Abstract
Adaptation is a ubiquitous phenomenon in the human visual system, allowing recalibration to the statistical regularities of its input. Previous work has shown that global scene properties such as openness and mean depth are informative dimensions of natural scene variation that are useful for both human and machine scene categorization (Oliva & Torralba, 2001; Greene & Oliva, 2006). A visual system that rapidly categorizes scenes using such statistical regularities should be continuously updated, and therefore prone to adaptation along these dimensions. Here we show that after adapting to one pole of a global scene property, observers show a scene property after-effect on subsequently presented test scenes. Observers were adapted to an RSVP stream of 100 images previously ranked by human observers to exemplify a pole of a global property (e.g. very large depth, Greene & Oliva, 2006). The images differed in their basic-level categories, other global properties, their objects and other low-level image features. After adaptation, observers were presented with a test scene and performed a binary classification along the global property dimension. Test images were drawn from the 25th, 50th and 75th percentiles of that global property to measure shifts in the psychometric function. Both poles of each global property were compared to a control RSVP stream of random images. All global properties showed significant main effects of adaptation (p[[lt]]0.01, ranging from 6% to 22%), with the magnitude of the adaptation effects inversely related to adaptation duration. Additionally, adaptation after-effects were preserved when the test image was presented 10 degrees away from the adapted location, suggesting the origin of these after-effects is not solely due to low-level adaptation. Finally, we show systematic modulation of observers' basic-level scene categorization performances after adapting to global properties, suggesting a strong representational role of global properties in rapid scene categorization.