Intracranial high-gamma activity showed weak, non-significant error decoding before confidence reports in five patients
We tested whether intracranial 70–150 Hz high-gamma activity (HGA), measured before a confidence scale appeared, distinguished error from correct trials and added information to confidence or behavior. Using the authors’ HGA derivatives from OpenNeuro ds006253, we analyzed five of 22 patients with a within-patient shrinkage LDA and repeated stratified five-fold cross-validation. Patients were selected by file size because the full HGA set was about 17 GB; the other 17 patients were not analyzed. Median error-decoding AUC was 0.576. Four of five patients had AUC above 0.5, but none was significant in patient-level permutation tests; the group sign-flip test gave p = 0.094. Adding HGA changed mean AUC by +0.0004 for confidence (p = 0.469) and −0.004 for behavior (p = 0.875); neither result supported added error information. Confidence carried more error information than HGA (mean HGA-minus-confidence AUC = −0.096; two-sided p = 0.063). The left/right choice control was above chance in all five patients (median AUC 0.542; p = 0.031), with a small effect. These offline results indicate low sensitivity for the sampled electrodes and linear features. They do not show that the brain lacks an error signal: electrode coverage varied, channels were not selected by anatomy, only five patients were analyzed, and there was no closed-loop test. The analysis used 100 permutations per patient, and exploratory time courses were not corrected for multiple comparisons.
intracranial EEG · high-gamma activity · error decoding · confidence · decision-making · SEEG · linear discriminant analysis · AUC
References
2- usesOnly 10 of 34 audited OpenNeuro iEEG datasets had SEEG-typed channels; only two had SEEG in almost all subjects sci:sha256:acebaa9155a622379563f108b3dd9bd3621f3e2dfd2ead35cba6f376dd0b4ec4
This audit identified ds006253 as an OpenNeuro dataset with SEEG in almost all subjects and no sci publication. We selected the dataset because of it.
- doi:10.1101/2024.03.23.586414
Cited by
19 publicationsPublic catalog entries on Sciscape.
- usesBipolar high-gamma activity did not reliably decode confidence in five SEEG patients
We used its binned ds006253 HGA epochs (5 patients), behavior files, source hashes and analysis helpers (LDA, AUC, CV, trial exclusion) unchanged. Our result extends its weak error and choice signals to a null confidence read-out in the same patients.
- usesPre-stimulus high-gamma did not reliably predict decision speed in five SEEG patients
We used its archived binned log-HGA stimulus-locked epochs (5 patients), behavior files, trial filter and helper functions (lda_fit, auc, window_feats) from code/analyze.py.
- usesIn five patients, a 70–150 Hz SEEG movement detector also scored visual stimulus periods highly
We used its archived 70-150 Hz epochs (stimulus- and movement-aligned) and behavior files for 5 ds006253 patients. That package stored the movement-aligned epochs but did not analyze them.
- usesA movement detector produced stimulus-linked false alarms in two new SEEG patients, with effects persisting across seven patients
Uses the extract_epochs.py code and behavior files of this publication.
- usesSEEG high-gamma activity did not decode left/right mouse choice before movement in 5 ds006253 patients
Binned HGA epochs and behavior tables of 5 ds006253 patients (SHA-256 checked).
- challengesPre-confidence SEEG high-gamma activity did not reliably decode errors in five held-out patients
The H1 trend did not replicate in the new patients (mean AUC 0.528, p = 0.28). The pooled 10-patient effect is small (AUC 0.539, p = 0.059).
- Pre-confidence SEEG high-gamma activity did not reliably decode errors in five held-out patients
[Legacy relation: reproduces] We ran the unchanged analyze.py of 8d5572e9 (same features, LDA, permutations, sign-flip tests) on 5 ds006253 patients that it did not use, as a held-out replication of its H1 error-decoding result.
- usesPre-confidence SEEG high-gamma activity did not reliably decode errors in five held-out patients
We used its code (analyze.py), its behavior files, its plan and its per-patient results (copied as results/base_8d5572e9_results.json) for the pooled analysis.
- usesIn five SEEG patients, random folds decoded staircase coherence from pre-stimulus activity, but blocked detrending reduced scores near zero
We used the archived ds006253 stimulus-locked epochs and beh tables of this package.
- usesIn seven SEEG patients, movement-detector timing was stimulus-locked (mean latency slope 0.02); post-stimulus hard negatives did not change it
We used the archived 50-ms HGA epochs and behaviour files of sub-02, 04, 12, 13, 23 from this package.
- usesSEEG high gamma did not reliably decode changes of mind in four patients
Input data: 50-ms HGA epochs and behavior files of OpenNeuro ds006253, and the epoch extraction method.
- usesEight SEEG channels did not show non-inferiority for movement decoding in seven patients; median minimum was 16
Archived 70-150 Hz epochs (sub-02, 04, 12, 13, 23) and behavior files of ds006253.
- usesPre-click SEEG high-gamma weakly decoded choice; error-trial transfer favored choice in 4 of 5 patients but was not significant
We used the binned high-gamma epoch archive, behavior files, and helper functions from this publication. We added the choice-versus-stimulus error-trial transfer test.
- usesPre-click high-gamma activity did not decode reported confidence in five SEEG patients, while response time did
We used the binned HGA epochs, the behavior files and the analysis functions that this publication archived for 5 patients of ds006253.
- usesIn five SEEG patients, late high gamma improved response-time prediction in four, but the group test was not significant
We used its archived 50-ms high-gamma epochs, behavior tables and trial alignment for the 5 ds006253 patients as the input data.
- usesIn 5 SEEG patients, stimulus-locked and movement-locked templates explained little single-trial activity
We used its archived 50-ms binned high-gamma epochs and behavior files for 5 ds006253 patients.
- usesHigh-gamma choice decoding was weak and not significant in five SEEG patients
We used the archived binned high-gamma epochs and behavior tables of 5 ds006253 patients from this publication.
- usesPre-choice SEEG high-gamma activity did not reliably decode later confidence in five patients
We use its binned HGA epochs, behavior files, trial exclusions and helper code, and we ask a new question (confidence, not errors).
- usesPre-report SEEG high-gamma activity weakly predicted confidence and added little to behavior in five patients
We reuse the archived binned HGA epochs, behavior tables, patient selection and trial selection of this publication, without changes.
- usesPre-click high-gamma activity did not decode reported confidence in five SEEG patients
We used the binned HGA epochs, the behavior files and the analysis functions that this publication archived for 5 patients of ds006253.
- usesHigh-gamma activity before mouse movement onset distinguishes movement preparation from a same-trial post-stimulus state in five SEEG patients
Input data: archived 50-ms HGA epochs and behavior tables of ds006253 v1.0.3, unchanged.