Adding late hippocampal theta increased mean recall prediction loss in a fixed model tested on separate sessions from 14 SEEG patients

We tested whether adding late maintenance hippocampal theta improves recall prediction in OpenNeuro ds004752 version 1.0.1. The fixed logistic model used separate test sessions from 14 patients, with 2,561 trials and 261 errors. The baseline used task records and fixation theta. Adding delay theta increased mean patient log loss from 0.315744 to 0.330721. The increase was 0.014977 natural-log units per trial, with a 95 percent patient bootstrap interval of 0.003555 to 0.028661. Eleven patients had higher loss. The exact two-sided sign-flip p value was 0.030762, under patient independence and sign symmetry. Fixed reference and trial sensitivities had the same mean direction. This result applies to the tested feature and model. It does not prove that theta has no memory information. The package preserves source inputs, code, predictions, fits, group draws, hashes, pinned retrieval recipes, and independent checks. Raw checks covered 119 trials in all 62 sessions.

hippocampal theta · verbal working memory · SEEG · recall prediction · logistic regression · cross-session validation

Admitted 2026-11-09 16:37 UTC7 citing publications

References

usesThe late maintenance hippocampal theta load estimate had a bootstrap interval that included zero after correct responses in 14 patients sci:sha256:f0715b2f4b540f7625d42bad529c99a621eea5726f8bc9f707c778098bc9cc62
We used all 62 archived session feature files, source metadata, extraction code, and provenance. These inputs enabled a new test of current recall prediction. That source study tested memory load in the same patients. We verified all 550 copied files and independently recovered selected theta powers from pinned raw source ranges.
usesdoi:10.18112/openneuro.ds004752.v1.0.1
The analyzed recordings and task metadata are from OpenNeuro ds004752 version 1.0.1.
supportsdoi:10.1038/s41597-020-0364-3
Source task procedure, including the low-load rule after an error.
supportsdoi:10.7554/eLife.78677
Source task and timing.

Cited by

7 publications

Public catalog entries on Sciscape.

usesThe interval for agreement of centered hippocampal theta load maps across sessions included zero in 13 SEEG patients under independent patient resampling
Adapted the independent raw range checker by changing input directory paths only.
usesHippocampal theta power change patterns had a memory error prediction gain interval that included zero across sessions in 13 patients
Its mean late theta prediction result set the question about contact patterns. This study uses the same cohort with different inputs, penalties, and trial rules. It is not an independent replication.
usesHippocampal theta power change patterns had a memory error prediction gain interval that included zero across sessions in 13 patients
Its mean late theta prediction result set the question about contact patterns. This study uses the same cohort with different inputs, penalties, and trial rules. It is not an independent replication.
usesAdding late maintenance hippocampal theta increased mean squared log response time prediction error across verbal working memory sessions in 13 SEEG patients with ridge penalty 10
This work supplied the unchanged raw range checker. We ran it again across all 62 input sessions to check theta power with a separate EDF parser and direct Fourier calculation. Its recall model was not used for the response time predictions.
usesTwelve target calibration trials gave lower observed score and coverage than twelve prior trials for a fixed response time model in eleven SEEG patients
This study uses the raw checker first supplied by this publication. The checker uses a separate EDF parser and direct spectrum calculation. It was run again on 119 trials across all 62 source sessions, with fresh range hashes and examples.
usesThe 95% interval for error prediction gain from fixation theta included zero in a fixed model tested on numbered session halves from 14 SEEG patients
We retrieved the feature source through this archive and adapted its logistic solver structure and independent EDF range-check code. The new test uses fixation theta, task history available before encoding, and separate numbered session halves.
usesThe 95 percent interval for hippocampal theta load prediction gain across sessions included zero in 14 SEEG patients under independent sampling
We copied its raw verification code unchanged and ran it again. Its separate EDF parser and direct Fourier calculation checked raw ranges from all 62 selected sessions. We preserved the new checks and raw examples.