It matters little, in effect, who is the victim, provided that there is one.
(Girard et al. 1978, p. 142)
The main Girardian claim being examined by this article is not so much that traits essential to human sociality would have presented early hominins with a novel fitness challenge in the form of increased conspecific violence, but rather that the scapegoat mechanism emerged as an adaptive response to this challenge.
Humans engage in a spectrum of victimization behaviors (Forgas et al. 2005). These have different names depending on the context, such as ostracization, social exclusion, bullying, or lynching. What these all have in common is a dynamic whereby one individual, or a small group of individuals, is singled out and becomes the target for hostility on the part of the rest of the community. Although analogous behaviors such as “mobbing” are observable in some other animals (Ostreiher 2003), these are invariably directed at intruders, or predators of another species. Humans are unusual in having a propensity to single out for attack, sometimes seemingly at random, not only members of other groups, such as in the case of headhunters (Rosaldo 1980), but also members of their own group, although, as noted above, a similar, albeit seemingly less arbitrary, phenomenon has been observed in both of our closest relatives, bonobos and chimpanzees.
Several other authors have addressed the issue of ostracization (Eisenberger 2003; Forgas et al. 2005; MacDonald and Leary 2005; Sweeting et al. 2006), but invariably do so from the perspective of the serious consequences for the ostracized. Girard in contrast focused on the effect on those doing the ostracizing, claiming that the human propensity for spontaneous victimization provided the framework for generating communal unanimity founded upon a shared sacred meaning. This claim is discussed below in the context of other hypotheses regarding the evolutionary origins of cooperation, morality, religion, ritual, and sacrifice, and on how hierarchies have, or at times seemingly have not, operated during hominin history.
Other Hypotheses on Religion and Morality
Numerous hypotheses have been advanced regarding the evolutionary origins of the uniquely human phenomenon of religion (Boyer and Bergstrom 2008; Bulbulia et al. 2008). Some of these consider how religious thinking might have emerged out of preexisting cognitive biases that had originally served other functions (Atran and Norenzayan 2004). For instance, it has been proposed that a hypersensitivity towards agency detection that would have been protective against predators, combined with the emergence of an ability to mentalize, would have resulted in a human tendency to ascribe agency to inanimate objects (Atran and Henrich 2010). Similarly, it has been suggested that while a propensity to better remember minimally counterintuitive information, coupled with a tendency to imitate exactly the behaviors of others, might have facilitated intergenerational transmission of skills and knowledge, it would also have allowed religious-type belief systems and associated rituals to become established through cultural selection (Boyd and Richerson 1985; Barrett 2008; Lewis and Laland 2012).
Others have emphasized the emotional (Dunbar 2020), and ritualistic (Rappaport 1999), rather than cognitive, aspects of religion. Dunbar argues that religion is important to the stability of human groups, with evidence that group fission is less, and thus group longevity greater, in communes with a religious rather than a secular ethos (Dunbar and Sosis 2018). Rappaport linked ritual to the emergence of symbolism and language, and argued that archaic religious rituals involved self-referent “indices” which were impervious to falsification, thereby helping to “ameliorate problems of falsification intrinsic to language.” Although this places a greater emphasis on communal practice, rather than on the cognitive aspects of belief, it has similarities with a cognitivist hypothesis that suggests that ritual sacrifices may be understood as credibility enhancing displays (CREDs) which would have facilitated the cultural transmission of belief systems (Henrich 2009).
Although some of these hypotheses disagree on whether religion is better thought of as an adaptation, or as a by-product of other traits (Sosis 2009), there is a consensus that it played a role in communal cohesion and the emergence of group norms and shared values (Durkheim 1915), with some linking it to the relatively large size of typical hominin groups (Dunbar 2008). It has also been proposed that the dramatic Holocene expansion of human community size, which has occurred in recent millennia, was linked to the emergence of a religious cultural variant that involved belief in all powerful deities, so called “Big Gods,” who take a direct interest in human moral behavior (Norenzayan et al. 2016; Lang et al. 2019). Similarly, Dunbar surmises that the “centerpiece” of the Neolithic revolution wasn’t agriculture as such, but rather the emergence of a form of religion that provided “mechanisms that allowed people to live together in cramped conditions without killing each other” (Dunbar 2020).
Another pertinent hypothesis is Tomasello’s account of the origins of human morality (Tomasello 2016). This considers the emergence of concepts such as sacred values and norms, and links these to human cooperation. Tomasello has proposed an interdependence hypothesis to explain the emergence of such cooperation, beginning with small scale mutual collaboration between individuals founded on a joint intentionality, which evolved into larger-scale group-level collaboration based on “collective intentionality” (Tomasello 2014).
