Abstract
Estimating average properties of ensembles is important when there is a need to act upon the ensemble, such as controlling the movement of a swarm of semi-autonomous agents. We explored observers' accuracy in perceiving both the center of the gestalt global shape and the average item position of ensembles of dots. Ensembles of 12, 18, or 24 dots were drawn within an invisible circle, with half on the perimeter. In separate blocks, we instructed participants to click on the Center of the Inferred Shape (i.e., the origin of the circle) or the Center of the Dot Positions (i.e., the average position). Baseline trials were constructed such that the two centers were identical. Test trials were constructed by modifying baseline ensembles, displacing single test dots to dissociate the Center of the Dot Positions from the Center of the Inferred Shape. On Baseline trials, participants were more accurate clicking the Center of the Inferred Shape (mean error 27.7 pixels) than the Center of the Dot Positions (17.3 pixels, p< .01), with no difference in response time (p>.05), suggesting that finding the center of the gestalt global shape is relatively easier. For Test trials in the Center of the Dot Positions condition, error increased with test dot distance from the center, but less so for perimeter test dots, suggesting greater focus on internal items. Conversely, for the Center of the Inferred Shape, perimeter test dots increased error, while test dots in the interior did not, suggesting greater perimeter item focus. Overall, participants were better at finding the Center of the Inferred Shape than the Center of the Dot Positions and this perception was more robust to changes to internal ensemble items, but more vulnerable to changes to the perimeter items.
Meeting abstract presented at VSS 2018