Two tests did not detect an eight-letter versus four-letter effect on late hippocampal 55–80 Hz changes after a correct prior response in 14 patients

We tested whether verbal memory load changes hippocampal 55 to 80 Hz power from fixation to late maintenance. We used public SEEG from OpenNeuro ds004752 version 1.0.1. The analysis used 2,561 trials from 62 sessions in 14 patients with epilepsy. Each trial followed a correct response. We compared eight letters with four letters. We tested mean log density and spectral excess above a fixed local power law. The patient mean density effect was -0.100 dB. Its 95 percent bootstrap interval was -0.511 to +0.300 dB. The excess effect was -0.011 dB. Its interval was -0.171 to +0.141 dB. Both Holm p values were 1.000. Neither test detected the specified load effect. Excess assumes the stated local power law. The exact tests require independent patient effects with a symmetric zero null. They do not prove absence of a memory signal. The package gives an inspected trial dataset, all planned checks, and explicit spectral contrast weights. A fresh raw session rebuild matched all 16 arrays. An independent reader and SciPy checked 300 spectra in that session. The full analysis reproduced eight result files byte for byte. Code, source hashes, and retrieval instructions are preserved. This is a secondary analysis of the source patients.

hippocampus · verbal working memory · gamma power · spectral excess · stereo-EEG · memory load