Any upscaling of group size would have been at the cost of reduced individual incentive and reduced information regarding the skills and reputation of partners, thereby increasing the problem of “free riders” and cheaters (Hardin and Garret 2003). Cosmides and Tooby (Cosmides and Tooby 2002) propose that much of human psychology is comprised of dedicated intelligence modules that emerged as specific adaptations to specific problems, and there is evidence for such an adaptation that produces punitive sentiments specifically targeted at free riders (Price et al. 2002).
Cooperative endeavors would have required not only a system for detecting and punishing cheating, but also a shared agreement as to what the behavioral norms were, and how to deal with transgressors (Fehr and Fischbacher 2004; Gardner and West 2004). Tomasello suggests that collective intentionality would have relied on the emergence of new cognitive processes that addressed the problems caused by having to cooperate with relative strangers. This, he claims, would have facilitated the emergence of a collective or cultural common ground, which came to form the basis of shared social norms and what we might call an objective morality (Tomasello 2016, p. 102).
Comparisons with the Scapegoat Hypothesis
What each of the above approaches have in common is the suggestion that codes of unanimously accepted behavioral norms became established through cultural selection and that this selection was mediated by fitness advantages associated with some form of enhanced group cohesion. The scapegoat hypothesis makes a broadly similar claim. Although it may not explicitly articulate the concept of cultural selection, this is, nonetheless, implicit in Girard’s account, as the idea of cultural contagion is a central theme of his work.
What others have described as a “group minded moral psychology” (Tomasello 2016, p. 136), “moral norms” (Atran and Henrich 2010), or “common knowledge” (De Freitas et al. 2019) are, arguably, the same phenomenon as what Girard described as “the sacred,” as they all involve values or beliefs that are shared by the entire social group and perceived as objectively true. Furthermore, Girard is not alone in linking this phenomenon to either punishment or victimization, with clear parallels between his scapegoat mechanism and Tomasello’s hypothesis regarding the punishment of transgressors and the associated emergence of social norms which in turn are linked to group cohesion and meaning. What Tomasello and others describe in terms of a cognitive tendency for third-party norm enforcement and the related creation of cultural institutions that allow punishments to be delivered impersonally or by the group (Dubreuil 2010, p. 189; Tomasello 2016, p. 104), is analogous to what Girard called “sacred violence,” which includes ritual violence, and punishments sanctioned by a deity.
Dunbar argues that early hominins were able to enjoy the advantages of larger group sizes by evolving novel ways of bonding, through the release of endorphins, which were more efficient at creating and maintaining bonds between group members than were the physical contact grooming methods used by other primates (Dunbar 2020). He surmises that the phylogenetically oldest of these uniquely human bonding methods is laughter, which he claims helped to bridge the “bonding gap” between conventional apes and archaic humans (Dunbar 2012). Laughter is a common associate of modern ostracization behaviors (Klages and Wirth 2014) and may also trigger endorphin release (Dunbar et al. 2012). The scenario envisaged by Girard, of spontaneous cathartic victimization, would, therefore, seem consistent with laughter-induced endorphin release playing an important role in early hominin group bonding. Dunbar also points out that the later emergence of a mental capacity to conceptualize four to five orders of intentionality would have been a necessary condition for the type of “storytelling” that is an essential part of communal religions involving social norms (Dunbar 2008). This is analogous to the Girardian concept of the role of mythology, founded on features of the scapegoat mechanism, such as double transference and misrecognition, which would have been similarly reliant on multiple orders of intentionality.
There are, however, some important differences between the scapegoat hypothesis and some of the approaches discussed above. Whereas all accounts envisage the emergence of a process whereby perceived transgressors are punished through some form of ostracization, Girard considered the primary selection pressure to have come from the problem of intragroup conflict, rather than that of free riders. Some other models emphasize the sensitivity of the cheater detection mechanism (Tomasello et al. 2012, p. 679), implying high specificity, with the hypothesized function primarily being the modification of the behavior of accused transgressors. This assumes that the individual accused of norm violation is invariably guilty. Although Girard also envisaged a high sensitivity to cheater detection, he considered the specificity to be of less relevance, as it is the endorsement of group values that he saw as being of greater functional importance. This is consistent with the argument that “bottom-up” social norms function better at addressing the problem of free riders than do “top-down” transgressor sanctions (Ostrom 1990).
