High-gamma speech–music differences did not repeat across film halves in SEEG depth contacts

This analysis asked whether speech-minus-music high-gamma (70–140 Hz, excluding 95–105 Hz) power differences in SEEG depth-contact pairs repeated across two halves of a naturalistic film. It used 579 adjacent bipolar pairs from 16 patients in the OpenNeuro ds003688 film task (version 1.0.7), reading seconds 10–20 of each of 13 alternating 30-second music and speech blocks. The planned split-half test used 15 patients with at least five pairs. The mean split-half Spearman correlation was 0.092 (8/15 positive; exact sign-flip p = 0.22), so the high-gamma pattern did not show reliable repetition or a stable group-level speech or music preference. Fifty pairs had permutation p < 0.05; four pairs in three patients passed within-patient false discovery rate correction. A secondary 8–30 Hz analysis showed a repeating split-half pattern (mean rho = 0.305, 12/15 positive, p = 0.00085). This lower-frequency result is not specific to speech or music because the conditions also differed in visual content, sound level, and attention. Electrodes had individual ACPC coordinates but no anatomical labels; only 10 seconds per block were analyzed, and 10 patients had 3–14 pairs. These limits reduce anatomical interpretation and power for small effects. The archived plan, code, derived features, byte-range hashes, results, and logs are included with the original files.

SEEG · intracranial EEG · high-gamma · speech · music · naturalistic stimulation · split-half reliability · OpenNeuro