The Grammar of the Cut

Lev Kuleshov, sometime around 1919, ran the experiment every editor eventually reruns by accident. He took one close-up of the silent star Ivan Mosjoukine, a face composed and unreadable, the eyes resting on nothing, and spliced it against three other shots in turn: a bowl of soup, a child in a coffin, a young woman on a divan. Audiences, the story goes, praised the actor’s range. Hunger in front of the soup. Grief at the coffin. Desire at the divan. The face never moved. The same frames of Mosjoukine ran each time. The soup did the acting. Whatever the audience saw lived in the seam between the shots, in the cut.

The footage is lost and the anecdote has hardened into legend, but no editor doubts it, because we feel it in our hands long before we read about it. Lay two shots end to end and a third thing appears that’s in neither. Reverse the order and the meaning flips. Trim four frames and a glance becomes a stare. At the bench, meaning belongs to the join.

Faced with this, we reach for language, almost without choosing to. The cut is a comma, the dissolve a semicolon, the fade to black a paragraph break. We say “the language of film” and “the grammar of editing,” and we teach beginners as if there were a syntax to learn and, now and then, a syntax error to mark in red. The metaphor is old and respectable, and it has earned its keep. It’s also flattering, and that alone is a reason to check how much of it is true, and what it costs us at the point where it gives out.

Start with the strongest version of the case, because the case is strong. What syntax and editing share is a principle: meaning can be manufactured out of arrangement. A word, Saussure argued, has no positive content of its own. Its value comes from its difference from the words it isn’t, and from its place in the sentence around it. “Dog bites man” and “man bites dog” use the same three words to opposite effect. The meaning lives in the order.

This is the editor’s daily experience raised to a principle. Eisenstein, who thought harder about the cut than almost anyone, refused Pudovkin’s gentler picture of montage as shots linked like bricks in a wall. For Eisenstein, montage was collision: two images strike each other and throw off a third meaning that belongs to neither. His model was the ideogram. The sign for “eye” set beside the sign for “water,” he wrote, means “to weep,” a concept made by the combination, an idea neither character holds on its own. The cut was where cinema stopped recording and started to think.

So when we call editing “syntactic” we’re pointing at something real. Cinema’s meaning is positional. It’s assembled from juxtaposition, and the parts carry less of it than the joins. If that were all the metaphor claimed, there’d be nothing to argue about.

The case deepens. Continuity editing, the invisible style behind nearly every narrative film you’ve watched without noticing, behaves uncannily like a grammar. It has rules. Keep the camera on one side of an imaginary line through the action (the 180-degree rule) so screen direction stays consistent. Cut in the middle of a gesture (the match on action) so the eye rides over the seam. Answer a glance with the thing glanced at. Alternate two framings across a conversation. These conventions get learned so early and so thoroughly that they’ve gone invisible, and like the rules of syntax they announce themselves mainly when broken. A continuity error registers as wrongness, a jolt, before anyone can diagnose it, the way a scrambled sentence trips the ear before the mind can say why.

And like grammar, the system can be violated on purpose. When Godard let the jump cuts stand in Breathless (1960), Jean Seberg shot from behind in a moving convertible, her head snapping from position to position, time visibly skipping, he was committing the exact fault every editor had been trained to fix. It was the cinematic equivalent of broken syntax in a poem, a rule flouted for rhythm and for nerve, to make you feel the medium’s grain. Enjambment does this to a line of verse. The jump cut did it to the shot.

Here the first crack shows, and it runs straight through the most popular version of the metaphor: the punctuation analogy. A comma has a fixed grammatical office. It separates clauses or sets them off, and a style manual can list the rest of its jobs. A cut has no fixed office at all. The same straight cut can mean “meanwhile,” “an instant later,” “three years on,” “look what she sees,” “compare these two,” or nothing beyond a change of view. Context supplies the meaning: the shots on either side and the story around them. A comma carries a function. A cut only catches one. The analogy holds at the level of feeling and breaks at the level of code, and that difference is about to become the whole story.

Push on the metaphor where it’s weakest and it gives way at one load-bearing joint: a shot is not a word.

