For fixed odor panels, amygdala response ranks reduced valence error in pooled human neurons and on average in three fMRI adults

We tested whether relative amygdala responses predict valence for odors excluded from training. Discovery used one released panel of 622 human amygdala units without donor labels. A fixed study test used separate fMRI data from three adults, with 160 odors per person. Sites were scaled with training odors. The reader kept only their response order. Error fell 18.22% in the 15-odor neuron panel and 3.61% on average in fMRI. Individual MRI changes were -6.87%, +15.86%, +1.85%. The conditional shared label test passed, p=0.0027. The 95% person Student interval for MRI mean skill was -24.87% to 32.10%. The result does not establish performance across the human population. The neuron release cannot support donor uncertainty. Separate cohorts, unequal odor panels, different rating targets, clinical sampling, and measurement filters limit scope. Intensity checks and all negative scores remain. This is a new response-order test in reused studies. Earlier work already established odor-valence coding. The package gives code, permitted neuron inputs, external MRI source pins, all results, figures and repeat steps. No common neural cause or causal effect is established.

olfaction · amygdala · single neurons · fMRI · response ranks · valence · odor identity · cross-validation