Twelve-subject SEEG analysis found weak modality-specific number patterns in odd/even splits, not confirmed by random splits

This extension of AutoSci 2ad72f35 added six subjects to a pre-specified test of whether 70–150 Hz response patterns across bipolar SEEG pairs were more similar within than between visual (Arabic, written, dots) and auditory (spoken, beeps) number stimuli. For 12 subjects from OpenNeuro ds005415 v1.0.0, the primary odd/even split gave mean D = 0.061 (95% bootstrap CI 0.020 to 0.104; exact two-sided sign-flip p = 0.021; 10/12 positive), meeting the pre-specified decision rule. The robustness analysis using the mean of 200 random splits gave D = 0.023 (95% CI −0.016 to 0.059; p = 0.27; 8/12 positive), and did not confirm the primary result. Because adjacent trials occur in both odd/even halves, slow signal changes could inflate within-modality agreement; evidence for modality-specific patterns is therefore weak. Visual and auditory mean patterns correlated positively (r = 0.139, 95% CI 0.051 to 0.224; p = 0.015), while within-modality agreement was low (mean r = 0.110). The onset rule for the six added subjects matched behavioral fixation times to FixationOnset events; unmatched rows were dropped and retained onsets were within 0.03 s of Cue events. Subject 80 was excluded because its EDF was truncated. The package preserves analysis code, plans, results, feature files, byte-range provenance, and hashes. The H2 responsive-pair log odds ratio used a 0.5 correction on very sparse tables and should not be interpreted.

SEEG · high gamma · number cognition · sensory modality · response patterns · split-half reliability · OpenNeuro