In many contemporary societies, some social roles and the behaviors they afford depend on the institutions that create them, regardless of the individuals who occupy them. A judge, a referee, or a cashier is treated primarily as a position within a larger institutional structure: whoever occupies the position is expected to behave in role-appropriate ways. The target article proposes an evolutionarily old “institutional stance,” according to which people make sense of others’ behavior based on institutional roles. We agree that social cognition is not solely, nor primarily, grounded in mental state reasoning and that an institutional stance captures an important feature of contemporary adult social cognition. However, we argue that role-filler independence – the intuition that anyone can occupy a role – is not an evolutionarily ancient default but a widely available innovation that helps solve coordination problems in large-scale societies. We connect this argument broadly to the implications of work on impartiality and suggest that another candidate for the foundation of social reasoning could instead be relationship cognition: abstract representations of relationships between particular individuals that include relationship-bound roles.

Roles can be thought of in at least two ways that are non-mutually exclusive: (1) role-filler-independent roles whose occupants are interchangeable and (2) roles embedded within interpersonal relationships (relationship-bound roles). Many of the authors’ examples – baristas, janitors, and cashiers – are contemporary jobs. The behaviors afforded by these roles are shaped by institutions situated in large-scale economic systems, which we argue are particularly salient examples of role-filler independent roles. Take their movie-theater example, where a boss no longer expects their employee to fetch things for them when the employee is “off the clock.” While intuitive to many readers, this example may not be intuitive to infants or young children or even people living in different economic systems. Indeed, even US adults likely do not “forget” their boss–employee relationship when off the clock. Alternatively, roles can be represented within interpersonal relationships. For example, people have expectations about how mothers (one role) behave toward their children (another role). However, while a child may have multiple caregivers whose roles have overlapping functions, children likely do not think of their caregivers as interchangeable. If a child’s father leaves or dies, he cannot simply be replaced with another caregiver without changing the representation. In this way, relational representations often include particular people in the relation. These two ways in which roles can function are not mutually exclusive. Even highly institutional roles can become relationship-bound (e.g., building a relationship with a particular barista), and conversely, relational roles (e.g., mother) can be conceived of abstractly in the absence of particular people. We agree that infants have abstract representations of relationship categories, which allow for generalization. However, we argue that they work with representations of particular people. Research demonstrates that infants track structured social relations, including dominance hierarchies, asymmetric interactions, and stable patterns of affiliation. However, these findings do not uniquely support representations of abstract roles with role-filler independence, the way the institutional stance may suggest. For example, dominance hierarchies in other species may arise from patterns of interactions among particular individuals and their particular ties, rather than abstract dominant and subordinate role slots. Indeed, children and infants generalize dominance relations asymmetrically, generalizing dominance more readily than subordination (e.g., the role of the “dominant” is not interchangeable; Mascaro et al., Reference Mascaro, Goupil, Pantecouteau, Depierreux, Van der Henst and Claidière2023; Hok et al., Reference Hok, Vasquez, Barakzai and Shaw2024). Moreover, infants expect different social interactions when social partners change (Kudrnova, Spelke, & Thomas, Reference Kudrnova, Spelke and Thomas2024; Pepe et al., Reference Pepe, Woo, Thomas and Powell2025; Thomas, Saxe, & Spelke, Reference Thomas, Saxe and Spelke2022). There is no evidence that infants change expectations of how particular dyads will interact when contexts change. While it is possible that infants can recognize institutions and would expect similar interactions even when a social partner is swapped out within the same institutional role, there is no such evidence. Thus, while existing infant research does not necessarily exclude the possibility that infants assume role-filler independence, the data can also be explained by representations that include particular individuals. While we think it is unlikely that the primitive foundation of social cognition includes role-filler independence, we agree that it plays an important role in shaping contemporary human adult cognition, which may allow for many new ways to relate, both positive and negative. For example, role-filler independence is closely related to other social tools such as impartiality. Thus, role-filler independence may be a representational format (focusing on the role rather than the individual). Impartiality may be a normative principle that uses this representational format. These types of social tools may emerge recurrently across historical and cultural contexts under conditions that require larger-scale coordination. In large-scale social environments – contemporary markets, bureaucracies, or contexts with many interactions with unfamiliar individuals – it is infeasible to maintain rich representations for every social partner. Institutional roles provide a solution by allowing expectations to generalize across individuals occupying the same functional position, rather than indexing across particular individuals and their relationships (for a similar argument on morality, see DeScioli & Kurzban, Reference DeScioli and Kurzban2013). Consistent with this view, both adults and older children, at least in the USA, expect formal rules to emerge preferentially in large groups, where identity-based relationship tracking may be more difficult (Hok, Thomas, & Saxe, Reference Hok, Thomas and Saxe2025a).Classic sociology and historical analyses converge on a similar conclusion. Weber’s account of bureaucratic institutions emphasizes the central role of impersonal, functionally defined positions in enabling scalable coordination in modern societies (Weber, 1921/1922/Reference Weber, Roth and Wittich1978). Likewise, historians have argued that notions of impartiality “forcefully emerged” in the seventeenth and eighteenth centuries, as cultural tools for managing perceptions of parochial bias in increasingly larger and more fragmented social worlds (Murphy & Traninger, Reference Murphy and Traninger2014). Furthermore, developmental work with children finds that notions of impartiality and procedural justice emerge fairly late in childhood: between 6 and 8 years (Schmidt et al., Reference Schmidt, Svetlova, Johe and Tomasello2016; Shaw & Olson, Reference Shaw and Olson2014; Hok et al., Reference Hok, Gerdin, Zhao and Shaw2025b). Together, this evidence suggests that an institutional stance does not form the foundation of social cognition. Rather, it may emerge as a social innovation to address issues that arise as humans form very large modes of organization. These inventions may be enabled by evolutionarily old capacities to represent relationships among particular individuals and by humans’ prolific ability to learn from others and invent new ways of relating.

Figure 1.

Predictions of the institutional stance versus existing infant evidence. The institutional stance predicts that expectations should track institutional roles rather than particular individuals. Existing infant work shows that expectations change when social partners change, but has not established role-filler independence. Circles represent individuals, background colors represent institutions, and links represent expected interactions.