Human languages are built on what the linguist André Martinet called double articulation, and it’s the quiet engine of their power. At the first level, a finite stock of meaningful units (words, or more precisely morphemes) combine into phrases and sentences. At the second, those meaningful units are themselves built from a smaller finite stock of meaningless ones: the few dozen phonemes of a language, sounds that mean nothing alone (the b in “bat” means nothing) but recombine to generate the whole lexicon. Two layers, both finite and discrete, both governed by rules. From a closed inventory of sounds, an open inventory of words; from a closed inventory of words, an unbounded set of sentences. And the lexicon exists. There’s a dictionary. The units are given before any particular utterance, fixed by convention, the same for every speaker.

Now look for the equivalents in film. There’s no phoneme of cinema, no meaningless minimal unit out of which shots are built. Worse, there’s no lexicon of shots: no dictionary to consult, no finite inventory of images with settled meanings shared across all films. Christian Metz, in a 1964 essay whose title asked whether cinema was a langue or a langage, followed the analogy to exactly this wall and stopped. A shot, he saw, is at minimum a statement (here is a man, walking left, at dusk) and often a whole paragraph. Any subject, any framing, any duration, any movement, none of it drawn from a closed set. A word is one of a countable many. A shot is one of an uncountable infinity, invented fresh each time. The unit the whole grammatical metaphor rests on turns out to be the wrong size and the wrong kind.

Metz drew the consequence in a phrase that has outlived most of his system: cinema is a langage sans langue, language without a language-system. In Saussure’s terms, langue is the shared code, the fixed set of conventions that precedes and enables any act of speaking; parole is the act itself, the particular utterance. A language has both. Cinema, Metz argued, has only parole, endless individual utterances with no settled code underneath that speakers hold in common before they speak. Films communicate, abundantly, but without a grammar in the strict sense, because a grammar is a property of a langue and cinema has none. Which means “grammar” was probably the wrong word from the start. Only “syntax,” in its loosest sense, the bare fact of meaningful combination, survives the audit. Everything “grammar” promised on top of that starts, here, to fall away.

And once it starts, it goes fast. Each thing we assumed a grammar would supply turns out to be missing, and each absence tells us something.

A grammar lets you do more than name. It lets you operate. Language can negate, quantify, suppose, and qualify; it has dedicated machinery for not, every, if, perhaps, because. Editing has almost none of this. There’s no cut for negation. You can’t splice the proposition “the dog did not bite the man.” You can only show a dog that fails to bite, which is a different thing, an image of restraint. You can’t easily cut “every soldier” as against “this soldier,” or “if she had stayed” as against “she stayed.” Film can smuggle these in through convention and context, but it has no morphemes for them, no reliable operators. Editing is presentational. It arranges appearances; it doesn’t assert claims that can be true or false. A sentence can lie. A shot can only show, and be framed.

Then there’s the shape of the thing. A sentence is a line, one word after another, a single stream the grammar threads. A shot is a bundle delivered all at once: image and motion, composition and color, faces and their performances, dialogue, ambient sound, music, all arriving together and all of it carried across the cut in the same instant. When you make an edit you’re joining two dense chords. The linear-string model that syntax assumes underdescribes this badly. The truer picture is orchestral, many lines sounding at once, the cut a downbeat that moves all of them together.

Then arbitrariness. Saussure’s first principle was that the linguistic sign is arbitrary: nothing about the sound “tree” resembles a tree, and the rules binding words into sentences are conventions you simply have to learn, different in every language. The conventions of continuity editing aren’t arbitrary in this way. They’re grounded in perception, in how the eye moves and how the mind stitches a coherent space out of fragments. The match on action works because it gives the eye a moving target to ride across the cut. The rule is motivated by the machinery of seeing. And these conventions are barely a century old, invented and revised within living memory of the medium, more like the shifting etiquette of a craft than the deep structure of a tongue. Worse for the metaphor, there’s no stable line between the grammatical and the ungrammatical. A “wrong” cut is only wrong until someone makes it right. The jump cut was an error before Godard made it a signature, and shaky handheld was a mistake before it was a style. Grammar draws a line, however contested, between the sentences a language permits and the ones it forbids. Editing draws no line that a bold enough editor can’t move.

