Schedule of Events | Search Abstracts | Invited Symposia | Symposia | Poster Sessions | Data Blitz

Neural mechanism underlying the construction of self-generated compositional imagery

Poster Session F - Tuesday, March 10, 2026, 8:00 – 10:00 am PDT, Fairview/Kitsilano Ballroom

Shitong Zhao^1,2 (), Qing Yu¹; ¹Institute of Neuroscience, Chinese Academy of Sciences, ²University of Chinese Academy of Sciences

Mental imagery enables us to flexibly generate internal representations that are not directly perceived by the visual system. Growing evidence indicates that mental imagery is a voluntary and generative process involving the prefrontal cortex (PFC), rather than simply “reverse perception.” However, the specific functional role of the PFC in self-generated imagery remains unclear. Here we investigated how the brain self-generates imagery content with a compositional structure. Participants played a Tetris-like game with fMRI recordings. They were asked to either mentally combine two specific shape features into an object with a predetermined contour (cue-induced imagery), or to volitionally decide which shapes and/or objects to imagine (self-generated imagery). This design allowed us to examine whether self-generated imagery signals in the PFC are feature-like, object-like, or both. Whole-brain activation analyses revealed clusters showing increasing activity that scaled with increasing self-generation demands in the frontoparietal network and default mode network, implicating the role of these networks in self-generated imagery. Moreover, whole-brain searchlight analyses (multivariate decoding and representational similarity analysis) provided converging evidence for representational differences between self-generated and cue-induced imagery. Superior representation of self-generated objects was observed in brain regions including the dorsomedial (dmPFC) and anterior lateral (alPFC) PFC. In contrast, feature representations engaged a distributed cortical network that remained comparable regardless of self-generation demand. This dissociation suggests that voluntary construction of self-generated compositional imagery relies primarily on PFC subregions to assemble high-level object representations, whereas the manipulation of detailed compositional features recruits a common imagery network.

Topic Area: THINKING: Other