Heritability: a handy guide to what it means, what it doesn’t mean, and that giant meta-analysis of twin studies

CT_Quartz-IQ_SPOTLIGHTby Jonathan M. Kaplan

Recently, Nature Genetics published a paper by Polderman et al. entitled “Meta-analysis of the heritability of human traits based on fifty years of twin studies,” [1]. It has already garnered a lot of attention: blog posts and various news sites are trumpeting its conclusions, and putting their own spins on the results. For reasons I’ll briefly discuss below, I found the paper to be very strange, but my main purpose in this post isn’t to criticize it. Rather, I wish to briefly explain, in broad terms, what “heritability” actually means, and perhaps more importantly, what it doesn’t mean. The twin studies analyzed by Polderman et al. attempted to estimate the heritability of various human traits. While one can, for reasons I’ll note briefly below, raise concerns about how accurately heritability can be estimated from twin studies in humans, the more important point is that heritability doesn’t measure what most people think it measures, and while twin studies may be answering some questions, they aren’t, and can’t, answer the questions that many people think they are answering.

I. An odd study

Before getting into heritability, the heart of the matter, I want to reflect briefly on a few features that make the new study somewhat puzzling. The first features I focus on here are related to the study’s design: a meta-analysis of twin-studies. The second set of features are a bit more general.

A. What are meta-analyses for?

Meta-analyses, as the term implies, analyze already published studies by combining the results of a number of individual studies focused on similar experimental outcomes together, in order to generate some overall measure of the results. There are two sorts of situations in which researchers regularly run meta-analyses: a) where previous studies on the issue at hand have reached conflicting conclusions, and b) where previous studies record results that fail to reach statistical significance (or are just barely significant). In both sorts of cases, the hope is that an analysis of all the studies published so far will generate results that are more robust than any individual study.

In the case of the Polderman et al. paper, however, neither purpose applies. It is well established that twin studies estimating heritability of traits in humans generally find a substantial heritable component for essentially all traits measured [2], nor is the statistical significance of these findings in doubt (although the methodology that generated those findings, and therefore the validity of the findings themselves, is certainly questionable). Performing a large-scale meta-analysis of this sort (over 2500 studies) is a herculean effort, but in this case, the reason why it was done is not explained.

B. What is the point of the conclusion?

One finding that the authors chose to mention in their abstract is that “across all traits the reported heritability is 49%.” This fact got picked up by many of the aforementioned blogs and media sites (“Nature versus Nurture a draw” is a typical headline; this, for reasons discussed below, is wrong on a number of levels). But the heritability of the least heritable traits measured was only around 5%, and the heritability of the most heritable traits over 80%. What information is gained by averaging the heritabilities of traits as disparate as “adult height,” “structure of the eye-ball,” “cognition,” and “social values”?

For comparison, imagine a similar meta-analysis of drug effectiveness that reported the average effectiveness of every randomized clinical trial of a drug over the past 50 years. Of what possible use would it be to know the average effectiveness of various drugs, designed to treat different diseases or disorders? All the different trials share is a vague area of interest (“drug effectiveness”) and a basic methodology (“randomized clinical trials”). That isn’t enough to support the coherence of a meta-analysis.

Certainly, no one is suggesting that the same genes underwrite the heritability of these different traits, nor that these wildly different traits are linked in any interesting ways. While the authors found interesting ways to visualize the data they gathered, for what purpose, in the end, the data was gathered and analyzed remains rather opaque.

C. What gets studied

It would be a mistake to suppose, as many readers of the review have, that about half of the variation in all human traits is genetic (a misreading the authors are at least in part responsible for). There are a number of human traits that are not studied in this way, because the results of such a study are a foregone conclusion. Certainly, the language you speak is a “trait,” and certainly it matters to who you are. But your primary language is determined more or less entirely by the household you grow up in. Once we limit our population to those who grew up speaking a particular language within a particular society, there is no doubt room for heritability to emerge in the acquisition of second languages, for example. But no one would be foolish enough to try to find the heritability of speaking, say, Japanese, taking a world-wide population as the relevant one. Similarly, while some food preferences are no doubt heritable, the cuisine you grow up with, again, surely plays a fundamental role in what foods you regard as “normal,” what you eat regularly, etc; and the foods we regard as normal and eat regularly are surely important traits.

The traits chosen for study matter, as do the traits not chosen. All one can say of a meta-analysis of all the studies done, is that if done well, it really does include all the studies done. The studies not done, of course, can’t be part of the analysis. But there are good reasons to think that the results for an arbitrarily chosen “trait” may not much resemble the traits that have historically been selected for study via these methods.

D. The “Missing” Heritability problem

The authors mention the “missing” heritability problem, but do not appear to take it seriously. For many traits studied, especially psychological traits, large-scale searches for the genes associated with the heritable variation have found at best only a very few genes, each with a tiny effect size, that together explain almost none of the variation that is supposed to be associated with genetic differences. This is in contrast to studies on (some) straightforward physical traits, like height, where genes associated with significantly more of the variation that is supposed to be associated with genetic differences have been found (but still, in no case has all, or even more than a bare majority, of the variation supposed to be associated with genetic differences been accounted for).

As these searches for candidate genes (GWAS – Genome Wide Association Studies) become more powerful, the maximum plausible effect size of each of the genes not found goes down; assuming that the genes interact additively – that is, that the effect of each gene on the trait in question is broadly independent of the other genes’ influences – the number of genes necessary to explain the variance goes up. For some heavily studied traits, the minimum number of genes necessary to explain the variance associated with genetic differences, assuming that the heritability estimates are accurate, is now approaching the thousands. It is, to say the least, somewhat implausible that hundreds of human psychological traits are each influenced by thousands of independent genes, each with a tiny additive effect.

Nevertheless, the authors confidently assert that for about 2/3 of traits studied, the pattern of variation is “consistent” with additive genetic variation, and that this “implies that, for the majority of complex traits, causal genetic variants can be detected using a simple additive genetic model.” And yet, causal genetic variants have not been detected using such models; serious searches using such models have failed, time and time again, to find causal genetic variants associated with more than a trivial fraction of the total heritability!

In “The mystery of missing heritability,” a paper that the authors of the meta-analysis in fact cite, the authors (Zuk, Hechter, Sunyaev and Lander) argue that heritability estimates may be inflated by interaction effects, and that the necessary sample sizes to detect these interactions are unreasonably high (so high that even a meta-analysis of the sort under discussion now would not detect the effects that they are suggesting might be responsible for “phantom” heritability). Now, Zuk and his co-authors may well be wrong – their suggestion might prove to be unworkable, or it may simply turn out to be the case that the world really is very strange and most of the missing heritability really is due to thousands of genes with tiny additive effects. But it is at least rather odd to have cited Zuk et al. and to have utterly ignored the most important conclusion that they drew!

II. Heritability: an overview

Heritability is a technical notion, and the implications of the way that it is defined are sometimes counterintuitive. Put simply, heritability is the proportion of the phenotypic variation in a trait of interest, measured in a given studied population and in a given environment, that is statistically co-varying with genetic differences (however measured) among individuals in the same population.

Note first that heritability is not a measure of “how genetic” a trait is. For heritability to make any sense at all as a statistic, the trait in question must vary in the population in question. So for humans, the heritability of “head number” is undefined – there is (almost) no variation in head-number for living humans (there are a vanishingly few conjoined twins that may count as exceptions; note that two-headedness is rather more common, albeit still rare, in for example snakes). Similarly, the ability to speak Jarawan among the Jawaran population also has an undefined heritability, because virtually all Jawarans speak the language. Since heritability is a measure of what is associated with variation in the trait, and not a measure of what causes the trait, the heritability of finger number in humans is essentially zero, and the vast majority of variation in finger number is environmental (traumatic amputations are the primary cause!).

Heritability can be calculated, in principle, by partitioning the variance (the squared deviation from the mean value) of the trait in question in the following way:

VP = VG + VE + VGxE + Ve

where VP is the total variance in the population, VG is the variance associated with genetic differences in the population, VE is the variance associated with shared environmental variation within the population, VGxE is the variance associated with gene by environment interactions, and Ve is everything else, which includes the variance associated with so-called “unique” environments, developmental ‘noise,’ independent epigenetic effects, and errors. VG is generally broken down into two parts: the variance associated with additive genetic variation (VG-additive), and the variance associated with non-additive genetic variation (VG-nonadditive, sometimes written as VGxG to suggest that it is associated with gene-gene interactions).

Given these definitions, Broad-Sense Heritability, H2, is VG / VP (the variance associated with genetic differences divided by the total population variance) and Narrow-Sense Heritability, h2, is VG-additive / VP (the variance associated only with the genetic differences that are statistically related to genes that behave in an additive manner, divided by the total population variance).

[Further, “C” is defined as the proportion of the variance associated with shared environments: VE / VP. “E,” somewhat confusingly, is defined as the proportion of the variance associated with non-shared environments: Ve / VP.]

Narrow-Sense Heritability is a fantastically important measure, if one is interested in breeding for a particular trait, or in the likely response of a trait to selection more generally. But the coarser Broad-Sense heritability is the measure usually associated with studies of heritability in humans.

III. Heritability and twin studies

In order to estimate heritability, one needs to be able to distinguish variation associated with environmental differences from variation associated with genetic differences. For non-human animals raised under controlled experimental conditions, this can be achieved by distributing organisms with known genetic variations into known developmental environments. But for humans, people that share similar genes (e.g., family members) also tend to share similar environments.

The basic trick exploited by the sorts of twin-studies reviewed in Polderman et al., then, is to compare monozygotic (aka “identical”) twins (“MZ twins”) to dizygotic (aka “fraternal”) same-sex twins (“DZ twins”). The assumption is that the family environment experienced by MZ and DZ twins will be relevantly similar, and therefore that the difference between how similar MZ and DZ twins are can be used to estimate how heritable the trait is.

There are a number of quibbles one can raise regarding the reliability of the estimates generated this way. MZ twins usually share a placenta; DZ twins never do. It is not clear that this makes a difference, but it might. MZ twins are sometimes, and perhaps usually, treated differently than DZ twins – dressed more similarly, assumed by both their parents and the other people that they meet to be more similar in a variety of ways, etc., which may also have a measurable effect.

Aside from these issues, there are other concerns regarding the proper interpretation of twin-studies. Should we regard the estimate of heritability (derived from doubling the difference between how similar MZ twins are to DZ twins) as an estimate of the broad-sense heritability, the narrow-sense heritability, something between the two, or something else entirely? Should we expect DZ twins to share none of the gene-gene interactions that influence the trait, or some fraction of those shared by MZ twins? It is sometimes said that DZ twins share half the genes shared by MZ twins; this is true in a sense, but it is of course also possible (in fact, it seems rather likely) that the alternative alleles available from each parent are not representative of the alleles available in the population as a whole. Again, whether these kinds of issues make a difference is fiercely debated.

The upshot of this is that people who think that twin-studies have serious limitations compared to the methods used in research on non-human animals and plants will be unimpressed with a meta-analysis of twin-studies. Several thousand studies, all with the exact same methodological problems and limitations, are no better than one such study.

IV: Heritability as local measure

Heritability is, famously, a local measure; the heritability of a trait is relative to a particular population and a particular range of developmental environments. Recall that heritability is the fraction of the variance in a trait associated with genetic variation. So anything that changes either the total variation associated with a trait, or changes the fraction of the variance associated with genetic variation, will change the heritability.

One way of increasing heritability is to reduce the variation in the environment for that population; at the extreme, a trait that is mildly heritable under ‘ordinary’ conditions can be made entirely heritable by reducing the environmental variation to near zero. If the environment doesn’t vary in the relevant way, all phenotypic variation associated with environmental variation will be eliminated, and the only variation left will be that associated with genetic differences.

Similarly, increasing the range of environments considered will often reduce heritability; if environments are added that are associated with differences in the trait, the heritability will decrease, as the overall variance is increased.

Changes in heritability due to changes in the population considered are also possible; if we reduce the amount of genetic variation in a population, the variation associated with the genetic component will decrease, and more of the variation left will be associated with whatever environmental variation exists. Similarly, if we increase the genetic variation that is associated with differences in the trait, the total variation will increase, and the total fraction associated with genetic variation will also increase.

Polderman et al.’s meta-analysis reports a heritability for “cognition” of around .57; consider the narrower part of this category “performance on standard IQ tests.” Assume, for the sake of argument, that the heritability of performance of standard IQ tests in contemporary U.S. populations is really around .6. (Contemporary estimates range from around .3 to around .8, give or take; there are good reasons to be suspicious of these estimates, some of which are mentioned below. But this is merely meant to be an illustrative example). IQ tests are designed to yield a mean of 100 and a standard deviation of 15. The total population variance is therefore 225, of which 135 (60%) is associated with genetic variation.

But it is well established that IQ test-taking performance increased steadily over time in many countries, the U.S. included. Reasonable estimates put the average IQ in the 1940s, as measured by 1990s tests, at perhaps 70 or 80. If we take 80 as our estimate, the adjusted standard deviation would be around 12 (note that there are complications here – scores did not improve across all tests equally, nor across all segments of the population equally; leave concerns about these issues aside for the moment – again, this is just an example).

