Human Kinship as a Green Beard (PDF)

1
Human Kinship as a Green Beard
Henry Harpending and Nathan Harris
University of Utah

DRAFT: March 2015
Prepared for Darwin’s Bridge, edited by Joe Carroll, Dan McAdams, and E.O. Wilson, Oxford
UP, in press.

2

Human Kinship as a Green Beard
Introduction
Humans have a lot of difficulty with ethnic diversity. Some have speculated that ethnic
discomforts are somehow based on or derived from genetic differences: in raw form the question
arises whether “racism is in our genes.” A recent article in the The Atlantic by Robert Wright
gives assurances that it could not be (Wright 2012). Wright’s assurances are based on a
widespread but not well supported argument about the human past: that we never encountered
people significantly different from us until the last few thousand years. Consequently, the
ecological conditions that could have favored ethnic discrimination never occurred in our
evolution. That argument demands a particular ecology and demography in our past, and it
demands that evolution takes place over very long time periods, hundreds of millennia or longer.
There are grounds to doubt both fundamental premises of the argument. One premise is that in
our past (whatever period that is supposed to occupy) humans were essentially a uniform carpet
of sessile groups with substantial ongoing gene exchange with neighboring groups. Another
premise is that the relevant past is tens to hundreds of thousands of millennia.
We know that there are cases in which genetically different ethnic groups experience no apparent
discomfort with each other, for instance the Parsi in India. Other Indians regard the Parsi as a
national resource (Nelson 2012). There are also familiar cases of ethnic antagonism between
genetically identical groups, for instance Mormons and non-Mormons in Utah.
The notion of “in our genes” deserves some attention, especially since “genetic” is often
contrasted with “learned” or “environmental.” We all learned our spoken language; many but not
all of us are reflexively afraid of snakes; and many of us learned calculus. Language never had to
be taught since humans are “prepared” to learn language. Many argue that we are also prepared
to fear snakes, but calculus is an entirely different matter. We had to sit and think hard about it
and spend hours doing exercises that molded what understanding we gained.
The issue in this paper is whether or not we are easily taught ethnic antagonism. Is it like
language, so easily learned that it never need be taught, or like calculus, which must always be
taught with great effort, or somewhere in between?
We could presumably answer this question with a properly designed experiment, but that is
never going to occur. We all correctly regard experimentation like that with humans as
unethical. An alternate approach, used in most of the sciences, is to make a model and see what
the model predicts. A good model should tell us whether or not we ought to have ethnic
antagonism built-in, i.e. easily learned, or never built-in, like calculus. It ought to tell us the
conditions under which one or the other might have occurred in our past, and best of all, it ought
to suggest to us ways to falsify the predictions of the model.
Genetic Theory

3

W.D. Hamilton (1964a; 1964b) created a revolution in behavioral ecology with a pair of papers
that defined and clarified what today is called indirect or extended fitness. If I take care of my
child, I not only augment that child’s Darwinian fitness but also my own, in proportion to the
amount of my genome shared with that child. The early version of Hamilton’s theory was
described in terms of the “coefficient of relationship,” the fraction of shared genes between an
actor and the recipient. In the case of my child this is conventionally thought to be one half.
While many were immediately enthusiastic about the application of this theory when it was first
published, especially entomologists who deal with strange variants of sexual reproduction, others
thought it was trivial and obvious, so they paid little attention. A colleague remarked to me in the
1970s, that “we don’t need inclusive fitness theory to understand why parental care evolved.”
The research of Daly and Wilson (1988) helped inclusive theory gain traction and respect in
anthropology and the social sciences. They demonstrated that children living in households with
adult males who were not their biological fathers were at much greater risk of severe
mistreatment and even murder.
It soon became apparent that “fraction of shared genes” is a slippery idea. Modern population
genetics bypasses the old distinction between “identity by descent” and “identity in state” since
evolution acts on the products of genes, and identical genes are simply identical or not,
regardless of the origin of their identity. It is true enough that I share half my autosomal genes
with my child, but if I am inbred then I share slightly more than one half since the inbreeding
means that there are correlations between the alleles at any locus in me. Hamilton himself came
to appreciate this and as a result extended his ideas beyond close family relationship to more
distant relationship, for example within ethnic groups in a multi-ethnic context (Hamilton 1975).
For diploid organisms, like humans, the difficulties with relationship are easily resolved by
reasoning about the coefficient of kinship between two organisms (Bulmer 1994) defined like
this: Pick a random allele from a locus in individual A, pick a random allele from the same locus
in individual B, and ask whether or not they are identical. Then ask what the frequency of the
allele is in the population as a whole, and call that p. Then the probability that the gene from B is
identical to the gene chosen from A is
pB = pAFAB + p(1-FAB)
leading to an estimate for F from this locus as
FAB= (pB-p)/(1-p).
This estimate can have bizarre properties. For some possible genotype configurations it is nearly
unbounded above 1 and below -1, while the “real” value of FAB must be between +1 and -1. On
the other hand averaged over hundreds of thousands of loci it returns correct results. Averaging,
the kinship between person A and person B at a locus is computed as
(pA − pB)2 /p(1 − p)

