Download The Evolution of Generosity: How Natural Selection Builds Devices

7KH(YROXWLRQRI*HQHURVLW\+RZ1DWXUDO6HOHFWLRQ
%XLOGV'HYLFHVIRU%HQHILW'HOLYHU\
0LFKDHO(0F&XOORXJK(ULF-3HGHUVHQ
Social Research: An International Quarterly, Volume 80, Number 2,
Summer 2013, pp. 387-410 (Article)
3XEOLVKHGE\7KH-RKQV+RSNLQV8QLYHUVLW\3UHVV
DOI: 10.1353/sor.2013.0021
For additional information about this article
http://muse.jhu.edu/journals/sor/summary/v080/80.2.mccullough.html
Access provided by University of Miami (3 Sep 2015 19:25 GMT)
Michael E. McCullough and
Eric J. Pedersen
The Evolution of
Generosity: How Natural
Selection Builds Devices
for Benefit Delivery
W H Y DO W E CE L E B RA T E D A R W I N DAY B U T NOT A R I S T O T L E DAY OR
Newton Day or Descartes Day? The reason Charles Darwin is special
enough to deserve a toast every Februaiy 12, quite simply, is because he
cam e up w ith what philosopher Daniel D ennett has called “the single
best idea anyone has ever had” (1995: 21): a purely naturalistic means
o f explaining why living things bear features th at m ake them appear
as i f they had been designed (West, Griffin, and Gardner 2007). Darwin’s
fundam ental contribution to science was to identify a mindless, physi­
cal process th at could produce com plex, functional design, w hich he
called natural selection.
To understand natu ral selection, it is useful to th in k o f genes
as replicating en tities that, through repeated generational cycles o f
replication, m utation, and selection for variants that confer beneficial
fitness effects on th eir bearers, com e to build “devices” around them ­
selves. These phenotypic design features, w hich are “devices” in the
sense o f having a com plex ordering o f interrelated and hierarchically
organized parts th at cooperate to efficiently and elegantly accomplish
some task, are how genes “take action” in the world, w hich eventuates
in their increased reproductive success. And, it is through the increased
social research
Vol. 8 0 : No. 2 : Sum m er 2013 387
reproductive success o f the genes th a t produce these design features
th at those features becom e species-typical (West, Griffin, and Gardner
2007; Dennett 1995; Dawkins 1976).
The reality o f n atu ral selection as a biological process m eans
th at genes (or com binations o f genes) th a t cause organism s to build
traits that increase those genes’ rates o f replication— so-called adapta­
tions (W illiams 1966)— them selves w ill becom e increasingly com m on
in any population o f genes. Consequently, organisms will, through the
process o f natural selection, com e to bear traits that appear as i f they
were designed to increase the organism ’s reproductive success.
NATURAL SELECTION A N D THE EVO LUTIO N O F BENEFITDELIVERY DEVICES
One o f the perennial sources o f fascination for evolutionary biologists
is the realization that even though genes are “selfish” (in the sense that
natural selection causes them to com e to encode recipes for traits that
cause the organisms in w hich they reside to act in ways that benefit the
genes’ own replication), the traits they create need n ot be (Dawkins,
1976): natural selection ’s action on genes can cause the evolution o f
traits that cause organisms to interact w ith other organisms in coopera­
tive, generous, and even self-sacrificial ways. This is because there are
many ways th a t delivering ben efits to oth er individuals in th e world
ultim ately boost the rep lication rates o f the genes th at create these
forms o f generosity. Here we use the term benefit-delivery devices to refer
to the class o f biological systems th at evolved through natural selec­
tion, at least in part, because o f the benefits they caused organisms to
deliver to other organisms.
W H Y BENEFIT DELIVERY IS AN E VO LU TIO NA RY PROBLEM
Ben efit-d eliv ery devices pose a p ro b lem for ev olu tion ary th eory
because organism s th a t possess such devices should, all else being
equal, experience less reproductive success than an organism th at does
n ot possess such devices. This is because delivering benefits to oth er
individuals prevents the b en efit deliverer from using those resources
388
social research
to advance its own reproductive efforts. Consequently, it should be
im possible for m utan t genes th a t cause organism s to deliver benefits
to others to overtake a population o f genes th at do n ot create such
benefit-delivery devices; fu rth erm o re, even i f a species did acquire
a gene th a t caused it to deliver b en efits to others, th a t gene would
easily be overtaken by a new m utant gene th at prohibited its bearer
from doing so.
Nevertheless, it is self-evident th at m any organism s (including
humans) do possess devices for benefit delivery (including the cognitive
ones th at m otivate us to care about other individuals), so a cardinal task
for evolutionary science, at least as far as cooperation is concerned,
is “exp lain in g how th e a ctu al is p o ssib le” (Rosenberg 1992, 19).
Evolutionary scientists who study cooperation try to determ ine how
natural selection builds benefit-delivery devices, w hich usually involves
tiying to identify the routes by w hich the genes th at give rise to those
benefit-delivery devices are “rewarded” through enhanced reproduc­
tive fitness. In general, evolutionary biologists divide the fitness bene­
fits that enable m echanism s for cooperation to evolve through natural
selection into two broad types: direct fitness benefits (that is, the organ­
ism enjoys b etter reproductive success because o f an allele [variant o f
a gene] that causes it to deliver benefits to others), and indirect fitness
benefits (that is, the organism ’s direct reproductive success is reduced
by possessing an allele th at causes it to deliver benefits to others, but
the reproductive success o f other individuals who also share the same
allele increases, thus prom oting the replication o f the allele overall).
The sum o f direct and indirect fitness benefits is referred to as inclusive
fitness (Hamilton 1964). In the rem ainder o f this article we will describe
first some o f the direct fitness benefits, and then the indirect ones, that
can lead to the natural selection o f benefit-delivery devices. We also
describe what is currently know n (or conjectured) about the circu it
logic governing the operation o f naturally selected cognitive m echa­
nism s that regulate benefit deliver. We close w ith a b rief treatm ent o f
the concept o f group selection and its value for social research on the
evolution o f benefit-delivery devices in humans.