Then generativity, and here it pays to go carefully. Chomsky’s central claim for syntax, from Syntactic Structures (1957) on, is that it’s recursive: a clause can be embedded inside a clause inside a clause with no fixed limit, so a finite set of rules generates an infinite set of sentences, most of them never spoken before. Editing has something adjacent (a flashback inside a flashback, a story nested inside a told story), but it’s closer to associative layering than to formal, unbounded self-embedding. There’s no closed rule set from which the whole space of possible films is generated. Editing’s productivity is the open, rule-light kind that belongs to an art. Chomsky’s engine is a different machine.

And last, most telling: no one is born ready to cut. Children acquire the grammar of their language without instruction, on a biological schedule, inferring rules of staggering complexity from scraps of evidence. That poverty of the stimulus is what led Chomsky to posit an innate faculty for language. Nothing of the kind happens with editing. We learn to read continuity the way we absorb any convention, by watching a great deal of film. Audiences parse the 180-degree rule because they’ve sat through ten thousand hours of films that obey it. This is the difference between a literacy and a faculty, and it puts editing firmly on the side of the cultural and the learned.

Strip all of that away and what’s left is where we started: meaning made by combination, and the steering of a viewer’s attention through time. That residue is real and it’s powerful. But the name has stopped fitting, and there’s a better one.

Take grammar away and the remainder points in two directions, and neither is linguistics.

The first is rhetoric. Classical rhetoric split the art of persuasion into canons, and the one called dispositio, arrangement, the ordering of material for greatest effect, is very nearly a job description for the editor. Which images exist is the shoot. In what order, at what length, with what emphasis, and above all with what left out: that’s the cut. Omission is one of editing’s deepest powers, and grammar barely has an equivalent. A film can simply decline to show, leaping a year or a death in a single cut and trusting the viewer to fill the gap, an ellipsis far more radical than any a sentence performs. The figures of rhetoric map onto cutting far better than the parts of speech ever did. A close-up of working hands standing in for the whole of a labor is synecdoche. Peter Zinner intercutting the baptism with the murders in The Godfather (1972) is antithesis and parallelism made visible. The editor is a rhetorician of attention.

The second direction, and the one practitioners reach for unprompted, is music. Ask editors how a cut works and they’ll talk about rhythm, beat, tempo, phrasing. That’s the vocabulary of music. Walter Murch, who cut both sound and picture and thought about the cut as carefully as anyone since Eisenstein, located its logic somewhere pre-linguistic. We blink, he observed, at the ends of thoughts; a blink is a tiny internal cut, the punctuation of our own attention, and an edit works when it lands where the viewer is already ready to blink, to close one thought and open the next. When Murch ranked what a good cut has to satisfy, he gave emotion 51 percent. Three-dimensional continuity of space, the thing that most resembles a rule, came last at 4. No grammar contains a clause that reads honor the feeling before the geometry. That’s an aesthetic and perceptual logic, the logic of where to land the beat and how long to hold the note. An action sequence cut shorter and shorter is an accelerando driving at a downbeat. The held frame before a cut is a rest, and rests carry weight. The cut’s true relatives are the breath and the musical phrase.

So the question is why we reached for grammar in the first place, and why we keep reaching.

Part of the answer is strategy. “The language of film” was a claim staked at a particular moment for a particular purpose. In the 1920s the Soviet montage theorists and the European avant-garde wanted cinema taken seriously as an instrument of thought, and calling its montage a language claimed for it the dignity of mind. In the 1960s, when Metz and his contemporaries built their semiotics, the motive was scholarly: to make film a fit object for rigorous study you treated it as a system with a code, the way linguistics had a code. In both cases the metaphor did real work, and in both cases it was partly true. The relational insight at its core, that the cut, like the sentence, makes meaning from arrangement, is durable, and no account of editing can do without it.

But the word smuggles in more than it can deliver. “Grammar” promises a code where there’s only a craft, and a clean line between right and wrong where there’s only the open question of whether a cut works. Held too tightly, the metaphor flattens the thing it set out to honor and turns the editor from an artist of rhythm and attention into a clerk checking utterances against a manual. The right relationship to it is, fittingly, the editor’s relationship to every rule of the craft: learn it until it’s invisible, then know exactly when to break it. Godard knew the cut he was breaking. That’s what made the break mean something.

Keep the metaphor, then, but hold it loosely, the way you hold any rule you mean to outgrow. Forget the comma. The cut is a breath, and it goes where the film needs air.