Genetics and Behavior
Genetics and Altruistic Behavior
Forrest O. Gulden
In a population of n individuals it is possible that there will be n different alleles at any single locus. If one of these individuals undergoes a mutation which creates a behavior that increases his reproductive success relative to others in the population than that behavior will spread; the individual will have more children who have more children and therefore the allele frequency will increase. Because of this there is a definite general mechanism for the development of a behavioral gene. This can be seen in several experiments.
Constraints on learning have been shown in experiments on rats performed by John Garcia. In the experiments, rats were easily able to associate a stimulus with a particular punishment (drink with illness and audible noise with pain) but were totally unable to make the converse associations (drink with pain, noise with illness). This is made clear only when the evolutionary history of the rat is considered. Unable to vomit, any rat which is able to 'sample' a food or drink and make value judgments on its edibility will leave more offspring than one which is perpetually sick as a result of eating tainted meat or drinking poisoned water. At the same time, any rat able to associate the warning calls of another rat with the potential for injury will leave more offspring than one unable to make this connection. This provides a path for understanding the Garcia experiment (through a genetic basis).
Physical development leading to differential behavior can be seen to be influenced by genetics also. The most frequently cited example of this is the disease Phenylketonuria (PKU). PKU is an inborn metabolic error resulting from the inability of an individual to produce adequate amounts of the enzyme phenylalanine hydroxylase due to mutation. Phenylalanine hydroxylase degrades the amino acid phenylalanine to tyrosine under normal circumstances. If not treated through a special diet early in life, the failure of this enzyme causes a build-up of phenylalanine in the body resulting in, among other symptoms, mild mental retardation. Here then it can be seen that an individual's genetic composition can alter development which can in turn alter behavior.
The expression of behavior is also influenced by the genetic make-up of an individual. The Mao A gene is one of two loci responsible for the degradation of serotonin. Serotonin levels in the blood have been correlated with aggression. A mutation in the Mao A gene can cause a partial inability to degrade serotonin. This can lead to a condition (known as Impulsive Aggressive behavior or IA) where the affected individual is unable to control his aggression at seemingly random times.
The examples given here are by no means the only ones known. Insects, animals, and humans are all studied in an effort to learn specifics of behavior. However, not all organisms are studied in the same manner. Insects are generally observed, euthanized, and genotyped. In addition, designed experiments can easily be run for multiple generations when dealing with insects (as a result of conveniently small sizes and short life spans). Animals, however, are slightly more complex. The life spans of most animals preclude multiple-generation experiments. Also, animals are much more complex behaviorally and genetically so more observation is needed and less genotypic matching is possible. It is also much more difficult to design experiments using animals due to the generally small maximum sample size.
Humans are even more difficult to study. Human researchers are much more in tune with differences in human behavior than they are with animal behaviors. In addition, there are more known human behavioral disorders and also more classifications of the differing causes of the behavior. Environment is also a very large part of any human behavior as not only the physical surroundings but also the social landscape must be taken into account. While this is true to some extent for all creatures it is much more so for humans. More, it is not possible to design experiments to test human subjects without serious ethical considerations. Finally, because of all of this, it is often necessary to isolate the particular trait of interest before studying it (ensuring that the trait of interest exists within the sample). This makes double-blind experiments impossible and leaves virtually all observations somewhat subjective.
Despite all of these problems, it is often easy to pinpoint a genetic basis for a behavior. Some behaviors are fairly easy to understand. But some behaviors are much more difficult to comprehend, and one of those behaviors is altruism. After years of research, however, it can be demonstrated that altruism is a behavioral trait capable of being spread through genetic means. The genetic models are then supported in clinical studies.
An altruistic act is defined as one that decreases the reproductive fitness of the actor while increasing the fitness of the recipient. There are different levels of altruism ranging from low cost (giving away something of little value to the actor but of high value to the recipient) to high cost (usually defined as self-sacrifice). There are also different types of altruism. "Pure" altruism is basically a gift where the actor is never repaid in any way by the recipient. False altruism involves an individual probing another with an altruistic act in hopes of receiving a bigger gift later, at which time the cooperation will end. Reciprocal altruism is what is seen most often and what is the easiest to explain. It occurs when an actor helps a recipient and is repaid by the recipient at a later time.
The problem with determining a genetic basis for altruism is that, in many situations, such a trait would be unlikely to spread. Assume that person X has a unique mutation that creates an altruism allele. This person will then act altruistically to those around him. This will increase the fitness of the other individuals in the population while decreasing the fitness of actor X. Therefore X will leave less offspring than the other individuals in the population. In the case of a unique mutation, this would result in continually decreasing frequencies of the altruism allele until it is eliminated from the population completely. Such a gene cannot spread in a population. Despite this problem, there are two recognized ways through which an altruism allele can increase in frequency within a population: through kinship and through reciprocity.