In Girard’s account, the actual guilt or innocence of the accused becomes immaterial because, provided that the group unanimously believe in their guilt, the process of norm enforcement works equally well with a victim who is innocent as with one who is guilty. In other words, accusations primarily function by modifying the beliefs and behaviors of the rest of the group, rather than by any effect on the accused. It is this focus on the role of the act of accusation, rather than the substance of the accusation, that, perhaps, best illustrates the difference between Girard and most other comparable theorists.
As with the potential, discussed above, for human desire to become divorced from functional utility, the mimetic nature of accusations would have resulted in a similar potential to become detached from any plausible guilt or innocence, resulting in a tendency to accuse individuals of a limitless variety of crimes, some of which they could not possibly have been guilty of, such as in the case of the medieval European phenomenon of witch trials. Such impossible accusations could not have functioned as means of instrumental behavioral modification, but instead may have functioned by attributing to the accused a malign disposition imbued with a mystique likely to generate myth.
A hypothesis that has strong parallels with Girard’s is that of DeScioli and Kurzban (DeScioli and Kurzban 2013), which seeks to account for the emergence of deontological (non-consequential) morality, contrary to the theoretical expectation that evolution should have favored consequentialism. They suggest that deontological thinking (which is arguably analogous to sacred thinking), evolved as a mechanism that facilitated bystanders to coordinate which side to take in a dispute, thereby curtailing what they call “third party discoordination” and the associated escalation of conflicts. Like Girard, they argue that moral cognition functions by coordinating condemnation, rather than by promoting any specific altruistic behaviors or direct beneficial consequences. This, they suggest, allowed the emergence in hominin societies of a delicate balance between despotism (akin to dominance hierarchies) and discoordination (akin to mimetic crises). Girard made similar claims regarding the scapegoat mechanism, but in addition posited that the victim’s ambiguous status, being both demon and deity, villain and king, together with the associated mythology, played an important role in maintaining such a balance.
Regardless of whether it is deontological or consequentialist, the human phenomenon of third-party punishment has been identified as being of importance to cooperation. Because there is a potential cost for the punisher, this has been described as an altruistic behavior, with debate as to how it may have emerged (Hill et al. 2009). As the greater the proportion of a population who act as punishers, the less the individual-level selection operating against them (Boyd et al. 2003), it has been suggested that, although the maintenance of third-party punishment behavioral traits in a population may be readily accounted for, it is harder to explain how such behaviors originally evolved (Gardner and West 2004). Girard’s concept of spontaneous unanimous hostility generated by mimesis provides a possible solution, as it describes the emergence of what would have been, ab initio, widespread, and thus potentially stable, third-party punishment behaviors.
Human sacrifice, although anathema to modern cultures and religions, appears to have been a widespread practice in premodern societies (Bremmer 2007; Acevedo and Thompson 2013) with a geographical distribution suggesting that it was not the result of local cultural aberrations, but rather was the product of a behavioral tendency ubiquitous in humans. Although Girard’s hypothesis does not offer any obvious explanation for some of the more puzzling, but seemingly pervasive, facets of the phenomenon, such as that of child sacrifice (Stager and Wolff 1984; Wilson et al. 2007), it is arguably consistent with other hypotheses that posit an adaptive function for the practice.
The social-control hypothesis suggests that human sacrifice was motivated by political control and intimidation (Winkelman 2014), and this is supported by evidence linking the practice with greater, and more stable, social stratification (Watts et al. 2016). This would seem consistent with Girard’s view that sacrificial rituals played a role in maintaining social differences. Arguably, Girard offers a parsimonious account of how such practices may have initially emerged from spontaneous acts of communal aggression which would, initially at least, have been devoid of any conscious political motivation or goal.
Whether sacrifice functioned by political intimidation or, as suggested by theorists such as Rappaport and Henrich, by reinforcing intergenerational transmission of unanimous communal values, Girard’s concepts of misrecognition and double transference offer an account of how sacred belief systems, symbolism, and language might have become associated with collective killing.
There are similarities between Girard’s scapegoat mechanism and the phenomenon of coalitions of aggression described above, to which Wrangham attributes self-domestication. These coalitions are characterized by the assailants having an overwhelming numerical advantage, thereby taking little personal risk (Wrangham 2019b). Judicial executions are a modern example of such proactive killing of selected, typically single, victims and some form of this seems to be ubiquitous in human societies (Otterbein 1986). Such coalitions necessarily involve individuals having some ability to discern each other’s intentions with sufficient accuracy, and the observation of a version of the phenomenon in chimpanzees and bonobos may, arguably, be consistent with it being associated with an ability for at least rudimentary second-order mentalizing (O’Connell and Dunbar 2003; Call and Tomasello 2011).
