How the Brain Trades Accuracy for Simplicity — and What Happens When It Can't

The Hollow Mask Illusion — in which a concave face is irresistibly perceived as convex — is one of the most robust demonstrations that perception is fundamentally inferential. The brain resolves ambiguous visual input by defaulting to the hypothesis with the highest prior probability: faces are convex, lighting comes from above. This Bayesian prior is so deeply embedded that explicit knowledge of the trick cannot override it. The illusion, far from being a failure, reveals the brain's core operating principle: minimize surprise by favoring simple, well-tested generative models over literal sensory fidelity.

The clinical significance emerges when this balance is disrupted. Individuals with schizophrenia frequently resist the Hollow Mask Illusion, perceiving the concavity accurately. Within a predictive processing framework, this suggests attenuated prior precision — sensory prediction errors are afforded excessive weight relative to top-down expectations. The perceptual system becomes more accurate in narrow terms but loses the contextual Scaffolding that makes experience coherent. Extended to severe autism, the same logic implies a system overwhelmed by unfiltered sensory detail, where the failure to deploy simplifying abstractions produces a world that is relentlessly unpredictable and demanding.

This reframes the Accuracy-Complexity Tradeoff — known formally as model evidence or free energy minimization — as something far more than an optimization problem. It becomes a characterization of mental health itself. A mind that over-regularizes hallucinates meaning where there is none. A mind that under-regularizes drowns in data. The functional mind lives in the tension between these extremes, and the difference between adaptive perception and clinical suffering may reduce to how precision is allocated across the hierarchy of prediction.