Top critical review
88 people found this helpful
Neo-Lysenkoism for dummies
on October 27, 2012
"Sophisticated people sneer at feel-good comedies and saccharine romances in which all loose ends are tied and everyone lives happily ever after. Life is nothing like that, we note, and we look to the arts for edification about the painful dilemmas of the human condition. Yet when it comes to the science of human beings, this same audience says: Give us schmaltz!"-Steven Pinker
1. MISUNDERSTANDING HERITABILITY
In the first chapter of The Genius in All of Us David Shenk writes that Charles Murray and Richard Herrnstein (henceforth, M&H), the authors of The Bell Curve, "fundamentally misinterpreted a number of studies, becoming convinced that roughly 60 percent of each person's intelligence comes directly from his or her genes." For at least two reasons, this claim is not a very auspicious start for a book that purports to challenge the vast research literature showing that genetic differences are important determinants of success. Firstly, no such claim is made anywhere in The Bell Curve. On the contrary, M&H write (p. 106) that "heritability describes something about a population of people, not an individual. It makes no more sense to talk about the heritability of an individual's IQ than it does to talk about his birthrate." Secondly, the fact that Shenk nevertheless thinks that M&H's thesis is that "60 percent of each person's intelligence comes directly from genes" indicates that Shenk does not understand what it means to say, as M&H do, that the HERITABILITY of a trait is such and such.
Heritability is a measure of the extent that DIFFERENCES among individuals have genetic causes, not a measure of the extent that genes have contributed to any individual's ontogenic development. For example, the heritability of height is at least 80 percent in first world countries, but that does NOT mean that 80 percent of any person's (or the average person's, as Shenk appears to interpret it) height is due to genes and the remaining 20 percent due to the environment. Heritability is a population statistic indicating the causes of VARIATION between individuals. Shenk doesn't understand this, and his discussion of genes and environments throughout the book is seriously handicapped by this fundamental mistake. Unfortunately, this is far from being the only instance of conceptual misunderstanding and misrepresentation in the book.
2. OVERSTATING GENOTYPE-ENVIRONMENT INTERACTION
Shenk continues by claiming that M&H are mistaken because they are unaware of genotype-environment interactions (GxE for short). This assertion would undoubtedly have surprised the late Richard Herrnstein who wrote about GxE at least as far back as in his 1973 book I.Q. in the Meritocracy, but let's leave that aside for the moment. Shenk writes that there's recently been a revolution in genetics: scientists used to think that genes cause phenotypes deterministically and in exactly the same way across all environments, but now they have realized that phenotypes in fact develop through complex interactions between genes and environments, or GxE. This account of the history of genetics is, of course, farcical, and bears no resemblance to reality. There's been no such revolution, and geneticists (as well as the general public) have always understood that interactions between genes and environment are crucial for the development of phenotypes. I'll write more about Shenk's bizarrely distorted view of the history of genetics later in this review, but I'll give one illustrative example here. He cites a 1957 study by Stanford anatomist William Walter Greulich which found that Japanese Americans grew up to be considerably taller than their cousins in the old country. Shenk suggests that Greulich was mystified by this result, but if you read the actual study, you'll see that Greulich sensibly attributed the difference to better nutrition in America. Astonishingly, Shenk appears to think that just about everybody was unaware of the link between malnutrition and stunted growth even as recently as the 1990s.
Over the last century, there have always been people who have maintained that it's a mistake to speak of nature and nurture as independent forces or to wonder whether it's genes or the environment that is more important in determining human differences. Joining this long line of people who don't understand quantitative and behavioral genetics, Shenk asserts that both genes and environment are needed and that it's wrong to speak of them as separate sources of influence. This is, of course, true and self-evident in a certain sense: an organism cannot exist if it doesn't have both a genome and an environment. Moreover, some individuals are, inevitably, born and raised in more favorable environments than others. However, this sort of GxE interaction poses no problems for quantifying the independent effects of genes and environment because heritability is not estimated at the level of an individual but rather at the population level. If you look at a single individual, trying to disentangle the effects of the environment and genes in producing a particular phenotype is indeed often a hopeless task. That's why quantitative geneticists step back from a single individual and instead look at a representative sample of individuals, asking whether similarities and differences in genotypes and environments correlate with phenotypic similarity among them. In animal and plant research, environments and genotypes can be directly manipulated (by using inbred strains and controlled environments, for example) in order to estimate heritability whereas in human research natural experiments (such as twinning and adoption) are utilized.
