Proceedings of the Annual Meeting of the Cognitive Science Society

Physical scene understanding requires not only detecting the current state of the world, but also predicting how the futurewill unfold. The need for such prediction is especially salient in the context of physical instability as when an object isteetering, about to fall off a surface. Here we asked whether such scenes automatically capture attention, such that themere presence of instability will impair performance on a central attention-demanding task. Observers viewed scenesin which an object (e.g. an open laptop) was either sitting stably, or was about to fall off a table. Observers simplycompleted a central Multiple Object Tracking (MOT) task (e.g which could appear on the screen of the depicted laptop).MOT Performance was indeed worse in the presence of physical instability, despite its task irrelevance, and even whenobservers failed to notice the physical stability vs. instability in the first place.