SEEG band-power drift persists with correct event timing but does not require word-locked activity

A re-test of an earlier SEEG encoding-pattern drift and contiguity analysis compared features cut from the same EDF bytes at events.tsv onset and at sample/fs. In the 22 subjects shared by both timing conditions, the onset analysis reproduced all archived subject values exactly. With sample/fs timing, drift (lag-1 pattern correlation minus mean correlation at lags 4–11) was +0.0304 [0.0216, 0.0400], positive in 24/26 subjects (Holm p = 1e-4). The paired sample-minus-onset difference was -0.0011 [-0.0080, 0.0063] (n = 22, p = 0.78), and drift across subjects correlated r = 0.77 between timings. The onset-timed epochs had almost no word-locked response, so the similar drift supports slow temporal autocorrelation of band power rather than a word-specific item-context signal. Adjacent-word encoding similarity did not predict consecutive recall: the sample-timed estimate was -0.0170 [-0.0334, -0.0008] (n = 14, sign-flip p = 0.080), with a direction opposite to the retrieved-context prediction. The analysis includes 26 pure-SEEG patients from session 0; the contiguity analysis has low power, and onset epochs are not a random control because they can fall in recall or distractor periods.

SEEG · free recall · band power · temporal autocorrelation · encoding patterns · event timing · retrieved context · contiguity similarity