The prehistory of violence and war: Moving beyond the Hobbes–Rousseau quagmire

This article presents and critically assesses the latest anthropological and archaeological research on the chronology, lethality, and frequency of violence and war in human history. Stepping back from the rhetorically polarized dispute between ‘Hobbesians’ and ‘Rousseauans’, the article examines the methods and findings of the latest research in a conceptually novel way, i.e. by dropping the existing and widely used polarized terms that have inevitably framed the literature so far. The article demonstrates that multiple sources of evidence point more in the direction of the modal human prehistoric social organization, i.e. nomadic hunter-gatherers, likely having warfare only in a minority of cases, or war even being virtually non-existent (with interpersonal violence being more common). The dispute over this claim so far is found to stem, at least in part, from the varying definitions of war and the grouping together of nomadic with complex foragers. More significantly, the disagreement is due to different sampling and sourcing techniques of different researchers, the biggest divide being between self-selection/systematic sampling and first-best/second-best sources. Important potential warlike exceptions are also noted and discussed in the article from multiple angles (Jebel Sahaba, Nataruk, Aboriginal Australia, etc.), as are the discovered precursors and enabling conditions of war, such as the complexification of (nomadic) hunter-gatherer societies with the transition to settled life.


Introduction
A decade ago, Gat (2012) published an influential review of the recent 'decline-of-war' literature (LeBlanc, 2003;Gat, 2006;Pinker, 2011) in the Journal of Peace Research, taking into account the level of violence both during prehistory and more modern times. His assessment was unequivocal: 'Hobbes was right, and Rousseau was wrong' about both the chronology and the lethality of violence in human history. Like Pinker's (2011) popular book on the topic, Gat's review (see also Gat, 2015) has ostensibly demonstrated violence to be quite prevalent in the human state of nature, declining only later with the rise of states, and declining further even later with the emergence of capitalism and democracy. In the last decade since Gat's review, new studies on the topic have emerged and past scholarly rivalries have reignited, further shedding light on this interesting and important question. Drawing on some of them, historians, biologists and psychologists have again published popular books on the topic. The general rhetorical stakes of the debate, again invoking Hobbes and Rousseau, remain the same as before. Interestingly, however, the conclusions of this most recent literaturecontra Gat's declaration of victory for Hobbes and defeat for Rousseau -seemingly fluctuate wildly between both. Sapolsky (2017: 327) concludes that: The HGs [i.e. hunter-gatherers] who peopled earth for hundreds of thousands of years were probably no angels, being perfectly capable of murder. However, 'war' [ . . . ] seems to have been rare until most humans abandoned the nomadic HG lifestyle. Our history as a species has not been soaked in escalated conflict. MacMillan (2020: 14) insists on the opposite, asking, 'So, the Rousseauian version of human history or the Hobbesian? The archaeological and historical evidence points firmly toward Hobbes, with war as a long-standing and integral part of human experience'. Bregman (2020: 91-92) denies this, asking 'What happens if we examine a society that has no settlements, no farming and no horses -a society that can serve as a model for how we once lived? You guessed it: [ . . . ] war is a rarity'. Finally, Pinker (2018: 199) basically repeats his conclusion from an earlier bestseller: 'In the absence of a government or powerful neighbors, tribal peoples tend to fall into cycles of raiding and feuding, with death rates exceeding those of modern societies, even including their most violent eras'. Pagel (2016: 181) characterizes a recent phylogenetic study by Gómez et al. (2016) as showing that 'Hobbes has landed a serious blow on Rousseau' -with the important qualification that he has, nevertheless, 'not quite knocked him out'. Hames (2019: 155) is somewhat less reserved, claiming 'Hobbesians with empirical data have already established that the frequency and intensity of hunter-gatherer warfare is greater compared with large-scale societies even though horticultural societies engage in warfare more intensively than hunter-gatherers'. There are other recent examples like these.
The goal of this article is to make sense of this conundrum through a critical and balanced assessment of the methods, definitions and findings of the recent research on the topic. The starting conceptual premise of the article is that, in order for our current understanding and future research to move beyond the polarized and by now obviously unproductive Hobbesian-Rousseauan split, scholars should stop framing it in this manner. There are two primary reasons for this. First, the terms themselves are unclear and tend to evoke misleading connotations, especially outside the specialist literature, of either pervasive violence and fear or near-total absence of violence in human prehistory, none of which is or ever was borne out in the empirical literature. Second and more importantly, the framing can undermine best scholarly practices as it puts the focus more on the individual researchers and their priors than on the actual findings -findings which might not easily fit under one rubric or the other. I will, therefore, refrain in the present article from invoking these and related terms.

