A fixed test did not confirm shared prediction of session changes in eight EEG people and three fMRI people

We tested whether early image responses predict session differences for images excluded from fitting. Separate public studies supplied scalp EEG and inferior temporal fMRI. Two EEG people supplied development. A fixed test used eight other EEG people and three fMRI people. Five disjoint repeat groups supplied predictors and two difference estimates. Mean error reduction was 1.043% in EEG, with a 95% person t interval from -0.0006% to +2.086%. The fMRI mean was -0.0163%, with interval -0.0867% to +0.0540%. The fixed shared rule failed. An added controlled check kept response templates fixed and varied independent artificial noise. Lower late noise gave positive first-calibration score means in all 11 templates. Mean gains were 12.12% for EEG templates and 8.45% for fMRI templates. Revised difference-score means were slightly negative in both groups. This is a method control, not evidence of neural change. Source EEG whitening and MRI repeat-consistency tuning used data before our holdouts. The small MRI cohort and different measurement spaces limit inference. We make no positive cross-modal neural or causal claim. The package preserves code, permitted compact inputs, source versions and terms, fixed plans, all outcomes, figures, independent audits and repeat instructions.

visual representation · representational drift · EEG · fMRI · session calibration · measurement noise · independent repeats · image holdout · negative result · reproducible benchmark