The evolution of lethal intergroup violence - Raymond C. Kelly #

Department of Anthropology, University of Michigan, 550 East University Avenue, 101 West Hall, Ann Arbor, MI 48109-1092
This contribution is part of the special series of Inaugural Articles by members of the National Academy of Sciences elected on April 20, 2004.
Contributed by Raymond C. Kelly, July 14, 2005

Recent findings and analyses in evolutionary biology, archaeology,
and ethnology provide a favorable conjuncture for examining the
evolution of lethal intergroup violence among hominids during the
2.9-million-year Paleolithic time span. Here, I seek to identify and
investigate the main turning points in this evolutionary trajectory
and to delineate the periodization that follows from this inquiry.

collective violence
armed conflict
war
hominid evolution

Detailed and well documented reports of intergroup attacks
and killings among free-ranging chimpanzee groups were
initially published during the period from 1979 to 1986 (1–4). In
subsequent years, data from additional research sites confirmed
that such episodes of lethal violence were characteristic of
chimpanzees as a species (5). These findings clearly raise an
important question with respect to the evolution of lethal
intergroup violence: Did the same constellation of causal factors
that gave rise to the chimpanzee pattern of coalitionary killing
of neighbors produce a parallel (and convergent) outcome
among Paleolithic hominids? Recent findings and analyses in
evolutionary biology, archaeology, and ethnology create a favorable
conjuncture for advancing our understanding of the
evolution of lethal intergroup violence. These data provide
grounds for evaluating Wrangham’s preliminary conclusion that
‘‘current evidence supports the hypothesis that selection has
favored a hunt and kill propensity in chimpanzees and humans,
and that coalitional killing has a long history in the evolution of
both species’’ (ref. 6, p. 1).
The crux of Wrangham’s explanation of the evolution of
coalitionary killing is that fitness is closely correlated with access
to food resources and that territorial enlargement thus enhances
fitness. In short, ‘‘fitness is correlated with territory size’’ (other
factors being equal) (ref. 6, p. 15). Unprovoked attacks on
members of a neighboring community thus convey a selective
advantage, provided that the costs to the attackers are low. The
mechanism by which aggression is rewarded is intercommunity
dominance. ‘‘If raiding leads to the wounding or death of a
neighboring male, the neighboring community’s competitive
ability is substantially reduced’’ (ref. 6, p.15). The dominant
community can thus freely encroach on the territory of its
neighbor whenever food resources within its own territory are in
short supply. The dominant community also may have an
advantage in recruiting reproductive females. However, the
capacity to translate additional females into increased fitness
would be contingent on the commensurate expansion of food
resources.† Thus, although intercommunity dominance ‘‘tends to
lead to increased fitness of the killers through improved access
to resources such as food, females, or safety’’ (ref. 6, p. 12),
territorial gain is the critical ingredient for the realization of this
fitness enhancement.
Wrangham applies his explanatory framework to an extensive
array of available data on intraspecific coalitionary killing of
adults among chimpanzees, bonobos, and wolves and is able to
account for variability in the frequency of intercommunity
attacks. For example, encounters between neighboring chimpanzee
communities are most likely to eventuate in attacks when
a single individual is spotted by a party of three or more males
who are able to overwhelm a lone victim while incurring little risk
of injury to themselves. Attacks are more frequent in those
portions of the chimpanzee geographical range where competition
for food resources is intensified by longer dry seasons.
Lethal intercommunity attacks are not reported for bonobos,
who rarely engage in the kind of dispersed feeding that exposes
lone individuals to fatal attacks by neighbors (ref. 6, pp. 12–18).
All in all, the available data largely support the imbalance-ofpower
hypothesis and the intercommunity dominance-drive
hypothesis formulated by Wrangham.
However, there is one apparent contradiction in Wrangham’s
explanatory framework. He documents border avoidance and
underutilization of food resources located in the border areas
between group territories. At the Taı¨study site, for example, ‘‘. . .
