One face coordinate supplied 99% of late MEG agreement and 36% of IT fMRI agreement in ten test rows per modality

We tested whether one class coordinate supplies more than half of visual distance agreement in both MEG and fMRI. Public targets came from one study of 15 adults and two matched photograph sets. Five rows defined the coordinate and reference maps. Ten other rows per modality tested the fixed rule. The coordinate was learned from late MEG only. It mainly separated faces from four other classes. We removed additive image effects and normalized each distance map. The coordinate supplied 98.87% of late MEG agreement and 35.87% of inferior temporal fMRI agreement. Pointwise 95% person intervals were 97.75–99.68% and 27.59–46.68%. Coordinate-minus-other agreement was 0.381152 in MEG and -0.006076 in fMRI. The primary Holm p values were 0.001953 and 0.966797. The shared majority rule failed. Coordinate contributions were positive in every test row in both modalities, but this descriptive result did not repair that failure. Rank distances and four visual controls retained the late point pattern. Person inference assumes the intended anonymous row order. Later bounds over all possible cross-photo person links retained the descriptive pattern. The shares concern measured residual-map agreement, not all neural information. This is a fixed evaluation of reused targets, not independent-study confirmation. Source processing, reference selection, image features, and measurement differences limit the scope. The package preserves code, source hashes and terms, plans, all results, figures, audits, and repeat instructions.

human vision · MEG · fMRI · face representations · representational geometry · cross-modal generalization · negative result · reproducibility