Skip to main content

[EXPERT: CONSTRUCTED EYE] Day 1 — Illusions as Instruments for Studying Perception

[EXPERT: CONSTRUCTED EYE] Day 1

Illusions as Instruments for Studying Perception

Course thesis: a visual illusion is not merely a picture that tricks a defective visual system. Properly controlled, it is an instrument that exposes the operations normally hidden inside successful seeing.

Place two physically identical grey squares on different surrounds and they may look markedly different. Arrange four luminance steps around a circle and a completely static image may appear to rotate. Nothing supernatural has happened, and the observer has not failed a simple act of looking. The perceptual system has produced an interpretation that is useful under many ordinary viewing conditions but diverges from the deliberately engineered stimulus.

That divergence is the beginning of a scientific question, not its answer. Saying that an illusion happens because “the brain fills in the gaps” explains almost nothing. Which signals differ? Which computations, neural populations, priors, adaptation processes, eye movements, or contextual statistics are implicated? Which parts of the explanation are observed, which are model-supported, and which remain disputed?

An illusion is a difference that must be located precisely 1. Stimulus Measurable image: luminance, wavelength, geometry, timing. 2. Processing Retinal encoding, context, adaptation, neural computation. 3. Percept Experienced colour, motion, depth, size, or organisation. 4. Report Choice, rating, adjustment, drawing, or verbal account. Scientific inference is not identical to the perceptual report. A report establishes an effect under stated conditions. Competing experiments and models are needed before attributing that effect to a particular mechanism. Original explanatory diagram for The Constructed Eye.
Original course diagram: the physical stimulus, perceptual processing, experienced percept, and observer report are related but not interchangeable levels of description.

Expert Objective

By the end of Day 1, you should be able to treat an illusion as an experimental contrast rather than a spectacle. You should be able to specify the physical stimulus, describe the perceptual effect without smuggling in a mechanism, propose at least two competing explanations, and design a control that would begin to distinguish them.

1. What Counts as an Illusion?

A practical definition is that an illusion is a systematic divergence between a perceptual experience or judgement and a specified property of the stimulus or scene. The word specified matters. “Reality” is too vague to serve as a scientific control. A grey patch can be physically specified by its display values or measured luminance; an angle can be measured geometrically; a static image can be verified as temporally unchanged. The percept can then be measured through matching, adjustment, forced choice, rating, eye tracking, physiological response, or other methods.

The comparison is never completely innocent. A display pixel value is not retinal irradiance. Retinal irradiance is not neural response. A verbal report is not the percept itself. This is why visual science relies on controlled manipulations and converging evidence rather than a single arresting image.

Visual illusions have nevertheless been extraordinarily productive. Eagleman’s review in Nature Reviews Neuroscience describes the historical study of systematic misperceptions, combined with methods for measuring and stimulating neural activity, as a rich guide for neurobiological frameworks and experiments. Illusions make hidden assumptions visible because they create situations in which ordinarily useful processing produces a measurable mismatch.

2. Error, Feature, or Rational Inference?

Calling an illusion an “error” is sometimes descriptively acceptable, but it can obscure the deeper question. A visual system cannot directly recover the world from the retinal image because many different scenes can produce similar retinal patterns. Illumination and reflectance can trade off. Size and distance can trade off. Object motion and observer motion can trade off. Vision therefore operates under uncertainty.

One influential family of accounts treats perception as inference. Geisler and Kersten discuss work in which visual illusions can emerge from Bayesian ideal-observer models: given noisy or ambiguous evidence and prior probabilities learned from ordinary environments, a perceptual estimate can be rational even when it differs from the artificial stimulus used in the laboratory. Brown and Friston similarly model the Cornsweet effect using prior beliefs about probable gradients of illumination and reflectance.

This does not mean that every illusion has been explained by the slogan “Bayesian brain.” A model must specify its evidence, priors, likelihoods, and predictions. It must outperform plausible alternatives. Ecological, direct-perception, neurophysiological, adaptation-based, and stimulus-specific accounts may explain different effects or different levels of the same effect. The expert task is not to choose a fashionable framework in advance; it is to identify what each framework predicts.

