Reassessing constant claims in EEG band boundaries and their source provenance

EEG band boundaries depend on how they are measured and which recordings yield usable estimates. This retrospective audit examines a public boundary-analysis preprint, a later local manuscript candidate and related code artifacts. A portable program checks published counts and means; no raw EEG was reprocessed. Three estimators retained 55, 158 and 171 ratios from a nominal 244 participants. Their means of 1.853, 1.826 and 1.685 are nearest e − 1 among four selected constants, but summarize different eligible subsets. A constructed example further shows that changing from the mean of individual ratios to a ratio of mean boundaries can change the nearest-constant label. Detector abstention must be distinguished from an unrecorded value before extending the target to the full cohort. Two dataset-specific derivative means are nearest 2, and the source reports free-mean models outperforming all constants for two estimators. The preserved figure contradicts the claimed small-sample preference for the golden ratio. The stated random-boundary test does not establish a calibrated permutation probability, and equivalent group means do not identify self-similarity. The local candidate also confuses e − 1 with e⁻¹ and changes a correlation sign without a recovered analysis trail. Missing detector code, participant measurements and exclusion records prevent a complete reanalysis of the reported measurements. The result is a corrective methods and source audit; a mechanistic comparison requires recovery of the complete measurement pipeline and calibration against ordinary, selection and nuisance explanations.

Corrective release with the revised manuscript and audit supplement. Related Research Square record: https://www.researchsquare.com/article/rs-8935579 . Earlier unsupported interpretations are withdrawn as detailed in the manuscript. No raw EEG reanalysis is claimed. This preprint is not peer reviewed.