FR1 memory signals at eight sites did not predict 50 Hz stimulation effects on recall

This pilot asked whether an encoding-related memory signal recorded at a bipolar pair predicts the change in recall when that pair receives 50 Hz open-loop stimulation during encoding. We joined FR1 per-word power features from session 0 to per-site recall effects in FR2/catFR2. Nine stimulation sites in seven patients were available; eight sites in six patients had an exact bipolar-pair match and were analyzed. At each matched pair, the primary read-side measure was the area under the curve (AUC) for spectral tilt (mean z-scored 80 and 140 Hz power minus mean z-scored 3, 5, and 8 Hz power) for recalled versus unrecalled words. Its association with the stimulation-related recall change was not significant (Spearman rho = -0.48, two-sided permutation p = 0.24). Two exploratory measures also did not predict the change: decoder importance percentile (rho = +0.10, p = 0.84) and 140 Hz power AUC (rho = -0.07, p = 0.88). Matched pairs had weak FR1 memory signals (tilt AUC 0.39–0.57; median decoder-importance percentile 0.44), so the clinical site choice did not select pairs with strong encoding signals. With eight sites, only |rho| of at least 0.71 gives p < 0.05. The sites are not independent within patients; the per-site recall change is noisy; FR1 and FR2 were separate sessions matched by contact name; decoder weights were fit and evaluated in-sample; and one patient had only 14 recalled FR1 words. This pilot supports a larger read-to-write test with more sites and a planned spread of FR1 signal strength, but does not establish that read-side signals cannot predict stimulation effects.

memory · intracranial EEG · stimulation · free recall · subsequent memory effect · spectral power · pilot study