A fixed loudness peak rate test failed across 20 iEEG patients and 16 fMRI adults with whole sound classes excluded

We tested whether neural measurements improve prediction of loudness peaks per second beyond nine acoustic summaries. Whole sound classes were excluded from fitting and tuning. Separate public studies supplied human signed iEEG LFP means and BOLD fMRI estimates. Five patients and four MRI adults supplied development. A local lock preceded reserved scores in 20 other patients and 16 other adults. Mean neural gains were 0.723 and 12.839 points of within-class constant-prediction error. Their 95% person Student limits were [-0.441,1.888] and [10.021,15.657]. Mean full skills were -0.634 and 19.853 points. The fixed shared positive rule failed, p=0.865728. All negative people and classes remain. Eight other MRI adults gave a 13.107-point gain on the same sound panel. Expanded temporal acoustic controls left negative mean gain estimates. A later audit over all 19 fixed gain checks gave expanded-control 95% family upper limits of 0.875 and 0.496 points under valid Student mean assumptions. This audit is exploratory. A later source-flag screen left nine electrical patients and uncertain gains. The package gives a 308-item rate benchmark, licensed compact inputs, code, all results, source pins and terms, figures and repeat steps. Separate cohorts, source screens, unequal site counts, stimulus duration, rate quantization and measurement filters limit scope. Conditional sound resampling keeps fits fixed. This result does not establish positive neural generalization across modalities, absent electrical rate coding, unique neuronal rate information or causality.

human neuroscience · intracranial EEG · fMRI · auditory cortex · loudness peak rate · source class transfer · negative result · prediction benchmark