The E vo lu tio n o f G e n ero sity
3 89
NATURAL SELECTION O F BENEFIT-DELIVERY DEVICES
THROUGH DIRECT FITNESS BENEFITS
In evolutionary biology, a behavior is “social” w hen the m echanism s
th at m otivate it evolved in part because o f the effects o f its behavioral
outputs on oth er organism s. W hen genes give rise to benefit-delivery
systems th a t m otivate generosity because those systems boosted the
replication rates o f the ancestral copies o f the genes that govern their
assem bly and operation, evolutionary biologists talk about mutually
beneficial cooperative traits. Mutually beneficial traits evolve by natural
selection because the beneficial effects o f those traits on other individu­
als in turn cause (by virtue o f second-order effects) an increase in the
reproductive success o f the genes th at build the trait w ithin actors. It
is worthwhile here to distinguish betw een systems that evolved for the
function o f delivering benefits to others (that is, systems whose func­
tion is to increase others’ reproductive fitness) and systems that evolved
for other functions, but w hich cause benefits to flow to other individu­
als incidentally in the course o f executing their evolved functions. For
exam ple, elephants m ake available to dung beetles a substance th at
dung beetles recognize as a benefit— elephant dung—but it would be
silly to conclude from this observation that elephants defecate for the
purpose o f delivering b en efits to dung beetles. Instead, the b en efit
is a b en efit only because dung b eetle’s evolved to m ake the m ost o f
elephants’ waste products (West, Griffin, and Gardner 2007).
Thus, it is wise to restrict our d efin ition o f benefit-delivery
devices to those th at have been designed via natural selection specifi­
cally fo r th e b en eficial effects they deliver to others (West, Griffin,
and Gardner 2007). That is, some behaviors th at are often m istaken
as exam ples o f “cooperation” or “generosity” or “altruism ” in hum an
behavior m ight m erely reflect (a) actors’ pursuit o f th eir own direct
fitness interests w ithout regard to their effects on others’ fitness, and
(b) others’ opportunism in reaping fitness benefits for themselves from
the actors’ behavioral by-products. Following someone hom e w hen you
have becom e disoriented is a costless enterprise for the person whom
you follow back to camp, for instance. She pursues her own self-inter­
390
social research
est by getting h erself back to camp—w ithout regard to your w elfare—
w hile you pursue yours by following the path she takes. Thus, readers
will do well to stay m indful o f the fact th at some apparent exam ples
o f benefit delivery are not produced by benefit-delivery devices at all.
Here, we hold such by-product exam ples o f “benefit delivery” aside and
focus instead on some o f th e ways in w hich benefit-delivery devices
can be explained in term s o f the direct fitness benefits they secure for
th e genes responsible for building them . These include reciprocity,
the extensions to reciprocity th at m ight have emerged from hum ans’
specialization in the cognitive niche, and group augmentation.
Reciprocity
The m ost com m only studied route by w hich benefit-delivery devices
could have evolved in hum ans via n atu ral selection is recip rocity
(Trivers 1971). The P riso n er’s D ilem m a, in w h ich two players are
presented w ith a ch oice to cooperate w ith, or d efect against, each
other, nicely illustrates how reciprocity can generate benefits. I f both
players cooperate, they each receive a m oderate payoff; i f b oth defect,
they each receive a sm all payoff; if one player defects and the oth er
cooperates, th e d efector receives a large payoff and th e cooperator
receives the sm allest possible payoff (the “su cker’s p ayoff”). Due to
th is payoff stru ctu re, d efectio n always yields th e h igh est average
payoff in a one-shot in teractio n (that is, an in teraction th at w ill not
be repeated). However, i f th e gam e consists o f m ultiple rounds o f iter­
ated play, m utual cooperation can lead to higher average payoffs than
m utual or alternating d efection (Axelrod and H am ilton 1981; Trivers
1971). W h en one thinks about these payoffs in the currency o f “life­
tim e reproductive success,” we can see how the prison er’s dilem m a
also illustrates how reciprocity m ight cause the evolution o f benefitdelivery devices via natural selection.
Axelrod and Hamilton (1981) demonstrated that the iterated pris­
on er’s dilem m a provides a gam e-theoretic m odel for the evolution o f
direct reciprocity. M echanisms th at deliver benefits to others via direct
reciprocity (as modeled in the iterated prisoner’s dilemma) can evolve
The E vo lu tio n o f G enerosity
391
by natural selection w hen the probability o f a successive round o f inter­
action betw een two individuals (that is, the probability that two indi­
viduals will interact again in th e future) exceeds the ratio o f the lifetim e
fitness cost o f the benefit delivery to the donor divided by its lifetim e
fitness value to the recipient. Results from m ore recen t models suggest
th a t a bias to cooperate, even w hen one faces cues o f an in teraction
being one-shot, should be expected to co-evolve w ith reciprocity. This
is because m istaking a one-shot interaction for a repeated interaction
is a less costly error than the reverse—th a t is, it is m uch m ore costly
to m iss out on th e long-term benefits th a t can be obtained through
repeated gains in trade than it is to be exploited in a single interaction
(Delton et al. 2011).
W ith a few p ossib le ex cep tio n s, recip ro city appears rela­
tively unim portan t for understanding cooperation in m ost non h u ­
m an anim als (Clutton-Brock 2009; W est, Griffin, and Gardner 2007).
However, reciprocity is arguably very im portant for understanding the
evolution o f hum an cooperation because m uch o f hum an social inter­
action involves long-term relationships am ong nonrelatives (such as in
the contexts o f hunting, foraging, territory defense, dwelling construc­
tion, food preparation, and child care). For exam ple, pooling risk and
effort for hunting by sharing the spoils w ith others, on the condition
th at recipients o f benefits at one point in tim e reciprocate in the future,
is m uch m ore efficient than solitary hunting to provision only oneself
and one’s close kin. This is because hunting is a relatively high-variance
m ethod o f food acquisition: am ong ex tan t hunter-gatherer groups,
individual probabilities o f failed hunts on any given day range from
40 percent to 96 percent (Gurven et al. 2012). Reciprocal cooperation
in such circum stances can reduce or elim inate the variance in individ­
ual food availability, leading to direct benefits for recipients (through
im m ediate acquisition o f food) and benefactors (by motivating recipro­
cal help at a later date).