Assume, again for the sake of argument, that the heritability of IQ test-taking performance was .6 in 1940 as well (our evidence for this claim is substantially weaker than for the same claim in the 1990s, and even there, as noted below, the number is at best an odd average of many different measured heritabilities in different subpopulations within the US, but again, this is meant to be an illustrative example – please just play along!).

What would happen if we combined an equal number of individuals from our 1940 population with our 1990 population, and thought of them as a single larger population? Using 1990’s tests as our standard, the first thing to note is that we no longer have a normal curve, but a curve with two ‘humps’ (one at 80 and one at 100). The second thing to note is that the standard deviation has increased (to around 16.6). The third thing to note is that is that there is a new ‘shared environment’ factor that is responsible for a significant portion of the variance – whether the person is from the 1940s or the 1990s explains a significant chunk of the variance in our total population!

In fact, the shared environment associated with the year in this example will be associated with about 36% of the total variance (1/2 [(100-90)2 + (80-90)2]) / 16.62). Once the variance associated with year is partitioned out, 60% of the remaining variance will be associated with genetic variation; heritability has therefore been reduced to around .38 [.6 * (16.62 – (1/2 [(100-90)2 + (80-90)2]) / 16.62)].

So what, in the end, is the “actual” heritability of IQ? The question makes no sense; heritability, as a measure, is always and only relative to particular populations at particular times in particular places. This problem is not merely hypothetical. Turkheimer et al (2003) [3] found that in the US populations that were studied, in relatively poor families, most of variation in IQ test-taking ability (about 60%) was associated with the shared familial environment, and almost none with genetic variation (the rest was associated with “unique” environments); in relatively affluent families, the reverse held, with most of the variation in IQ test-taking ability (70%) being associated with genetic variation, and almost none associated with shared familial environment (these findings have been relatively robust in the US context). What, then, is the heritability of IQ in these populations in the U.S.? Should we take the average? Should we adjust for the frequency of the tested SES’s in the population? Does the question even make sense?

Another, slightly more fanciful example, may drive the point further home, adapted very loosely from an example of James Flynn [4].

Imagine a population in which no one plays basketball. It isn’t a sport anyone is familiar with. Imagine that I then test every young adult in this population for basketball playing ability (after explaining the rules, etc.). Much of the variation will likely be broadly genetic (e.g., height will make a huge difference, and within populations today, variation in height is mostly associated with genetic variation; note as well that between populations, however, variations in height can be largely environmental). If I compare MZ to DZ twins, I’ll very likely find that the heritability of basketball playing ability is quite high in this population.

Now imagine that I take, randomly, half the kids from that population, and train them extensively in basketball, whether they like it or not (note that this would be very odd, and also rather mean – forcing people with no particular interest in or talent for basketball to practice for hours a day, to do all sorts of sport-specific strength training, etc.). If I now consider them part of the same population, and measure the heritability of basketball playing ability across the population, heritability will be very low – the vast majority of the difference in ability will come down to whether the people were in the highly trained group or not. (Note that within each sub-group – within the trained and within the untrained – there will very likely be genetic variation associated with differences in abilities, but when we consider the overall population, the differences between the trained and the untrained subgroups will swamp everything else.)

Now, what about a society that cares enough about basketball that to be ‘good’ – good enough to play on a school team, etc. – you have to be really good, because there is a lot of competition, because lots of people try out. Everyone plays a little when they are very young, because it is an important sport that everyone is interested in. And when young, most of the small differences in ability will be down to odd little differences, some of which will likely be heritable within that population – difference in body type, reaction speed, perhaps interest! But then, small differences in abilities and interest will get magnified – those who start out not very good and not very interested are unlikely to pursue it much, very unlikely to get specialized training, etc. Those who start out with some ‘natural’ talent and some interest are likely to be recognized, rewarded, and eventually highly trained.

So, in such a society, any small differences in ability and interest that are related to genetic variation will be greatly magnified. But here is the wrinkle. Any of those differences that are related, however distantly, to those early differences in heritable traits, will show up in an analysis as “genetic.” MZ twins will tend to share the same training regime (or lack thereof) rather more often than DZ twins, because they will tend, more often, to share those small variations that make them either more or less likely than average to pursue basketball. But on one plausible view, what’s doing most of the work in creating differences in abilities is training and practice – not ‘genes’! The trait will be highly heritable, but differences in ability will be mostly down to environments – environments selected (in part self-selected, and in part imposed by others) at least in part because of genetic endowments.

So, in this population, is basketball playing ability mostly genetic, or mostly environmental? The question makes no sense – or rather, depending on how one interprets it, one can defend either answer, or neither, equally well.

It is worth, at this point, stressing another oddity of twin studies. When “shared environment” is spoken of in twin-studies, it means explicitly family environment. Other environments are not considered. So a twin-study would either identify the training regime (or lack of it) noted above as genetic – part of heritability – or if a portion of it was not associated with any genetic differences, it would be chalked up to “unique” environmental effects. But there is nothing “unique,” in the usual sense of the word, about basketball training, either in our hypothetical example or in the real world. Unique simply means, in this context, not part of the familial environment that is shared equally by MZ and DZ twins – nothing more.

V. The Norm of Reaction: another approach to variation

Theodosius Dobzhansky, one of the founders of modern genetics, argued that the correct way of thinking about the relationship between an organism’s genotype and its phenotype was through the lens of the reaction norm. The partial reaction norm for a particular trait and a particular genotype is the way that that trait develops, given that genotype, over a defined range of environments.

The following example is from a paper by Pigliucci and Marlow [5]. The graph shows the reaction norms for bolting time as a function of exposure to increasing lengths of the growing season in 16 populations of Arabidopsis thaliana (a small plant, in the mustard family, that has become one of the model organisms used in biological research).

arabidopsis

Note that some populations have average genotypes that are broadly unresponsive to season length, and in others, season length matters enormously to bolting time. Note that some populations have a shorter bolting time than others given one season length, but a longer time than others given a different season length.

Genotype doesn’t determine the phenotype; rather, we can think of it as determining how the organisms will develop given a particular developmental environment. But of course, while it is possible to compare how one trait develops, given a particular set of genotypes, over a few different environments, it is impossible to determine the complete norm of reaction for even a single genotype (a complete norm of reaction would have an essentially infinite number of dimensions – one for each way in which the environment can vary!).

As Dobzhansky put it:

The norm of reaction of a genotype is at best only incompletely known. Complete knowledge of a norm of reaction would require placing the carriers of a given genotype in all possible environments, and observing the phenotypes that develop. This is a practical impossibility. The existing variety of environments is immense, and new environments are constantly produced. Invention of a new drug, a new diet, a new type of housing, a new educational system, a new political regime introduces new environments. (Evolution, Genetics, and Man 1955 pp. 74-75).

From the perspective of reaction norms, questions about “nature” and “nurture” are ill-formed. Rather, the right questions to ask are more of the form “how does this genotype respond to changes in this environmental variable?” and “how does the response of this gene to changes in the environment depend on or vary with the rest of the organism’s genes?” For some traits, the norm of reaction will be basically flat against most reasonable developmental environments – as noted above, there are essentially no genetic variations that, in any reasonable range of environments, regularly produce a living human with other than a single head. Most human genotypes, in most developmental environments, produce humans with 10 digits on their hands – again, for most developmental environments regularly encountered, a reaction norm plotting the number of fingers against the environmental variation will be flat. For other traits, the trait will develop differently in different environments.

The only way to determine the reaction norm of a trait, given a particular genotype, for a particular range of environments, is to raise genetically identical clones in the variety of different environments in which one is interested. Partial norms of reaction (like the one by Pigliucci and Marlow, above), looking at the response of particular genetic variants averaged against varying genetic backgrounds need not use clones, but still require that organisms be able to be sorted into developmental environments based on their genetic endowments. Needless to say, it is (ethically and practically) impossible to generate reaction norms for human traits!

VI. Some final thoughts

So, is it fair to say, as many commentators on the Polderman et al. study have, that the “nature / nurture” debate is over, and that it is about 50/50? Hardly. In part, this is because heritability studies are simply ill-designed to answer ‘nature/nurture’ questions. Is our ability to read and write mostly “nature” or mostly “nurture”? The question, as stated, is too ambiguous to answer. Humans are the only animals we know of that can read and write, and our ability to perform those tasks clearly has something to do with our nature – with the kinds of creatures that we are. But particular environmental contexts are necessary for those skills to develop – “nurture.”

In the end, heritability doesn’t tell us much that we ought to want to know. It doesn’t tell us whether a trait will be easy or hard to change. It doesn’t tell us what developmental resources are necessary for the trait to develop normally, nor how changes in those resources will change the development of the trait in question. It is simply, at best, a snap-shot of how much of the variation in the trait that there happens to be now is statistically associated with the genetic variation there is in this population, in this range of environments, with this particular distribution of genotypes into those environments. This does not make it meaningless, or useless. But it does put severe limits on what can be deduced from it.

_____

Jonathan M. Kaplan is a philosopher at Oregon State University. His main areas of interest are the philosophy of biology and political philosophy.

Further readings:

Lewontin, Richard C. “Annotation: the analysis of variance and the analysis of causes.” American journal of human genetics 26.3 (1974): 400.

Downes, Stephen M., “Heritability,” The Stanford Encyclopedia of Philosophy (Summer 2015 Edition), Edward N. Zalta (ed.).

[1] Meta-analysis of the heritability of human traits based on fifty years of twin studies, by T.J.C. Polderman et al., Nature Genetics, 18 May 2015.

[2] See: Eric Turkheimer, 2000. “Three Laws of Behavior Genetics and What They Mean.” Current Directions In Psychological Science. 9(5): 160-164.

[3] Socioeconomic status modifies heritability of IQ in young children, by E. Turkheimer et al., Psychological Science, November 2003.

[4] The Flynn Effect: Modernity Made Us Smarter, by S. Mirsky, Scientific American, 20 August 2012.

[5] Differentiation for flowering time and phenotypic integration in Arabidopsis thaliana in response to season length and vernalization,” Oecologia (2001) 127:501–508.

Advertisements


Categories: essay

Tags: , ,

50 replies

  1. A wonderful analysis! Thanks for a reminder of how cautious we have to be with in-house research reports and popular press summaries. Over my head in places but clearly all good sense.

    Like

  2. I’m reminded of Churchill’s view of statistics; “There are three kinds of lies. Lies, damn lies and statistics.”

    This does go to the fact that knowledge is not objective, but is a function of framing/perspective and that combining different views, without some specific focus or goal, creates more noise than signal. This would go against the premise of an objective “God’s eye view” that is not only at the heart of monolithic theologies, but mathematical universe beliefs as well. In Tegmark’s language, the “bird’s eye view.”

    One way to describe the nature/nurture relationship would be with nature as a given spatial structure at a given point in time, while nurture is the evolution of that frame over time, in a feedback relationship with its environment.

    Like

  3. Very informative and much appreciated post.

    Being employed in the field I certainly disagree with the Churchill quote on statistics. Those looking to support a particular thesis will use data to that end. Proper use of statistics which appropriately reveals it’s assumptions and quantifies the associated uncertainty and points to appropriate interpretation (as this post does) can limit the ways in which data can be twisted to support any given ideology.

    Liked by 1 person

  4. This made me think of the recent story on a new self-healing robot technology*. So even our future robot “descendants” will not be determined by their “genotypes”. 🙂

    * http://www.livescience.com/50988-damaged-robot-heals-itself.html

    Like

  5. This essay is lovely example of the great care we must bring to the understanding of complex research.
    Herein lies the problem. Lay people like myself could never bring an informed analysis to claims like that made in the original paper. As it is,science is beginning to sex-up its reporting, see this NYTimes article, Academics Seek a Big Splash (http://nyti.ms/1RFomJg). Many scientists are being encouraged to court media attention. That can be seen as a good thing:

    1) media attention raises the profile and influence of science;
    2) it helps to ensure continued funding;
    3) it informs the public;
    4) it helps continued recruitment of young scientists

    or a bad thing:

    1) it shapes the work of science as it tends then to pursue high profile/high impact science;
    2) popular media have an influence in the way they choose what to publish or not publish;
    3) reporting is oversimplified and important nuance is lost;
    4) it may even encourage bad science.

    The author of this essay was very circumspect in his wording, calling the original paper an odd study and asking what its purpose really was. He pointed out:

    It has already garnered a lot of attention: blog posts and various news sites are trumpeting its conclusions, and putting their own spins on the results. For reasons I’ll briefly discuss below, I found the paper to be very strange

    Seen in the light of the NY Times article(referenced above), perhaps it was not so strange. Perhaps this is an example of the problem highlighted by the NY Times. As this author said “It has already garnered a lot of attention: blog posts and various news sites are trumpeting its conclusions, and putting their own spins on the results.“.

    Lord Acton famously said all power tends to corrupt. A corollary to that might be that all media attention tends to corrupt.

    And yet, we, the lay public, need to be informed and guided about progress in science. How are we to navigate between the Scylla and Charybdis of self-serving publicity and obscure papers? It is in no one’s interests for there to be an uninformed public.

