Last Year at Marienbad · essays & theory

A reading · through the lens of theory

Stand in the garden. By the angle of the light it is late afternoon: the cone-shaped hedges and the stone figures throw long shadows across the gravel. The people standing among them throw none. Look again and you cannot un-see it. The film has just told you, without a word, that what you are watching is not a record of anything that happened. The shadows are the place where the dream forgets to cover its tracks.

A man tells a woman they met here last year. She does not remember. He describes a room, a promise, a statue, a near-accident with a broken heel; the camera obliges, supplies the room, and then the room is wrong, the gown changes from white to black between two shots of what seemed one continuous conversation. He corrects himself. The image corrects itself. Nothing settles.

This is the film Deleuze reached for when he needed to show what cinema looks like once it stops measuring time by movement and starts presenting time directly. Most movies run on what he called the sensory-motor link: someone sees a situation, acts on it, changes it, and time passes as the sum of those actions. Marienbad cuts that wire. X cannot make A remember by acting; A cannot escape by acting. They can only watch, insist, endure. She becomes what Deleuze calls a seer — a character to whom things happen as images rather than as events she can answer back.

What rises in the gap is the crystal-image. Deleuze's idea is simple to state and hard to shake: a moment where the actual — the present you're in — and the virtual — memory, dream, the mirror-double of the present — become impossible to tell apart. Usually a film flags memory for you. A dissolve, a wash of soft focus, a title card reading LAST YEAR. Marienbad refuses every flag. "Last year" and "this year" are cut in the same hard CinemaScope light, the same baroque rooms, the same voice. You are inside the crystal, where present and past reflect each other with no surface between them. Resnais and editor Henri Colpi had rehearsed this in Hiroshima Mon Amour, where a woman's narrated memory was steadily undercut by images that would not confirm it. Here they take the training wheels off. No Nevers, no anchoring city, nothing dated. Just indeterminacy, total.

Now sharpen it, because "memory film" is too soft a label and the film deserves better. Robbe-Grillet wrote the script and read it his own way: there is no last year, only the present moment of telling. Resnais leaned toward memory and desire. The structure both men built does something neither reading alone catches, and Deleuze gives it a name — crystalline description. Ordinary description points at a thing already there. Crystalline description conjures the thing and then wipes it out, so the only reality left is the describing itself. X says "a statue"; the statue appears; X says it stood elsewhere, or meant something else, and the first statue is gone. The words make the world and unmake it in the same breath.

And the two stories — we met, we did not; she consents, she recoils; white gown, black gown — are not a mystery with a buried solution. They are what Deleuze, taking the idea largely from Robbe-Grillet, calls peaks of present: incompossible presents held at once, each fully true while it lasts, none of them cancelling the others. The match game that M keeps winning belongs here too, by contrast — a closed system that always resolves, set against a love that never can. All of this is the powers of the false in its purest cinematic form: narration that drops the obligation to be true and makes contradiction itself the engine. Film had always rested on a quiet contract, that the camera does not lie. Marienbad voids the contract on principle and discovers that a film can be wholly compelling without being true about anything at all.

Notice that the voice works against the image, not with it. X's narration describes things the pictures confirm, things they flatly contradict, and things they decline to show. Sound and image come loose from each other; the image stops being self-evident and has to be read rather than simply seen. You are deciphering a discrepancy. And the grand corridors, shot wide so the figures sit marooned at the far ends of the frame, drain of their function as places where anything happens. They become what Deleuze calls any-space-whatevers — beautiful, emptied of coordinates, a setting for nothing.

The lineage is exact. Cocteau's living corridor in La Belle et la Bête taught the camera to glide past frozen, posed bodies; Ophüls's crane-and-dolly through the ballrooms of Letter from an Unknown Woman gave Sacha Vierny the grammar of obsessive memory as fluid architectural movement; Lola Montès stranded a woman in CinemaScope as the passive object of competing stories, and Seyrig's posture is its direct heir. From Citizen Kane comes the deep focus that holds ornament and distant figure in equal sharpness. Downstream, the debt is everywhere: every puzzle film that rewards a second viewing, every story that treats memory as unreliable by design, the whole mind-game tradition, traces back to the permission this film granted — that a film need not resolve, that unresolvability can itself be the form.

So the question the film dangles — did they meet? — turns out to be the wrong question, and learning that is how you learn to watch it. There is no fact under the surface to recover. There is only the crystal, turning, showing you each face of a present that refuses to become past. Watch it again. Watch the shadows.