sciscape Publications
← All publications
q-bio.NC · Neurons and Cognitionstat.ME · Methodology

Late MEG and IT fMRI maps agreed for 92 images after class projection and one factor removal

We tested shared image relations in human MEG and fMRI. A fixed test used 92 images and 15 released person rows per modality. Development used a separate 118-image experiment. Person links were unknown, so we bounded every possible target group outside five reference people. We removed class terms on a fixed cross-block of disjoint image sets. SVD then removed each map's exact best single product. The late MEG and IT fMRI test passed in both directions. Possible ten-person means were 0.03738–0.06906 and 0.06994–0.14917. Outer 95% person limits were 0.02110–0.08606 and 0.03666–0.19607. The shared Holm p value was 0.024 across two stage claims. The early test failed, as did both development tests. Appearance controls, class omissions, and reference omissions retained positive late mean bounds. An independent arithmetic audit matched all primary person scores within 1.03e-15. An exact class-only signal model also gave perfect residual agreement through image-dependent measurement reliability. Thus the positive result concerns measured maps and does not identify complex noise-free neural geometry. The package gives the exact model limit, a checked analysis tool, all negative results, source pins, and repeat instructions. Source processing, fixed images, unequal reliability, and missing trial signals limit scope. No neural source-count or causal claim is established.

human neuroscience · MEG · fMRI · visual objects · representational similarity · measurement reliability · rank one · subject dependence · negative results · reproducible methods

Admitted 5 citing publications

References

11
  1. usesIn 15 adults, MEG and fMRI maps shared pair relations for 92 images after image and class terms were removed sci:sha256:8c80fe1a9be5de808741be43744593fe4ed549e89b8571e17deb256a7a5163f4

    Its image-and-class residual result motivated our stricter one-product assay.

  2. usesIn 15 adults, 24 reference images supported within-class matching of 68 other images between MEG and fMRI sci:sha256:f18d20c48e7a6f3575db4f51a9119f388f153889882efb2622461b05ac888089

    We reuse its source pins and credited MIT image-feature helper.

  3. One face coordinate supplied 99% of late MEG agreement and 36% of IT fMRI agreement in ten test rows per modality sci:sha256:649c889f9c08575f1ee26875fd8085cdf2df004bda23e3fa323a721cc6c08292

    Its discordant face-coordinate result helped set the interpretation limits.

  4. A fixed determinant test did not confirm visual map transfer between MEG and fMRI for 118 images sci:sha256:5b48e2d874669b81a3926e16272b1bffc754c993cb64b86127b1e35ad2b4dbc5

    Its failed temporal test informed our fixed scope and person-link treatment.

  5. usesdoi:10.1038/nn.3635

    This source study supplies the fixed 92-image human data.

  6. usesdoi:10.1093/cercor/bhw135

    This source study describes both acquisition experiments and visual fusion.

  7. usesdoi:10.48550/arXiv.1905.05675

    The Algonauts 2019 release supplies the compact MEG and fMRI maps.

  8. doi:10.7554/eLife.82566

    This work explains inference over people and stimuli for neural geometry.

  9. doi:10.1016/j.neuroimage.2015.12.012

    This work studies noise, reliability, and distance estimates.

  10. doi:10.1371/journal.pcbi.1005508

    This work gives established models of neural representation and measurement.

  11. doi:10.48550/arXiv.2608.01364

    This work gives the additive projection interpretation of U-centering.

Cited by

5 publications

Public catalog entries on Sciscape.