APPENDIX ONE: THE MEANING OF GENETIC CAUSATION
by Elliott Sober
Genes do two things. They provide a mechanism of inheritance, and they influence how organisms develop. When genes do the former, they effect a connection between generations--parents pass genes along to their children. When genes do the latter, they participate in processes that occur within a generation; they affect how a fertilized egg--a single cell--divides and differentiates, and eventually becomes an adult, who has numerous traits that were not present at conception.
In saying that genes provide a mechanism of inheritance, the point of the indefinite article is to highlight the fact that there are nongenetic pathways whereby parents influence the traits of their children. A child who hears English spoken while growing up will come to speak English, but this is not because English-speaking parents transmit a gene for speaking English to their offspring. Imitation and learning can lead children to resemble their parents. Cultural context can engender additional similarities that are not genetically mediated; when children inherit money from their parents, this is neither genetic transmission nor learning.
When we turn to the second role that genes play, the indefinite article is again appropriate. Genes are a cause of the traits that organisms develop, but there are others. These nongenetic causes are lumped together under the heading of "environment." Genes contribute to an individual's being tall, but so does the amount of nutrition consumed when young. In discussing how development proceeds, it is important not to equate "genetic" influences with "biological" influences. How many calories you eat is a "biological" influence on your height, but this is an environmental, not a genetic, factor. To say that "biology" is relevant to a trait is not the same as saying that "genetics" is relevant. There is more to biology than genetics!
This point is easy enough to grasp in connection with an example like height, but it seems to get cloudy for many people when they think about psychological phenotypes. For example, in one controversial study, Simon LeVay (1993) performed autopsies on the hypothalmuses of a number of homosexual and heterosexual men, and claimed to find that a certain region--the INAH-3 nucleus--was smaller in the former group than in the latter. LeVay's claims have been disputed for many reasons; for example, the gay men examined had all died of AIDS, so the question arises as to whether the size of their nuclei was affected by the disease or by the treatments the individuals received. However, the point I want to make about this research concerns the fact that many people described LeVay's work as showing that there is a genetic basis of homosexuality. No such conclusion follows, even if LeVay is right that the two groups exhibit this neurological difference. A difference in phenotype may be due to a genetic difference, an environmental difference, or both; the point does not change when one focuses on the phenotype of brain structure, or on a psychological phenotype like sexual orientation.
The two roles played by genes--in inheritance and in development--are illustrated in the accompanying figure. Parents transmit genes to offspring during reproduction (R); genotypes influence phenotypes during development (D); and parental phenotypes influence offspring phenotypes when children learn (L) from their parents. The figure also depicts the fact that genes are not the only players in the processes in which they participate. Parent/offspring resemblance may be due to shared genes, to the fact that parents teach their children, or to the fact that both develop in the same environment. And the traits that an organism develops are a joint product of the genes the organism possesses and the environment in which the organism lives.
parents' phenotypes ----------(L)--------> offspring's phenotype
parents' genotypes ----------(R)--------> offspring's genotype
Of the two roles that genes play, one is much better understood than the other. The rules of genetic transmission are pretty clear. Genes reside in chromosomes, which are found in the nuclei of cells. Human parents have forty-six chromosomes in each cell of their bodies; these chromosomes come in pairs, so there are twenty-three pairs. When parents produce sperms and eggs, each sperm and each egg receives one chromosome from each pair; each possesses twenty-three singleton chromosomes. When these sex cells come together to form a fertilized egg, the genetic complement of twenty-three pairs is created anew. There is a great deal more to the physical processes involved in the formation of sex cells and in the way that sex cells come together to form a new individual, and some of these details are the focus of continuing research. Yet, in outline, the way genes behave in the process of reproduction is a known quantity.
Matters are very different when we ask how genes participate in the process of development. The initial stages of this process are well understood. Genes are pieces of DNA. Two pieces of DNA are physically different when they contain different sequences of nucleotides. Development begins with different pieces of DNA constructing different pieces of messenger RNA, which then are used to construct amino acids, the building blocks of proteins. These proteins then help construct still larger and more complex structures. A zygote has one cell. Genes cause the cell to divide and the daughter cells to specialize, so that the end product is an organism with many specialized cells--brain, liver, heart, etc.--that form up into specialized organ systems. In general, the more downstream developmental processes are, the less they are understood: a great deal is know about how genes produce proteins, but we still are vastly ignorant about how genes construct working livers and brains.
I. Three Modes of Intervention
In spite of the fact that development remains a mystery in many respects, the current molecular revolution in biology is vastly expanding our understanding of how genes contribute to developmental outcomes. Scientists already are able to intervene into human lives in various ways, and the precision and pervasiveness of these interventions are bound to increase as we learn more. We can use the previous diagram to identify three types of intervention that knowledge of genetics makes possible. To see how an individual's phenotype might be modified, consider the arrows that point, directly or indirectly, to that item in the figure. Two pathways of intervention were in use before the molecular revolution of the 1980s, though new knowledge will expand their domains of application. A third is still in its embryonic stage.