Four explanatory lenses can address the same illusion Stimulus / computational account Which image statistics or local sequences predict the effect? What computation transforms them? Test: manipulate one image parameter at a time. Neurobiological account Which cells, pathways, latencies, eye movements, or physiological responses covary with the effect? Test: link controlled stimulus changes to responses. Empirical / inferential account What environmental statistics or prior experience would make this percept useful or probable? Test: derive quantitative predictions from the model. Ecological / task account Which useful information is the observer detecting, and how does the artificial display alter it? Test: vary task, action, and viewing environment. Original course diagram. These lenses may compete, complement one another, or operate at different explanatory levels.
Original course diagram: a perceptual effect does not arrive with its mechanism attached. Explanations must be converted into distinguishable predictions.

3. Kitaoka and Lotto as Two Productive Entry Points

Akiyoshi Kitaoka is a professor of psychology at Ritsumeikan University who studies geometric, lightness, colour, and motion illusions, as well as perceptual completion and transparency. His official illusion pages function simultaneously as a research catalogue, design laboratory, and public archive. His best-known work, Rotating Snakes, created in 2003, is an optimized Fraser-Wilcox illusion: a static arrangement that commonly produces apparent rotation, especially away from direct fixation.

The importance of Kitaoka’s practice is not merely that the images are compelling. His designs isolate and recombine visual variables. A sequence of luminance and colour steps, repeated around a geometry, becomes a controllable probe. Change the sequence, contrast, size, eccentricity, or arrangement and the percept changes. The picture becomes an experimental apparatus.

Beau Lotto’s work, including research with Dale Purves, offers a complementary route. Their empirical account of colour perception argues that visual responses to spectral stimuli are shaped by the historical relationship between those stimuli and their probable sources. The key artistic consequence is radical but precise: colour appearance cannot be understood from a swatch alone. Context is not decoration added around the “real” colour; it is part of the evidence from which colour appearance is constructed.

These approaches meet in a shared refusal to treat seeing as transparent access to physical measurements. They differ in emphasis. Kitaoka often begins by designing exceptionally strong, manipulable effects. Lotto and collaborators emphasize how perceptual meanings emerge from empirical relationships and uncertainty. Across the course, neither will be treated as a final authority. Their work will be compared with competing models and evidence.

Evidence and Competing Explanations

Consider apparent motion in a static peripheral-drift display. The observation is straightforward: under some viewing conditions, a static pattern appears to move. Several facts and hypotheses must then be separated:

  • Established stimulus fact: the digital image is static; its pixels do not change over time.
  • Observer report: many observers report apparent motion, often stronger in peripheral vision.
  • Supported stimulus relationship: particular luminance arrangements and contrast relationships strongly affect the illusion.
  • Supported physiological association: studies have linked microsaccades and blinks with reported illusory rotation, but association and complete causal explanation are not identical.
  • Competing or partial mechanisms: visual latency differences, transient responses, motion-energy signals, eye movements, adaptation, and predictive processes may contribute at different stages.
  • Unresolved question: no short verbal formula fully explains why every observer, pattern, scale, and viewing condition produces its particular strength and direction of apparent motion.

This hierarchy is essential. Illusion discourse often jumps directly from “I see motion” to a confident neural story. Research-grade work preserves the gap between effect and explanation, then uses experiments to narrow it.

4. Psychophysics: Turning Seeing Into Evidence

Psychophysics studies relationships between physical stimuli and perceptual judgements. Its power lies in disciplined comparison. Instead of asking only, “Do you see the illusion?”, we can ask how much contrast is needed before the effect appears, which of two patterns appears stronger, what adjustment makes two regions look equal, how the effect changes with eccentricity, or how reliably an observer discriminates alternatives.

For an artist, this transforms intuition into an experimental practice without eliminating aesthetic judgement. You can still ask whether an image feels unstable, luminous, or spatially ambiguous. But you must also construct controls. If two versions differ in colour, contrast, scale, and repetition simultaneously, you cannot know which variable mattered. A strong first experiment changes one parameter and holds the others constant.

A minimal controlled self-experiment A: reference Target grey = identical in A and B B: one changed variable Only the surround luminance changes compare Record: viewing distance, display, ambient light, fixation, order, duration, response, uncertainty. Do not infer a neural mechanism from a single introspective comparison. Original course diagram: identical target values can support different appearances when contextual evidence changes.
Original course experiment: a simultaneous-contrast comparison. The target rectangles are physically identical; only their surrounds differ.

Digital Experiment

Question: How does surround luminance change the apparent lightness of a physically identical target?

