A fixed test supported animal transfer from mammals in IT fMRI but not in medial temporal spiking

We tested whether a readout trained on mammal images separates other vertebrates from selected nonanimal objects. Public studies supplied separate human intracranial spiking and IT fMRI cohorts. Three electrical patients supplied development. A local code lock preceded tests in 12 other patients and 15 fMRI adults. Mean AUC was 0.518204 and 0.794000. Their 95% person Student intervals were 0.477120–0.559288 and 0.711600–0.876400. The fixed shared claim failed, with one-sided conjunction p=0.175201. Appearance-adjusted means were 0.514592 and 0.768000. A later image-only orientation reader reached AUC 0.790 on the fMRI panel. This does not prove that image appearance explains the neural responses, but it limits semantic claims from transfer alone. An image audit corrected our interpretation of bark: this source category contains sailing ships. Removing it left electrical mean AUC 0.518529 with an interval across chance. All original and later results remain. Independent arithmetic reproduced the AUCs. Timestamp reconstruction reproduced 854183 finite source counts exactly and checked image order. Different cohorts, brain regions, sparse clinical sampling, unequal trial averaging, image panels and measurement filters limit the comparison. No common neural cause, absence of an animal code, or causal regional difference is established. The contribution is a new transfer benchmark and an audited negative cross-modal test. The package gives code, source versions and terms, plans, all results, figures and repeat steps.

human neuroscience · intracranial spiking · fMRI · animal representation · transfer testing · source audit