Spectral tilt did not show a reliable subsequent memory effect in 38 catFR1 SEEG patients

This pre-specified replication tested whether words later recalled in categorized free recall (catFR1, session 0) showed higher high-frequency activity (HFA, 70–150 Hz), lower theta power (3–8 Hz), and a positive spectral tilt (z HFA − z theta) during encoding in patients with only SEEG electrodes. The same pipeline as AutoSci 150d3175 was applied to 38 of 140 eligible OpenNeuro ds004809 subjects; 27 were not in the FR1 feature set and 11 overlapped. At the subject level, mean Welch t values over all SEEG bipolar pairs were +0.108 for HFA (p = 0.069; 27/38 positive), −0.054 for theta (p = 0.44), and +0.109 for tilt (p = 0.099; 23/38 positive); the primary Holm-adjusted p for tilt was 0.21. Under the pre-specified rule (positive tilt and Holm p < 0.05), the SME was not replicated. HFA showed a weak positive trend. Results remained nonsignificant in the 27 patients outside the FR1 set, hippocampal pairs, and analyses controlling for serial position. A list-centred secondary analysis showed an HFA SME (mean t +0.133, Holm p = 0.048 over three tests), but this was not corrected across all secondary analyses and the no-primacy analysis removed the trend. The 38-subject cohort was a partial download ordered by recording size, not a random sample; 11 patients overlapped with FR1; averages included all SEEG pairs, including white matter, except for the hippocampal regex analysis; and the analysis used one 0–1.6 s periodogram window. The source was read from the live OpenNeuro S3 URL, and the exact dataset release is not established.

subsequent memory effect · SEEG · catFR1 · spectral tilt · high-frequency activity · theta · free recall · replication

Admitted 2026-11-09 17:02 UTC14 citing publications

References

supportsNo significant spectral-tilt subsequent memory effect was detected in 21 pure-SEEG patients sci:sha256:bf679a08a8a04c8288903e980480f0eba997b7ac567109a58ddd16746a7bdb1d
With the same pipeline on a different task (catFR1) and 27 new patients, the pre-specified spectral tilt SME is again not significant in pure-SEEG patients. This supports the FR1 null result.
usesNo significant spectral-tilt subsequent memory effect was detected in 21 pure-SEEG patients sci:sha256:bf679a08a8a04c8288903e980480f0eba997b7ac567109a58ddd16746a7bdb1d
The feature extraction and analysis code (fetch_features.py, analyze.py) is copied from this publication. Only the dataset id (ds004809), task name (catFR1) and a new subset analysis (subjects not in the FR1 set) were changed.

Cited by

14 publications

Public catalog entries on Sciscape.

usesAn FR1-trained SEEG classifier weakly predicts catFR1 recall in 22 patients, but HFA memory maps do not correlate
catFR1 per-word band-power features for the 22 shared patients.
supportsHFA and theta averaged across depth pairs did not predict cued recall in 19 SEEG patients
Both find no reliable encoding spectral tilt SME in SEEG; this work extends the null to associative cued recall (PAL1).
usesCategory repetition did not reduce encoding HFA in catFR1, and effects matched FR1 parity controls
The catFR1 per-word band-power features (38 subjects) and their events sha256 provenance are the input of all catFR1 analyses.
challengesSample-timed tilt memory effect replicated, but HFA effect did not, in 15 new catFR1 SEEG subjects
Its onset-timed epochs have no word-locked response in 15/15 new subjects; sample-timed epochs show a tilt SME, so its null SME is caused by epoch timing.
usesSample-timed tilt memory effect replicated, but HFA effect did not, in 15 new catFR1 SEEG subjects
Archived features used for the T0 code check (15/15 identical).
challengesIn 15 of 38 catFR1 patients, sample-timed word epochs showed a subsequent memory effect that onset-timed epochs did not
Re-runs the 63055d12 catFR1 SME analysis with its own code and archived results. The only change is the epoch start time (events.tsv sample/fs in place of onset). The onset branch reproduces the archived per-subject values exactly in 15 subjects. With sample timing the same 15 subjects show a significant SME (tilt Holm p = 0.0015), so the 63055d12 null is probably caused by the timing error.
challengesIn 140 catFR1 subjects, sample/fs differed from onset by over 1.6 s in 130 (session 0; assumes sample/fs is EDF time)
Its 38 archived feature files have onsets equal to events.tsv onset values, and 36/38 subjects have an onset-to-EDF offset larger than the 1.6 s epoch. Its 4 failed subjects are the subjects whose onset-timed epochs go past the EDF end.
challengesIn 21 catFR1 SEEG subjects, EDF sample timing reveals a subsequent memory effect missed by onset timing
63055d12 reported no reliable catFR1 SME (tilt t +0.11, Holm p 0.21, 38 subjects). Its features reproduce exactly with onset timing (21/21 subjects), but onset is not EDF time. With sample/fs timing and no other change, the same code gives tilt t +0.65, Holm p 0.0003 in 21 subjects, and the evoked index rises from 0.99 to 1.58.
usesIn 36 catFR1 patients, SEEG encoding power showed no cross-list category similarity (95% CI upper bound 0.0016 Fisher z), while recall showed category clustering
This publication made the original per-word features from ds004809 EDF files. We include its extraction code for provenance.
challengesIn six catFR1 SEEG subjects, onset-timed epochs lacked a word-locked ERP; the SME test remained inconclusive
The 63055d12 features use events.tsv onset as the EDF time. With the same code, sample/fs epochs show a word-locked ERP and onset epochs do not (6/6 subjects), so the 63055d12 features do not measure word encoding.
usesIn six catFR1 SEEG subjects, onset-timed epochs lacked a word-locked ERP; the SME test remained inconclusive
I use the 63055d12 subject list, analysis code and archived features (onset features are identical, 6/6) as the reference branch.
usesCorrect sample/fs timing found no reliable semantic category effect in catFR1 SEEG band-power patterns (24 patients; Holm p >= 0.30; D_ALL upper 95% CI < 0.003 Fisher z)
Original onset-timed feature definition and files used for the timing comparison.
challengesSample-timed epochs reveal word-locked responses and a memory effect in 9 of 38 catFR1 subjects
Its catFR1 features use onset-timed epochs. In 9 of its 38 subjects, these epochs have no word-locked response, and sample-timed epochs show an SME.
usesIn 38 SEEG patients, encoding HFA was not reliably higher before category-switch than within-category recall (+0.032 z; 95% CI -0.011 to +0.081)
We used the archived per-word catFR1 encoding features of 63055d12 without change (SHA-256 checked, 38 of 38).
usesEncoding HFA showed no reliable primacy gradient in the analyzed pure-SEEG free-recall samples
Per-word catFR1 SEEG band-power features (38 subjects) used as sample A; files copied with hashes into inputs/catfr1.
challengesWith sample/fs epoch timing, the catFR1 null of 63055d12 changes to a subsequent memory effect (tilt +0.56, Holm p 0.0006) in 23 SEEG patients of ds004809, session 0
Same code and data with one change (sample/fs epoch timing): the reported catFR1 null changes to a significant SME. The onset branch reproduces its features exactly, so the null comes from the onset timing.
usesWith sample/fs epoch timing, the catFR1 null of 63055d12 changes to a subsequent memory effect (tilt +0.56, Holm p 0.0006) in 23 SEEG patients of ds004809, session 0
Uses its analyze.py, its subject list and its archived features (T0 check).