Kuleshov Effect

What is the Kuleshov effect?

The Kuleshov effect is a basic law of film editing: an audience never judges a shot alone, it judges it by the shots around it. Put a calm close-up of an actor next to a plate of food and viewers see hunger. Put the same close-up next to a child in a coffin and they see grief. The face never changed; the context did.

The name comes from Lev Kuleshov, a Soviet director and teacher who ran a film workshop in Moscow after the revolution. In experiments usually dated to the late 1910s and early 1920s, he and his students, Vsevolod Pudovkin among them, cut one close-up of the star Ivan Mozzhukhin against different images. The best-known account lists three: a bowl of soup, a dead child in a coffin and a woman reclining on a divan. Audiences, the story goes, praised the actor's range: hunger, sorrow, desire. Kuleshov did not invent the phenomenon, since editors of silent film had been leaning on it from the first cuts, but his workshop demonstrated it deliberately and turned it into a teaching principle.

Two details matter if you quote the experiment. The original footage is not known to survive, and the accounts Kuleshov and Pudovkin wrote years apart disagree on the specifics, so treat the three pairings as a well-told reconstruction rather than a lab record. And the lesson was never that acting doesn't matter. The lesson is that a shot works like a word, and the sentence is written in the cut.

Creative geography

The same workshop ran a companion test, usually called creative geography. Two actors were filmed in different parts of Moscow, and the shots were cut so they seemed to walk toward each other, meet and look up at a building that was really the White House, taken from existing footage. It is the Kuleshov effect applied to space instead of emotion: viewers assume that shots placed side by side share one place and one moment. Every eyeline match and every cross-cutting sequence depends on that assumption.

Why do filmmakers use the Kuleshov effect?

The audience does the acting. A face that holds back invites the viewer to fill it in, and what people fill in themselves feels truer than what is shown to them. Alfred Hitchcock demonstrated this on himself in a 1964 television interview: the same mild squint reads as kindly after a shot of a mother with a baby and as leering after a young woman in a bikini. Rear Window (1954) stretches the idea across a whole film: Jeff sits in his chair, and our reading of his face depends on which window across the courtyard we cut to.

A reveal lands twice. Cutting to the reaction shot before the thing itself primes the emotion. In Jurassic Park (1993), Spielberg holds on the paleontologists' stunned faces before showing the first dinosaur, so the awe is already in the room when the creature appears.

Geography can lie. Creative geography lets an editor place people where they never were. In The Silence of the Lambs (1991), cross-cutting between an FBI team at a front door and a doorbell ringing inside the killer's house makes us assume they are at the same address. They are not, and the reveal works because the audience trusted the cut.

Performances get shaped in post. Because the neighbor shot supplies the meaning, an editor can move a neutral look from one scene to another, or swap the POV shot it answers, and change what a character seems to feel. Editors often keep a bin of unused looks for exactly this.

Sound counts as the second shot. A neutral face under a low drone reads as dread; the same face under a children's choir reads as nostalgia. Music and off-screen sound work as the insert when no picture follows.

How to shoot and edit for the Kuleshov effect

The effect is made in the edit, but it has to be planned on set. The face and the insert are two halves of one idea, so shoot them as a pair.

1. Direct a neutral, not an empty face. A blank stare looks dead on screen. Give the actor a private task, like recalling a phone number or listening to a sound off set, so the eyes stay alive while the face stays unreadable. Roll 5 to 10 seconds per take and grab a few extra wild looks at the end of each setup.
2. Keep the framing plain. A close-up or medium close-up on a 50 to 85mm lens (full-frame) at eye level lets the face read without the angle commenting on it. A strong high or low angle adds its own meaning and competes with the insert.
3. Set a precise eyeline. Put a mark just off the lens on the side the insert will be. If the insert is a plate on a table, the eyes drop about 15 to 20 degrees; if it is a person across the room, they sit near lens height. Tape a mark on a C-stand arm so every take hits the same spot.
4. Shoot the insert from the character's position. Put the camera at the actor's eye height, tilted to match that eyeline, with a similar focal length. The object must read in one second: one clear subject, uncluttered background.
5. Match light and color. The key should come from the same side in both shots, at the same color temperature. A mismatch tells the viewer the two shots were made in different places, which breaks the link unless that is what you want.
6. Cut in the A-B-A order. Face, insert, face. The second face is where the emotion appears, so hold it 2 to 3 seconds, longer than feels natural, and let the viewer project. Cutting out of the first face on a small eye shift or blink sells the connection.
7. Bridge with sound. Keep room tone continuous across the cut, or start the insert's sound under the face as an L-cut, so the ear confirms the two shots belong to one moment.

