The Evolution of Altruism
By ERIC STRONG
ew theories in the history of science have revolutionized an entire field as Darwin’s theory of evolution
by natural selection has done for biology. Natural selection is the process by which an environment
affects the genetic composition of a population over many generations. First, genetic mutations within
the individuals of a population arise through errors in DNA replication. Some fraction of these mutations
will lead to measurable differences in the traits comprising the affected individual. Those traits that
increase their host’s reproductive success will ensure their greater representation in successive
generations, thus slowly eliminating those traits that are less able to aid in an individual’s ability to
reproduce. As generations pass, this process can significantly alter the relative proportion of traits
within a population, causing the population to become more adapted for a particular environment. Despite
the overwhelming success Darwin’s theory has had in explaining a wide variety of natural phenomena, great
debate continues over the theory’s application in explaining the evolution of an aspect of animal
behavior known as altruism.
Altruism is the deliberate sacrifice of a portion of an individual’s reproductive capacity in order to
increase that of another. This reproductive capacity is more often described as an individual’s genetic
fitness, and is precisely defined as the contribution an individual makes to the gene pool of succeeding
generations relative to the contributions of other individuals within a population. Thus, an altruist is
defined as an individual who decreases his own genetic fitness to increase the fitness of another. The
concept of altruism is best understood through example: an African wild dog voluntarily “babysitting” the
pups of a pack, while the pack’s hunters search for food ; a bird giving an alarm call to warn others of
an approaching hawk, and thus drawing attention to itself in the process ; a man jumping into a swimming
pool to save a drowning stranger. While these acts obviously require different levels of sacrifice on
the part of the altruist, on average we can expect all of them to decrease his number of expected
This is where the paradox of altruism arises. If, by definition, altruism reduces an individual’s
fitness, we should expect Darwin’s natural selection to select against the altruistic trait and
eventually reduce its representation within a population to zero. Even if a population existed that
contained only altruists from the beginning, it would be vulnerable to subversion from within, whereby a
single, mutant selfish individual could exploit the altruistic tendencies of his neighbors and eventually
drive the altruistic trait to extinction.
Although the problem of altruism was largely ignored by early evolutionary theory, over the past several
decades it has risen to become a central issue in the debate over the level at which natural selection
operates - whether that be the level of the gene, individual, kin group, or even an entire population.
Numerous theories have been proposed to explain this phenomenon, several of which are discussed bellow.
The first mechanism by which altruism can evolve does so under the process of group selection
The following model of E.O. Wilson’s is a good example. Consider a population of N individuals. Within
this population, their exists two types of individuals: an altruistic type (A) and a selfish type (S) who
exist in proportions p and (1-p) respectively. Due to the altruistic nature of A-type individuals, each
member of the population will experience an increase in offspring equal to the product of a value b (a
measure of relative offspring gain) and the proportion of the population’s “total available altruism” that
the member will benefit from. For S-types, this value will equal b* [Np / (N-1)], while for A-types it
will equal b* [Np-1 / (N-1)], the difference reflecting the fact that altruists cannot benefit from their
own altruism. If we then assign values to c (A-types’ decrease in offspring due to their altruism) and
X (the number of offspring each individual can expect in the absence of altruism), we can then express
the average offspring expected, W, of each type in the presence of altruistic behavior.
Since b and c are always positive, and N is always greater than 1, we can immediately see that selfish
individuals will always have more offspring within a given population than altruistic individuals. A
specific numerical example is shown below using a hypothetical population of 100 individuals in which
half are altruistic and half are selfish. N’ is equal to the total number of individuals in the next
generation and p’ is the proportion of A-type individuals in the next generation.
In this example, we can see how the total percentage of altruistic individuals will decrease after one
generation from 50% to 48%. Iterating this same mathematical procedure over many generations, we can see
how the percentage of altruistic individuals within this population will soon become very small. If the
population is near its carrying capacity and is at steady state, this example suggests that the
altruistic trait within this population will disappear altogether. This is merely a mathematical model
to describe the apparent paradox discussed above - how can altruism possibly evolve if nature always
selects against it?