First, understanding which genes contribute to which traits allows people to control which genes are passed along from parents to offspring. This was the dream of the old eugenics movement. It lives on in the modern guise of genetic counseling. For example, potential parents can be tested for whether they carry a single copy of the Tay-Sachs gene; if each of them does, there is a 1 in 4 chance that their child will have two copies and so will have the disease. Parents, thus informed, may decide not to have children, or to take the chance and abort the fetus if it turns out to have two copies. This intervention does not change the phenotype that a given individual comes to possess; it merely affects which individuals are conceived and allowed to be born. The use of this screening technique has significantly reduced the frequency of Tay Sachs disease--not by curing it, but by preventing individuals who would have had it from coming into existence.
Much of what is known about how to detect the presence of a gene and to predict its phenotypic effects predates the molecular revolution. Notice that it is not essential to understand the developmental details to put this intervention into practice. All that is needed is a test for whether parents have one copy of a gene that produces the target phenotypic condition when the gene is found in double dose. Powerful molecular techniques will greatly enhance this domain of knowledge--more traits in offspring, both good and bad, will be predictable (in that probabilities will be assigned to their occurrence) from genetic tests performed on parents.
The second pathway of intervention involves manipulating the environment. As we learn more about what genes do, we also learn more about how environments may be modified to compensate for defective genes. For example, PKU syndrome (phenylketonuria) is a genetic condition in which people are unable to digest phenylalanine. If they receive a normal diet that contains this amino acid, phenylalanine accumulates in their system, resulting in a severe mental retardation. But if they grow up in a different environment--one in which the amino acid is not present in their diet--they develop normally. Cystic fibrosis provides a second example. It is a genetic disease. People with cystic fibrosis often die of lung infections that occur because the mucus in their lungs is a breeding ground for bacteria. It has been found that the enzyme DNAase is able to break down the DNA in this mucus, thus reducing its viscosity. Here molecular biology has provided a new environmental manipulation, one that promises to improve the lives of many CF patients.
The third pathway of intervention is very much in its infancy, but it can be expected to become more and more available, across an ever widening range of conditions. If individuals lack a gene that is needed to produce a given protein, then gene replacement therapy may allow working genes to be inserted into their cells that do the job. Likewise, if individuals have a gene that produces a destructive protein, new genes may be added that remove the protein, or prevent the protein from being constructed in the first place.
One version of the condition called severe combined immune deficiency (SCID) results when two copies of a recessive mutation are present; the result is that a protein needed by the immune system is not produced. An experimental procedure has been tried out on a small number of children; they received DNA that causes the needed protein to be produced. The results have varied, but immune response has consistently improved, sometimes quite dramatically. Another example is a genetic condition that results in drastically elevated levels of LDL cholesterol. A young woman with hypercholesterolemia, who as a result had a history of severe coronary disease, had her LDL levels significantly reduced by a gene replacement therapy that involved modifying the genes in some of her liver cells. These two examples are still experimental procedures, so their general efficacy is at present unknown. It remains to be seen how many other phenotypes may be modified by similar techniques. Still, it seems clear that molecular biology will provide an ever increasing number of such interventions.
Early geneticists had no knowledge of the physical makeup of genes. The rediscovery of Mendel's work around 1900 led scientists to think of genes as hypothetical entities, known only by their effect on the phenotypes of organisms. The pea plants that Mendel studied produce smooth and wrinkled peas in predictable proportions; this allowed the inference that inside of those plants are entities--"factors"--that combine in certain ways and somehow cause phenotypes to appear. It was a major advance when Thomas Hunt Morgan and his school were able to establish in the early part of this century that genes are located in chromosomes. Even so, it remained true that the existence and nature of genes was something that could be gleaned only from the observable phenotypes of organisms.
The way in which genes actually produce phenotypes remained a mystery and science had no independent access to their physical makeup. The great leap forward effected by molecular biology is that the physical nature of genes and their contribution to developmental processes can be studied more directly. Genes are no longer thought of as hypothetical entities; rather, they are "factories." They start and stop work because of identifiable physical signals and they manufacture products, whose physical makeup can be analyzed.
In discussing these three possible types of intervention, which aim to influence the phenotypes that individuals develop, I have chosen diseases as my examples. Indeed, medical applications of this sort are pretty much the exclusive focus of present-day genetic counseling, genetically informed environmental manipulation, and gene replacement therapy. Still, what can be done for abnormal phenotypes may also be possible for normal ones. Just as science may provide techniques for changing a diseased phenotype into a healthy one, it also may be able to change a normal phenotype to one that is "better" than normal.
For example, if some interventions can prevent certain sorts of mental retardation, it is possible in principle that other interventions can change normal intelligence to above average intelligence. Not that it is easy to figure out how this type of intervention might proceed. Perhaps it is a simpler scientific task to identify genes that contribute to diseased phenotypes than it is to figure out how normal genes might be modified to produce "enhanced" phenotypes. And, of course, there may be ethical objections to interventions of this type. Still, the ability to intervene in this way is on the horizon as scientific knowledge expands. The above diagram applies just as much to "superior" phenotypes as it does to average or defective ones.
I place the term superior in quotation marks to signal the fact that it raises problems. Who is to say which traits are better? The Nazis thought in these terms, but their odious form of "racial biology" has long been exposed as the nonsense it always was. Although racists will be quick to use the term "superior," it is important to see that the issue I am pointing to does not disappear just by disavowing racism. When nonracists say that disease is less desirable than health, they are applying the terminology of better and worse.
