In 18 SEEG patients, the encoding recall score separated recalled from unrecalled words but did not predict recall order or latency

We tested whether the W_chan SEEG encoding-period recall decoder from SCI publication c4b8c884 measured graded memory strength among later-recalled words. In 18 patients performing FR1 free recall (OpenNeuro ds004789 v3.1.0, session 0), its score did not predict output order after controlling for list and serial position (median partial r = -0.15; one-sided exact Wilcoxon p = 0.18; bootstrap 95% CI, -0.28 to +0.12) or recall latency (median partial r = -0.11; p = 0.25). The score was higher for recalled-late than not-recalled words (+0.45 SD; 16/18 patients; p = 2.7e-5), but not for recalled-early than recalled-late words (+0.01 SD; p = 0.60). A behavioral check found that serial position predicted output order (median partial r = +0.30; p = 0.0013). These results are consistent with a mostly binary encoding-success signal in this dataset, not a graded memory-strength measure among recalled items. Failure to reject the null does not prove no effect; small graded effects remain possible. The analysis used archived per-word features and was offline, limited to one task and session.

SEEG · memory · free recall · brain-computer interface · decoder · memory strength · output order · recall latency