Abstract
Object recognition relies on a network that codes object identities across diffuse regions of the visual field, often leading to uncertainty about the location of any given object. One solution to this problem may be to isolate a single spatial position with selective attention, allowing recovery of the object identity at that position. If so, how do we perform visual tasks that require us to compare spatial positions between objects? While such spatial relationship judgments lead to a feeling that we select multiple objects at once, we predict that they will require isolating each object over time. We tracked eye movements as participants encoded spatial relations between objects in the simplest displays involving just two objects (e.g., a green circle on the left, a red on the right). Participants compared the spatial relationship within an encoding display to a later test display. The position of the object pairs varied across the two displays, requiring encoding of relative positions within the group, rather than absolute positions on the screen. Eye movements revealed that during encoding of the relationship, participants systematically selected the two objects in a left-to-right sequence, even when they were explicitly instructed to maintain fixation. A control experiment showed that when the spatial relationships were made irrelevant by changing the task to a same-different identity judgment, the left-to-right bias disappeared. This sequential selection during relation judgments raises a new puzzle: If we must select one object at a time, how can we recover the relative spatial location between multiple objects? We speculate that the attention shift itself is used to produce the relation judgment: extracting the vector created by the shift of attention (rightward) would provide the relative location of the object that was selected second (red).