One bit per auditory site gave mean four-class accuracy above chance in 20 iEEG patients and 16 fMRI adults
We tested whether a fixed-count binary spatial code can classify new sound rows in separate human studies. The electrical input was signed intracranial local field potential means. The MRI input was BOLD response estimates. Each sound kept only which sites were in its upper half. We fitted a separate reader for each person. Every coded sound within a person had the same value distribution. Four people per study supplied development. A local code lock preceded tests in 20 other patients and 16 other adults. Mean four-class balanced accuracy was 31.82% and 60.78%, against 25% chance. Pointwise 95% person Student intervals were 28.55–35.08% and 58.83–62.73%. A shared mean, majority and common-label reference test passed two-claim Holm correction, p=0.002577. A fixed five-point mean loss limit failed, p=0.945945. Binary reading lost 7.51 points in iEEG and 0.62 points in MRI. Site shuffles gave means near chance. Six-site and cross-estimate checks retained positive mean scores. Another eight-person MRI acquisition gave 61.77%; person overlap was not resolved. Post-result sound refitting gave wider ranges. A later matched electrical baseline check gave 25.14%. Main minus baseline accuracy was +6.68 points, with 95% person interval +2.56 to +10.79 points. Acoustic descriptors separated the labels well. An exact one-scalar model gave perfect binary class reading. The result supports a compact reading method in the selected response summaries. It does not establish a category code independent of acoustic cues, binary neurons, source count, or causality. Reused public cohorts, source screens, sparse clinical sites, distinct sound panels and measurement filters limit scope. The package gives code, permitted inputs and packets, source versions and terms, fixed plans, every negative score, figures and repeat instructions. An integer solver matched all binary decisions. A fresh source rebuild matched all 110,682 included electrical means exactly. A clean replay matched 39 result, packet, data and figure files.
auditory cortex · intracranial EEG · fMRI · spatial order · binary code · sound classification · subject holdout · reproducibility
References
11- usesUp to four class examples reduced mean response pattern error after two amplitude controls in 20 iEEG patients and 16 fMRI adults sci:sha256:7022157afe1bb9ba7eea43f7a092273baa1257440813810284a25982d5525667
We reuse its verified licensed electrical inputs, source pins and MIT preparation and source-audit code.
- usesSite ranks gave mean speech versus animal AUC above 0.5 after training without speech in 20 iEEG patients and 20 fMRI adults sci:sha256:a5ef832c05cfd2fb1316e3336a3b3fbb4edfebf33108f70c527b561553e3cff7
Its auditory site-rank result motivates the stricter fixed-count binary code test. We test a new four-class endpoint on held-out sounds.
- usesOne bit per site gave sound match scores above chance in selected panels from six SEEG patients and 24 fMRI adults sci:sha256:4ee8594db22f3da99b83d6ec4e88c4f2e77b400cd193c2ba55b5f7c81946266c
Its one-bit within-group sound matching result motivates a test without a known test class or test-sound template.
- usesdoi:10.5281/zenodo.20599001
Version 1.1 public electrical source, with CC BY 4.0 terms and pinned source checks.
- usesdoi:10.1016/j.cub.2026.07.063
The human intracranial study supplied the source data and its measurement context.
- usesdoi:10.1152/jn.00588.2020
The primary separate fMRI study supplied B2021 response estimates.
- usesdoi:10.1016/j.neuron.2015.11.035
The additional fMRI acquisition supplied a fixed descriptive check.
- usesdoi:10.1371/journal.pbio.3002366
Its public author archive supplied version-pinned fMRI arrays and source processing details.
- usesdoi:10.1016/j.cub.2022.01.069
Its pinned repository supplied the MRI class reference and acoustic descriptors.
- doi:10.1080/02643294.2022.2085085
- doi:10.7554/eLife.82566
It explains inference over both people and conditions; our naive resampling ranges are stated as sensitivity checks.
Cited by
14 publicationsPublic catalog entries on Sciscape.
- usesNormalized spatial quantiles separated speech or song from other sounds in six SEEG patients and 16 fMRI adults
Its site-identity-only reader motivates the complementary spatial-distribution restriction. Electrical sources and labels differ; this is not an independent replication of its endpoint.
- usesMaps fit to other sound classes predicted fixed animal sound patterns in 20 iEEG patients and 16 fMRI adults
Licensed electrical window means, baseline, chunk map and MIT source-audit helpers.
- usesSound labels lowered error for omitted acoustic clusters in 18 iEEG patients and two fMRI groups
Electrical compact extraction, source chunk pins and MIT HTTP range reader.
- usesAt 20 calibration sounds, matched class examples lowered association forecast error in 14 iEEG patients and 16 fMRI adults
Earlier electrical extract, chunk map and MIT range-reader code.
- usesA fixed sound group forecast failed a shared test in 14 SEEG patients and 16 fMRI adults
Upstream electrical extract and compressed source range map.
- usesFor fixed auditory tables, exact sampling risks favored a common estimate over means from four sounds per class in 19 of 21 iEEG patients and 0 of 24 fMRI adults
Inherited public byte-range reader and source extraction map.
- usesA fixed test did not confirm a shared sound class effect in 14 iEEG patients and 16 fMRI adults
This credited source chain supplies licensed numerical extracts or descriptor measurements through SCI 973fadb1.
- usesWith 165 sounds per study, matching repeats gave higher spatial forecast scores in six SEEG patients and eight fMRI adults
We reuse its licensed compact electrical means and baseline, with exact source records and credit.
- usesA fixed 32-sound selection rule failed its shared accuracy test in 21 iEEG patients and 16 fMRI adults
Licensed electrical derivative and source chunk-map ancestry.
- usesWith 32 calibration sounds, acoustic prior transfer failed a shared test in 20 iEEG patients and 16 fMRI adults
Upstream original-source chunk map and licensed electrical/baseline summaries, obtained through SCI 479585e9 and cf4efd8f.
- Fixed tests did not confirm shared sound responses below baseline in six SEEG patients and 16 fMRI adults
Related binary spatial reading benchmark reviewed before the question and code lock. Its relative bits differ from our baseline-sign test.
- usesFixed tests did not confirm shared animal–voice transfer in both directions in 16 iEEG patients and 16 fMRI adults
Verified licensed electrical means, duration-matched baseline, labels, source pins, MRI preparation and source audit code.
- usesTwo response comparisons gave mean four-group accuracy above chance for fixed sounds in 20 iEEG patients and 16 fMRI adults
Licensed electrical extract, source audit code and rank-reader precedent.
- usesAt 20 calibration sounds, mean class gains over acoustic neighbors were larger in association than early auditory parcels in 14 iEEG patients and 16 fMRI adults
Original electrical extract donor and exact source chunk map, via cf4efd8f.