75% of time was spent in the central 35% of the [territorial]
range’’ (ref. 6, p. 11). Although this substantial day-to-day
frequency of border avoidance shows that coalitionary attacks by
neighbors are anticipated (and constitute a pattern), such border
avoidance also entails a reduction in the effective size of group
territories as repositories of food resources, because 65% of
these resources cannot be exploited in complete safety.
Available data are insufficient to precisely quantify the magnitude
of resource underutilization at all chimpanzee study sites.
However, the Taı¨ data (cited above) clearly indicate that a
substantial percentage of the available food resources present
within a group territory are not exploited in this instance, and
border avoidance is established by Wrangham as a general
phenomenon.
If fitness is enhanced by territorial enlargement, then fitness
would be reduced by a pattern of lethal intercommunity attacks
that curtails resource availability along borders. Wrangham
assesses the costs of aggression in terms of the risk of serious
injury to the members of a party of chimpanzees carrying out an
attack. Resource underutilization does not enter into his costbenefit
calculations or his fitness assessments. If the posited ‘‘. . .
benefits to be gained by raiding or killing neighbors’’ (ref. 6, p.
15) flow from territorial expansion that, in actuality, only adds
a buffer zone, rather than food resources, then the value of this
benefit would need to be recalculated. Factoring in a resourcereduction
penalty for initiating aggression, and rendering one’s
own border zone unsafe as a result, also alters the cost side of the
equation. The circumstances under which the potential benefits
of coalitionary attacks definitively outweigh the costs thus
appear more narrowly circumscribed than Wrangham posits.
Alternative adaptive strategies also may be superior in a broader
array of contexts. Mutual nonaggression would clearly convey
both mutual benefits and reduced costs. More specifically, the
reported bonobo pattern of peaceful interactions between individuals
of neighboring communities, in which ‘‘. . . they rest,
travel, copulate, play, and groom together’’ (ref. 6, p. 17), entails
an absence of border avoidance that should convey a selective
advantage. The identification of a potential advantage linked to
nonaggression calls into question Wrangham’s conclusion that
selection favors lethal coalitionary attacks on neighbors and that
these selective forces would produce a convergent evolutionary
outcome among humans. As noted earlier, my principal concern
is to assess the question of convergence.
Recent analyses of intercommunity interactions among unsegmented
hunting and gathering societies‡ that are organizationally
comparable with those that existed during the
Upper Paleolithic period (35,000 –10,000 B.P.) provide
grounds for considering the applicability of the imbalance-ofpower
and intercommunity dominance-drive hypotheses (7).
The Andaman Islanders (ca. 1858) constitute an especially apt
case in that resource scarcity is reported. In other words, the
condition of intensified food competition is present (as in the
portion of the chimpanzee geographic range where intercommunity
attacks are most frequent). Conflict between neighboring
groups over access to resources is more frequent among
the Andamanese than among comparable hunting and gathering
societies in which resources are reportedly plentiful
relative to population (ref. 7, p. 142). The cooccurrence of
resource scarcity and conflict over resources is consistent with
Wrangham’s explanatory framework.
However the cost–benefit relationship is very different from
that described for chimpanzees, in that trespassing entails considerable
risk of injury. Intrusion into the territory of a neighboring
group involves entering an area that is better known to its
owners than to the intruders. More importantly, the owners are
hunters who stalk game with weapons that kill at a distance.
There is no guarantee that intruders will spot members of the
owning group before they themselves are spotted. Under these
circumstances, an undetected lone hunter of the owning group
is not at a disadvantage in relation to a larger party of intruders.
He may stalk them and kill one individual before fleeing to alert
others. Given superior knowledge of the territory, this course of
action entails little or no risk.
