Abstract
Responses of single sensory neurons to stimuli are ‘noisy’, varying substantially across repeated trials of identical stimulation. Intriguingly, these individual ‘noise responses’ (NR)—deviations from their means—are not isolated; rather they are highly correlated, referred to as ‘noise correlation’ (NC). From a computational viewpoint, the presence and nature of NC exert great impacts on the information processing capacity of neurons as they encode sensory events as a population, decode those encoded neural responses, and contribute to perceptual choices for action. Regarding the origin of NR, on the other hand, there has been growing evidence pointing to its tight linkage with ‘spontaneous responses’ (SR)—fluctuations of neural activity in the absence of external input or tasks. To investigate the functional structure of NC and its relationship with ‘correlations in SR’ (SC), we defined population receptive fields (pRFs) of unit volumes of gray matter (UV) in human early visual cortex and computed NRs and SRs using fMRI. NC increased with an increasing degree of similarity in pRF tuning properties such as orientation, spatial frequency, and visuotopic position, particularly between UV pairs close in cortical distance. This ‘like-to-like’ structure of NC remained unaltered across scan runs with different stimuli, even among between-area UV pairs. SC was higher than NC, and its functional and temporal structures were quite similar to those of NC. Furthermore, the partial correlation analysis revealed that NC between a given pair of UVs was best predicted by their SC than by any other factors examined in the current study.