    Liked by 1 person

  6. seth,

    I assume Churchill made that remark tongue in cheek. It goes to the point, which you make as well, that information is very much a function of the context and not all intentions are obvious.

    labnut,

    Looked at in a different light, media is a form of measurement and no measurement is objective. It becomes part of and affects what it measures. Also going back to the point that information is context dependent.

    Some might say math is objective, but math is extremely reductionistic, in order not to be too subjective. Basically a form of generalization.

    The solution is what it has always been; Trial and error. Occasionally it does seem that corruption and confidence schemes prevail over careful consideration, but sometimes that is a sign of larger issues, frames, cultures, beliefs, etc, becoming unstable. The world operates on many levels at once and those who don’t keep their eyes open to more than their immediate area of focus, tend to be unpleasantly surprised on occasion.

    Nature and nurture edit out the unfit and the unlucky.

    Like

  7. It is, to say the least, somewhat implausible that hundreds of human psychological traits are each influenced by thousands of independent genes, each with a tiny additive effect.

    For clarification, is that saying that it is implausible that the genes are independent and their effects additive, or is it saying that it’s implausible that thousands of genes are involved?

    If it’s the former then, agreed, surely all gene effects have to be complex convolutions (rather than linear additions) and surely more or less everything is {G X G X E}? I’m not sure I understand how any gene’s effects on traits could be independent of other genes, and thus be an “additive” effect.

    Similarly, on complex psychological traits (or complex anything traits) wouldn’t one need interactions of large numbers of genes just from an information-content point of view, that anything complex needs a lot of specification?

    Like

  8. Reblogged this on peakmemory and commented:
    Heritability:
    ” It is simply, at best, a snap-shot of how much of the variation in the trait that there happens to be now is statistically associated with the genetic variation there is in this population, in this range of environments, with this particular distribution of genotypes into those environments. This does not make it meaningless, or useless. But it does put severe limits on what can be deduced from it.”

    Like

  9. It has been brought to my attention that Eric Turkheimer, whose work I cite in the entry above, has two recent blog posts that are relevant to this discussion (one quite directly so).

    This is his blog post on the giant meta-analysis of twin studies that this post addresses. Readers will find that some my criticisms are broadly similar to some of his:
    http://ericturkheimer.blogspot.com/2015/05/the-heritability-of-everything.html

    This is a blog post that takes as its start a recent debate in the literature, but it is mostly about how Turkheimer thinks one can (and can’t) use twin-studies, etc., and to what purposes one ought (and ought not) invoke his research:
    http://ericturkheimer.blogspot.com/2015/06/the-criminology-debate.html

    Turkheimer is an “insider” to these debates — as he notes, he does “twin studies for a living.” Obviously, I don’t agree with everything he says (or exactly how he says it!) but I suspect many readers will be interested in the perspective he brings to this kind of work.

    Liked by 2 people

  10. This essay describes research on heritability, along with odd comments, such as complaining about the purpose of the research or how it does not tell us what we ought to want to know.

    So please tell us your purpose in this attack. Is there something wrong with studying heritability? Are you suspicious that it will all be proved wrong someday? Or is this knowledge that people should not have? Is there something dangerous or offensive about getting data on nature-v-nuture questions? Why are you so hostile to what seems like perfectly good research?

    Like

  11. Coel:
    “For clarification, is that saying that it is implausible that the genes are independent and their effects additive, or is it saying that it’s implausible that thousands of genes are involved?”

    The former. That “surely more or less everything is {G X G X E}” is what Polderman et al seem to deny; they do seem to suggest that is all just simple “linear additions.” That’s what I find implausible, given our current evidence, etc. See Turkheimer’s post on Polderman for more on this issue.

    Liked by 1 person

  12. Seth,
    Being employed in the field I certainly disagree with the Churchill quote on statistics

    You are right. That unfortunate quote has been used to bash statisticians for a long time. It is an injustice, given the importance of mathematical statistics. It certainly has a Churchillian ring to it but he was not the originator. That is a matter of some debate but a Google books search shows the phrase was already popular in the 1890s. John Dryden used a very similar phrase in 1682. He said “Now here are three damned lies crowded together into a very little room“, The Vindication of the Duke of Guise from The Works of John Dryden, Vol VII. He was defending himself against a charge of plagiarism.

    I think it is likely Dryden’s phrase was gradually expanded until it reached its present form, with its many variations.

    Like

  13. I suppose I should mention that several of the authors are (former) colleagues and/or collaborators, so I would regard myself as an “insider” too. Curiously, I was thinking yesterday that I might offer Massimo an essay on the classical twin study as an example of a “good enough” method – one with obvious shortcomings but nevertheless useful…

    “[Heritability] doesn’t tell us whether a trait will be easy or hard to change”: Well it actually tells us exactly that about selectability ie the population mean for a highly heritable trait should respond strongly to selection by breeders. Breeders to some extent dislike plasticity, although obviously they would like there to be genotypes that are superior in a range of environments.

    The other reason geneticists are interested in a “heritability number” is that it predicts statistical power to detect genetic effects for a sample taken from that population (experiencing the same mix of environmental influences).

    Re Coel’s question about additivity, this refers to the effects of variation in the gene on a phenotype. The actual structure of the protein (say) that the gene encodes will be complex, but variants of the gene that are relatively frequent in the population will usually have small effects (variants that lead to large phenotypic effects are uncommon, and more often than not lethal or disadvantageous). Nonadditivity can either be within a locus (dominance, diploids can have 0,1,2 copies of a given variant and the relationship with the phenotype can be nonlinear) or between loci, epistasis. The classical twin study has the advantage that large amounts of epistasis will give rise to monozygotic (MZ) twins being far more similar to one another than dizygotic (DZ) twins.

    The “missing heritability” is this: from a twin design we can suggest the heritability of a trait such as height is 80%, but when we measured SNP genotypes and carried out multiple regression predicting an individual’s height, we didn’t do very well, obtaining a realized heritability of only 20-30% using say the strongest 1000-2000 loci. So either there is epistatic interaction between these top loci, or there are lots of tiny effects, that you need large samples of genotyped individuals to detect, or the twin method is too flawed. The seminar I am going to tomorrow is:

    http://www.qbi.uq.edu.au/content/estimating-genetic-variation-human-complex-traits-and-common-diseases-using-whole-genome

    ESTIMATING GENETIC VARIATION FOR HUMAN COMPLEX TRAITS AND COMMON DISEASES USING WHOLE GENOME SEQUENCE DATA

    Abstract: Understanding the extent to which variation in DNA sequence explains variation in phenotype is
    of fundamental importance to genetics, and quantifying the relative contributions of common and
    rare variants to phenotypic variation is of great importance for experimental designs that aim to
    disentangle the genetic architecture of complex traits (e.g. obesity and schizophrenia). We developed a
    new method to estimate heritability for complex traits from unrelated individuals using
    whole-genome sequencing (WGS) data. We demonstrate by simulations based on real WGS data that ~97% and ~72% of variation at common and rare sequence variants, respectively, can be captured by SNP array
    data imputed to a sequenced reference panel. Using our method, we estimate from 44,126 unrelated individuals of European descent that all ~17M imputed variants explain 56% of variance for
    height (a model trait) and 27% for obesity (as measured by body mass index, BMI). We find evidence that
    both height and BMI have been under natural selection. We conclude that narrow-sense heritability is
    likely to be between 60 and 70% for height and between 30 and 40% for BMI. Therefore, missing
    heritability is small for both these traits. For further gene and variant discovery of complex traits,
    a design with SNP arrays followed by sequence imputation is more cost-effective than WGS at current
    prices. Our results have implications for gene discovery for common diseases, including neurological and
    psychiatric disorders.

    If you are interested, there is some didactic material on my website, which should be fairly easy to find.

    Liked by 1 person

  14. schlafly,

    “Is there something wrong with studying heritability?”

    Have you read the article? Because if you did, you would know the answer.

    “Is there something dangerous or offensive about getting data on nature-v-nuture questions?”

    No, unless that data is entirely unreliable and besides the point, in which case it is potentially dangerous.

    “Why are you so hostile to what seems like perfectly good research?”

    Because it ain’t even remotely perfectly good?

    davidlduffy,

    “Well it actually tells us exactly that about selectability”

    As Jonathan says, only narrow sense heritability does that, and that’s not what you get in human studies.

    “The other reason geneticists are interested in a “heritability number” is that it predicts statistical power to detect genetic effects for a sample taken from that population”

    I’m not sure what you mean here. Since heritability is a highly local measure (i.e., dependent not only of the range of genotypes sampled, but also on the environments), what sort of “genetic” effects is it supposed to predict? Or are you talking about gxe effects?

    Liked by 2 people

  15. First, it’s “nice” to note that Schlafly and DavidDuffy have dragged their wrongness about race and racialism over from a couple of previous essays to this one. You’re still wrong.

    I do like aspects of this discussion, especially the look at Genetic-wide complex trait analysis as detailed in this piece:

    http://www.independentsciencenews.org/health/still-chasing-ghosts-a-new-genetic-methodology-will-not-find-the-missing-heritability/

    It’s clear that even Genetic-wide complex trait analysis, at this time, has relatively little to tell us. (I found this when Googling about GCTA.)

    I think it shows the failings to account for prenatal shared issues with monozygotic twins, and how these can differ from three different classes of MZ twins.

    Let’s note that MZ twins can have:
    1. Separate placentas, separate amniotic sacs (earliest separation into two fetuses, not common)
    2. Shared placentas, separate amniotic sacs (most likely)
    3. Shared placentas AND amniotic sacs (latest separation)

    Obviously, these are different amounts of shared environments, especially the third. If a twin study doesn’t have advance knowledge of which pre-natal situation a pair of MZ twins came from, the study’s going to be bollixed up right there.

    Next, how well are twin studies screening for chimerism? Maybe MZ twins actually had a DZ triplet in the womb at one time.

    Beyond that, we know that MZ twins’ brains aren’t genetically identical anyway, perhaps in part due to chimera issues:

    http://socraticgadfly.blogspot.com/2012/02/identical-twins-arent-so-identical-not.html

    So, unless twin studies are allowing for a lot of this, beyond this issue of single-gene vs multi-gene, even to the point of whole-genome genetic traits, they’re not that scientific.

    In other words, though we’ve advanced beyond the outright falsification of studies by the likes of Cyril Burt, we’ve still got a ways to go on methodology, let alone on verifiable findings. Thanks to Jonathan Kaplan for his work on letting us know where the ball is at, and where it needs to be rolling.

    Jonathan, if you could answer the issues above, about whether twin studies screen at all for the different types of MZ twins, or for chimerism, I’d appreciate it!

    Liked by 1 person

  16. The article is pretty much a smack down of the Polderman et al. paper. It is unclear why Polderman et al. bothered. The abstract implies that they were looking for ‘a grand key’ to the gene mapping process (“This study… will guide future gene-mapping efforts”), but it’s hard to see how mere statistical correlation will get us that.

    On the broader issue of the relationship between heritability, genetics, and behavior, I’ve long maintained that any studies on such issues need to be rigorously examined for the socially-biased assumptions we make about behavior. Without such examination and consequent qualifications, such studies will likely prove worthless.

    One of the studies Turkheimer links to, http://onlinelibrary.wiley.com/doi/10.1111/1745-9125.12061/full , is a fairly reasonable attempt to maintain openness on the issue of relating behavioral genetics and social causative studies, while defending twin-studies as a means of bolstering epigenetics studies on environmental effects on gene activation. But towards the end of the study, Moffit and Beckley make an odd remark, that, given its casual appearance, is apparently widely accepted in the field – that cigarette smoking is strongly associated with crime. This stuck in my craw. In 1950, it was estimated that more than half the population of the US smoked cigarettes. * If the Moffit and Beckley remark is taken as a necessary correlate, then should we assume that in 1950 half the population was involved in crime? Obviously the correlate requires considerable historical and social qualification to be in anyway meaningful. Today only some 20% of the population smokes cigarettes; what percentage of these smokers are active criminals? No doubt engaging in any risk-taking behavior will leave one more open to other risk-taking behaviors; yet there’s an important question about whether any such behavior necessarily or only incidentally leads to any other specific behavior. Those more likely to smoke cigarettes are possibly more likely to eat at McDonald’s once a day, and any good nutritionist will tell us that this is risk-taking behavior; but it’s not a crime.

    As I have repeatedly insisted in such discussions – what constitutes “crime”? Crime is socially, culturally, and politically defined – it’s not hanging out there in the environment like mosquitoes, nor is it an inevitable natural behavior of the body like defecation. Until well into the 1980s, homosexuality was quite literally criminal behavior; now gay marriage is becoming widely legal behavior. In 1980, homosexuality was considered ‘risk-taking’ behavior, and psychiatrically diagnosable. Now that diagnosis has been considerably redefined, and will probably be further redefined in the future. Yet in 1980, behavioral-genetic criminology research would have needed to account for it (in the same era that behavioral psychologists were supposedly developing ‘therapy’ to ‘modify’ homosexuals into heterosexuals).

    Frankly, any genetics research on human behavior should be approached cautiously and skeptically. Tendencies do not make certainties; and if they don’t then any such research is going to have considerably limited application.

