Three fixed acoustic tests did not confirm a shared gain in six iEEG patients and 25 fMRI listeners

We tested whether added spectral features improve later prediction of human auditory responses. Separate public studies supplied six reserved iEEG patients and 25 fMRI listeners who heard different stories. Three other iEEG patients supplied exploration. Fixed tests used spectral diversity, 50 ms band rises, and a regularized rise model. Flexible marginal sound responses and signed and pooled change controls were included. All three shared positive claims failed Holm correction, with p=1. The regularized model gave mean gains of +0.0334 and -0.0008 R-squared percentage points. Their 95% person t intervals were -0.0140 to +0.0808 and -0.0524 to +0.0509 points. Post-analysis upper limits covered all six primary mean effects as one family. Under independent normal person scores, the regularized-model limits were 0.0986 and 0.0637 points, with at least 95% joint coverage. No practical equivalence margin was set in advance. The result limits one fixed prediction procedure. It does not establish absent onset coding, a shared neural mechanism, or causality. Clinical sampling, source selection, different stories, spatial support, full-run processing and hemodynamic filtering limit scope. The packet gives compact inputs, source versions and terms, plans, all failures, code, figures, loss audits and repeat instructions. Fresh checks matched eight audio frames, 50 fMRI series and one full electrical derivative reconstruction.

human neuroscience · intracranial EEG · fMRI · speech · acoustic encoding · spectral change · prediction · negative results