Foundations of Visual Mesmerization in Real-Time 3D Graphics

Full report

On this page

The capacity of real-time 3D graphics to evoke visually mesmerizing, trance-like, or deeply absorbing experiences relies on the precise, calculated manipulation of human perceptual systems. Colloquially, terms such as "hypnotic" or "mesmerizing" describe states of profound visual fascination and sustained attentional absorption. Scientifically, however, these phenomena emerge from a confluence of neurological adaptations, sensory entrainment, and cognitive flow states. It is critical to distinguish this sustained attentional capture from clinically defined hypnosis; there is no empirical evidence to suggest that viewing a shader or 3D geometry induces true clinical hypnosis, which involves heightened suggestibility and altered states of consciousness orchestrated by a trained clinician. Instead, digital graphics induce states of hyper-focus, motion aftereffects, and visual flow by exploiting the biological hardwiring of the retina, the visual cortex, and higher-order predictive processing networks.

This investigation establishes a comprehensive framework for engineering visual absorption in interactive digital environments. By deconstructing the perceptual mechanisms underlying spirals, tunnels, infinite recursion, and non-Euclidean geometry, this analysis delineates what is neurobiologically established, what is experimentally plausible within cognitive science, what remains speculative, and what is merely artistic convention. Furthermore, it provides a rigorous technical analysis of how these perceptual exploits can be mathematically formulated and optimally rendered using modern graphics APIs such as WebGL and WebGPU.

Scientific Foundations of Visual Absorption

To engineer visual absorption through procedural graphics, one must first deconstruct the biological pathways that process geometry, motion, and top-down cognitive prediction. The human visual system is not a passive camera; it is an active inference engine driven by evolutionary imperatives to detect motion, estimate depth, and recognize patterns.

Retinocortical Mapping and Klüver Form Constants

During the 1920s, the neurologist Heinrich Klüver systematically documented recurring geometric patterns experienced by individuals during altered states of consciousness, classifying them into four fundamental "form constants": (I) tunnels and funnels, (II) spirals, (III) lattices and checkerboards, and (IV) cobwebs1. For decades, these forms were considered arbitrary hallucinations. However, research by Jack Cowan, Paul Bressloff, and colleagues demonstrated that these geometric structures are a direct mathematical consequence of the neurological architecture of the primary visual cortex (V1)4.

The topological transformation from the retina's visual field coordinates to the cortical surface in V1 can be approximated by a complex logarithmic mapping5. Let [Figure omitted from source export] and [Figure omitted from source export] represent the polar coordinates (eccentricity and azimuth) of a stimulus in the visual field. The mapping to the Cartesian coordinates [Figure omitted from source export] of the V1 cortical sheet is described by the equation:

[Figure omitted from source export]

Where the complex representation of the visual field is [Figure omitted from source export], and the cortical representation is [Figure omitted from source export]7. Expanding this formulation yields:

[Figure omitted from source export]

Under this tangent map of the complex logarithm, concentric circles and radial rays in the visual field become horizontal and vertical parallel stripes of neural activity in V14. Logarithmic spirals, which can be described by the equation [Figure omitted from source export], are transformed into oblique, parallel lines of constant slope in the cortex1.

The visual mesmerization associated with spirals, tunnels, and concentric patterns—often treated merely as an artistic convention for "hypnosis"—is in fact a well-established neurobiological phenomenon. These geometries are visually compelling because they directly mirror the functional architecture of V1 Hubel-Wiesel hypercolumns9. Viewing these geometries overstimulates the spatial frequency preferences of the cortex. Through a neural Turing mechanism—where localized excitation and lateral inhibition interact—the brain generates spontaneous pattern formation when its resting state becomes unstable4. Graphics that utilize these specific mathematical forms capture sustained attention by efficiently saturating the brain's baseline pattern-recognition circuitry, creating a resonant feedback loop1.

Motion-Sensitive Visual Pathways and Adaptation

Sustained visual absorption is heavily driven by motion adaptation within the middle temporal visual area (MT, also known as V5)12. Area MT is highly specialized for detecting visual motion and contains direction-selective neurons that map the velocity and trajectory of visual stimuli14. When a viewer fixates on a continuously moving stimulus—such as an expanding digital tunnel or a rotating spiral—these direction-specific neurons undergo sensory adaptation, reducing their firing rate over time16.

