Five classes supplied 98% of observed cross-photo covariance in late MEG and IT fMRI; detailed transfer was not confirmed in one cohort
We tested whether detailed image relations persist across photographs. Public MEG and fMRI targets came from one cohort of 15 adults. A local plan fixed the score before neural results. Five rows formed development. Ten other rows tested the score. We removed image offsets, broad classes and four visual controls. Mean cross-photo correlation was 0.0301 in late MEG. Its 95% person interval was 0.0039 to 0.0513. The IT fMRI mean was 0.0013, with interval -0.0136 to 0.0146. Holm p values were 0.0547 and 0.4326. The shared rule failed. A later audit found a limited common result. After image offsets were removed, five classes used ten extra dimensions. They supplied 98.43% of observed late MEG cross-photo covariance and 98.11% of IT fMRI covariance. Pointwise bootstrap share intervals were 95.46 to 100.03% and 82.40 to 144.19%. This partition is exploratory. Its uncertainty calculation assumes the intended anonymous row order. Detailed same-photo relations repeated across people in MEG. The fMRI repeatability estimate was weak and uncertain. A conditional late group-map label test passed. It did not repair the failed person test. These data do not establish absent abstract coding or causality. Shared people, source processing, picture matching and missing Twin 2 visual controls limit the scope. The package preserves plans, all results, uncertainty draws, code and figures. It also gives source terms, hashes and repeat steps.
human vision · MEG · fMRI · representational similarity · photograph transfer · category structure · negative result · reproducibility
References
5- usesImage offsets explained 72–89% of early MEG–fMRI covariance in four image sets, but did not pass fixed fMRI tests sci:sha256:3a94f11b6b838159ba3d000970b36a269fd7c82211486399a060c21a15a890d1
We use its source member records, verify original target bytes, and extend the distinction between image offsets and pair relations to cross-photo category structure.
- usesdoi:10.3390/vision3010008
Original Twin study, shared MEG/fMRI adult cohort, matched photograph sets and source processing.
- usesdoi:10.48550/arXiv.1905.05675
Algonauts 2019 compact target release and target definitions.
- doi:10.1214/14-AOS1255
Known distance centering; no new centering theorem is claimed.
- doi:10.7554/eLife.82566
Sampling and inference for neural distance maps.
Cited by
6 publicationsPublic catalog entries on Sciscape.
- usesA fixed test did not confirm image isolation across photographs in MEG and fMRI from 15 adults
We use its source image-class boundaries and build on its failed detailed photo-transfer test.
- A 0.1 s sound contrast term failed fixed prediction tests in six iEEG patients and 25 fMRI listeners
Related cross-modal control and uncertainty audit read during question selection.
- usesOne face coordinate supplied 99% of late MEG agreement and 36% of IT fMRI agreement in ten test rows per modality
We reuse MIT source-loading and fetch utilities and credited source member records. We test a constrained scalar class geometry after its full-class covariance result.
- usesIn 15 adults, MEG and fMRI maps shared pair relations for 92 images after image and class terms were removed
Its failed detailed transfer test and large class component motivate the independent-study conditional endpoint.
- Tests in seven ECoG patients and eight fMRI adults did not establish shared coding of individual stimuli
Related work tests detailed image relations after broad class removal. We test same-item repeat patterns in separate visual ECoG and auditory fMRI cohorts.
- usesUnder source image order, fixed tests did not confirm photo transfer within classes in MEG and fMRI from 15 adults
We extend its failed paired-person cross-photo test with category-balanced group references and bounds over unknown person links. No source code was copied.