M&H noted in The Bell Curve that the more equal the environmental circumstances in America get, the more strongly differences in intelligence will be determined by genes. Shenk writes that they're mistaken and don't understand how genes work. Here as in many other places in the book, one wonders at the size of David Shenk's ego. He thinks that until recently, almost nobody doing research on these topics realized any of the simple ideas he presents in his book. Had he actually bothered to familiarize himself with the research on which M&H draw, he would have understood that the claim of greater importance for genes in more equal environments is a truism. It follows from the fact that at the population level genetic and environmental sources of variation are complements: when one decreases, the other increases, and vice versa. Thus if (trait-relevant) environmental variation decreases in society, then genetic variation will, by mathematical necessity, determine a larger share of phenotypic variation.
So, it's trivially true that both genes and environments are needed, and that some environments are better for everyone than others. GxE in this sense does not pose great problems for behavioral genetic analysis. However, there is a special sense in which GxE may complicate heritability estimation. This is when the same environment affects different genotypes differently, or, to put it another way, if the sensitivity of different genotypes to the same environment varies. This is usually referred to as 'statistical GxE' to differentiate it from the common, trivial GxE discussed above. Statistical GxE concerns such questions as whether genotype G1 is superior to genotype G2 universally, that is, across all known environments, or whether there are environments where G2 is less inferior, equal or, especially, superior to G1. For example, it could be that in environment E1 individuals with G1 will have a higher phenotypic value P than those with G2, whereas in environment E2 the rank order of individuals on P will be reversed. If strong GxE interaction effects in the statistical sense are present, then it will indeed be impossible to give a meaningful answer to the question of the relative importance of genes and the environment, and heritability estimates may become biased (depending on the research design).
The book discusses a famous example of statistical GxE, the Cooper and Zubek rat study. This old study from the 1950s is often cited because there are very few, if any, other comparable examples of statistical GxE in the literature. Cooper and Zubek had two inbred strains of rats, one of them "maze-bright", the other "maze-dull." They found that the bright rats beat the dull ones in a maze task when both were reared in a normal environment, but the two strains performed similarly when both were reared in either restricted or enriched environments. Shenk interprets this study as a seminal confirmation of GxE. However, in their paper Cooper and Zubek warn against making strong interpretations of the results until the study is replicated. Among other things, they note that the maze task may have been too easy to differentiate between the two strains in the enriched condition; if one of the strains was actually brighter, it could not have been detected. Tellingly, Cooper and Zubek's results have never been reproduced.
The fact that it's theoretically possible that statistical GxE is an important source of individual differences does not mean that it actually is. Whether statistical GxE effects exist for any particular trait is an empirical question. Behavioral genetic research on a wide range of cognitive, personality, and other human traits indicates that statistical GxE is at best a trivial source of variation, accounting for at most a few percent of population variation, and that it therefore cannot bias heritability estimates. For example, twin and adoption studies have demonstrated that statistical GxE effects for IQ are either very small or non-existent. However, the most powerful and direct refutation of the idea that statistical GxE interactions bias estimates of IQ heritability has come from recently developed DNA-based methods of heritability estimation. These new studies show incontrovertibly that the contribution of additive genetic effects to IQ variation is around 50 percent (and this figure is certainly an underestimate of total heritability because the method ignores the effects of rare genetic variants, dominance, and epistasis). The heritability of IQ in childhood is about 40 percent, and it increases linearly with age, reaching about 80 percent in adulthood. These estimates are not biased by any statistical GxE effects. (While it is less relevant for the sorts of phenotypes Shenk's book deals with, there is one domain of human behavior where some evidence exists for substantial statistical GxE effects. A handful of studies have indicated that there are mental and behavioral disorders, e.g., depression, whose occurrence may, to some extent, be dependent on interactions between vulnerable genotypes and bad environments.)