This paradox can be resolved with a slight modification of our model. In this case, we have two
independent groups of equal size, differing only in their relative proportions of altruistic and selfish
individuals. This example is shown here, using the same values for X, b, and c.
Although the proportion of altruistic individuals declines in each group when taken separately, if both
groups are examined together, their overall proportion actually increases. This counterintuitive result
arises from the condition that the group with the greater number of altruists is also the most productive.
Thus, as long as a population contains multiple groups with differing proportions of altruistic and
selfish individuals, and the groups occasionally mixed for reproductive purposes, the total proportion of
altruists could increase indefinitely. Despite this model which demonstrates how an altruistic trait
could increase its representation if already established, group selection theories fail to address the
issue of how such a trait could arise in the first place. For this, we must turn our attention to
another mode of selection.
One of the great problems of Darwin’s theory of evolution was its apparent inability to explain the
development of sterile worker castes among the social insects. How could such a caste evolve if its
members could leave no offspring to propagate the sterile worker traits? This problem was overcome by
invoking a special category of natural selection known as kin selection. The basic premise of kin
selection is as follows: An individual can maximize the representation of his genes in succeeding
generations by either increasing his own personal genetic fitness or by increasing the fitness of his
relatives, as relatives are likely to share a great number of his genes. To fully understand kin
selection, it is first necessary to introduce the terms relatedness and inclusive fitness. The degree
of relatedness between two individuals is simply the fraction of genes they are likely to share due
to common ancestry. For example, the relatedness between a parent and child is ½ since a child will
inherit half of his genes from each parent. The relatedness between two siblings is also ½, between a
grandparent and grandchild is ¼, and between first cousins is 1/8. Inclusive fitness is then defined as
“the sum of an individual’s own fitness plus the sum of all the effects it causes to the related parts
of the fitnesses of all its relatives” (Wilson, 1980). Thus, although a sterile insect may have a zero
personal fitness, its inclusive fitness can actually be quite high if it devotes its life to providing
for its relatives.
One can then use the ideas of relatedness and inclusive fitness to determine when altruism between kin
might arise in nature. For example, suppose you are a blackbird who suddenly notices the approach of a
hawk. If you give off a warning call to your neighbors, even if the call alerts the hawk of your
location and you eventually perish, your waning could save a reasonable number of close relatives in the
surrounding area. Therefore, the gene (or set of genes) encoding for the alarm call trait could still be
successful under the pressures of natural selection since your inclusive fitness would be greater than
that of individuals who do not warn relatives of approaching predators. But what exactly is meant by a
“reasonable” number of close relatives?
W.D. Hamilton solved for this mathematically, deriving what is now known as Hamilton’s rule. Hamilton’s
rule states that in order for altruism to occur between two individuals, the condition k>(1/r) must be
met when k is the ratio of the relative’s gain in fitness to the altruist’s loss of fitness, and r is the
degree of relatedness. In the case of two brothers, one brother will give his life for the other, if
that sacrifice will more than double the representation of the other’s genes in the next generation
(since r=½). This rule can also be applied to situations involving more than two individuals such that
an individual will also give his life to save the lives of two of his brothers, four of his grandchildren
(r=¼), or eight of his first cousins (r=1/8). These are extreme cases where the altruist’s personal
fitness is reduced to zero (by death), but Hamilton’s rule still holds true for more subtle cases of
To better understand, let’s take a look at the following example. You are a rabbit searching for food in
the woods when you come upon 9 carrots. Let’s suppose that you can only eat 4 carrots, and that each
carrot consumed gives the consumer an arbitrary payoff of 2 units. If you keep the discovery secret and
consume the carrots alone, your total payoff will be 4*2=8 units. However, if in the vicinity are your
brother and your brother’s friend (no relation to you) and you call them over to share in your discovery,
now the carrots will be split evenly among all three. Now the total inclusive payoff to your genes will
be (3*2)(1) + (3*2)(½) + (3*2)(0) = 9 units. Despite your apparent loss of one carrot, and a
non-relative’s gain of 3 carrots, in this particular case, it will still benefit your genes to give the
food call and share your find.
