Nature vs. Nurture
...d because he is at least on our side against the creationists. If that is not character assassination, it is hard to see what qualifies. So much for fair play and scientific integrity[iii], but as I mentioned at the beginning, this is not really a scientific debate, and the reader can and should appreciate the contrasting personalities involved in it as well as their philosophies and motives. The point I am trying to make here is not simply to indulge in a demonstration of how ugly things can get when science has direct social implications; rather, this is a salutary exercise in the psychology and sociology of science itself. One thing is clear: each side has done a marvelous job at poking large holes in the flanks of the other, of which peculiar situation I am clearly taking advantage here. Let me briefly discuss how some of the apparently sound arguments used by each camp can in fact been dismissed quite easily, and how the insistence upon such arguments on the part of either nurturists or naturists indicates poor understanding, less than objective analysis, or both. One of Gould’s major attacks against genetic determinism is based on what he calls the “reification” fallacy. Simply put, the fact that a statistical correlation exists among two or more variables does not imply that that correlation corresponds to a physical entity or demonstrates the existence of a real phenomenon underlying the correlations. So, for example, genetic determinists interpret the fact that the scores from different types of intelligence tests correlate in multivariate space as an indication of the existence of an underlying factor describing general intelligence, which they call “g”. g, Gould objects, is simply a statistical artifact of the fact that the different tests are designed in a similar way, so that of course individuals who score high on one test will score high on another, while individuals who don’t do as well on one test will fail another one too. This may be true, except that Gould then does not apply his preaching to his own scientific research. He published several papers (Gould 1984; Gould 1989) describing the (statistical) covariation among morphological traits measured in land snails of the genus Cerion. In that body of work, he argues that these correlations show the existence of several “constraints,” i.e. limitations on the future evolutionary trajectories of these populations of snails. Furthermore, he goes on to name such constraints! Is that not a perfect example of “reification”? Obviously, the practice is either acceptable or questionable to an equal measure in both cases, not just when it is not convenient for Gould’s position. In the opposite camp, naturists have always argued that their incontrovertible evidence is based on rigorous studies of twins reared together or apart. The statistical arguments are quite complex, but they boil down to the fact that twins have a high degree of genetic similarity, compared, for example, to fraternal siblings, or to unrelated individuals – such as adopted children. If twins show a higher correlation of their IQ scores than the control groups, this is taken as prima facie evidence for genetic determination of intelligence. Not so easily. First of all, this approach ignores the fact that the real problem is how to determine human reaction norms (see below), not to simply estimate heritabilities. Second, it is an astoundingly naïve reading of the data themselves. For example, it underestimates the fact that the environment in which twins or other siblings are raised can play a major role. If twins are reared together, their environments are also going to be very similar. In statistical terms, this causes a complete confounding effect of genetics and environment. Furthermore, even if the twins are separated at birth, this does not guarantee that their environments will be dissimilar, thereby allowing a statistical decomposition of nature and nurture effects. Indeed, most adoption agencies take great care in assigning children to families of social, economic, and cultural status similar to the ones characterizing the biological family (Levins and Lewontin 1985; Lewontin 1992) . Attempts at counteracting this problem by using some measures of the environment as “covariates” in the analyses are technically sound, but again rather simplistic. Does anyone really believe for a moment that just measuring income is going to take a great portion of the environmental differences away (Murray 1998) ? Twin studies are not very useful for the simple reason that they control for only one of the two factors, the genetics but not the environmental. Thus, while yielding evidence of some sort of genetic basis to human behavior, they provide us with no information on the all-important genotype-environment interactions. A technical paper written on similar grounds but dealing with plants or animals would be mercilessly rejected by any journal editor worth her salt as the unacceptable work of a dilettante. A perfect example of how unreasonable and unscientific both parties are in this controversy is given by their arguing over the relationship between intelligence and brain size. As you would expect, the nurturist camp claims that there is no relevant connection between the two, while the genetic determinists insist that brain size has a lot to do with