Qihan Wu¹ (), Jonathan Flombaum¹; ¹Johns Hopkins University

A composite assembles a representation of an object from different kinds of variables. Some variables describe the parts of the object within an internal frame of reference [handle at -4 on the object’s principal axis (OPA)]. Others describe the object’s location in space [hammer at 3,5 on the table]. And some describe the current relationship between the prior two types [OPA aligns to the negative vertical of the table, tilted clockwise by 20]. That brittleness of this last variable type has been cited to explain the specificity of orientation perception deficits in neuropsychological cases, and nonuniform error types in working memory tasks. We predict that the same susceptibility to error should make it uniquely difficult to track the end of a rotating object. We placed two identical discs on opposite sides of an invisible square. A second pair of discs was arranged in the same way on the perimeter of a larger square that shared the same origin. Two discs (one from each pair) were designated as the tracking targets. Participants tracked the targets as the squares rotated around their center point. (The motion included occasional changes in direction). Performance was significantly lower in this rotation condition compared to a highly similar, but nonrotational (instead translating) motion around the perimeters of the squares. In Experiment 2, a visible line connected the discs in each pair, now impairing translational motion. This is consistent with difficulty in the task caused by a reliance on object representations, a reliance that can be caused by visible connections and by the implied connections of rotation. In a third experiment we predicted trial-by-trial errors by identifying configuration changes that demand re-expression of relational variables in an object composite. These experiments support the hypothesis that object representations are built from composites whose formats can constrain visual processing.