A pre-specified phase-randomization surrogate test of a second-order symbolic statistic in stage 2 sleep EEG/EOG/EMG: a negative result and its limits

This preprint reports a retrospective, prespecified, falsification-oriented analysis of a second-order symbolic statistic in stage N2 sleep EEG/EOG/EMG from Sleep-EDF Database Expanded. The frozen analysis included 19 participants and 38 recordings, using 1,999 surrogates per recording per scheme. The group-level G1 test did not meet the prespecified excess criterion under the evaluated phase-randomization surrogate nulls (empirical p-values: 0.8773, 0.1264, and 0.1152 for the sensitivity analysis). The result is negative within the prespecified statistical framework and does not establish a biological mechanism, consciousness, causality, or clinical validity. Full G2 diagnostics have not been run. This is preprint version 2.7-R2, not peer reviewed. The associated software and provenance records are available through the Brain Model GitHub repository and its Zenodo software release.