If we try to say why it is better to be healthy than diseased, we inevitably are led to say that some unusual phenotypes would be better to have than the average phenotypes found in human populations. Mental retardation is less desirable than normal mental function, in part because retardation drastically contracts the range of worthwhile lives a person might be able to lead. However, this reasoning leads to a further conclusion--that enhanced mental function can be more desirable than average mental function if it expands the range of worthwhile options a person has. Of course, this does not mean that these two comparisons are equally weighty--that avoiding retardation and being average is exactly as important as avoiding being average and being a genius. The point is that if disease is worse than health, then it is hard to resist the conclusion that some rare traits would be better to have than traits that are "normal" in the sense of being average.
My claim is that from the point of view of biology, it makes sense to think of disease, health, and enhanced function as all falling on a single continuum. If intervening in the lives of individuals allows science to move them from disease to health, science may also be able to move individuals from being merely healthy to having enhanced function. The difference between a diseased and a healthy person may be due to their different genes, different environments, or both; by the same token, the difference between a normal healthy person and someone who has enhanced function in some respect (such as better lungs, better eyesight, better memory) may be genetic, environmental, or both.
II. Four Key Questions
I now want to analyze more carefully the causal language that is used to describe the impact of genes and environment on an individual's phenotype. The preceding figure was full of arrows that represent causal influence. What does it mean to say that genes are more important than environment (or vice versa) in the development of some phenotype? I said earlier that PKU and cystic fibrosis are "genetic" diseases, but then I described environmental interventions that prevent their development or ameliorate their symptoms. What does "genetic" mean in this context? I also said that speaking English is not a genetic trait, in that people acquire whatever language they hear while they grow up. But it also is true that individuals without genes do not end up speaking English, because they do not grow up at all; and a chicken exposed to spoken English will not end up speaking English, presumably because of the genes it has. So what does it mean to say that speaking English is environmental, not genetic?
To address these issues, it will be useful to consider four questions about the role that genes play in causing some trait that an individual has:
! Do genes causally contribute to the trait?
! How much do genes, as opposed to environment, contribute to the trait?
! Which genes contribute to the trait?
! How do these genes contribute to the trait?
These questions will have different answers, depending on the trait under discussion; what is true for speaking English is not true for Huntington's disease. The questions are listed in a certain order; typically, to answer a question further down the list, you have to answer earlier ones. In this sense, later questions are harder than earlier ones. My goal here is to clarify what these questions mean; I will not try to provide factual answers to these questions in connection with any particular trait.
All four questions assume a division between genes and phenotype on the one hand and between genes and environment on the other:
Phenotype and environment are defined in biology as "garbage can categories"--an organism's phenotype is all the traits it has other than its complement of genes. So all your morphological, physiological, behavioral, and psychological traits are part of your phenotype. The fact that someone is 5' 8" tall, has brown eyes, high blood pressure, a certain blood type, lives in the Midwest, and likes jazz are all phenotypic traits. Similarly, an organism's environment includes any trait that it has other than the genes inside its body. How many calories you consume is part of your environment; your caloric intake, plus your genes, together influence how tall you grow.
Even things that are inside the organism's own body can count as part of the "environment" as far as the development of a given phenotype is concerned. For example, if an individual is born with some degree of immunity to a certain disease because it received antigens from its mother contained in the extrachromosomal part of the egg, these antigens are counted as part of the organism's environment. The antigens, plus the organism's genes, influence whether and how much immunity it will exhibit. Possessing a set of antigens might be considered a part of the environment as far as the development of immunity is concerned, but having those antigens is also a phenotypic trait that the organism possesses. Now let's consider the four questions.
Question (1): Do genes causally contribute to the trait?
A set of genes can causally contribute to a trait without it being true that those genes suffice for the trait. And it need not be true that those genes are necessary for the trait to be present. This is a general point about what the word "cause" means. Striking a match causally contributes to the match lighting, but the striking is not by itself sufficient; oxygen has to be present and the match has to be dry. Similarly, you can get the match to light without striking it--for example, by using a magnifying glass to focus light on the head. Here are two illustrative examples:
! The defective gene in PKU syndrome causally contributes to the syndrome. However, a particular environment has to be in place--the presence of phenylalanine in the diet. So the gene is not sufficient for the syndrome.
! There is a defective gene that causes breast cancer. However, this gene is present in only a small percentage of women who contract breast cancer. The gene is not a necessary condition for the cancer.
These examples also show that when a phenotype has a genetic cause, this does not mean that the phenotype cannot be modified by changing the environment. As noted before, PKU syndrome can be avoided by modifying the diet. And it is possible that therapies will be discovered for women who have the gene for breast cancer that prevent the gene from doing its destructive work. Discovering genetic causes can provide insights into the types of environmental manipulation that would be efficacious.
Another point about question (1) is that it has to be understood in the right way. An organism will not develop at all if it has no genes. However, this obvious fact does not mean that genes automatically "contribute" to every phenotype we might consider. The reason is that "genetic contribution" means that genetic differences make a difference in whether you have a particular phenotype. For example: an organism cannot develop to the point where it is able to speak a language if it has no genes. However, this leaves open whether the genes you have make a difference in whether you will speak English rather than Chinese. Apparently, your genes make no difference in this regard:
Language heard as a child
G1 English Chinese
Genes you might have G2 English Chinese
G3 --- ---
The genes G1 and G2 each permit normal development, in which case the language you speak is determined purely by your environment. But suppose that G3 is a genetic configuration that causes a severe retardation that prevents a person from acquiring any language at all. If so, your genes affect whether you will learn a language or not, but they do not affect whether you will learn English or Chinese. Notice that this is entirely consistent with the fact that there are genetic differences between English speakers and speakers of Chinese. However, these differences do not make a difference; this is correlation without causation.