If the on-screen motion abruptly stops, the unstimulated neurons tuned to the opposite direction of motion temporarily dominate the neural baseline. This creates the Motion Aftereffect (MAE), commonly known as the waterfall illusion, where static images appear to contract, expand, or rotate in the opposite direction of the original stimulus13. Functional MRI (fMRI) studies confirm that area MT shows elevated activity during the perception of illusory motion in static stimuli following adaptation, proving that this effect is deeply physiological rather than purely psychological13. The compulsion of mesmerizing graphics often relies on maintaining motion just long enough to edge the viewer toward adaptation, creating a state of sensory entrainment where looking away from the screen results in the physical environment appearing to warp.

The depth of this adaptation can be manipulated through contrast and color. Visual motion relies heavily on the magnocellular pathway, which is highly sensitive to luminance contrast but color-blind16. If a graphics programmer utilizes isoluminant colors—colors with vastly different hues but mathematically identical brightness—the motion-sensitive magnocellular pathway is starved of data. The brain is forced to process the motion through the parvocellular pathway, which is specialized for color and fine detail but poor at motion tracking. This creates an eerie, sluggish, or "floating" perception of motion, a phenomenon extensively documented by perceptual psychologists16.

Predictive Processing, Optic Flow, and Time-to-Contact

The brain does not passively receive sensory data; it acts as an active inference machine. According to the predictive coding framework, the brain constantly generates top-down predictions about the sensory environment and updates its internal models based on prediction errors—the difference between expected and actual sensory input19. Within Karl Friston's free-energy principle, biological systems fundamentally seek to minimize these prediction errors, a concept mathematically formalized as variational free energy19.

Optic flow—the apparent motion of the visual scene caused by the relative motion between an observer and the environment—is a critical component of predictive processing. David Lee's tau ([Figure omitted from source export]) hypothesis postulates that the visual system directly calculates the time-to-contact (TTC) with approaching objects using the relative rate of expansion of the object's retinal image22. The optical variable [Figure omitted from source export] is defined as:

[Figure omitted from source export]

Where [Figure omitted from source export] is the angular size of the object and [Figure omitted from source export] is its rate of expansion22.

When 3D graphics simulate continuous forward motion through a tunnel or a dense starfield, they artificially trigger the [Figure omitted from source export] variable. Because the observer is physically stationary, a minor sensory conflict occurs25. However, the brain's predictive models become deeply absorbed in continually processing the simulated TTC to avoid a perceived collision26. Geometric visual illusions constructed from these mechanics exploit the predictive processing architecture by generating unresolvable prediction errors. The brain predicts motion, depth, or collision where none physically exists, holding the viewer in a loop of continual, unresolved perceptual updating27.

It is important to note the nuanced debate within the scientific community regarding predictive coding. While proponents argue that perception is effectively a "controlled hallucination" aimed at minimizing surprise19, dissenting researchers suggest that what is labeled as "predictive coding" may sometimes just represent classical repetition suppression or efficient data compression30. Furthermore, some researchers argue that visual illusions persist even when top-down explicit knowledge confirms they are illusions, suggesting that low-level sensory prediction errors can override high-level cognitive priors28. Regardless of the underlying ideological debate, engineering graphics that force the visual cortex to rapidly test and fail its spatial predictions reliably induces profound visual absorption.

Artistic Precedents: Op Art and Perceptual Psychology

Long before real-time computer graphics, the Optical Art (Op Art) movement of the 1960s manipulated human optics to create sensations of movement, vibration, and depth on static two-dimensional canvases. Pioneers such as Bridget Riley and Victor Vasarely engineered high-contrast geometric works that effectively short-circuited the visual cortex, laying the aesthetic and physiological groundwork for modern procedural shaders33.

Op Art functions at the intersection of geometric design and neurobiology. One primary mechanism is the exploitation of microsaccades—rapid, involuntary eye twitches that occur during visual fixation35. When viewing high-contrast, repetitive grids or undulating lines, microsaccades cause the sharp black-and-white boundaries to shift constantly across adjacent retinal photoreceptors33. This rapid shifting produces phantom motion signals, interpreted by the visual cortex as shimmering, vibration, or rolling waves36.

