Designing something an audience accepts
Perception is not a camera. It is a prediction machine constantly proposing the most likely world consistent with the evidence — which means an illusion is not a lie told to the eye. It is a plausible story the eye is allowed to finish, and the whole craft is about not handing over the evidence that makes a different story more likely.
The premise
The system is filling in constantly, and it does not tell you
Almost everything a person sees is inferred. There is a hole in each retina where the optic nerve leaves and nobody perceives a hole. Detail collapses within a few degrees of the centre of gaze and the world does not appear blurry at the edges. Colour is reconstructed from three overlapping detectors that do not measure wavelength. The system fills gaps continuously, confidently, and without flagging that it did.
That is the resource. An illusion supplies a small amount of carefully chosen evidence and lets the machinery build the rest, and the reason the good ones feel solid rather than clever is that the audience did most of the construction themselves. It is also why the failures are so total: once a competing piece of evidence arrives, the machinery reruns and lands somewhere else, and there is no partial credit.
Practically this yields one governing question, which is worth asking of every effect before it gets built. What is the cheapest piece of evidence that would make the wrong answer more likely than the right one? Usually it is a sightline, a reflection, a shadow that falls the wrong way, or a second look.
Worked example one
Pepper’s ghost is a contrast problem wearing a geometry costume
A pane between the audience and the scene; something hidden, usually below, lit hard; its reflection appears standing in the space. The geometry is genuinely easy — the image sits as far behind the pane as the object sits in front of it, mirrored about the plane of the glass — and it is not where these fail.
They fail on brightness, because the audience is receiving two light paths on the same retina at once. The ghost is the object’s luminance multiplied by the pane’s reflectance. The set is its own luminance multiplied by the pane’s transmittance. Whether the figure reads as solid or as a smear on a window is the ratio between those two numbers and nothing else.
Which gives you exactly two moves, and people reliably reach for the wrong one first. You can light the hidden object harder, which costs fixtures, power, heat and a performer standing in a punishing amount of light. Or you can darken what sits behind the ghost, which costs a conversation with the designer. Plain uncoated glass reflects about 8%, so it demands a very dark background; purpose-made stage foil runs far higher and buys back a lit set. The calculator does the ratio and tells you what each move would need to be.
Worked example two
Forced perspective, and the front row that ruins it
Two things look the same size when they subtend the same angle at the eye, and angle is size over distance. So an object twice as far away has to be twice as big to match. That is the entire arithmetic, and it is one line.
The interesting part is where it stops. Angular size is one depth cue among many, and the others are not obliged to agree with it. Binocular disparity — the difference between what two eyes see — resolves depth well out to roughly ten metres, and inside that range it simply overrules the size cue. Motion parallax does the same job for anybody who moves their head at all.
So forced perspective is not really a statement about objects. It is a statement about the viewer: it works for one eye, at a distance, held still. Which is precisely a camera. A set that is flawless in the production photograph and collapses for the first three rows has not failed; it was only ever built for a viewpoint the audience was not sitting in, and nobody said so out loud.
| Technique | What the audience supplies | The evidence that kills it |
|---|---|---|
| Pepper’s ghost | That a bright figure in space is a physical object | Any light on the pane itself, or a background bright enough to show through |
| Forced perspective | That equal angular size means equal size | A second eye, or a moving head, inside the stereo range |
| Projection mapping | That light on a surface is the surface’s own colour | Moving off the calibration viewpoint; ambient light raising the black level |
| Scenic reveal | That what appeared was not there before | A sightline into the wing, or a shadow cast from the wrong side |
| Body double / swap | Continuity of a person across a cut in attention | Being looked at continuously by anyone the misdirection missed |
| Levitation | Absence of support | A specular highlight on the support, or a black that is not as black as its surround |
Attention
The audience does not see what you did not point at
Misdirection has a bad reputation as a word, as though it were about deceiving somebody. Mechanically it is about allocation: attention is a limited resource with a single high-resolution centre, and the craft is deciding where it goes rather than preventing somebody from looking.
The effects it exploits are well documented and slightly unnerving. Inattentional blindness means an unattended object can be looked directly at and not reported. Change blindness means a substantial alteration made during a saccade, a blink or a cut is very often missed entirely. And attention follows a small, reliable set of pulls: movement, faces, high contrast, the direction other people are looking, and anything that changes.
Which turns misdirection into ordinary technical work. A followspot is an attention instrument. So is a blackout, so is a music sting, so is a performer looking stage left. The most robust reveals are built on a moment when attention was legitimately somewhere else for a reason the audience would endorse — which is also why an audience that has been given nothing to watch will find the thing you did not want them to.