Question (2): How much do genes, as opposed to environment, contribute to the trait?
When two causes both contribute to an effect, how are we to say how much each contributed, or which was more important? Consider an example due to Richard Lewontin (1974). Suppose two bricklayers build a wall. Each brings his own mortar, bricks, and tools to the site; they start at opposite ends, the one building from the right, the other from the left. We could determine which bricklayer contributed more to the resulting wall by counting the number of bricks in the wall that each had placed there. However, suppose that one bricklayer made the bricks and mortar and brought all the necessary equipment to the site, while the other used these items to put up the wall. Now it is impossible to say which bricklayer contributed more, or to say how much each contributed in terms of some uniform measure.
The second version of the bricklayer example resembles the problem we face when we try to say whether genes or environment contributed more to a given phenotype. If Sally is 5' 8" tall, it will not be true that her genes built 4' of her and her environment built the rest. Rather, her genes and her environment made different kinds of contribution, the net result of which was that she grew to a certain height.
Does this mean that question (2) is meaningless--that it makes no sense to ask whether genes or environment are more important causes of the resulting phenotype? No. It just means that we must pose the question in a different way. Rather than asking about the relative contributions of genes and environment to the phenotype of a single individual, we change the object of the preposition. We ask how genetic differences and environmental differences contribute to the differences in phenotype found in a population.
To see what this means, consider a simple experiment designed to show how genetic variation and environmental variation contribute to variation in height in a population of corn plants. In our experiment, the plants have either genotype G1 or G2; their environments provide either one unit of fertilizer (E1) or two units of fertilizer (E2). Thus, there are four "treatment cells" that a corn plant might occupy; it might be in E1G1, E1G2, E2G1, or E2G2. Suppose we plant four plots of corn, each containing a hundred corn plants, one for each of the four treatment combinations. How will the plants in one treatment cell differ in average height from the plants in others? Here are four possible observational outcomes that might be obtained:
G1 G2 G1 G2 G1 G2 G1 G2
--------- --------- --------- ---------
E1 1 1 E1 1 2 E1 1 3 E1 1 4
E2 4 4 E2 3 4 E2 2 4 E2 1 4
--------- --------- --------- ---------
(i) (ii) (iii) (iv)
In outcome (i), the genetic factor makes no difference; whether the plants have genotype G1 or G2 does not affect their height; it is the environmental factor--the amount of fertilizer the plants receive--that explains all the observed variation. Outcome (iv) is the mirror image of (i). In (iv), the fertilizer treatment makes no difference; the genetic variation explains all the variation in height. Outcome (iv) illustrates the idea of genetic determinism--the general thesis that an individual's phenotype cannot be modified by changing its environment. Outcome (i), on the other hand, illustrates the idea of radical environmentalism--the view that the only factor that influences an individual's phenotype is its environment. Outcomes (i) and (iv) thus support monistic explanations of variation in plant height; each suggests that only one of the factors considered made a difference in the observed outcome.
Outcomes (ii) and (iii), on the other hand, would support pluralistic conclusions. Each suggests that genetic and environmental factors both made a difference. However, they disagree about which factor mattered more. In outcome (ii), changing the fertilizer treatment yields two units of change in height, whereas changing from one genotype to the other produces only a single unit of change. In this case, the environmental factor makes more of a difference than the genetic factor. By the same reasoning, we can see that outcome (iii) suggests that genetic variation was more important than the environmental factor considered.
Although the four possible outcomes described so far differ in various respects, they have something in common. In each of these data sets, the change effected by moving from G1 to G2 does not depend on which environmental condition one considers. Similarly, changing from E1 to E2 has the same impact on plant height, regardless of which genotype the plant possesses. Results (i)-(iv) are thus said to be "additive" (or to show no "gene-environment interaction"). This is not the case for the following possible results:
G1 G2 G1 G2
E1 1 7 E1 1 1
E2 7 4 E2 1 4
In outcome (v), going from one unit of fertilizer to two increases height for plants with genotype G1, but reduces height for plants that have genotype G2. In (vi), changing genotype has an effect on plant height within one fertilizer treatment, but not within the other.
Gene-environment interactions of the sort depicted in (v) and (vi) occur frequently in nature. Consider for example, the accompanying figure from an introductory genetics textbook (Griffiths, Miller, Suzuki, Lewontin, and Gelbart 1993, p. 14). It depicts the result of taking cuttings from each of seven plants in the species Achillea millefolium and growing them at high, medium, and low elevations. Notice that different environments have different effects on plant height; moving from medium to high elevation makes type 1 cuttings grow taller, but reduces the height of cuttings from type 4; and even though moving from low to medium elevation reduces plant height in both types 2 and 3, it does so by different amounts.
