The Human Brain Reads the Same Local Structure as Earthquakes and Language Models

Abstract

The Watabe-Claude Method embeds a scalar time series into 4D via a width-4 Takens window and reads the 24-cellpolytope occupancy; the signed-permutation group B4 decomposes that occupancy into six orthogonal isotypicblocks. A prior study showed that one block, ax_a, reads a direction-symmetric local-transition / autocorrelationstructure in two very different domains -- local transitions in LLM token-representation sequences (reversal-invariant)and short-lag magnitude autocorrelation in single earthquake aftershock sequences (AR(1)-reproducible,reversal-invariant). Here we apply the identical pipeline and the identical mechanism-dissection diagnostics to a thirdinformation system: the human brain, via scalp EEG (PhysioNet EEGMMIDB, 109 healthy subjects, 64 channels,160 Hz). The brain is on the same side. Across all 109 subjects the ax_a component is strongly order-dependentunder a pre-windowing block-shuffle null (eyes-open Stouffer z=+83, 90/109 subjects positive; eyes-closed z=+122,97/109). The structure is direction-symmetric (time-reversal preserves it entirely, real z +9.9 vs reversed +10.2 onthe pilot cohort) and local rather than a global-drift artifact (a trend-only surrogate destroys it; the effect survivesfirst-differencing). Two neurophysiological controls establish that the detector is reading a genuine brain rhythm, nota signal-processing artifact: the effect strengthens when the eyes close (the Berger alpha effect), and it is occipitallydominant (posterior z=+18 versus frontal z=+8), the canonical topography of the alpha rhythm. Unlike earthquakes,a pure lag-1 AR(1) surrogate reproduces only ~43% of the effect, so the brain's ax_a structure is richer than a singleautocorrelation lag -- consistent with rhythmic, higher-order temporal organization. We read this not as a grand 'brainequals AI' claim but as a concrete, falsifiable scope result: the same geometric detector responds to adirection-symmetric local temporal structure wherever the data-generating process produces one, be it a fault zone,a language model, or a resting cortex. Text CC-BY-4.0; code AGPL-3.0-or-later.