Thomas W. Brewitt¹ (), Vivian C. Paulun¹, RT Pramod¹, Nancy Kanwisher¹; ¹MIT

The Physics Network (PN) is a set of brain regions in the fronto-parietal cortex engaged when people make physical rather than descriptive judgments on dynamic visual stimuli, and representing physical properties and relations of objects (“Things”) and non-solid substances (“Stuff”). Prior studies on this network have focused exclusively on the processing of visual information, whereas we also make physical inferences based on sounds, such as the ‘boing’ of a bouncing ball, the ‘splash’ of water, or the ‘clink’ of wine glasses. Here, we test whether the PN responds to audio recordings of physical events, or whether it operates primarily on visual input. We compiled short (2 sec) naturalistic audio and video clips of three different types: 1) “Things”, e.g., rolling dice, 2) “Stuff”, e.g., water sloshing, and 3) social scenes, e.g. people laughing. These stimuli were presented in a 2 (modality) x 3 (condition) blocked fMRI design while subjects (N=4) performed a 1-back task. The independently localized PN showed a significantly stronger response to videos of physical events (both Stuff and Things) than social events (p<0.05), replicating and extending previous findings. Crucially, we found that the response to audio clips of physical events was significantly lower than to video clips showing similar physical events (p<0.005), and indeed no higher than the fixation baseline. There was no difference between the auditory “Things” and “Stuff” conditions. This pattern of responses was robust and present in each participant individually. Our findings show that the PN doesn’t respond to the sounds of physical events, suggesting that the PN is not a general system for amodal intuitive physics, but that it is specifically engaged in visual physical scene understanding.