- Building for the drawing rather than for the seat. The plan view is a viewpoint nobody occupies. Check from the extreme seats, the boxes, and the standing rail before anything gets built.
- Forgetting the camera. An IMAG or broadcast feed is another viewpoint with different rules — longer lens, one eye, and a crop the audience does not have. Effects that work in the room can be exposed on the screen beside it.
- Lighting the mechanism. Black scenic is not black; it has a reflectance, and enough spill will find it. A support that is invisible in the plot goes visible when somebody adds a sidelight in tech.
- Second looks. Most illusions survive one look and not two. A reveal held too long is a reveal being examined.
- Assuming the audience is where you left them. In immersive and promenade work the viewpoint is not fixed at all, and every technique on this page assumes it is. That is the central technical problem of the form.
The line
Fooled with consent is a different experience from fooled without
One distinction is worth keeping sharp, because everything on this page is a technique for making people believe something untrue.
An audience at a show has agreed to it. That agreement is the whole basis of the pleasure — they know they are being worked on, they want to be, and the enjoyment includes admiring the work. That is a collaboration in which the audience is a participant rather than a mark, and the more skilful the technique the more they enjoy losing.
The same techniques stop being that the moment the consent is not there. An effect nobody was warned about, a startle deployed on an audience that did not sign up for one, a simulated emergency, a use of somebody’s likeness they did not agree to — the mechanism is identical and the experience is not. It is worth being able to say, of any effect, which side of that line it is on, and to notice that the answer sometimes depends on a sign at the door that somebody else is responsible for printing.
Where this goes next
The mechanism underneath
The person on the other end has the perceptual thresholds this page exploits — fusion, the cross-modal window, attention. How each sense tells things apart has the depth cues, including the eight that forced perspective is competing against. Being somewhere covers the other kind of illusion — owning a body and causing things — and why latency destroys it. And the two calculators here are Pepper’s ghost contrast and forced perspective.
4 questions on the mechanism rather than the trivia. Nothing is scored and you can answer again — the explanation under a wrong answer is the point of it.
-
1Your Pepper’s ghost looks like a smear on a window rather than a solid figure. The geometry is correct. What is wrong?
-
Distance sets where the image appears, not how solid it looks.
-
The angle sets where the image appears. If the position is right, the angle is right.
-
Right. The audience receives two light paths on one retina: object times reflectance, against background times transmittance. Plain glass reflects about 8%, so it demands a very dark background before the ghost reads as solid.
-
Thickness affects double-imaging on cheap glass but does nothing for solidity, which is purely the brightness ratio.
-
-
2Your ghost needs more contrast. Which move is usually cheaper?
-
This is the one people reach for first and it is the expensive one: more fixtures, more power, more heat, and a performer standing in a punishing amount of light.
-
Right. It costs a conversation with the designer rather than a truck of fixtures, and it moves the same ratio. There are only two moves and this is the one that gets tried last.
-
Distance does not change the ratio of two luminances arriving together.
-
Size decides how much of the object you can see, not how solid it looks.
-
-
3A forced-perspective set is flawless in the production photograph and collapses for the first three rows. Why?
-
Right. Angular size is one depth cue among several and the others are not obliged to agree. Binocular disparity simply overrules it up close, and motion parallax does the same for anyone who moves their head. The set was only ever built for a viewpoint the audience was not in.
-
The arithmetic can be perfect and the effect still fails, because the arithmetic only covers one cue.
-
Lighting can help or hurt, but the front-row failure happens under any lighting because it is stereoscopic.
-
Lens choice matters, but the effect fails from the front row even with a matched focal length.
-
-
4What is the most reliable way to keep an audience from looking at the thing you do not want them to see?
-
Movement is one of the strongest attention pulls there is. Speed attracts the eye rather than escaping it.
-
True on average and useless as a design: the effect has to survive the one person who was.
-
Useful, and fragile: black scenic has a real reflectance and enough spill will always find it eventually.
-
Right. Attention is allocation, not prevention. The most robust reveals sit on a moment when attention was legitimately elsewhere — which is also why an audience given nothing to watch will find exactly the thing you were hiding.
-
Answers are kept in this browser on this device. Nothing is sent anywhere.
- Why is my ghost see-through?
- Why does it fail up close?
- Where will they look?
Where to go next
These pages are a chain rather than a library — the next one along usually answers the question this one raised.