Characteristically, Shenk does not realize that the term GxE is (admittedly confusingly) used in the genetics literature in these two different, "common" and "statistical", technical senses and that the implications of the two are very different. After discussing the rat study, he lists other examples of GxE, such as how temperature around eggs determines the sex of turtles and snakes, or how the color of a grasshopper's skin is influenced by the environment. However, these other examples are just ordinary environmental effects that have nothing to do with statistical GxE. For Shenk, GxE is a buzzword that needs not be precisely defined and can mean just about anything. At times the book's claims about GxE are so expansive that it appears, as one commentator put it, that Shenk would be genuinely surprised if his own children looked anything like him or had any mental traits in common with him.
3. FUNHOUSE VERSION OF THE HISTORY OF GENETICS
The most curious thing about Shenk's book is how he has managed to get the history of genetics so wrong. For example, in his discussion of several works published in the 1990s, he claims that the authors attribute such a large role to genes in behavioral differences only because in the 1990s no one knew about interactions between genes and environment. However, the idea of GxE interaction (in the non-statistical sense) was formally introduced by the German zoologist Richard Woltereck in 1909. He called it the norm of reaction, and it has long been covered in standard textbooks. The absurdity of the claim that GxE in this sense was unknown in the 1990s becomes apparent when one considers that it would suggest that in the 1990s scientists, not to mention the general public, did not understand that, say, crop yields could be improved with fertilizers and pesticides.
The first investigation into GxE in the statistical sense was R.A. Fisher's 1923 experimental study on the effects of different manurial treatments on several varieties of potato. He concluded that there was no statistical GxE (he called it "non-linear interaction of environment and heredity"). Fisher always maintained that statistical GxE effects were unimportant because they were so rare. The case for the importance of statistical GxE was first made by the British zoologist Lancelot Hogben in the 1930s. He drew particularly on a study of fruit flies that showed evidence of (weak) statistical GxE. In contrast to Fisher, Hogben conjectured that statistical GxE effects are widespread in nature, making it an important third source of population variation besides genes and the environment. Later scientific disputes about GxE, such as the IQ debate of the 1970s in which Richard Lewontin, David Layzer, and others were pitted against Arthur Jensen, were essentially reprisals of the original Hogben-Fisher debate, with little more empirical evidence to back up the notion that statistical GxE could be important for human behavioral differences. As noted above, research since the 1930s has amply vindicated Fisher's skepticism about the practical significance of statistical GxE. Books about how everybody can be smarter and better than the average usually draw heavily on the arguments of Lewontin et co., but Shenk's doesn't because he seems to be almost completely oblivious to this long intellectual history, laboring as he is under the misconception that GxE was discovered in the 1990s or so.
Far from being some new revolutionary idea, GxE has been in the conceptual toolbox of geneticists ever since the beginnings of the science. Some vulgar Mendelians of the early 20th century may have had ideas of strictly deterministic genes, but Shenk's argument that this was the reigning orthodoxy in behavioral and developmental genetics until recently is utterly absurd.
4. THE FALLACY OF "DELIBERATE PRACTICE"
Shenk is infatuated with Anders Ericsson's theory of "deliberate practice", according to which the amount and quality of practice determines who will or will not excel in any given field, and innate differences have almost nothing to do with it. Shenk writes that the trick to becoming an expert or a top performer is to be extraordinarily motivated, unfazed by disappointment and lack of progress, and able to delay instant gratification. But this is simply a description of what high achievers are like. It says nothing about WHY they are like that. Just giving a name like deliberate practice to something that high achievers do doesn't explain anything. To think it does is the nominal fallacy.
Shenk writes that not only is it possible to become a high achiever in any field by practicing for 10,000 hours, but also that it is not possible to do it in less time. He cites research on chess players by Guillermo Campitelli and Fernand Gobet who indeed found that titled chess masters needed an average of about 10,000 hours to reach that level. However, note the word 'average' there. Shenk has fallen for the ecological fallacy, that is, making inferences about individuals based on the aggregate properties of the group to which they belong. Campitelli and Gobet have also found that there's a lot of individual variation in how much practice is needed to achieve master level in chess. Some players reach that level in as few as 3,000 hours of practice, whereas others need much more than 10,000 hours. Still others fail to reach master level even after 30,000 hours. The same pattern holds everywhere, not just in chess: some learn things much easier and much quicker than others.
