Lev Kuleshov, sometime around 1919, performed the experiment that every film editor eventually reruns by accident. He took a single close-up of the matinee actor Ivan Mosjoukine — a face composed and unreadable, the eyes resting on nothing — and spliced it, in turn, against three other images: a bowl of soup, a child in a coffin, a young woman draped across a divan. Audiences, the story goes, praised the actor’s range: his restrained hunger before the soup, his grief at the coffin, his desire at the divan. The face had not moved. The same frames of Mosjoukine ran each time. The soup did the acting. Whatever the audience saw was not in either shot but in the seam between them — in the cut.
The original 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 meet it in a book. You lay two shots end to end and a third thing appears that is in neither one. Reverse the order and the meaning inverts. Trim four frames and a glance becomes a stare. Meaning, at the bench, is not a property of the shot; it is a property of the join.
Faced with this, we reach — almost helplessly — for language. We say the cut is a comma, the dissolve a semicolon, the fade to black a paragraph break. We speak of “the language of film” and “the grammar of editing,” and we teach beginners as though there were a syntax to be learned and, now and then, a syntax error to be marked in red. The metaphor is old, respectable, and productive. It is also flattering, and that alone is reason to ask how much of it is true. How far does the analogy between editing and syntax carry — and what does it quietly cost 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 not a vocabulary or a rulebook but a principle: that 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 is not, and from its place in the sentence around it. “Dog bites man” and “man bites dog” deploy identical lexical atoms to opposite effect; the meaning lives in the order, the relation, the position. The unit is subordinate to the structure that holds it.
This is the editor’s daily experience, raised to a principle. Eisenstein, who thought harder about the cut than almost anyone, insisted that montage was not the linking of shots like a chain of bricks — that was Pudovkin’s gentler model — but their collision, a dialectic in which two images strike each other and throw off a third meaning belonging to neither. He took his model from the written character, where the sign for “eye” set beside the sign for “water” yields not “wet eye” but “to weep” — a concept produced by combination, an idea no single pictograph contains. The cut, for Eisenstein, was the place where cinema stopped recording and began to think.
So when we call editing “syntactic,” we point at something real and important: cinema’s meaning is positional, relational, emergent. It is assembled from juxtaposition rather than carried by its parts. If that were all the metaphor claimed, it would be unimpeachable.
The case deepens. Continuity editing — the invisible style that governs nearly every narrative film you have ever 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 that screen direction stays consistent. Cut on the middle of a gesture, the match on action, so the eye glides over the seam. Answer a glance with the thing glanced at, the eyeline match. Alternate two framings across a conversation, shot/reverse-shot. These conventions are learned so early and so thoroughly that they have gone invisible, and like the rules of syntax they announce themselves mainly in the breach: a continuity error registers as wrongness — a jolt, a stumble — before it is consciously diagnosed, the way a scrambled sentence trips the ear before the mind can say why.
And like grammar, this system can be violated on purpose. When Godard let the jump cuts stand in Breathless — Jean Seberg’s head snapping from position to position in the back of a convertible, time visibly skipping — he was committing, deliberately, the very thing every editor had been trained to repair. The effect was the cinematic equivalent of broken syntax in a poem: a rule flouted not from incompetence but for rhythm, for nerve, to make you feel the medium’s own grain. Enjambment does this to a line of verse; the jump cut did it to the shot.
Here, though, the first hairline crack appears, and it runs straight through the most popular form of the metaphor — the punctuation analogy. The trouble is that a comma has a fixed grammatical office. It coordinates, it separates, it sets off a clause, and its functions can be listed in a style manual. 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 in particular beyond a change of view. Its meaning is supplied entirely by context — by 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 distinction is about to become the whole story.
Push on the metaphor where it is weakest and it gives way at a single load-bearing joint: a shot is not a word.
Human languages are built on what the linguist André Martinet called double articulation, and it is 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 still smaller, finite stock of meaningless ones: the few dozen phonemes of a language, sounds that signify nothing alone — the b in “bat” means nothing — but that recombine to generate the entire lexicon. Two layers of articulation, both finite, both discrete, both rule-governed: from a closed inventory of sounds, an open inventory of words; from a closed inventory of words, an unbounded set of sentences. And crucially, the lexicon exists. There is a dictionary. The units are given in advance of any particular utterance, fixed by convention, the same for every speaker.
Now look for the equivalents in film, and they are not there. There is no phoneme of cinema, no meaningless minimal unit out of which shots are built. More damaging, there is no lexicon of shots — no dictionary you could consult, no finite inventory of images with settled meanings shared across all films. Christian Metz, who in the 1960s tried more rigorously than anyone to constitute a semiotics of cinema, followed the analogy to exactly this wall and stopped. A shot, he saw, is not a word. It is at minimum a statement — here is a man, walking left, at dusk — and frequently a whole paragraph. It is infinitely variable: any subject, any framing, any duration, any movement, none of it drawn from a closed set. Where a word is one of a countable many, a shot is one of an uncountable infinity, invented fresh each time. The unit on which the entire grammatical metaphor rests 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, he concluded, is a langage sans langue — language without a language-system. Borrow Saussure’s distinction: langue is the shared code, the fixed system of conventions that precedes and enables any individual act of speaking; parole is the act itself, the particular utterance. A language has both. Cinema, Metz argued, has only parole — endless individual utterances — and no langue beneath them, no settled code that speakers hold in common before they speak. Films communicate, abundantly; but they do so without a grammar in the strict sense, because a grammar is a property of a langue, and cinema has none. Which means “grammar” was very likely the wrong word from the beginning. Only “syntax,” in its loosest sense — the bare fact of meaningful combination — survives the audit. Everything the word “grammar” promised on top of that begins, here, to fall away.