If two hunting parties become aware of each other’s presence
at the same time, the numerically smaller group withdraws and
yields the contested food resource to the stronger party. However,
if one party secures the advantage of remaining undetected,
it employs this advantage to ambush the other party. The relative
size of the two parties is not a factor under these conditions of
asymmetrical detection. Moreover, substantial casualties are
inflicted on these occasions. Thirteen encounters reported in
colonial records (8) resulted in five serious wounds and six
deaths, yielding an 85% overall casualty rate (ref. 7, p. 100). In
1863, Andamanese guides thus cautioned early British colonial
exploratory patrols that proceeding beyond a certain point
would inevitably provoke a lethal attack.
Jacko pointed to my heart and represented the act of a
savage aiming at me with his bow and arrow, of the arrow
piercing my heart and my falling wounded, closing my
eyes, and expiring. Topsy also pathetically enacted the
death scene, and both waved their hands deprecatingly
in the direction disapproved of and entreated me not to
proceed further but to return [to base].
Ref. 8, p. 435
This exploratory patrol was quite large in relation to Andamanese
hunting parties. Firearms amplified this numerical advantage.
Nevertheless, Jacko and Topsy clearly did not think that
this superiority in numbers and weapons would render the risks
insignificant. It is implicit in their pantomime that trespassing is
a capital offense. Intruders are presumed to be engaged in game
theft and are shot on sight.§ The first sign of contact with
territory owners is thus likely to be an arrow fired by an unseen
archer. In the region of most intense conflict, Jarawa men wore
bark torso armor extending from the armpits to the hips as
protection against being ambushed during the course of routine
subsistence activity (7).
These ethnographic data show that the potential costs of
intrusion are extraordinarily high. The central premise of
Wrangham’s imbalance-of-power hypothesis ‘‘ that one party can
attack the other with impunity’’ (ref. 6, p. 1) is not met.
Moreover, there is no realistic prospect for achieving intercommunity
dominance that would enable one group to encroach on
the territory of a neighbor at will. The consequences of intergroup
hostility are thus stalemate and an analogue of a war of
attrition in which there is a 0.02% mortality per annum from
spontaneous lethal conflicts over resources, with 85% of this
mortality being among males (ref. 7, pp. 100 and 158). Such
conflicts are generally not sought but can occur when two groups
of hunters inadvertently encounter one another in the course of
the food quest at the border of their respective territories.
As noted among chimpanzees, the risk of encountering hostile
neighbors encourages border avoidance. The concomitant underutilization
of food resources also reduces the effective size of
group territories. Thus, in the southern portion of the Andaman
Islands, where endemic hostile intergroup relations obtain, the
population density (of 2.0 persons per square mile) is only 73%
of that attained in North Andaman, where periodic peacepromoting
joint gatherings of neighboring bands entail mutual
exploitation of food resources located in border areas. On these
occasions, two bands (of about 45 individuals each) camped
together at a site located in the zone between their respective
territories that might otherwise have been contested and therefore
avoided because of the possibility of ambush. The resources
of this zone were exploited through joint hunting, fishing, and
gathering expeditions, with the proceeds of these endeavors
shared and consumed at feasts. The general atmosphere of these
gatherings has been described as one of amiable rivalry in which
an effort is made to put aside residual ill feeling arising from past
quarrels. Gifts were exchanged, including bows, arrows, arrowheads,
adzes, baskets, pigments, and shells for personal adornment.
Each individual donor endeavored to outdo his or her
counterpart in generosity. In the course of the festivities, men
performed dances to the accompaniment of female singing (7).
These gatherings provided a context for courtship leading to
intermarriage. Kin ties between neighboring bands arising from
marriage were reinforced by reciprocal visiting and a cultural
practice of adoption whereby a child of one band was raised by
a friend or relative of the parents residing in another band.¶ This
repertoire of peace-promoting practices facilitated more complete
utilization of food resources in North Andaman, resulting
in a higher population density (of 2.75 persons per square mile).