Shenk cites an article by Anders Ericsson published in The International Journal of Sport Psychology where it is claimed that with the exception of body size, genetic differences do not contribute to elite sports achievement. He doesn't cite the article by Klissouras and colleagues, published in the same issue of the same journal (Vol. 38, Issue 1), who report, drawing on twin studies, that "individual differences in most functional abilities, morphological characteristics, motor attributes, personality and cognitive traits linked to superior sport performance are substantially influenced by genetic factors" and that "genetic influence is so ubiquitous and persuasive that we ask not what is heritable but what is not heritable." Stubbornly ignoring such confounding genetic variables indicates that one is only interested in promoting an agenda.
The book's account of the development of excellence is purely descriptive, ending where it should properly begin. Why is it that some people have the energy and the will to practice day in and day out, year after year? Are there ways to becoming more conscientious and motivated? Why do some people reach a top level of performance so much faster that others who work equally hard? Why are there large differences among top performers in any field, independent of the amount of practice? Most of these questions are not even raised in the book.
One of the most replicated and most interesting findings from behavioral genetics is that for IQ, personality, and many other complex traits, heritability increases with age, while the effect of the childhood environment pretty much completely fades away by adulthood. It is believed that this is because genetic differences express themselves only gradually during development, and/or because growing up, people increasingly select environments that better match their natural inclinations, resulting in positive or negative feedback loops. This would tie nicely into Shenk's discussion of how child prodigies often fail to grow up to be high achievers in adulthood, but ignorant as he is of (and hostile to) behavioral genetics, this connection is never made in the book. In general, it's of course expedient for Shenk to ignore the fact that genetic factors become more, not less, important with age, as this finding is difficult to fit together with his overarching interactionist argument.
Charles Murray, who serves as an archetype of the wicked hereditarian in Shenk's book, has argued that the smart and successful are so mainly due to their luck in the genetic lottery, and should therefore be educated in the spirit of noblesse oblige. In contrast, Shenk maintains that genes have little or nothing to do with success, and that anyone who has failed to succeed has only themselves to blame; the successful are simply morally superior. I think Murray's view not only makes much more sense but is also more humane than Shenk's. Where does Shenk think that the personal qualities that he says lead to excellence -- such as perseverance and zeal -- come from? He seems to believe that people freely choose what their personality is like. This presupposes an extent of control over one's own behavior that people simply don't have.
5. UNDERESTIMATING THE POWER OF IQ
The measurement of intelligence is perhaps the greatest success story of psychology. IQ tests are highly reliable instruments measuring a stable human characteristic which has a predictive power stronger and more universal than that of any other social science variable. Shenk grudgingly admits as much, but then argues incoherently that it does not matter and that IQ is in fact highly malleable. The stability of IQ is an empirical fact. To take an extreme example, in a unique study by Ian Deary and colleagues, the correlation between IQ measured at ages 11 and 80 in a Scottish sample was 0.63. With statistical corrections for attenuation and range restriction, the correlation was as high as 0.80, which represents a truly remarkable level of stability over 69 years, especially as IQ is not yet fully developed at age 11 and by age 80 many people suffer from neurodegenerative diseases.
Here, as elsewhere, Shenk is hobbled by his inability to conceptualize things at different levels of analysis. At the level of a particular individual, IQ may indeed show substantial instability, particularly in childhood. The reasons for this are many and include, for example, individual differences in childhood developmental trajectories, accidents of test situation, illnesses and injuries, training effects, and random measurement error in general. However, most people do not experience large score discrepancies between tests, which ensures stability at the population level.
Shenk claims that the large shifts in IQ that some individuals experience mean that we know how to change IQs. This is not correct. Despite much effort, we still do not know how to produce intelligence gains of significant magnitude. We know that test scores may be increased by suitable training, but such coaching does not make people smarter and over time these hollow gains fade away completely. We know that while genetic factors explain the majority of IQ differences, environmental effects play a role, too, but what those effects might be remains a mystery.
