Chapter 3 – Emotional underpinnings of war: An evolutionary analysis of anger and hatred
Aaron N. Sell
Assistant Professor of Psychology and Criminology
Anthony C. Lopez
Associate Professor of Political Psychology
Washington State University
Citation: Forthcoming, in Ireland, C., Ireland J., Lewis, M. & Lopez A. (Eds.) The International
Handbook on Collective Violence: Current Issues and Perspectives, Routledge.
Warfare is an instance of collective action in which members of one group impose physical damage
upon one or more members of another group (Pietraszewski, 2016; Glowacki, Wilson and
Wrangham, 2017; Lopez, 2019). Although the stated rationales for warfare are myriad and not
mutually exclusive, ranging from sex and sorcery, to revenge and resources (Keeley, 1996; Gat,
2006), these goals are often facilitated by one or a combination of emotions that are designed to
organise motivation, reasoning, and behaviour in ways that would have been ancestrally adaptive.
Although many emotions likely facilitate political violence generally and warfare specifically, we
focus on two in particular: anger and hatred. These two emotions predictably and distinctly shape
the character of collective violence at multiple stages of fighting, including: the triggers of
aggression, the selection of targets, the pace and character of escalation and brinkmanship, and the
possibility of surrender and reconciliation. Since large-scale social phenomena (such as warfare)
must rest upon ecologically valid microfoundations of individual psychology, we begin by
elaborating the adaptationist approach to understanding the evolved function of anger and hatred.
Once the contours of this framework are outlined, we describe some of the ways in which this
psychology is manifest in contexts of modern warfare.
Evolution and the adaptationist program
Life on Earth is the product of natural selection (Darwin 1859; Dawkins 1996). The differential
reproductive success of different genes over long periods of time led to organisms acquiring
functional design that aided the replication of their genes in ancestral environments. Put
unemotionally, animals are “gene machines” – highly ordered mechanisms composed of chemical
matter (mostly carbon and water) that inherited their current form because their ancestors had
mutations that increased their probability of reproducing their genetic code (Dawkins 1976).
Taking evolution seriously led biologists to create a program for understanding animal design and
behaviour – the adaptationist program (Williams 1966). This program involves matching evolved
design (e.g. fur) with the way in which it aided reproduction for that animal’s ancestors (e.g.
conserving body heat). This initial match is treated as a hypothesis and is supported only if multiple
independent features of the “adaptation” are improbably well-designed for performing that
function. For example, the fur of aquatic mammals is improbably well tailored to underwater
thermal regulation – the hairs are shorter, denser, and flatter in ways that enable air to be trapped
in the fur and act as a buffer to the colder water (Liwanag et al. 2012). The more features of the
adaptation that are consistent with the posited function, the more confident we can be that the
adaptation evolved for that function.
Evolutionary psychologists use this adaptationist program to discover the evolved function of
human cognitive systems (Buss 2005; Cosmides & Tooby 1994, 1995, 2000; Tooby & Cosmides
1989). They can then use those functions to “map the mind” – that is, to identify the myriad
components of human nature, such as: the way the mind identifies kin (Lieberman, Tooby &
Cosmides 2007), the function of emotions such as pride (Sznycer et al. 2017) and forgiveness
(McCullough 2008), as well as coalitional dynamics such as our attentiveness to and evaluation of
leaders (Antonakis and Dalgas, 2009; Garfield, von Rueden and Hagen, 2019).
The recalibrational theory and the evolved function of anger
Anger is a universal human emotion, experienced in a similar fashion across cultures and
individuals (Ekman 1973; Sell 2011; Sell et al. 2017). It bares the hallmarks of being an evolved
adaptation, including early ontogenetic development (Galati et al. 2003), universal design (Ekman
1973; Scherer & Wallbott 1994; Sell et al. 2017), and most importantly functional design (Sell et
al. 2017). To understand what anger is and why it works the way it does, we have to understand
why anger evolved, i.e. why did human ancestors who experienced anger the way that we do
outreproduce those who inherited different variants of anger? Framed differently, what was the
problem that anger solved for our ancestors?
The recalibrational theory of anger (Sell 2011; Sell, Tooby & Cosmides 2009; Sell et al. 2017)
posits that the evolved function of anger is to bargain for better treatment. When successful, anger
“recalibrates” the target so that they put greater concern on the angry individual’s welfare in
current and future decisions. For example, a man is rude to his colleague, the colleague becomes
angry at him and confronts him about his behaviour. His rudeness is illustrated to him, and he
apologises, recalibrates, and treats his colleague better in the future. In this way, angry individuals
in our evolutionary past protected their own interests and outreproduced those who tolerated poor
treatment from others.
To understand this recalibration process, we must understand what is actually being recalibrated.
What part of the mind is changed when a person concedes a fight and apologises? According to
the recalibrational theory, anger has its effects by upregulating the “welfare tradeoff ratio” toward
the angry individual. Welfare tradeoff ratios are hypothetical mental barometers that store the
extent to which we will value another individual’s welfare compared to our own when making
decisions that affect us both (Tooby et al. 2008). For example, a student asks their professor to
look over a paper they have due in another class. Whether he helps or not depends on how much
he values the student’s welfare (i.e. his welfare tradeoff ratio (WTR) toward the student). Conflicts
of interest thus indicate how much an individual values another’s welfare. For example, if a hungry
man skips dinner to comfort a friend, it indicates a high value WTR toward that friend. If a wealthy
individual refuses a hungry orphan a slice of bread, it indicates a low WTR toward the child (see
Sell et al. 2017).
Of course, setting a welfare tradeoff ratio above zero (i.e. caring about another’s welfare at all)
comes at a cost to that individual, by definition. Evolution would – generally – design animals to
put no weight on the welfare of others. But there are known exceptions to this wherein an organism
can reproduce more by accepting temporary costs (Nowak 2006). Most relevant to anger, animals
will frequently accept costs and forgo benefits when competing with more formidable animals, i.e.
animals that can impose greater costs typically via aggression. Because the costs of punishment
can be greater than the costs of deferring to the more dominant animal (Clutton-Brock & Parker
1995) weaker animals will (in general) evolve to cede resources to stronger animals. This is the
source of dominance hierarchies (Archer 1988).
