Sixteen acoustic directions failed a shared prediction gain test in 20 iEEG patients and 16 fMRI adults

We tested whether acoustic directions from one neural measure improve response forecasts in another. Separate public studies supplied signed iEEG field means and BOLD fMRI. Five patients and four MRI adults supplied development. A local lock set 16 directions before tests in 20 other patients and 16 other adults. The shared rule required gains over acoustic PCA and a source-design basis, plus skill over class-constant prediction, in both studies. It failed, p=1. PCA gains were -0.384 [-1.483,0.715] and +2.203 [1.805,2.601] points, with 95% person Student intervals. Skills were +3.527 and +13.328 points, with positive person intervals. The MRI PCA-gain range crossed zero when reference people varied. Shuffled source labels also gave useful recipient forecasts. We provide the exact class-shuffle covariance, a tested source-design reference, and all negative results. Separate cohorts and corpora, clinical coverage, source screens, unequal filters and uncertain feature extraction equivalence limit scope. No positive shared PCA gain or common neural cause was established. The package preserves permitted inputs, new statistical summaries, code, provenance, figures and repeat steps.

auditory cortex · intracranial EEG · fMRI · acoustic encoding · transfer learning · negative result · source design · subject dependence