Although reciprocity can provide m utual direct benefits through
gains in trade, it also exposes cooperators to the possibility o f exploi­
tation—that is, to interacting w ith individuals who will take benefits
3 92
social research
w ithout reciprocating them at a later tim e (Cosmides 1989; Trivers
1971). The dangers o f exploitation m aybe particularly acute in humans,
for hum ans possess zoologically unusual cognitive capabilities such
as language, m ental tim e travel, the ability to infer oth ers’ goals and
psychological states, and th e ability to in fer the social consequences
o f their (and oth ers’) actions. These cognitive capacities m ight m ake it
easier for humans to evolve m echanism s for exploiting other individu­
als’ cooperative dispositions.
W ere such exploitive abilities to evolve, they would set the stage
for evolutionary arm s races in w hich natural selection m ight conse­
quently favor the evolution o f m echanism s th at im p lem en t defen­
sive tactics for resisting or deterring exploitation. To the exten t th at
n atu ral selection has indeed produced adaptations for exploitation
and counter-exploitation in reciprocal interaction, all neurologically
in tact hum ans can be expected to possess the full suite o f exploitive
and counter-exploitive m echanism s. Such subsidiary cognitive m ech­
anism s m ight include perceptual systems th at identify likely reciprocators (for exam ple, kin detection; Lieberm an, Tooby, and Cosmides
2007) th a t in h ib it individuals from helping likely nonreciprocators
(Cosmides 1989), and that m otivate the punishm ent or term ination o f
interactions w ith nonreciprocators (West, Griffin, and Gardner 2007).
Additional m echanism s m ight include capacities for rem em bering the
outcom es o f cooperative in teraction s (for exam ple, classifying those
actions with respect to w hether they violated social contracts), as well
as m otivational systems for im pelling cooperation, reciprocal return o f
benefits, and the punishm ent o f cheaters (Trivers 1971).
It is worth keeping in m ind that the prospect th at a given bene­
fit-delivery system evolved via reciprocity (that is, by virtue o f the
fact that benefit/cost ratio exceeded the probability o f repeat encoun­
ters betw een the actors in th e environm ent in w hich the m echanism
evolved) tells us nothing, in and o f itself, about how th e resultan t
cognitive m echanism s im plem ent their functions. A reciprocity-based
benefit-delivery system could (theoretically) simply entail, for instance,
a cognitive system th a t com pares on e’s own language or accen t to
The E vo lu tio n o f G e n ero sity
393
the language or accent o f a potential reciprocity partner— since facil­
ity w ith the sam e language was likely very highly correlated w ith
geographic locale, and thus, the likelihood o f m eeting again (Pagel and
Mace 2004)— and th at m otivates benefit-delivery toward individuals
whom one determ ines to speak in a language or w ith an accent th at is
suitably sim ilar to one’s own (Nettle and Dunbar 1997). It is the jo b o f
evolutionary behavioral scientists to em pirically determ ine the circuit
logic that governs the operations o f these systems; the theories o f social
evolution do not them selves prescribe how they should be cognitively
or neurally instantiated (Scott-Philips 2007).
OCCUPYING THE C O G N ITIVE NICHE FAVORS THE
EVO LUTIO N OF INDIRECT RECIPROCITY
Tooby and DeVore (1987) proposed th at hum ans evolved to exploit a
“cognitive n ic h e ”— th at is, a way o f life for w hich natural selection
resulted in problem -solving m echanism s th at enable organism s to
form ulate in real tim e (rather than over evolutionary time) novel strate­
gies for surm ounting the fixed defenses o f other organisms (for exam ­
ple, the toxins in otherw ise edible plants or the defensive weaponry
o f potentially edible animals), and in so doing, extract resources from
them . Pinker (2010), in particular, proposed th at hum ans succeed in
exploiting the cognitive niche due to a suite o f zoologically exceptional
cognitive adaptations that include technological problem-solving and
gram m atical language. W ith these adaptations in place, new adaptive
space opens up in w hich new variations on reciprocity can plausibly
evolve.
For exam ple, once language has evolved, in form atio n about
third parties’ previous cooperative behaviors can be efficiently trans­
m itted orally from someone who has acquired direct (or even indirect)
knowledge bearing on this issue (“Maggie is ju st so stingy . . . ”) to third
parties who lack such knowledge. If people subsequently were to come
to possess, via natural selection, psychological m echanism s that m oti­
vated them to adjust th eir cooperative behavior on the basis o f such
reputational inform ation, they could avoid the costs associated w ith
394
social research
providing benefits to nonreciprocators. Indeed, people do use thirdperson inform ation to preferentially cooperate w ith those who have
cooperated w ith others, but only w hen they lack first-person experi­
ence (once first-person experience is acquired, third-party inform ation
becom es irrelevant; Krasnow et al. 2012). This sets the stage for the
evolution o f so-called indirect reciprocity (Nowalc and Sigmund 1998),
w hich results in individuals who appear as if they are m otivated to
reward individuals for their previous cooperative behavior toward other
b eneficiaries—w hen w hat they are really doing is stream lining their
own efforts to b en efit from reciprocal in teraction by directing ben e­
fits toward individuals who have proven them selves to be trustworthy
reciprocators.
GROUP AUGMENTATION
In som e anim al species, m em bers o f stable social groups provide
care (that is, deliver benefits) to offspring th at are n eith er th eir own
offspring nor th e offspring o f close genetic relatives (Clutton-Brock
2002). Although such behavior at first seem s puzzling, a closer look
at the consequences o f reproductive support reveal several routes by
w hich the genes th at build the m echanism s th at m otivate such behav­
iors could be favored by natural selection. First, providing reproductive
support can increase helpers’ direct fitness by increasing their chances
o f survival (Grinnell, Packer, and Pusey 1995) or yielding allies if they
m igrate from their natal groups to begin th eir reproductive careers else­
w here (Clutton-Brock 2002). Second, by contributing to group survival,
nonreproductive m em bers can increase the likelihood th at helpers will
be available to aid th eir own reproductive efforts if they should ever
becom e one o f the group’s reproductive m em bers (Clutton-Brock 2002).
H um ans are n o t coop erativ e breed ers, o f cou rse; an cestral
hum an societies are n ot characterized by castes o f nonreproductive
helpers. Even so, group size its e lf m ight nevertheless have b een an
im p o rtan t source o f d irect fitn ess b en efits for an cestral hum ans.
