Abstract
Humans can easily judge not only the color or texture of a surface but also its emotional value such as pleasantness or disgust. Recent psychophysical studies suggest that human observers can infer the emotional value of natural surfaces from low-level image statistics and that observers can do so quickly and somewhat independently from the recognition of material category (e.g., Motoyoshi & Mori, 2016). To investigate the neural dynamics involved in such visual emotional processing, we measured visual evoked potentials (VEPs) for 150 images of natural surfaces presented for 500 ms separated by blanks of 750 ms. At each temporal point, we computed the correlation between VEP and the subjective rating for each surface that ranged from unpleasant (-4) to pleasant (4). The analysis revealed that large occipital VEPs (O1/O2) correlate with positive emotional values at 150-300 ms from stimulus onset and that frontoparietal potentials (Cz) correlate with negative emotional values at latencies less than 120 ms. In line with previous psychophysical data (Motoyoshi & Mori, 2016), an analysis of low-level image statistics showed that negative emotional values were associated with high luminance/color contrast at middle spatial frequencies and with high cross-orientation energy correlation at high spatial frequencies. Image statistics were partly correlated with emotion-related components in occipital (O1/O2) and parietal (Cz) cortex. The results suggest that emotion-related low-level image features of visual textures produce neural responses related to negative emotions in frontoparietal regions even more rapidly than in the early visual cortex.