If fitness is augmented by territorial enlargement (as a general
rule of selective advantage), then peace-promoting practices are
adaptive and the maintenance of hostile intergroup relations is
maladaptive under conditions commensurate with those found
in the Andaman Islands. When relations with neighbors are
friendly, an average-sized territory of 16.4 square miles is
capable of supporting a band of 45 individuals. Hostile relations
shrink the utilizable area to the equivalent of 12 square miles,
with a peripheral zone about one-third of a mile wide subject to
border avoidance. This reduced territory is sufficient to support
a band of only 33 individuals.
When group territories are extensive and population densities
are low, the percentage of territorial area rendered unexploitable
by border avoidance is correspondingly reduced and the cost
of hostile relations with neighbors would be less than that
documented for the Andamanese. For example, a hunting and
gathering band may require a territory of 100 square miles in a
less favorable arctic or desert environment. Under these circumstances,
avoidance of a one-third-mile-wide border zone would
reduce a group’s effective territory by 11.4%. With a 200-squaremile
territory, the comparable figure is 8.1%. However, low
population density also entails little or no competition for food
resources and infrequent encounters between members of neighboring
groups so that conflicts rarely occur and those that do are
peacefully resolved (ref. 7, pp. 133 and 142).
The selective factors that give rise to coalitionary killing of
neighboring group members among chimpanzees do not have the
same effects in the case of unsegmented foraging societies.
Fitness does correlate with territory size, or more accurately,
effective territory size (provided that all proximate territories
are occupied). Moreover, intergroup competition is intensified
at higher population densities. But the attainment of intercommunity
dominance is thwarted by a combination of factors that
increase the costs to attackers. First, a party of hunters from one
group cannot be certain that they are able to make an accurate
assessment of the strength of the opposition. There is no
certainty that a single individual who has been sighted is indeed
alone, because hunters often work in pairs or teams that stalk
game silently and employ ambush techniques. It is consequently
unclear whether a favorable imbalance of power obtains. There
is substantial risk that costs may be underestimated. Second,
territory owners have intrinsic advantages over intruders that are
akin to the advantages enjoyed by combatants who have set up
an ambush in territory well known to them but not to opponents.
Hunters lying in wait along game trails can readily kill intruders.
Third, a hunter is a dangerous quarry because he possesses
weapons that kill at a distance. Weapons augment the lethality
of combat and amplify the costs of assessment errors made by
attackers. These errors include the potentially mistaken deductions
that the party seeking to initiate an attack has remained
undetected (and enjoys the advantage of surprise) and that
attackers have a numerical advantage (a deduction based on the
assumption that all members of a targeted hunting party have
been spotted).
If entirely accurate assessments could be made, weapons
would amplify the effects of both numerical superiority and
surprise, and the consequences of augmented lethality would
accrue to attackers. In other words, the decisive factor in the
balance of power between intruders and territory owners is
detection, not weaponry. (That asymmetrical detection outweighs
superiority in weapons and numbers and definitively
determines the outcome of territorial incursions is precisely the
point Jacko and Topsy sought to make.) Moreover, the risks of
both asymmetrical detection and assessment errors are intrinsically
greater for intruders as they endeavor to move into
unfamiliar territory in which hunters employing ambush techniques
may well be operating from concealed positions. The
necessity of movement in initiating an attack is itself a disadvantage
with respect to detection. Moreover, territory owners
invariably sound the alarm if combat is joined. Other hunters in
the vicinity converge on the area and may potentially cut off the
retreat of intruders. In sum, projectile weapons amplify the
casualties (and costs) of initiating attacks on neighbors as a result
of the intrinsic advantages enjoyed by defenders.
Although the differences are striking, there also are a number
of notable similarities between chimpanzees and bonobos (on
the one hand) and the members of unsegmented foraging
societies (on the other) in the realm of relations between groups.
The potential for lethal intergroup violence is an ambient
condition of existence in both cases, and we can conclude that
this potentiality has been an integral contextual feature of
human (hominid) evolution from the beginning of the Paleolithic
period to the ethnographic present. Three responses to this
condition are noted in both cases: (i) avoidance, (ii) positive
engagement in friendly relations with neighboring local groups,
and (iii) aggression that may result in territorial gain or loss.