For exam ple, studies w ith non h um an anim als suggest increases in
group size can in crease th e fitn ess o f all group m em bers because
The E vo lu tio n o f G enerosity
395
larger groups can be b etter at obtaining (and retaining) food (CluttonBrock 2002), spotting and defending against predators (Clutton-Brock
2002), defending breeding sites against takeovers (Port, Kappeler, and
Joh n ston e 2011), and rearing offspring. Because o f the direct benefits
to individuals in the group th a t m ight accrue simply by virtue o f the
group’s expanding, it can b e b en eficia l for helpers to provide care
for unrelated offspring in order to help in increasing the size o f the
group. In addition to direct benefits from so-called group augm enta­
tion, provisioning b en efits to young can also be repaid via “delayed
reciprocity,” or benefits provided to the helper by offspring who have
reached adulthood. Although th e im portance o f group augm entation
as an evolutionary route to th e evolution o f h u m an s’ cooperative
instincts is currently unknow n, it has received less atten tion th an it
m erits (Port, Kappeler, and Joh n ston e 2011).
NATURAL SELECTION O F BENEFIT-DELIVERY DEVICES
THROUGH INDIRECT FITNESS BENEFITS
The classical Darwinian notion o f natural selection has seen only one
m ajor th eoretical refinem en t. This refinem en t resulted from efforts
by biologists to understand how traits that reduce their bearers’ direct
reproductive success can evolve. The existence o f sterile worker castes,
for exam ple, along w ith less extrem e instances o f organisms that pay
fitness costs in order to provide fitness benefits to others, caused evolu­
tionists to eventually realize that the scope o f natural selection was
broader than they had originally realized: natural selection’s partiality
for a gene is determ ined n o t only by how that gene affects the repro­
ductive success o f the individual organism in w hich it resides, but also
by how it affects the reproductive success o f replicas o f itse lf that are
locked inside other organism s— namely, genetic relatives o f the donor
individual. Decades later, Hamilton (1964) showed m athem atically that
natural selection builds com plex functional design by favoring m utant
genes that cause traits th at boost those genes’ inclusive fitness, w hich is
the sum o f direct fitness (a genetic variant’s success in creating replicas o f
itself) and indirect fitness (its success in assisting other copies o f itse lf—
396
social research
located inside offspring and siblings and cousins and other relatives—
to replicate).
H am ilton ’s m a jo r in sig h t— la ter dubbed H am ilton’s rule— was
th a t a new genetic m utant th a t reduces its b earers’ direct reproduc­
tive success can evolve by natural selection if the lifetim e reproductive
b en efit b it confers to all recipients, discounted by the average degree
o f relatedness r betw een th e actor and all recipients (which is really
shorthand for “the probability that the recipients o f the benefit share
the new genetic variant that is responsible for causing the benefit-delive iy ”), exceeds the lifetim e reproductive cost c that the trait imposes
upon the actor— th a t is, w hen rb - c > 0. Traits th at evolve by adher­
ing to Ham ilton’s rule can be said to have evolved through “kin selec­
tion ,” as com m on ancestry (kinship) is the m ost com m on reason why
two individuals share a genetic variant. Because they evolve through
the accum ulation o f genes th a t cause organisms to provide benefits for
other individuals th at the donors could otherwise use to pursue their
own reproductive agendas (with benefits and costs referring to lifetime
fitness consequences), kin-selected traits can rightly be said to evolve
through an “altruistic” pathway o f natural selection (West, Griffin, and
Gardner 2007).
Adaptations for Parental Care as Altruism Devices
Some o f the clearest illustrations o f altruistically designed biological
devices are adaptations for parental care (Dawkins 1979, 1976), and
here we dwell on them in some detail. “Parental care” refers to the full
suite o f anatom ical, physiological, and behavioral adaptations whose
function is to cause parents to invest in the fitness o f th eir offspring
(Clutton-Brock 1991). For m any biologists who study parental care, it
seem s intuitively wrong to th in k o f adaptations for parental care as
“altruistic.” They often argue that w hen an organism provides parental
care it is simply pursuing th e m axim ization o f its own direct fitness,
w ith d irect fitn ess being defined as th e num ber o f the organism ’s
offspring that reach reproductive maturity. Nevertheless, evolutionary
geneticists have good reasons to think th at the production o f zygotes
The E vo lu tio n o f G enerosity
397
(the cell th a t results from th e u nion o f egg and sperm) is a b etter
measure o f an individual’s fitness than is the num ber o f their offspring
th a t m ake it to m aturity (Hunt and Hodgson 2010). As a side benefit,
we find it is clarifying to count fitness in zygotes rather than in sexu­
ally m ature adults because it is a less parento-centric view o f sexual
reproduction. Most organism s do n ot provide any parental care, and
relatively few provide m ore than hum ans do. Consequently, one should
not be surprised if adaptations for parental care require a special kind
o f biological explanation. W e th in k th at explanation is the altruistic
route to com plex functional design.
In this way o f thinking, m am m alian lactation is an iconic exam ­
ple o f an altruistically designed benefit-delivery device (Oftedal 2012).
Human m others, for instance, whose infants (in traditional societies
and, one presumes, ancestral societies) feed exclusively on breast m ilk,
m ust increase th eir caloric intake by 670 kilocalories per day because
o f the high m etabolic costs o f m ilk production (Dewey 1997). In addi­
tion, there are opportunity costs: breastfeeding stimulates the produc­
tion o f prolactin, a horm one that both stim ulates m ilk production and
prevents the release o f the horm ones (called gonadotropins) th at cause
w om en’s m enstrual cycles to resume. Thus, while hum an infants are
nutritionally dependent on th eir m o th er’s m ilk (a period th at lasts
betw een 2 and 3 years am ong w om en from nonindustrial societies;
Sellen 2001), wom en are typically not ovulating, w hich is to say, they
are delaying the pursuit o f increasing th eir direct reproductive success.
Moreover, it is insufficient m erely to produce m ilk: m am m alian
m others also m ust have a behavioral repertoire that makes them both
able and w illing to tran sfer th at m ilk to th eir offspring. There is an
im portant but often underappreciated behavioral com ponent to success
in breastfeeding, which includes learning to read infants’ signals that
they are ready to eat, specialized learning systems that stim ulate m ilk
let-down in response to in fan ts’ cries, developing skill with stim ulating
babies who are too sleepy to eat and soothing babies who are too fussy
to eat, and others (Mulford 1992). And ideally, all o f this learning should
happen very soon after the in fan t is born. For this reason, natural selec­
398
social research
tion, through the altruistic design route, plausibly led to the evolution
o f specialized learning m echanism s, physically instantiated through a
com plex netw ork o f neural and endocrine pathways, whose function is
to enable m others to becom e proficient at breastfeeding quickly.
