In her thought-provoking commentary, Thomas draws a line between social relationships and social groups. We appreciate Thomas’s efforts to characterize how infants initially represent the social world, but we question whether separating relationships and groups is necessary or correct. Indeed, in looking broadly across literatures in the social sciences to define what a social group entails, sociologists and social psychologists often center – or at least reference – social relationships. For example, Forsyth (Reference Forsyth2019) defines a group as “two or more individuals who are connected to one another by and within social relationships.” Sociologists have also traditionally classified groups into types that vary according to features such as degree of intimacy and longevity (e.g., “primary groups” such as families vs. “secondary groups” such as work colleagues; Cooley, Reference Cooley1909). And, social scientists from many sub-disciplines recognize that groups often have hierarchical structures (e.g., Heck, Shutts, & Kinzler, Reference Heck, Shutts and Kinzler2022). Thus, it is not clear that there should, in principle, be a hard line between relationships and groups; a more parsimonious account may view social relationships as embedded within social groups.

More proximate to the psychology of infants and young children, empirical studies suggest that young children’s developing representations of social groups include social relationships as a central component. Rhodes (Reference Rhodes2013) argues that categorizing humans into groups invites inferences related to 1) natural kinds and 2) social obligations. Another way to frame this distinction may be that a “social group” licenses two different downstream consequences: inferences about common attributes and inferences about social relationships (Liberman, Kinzler, & Woodward, Reference Liberman, Woodward and Kinzler2017a). Indeed, existing evidence with infants suggests that (at least rudimentary) aspects of both kinds of social-group inferences are observable. Infants’ reactions to speakers of the same versus different languages provide a useful example. These studies suggest that infants both infer positive social relationships between same-language speakers (Liberman et al., Reference Liberman, Woodward and Kinzler2017b) and make inductive inferences about such individuals’ common attributes (e.g., that two speakers of the same language will share food preferences; Liberman et al., Reference Liberman, Sullivan, Woodward and Kinzler2016). These two classes of inferences (i.e., “relationships” and “attributes”) may also inform one another: for instance, infants predict that people who share similar food preferences are also more likely to affiliate (Liberman et al., Reference Liberman, Kinzler and Woodward2014; Liberman et al., Reference Liberman, Kinzler and Woodward2021), and that people who affiliate are more likely to share food preferences (Liberman et al., Reference Liberman, Sullivan, Woodward and Kinzler2016).

Further, some research suggests that early inferences of shared relationships may be even more easily licensed by social group information than are those of common attributes. For example, in one study (Shutts, Roben, & Spelke, Reference Shutts, Roben and Spelke2013), four-year-old children were presented triads of unfamiliar children. In each triad, two people matched in either gender or race, and one person differed in either gender or race (e.g., on one trial, children saw two Black girls and one Black boy). When asked to guess which two people were friends, children selected the people who matched in gender or race (e.g., the two Black girls; see also Roberts et al., Reference Roberts, Williams and Gelman2017). When asked which two people shared the same attribute (e.g., liking to play a particular game), children were equally likely to select children from the same vs. different social groups. Thus, when reasoning about social groups, young children can privilege social relationships over shared properties.

The target article lays out a possibility that human social group notions are built from representations of social relationships per se. Alternatively, in studying infants’ reasoning about the nature of social connections, Thomas and her colleagues may be studying early social group cognition. In this case, Thomas’s work could be seen as indicating that some of the social group phenomena revealed most often in studies of children and adults are also present in infancy – and that realization could guide further studies building on the creative methods Thomas has developed, but linking signature phenomena identified in children and adults.

Ultimately, of course, older children and adults come to know many more things about social groups than infants and young children do, including which properties (stereotypes) are associated with particular kinds of people (Sierksma, Brey, & Shutts, Reference Sierksma, Brey and Shutts2022) – and the nature and complexity of dyadic relationships within a group certainly grows. But the heart of social groups across development appears to include social relationships as a critical downstream consequence. Indeed, adults themselves have closer and more numerous social relationships with those who match them in gender, race, and other social group dimensions (McPherson, Smith-Lovin, & Cook, Reference McPherson, Smith-Lovin and Cook2001).

Whether Thomas’s work should be conceptualized as focusing on social relationships or social groups, there is one thing that is clear at present: we are likely to learn a lot about the origins of humans’ extraordinary social capacities through the clever studies and methodology of Thomas and her colleagues in the coming years. We applaud these efforts and look forward to additional work testing ideas Thomas has advanced.