Hippocampal SEEG band power showed no corrected monotonic load effect during maintenance in 14 patients

In session 1 of the OpenNeuro ds004752 verbal working-memory dataset, we tested whether hippocampal SEEG band power changed monotonically with memory load (4, 6 or 8 letters) during maintenance. The pre-specified analysis used correct, artifact-free trials from 14 patients with hippocampal bipolar pairs, z-scored log power in five bands, and exact two-sided sign-flip tests with Holm correction. No band showed a corrected maintenance load effect (theta mean rho +0.053, Holm p = 0.29; high gamma mean rho -0.065, Holm p = 0.22). Response time increased with load (mean rho 0.30, p = 0.0002, n = 15), and hippocampal power was higher during maintenance than fixation in all five bands. During encoding, alpha and beta power increased with load; this may reflect visual or attention load because the displayed letter set differed. Only one session per patient was analyzed, and the held-out session-2 replication was not completed. Effects smaller than about |mean rho| 0.07 cannot be excluded. The analysis tests monotonic rank relations in band power and does not test non-monotonic effects, single-pair effects, phase, coupling, or neuron firing.

verbal working memory · hippocampus · stereo-EEG · band power · memory load · maintenance · epilepsy · sign-flip test