A fixed graph test did not confirm more fragmentation with propofol in seven iEEG patients and 23 fMRI adults

We tested whether strong measured links stay in separate groups during a source propofol state after additive site and distance terms are removed. Separate public human studies supplied 10 Hz iEEG cross-spectra and cortical BOLD correlations. Development used four patients and 26 adults. A local code lock preceded tests in seven other patients and 23 adults from a second MRI site. A rank-based component-curve score uses all edge thresholds. Mean changes were +0.127 and -0.372 rank percentage points. Their 95% person t intervals were [-0.309, +0.564] and [-0.779, +0.036]. The fixed shared positive rule failed, p=0.984887. Three patients and 16 MRI adults had negative changes. The MRI bootstrap interval was below zero while its t interval crossed zero. All fixed controls and exclusions remain. A fixed common-input Gaussian model changed the same adjusted score without direct links between readouts. An independent solver reproduced all 120 primary state scores from 60 people. Hypothetical precision simulations retain the same latent covariance. The contribution is a checked graph tool and a complete test benchmark, with permitted inputs, code, source versions and terms, all outcomes, figures and repeat steps. Sparse clinical sites, separate cohorts, unequal precision, drugs, source preprocessing and vascular effects limit scope. No new positive neural effect across modalities or causal disconnection is established.

human neuroscience · intracranial EEG · fMRI · propofol · graph filtration · spanning forest · negative result · reproducibility