The contribution of different contextual informational sources in visual object recognition

Abstract

ABSTRACT The contextual scene in which an object appears may allow it to be recognized with lower resolution. A contextual scene typically contains multiple potential sources of information that may facilitate such recognition including the contextual scene category, the location of the object within the scene and the relative size compared to nearby objects. Here, we assessed the contribution of different sources of contextual information as well as familiarity in facilitating low-resolution recognition. Participants increased the resolution of a photograph of an object until they could identify it. The threshold resolution for accurate identification was compared across three contextual conditions: (1) No Contextual Scene (2) Contextual Scene only (3) Contextual Scene and Location. Additionally, we assessed the effect of a standard size cue shown alongside the quantized object. We compared performance for photographs of objects taken from the subject’s own vs. participants who had no exposure to these scenes prior to the experiment. Finally, each object was rated along multiple dimensions (e.g. typicality, movability goodness of fit with the contextual scene). Results showed interactions between context condition, level of familiarity and the objects' rated dimensions, suggesting that people flexibly engage with multiple sources of contextual information during object recognition.