Furthermore, lateral inhibition—a process where stimulated retinal cells inhibit the activity of neighboring cells to sharpen edge contrast—creates illusory dark and light spots and enhances the perceived depth of 3D forms encoded in 2D plains36. Vasarely's Vega series utilized convex and concave distortions within rigid grids to simulate pulsating, three-dimensional spheres bursting from flat surfaces34. Riley's exploration of Gestalt grouping principles allowed her to arrange discrete, high-contrast units in ways that the brain automatically grouped into moving waves37. In modern 3D graphics, these exact principles are translated into shader programming. High-frequency contrast, moiré patterns, and repetitive geometric structures in digital displays directly inherit the perceptual legacy of Op Art, upgrading it with true kinetic temporal dynamics.

Graphics Techniques and Perceptual Mechanisms

To construct a visually absorbing experience, 3D graphics developers utilize a specific vocabulary of visual mechanisms. The efficacy of these mechanisms depends heavily on implementation methodology, field of view, color theory, and temporal dynamics.

Spirals, Tunnels, and Continuous Zooms

The visual experience of a continuous zoom or spiral is one of a hypnotic, infinite descent into a central void, or conversely, a perpetual outward expansion. As established by the retinocortical map, tunnels and spirals map directly to straight, parallel lines in the visual cortex, deeply resonating with the brain's foundational edge-detection architecture1. Furthermore, they artificially stimulate optic flow and the [Figure omitted from source export] variable, continuously engaging the brain's spatial navigation and collision-prediction systems5.

This effect is most potent when it encompasses the full field of view (FOV). Central fixation anchors the viewer's focus on the vanishing point, while the peripheral motion dictates the sensation of vection—the powerful illusion of self-motion. Movement speed is a critical modulating factor. Low to moderate speeds are highly absorbing, mimicking the relaxed flow of a hypnagogic state and favoring the onset of the motion aftereffect14. Conversely, high speeds generate a sense of urgency or alarm due to a rapidly decreasing time-to-contact, shifting the brain from a state of trance to a state of high physiological arousal24.

Contrast and color drastically alter the effect. High-contrast alternating bands maximize the firing rate of edge-detecting neurons36. Utilizing isoluminant colors shifts the processing burden, creating the aforementioned floating motion18. While continuous zooms can be implemented using long tube meshes or sequenced geometry, they are vastly superior when executed as screen-space shaders or via raymarching. In a fragment shader, one can convert Cartesian screen coordinates [Figure omitted from source export] to polar coordinates [Figure omitted from source export] to construct continuous zooms and logarithmic spirals using simple modulo arithmetic. This approach carries a very low performance cost, making it highly viable for mobile and browser limitations. However, developers must be wary of photosensitivity concerns; high-contrast spirals rotating at specific frequencies can trigger photosensitive epileptic responses.

Radial Flow, Rotational Optic Flow, and Looming

When viewing radial flow, the observer sees particles, vertices, or geometric patterns expanding radially from a center point, heavily simulating the view of a spacecraft moving at warp speed or a rapid descent through a debris field. Looming taps into ancient evolutionary threat-detection and navigation systems; the radial expansion of optic flow serves as the primary visual cue for forward locomotion39.

Because peripheral vision is highly sensitive to motion and essential for perceiving radial flow, this mechanism demands a wide FOV. Narrowing the FOV breaks the illusion of vection and reduces the effect to a mere 2D animation. As with tunnels, the speed of radial expansion dictates the time-to-contact variables22. Fast looming triggers an involuntary flinch or bracing response, while slow, rotational optic flow induces a sense of drifting.

The optimal implementation for radial flow is a massive particle system operating entirely on the GPU via compute shaders. If mathematical textures are preferred over discrete particles, log-polar mappings are utilized within a fragment shader. If [Figure omitted from source export] are the screen coordinates centered at [Figure omitted from source export], the log-polar transformation is calculated as:

[Figure omitted from source export]

[Figure omitted from source export]

Rendering a simple grid using the transformed coordinates [Figure omitted from source export] instantly produces a flawless infinite zooming tunnel or radial flow field. In WebGL, computing millions of radial particles incurs a massive performance cost due to CPU-bound vertex updates, severely limiting mobile performance. WebGPU compute shaders resolve this bottleneck. Safety constraints for looming involve Visually Induced Motion Sickness (VIMS), as the intense peripheral flow creates severe sensory conflict with the static vestibular system25.