Friendly relations facilitate full exploitation of the zone along
territorial borders, whereas hostile relations lead to border
avoidance and underutilization of the food resources present in
this zone. Hostile relations are conditioned by resource competition
and also vary in intensity in accordance with the degree of
resource competition (other things being equal). The subjects
involved in acts of aggression against neighbors are nearly always
adult males who are the group members that most frequently
exploit food resources located in the zone along territorial
borders. Attacks are invariably opportunistic, in that the assessment
of a significant advantage (in numbers or in asymmetrical
detection or both) is always a precondition for initiating violence
that is lethal in intent. Nearly all attacks occur in close proximity
to territorial borders. Attacks on base camps where group
members sleep are nonexistent among chimpanzees and rare to
nonexistent on the part of unsegmented foraging societies.
Elaboration of the means of maintaining friendly relations and
the capacity for sharing access to resources with neighbors set
humans apart from chimpanzees and bonobos. Indeed, it might
be said that the members of unsegmented foraging societies
increase their fitness through ‘‘improved access to resources such
as food, females, or safety’’ (ref. 6, p. 12) by eschewing efforts to
achieve intercommunity dominance in favor of egalitarian relations
of friendship, mutuality, and sharing. This course of action
is a wise adaptive choice because dominance is characteristically
unattainable, and the only effective means of increasing territory
size is to fully utilize border zones. The capacity to maintain
friendly relations that allow for access to a neighbor’s territory
during lean years is particularly important in environments
where there are localized year-to-year fluctuations in resource
availability. These cooperative relations clearly played a key role
in the Upper Paleolithic expansion of human populations across
the globe into every environmental zone in which terrestrial
mammals of any kind are capable of existing (9). Intergroup
cooperation facilitates the rapid colonization of open environments.
Under these circumstances, territory size is not a relevant
variable and fitness does not correlate with it. Fitness instead
correlates with the social group’s reproductive rate, which is
primarily a function of the ease of obtaining reproductive
females from neighboring groups (based on a past history of
positive relations).
We may now turn to the archaeological record to ascertain the
earliest known use of projectile weapons that kill at a distance.
At present, that date has reliably been established as about
400,000 years ago. Wooden spears 1.82–2.60 m in length that
‘‘resemble modern javelins’’ have been recovered from a site
near the Scho¨ningen brown coal mine in Germany.
Found in association with stone tools and butchered
remains of more than 15 horses, the seven spears. . .
strongly suggest that systematic hunting, involving foresight,
planning, and appropriate technology, was part of
the behavioral repertoire of premodern hominids. The
use of sophisticated spears at these remote times necessitates
rewriting of many current theories on early
human behavior and culture.
Ref. 10, p. 8
There is considerable continuity in the stone tools manufactured
by Homo erectus during the period from 1,000,000 to
300,000 years ago. Temporal homogeneity in the stone component
of the tool kit provides grounds for suggesting that this same
continuity applied to wooden implements as well. Wooden
spears like those described above may then have been used 1
million years ago. Large game made up a significant component
of the diet during this period, and it is reasonable to assume that
spears were used in game procurement and in driving off
scavengers. Grahame Clark (11) provides a concise summary of
the diet and probable tool kit of the H. erectus inhabitants of
Zhoukoudian Cave in northern China, a site dated at 500,000–
350,000 B.P.
To judge from the animal remains associated with him,
Peking man depended largely on venison, since twothirds
of them belong to two species of deer, namely
Euryceros pachyosteus and Pseudaxis grayi. His victims
also included elephants, two kinds of rhinoceros, bison,
water buffaloes, horses, camels, wild boars, roebucks,
antelopes, and sheep, not to mention such carnivores as
saber-toothed tigers, leopards, cave bears, and a huge
hyena. How he managed to secure this varied selection
of game we can only speculate. No specialized projectileheads
have survived in the archaeological record, but to
judge from evidence from elsewhere, he would have had
available wooden spears with the tip hardened in fire,
and it seems likely in view of the character of some of his
victims that he would have used primitive pit traps.