intelligence (both camps admit that the brain’s fine structure is more important than simple size). The latter position is backed by the now incontrovertible correlation between brain size and IQ: about 0.44. This sounds impressive, but it is much less so once one realizes that a correlation of 0.44 between two variables means that one explains only 21% (0.442) of the variance of the other. That is, while brain size is indeed associated with intelligence (or at least that component of intelligence measured by IQ tests – and that is an entirely different controversy which I will not even attempt to touch here), about 80% of the variation in IQ scores in the population does not depend on brain size. It seems to me that there is ample space in that 80% for environmental effects, but of course this finding is trumpeted by Rushton (1998) as a decisive blow to the nurturists. It does not come even close to that effect. Back on the other side of the fence, Gould (in “The Mismeasure of Man”) takes the preposterous position of flatly denying even the existence of such correlation. He builds implausible scenarios of respected scientists of the 19th and 20th century “leaning” on their balances, or stuffing just the right amount of measuring pallets in skulls to come up with the “right” results. Even if we conceive that as a possibility, Gould himself acknowledges that the correlation would not disappear, but encourages us to ignore such a “minor” detail. Gould goes even so far as to make the argument that since clearly intelligent individuals, such as the French anatomist Cuvier and Nobel Prize novelist Anatole France, had brains that span almost the whole gamut of human values (1830 and 1017 grams respectively), brain size doesn’t have anything to do with intelligence. This is such a ridiculous claim that, made in a different context, would undermine a scientist’s credibility. The fact that two variables are correlated, and even causally connected, does not imply that you cannot pick two arbitrary data points and use them to contradict the general trend. Since the brain size-IQ correlation explains only about 20% of the variance, it is no surprise that some very small brains belong to intelligent people (or vice versa, some very large brains to particularly dumb subjects). This does not negate the validity of the correlation by any stretch of the imagination. For example, historian and social scientist Frank Sulloway (1997) demonstrated quite clearly a very strong statistical relationship between mental attitudes and birth order. First-born individuals tend to be much more conservative and refractory to innovations than later born ones. Sulloway also advanced good arguments for why this link is probably causal, and not just statistical. The explanation may have to do with the idea of family niches: first and later born siblings literally fill different niches in the eyes of their parents, and the resulting family dynamics mold much of their character and actions. Sulloway noted that this effect is more important (statistically, that is it explains more variance) than other traditional explanations such as social or economic status, or gender. Sulloway’s data, however, are perfectly consistent with a large role of these other factors, as well as with a role played by innate (genetic) determinants, simply because birth order does not explain the whole variance in behavioral responses. It should be clear to every graduate student that biological and social studies can only thrive on the understanding of general trends, and that there are exceptions to be found everywhere. However, these exceptions do not negate the rules. Biology is not pure mathematics and to pretend otherwise, as Gould selectively does in the case of the nature-nurture debate, is questionable to say the least. What do we actually know about the biological basis of human characteristics? The discussion in this essay does not have to be taken as a nihilistic statement on the state of our knowledge of genetic and environmental effects on human characteristics. We do indeed know a lot about it, but we do not know even close to as much as the nurturists and the genetic determinists actually claim or imply. On the genetic front, there is little doubt that genes control (affect would be a better term) the development of the brain. Since the brain and the peripheral nervous system determine human behavior, then in some sense it is undeniable that we behave because of our genes. Indeed, it would be foolish to deny this possibility where humans are concerned, since it has been ascertained for every other animal that has been sufficiently studied. A large and fascinating literature on brain damage provides direct and incontrovertible evidence that alterations in the physical structure of the brain (which can be brought about by accident or by mutations) directly affect all sorts of human behaviors, including subtle personality traits (Gazzaniga 1998; Ramachandran and Blakeslee 1998; Damasio 1999) . Nurturists routinely make fun of the idea of genes affecting homosexuality or religious beliefs. But there is no doubt that both homosexuality and religious beliefs do have a physical basis in the brain (and how could it be otherwise, since they are human behaviors?), and are therefore alterable by genetic mutations, at least in principle. The topic of