An exact spectral control did not support a shared N3 rise in covariance variation in 26 SEEG patients and 16 fMRI participants

We tested whether N3 sleep increases temporal variation in distant cortical covariance beyond the full recorded cross spectrum. Public data came from separate human SEEG and fMRI studies. A fixed patient split gave 30 SEEG discovery patients and 26 reserved patients. The fMRI test used 16 other people. We derived and checked an exact finite-record mean for a common Fourier phase control. It keeps all auto spectra and cross spectra. The outcome was fractional excess variation of block covariance contributions. The reserved SEEG N3-minus-wake mean was +0.0482, with 95% person interval -0.0217 to +0.1230 and Holm p=0.2122. The fMRI mean was -0.1998, with interval -0.7194 to +0.3174 and Holm p=0.8076. The shared increase rule failed. Planned controls did not establish a common increase. A later method check found that a correctly centered Gaussian reference exceeded the exact phase mean by about 0.2% in SEEG and 1% in fMRI. This small difference did not explain the failed prediction. Simulations, exact phase quadrature, source spot checks, and a full numerical repeat passed. The package supplies code, the derivation, plans, source versions and hashes, all person results, figures, and repeat steps. Clinical sampling, unequal time scales, missing fMRI covariates, prior processing, and short records limit inference. No positive biological effect generalized across these modalities, and no causal claim is made.

human neuroscience · SEEG · fMRI · N3 sleep · covariance variation · Fourier phases · negative result · reproducible methods