Milk delivery is only one elem ent o f parents’ food provisions for
their offspring (in modern and traditional societies alike, parents subsi­
dize th eir offspring’s caloric intake throughout adolescence; Kaplan
et al. 2000), and food provision itse lf is m erely the tip o f the parental
care iceberg. For exam ple, the anthropologist Melvin Konner (2010) has
enum erated many additional evolved functions o f hum an motherhood.
These include tem perature hom eostasis; protection from predators,
enem ies, and organisms such as insects, scorpions, and snakes; transfer
o f specific im m unity; and th e nongenetic transm ission o f behavioral
and psychological dispositions. Parents also invest in th eir children’s
reproductive success by helping them to acquire the hum an capital—
specialized storehouses o f know ledge, skill, or expertise— th at will
m ake them valuable to potential m ates, potential friends and allies, and
m em bers o f the wider com m unity (Geary 2010). The cognitive m echa­
nism s th at enable m others (and fathers) to provide these benefits to
offspring require natural selection explanations every b it as m uch as
adaptations for food delivery do.
Kin Discrimination Systems
The high degree o f relatedness betw een m others and th eir offspring
m eans that it is relatively easy for natural selection to build altruistic
adaptations for m aternal care, but another im portant factor is the high
availability o f so-called kin discrim ination cues, w hich enable altruistic
genes to “know ” (in a purely m etaphorical sense) w here to direct their
help. Kin discrim ination— the ability to recognize (nonconsciously, o f
course) one’s own relatives and then regulate behavior in response to
th at recognition— is a feature o f hum an nature that is so thoroughly
taken for granted (because language enables us to label certain individ­
uals as our fathers, m others, siblings, grandparents, and so forth) that
we fail to recognize that altruistically designed systems for caring— in
The E vo lu tio n o f G e n ero sity
399
hum ans and m any other species— often rely deeply on a netw ork of
kin-discrim ination cues to regulate their operation. By relying on such
cues, evolving altruistic systems can reduce the costs o f parental care
per unit o f benefit transferred to offspring, which, per H am ilton’s rule,
enhances evolvability (West, Griffin, and Gardner 2007).
Many cues for offspring recognition are available to m others. As
a soon-to-be-born m am m al descends through its m oth er’s b irth canal,
for instance, th e stretch o f th e m uscle fibers th at line th e uterin e wall
send a volley o f neural signals to the brain th at stim ulate the pituitary
gland to release prolactin and oxytocin. These horm ones, in turn, not
only stim ulate the further progress o f labor and delivery, but also the
activation o f m ilk production and the b ra in ’s behavioral parenting
circuits (Russell and Leng 1998). After infants are born, hum an m oth­
ers alm ost instantaneously appear to begin m aking use o f visual, audi­
tory, and even olfactory cues to th eir offspring (some o f w hich can be
updated as children age and develop) for the purpose o f kin recogni­
tion (Tal and Lieberm an 2007). It is also likely th at hum ans, like m any
oth er m am m als (and birds and fish), com e to recognize offspring on
the basis o f co-residence— perhaps m ediated by cognitive system s
th at im plem ent the logic “any young in m y nest are m in e” (Penn and
From m en 2010, 64).
In discussing altru ism thus far, we have focused m ainly on
m other-child relation sh ip s, but a ltru istic benefit-delivery systems
have very likely also evolved to regulate benefit-deliveiy from fathers,
siblings, grandparents, aunts and u ncles, and o th er relatives (Sear
and Mace 2008; Gurven et al. 2012). In m ost animals, m others have a
distinct advantage over fathers and other relatives in offspring recogni­
tion because o f their heavy physiological investm ents in the growth and
developm ent o f the offspring following egg fertilization. These invest­
m ents m ake available to m others (but not fathers and other kin) a vari­
ety o f kin discrim ination cues that m aternal altruistic benefit-deliveiy
systems can put to use. Although some evidence suggests that fathers
assess kinship by relying on visual or olfactory cues o f genetic sim ilar­
ity (Alvergne, Fauire, and Raymond 2009), it is also possible th at fathers
400
social research
possess com putational neurocircuitiy th at solves offspring-recognition
problem s in a different way.
Holding aside the w onders o f m odern fertility m edicine, m en
cannot be the fathers o f the children o f wom en w ith w hom they did not
have sex, so a reasonable first step for simplifying the offspring recogni­
tion problem would be for m en to rule out o f further consideration all
w om en w ith w hom they have never had sex. Second, because ances­
tral m en cannot be fathers o f children who are not also the children o f
wom en w ith w hom they had sex, they m ight also observe the relations
betw een the wom en with w hom they have had sex and the young chil­
dren in their midst. Young children who receive m aternal care from a
wom an w ith w hom the m an had sex approxim ately 10 m onths prior to
the in fan ts’ births are likely (holding aside other considerations such
as the w om an’s likelihood o f having had sex w ith oth er m en, w hich
m ight also be relevant) to be the m an’s children (Lieberman, Tooby, and
Cosmides 2007; Tal and Lieberm an 2007).
Sim ilarly, child ren apparently recognize th e ir siblings based
largely on environm ental cues acquired during childhood experience,
including the num ber o f years o f co-residence and (for older siblings)
w atching one’s own m other provide m aternal care to another individ­
ual (Tal and Lieberm an 2007; Lieberman, Tooby, and Cosmides 2007).
DOES GROUP SELECTION EXPLAIN ADAPTATIONS FOR
BENEFIT DELIVERY?
