[EXPERT: CONSTRUCTED EYE] Day 17 — Competing Neural Explanations and Current Evidence

Competing Neural Explanations and Current Evidence

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

Day 17

Demonstration: Two identical ellipses appear differently colored when placed on contrasting backgrounds—a canonical test for neural explanations of contrast illusions.

Expert Objective

Today’s aim is to scrutinize the competing neural models—lateral inhibition, normalization, top-down prediction, and recurrent integration—that underpin our current understanding of complex visual illusions. Advanced artists working with visual phenomena like those in the work of Akiyoshi Kitaoka and Beau Lotto must appreciate how these conceptual frameworks contest and explain psychophysical data, and which explanatory gaps remain unresolved.

Evidence and Competing Explanations

Observed Effects: Illusions such as simultaneous contrast, the checker shadow illusion and chromatic assimilation exhibit clear perceptual shifts not directly traceable to local stimulus energy. Kitaoka’s and Lotto’s demonstrations often combine gradients, edges, and contextual cues to defeat simple low-level causes. Key psychophysical results:

• The magnitude of color or brightness illusion varies with the spatial frequency and arrangement of adjoining stimuli (Kingdom, 2011).
• Neural activity in early visual cortex (V1–V2) correlates with some, but not all, of the reported perceptual modulations (Rossi & Paradiso, 1996; Boyaci et al., 2007).
• fMRI and TMS studies reveal later-stage areas (V4/LOC/IT) are required for full illusory experience, implicating feedback mechanisms (Bouvier & Treisman, 2010).

Supported Mechanisms:

• Lateral inhibition: Explains strong edge-dependent illusions but fails with weak-gradient contexts (Shapley & Enroth-Cugell, 1984).
• Normalization: Population-based models (Carandini & Heeger, 2012) robustly fit psychometric curves but are less effective for high-level context effects.
• Recurrent neural interactions: Supported by latency measurements (Lamme & Roelfsema, 2000), these dynamics can implement both normalization and feedback.
• Top-down modulation: Predictive coding and expectation effects are now directly measurable via cross-temporal decoding (Kok et al., 2016).

Competing Explanations and Unresolved Questions:

• Does normalization alone suffice to explain illusions with strong semantic context (e.g., shadow illusions), or must feedback from object-recognition layers be invoked (Teufel & Fletcher, 2020)?
• Can opponent process models predict complex results (e.g., color assimilation within pattern-rich environments)? Competing evidence suggests standard models cannot account for all effects unless hierarchical or recurrent layers are explicitly modeled (Zaidi et al., 2012).
• What are the temporal windows: Which illusions manifest in the earliest feedforward pass (as measured by EEG/MEG), and which require reentrant processing?
• How do these models account for individual differences in perception, such as in the “#theDress” phenomenon (Lafer-Sousa et al., 2015)?

Classic lateral inhibition test: The central squares are physically identical, yet appear lighter/darker. Observe persistence of the illusion as context varies—a point where simple lateral inhibition is challenged by evidence from normalization and feedback models.

Studio implication: Advanced patterns, like those found in Kitaoka’s “Perpetual Motion” illusions, selectively reveal which low-level models break down first, making them essential for both artistic exploration and neurobiological falsification.

Contextual modulation: Perceived color of the disk shifts under complex gradients in ways basic lateral inhibition cannot explain, demanding higher-order or recurrent models.

Digital Experiment

1. Stimulus: Use the above SVGs or generate identically coded patches with known RGB values.
2. Procedure: Slowly sweep your gaze horizontally while focusing on the central element. Observe any transient or persistent changes in brightness or color.
3. Control: Keep ambient light and monitor settings fixed for all trials.
4. Variables: Change only the arrangement/contrast of adjoining stripes or backgrounds. Avoid image compression.
5. Limitations: This self-experiment cannot distinguish feedforward from feedback effects. Neural mechanisms inferred remain hypothetical, as only behavioral outcomes are observed.

Retrieval Question

Summarize two pieces of empirical evidence that challenge the sufficiency of lateral inhibition as a sole mechanism for visual illusions in contemporary vision science.

Sources

• Carandini, M. & Heeger, D. J. (2012). Normalization as a canonical neural computation. Nature Reviews Neuroscience.
• Kingdom, F. A. A. (2011). Lightness, brightness and transparency: Different processes but common mechanisms? Vision Research.
• Rossi, A. F., & Paradiso, M. A. (1996). Neural correlates of perceived brightness in the retina and primary visual cortex. Science.
• Lamme, V. A. F. & Roelfsema, P. R. (2000). The distinct modes of vision offered by feedforward and recurrent processing. Trends in Neurosciences.
• Zaidi, Q., et al. (2012). Neural Loci of Color Afterimages. Vision Research.
• Kok, P., et al. (2016). Selective Activation of the Deep Layers of the Human Primary Visual Cortex by Top-Down Feedback. Current Biology.
• Teufel, C., & Fletcher, P. C. (2020). Forms of prediction in the nervous system. Trends in Cognitive Sciences.
• Lafer-Sousa, R., et al. (2015). Individual differences in color perception revealed by The Dress photograph. Current Biology.