A test of adaptation mixtures that start at zero did not show shared violations in five ECoG patients and six fMRI adults

We tested a bound on visual responses in separate human ECoG and fMRI studies. The model allows any positive mixture of adaptation processes. Each process starts at zero and has the same build and recovery rate. Three ECoG patients defined one contrast. An exact rational polynomial certificate proves that all model time constants obey its bound under the ideal pulse schedule. Five other ECoG patients and six fMRI adults tested the fixed contrast. Four of five ECoG scores were negative (one-sided sign p = 0.1875). The mean was -0.06277, with a 95% person-bootstrap interval from -0.09853 to -0.02214. Eight fMRI records contained six people with unknown links between two repeated people. Across all 72 possible links, zero to two people had negative scores (p = 0.890625 to 1). Mean fMRI scores ranged from +0.04103 to +0.07336; the interval union included zero. The fixed shared rule failed. We also supply an exact tolerance for the saved ECoG sample schedule. A simulation shows that a nonzero initial adaptation state can violate the original bound. Prior run-order concerns affect the condition with the largest contrast weight. No shared neural mechanism, equal biological time constant, or causal modality effect is established. The package preserves the model proof, code, unchanged public inputs, source terms, plans, all negative results, uncertainty checks, figures, and repeat steps.

visual cortex · adaptation · ECoG · fMRI · positive mixtures · model falsification · Bernstein certificate · independent confirmation