4
where pA and pB are the frequencies in the two individuals, with possible values 1, 0.5, and 0,
and p is the population average (Yang et al. 2010; Davies et al. 2011). With many loci, hundreds
of thousands in the present case, one sums the numerators of the above expression of loci, sums
the denominator, and computes the ratio. Yang et al. (2010) and Davies (2011) give the
computational formulae.
I have used single nucleotide polymorphism genotype data from the Human Genome Diversity
Project (HGDP) project (Li et al. 2008) to examine kinship between individuals in several
populations from the database. This was a loosely coordinated project to collect genetic samples
from several dozen human population around the world. In each population, a few dozen or more
individuals were sampled. Samples were included from large cosmopolitan populations, like the
French, Japanese, and Mormons from Utah, as well as from small isolated tribal populations like
the Surui and Karitiana of the Amazon basin. Individual populations samples are of more than 20
and fewer than 50 individuals. There are several recently admixed populations: a sample of
Uigurs from central Asia and a sample of American Blacks from the US Southwest.
Pairwise kinship between two individuals immediately informs about the genetic “interest” that
one individual should share with the other. If the social and economic system is such that
individuals can reward (or harm) each other according to their genetic interests, then the
prediction is that dispositions for discrimination and nepotism, according to shared kinship,
could have evolved, such that discrimination and nepotism would be easily learned or “natural.”
Two important questions to ask are (1) whether there are enough differences in kinship between
individuals of unknown genealogical relationship in populations to have favored cryptic kinship
recognition by natural selection, and (2) whether we have mechanisms available to us that could
detect cryptic kinship.
Cultural Background
Since the 1970s, many evolutionary social scientists have believed that technologically primitive
groups, especially foragers, offer clues to our species’ proclivities and potentials. The reasoning
was that since we were hunter-gatherers for 99% of our evolution, we must have been somehow
“shaped” by evolution to be like them. Some have called this the search for “Mr. Natural,” while
others refer to this hypothetical ancestral state as the “environment of evolutionary adaptedness”
or EEA. Although the rationale is subject to criticism, the search produced a lot of useful and
interesting ethnography, ethnography addressing evolutionary rather than simply social
questions. Where does food come from? How does reproductive competition work itself out?
What determines the fitness of individuals?
In retrospect that are several weaknesses with EEA theory. First, what does “99% of our
existence” mean? If clearly modern humans appear at least 40,000 years ago and if many of our
ancestors have practiced agriculture for 10,000 years, then they were foragers for 75% or our
existence, not 99%. The 99% argument requires that we are living relics of life as Homo erectus
or even some flavor of Australopithecines, for the last million years.
Second, EEA theory needs there to have been some identifiable EEA, but the archaeological and
ecological record of the past million years shows much diversity, and its commonalities fail to

