A tale of three types of internal object representations in the human brain

How do internal brain representations bridge seeing an object and thinking about it after it disappears? Both object knowledge and mental imagery are involved in this process, engaging overlapping perceptual regions, yet whether their neural codes are shared or distinct remains unknown. We compared people with ("visualizers") and without voluntary visual imagery ("aphantasics") using fMRI, to examine experience of imagery sensation, and a multimodal deep neural network model, to examine representational contents (encoding text vs. image). We found distinct types of internal representations: (1) the left lateral occipitotemporal cortex (LOTC) encoded visual-structured knowledge linked to imagery sensation; (2) the bilateral fusiform gyrus, left dorsal LOTC, and right inferior frontal gyrus encoded language-structured knowledge independent of imagery sensation; and (3) the left superior parietal lobule maintained visual representation without prior knowledge, also independent of imagery. These findings reveal functionally and computationally distinct neural mechanisms that bridge seeing and thinking of objects, differing in their reliance on knowing and internal experiencing.