Abstract
An object that is semantically consistent with its scene context (e.g., a volleyball at a beach) is typically recognized better than an inconsistent object (a printer). This type of object-to-scene consistency effect has been reported in a number of behavioral and event-related potentials (ERPs) studies. Moreover, there is behavioral evidence that object-to-object relatedness modulates the recognition of objects in scenes. Here, we used ERPs to assess the contextual influence that two local objects may exert on each other and how they may interact with the global background scene context. Specifically, we looked at the N400 component, a known marker of semantic access in language and scene perception. Thirty-two participants were presented with two objects superimposed on a scene for 2000 ms, on the left and right of a central fixation point. We manipulated both the semantic object-to-scene consistency and object-to-object relatedness, resulting in four conditions: 1) both objects are consistent with the scene and related to each other (CON_REL), 2) one object is consistent with the scene while the two objects are unrelated (CON_UNREL), 3) both objects are inconsistent with the scene and related to each other (INCON_REL), 4) both objects are inconsistent with the scene and unrelated to each other (INCON_UNREL). Participants completed a repetition detection task, requiring them to attend to both the objects and the scene. In the N400 time window, the INCON_UNREL condition evoked a more negative potential than all other conditions whereas all other possible comparisons were not significant. These data suggest that – in such a paradigm – one consistent relation of an object with either the scene or its object-to-object neighbor is enough to eliminate the inconsistency effect in the prototypical time window.