The Grammar of the Frame

Nobody sat down and wrote the grammar of film. It accumulated, in the first few decades of the twentieth century, out of thousands of practical problems, and by the time sound arrived most audiences could read it without knowing they were reading anything.

The basic vocabulary is a matter of distance. A wide shot shows a place and where everyone is in it. A medium shot, roughly waist up, shows behaviour: what people do with their hands, how they stand towards one another. A close-up shows a face, and a face on a cinema screen is enormous, large enough that a thought crossing it becomes an event. Directors move between these sizes the way a writer moves between a paragraph and a single word set on its own line.

Then there are the rules of syntax. The best known is the 180-degree rule. Draw an imaginary line between two characters in conversation. Keep the camera on one side of it and one character will always look to the left of the frame and the other to the right, so that when the film cuts between them the viewer never loses track. Cross the line carelessly and they appear to swap places, and something in the audience lurches without knowing why. Shot/reverse shot, the standard way of filming a conversation, depends on it: two separate set-ups, often filmed hours apart, edited into one exchange. Eyeline matches complete the illusion. Show a character looking off screen, cut to an object, and the viewer assumes that is what they are looking at, even if the object was filmed in another country.

None of this is law. It is convention, and conventions exist partly to be broken on purpose. Directors who want a scene to feel unstable will jump the line deliberately. But you cannot break a rule meaningfully until the audience has absorbed it, which is why it matters that the rule is there.

Some of the grammar removes the safety net of editing altogether. Orson Welles opened Touch of Evil (1958) with a single crane shot of more than three minutes. A bomb is placed in the boot of a car; the car drives slowly through a border town; the camera rises and falls and drifts, picking up the two leads on foot, losing them, finding them again, until the car explodes just after the shot ends. Every actor in that shot has to hit marks and lines in sequence, with traffic, extras and a moving crane, because a mistake in the last minute means going back to the beginning. A long take turns a film shoot back into something like a stage performance, with the camera as the only audience member.

Welles had already done something comparable inside the frame. In Citizen Kane (1941) his cinematographer Gregg Toland used deep focus, keeping the foreground and the distant background sharp at the same time. In one celebrated scene, adults negotiate a boy's future in the foreground while, framed in the window far behind them, the boy plays in the snow. The film does not cut to tell you what matters. You look where you choose, and the actors at the back of the frame have to stay alive even though the scene is not, at that moment, about them.

Other devices are closer to special effects. The dolly zoom, worked out for Hitchcock's Vertigo (1958) and usually credited to the cameraman Irmin Roberts, moves the camera backwards while zooming in, so that the subject stays the same size and the world behind them seems to stretch away. Angles do quieter work. Shoot a character from below and they loom, powerful or threatening; shoot from above and they shrink.

The point for actors is that each of these choices is a sentence they are standing inside. In a wide shot the body carries the scene, and a performance that is all eyes will vanish. In a close-up the same gesture that looked natural from across the room becomes grotesque. On a long take there is nowhere to hide and nothing to fix later. In deep focus you may be doing your most important work while the audience is looking at someone else.

Actors who move from stage to screen often say the hardest adjustment is not the camera itself but the knowledge that the frame keeps changing size around them, and that the right performance is a different size each time.