Cortical activity is highly variable across time, but the origin of this variability remains unclear. Increasing evidence suggests that not only local stochastic processes (i.e. noise) cause neural response variability; rather, a large share of cortical variability may arise due to more global cognitive processes, such as arousal or attention, affecting the responses of whole populations of neurons. We used functional MRI to investigate the nature and potential source of trial-by-trial activity fluctuations in human visual cortex. Participants viewed annular gratings in a range of orientations. The measured BOLD activation patterns varied from trial to trial, even when the orientation of the stimulus was held constant. To examine the structure of across-trial variability in cortical responses, we computed voxel-by-voxel correlation matrices from activity in areas V1, V2, and V3 combined. This revealed that voxels of similar orientation preference share more response variability than voxels with dissimilar tuning properties. The strength of this effect diminished with increasing Euclidean distance between pairs of voxels, but interestingly, remained significant up to distances of more than 5 cm. Crucially, spatial proximity alone was a poor predictor of shared response variability, ruling out explanations based on spatial smoothing artifacts of the imaging method (such as those due to the point spread function, or small movements by the participant). The global nature of these spontaneous fluctuations suggests long-range feature-tuned connections across retinotopic space.
展开▼
