Subjects can visually track several moving items simultaneously, a fact that is difficult to explain by classical attention models. Previous work revealed that building a global shape based on the spatial position of the tracked items improves performance. Here we investigated the involved neural processes and the role of attention. A task-irrelevant probe stimulus was presented during multiple objects tracking at a fixed spatial location. Depending on the tracked item’s trajectories the probe appeared either outside, inside, or on the edge of aforementioned global shape. Event-related potentials to the probe stimulus revealed two subsequent stages of attentional selection during multiple object tracking. After 100 ms attention was deployed on the edge/boundary of the figure formed by the tracked items. In the following 80 ms, attention spread from the outline to the full figure. These findings clarify the eminent contribution of attentional mechanisms in multiple objects tracking.