The interaction effect depicted in (vi) shows that it can be overly simplistic to ask whether a phenotype is influenced by the environment. Notice that genotype G2 produces a different phenotype when the environment is changed, but genotype G1 does not. The same point holds when we ask whether a phenotype is influenced by genetic factors. Organisms reared in environment E2 produce different phenotypes depending on what genotype they have; organisms reared in environment E1 do not. We can look at the data in (vi) and make the summary statement, "the phenotype depends on both genotype and environment," but a more fine-grained analysis shows that genes make a difference in some contexts but not in others, and that the same is true of the environment considered.
This is not a merely academic fine point. An important discovery in AIDS research is that this disease is able to enter host cells by connecting with a protein found on the surface of CD-4 immune cells. Most human beings produce this protein, and so are vulnerable to the disease. However, it has recently been discovered that a small number of individuals possess two copies of a mutation in the normal gene CKR-5, which causes this protein not to be produced. These individuals are otherwise healthy; they also seem to be immune to the disease. Notice what this implies about whether AIDS is an environmental or a genetic condition. Before this mutation was discovered, it would have been natural to think that the disease is caused purely by an environmental insult--the presence of a virus; the difference between people with the disease and people without it is completely ascribable to the environment. But this recent discovery complicates the picture. For some people, not having AIDS is attributable to their genes, not to their environment. For many others, however, the disease remains an environmentally induced phenotype.
The simple 2-by-2 experiment we have described illustrates the basic idea behind the methodology that statisticians call the analysis of variance (ANOVA). It allows you to compare how much genes and environment contributed to a phenotypic outcome. A conclusion might be reached that "70% of the variation is explained by environmental differences and 30% by genetic differences." ANOVA quantifies the relative contributions that genes and environment make to phenotypic variation. It is important to bear in mind that this type of analysis is highly specific to the range of genotypes and environments considered. Even if changing from E1 to E2 has only a negligible effect on plant height, it is an open question whether changing to a new environment E3 would make a larger difference. Likewise, even if genotypes G1 and G2 perform similarly across a given range of environments, it is quite possible that genotype G3 would behave quite differently in those same environments.
This point is relevant to understanding how the following two findings are quite compatible: a significant proportion of the variation in intelligence (as measured by IQ) found among individuals is attributable to genetic differences (Bouchard et al. 1990); and intelligence has been increasing steadily in 14 countries, including the United States, for several decades (Flynn 1987). When quantitative geneticists use ANOVA techniques, they typically study the range of genes and environments that are available at the time of their study; the results they obtain say nothing about whether the range of genes or environments has changed or will do so in the future, nor do the results predict what will happen should such changes occur. Even if the variation in height now found in a population is explained completely by genetic variation, it remains perfectly possible that the population will have a greater average height several generations later because nutrition has improved. The fact that genes significantly affect phenotypic variation does not mean that environmental interventions will be for naught.
I began my discussion of the 2-by-2 experiments on corn plants by saying that it makes no sense to apportion causal responsibility between genes and environment with respect to a single individual (Sally); this was why I shifted focus to the question of explaining patterns of variation in a population of individuals. However, it might appear that the ANOVA technique just described does allow one to explain the traits of individuals. Consider, for example, the data given in (ii); below I have presented this set of data again, this time noting in the margins the average heights of individuals sharing the same environment, the average heights of individuals sharing the same genes, and, in the lower-right hand corner, the grand mean--the average height of all the individuals in the experiment:
E1 1 2 1.5
E2 3 4 3.5
2 3 2.5
Consider one of the individuals in the lower-right hand cell of the 2-by-2 table of data who is 4 units tall. The average height in the whole population is 2.5. The difference between this individual's height and the grand mean (4-2.5=1.5) can be decomposed into a difference between the average height of individuals with the same genotype as this individual and the grand mean (3-2.5=0.5) and the difference between the average height of individuals living in the same environment and the grand mean (3.5-2.5=1.0). Does this show that this individual's being 1.5 units above average in height is explained by her genotype making her 0.5 units above average and her environment making her an additional 1 unit above average? If so, her genes account for one third of her overshooting the average and her environment for two thirds.
I do not disagree. However, this is not the same thing as explaining why the individual is 4 units tall. Notice that if this individual had the same genes and the same environment, but had lived in a population of different composition, her deviation from the mean would have been different, as would the breakdown between her genetic deviation and her environmental deviation. Yet the individual would still be 4 units tall; the genetic and environmental processes that determine her height would have been the same. ANOVA is an irreducibly population-level analysis; it can be used to describe how individuals are related to the populations they inhabit, but this is an entirely different matter from explaining why they possess the phenotypes they do.
One other conceptual point needs to be made about the logic of ANOVA studies. This concerns the relation of within-population differences to between-population differences. One of the most controversial subjects in the continuing debate about nature and nurture is the question of whether the observed difference between white and black Americans in IQ has a genetic basis. Murray and Hernnstein's The Bell Curve (19//) is the most recent attempt to muster evidence for this conclusion; theirs was not the first effort, nor should we expect it to be the last. The point I want to make here is that their conclusion about group-level differences would not follow even if variation within the two groups had a significant genetic component.