One might be tempted to ask, how is it that a rabbit can either determine the complex kin relationships
between other neighboring rabbits, or perform the necessary mathematical calculations to determine
whether it should or should not behave altruistically in a given situation? The obvious answer is that
it can’t. Genes are not clairvoyant and cannot necessarily detect copies of themselves in other
individuals. However, they can be programmed to act favorably to other individuals who are more likely
to contain copies of those same genes. There are multiple ways by which this can be accomplished. The
first is known as the “green beard effect” as coined by Richard Dawkins. If a gene arose in a
population that both gave individuals a conspicuous physical trait (such as a green beard) and a
tendency to act favorably to others with that same trait, mutual altruism between green bearded
individuals could evolve. In this particular case, one can see that it is not necessary for
individuals to determine whether their identical genes are due to common descent or simply coincidental
random mutation, as such a distinction would be a waste of time from the altruistic gene’s point of view.
Although the green beard effect is a theoretical possibility, even Dawkins admits it is somewhat
unlikely that a single gene will produce both a very distinguishable trait (such as a green beard) and
Another way by which individuals are more likely to act altruistically to others who were genetically
similar is by associating familiarity with kinship. Mothers can certainly know the exact relationships
to their children, but siblings can infer the relationships amongst themselves also, since individuals
who are raised together are more likely to be kin than non-kin. Related to this method of kin
recognition is one dependent on similar location between individuals. For instance, if you are a member
of a population that does not move around much, it is likely that many of your neighbors are also kin.
Thus, a gene encoding for altruism could act not necessarily on kin, but rather more indiscriminately on
any of your close neighbors. This kind of mechanism can be used to explain altruism within troops of
monkeys and pods of whales, since any two members in these groups are likely to be highly related.
Although these approximate forms of kin recognition are likely to be the prime forces in kin selection, it
should not be ignored that studies have shown that in certain cases closer relatives are sometimes treated
preferentially over non-relatives, independent of other environmental factors such as proximity. An
explanation of this phenomenon is not yet clear.
Although kin selection provides a powerful explanation of altruism in much of the natural world, there
are occasions (particularly in higher mammals) where this form of selection fails to explain altruism
occurring between two unrelated individuals. To understand these situations, Robert Trivers has
introduced the idea of reciprocal altruism. Reciprocal altruism is the process by which one individual
will commit an altruistic act towards another in exchange for the second individual returning the
altruistic favor at some later time. The end result of this exchange leaves both with an increased
Let’s suppose a man is drowning, and an unrelated woman from nearby jumps in to rescue him. If not
helped, the man has a ½ chance of drowning, as opposed to the woman who has a 1/20 chance of drowning if
she attempts to save him. One might be inclined to conclude that she will never attempt a rescue, since
the immediate effect would be a decrease in both her personal fitness and her inclusive fitness (since
the man is not a relative). However, if at some later point in time the woman is drowning, and the man
is nearby, he might be inclined to rescue her in return for her original help at the same personal risk
(1/20) that the woman experienced when she saved him. We can now see that both people have exchanged a
½ chance of death for a 1/20 chance. Thus, a population full of altruistic individuals who enter into a
series of reciprocal exchanges will increase their total genetic fitness as compared to competing
population devoid of such altruism.
A problem with this idea arises however, when one asks why does the man in our above example bother to
reciprocate? If his life has already been saved, what reason does he heave afterwards to continue to
risk it unnecessarily? In human societies, this answer is simple - the cheater will be punished by both
the individual he failed to reciprocate with and also be the greater society. The cheater will
experience a greater decrease in fitness due to cheating than he would by reciprocating, and natural
selection will select against him. However, in most species, the level of recognition and memory
required for this type of selection is not possible, and this is the main reason reciprocal altruism is
relatively infrequent in the natural world as compared to altruism due to kin selection.
Perhaps the most discussed example of altruistic behavior in animals is that dealing with alarm calls.