    —–
    * http://www.smokingstatistics.org/Smoking_Statistics_Since_1950.asp

    Like

  17. Heritability informs us about what is means to be humans. People have wondered about nature/nurture interactions for millennia, and now we have some hard data to prove or disprove what others have suspected.

    If you want tall kids, do you find a tall spouse, feed your kids vitamin supplements, pray to the tallness gods, or have the schools teach books about tall people? Science can answer this question, as well as similar questions about a lot of other human traits.

    Suppose immigrants are moving into your neighborhood and you are wondering whether they will assimilate? Chances are that the kids will have most of the heritable traits of the parents. Public school brainwashing can only do so much.

    When I am an old fart telling kids to get off my lawn, I am going to want to know whether their bad behavior is caused by poor breeding or poor training.

    Of course additivity is just a first-order approximation. Did anyone ever say anything else? We are just starting to understand the genes. Heritability studies have upset conventional wisdom as much as Wegener on continental drift, and you guys should be celebrating scientific advances, not denouncing them.

    Like

  18. Schafly,
    you guys
    Who’s ‘you guys’? Name the names.

    should be celebrating scientific advances, not denouncing them

    We celebrate good science and denounce bad science. It is called critical thinking and all skeptics should practice it. When someone demands I accept something without critical examination I know I am dealing with an ideologue. The name of the game is called question, take apart, look for flaws in data or reasoning. Only once it has passed the most rigorous critical examination does it deserve to be called good science. Do you suggest we should abandon the good and wholesome practice of critical examination?

    I am going to want to know whether their bad behavior is caused by poor breeding or poor training

    Then you should take up horse or dog breeding.

    Chances are that the kids will have most of the heritable traits of the parents

    That sound like an oxymoron. It is true by definition. Heritable traits are inherited, to varying degrees. The ‘most’ is the debatable part where science is trying to refine our knowledge.

    Heritability informs us about what is means to be humans

    Does it? When I want to know what it means to be human I read our history, I study our societal interactions, I study our culture and I examine our moral systems. That informs me what it means to be human.

    Liked by 2 people

  19. It’s really a little hard to take these sets of criticisms seriously. You really do seem to be cherry picking every possible angle you can in order to discredit the findings of this massive meta-analysis.

    Look, even Turkheimer — as famous a proponent of the importance of the environment on social traits as there is, and one whom you quote approvingly when it favors your thesis — pretty much has thrown in the towel on the heritability of IQ and g, based on a well known GCTA study (though he does believe those studies gain us no real insight into the exact causal mechanisms in which the “missing heritability” consists):

    “Thanks to the Visscher program of research, it should now be impossible to argue that the whole body of quantitative genetic research showing the universal importance of genes for human development was somehow based on a sanguine view of the equal environment assumption in twin studies, putting an end to an entire misguided school of thought among traditional opponents of classical quantitative (and by association behavioral) genetics.”

    http://www.docstoc.com/docs/155493190/_2011_-Still-missing-Research-in-Human—University-of-Virginia

    Likewise, since the Lander et. al. paper from 2012 which you cite, it would seem Eric Lander has thrown in the towel regarding whether common variants can explain missing heritability. See the following link for some details and further links:

    http://infoproc.blogspot.com/2014/12/measuring-missing-heritability.html

    You really should try to keep up with the science rather than trotting out the same, now rather quaint, objections.

    Like

  20. Schlafly:

    “So please tell us your purpose in this attack.”

    I did not write this as an attack on twin-studies per se, but rather as a caution about the limits of heritability as a concept. I did (briefly) criticize the Polderman et al article, on the grounds that a meta-analysis of every twin-study ever, while a monumental undertaking, was an at best odd undertaking, given that a) that pretty much all twin-studies yield non-zero heritabilities is well-understood and accepted even by the harshest critics of twin-studies, and b) that averaging the heritabilities of wildly disparate traits, measured in different populations at different times, yielded a number that seemed broadly meaningless.

    “Is there something wrong with studying heritability?”

    Of course not. Studies designed to estimate heritability in non-human organisms remain important for a variety of different reasons. I have some doubts about how many of those reasons still hold when applied to estimating heritability in humans, but that is a different issue.

    “Are you suspicious that it will all be proved wrong someday?”

    Given the locality of heritability estimates, I don’t know what you are asking. In many cases, even if we were able to measure heritability with perfect accuracy, changing the developmental environment, or changing the population considered, would change the heritability. In other cases, it wouldn’t. We have no way of knowing what kind of case we are looking at w/o knowing (roughly) the norm of reaction for those types of genotypes against the types of environments we are interested in, or at least no way of knowing unless we actually change the environment and retest. So, leaving aside legitimate concerns about the difficulty of accurately estimating heritability in humans via twin studies, the question remains: what work does the estimate do? That’s different than worrying that all the estimates are “wrong,” whatever that would mean for an estimate that is a snapshot of a particular population at a particular time, etc. (I do suspect that many of the estimates we currently have are in some way problematic – not wrong, per se, though they might be that, too, but not measuring what we think is being measured, even within the narrow confines of measuring heritability. But that’s neither here nor there – how accurate the estimate is doesn’t matter to the concerns I’ve raised. That is part of the point.)

    “Or is this knowledge that people should not have?”

    If you had a magic device that, given a particular population, distributed in a particular way in a particular environment, informed you of the precise heritability of a trait, given a particular operationalized measure for that trait, what would you gain when you applied this to human traits? What use would you make of this information? It isn’t “knowledge that people should not have” — it is knowledge that, for the reasons I sketch above, I think is less relevant to most of the questions we want answered than some people think. (There are some places where it might be very useful, but not in many of the sorts of cases that some people think it would be.)

    “Is there something dangerous or offensive about getting data on nature-v-nuture questions?”

    For the reasons sketched above, I argue that heritability estimates do not provide data on “nature-v-nurture” questions, and that it is a mistake to interpret them as providing that kind of information. I did suggest that other research programs pointed towards projects that looked more like they answered what we usually think of as “nature-v-nurture’ questions. That would have been an odd thing to do if I thought such questions were dangerous or offensive per se.

    “ Why are you so hostile to what seems like perfectly good research?” If you think that outlining the limitations of the heritability concept is being “hostile” to research, may I suggest that you rethink your position? The limitations, outlined above, of heritability as a concept are well-understood and acknowledged by both those people who engage in behavior genetics research via twin-studies, and those who criticize that research. None of this is controversial. It is, however, not as well appreciated as it might be. On the other hand, if you object to my criticisms of combining the results of lots of different twin-studies together and estimating the average heritability of the wildly different traits that were measured, I ask you to reconsider my analogy regarding e.g. an estimate of ‘average’ treatment effectiveness from every randomized clinical trial. Why is this case different in kind?

    “Heritability informs us about what is means to be humans.”

    Wow. No, it doesn’t. Not at all. The ability to learn a complex human language is surely part of what makes us human. To a first approximation, all humans are sophisticated language users – by far the most sophisticated language users we know! Heritability is just about undefined — there are essentially no even remotely common genetic variants associated with the ability to learn language. Nearly everyone can do, and does it. Fortunately, there are also no even remotely common environments in which this doesn’t take place. Looking for the heritability for the traits that “make us human” would yield nonsense answers. No one does that, because it is stupid.

    “People have wondered about nature/nurture interactions for millennia, and now we have some hard data to prove or disprove what others have suspected.”

    Again, for the reasons sketched above, heritability does not answer questions about nature/nurture interactions, as they are usually understood. Misunderstanding the results of twin studies that attempt to estimate heritability as being informative about nature/nurture is the problem the above addresses. And again, the conceptual limitations I point towards are not controversial, nor mysterious.

    “If you want tall kids, do you find a tall spouse, feed your kids vitamin supplements, pray to the tallness gods, or have the schools teach books about tall people? Science can answer this question, as well as similar questions about a lot of other human traits.”

    Height is highly heritable within populations, but between populations, height differences can be mostly environmental in origin. If you want tall kids, would you be better advised to pick a tall spouse in a horrible environment in which the child would likely develop w/o adequate nutrition, a high parasite load, etc., or would you be better advised to pick an environment with excellent nutrition and other health-related features, and an average spouse? Answer: It depends! (Well, obviously, if you can have *both* then go for both! But if it is a forced choice…) And the answer is given by looking at the way that different genotypes respond to different environments, not by the measure of heritability within any given environment.

    “Suppose immigrants are moving into your neighborhood and you are wondering whether they will assimilate? Chances are that the kids will have most of the heritable traits of the parents. Public school brainwashing can only do so much.”

    OK, I am at a bit of a loss here, because the obvious interpretation of your claim seems so at odds with how the world works, that I worry that I must be misunderstanding you. You do realize that when children from one country are adopted into families in another country, they pretty adopt all the cultural traits of their adopted country (“assimilation”)? They grow up speaking the same language, following the same norms of politeness etc, eating the same foods, etc. There are no populations of people for which that is not true. So “assimilation” into a culture is entirely dependent on the culture in which one grows up — the shared familial and ‘unique’ environments explain essentially all of the differences in what culture we assimilate to. To deny that would be to make an obviously absurd claim. So the variation in culture assimilated to is entirely explained by the environment — heritability is zero.

    Now, do I suspect that if we did an adoption study were one MZ twin was adopted into country A, and the other into B, and did the same with DZ twins, that the MZ twins would be more similar to each other than the DZ twins on many interesting measures? Sure. But with respect to language spoken, standard cultural practices, food preferences, etc etc etc. – all the things we mean by “assimilation!” – what matters is what culture you develop in. That swamps everything else. So heritability is total red-herring when we think about whether it will be possible for someone to “assimilate” into a new culture.

    (Similarly, whether a family can assimilate – as a family! – also quite obviously has very nearly nothing to do with whether the traits in question are ‘heritable.’)

    “When I am an old fart telling kids to get off my lawn, I am going to want to know whether their bad behavior is caused by poor breeding or poor training.”

    Heritability doesn’t tell you whether something is the result of “bad breeding” or “poor training.” I politely remind you that the rates of violent crimes between populations, or within populations over time, vary by more than an order of magnitude. Even if criminality (whatever that means) is highly heritable within a population, this says nothing about whether it is “caused by” bad breeding or poor training, and still less about what our response should be if we want to reduce the prevalence of, and the harms associated with, crime and violence.

    Liked by 6 people

  21. Dear Massimo – “narrow heritability…not what you get in human studies”. This is in fact exactly what you get, and underlies the second of the aims of the Polderman paper. Narrow = additive, broad=additive+nonadditive.

    Dear Socratic – the papers that Charnley quotes find that the GCTA approach tends to underestimate heritability, when compared to the related models they replace it with. We routinely include the set of principal components estimated from the marker data as covariates – these capture the genetic differences between subpopulations.

    With regards to twin chorionicity, I have tested these effects for a variety of traits (Duffy DL (1993). Twin studies in medical research. Lancet 341:1418-9), and do not find any large confounding effects. Effects of fetal environment paradoxically can decrease MZ twin resemblance, eg discordance in birth weight lead to lifelong differences in height and differing risk of atherosclerotic disease. Obviously this will cause your heritability estimate to fall.

    Weaknesses in these types of analysis lie elsewhere, as intimated by the Turkheimer essays. One of the largest genetic risk factors for lung cancer found by GWAS are variants in the CHRNA3 locus. This encodes a nicotinic cholinergic receptor and “polymorphisms in this gene have been associated with an increased risk of smoking initiation” ie the genetic determination of risk is via a behaviour. Obviously this is not a problem for behaviour geneticists, but it is annoying if you are looking for new insights into carcinogenesis. Generally, the effects of the proximate cause (smoking in this example) have to be very large, because the gene variant’s effect on the intermediate variable is small. Coming back to genes for IQ – some of those identified to date are known to influence both brain volume and height, so it is easy to speculate that effects might be mediated by social factors.

    The latter is why epidemiologists are interested in the study design called Mendelian Randomization. If gene variation affects one trait that is a risk factor for a second trait (ie correlated), then you can test causation in an non-experimental way (genotype is assigned randomly, pace ethnic stratification). Genetic linkage analysis allows even stronger conclusions to be drawn.

    Circling back to “effective heritability” – this is the predictive accuracy of a set of SNPs for a phenotype. It is obvious to me that this is a worthwhile thing to measure. We can and do add in measured environmental predictors and the GxE interaction terms, and see how much improvement in accuracy they bring. And it is nice to have an alternative estimate from a family-based design (eg twin study) to check against.

    Like

  22. David —

    “‘[Heritability] doesn’t tell us whether a trait will be easy or hard to change’: Well it actually tells us exactly that about selectability ie the population mean for a highly heritable trait should respond strongly to selection by breeders.”

    Massimo’s comment re: this being true of narrow rather than broad heritability is right, as far as it goes, but I want to stop you a bit earlier. No sane person gives a damn about how strongly human traits will respond to selection by breeders. That is not a question any reasonable person wants answered, and it is simply weird to pretend that it might be. (The question of how strongly a trait in humans might respond to *natural* selection is less crazy, but even narrow-sense heritability provides rather less information about that than one might wish, over time scales that one might care about when thinking about e.g. human evolution.)

