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

The display and task are simple to describe, but in one condition, the task is surprisingly difficult to perform. In the easy condition, the perimeter of a square is constructed from adjacent red discs, four per side. A target disc turns white momentarily. The participant must reidentify the target at the trial end, which follows a period during which the square rotates around its center point. Average performance was 92%. The second condition is identical, except the red discs translate at a constant linear speed, tracing the perimeter of a stationary square. Performance is significantly worse, 84% on average. Two control experiments replicate the findings, and demonstrate that translational motion is the specific cause of the tracking deficiency compared to rotation. We explain the results through appeal to internally referenced representations, like those that underlie the perception of elongated objects. Specifically, the presence of a visible square causes the individual discs to be localized with (internal) reference to the square as a whole—something like "it's the second one from the left on the north side." Translating motion conflicts with this representation because the "north" side remains stationary though the target does not remain there. But with rotation, the internally referenced description remains accurate throughout the motion period, and tracking can progress by updating the square's changing alignment to an external reference. In a fourth experiment, connections between discs alleviate the challenge of translation, by exchanging the square for representations of barbell objects. We also report experiments with opposite effects: a cost for rotation compared to translation in situations where the representational conflict runs in the opposite direction. Together, the experiments reveal how composite and internally referenced representations support and constrain visual processing broadly, including the perception of objects, orbits, and groups.