For 455 objects, all 256 source maps gave positive adjusted taxonomy scores in eight EEG and three fMRI adults

We audited two alternative WordNet senses for each of eight objects. Separate studies supplied eight EEG adults and three fMRI adults. All 256 source maps gave positive adjusted taxonomy scores in each person for 455 objects. Exact group ranges were 0.03244 to 0.03461 for EEG and 0.03370 to 0.03763 for fMRI. These bounds concern input coding in the observed panel, not physiological uncertainty. Unions of pointwise 95 percent person t limits were [+0.02259, +0.04475] and [+0.01077, +0.06081]. Their validity needs a suitable person-mean model. The coding audit is exploratory. The first fixed finite-panel reference passed, Holm p = 0.001. It conditions on the observed people and objects. Two stronger fixed tests did not establish a positive relation in new people. A local code lock preceded those tests. The first used 447 objects, eight EEG people and three MRI adults. Its mean partial correlations were 0.03366 and 0.03747. The second used 26 objects, reused EEG people and 18 adults from a separate MRI study. Its means were 0.04938 and -0.00167. The stronger shared Holm probabilities were 0.25 and 0.57. Three MRI people cannot give an exact one-sided sign-flip probability below 0.125. All negative results remain. Prior SCI work already tested taxonomic geometry in SEEG and fMRI. The distinct contribution is a source sense audit, an exhaustive coding tool and a new controlled benchmark. Cohort, task, image, sampling and measurement differences limit scope. No common neural cause or causal effect was established. Code, permitted inputs, provenance, all results, figures and repeat steps are supplied.

taxonomy · EEG · fMRI · WordNet · representational similarity · source audit