I previously argued that a “Theory of Institutions” (Noyes, Keil, & Dunham, Reference Noyes, Keil and Dunham2020) develops during middle childhood and constitutes a conceptual change – i.e., a transformation and expansion of children’s conceptual repertoire. Jara-Ettinger and Dunham take up the idea that there is a Theory of Institutions, but attribute that theory to infants and young children.

Jara-Ettinger and Dunham elide a critical distinction between roles in the very broad sense and institutional roles. An institution is (or defines) a self-reproducing structure that retains its identity across changes in role-occupiers. For example, the US national women’s soccer team has existed for over 40 years, even though the members today are not the members of 1985. Contrast a soccer team with a friendship. We can refer to being a friend as a role, but a friendship is not an institution. The identity of a friendship is not independent of the people in it. When I became friends with Jonah, a new friendship was born. If Jonah stops being my friend, I cannot preserve that friendship by substituting him with someone else.

Infants and young children make sophisticated inferences about social interactions and relationships. They might infer, for example, that if Amy dominates Bill and Bill is friends with Carl, then Amy dominates Carl too (Mascaro et al., Reference Mascaro, Goupil, Pantecouteau, Depierreux, Van der Henst and Claidière2023). But no degree of sophistication transforms patterns of domination and friendship into an institution – in the sense of a self-reproducing structure that retains its identity across changes in role-occupiers.

In Noyes and Dunham (Reference Noyes and Dunham2020), I presented children with the following scenario: a target is friends with the members of group A (but not B), but has fulfilled the joining procedure for group B (but not A). Four-year-old children say the target is a member of group A, the group of their friends, whereas older children and adults say the target is a member of group B. Without the concept of an institution, young children cannot represent a group as numerically distinct from its members. In turn, they can only render a group as an aggregate of agents. To be a group, for young children, is to cooperate, share, and play together. The aggregate can expand or shrink as agents start and stop cooperating, sharing, and playing together, but there is no self-reproducing structure that is numerically distinct from the aggregate (Noyes et al., Reference Noyes, Keil, Dunham and Ritchie2023).

Let me return to the explanation, which is the focus of the target article. Jara-Ettinger and Dunham argue for two modes of explanation of human behavior: We can explain behavior via mental states or institutions. Everyone agrees that children can make sense of agent behavior by drawing on knowledge about things other than mental states. Some of that knowledge includes generalized expectations about people in role categories, such as “soccer player.” Although these categories refer to institutional roles in the adult language, we cannot assume that these words are backed by institutional concepts in early childhood. Children might say, “she kicked the ball because she is a soccer player”; Jara-Ettinger and Dunham take this as evidence for an “institutional stance.” But, if the explanation rests on generalized expectations about agent goals, rather than how the agent is situated in an institution, then children’s reasoning may be accomplished by a Theory of Agents alone. To evidence a Theory of Institutions, we need to examine cases where a Theory of Agents furnishes the wrong explanation of an institutional phenomenon, such that the right explanation is diagnostic of a Theory of Institutions.

One test case that serves this purpose is institutional action, which includes signing an executive order, officiating a wedding, and scoring a goal. An action is institutional when the agent performing the action must do so in their capacity as an institutional role-occupier in order to achieve the expected outcome, which is a change to some institutional fact. For example, pushing a ball across a threshold can obligate the other team to concede defeat and the governing body to award a trophy. Institutional actions are an ideal case for testing whether children have a Theory of Institutions: One cannot understand why an institutional action achieved its outcome based on an analysis of agent behavior alone. Without the concept of institution, one cannot grasp that a soccer team is a separate entity from the agents; in turn, one cannot grasp that scoring a point depends on how an agent is situated within the structure of the soccer team. If an “institutional stance” is needed anywhere, it is needed here.

I examined children’s reasoning about institutional actions in Noyes et al. (Reference Noyes, Keil and Dunham2020). I taught children that three people belonged to a role that kicks balls into trees to score points for their team. Then, I told the children that the team removed the original three people from the role and inserted three new people. I then asked children what would happen if one of the removed individuals kicked a ball into a tree. With a Theory of Agents alone, a child may see the behavior as weird or even wrong. But the child cannot see the behavior as ineffectual. The action did have the desired effect – it went across the threshold! The action is only ineffectual with respect to institutional facts. Therefore, to see the action as ineffectual requires a Theory of Institutions. I asked the children: Would kicking the ball into the tree score a point for one of the teams? Nine-year-old children said it would not. But four-year-old children said it would. They failed to see occupying the institutional role as explanatorily relevant. The same pattern of responding occurred for all the institutional actions I tested.

Jara-Ettinger and Dunham may interpret these data as the continued enrichment of children’s Theory of Institutions. But, if children’s ontology excludes institutional objects – i.e., self-reproducing structures that retain their identity across changes in role-occupiers – then there is no Theory of Institutions to be enriched. During middle childhood, a conceptual change unlocks a whole new ontological category and makes intelligible a whole new class of phenomena, including institutional groups and actions.