Decoding Natural Scenes

Abstract: Human beings are surprisingly adept at extracting the gist from natural scenes, even with
extremely brief exposures. How does the brain so quickly categorize such complex and cluttered
stimuli? We use both behavioral methods and fMRI pattern analysis to address this question.
Specifically, using statistical learning techniques borrowed from computer science, we have been
able to decode scene category throughout visual cortex, including primary visual cortex and the
parahippcampal place area (PPA). Moreover, across a range of experiments we have found that
decoding, particularly in later visual areas, mirrors human behavior. These data not only
provide insight into the brain regions involved in natural scene categorization but also the features
critical to our ability to categorize the scenes