Blank space, soft edges and broken lines do more cognitive work than many neon overlays ever will. The visual cortex is built as a prediction engine, not a camera; it constantly infers the most likely scene from fragmentary input, then back‑fills detail. When an image withholds structure at the right spots, that inference machinery spins up, generating a vivid, internally authored version of reality.
Neuroscience describes this as predictive coding and Bayesian inference: higher cortical areas send expectations downward while raw signals travel upward. When an artwork offers strong cues to edges or light but removes full contours, those top‑down priors dominate, creating illusory contours and phantom textures. Gestalt perception laws, such as closure and good continuation, emerge from this circuitry, pushing the system to complete shapes, faces and depth even where no pixels exist.
Paradoxically, adding more effects can reduce this impact by resolving ambiguity and lowering sensory uncertainty, leaving less room for internal hallucination. The most striking images operate closer to an entropy sweet spot: enough noise to invite inference, enough structure to stabilize it. That balance turns a static file into an active perceptual negotiation, where the viewer’s own neural activity becomes the final brushstroke.