  1. Create two equal mid-grey rectangles using exactly the same digital value.
  2. Place one on a dark surround and one on a light surround. Keep target size, position, edge sharpness, and viewing duration constant.
  3. View them side by side at a fixed distance. Record which target appears lighter and rate the strength of the difference from 0 to 5.
  4. Reduce the difference between the surrounds in several steps. Record the point at which the effect becomes weak or uncertain.
  5. Reverse left and right positions to test whether spatial order changes your judgement.
  6. Add a thin neutral border around both targets and repeat. Record whether separating the target from the surround changes the effect.

Controlled variables: target values, target dimensions, display, ambient illumination, viewing distance, exposure duration, and instructions.

Manipulated variables: surround luminance, target position, and border condition.

Dependent measure: forced choice or ordinal rating of apparent target lightness difference.

Limitation: this structured self-experiment establishes how your reports change under controlled conditions. It does not by itself identify retinal, cortical, Bayesian, empirical, or ecological mechanisms.

5. What an Artist Can Learn That a Demonstration Cannot Teach

An illusion demonstration says: look, these identical greys appear different. An expert investigation asks how much they differ, under what conditions, for whom, and which formal decisions amplify or suppress the result. It asks whether the effect survives changes in scale, edge treatment, viewing distance, colour, attention, and reproduction medium. It distinguishes an artwork that contains an illusion from an artwork whose structure depends on an understood perceptual operation.

This distinction will shape the entire course. We will reconstruct effects, but reconstruction is only the first step. We will parameterize them, create controls, document observations, and compare explanations. The eventual capstone must not merely resemble Kitaoka or illustrate a claim by Lotto. It must present an original perceptual problem with a transparent experimental logic and a defensible account of what is known and unknown.

Retrieval Question

Without referring back to the article, explain the difference between these four statements:

  1. The stimulus is physically static.
  2. The observer reports apparent motion.
  3. Microsaccades covary with reports of apparent motion.
  4. Microsaccades completely explain the illusion.

Which statements are observations, which are relationships supported by experiments, and which would require substantially stronger causal evidence?

Key Takeaways

  • Illusions are scientifically valuable when they create controlled divergences between specified stimulus properties and perceptual reports.
  • A perceptual effect does not arrive with its mechanism attached.
  • Bayesian, empirical, neurobiological, ecological, and stimulus-level accounts must be judged by their distinct predictions, not their slogans.
  • Kitaoka’s illusion designs function as parameterizable experimental objects; Lotto’s work foregrounds the role of context and empirical history in appearance.
  • For advanced artistic research, controls and documented uncertainty are as important as visual intensity.

Sources

Next lesson: The eye, retina, visual cortex, and perceptual construction—what can and cannot be inferred from the anatomy of seeing.

Labels:

Comments

Post a Comment

Popular posts from this blog

The Impact of AI on the Art World: A Double-Edged Sword

Art, in its many forms, has been a cornerstone of human culture for millennia. From the cave paintings of our ancestors to the digital art of today, it has always been a medium for expression, communication, and reflection. However, the advent of Artificial Intelligence (AI) has brought about a seismic shift in the art world, particularly in the realms of realism and photorealism. While AI has the potential to revolutionize these art movements, it also poses significant challenges and threats.  ## The Rise of AI in Art AI's involvement in art is not a new phenomenon. In fact, it has been gradually infiltrating the art world for several years now. AI algorithms can now generate artwork, and some of these pieces have even been sold at prestigious auction houses.  (images created at random using Midjourney) However, the impact of AI on art is not limited to the creation of new pieces. It also extends to the way we interact with and interpret art. For instance, AI can anal...
Post a Comment
Read more

The Art of the Intimate Scale | Chapter 1 of 12: What is Miniature Painting?

  The Art of the Intimate Scale Chapter 1: What is Miniature Painting? Definitions, Scale, the RMS Canon & Why the Word Has Nothing to Do With Small Chapter 1 of 12 The word "miniature" has deceived the world for five centuries. It does not mean small. Understanding this single etymological fact is the foundation of everything that follows — and the single most powerful opening line you own as RMS President. The Etymology: Minium, Not Minor The term derives from the Latin miniare — to paint or illuminate with minium , the bright red lead pigment (lead tetroxide, Pb₃O₄) used by medieval scribes to draw the decorative initial letters and ornamental borders of manuscripts. The artisan who did this work was a miniator . The resulting decorated letter was a miniatura . The confusion with "small" arose from a happy accident: portrait miniatures were indeed tiny objects. By the time the word entered English usage in the 16th century, the two concepts had become fused ...
Post a Comment
Read more