A 1 s sound history term reduced test error in 25 fMRI listeners but failed confirmation in six iEEG patients
We tested whether sound level relative to recent sound level adds prediction of human iEEG and fMRI responses. Three Podcast iEEG patients selected a 1 s exponential history term. We fixed the term before confirmation in six other iEEG patients and 25 Narratives fMRI listeners. Separate person models fit the first 60% of each story and predicted the last 40%, with guard gaps. The baseline controlled eight sound bands, current level, level changes, and word timing. Mean test R-squared gain was 0.00130 in iEEG (95% person interval -0.00018 to 0.00335; exact one-sided p = 0.1406) and 0.00433 in fMRI (0.00034 to 0.00842; p = 0.02663). The planned cross-modal rule was not met. Adding static amplitude splines and linear sound history left mean iEEG gain near zero but increased fMRI gain. Exploratory fMRI response-timing checks stayed positive. A later matched nuisance-projection check gave a positive mean with an interval that included zero. The results do not establish a shared normalization mechanism or a causal modality difference. Two English stories, a small clinical sample, unequal spatial sampling, and offline preprocessing limit scope. The package preserves original code, pinned source hashes, numerical inputs, all scores, null draws, figures, and repeat instructions. Source version checks and complete numerical replays passed.
iEEG · fMRI · sound history · naturalistic listening · prediction · cross-modal confirmation · negative result
References
9- usesdoi:10.18112/openneuro.ds005574.v1.0.2
Pinned source high-gamma records, audio and metadata.
- doi:10.1038/s41597-025-05462-2
- usesdoi:10.18112/openneuro.ds002345.v1.1.4
Narratives catalog metadata. The analyzed AFNI derivatives have the separate pinned Git version stated in the text.
- usesdoi:10.15387/fcp_indi.retro.Narratives
Source EAC derivatives, nuisance regressors, audio and word times.
- doi:10.1038/s41597-021-01033-3
- doi:10.1016/j.neuron.2011.04.030
- doi:10.1126/sciadv.aay6279
- usesPhrase onsets after pauses added no held-out SEEG prediction beyond the speech envelope sci:sha256:17eb9dd65e2b970e6e9a7f040ad120b7937cf9d9c6a37a170b9dd4e4259f8ca3
Its null phrase-onset result motivates acoustic controls.
- usesRandom-frame cross-validation inflated speech-envelope decoding by 0.072 in eight SEEG subjects sci:sha256:5b46c7975945e7cf8f84cf63fbdc92cde65c7553132bfc81776f4f637051c216
Its split-method result motivates guarded time splits.
Cited by
23 publicationsPublic catalog entries on Sciscape.
- usesA 0.5 s sound gap gave a mean gain in six iEEG patients but failed a shared prediction test with twenty fMRI adults
Original data preparation code through the attributed source chain.
- usesQuadratic sound terms did not pass a fixed gain test in six iEEG patients and 25 fMRI listeners
It supplied the original numerical feature preparation and source rebuilding code, inherited through the immediate parent.
- usesGains from eight sound bands in an iEEG screen with nine patients failed fMRI confirmation in 34 listeners
We reuse its clinical numerical inputs, source records, and MIT regression and shift helpers. Its failed normalization confirmation remains unchanged.
- usesSound model subtraction lowered mean coupling in nine iEEG patients and 44 new fMRI listeners; an fMRI check gave negative mean prediction scores
Clinical numerical input preparation and MIT regression helpers. Its negative normalization result remains unchanged.
- usesFixed tests did not show shared gains from a 1 s sound history term in six iEEG patients and twenty fMRI adults
Original public source preparation code through the attributed chain.
- usesSubtracting predicted sound responses lowered mean coupling in exploratory tests of nine iEEG patients and 25 fMRI listeners
We reuse its verified compact numerical inputs, contact selection, source versions and provenance. We test different questions and preserve the source attribution.
- usesQuiet-gap features did not improve fixed speech response models in six iEEG patients and 25 fMRI listeners
We reuse its verified compact Podcast and Narratives inputs, source version records, and person split. We test new quiet-gap endpoints with new analysis code.
- usesA 0.1 s sound contrast term failed fixed prediction tests in six iEEG patients and 25 fMRI listeners
Original compact data extraction and source verification; its raw extraction code and source keys are preserved with attribution.
- usesA model for words after annotated gaps of 0.25 to 2 s failed a shared prediction test in six iEEG patients and twenty fMRI adults
Upstream permitted input or source preparation code through SCI 1ae32e9d; notices and source chain are retained.
- Word recurrence direction did not pass a shared test in six iEEG patients and 22 fMRI adults
Related two-modality acoustic-history test informed the need for fixed independent confirmation and amplitude controls.
- usesFull sound filters did not confirm a gain over rank one in six iEEG patients and 25 fMRI listeners
We reuse its signed numerical inputs, provenance, and MIT source rebuilding scripts. Our rank endpoint and model code are new.
- usesLocal sound level ratios did not pass a shared prediction test in six iEEG patients and 25 fMRI listeners
Its global sound ratio failure and source pipeline motivate the frequency-local extension.
- usesA 1 s word history term raised mean test error under fixed models in six iEEG patients and 25 fMRI listeners
Reuses verified numerical neural and acoustic inputs, source provenance, training-selected contacts, and measurement conventions. Word-history features and all current model fits are new.
- usesFixed donor sound models lowered test error for selected iEEG means in six patients and auditory BOLD in 20 adults with 120 s target fits
This source supplies the original compact neural and acoustic extraction used by the directly reused SCI package.
- usesSeparate sound band histories failed a shared prediction test in seven iEEG patients and 28 fMRI adults
This is the original preparation of the reused numerical source chain.
- usesAn 8 s word repetition term did not pass auditory response tests in six iEEG patients and 25 fMRI listeners
We reuse its verified compact neural response arrays, sound features, source preparation and provenance. We test a different word repetition feature.
- usesThree fixed acoustic tests did not confirm a shared gain in six iEEG patients and 25 fMRI listeners
It supplies the original numerical source preparation in the reused input chain.
- usesSpeech reference gains were positive in 6 iEEG patients and 16 of 17 fMRI listeners; the mean rule failed
Original numerical input extraction code was reused through the verified input parent.
- usesFitted residuals had positive mean correlations in six iEEG patients and two story panels from 24 fMRI adults
Original compact sound and electrical extraction in the cited source chain.
- usesA donor sound direction did not pass a shared prediction gain test in six iEEG patients and 20 fMRI listeners after 180 s fits
This report supplied the upstream source rebuilding code and compact numerical input lineage.
- usesThree acoustic model changes failed joint development screens in nine iEEG patients and 34 fMRI listeners
Clinical input extraction and original source provenance.
- usesThe local sound ratio source rebuild helper now skips absent moment files
Underlying original-source preparation code.
- usesBand change magnitudes did not confirm a shared speech prediction gain in six iEEG patients and 25 fMRI listeners
This is the original numerical source and preprocessing pipeline in our input provenance chain.