Visual segmentation of spatially overlapping subsets I Utochkin |
|---|
Cognitive Research Laboratory, The Higher School of Economics, Russian Federation
|
| In everyday perception we often see multiple objects forming heterogeneous spatially overlapping subsets (such as berries and leaves on a bush) and are able to distinguish between these subsets. In three experiments I studied the limitations of this subset segmentation ability and the role of attention in this process. Observers had to enumerate the number of briefly flashing spatially-overlapped color subsets of 6, 12, or 36 dots (1 to 6 colors in total). In all experiments, 1 or 2 subsets were enumerated with almost same speed and accuracy, while all other numbers yielded substantial increment in error rate and reaction time. This indicates that 2 subsets can be segmented in parallel, and once this limit is exceeded serial shifts of attention are required for segmentation. I also found that segmentation benefits from large sets and this doesn’t depend on spatial arrangement of items in the visual field (Experiment 2).This provides evidence in favor of parallel collecting abstract statistics within each subset that eventually makes subset representations more discriminable. Finally, the evidence was found that observers are able to use an “all-colors” internal template when possible that helps in segmentation when large numbers of subsets are presented (Experiment 3). |
| Up Home |
|---|