Thus, welfare tradeoff ratios appear calibrated by relative formidability such that those with greater
fighting ability expect and demand better treatment (Lukaszewski et al. 2016; Sell, Tooby &
Cosmides 2009). The fact that welfare tradeoff ratios are calibrated by formidability offers anger
an opportunity to fulfil its function - if WTRs respond to estimates of formidability and the
willingness to impose costs, then anger can demonstrate formidability and the willingness to
impose costs as a means of recalibrating a low WTR in the target of anger (i.e. the person one is
angry at). This, as we argue below, typifies anger-based aggression.
In sum, the evolved function of anger is to recalibrate the status-setting mental mechanisms of
individuals who indicate that they value your interests insufficiently. Its evolved purpose is to
recalibrate the target, and restore acceptable treatment. This is not the function of hatred (see Table
The neutralization theory of hatred
An analysis of hatred indicates contrasting design to anger. Rather than attempting to recalibrate
a target and restore deference and cooperation, hatred appears designed to limit or eliminate a
target whose existence is damaging to the hateful person. We will refer to such a person as an
“enemy”. Our definition deviates somewhat from the dictionary definition of “enemy” which
implies mutual hostility between the parties. Our definition of enemy is specifically a person
whose existence is damaging to the hateful person, and can include someone who is not hostile in
return (e.g. a romantic rival who nonetheless like you can still be your enemy).
An enemy can be precisely defined as an individual for whom one has a negative association value.
An association value is the net fitness consequences one individual has on another. High
association values are typically held by friends and family members who improve your welfare by
monitoring social environments, sharing rare resources, dispensing knowledge, combining labour,
dispersing risk, and so forth. An enemy is one who has a negative association value: their existence
is (on average) bad for you. The existence of enemies in our ancestral past would have selected for
a cognitive adaptation, a part of human nature, designed to deal with their existence in ways that
increased our ancestors’ reproduction. Just as with anger, the design of the emotion hatred should
match the nature of the problem it evolved to address. What could our ancestors do to minimise
the consequences of their enemies? By hypothesis, we propose that a general solution evolved to
deal with enemies, which is to heap costs upon them. These increasing costs could have had up to
three beneficial effects for the hateful person:
1) the target can respond to their worsening conditions by removing themselves (physically or
socially) from the hateful person and thus minimise their interactions and the opportunities to
impose costs on the hateful person,
2) the target can be rendered less powerful (via physical, reputational, sexual, or other mediums)
and thus less able to impose costs on the hateful person, or
3) the target can die, rendering them incapable of actively imposing costs on the hateful person.
How are these cost impositions accomplished? We argue in this chapter that hatred calibrates the
hateful person’s welfare tradeoff ratio toward the target to be negative, i.e. costs to the hated person
are treated as benefits to the hateful person. As discussed with anger, welfare tradeoff ratios
function to make tradeoffs during conflicts of interest. For example, the higher a welfare tradeoff
ratio Bob has toward Nancy, the more Bob will accept costs to benefit her, and the less Bob will
impose costs on Nancy to benefit himself. If Bob hates Nancy (i.e. adopts a negative welfare
tradeoff ratio) he will treat any benefit to Nancy as a reason not to do the action, and will treat a
cost to Nancy as a reason to do it. In other words, negative welfare tradeoff ratios sanction spiteful
acts (e.g. hurting yourself just to hurt the other person, and denying yourself a benefit as long as it
hurts the other person). The magnitude of the welfare tradeoff ratio indicates the extent to which
they will accept costs to hurt the hated other. For example, an extremely negative WTR was
exemplified by Captain Ahab at the end of Moby Dick [a classic book detailing an obsessive quest
of Captain Ahab for revenge on Moby Dick, the giant white sperm whale that had previously bit
off his leg at the knee]. Here Captain Ahab gives all he has (his last breath - a precious thing) to
spit at the whale (a trivial cost to the hated other).
Highly negative WTRs presumably have many effects similar to highly positive WTRs; e.g.
person-specific memory is probably enhanced such that individuals remember information about
both their loved friends and hated enemies. On the other hand, some effects are likely opposite;
e.g., while a high positive WTR engages empathy such that the suffering of one’s loved child is
“felt” as suffering to the parent, a highly negative WTR reverses the polarities such that the
suffering of one’s hated enemy is felt as enjoyment to the hateful person.
In addition to sanctioning actual cost infliction, hatred should be designed to deploy social forces
against the target. In short, hatred is predicted to motivate the spread of information detrimental
to the well-being and social status of the target of hatred. This serves both to lessen the hated
person’s status but also recruit allies for group-based aggression. Interestingly, there is no
particular need for the hateful person to be honest when they spread gossip and rumours about the
target of hatred. This is unlike anger-based arguments, which are often rebutted by the target and
which function primarily in cooperative relationships, wherein truth-telling between partners is of
value to both participants (i.e. lying to win an argument with your spouse will likely have
detrimental repercussions later). Furthermore, anger-based arguments even outside of cooperative
relationships can risk costly escalation when one party is found to be lying. However, in hatred-
based information warfare the target’s association value is negative so there is little danger in
sacrificing future cooperation, and escalations are less concerning because the target is already an
enemy. In short, we should expect a hunger for negative information about a hated target and no
substantial concern about the truth of that information. In fact, insistence on the truth of a charge
against a hated individual will likely imply a lack of hatred that can be disconcerting (e.g. someone
standing up for Hitler against claims that he was an animal-abuser may sound insufficiently anti-
Nazi even though he is innocent of those charges). In total, the negative WTR that is adopted under
hatred leads the hateful individual to damage the welfare of the target by denying them benefits
and actively imposing costs. It functions to neutralise the enemy and reduce their power rather
than recalibrate them and restore cooperative treatment.
Triggers of anger and hatred
Anger and hatred have distinct functions, and therefore, they are predicted to have distinct triggers.
