From visual features of moving objects to subjective impressions of causality

In daily life, we instinctively attribute causes and effects to objects in our environment based on visual information, though the precise mechanisms underlying the perception of causality remain unknown. Previous work on causality perception employing Michotte’s launching paradigms (or variations thereof) implied a correlation between the physical plausibility of a stimulus and behavioral causality judgments, given an observer’s sensory uncertainty around the objects’ moving trajectories. To examine the relationship between sensory uncertainty, physical plausibility and perceptual causality, we rendered launching events with varying levels of physical realism and measured both observers’ sensory uncertainty around moving trajectories and their respective causality judgments. Our experimental conditions included horizontal and oblique launch directions at fast and slow speeds to systematically manipulate uncertainty; in addition, we added physically implausible moving trajectories to the causality judgment task. We fitted psychometric functions to the responses and computed uncertainty measures from the slope of the psychometric fits. Overall, our results did not suggest a simple relationship between an individual’s uncertainty around moving trajectory and their respective causality judgments. Instead, we found substantially higher uncertainty for causality judgments than for angle judgments with a stronger interaction of moving speed and launch direction, implying that causality perception was modulated to a greater extent by the temporal dynamics of the display than angle perception. That interpretation is consistent with previous research on the launching effect, indicating that timing is a prerequisite to visually integrate the individual objects’ motion into the perception of one continuous motion. We speculate that observers’ subjective causality reports were more related to the degree to which they could postdict a plausible sequence of events rather than a prediction about the physically plausible trajectories.