    Now, we could quibble about whether Polderman et al really provide strong evidence that most of the heritability found in twin studies in most of traits studied is additive (the view you endorse). And one might suggest that if every trait shows high heritability, and that if that heritability is down to many many genes with small additive effects in most cases, then (weirdly) every gene must be involved (additively!) in some huge number of (broadly uncorrelated) traits (true universal pleiotropy), in which case, one might think, selection might in fact be more highly constrained than heritability estimates would imply… But the arguments I suggested above don’t rely on any that, and it isn’t an argument I care much about in this context. (Nor, as my original post made clear, do I really care to argue about the accuracy of particular estimates of heritability — that just doesn’t matter to the point I’m making.)

    Rather, when I stated that heritability doesn’t tell us whether a trait will be easy or hard to change, I meant just that. If I’m interested in changing the distribution of some trait in a population, the heritability of that trait alone provides me with precisely no information about how easy or hard that will be to do. Period. The usefulness of narrow-sense heritability for predicting the response to selection within a given environment notwithstanding, heritability tells us nothing about how easy or hard it will be to change the trait more generally.

    Yet another quick example: The heritability of “anti-social” and “criminal” behaviors in the U.S., as measured by twin studies, was pretty much the same in 1980 as it was in 2010. But the homicide rate in the U.S. in 1980 was more than double what it was in 2010, the violent crime rate almost double, and the rate of property crimes about double. And if you think that difference was because there was a shift in the distribution of crime alleles in the U.S. population between 1980 and 2010, I’ve got a bridge you might be interested in buying… (Do I buy the leaded gas theory? For the purposes of this, I really don’t care what caused the change — the fact is, the frequency of the behaviors in question changed radically, while the heritability of the behaviors in question remained just about the same, and no change in the distribution of genes is gonna help explain that.)

    Now, we can play this game for any number of traits, but not all traits. Just because a trait varies, doesn’t mean we can do much about it — even (gasp!) if the heritability turns out to be relatively low. Conversely, just because the variation is highly heritable, doesn’t mean we can’t change it quite radically, and perhaps quite easily.

    Again, none of this is to say that estimating the heritability of human traits is necessarily a waste of time, or useless. But don’t mistake a heritability estimate for a measure of how easy or hard it would be to change the trait in question. That isn’t what heritability — narrow or broad — measures.

    To come back with “but narrow-sense heritability does predict the response to selection in a constant environment, at that particular time!” is to completely miss the point.

    There is simply no such thing as “the predictive accuracy of a set of SNPs for a phenotype” unless you specify at least the (range of) developmental environment(s). Now, if your interest is not “human nature” but “what alleles, in the range of developmental environments usually encountered in middle-class suburban families in this part of the world in the 2010s, are associated with childhood asthma,” that fact may not matter much. But just because your particular interest specifies the developmental environments that matter to you doesn’t mean that those developmental environments are the only ones.

    (With respect to the brief note re: fetal environments perhaps resulting in underestimates of heritability under some conditions, one thing that has gone weirdly unnoticed among those who interpreted my original post — which, again, is mostly about the meaning of ‘heritability’ and the limitations that follow from that — as wholesale attack on twin-studies and behavior genetics is that the list of concerns I noted about the accuracy of twin-studies included both elements that might tend to result in over-estimates, and elements that might tend to result in under-estimates. I was merely listing some of the concerns people have raised about the limitations of contemporary ways of estimating heritability in humans, and noting that if these are problems, then the problems don’t go away when one combines thousands of studies with the same problems.)

    Liked by 3 people

  23. Jonathan, you keep trying to pick away at these nature/nurture distinctions, but then you acknowledge them when you say “Well, obviously, if you can have *both* then go for both!” Okay, but first I have to know what is nature and what is nurture so I can go for both.

    You also acknowledge them when you say kids “adopt all the cultural traits”. Here, I assume that cultural means non-heritable. So they bring the heritable traits from the old country, and acquire the non-heritable ones during development.

    You say that heritability is zero for “following the same norms of politeness” and “food preferences”. You may be right, but this isn’t obvious to me. How would you know, except to look at some study like the ones you are attacking? Is there data to back that up, or are you just guessing? And if this is so important, shouldn’t you be in favor of heritability studies so that you can draw conclusions like this?

    Yes, you point to some obscure technical issues, such as whether the MZ twins share a placenta. These are of some interest to researchers in the field, but not likely to change the overall conclusions much.

    You say that learning language is a universal human trait. Yes, but some learn it much better than others, and talent in this area is heritable.

    “Studies designed to estimate heritability in non-human organisms remain important”

    This sounds a little like creationists who argue that micro-evolution applies to dogs and finches, but evolution does not apply to humans. How does that work? Heritability is essential for understanding the evolution of all life on Earth — except for humans?

    The creationists say that humans are different because we have eternal souls, took bad serpent advice in Eden, got almost wiped out by Noah’s Flood, and were saved by Jesus Christ. But that does not appear to be your argument here. Instead you seem driven by some leftist-Marxist-egalitarian-materialist political view. Recently on this site you have been quibbling about human skull measurements and denying human races (“our folk conception of race … does not pick out populations of biological interest”). You appear to be driven by some non-scientific ideological beliefs that are contradicted by heritability. Do you want to spell those out for us?

    “No sane person gives a damn about how strongly human traits will respond to selection by breeders. That is not a question any reasonable person wants answered, and it is simply weird to pretend that it might be.”

    Darwin openly speculated about such questions, and so have a lot of other sane people. Scientists are nearly always interested in finding ways to get answers to questions. There are 1000s of scientists interested in such questions for fruit flies. Why not humans?

    Again, what exactly is the ideology that prefers ignorance about human nature?

    “If I’m interested in changing the distribution of some trait in a population, the heritability of that trait alone provides me with precisely no information about how easy or hard that will be to do. Period.”

    If the trait is highly heritable, like height in the USA, then that fact tells us that change will be very difficult unless you figure out some easy way to get tall people to breed more, and short people to breed less. That seems like information to me.

    Again, you don’t explain why you are so eager to deny obvious consequences of heritability.

    Like

  24. Dear Jonathon – “There is simply no such thing as ‘the predictive accuracy of a set of SNPs for a phenotype’ unless you specify at least the (range of) developmental environment(s)”. A little thought would make you realize this broad statement is equally true for any trait and any putative cause you would care to study, since we can never measure and control all the possible confounders and interacting variables. At a trivial level “There is simply no such thing as ‘the predictive accuracy of a social sciences type predictor’ unless you specify at least the (range of) individual genotypes on which it will be acting” is just as bad.

    Obviously, we use a large enough sample and report average effects of an cultural intervention or a locus, whether it is a method of teaching reading or KIAA0319 genotype. It is nice to refine our understanding, and says things like “teaching method X is less effective in individuals carrying genotype Y”, but it doesn’t make the averaged effect less real and is what random effects models are all about.

    WRT constancy of heritability of criminality in time – you admirably summarized the peculiarities of interpreting ratios such as this. Although the Polderman et al meta-analysis concentrates on h2, all of the studies cited can as easily provide estimates of the genetic and environmental variances.

    If heritability remains unchanged despite changes in population trait means, the simplest interpretation is that the environmental effects underlying the change have affected all genotypes equally (“all boats”), so that the relative genetic differences between individuals are constant. This comes back to what Mayr called “population based thinking” – variance components analysis of twin correlations and covariances is driven by individual differences, not the actual values (rather “structuralist” in one sense).

    Similarly, if you remove environmental inequalities in a population, say by introducing good quality universal higher education, then you reduce environmental variance, and so in the absence of GxE, as you suggested earlier the heritability of educational achievement will rise, as remaining differences between individuals are more likely to be genetic in nature.

    Regarding gene-environment interaction, the twin design can detect this as a difference in the total
    trait variance in MZ versus DZ twins – though you need large sample sizes and a nicely distributed trait.

    Liked by 1 person

  25. Schlafly – “Again, you don’t explain why you are so eager to deny obvious consequences of heritability”

    What I see here is not a denial of anything, just a dislike of poorly designed research and poorly reported findings.

    Liked by 1 person

  26. Hi Schlafly – “If the trait is highly heritable, like height in the USA, then that fact tells us that change will be very difficult”. This isn’t quite right, because as I explained earlier, a high heritability merely means that differences between individuals in the population currently are largely due to genetic differences. The differences in availability of adequate nutrition that we saw historically due to social class – leading to the upper classes being taller – have been minimized to some extent. We instead see the reverse problems with diabetogenic diets and excess calories in lower SES individuals. But if you look across Europe, there are big rapid recent changes in average height – one Portugese study found a 6 cm increase in eight years – reflecting the economic upturn. The heritability may or may not have changed proportionately. In Ireland there has been a 23 cm increase over 50 years in 14 y.o.’s.

    One of our interesting examples is the heritability of tonsillectomy. This used to be low in the ’50s and ’60s, when everyone had their tonsils out, but has risen now that only severe cases are operated on.

    The other point is the actual size of the genetic variance. One can have a high heritability, but the variability of the trait can be low. This is also seen with measures like the genetic correlation between variables. The total (phenotypic) correlation between X and Y might be very low, but entirely genetic. In this case, the usual definition of a genetic correlation would say that the genetic correlation is 100%. Again, looking at the actual variances and covariances is the way to go. I’ll stop lecturing now 😉

    Like

  27. Jonathan Kaplan has reminded me of other things related to heredity study, per comments here and elsewhere, going beyond issues of methodology reliability and to the main point of his second comment.

    Just what is heritable, how much is it, and how can we tell, vs. a constellation of environmental effects?

    Even something “purely physiological,” and seemingly “simple,” like height, we may not be able to perfectly answer. And that’s not what most laypeople really want to know, anyway.

    No, with a, not a hat tip but something else to a couple of commenters, it’s psychological traits and tendencies, and their heritability, that drive most interest.

    Let’s just seize the horns, per a couple of previous essays and those couple of commenters, on the biggie: Intelligence.

    Since there is no g (down, to a couple), just what is heritable, and how can we test for that?

    That shows that this issue starts with definitions. One can’t ask these questions of heritability without properly defining what’s supposed to be heritable.

    The move to establish multiple intelligences, while good psychologically and sociologically in moving beyond a fixation with g, and with IQ tests, may have made things worse.

    What’s supposed to be heritable in kinesthetic intelligence? In emotional intelligence?

    Take the former, first. Is muscle firing speed part of that? Is proprioception?

    Let’s say yes, for the sake of argument.

    How do you show how to exclude the ability to train oneself in improvement in the former through work at a gym? How do you investigate the possibility that the latter is learnable?

    Even more with emotional intelligence. Let’s say that we accept that facial reading skills are accepted as part of this. How do we test for how much of that is heritable versus socially learned? After all, by the time a child is old enough to reliably report “Face X looks like disgust” or whatever, he or she is old enough to be doing social learning on this.

    Otherwise, Jonathan appears to have already answered some objections by DavidD and all by Schlafly, who may have brought objections with him from “the old country.”

    What Jonathan hasn’t directly said is that some of the environmental controls David seemingly is talking about don’t seem to be issues that can be semi-controlled for, post hoc, by whom is selected for a study, but rather would have to be consciously controlled in advance. And that, of course, leads to ethical issues.

    David even admits this:

    we can never measure and control all the possible confounders and interacting variables.

    Indeed, that then leads to presumptions which may or may not be warranted as to which confounders and interacting variables are more important, whether or not we can work around them through study size, metadata and other things or not, and more.

    Liked by 1 person

  28. candid,

    “It’s really a little hard to take these sets of criticisms seriously”

    It’s really hard to take a comment that begins like that seriously, but I’ll try.

    “Thanks to the Visscher program of research, it should now be impossible to argue that the whole body of quantitative genetic research showing the universal importance of genes for human development was somehow based on a sanguine view of the equal environment assumption in twin studies”

    But, of course, Jonathan has not been claiming that at all. His claim is that heritability is a local measure, and that it tells you nothing about human reaction norms. He further claims that without characterizing such reaction norms we cannot move beyond the generic assertion — which is certainly the case — that many human traits are influenced by both genes and environments. In other words, it’s not that heritability measures are wrong, so much that they are largely uninformative.

    “You really should try to keep up with the science rather than trotting out the same, now rather quaint, objections”

    And perhaps you really should try to understand what such objections are before replying to points that the author did not raise.

    david,

    “”narrow heritability…not what you get in human studies”. This is in fact exactly what you get”

    Uhm, no. In order to get narrow-sense heritability one has to conduct the kind of genetically and environmentally controlled experiments that are, in fact, impossible in humans. While twin studies are certainly interesting, they still don’t get at the sort of partitioning of phenotypic variance that is possible in experimental organisms.

    Concerning SNPs, of course it is useful to find genetic variants statistically associated with certain phenotypes. And as you say, sometimes one can begin with that work and proceed to elucidate the molecular mechanisms. But, again, this is somewhat beside the point, since nobody here is denying that phenotypes are influenced by genes. The crucial question is what heritability tells us (next to nothing), and the degree of gene-environment interaction (i.e., the population’s variability for the shape of the reaction norms). The latter can’t be assessed by correlational studies, no matter how sophisticated the molecular marking.

