Social concepts rely on a domain-general anterior-temporal hub and social spokes in ventral prefrontal cortex and insula

Technical Summary: Social Concepts Rely on a Domain-General Anterior-Temporal Hub and Social "Spokes" in Ventral Prefrontal Cortex and Insula

Problem and Background
A persistent debate in cognitive neuroscience concerns the neural architecture of social conceptual knowledge. Two primary hypotheses compete: (1) the Social Knowledge Hypothesis, which posits that specific subregions of the anterior temporal lobes (ATLs), particularly the right superior ATL, are dedicated to social concepts; and (2) the Domain-General Hub Hypothesis, which argues that the ATLs constitute a transmodal hub supporting all conceptual domains, with functional specialization arising only from graded connectivity to modality-specific "spokes."

Adjudicating these views has been challenging due to confounding factors in previous studies, including differences in stimulus properties (e.g., abstractness, frequency), task demands, and loose definitions of "social" concepts. Furthermore, patient studies often fail to control for the distinct neuropsychological profiles of different Frontotemporal Dementia (FTD) subtypes: Semantic Dementia (SD), characterized by bilateral ATL atrophy, and behavioural-variant FTD (bvFTD), characterized by frontoinsular atrophy.

Methodology
The authors employed a transdiagnostic approach using three methodological innovations to control for confounding variables:

1. Novel Assessment Task: A 35-item, four-alternative forced-choice word-to-picture matching task was developed. It included "social" items (words denoting classes of people with implied age/gender, e.g., "toddler," "widow") and "non-social" items (man-made objects, e.g., "screwdriver," "shed"). Crucially, items were matched for log word frequency, concreteness, and contextual variability. Distractors were semantically proximate to the target to ensure high semantic precision was required.
2. Mixed-Effects Modeling: The authors applied binomial linear mixed-effects models to behavioral data from 45 FTD patients (21 SD, 24 bvFTD) and 19 controls. The models regressed out item-level factors (frequency, concreteness, semantic diversity) and patient-level factors (overall semantic impairment, executive function, global cognition) to isolate the unique effects of diagnosis and stimulus type.
3. Mixed-Linear Lesion-Symptom Mapping: A novel extension of the mixed-model approach was applied to structural MRI data. Instead of using binary diagnosis groups, the models replaced diagnosis with continuous grey matter intensity (GMI) at every voxel. This allowed the identification of cortical regions where pathology predicts a disproportionate impairment on social versus non-social items, independent of other confounding factors.

Key Results

• Behavioral Findings: After controlling for semantic impairment, word frequency, and global cognitive/executive function, the interaction between diagnosis and stimulus type was significant.

  • SD Patients: Showed equal impairment on social and non-social concepts.
  • bvFTD Patients: Showed a disproportionate impairment on social concepts compared to non-social concepts.
  • This pattern contradicts the Social Knowledge Hypothesis (which predicted worse social performance in SD due to ATL damage) and supports the "Hub + Spoke" view.

• Lesion-Symptom Mapping: The analysis identified specific cortical regions where atrophy predicted the differential impairment (social > non-social deficit):

  • Significant Regions: Left insula, left lateral orbitofrontal cortex (OFC), bilateral superior/middle frontal gyri, anterior cingulate cortex, and supplementary motor area.
  • Non-Significant Regions: No subregions of the ATL were identified as differentially important for social concepts. While ATL atrophy correlated strongly with general semantic impairment, it did not predict a specific deficit in social concepts over non-social ones.

Key Contributions

1. Resolution of Competing Hypotheses: The study provides empirical evidence supporting a "Hub-and-Spoke" model where the bilateral ATL serves as a domain-general semantic hub for all concepts (social and non-social), while ventral prefrontal and insula regions act as "social spokes" that contribute preferentially to social knowledge.
2. Methodological Rigor: By using mixed-effects models to regress out psycholinguistic and neuropsychological confounds, the authors demonstrate that the observed social-specific deficits in bvFTD are not artifacts of task difficulty or general executive dysfunction.
3. Lesion Mapping Innovation: The introduction of mixed-linear lesion-symptom mapping allows for the identification of brain-behavior relationships that are specific to conceptual domains (social vs. non-social) rather than just general diagnostic categories.

Significance and Claims
The authors claim that their results reconcile the social cognitive and semantic cognition literatures. They conclude that:

• The bilateral ATLs are critical for the representation of all conceptual knowledge, including social concepts. There is no evidence for a dedicated social module within the ATL.
• The disproportionate impairment of social concepts in bvFTD arises from pathology in the "social spokes"—specifically the insula and ventral prefrontal cortex (OFC). These regions likely support social concepts through mechanisms related to interoception, affect, reward valuation, and semantic control, which are more heavily engaged by social than non-social concepts.
• The findings suggest that the "social knowledge" observed in neuroimaging studies likely reflects the engagement of these connected frontal/insular regions rather than a specialized ATL subregion.

The paper remains modest regarding the specific mechanisms of the "spokes," noting that while the insula and OFC are implicated, the precise nature of their contribution (e.g., interoceptive integration vs. value computation vs. semantic control) requires further investigation. Additionally, the authors note that the study focused on social roles and that future work is needed to determine if these findings generalize to other types of social concepts (e.g., abstract emotions).