The Foreground Bias: Initial scene representations dominated by foreground information

Abstract

Researchers have often posited that scene representations have a hierarchical structure with background elements providing a scaffold for more detailed foreground elements (Brooks, Rasmussen, & Hollingworth, 2010; Davenport & Potter, 2004; Henderson & Hollingworth, 1999). To further investigate scene representation and the role of background and foreground information, we introduced a new stimulus set: chimera scenes, which have the foreground set of objects belonging to one scene category, and the surrounding background structure belonging to another category. Across three experiments, we examined the contribution of each scene plane to the initial understanding of scenes when rapidly presented. In the first experiment, participants categorized either Normal or Chimera scenes (i.e., scenes with background and foreground from different semantic categories). Results revealed a Foreground Bias, in which participants initially processed the foreground information at the exclusion of background information. Interestingly, this bias persisted in Experiment 2 when the initial fixation position within the scene image was modified such that even when fixating the scene background, participants continued to show a Foreground Bias. This was true for the shortest presentation duration (50ms) and dissipated once the presentation duration exceeded 100ms. In Experiment 3, we changed the task to further emphasize the scene category information (for a more global oriented task), but found that the Foreground Bias persisted. We conclude that the Foreground Bias arises from initial processing of scenes for understanding and suggests that attention is initially focused on the foreground and over time expands to include background information. Implications for scene gist perception and scene representation theories will be discussed. Meeting abstract presented at VSS 2018