A fixed sound change test did not confirm shared gains in 18 ECoG patients and five fMRI adults

We tested whether common changes across sound bands improve neural music prediction beyond changes within each band. Separate public studies supplied ECoG from 29 patients and fMRI from five adults. Six patients supplied development. A fixed test retained 18 other patients; five failed the training quality rule. The fMRI test used 20 music clips excluded from fitting. Mean normalized error gain was +0.000221 in ECoG and -0.000914 in BOLD. The 95% person intervals were -0.000250 to +0.000692 and -0.001801 to -0.000027. Joint person and common music block ranges crossed zero in both measures. Gains were positive in 9/18 and 0/5 people. The shared positive gain sign p was 1. Stronger level controls and spatial mean removal did not establish a shared gain. An electrical advantage over a shifted feature did not transfer to BOLD. All negative outcomes remain. This result limits one acoustic feature and fitting procedure. It does not show absent neural integration across sound bands or a causal modality difference. The package gives code, source versions and terms, fixed plans, a logged implementation repair, all losses and models, figures, and repeat steps. Original source arrays and music stay external.

human neuroscience · ECoG · fMRI · music · acoustic interactions · out of sample prediction · negative result