Abstract
How does the visual system extract shape and feature information from noisy input? Textures in our visual environment are defined by various spatial properties, including spatial structure, which may affect the extraction of shapes and features. Classic texture segmentation experiments have used pixel-to-pixel dependence to manipulate visual spatial structure, but the results can be fully explained by highly stereotypical visual objects (shapes and lines) rather than structure itself. Here, we implement two novel approaches to test perception of visual spatial structure using 4-alternative forced-choice tasks. We hypothesized that increased spatial structuring would lead to improved shape segmentation and extraction of orientation, affecting threshold and performance in segmentation tasks. In experiment 1, we tile a pattern aperiodically across the visual field and systematically manipulate tile size: the smaller the tile, the more frequent the repetition across the screen, leading to higher spatial structuring. Participants report the shape of a tiled target embedded in the background. In experiment 2, we manipulate spatial structure by varying the displacement distance of dots from a regular lattice. Each dot is first placed onto a standard grid and jittered in a random direction by one of five possible fixed distances. Participants indicate in which quadrant a cued orientation was most strongly represented. Our results show the expected correspondence between behavioral variables and spatial structuring. In experiment 1 (tiling), participants’ accuracy and response speed increase as tile size decreases. Similarly, in experiment 2 (spatial jittering), reaction times increase with dot displacement distance, following a classic psychometric function centered on similar values across participants. These results provide new insights into humans’ ability to discriminate spatial structure and suggest that visual spatial order aids both shape segmentation and feature extraction.