Continuity editing

Why does an audience read two different camera angles as one continuous space?

Two people talk across a kitchen table. The screen shows one of them; then, between one frame and the next, every pixel is replaced and we are looking at the other from the opposite side of the room. Nothing in ordinary experience does that. It ought to be alarming, and instead nobody in the cinema so much as blinks — they report having watched one room, continuously.

The stock answer is that we have learned the convention: a century of film has taught audiences what a cut means. That is worth testing rather than accepting, because if it were the whole story, a person who had never seen a film should be thoroughly lost.

Researchers have run something close to that experiment. Stephan Schwan and Bilge Ildirar showed short edited sequences to adults in remote Turkish villages with essentially no exposure to film, and the answer came back split. These viewers frequently did not read a cut as continuing a scene; they described separate pictures. But when the sequence depicted an action they knew intimately — preparing food, working with familiar tools — they followed it across the cuts easily. Learning matters, then, but what did the work was not familiarity with cinema. It was familiarity with the event.

Now a stranger question. When has your own vision ever been continuous? Your eyes jump three or four times a second, and during each jump vision is suppressed and the whole image is replaced by one from a different angle. You do not experience your visual field as flickering between snapshots. The machinery for stitching a discontinuous stream of views into one stable place is not something cinema invented; it is running all the time. A cut, on this reading, is a glance you did not make.

Which raises the decisive question: what does that machinery carry across the gap? If it stored a picture, a mismatched cut would jar like a wrong note. It plainly does not. Daniel Levin and Daniel Simons filmed short scenes in which the central actor was swapped for a different person across a cut, along with wholesale changes of props, and most viewers noticed nothing. Vision retains the gist and a handful of attended things: who is where, who is looking at whom, what motion was under way. The rest is not remembered wrongly. It is simply not held.

So the continuity is not reconstructed from evidence. It is hypothesised, and then left unchallenged because almost nothing is checked. An editor's craft is therefore not building a space but arranging that the cheapest available hypothesis is the intended one. An eyeline match fixes where a character is looking, one of the few things attention holds. Keeping the camera on one side of an imaginary axis of action — the 180-degree line — preserves screen direction, so whoever was on the left stays there. Cutting on a movement lets the second shot complete a gesture the first began, and lands the join when attention is committed elsewhere and change detection is at its worst.

Notice what crossing that line does and does not do. It does not create an impossible space; the geometry is perfectly constructible. It creates a space consistent with a simpler hypothesis — the pair have swapped places, or one turned around — and the viewer takes that one. The rule is not a law of optics. It keeps the intended reading the lazy reading.

Three refinements connect those steps. First, "the audience is fooled" is the wrong description. Nothing false is believed. The viewer builds the space the shots jointly support, using the only evidence on offer — and the evidence is thin because human vision is always thin, not because film is deceptive.

Second, this is why continuity errors are so cheap. They survive into finished films by the thousand, despite supervisors paid to catch them, because the audience is not comparing images. Only a mismatch in something attended — a hand that changes which glass it holds mid-reach — reliably breaks through.

Third, and honestly: how much of this is perceptual and how much learned is genuinely unsettled. The attentional account sketched here, developed most fully by Tim Smith, is a leading theory rather than a settled result, and the cross-cultural work above shows learning effects a purely perceptual story cannot absorb. The safest summary is that cinema's rules are conventions fitted to a perceptual system already stitching discontinuous views together — which is why they were found so quickly and travel so well.

• Why a jump cut within a single angle feels more violent than a cut between two very different ones.
• What happens in a long unbroken take, where the space must survive without any cuts to protect it.
• How the same attentional logic is exploited in stage magic and in the edited "gotcha" interview.

Every term the collection defines is gathered in the glossary.