Visual perception of actions and objects engages distinct cortical systems. Object recognition relies heavily on the ventral occipitotemporal cortex (VOTC), while action perception involves a network spanning parietal, frontal, and dorsal temporal regions. The VOTC displays a well-defined object domain organization, with specialized areas responding to faces, tools, and other object categories. However, the organization of the action perception system and its domain-specific communication with the ventral object recognition system remain less understood. Previous research suggests a domain-related interaction between object and action perception systems; for example, manipulable objects activate the inferior parietal lobe (IPL), implicated in manipulation knowledge, and faces activate the posterior superior temporal sulcus (pSTS), involved in social interaction. However, these studies often involved object stimuli, raising the question of whether the observed effects were driven by object properties. It's unclear whether action perception is organized by parallel processing domains (social vs. object purpose) beyond animate/inanimate distinctions. This study addresses this gap by testing the hypothesis of a domain-organized action perception pathway, using action stimuli devoid of object-domain differences, to examine both regional activities and task-based functional connectivity (FC) patterns. Participants watched videos of a human figure performing social-communicative and manipulation actions on identical meaningless shapes while undergoing fMRI. The goal was to determine if these two action types elicited distinct dorsal action perception system activations and whether these activations communicated differently with the ventral system.

Existing research demonstrates distinct neural pathways for object and action processing, with the ventral stream primarily responsible for object recognition and the dorsal stream for action. While the interaction between these pathways is well-documented, the role of domain-specific organization within the action perception system and its dynamic communication with ventral object regions remains less clear. Studies have shown that different object domains activate corresponding regions within the dorsal stream; manipulable objects activate the IPL, while faces and animals activate the pSTS. Resting-state functional connectivity studies have also revealed a domain-like organization between ventral object and dorsal action systems, with tool-preferring ventral areas connected to frontoparietal manipulation regions, and face-preferring areas connected to pSTS social cognition regions. However, these findings are indirect evidence for the domain-specific organization of the action perception system itself and its dynamic communication with the ventral object system. Existing studies tend to involve object stimuli, potentially confounding the effects of object-domain properties. This study aims to address this limitation by using stimuli devoid of object-domain properties, focusing solely on the domain-specificity of actions.

Forty-four right-handed participants (36 included in final analyses) with normal or corrected-to-normal vision participated. A long-block fMRI design was used, presenting videos of a human cartoon figure performing two types of actions: social-communicative actions (e.g., waving) and manipulation actions (e.g., folding) on a set of identical meaningless shapes. Three groups of participants viewed videos with different action-shape correspondences to ensure group-level shape matching. The experiment included four runs, each containing blocks of social-communicative, manipulation, and navigation (control) action videos. Participants were instructed to remember the action-shape correspondence and report the action associated with the shape after scanning. fMRI data acquisition employed a Siemens Trio Tim 3-T scanner. Preprocessing included motion correction, normalization to MNI space, spatial smoothing, linear trend removal, band-pass filtering, and regression of nuisance covariates (head motion, white matter, and CSF signal). Global signal regression was also performed as a validation analysis. Whole-brain univariate analysis was conducted using SPM12 to identify brain regions showing differential activation for the two action types. A leave-one-participant-out approach was used to define the action system ROIs based on the univariate analysis. Functional connectivity (FC) analyses calculated the FC between action system ROIs and ventral object perception regions (defined using Neurosynth meta-analyses for FFA and LOTC, and a whole VOTC mask from a previous study) under different action perception conditions. Repeated-measures ANOVAs tested for domain-specific interaction effects. Finally, correlations were computed between domain-specific FC strength and local activation strength across VOTC voxels.

The study yielded two main findings. First, social-communicative and manipulation actions elicited distinct patterns of activation within the dorsal stream. Social-communicative actions resulted in stronger activation in the bilateral pSTS and right precentral gyrus, while manipulation actions showed greater activation in the bilateral supramarginal gyri, IPL, SPL, precentral gyri, and other frontal regions. Second, domain-specific functional connectivity was observed between the dorsal action perception system and the ventral object processing stream. During social-communicative action perception, enhanced FC was observed between the social-communicative action system and the bilateral FFA. Conversely, during manipulation action perception, enhanced FC was found between the manipulation action system and the left LOTC. A whole VOTC analysis revealed a cluster in the right ITG/FG showing significant interaction effects, with stronger connectivity to one action system during the corresponding action condition. Crucially, a significant correlation was observed between the FC with the action system and the local activity strength across VOTC voxels.

The results strongly support the hypothesis that action perception is organized by action domains (social vs. manipulation) that parallel the domain organization in the ventral visual pathway. The domain-specific activation patterns in the dorsal stream demonstrate that the action perception system is not simply responding to movement, but is sensitive to the goal and type of interaction. The domain-specific FC patterns observed between the dorsal and ventral streams indicate a dynamic interaction that is tuned to the type of action being perceived. This aligns with the connectivity-constrained domain representation hypothesis, extending the notion of domain organization to include human interaction domains (social and manipulation). The lack of significant VOTC activation differences between action conditions, which contrasts with some previous studies, may reflect the absence of object-related bottom-up inputs in the current design. The significant correlation between FC strength and local VOTC activation strength suggests that the ventral system's activity is influenced by inputs from the dorsal action perception system, rather than the other way around.

This study provides strong evidence for a domain-specific organization of the action perception system, mirroring the domain organization in the ventral visual pathway. Social and manipulation domains emerge as overarching principles governing both object and action perception. The domain-specific functional connectivity between these systems underscores a dynamic interaction shaped by the type of action perceived. These findings emphasize the crucial role of human interaction domains in shaping perception and suggest avenues for future research investigating the causal mechanisms underlying this interaction.

One limitation is the use of cartoon stimuli, which may not fully capture the complexity of real-world actions. The relatively simple stimuli could have influenced the observed brain activity. Additionally, the study relied on correlational data, making it difficult to establish definitive causal relationships between brain regions. Further research using more naturalistic stimuli and techniques allowing for causal inference would strengthen the conclusions.