Intuitive knowledge of object acoustics enables perceptual separation of physical variables from impact sounds.

Humans can infer object material and mass from collision sounds, but their accuracy drops when both colliding objects’ resonance distributions deviate from real-world measurements.

• Why it matters: The inverse problem of inferring physical parameters from impact acoustics is unclear, especially when multiple objects contribute overlapping sound sources.
• What they did: The team measured impulse responses from hundreds of everyday objects, extracted resonance statistics, built a probabilistic generative model, and used it to synthesize and manipulate impact stimuli for perceptual judgment experiments.
• The result: Listeners’ property judgments were accurate with single-variable conditions, yet impaired when both objects varied and resonance distributions differed from real-world object data, implying internalized statistical regularities for separating concurrent acoustic contributions.

AI summary of the abstract · may contain errors · not medical advice · read the paper.