Temporo-Parietal Network Model for 3D Mental Rotation

Abstract

The ability to rotate objects mentally has been suggested to be related to recognition of visual objects presented from non-canonical viewpoints. However, the neural mechanism underlying this ability is still unclear. In this paper, a global neural network model is proposed. This model consists of two subsystems, a parietal network for mental rotation, and inferior temporal network for object recognition. In this model, it is assumed that mental rotation is realized by a process in which the egocentric representation of objects in the intraparietal sulcus is rotated by motor signals that are internally generated in the premotor cortex. The rotated information is sent downward to the visual cortex as a rotated visual image; meanwhile, object recognition is achieved by a matching process with target object images in the inferior temporal cortex. The parallel distributed processing of this model achieves robust object recognition from various viewpoints including the non-canonical view.