Anger should be deployed at individuals who indicate that their WTR toward the angry person is
lower than some threshold of acceptability. This threshold is presumably set by factors that relate
to bargaining power; e.g. stronger men expect higher WTRs from others (Sell, Tooby & Cosmides
2009) and deploy anger over a greater range of offences, as do juvenile boys and girls (Sell, Eisner
& Ribeaud 2016). While we focus on formidability-based WTRs in this chapter, it is important to
note that the cooperative nature of human relationships also means that bargaining power and
WTRs can be set by many other factors, e.g. the mutual regard that spouses have for one another
is rarely based on their ability to physically harm each other. Another example: more physically
attractive women (and to a lesser extent men) have also been shown to demand higher WTRs and
deploy anger over a greater range of offences (Sell, Tooby & Cosmides 2009; Sell, Eisner &
Indications of one’s WTR can come from many channels. In fact, argument analyses suggest at
least twelve computationally distinct triggers of anger (Sell 2014). The most common is a cost
infliction; specifically, the target of anger imposes a cost or refuses a benefit to the angry person,
e.g. a roommate refuses to pick you up from class in their car, or a colleague takes your headphones
without asking (see Sell et al. 2017). Others can be symbolic or material indicators that correlate
with less respect (i.e. a lower WTR). For example, insults that imply that one is unworthy of value
(e.g. that the target is unattractive or unintelligent or emits foul odours), or failure to remember
information about the individual (e.g. forgetting an anniversary), and so on (see Sell et al. 2017).
In each case, anger is triggered because of information that implies the target of anger holds a low
WTR toward the angry person.
Hatred, on the other hand, responds to cues of a negative association value; i.e. when the totality
of another’s impact on your life is negative, or colloquially, your life would be better off without
them. Commonly, such a person is hated because they place little value on your welfare and thus
anger and hatred can both be active. For example, a bully who treats you disrespectfully at work
can be both the subject of your anger – as a bargaining strategy to recalibrate their WTR toward
you - and hatred - as a strategy for getting rid of them from your social world. If the anger is
successful and the bully apologises and ends their mistreatment, then hatred too may switch off
because the bully is not likely to negatively impact your life in the future.
In most cases of anger, however, hatred is not active. This is known because most cases of anger
occur within the context of valued relationships –often characterised by love (Averill 1982). In
these cases, anger functions to bargain for better treatment, but the association value with the target
of anger is still positive. Hatred is caused by a negative association value, which is an average of
another’s impact on your welfare, e.g. someone who abuses your child but gets you a great deal
on car insurance is still an enemy. This means that hatred can be triggered even by multiple
justifiable impositions, while anger may not; e.g., a needy and incompetent colleague who
constantly requires help and never seems to improve. Each individual imposition may be justified:
after all, you are better able to solve the problem than they are - but taken together your life is
made worse for their existence. As noted by Sell (2012), this could explain the abuse of elders in
assisted living conditions, some of whom require costly and repeated help to maintain their lives
(Lachs & Pillemer 2004).
Hatred evolved because it increased our ancestors’ reproduction by preventing future harms (i.e.
there is no logical reason to punish past harms unless it impacts the future). This means that if a
person with an innocuous history is likely to produce harms sufficient to turn their association
value negative in the future, hatred will likely be triggered. This is one explanation for some
classic work in social psychology showing that individuals will dislike their own victims, i.e. by
hurting someone unjustly you have made an enemy because their future behaviour toward you will
likely be negative (Schopler & Compere 1971).
Aggression borne of anger and hatred
The recalibrational theory of anger states that the primary evolved function of anger is to
recalibrate the target’s WTR. This is because, for some human relationships, the WTR is calibrated
by estimates of formidability, the anger system should be designed to deploy demonstrations of
fighting ability and willingness to aggress as a low-cost means of winning the conflict and
recalibrating the target. There are abundant animal precedents for this kind of aggression, called
“ritualized aggression” “conflicts of assessment” or “sequential assessment” (Alcock 2009; Archer
1988; Sell 2011). From an evolutionary biological point of view, ritualized aggression is an
interaction wherein two animals attempt to assess each other’s fighting ability so as to avoid the
costs of a real fight. The early stages of ritualized aggression are typically perceptual evaluations
during which animals flex and enhance their cues of fighting ability (e.g. baring fangs,
piloerections). For example, a breed of cichlid (a fish) will engage in staring contests and parallel
swimming during which they evaluate the fighting ability of their opponent (Enquist & Leimar
1983). If one of the fish indicates submission, the aggression ends. If not, the fight can escalate
to more accurate but dangerous indicators of fighting ability (e.g. the fish can lock jaws with one
another and shake, sometimes damaging their mouths, a process called “mouth-wrestling”). Only
in the later stages of assessment are these bouts potentially lethal. These contests evolved from
mutualism: both contestants benefit by not paying the full costs of aggression, which likely
involves injury to both the winner and loser.
Anger-based aggression is strikingly similar to these contests. A typical bout of anger-based
aggression involves assessment, posturing, and the deployment of the universal “anger face” (Sell
2011). The anger face itself is a relatively recent evolutionary adaptation – distinct from
chimpanzee aggressive faces and part of evolved human nature (e.g. it is present in blind children,
see Sell, Cosmides & Tooby 2014). It is composed of at least seven distinct facial muscle
movements that independently increase the features of the fact that distinguish physically strong
from physically weak men (Sell 2014). For example, stronger men typically have wider nostrils,
a lower browridge, and thinner lips, and so the anger face flares the nostrils, pushes flesh over the
browridge in ways that make it appear larger, and bends the lips inward to shrink their apparent
size (Sell Cosmides & Tooby 2014, see also Marsh, Adams & Kleck 2005). This face evolved
because those who deployed it were able to intimidate their opponents and achieve recalibration
earlier and at lower cost than those who deployed different faces. Similar processes presumably
explain the human posturing that takes place in the early stages of anger-based aggression (e.g.
chest inflated, legs together to maximise height, head tilted up to give the illusion of height,
When posturing does not resolve a conflict, it can escalate to actual aggression, though typically
sublethal aggression. Pushing and shoving, for example, is extremely unlikely to incapacitate a
target, but can indicate relative strength. These escalating patterns to aggression are extremely
common in human fighting (Luckenbill 1977; Daly & Wilson 1988) and appear to be a feature
particular to bargaining aggression (Sell 2011). Some cultures have formalised these rituals in
ways that remove the ambiguity of when to escalate, e.g. see Chagnon’s descriptions of Yanamamo
fighting (1983), or William’s description of Southern dueling (1980). Regardless of these formal
examples, there appears to be an instinct for the rules of assessment-based combat such that
“cheating” invalidates the outcome (Romero, Pham & Goetz 2014). The concept of “cheating” at
aggression is sensible only if there are implicit rules to combat. These rules mean that aggression
is designed to do something other than simply destroy one’s opponent. In the case of anger-based
aggression, a “cheap” or “cheating” move is one that reduces the accuracy of the bout such that
the putative winner may not be the most formidable (e.g. a “sucker punch,” challenging someone
when they are already injured, and so on). These escalating fights work, when they work, by
demonstrating one’s formidability to the target in ways that lead them to upregulate their WTR.