The theory of kinship is based on relatedness. Relatedness is operationally defined as, "the conditional probability that if individual p has a particular allele then so too does individual q." This refers only to identical-by-descent situations, not to allele frequency arguments. Altruism can spread solely through kinship if there is only minimal decrease in actor fitness for the first few generations and continual grouping of altruistic individuals. Put simply, if actor X is still able to have children, and then those children have children and grandchildren who remain in close proximity to one another, eventually the majority of all altruistic acts will be directed to others who are related to the actor. This will be so due to proximity arguments (individuals act more altruistically to those they interact with frequently) and due to the ability to recognize kin (which will be discussed shortly).
However, not all kin are created equal. This can be shown through relatedness calculations. A mother donates one half of her genome to her child. The odds of finding one of the mother's alleles in the child now stand at 1 in 2, depending on whether the allele was one of the 50% transferred to the egg. Therefore the relatedness between mother and child is ½. The same is true for a father-child relationship.
Siblings are much the same. The odds that a mother passes an allele to a son are 1 in 2. The odds that a mother passes the same allele to a daughter are 1 in 4. This is because ½ of the mother's genome is inherited and the odds that any particular gene will be transferred are 1 in 2, and ½ * ½ = ¼. The same is true from the father's perspective. Because of this the relatedness between siblings is ¼ (the conditional probability of having a gene from the mother in both siblings) + ¼ (the conditional probability of having a gene from the father in both siblings) = ½. By the same manners used above it can be seen that grandparents and grandchildren are related by ¼, aunts and children are related by ¼, cousins are related by 1/8, and so on.
If altruism is to evolve from kinship, it would be beneficial to the altruistic actor to be able to discriminate between offspring and siblings versus cousins and aunts versus unrelated individuals. This is because increasing the reproduction of offspring, who are related to the actor by ½, is more beneficial to the actor's genome than increasing the reproduction of cousins, who are only related by 1/8. In general, the formula for determining when it is genetically beneficial to be altruistic is based on when rB > C (where B is the benefit to the recipient, C is the cost to the actor, and r is the relatedness between the individuals). Anytime this is so the actor will leave more 'reproductive units' through the recipient than he himself would on his own. Because of this the allele can spread.
Reciprocity further helps the spread of an altruism allele. Consider a situation where two individuals are given the opportunity to either cooperate with one another or defect on one another. The payoff matrix for their actions is shown below. In the matrix the payoff values are given with respect to player A.
If player A somehow 'knew' that player B was going to cooperate, it would 'pay' player A to defect. This is because the payoff for defection is higher than that for cooperation when B cooperates. By the same token, if player A somehow knew that player B was going to defect, it would again pay player A to defect, as again the payoff would be greater. However, given knowledge of the payoff matrix both players will realize this and both will defect, giving each a payoff of one unit. This is the rational strategy for the players to use, yet it yields a lower payoff than if both would cooperate. Hence there is a dilemma; the best way for player A to maximize his success is to maximize the success of player B. A series of experiments were done to test this assumption.
In the first experiment the payoff matrix was sent to sixty-two economists, sociobiologists, and psychologists from around the world. Each was to submit a computer program to 'compete' against others in the prisoner's dilemma game. In addition, "all defect" and "all cooperate" were added, rounding the field to sixty-four. Each program faced every other program a large number of times (a variable number due to assumptions not stated here). Each program also faced itself. The winner(s) was chosen based on the payoff values gained through the competition. Several programs did quite well, but the most impressive was a simple program called 'Tit - for - Tat,' or TFT. TFT played 'cooperate' on the first turn and then mirrored its opponent's turn every successive match (e.g. if B defected on turn one TFT would defect on turn two). In essence, TFT was reciprocal in nature.
The second experiment done was an extension of the first. In this case a larger number of strategies were compared in the same manner as in the first experiment. However, instead of declaring a winner after the first round, the number of points gained was correlated to reproduction units. Therefore the more points a strategy won the more the strategy was represented in the next round of the game. This was done many times for up to fifty generations. One of the eventual winners was TFT. From this it was determined that TFT is an Evolutionarily Stable Strategy (ESS). An ESS is a strategy which, when employed by a member of a population, cannot be beaten by any other strategy in a population (neglecting the effects of drift). Why is this so?
To be evolutionarily stable, a strategy must be able to accomplish three things. First, the strategy must be able to initially reproduce within a population; it must be able to gain a foothold. Secondly, a strategy must be able to compete successfully enough to eventually be used by a large proportion of the population. Finally, the strategy must be able to withstand invasion by opposing strategies when at high frequencies. This is the case for TFT in many situations. When TFT is introduced into a population and given time to spread (e.g. the altruistic acts are negligible and do not immediately terminate the lineage) it can reside in a population successfully at low frequencies. If the strategies of high frequency consist of a majority of defectors then they will perform at a low rate against TFT and at a low rate against themselves (with a average payoff near one). TFT meanwhile will perform at a low rate against them (a payoff near one) and a higher rate against itself (a payoff near three), increasing the spread of TFT. When TFT is at a high frequency then it will be immune to invasion by majority cooperators (which will do no better than TFT in the population) and also immune to majority defectors (which will do worse against TFT than TFT does against itself). Please note that there is no cause to worry about TFT invading an all-cooperator society as all-cooperate is not an ESS (it is very susceptible to all-defect). From this analysis it seems apparent that any individual who adopts a strategy of TFT will spread the allele throughout the population (given circumstances not discussed here). Together with kinship this allows for the evolution of altruism.
