[EXPERT: CONSTRUCTED EYE] Day 2 — The Eye, Retina, Visual Cortex, and Perceptual Construction

The Constructed Eye: Retina, Visual Cortex, and Perceptual Construction

Day 2 of our intensive masterclass explores how the physical apparatus of vision and its neural processing stages interact to produce our constructed visual reality, emphasizing neural constraints and theoretical challenges for artists and perception scientists alike.

Expert Objective

This lesson equips advanced artists to deliberately engage—rather than merely experience—the neurologically grounded illusions fueling the work of Akiyoshi Kitaoka and Beau Lotto. By dissecting retina-cortex interactions, you will learn to manipulate image features within strict neuro-visual constraints, and to critique claims about visual filling-in and constructive perception using laboratory evidence, not intuition.

Observed Phenomena

• Retinal mosaics (e.g. cone density gradients, ON/OFF ganglion cell mosaics) systematically distort sampled visual input (Curcio et al., 1990; Field & Chichilnisky, 2007).
• Retinotopic mapping in primary visual cortex preserves spatial layout but introduces scale, crowding, and boundary effects (Dumoulin & Wandell, 2008).
• Illusory contours and brightness illusions (Kitaoka, Lotto) depend on specific geometrical and contrast arrangements, not an abstract “gap-filling” function (Pessoa & De Weerd, 2003).
• Physiological blind spot filling-in is position-, context-, and luminance-sensitive (Komatsu, 2006); not generalized, but highly constrained by feedforward and feedback processes.

Evidence and Competing Explanations

Peer-reviewed evidence robustly demonstrates that the retina actively preprocesses input—via contrast gain control, edge enhancement, simultaneous suppression, and asynchronous spike patterns—before the cortex receives any percept. Masland (2012) details how dozens of functionally distinct ganglion cell types segment visual input for different perceptual purposes, supporting a multiplexed, not unitary, construction.

• Supported mechanisms: Patchy thalamo-cortical integration and lateral inhibition enable the assembly of apparent motion, edges, or filled-in brightness by weighted summation of physically discontinuous signals (Wandell, 1995; Komatsu, 2006).
• Competing explanations: Predictive coding models (Friston, 2003; Rao & Ballard, 1999) propose that recurrent cortical feedback shapes the final percept, so that retinal signals are modulated by top-down predictions—explaining context-dependent illusions (Lotto & Purves, 1999).
• Unresolved questions: The precise neural pathway and timecourse for “perceptual filling-in” (such as in boundary completion or color spread) remain debated, with evidence for both early (retinal/LGN) and late (V1/V2/V4) contributions (Komatsu, 2006; Pessoa & De Weerd, 2003).

For advanced artists, the actionable insight is this: Apparent edges, shapes, and lightness emerge from an ensemble of neural computations, not from a blank canvas "gap-filler." Studio techniques exploiting these ensemble effects (Kitaoka’s gaborized images, Lotto’s interactive installations) depend on precise pattern-scale and local contrast, mirroring ganglion- and cortex-scale computations.

Digital Experiment: Retinal Sparse Sampling and Contour Perception

1. Setup: Display a grid of black squares (visual angle < 2°) on mid-grey. Remove every third square along rows and columns to simulate foveal cone mosaic under-sampling. Place a continuous dark bar across the grid, intersecting many gaps.
2. Observation Protocol: View from a fixed arm’s length. Note whether your visual system joins the interrupted bar across the sparse grid, or if discontinuities dominate. Repeat under high and low luminance to probe light-adaptation effects.
3. Controlled Variables: Square size, viewing distance, bar width, luminance levels.
4. Limitations: While the phenomenon mimics retinal sampling limitations, it cannot precisely replicate neural receptive field diversity or dynamic eye movement effects.

This experiment demonstrates that apparent continuity (e.g., an unbroken bar) can survive gross sampling gaps, but the perceived strength and smoothness of the contour depend systematically on bar width, local luminance contrast, and grid scale—reflecting known properties of retinal and cortical integration (Field et al., 1993; Geisler, Perry, Super & Gallogly, 2001).

Retrieval Question

Challenge: Given an image in which edge continuity is ambiguous (e.g., a complex Kitaoka-style geometric arrangement with local interruptions), how would you diagnose whether the effect arises primarily from retinal preprocessing or cortical prediction? Outline what psychophysical and anatomical variables you would manipulate and what outcome would support each theory.

Sources

• Pessoa, L., & De Weerd, P. (2003). Filling-in: From perceptual completion to cortical reorganization.
• Masland, R.H. (2012). The Neuronal Organization of the Retina.
• Komatsu, H. (2006). The neural mechanisms of perceptual filling-in.
• Dumoulin, S.O., & Wandell, B.A. (2008). Population receptive field estimates in human visual cortex.
• Field, G.D. & Chichilnisky, E.J. (2007). Information processing in the primate retina.
• Friston, K. (2003). Learning and inference in the brain.