Abstract
What information is available from a brief glance at a scene? While previous efforts to answer this question have focused on scene categorization or object detection, real-world scenes contain a wealth of information whose perceptual availability has yet to be explored. Here we used image exposure thresholds to compare several basic-level categorizations with global-property categorizations: tasks that reflect ecological properties describing spatial and functional aspects of a scene space. In separate experimental blocks, observers performed yes-no forced-choice tasks on either global properties (e.g. openness, naturalness) or basic-level categories (forest, desert, etc). All target images were masked with an RSVP sequence of textures to fully mask visual features. Thresholds on all tasks were remarkably short: observers achieved 75% correct performance with presentations ranging from 19ms to 67ms, reaching maximum performance after 100ms. Global-property tasks had significantly shorter thresholds than basic-level tasks, suggesting that there exists a time during early visual processing where a scene may be classified as open or navigable, but not yet as a mountain or lake. We explored this possibility in a second experiment: observers were shown a briefly presented scene (30ms masked) and then given four image alternatives. In addition to the target, the three distractors were chosen from a 2×2 in which images could share a global property, a category or neither with the target. We compared the error distributions made in this task, and found that observers were significantly more likely to choose the distractor sharing a global property with the target category than alternatives that shared the category, providing evidence that global properties were more completely processed than basic-level categories in this short presentation time. Comparing the relative availability of visual information reveals bottlenecks in the accumulation of meaning. Understanding these bottlenecks provides critical insight into the computations underlying rapid visual understanding.
Funded by an NSF-GRF to MRG and a National Science Foundation Career award to A.O (IIS 0546262).