Aggression deployed by hatred is predicted to be simpler than anger-based aggression. While
anger-based aggression is akin to bargaining aggression in other animals, hatred-based aggression
should look more like predatory aggression. While a predator hunts, it is not attempting to
recalibrate territorial boundaries or change the dominance hierarchies stored in the prey animal: it
is efficiently attempting to injure or kill the prey. Hatred-based aggression is similarly functional:
it is designed to impose damage in a cost-effective way on the target. This predicts several
discrepancies with anger-based aggression. Anger-based aggression is typically “loud” with
attention-grabbing vocalizations, facial displays, posturing, yelling, and low-cost demonstrations
of formidability like hitting inanimate objects. Hatred-based aggression is predicted to be
relatively quiet, like a predator stalking prey. Signaling the aggressive intent is a crucial part of
anger-based bargaining aggression because it allows the target to recalibrate without having to pay
the mutual costs of aggression and it sets the stage for a “fair” fight to determine who is more
formidable. Signaling is a mistake in hatred-based predatory aggression because it allows the
target to prepare to defend themselves.
As hatred is not trying to recalibrate the target, it does not need to gather the attention of the target
or signal information to them. Rather, information (true information at least) should be denied the
hated person so as to render them less competent. For this reason, there does not appear to be a
“hate” face in the same way that there is an anger face.
For example, when Leon Gary Plauche
shot the man who was accused of abusing his son - a crime caught on camera - his face appeared
plain and unemotional. The function of the anger face is to bargain with the target by exaggerating
cues of fighting ability in the face, but hatred does not need to interface with the cognitive
mechanisms in the target or bargain for better treatment. It is designed to hurt the target
surreptitiously, if possible.
Hatred-based aggression is also predicted to be shorter in duration, much as predatory aggression
is in the wild. While anger-based aggression involves prolonged escalating aggression with amble
time for either party to recalibrate and concede, hatred-based aggression is not designed to allow
recalibration because the problem isn’t a lack of WTR calibration but is rather the existence and
Combining an anger face with a smile will sometimes produce a sadistic face - as if someone is aggressing against
another and is happy about it. We consider this face and its connection to hatred an interesting source of future
power of the opponent. For this reason, hatred-based aggression should not escalate in a given
bout but be at maximum from the start and rapidly deployed without signal (again, Leon Gary
Plaunche’s assassination of his enemy is a clear example of this).
Anger and hatred-based aggression respond differently to signs of weakness in the target.
Bargaining aggression often conforms to implicit or explicit rules about allowing a defeated
opponent to retreat. These limitations, like the ritualised aggression of other animals, allow for
accurate estimates of formidability so as to trigger recalibration. Yet, hatred is functioning to
inflict damage on the target, not to recalibrate it, and so any temporary weakness or vulnerability
in the target is an opportunity to inflict costs with less retaliatory risks. As such, hatred should
motivate increased aggression toward an enemy that signals weakness, an unwillingness to fight,
or other indicators that cost-infliction will not lead to retaliation. The instinct to crush an enemy
when they indicate weakness has been described under the term “forward panic” (Collins 2008).
Ending conditions of hatred and anger
Emotions are temporary modes of operation (Cosmides & Tooby 2000). As such, they have
endpoint conditions that shut them off. Two such endpoint conditions have been posited for anger:
1) a contradiction of the inputs, i.e. you realise you are angry when you shouldn’t be, 2) anger
fulfils its function and the target recalibrates, i.e. the target apologises and you aren’t angry
anymore (see Sell 2011). The first can occur during anger-based arguments when individuals
present evidence and reframe social conflict in ways that change the angry person’s estimate of
their WTR. Indeed, research shows that targets of anger will select arguments that indicate that
There may be an exception to the rapid deployment of maximum aggression in the form of prolonged torture
when the hateful person is protected from retaliation. This would presumably be an evolutionary byproduct (i.e. a
functionless accident of design, like the bellybutton or the fact that humans are flammable), stemming from the
fact that hateful people experience vicarious enjoyment at their enemy’s suffering.
their actions were consistent with a high WTR; e.g. “it didn’t hurt you that bad, and I had a good
reason…” or that their relative bargaining power is in their favour, “Fine…but who is going to
loan you money next week when your rent is due?” If the target of the anger’s arguments work
and show that their behaviour was consistent with a high WTR, or that the angry person was
demanding too much deference, then the episode will be terminated. The other endpoint condition
of anger occurs when the target actually recalibrates and anger fulfils its function. Research shows
conclusively that anger can, in many circumstances, be switched off by a sincere verbal apology
even in the absence of any tangible recompense (Frantz & Bennigson 2005). This is consistent
with its function to recalibrate WTRs, i.e. the point is to change the target’s mind.
For hatred, a logical analysis of its function suggests two means of switching off hatred. The first,
like with anger, is evidence counteracting its triggers; i.e. discovering that you hate someone who
actually has a positive association value. Put colloquially, love can trump hate. If a hateful person
discovers that their enemy had been responsible for naming them for an award, or donating
anonymously to their child’s healthcare fund, or merely having said some nice words about the
hateful person in a conversation that was thought to be private, then the predicted association value
can become positive and hatred terminated. Whether this strategy for terminating an incident of
hatred works depends, presumably, on the magnitude of the negative association value (i.e. were
they hated a lot or a little). The second way to end an incident of hatred is for it to fulfil its function
and weaken, ostracise, or kill the target, at which point further hatred would be a waste of effort.
While this prediction can be logically derived from our posited function of hatred, our intuition is
that it may not be correct. Humans have a peculiar tendency to experience social emotions toward
the dead, e.g. gratitude, anger, hatred, love. Nonetheless, when a hated target is depowered, this
can satiate hatred if it removes the negative externalities caused by the hated person; e.g. a boss
who is hated by the employees is demoted to a harmless entry-level position and then pitied.
Strategy preferences - who gets angry and who becomes hateful?
