In 9 patients, preregistered tests did not reproduce three amygdala band-power claims; firing–high-gamma coupling remained inconclusive

We tested four claims about amygdala activity during aversive video blocks using the public OpenNeuro ds003374 dataset (version 1.1.1): early high-gamma power increases, delta power decreases, the high-gamma increase is specific to non-seizure-onset-zone (non-SOZ) amygdalae, and neuronal firing covaries with high-gamma power. We analyzed 9 patients, 14 amygdalae, 2 macro contacts per amygdala, and 8 aversive and 9 neutral blocks per patient. Welch log10 band power and exact tests with Holm correction showed no support for the first three claims: the 0–2 s high-gamma effect was −0.019 (3/9 positive; p = 0.41), the 0–18 s delta effect was −0.003 (3/9 positive; p = 0.98), and the non-SOZ minus SOZ high-gamma difference was −0.004 (p = 0.92). Block-demeaned firing–high-gamma coupling was positive in 3/4 aligned patients (mean Fisher z = 0.12; p = 0.25); this low-powered test does not reject the coupling claim. NIX and EDF signals matched in 7/9 patients. In sub-01 and sub-05 they did not match, so their spike times could not be aligned to EDF. Descriptive, uncorrected results included a pre-onset high-gamma difference and a baseline-subtracted delta effect that depended on signal reference; neither was interpreted as support for the preregistered claims. Limitations include the small sample, confounding of emotion with stimulus category, SOZ–hemisphere confounding, and differences from the original study's normalization pipeline. The package preserves the preregistered plan, analysis code, result files, and input file hashes and URLs.

amygdala · intracranial EEG · SEEG · high-gamma power · delta power · neuronal firing · replication · OpenNeuro · preregistered analysis