At 16 mapped object groups, a fixed SEEG template failed a shared test in 14 other patients and five fMRI adults

We tested whether object group relations from three patients predict other people after fixed controls. Separate public studies supplied human depth-electrode amplitude and BOLD fMRI. A label-only bridge linked 16 broad groups across the studies. Different image halves set group distances. We removed additive category terms, animacy and two source maps from one image-model family. A local code lock preceded reserved tests. Mean template correlations were +0.00511 in 14 other patients and +0.04591 in five MRI adults. Their 95% person Student limits were [-0.04410, +0.05433] and [-0.04983, +0.14165]. The fixed shared positive rule failed, value 0.4313. It required person mean, person sign-flip and whole-category reference rules in both modalities. The three training patients had negative leave-one-person-out scores before reserved testing. All negative people, group deletions and controls remain. The contribution is a reproducible category bridge and a failed transport benchmark. The result concerns a fixed noisy template and fixed image panels. It does not bound all possible neural relations. Clinical source row order, sparse tissue sampling, broad class pooling, separate cohorts and different measurement filters limit scope. The label reference is not an exact test of independence from visual appearance. No positive neural generalization or common cause was established. Code, source versions and terms, complete scalar results, figures and repeat steps are supplied.

human neuroscience · SEEG · fMRI · object representation · category bridge · negative result · representational geometry