Acoustic fits without voice responses gave uncertain contrast gains in 20 iEEG patients and positive gains in 16 fMRI adults

We tested whether acoustic models fit without voice responses predict the spatial voice-minus-nonvoice contrast. Separate public studies supplied signed intracranial EEG and BOLD fMRI. Four people per modality supplied development. A fixed test used 20 other patients and 16 other adults. Model fitting and penalty choice used only nonvoice responses. Held-out nonvoices supplied the contrast baseline. The score S compares squared contrast error with a zero spatial-pattern reference after site scaling and mean removal. Mean S was -0.002924 in iEEG, with a 95% person Student interval of -0.022870 to +0.017021. It was +0.092795 in fMRI, with interval +0.069751 to +0.115840. Scores were positive in 12/20 and 16/16 people. The shared positive test failed, p=0.251722. Electrical window, parcel-weight, and five-site checks remained uncertain. A separate-estimate BOLD check stayed positive. Later test-clip and person resampling retained the same uncertainty pattern. Fifteen electrical patients had positive map cosines; three still had higher contrast error. A simulated population with only acoustic mean responses also often failed the contrast test. Thus failure cannot identify a non-acoustic voice code. This is a new endpoint in reserved people from reused studies. Source selection, clinical sampling, tasks, features, and signal resolution differ. No positive cross-modal or causal claim is made. The package gives permitted electrical inputs, source versions and terms, code, all results, figures, local plans, audits, and repeat steps. Original BOLD arrays stay external.

human auditory cortex · voice contrast · intracranial EEG · fMRI · acoustic encoding · category transfer · negative result