homosexuality is exceedingly controversial and emotional, but it does represent a great example of why the nature-nurture controversy is important, and at the same time of why neither camp is “right”. It is quite clear that homosexual behavior can be acquired or lost, so in some sense there must be environmental components to it. At the same time, the level of exposure to male or female hormones in the womb also very significantly affects gender preferences and sexual behavior of the adult individual (for a lively tour of the implications and references see: Moir and Jessel 1992) . While hormonal levels can be considered an environmental factor, the way in which the brain responds to them is definitely genetically hardwired (because it depends on the presence and expression of hormonal receptors). Perhaps one of the most emotionally charged (and actively avoided by most scientists) areas of inquiry concerns religion (Larson and Witham 1997; 1998; Pigliucci 1998) . It would be foolish to claim that the environment, given all we know about the psychology of individual humans and the sociology of human societies, does not preponderantly affect religious beliefs. Nevertheless, Ramachandran (1998) reports that micro-seizures in the temporal lobes are associated with unusually vivid religious experiences, often complete with sounds, visual stimuli, and a sense of sudden cosmic “understanding”. It is easy enough to imagine genetic mutations making some individuals more prone to such seizures, and therefore genetically affecting their propensity for religious beliefs. In both the cases of homosexuality and of religious beliefs, it doesn’t make any sense to take an ideological posture and defend it regardless of the evidence. However, the data also do not allow us to disentangle the relative contributions of nature and nurture in a neat and simple way, they only lead us to conclude that both components must be there. A recent example of a scientifically sound investigation of the biological basis of complex human behaviors is the study of the peculiarities of Einstein’s brain when compared to appropriate controls from the general population (Witelson et al. 1999) . As it is well known, Einstein was gifted with an unusual mathematical ability; it is also well known that in humans increased mathematical ability is associated with an expansion of the inferior parietal region of the brain. Witelson and colleagues found that Einstein’s brain was about 15% larger than the controls precisely in that region and therefore concluded that they may have pinpointed at least part of the biological basis of the great physicist’s rare ability. This, of course, is not to say that Einstein’s mathematical genius was written in his genes: a change in brain structure can be brought about by mutations or by environmental influences (or both). Yet, it is uncanny to see how several immediate reactions in the mass media were along the lines of “reductionist” science attempting to “diminish” the contribution of Einstein to human thought. How would the demonstration that such contributions came from his brain in any way diminish Einstein’s figure? Where else could his genius have been coming from? Along similar lines, Tang and coworkers (1999) genetically engineered mice to overexpress the NMDA (N-methyl-D-aspartate) receptor 2B in their forebrains. NMDA receptors are synaptic coincidence receptors that are known to play a fundamental role in memory formation and learning. Not surprisingly, the transgenic mice showed enhanced cognitive abilities, perhaps opening an important window toward both the understanding of the complex phenomena of learning and memory and the treatment of debilitating human diseases affecting them. Since a gene codes the NMDA 2B receptor, this is a pretty clear demonstration that higher cognitive abilities in humans are indeed under genetic control. It does not follow, however, that this single gene (or any other gene, for that matter) is all there is to know in order to understand cognition. A clear, simple, and largely unemotional example of why both nurturists and genetic determinists are partially right, but neither can address the most important questions, is given by the environmentally curable genetic disease known as Phenilketonuria (PKU). PKU is caused by a simple “inborn metabolic error”, i.e. a mutation that does not allow the formation or proper functioning of an enzyme that metabolizes the amino acid phenylalanine. Often (though not always)[iv] this results in accumulation of the amino acid in the brain during development, which in turn causes a host of phenotypic and behavioral effects, including severe mental retardation. The link between the genetic level (the mutation and consequent enzymatic defect) and the phenotype/behavior is unusually straightforward, and therefore offers an ideal example of how genes can affect behaviors, although no genetic determinist would go so far as to say that this is a gene for normal intellectual development. On the other hand, a very simple dietary change (i.e., a change in behavior) can completely neutralize the genetic effect: an individual can grow up normally by carefully avoiding phenylalanine, which is why many soda cans carry a warning label for PKU patients. This is the most extreme and one of the few clear cases of plasticity one can document in