    “Regarding gene-environment interaction, the twin design can detect this as a difference in the total trait variance in MZ versus DZ twins”

    Again, not exactly. They can at best detect part of the gxe interaction, they don’t give us any magical access to the shape of the underlying reaction norm.

    schlafly,

    “first I have to know what is nature and what is nurture so I can go for both”

    Even putting the question that way is a misunderstanding of the issue. Genetic effects have to take place within the context of an environment, so to attempt to neatly separate the two is senseless. Perhaps an analogy by Lewontin may help: imagine you build a house using bricks and lime. Someone then wants to understand how you built the house. To achieve that he asks you the exact partition of the total weight of the house: how many pounds of bricks, and how many pounds of lime? I hope you realize that that question does have an answer, but it is an answer that doesn’t even begin to get you an understanding of how to build a house, because the latter is the result of a very specific, interweaving pattern of bricks and lime.

    “you point to some obscure technical issues, such as whether the MZ twins share a placenta. These are of some interest to researchers in the field, but not likely to change the overall conclusions much”

    Those aren’t obscure issues, they are crucial. And they ought to change the overall conclusion.

    “This sounds a little like creationists who argue that micro-evolution applies to dogs and finches, but evolution does not apply to humans”

    This sounds to me like a horribly uncharitable reading of what Jonathan has actually written.

    “You appear to be driven by some non-scientific ideological beliefs that are contradicted by heritability. Do you want to spell those out for us?”

    You mean like the belief that one should accept only methodologically sound scientific results, especially when it comes to issues of social import? Yeah, sign me on to that sort of non-scientific (it’s called “ethical”) ideological belief.

    “There are 1000s of scientists interested in such questions for fruit flies. Why not humans?”

    You are equivocating here, most likely willfully: Jonathan isn’t saying that it is insane to wonder about the genetic bases of human traits; he is saying that it is insane to think of humans in terms of selective breeding.

    “If the trait is highly heritable, like height in the USA, then that fact tells us that change will be very difficult unless you figure out some easy way to get tall people to breed more, and short people to breed less. That seems like information to me.”

    Because you don’t understand the meaning of heritability. Even a highly heritable trait may be highly plastic in response to environmental conditions. For instance, phenylketonuria: it is caused by a single genetic mutation, and in “standard” environments it causes horrible developmental abnormalities. But change the environment to one that does not contain phenylalanine and much of the phenotype reverts to normal.

    Liked by 4 people

  29. I fully support the author of this paper, and though he must maintain some diplomacy, I do not. Thus I can state that in relation to the complexities of reality itself, the human is surely just an idiot . But to the extent that the human conveniently ignores its limitations, as he has noted, its idiocy becomes magnified. I’ve only just recently begun to think of philosophy as a means through which to assess our conceptions of reality (in which case philosophy does concern more than various still failed topics or reality), so this article does come at an opportune time for me.

    Liked by 1 person

  30. SocraticGadfly —

    Sorry not to have responded sooner. The easy answer is that the vast majority of twin studies have not controlled for any of the factors you note (shared v unshared placenta, chimerism, etc). Indeed, until relatively recently no twin studies (by necessity) genotyped to ensure that the MZ/DZ status was reported correctly.

    All that said, I am somewhat skeptical that violations of the “equal environment assumption,” narrowly understood, are going to explain much of the variance that twin-studies assign to heritability. These concerns do mean we shouldn’t take any particular estimate particularly seriously, but again, for the reasons I suggest in the post, for most things it just doesn’t matter if the ‘true’ broad-sense heritability of some trait, in this particular population at this particular time, is .3 or .7 — there may be exceptions, and if someone finds a case where it does matter (to something other than raw curiosity) then arguments about how accurate a particular estimate is will of course be important. But most of the back-and-forth about how accurate heritability estimates are seem to me to miss the point. (The exchange in Criminology that Turkheimer riffs off of in his post strikes me as very odd in this way — a fairly vicious back and forth about whether the heritability estimates from twin-studies are sufficiently accurate, but with no clear sense of what they could be sufficiently accurate for! That is — no clear sense of what use would get made of the numbers, such that the accuracy of those numbers was the primary thing at issue. Taking place in the journal Criminology, one would expect someone to say “and this is why knowing this will help us reduce crime.” But that would, for the reasons I suggest in the post and reiterate in previous comments, be incorrect, no matter what the ‘right’ number was!)

    I do suspect that the failure to find genes that can account for more than a tiny fraction of variance associated with heritability in most traits studied is telling us something — but what it is telling us remains unclear. I am, for the reasons I suggested above, skeptical of the claim that it is telling us that every one of the thousands of traits studied is influenced, in a nice additive way, by some significant fraction of our entire genome. But that isn’t actually impossible — stranger things have turned out to be true! If that’s the case, then I do think it will require a rethinking of some of the ways we think about e.g. quantitative genetics. None of the other explanations (weird interaction effects, genes exerting their influence through the developmental environments encountered, etc.) strike me as obviously best, nor indeed are they mutually exclusive. What we need to figure out, as Massimo suggests in his most recent comment, are the developmental pathways that actually produce the variations in the particular trait we are interested in, if the variation in the trait is really our interest, or more generally, produce the traits in question. When we figure that out, we’ll know (or be able to figure out) what’s going on with heritability. But figuring out the heritability simply isn’t informative about the developmental pathways, and it is the latter that we really want to know about.

    Another point is perhaps worth raising. My post lays out the limitations of the heritability concept, and explains why twin-studies that estimate heritability don’t provide the kind of information that they are sometimes thought to, and don’t permit the kinds of predictions or explanations that people sometimes attribute to them. One might interpret this as implying that I think all twin-studies are crap. But please note well that this is not the position I take in this blog post, and that interpretation is not supported by the claims I in fact make. One might go further and interpret my comments are implying that I think behavior genetics, as a field, is all total crap — those familiar with just the title of my first book, or who failed to read it at all carefully, might feel justified in reading this post through that lens. But again, this is not the position I take in this blog post, nor is it the position I have actually defended in my work (somewhat inflammatory titles notwithstanding).

    To put a somewhat more forceful spin on a comment of Turkheimer’s on his blog post on the Criminology debate (his point 4, for those keeping track), we should recognize that twin studies designed to estimate heritability, and nothing more, are now a useless exercise. The result is a foregone conclusion (h^2 will come out between .25 and .75, the vast majority of the time), and there is nothing that can be done with the result (no policy implications follow from it, no suggestions about treatment for diseases, no hints about how to find the genes that supposedly underwrite the heritability, etc etc.). If you are thinking of doing a twin study to estimate the heritability of a non-silly trait, and nothing more, you are wasting your time. (I am not calling for ‘censorship’ here — your time is yours to waste. But I will strenuously object if you misrepresent to funding agencies the implications of your research — e.g., claiming it actually has any — and I will strenuously object if you misrepresent the meaning of your work in the popular press…)

    But there are other things that can be done with twin-studies, aside from just estimating heritability. Whether these other projects produce results that are accurate enough to guide further research, or produce results that can be applied to interesting explanatory and predictive hypotheses, etc., depends critically on the particular project — nothing in general can be said either for or against them. Except perhaps to remind ourselves that, as usual, any work on humans is really hard to do well, and we need to be vigilant about moving from “x, which is how we determine y in non-human organisms, is impossible to do with humans,” to “therefore we can definitely learn about y in some other way.” Maybe we can, and maybe we can’t. And if we can’t, we shouldn’t pretend that we can.

    Liked by 3 people

  31. Massimo,

    I really do wish I could see some sign that you or Jonathan truly understand the now quite famous paper that Turkheimer was commenting on:

    http://www.nature.com/mp/journal/v16/n10/full/mp201185a.html

    To begin with, of course Jonathan was trying to undermine the results of twin studies, among other things — this is the point of his bringing up placentas and the like, as if they might represent a genuine objection. What, again, the paper Turkheimer was discussing demonstrated is that even approaching the heritability question from nearly the opposite point of view, namely looking at the gene variants related to pairs of UNRELATED individuals, the same results come up. This not only verifies the heritability estimates of twin studies but also makes most of the far fetched “problems” Jonathan raises seem only less plausible. Heritability is clearly being based on a large representative sample across the full population, with both genes and environment being generalized across that full population, and not on some set of genes and environments (shared or non-shared) that might be peculiar to twins.

    It’s worth bearing in mind what this means for some of Jonathan’s objections. The study found high heritability across the full population and across all environments and genes. What that means is that if there are indeed any significant “norms of reaction” that affect the heritability in that population, then they would downgrade the overall heritability number; this is likewise true for any greatly impoverished environment in which heritability would go down (as per the Turkheimer study Jonathan describes). That is, excluding these presumably unusual situations, the heritability would still be far higher than the already very high figure found in the study. A far more plausible account is, of course, that these phenomena have rather minimal effect in the larger population, and are fairly rare in occurrence.

    The study also reaffirms the importance of additive effects, and therefore of narrow sense heritability. Moreover, it verifies the exceedingly small effect of individual genes, which, again, contra Jonathan’s concerns, would show why we would not yet have our hands on many specific genes that contribute to g. (Jonathan even says, “It is, to say the least, somewhat implausible that hundreds of human psychological traits are each influenced by thousands of independent genes, each with a tiny additive effect.” Really, is this even an argument?)

    Frankly, the arguments both you and Jonathan trot out were raised over 30 years ago. These hoary chestnuts were always stretches even at the time. I should think that the science has progressed enough that they can be put to rest. Indeed, those who are making the genuine scientific progress, such as Visscher, are precisely the people who don’t take them as serious scientific objections. Your point of view has been, one would think, demonstrably nothing more than an impediment to scientific progress.

    Like

  32. Turkheimer et al (2003) [3] found that in the US populations that were studied, in relatively poor families, most of variation in IQ test-taking ability (about 60%) was associated with the shared familial environment, and almost none with genetic variation (the rest was associated with “unique” environments); in relatively affluent families, the reverse held, with most of the variation in IQ test-taking ability (70%) being associated with genetic variation, and almost none associated with shared familial environment (these findings have been relatively robust in the US context).

    No, the findings have certainly not been robust. Turkheimer’s study is a clear outlier in this literature. No other study has found the SES interaction to be anywhere that strong. Turkheimer’s unusual results may stem from the fact that his underpowered study was based on a sample of small children whose IQs were measured 50 years ago.

    The most sophisticated analysis of this issue is by Kirkpatrick et al. (2014). They found, using a contemporary US sample about eight times larger than Turkheimer’s, with a wider range of SES differences, that the heritability of IQ increased only modestly with SES in adolescents. Specifically, in the children of the poorest and least educated parents heritability was about 0.55, while it was about 0.63 in the children of the richest and best educated parents. Heritability averaged across the sample was 0.62. The SES interaction is so small that little is lost by using the average value for everybody. The generalizability problem described by Kaplan doesn’t really exist.

    Like

  33. PLEASE USE THIS VERSION OF COMMENT, with correctly closed “bold” formatting:

    Dear Candid Observer, I wish you would accept that Massimo and Jonathan have already more than addressed your claims, especially Jonathan, in depth, even before his last comment in response to me.

    As for these objections being raised, let alone, as you imply, answered 30 years ago, erm, not even wrong, per Pauli.

    That’s because, among other things, epigenetics wasn’t a blip on a radar screen 30 years ago and gene-environment interaction, at least related to the modern level of study, wasn’t either. For that matter, genetic sequencing wasn’t a blip then either. Per the link I posted in my first comment, a lot of SNP claims have been categorically rejected.

    I’m pretty sure you can see that. Whether you want to, or choose consciously not to, is a different matter. I’m not sure if you’re singing exactly the same chorus, or same vocal range, as a couple of other members of the right wing of the choir, so I’ll pencil you in as a countertenor, OK?

    And, shock me that your paper, directly re my previous comment, is about “g.”

    Jonathan gets somewhat to my second comment, and perhaps more to my first, with this:

    The exchange in Criminology that Turkheimer riffs off of in his post strikes me as very odd (on accuracy of heritability estimates) — a fairly vicious back and forth about whether the heritability estimates from twin-studies are sufficiently accurate, but with no clear sense of what they could be sufficiently accurate for! That is — no clear sense of what use would get made of the numbers… Taking place in the journal Criminology, one would expect someone to say “and this is why knowing this will help us reduce crime.”

    Indeed, this is a limitation.

    It’s more a limitation yet when we take in mind my second post, that I assume that behavioral geneticists on a number of issues, like intelligences, don’t even agree yet on what’s being searched for.

    Then there’s this from Kaplan:

    (W)e should recognize that twin studies designed to estimate heritability, and nothing more, are now a useless exercise. The result is a foregone conclusion (h^2 will come out between .25 and .75, the vast majority of the time), and there is nothing that can be done with the result.

    Correct me if I’m wrong, but it sounds like you’re saying that all most heritability twin studies do is draw us a picture of a bell curve. Gee, bell curves on subjects like intelligence. Does that … erm, “ring a bell” in a certain part of the choir?

    Off on yet another related tangent. I’ve said before that Ev Psych is probably in the Early Bronze Age. I think that’s in part because this from Kaplan:

    If you are thinking of doing a twin study to estimate the heritability of a non-silly trait, and nothing more, you are wasting your time.