Recursive Geometry, Infinite Repetition, and Symmetry

Recursive geometry presents the viewer with fractalline structures, repeating monoliths, or infinite grids that stretch endlessly toward a fog-covered horizon. Symmetrical and repeating patterns provide the brain with highly predictable sensory data. In the predictive processing framework, highly predictable environments reduce computational load, allowing the observer to enter a relaxed, trance-like flow state where prediction errors are minimized and cognitive friction drops19.

This effect requires a deep depth of field and is highly effective across the full FOV. Movement through infinite repetition must be slow and deliberate; rapid movement through dense lattices causes visual noise and destroys the structural predictability that makes the effect compelling. High-contrast lighting with deep, soft shadows enhances the perceived vastness of the space.

Implementing infinite repetition through standard rasterized geometry is computationally prohibitive, as it requires instantiating millions of polygons. Raymarching with Signed Distance Functions (SDFs) is the definitive, industry-standard method for this mechanism40. Instead of instantiating distinct objects, the entire infinite space is represented by mathematically folding the spatial domain. In GLSL or WGSL, the domain repetition function uses a simple modulo operator to fold space around a period [Figure omitted from source export]:

 

 

 

OpenGL Shading Language

vec3 opRep(vec3 p, vec3 c) {     return mod(p \+ 0.5 \ c, c) \- 0.5 \ c; }

When this modified coordinate is passed into a basic SDF—such as a sphere or box—the rendering engine evaluates an infinite lattice of that geometry simultaneously42. While memory-efficient, raymarching heavy recursive scenes carries a high per-pixel computational cost, which can lead to thermal throttling on mobile browsers. However, it carries very low risks regarding VIMS or photosensitivity, provided the geometry does not flicker.

Pseudo-Depth, Stereoscopic Cues, and Figure-Ground Ambiguity

Viewers exposed to pseudo-depth see two-dimensional patterns that appear to contain massive three-dimensional depth, or alternating shapes that flip between foreground and background. The visual system constructs depth using binocular disparity, but in standard browser environments lacking VR headsets, it relies heavily on monocular depth cues and Gestalt grouping22. Conflicting cues—such as a figure-ground ambiguity where the foreground and background are interchangeable—force the visual cortex to rapidly alternate between two stable neural states, creating intense, sustained visual fascination28.

This effect works exceptionally well in central vision, where visual acuity is highest. Movement speed should be static or glacial, allowing the viewer's brain time to resolve and re-resolve the optical conflict. Color and contrast are paramount; depth can be artificially induced by rendering "deeper" layers with atmospheric perspective (lower contrast, darker colors) and "shallower" layers with high contrast and brightness.

These effects are best implemented using fragment shaders and post-processing stacks. Sinusoidal warping applied to UV coordinates can create undulating, breathing surfaces that mimic depth without requiring a z-buffer. The performance cost is generally very low, making it ideal for mobile devices, and it presents minimal safety risks aside from minor ocular fatigue from prolonged staring.

Motion Parallax and Focal Anchoring

Motion parallax occurs when the viewer perceives objects closer to the camera moving faster than objects in the distance, providing a profound sense of scale and immersion. A focal anchor—such as a static crosshair or a slow-moving central object—gives the fovea a resting point while the peripheral motion dictates the speed of the environment. This is compelling because it perfectly mimics real-world spatial navigation, engaging the brain's proprioceptive predictions20.

To maintain a mesmerized state, the camera's velocity and rotation must be perfectly smooth. Jerky or erratic camera movements immediately break the flow state and induce VIMS. This is mathematically achieved using Catmull-Rom splines for the camera path, combined with a Frenet-Serret frame to calculate the camera's orientation44. A Catmull-Rom spline guarantees that the camera passes smoothly through a defined set of control points without sudden changes in acceleration44.

The Frenet-Serret frame ensures the camera correctly banks and tilts into curves. For a parametric curve [Figure omitted from source export], the Tangent ([Figure omitted from source export]), Normal ([Figure omitted from source export]), and Binormal ([Figure omitted from source export]) vectors are calculated as:

[Figure omitted from source export]

By strictly aligning the camera's forward vector with [Figure omitted from source export], the up vector with [Figure omitted from source export], and the right vector with [Figure omitted from source export], the viewer experiences seamless, rollercoaster-like navigation45. This technique relies heavily on traditional geometry and rasterization. The performance cost varies depending on the scene's polygon count. The primary limitation is the high potential for VIMS if the spline contains overly tight radii or rapid rotational shifts25.