humans. So, one has to conclude that PKU is both genetically determined and environmentally plastic. The crucial question, however, still remains: what are the reaction norms of PKU genotypes? There surely are several PKU genotypes, not only because the mutation can be caused by different alleles at the main locus, but also because the effects of the mutation depend on the other genes that interact with that locus. Furthermore, there may be plenty of other environmental circumstances that affect the degree of occurrence and gravity of PKU symptoms. We simply do not know, and we will not know any time soon given the obvious limitations on experimental research in human genetics. Summarizing, we can (and in many instances we do) know that a trait is affected by changes in the genetic makeup of a person. Similarly, we can and do know that a given characteristic or behavior can be altered by changes in the environment. What we do not know is if and how these two effects interact with each other. Unfortunately, this is the single most crucial thing there is to know about the nature-nurture controversy. Why we don’t know what really matters The fact that the quantitative genetics of human behavior is far less advanced than most “experts” would like the public to think is a logical consequence of what we have been discussing in this book about genotypes, environments, and phenotypic plasticity. It can be summarized in a few simple, yet apparently incredibly difficult to sink in, sentences. First of all, what we would really like to know is how genes and environments interact to produce a certain phenotype or behavior in an individual. Given that this has so far proven to be a very difficult and elusive goal, human quantitative geneticists have shifted the target to a general (statistical) understanding of variation within and among human populations. Enter the venerable concept of heritability, i.e., the ratio between the genetic and the phenotypic variance for a particular trait, measured by a variety of breeding designs. The result of this shift is that, assuming a given trait turns out to have a heritability of 70%, one cannot in any meaningful way say that 70% of that trait in a particular individual is due to her genes and the remaining 30% to the environment. All one can say – under ideal circumstances – is that 70% of the phenotypic variation in the population is associated to differences in the genetic constitution of its individuals. The problem is that for humans the circumstances are never “ideal,” since measurements cannot be taken under the same set of environmental conditions. This is Lewontin’s mantra: heritability is only a local measure; it estimates the amount of genetic variation for a quantitative trait only within a particular population (set of genes) and in a particular environment (Lewontin 1974, Figure 1) . It cannot be reasonably extrapolated either to other populations or to other environments without further empirical research. Using again the concept of reaction norms, it is rather easy to understand where the problem lies. The extreme genetic determinist position (which, of course, nobody really espouses, but which provides a convenient conceptual endpoint) can be visualized as in Figure 2. The three reaction norms are flat, indicating that there is no effect of the environment whatsoever. They are also widely spaced apart, suggesting that genetic differences do result in major phenotypic/behavioral differences. If this were indeed the case, no social program of any kind would be worth a dime. However, we know that this is not the real scenario, because we have plenty of evidence supporting the existence of some significant environmental effect on human behavior. The extreme nurturist hypothesis is instead the one depicted in Figure 3. Here the reaction norms are all sloped, indicating a dramatic effect of the environment. They are also so close together in a bundle that there basically are no differences among genotypes. I hope that even the staunchest supporter of humane social policies, however, will admit that this scenario is as unlikely and inconsistent with the scant data we have as the previous one. What is the real shape of human reaction norms? I would guess something approaching Figure 4: some genotypes are almost not responsive to environmental changes, others respond significantly, and still others are dramatically affected by the environment. Furthermore, the differences among genotypes are large in some environments (e.g. at the left and right extremes of the graph), but small in others (e.g. in the center of the hypothetical environmental range). If this were the case, the truth would indeed lie somewhere in the middle, though such middle would not be a simple average of genetic and environmental effects. However, the fact of the matter is that, again, we do not know. While we can confidently exclude both extreme scenaria, there are an infinite number of potential configurations in the middle. Since we cannot ethically grow genetic replicates of human beings under controlled and diverse environmental conditions, we simply do not know what the patterns of plasticity for human cognitive traits are. Scientists should acknowledge this and move on. But the temptation is apparently too great for us not to sin. A characteristic example of how difficult it is even