A fixed personal sound forecast failed a shared A4 test in 12 iEEG patients and 16 fMRI adults

We tested whether a person's sound-category distance profile predicts that person's profile for different sounds. Separate public human studies supplied signed intracranial EEG and BOLD fMRI. The test used auditory parcel A4, hemisphere-matched group references, disjoint sound panels and separate calibration on each side. Five patients and four adults supplied references. Fixed tests used 12 other patients and 16 other adults. The personal rule added one quarter of the predictor profile after removal of a group direction. Mean squared-error gains were -0.002467 and 0.006797. Person 95% intervals were [-0.008408, 0.003473] and [0.001824, 0.011769]. Only 2/12 and 12/16 people improved. Person-identity shuffle p values were 0.4967 and 0.0002. Mean gains divided by baseline errors were -4.41% and 21.58%. Full resamples of people, reference people and sounds gave gain ranges [-0.049546, 0.013722] and [-0.006076, 0.023698]. The shared rule failed, conjunction p=0.996826. Later electrical responses gave a positive planned sensitivity; it did not replace the failed test. A model with one common population of three site types gave positive mean personal gains in 200/200 realizations without personal tuning parameters. This limits trait claims from fixed-site prediction. Sparse clinical sampling, different cohorts and clips, source selection, within-parcel placement and measurement effects limit scope. We establish neither shared personal neural traits nor a causal modality difference. The package preserves code, permitted inputs, exact source versions and terms, plans, all results and draws, figures and repeat steps.

human neuroscience · intracranial EEG · fMRI · auditory cortex · individual differences · stimulus sampling · negative result · reproducibility