Impossible and Non-Euclidean Geometry

Impossible geometry presents architecture that defies physical reality, such as spaces that are larger on the inside, portals seamlessly connecting distant locations, or hyperbolic tilings (reminiscent of M.C. Escher's limits). This is compelling because it completely shatters the brain's highly tuned predictive models28. Because the geometry violates Euclidean rules, the brain cannot resolve the spatial layout, forcing continuous, hyper-focused cognitive engagement as the observer attempts to make sense of the visual data.

This effect demands a full FOV to allow the observer to look around corners that shouldn't exist. Movement should be slow, allowing the cognitive dissonance to build. Raymarching excels at non-Euclidean rendering43. By using the Poincaré disk model or Beltrami-Klein models, developers can simulate hyperbolic space directly in a shader46. In hyperbolic space, angles remain conformal (true to Euclidean perception), but parallel lines diverge and distances compress exponentially near the boundary46. Raymarching in this space requires stepping the ray not along straight Euclidean vectors, but along hyperbolic geodesics (circular arcs)46.

The performance cost of hyperbolic raymarching is extraordinarily high. The geodesic math requires complex matrix transformations at every step of the ray, restricting this technique primarily to high-end desktop hardware and making it largely unviable for mobile browsers without severe resolution downscaling. It also carries a risk of cognitive fatigue and spatial disorientation.

Mathematical Formulations for Procedural Rendering

Beyond domain repetition and non-Euclidean geodesics, achieving a hypnotic, organic quality in procedural graphics relies heavily on blending mathematics. Hard geometric intersections break the illusion of a fluid, mesmerizing space. This is resolved using a "smooth minimum" function (smin) applied to SDFs, a technique famously refined and popularized by Iñigo Quílez50.

A polynomial smooth minimum blends two distance values, [Figure omitted from source export] and [Figure omitted from source export], over a configurable blending factor, [Figure omitted from source export]. The GLSL implementation is as follows:

 

 

 

OpenGL Shading Language

float smin( float a, float b, float k ) {     float h \= clamp( 0.5 \+ 0.5 \ (b \- a) / k, 0.0, 1.0 );     return mix( b, a, h ) \- k \ h \* (1.0 \- h); }

This single mathematical operation transforms harsh, clipping geometric intersections into mesmerizing, continuous organic forms that appear to melt, fuse, and separate like cellular structures or liquid metal50.

Furthermore, to create rhythmic, breathing animations that drive visual entrainment, developers utilize polynomial impulse functions rather than standard sinusoidal waves. An exponential impulse peaks quickly and decays slowly, providing a more natural, heartbeat-like rhythm to visual expansion and contraction:

 

 

 

OpenGL Shading Language

float expImpulse( float x, float k ) {     float h \= k \ x;     return h \ exp(1.0 \- h); }

By substituting raw trigonometric time variables with specialized impulse functions, the motion of the graphic aligns more closely with biological rhythms, enhancing the sense of deep absorption50.

Implementation Implications: WebGL vs. WebGPU

Delivering these complex mathematical mechanisms seamlessly in a browser requires strict performance optimization. A stuttering frame rate destroys visual absorption, snapping the viewer out of the flow state and disrupting the [Figure omitted from source export] variable optic flow calculations. The industry is currently transitioning from WebGL to WebGPU, representing a paradigm shift in how these mesmerizing effects are engineered53.

WebGL, based on the aging OpenGL ES standard, relies on a global state machine53. Every change to a texture, shader, or buffer requires synchronous CPU-side commands to bind and unbind resources. This leads to massive CPU overhead and single-threaded bottlenecks54. For simple raymarching and SDF rendering—which push heavily on the fragment shader while leaving the vertex shader largely idle—WebGL performs adequately because the geometry is usually a single fullscreen quad56.

However, WebGL lacks native Compute Shaders55. If a developer wishes to render one million radially expanding particles to simulate an overwhelming looming effect, WebGL forces the CPU to calculate the physics and upload the vertex data every frame. This throttles the system and destroys the illusion54.

WebGPU is a low-level, stateless API inspired by Vulkan and Metal53. It replaces global state mutations with immutable Pipeline Objects and Bind Groups, drastically reducing API overhead and allowing multi-threaded command recording55. More importantly, WebGPU introduces Compute Shaders via the WebGPU Shading Language (WGSL)55. Compute shaders allow developers to leverage the GPU for general-purpose parallel processing55.

