Tests with one sound class left out did not confirm shared gains from seven loudness terms in 21 iEEG patients and 24 fMRI adults

We tested whether seven measures of loudness change improve auditory response forecasts for a sound class absent from training. Separate public studies supplied signed intracranial field means and BOLD fMRI. Four patients and four MRI adults supplied development. A local code lock preceded new endpoint tests in 21 patients, 16 B2021 adults and eight NH2015 adults. Each model used nested tests that left out whole sound classes. Two fixed claims compared 16 acoustic means with 23 terms, and 81 other terms with all 88 terms. Each claim required positive mean added gain and full-model skill in all three groups. Both claims failed after Holm correction, p=0.304688. With 81 other terms, mean gains were 0.395, 3.056 and 0.776 error points. Their pointwise 95% person Student limits were [-0.485,1.276], [1.886,4.226] and [0.223,1.329]. The electrical gain also had a range that crossed zero when people and sounds varied. All negative people, classes and controls remain. A post-result four-claim family supported rich-model mean skill in all studies, Holm p=0.015625. This is exploratory and does not confirm added value from the seven terms. This is a new test in reused public people. It is not independent confirmation of the earlier exploratory anatomy result. Separate cohorts, different sounds, task, feature extraction, source screens and signal filters limit scope. No shared neural mechanism or causal effect was established. The package gives licensed compact inputs, source pins and terms, all results, figures and repeat code.

human auditory cortex · intracranial EEG · fMRI · sound class transfer · loudness · negative result · nested validation