When an individual is the victim of a cost, they have evidence that the offender has a low WTR
toward them (i.e. the offender was willing to impose that cost) but also that the offender may have
a negative association value (i.e. their life is now worse off because of the cost imposed by the
offender). This means that anger and hatred may fire simultaneously, which we suspect is a
common occurrence. That said, some individuals may be better able to bargain directly with the
target via anger and other individuals may be better able to neutralise negative externalities by
surreptitiously imposing costs on the target. If these asymmetries existed in our evolutionary past,
then we should expect modern humans to calibrate their retaliatory strategies to the ancestral cues
that determined which strategy was best given their circumstances. For example, the efficacy of
anger depends primarily on bargaining power. Those with more power, can recalibrate others by
demonstrating it. Those with less will find their attempts at bargaining frequently fail. Thus,
research shows that individuals with more bargaining power (e.g. coalitional strength, individual
fighting ability, and sexual attractiveness) are, on average, more prone to anger and aggressive
bargaining than others (Sell, Tooby & Cosmides 2009; Sell, Eisner & Ribeaud 2016). Hatred may
be an emotion more common among the weak and others who cannot directly bargain for better
treatment. For example, Mackie, Devos and Smith (2000) find that appraisals of in-group strength
(in contrast to weakness) elicited anger (instead of fear), as well as greater willingness to confront
and attack a threatening out-group.
Key points and summary
The features of anger indicate its evolved function. Anger is a temporary state activated primarily
by cues that indicate that someone holds a low welfare tradeoff ratio toward the angry person.
Anger then generates behavioural patterns that gather the target’s attention and feeds them
information relevant for the calibration of WTRs. The early stages of this bargaining are typically
verbal (Sell 2012), and involve information exchange about the events that triggered anger (e.g.
“you were being so rude!”) and interrogations of the target of anger to determine if the WTR
apparent in their actions truly represents their actual concern for the angry person’s welfare (e.g.
“what were you thinking?” “do you know who I am?” “do you even know how mean you’ve
been?”). If recalibration is not evident, anger can orchestrate bargaining tactics designed to
recalibrate the target’s WTRs. In the case of aggressive bargaining, this is done by enhancing cues
of formidability in the body and face, and then engaging in an escalating pattern of aggression
designed to prove to the target that they would be better off recalibrating than continuing to suffer
from the aggression.
The primary purpose of anger is to recalibrate the target and restore compliance and/or
cooperation. This is why it is most commonly deployed in the context of cooperative, even loving,
relationships (Averil 1982). The same is not true of hatred. Hatred is an orientation toward an
enemy, defined as an individual with a negative association value. Association value is a summary
variable of the average predicted future consequences of that person’s existence on one’s welfare.
If John has a negative association value for Daryl, it means that Daryl’s life would be better without
John. Such a circumstance would have recurred frequently among our ancestors, and as a result,
hatred appears to have evolved to navigate this problem. The primary way hatred does that is to
set a negative welfare tradeoff ratio toward the hated other; i.e. to positively value costs to that
person. Hatred then generates thoughts, fantasies, and plans of hurting that individual, often in
ways that prevent detection, and remains vigilant for circumstances that enable cheap, efficient
means of damaging the enemy. Any aggression generated by hatred will be akin to predatory
aggression in animals: quick, highly damaging, and deployed using every available advantage.
The primary purpose of hatred is to neutralise the enemy.
Table 1: Functional differences between anger and hatred
People who lack enough concern
for your welfare; e.g. disrespect
People whose existence is bad for you
Recalibrate target, restoring
compliance and/or cooperation
Diminish negative externalities by
depowering or eliminating the target
Cues of low WTR
Cues of negative association value
Arguments designed to
recalibrate the target of anger
Surreptitious recruitment of allies
Harm the target’s reputation
Attention gathering displays
displays (e.g. anger face/posture)
Bargaining (i.e. low-cost
demonstrations, followed by
rule-based escalation if
Predatory. High-damage timed to
when the target is vulnerable
1. Realize target has high WTR
2. Target recalibration (i.e.
1. Association value becomes positive
2. Target neutralised
Preferred strategy for those with
high relative bargaining power
Preferred strategy for those with low
relative bargaining power
Anger, hate, and warfare
Warfare can be defined simply as collective action with the purpose of hurting members of other
groups, and its occurrence varies extensively across time and space in terms of motivations,
instruments and scale. The application of anger and hatred to warfare relates to questions of
motivation specifically, and we leave as an empirical question the issue of whether and how
variation in the instruments and scale of war might affect the expression of these emotions. For
example, some have argued that democracy relative to autocracy may be unusually susceptible to
jingoism and vengeance (Levy 1988). Others have argued that a capacity for hatred and revenge
is what makes possible the irrational commitment to nuclear retaliation upon which stable
deterrence depends (McDermott et al. 2017). In the remainder of this chapter, however, we focus
on the more general argument that the form of warfare reflects its underlying motivation. In the
case of anger and hatred, this generates at least two types of warfare: rule-based bargaining war,
and unrestrained total war.
Anger-based bargaining aggression and the limits of war
Many features of animal bargaining aggression appear prominently in human warfare. The key
feature of anger-based aggression is that it functions to send bargaining signals that shape
preferences, and does not seek the destruction of the target. Thus, generally speaking, warfare in
these instances is “lethal but limited.” In other words, the delivery of costs should be limited in
such a way as to facilitate restructuring of adversary preferences. What are some of these limits?
Scholarship on the limits of war is perhaps as old as war itself, as evidenced by ancient
philosophical inquiries into “just war” as well as modern investigations of “limited war” (Osgood,
1957; Walzer, 2015). In practice, some limits are specific and codified, such as the Chemical
Weapons Convention and the Geneva Accords. Other limits are clearly more ancient, such as the
principle of diplomatic immunity, which has clear analogues in rules against killing the messenger
or a guest in one’s home, and are also common among foraging peoples (Boehm 1984). Rules such
as these that place limits on the methods of warfare are useful in only one context: when it is both
possible and desirable to adjust the preferences of others. These limits can be organised in terms
of the conduct and termination of war.
