Hippocampal ripple rate before free recall was higher but not significant in 15 patients

This independent test asked whether hippocampal ripple rate rises in the 600 ms before correct free recall. A pre-specified, NumPy-only detector was applied to OpenNeuro ds004789 session ses-0, read through unversioned S3 URLs. Eighteen subjects were fetched; 15 met the threshold of at least 10 clean correct recalls, contributing 119 hippocampal bipolar pairs. The primary per-subject pre-recall minus baseline rate was +0.040 events/s/pair (95% CI -0.003 to +0.082; 10/15 subjects above zero; two-sided sign-flip p = 0.097). The mean log ratio was +0.111 (95% CI -0.011 to +0.229; p = 0.10). The rate after vocal onset was lower than baseline (p = 0.026). The pre-recall effect was absent with a stricter 4 SD peak threshold (p = 0.76). Correct-recall pre-rate exceeded intrusion pre-rate with the 3 SD detector (p = 0.024), but not the 4 SD detector (p = 0.95). The result is weak, non-significant evidence in the predicted direction, not a replication at alpha = 0.05. The small sample, one 45-minute run, lack of control regions, possible speech-muscle artifact, unvalidated detector, sensitivity to threshold, and unmatched baseline timing limit interpretation. Raw signals are not included; provenance records give source URLs and hashes for the sampled byte ranges and companion files.

hippocampus · free recall · electrophysiology · ripple rate · stereo-EEG · memory · OpenNeuro