Whether that feels like a swindle or a gift depends entirely on the collateral — on whether everything before the final shot was rigorous enough that the withheld answer reads as the film’s considered position rather than the writer’s shrug. Most films that leave you hanging have nothing in the vault. A small number are so precisely built that the missing answer becomes the most solid thing in them.

The distinction matters because “ambiguous ending” has become a compliment by default, a way of describing anything that stops without a summary. It is worth separating the films that decline to answer from the films that never had an answer, because the two produce identical final shots and opposite experiences on a second viewing.

Rigour is the collateral

Test any unresolved ending by rewinding. A film that earned it will have been laying evidence the whole time, evidence that supports two readings simultaneously and never cheats to favour either. A film that did not will fall apart, because the withheld answer was withheld from the screenwriter too, and the earlier scenes were written by someone who did not yet know what they meant.

John Carpenter’s The Thing (1982) passes this test with room to spare. Bill Lancaster’s screenplay, adapted from John W. Campbell Jr’s 1938 novella Who Goes There?, is essentially a logic problem: twelve men, an organism that perfectly imitates its host, and a series of tests that each narrow the field while introducing a new hole. Rob Bottin’s effects work exists to make the imitation credible at the level of biology, which is what allows the paranoia to operate at the level of argument. Every scene either constrains the solution or widens it. When the film stops without telling you who is infected, it stops on a question the preceding hundred minutes have made genuinely undecidable — the evidence is all there and it genuinely points both ways. Universal released it on 25 June 1982, two weeks after E.T., and audiences declined the invitation; the film’s reputation was rebuilt on video by people who could rewind. I came to it on a rented tape in the mid-90s, years after the fact, and the thing that struck me was how much of it is people talking. Our full revisit goes deeper on what Carpenter kept from Campbell.

Peter Weir’s Picnic at Hanging Rock (1975) took the opposite route to the same place. Joan Lindsay’s 1967 novel had a final chapter explaining the disappearance; her publisher cut it, and it appeared posthumously in 1987 as The Secret of Hanging Rock. Weir built the film on the amputation. Russell Boyd shot through bridal veil stretched over the lens to give the daylight its narcotic bloom, Zamfir’s pan flute does the work a score would normally do in a horror film, and the whole picture is organised around geological time looking back at Victorian propriety and finding it temporary. The refusal is the argument. The revisit walks through how Weir constructs a vanishing.

The mechanics of withholding

Craft is what separates the two categories, and it is more specific than “atmosphere”.

Jack Clayton’s The Innocents (1961) is the cleanest technical demonstration in the genre. Adapting Henry James’s The Turn of the Screw (1898) with a screenplay that passed through William Archibald and Truman Capote, Clayton had to preserve James’s central undecidability: are the ghosts real, or is the governess unwell? Freddie Francis solved it optically. Shooting in CinemaScope, he used deep focus and heavy black borders so that the frame’s edges swallow themselves, then lit the apparitions so they can be read as figures or as tricks of light, and staged them where Miss Giddens is always looking. Nobody else ever independently confirms a sighting. The film never once cheats by putting the audience somewhere Giddens is not. That discipline is expensive — it costs you every easy scare — and it is why the film still works. More on how the film’s ambiguity outperforms its shocks.

Michael Haneke’s Caché (2005), which won him Best Director at Cannes, hides its resolution somewhere almost nobody looks: in a static wide shot held long enough that most of the audience reads it as a closing card. The information is on screen, in focus, unremarked. Haneke’s bet is that a viewer trained by conventional coverage will not know where to point their eyes without a cut telling them. He is usually right, which is the point of the film.

Nicolas Roeg’s Don’t Look Now (1973) sits slightly outside this argument and clarifies it. Its ending is brutally definite. What is ambiguous is everything leading to it — Graeme Clifford’s editing splices premonition into the present tense so that the audience is receiving the same disordered signal the protagonist is, and misreading it the same way. The film withholds nothing at the end because it has been withholding continuously throughout. The editing is the whole film.

Who is allowed to do it

Ambiguity is also a budget question, and pretending otherwise is dishonest. Studios have hated the unresolved ending since the preview-card era, for the sound commercial reason that a confused audience recommends nothing. The films above mostly escaped through a side door. Weir made Picnic at Hanging Rock for the newly funded Australian film industry, outside Hollywood’s testing apparatus entirely. Clayton had Twentieth Century-Fox money for The Innocents and a source text prestigious enough that fidelity to James’s undecidability was a defence rather than a risk. Haneke worked in a French production culture where final cut sits with the director as a matter of course.

