The Golden Ratio in Neuroaesthetics: From a Fragile Universal to a Compressible Form

Beauty is compressibility: the golden ratio isn't a magic number, it's just one easy-to-describe shape among many that a brain tuned to find short descriptions naturally likes.

The paper starts by taking seriously a claim that has floated around since the Renaissance — that the golden ratio φ ≈ 1.618 is a privileged aesthetic constant, hardwired into human (and maybe animal) perception. Then it systematically dismantles that claim using the empirical literature. Behavioral studies, starting with Fechner in the 1870s and running through careful modern replications, show that rectangle preferences cluster broadly across aspect ratios from 1.3 to 1.7. φ sits in that range, but so do √2, 3:2, and 5:3. No methodology has cleanly isolated φ as a unique attractor. Neuroaesthetic imaging studies (fMRI, MEG) find that canonical, harmonically proportioned forms activate occipito-temporal and parietal cortex differently from distorted forms — but the manipulation is always "canonical vs. scrambled," never "φ vs. other simple ratio." The neural signal tracks compressible regularity, not φ specifically.

What the evidence does support is a general preference for stimuli that are easy for the visual system to parse: symmetric, self-similar, low-parameter, or otherwise algorithmically simple. Three independent theoretical frameworks — processing fluency, predictive coding, and ecological affordance — all converge on the same underlying idea: the brain rewards itself for finding short descriptions of its input. The paper formalizes this within BCOM's Kolmogorov Theory (KT) framework, which models the brain as a "bounded compressor" evolved in a world whose structure is generated by compact, compositional physical laws (think symmetry groups, rigid-body transformations, fractal textures). The brain's generative vocabulary is pre-matched to that structure.

The core formal contribution is the notion of aesthetic affordance: a stimulus scores high when the agent's existing generative model can describe it with a short program. φ earns its modest reputation honestly — it has a genuine self-similarity property (a golden rectangle tiles recursively into a square plus a smaller golden rectangle), which makes it compressible under scale-invariant generators. But that's a property shared by any fractal or self-similar structure, not a unique privilege. Cultural reinforcement does the rest of the mythologizing.

The paper explicitly positions itself as the spatial/compositional companion to WP0081, which handled the temporal axis of aesthetic experience — how structured prediction-error flow over time drives neural plasticity. Together they sketch a two-axis KT account: a stimulus is beautiful when it is short for the agent in space (high aesthetic affordance) and generates well-structured prediction errors in time (the Goldilocks zone). The practical upshot for BCOM's algorithmic-art therapeutic platform is a design principle: build stimuli from low-dimensional, group-structured generators so they score high on both axes simultaneously.