Human see, human do? Viewing tool pictures evokes tool-use action information in hand-selective occipitotemporal cortex

Abstract

Amongst all the objects we encounter, tools are unique because they are tightly linked to predictable actions. Neuroimaging studies have reported selective responses in occipitotemporal and parietal cortices when viewing pictures of tools or hands. Whether these responses contain information about specific tool-use actions (e.g., rotation for keys), and whether such information shares a neural format with action execution, remains unclear. Here, we used fMRI and multivoxel pattern analysis to ask whether specific actions associated with viewed tools could be decoded and whether action-specific information generalised across viewing and acting. Participants (N=18; 11 females) viewed tool pictures and, in separate runs, pantomimed tool-use actions in response to tool names. Familiar tools differed in their associated action (rotate or squeeze) but were matched for grip type. Viewing tool pictures elicited tool-use action information in lateral occipitotemporal cortex (LOTC), parietal cortex, and even somatosensory cortex. Notably, hand-selective LOTC was the only region in which action decoding exceeded tool-identity decoding during both viewing and pantomime, suggesting sensitivity to learned relationships between objects and characteristic hand actions. However, neither region-of-interest nor whole-brain searchlight analyses yielded reliable cross-task decoding. Thus, although overlapping cortical regions contained information about tool-use actions during perception and production, we found no evidence that this information was represented in a shared multivoxel format. These findings suggest that seeing familiar tools evokes distributed information about their associated actions and sensory consequences, but that such information may be expressed in task-dependent formats rather than simply reinstating the neural patterns engaged during action production.