The models used above demonstrate that it is possible for altruism to evolve within a population. However, these models do not demonstrate what an 'altruism gene' would do or how it would function. For this another model is needed. This model is provided by Hamilton via the Green Beard Gene. Imagine that there is an allele in a human which causes the growth of a green beard. Imagine also that this allele is able to cause the individual with the allele to recognize others containing the allele (via the green beard). Finally, this allele would be able to cause the individual to act preferentially towards others with the same allele. In this situation one gene would be responsible for all effects. As unlikely as this is, such a situation has been found in a species of fire ant. And while it is virtually impossible that such a situation exists in humans, it is useful to use this model, along with the two previous ones dealing with kinship and reciprocity, as a means to discuss what exactly an individual must be able to do for altruism to exist.
From the kinship model it is apparent that kin discrimination and kin recognition are important in the evolution of altruism. This is done through many means including shared phenotype (based on visual, olfactory, or auditory cues), social learning ("introduced" to kin as kin), and association (e.g. lives in same house so is brother). From the reciprocity model it becomes apparent that it is important to be able to recognize reciprocators. This is done through, among other means, probing (testing for returned altruistic behavior), learning (remembering who is not altruistic), and symbols (assessment cues such as mannerisms or attire on first meeting). From the green beard model (and especially the fire ant study not discussed in this paper) it is clear that the molecular basis should also be considered. Included in this are the questions of what biochemical processes are altered by different perceptions, what proteins and genes cause these alterations, and what other functions do these genes and gene products serve.
There is evidence that these models do point towards appropriate searches for altruism alleles in humans. The most convincing argument for this comes in a paper by McGuire et al on altruism and mental disorders. In this paper the authors discuss individuals who show decreased levels of altruism as a result of mental disorders. The findings of the paper were that reduced altruistic behavior may be an evolved strategy, a consequence of dysfunctional recognition systems, and/or a secondary response to an increase in symptoms (McGuire et al).
The first conclusion, that of reduced altruism as an evolved strategy, has already been briefly discussed in this paper under the heading of false altruism. In a society of cooperators a defector would profit based on the payoff matrix shown earlier in this paper and would therefore spread if the defector were able to avoid the detection systems of the reciprocators. This could be done by reducing the likelihood of repeat interactions (e.g. finding new helpers as others become less inclined to help) or by manipulating the symbols used by reciprocators to locate defectors. In essence, these individuals are, "effective manipulators of the symbol, experience, and probe systems," (McGuire et al). These individuals are fairly common, often go undetected, and are referred to as Primary Adaptive Uncooperatives.
Dysfunctional Uncooperatives are the second set of individuals studied. These individuals are unable to understand the rules of reciprocity and are unable to make cost-benefit assessments accurately. Generally, this results in an individual who devalues the help given by others while over estimating the cost incurred by himself. This leads to reduced cooperation and decreased social capabilities. This is less likely to go unnoticed (as the Primary Adaptives were able to) and is found often in mental patients.
Secondary Uncooperatives are those who view themselves under large amounts of distress and as such alter their reciprocity budget. Generally these individuals feel that their immediate future is in danger. As a result of this they budget less of their energy towards reciprocal exchanges and show decreased levels of altruism. Depression is an example of this type of uncooperativeness. Support for this theory arises from the fact that there is a strong correlation between the level of symptoms and the level of altruistic behavior (the greater the symptoms the less the altruism).
This paper has attempted to demonstrate that many aspects of behavior are genetically influenced and that altruism can evolve within a population as a result of kinship and reciprocal interactions. It is possible to use these models to develop another model, that of the Green Beard. This can then be used to predict the capabilities of altruism alleles. Finally, it has been shown that many of the predictions made by these models are supported by clinical research.
Goldin et al. "Segregation and Linkage Studies of Plasma Dopamine-Beta-Hydroxylase (DBH), Erythrocyte Catechol-O-Methyltransferase (COMT), and Platelet Monoamine Oxidase (MAO): Possible Linkage between the ABO Locus and a Gene Controlling DBH activity." American Journal of Human Genetics, 34:250 - 262, 1982.
Krebs and Davies. An Introduction to Behavioural Ecology, third edition. Blackwell Science Ltd, Osney Mead, Oxford, 1993.
Kurland, Jeffrey. Anthropology 464. Spring Semester 1999, Penn State University.
McGuire et al. "Altruism and Mental Disorders." Ethology and Sociobiology 15: 299 - 321, 1994.
Sociobiology and Human Nature. Edited by Gregory et al. Jossey-Bass, Inc, San Francisco CA, 1978.