As a predator approaches an area occupied by more than one individual of a prey species, the individual
who first detects the predator may give an alarm call to alert its neighbors, drawing attention to
itself in the process. As drawing attention to oneself in this situation is obviously disadvantageous,
an explanation for this phenomenon has been actively sought. P.W. Sherman hypothesizes that any given
alarm call may have one (or a combination of several) of the following six functions:
The alarm call may cause the caller’s neighbors to either aggregate or act nervously, thus actually drawing attention away from the caller.|
The call may discourage prey from a pursuit by alerting it that it has been spotted and has lost the element of surprise.|
Alarm calling might reduce the probability of later attacks by the same predator, if a predator is more likely to hunt a particular species of prey that has given it success previously.|
An individual might give an alarm call if those benefited are likely to return the favor in a form a reciprocal altruism.|
Alarm calling might evolve if prey populations consisted of multiple groups with differing proportions of individual with alarm calling tendencies as a result of group selection. (Similar to the previously discussed example.)|
Although a caller jeopardizes its own life in the face of immediate danger, if its neighbors consist of relatives, the call may aid in their escape, thus increasing the caller’s inclusive fitness.|
We can immediately conclude that theoretical functions 1, 2, and 3 would not classify as altruism because
the caller would also benefits from the action. Although function 4 may sound reasonable, Trivers raises
several objections to this possibility. First, it is difficult to see how a group of individuals would
identify and discriminate against a single non-reciprocating “cheater”. Second, no evidence has been
found to support the idea that animals in the wild refrain from sounding alarm calls because their
neighbors fail to reciprocate. Function 5 is nearly identical to the mathematical model of group
selection discussed above. This leaves function 6 to examine more in depth.
Sherman has shown kin selection to be a powerful force in the evolution of altruism in Belding’s ground
squirrels. After studying a population of approximately 1100 individual squirrels over 3000 hours of
observation, Sherman found that female squirrels who were currently in their reproductive years were
both more likely to give an alarm call than would be predicted by chance alone, and more likely to give
an alarm call than a control group of non-reproductive females. Although no significant disagreement
was found between reproductive and non-reproductive males, Sherman argues that this is merely the
result of both the matrilineal kin structure of Belding’s ground squirrels and the observation that the
female is the more parental sex in this species.
As best put by E.O. Wilson, “Other than suicide, no behavior is clearly more altruistic than the surrender
of food.” Yet this form of altruistic behavior occurs frequently in a wide variety of animal species.
Both gibbons and chimpanzees have been observed to offer food to others after solicitation, and African
wild dogs are known to carry fresh meat back to the “babysitter” of their cubs after a successful hunt.
One of the most well documented studies of food sharing however was done by Gerald Wilkinson on the
vampire bat. Vampire bats feed on blood, and although a single bat will not always be successful in
finding prey on any given night, if it does, its payoff will be big, and often bats will suck up a
surplus of blood. Upon the bats’ return to the group, it is often seen that the successful hunters will
regurgitate blood for the consumption of their less fortunate neighbors. While this sharing of food is
often between relatives (usually a mother and child), there are numerous occasions when the two
individuals involved in the exchange are not related. Wilkinson hypothesized that such generosity could
be the result of reciprocal altruism if the following three conditions were met:
Contact between members of a group of bats must be long enough to ensure many such exchanges |
The gain of receiving aid must be greater than the cost of donating.
Altruistic donors must be able to recognize those bats that had previously failed to reciprocate, and refuse to feed them.
The first and second conditions were shown to be true from simple observation of wild bat populations.
Wilkinson demonstrated that the third condition was also met. He selectively removed certain bats from
the population and starved them. When the bats returned, the population was significantly more likely
to feed these bats (who had likely shared food with them in the past) than they were with a control
group of starved bats who were introduced to the population despite no previous contact with them. One
can therefore conclude that food sharing in vampire bats is at least partly the result of reciprocal
Altruism in Humans
It is only logical that one might now be inclined to take these theories of altruism and apply them to
our own species, especially considering that man is the most altruistic animal of all. Altruism
pervades every aspect of our society, whether in the form of food sharing, helping the sick, gift
giving, or even the sharing of knowledge through education. The use of money may even be seen as the
prime example of reciprocal altruism in modern man, as money itself has no intrinsic value other than
the assurance that another person will trade valuable goods for it at some point in the future.
