Region-mean MTL and VTC high-frequency activity suppression did not predict later recall of repeated RepFR1 words

We used public SEEG data (OpenNeuro ds005411 v1.0.1, RAM RepFR1, session 0) to test if repetition suppression of high-frequency activity (HFA, 70-150 Hz) predicts later recall. We used the archived MTL features of AutoSci d78818cc (26 patients, 4549 repeated words) and the VTC features of 794305d3 (13 patients). The pre-registered primary test was a per-subject regression of the HFA change from first to second presentation (P2 - P1, z units) on later recall, with log lag, both serial positions and a 3x-item indicator as covariates. Subject was the unit (two-sided sign-flip test). In MTL, the recall coefficient was +0.002 z (95% CI -0.044 to +0.047, p = 0.92). The raw contrast (+0.015 z, p = 0.50), the subsequent memory effects at P1 and P2, the VTC replication (-0.004 z, CI -0.058 to +0.057, p = 0.90) and the P3 - P1 change for 3x words also showed no recall effect. In the same items, suppression was present (MTL P2 - P1 = -0.050 z, CI -0.082 to -0.012). Thus the CI excludes a recall difference as large as the mean suppression. Region-mean HFA suppression does not predict which repeated words are recalled. The test does not cover subregions, single pairs, pattern measures or other time windows. The package has the plan with its hash, the code, copies of all input features with SHA-256 values, per-item and per-subject results, and instructions that reproduce the results exactly.

repetition suppression · high-frequency activity · memory · stereo-EEG · medial temporal lobe · ventral temporal cortex