A fixed half-energy test failed for speech/song directions and nine voice clips in six SEEG patients and 16 fMRI adults

We tested whether one learned speech/song direction retains more than half of a nonspeech human voice mean contrast. Separate public studies supplied SEEG high-gamma amplitude and BOLD fMRI. One repeat fitted a direction without voice clips. Two other repeats supplied total and projected contrast products. Three rotations gave one score per person. A local plan preceded new scores in six patients and 16 MRI adults. The primary shared test failed, p=0.983741. Mean projected minus half-total energy was -0.006778 [-0.012718,-0.000838] in SEEG and +0.001210 [-0.006548,+0.008968] in fMRI. Brackets give 95% independent-person Student limits in squared calibration-SD units per site. Ratios of mean energies were 0.306 and 0.508. An added study of eight MRI adults gave 0.490. These ratios have descriptive status. Mean projected and residual scores had positive pointwise limits in both primary studies. Those components do not rescue the failed claim. Keeping spatial means changed concentration. Joint person and sound resampling crossed zero for the SEEG projected score and half margin. Known models gave residual energy with parallel true contrasts, and positive scores with shared repeat errors or future-based selection. Thus no common majority-energy transfer was confirmed, and residuals cannot count neural sources. Clinical sampling, source screens, measurement filters, separate cohorts, fixed sounds and the chosen metric limit scope. Code, source pins and terms, all results, negative checks, figures and repeat steps are supplied. Original neural arrays stay external.

human neuroscience · SEEG · fMRI · voice · speech · cross-validation · response projection · negative results · measurement error