Cross-task slopes reject a one-step-per-event model but do not separate elapsed time from word count

We compared serial-position changes in high-frequency activity (HFA, 70–150 Hz) and theta power (3–8 Hz) across four SEEG word-encoding tasks to test whether slopes follow the number of presented events or elapsed time. The analysis used archived per-subject ordinary least-squares slopes, expressed as standard deviations (SD) per position, for FR1 (51 patients), catFR1 (23), pyFR (18), and PAL1 (18). Median onset-to-onset intervals were 2.50, 2.50, 2.63, and 5.93 s, respectively. The ratio of the PAL1 mean slope to the pooled short-interval mean was 1.83 for HFA (95% CI 1.10–2.78; p=0.011 for R>1) and 2.31 for theta (95% CI 1.48–3.29; p=0.0009). The one-step-per-event prediction (R=1) was rejected for both bands. The elapsed-time prediction (R=2.33) was not rejected (Holm-adjusted p=0.28 for HFA and 0.48 for theta). Between-task heterogeneity was lower in per-second than per-position units, with p=0.057 for HFA and p=0.68 for theta in per-second units. Each PAL1 position contains a pair of words, so a word-count prediction of R=2 lies within both confidence intervals; the data therefore do not distinguish elapsed time from word count. The within-subject analysis of nine patients was inconclusive. Different task demands, epoch windows, patient groups, and the assumption that channel-wise z-scoring yields comparable scales across tasks limit interpretation.

SEEG · serial position · high-frequency activity · theta power · word encoding · elapsed time · item count