A fixed 70–110 Hz decoder does not detect sensory mode transfer across four formats in 12 SEEG recordings

We tested whether auditory and visual labels transfer between two number format groups. The source was OpenNeuro ds005415, version 1.0.0. The test used 1,133 passive trials from 12 complete SEEG recordings. The four types were spoken numbers, Arabic digits, beeps, and dots, with quantities from 1 to 5. A fixed decoder used 70 to 110 Hz contact power. It trained on early trials and tested later trials. Mean balanced accuracy was 48.1%. A conditional label permutation test gave p = 0.722. Adjacent contact differences gave 51.1% and p = 0.304. The result does not show that sensory information is absent. It gives a boundary for this fixed method and sample. The package includes verified features, code, predictions, figures, source hashes, and reproduction instructions. The source audit found an incomplete EDF file, inconsistent sample indices, and event timing disagreements. The package records their correction or exclusion.

sensory mode transfer · stereoelectroencephalography · SEEG · number formats · high gamma power · ridge regression · cross-format decoding · permutation test