Consider the following example, described by Lewontin (1970). Suppose a heterogeneous collection of corn seeds is planted in a single environment E1. Since all the plants experience the same environment, all the variation in height must be due to genetic differences. Now suppose that a second experiment is performed in which the same collection of seeds is planted in a quite different environment E2; once again, all the phenotypic variation in this experiment will be due to genetic differences. However, what should we say about the difference between the average height exhibited in the first experiment and the average height obtained in the second? This difference will be due entirely to the environmental difference between E1 and E2, since the two experiments tested the same range of genotypes. The fact that genetic differences explain phenotypic variation within two groups does not mean that genetic differences explain phenotypic variation between those two groups.
I hope this discussion makes clear what it means to compare the impact of genetic and environmental variation on variation in some phenotype. However, it leaves a related question unanswered. How are we supposed to find out what the contributions of genes and environment are if we cannot run a controlled experiment of the kind just described? Manipulations of the type just described are fine for corn plants, but ethics prevents us from cloning individuals and rearing them in identical environments. So how is it possible to say anything meaningful about the relative contributions of genes and environment to human height, much less to more subtle traits such as intelligence, aggressiveness, sexual orientation, or risk-taking?
A variety of techniques are used in the field called quantitative genetics to answer this question--ones that involve assembling observations about individuals as they are. One approach is to study genetically identical twins who are reared apart. If two identical twins differ in height, this must be due to the fact that their environments were different, since they have the same genes. Suppose pairs of identical twins were more similar in height than pairs of unrelated individuals are. This would suggest that genes make a difference in how tall an individual grows.
Such methods are subject to various pitfalls. For example, in twin studies, it is important to investigate how much the environment of one twin differs from the environment of the other. Suppose that identical twins who are reared in separate households nonetheless receive the same number of Cheerios, and that this environmental factor is a powerful influence on how tall an individual grows. It will then turn out that twins closely resemble each other in height, but this is not simply because they have the same genes.
Bouchard et al. (1990) make this point in their report on the Minnesota twin studies, which involved comparing a set of identical twins who were reared together with about 100 pairs of identical twins who were reared apart. These individuals were subjected to about a week of psychological and physiological tests. Remarkably, the two groups each showed significant degrees of similarity; even more remarkably, they did not differ much in terms of how similar they were. For example, twins reared together are rather similar in how they scored on tests of "religiosity" (they exhibited a correlation of 0.49); twins reared apart also were quite similar (they exhibited a correlation of 0.51). Bouchard et al. (p. 227) point out that the twins reared apart in their study almost never grew up in extreme poverty and were almost never raised by illiterate parents. This means that these individuals had environments that were at least somewhat similar, and this fact has to be taken into account in interpreting the results. Some of the similarity found in identical twins (whether reared together or reared apart) will be nongenetic.
There are other caveats that apply to empirical efforts to estimate the genetic and environmental contributions to patterns of variation in some phenotype of interest. It is true that a number of studies in the past were sloppy. In addition, it would be naive to think that quantitative geneticists now do their studies without ever making mistakes. However, the fact remains that scientists learn from the errors of their predecessors; scrupulous studies are now being executed and they will become more numerous with the passage of time. The mistakes of the past should not lead us to dismiss twin studies and related inquiries a priori.
Question (3): Which genes contribute to the trait?
Twin studies and similar methodologies in quantitative genetics can tell you that genes make a difference in the chances of having some trait; however, they fail to tell you which genes matter or how they have this effect. You may find that identical twins resemble each other in height far more than unrelated individuals do, but which genes on which chromosomes influence height and how do they do so? An answer to question (2) can leave you entirely in the dark concerning questions (3) and (4). This is both the strength and the limitation of quantitative genetics studies. They can go forward in total ignorance of developmental details; this is why the field of quantitative genetics developed before the molecular revolution of the 1980s and, indeed, before Watson and Crick discovered the chemical structure of DNA in the 1950s. Quantitative genetics is genetics without much attention to genes.
I now want to describe a study that aimed at providing insights into question (3). It was quite controversial because the scientists were not studying something innocuous like height in corn plants. They studied sexual orientation in human beings. This is the study by Dean Hamer and his associates (1993), in which they claim to have found a region of the X-chromosome in males that causally contributes to homosexuality. It remains to be seen whether the study will be replicated and exactly how it should be interpreted. However, the point of interest here is to understand the logic of Hamer's investigation.
Recall an elementary fact about human genetics depicted in the diagram below. Males are XY and females are XX. This means that males inherit their X chromosomes from their mothers. Suppose the mother is heterozygous at some location on her pair of XX chromosomes. This means that she has one copy of gene A next to a copy of gene B at that location. Now consider a population of such heterozygote mothers. Imagine that each has two sons, who are not identical twins. What is the probability that they will both have A or both have B on their one and only X chromosome? The probability that son #1 has A is 1/2 and the probability that son #2 has A is also 1/2, so the probability that both have A is 1/4; similarly, the probability that son #1 has B is 1/2, which is also the probability that son #2 has B, so the probability that both have B is 1/4. This means that the probability of them matching (i.e., that both have A or that both have B) is 1/4+1/4=1/2:
XX ----------------------- XY
Son #1 Son #2
Suppose you wanted to know whether the genes at this locus on the X chromosome, when present in males, affect their probability of having blue eyes. To find out, you might assemble 40 pairs of brothers, each of whom has blue eyes, and each of whom has a mother who is an AB heterozygote at some locus on her XX chromosomes. If these genes have no bearing on eye color, you'd expect that 1/4 of the brother pairs will match for A, 1/4 will match for B, and 1/2 will not match at all. In other words, if these genes do not affect eye color, you'd expect blue-eyed brothers to match at this locus about half the time. On the other hand, if you found that a very high proportion of the brother pairs match, this would be evidence that the locus in question influences eye color.
