Causation without realism.

Current theories of causality from visual input predict causal impressions only in the presence of realistic interactions, sequences of events that have been frequently encountered in the past of the individual or of the species. This strong requirement limits the capacity for 1-shot induction and, thus, does not sit well with our abilities for rapid creative causal learning, as illustrated, for example, by the effortless way we adapt to new technology. We present 4 experiments (N = 720) that reveal strong causal impressions upon first encounter with collision-like sequences that the literature typically labels "noncausal." Our stimuli include both the commonly used computer-based animations and edited video sequences. Besides direct reports, we present evidence based on goal-oriented behavior that makes sense only in the presence of strong causal assumptions. Finally, we document impressions of causality in highly unrealistic sequences involving, for example, instantaneous shape or size change. In the case of the more realistic clips used in the past, causal ratings abruptly decline and approach the findings of previous work, only after a canonical collision (launch event) is presented. We argue that previously used experimental procedures conceal order effects because of participants adapting to the task and reinterpreting its demands. We discuss ways to account for this adaptation whereby people either focus on experiences of perceptual causation or take realism into account even when asked for impressions of causality.