It is easy to see how altruism in man could have evolved through kin selection. Bands of early hunter-
gatherers during prehistory were almost certainly composed mainly of close kin. Also, through language
and an increased mental capacity, early man had a much greater ability than other primates at not only
recognizing kin, but also at distinguishing between subtle differences in degrees of relatedness. As
evidence of the operation of altruism by kin selection, J. Rushton has showed that humans are more
likely to favor not only those they definitely recognize as kin, but also those with whom they share
genetic traits. Through a series of extensive studies, he demonstrated that on average there was
greater genetic similarity between friends then between strangers, and he also made the remarkable
discovery that sexually interacting couples were more likely to share similar blood markers than would
be expected by chance alone. The idea that human behavior between two individuals may be governed by
genetic similarity is profound, and its implications stretch far beyond a discussion of altruism.
Although there is no doubt that kin selection plays an important role in the formation of altruism in
human societies, reciprocal altruism is generally more prominent. Our species easily fits the
conditions for the rise of reciprocal altruism: long lasting relationships, an increased memory to
distinguish reciprocators from non-reciprocators, and a method of punishing non-reciprocators. It has
been hypothesized that some of man’s more complex emotions may have evolved to improve upon the system
of reciprocal altruism. Gratitude and sympathy may increase one’s chances of receiving altruism by
implying an increased chance of reciprocation, while guilt serves to discourage the non-reciprocator
and causes him to demonstrate that he does not plan to refuse reciprocation in the future. On top of the
development of emotions, Trivers even suggests that in a complicated form of coevolution, the combined
selective pressures of finding subtler ways of cheating one’s neighbors and increasing one’s ability to
detect such subtle cheaters may have contributed to the expansion of man’s mental capacities and led to
his current state of high intellectual ability.
Altruism is one of the great mysteries of social behavior in animals, as it appears to contradict our
understanding of natural selection. Even one hundred years after the birth of Darwinism, scientists are
still continuing to debate the causes and effects of altruistic behavior. Whether the mathematical
models of group selection, the instinctive qualities of kin selection, or the trusting attributes of
reciprocal altruism, are the prime explanations of the development of this behavior is still largely
unknown. In the end, it will probably be found that it is the combination of all three possibilities that
plays a significant role in the natural world.
*Note: I have chosen to use the controversial and often misunderstood term “group selection” here for
simplicity to refer to any process of selection which operates at a level above that of the kin group.
A more precise term to describe the following hypothetical mechanism of evolution of altruism would be
interdemic selection, which is also sometimes referred to as interpopulation selection. For further
clarification of this distinction, see J. Maynard Smith’s 1975 paper “Group Selection” in the Quarterly
Review of Biology and E.O. Wilson’s textbook, Sociobiology.
Begon, M, Harper, JL, Townsend, CR. Ecology. 1996.
Barlow, C. From Gaia to Selfish Genes. 1997.
Dawkins, R. The Selfish Gene. 1989.
Hamilton, WD. “The genetic evolution of social behavior (I&II;)”. Journal of Theoretical Biology. 1964, 7:1-52.
Maynard Smith, J. “The theory of games and the evolution of animal conflict”. Journal of Theoretical Biology. 1974. 47:209-221.
Maynard Smith, J. “Group Selection”. The Quarterly Review of Biology. 1975. 51:277-283.
Rushton, JP. “Genetic similarity, human altruism, and group selection”. Behavioral and Brain Sciences. 1989. 12:503-518.
Sherman, PW. “Nepotism and the Evolution of Alarm Calls”. Science. 1977. 197:1246-1253.
Simon, HA. “A mechanism for social selection and successful altruism”. Science. 1990. 250:1665-1668.
Trivers, RL. “The evolution of reciprocal altruism”. The Quarterly Review of Biology. 1971. 46:35-55.
Wilkinson, GS. “Reciprocal food sharing in the vampire bat”. Nature. 1984. 308:181-184.
Wilson, DS. “Levels of Selection: An alternative to individualism in biology and the human sciences.” Social Networks. 1989. 11:257-272.
Wilson, EO. Sociobiology. 1980.