Pietraszewski provides a compelling computational framework for understanding how people represent and reason about social groups, but the approach requires logical extensions. For example, the model does not factor in the nature of bonds between ingroup members, and how those associations might be communicated to others. We should expect machinery for processing signals of groups that could create or update existing representations. Social connections can influence a group's perceived entitativity, and consequently how others will potentially interact with it (e.g., whether to retaliate on another's behalf or not). Research has revealed possible strategies of how groups can either collectively signal their existence, or reveal themselves through by-product cues arising in different modalities. Here, we will describe some recent research exploring acoustic signals and cues of affiliation and group membership.

The question of how people might detect groups based on observable behavior requires making the distinction between adaptive signals and by-product cues (Maynard Smith & Harper, Reference Maynard Smith and Harper2003). Signals are communicative adaptations designed to affect the behavior of other organisms, and evolve in tandem with receiver adaptations. Signaling systems generally evolve by conferring mutual benefits to senders and receivers on average. Conversely, cues are detectable by-products of a phenotype that are not designed to convey information, but they can nevertheless shape receiver perceptual systems. Cost–benefit trade-offs underlie the dynamics of whether cues are maintained by selection; that is, they are typically associated with necessary components of physiological, cognitive, or behavioral systems that serve other purposes. For example, deceptive signaling is often associated with a variety of effects that are difficult for senders to conceal because of cognitive load and/or emotional effects (DePaulo et al., Reference DePaulo, Lindsay, Malone, Muhlenbruck, Charlton and Cooper2003).

Most generally, observable social interactions provide many cues of group structure. Group detection is a flourishing area of research in computer vision, revealing detectable structure in small groups in crowds (Wang, Chen, Nie, & Li, Reference Wang, Chen, Nie and Li2018), conversationalists (Vascon et al., Reference Vascon, Mequanint, Cristani, Hung, Pelillo and Murino2014), and even rapport between interlocutors (Müller, Huang, & Bulling, Reference Müller, Huang and Bulling2018), among others. Definitions of groups by researchers in this domain typically suffer from the vagueness described by Pietraszewski. Nevertheless, correlates of social group behavior are often not present by design but as by-products of social interaction patterns. Other markers are by design however, and will culturally evolve if certain social ecological conditions are met related to the evolutionary stability of cooperative interaction between the relevant individuals (McElreath, Boyd, & Richerson, Reference McElreath, Boyd and Richerson2003).

In the domain of auditory communication, a variety of possible signals of social groups have been examined, including nonlinguistic vocalizations such as colaughter, linguistic phenomena such as speech accents, and group musical behavior. For instance, listeners across two dozen disparate societies were able to reliably detect friends and strangers from very brief (<2 s) recordings of dyadic colaughter (Bryant et al., Reference Bryant, Fessler, Fusaroli, Clint, Aarøe, Apicella and Zhou2016). Colaughter is better than cospeech at revealing affiliation (Bryant, Wang, & Fusaroli, Reference Bryant, Wang and Fusaroli2020), and infants as young as 5-months are sensitive to this information (Vouloumanos & Bryant, Reference Vouloumanos and Bryant2019). Moreover, the acoustic features of human colaughter implicate a broadcast function as it is often loud, abrupt, conspicuous, and repetitive (Bryant et al., Reference Bryant, Wang and Fusaroli2020).

Production and perceptual data suggest that colaughter could constitute a signal of affiliation – one that often operates beyond the awareness of the signalers. Group interactions can not only result in behaviors that function to communicate group structure without the knowledge of the signalers, but also evolutionary processes can shape communication systems to have features that ultimately serve to cue social categories. For example, speech accents arise when adult second-language learners fail to acquire phonological and morphosyntactic idiosyncrasies of a given language. Accents are highly detectable, and are a fixed aspect of social categorization, not a by-product of coalitionary reasoning (Pietraszewski & Schwartz, Reference Pietraszewski and Schwartz2014a, Reference Pietraszewski and Schwartz2014b).

Another domain where group signaling occurs is that of music. Human music is plausibly linked evolutionarily to the coordinated vocalizations of many species signaling territory boundaries (Hagen & Bryant, Reference Hagen and Bryant2003; Hagen & Hammerstein, Reference Hagen and Hammerstein2009), and rooted in emotional signaling and the coordination of group affect (Bryant, Reference Bryant2013, Reference Bryant2014). One important mechanism for coordinating signals effectively is physical entrainment, allowing for social groups to collectively produce rhythmic displays (Phillips-Silver, Aktipis, & Bryant, Reference Phillips-Silver, Aktipis and Bryant2010; Ravignani, Bowling, & Fitch, Reference Ravignani, Bowling and Fitch2014). Rhythmic production and linked perception-action mechanisms potentially constitute core musical adaptations (among possible others) that underlie the cultural evolution of music, affording complex social group signaling at multiple levels (Mehr, Krasnow, Bryant, & Hagen, Reference Mehr, Krasnow, Bryant and Hagen2021).

Auditory communication can be an especially effective medium for groups to advertise. Many nonhuman animals exploit the sonic environment for this purpose, including social carnivores and primates (Hagen & Hammerstein, Reference Hagen and Hammerstein2009). For example, a single covocalizing pair of wolves has been shown to be perceived as a larger group by listeners (Harrington, Reference Harrington1989), a phenomenon coined the Beau Geste effect (Krebs, Reference Krebs1977). Currently, in our lab we are exploring differences across human vocal modalities (e.g., laughter, yelling, and speaking) in how listeners judge group size, that could point to possible vocal adaptations in humans for communicating about social groups. Early results suggest, unsurprisingly, that yelling affords perceptions of larger group sizes, and that colaughter seems to reduce size estimates, making groups sound more integrated. These differences may reveal distinct social signaling functions of varying kinds of human covocalizations and could be understood in the context of the computational framework presented by Pietraszewski. For example, if collective yelling revealed strength in defending a resource, bystanders may engage such a group with a primed expectation of conflict, and assess risk appropriately. Alternatively, if affiliative laughter signals a different social ecology, it could encourage others to more readily join a group, given the relatively lowered threat. Particular patterns of signaling may often characterize specific role arrangements in group interactions, with the associations reliably learned in social agents.

We propose that a thorough treatment of social signaling across all communicative channels is required for a comprehensive computational theory of social groups, and that signaling adaptations must be conceptually separated from the reliable production and detection of by-product cues. We have provided some preliminary examples of recent research in vocal communication and the evolution of music that point in this direction.