Dominant Peaks, Fragile Metrics: Separating Architectural Signal from Methodological Artifact in Phi-Lattice EEG Analysis

The golden ratio (φ = 1.618) has been proposed to organize brain rhythms, but prior evidence relied on aggregate spectral metrics vulnerable to analysis choices. This paper applies a progressive diagnostic to those metrics — systematically demonstrating that structural scores, enrichment measures, and age correlations are dominated by FOOOF extraction parameters — before identifying what survives. The single strongest oscillation in each canonical frequency band falls closer to φ-predicted positions than to positions predicted by eight alternative mathematical constants, replicating across three datasets totaling 919 subjects (LEMON N=202, Dortmund N=608, EEGMMIDB N=109). Per-subject mean lattice distance converges to d̄ = 0.069 across all three datasets. Under coverage-normalized degree-3 positions, φ ranks 1st among nine logarithmic bases in every dataset (d̄/d̄_null = 0.90–0.92, p < 0.001), with alignment age-invariant (r = −0.004, Dortmund ages 20–70), fatigue-invariant, and nearly doubling under eyes-closed conditions (d = 0.75 vs 0.40). Critically, φ's rank-1 status requires per-octave aperiodic modeling; under standard whole-range FOOOF, φ ranks 7th of 9 (p = 0.83). In aggregate cross-band analysis, alignment concentrates at three positions: noble₁ (u = 0.618), attractor (u = 0.500), and boundary (u = 0.000); individual frequency bands may show distinct position preferences. Higher-degree positions carry anti-signal — peaks actively avoid them (d = −0.38 to −0.49). The monotonic attenuation from d ≈ 0.40 (4 positions) to d ≈ 0.07 (8 positions) confirms signal concentration at the core positions. A five-year longitudinal follow-up (N = 208) reveals a clean dissociation: population-level alignment is a constant (group d̄ = 0.031–0.033), but individual alignment shows no stability (ICC = −0.25 to −0.36). The lattice describes species-level architecture, not an individual trait. A cross-dataset sign flip resolves the anchor frequency: in Dortmund, where f₀ = 7.83 Hz and IAF/2 ≈ 5.0 Hz lie in opposite directions from theta, eyes-closed theta converges on f₀ (d = −0.385, p < 10⁻¹⁹) while diverging from IAF/2 (d = +0.351, p < 10⁻¹⁶), excluding all IAF-derived models. Yet this architecture predicts nothing about cognition: 100+ tests across three metrics in LEMON yield zero FDR survivors (80% power to |r| = 0.20). The brain's frequency architecture carries a golden-ratio signature anchored to a geophysical frequency, but that signature is a structural property of the species, not a functional lever of the individual.