Group selection is the idea th a t natural selection can w ork by en h an c­
ing th e fitness o f groups o f individuals rath er th an acting exclusively
on individuals. In th e sc ie n tific lite ra tu re, som e sch olars p resen t
group selection as an alternative to standard inclusive fitness m axi­
m ization m odels o f n atu ral selection in order to explain the evolu­
tion o f m echanism s for hum an b en efit delivery (Fehr and Fischbacher
2 0 0 3 ; Sober and W ilson 2011). C urrent group se lectio n th eo rists
recog n ize th a t in clu sive fitn ess m odels are n ecessary in g en eral
and th a t th ey ex p lain m an y coop erativ e p h en o m en a, b u t th ese
scholars also m ain tain th a t inclusive fitn ess theory can not account
The E vo lu tio n o f G e n ero sity
401
fo r all o f th e types o f coop eration found in hum ans. For in stan ce,
group se lectio n has b een invoked to exp lain coop eration in oneshot in teraction s (in teraction s th a t w ill n o t be repeated and h en ce
can not be reciprocal; Gintis 2000), costly punishm ent in public goods
scenarios (Fehr and G achter 2002), costly punishm ent by third-party
w itnesses o f unfairness (Fehr and Fischbacher 2004), and even costly
punishm ent o f an individual who has directly harm ed the punisher
(Fehr and Fischbacher 2003).
In all o f these cases, proponents o f group selection models argue
that the behaviors in question provide benefits to others (for example,
punishm ent deterring a transgressor from harm ing a stranger in the
future) at a n et cost to the individual perform ing them (for exam ple,
pu n ish m en t is in h eren tly costly to th e pu nisher as it takes tim e,
energy, and could incite retaliation) and thus cannot evolve through
the standard Darwinian design pathways. Importantly, group selection
proponents also tend to be skeptical o f claims that behaviors like these
would have been adaptive for individuals under ancestral (though not
modern) conditions, or that they would reveal them selves to be adap­
tive even today if the fitness benefits were reckoned over the course o f
a lifetim e (rather than over the course o f a one-hour laboratory session),
or th a t th eir salutary effects on others are by-products rath er th an
evolved functions (Chudek, Zhao, and Henrich in press). Space lim ita­
tions prevent form al consideration o f these specific em pirical claim s,
so we lim it our attention here to describing the various ways in which
group selection is invoked and some o f the lim itations o f these usages.
Types o f Group Selection
It is im portant to note th a t group selection is n ot a singular, form al
theory. Rather, the term has been applied to several different processes
th at can usefully be classified as exam ples o f either “old” group selec­
tion or “new ” group selection (West, El Mouden, and Gardner 2011).
Old group selection models propose th at group-level adaptations would
result from d ifferen tial survival am ong groups—th a t is, th a t traits
would be selected to m axim ize group success (West, El Mouden, and
402
social research
Gardner 2011). For exam ple, Wynne-Edwards (1962) argued that groups
o f organisms th at sought solely to m axim ize their own individual inter­
ests would exhaust ecological resources and therefore face extinction,
whereas groups o f organisms that exercised restraint in pursuing their
individual interests would n ot deplete th eir resources, and thus would
be m ore likely to survive. Through a persistent process o f differential
group success on th e basis o f individual restrain t, Wynne-Edwards
argued, behaviors could evolve th at ben efitted the group as a whole
even if costly to the individual. Old group selection models therefore
rely on the proposition that groups o f individuals, and not m erely indi­
viduals, can be the vehicles on which natural selection acts.
The theoretical and em pirical w ork th at em erged in the wake
o f Wynne-Edwards’s proposal showed th at old group selection, w hile
possible in theory, could only w ork under a very narrow range o f condi­
tions, and thus was veiy likely to be unim portant in explaining adap­
tation (and basically im possible in hum ans; W illiam s 1966; W est, El
Mouden, and Gardner 2011). For exam ple, group-level adaptations can
be favored by natural selection w hen all o f the individual organisms
w ithin a group are genetic clones, w hich can occur w hen relatedness
am ong group m em bers is high (that is, m igration is essentially zero)
and there is com plete suppression o f reproductive com petition w ithin
groups (West, El Mouden, and Gardner 2011; Gardner and Grafen 2009).
Because these conditions are rarely fulfilled in nature— and not at all
in the case o f hum ans—group-level adaptation is rarely invoked inten­
tionally in the contem porary literature, although confusion arising
from “new ” group selection models does som etim es result in research­
ers rehashing some old group selection ideas (West, El Mouden, and
Gardner 2011).
Unlike old group selection models, “new ” group selection models
do n ot focus on group-level adaptations. Rather, new group selection
models propose th at natural selection operates at m ultiple levels, and
that traits can be selected for if their benefits at the level o f the group
(between-group level) outweigh the benefits at the level o f the individ­
ual (within-group level) (Sober and W ilson 2011, 1998). Some models
The E vo lu tio n o f G enerosity
403
also posit that direct com petition betw een groups results in selection
for cooperative behaviors, such as w hen groups battle for territories
(that is, groups w ith a g reater num ber o f cooperators are expected
to outcom pete groups w ith fewer cooperators; W est, El Mouden, and
Gardner 2011). O ther m odels th at can be categorized as new group
selection models incorporate culture to explain differential success o f
groups, w hereby groups th a t express a particular cultural trait outcom ­
pete groups th at do not (Boyd and Richerson 2005; W est, El Mouden,
and Gardner 2011). These various new group selection models, collec­
tively, are featured prom inently in the current literature on the evolu­
tion o f benefit delivery in humans.
Is Group Selection Useful?
Group selection m odels can provide a ten able selectio n ist logic by
w hich some adaptations for ben efit delivery evolve (Gardner in press).
So why do m ost evolutionary biologists resist using group selection
m odels? There are four reasons. First, group selection models do not
provide any explanatory pow er above and beyond the power th at is
already built into standard inclusive fitness theory (that is, accounting
for direct and ind irect benefits). More strongly, new group selection
models already are incorporated into standard inclusive fitness theory—
though th ey rely on d ifferen t m ath em atical expressions o f n atu ral
selection, they yield identical predictions (Gardner in press; Gardner
and Grafen 2009).
Second, despite many claims to the contrary, no one has proposed
an em pirical or theoretical exam ple o f a benefit-delivery system that
can be shown to evolve by m ulti-level selection and th at cannot also be
shown to evolve through standard inclusive fitness considerations (for
a list o f examples, see table 5 o f W est, El Mouden, and Gardner 2011).
The converse is not true: som e benefit-deliveiy systems th at are explica­
ble w ith an inclusive fitness m axim ization approach to understanding
natural selection cannot be explained using the analytical tools asso­
ciated w ith m ulti-level selection. Third, the m athem atics required to
analyze the evolution o f biological phenom ena w ith group selection
404
social research
models tends to be m uch m ore unwieldy than the m ath th at is required
to analyze them w ith inclusive fitness models.