Ref. 11, p. 27
The selective factors that favored coalitionary killing of neighbors
may have remained in play until as late as 1 million years
ago. The precise chronology of the persistence of these selective
factors during the Lower Paleolithic remains an open question
at present. However, the development of the throwing spear,
used in conjunction with ambush hunting techniques, ushered in
an era in which the enhanced lethality of weaponry amplified the
costs of assessment errors, and the necessity of movement also
placed intruders at a comparative disadvantage with respect to
both detection and assessment. Moreover, asymmetrical detection
rather than a numerical imbalance of power determined the
outcome of hostile encounters. This reconfiguration of the
decisive factors in lethal conflict not only raised the stakes (or
potential costs) for would-be aggressors but also rendered the
benefit of intercommunity dominance unattainable. Because
superior numbers were not invariably decisive in encounters
between hunting parties, an initial success would neither materially
reduce the stakes for aggressors in subsequent attacks nor
make it possible to freely encroach on the territory of a
neighboring group that had sustained a casualty. Under these
circumstances, aggression resulted in stalemate and a condition
analogous to a war of attrition rather than territorial gain.
These developments marked a major turning point in the
evolution of lethal intergroup violence and in the character of
interrelations between neighboring groups. Although fitness
continued to be related to territory size (for food-limited populations
in occupied environments), selective circumstances no
longer favored aggression as a means of achieving territorial
gain. Conflict avoidance and the development of intergroup
relations of friendship, mutuality, sharing, and cooperation were
favored instead. Intragroup cooperation was elaborated in conjunction
with the teamwork entailed by large game hunting and
was further reinforced by mechanisms for sharing large animals
jointly killed by a hunting party. Development of these practices
provided a template for establishing positive relations between
neighboring social groups that could readily be realized when
two bands temporarily camped and hunted together. Conflict
avoidance dictated that groups move apart to alleviate resource
competition whenever possible. This response to crowding facilitated
migration out of Africa into the temperate and subtropical
zones of Europe and Asia, vastly expanding the geographic
distribution of H. erectus. In other words, the
development of the throwing spear altered the means of production
as well as the social relations of production, distribution,
and consumption within groups in fundamental ways that also
transformed intergroup relations and influenced subsequent
hominid evolution.
This period of Paleolithic warlessness, grounded in low population
density, an appreciation of the benefits of positive
relations with neighbors, and a healthy respect for their defensive
capabilities, lasted until the cultural development of segmental
forms of organization engendered the origin of war (7). This
organizational transformation facilitated the mobilization of all
adult male group members and their participation in preplanned
dawn raids on settlements in which the tactical advantages of
surprise and numerical superiority could be brought to bear. At
this juncture, the unit involved in combat is a raiding party (a
military organization) rather than a hunting party (an economic
organization), and the location of combat shifts from the border
zone to the sleeping quarters at the core of a group’s territory.
At the same time, the intrinsic military advantage shifts from
defenders to attackers. All of the attackers are combatants,
whereas less than half of those under attack are armed. Attackers
characteristically inflict numerous casualties while suffering few
or none. This outcome is a consequence of weaponry amplifying
the advantages of surprise and numerical superiority.
The earliest conclusive archaeological evidence for attacks on
settlements is a Nubian cemetery (site 117) near the present-day
town of Jebel Sahaba in the Sudan dated at 12,000–14,000 B.P.
(7, 12). War originated independently in other parts of the world
at dates as late as 4,000 B.P. (13). Otterbein argues that
agriculture was only able to develop initially at locations where
ambushes, battles, and raids were absent (14).
