SEEG high-gamma decoding did not show cross-modal transfer of number format in five subjects, with onset and coverage uncertainty

We tested whether a shrinkage linear discriminant decoder could distinguish symbolic from non-symbolic number stimuli within visual and auditory modalities and transfer between them using stereo-electroencephalography (SEEG) high-gamma power. In five analyzed subjects, the pre-registered cross-modal transfer endpoint failed its support rule (p < 0.05 and AUC > 0.5) in every subject; transfer AUCs were 0.40–0.55 (p = 0.24–0.90). Within-modality decoders also failed. Two of seven subjects were excluded after data checks. An added alignment and coverage check found t > 4 in only 2 of 769 bipolar pairs, so the analysis did not confirm a stimulus-locked high-gamma response. The result cannot distinguish absence of a shared code from incorrect stimulus onsets or lack of task-responsive contacts, and it is not evidence that no modality-invariant code exists. The saved analysis plan, code, features, per-subject results, figure, metadata tables, and source-file manifest are preserved. The raw EDF files are not included; their OpenNeuro download location and recorded ETags are provided in the report.

stereo-electroencephalography · SEEG · high gamma · cross-modal decoding · number cognition · linear discriminant analysis · negative result · OpenNeuro