A striking take on mass inferences from collisions

Abstract

Our everyday activities – e.g., packing groceries, building sandcastles – require knowledge about the physical properties of the objects and surfaces in our environments. We often learn about properties such as weight, hardness, and slipperiness by interacting with the world, but we also learn about latent physical properties by observing events. Among the best-studied examples is people’s ability to infer objects’ relative masses based on seeing them collide, and work on this front has shown that people’s inferences sometimes show marked biases. For example, when an incoming object strikes a stationary one, people consistently overestimate the mass of the incoming object. Why? Various accounts attribute the effect to heuristics that fail in certain conditions, simplified stimuli that lack realistic cues, or a characteristic of rational Newtonian decisions made under sensory uncertainty. Here we sought to disentangle these competing accounts by asking observers to judge the relative weights of bowling balls in videos of real collisions. Balls ranged from 6 to 16 pounds; subjects saw all possible pairings and reported whether the incoming ball or the initially static one was heavier. Under these realistic conditions, participants still consistently overestimated the incoming ball’s weight. We found reliable individual differences in the size of this bias, and in overall performance in weight discrimination. While weight discrimination performance was predictive of intuitive physics abilities in general (assessed with an independent task battery), bias in weight estimation showed no relationship, suggesting it is not a hindrance in other tasks. In a final experiment, we occluded cues in the videos that the purported heuristics would rely on, and we found it had little impact on overall weight discrimination or bias. Our results argue against accounts based on heuristics or oversimplified stimuli and support the idea that the bias is a characteristic of a well-tuned Newtonian system.