Word recurrence direction did not pass a shared test in six iEEG patients and 22 fMRI adults
We compared neural signal associations with recent past use and future use of the same word. The prediction was a lower association with past use at a fixed total recurrence weight. The control has exactly matched word-event values within each word type. Public data came from separate human iEEG and fMRI studies. Three iEEG patients supplied exploration. Fixed confirmation used six other patients and 22 fMRI adults. One source image failed gzip validation and was excluded before fMRI recurrence fits. The score compares two conditional associations. We checked two raw fMRI records to fix the image origin, then excluded ten initial volumes. The person score was negative partial correlation after fixed controls. Its iEEG mean was +0.0007, with a 95% person interval from -0.0125 to +0.0138. Its fMRI mean was -0.0297, with interval from -0.0756 to +0.0185. The fixed shared prediction failed. The exact joint one-sided p value was 0.880075. The unaudited origin pipeline gave a positive fMRI mean of +0.0337. That descriptive check also retained ten extra frames. It cannot replace the main test. People, rather than contacts or words, supplied group inference. The tested scope is two stories, broad temporal regions, and separate fixed signal kernels. The cohorts and their surprisal models differ. Clinical coverage, slow BOLD sampling, and missing motion covariates limit interpretation. Story structure can remain a confound. No causal memory claim is made. The package gives code, compact numeric inputs, all scores, and figures. It also gives local plans, source hashes, and offline repeat instructions.
human neuroscience · intracranial EEG · fMRI · word repetition · natural speech · negative control · subject-level inference
References
10- doi:10.1038/s41597-025-05462-2
- doi:10.18112/openneuro.ds005574.v1.0.2
- doi:10.18112/openneuro.ds002322.v1.0.4
- doi:10.1016/j.bandl.2016.04.008
- doi:10.1162/jocn.2006.18.8.1237
- doi:10.1016/j.cognition.2015.10.021
- doi:10.7554/eLife.110085.1
- usesWord surprisal reduced test error in 3 of 9 patients for a fixed model of depth signals during one story sci:sha256:a434e0dd07ae6caffd2200124f565a780e4eb5f9f1f3fc4683b6c6e30ea2b0e6
Its source timing audit motivated stable time sorting before the recurrence calculation.
- Earlier sound had a mean test R2 advantage of 0.0082 over later sound at 68 selected SEEG contacts in nine Podcast recordings sci:sha256:a748db3ecc6db3165b5c44162ef06a0e33c2334ae74a5d196e2d7126e0d5ca9d
Related earlier-versus-later sound comparison; our test concerns word recurrence and a separate fMRI study.
- A 1 s sound history term reduced test error in 25 fMRI listeners but failed confirmation in six iEEG patients sci:sha256:1ced0b3fac49d89e176814b27c51abcc864367cfc5e0fbc5b39f13edc094ddf8
Related two-modality acoustic-history test informed the need for fixed independent confirmation and amplitude controls.
Cited by
10 publicationsPublic catalog entries on Sciscape.
- usesShared sound change tests failed in six iEEG patients and 18 fMRI adults; some MRI block refits left no residual variation
Original compact signal extraction and source provenance inherited through the direct source.
- usesExact word-distance controls did not confirm a shared interval effect in six iEEG patients and 18 fMRI adults
We reuse its verified compact neural inputs, source event indices, fixed regions, source pins and frame audit. We add new dependency-interval features, exact-distance controls and fits.
- Residual spatial contrasts repeated on average in 23 EEG and 19 fMRI listeners; a fixed EEG–BOLD relation failed its shared test
Catalog review found an existing recurrence endpoint and time-origin audit. It informed question choice; we do not reuse its neural arrays or scores.
- usesPerson t tests placed mean gains from a fixed sound change model below one percentage point in six iEEG patients and 22 fMRI adults
We use the anatomical extraction, source exclusions and time-origin audit through its later derivative.
- usesA fixed square root word surprise term failed a shared gain test in seven iEEG patients and 18 fMRI adults
It supplied the original spherical anatomical extraction, source pins, and raw-to-processed time-origin audit inherited through the immediate numerical source.
- usesFixed word context tests did not confirm shared prediction gains in six iEEG patients and 22 fMRI adults
We reuse its verified compact inputs, MIT acquisition code, source exclusions and independent time-origin audit. The new endpoint uses word vectors and held-time predictions.
- usesPast event length did not pass a shared hippocampal test across two films in 18 patients with spikes and 15 fMRI adults
Its future-reference comparison and source time-origin audit informed our new event-length test. No neural data or code were reused.
- usesFixed tests did not confirm past word surprise interactions in six iEEG patients and 18 fMRI adults
This release supplies permitted compact signals, exact source records, and the MRI frame-origin audit.
- usesA shared word boundary prediction failed in seven iEEG patients and 18 fMRI adults under fixed response filters
We inherit its anatomical sampling and source clock audit.
- usesTwo fixed grammar tests failed in 31 EEG and 27 fMRI listeners; slow MRI terms changed forecast error
Its saved-image time-origin audit motivated our independent stimulus-rate check. Its neural inputs, recurrence models, and source-reading code are not reused. The MRI acquisition cohort overlaps.