Fourth, and perhaps m ost im portant, even w hen using group
selectionist models to explain the evolution o f benefit-delivery systems,
one reaches the conclusion th a t organism s still evolve in such a way
as to pursue th eir inclusive fitness interests. That is, even w hen one
uses the (often) cum bersom e group selection approach to modeling the
selection dynamics th at lead to benefit-deliveiy devices, those devices
still reside in individuals who w ill appear as i f they are designed to
optimize their inclusive fitness. Group selection may provide an alter­
native view o f selection, but it does not lead to new understandings o f
what results from natural selection. This is one source o f the elegance
o f inclusive fitness theory: it seamlessly bridges the process o f natural
selection (genes acting in th e world to create phenotypes th at boost
th eir own rates o f replication will increase in the population) and the
product (individual organism s th a t appear designed to pursue the
enhancem ent o f their own inclusive fitness). Group selection views do
not unite process and product in such an elegant fashion (Gardner in
press).
CONCLUSION
Human generosity is not an illusion, and m uch o f the generosity that
we see betw een individuals and w ithin societies today is motivated by
biological systems that plausibly evolved by natural selection through
the causal pathways we have outlined here. But is evolution the only
gam e in town for explaining hum an generosity? Far from it. Much o f
th e generosity we see in th e contem porary world is probably caused
not by biological systems designed for generosity acting alone. Instead,
these m echanism s often in tera ct w ith unique innovations th at have
arisen in the m odern world— for exam ple, laws, charitable organiza­
tions, taxation, and social insurance—that were intentionally designed
to goad humans into greater generosity than could be leveraged on the
basis o f naturally selected benefit-delivery devices alone. Some o f these
innovations m ight n ot even require the action o f naturally selected
The E vo lu tio n o f G enerosity
405
benefit-delivery m echanism s at all. All they m ight require, instead,
is for hum ans to apply th eir capacities for abstract reasoning (Singer
1981) and their evolved inclinations to avoid sticks (such as the form al
penalties associated w ith tax dodging or failing to return found prop­
erty; W est 2003; W ebber and Wildavsky 1986), pursue carrots (as soci­
eties grow larger, and as com m unication goes from oral to w ritten to
electronic, the potential reputational benefits o f generosity increase
exponentially), and follow fashionable social trends (note the rise o f
charitable voluntary associations in V ictorian England; H im m elfarb
1991). A com plete understanding o f hum an generosity in the modern
world, therefore, w ill require not only the careful application o f the
insights o f evolutionary science, but also a careful inspection o f the
insights that can com e from histoiy, political science, econom ics, and
the other social sciences.
ACKNOW LEDGEM ENTS
We than k Claire El Mouden and Robert Kurzban for com m ents on a
previous draft.
REFERENCES
Alvergne, A., C. Faurie, and M. Raymond. 2009. “Father-Offspring
Resem blance Predicts Paternal Investm ent in Humans.” Animal
Behavior 78: 61-69.
Axelrod, R., and W. D. Hamilton. 1981. “The Evolution o f Cooperation.”
Science 211; 1390-1396.
Boyd, R„ and P. J. Richerson. 2005. The Origin and Evolution o f Cultures.
Oxford: Oxford University Press.
Chudek, M„ W. Zhao, and J. Henrich. In press. “Culture-Gene Coevolution,
Large-Scale C ooperation and th e Shaping o f Human Social
Psychology.” In Signaling, Commitment, and Emotion, edited by B.
Calcott, R. Joyce, and K. Sterelny. Cambridge: MIT Press.
Clutton-Brock, T. H. 1991. The Evolution o f Parental Care. Princeton:
Princeton University Press.
----------. 2002. “Breeding Together: Kin Selection and M utualism in
406
social research
Cooperative Vertebrates.” Science 296: 69-72.
--------- . 2009. “Cooperation betw een Non-kin in Animal Societies.” Nature
462: 51-57.
Cosmides, L. 1989. “The Logic o f Social Exchange: Has Natural Selection
Shaped How Humans Reason? Studies w ith the Wason Selection
Task.” Cognition 3 1 :1 8 7 -2 7 6 .
Dawkins, R. 1976. The Selfish Gene. Oxford: Oxford University Press.
--------- . 1979. “Twelve Misunderstandings o f Kin Selection.” Zeitschrift fixr
Tierpsychologie 5 1 :1 8 4 -2 0 0 .
Delton, Andrew W., Max M. Krasnow, Leda Cosmides, and John Tooby.
2011. “Evolution o f D irect Reciprocity under U ncertainty Can
Explain Human Generosity in One-Shot Encounters.” Proceedings o f
the National Academy o f Sciences 108 (32): 13335-13340. doi: 10.1073/
pnas.1102131108.
Dennett, D. C. 1995. Darwin’s Dangerous Idea: Evolution and the Meanings o f
Life. New York: Simon and Schuster.
Dewey, K. G. 1997. “Energy and Protein Requirements during Lactation.”
Annual Review o f Nutrition 1 7 :1 9 -3 6 .
Fehr, E., and U. Fischbacher. 2003. “The Nature o f Human Altruism.”
Nature 425: 785-791.
--------- . 2004. “Third-Party Punishm ent and Social Norms.” Evolution and
Human Behavior 25: 63-87.
Fehr, E., and S. Gachter. 2002. “Altruistic Punishment in Humans.” Nature
4 1 5 :1 3 7 -1 4 0 .
Gardner, A. In press. “Adaptation o f Individuals and Groups.” In From
Groups to Individuals: Evolution and Emerging Individuality, edited by F.
Bouchard and P. Humeman. Cambridge: MIT Press.
Gardner, A., and A. Grafen. 2009. “Capturing the Superorganism : A
Formal Theory o f Group Adaptation.” Journal o f Evolutionary Biology
2 2 :6 5 9 -6 7 1 .
Geary, D. C. 2010. Male, Female: The Evolution o f Human Sex Differences.
Washington, D.C.: American Psychological Association.
Gintis, H. 2000. “Strong Reciprocity and Human Sociality.” Journal o f
Theoretical Biology 2 0 6 :1 6 9 -1 7 9 .
The E vo lu tio n o f G enerosity
407
Grinnell, J., C. Packer, and A. E. Pusey. 1995. “Cooperation in Male Lions:
Kinship, Reciprocity or Mutualism?” Animal Behaviour 49: 95-105.
