Deep sleep did not confirm a shared rise in magnitude memory above a phase model in 37 SEEG patients and 16 fMRI records

We tested whether deep sleep increases local signal magnitude persistence above a phase model in separate human SEEG and fMRI studies. Normal rank scores remove differences in measured marginal values. A temporal fourth-moment score removes the Gaussian lag-correlation term. Its phase controls keep each circular power spectrum. Fourteen SEEG discovery patients gave an uncertain positive mean. The fixed test then gave mean N3-minus-wake change +0.00251 in 37 other patients, with 95% person bootstrap interval -0.01136 to +0.01610. The fMRI release had 16 records and two tested identity schemes. Its mean ranged from -0.05235 to -0.05121; the bootstrap interval union was -0.10850 to -0.00474. The t-based union included zero. Primary positive sign-test p values were 0.74431 and at worst 0.98936; both Holm values were 1. The shared positive rule failed. We also derived and checked an exact phase-model mean for this score. It is negative for every nonconstant finite signal. Human phase means were about -0.012 to -0.014 in SEEG and -0.157 to -0.158 in fMRI. Thus, zero is not the conditional finite-spectrum reference. The exact formula supports a fast inspectable tool, not a claim of a shared neural mechanism. Simulations show that rank scoring, detrending and nonstationarity can change the score. Short records, clinical sampling, unresolved fMRI identities, missing nuisance traces and unequal measurement scales limit inference. The package gives code, plans, source versions and hashes, person and site results, negative checks, figures, and repeat steps. Restricted MNI waveforms remain external under their source terms.

sleep · SEEG · fMRI · temporal fourth moment · magnitude persistence · phase surrogate · independent confirmation · negative result