Under source image order, fixed tests did not confirm photo transfer within classes in MEG and fMRI from 15 adults

We tested whether detailed visual relations remain stable when both the photograph and the person change. Public MEG and fMRI targets came from the same 15 adults in separate sessions. Two sets of 78 pictures had matched verbal descriptions. We used the intended source image order. Within each of five classes, we removed additive image effects and averaged normalized maps from five reference people. We scored the other photo set in both directions. Anonymous rows did not establish person links. We therefore bounded all possible groups of ten target people outside each reference. Late MEG mean ranges were 0.02268 to 0.05892 and -0.00770 to 0.04810. IT fMRI ranges were -0.01813 to 0.04394 and -0.03296 to 0.01638. All four outer 95% mean interval ranges crossed zero. The shared transfer rule and the late-stage advantage rule both failed; their Holm values were 1.0. One MEG image-label test gave p=0.0167 but did not pass its person rule. A same-photo check found weak detailed agreement in these fMRI targets. Thus, failure does not establish absent invariant neural information. This new endpoint extends an earlier failed paired-photo test. It is not independent acquisition confirmation. The package gives code, exact source records, all scores and label draws, figures, audits, and repeat steps. Source arrays remain external under their own rights. Source image order, fixed references, limited target quality, visual confounds, and measurement differences limit scope. No positive cross-modal neural generalization or causal claim is made.

MEG · fMRI · human vision · image classes · photograph transfer · representational geometry · person dependence · negative result · reproducibility