Antonioni’s Blow-Up (1966) is the instructive case, because it snuck an unresolved ending through the American system by being a foreign object inside it. MGM financed it, the Production Code refused it a seal, and MGM released it anyway through a subsidiary — a manoeuvre that helped kill the Code within two years. It won the Palme d’Or in 1967 and turned a profit. The film’s photographer never establishes whether he saw a murder; the closing mime tennis match asks him to accept a ball that does not exist, and he does. That gesture is the thesis of this entire argument compressed into thirty seconds: the film hands you nothing and asks whether you will play. Audiences did, in large enough numbers to make the studio’s nerve look retrospectively clever.

The counterweight is that when the system does permit ambiguity now, it usually permits it as a marketing asset. A closing shot engineered for argument generates discourse, and discourse is free advertising. That is a different animal from Clayton’s optical discipline, and it should be graded differently, because the film that withholds an answer in order to trend has found a distribution channel rather than solved a problem.

The case against my own argument

Here is the honest problem. Ambiguity is unfalsifiable praise. A film that means nothing and a film that means two things look identical from the outside, and critics — including this one — have a standing incentive to read the second where only the first exists, because the second makes for a better essay.

Worse, the definite ending often haunts harder. George Sluizer’s Spoorloos (1988) resolves completely. It tells you exactly what happened, in detail, and that specificity is precisely what makes it the most upsetting ending in European thriller cinema. When Sluizer remade it himself in Hollywood in 1993 with Jeff Bridges and Kiefer Sutherland, the change that gutted it was a rescue — a resolution swapped for a different resolution, both definite, one enormously worse. Ambiguity had nothing to do with the failure. The two versions are a controlled experiment in nerve, and the pattern repeats across the whole remake economy, where the softening is structural rather than cowardly and the transpacific versions flatten what they cannot license.

The Coens’ No Country for Old Men (2007) takes the Best Picture Oscar for an ending that is not ambiguous at all. Every plot question is answered; the film simply answers them off-screen and then declines to stage the catharsis. That is a refusal of form, and the industry mislabels it as ambiguity because it produces the same confused lobby chatter. The Coens dismantle the thriller’s comforts by fulfilling its plot exactly.

So the thesis needs narrowing. Ambiguity earns its keep only when the unanswered question is the subject. Where the question is decoration, a firm answer would serve better. The twist-ending tradition understood this — its economy depends on a reveal that lands with total finality — and the films that borrow its shape without its rigour are the ones that leave you cold.

What the good ones want

The films that survive on this ground share one trait: they know the answer and have decided the answer is less interesting than the condition of not having it. Kiyoshi Kurosawa’s Cure (1997) is the purest example outside the anglophone canon, a serial-killer procedural in which the detective’s emptiness is the real infection and the case file is a distraction. Kurosawa builds dread from long lenses and dead air rather than from concealment. David Fincher’s Zodiac (2007) is the mainstream version — a two-and-a-half-hour procedural about a case that history genuinely did not close, where the ambiguity is imported from reality and the film’s honesty consists of refusing to improve on it. Fincher makes obsession the subject and the unsolved case the cost.

That is the whole test, and it is a demanding one. Does the film know? If it knows and stays quiet, you are in the hands of somebody who has thought about what you will do with the silence. If it does not know, the silence is just a room with nobody in it, and no amount of pan flute will furnish it.

Spoilers below

The Thing ends with MacReady and Childs sitting in the burning wreckage of the camp, sharing a bottle, each unable to verify the other, waiting to freeze. An alternate ending exists on home video in which MacReady is rescued and cleared by a blood test; it is a considerably worse film in ninety seconds. Carpenter has been consistent for four decades in declining to say which man, if either, is the organism — and the film’s construction means he could not tell you without breaking it, because the evidence was designed to balance.

Caché closes on a long static shot of a school’s front steps. Somewhere in the crowd, two characters who have no reason to know each other are talking. The film’s central mystery has an answer and it is right there, unmarked, for whoever was still watching the frame rather than waiting for the credits.

Spoorloos ends with Rex learning what happened to Saskia by having it happen to him. He wakes underground, in the dark, in a box. The film cuts to a newspaper. Sluizer’s 1993 remake gives Jeff Bridges’s character a shovel, an escape and a villain who loses. It is the same director, the same story and a total collapse.

The Shining (1980) is the outlier worth knowing about. Kubrick shot a hospital coda, played it at the premiere, then had it physically cut from prints within days of release. What survives is the 1921 photograph, which answers nothing and implies everything — a director removing his own explanation once he saw an audience receive it. That is the loan being repaid in public. The hotel’s geography does the rest of the work.