A radial optic flow simulation of 100,000 particles that takes 30ms per frame on the CPU under WebGL can be processed in under 2ms using a WebGPU compute shader—a 150x improvement in performance and scale54. For complex raymarching scenes, WebGPU compute shaders can pre-calculate spatial data or simulate 3D noise fields into a storage buffer, which the fragment shader then samples. This alleviates the heavy for-loop iteration constraints often encountered on specific hardware architectures. For instance, benchmarking reveals that while older hardware like Intel UHD Graphics Gen9 may still perform faster on highly optimized WebGL rasterization, modern architectures (such as Apple M2 or AMD Radeon GCN5) see massive frame rate improvements and draw-call reductions when transitioning to WebGPU59. Furthermore, for implementations attempting to use neural networks to drive procedural generation in real-time, WebGPU delivers 3x to 8x faster GEMM (General Matrix Multiply) operations for large matrices compared to WebGL59.

Safety Constraints and Accessibility

Manipulating human optics to induce trance-like states carries distinct physiological risks that must be actively mitigated by developers.

Photosensitive Epilepsy (PSE): High-contrast geometries that strobe or move rapidly can trigger seizures in susceptible individuals. The critical flicker fusion rate spans from approximately 3 Hz to 30 Hz. Continuous zooms, infinite patterns, or Op Art-inspired moiré grids must not alternate between high-contrast light and dark states at these frequencies. Implementing contrast reduction (lowering pure white to a gray value) and restricting the spatial frequency of repeating patterns are mandatory accessible design practices.

Visually Induced Motion Sickness (VIMS): Generating powerful optic flow while the user is stationary creates a profound sensory conflict. The visual system perceives forward motion through the [Figure omitted from source export] variable, while the vestibular system in the inner ear perceives stasis25. To minimize VIMS, developers must:

1. Provide a static focal anchor in the center of the screen (such as a reticle, vignette, or a static character) to give the fovea a resting point that aligns with the vestibular system's reality.

2. Avoid uncontrolled rotational optic flow. Forward and backward translation is much better tolerated by the human nervous system than continuous rolling on the z-axis.

3. Ensure a highly stable frame rate. Dropped frames disrupt the visual cortex's predictive mechanisms and rapidly induce nausea.

Prioritized Recommendations

The following table categorizes the core techniques for generating visual mesmerization, ranked by their overall utility, cost, and stability in real-time browser environments.

The creation of visually mesmerizing, attention-capturing 3D graphics is not an act of magic, but a precise engineering discipline operating at the nexus of mathematics, rendering technology, and neurobiology. By understanding how the retinocortical map transforms raw photons into neural stripes, how the MT/V5 area adapts to motion to create aftereffects, and how the brain minimizes predictive errors when faced with infinite repetition, developers can author procedural experiences that seize and hold human attention. Translating these biological vulnerabilities into real-time experiences requires advanced graphics programming. Techniques such as Signed Distance Functions with smooth minimums, log-polar space folding, and Catmull-Rom camera framing provide the necessary geometric foundation. Simultaneously, the architectural shift from WebGL to WebGPU ensures that the immense computational overhead of these mathematical illusions does not compromise the smooth frame rates required to maintain the psychological flow state. Ultimately, visual mesmerization in 3D graphics is achieved when computational perfection seamlessly aligns with the limitations and expectations of human perception.

Works cited

1. What geometric visual hallucinations tell us about the visual cortex, https://scispace.com/pdf/what-geometric-visual-hallucinations-tell-us-about-the-sk8cffonp5.pdf

2. Form constant \- Wikipedia, https://en.wikipedia.org/wiki/Form\_constant

3. Form constant \- bionity.com, https://www.bionity.com/en/encyclopedia/Form+constant.html

4. A Math Theory for Why People Hallucinate \- Quanta Magazine, https://www.quantamagazine.org/a-math-theory-for-why-people-hallucinate-20180730/

5. User:Jack D. Cowan/Proposed/Models of visual hallucinations, http://www.scholarpedia.org/article/User:Jack\_D.\_Cowan/Proposed/Models\_of\_visual\_hallucinations

6. He and they approached hallucinations mathematically, https://improbable.com/2014/07/27/he-and-they-approached-hallucinations-mathematically/

