The classic animation experiment by Heider and Simmel (1944) demonstrated our strong tendency to impose narrative on our perception and memory of interactions between simple geometric shapes. In their most famous animation with three simple shapes, observers almost inevitably interpreted them as rational agents with intention, desires and beliefs (“That nasty big triangle!”). Much work on dynamic scenes has identified basic visual properties that can make shapes seem animate. Here, we investigate the limits on the ability to use narrative to understand animated scenes. We created 30-second Heider-style cartoons with 3 to 9 items. Item trajectories were generated automatically by a simple set of rules, but without a script. In Experiments 1 and 2, ten participants wrote short narratives for each cartoon. Next, different participants were shown a cartoon and then presented with one of the five highest scoring narratives: either one generated for that specific cartoon or one generated for a different cartoon having the same items. Participants made the decision about how well the description fit the cartoon on a scale from 1( clearly does not fit) to 5(clearly fits). In Experiments 3 and 4, we investigated whether performance improved without narratives or when a direct choice between a matched or mismatched narrative was given. ROC curves generated from the rating scale data show that our ‘Heider capacity’ falls off dramatically after 3 items, suggesting a limit related to the visual working memory and/or motion tracking limits. Such limits may impact interpretation and recall of real-world dynamic scenes.