5
answer the most basic questions about this social environment. For example, were males in the
EEA paternal males who provisioned their own offspring? Were they instead competitive
violence-prone gaudy males who provided little in the way of subsistence for women and
children? In Europe in the several millennia before the last glacial maximum, we find upper
Paleolithic remains with many signatures of the latter, and Mesolithic remains looking like the
former.
Two candidates for our EEA, for “Mr. Natural,” emerged in the textbooks of the 1970s. One was
!Kung Bushmen of the northern Kalahari in Africa (Lee & DeVore 1976), and the other was the
Yanomamo of the Amazon basin (Chagnon 1968). They could hardly have been more different.
The Bushmen were laid back, relaxed people seemingly tailor-made for the popular culture of
the time—Kalahari hippies with beads and sandals even. The males were not belligerent, not
loud, and they put in a lot of time in resource acquisition that was distributed not only to their
immediate families but also to every member of the local band. Women gathered and
provisioned their own children and husbands with the harvest. Nuclear families were durable.
Polygyny was certainly possible but it was ordinarily not real reproductive polygyny: it was
more likely an older widow living with her sister and her brother-in-law as a “co-wife.” Marital
distances (distances between birthplaces of husbands and wives) were large, so the population
was outbred. Groups separated by hundreds of kilometers displayed only small genetic
differences. Population was stationary, a significant finding in that era of concern about world
population. They were no threat to the environment. Settlements, called “bands,” were not
persistent social units, but families were. Yellen and Harpending (1972) show maps of where
families of one band had lived in the last year, with no apparent correlation between one family
and the next.
The Yanomamo were in many ways opposite to the !Kung. They were engaged in chronic local
warfare and raiding, so it seems no one was very relaxed. Pair bonds were unstable and males
were not focussed very much on provisioning. With their transient gardens (swiddens), they were
a real threat to the environment since abandoned swiddens in tropical environments take a long
time to recover. The rate of reproduction was so high that the population was approximately
doubling each generation. The Yanomamo were predatory expansionists with gaudy loud violent
males. There was a substantial payoff to male violence. Males who had killed someone enjoyed
more reproductive success than males who had never killed anyone (Chagnon 1988). Marital
distances were small. Villages were persistent clusters of related people. A consequence is that
villages were genetically distinct from each other, even with the small set of genetic markers
available at the time.
These two models for our EEA provide dramatically different models of the social contexts for
the evolution of ethnic antagonism and kin nepotism. In a Bushman world, with its stationary
populations and high outbreeding, there is no ethnic differentiation. Two people separated by
several hundred kilometers are scarcely more different from each other than two people from the
same settlement. In a Yanomamo world, on the other hand, the landscape is one of kaleidoscopic
expansions and extinctions. In a lifetime, people from one group would ordinarily have violent
interactions with people from very different groups. In their studies of the Upper Midwest before
European contact, George Milner and his colleagues provide an excellent archaeological

6
example of this world (Milner et al. 1991). By the time of Columbus, struggles with invading
Oneota had led to near depopulation of the Upper Midwest.
Such expansion, disruptions, and extinctions are familiar to us from history: the European
invasion of the New World, the Bantu and Han expansions, the Indo-Europeans, the Mongols,
and others. We have no good reason to think that the same pattern did not occur further back in
human history. Patterns are provided by the Oneota case, the Numic expansion in the American
West, the arrival of Navaho and Apache in the American Southwest, and the expansion of Inuit
at the expense of the Dorset in the arctic.
Evidence from the Genome
The HGDP contain several samples from large cosmopolitan populations like Mormons from
Utah, French, Japanese, and two Chinese samples); and other samples from small isolated
populations and from ethnic groups within larger polities. Details of the samples are not
available because of concerns about protecting the privacy of those who furnished the samples.
We are forced to assume that the samples are not biased in any obvious ways, but we will see
that the occasional pairs of relatives to turn up in them. To determine the extent of cryptic
kinship, and thus to see whether natural selection could favor recognition of kin, we can simply
look at histograms of kinship between all possible pairs of individuals in the HGDP samples. I
first consider pairwise kinship in the samples of French and Japanese.