7. (A) The retino-cortical map generated by the complex logarithm. (B), https://www.researchgate.net/figure/A-The-retino-cortical-map-generated-by-the-complex-logarithm-B-Action-of-rotations\_fig7\_8883777

8. What Geometric Visual Hallucinations Tell Us about the Visual Cortex, https://case.edu/artsci/math/thomas/BressloffCowanGolubitskyThomasWiener2002NeuralComp.pdf

9. Geometric visual hallucinations, Euclidean symmetry and the ... \- PMC, https://pmc.ncbi.nlm.nih.gov/articles/PMC1088430/

10. Geometric visual hallucinations, Euclidean symmetry and ... \- PubMed, https://pubmed.ncbi.nlm.nih.gov/11316482/

11. Action of the retinocortical map on the funnel form constant. (a, https://www.researchgate.net/figure/Action-of-the-retinocortical-map-on-the-funnel-form-constant-a-Image-in-the-visual\_fig1\_12018383

12. Visual motion aftereffect in human cortical area MT revealed by, https://nmr.mgh.harvard.edu/node/3239

13. Visual motion aftereffect in human cortical area MT ... \- PubMed, https://pubmed.ncbi.nlm.nih.gov/7753168/

14. Visual motion aftereffect in human cortical area MT revealed by, https://www.researchgate.net/publication/15446360\_Visual\_motion\_aftereffect\_in\_human\_cortical\_area\_MT\_revealed\_by\_functional\_magnetic\_resonance\_imaging

15. NIH Public Access \- Computer Science and Engineering, https://cse.msu.edu/ei/seminarreadings/ArashSS09\_2.pdf

16. Functional Analysis of Human MT and Related Visual Cortical Areas, https://www.hms.harvard.edu/bss/neuro/bornlab/lab/papers/tootell-etal-born-mtfmri-jns1995.pdf

17. The motion aftereffect \- PubMed, https://pubmed.ncbi.nlm.nih.gov/21227087/

18. Interaction of motion and color in the visual pathways, https://www.allpsych.uni-giessen.de/karl/publications/Gegenfurtner\&Hawken\_1996.pdf

19. Predictive Coding and the Predictive Brain: A Deep Dive \- People-Shift, https://people-shift.com/articles/predictive-coding-and-the-predictive-brain-a-deep-dive/

20. Predictive coding \- Wikipedia, https://en.wikipedia.org/wiki/Predictive\_coding

21. Predictive coding under the free-energy principle \- FIL | UCL, https://www.fil.ion.ucl.ac.uk/\~karl/Predictive%20coding%20under%20the%20free-energy%20principle.pdf

22. Visual processing of the impending collision of a looming object, https://jov.arvojournals.org/article.aspx?articleid=2121255

23. Lee's 1976 paper \- ResearchGate, https://www.researchgate.net/profile/David-Lee-177/publication/236849987\_Tau\_in\_Action\_in\_Development/links/64524160809a53502147df64/Tau-in-Action-in-Development.pdf

24. Unifying Time to Contact Estimation and Collision Avoidance across, https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1002625

25. VR Motion Sickness Explained: Causes, Fixes, and How VIVE Helps, https://blog.vive.com/us/vr-motion-sickness-explained-causes-fixes-and-how-vive-helps/

26. Using Time-to-Contact Information to Assess Potential Collision, https://pmc.ncbi.nlm.nih.gov/articles/PMC2572212/

27. Why visual perception is a decision process \- RUB Newsportal, https://news.rub.de/english/press-releases/2020-05-12-neuroscience-why-visual-perception-decision-process

28. Motion Illusions Generated Using Predictive Neural Networks ... \- arXiv, https://arxiv.org/html/2112.13243v2

29. How Your Brain Predicts Reality | Flow, Intuition & Mental Flexibility, https://www.youtube.com/watch?v=CRqIS7vfIm0

30. A brief critique of predictive coding | Romain Brette, https://romainbrette.fr/a-brief-critique-of-predictive-coding/

31. A Long and Winding Road towards and away from Predictive Coding, https://communities.springernature.com/posts/a-long-and-winding-road-towards-and-away-from-predictive-coding

32. Predictive Processing and Some Disillusions about Illusions, https://www.researchgate.net/publication/356506727\_Predictive\_Processing\_and\_Some\_Disillusions\_about\_Illusions

