[EXPERT: CONSTRUCTED EYE] Day 11 — The Chubb Illusion and Statistical Context

The Chubb Illusion and Statistical Context

Course: The Constructed Eye: Visual Illusion, Perception Science, and the Work of Akiyoshi Kitaoka and Beau Lotto
Day: 11

Demonstration of the Chubb illusion: The central squares are identical in luminance but appear lighter or darker depending on the contrast of their surround. Advanced control of such context effects is crucial for perceptual painters.

Observed Effect: The Chubb illusion, first systematized by Chubb, Sperling, & Solomon (1989), is a brightness and contrast illusion where a textured region seems lighter or darker depending on the spatial and contrast statistics of its surround. Critically, artists leveraging variable markmaking or backgrounds frequently encounter such context-driven shifts, often ascribed to 'contrast adaptation' in studio lore but rooted in physiology.

Expert Objective

Build a nuanced technical model for how context statistics—especially contrast and noise—alter the apparent lightness and contrast of image regions, referencing psychophysics and early visual processing. Evaluate implications for constructing pictorial depth or illusionary volume in advanced studio practice.

Evidence and Competing Explanations

Contrast and noise-driven context: The left target is embedded in a high-contrast, noisy surround and appears less salient; the right in a quieter surrounding, sharpening its perceived contrast. This reflects Chubb, Sperling & Solomon (1989) and subsequent findings (Kingdom, 2011).

Supported Mechanism: The most empirically supported account posits a normalization mechanism: the visual system encodes the contrast of a patch relative to a weighted pool of surrounding contrast (contrast gain control). This adaptive normalization occurs at early cortical stages (Carandini & Heeger, 2012). Notably, this is not a generic 'filling in' but a computation based on local population coding, evolving from retinal/globular-cell inputs.

Competing Explanations: While surround suppression and normalization dominate, some argue for higher-level grouping (e.g., Gestalt context) or Bayesian priors adjusting expectations of surface reflectance (see Lotto & Purves, 2000). However, the latter is less predictive for fine-grained studio manipulations.

Unresolved Questions: The precise size of the 'normalization pool' and how it interacts with material boundaries, textures, or painted contours in real artworks is debated. Few studies extend beyond simple geometric targets to mark-rich complex surfaces as encountered in contemporary practice.

Digital Experiment

Self-experiment setup: Adjust viewing so that peripheral vision encompasses both grey squares. Note any difference in perceived brightness as you alternate focus rapidly, observing contextual contrast effects.

• Controlled variables: Central patch luminance, global white balance, surround noise/contrast amplitude, adaptation time.
• Protocol: Allow at least 10 seconds of adaptation, then look between squares and describe perceived differences in lightness/contrast.
• Limitations: Self-experiment cannot isolate neural locus or control for distraction, physiologic adaptation, or reflective glare.

Retrieval Question

Question: In the Chubb illusion, what physiological mechanisms create the contextual alteration in perceived contrast, and how would you verify this in a studio setting?

Show Answer

The contextual effect derives from normalization pooling across spatial regions, with early cortical networks computing relative contrast (Carandini & Heeger, 2012). In studio practice, precise grayscale targets should be viewed on varying ground/noise textures, ideally under controlled illumination—results can be compared to published psychophysical values using printed or screen-viewed stimuli, but only controlled psychophysical lab setups can pinpoint neural mechanisms conclusively.

Sources

• Chubb, C., Sperling, G., & Solomon, J.A. (1989). Texture interactions determine perceived contrast. Proceedings of the National Academy of Sciences, 86(23), 9631–9635. Read
• Carandini, M., & Heeger, D.J. (2012). Normalization as a canonical neural computation. Nature Reviews Neuroscience, 13(1), 51–62. Read
• Kingdom, F. A. A. (2011). Lightness, brightness and transparency: Illusions and models. Vision Research, 51(7), 722–736. Read
• Lotto, R.B., & Purves, D. (2000). An empirical explanation of the Chubb illusion. Journal of Cognitive Neuroscience, 12(5), 776–787. Read