Movie viewing presents a complex cognitive task involving directed attention and mentalizing. Attentional selection, often measured via eye movements, is influenced by social cues, with faces and gaze direction being stronger predictors than low-level visual features. This "social orienting" is distinct from higher-order cognition involved in mentalizing—inferring mental states of others. Movie viewing demonstrates robust shared neural responses across viewers, a phenomenon termed neural typicality, measured by inter-subject correlations (ISCs) of fMRI data. This neural synchrony reflects shared interpretation of narratives and even predicts friendship. Individuals with autism spectrum disorder (ASD) present social deficits impacting both social orienting and mentalizing. While high-functioning ASD individuals show subtle deficits in these areas, measurable differences in eye movements and neural responses during movie viewing offer a valuable avenue for investigating these neural underpinnings. Previous research has primarily focused on correlating neural typicality with behavior directly related to the viewed movie, leading to a fragmented understanding of the neural structures underlying complex social orienting. This study aims to utilize the rich environment of movie viewing to elucidate the neural networks associated with social orienting and mentalizing in TD and ASD individuals by incorporating independent measures of eye movement and neural typicality. By comparing these measures across groups, a more comprehensive understanding of the social brain’s components can be achieved.

Existing research highlights the involvement of various brain regions in social cognition, forming the "social brain." However, the precise delineation of the interacting systems remains unclear. Studies have shown that eye movements are influenced by social cues such as faces and gaze, which are more effective in predicting attentional focus than low-level visual features. Neural synchrony, or neural typicality, as measured by ISCs, reflects shared processing and comprehension during movie viewing, impacting memory and even social bonding. In ASD, social deficits manifest as difficulties in mentalizing and social orienting, though differences may be subtle and challenging to detect experimentally. Movie viewing provides a robust and naturalistic setting to study these differences, with both behavioral and neural responses exhibiting individual variability. Prior research using movies has yielded mixed results in comparing neural typicality between TD and ASD groups due to small sample sizes and inconsistent methodologies. Similarly, studies investigating the neural correlates of social orienting often employ simplistic stimuli, resulting in a fragmented view of the underlying neural networks.

Sixty-two TD participants (24 female) and 36 ASD participants (all male) aged 15–30, with matched age, IQ, and motion in a subset, underwent both eye-tracking and fMRI sessions. The eye-tracking session involved viewing 24 short movie clips while eye movements were recorded. A separate 9.5-minute movie clip was used for fMRI acquisition. Eye movement typicality was quantified by the Euclidean distance of an individual’s scan path from the average scan path of the TD group for each movie. Neural typicality was calculated using voxel-wise ISCs during fMRI movie viewing, again comparing each individual to the TD group average. The Social Responsiveness Scale (SRS) measured social impairment in ASD participants. Data were analyzed using custom MATLAB software and AFNI, with statistical analyses employing t-tests, correlation analyses, and permutation-based cluster size correction for multiple comparisons. To evaluate the stability of eye movement typicality, the 24 clips were split into two sets, and the correlation of typicality between the sets was examined. Group differences in neural typicality were assessed through t-tests, both at the voxel and network levels. The correlation between eye movement typicality and neural typicality was explored to identify the neural underpinnings of social orienting. Finally, overlap and functional distinctions between networks were investigated using conjunction maps and analyses of within- and across-network correlations.

Eye movement typicality showed a high degree of stability across different movie clips for both TD and ASD groups. ASD participants, however, showed significantly less typical eye movements and greater variance compared to the TD group. Neural typicality during fMRI movie viewing revealed widespread ISCs across the cortex, consistent with prior studies. A whole-brain analysis revealed a strong anticorrelation between eye movement typicality and neural typicality in various brain regions, constituting a "social orienting network" encompassing bilateral STS, IFG, anterior insula, cingulate cortex, MPFC, hippocampus, putamen, and caudate. This network was observed in both TD and ASD participants. However, significant differences in neural typicality between TD and ASD groups were primarily observed in right TPJ, middle temporal gyrus, and other regions implicated in mentalizing (posterior STS, IFG, anterior insula, cingulate cortex, putamen, caudate), forming a distinct "network 2." While many individual voxels within the social orienting network didn't show group differences, network-level analysis revealed significantly lower neural typicality in ASD participants. Conjunction analysis showed partial overlap between the two networks, with some regions specific to each network. Further analysis demonstrated greater within-network connectivity compared to across-network connectivity for both networks and groups, supporting their functional distinction. Finally, neural typicality within network 2, but not the social orienting network, showed a significant negative correlation with SRS scores in ASD participants.

The findings demonstrate a double dissociation between two distinct neural networks underlying social orienting and mentalizing. The social orienting network, identified through the correlation between neural and eye movement typicality, is involved in the lower-level process of attentional selection towards social cues, a process that is present in both TD and ASD individuals. Network 2, identified by group differences in neural typicality, is associated with higher-order social cognitive processes, especially mentalizing, which shows significant differences between the groups. The partial overlap between the two networks underscores the dynamic interaction between these processes. The distinct patterns of connectivity and correlations with SRS scores further support the functional distinction between these networks and their relevance to the social cognitive deficits seen in ASD. While the study's results strongly suggest these two distinct networks, it would benefit from including an assessment of eye-tracking measures to directly assess the mentalizing aspects. The methodology employed in this study, using both eye-tracking and fMRI, along with rigorous statistical analysis, enhances the understanding of the neural substrates underpinning social cognition in ASD.

This study provides compelling evidence for distinct neural networks supporting social orienting and mentalizing, with notable differences observed in ASD. The social orienting network exhibits similar function across TD and ASD groups, while network 2, implicated in mentalizing, shows significant differences. Future research should explore the dynamic interplay between these networks and investigate the development of these networks during childhood. Further investigation into potential compensatory mechanisms in ASD could also be fruitful.

The study's sample size, though relatively large compared to previous research, could benefit from an even larger cohort. The lack of complete brain coverage in the fMRI data limits the extent to which conclusions can be drawn about other regions potentially involved in social orienting and mentalizing. The reliance on specific movie clips could limit the generalizability of findings. The use of only male ASD participants also limits the generalizability of the findings.