33. Bridget Riley: The Female Artist Who Creates Optical Illusions, https://www.thecollector.com/bridget-riley-op-art-optical-illusions/

34. 6.7: Op Art (1960s – 1970s) \- Humanities LibreTexts, https://human.libretexts.org/Bookshelves/Art/Art\_History\_and\_Theory/A\_World\_Perspective\_of\_Art\_History%3A\_1400CE\_to\_the\_21st\_Century\_(Gustlin\_and\_Gustlin)/06%3A\_The\_Art\_of\_Engagement\_(1940-1970)/6.07%3A\_Op\_Art\_(1960s\_\_1970s)

35. The Neuroscience of Op Art | Artsy, https://www.artsy.net/article/the-art-genome-project-the-neuroscience-of-op-art

36. Op Art: History, Illusions, and Key Artists | GeoPatternLab, https://geopatternlab.com/blog/op-art-history-illusions-and-key-artists

37. Bridget Riley and the Birth of Op Art \- Unit London, https://unitlondon.com/artists/bridget-riley/

38. Bridget Riley and Op Art | Art UK, https://artuk.org/learn/learning-resources/bridget-riley-and-op-art

39. Time-to-contact judgment in the locomotion of adults and preschool, https://pubmed.ncbi.nlm.nih.gov/8228839/

40. An Introduction to Ray Marching | Michael Crum | Portfolio, https://michael-crum.com/raymarching/

41. SDF Raymarching, https://zone.dog/braindump/sdf\_marching/

42. Ray marching \- Wikipedia, https://en.wikipedia.org/wiki/Ray\_marching

43. Immersive Visualization of Flat Surfaces Using Ray Marching \- arXiv, https://arxiv.org/html/2506.07558v1

44. Catmull-Rom Spline \- Lighthouse3d.com, https://www.lighthouse3d.com/tutorials/maths/catmull-rom-spline/

45. Lab 1 \- Cubic Spline Curves, https://people.engr.tamu.edu/sueda/courses/CSCE489/2020F/labs/L01/index.html

46. Poincaré Hyperbolic Disk \-- from Wolfram MathWorld, https://mathworld.wolfram.com/PoincareHyperbolicDisk.html

47. Hyperbolic Geometry \- EscherMath, https://eschermath.org/wiki/Hyperbolic\_Geometry.html

48. The Poincaré Models of the Hyperbolic Plane Plane, https://new.math.uiuc.edu/public402/models/conformal.pdf

49. Virtual Reality, http://www.segerman.org/XR.html

50. Inigo Quilez \- memalign, https://memalign.github.io/m/iq-ulf/functions.htm

51. glslify/glsl-smooth-min: Smooth minimum functions for GLSL \- GitHub, https://github.com/glslify/glsl-smooth-min

52. Painting with Math: A Gentle Study of Raymarching, https://blog.maximeheckel.com/posts/painting-with-math-a-gentle-study-of-raymarching/

53. WebGPU vs. WebGL: What are the main differences? \- Design4Real, https://design4real.de/en/webgpu-vs-webgl/

54. WebGL vs. WebGPU Explained \- Three.js Roadmap, https://threejsroadmap.com/blog/webgl-vs-webgpu-explained

55. From WebGL to WebGPU | Chrome for Developers, https://developer.chrome.com/docs/web-platform/webgpu/from-webgl-to-webgpu

56. Ray Marching and Signed Distance Fields, https://bentonian.com/Lectures/FGraphics1819/1.%20Ray%20Marching%20and%20Signed%20Distance%20Fields.pdf

57. WebGL vs. WebGPU \- FenixFox®Studios, https://fenixfox-studios.com/content/webgpu\_vs\_webgl/

58. WebGPU vs WebGL: Why the Industry Is Moving On, https://blog.openreplay.com/webgpu-vs-webgl-industry-moving/

59. WebGPU vs. WebGL: Performance Benchmarks for Client-Side, https://daily.dev/posts/webgpu-vs-webgl-performance-benchmarks-for-client-side-inference-bxfaw5zzt

60. We tested WebGPU vs WebGL in a real Three.js scene. The ... \- Reddit, https://www.reddit.com/r/threejs/comments/1vbj3ca/we\_tested\_webgpu\_vs\_webgl\_in\_a\_real\_threejs\_scene/