Figure 1 Kinship among French. The top panel is a histogram of kinship between all pairs of
individuals including kinship with self, shown by the small number of comparison at F ≈ 0.5.
The middle panel shows, for each person in the sample, kinship with his closest kin in the
sample. Since overall mean pairwise kinship must be zero, and since kinship with self around

7
F=0.5 dominates small samples such as this, kinship in the middle panels has been adjusted to a
mean of zero.

Figure 2 Kinship among Japanese. See the caption of figure 1 for details about the panels.
Recalling that kinship with self is roughly 0.5, so that kinship with a child is 0.25, a grandchild
0.125, a grandchild’s grandchild is 0.006, and so on, it is immediately apparent from figures 1
and 2 that there is not enough diversity in kinship with other members of the same group to have
ever favored cryptic kin recognition or spontaneous nepotistic groups. In these large
cosmopolitan groups, the “best” a person can do (or “worst” if one is to harm someone) is so
small in absolute value that there is essentially no fitness payoff to cryptic kin recognition. In
other groups in the HGDP sample like Utah Mormons or Han Chinese, the pattern is the same.
Things change in diverse societies. Figure 3 shows the distribution of kinship between all pairs
of individuals in a (synthetic) society created by simply pooling the French and Japanese
samples. Here a random person is related to you either as your grandchild or as your negative
grandchild according whether that person is of your own ethnicity or of the other. Aiding
someone like you is equivalent to aiding your grandchild. Worse, harming someone unlike you is
also like aiding your grandchild. In a group like this, ethnic nepotism would be positively
selected, fostering all the familiar difficulties humans have with diversity.

8

Figure 3 Kinship between pairs in a synthetic population of pooled French and Japanese. See the
caption of Figure 1 for details.
The picture so far is that in large homogeneous societies, kin recognition would have no fitness
payoff; in diverse societies, the fitness payoff to nepotism and ethnocentrism would be
immediate and substantial. Is this relevant to understanding our species’ propensity for ethnic
antagonism? A possibility is that ethnic antagonism is relatively new and that it is an evolved
response to life in ancient cities since large urban centers arising in the context of wide trade
networks could easily have generated ethnic mixes that mirror our synthetic mixture of French
and Japanese.
If long-range movements did in the past generate societies like our synthetic mixture, how long
did the diversity persist? There are several relevant populations to examine in the HGDP sample.
First, Uyghurs are a central Asian group descended from Tocharians, the easternmost extension
of Indo-European speakers in the Tarim basin of northwest China. They are today roughly half
descended from these Europeans and half from east Asian peoples. The second population of
interest is a groups of African-Americans from the US Southwest.

9

Figure
4. Ancestry
components
in recently
two recently
mixed
populations.
Figure
4 Ancestry
components
in two
mixed
populations.
EachEach
horizontal line shows
horizontal
line
shows
the
estimated
ancestry
of
a
single
individual
from
each
of
the estimated ancestry of a single individual from each of the ancestral populations.
This figure is
the ancestral populations. This figure is from here.
from blogs.discovermagazine.com/gnxp/.
Figure 4 shows the estimated ancestry of individuals in each of these populations
In each part of the figure individuals are represented by horizontal lines, the
Figure
4 shows
the shows
estimated
ancestrycontributions
of individuals
in each
of these populations.
colors
of which
the ancestral
to that
individual.

In each part of
the figure, individuals are represented by horizontal lines, the colors of which shows the
The Uyghurs have been mixed for a millenium or so. They are an old admixed
ancestral
contributions
to that
The
Uyghurshave
havebeen
beenhomogenized.
mixed for a millenium or so.
population,
old enough
thatindividual.
ancestries of
individuals
TheyEvery
are anindividual
old admixed
population,
old
enough
that
ancestries
of
is around half European and half east Asian. What individuals
this means have been
homogenized.
is around
halfthere
European
and half
eastdiversity
Asian. What this means is
is that theyEvery
have aindividual
“flavor of their
own” and
is not much
kinship
generated
their mixed
ancestry.
canthere
see this
by looking
at the distribution
that they
have by
a “flavor
of their
own”Weand
is not
much kinship
diversity generated by their
of ancestry.
pairwise kinship
in see
the this
Uyghur
sample inatFigure
5.
mixed
We can
by looking
the distribution
of pairwise kinship in the Uyghur
sample in Figure 5.

9