With linear acoustic controls, late sound maps from one ECoG panel agreed more with BOLD in 20 adults; an eight-adult check was uncertain

We tested whether late sound maps have more BOLD agreement than early maps after full linear acoustic control. A public ECoG release supplied one fixed panel of 32 distinct selected profiles. A separate fMRI study supplied 20 adults. The same 165 clips had verified neural sound IDs and names. Electrical windows were 0.1-0.5 s and 1.3-1.7 s. We removed 12 category directions, 69 acoustic directions and the span from independent sound error variances. A local plan and code lock preceded new scores. Mean late cosine was 0.049984; mean late minus early was 0.035756. Their 95% person intervals were [0.035231,0.064736] and [0.018342,0.053169]. All 20 late scores and 17 differences were positive. The fixed mean-and-majority rule passed, p=0.001288414. Category control alone favored early maps in all 20 people. The historical eight-person matched-window check was uncertain. One MRI repeat and two category-omission majority checks also failed. All remain. A constructed model gave a late advantage with identical true maps and unequal measurement noise. An exact corrected-kernel sensitivity allowed a mean-order tie at 4.5850% relative error. This is not a measured noise bound. Missing electrical patient labels and repeats prevent new-patient uncertainty or latent timing claims. Separate cohorts, source selection, fixed sounds, task and measurement filters limit scope. The package gives code, source pins and rights, all results, figures, independent arithmetic and source audits, and repeat steps. Original source arrays stay external. No common neural cause or broad population transfer is established.

human neuroscience · ECoG · fMRI · natural sounds · representational similarity · acoustic control · measurement noise · independent study · negative results