Semantic Consistency in Boundary Extension

Two experiments explored the effect of semantic consistency on boundary extension by presenting images depicting scenes with either a semantically consistent or inconsistent object-background relation; Experiment 1 presented these images for brief durations and Experiment 2 presented these images for long durations. The results of these experiments revealed no difference in boundary extension for brief image durations in Experiment 1 and a significant difference in boundary extension for long image durations in Experiment 2, with semantically inconsistent scenes eliciting less boundary extension than semantically consistent scenes. These findings are interpreted as evidence for boundary extension being context-independent early in scene perception and context-dependent late in scene perception. I suggest that spatial computation initially operates on object-background spatial relations alone and does not include semantic information, whereas spatial computation later operates on both object-background spatial relations as well as semantic information about the scene. For semantically inconsistent scenes, the competing semantic information from the context depicted in the background and the context implied by the object interfere with further spatial computation later in scene perception.