Conduct of war. One prominent limit that emerges in bargaining aggression involves the targeting
of “innocents”; that is, bargaining aggression should differentially target an outgroup’s coercive
abilities (e.g. combatants, not civilians; weapons, not warehouses). Therefore, modern efforts to
limit the targeting of civilians are likely the product of an evolved intuition for bargaining
aggression. For this aggression to be effective in restructuring an adversary’s preferences, it must
be discriminant, at least to a degree, even if it simultaneously carries the promise of greater pain –
which it often must for coercive leverage. The point is not that bargaining aggression should never
target non-combatants. Indeed, under some circumstances, targeting civilians can uniquely convey
the costs of war so as to incentivize a restructuring of enemy preferences. This bargaining logic is
evident, for example, in the decision of the United States to drop two atomic bombs on Japan in
WWII, and in General Sherman’s infamous March to Sea during the American Civil War. In each
of these cases, the targeting of civilians was not meant to cause pain for its own sake, but rather to
compel a revision of relative bargaining power, it being clear that a cessation of aggression (i.e.
the further targeting of civilians) was conditional upon it. This is in marked contrast to hate-based
aggression, in which the targeting of civilians is intrinsically purposeful, unconditional, and
unrestrained. This principle of discriminant violence spawns and explains a host of limits on the
conduct of war, including the non-targeting of civilians as well as prohibitions against the use of
certain weapons that are by nature non-discriminating, such as nuclear, chemical, and biological
weapons. Such weapons are tellingly lumped together under the label of Weapons of Mass
The political science literature on “limited war” helps to convey the nature of these restraints in a
useful way. For example, McClintock defines limited war as a conflict “short of general war to
achieve specific political objectives, using limited forces and limited force” (McClintock, 1967).
Stoker (2016), with reference to Clausewitz and Corbett, defines limited war more narrowly in
terms of its objectives; that is, it is anything short of the overthrow of an enemy government. We
are agnostic on whether the label “limited war” should rest on criteria specific to aims or methods;
rather, we argue that the method of violence is likely to be shaped in character by its underlying
motivation. In other words, the form of violence mirrors its ultimate function. In the case of anger-
based aggression, this function is bargaining. Therefore, we are closer to Schelling, who notes that
“[limited] war is always a bargaining process…The critical targets in such a war are in the mind
of the enemy…” (1966, p. 142). Furthermore, and to emphasise the limits of force in bargaining
aggression, Schelling also observes that “Coercion depends more on the threat of what is yet to
come than on damage already done…the object is to make the enemy behave…not to destroy the
subject altogether” (1966: 173).
Termination of war. To avoid the full costs of unlimited war, bargaining aggression must cease
when cues of recalibration are received. This involves at least two steps: One is allowing the target
the opportunity to signal recalibration (holding the head of one’s adversary under water is a poor
method for determining whether preference adjustment has occurred), and the second is cessation
of aggression itself. As Schelling illustrates, “the very idea of ‘surrender’ brings bargaining and
accommodation into warfare” (1966: 126). Bargaining aggression should therefore be
characterised by restraint at the treatment of prisoners (e.g. inhibition toward trophy taking,
torture), and surrender should be meaningful and lead to a cessation of aggression rather than
seized upon as opportunity for ambush and complete route.
For example, Harkavy explains that part of Kissinger’s strategy for a resolution to the 1973 Arab-
Israeli war entailed allowing the Arabs a “limited victory, even if largely mythical,” which was
meant to mitigate the feelings of shame and humiliation that Kissinger expected in the case of a
complete routing at the hands of the Israelis (2000, p. 348). Although such a routing would be
useful for eradication, it would be counterproductive as a bargaining strategy. This same principle
was similarly at work thousands of years prior, when Archidamus of Sparta offered the following
“…when they see that our actual strength is in keeping pace with the language that we use,
they will be more inclined to give way, since their land will still be untouched and, in making
up their minds, they will be thinking of advantages which they still possess and which have
not yet been destroyed. For you must think of their land as though it was a hostage in your
possession, and all the more valuable the better it is looked after. You should spare it up to
the last possible moment, and avoid driving them to a state of desperation in which you will
find them much harder to deal with” (Finley 1972, 82).
Bargaining aggression between groups should generate a form of aggression that seeks preference
restructuring and is limited toward that end in both the conduct of war and war termination, when
the treatment of prisoners, respect for enemy leaders and nonexploitation of weakness and
surrender can facilitate an end to hostilities consistent with recalibrated preferences. This is similar
to Blainey’s (1973) classic argument that war is fundamentally a disagreement about relative
power, in which peace is more likely after decisive wars, in which relative (bargaining) power is
clear, than after stalemates. It should come as no surprise, therefore, that bargaining aggression is
often preceded by pre-battle displays of coalitional power, or what international relations theorists
would classify as “swaggering” (Art, 1980). In contrast, hateful adversaries gain nothing and
potentially lose much by seeking to advertise offensive capabilities in advance of attacking. Anger
is the realm of bluster and bluff, while hatred is quiet and unyielding.
Hate-based aggression and the concept of total war
In contrast to anger, hate-based aggression has no bargaining function, and therefore, wars
characterised by this motivation are more likely to be total rather than limited. Whereas restraints
on the methods of violence are necessary to facilitate bargaining, hate seeks the extermination of
an outgroup rather than its preference restructuring. While anger operates to differentiate and
discriminate among an outgroup population, hate activates heuristics of outgroup homogeneity
and social substitution. In their discussion of hate, Michener (2012) labels the heuristic simply as
‘one did it, they all did it’. In total war there is little incentive to avoid the most damaging attacks
because the enemy will likely abandon those rules as well. Similarly, surrender is a meaningless
gesture in wars of extermination, as are related overtures of apology and accommodation. Thus,
behaviours such as the torture of prisoners, assassination of leaders, and the targeting of civilians
are reciprocally likely to send signals that one’s adversary seeks extermination rather than
The cue structure of negative association value combined with perceived intransigence are on
dramatic display in the context of forms of violent extremism. For example, in their extensive
review of ISIS rhetoric and propaganda, Baele et al. (2019) demonstrate that a persistent and
foundational theme of stated grievances contains the assertion that anti-Western sentiment is not
rooted in policy but identity. It is worth reviewing this language directly:
“… although some might argue that your foreign policies are the extent of what drives our
hatred, this particular reason for hating you is secondary. […] The fact is, even if you were
to stop bombing us, imprisoning us, torturing us, vilifying us, and usurping our lands, we
would continue to hate you because our primary reason for hating you will not cease to
exist until you embrace Islam” (2019: 11)
ISIS does not view itself as bargaining with the West; more to the point, it does not see bargaining
itself as possible or desirable. Rather, it views the West as a civilization of humiliation and
subjugation. Warfare motivated by such hatred will not be characterised by bargaining aggression,
but rather unrestrained aggression, e.g. genocide, torture, civilian targeting, chemical weapons,
absence of displays of open war, attempts at assassinations, and so forth.