The evolution of lethal intergroup violence thus encompasses
three major periods: (i) the era of coalitionary killing, (ii) the era
of intrinsic defensive advantage, and (iii) the era of war. An
advance in weapons technology (the javelin-like throwing spear)
engenders the first transition, whereas an advance in military
organization and tactics produces the second. The decisive
significance of these factors is expectable in light of what we
know from recorded military history. However, the duration of
the first two eras extends over hundreds of thousands of years.
The protracted character of these eras is consistent with the slow
pace of technological and organizational change during the
Paleolithic period.
The main objective of this paper is to specify the major turning
points in the evolution of lethal intergroup violence, to delineate
the periodization that results from this specification, and to
broadly characterize each era. The dating of these periods can be
expected to be refined as additional archaeological data come to
light. Moreover, the key transitions did not occur simultaneously
in every world area so that regional, rather than global, chronologies
are required, especially with respect to the origin of war.
I thank Kent Flannery for assistance in locating archaeological sources
and Joyce Marcus for her useful comments on an earlier draft.

Notes

#E-mail: rck@umich.edu.
†This point can be further elaborated. It is possible that adding females without increasing
territory could augment fitness, but this fitness enhancement would transpire only if food
resources were more than adequate. However, adding territory would not augment
fitness under such conditions. In other words, Wrangham’s argument presupposes that
chimpanzees are a food-limited (rather than disease-limited) population, because intercommunity
dominance would otherwise convey no selective advantage. See ref. 15 for a
further discussion of the interrelationship between the food-limited status of a population
and the adaptive significance of territorial gain.

‡Unsegmented societies are defined by eight organizational criteria (7). In brief, the social
organization of unsegmented societies is limited to the bare minimum: these societies
manifest only those social groups that are cultural universals (present in every society) and
nothing beyond this minimal set. Local groups are comprised of independent families
(including a man, one or more spouses, and their children). These communities are
autonomous, i.e., there is no level of organization beyond the local group. Egocentric
bilateral kin networks or kindreds link individuals to relatives in other communities. Thus,
family, local group, kindred, neighborhood, and language group (‘‘one-talks’’) are
present. Descent groups, characteristic of segmental organization, are absent.

§The armed conflict that occurs between hunting parties in unsegmented societies differs
from warfare in that only the perpetrator of an offense is targeted. Likewise, a homicide
in an unsegmented society may result in an attempt to execute the perpetrator. Capital
punishment and warfare share a number of features (ref. 7, p. 7) but differ in one critical
respect: in war, social substitution governs the targeting of individuals for lethal violence.
These two forms of lethal violence can readily be distinguished by a simple example
illustrating the choices that confront a member of a typical hunting and gathering band
composed of a half-dozen related families. If a member of a neighboring group kills an
individual’s sister, there are two alternative retaliatory responses: the aggrieved brother
of the woman may kill the killer, or he may kill the killer’s sister (or parent or child or any
member of the killer’s social group). In unsegmented foraging societies, only the first
course of action is considered morally justified. The second course of action would not be
considered to resolve the situation in any way and indeed would constitute an independent
event. Although some kinsmen and coresidents might potentially assist an individual
in carrying out a capital punishment execution, they would not assist in the second course
of action nor dependably come to the defense of the perpetrator of a homicide. The
transition from capital punishment to war thus requires development of the cultural
concepts of injury to the group and group responsibility for the inflection of injury (7).
These concepts render any member of the killer’s collectivity a legitimate target for blood
vengeance and, at the same time, mobilize a vengeance party.
The development of segmental forms of organization leads to revenge-based raiding
initiated in response to an initial homicide stemming from a commonplace interpersonal
conflict. Such conflicts are characteristically unrelated to resource scarcity and are not
density dependent. As a result, the correlation between the frequency of lethal intergroup
violence and the degree of resource scarcity that obtains among unsegmented societies is
not applicable to segmental societies (7).
¶In unsegmented societies, it is not uncommon for living parents to give a child to kin or
friends of another local group to be adopted and raised by them. Among the Andamanese,
nearly all children older than 7 years of age are raised by adoptive parents (7). This
cultural practice is obviously conducive to mutual nonaggression between local groups
linked by such child transfers.