Critical assessment of the anthropology of violence and war
Before delving into the evidence, a note on definitions is required. An important conceptual distinction to be noted is between person-on-person (i.e. interpersonal) violence, such as homicide and murder at the lethal extreme, and group-on-group (i.e. intergroup) violence. War is obviously an instance of the latter, not former. The involvement of multiple people is required for a violent act to be war or at least minimally war-like (in a small-scale society the number is usually low in an absolute sense, but in larger or even state societies a much higher absolute number is required) and the conflict needs to happen between distinct groups, i.e. human collectives divided along social lines. Thus, it is somewhat hard to firmly classify feuds and raids -forms of HG violence which can be both groupish and war-like as well as more interpersonal in nature -as instances of either war-or non-war violence. This definitional ambiguity cannot be avoided. Pinker's (2011: 49) famous figures -drawn from Bowles' (2009) nomadic HG sample and analysiswhich purported to reveal that prehistoric warfare was highly lethal, were recently rebutted by Fry & Söderberg's (2013a) larger and ostensibly more accurate sample. The weak points in Bowles' sample, which shows that war in nomadic HG societies on average accounts for a whopping 14% of all deaths, are at least twofold. First, not all of his cases are nomadic HGs. Second, in two cases (the Ache and the Hiwi) all 'war deaths' are actually caused by frontiersmen ranchers, not nomadic hunter-gatherers, which somewhat reduces the percentage death from 14% to 9% (Fry, 2013; for other points of contention see Fry & Söderberg, 2014).
Dispelling the notion of pervasive and chronic warfare, the sample used by Fry & Söderberg (2013a: 272) shows that roughly 'half of the societies had no lethal events that involved more than one perpetrator'. Earlier, Fry (2007) has similarly argued that a subsample of 21 societies from the 35 included in the Standard Cross-Cultural Sample (SCCS) of nomadic hunter-gatherers, complex hunter-gatherers and equestrian foragers reveals that the majority of nomadic HG societies do not fight, while all of the societies belonging to the last two types have experienced war.
Popular social-scientific sources cited in the Introduction, like Sapolsky (2017) and Bregman (2020), as well as other scholarly reviewers, for example, the sociologists Malešević (2014) and Mann (2018), rely primarily on these two studies when arguing that war was likely absent in prehistory. However, what they do not report is that the challenges can themselves be, and have been, challenged. For instance, Hames (2019) has noted methodological as well as statistical problems with Fry & Söderberg (2013a), while Wrangham & Glowacki (2012) point to issues with Fry (2007).
Methodologically, Fry & Söderberg restrict their sample (unduly, according to Hames), excluding other reliable sources from it. More specifically, to calculate cases of lethal aggression for the whole sample they rely solely on the Principle Authority Sources (PAS), which are the earliest high-quality descriptions of the societies in the SCCS. But this is problematic, claims Hames, because one can use the SCCS (as one should) without relying solely on PAS sources. Moreover, some PAS sources 'contain little detail on violence [ . . . ] even though they may have rich and extensive information on standard ethnological topics' (Hames, 2019: 162). Lastly, some lethal cases described by PAS sources are caused by modern outside influence and are, therefore, irrelevant for inferring the state of violence in prehistory.
However, Fry & Söderberg rely on PAS sources precisely because they are of high quality, thus being informationally most reliable descriptions, and because they are the earliest (high quality) descriptions, thus least likely to provide post-contact lethal cases, which would be spurious for purposes of prehistorical violence inference. They are also aware that outside influence is a potential problem in certain cases, even with PAS sources, but report that they 'find few examples of this effect in the early ethnographic sources used in this study' (Fry & Söderberg, 2013b: 2). Lastly, Hames' proposed solution to the potential problem faced by Fry & Söderberg's analysis, that is, 'it is up to a researcher to select those sources she or he believes to be reliable and accurate, and by citing those sources in requisite detail to allow others to evaluate the quality of those sources' (Hames, 2019: 162), is understandable in principle. Especially so in light of the existence of other (non-PAS) reliable sources pointed to by Hames, such as used by Kelly (2002) or presented by Lee (1979), which demonstrate a much higher count of lethal events in the case of hunter-gatherers in Andaman Islands and the !Kung compared to Fry & Söderberg's count. At the same time, importantly, selective personal picking and choosing of sources as recommended by Hames can lead to increased subjectivity and reduced rigor in research.
