Abstract
To perceive 3D motion, 2D motion signals from the two eyes must be combined. Previous work has shown that 2D motion direction can reliably be decoded from MRI BOLD activity throughout visual cortex. 3D motion presents a special case for motion encoding and decoding, since 2D motion direction can be inferred from monocular signals, but 3D motion requires that signals from both eyes are combined. Here we investigated if 3D motion direction can be reliably decoded from visual cortex. We measured the BOLD response (3 mm isotropic resolution) of human visual cortex to binocular stimuli moving in one of eight directions (leftward/rightward, approaching/receding, and the 4 intermediate directions), while observers performed a fixation task. The approaching/receding stimuli avoided the possibility of being decoded from spatiotopic cues, known to facilitate decoding of 2D motion direction (Wang, Merriam, et al, JoN 2014). We subjected the BOLD response to MVPA decoding analyses (ROI-wise or searchlight-wise). We found that the BOLD response is highly informative about 2D direction across visual cortex (V1, V2, V3, MT+, and beyond), replicating previous findings. BOLD response is less informative about 3D direction, with information concentrated in later areas (V3, MT, IPS). Interestingly, decoding errors for the direction of 3D motion mirror the typical perceptual errors in the perception of 3D motion (Fulvio, Rosen & Rokers APP 2015). Specifically, approaching and receding directions tend to be confused both perceptually and by the decoder. These results suggest that 3D motion is encoded by later visual areas. The limitations in spatial resolution of the BOLD signal are consistent with poor 3D motion decoding in V1. Finally, the neural and behavioral confusion of 3D motion stimuli may have a common basis.
Meeting abstract presented at VSS 2016