This generalization is based on detailed analysis (7) of the eight unsegmented foraging
societies in Ross’s (16) sample of 93 societies [which is a half sample of Murdock’s and
White’s (17) Standard Cross-Cultural Sample]. Attacks on settlements per se are not
reported for the Mbuti, Semang, Copper Eskimo, !Kung, Yahgan, and Slave. The Slave were
subject to such attacks by the segmental Cree but did not carry them out. The Ingalik were
likewise subject to raids on their settlements by their segmental neighbors (the Koyukon
and Kolchan) in the 1840s, and they responded in kind. However the Ingalik did not carry
out raids against the settlements of fellow tribesmen or those of their unsegmented
neighbors, the Kuskowagamuit. Among the !Kung and Yahgan, the family of a homicide
victim may go to the settlement of the perpetrator and attempt a capital punishment
execution, but only the perpetrator is targeted, so this endeavor does not constitute an
attack on a settlement per se. Attacks on settlements are reported for the Andamanese,
who represent a transitional case with respect to the origin of war. However, no such
attacks are recorded in colonial records that cover the period from 1863 to 1899. These
records contain detailed accounts of numerous border conflicts during this early contact
period. In all, these data indicate that, for unsegmented societies, attacks on settlements
may be characterized as rare to nonexistent. Events that occur less than once a generation
are considered rare in the comparative cross-cultural study of lethal violence. Such events
are generally not within the life experience of the members of these societies.

References
1. Goodall, J., Bandora, A., Bergmann, E., Busse, C., Matama, H., Mpongo, E.,
Pierce, A. & Riss, D. (1979) in The Great Apes, eds. Hamburg, D. A. &
McCown, E. R. (BenjaminCummings, Menlo Park, CA), pp. 403–420.
2. Nishida, T. (1979) in The Great Apes, eds. Hamburg, D. A. & McCown, E. R.
(BenjaminCummings, Menlo Park, CA), pp. 73–121.
3. Nishida, T., Haraiwa-Hasegawa, M. & Takahata, Y. (1985) Z. Tierpsychol. 67,
284–301.
4. Goodall, J. (1986) The Chimpanzees of Gombe: Patterns of Behavior (Harvard
Univ. Press, Cambridge, MA).
5. Wrangham, R. & Peterson, D. (1996) Demonic Males: Apes and the Origins of
Human Violence (Houghton Mifflin, Boston).
6. Wrangham, R. (1999) Yearb. Phys. Anthropol. 42, 1–30.
7. Kelly, R. C. (2000) Warless Societies and the Origin of War (Univ. of Michigan
Press, Ann Arbor).
8. Portman, M. V. (1899) A History of Our Relations with the Andamanese (Office
of the Superintendent of Government Printing, Calcutta).
9. Whallon, R. (1989) in The Human Revolution: Behavioral and Biological
Perspectives on the Origins of Modern Humans, eds. Mellars, P. & Stringer, C. B.
(Princeton Univ. Press, Princeton), Vol. 1, pp. 433–454.
10. Thieme, H. (1997) Past 26, 7–8.
11. Clark, G. (1977) World Prehistory (Cambridge Univ. Press, Cambridge,
U.K.).
12. Wendorf, F. (1968) Prehistory of Nubia (Southern Methodist Univ. Press,
Dallas) Vol. 2, pp. 954–995.
13. Flannery, K. V. & Marcus, J. (2003) Proc. Natl. Acad. Sci. USA 100, 11801–
11805.
14. Otterbein, K. F. (2004) How War Began (Texas A&M Univ. Press, College
Station).
15. Kelly, R. C. (1985) The Nuer Conquest (Univ. of Michigan Press, Ann
Arbor).
16. Ross, M. H. (1983) Ethnology 22, 169–192.
17. Murdock, G. P. & White, D. R. (1969) Ethnology 8, 329–369.