Moving on to the sample used by Fry (2007), it includes only nomadic and seminomadic HG/forager societies (n ¼ 21) from the SCCS, deliberately excluding other non-nomadic and more complex, later types of HG (i.e. settled and equestrian foragers) generally appearing only after around 10,000 BC, which otherwise together make for 35 foraging societies in the SCCS (21 nomadic þ 13 non-nomadic). Fry's nomadic HG sample shows that, of all the 21 societies, the majority (13, or 62%) are not engaged in warfare. This is significantly more peace than the 0% cases of peaceful societies among the complex HG and equestrian forager groups. As Fry (2018: 256) concludes, 'most nomadic forager band societies do not make war'.
Notably, however, this nevertheless still means that a large minority -more than a third or 38% -of nomadic HG cases are coded as 'warring' by Fry himself. This is not evidence of 'absent' or 'exceedingly rare' war in social types close to the typical prehistoric society. Fry (2018) warns that such nomadic HG war is (1) less severe in comparison to complex foragers, is (2) sometimes explained by contact with outsiders, and (3) tends to stem from interpersonal motives (adultery, insults, etc.), which suggests that at least some cases are irrelevant for inferring the likelihood of prehistoric nomadic HG war (due to reason 2) or simply do not necessarily constitute group-versus-group conflict, in other words war (due to reason 3). These are judicious observations that have to be taken into account, but they do not explain away all cases of nomadic HG warring.
Wrangham & Glowacki (2012: 21) note a different potential issue, namely that the majority of nomadic HG cases in Fry's 2007 book (13 out of 21) have had interactions with non-HG and state societies 'in ways suggesting that they were militarily and/or politically subordinate'. This could, importantly, mean that at least some if not most of them are non-warring not due to the inherent nomadic HG tendency for peace but due to an outside interaction effect. An effect of which Fry is otherwise correctly wary and has himself used to criticize Bowles (2009). Fry (2018) has suggested that Wrangham & Glowacki make a mistake in this assessment, because the sample of 21 societies they refer to is not, as one might think, Fry's (2007)  Wrangham & Glowacki's analysis is justified in warning against simple inference from the existence of peaceful (even externally pacified) nomadic HGs to the existence of a general tendency of nomadic HG to be non-warring in their natural and prehistorical environment. And, indeed, according to sources cited by Wrangham & Glowacki, of at least the six non-warring societies appearing on both lists, four (!Kung, Hadza, Mbuti and Semang) have had interactions with nonnomadic HG neighbors (whether they are militarily/ politically subordinate is a different question, especially because according to their own sources, two have also had recent fights with outsiders), one (Yahgan) has little pre-contact ethnography, and one (Copper Inuit) is extremely isolated, so has no opportunity for war. However, this does not rebut Fry's (2007) SCCS sample and findings. First, the overlap between both samples is sparse, making direct comparison impossible. Second, Wrangham & Glowacki compiled both their own sixcase sample (see Table 1 in Wrangham & Glowacki, 2012) and the partial version of Fry's 2007 sample (see Table 3 in Wrangham & Glowacki, 2012) in a not wholly appropriate way. They self-selected cases (e.g. only six of the 21 SCCS nomadic HG cases are used in Table 3) and, for purposes of analysis, they sometimes relied on secondary instead of primary sources (see also Fry, 2018: 260).
In summary, contemporary studies based on more robust methodological principles suggest a much lower (though not vanishingly small) lethality and tendency of nomadic HGs to engage in war than suggested by popular accounts such as Pinker (2011).
