Four linear axes failed a shared prediction test against a scalar curve in human SEEG and fMRI sound responses

We tested whether four linear response axes predict sound responses better than a fitted scalar curve. Public studies supplied separate human SEEG and fMRI cohorts. Two patients and four MRI adults supplied development. A local code lock preceded new scores in six patients and sixteen other MRI adults. Five folds excluded each target sound from fitting in all three response estimates. Two estimates then predicted the third. Mean gains over the curve were 0.126 reference-error points in SEEG and 1.965 in main fMRI. Their 95% person Student limits were [-0.891,1.143] and [1.213,2.717]. Eight additional MRI records gave 4.503 [3.521,5.486]. The shared gain-and-skill rule failed, p=0.38136. A planned control without extra SEEG trial baseline subtraction had positive mean intervals. An added exploratory check found that subtraction raised measured repeat disagreement by 7.96 times, with person limits [6.52,9.40]. A noiseless scalar example also gave positive model gains. Thus the model comparison does not count neural drives. Source selection, small clinical samples, separate cohorts and different filters limit inference. No positive cross-modal generalization or causal modality effect was confirmed. Code, source pins, all outcomes, figures and repeat instructions are supplied. Original neural arrays stay external.

human neuroscience · SEEG · fMRI · auditory cortex · natural sounds · principal curves · response prediction · baseline subtraction · negative result