    Per him, shows not that BG is “crap,” but that, it’s probably not much more advanced than the Early Bronze Age.

    Like

  34. Sorry for the mixup in my last comment — by “paper” I meant that I support the Johnathan Kaplan post here.

    Like

  35. In discussions about biological vs. environmental (nature vs. nurture) bases of “traits” of humans, what seems to be mostly missing is the possibility that some traits could be created (ex nihilo) by the individuals themselves — assuming, that is, humans are creative animals.

    Like

  36. Dear Massimo – “to get narrow-sense heritability one has to conduct the kind of genetically and environmentally controlled experiments”. This is just incorrect – “narrow sense” = “additive genetic”, and this is frequently estimated in nonexperimental settings outside of human genetics eg field studies in population and ecological genetics. The relevant environmental features do not have to be controlled by an experimenter, it just makes one’s conclusions much cleaner if one can. There are natural experiments, like being a twin. They are not perfect, but they can address some questions that we think are worthwhile.

    Dear Jonathan – “I am…skeptical of the claim that it is telling us that every one of the thousands of traits studied is influenced, in a nice additive way, by some significant fraction of our entire genome.”:
    It is the GWAS studies that suggest that, of course. The results from a twin study could just as easily be due to a single locus. There are over 100 loci affecting sensory bristle number in fruit flies known so far, and I hardly think human traits will be simpler. Directional artificial or natural selection causes movement in allele frequencies at a large number of loci simultaneously. It is also the case that complex interactions at the biological level appear at the population level to be additive (eg Hill et al 2008). Note also (this to other readers) that loci are variable sites, not the 20000 genes – most of the GWAS hits are outside the structural genes, in regulatory regions, and affect complex control networks (in those cases where we have been able to work them out).

    Dear SG, although we do have examples of transgenerational epigenetic inheritance, the fact is that they do not make important contributions to inheritance of most traits in humans. The vast bulk of interest in epigenetic modification of DNA is as just another phenotype influenced by genetic and environmental factors. Many of the loci found by GWAS act on phenotypes by altering amounts of other gene products (expression). The turning on and off of genes by these regulatory elements is associated with chromatin and methylation epigenetic alterations at the target gene.

    As to replicability of SNP associations, I can assure you that most of the ones I have worked on have replicated nicely. The main problem is study power – most of the effects sizes require large samples to detect them. Ioannidis is always fun to read (another useful gadfly!), the history of candidate gene studies is not pretty, but the big collaborative studies continue to be funded because they find things. As to sequencing, at the seminar I mentioned above it was suggested the effects of 98% of common and 76% of rare variants can be detected just using medium density SNP arrays and imputation.

    Like

  37. candid,

    “This not only verifies the heritability estimates of twin studies”

    You insist in misreading Jonathan, even though he has repeated his point several times: yes, heritability estimates are consistent and repeatable. They are also close to useless. *That* is the issue.

    “with both genes and environment being generalized across that full population”

    Biologically, that’s utter nonsense. One cannot “generalize” genes and environments, because the issue at hand is the shape and variability of human reaction norms. These are known, in experimental organisms, to vary enormously with different genotypes and environments. If one “generalizes” (i.e., averages out) one is missing much of what is interesting, coming up with exactly the sort of well known and lame results that research on heritability keeps yielding: the “average” heritability of a trait varies between .4 and .6, and it is distributed normally. Duh.

    “What that means is that if there are indeed any significant “norms of reaction” that affect the heritability in that population”

    No, it doesn’t mean that. We have no idea how human reaction norms affect heritability, although it is definitely the case that in experimental organisms the shape and variation of reaction norms can make the heritability for the same trait vary between 0 and 1. (I know because I did some of those experiments.)

    “the arguments both you and Jonathan trot out were raised over 30 years ago”

    Indeed, and they keep being either ignored or sidestepped by people obsessed with what should be obvious is an increasingly useless statistic.

    WS,

    “Specifically, in the children of the poorest and least educated parents heritability was about 0.55, while it was about 0.63 in the children of the richest and best educated parents”

    Besides the fact that the difference between .55 and .63 is rather irrelevant, doesn’t that strongly suggest an environmental effect on heritability? Or do you think that the children of rich and educated parents are significantly genetically different from those of poor and uneducated ones?

    david,

    “”to get narrow-sense heritability one has to conduct the kind of genetically and environmentally controlled experiments”. This is just incorrect – “narrow sense” = “additive genetic”, and this is frequently estimated in nonexperimental settings outside of human genetics eg field studies in population and ecological genetics. The relevant environmental features do not have to be controlled by an experimenter, it just makes one’s conclusions much cleaner if one can.”

    Sorry, but no. It is well known that estimates of “additive” genetic variance obtained by indirect methods are at odd with those estimated by, for instance, selection experiments. This makes no sense unless those different methods are in fact not estimating the same thing, or not entirely.

    Indeed, let me go one step further: the whole additive vs. (generic) non-additive discussion is a leftover from the early days of quantitative genetics, and it is based on highly unrealistic statistical assumptions. It is not an estimate of actual genetic effects, since genes simply don’t enter into the picture; the numbers refer to phenotypic measurements, not genetic ones (they are statistically *associated* with presumed genetic differences, but that’s not the same thing).

    The best way to calculate narrow sense heritability is to do selection experiments, which is obviously out of the question in humans. But even that wouldn’t really get at what ought to be the interesting issue: the shape and variability of human reaction norms.

    So we are left with a lot of work, over decades, that has arrived at the following entirely trivial and not particularly useful conclusions: most human traits are affected both by genetic factors and by environmental ones. Wow, who would have thought?

    Liked by 4 people

  38. Besides the fact that the difference between .55 and .63 is rather irrelevant, doesn’t that strongly suggest an environmental effect on heritability? Or do you think that the children of rich and educated parents are significantly genetically different from those of poor and uneducated ones?

    Yes, it does suggest an environmental effect on heritability. That’s the point of the study. Behavioral geneticists usually explain such effects in terms of there being greater opportunities for trait development in higher-SES families. However, contrary to what Kaplan wrote in his post, the h^2 difference between SES levels is small.

    The children of rich and educated parents are, of course, significantly genetically different from those of poor and uneducated ones. For example, many of the genetic variants that are associated with higher parental SES are also associated with higher IQ in children.

    Like

  39. Massimo said:
    most human traits are affected both by genetic factors and by environmental ones. Wow, who would have thought?

    That sums it up rather nicely.

    I want to suggest that another way to look at the problem is to see the human being as a work in progress. If we see where we came from, comparing it with where we are, we can see an important trajectory which helps us to make sense of the subject.

    I will start with an analogy, based on two of my dogs, both excellent runners. One, a Border Collie, is obedient and highly trainable. I can safely run through city streets with her at my heel. I cannot do this with my other dog, a Jack Russell terrier. He can never be let off the leash in city streets because of his uncontrollably impulsive behaviour. Then there is my third dog, a cross between a Border Collie and Rottweiler. Her behaviour is a blend between the loving(to my family) nature of a Border Collie and the vicious, fierce nature of a Rottweiler. I cannot trust her in the presence of a stranger.

    These behaviours are typical of their breeds and they are certainly genetic in origin, the consequence of a long breeding programme. In effect, my dogs’ behaviours are constrained by genetic straitjackets. Their behaviour is predictable and they cannot choose to be otherwise. I(their environment) have limited ability to shape these essential aspects of their behaviour.

    This is where we were as a species, with our behaviour constrained by genetic straitjackets. Today we are different by orders of magnitude. We make free choices, shape our own behaviour and acquire new behaviours, as evidenced by our vast and creative cultural output. There is an immeasurable difference between the creative output of my dogs and the creative output of our species. Somehow, the genetic straitjacket that constrained our cognitive apparatus has been loosened and we are no longer just our genes, from the behavioural point of view. Genetic essentialism is certainly true of my dogs and it was true of our species in the past. But if one considers the trajectory of our development it is to a large extent not true today.

    Today our behaviour is the result, in part, of genetic influences, in large part, the result of environment and equally the result of our ability to exercise free choices(Turkheimer calls it the free will coefficient). But the really important thing is that our development is on a trajectory from the highly constrained behaviour of a distant past, to today’s partly constrained behaviour, and if the trajectory is continued, to a future where our behaviour is largely free of genetic constraints. We are becoming a self-modifying species. That will be true freedom and there is no reason to think that our progress along this trajectory has stopped.

    Genetic essentialism is a lost cause, though it remains attractive to scientismists and free will denialists.

    See also http://bit.ly/1GmAQDP, http://bit.ly/1Qt4UNi and http://bit.ly/1Kcfc4X

    Liked by 2 people

  40. Candid Observer: “I really do wish I could see some sign that you or Jonathan truly understand the now quite famous paper that Turkheimer was commenting on…”
    And I really do wish I could see some sign that you’d actually read the post I’d written, and not some fantasy post you’d made up, but I fear I at least am doomed to be disappointed.

    For those people that are interested in the details of traditional GWAS studies versus GCTA studies, the following may be of some limited interest. I suspect everyone else will be justifiably bored.

    Traditional genome-wide association studies attempted to find genetic variations associated with some variation in a phenotypic trait across (wait for it!) the whole genome. Of course, no studies really quite do that — despite the rapidly decreasing sequencing costs and increasing computer speeds, actually comparing entire sequenced genomes is still impossible, from a practical standpoint. So you use a bunch of markers across the genome (for a variety of reasons SNPs — Single Nucleotide Polymorphisms — are by far the most common kind of genetic variant to use in GWAS). Early efforts had problems w/ high false-positive rates, and given their low power, poor replication ability even of ‘true’ hits. They’ve gotten much better, though how much better is still debated. To a first approximation, if you find that a particular genetic variant is associated with a particular phenotypic variant more often than one would expect by chance, you mark it as a ”possible,” and then retest in a new population sample to see if it comes up again. If it does, that suggests that it is really associated with the phenotype in question, and not just by chance. (This is a very “rough and ready” description of how this works, and people interested in the details should consult the primary literature.)

    The hope with these studies was that you’d find genomic regions associated with the phenotype, and then start to be able to figure out, mechanistically, what those regions did — what proteins they coded for, what regulatory functions they were involved with, etc., — and actually have an entry into understanding the development of complex traits. That would be pretty cool.

    The problem, as noted above, was that many of these studies failed to find genetic variation associated with more than a trivial portion of the genetic variation that was supposed to be there. When GWAS was young, and sample sizes were small, this was only a little surprising — many people had hoped that alleles associated with more than a few percent of the variance in the trait would be common, but since the studies didn’t have sufficient power to detect anything much smaller than that, people assumed that actually more genes with smaller effects must be what was going on. However, as power has increased — and in some heavily studied areas has gotten to the point where the studies have a good chance of finding any SNP associated with more than 1/10 of 1% of the variation — and have still not succeeded in finding genes associated with more than a trivial portion of the variance of the traits, some researchers (not all!) have become concerned that the standard “well, the genes must just have *even smaller* effects!” line is getting a little implausible. I’ve explained (briefly) my reasons for being on the ‘getting a little implausible’ side, but I freely acknowledge that no one (to the best of my knowledge) has a knock-down argument that it is *impossible*. It’s just getting a little weird, at this point.

    GCTA studies, an example of which is what CO links to above and accuses Massimo and me of not having understood, work on a somewhat different principle. (For those interested — and again, unless you have a good reason to care about this, I’ve no idea why you would be — the classic paper in which the basic technique is explained, and some of the rather complicated technical details laid out, can be found here: http://www.sciencedirect.com/science/article/pii/S0002929710005987 — no, I don’t pretend to have followed all the ins and outs of their statistical arguments or the coding details, as this isn’t, for reasons that should be clear to anyone who actually read my post of my comments so far, something I really care all that much about.) Again, the following is a very ‘rough and ready’ first-approximation description of the process — readers interested in the details are well-advised to consult the primary literature. Basically, though, these studies take the set of genetic variants associated with each genome in the database, and compare them, generating an index of how similar they are — roughly, a estimate of how ‘related’ the two genomes are to each, where this ‘relatedness’ is virtual rather than actual (e.g. “it is ‘as if’ they were nth degree cousins”). They *exclude* any comparisons between genomes that are *too* related to one another, on the grounds that they might *actually* be related (actually from the same real family), and that this would introduce possible confounding factors (like shared environments) into the analysis. Additional steps are taken to remove the influences of population structure, though what these steps are vary by who is doing the study and how, and remain an area of contention. How “related” individuals are is then compared to how similar they are with respect to the phenotype of interest, and the heritability of the trait estimated via an extension of the reasoning explained above.

    There are a number of ways these estimates can go wrong, but like twin-studies, some of these ways of going wrong will tend to result in over-estimations of heritability, and some in under-estimations, and the arguments about which are more important seem to generate more heat than light, at least as far as I can tell, as an outsider to these debates.

    One Interesting feature of this research is that sometimes GCTA studies come up with estimates of heritability that are close to those that more traditional methods (e.g. twin studies) come up with, and sometimes they don’t. Detractors of twin-studies (both in general, and of their ability to generate accurate estimates of heritability) focus on the times they don’t. Supports of twin-studies as a way of estimating heritability focus on the times that they do. Again, lots of heat, surprising little light.