for renowned biologists to get over what Feldman and Lewontin (1990) called “the heritability hang-up” can be found in works by sociobiologist Edward O. Wilson. In his “Consilience: the Unity of Knowledge” (1998) he explains his theory of gene-culture evolution, i.e., how genes and environments co-evolve to shape species’ destiny, humanity in particular. The chapter entitled “From Genes to Culture” is an odd mix of an enlightened vision of the problem and a stubbornly orthodox one, as if Wilson couldn’t make up his mind. After admitting that “all biologists speak of the interaction between heredity and environment”, he goes on to introduce the concept of reaction norm. He correctly points out that: Redefined with the more precise concepts of genetics, nurturists can now be seen to believe that human behavioral genes have very broad norms of reaction, while hereditarians think the norms are relatively narrow. In this sense the difference between the two opinions is thus one of degree, not of kind. It becomes a matter that can be settled and agreed upon empirically. Wilson must know that the empirical way to settle this matter would be to selectively breed humans and raise them under controlled conditions, an ethical impossibility. Nevertheless, he expresses some optimism that the matter will eventually be settled through the accumulation of data while not providing a clue as to how this would be done. Wilson also brings up the concept of heritability. Even though he expressly admits that there is a genotype-environment correlation that makes the use of heritability measures inadequate, and that heritability is ‘flexible’ (i.e., it depends on the environment), he calls these “peculiar twists”. While coming so close to destroying (or at least greatly reducing) the very meaning and usefulness of heritability, he still affirms that “the measure has considerable merit, and in fact is the backbone of human behavioral genetics”. He proceeds by imagining a series of scenaria in which we would be able to measure heritability in different cultural (and therefore environmental) contexts, predicting that in some cases heritability would increase, in other it would decrease. Let me be clear on this point: such predictions have no foundations whatsoever given current procedures in empirical quantitative genetics. The only way to make them in a reliable way is either to do the experiment, which is both ethically impossible and technically extremely difficult, or to understand exactly how genes and environments interact from a mechanistic standpoint. The latter objective is not as of now even within the range of the most powerful telescope that any scientist may hope to use to divine the future. Worse, immediately after providing the reasons for being extremely cautious about heritability measures, Wilson says: Nurturists have also traditionally thought that the heritability of intelligence and personality traits is low, while hereditarians have considered it to be high. That disagreement has largely been solved. In contemporary Caucasians of Europe and the Unites States at least, heritability is usually in midrange, with its exact value varying from one trait to another. It is astounding to see this kind of reasoning on the part of a biologist who demonstrably knows better. Unfortunately, he is not alone. I think it would be of great interest to psychologists and sociologists to try to explain how one can have all the pieces of the puzzle in mind, even put them together in almost the right way, and still fail to perceive the emerging picture. As it is clear when one considers the environmental plasticity of heritability measures (see below; and as Wilson himself freely acknowledges), it just does not make any sense to use heritability across environments. Since the Caucasians to whom Wilson is referring were raised under different environments, the resulting estimates of heritability of IQ are a hopelessly confounded hodge-podge of nature and nurture. They emerge from genotype-environment interactions during the entire lives of the organisms concerned, not just from the genetic component as it is simplistically assumed. The only confidence limits that can be reasonably put around such “estimates” are zero to one hundred percent, not exactly an example of good empirical science. The environmental dependence of heritability is by no means a mere theoretical possibility. In many organisms in which the experiment can and has been done, a strong dependence of heritability on the environment has been found for many (albeit not all) traits. For example, the work of Mazer and Schick on wild radish (Figure 5) clearly shows that the heritabilities of three traits (flowering time, petal area, and pollen volume) are highly dependent on the density of conspecifics, a crucial environmental parameter in plants. More importantly, there is no regular pattern describing such interdependence: heritability is lowest under high density for flowering time, but under low density for petal area, and under medium density for pollen volume. Therefore, one cannot even say that a particular environmental range is likely to yield high heritabilities while another is associated with low heritabilities. It depends on the character (and, probably, on the population in which the study is conducted). This may be a frustrating aspect of biol...