Critical assessment of the archaeology of violence and war Ferguson (2013a) has gone through all the 21 cases which Pinker had taken from Keeley (1996) and Bowles (2009) to compile his key list for estimating the lethality of prehistoric war. Ferguson shows how, of the original 21 cases, seven have to be excluded either due to containing only a single instance of violent death (which is insufficient evidence for war of any kind) or due to erroneous duplication. Of the remaining 14 cases, Ferguson argues, multiple ones are not representative because they are 'highly unusual', thus 'distorting war's antiquity and lethality'. In a much broader archaeological review that considered all the available archaeological evidence for Europe and the Near East, Ferguson (2013b) points out the absence of war for the Middle and upper Palaeolithic, and he paints a less bellicose picture for the Mesolithic (down to 3.7-5.5% war casualties as compared to Pinker's 15%). Haas & Piscitelli (2013: 176) concur. As they claim in their review of the evidence, 'In looking for archaeological indicators of warfare, a fairly [clear] line can be drawn at approximately 8,000 BC (10,000 years ago). After 8,000 BC there is evidence of significant, though localized, warfare in several parts of the world'. This is an important and strong challenge to the claim that we positively know (based on archaeological data) that war had been pervasive in pre-Neolithic times. There is little, if any, reliable indication of war on skeletal remains from before 8,000-10,000 BC (Haas & Piscitelli, 2013).
In response to this, three points can be made. First, researchers point to a few prominent and ostensibly pre-Neolithic cases such as Jebel Sahaba (located in the Nile Valley) and Nataruk (located in Turkana City, Kenya) that perhaps indicate collective violence. The authors of the recent report on skeletons from Nataruk themselves claim that 'the deaths at Nataruk are testimony to the antiquity of intergroup violence and war' (Mirazón Lahr et al., 2016a: 397). This has been challenged by Stojanowski et al. (2016: E8), who claim that 'most of the observed cranial damage is inconsistent with blunt force trauma'. Mirazón Lahr et al. (2016b: E11) replied to the challenge, maintaining that 'A case of inter-group conflict remains the best explanation of the events at Nataruk'.
The case of Jebel Sahaba is also open to interpretation. A recent full reanalysis of the case by Crevecoeur et al. (2021) uncovered new findings, such as many previously undocumented healed and unhealed lesions and the repetitive nature of the trauma. Most importantly, the claim that this was a single warfare event is not supported. Still, the researchers agree that 'the projectile nature of at least half of the lesions suggests intergroup attacks, rather than intra-group or domestic conflicts', and that the various pieces of evidence 'indicate small episodes of recurring violent events such as raids or ambushes against this community' (Crevecoeur et al., 2021: n.p.). In any case, it has to be admitted that such events are exceedingly rare in the available archaeological record. And the further back in time one goes, the less evidence there is (Ferguson, 2013b). It is thus hard to see how such exceptional cases could provide solid systematic insights into the ubiquity of prehistorical war.
Second, some researchers (e.g. Hames, 2019) make the point that the paucity of pre-Neolithic evidence for war does not mean that we can therefore positively conclude war only emerged with settlements, agriculture and states, instead of being already present in nomadic hunter-gatherer societies (existing either before the 8,000 BC mark or after). Indeed, this would be a non-sequitur. Such fallacious reasoning can occur due to equivocation of 'after 8,000 BC' with 'settlements', 'agriculture' or 'states'. The conclusion from one to the other would perhaps be reasonable if most (or all) societies after the Neolithic revolution rather quickly transformed into settled, agricultural or state societies. Then the link between evidence for, as Haas & Piscitelli (2013: 176) say, 'significant [ . . . ] warfare' after 8,000 BC and a concurrent move away from hunting and gathering specifically would have been tight.
But there are societies after 8,000 BC that have maintained their prehistoric form in subsequent millennia. This means that archaeological evidence for 'significant war' only after -but not before -the Neolithic revolution might still be consistent with the claim that war is not foreign to nomadic foragers. Take the example of Australia (or North America). In Australia, millennia have passed since 8,000 BC without the Aboriginal tribes losing their original social form, which is nomadic hunter-gathering. If there is archaeological evidence for war in Australia after 8,000 BC, but not before, this is clearly not evidence for war emerging only with the transition from nomadism to settled life, as this was absent even in the post-8,000 BC era in the case of Australia. In fact, there is archaeological evidence for warfare in both Australia and North America (Allen, 2014;Gat, 2015) after the Neolithic revolution, even though societies in these regions ostensibly remained pristine. As Hames (2019: 167) notes, 'Haas's 10,000-year mark may be relevant in certain Old World contexts, [but] it is completely irrelevant for other parts of the world'.
