Abstract
Objects in our environment tend to be grouped in typical visual settings. These settings may be represented in contextual frames that contain information regarding the identity of objects in specific scenes, and the spatial relations between them. To what extent are the semantic and spatial information linked to each other within a unified contextual frame, and to what degree do these contextual components independently contribute to object identification?
We investigated this question using a priming task in which the spatial and semantic relations of prime and target were independently manipulated. In the spatial dimension, target and prime were either properly or improperly positioned (e.g., an ashtray appearing below or above a cigarette, respectively). In the semantic dimension, target and prime were either semantically related or unrelated (e.g., an ashtray and a cigarette, vs. an ashtray and an ice cream cone, respectively). Prime and target were successively presented, each for 250ms. The prime always appeared in the center of the screen, while the target appeared either above or below it. Subjects judged whether the target was a real or a nonsense object (the prime was a real object in all cases).
Results revealed a strong spatial priming effect, presumably mediated by an automatic shift of attention to the location implied by the prime. In addition, spatial and semantic factors interacted with each other, showing a larger spatial priming effect in the semantically-related than the semantically-unrelated condition. These results suggest that contextual effects are mediated by a crude mechanism that directs attention to the most likely location of an object, as well as a more fine-tuned mechanism (or contextual frame) that generates a specific association involving both semantic and spatial dimensions of object representation. FMRI data corresponding to these findings will also be presented. Supported by NIH R01NS050615 and J.S. McDonnell Foundation 21002039.