sciscape Publications
← All publications
q-bio.NC · Neurons and Cognitionstat.ME · Methodology

A two-pattern sound forecast did not pass a shared 1% gain test in 17 iEEG patients and 16 fMRI adults

We tested whether two spatial patterns predict new sound examples better than one. Separate public human studies supplied signed intracranial EEG and BOLD fMRI. Four people per study supplied development. A local code lock preceded the new test in 17 other patients and 16 other adults. Five sound classes had disjoint training and test clips. Mean error reductions were 1.115% in iEEG and 8.801% in fMRI, relative to grand-center baseline error. Person 95% intervals were -0.266% to 2.495% and 7.102% to 10.500%. Ranges from common-sound and person resampling were -2.747% to 1.812% and 4.656% to 10.574%. The fixed shared 1% gain rule failed. All negative outcomes remain. A separate exact model has one scalar input and six strictly increasing site functions. It gives a 16.694% two-pattern gain after the same scaling and spatial mean removal. Thus, extra useful linear patterns do not count independent stimulus inputs. This model is not fitted to people. Unequal clinical coverage, different clips, source selection and measurement filters limit scope. The contribution is a reproducible prediction benchmark, a scope diagnostic and a failed cross-modal test. It does not establish shared neural rank or causality. The package gives code, permitted inputs, exact source records, all outcomes, figures and repeat steps.

auditory cortex · intracranial EEG · fMRI · low rank · cross-validation · negative result · stimulus sampling

Admitted 1 citing publication

References

12
  1. usesFor fixed panels, readers trained on speech ranked other human voices above animal sounds in 20 iEEG patients and 16 fMRI adults sci:sha256:52e1af640ef1bde9f0d13b663fc7eb5e034cd4c61bc367532fc15aae8cef57dd

    We reuse its licensed electrical inputs, source records and BOLD extraction code. Its one-pattern audit motivates this new forecast endpoint.

  2. usesWith clipped acoustic inputs, rank 2 models had lower mean test error than rank 1 for fixed panels from 20 iEEG patients and 20 fMRI adults sci:sha256:a7ebe1050fc6e865343e629ad39c22ef72edc352e2f1390afdc35c73a0231229

    We adapt its MIT source-range reader and rebuild the licensed electrical input from public author bytes.

  3. Full sound filters did not confirm a gain over rank one in six iEEG patients and 25 fMRI listeners sci:sha256:5d8e1588cef40938c49cc8f56af0afd1a96e9f2eca79d1555dffc517f4f83ecc

    Earlier SCI work tested prediction beyond a rank-one sound filter in different studies. Our test uses spatial class means and disjoint sound examples.

  4. usesdoi:10.5281/zenodo.20599001

    Version 1.1 of the Hect et al. processed human iEEG data, CC BY 4.0.

  5. usesdoi:10.1016/j.cub.2026.07.063

    Source electrical study and preprocessing description.

  6. usesdoi:10.1152/jn.00588.2020

    Source B2021 human fMRI study.

  7. usesdoi:10.1371/journal.pbio.3002366

    Public author archive supplies B2021 response arrays and site metadata.

  8. doi:10.5281/zenodo.8349726

    Archived code for the fMRI author release; actual neural archive bytes are pinned by hash.

  9. usesdoi:10.1016/j.cub.2022.01.069

    Pinned author repository supplies sound labels and acoustic features; these original files remain external.

  10. doi:10.1016/j.neuron.2015.11.035

    Prior evidence for distinct auditory response patterns. We do not claim their first discovery.

  11. doi:10.1016/j.neuroimage.2013.02.062

    Prior work on cross-validated estimates of response dimension. Our forecast is an application, not the first dimension estimator.

  12. doi:10.1038/s41467-021-21696-1

    Prior primary work distinguishes linear and nonlinear dimension. The scalar example is a specific quantitative scope check, not a new general principle.

Cited by

1 publication

Public catalog entries on Sciscape.