Facial expression has evolved as a solution to the primate group living problem. Involving both behavioural and cognitive processes, primate facial expression has diverged in complexity from those of mammalian relatives. We are only just starting to understand how facial expression functions in the form and maintenance of primate social relationships, but the evidence we have so far suggests that the evolution of facial expression has been driven by the need to bond (Schmidt & Cohn, Reference Schmidt and Cohn2001). Dunbar’s theories of group cohesion are therefore key to understanding primate (including human) facial expression.

Dunbar proposes that primates need structural, behavioural, and cognitive solutions to avoid group fragmentation (strategies that are not mutually exclusive). Facial expression is a prime candidate solution, operating across these strategies in three ways. First, facial expressions act as cues to an individual’s future behaviour (Waller, Whitehouse, & Micheletta, Reference Waller, Whitehouse and Micheletta2017). Whether this is underpinned by cognitive reading of intentions or simpler associative learning is an interesting question, but the function is the same. Knowing how others are likely to act next allows individuals to prepare an optimal response, synchronise behaviour, and (importantly) avoid conflict. Illustrating this, primates predict a peaceful outcome when they see others using any facial signals, including facial threats (Waller, Whitehouse, & Micheletta, Reference Waller, Whitehouse and Micheletta2016), presumably as the sharing of the potential willingness to engage in conflict allows all parties to act promptly to diffuse the aggression. Second, the facial expressions of others can be used to understand third parties and their relationships. There has been a historical focus on thinking of facial expressions as revealing information within a dyadic context (when directed towards ‘you’), but their potential utility within third-party dynamics is in many ways much greater. Observing the nonverbal behaviour exchanged between others can be incredibly informative about the nature of their relationship (Bergman & Sheehan, Reference Bergman and Sheehan2013; Dunbar et al., Reference Dunbar, Robledo, Tamarit, Cross and Smith2022). Again, bystanders can then use this information to avoid potential ensuing conflict, take heed of blossoming alliances, identify the social centrality of other individuals, and manage their own relationships with those parties. In particular, this could be a powerful mechanism to keep track of the ‘weak’ ties in the social group (indirect social connections), which Dunbar suggests is important. Third, Dunbar argues that one-trial learning gives anthropoid primates the ability to generalise from single experiences, allowing a more nuanced management of social relationships. We agree, and within the facial expression domain, this is particularly important, as an isolated glance, baring of teeth, or eyebrow raise between parties can be significant and meaningful; it does not need to be repeated to be indicative of a relationship.

We argue that facial expression, via these processes, acts as a bonding mechanism. This function has been evidenced through analysis of individual differences, whereby individuals who use facial expression well (either as a producer or observer) leverage social advantages. Macaque groups led by facially expressive males have greater social cohesion, and expressive male macaques occupy more central positions in their social groups (and are thus recipients of social advantages) (Whitehouse et al., Reference Whitehouse, Clark, Robinson, Rees, O’Callaghan, Kimock, Witham and Waller2024). Similarly, facially expressive humans are more likeable and can be better at negotiating during conflict (Kavanagh, Whitehouse, & Waller, Reference Kavanagh, Whitehouse and Waller2024). Dunbar points out that demonstrating how individual differences in specific skills correlate with relationship management is difficult, and we agree; but this body of evidence in relation to facial expression is building. Other forms of communication, such as vocalisation and gesture, could function similarly, but unlike facial expression, there is little data at present showing how individuals differ in their communicative capacity in these other domains and whether they leverage this to navigate their social worlds better.

Human and non-human primate facial expression has been of interest to scientists for decades, featuring heavily in Darwin’s early comparisons of behaviour across species (Darwin, Reference Darwin1872). Why, then, are studies relating to its social cohesion function in relative infancy? We argue that the dominant emotional framework (Ekman, Reference Ekman1992) in which facial expression is commonly approached tends to distract from this approach. If facial expressions are seen only as ‘expressions’ of subjective feeling state (which is currently heavily debated: Barrett et al., Reference Barrett, Adolphs, Marsella, Martinez and Pollak2019), it is hard to envisage (and test) how they operate as solutions to the group living problem. In contrast, if they are approached as predictive (Fridlund, Reference Fridlund1994) and adaptive (Williams, Reference Williams2002) signals, this better explains how they might help individuals avoid conflict, cooperate, and synchronise behaviour.

One outstanding question is whether human facial expression is different from that of other primates. Did humans leverage the primate facial expression mechanism to break a glass ceiling in group size? Humans produce approximately 100 facial movements per minute during conversational social interaction (Rollings et al., Reference Rollings, Kavanagh, Balabanova, Keane and Waller2024), which is higher than other primates (e.g., 67 facial movements per minute in rhesus macaques during affiliation: Whitehouse et al., unpublished data), but there is no current evidence that absolute function differs. One possibility, however, is that the evolution of language has shifted the function of primate facial expression in humans, by entwining facial movement with language and creating a more multimodal, communicatively rich social interaction where intentions, goals, and predictive cues are shared accurately and rapidly (Jack & Schyns, Reference Jack and Schyns2015). The reality is that the discrete, basic universal expressions (happy, angry, sad, disgusted, surprised, fearful) are rarely used in spontaneous social interaction (with the exception of smiling which may not necessarily indicate actual happiness: Kavanagh et al., Reference Kavanagh, Whitehouse and Waller2024) and instead facial movements pepper conversations allowing individuals to predict each other’s responses (Emmendorfer & Holler, Reference Emmendorfer and Holler2025). These real-world behaviours need to be studied within an evolutionary framework, with explicit comparison to those of other primates, if we are to understand whether this represents a step-change social bonding mechanism. Dunbar’s theories about the mechanisms of group cohesion are therefore important to help generate and test predictions about human facial behaviour, particularly to understand the differences between humans and other primates, and not just the similarities.