A leading framework for the causes of war in international relations is the bargaining model of
war (Fearon, 1995; Powell, 2006), which views war as the rational outcome of issue
incompatibilities between actors. Importantly, however, this model of war begins with the
overarching claim that political actors should rationally always be motivated to seek negotiated
settlements that avoid war. According to the model, this is in part a consequence of two
assumptions: states are risk averse and war is costly for the actors. However, when hate motivates
warfare, rather than preferring to seek negotiated settlements that avoid the costs of war – as
Fearon’s (1995) bargaining model assumes – actors will instead be characterized by an inversion
of these assumptions; namely, they will view violence as intrinsically beneficial and peace as
Just as interpersonal retaliations can be motivated by two somewhat distinct styles of aggression
– anger-based bargaining aggression and a hatred-based predatory aggression – so can warfare be
characterised as either rule-based bargaining war or unrestrained total war. As warfare is ultimately
the result of many different human minds making many decisions, and because anger and hatred
can fire in unison, these types of war will often coincide; e.g. troops abandoning rules of
engagement to get revenge on the battlefield. However, the distinctions between these two types
of war are extremely important because they are predicted to respond to different strategies in very
different ways. In particular, rule-based warfare is likely to respond to concessions and peace talks,
because this type of war appears designed to resolve conflicts of interest at relatively minimal cost.
However, those engaged in hatred-based unrestrained warfare will likely see attempts to negotiate
as indicators of weakness on the opponent’s behalf, and this may be a signal that it is a good time
to escalate. Similarly, the value of negotiating at all depends on whether one is facing an opponent
who is willing to bargain or one who is only interested in getting more information so as to better
harm you. While this should make us cautious of negotiating with hateful enemies, it does not
doom us to unrestrained war. Hatred can, in some circumstances, be counteracted by changing the
perceived association value with the enemy, e.g. uniting against a common enemy, benefiting from
gains in trade, or more direct acts of charity. Finally, nations are not monolithic, and an army
directed by hateful rulers can be usurped by moderates willing to bargain.
Alcock, J. (2009) Animal Behaviour: An Evolutionary Approach, 9th edition, Sunderland, MA, US:
Antonakis, J. and Dalgas, O. (2009) ‘Predicting elections: child’s play!’, Science, 323(5918), pp.
Archer, J. (1988) The Behavioural Biology of Aggression (Vol. 1). CUP Archive.
Art, R. J. (1980) ‘To what ends military power?’, International Security, 4(4), p. 3.
Averill, J. R. (1982) Anger and Aggression: An Essay on Emotion. Springer Science & Business
Baele, S. J. et al. (2019) ‘ISIS’s Clash of Civilizations: constructing the “West” in terrorist
propaganda’, Studies in Conflict & Terrorism, https://doi.org/10.1080/1057610X.2019.1599192.
Blainey, G. (1973) The Causes of War, London: Macmillan.
Boehm, C. (1984) Blood Revenge: The Anthropology of Feuding in Montenegro and Other Tribal
Societies. Lawrence, KS: University Press of Kansas.
Buss, D. M. (2005) The Handbook of Evolutionary Psychology. John Wiley & Sons.
Chagnon, N. A. (1983) Yanomamö: The Fierce People. Holt McDougal.
Clutton-Brock, T. H., & Parker, G. A. (1995) ‘Punishment in animal societies’, Nature, 373(6511):
Collins, R (2008) Violence: A Micro-Sociological Theory, Oxford: Princeton University Press.
Cosmides, L. & Tooby, J. (1994) ‘Origins of domain-specificity: the evolution of functional
organization’, In L. Hirschfeld & S. Gelman (Eds.), Mapping the Mind: Domain specificity in
cognition and culture, New York: Cambridge University Press.
Cosmides, L. & Tooby, J. (1995) ‘From evolution to adaptations to behaviour: toward an
integrated evolutionary psychology’, In R. Wong (Ed.), Biological perspectives on motivated
activities, Norwood, NJ: Ablex.
Cosmides, L. & Tooby, J. (2000) ‘Evolutionary psychology and the emotions’, In M. Lewis & J.
M. Haviland-Jones (Eds.), Handbook of Emotions, 2nd Edition, pp. 91-115, NY: Guilford.
Daly M, & Wilson M (1988) Homicide, Transaction Books.
Darwin, C. (2004) On the Origin of Species, 1859, Routledge.
Dawkins, R. (1976) The Selfish Gene, New York: Oxford University Press.
Dawkins, R. (1996) The Blind Watchmaker: Why the Evidence of Evolution Reveals a Universe
Without Design, WW Norton & Company.
Ekman, P. (1973) ‘Cross-cultural studies of facial expression’, In P. Ekman (Ed.), Darwin and
facial expression: A century of research in review, pp. 169–222, New York: Academic Press.
Enquist, M., & Leimar, O. (1983) ‘Evolution of fighting behaviour: decision rules and assessment
of relative strength’, Journal of theoretical Biology, 102(3), 387-410.
Fearon, J. D. (1995) ‘Rationalist explanations for war’, International Organization, 49(3), pp.
Finley, M. I. (1972) History of the Peloponnesian War, Penguin Classics.
Frantz, C. M., & Bennigson, C. (2005) ‘Better late than early: the influence of timing on apology
effectiveness’, Journal of Experimental Social Psychology, 41(2), 201–207.
Gat, A. (2006) War in Human Civilization, Oxford: Oxford University Press.
Galati, D., Sini, B., Schmidt, S., & Tinti, C. (2003) ‘Spontaneous facial expressions in congenitally
blind and sighted children aged 8–11’, Journal of Visual Impairment & Blindness, 97(7), 418–428.
Garfield, Z. H., von Rueden, C. and Hagen, E. H. (2019) ‘The evolutionary anthropology of
political leadership’, The Leadership Quarterly, 30(1), pp. 59–80.
Glowacki, L., Wilson, M. L. and Wrangham, R. W. (2017) ‘The evolutionary anthropology of
war’, Journal of Economic Behavior & Organization, https://doi.org/10.1016/j.jebo.2017.09.014.
Harkavy, R. E. (2000) ‘Defeat, national humiliation, and the revenge motif in international
politics’, International Politics, 37(3), pp. 345–368.
