In selected panels, rank readers kept mean speech versus animal AUC above 0.5 after each parcel omission in 15 iEEG patients and 20 fMRI adults

We tested whether voice-related information remains in recorded auditory inputs when any one sampled atlas parcel is omitted. Separate public studies supplied human intracranial EEG and fMRI. Four electrical patients supplied development. A fixed test of a new outcome used 15 other patients and all 20 fMRI adults. Each reader learned voices without speech versus other sounds, then scored speech versus animal sounds. Inputs were within-sound response ranks. We refitted after each parcel omission and took each person's lowest test AUC. Mean lowest AUC was 0.63922 in iEEG and 0.92471 in fMRI. The 95% person Student intervals were 0.55460 to 0.72383 and 0.88123 to 0.96819. Scores exceeded 0.5 in 14 and 20 people. The single shared test passed, p=0.001670. Common-label references recomputed every minimum. Crossed person-and-test-clip percentile intervals were 0.50906 to 0.72941 and 0.84412 to 0.97500, with training clips and fitted readers fixed. Mean costs from intact readers were 0.15686 and 0.04706 AUC. With at most two sites per parcel, the iEEG mean was 0.55225 and its 95% person interval was 0.48686 to 0.61765. Early electrical timing and two-parcel omissions were also uncertain. All negative and discordant outcomes remain. One scalar source also gave perfect worst-omission transfer. Thus the result supports redundancy in these processed measurements. It does not show independent neural sources, biological compensation, a non-acoustic code, or causality. Clinical coverage, field spread, reference and baseline processing, source selection, unequal sampling and different sound panels limit the comparison. These are new outcome tests in reused cohorts. They do not confirm the result in new acquisition cohorts. The package supplies code, permitted inputs, source versions and terms, plans, all scores, figures, audits and repeat instructions.

human auditory cortex · iEEG · fMRI · response ranks · regional input omission · sound transfer · measurement redundancy · independent cohorts