Nakwon Rim¹, Omid Kardan², Sanjay Krishnan¹, Wilma A. Bainbridge¹, Marc G. Berman¹; ¹University of Chicago, ²University of Michigan

Past work has suggested that perceiving natural scenes requires less cognitive resources compared to perceiving urban scenes, which may explain the cognitive benefits attained after interacting with natural environments. While studies have shown that natural environments have restorative benefits, the hypothesized mechanisms have not been rigorously tested. Here, we investigated whether perceiving natural scenes may consume less cognitive resources. First, we conducted a continuous recognition task to probe the memorability of images and found that natural images are less remembered, which suggests that fewer cognitive resources are used to process natural stimulation. Next, using a Canny edge detection algorithm, we analyzed the number and significance of edges in the images and found that the proportion of "strong" edges (edges with higher gradient magnitude) is smaller in natural images. This association suggests that the number of edges that are essential to capture the information of a scene is smaller for natural images, which aligns with theories that perceiving natural images consumes less cognitive resources. Finally, we analyzed the spectral properties of the images by applying a discrete cosine transform to 8 x 8 pixel tiles (similar to JPEG compression). We found that natural scenes have a larger proportion of their spectral energy in high-frequency coefficients compared to man-made scenes. Since the human visual system may be less sensitive to high-frequency information, this result implies that natural images have less information that will be processed by humans, which is consistent with the idea that perceiving natural scenes will be less taxing. In conclusion, we found that natural scenes are less memorable, have a lower proportion of strong edges, and contain more high-frequency information compared to man-made scene images. These findings are consistent with theories that posit that natural images are less taxing to process.

Acknowledgements: This research was supported by National Science Foundation Smart and Connected Communities grant (NSF S&CC) #1952050 to MGB.