Keeley, L. H. (1996) War before Civilization: The Myth of the Peaceful Savage, New York: Oxford
Levy, J. S. (1988) ‘Domestic politics and war’, The Journal of Interdisciplinary History, 18(4):
Lieberman, D., Tooby, J. & Cosmides, L. (2007) ‘The architecture of human kin detection’,
Nature, 445 (7129), 727-731. doi: 10.1038/nature05510
Liwanag, H.E.M., Annalisa, B, Costa, D.P., Masako A., Terrie M.W. (2012) ‘Morphological and
thermal properties of mammalian insulation: the evolution of fur for aquatic living’, Biological
Journal of the Linnean Society, 106(4), 926–939. https://doi.org/10.1111/j.1095-
Lopez, A. C. (2019) ‘Making “my” problem “our” problem: warfare as collective action, and the
role of leader manipulation’, The Leadership Quarterly, doi: 10.1016/j.leaqua.2019.05.001.
Luckenbill, D. F. (1977) ‘Criminal homicide as a situated transaction’, Social problems, 25(2),
Lukaszewski, A. W., Simmons, Z. L., Anderson, C., & Roney, J. R. (2016) ‘The role of physical
formidability in human social status allocation’, Journal of Personality and Social
Psychology, 110(3), 385.
Marsh A. A., Adams R. B. & Kleck R. E. (2005) ‘Why do fear and anger look the way they do?
Form and social function in facial expressions’, Personality and Social Psychology Bulletin, 31(1),
McClintock, R. (1967) The Meaning of Limited War, Houghton Mifflin.
McCullough, M. (2008) Beyond Revenge: The Evolution of the Forgiveness Instinct, John Wiley
McDermott, R., Lopez, A. & Hatemi P. (2017) ‘Blunt not the heart, enrage it: the psychology of
revenge and deterrence’, Texas National Security Review, 1(1): http://hdl.handle.net/2152/63934.
Michener, W. (2012) ‘The individual psychology of group hate’, Journal of Hate Studies, 10(1),
Nowak, M. A. (2006) ‘Five rules for the evolution of cooperation’, Science, 314(5805), 1560-
Osgood, R. E. (1957) Limited War: The Challenge to American Strategy, 5th ed. Edition, Chicago:
University of Chicago Press.
Pietraszewski, D. (2016) ‘How the mind sees coalitional and group conflict: the evolutionary
invariances of n-person conflict dynamics’, Evolution and Human Behavior, 37(6): 470-480.
Powell, R. (2006) ‘War as a commitment problem’, International Organization, 60(1), pp. 169–
Romero, G. A., Pham, M. N., & Goetz, A. T. (2014) ‘The implicit rules of combat’, Human
Nature, 25(4), 496-516.
Schelling, T. C. (1966) Arms and Influence, Yale University Press.
Scherer, K. R., & Wallbott, H. G. (1994) ‘Evidence for universality and cultural variation of
differential emotion response patterning’, Journal of Personality and Social Psychology, 66, 310–
Sell, A. (2011) ‘The recalibrational theory and violent anger’, Aggression and Violent Behaviour,
Sell, A. (2012) ‘Revenge can be more fully understood by making distinctions between anger and
hatred’, Commentary on McCullough, Kurzan & Tabak’s ‘Cognitive Systems for Revenge and
Forgiveness’, Behavioural and brain sciences, 36(1) 36-37.
Sell, A. (2014) ‘Twelve triggers of anger and how they invalidate all major theories of anger and
aggression’, International Society for Research on Aggression, Georgia State University, July 19th.
Sell, A., Cosmides L. & Tooby, J. (2014) ‘The human anger face evolved to enhance cues of
strength’, Evolution and Human Behaviour, 35(5), 425-429.
Sell, A., Eisner, M. & D. Ribeaud (2016) ‘Bargaining power and adolescent aggression: the role
of fighting ability, coalitional strength, and mate value’, Evolution and Human Behaviour, 37(2),
Sell, A., Sznycer, D., Al-Shawaf, L., Lim, J., Krauss, A., Feldman, A., Rascanu, R., Sugiyama, L.,
Tooby, J. & L. Cosmides (2017) ‘The grammar of anger: Mapping the computational architecture
of a recalibrational emotion’, Cognition, 168, 110-128.
Sell, A., Tooby, J., & Cosmides, L. (2009) ‘Formidability and the logic of human anger’,
Proceedings of the National Academy of Science, 106(35), 15073-78.
Sell, A., Cosmides L. & Tooby, J. (2014) ‘The human anger face evolved to enhance cues of
strength’, Evolution and Human Behaviour, 35(5), 425-429.
Stoker, Donald. (2016) ‘Everything You Think You Know About Limited War is Wrong’, War on
the Rocks, December 22.
Sznycer, D., Al-Shawaf, L., Bereby-Meyer, Y., Curry, O. S., De Smet, D., Ermer, E., Kim, S.,
Kim, S., Li, N. P., Lopez Seal, M. F., McClung, J., O, J., Ohtsubo, Y., Quillien, T., Schaub, M.,
Sell, A., van Leeuwen, F., Cosmides, L., Tooby, J. (2017) ‘Cross-cultural regularities in the
cognitive architecture of pride’, Proceedings of the National Academy of Sciences, 114(8), 1874-
Sznycer, D., Tooby, J., Cosmides, L., Porat, R., Shalvi, S., & Halperin, E. (2016) ‘Shame closely
tracks the threat of devaluation by others, even across cultures’, Proceedings of the National
Academy of Sciences, 113(10), 2625–2630. doi: 10.1073/pnas.1514699113
Tooby, J. & Cosmides, L. (1989) ‘Evolutionary psychologists need to distinguish between the
evolutionary process, ancestral selection pressures, and psychological mechanisms’, Behavioural
and Brain Sciences, 12(4), 174-175. doi: 10.1017/S0140525X00025577
Tooby, J., Cosmides, L., Sell, A., Lieberman, D. & Sznycer, D. (2008) ‘Internal regulatory
variables and the design of human motivation: A computational and evolutionary approach’, In
Elliot, Andrew J. (2008) Handbook of approach and avoidance motivation, pp. 251-271, New
York, NY, US: Psychology Press.
Walzer, M. (2015) Just and Unjust Wars: A Moral Argument with Historical Illustrations, New
York: Basic Books.
Williams, G. C. (1966) Adaptation and Natural selection: A Critique of Some Current
Evolutionary Thought, Princeton University Press.
Williams, J. (1980) Duelling in the Old South. Vignettes of Social History, College Station, TX:
Texas A & M University Press.