The book discusses Hart and Risley's oft-cited study where it was found that children whose parents talk more to them tend to have higher IQs. Shenk is convinced that this correlation represents a causal effect. He quotes the American Psychological Association's 1996 report on intelligence research where it was noted that the results of Hart and Risley's study are causally ambiguous because such correlations may be confounded by genetic effects. Shenk contemptuously dismisses this argument, and asserts that the correlation must be causal because of GxE. It seems that for him, GxE is a magic word that can miraculously convert any correlation into a causal effect, without any need for evidence of causality. The APA report correctly notes that causal interpretations of correlations between IQ and the family environment are difficult to reconcile with evidence from twin and adoption studies. Research shows that after early childhood the shared family environment has a negligible influence on IQ differences, outside of extremes of environmental deprivation. What renders any hypothesis about strong family effects on IQ prima facie implausible is the fact that the IQs of siblings from the same family vary widely, and that the IQs of unrelated people raised together are not more similar than those of two randomly selected people from the same population.
The book also describes Robert Sternberg's "triarchic" intelligence theory. Sternberg's major claim is that IQ tests measure only "academic intelligence", whereas real-life success is dependent on "practical intelligence." In support of this theory, Sternberg has presented evidence showing that, for example, highly experienced manual workers are superior at the sort of mental calculations needed in their jobs than higher-IQ but inexperienced white-collar workers. Shenk asserts that these sort of findings "pose a serious challenge to research psychologists adhering to traditional definitions of intelligence." Of course, he doesn't actually discuss what such research psychologists think about Sternberg's theory.
Perhaps the most thorough refutation of Sternberg's theory was published in issue 4, 2003, of the academic journal Intelligence. In their articles, Linda Gottfredson and Nathan Brody showed that all of Sternberg's findings are explainable in terms of traditional psychometric intelligence theory. They note, for example, that Sternberg's examples of "practical intelligence" represent narrow skills honed through lots of practice. Research has established that general intelligence is very difficult to train, but narrow, task-specific skills and abilities are of course trainable. Shenk himself gives a good example of this in his discussion of the experiment where a man was able to greatly improve his memory for sequences of numbers through intensive practice, but this did not improve his other skills at all, not even the ability to remember sequences of letters. The claim of IQ, or general mental ability, to being a valid operationalization of intelligence rests precisely on the fact that it is not a narrow trained skill but a highly general, context-independent capacity. Higher IQ makes it easier to solve novel problems and to acquire every kind of knowledge and expertise. As a result, those with higher IQs tend to develop a broader and deeper expertise and have careers commensurate with that, with no need to learn skills required in some particular low-skilled job. Sternberg claims that the greater the overlap in the specific skills measured by a test and those needed in a job is, the greater is the test's predictive validity for that job. In fact, the contrary is true: tests of general mental ability with no job-specific content are clearly superior predictors of job performance than tests assessing specific job-related skills.
In his reanalysis of Sternberg's data, Brody dealt a coup de grâce to Sternberg's research program by showing that the intelligence tests Sternberg had developed in order to measure three supposedly independent domains of intelligence were in fact correlated with each other. Damningly, the predictive validity of the tests was entirely a function of the extent that they measured general mental ability.
Shenk asserts that "[i]ntelligence scores of infants are not predictive of future scores or life success." This is a key claim in his critique of IQ, but alas, it's not true. For example, Joseph Fagan and colleagues developed a test of visual memory (which is a pretty good proxy for general mental ability at any age) and then showed in a longitudinal study that the disattenuated correlations of six-to-twelve-month old infants' visual memory skills with adult IQ and academic achievement were 0.59 and 0.53, respectively ("The prediction, from infancy, of adult IQ and achievement." Intelligence, Volume 35, Issue 3, May-June 2007, Pages 225-231). Thus infant test scores do predict adult IQ and achievement. Cognitive inequality is evident already in the cradle.
The book's treatment of the Flynn effect, the observed increase in mean IQ scores across generations, is predictably misinformed. The book appears to argue that the Flynn effect represents an actual increase in intelligence, but statistical analyses show that measurement invariance does not hold in between-generation IQ comparisons. This means that IQ tests do not give unbiased estimates of the IQs of older cohorts when compared to those of younger cohorts. This is somewhat similar to giving an English-language IQ test to a non-native speaker: the obtained IQ score will be downwardly biased and will not reflect the true level of ability. The past is a foreign country in the case of IQ, too. It may be that people have indeed got smarter over the last few generations -- just as the average height has increased -- but this cannot be established based on simple IQ comparisons.
Shenk makes much of the fact that relatively few of the participants in Lewis Terman's classic study of intellectually precocious children made significant creative achievements as adults. He concludes that IQ is thus not associated with creativity. However, as Murray and Herrnstein demonstrated in The Bell Curve, the social and economic value of high cognitive ability increased enormously during the 20th century. Terman's "Termites" were born around 1910, and in those days it was normal for highly intelligent individuals to end up in unremarkable jobs. In the second half of the 20th century the intellectual sorting of American society rapidly intensified as universities and employers started seeking out high ability individuals all over the country. The results of this change can be seen in the Study of Mathematically Precocious Youth, or the SMPY, an ongoing longitudinal study similar to Terman's but started in 1971 (despite the name, it is not limited only to the mathematically gifted). The results of the SMPY are quite different from Terman's study. What the SMPY has established is that intelligence test scores measured in childhood are robust predictors of later creative accomplishments, including literary works and scientific articles published, PhDs and patents earned, and attainment of tenure-track positions at top research universities.
6. CONFIRMATION BIAS RUN AMUCK
Throughout the book, Shenk's discussion of relevant literature is highly selective and one-sided. He attempts to defend himself by saying that he has a particular viewpoint that he wants to promote and that he relies on those studies and researchers he has found to be the most credible. While I don't mind books making a bold case for something, I think it's reasonable to expect a partisan author to engage the evidence and arguments of the other side and try to refute them. Shenk makes no such attempt. He has not talked to those with views different from his, nor familiarized himself with their work. He misunderstands basic concepts and ideas, constructing straw man versions of opposing arguments to be knocked down. He has simply followed his confirmation bias by seeking out people and studies that lend support to the views he had from the outset. He accepts even the most implausible claims of his favorite authors, while ignoring other views and contrary evidence.
One example is his criticism of the Minnesota Twin Study which found that identical twins separated in infancy grew up to be remarkably similar in numerous ways. He tries to dismiss these results by bringing up the well-worn criticism that many of the twin pairs had been reunited long before the study, arguing that this, rather than shared genes, was what explained the similarities. What he doesn't mention is that the authors of the Minnesota study have responded to this criticism by analyzing whether those twins who had known each other longer were more similar than those who had only recently met. The answer was negative, refuting the criticism. Shenk's criticisms of twin studies are uninformed throughout, and he does not seem to even realize that the results of twin studies have been corroborated by other research designs, such as those based on adoptees and extended kinships.
Shenk repeatedly acclaims that genes are not meaningless, that he's not claiming that humans are blank slates. However, this is mere obfuscation, because the entire book is an attack on the idea that genes are a significant cause of human differences. Shenk's neo-Lysenkoism reaches a crescendo in the last chapter dealing with epigenetics. He claims, among other things, that you can make your future kids innately smarter by being intellectually active. Needless to say, there is no evidence for such an absurd suggestion. In recent decades, cumulating behavioral genetic evidence has dismantled the intellectual edifice of anti-hereditarian thinking. Preposterous, unsupported claims made about epigenetic effects are the last, desperate refuge of this school of pseudo-science.
To sum up, Shenk is hopelessly confused about several key concepts; is astonishingly ignorant of the history of the nature/nurture debate; ignores or misrepresents research that contradicts his claims; places enormous trust on anecdotes and the results of never-replicated experiments; ignores entire research traditions and the best, methodologically most sophisticated studies; and thinks that if he gives a name to a phenomenon he has explained it (i.e., the nominal fallacy). The book is permeated by relentless optimism about human potential but the ideas in it do not support the optimism. Shenk's toxic combination of arrogance and ignorance and the fact that many people have bought into his brand of snake oil make me a bit more pessimistic about humans.