Representational objects such as animations are widely used in infancy research, yet little is known about what infants make of animations as a stimulus category when they interpret them. In the first set of studies (Experiments 1-3), we ask whether 19-month-olds take what they see on the screen to be happening here and now, or whether they think that on-screen events are decoupled from the immediate environment. We find that infants do not expect an animated ball falling on a screen to end up in real boxes below the screen, even though they can track the ball (i) when the ball is real or (ii) when the boxes are also part of the animation. In Experiment 4, we ask whether infants use a simple rule according to which a screen is a spatially bounded physical container that does not allow objects to pass through. They do not. When two location cues are pitted against each other, infants individuate the protagonist of an animation by its virtual location (the animation of which it was a part of) as opposed to its physical location (the screen on which the animation was presented). Taken together, the results from the 4 experiments indicate that 19-month-olds reject animation-reality crossovers but accept the depiction of the same animated environment on multiple screens. This pattern of findings is consistent with the possibility that 19-month-olds understand external representations.