In a dialogue scene this is everyday craft: the listener's close-ups in a shot reverse shot exchange are small Kuleshov pairs, and choosing which look follows which line is most of the job.

Does the Kuleshov effect really work?

Yes, with a caveat. When researchers tried to replicate the experiment, the results were more modest than the legend. A 1992 re-creation by Stephen Prince and Wayne Hensley found little evidence that context changed how viewers judged a neutral face. Later studies with more controlled designs, including a brain-imaging study by Dean Mobbs and colleagues in 2006 and a 2016 study by Daniel Barratt and colleagues, found that context does shift the emotion people attribute to a face, but the shift is gradual and stronger for some emotions than others.

For a filmmaker the practical reading is simple. Context nudges an ambiguous face; it does not turn a clear smile into grief. The more neutral and alive the face, the more the neighbor shot can steer it. And in a real film the effect is stronger than in a lab, because story, music, performance and the shots before all stack up behind each cut.

Kuleshov effect vs. juxtaposition

Both come from the same insight, that two images together mean something neither means alone, but they point it in different directions.

Juxtaposition sets two contrasting images against each other so the collision produces an idea, and the viewer is meant to notice the collision. Sergei Eisenstein's Strike (1925) cuts the shooting of workers against the slaughter of a bull: nobody believes the bull is in the same place; the pairing is a metaphor. Juxtaposition can even happen inside one frame, through what shares the composition.

The Kuleshov effect usually hides the cut. The viewer believes the character is looking at the thing, in the same space and time, and assigns the emotion to the character, not to the film. It depends on continuity, so eyelines, light and sound have to match, while juxtaposition is free to break them.

In practice the two sit on a scale that runs from invisible continuity editing to open montage. A reaction shot that borrows grief from a coffin sits at the Kuleshov end; a cut from a crowd to a herd of sheep sits at the juxtaposition end. Soviet filmmakers of the 1920s explored the whole range, which is why both ideas are core to film theory.

Watch the order of the shots, not the acting. In each clip a face that gives little away sits next to what it looks at, and the neighboring shot decides what that face feels. Notice how little the faces move before and after the cut.

AI makes the Kuleshov effect cheap to test: generate one neutral face, cut it against several inserts and watch the meaning move. The hard part is neutrality, because image models like to add a smile or a frown. Nano Banana 2 Lite builds the face, GPT Image 2.5 builds the matching point-of-view insert, and MiniMax H3 Max Turbo keeps the face alive for a few seconds without acting. Lay the pairs out in order on one FlashBoards board and read them left to right like a cut. The fourth prompt re-shoots a real film shot from a text description alone, and the original plays next to the result so you can compare.

Common mistakes

• The model adds an emotion. Neutral expression often returns a polite smile. Describe relaxed muscles: mouth closed and soft, brows level, eyes calm.
• The eyes hit the lens. A stare into camera breaks the link to the insert. Say exactly where the gaze lands and on which side.
• Face and insert don't match. Light from opposite sides or different color temperatures makes the cut read as two places. Repeat the light direction and background in both prompts.
• The insert looks like a product shot. Flat top-down framing doesn't read as someone's view. State the viewer's eye height and the downward angle.
• The video starts acting. Without guidance the face drifts into a frown or smile. Write small physical beats (blink, eye shift, breath) and state what the face is doing: relaxed, lips closed.

Further reading

Lev Kuleshov, Kuleshov on Film, ed. and trans. Ronald Levaco (1974). Vsevolod Pudovkin, Film Technique and Film Acting. Daniel Barratt et al., "Does the Kuleshov Effect Really Exist? Revisiting a Classic Film Experiment on Facial Expressions and Emotional Contexts," i-Perception (2016).