The execution hypothesis contrasts in several ways with Girard’s scapegoat mechanism, perhaps the most significant of which concerns the type of aggression envisaged. Wrangham describes coalitionary aggression as proactive. Indeed, the execution hypothesis essentially posits that, in both humans and bonobos, reactive aggressive traits were selected against by targeted proactive aggression. In contrast, although Girard does not make the distinction between proactive and reactive aggression, he nevertheless clearly envisages the initial scapegoat killing as a collective act of reactive aggression. He argues that lynch mobs do not plan their actions but rather “know not what they do” (Girard 2017, p. 86; English Standard Version Bible 2001; Luke 23:34). However, subsequent reenactments and deliberate human sacrificial rituals would seem to entail proactive aggression and do have some similarities with execution (Smith 2000).
Wrangham and Girard take contrasting views on the primary role played by language. Wrangham focusses on how sophisticated communication would have allowed for complex conspiratorial planning (Wrangham 2019a), whereas Girard, like Dunbar, considers that the more important role played by language related to mythologizing, storytelling, and the intergenerational transmission of sacred values. Arguably, this is a difference in emphasis, with language potentially having played both roles.
They also differ in terms of the hypothesized consequences. Wrangham’s execution hypothesis assumes that individuals who exhibited certain behavioral traits were more likely to be selected as victims and that as a result, the genetic basis for such traits was selected against, whereas because Girard regards the identity of the victim to be arbitrary, and of little relevance to the pacifying function, he does not consider the possibility of selection against any particular trait. In other words, Wrangham emphasizes the genetic consequences, Girard the cultural, although these two distinct consequences might be regarded as having been complementary rather than contradictory.
As alluded to above, some anthropologists have suggested that hierarchical social structures have had a U-shaped trajectory during human evolution (Knauft et al. 1991; Boehm 1999; Dubreuil 2010), and that in contrast to most apes and to modern humans, earlier hominins and hunter-gatherers appear to have lived in more egalitarian groups. It has been suggested that such egalitarianism was the product of a system of “reverse dominance” (Boehm et al. 1993) which emerged from the tendency of subordinate males in the case of chimpanzees, or females in the case of bonobos, forming alliances against dominant males. Boehm argues that, in hominins, this evolved into a communal intolerance of dominants, or “upstarts,” giving rise to what he calls “the domination by the rank and file,” resulting in groups acting as “unified moral communities” (Boehm 1999).
There are parallels between this and Girard’s scapegoat mechanism such that both may be describing the same development. What appears as egalitarianism may in effect have been a binary hierarchy founded on the difference between the community as a whole and any individual who was either overly dominant (Boehm), or perceived (possibly erroneously) to be transgressing in some way (Girard). Boehm described this as a “very special type” of hierarchy and proposed a definitive attitudinal switch from “don’t fool with me” to “don’t fool with us.” This is resonant with the Girardian idea of a switch from “all against all” to “all against one.”
Despite such similarities, an important difference is that Boehm proposes an explicit egalitarian ethos whereby members of a community “deliberately take charge of their own fate” and may even have known what they were doing. In contrast, Girard proposed a more spontaneous and unconscious dynamic, an important part of which would have been the generation of mythology, thereby requiring that the community did not understand their own actions.
Social hierarchical structure has reemerged, and did so contemporaneously with the appearance of larger, more complex human societies at the beginning of the Holocene era, about 12,000 years ago. Boehm’s account suggests that this might be understood in terms of the dominants eventually regaining the upper hand in their struggle with subordinates. Dubreuil has disputed this interpretation (Dubreuil 2010), arguing instead that modern human hierarchies differ qualitatively from the earlier dominance-based variety in that they are characterized by symbolism and political institutions, which he links to the expansion of the temporoparietal area during the transition from Homo heidelbergensis to Homo sapiens, some 160,000 years ago, and to the emergence of sufficient attentional flexibility “to look simultaneously at a person as a man or as a president” (Dubreuil 2010, p. 136). Such an ability to hold ambiguous perspectives on individuals is analogous to Girard’s account of victims acquiring an ambiguous status of being simultaneously both malign and benign. Thus, Dubreuil and Girard both posit that modern political institutions have their origins in status ambiguity of individuals, although Dubreuil does not ascribe any specific role to victimization in generating this.