    Again, for the reasons given above, my interest is not primarily in how accurate any particular estimate of heritability is, but of what heritability estimates in fact mean, and what uses can and can’t be legitimately made of them. So my interest in new methods of estimating heritability is pretty low. (cf Lewontin’s final line in his classic “Analysis of Variance and Analysis of Causes.)

    My interest in “traditional” GWAS studies was the promise that they would find the regions associated with differences, and permit researchers to begin to actually uncover the causal pathways involved in those differences — that’s actually kind of exciting work. That they have so far failed to live up to their promise is a interesting fact, and means that much the work of figuring how, in humans, to begin to unravel those causal pathways remains to be done. Since GCTA studies do not identify particular genes or regions, they are not useful for this project, and much less exciting (read: more or less completely useless) if what one was looking for was a way to do what I, and for example Massimo, regard as the important work of figuring how genes and other developmental resources, including environments, interact to produce the phenotypes and phenotypic differences that we care about.

    Other people are interested in different projects, of course, and my understanding is that some practical applications of GCTA analyses have actually been found (e.g. some agricultural contexts). My general lack of interest in the details of rice-breeding has prevented me from exploring these uses in much detail, alas. However, they do seem to involve extensive experimental manipulations (particular hybridization crosses, etc.) that make those uses less than plausible for studies involving humans. Given the lack of causal information afforded by GCTA studies, I remain skeptical that they have practical applications in the human case, but I may well be wrong about that. Their use in generating heritability estimates, however, are simply another way of generating an estimate of a number that is, as I noted above, of much less use than it is sometimes thought to be.

    As far as some of the other criticisms go (e.g., what an appropriate review of the complete literatures on the ways that SES and measures of cognitive ability interact with h^2 and C, etc. might look like) they too seem to have (rather deliberately) missed the point, and as I can only repeat that point so many times before it gets tiresome, I’m not going to bother.

    Liked by 1 person

  41. Massimo,

    One of the points I had been trying to make is a simple one: if heritability is, averaged across the population, quite high – say .6 for a cognitive trait – then such things as differential “norms of reaction” will be possible only if there are, in the different settings, both heritabilities higher than .6 and heritabilities lower than .6. (I guess in principle it could even be more extreme, so that genes positively correlated with a trait in one setting are negatively correlated in another – but I don’t think there’s a serious argument for that in any of the cases we might be concerned about). But if the heritability is already as high as .6 on average, how much room is there for zero or very low heritabilities in that population? You can push down heritability in one setting, but then it has to pop up in another. Turkheimer’s study seemed to show some circumstances under which heritability was close to zero, but as WS pointed out, further studies have pretty effectively contradicted that result.

    In short, there’s always talk about “norms of reaction”, and significantly different heritabilities in different settings, but never any convincing case made of its impact in the cases we care about.

    This is in fact a more general complaint about the environmental side of this debate – all kinds of theoretical “problems” get raised, but there’s never any delivery on the actual existence of these potential factors.

    This applies to the gene-environment vicious circle which Flynn raises, and which Jonathan brings up. Yes, in principle, small effects by genes might, through a feedback mechanism with the environment, increase those differences substantially. But where’s the evidence this ever happens with regard to cognitive traits like IQ? Do children who start out kindergarten later, who should seem on average distinctly smarter than their peers, some of whom may be a full year younger, actually do better than their younger peers over the long run on IQ? I certainly have heard of no study confirming that expectation – in fact there appears to be no correlation between birth date in the year and IQ. Nor is there any other evidence of such a connection.

    I would add to this list of purely speculative “problems” with heavily genetic accounts of these traits the whole “epigenetics” thing. Is there any direct evidence that for the cases of actual interest, such as IQ, that “epigenetics” has any effect whatsoever? No, certainly not. But it’s more sand in the eyes for the proposition that IQ is dominantly genetic in origin – and let’s face it, for the environmentalists, that is the real point.

    But, as I said earlier, those interested in truth and science simply ignore the contrived “problems”, conduct their science, and make steady progress despite the deliberate distractions.

    Like

  42. Labut —

    On a completely different topic, my understanding is that the “personality” that Rottweilers were bred for was NOT “vicious, fierce” but rather “Devoted, Good-natured, Alert, Steady, Self-assured, Obedient, Calm, Confident, Courageous, Fearless” (FCI and AKC standards are similar here, I’m giving to understand). They are larger dogs, and powerfully built, and therefore (much) more dangerous if they do attack than small dogs, but the breed as a whole is generally less likely to attack than are many smaller dog breeds. From a breed perspective, your Jack Russell is *far* more likely to exhibit aggressive behavior than a Rottweiler! (Seriously, look up the stats for passing various aggressive behavior tests, etc. — Jack Russell’s are little terrors from that perspective, but of course wonderful dogs in many many other ways!) It’s just that it doesn’t matter quite so much if a Jack Russell (tries to) bite someone!

    But of course, different individual dogs within a breed have different personalities and tendencies, and the breed tendency is only a part of the picture (albeit an important part!). The genetic “straightjacket,” even in dogs, binds some things tightly, and others not tightly at all. (Similarly, humans are ‘bound’ to become sophisticated language users under almost all developmental conditions we regularly encounter, and almost all our available genotypes — we our ‘bound’ quite tightly by our nature in that way! But what uses we make of that ability varies widely, of course…)

    But it is true that anyone who thinks genetics, or even more generally selection for certain behavioral tendencies, simply can’t influence behavior doesn’t understand either, and the extensive experiments with dog breeds reveals just how robust certain selected behavior tendencies can be across various changes in the developmental environment. This is not, or should not be, controversial. But heritability estimates don’t tell you that. Extensive and carefully done cross-fostering experiments tell you that! (And of course also tell you what traits *aren’t* stable across changes in the developmental environment!) Heritability estimates are rather besides the point in these cases. And that, from the standpoint of the point of the post, is what is important here!

    Liked by 3 people

  43. Hi all, but especially Jonathon Kaplan andMassimo,

    This is a very small world indeed… as a matter of full disclosure my professor and lab are in direct collaboration with the group that authored the paper being reviewed. I do not know the first authors at all but I know the last author and lead professor Danielle Posthuma a little, work daily with some of her PhD students, and in fact had a lecture from her just a few hours ago that included discussion of this very article! I wish I had read this essay before going to class, as I could have asked more intelligent questions ☺

    To be honest I have my own issues with both twin studies and heritability estimates, particularly when it comes to behavior. Unless twin studies involve people separated at birth there are too many confounding factors to get from genes to behavior. Outside of chemical environmental issues (in utero or ex), twins are usually treated “special” which can affect how they act. And it is not always that they are made to be identical. There are also people who drive them to be different. And in any case the idea they share 100% of the same environment seems like a flawed assumption. This point was raised during today’s discussion and she seemed to acknowledge there are concerns about it, while downplaying significance.

    I was also dubious about averaging heritability of all traits. I asked her about this but did not get a clear answer (we had to move on and I’ll ask her again later).

    Finally I found some traits considered heritable (and this is not just their report) rather sketchy. During her talk, Danielle gave “education level attained” as something investigated as heritable. I sort of came out of my chair on that one. I raised the question whether they controlled for family income (and frankly there is more to this that could be considered). The answer would seem to be no, since it included studies from different countries (which treat education finance very differently). She agreed that could be an issue.

    I should be clear that she downplayed hyperbolic headlines coming out about their paper, and dispelled some public/common misunderstandings along the lines that Jonathon laid out. The nature/nurture debate is not over except to say that both are involved. Their exact contribution is not 50% and the real ratio needs to be determined case by case by further data/experimentation.

    Currently the Posthuma lab is working with my own lab to investigate/connect results suggested by GWAS to brain function relevant to certain clinical phenotypes.

    Her work is not my field at all so I can’t answer questions myself. However she is very nice and if anyone has specific questions I can try to get answers. I can also try to alert her (and her students) to this essay, but she is pretty busy and I think discussion will close before she could answer.

    Liked by 2 people

  44. candid,

    “if heritability is, averaged across the population, quite high – say .6 for a cognitive trait – then such things as differential “norms of reaction” will be possible only if there are, in the different settings, both heritabilities higher than .6 and heritabilities lower than .6”

    I’m sorry, but based on this it is now clear to me that you simply do not understand what a reaction norm is, or what the relationship between a reaction norm and heritability is. Averaging heritabilities, given that they are *by definition* local measures, sensitive to both genotype and environment, makes no sense whatsoever. To then claim that these averages tell you anything at all about the underlying reaction norm is simply not true.

    “if the heritability is already as high as .6 on average, how much room is there for zero or very low heritabilities in that population?”

    Who cares? I believe Jonathan is correct at this point: you are reading your own version of his essay, bot what he actually wrote, or what I tried to explain in my replies.

    “there’s always talk about “norms of reaction”, and significantly different heritabilities in different settings, but never any convincing case made of its impact in the cases we care about.”

    Because reaction norms cannot be studied in the case of humans! But we have a huge literature on this in experimental models, that’s why these issues are raised, are known, and are important. My own book on this is pretty old now, but here it is, just in case: http://goo.gl/I7Vtra

    holmes,

    “I wish I had read this essay before going to class, as I could have asked more intelligent questions”

    Wow, a small world indeed!

    “During her talk, Danielle gave “education level attained” as something investigated as heritable. I sort of came out of my chair on that one”

    I expect that!

    “I raised the question whether they controlled for family income (and frankly there is more to this that could be considered). The answer would seem to be no … She agreed that could be an issue.”

    Yeah, that’s a pretty serious understatement…

    “The nature/nurture debate is not over except to say that both are involved. Their exact contribution is not 50% and the real ratio needs to be determined case by case by further data/experimentation”

    As I explained before, that’s precisely what I find objectionable, and – frankly – rather naive from a professional biologist: the “ratio” is an entirely meaningless statistic, as was pointed out by Lewontin back in 1974. See above, my example (straight from Lewontin) of the bricks and the lime to make a house.

    Liked by 1 person

  45. Hi Massimo, to be fair I should have put “real ratio” in scare quotes. I don’t think she used that phrase. The concept she outlined was finding the actual contributions of genetic and environmental factors, which is not necessarily understandable as a ratio (as you rightly suggest).

    As I understand it the techniques used in her group are meant to uncover potentially fruitful lines of research, rather than definitive estimates of contribution. This is especially true given the large numbers of genes, with small statistical connections to phenotype. Hence their recent connection with my lab.

    Like

  46. Jonathan Kapland —

    Very informative and clearly presented. I especially liked your adapted basketball example.

    And your comments too. I’m sure that like me a lot of people learned a lot from them in many ways : to solidify, correct, or fill in holes in their understanding of your article.

    I get the impression people, including me, easily mix up ideas of inheritance/heritance and heritability, in that context the following helped drive the point home :

    “Note first that heritability is not a measure of “how genetic” a trait is”

    “Put simply, heritability is the proportion of the phenotypic variation in a trait of interest, measured in a given studied population and in a given environment, that is statistically co-varying with genetic differences (however measured) among individuals in the same population”

    “Since heritability is a measure of what is associated with variation in the trait, and not a measure of what causes the trait, the heritability of finger number in humans is essentially zero, and the vast majority of variation in finger number is environmental (traumatic amputations are the primary cause!)”

    And from Turkheimer :

    “everything is heritable”

    “The heritability of criminality doesn’t mean that it has somehow turned out to be ‘biological’ ”

    “A heritability study isn’t a thing. A heritability is a descriptive statistic, an effect size, not a kind of study”

    Like

  47. Candid_Observer —

    “Is there any direct evidence that for the cases of actual interest, such as IQ, that “epigenetics” has any effect whatsoever? No, certainly not”

    Epigenetic Marks Lay Foundations for a Child’s Future Abilities

    ( http://ije.oxfordjournals.org/content/early/2015/04/22/ije.dyv052 )

    Like

  48. “… the more important point is that heritability doesn’t measure what most people think it measures,”

    The problem with science is measurability. Nature is truly (philosophically) immeasurable.
    Copenhagen anyone? =

    Like

  49. Hi Jonathan,
    I concur, thanks for that info, you know your breeds better than I do!

    Heritability estimates are rather besides the point in these cases. And that, from the standpoint of the point of the post, is what is important here!

    Quite so. Thanks for a really valuable essay.

    Like

  50. One more shout-out to Turkheimer’s blog. In a recent post (still nominally on that Criminology debate) he discusses why, in his view, twin-studies (and e.g. family studies & studies of siblings more generally) can be very valuable tools, even if one doesn’t care about what the “right” estimate of heritability actually is. It is, I think, a compelling argument that one can both be a supporter of doing twin-studies, and of the tools of behavior genetics, without being committed to thinking that arguments over the exact proportion of variance associated with broad- or narrow-sense heritability make a whole lot of sense.

    Again, I may not agree with everything in it, but think that readers interested in these issues (including some of those raised by the comments above and in the context of my original post) would be well-advised to read it!

    http://ericturkheimer.blogspot.com/2015/06/b-b-b-behavior-genetics-and-criminology.html

    Liked by 2 people

%d bloggers like this: