N3 cortical lag products exceeded a reversible Gaussian spectral control in 43 SEEG patients and 16 fMRI records

We tested whether deep sleep reduces repeated cortical lag direction in two separate human studies. The fixed primary prediction failed. The N3-minus-wake cross-block product changed by +0.002960 in 43 SEEG patients (95% person interval +0.000590 to +0.005533) and -0.009337 in 16 fMRI records (-0.036097 to +0.016128). A planned spectral diagnostic gave a negative product even in reversible Gaussian signals. After the SEEG result, we fixed one follow-up before any fMRI outcome was read: N3 measured products exceed this diagnostic mean. SEEG gave exploratory excess +0.004596 (+0.002986 to +0.006143). The independent fMRI test gave +0.054936 (+0.035000 to +0.073953), exact sign-flip p=0.000336 and Holm p=0.001007 across the two original tests and the new test. It remained positive when records with identical initials were merged. The control retains each full record's real cross spectrum in expectation and removes its imaginary part. Thus the shared finding is a deviation from one reversible stationary Gaussian model. Non-Gaussian activity, nonstationarity and model error can also cause that deviation. The result does not establish a common physical lag, entropy production or causality. Different cohorts, electrode coverage, BOLD dynamics, short blocks and incomplete upstream fMRI processing limit transfer. The package preserves two local plans, code, source hashes, all person results, model draws, figures and repeat steps. Every biological summary, geometry record and inclusion decision matched a source repeat.

human neuroscience · SEEG · fMRI · sleep · lag asymmetry · reversible Gaussian control · independent confirmation