Importantly, we find good correspondence of the components between the MEG, EEG, and fMRI datasets. This indicates the PLSR identified robust neural representations which are at least partially shared over subjects, imaging modalities, and experimental paradigms.