Hamer assembled 40 pairs of brothers whose score on a variety of self-report sexual orientation scales placed them at the homosexuality end of the spectrum, and whose mothers were heterozygotes for a marker placed at a certain region of their two X chromosomes. Hamer found that 33/40 of the brother pairs matched for the markers in question. This is evidence for the claim that genes in this region influence sexual orientation.
Suppose that these 33 brother pairs all have gene A. It would not follow that this gene suffices for homosexuality; there may be individuals, not in the study, who have the gene, but are not homosexual. Nor would it follow that the gene is necessary for that phenotype to exist; it may be that this gene is present in only a small percentage of male homosexuals. An additional possibility is worth noting; it also is possible, given Hamer's finding, that the effect of gene A on sexual orientation is environmentally mediated. Let us consider a fanciful example to see why. Suppose that this gene makes boys have a freckle on their forehead and that this causes their parents to treat them in a certain way. Because of the way their parents treat them, boys with gene A tend to become gay. This may lead you to think that homosexuality in this scenario is environmental, not genetic. However, geneticists understand the contrast between genes and environment in such a way that the opposite conclusion is the one they draw. When genes cause individuals to be reared in special environments, the resulting phenotype is said to be due to genes, not to environment.
Christopher Jencks et al. (1972) made this point in their book Inequality when they discussed the effects of racism on IQ. They used the following hypothetical example. Suppose that red hair is caused by genes, and that the society we live in leads people to abuse red-headed children. If abuse lowers IQ, then red-heads who live with their biological parents and red-heads who live with adoptive parents will suffer the same degree of abuse. The change in environment will not produce a change in the phenotype, because everyone in the society abuses red-heads. The conclusion will then be drawn that IQ is genetic; the gene for having red hair is also a gene for lower IQ.
The analysis would be different if red-headed children were abused by their biological parents, but not by others. Then an adoption study would show that a difference in environment makes a difference in the IQ of red-headed children. Consider the following two causal pathways:
(a) Genes ---> Red hair ----> Abusive environment ---> Lower IQ
Genes ---> Red hair
(b) Lower IQ
In (a), the lower IQ of red-haired children is said to have a genetic cause. In (b), the lower IQ of red-haired children is said to be caused by their environment, not by their genes. The standard methodology in quantitative genetics says that the environment makes no difference in (a), but does make a difference in (b).
The possibility illustrated by this example is not merely hypothetical. As noted before, the Minnesota twin study found that identical twins reared apart are remarkably similar on a variety of measures. Bouchard et al. (1990, p. 227) speculate that this may be because "their identical genomes make it probable that their effective environments are similar....Infants with different temperaments elicit differing parenting responses...children and adolescents seek out environments that they find congenial." The suggestion is that genes make an important difference in the phenotypes measured because different genes lead individuals to experience different environments. If so, the formula should not be "nature versus nurture," but "nature via nurture" (Bouchard et al. 1990, p. 228).
Question (4): How do these genes contribute to the trait?
Hamer's study, even if it is replicated, throws no light on the developmental pathway from a gene to a trait. Question (3) has been answered, but question (4) has not. In a sense, question (4) is the hardest of the four. You have to do all the work involved in answering questions (1)-(3), and then you have to tackle a difficult developmental question besides. The reason that question (4) is difficult is that gene/environment pairs do not directly produce phenotypes like height or sexual orientation or intelligence. Rather, genes produce gene products, which are chemicals. How these gene products interact with other gene products and with the environment is typically quite complicated. Somewhere downstream, the observable phenotype results.
There are examples of phenotypes in which the pathway from gene to phenotype is fairly well understood. Sickle-cell disease is caused by two copies of a recessive gene. People with two copies produce abnormal hemoglobin, which is the molecule inside of red blood cells that carries oxygen from the lungs to the rest of the body. When these individuals are oxygen-deprived, for example as a result of working hard, their red blood cells become sickle-shaped, thus blocking blood flow in capillaries. It is worth noting here that this fairly detailed grasp of developmental details has not yet resulted in the discovery of a viable gene replacement therapy. Knowledge and power do not always move exactly in tandem; some diseases are less well understood than sickle-cell anemia, but more effective therapies have been obtained nonetheless.
Examples like sickle-cell disease, in which science has discovered how genes contribute to the production of phenotypes, should not obscure the fact that when we shift to phenotypes such as sexual orientation, intelligence, risk-taking, and so on, question (4) represents a region of vast ignorance. As we descend from (1) and (2) to (3) and then to (4), the amount of knowledge we have at each stage vastly shrinks. There are many, many studies that purport to answer questions (1) and (2). As we have already seen, the methods of quantitative genetics permit these questions to be addressed even when science knows nothing about which genes might influence the phenotype under study and how they do so. However, when we look at questions (3) and (4), the range of phenotypes studied by science radically contracts. Typically, progress has been made for phenotypes, like sickle-cell anemia, that are diseases and are influenced by a small number of genes. It is a task for future science to try to achieve a more complete grasp of the developmental pathways that lead to nondiseased phenotypes.
Should we expect genes to be important in the explanation of normal function, just as they are relevant to the explanation of some, but not all, diseases? Let us be clear about what this question means. Genes help explain the difference between sickle-cell anemia and normal hemoglobin function. Is the same true of the difference between two varieties of normal hemoglobin function? Is it true for differences in intelligence, shyness, taste in music, etc., that all are normal? "Normal" is a vague term; it encompasses an array of phenotypes that are said to "fall in the normal range." Will genes be relevant to explaining differences within the normal range, just as they are often relevant to explaining differences between normal and abnormal?
It is a striking fact about work in quantitative genetics that genes are said to make at least some difference in virtually all phenotypes that have been studied (Bailey 1996). Even when traits like religiosity and political affiliation are studied by quantitative geneticists, they end up concluding that genetic differences among individuals matter, at least somewhat (Bouchard et al. 1990). Their question concerns how much genes matter, as compared to the environment. Not all these studies have been well executed, but quantitative geneticists keep getting better and better at avoiding mistakes. Even after all fallacies are removed and dubious interpretations are seen for what they are, we may have to live with the fact that the genetic contribution to any phenotype of interest is never precisely zero. This does not mean, of course, that genes make all the difference, or that they are more important than environment. Again, it is crucial to remember that genetic causality does not rule out environmental causality. Both influences exist. However, if genes contribute to variation in some phenotype of interest, more detailed questions remain to be addressed concerning which genes affect the phenotype in question, and how they do so; these are matters that future science will endeavor to answer.
It is often convenient to save breath or ink by neglecting to make explicit the relational character of certain claims. If I say that someone is a good musician, you will know roughly what I mean. But if you press for precision, I will have to say what benchmark of comparison I am using. The person in question may be good compared with the average player, mediocre compared with gifted amateurs, and decidedly poor when compared with the creme de la creme. This point is not news to anyone, but is part of our ordinary understanding of what such statements mean.
When we say that X is a genetic condition, full stop, it is easy to think that a complete statement has been made. The same conclusion seems natural when we say that Y is an environmentally acquired phenotype. However, such statements are no less elliptical than the statement that someone is a good musician. A condition has a significant genetic or environmental component only relative to a range of genes and a range of environments. Change the range of genes or the range of environments, and the conclusion about the condition may change as well.
In addition to emphasizing this fact about the relativity of claims concerning genetic causation, I also have tried to emphasize the difference between quantitative genetics and developmental genetics. Quantitative genetics studies have existed for about a hundred years; they long predate the molecular revolution. These studies can tell us whether genetic variation helps explain the phenotypic variation found in a population, but they are utterly silent on the question of which genes affect the phenotype and how they do so. Developmental genetics is the science that will make tremendous strides in coming years because of new techniques that allow scientists to investigate what genes do.
In fact, these two conclusion are not unrelated. A large measure of the power of molecular techniques stems from the fact that they allow scientists to modify the environments in which genes do their work. In addition, gene replacement therapy holds out the possibility that individuals can be furnished with new genes. Thus, interventions stemming from developmental genetics will have the effect of revising the range of environments and the range of genes that are available in human populations. As a result, it is not a mere flight of fancy to suspect that quantitative geneticists who now study a trait will reach conclusions quite different from the ones their successors will draw in the populations of the future. Heretofore, geneticists have studied the world of human variation; we now face a new age in which, for better or worse, they will possess the power to change it.
My thanks to Michael Bailey, James Crow, and Hill Goldsmith for comments on an earlier draft.
Bailey, J. (1996): "Can Behavior Genetics Contribute to Evolutionary Studies of Behavior?" In Crawford and Krebs (eds.), Evolution and Human Behavior--Ideas, Issues, and Applications. New York: Erlbaum.
Bouchard, T., Lykken, D., McGue, M., Segal, N., Tellegren, A. (1990): "Sources of Human Psychological Differences--The Minnesota Study of Twins Reared Apart." Science 250: 223-228.
Flynn, J. (1987): " ." Psychological Bulletin 101: ///-///.
Griffiths, A., Miller, J., Suzuki, D., Lewontin, R., and Gelbart, W. (1993): An Introduction to Genetic Analysis. New York: W.H. Freeman.
Hamer, D., Hu, S., Magnuson, V., Hu, N., Pattatucci, A. (1993): "A Linkage Between DNA Markers on the X Chromosome and Male Sexual Orientation." Science 261: 321-327.
Jencks, C., Smith, M., Acland, H., Bane, M. Cohen, D., Gintis, H., Heyns, B., and Michelson, S. (1972): Inequality--A Reassessment of the Effect of Family and Schooling in America. New York: Basic Books.
LeVay, S. (1993): The Sexual Brain.
Lewontin, R. (1970): "Race and Intelligence." Bulletin of the Atomic Scientists, March, 2-8. Reprinted in N. Block and G. Dworkin (eds.), The IQ Controversy, New York: Pantheon, 1976, pp. 78-92.
Lewontin, R. (1974): "The Analysis of Variance and the Analysis of Causes." American Journal of Human Genetics 26: 400-411.
Murray and Herrnstein, (19//): The Bell Curve.