Abstract
Our visual perception is usually determined by an interaction between externally driven “bottom-up” sensory information and internally generated “top-down” signals guiding interpretation of the sensory input and reflecting prior knowledge and intent. Mental imagery, which relies entirely on this top-down signal, provides a means to investigate the nature of top-down signals independent of bottom-up perceptual processing. While prior studies have suggested similar neural substrates for imagery and perception, the specificity of imagery representations and their distribution across visual cortical areas remain unclear. Here we used functional magnetic resonance imaging and multi-voxel pattern analysis in an ungrouped event-related design to conduct a detailed comparison of the neural activation patterns to ten individual objects throughout visual cortex during both visual imagery and perception. We found that visual imagery as well as perception activated a precise and fine-grained representation allowing decoding of individual imagined objects throughout ventral visual areas. Further, there was a close correspondence between the imagery and perceptual representations, allowing cross-decoding between them. However, while there was a posterior-anterior gradient of performance with better perceptual decoding in early visual cortex compared with object-selective regions, imagery decoding was similar across ventral visual areas. Moreover, the structure of representations across visual areas were more similar in imagery than perception, suggesting less transformation of visual representations between areas during imagery. Finally, the correspondence between imagery and perception was correlated with independent behavioral measures of participants' vividness of mental imagery. These results suggest that while perception and imagery share similar neural substrates, even at the individual object level, they remain distinct particularly in terms of the distribution and similarity of representations across visual areas.
This work was supported by the US National Institutes of Health Intramural Research Program of the National Institute of Mental Health.