Gurven, M., J. Stieglitz, P. L. Hooper, C. Gomes, and H. Kaplan. 2012.
“From the W omb to the Tomb: The Role o f Transfers in Shaping
th e Evolved H um an Life H istory.” Experim ental Gerontology
4 7 :807-813.
Hamilton, W. D. 1964. “The Genetical Evolution o f Social Behaviour. I, II.”
Journal o f Theoretical Biology 7 :1 -5 2 .
Himmelfarb, G. 1991. Poverty and Compassion: The Moral Imagination o f the
Late Victorians. New York: Knopf.
H unt.J., and D. Hodgson. 2010. “W hat Is Fitness, and How Do We Measure
It?” In Evolutionary Behavioral Ecology, edited by D. E. W estneat and C.
W. Fox, 4 6 -7 0 . New York: Oxford University Press.
Kaplan, H. S., K. Hill, J. Lancaster, and A. M. Hurtado. 2000. “A Theory o f
Human Life History Evolution: Diet, Intelligence, and Longevity.”
Evolutionary Anthropology 9 :1 5 6 -1 8 4 .
Konner, M. 2010. The Evolution o f Childhood: Relationships, Emotion, Mind.
Cambridge: Harvard University Press.
Krasnow, M. M„ L. Cosmides, E. J. Pedersen, an d j. Tooby. 2012. “W hat Are
Punishment and Reputation For?” PLoS ONE 7: e45662.
Lieberman, D., J. Tooby, and L. Cosmides. 2007. “The Architecture o f
Human Kin Detection.” Nature 445: 727-731.
Mulford, C. 1992. “The Mother-Baby Assessment (MBA): An “Apgar Score”
for Breastfeeding.”Journal o f Human Lactation 8: 79-82.
Nettle, D„ and R. J. M. Dunbar. 1997. “Social Markers and the Evolution o f
Reciprocal Exchange.” Current Anthropology 38: 93-99.
Nowak, M. A., and K. Sigmund. 1998. “Evolution o f Indirect Reciprocity
by Image Scoring.” Nature 393: 573-577.
Oftedal, O. T. 2012. “The Evolution o f Milk Secretion and Its Ancient
Origins.” Animal 6: 355-368.
Pagel, M„ and R. Mace. 2004. “The Cultural W ealth o f Nations.” Nature
428: 275-278.
Penn, D. J., and J. G. From m en. 2010. “Kin Recognition: An Overview
o f Conceptual Issues, M echanisms and Evolutionary Theory.” In
408
social research
Animal Behaviour: Evolution and Mechanisms, edited by P. Kappeler,
55-86. Heidelberg: Springer.
Pinker, S. 2010. “The Cognitive Niche: Coevolution o f Intelligence,
Sociality, and Language.” Proceedings o f the National Academy o f Sciences
107 (Supplement 2): 8993-8999. doi: 10.1073/pnas.0914630107.
Port, M., P. M. Kappeler, and R. A. Johnstone. 2011. “Communal Defense
o f Territories and the Evolution o f Sociality.” The American Naturalist
178 (6): 787-800.
Rosenberg, A. 1992. “Altruism: Theoretical C ontexts.” In Keywords in
Evolutionary Biology, edited by E. F. Keller and E. A. Lloyd, 19-28.
Cambridge: Harvard University Press.
Russell, J. A., and G. Leng. 1998. “Sex, Parturition, and Motherhood with­
out Oxytocin?”Journal o f Endocrinology 157: 343-359.
Scott-Philips, T. C. 2007. “The Social Evolution o f Language, and the
Language o f Social Evolution.” Evolutionary Psychology 5: 740-753.
Sear, R., and R. Mace. 2008. “W ho Keeps Children Alive? A Review o f the
Effects o f Kin on Child Survival.” Evolution and Human Behavior 29 (1):
1-18. doi: http://dx.doi.Org/10.1016/j.evolhumbehav.2007.10.001.
Sellen, D. W. 2001. “Comparison o f Infant Feeding Patterns Reported
for Nonindustrial Populations w ith Current Recom mendations.”
Journal o f Nutrition 131: 2707-2715.
Singer, P. 1981. The Expanding Circle: Ethics and Sociobiology. New York:
Farrar, Straus, and Giroux.
Sober, E., and D. S. W ilson. 1998. Unto Others: The Evolution and Psychology o f
Unselfish Behavior. Cambridge: Harvard University Press.
--------- . 2011. “Adaptation and Natural Selection Revisited.” Journal o f
Evolutionary Biology 2 4 :462^168.
Tal, I., and D. Lieberman. 2007. “Kin D etection and the Development
o f Sexual Aversions: Toward an Integration o f Theories o f Family
Sexual Abuse.” In Family Relationships: An Evolutionary Perspective,
edited by C. A. Salmon and T. K. Shackelford, 205-229. New York:
Oxford University Press.
Tooby, J., and I. DeVore. 1987. “The Reconstruction o f Hominid Evolution
through Strategic Modeling.” In The Evolution o f Human Behavior:
The E vo lu tio n o f G enerosity
409
Primate Models, edited by W. G. Kinzey, 1 8 3 -2 3 7 . Albany: State
University o f New York Press.
Trivers, R. L. 1971. “The Evolution o f Reciprocal Altruism .” Quarterly
Review o f Biology 46: 35-57.
Webber, C„ and A. Wildavsky. 1986. A History o f Taxation and Expenditure in
the M odem World. New York: Simon and Schuster.
W est, M. D. 2003. “Losers: Recovering Lost Property in Japan and the
United States.” Law and Society Review 37: 269-424.
West, S. A., A. S. Griffin, andA. Gardner. 2007. “Evolutionary Explanations
for Cooperation.”Journal o f Evolutionary Biology 17: R661-R672.
W est, S. A., C. El Mouden, and A. Gardner. 2011. “Sixteen Common
Misconceptions about the Evolution o f Cooperation in Humans.”
Evolution and Human Behavior 32: 231-262.
W illiam s, G. C. 1966. Adaptation and Natural Selection: A Critique o f Some
Current Evolutionary Thought. Princeton: Princeton University Press.
Wynne-Edwards, V. C. 1962. Animal Dispersion in Relation to Social Behavior.
London: Oliver and Boyd.
410
social research