The capacity limits of mental simulation

People have severe capacity limits when they track objects in direct perception. But how many objects can people track in their imagination? In seven pre-registered experiments (N=241 total), we examined the capacity limits of mentally simulating the movement of objects in the mind's eye. In a novel \taskname task, we had participants continue the motion of animated objects in their mind up to a pre-defined point. When tracking one object in the imagination (Experiment 1a), participants gave estimations well in line with ground truth. But, when imagining two objects (Experiment 1b), behavior altered substantially: responses when tracking two objects in the imagination were fit best by the predictions of a Serial Model that simulates only one object at a time, as opposed to a Parallel Model that simulates objects in tandem. The serial bottleneck is not due to response/motor limitations (Experiment 2), and is reduced -- but not eliminated -- by adding extremely strong grouping cues (Experiment 3). Additional studies validate that the serial effect is not due to noise, exists in both realistic and hyper-simplified physics, and is unaffected by motivation (Experiments S1-S3). Altogether, we find that the capacity of moving entities in the imagination is likely restricted to a single object at a time.