And once it begins to fall away, it goes quickly. Each thing we assumed a grammar would provide 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 is no cut for negation. You cannot 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, not a denial. You cannot easily cut “every soldier” as against “this soldier,” or “if she had stayed” as against “she stayed.” Film can sometimes smuggle these in through convention and context, but it has no morphemes for them, no reliable operators. The lesson is fundamental: editing is presentational, not propositional. It arranges appearances; it does not assert claims that can be true or false. A sentence can lie. A shot can only show, and be framed.
Then there is the shape of the thing. A sentence is a line — one word after another, a single stream the grammar threads. A shot is not a unit on a line; it is a bundle delivered all at once: image and motion, composition and color, faces and their performances, dialogue, ambient sound, and music, all arriving together and all of it carried across the cut at the same instant. When you make an edit you are not joining two words; you are joining two dense chords. The linear-string model that syntax assumes simply underdescribes this. The truer picture is orchestral — many lines sounding at once, the cut a downbeat that moves all of them together.
There is the question of 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 must simply learn, differing from language to language. The conventions of continuity editing are not arbitrary in this way. They are grounded in perception — in how the eye moves, how attention transfers, how the mind stitches a coherent space from 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 apparatus of seeing, not decreed by an arbitrary code. And these conventions are barely a century old — invented, refined, and revised within living memory of the medium itself, more like the evolving etiquette of a craft than the deep structure of a tongue. Worse for the metaphor, there is 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 such line that a bold enough editor cannot move.
There is the matter of generativity, and here one should tread lightly. Chomsky’s central claim for syntax is that it is recursive: a clause can be embedded inside a clause inside a clause, with no fixed limit, so that a finite set of rules generates an infinite set of possible sentences, most never spoken before. Editing has something that looks adjacent — a flashback within a flashback, a story nested inside a told story — but it is closer to associative layering than to the formal, unbounded self-embedding of syntax. There is no closed rule set from which the whole space of possible films is generated. Whatever editing’s productivity is, it is not Chomskyan recursion; it is the open, rule-light productivity of an art, not the bounded-but-infinite productivity of a formal system.
And last, most telling: no one is born ready to cut. Children acquire the grammar of their language spontaneously, without instruction, on a biological schedule, inferring rules of staggering complexity from fragmentary evidence — the poverty of the stimulus that led Chomsky to posit an innate faculty for language. Nothing of the kind happens with editing. We learn to read continuity by watching a great deal of film, the way we absorb a convention, not the way we grow a capacity. Audiences parse the 180-degree rule because they have sat through ten thousand hours of films that obey it, not because the human mind arrives equipped to. This is the difference between a literacy and a faculty, and it places editing firmly on the side of the cultural and the learned rather than the deep and the given.
Tally the losses. We have given up the lexicon, the logical operators, the linearity, the arbitrariness, the line between right and wrong, the recursive engine, and the native faculty. What remains of the original metaphor is the one thing we began with: meaning made by combination, plus rhythm, plus the management of a viewer’s attention through time. That residue is real and powerful. But it is no longer recognizably grammar. It is something else — and the something else has a better name.
If editing is not grammar, what is it? The remainder points in two directions at once, and neither is linguistics.
The first is rhetoric. Classical rhetoric divided the art of persuasion into canons, and the one called dispositio — arrangement, the ordering of material for greatest effect — is very nearly a definition of the editor’s job. Not which images exist, that is the shoot, but in what order, at what length, with what emphasis, and above all with what left out. 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. Cross-cutting between two opposed scenes — Coppola intercutting a baptism with a string of executions — is antithesis and parallelism made visible. The editor is an arranger and an emphasizer, a rhetorician of attention, not a grammarian of images.
The second direction, and the one practitioners reach for unprompted, is music. Ask editors how a cut works and they will talk about rhythm, beat, tempo, phrasing — the vocabulary of music, not of language. Walter Murch, who cut both sound and picture and thought about the cut as carefully as anyone since Eisenstein, located its logic in something pre-linguistic entirely. 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 falls where the viewer is already prepared to blink, ready to close one thought and open the next. When Murch ranked what a good cut must satisfy, he put emotion first and overwhelmingly: by his own reckoning, more than half of what makes a cut right is whether it is true to the feeling of the moment, with the continuity of screen space — the thing that most resembles a “rule” — dead last and worth a fraction of the rest. No grammar contains a clause that reads honor the feeling before the geometry. That is an aesthetic and perceptual logic, the logic of where to land the beat and how long to hold the note. A long take is a sustained tone; an action sequence cut ever shorter is an accelerando driving toward a downbeat; the held pause before a cut is a rest, and rests carry weight. The cut’s true relatives are the breath, the heartbeat, and the musical phrase — not the comma.
Why, then, did we ever reach for grammar — and why do we go on reaching?
The plainest answer is partly strategic. 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 not merely a fairground novelty, and to call its montage a language was to claim 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 and useful 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 real and 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 is only a craft, a rulebook where there is mostly a feel, a clean line between the correct and the incorrect where there is only the open question of whether a cut works. Held too tightly, the metaphor flattens the thing it set out to honor — 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 own relationship to every rule of the craft: learn it completely, internalize it until it is invisible, and then know exactly when and why to break it. Godard knew the cut he was breaking. That is what made the break mean something.
So keep the metaphor, but hold it loosely, the way you hold any rule you intend to outgrow. The cut is not a comma. It is a breath — and you place it not where a grammar permits, but where the film, like a living thing, needs to draw air.