It remains disputed, however, whether North American foragers were really nomadic and not settled complex HGs (Fry, 2007 also disputes some of the aforementioned evidence of Aboriginal warfare, but is critiqued by Gat, 2015). More importantly, archaeological skeletal research on the sequential origin of war on Kodiak Island in the North Pacific, eastern North America, Valley of Oaxaca and Japan (see sources in Fry, Keith & Söderberg, 2020) robustly points in a different direction. In all these regions, archaeology demonstrates warfare originating only after the transition from nomadism to settled complex life, regardless of whether the transition happened shortly after the 10,000-year mark or only much later. Before the transition, there is evidence of homicides (eastern North America), but not evidence of war-related deaths, weapons, or fortifications -with the Kodiak Island being a possible partial exception as minimal warfare before the transition is not ruled out (see Fry, Keith & Söderberg, 2020). Suggestively, virtually all these regions (and those already mentioned) evince a distinctive pattern of increasing war with increasing social complexity and later timeframes.
Third and closely related, there is the claim that the absence of evidence of war in deep prehistory is not the same as evidence of war's absence in that period. More specifically, Gat (2015) and Hames (2019) point out that, even if war had actually been ubiquitous in the nomadic pre-10,000 BC period, archaeological evidence of it would probably be lacking due to the nomadic nature of prehistorical societies. Because they were so small and mobile, they would not have had large and durable fortifications or concentrated cemeteries, which we could now discover and use to infer how warring they were and what the lethality of conflicts was.
This last objection is understandable in principle. Nevertheless, given the recent examinations of the whole of the archaeological archive of skeletal human remains around the world (e.g. Haas & Piscitelli, 2013), which do not show any systematic signs of warfare before the 10,000-year mark, but do so quickly afterwards, it does not seem that we are confronted simply with absent evidence for a certain period but rather precisely with evidence of absence.

Conclusion
In conclusion, although interpersonal violence is not foreign to the modal prehistoric human social organization, current anthropological and archaeological research points more in the direction of human nomadic HG prehistory not being soaked in warfare, war instead being a minority affair (going by the systematic ethnographic samples of existing nomadic HG societies) or even virtually non-existent (going by systematic archaeological evidence). The following concluding points summarize, draw together and discuss the examined evidence from the previous two sections.
1. Has interpersonal violence been a part of human prehistory, predating state societies and even the Neolithic revolution? Yes, there is no contemporary dispute about this question. An important recent study (Gómez et al., 2016) demonstrates that human violence has a phylogenetic component, with the value (2% of human deaths caused by interpersonal violence) similar to the 'phylogenetically inferred values for the evolutionary ancestor of primates and apes' (Gómez et al., 2016: 233). 2. Are nomadic HGs capable of war? Yes, they are. This is known primarily because some (although a minority) of modern but preserved nomadic HGs are warring. 3. How likely were prehistoric nomadic HGs to engage in war? Inferring from careful anthropological analyses, the answer is that they would not have been very likely to do so. Nevertheless, the same anthropological evidence shows that (at least a sizable minority of) nomadic foragers do make war and not only because of outside contact, which means that prehistoric nomadic HG warfare was certainly possible and likely in certain cases (Aboriginal Australia is likely another important case in this regard, see Gat, 2015). In fact, whether (nomadic) foragers are violent or peaceful seems not to be the appropriate question. It is obvious that they can be both. 4. A better question, then, is simply what conditions trigger them to more likely act violently or peacefully. Multiple studies show that increasing population density, hierarchy, settled life, and so on, are among the key conditions that make war among foragers comparatively more likely. 5. Does archaeological evidence point to any systematic presence of war in nomadic HG prehistory (i.e. before the 10,000-year mark)? There are no systematic signs, only a few sporadic cases. Instead, several world regions evince a distinctive archaeological pattern of war developing and intensifying only with the complexification of (nomadic) HGs. 6. Does the anthropological and archaeological evidence demonstrate a positive and comparatively frequent and intense war in other pre-state or even pre-agricultural societies, such as complex and equestrian foragers? Yes. 7. Overall, a distinct 'inverted-U-shaped' pattern of the evolution of violence can be seen (on this, see already Fry, 2013), with prehistoric violence being no more lethal than what human phylogeny predicts (2% lethality), the rate exceeding the phylogenetic value in the Iron and Medieval age, and then decreasing again, this time even below the 2% rate, with the Modern and Contemporary age (Gómez et al., 2016: 236). Prehistoric tribes had higher violent lethality than prehistoric bands (but both are at the phylogenetic value), later chiefdoms were much more lethally violent (exceeding the value, as is the case for contemporary bands and tribes), with historic and contemporary states then descending below the value (Gómez et al., 2016: 236).