Home » Heritability
Category Archives: Heritability
Twin and adoption studies have been used for decades on the basis that genetic and environmental causes of traits and their variation in the population could be easily partitioned by two ways: one way is to adopt twins into separate environments, the other to study reared-together or reared-apart twins. Both methods rest on a large number of (invalid) assumptions. These assumptions are highly flawed and there is no evidential basis to believe these assumptions, since the assumptions have been violated which invalidates said assumptions.
Plomin et al write (2013) write: For nearly a century, twin and adoption studies have yielded substantial estimates of heritability for cognitive abilities.
But the validity of the “substantial estimates of heritability for cognitive abilities” is strongly questioned due to unverified (and false) assumptions that these researchers make.
The problem with adoption studies are numerous, not least: restricted range of adoptive families; selective placement; late separation; parent-child attachment disturbance; problems with the tests (on personality, ‘IQ’); the non-representativeness of adoptees compared to non-adoptees; and the reliability of the characteristic in question.
In selective placement, the authorities attempt to place children in homes close to their biological parents. They gage how “intelligent” they believe they are (on the basis of parental SES and the child’s parent’s perceived ‘intelligence’), thusly this is a pretty huge confound for adoption studies.
According to adoption researcher Harry Munsinger, a “possible source of bias in adoption studies is selective placement of adopted children in adopting homes that are similar to their biological parents’ social and educational backgrounds.” He recognized that “‘fitting the home to the child’ has been the standard practice in most adoption agencies, and this selective placement can confound genetic endowment with environmental influence to invalidate the basic logic of an adoptive study (Munsinger, 1975, p. 627). Clearly, agency policies of “fitting the home to the child” are a far cry from random placement of adoptees into a wide range of adoptive homes. (Joseph, 2015: 30-1)
Richardson and Norgate (2005) argue that simple additive effects for both genetic and environmental effects are false; that IQ is not a quantitative trait; while other interactive effects could explain the IQ correlation.
1) Assignment is nonrandom. 2) They look for adoptive homes that reflect the social class of the biological mother. 3) This range restriction reduces the correlation estimates between adopted children and adopted parents. 4) Adoptive mothers come from a narrow social class. 5) Their average age at testing will be closet to their biological parents than adopted parents. 6) They experience the womb of their mothers. 7) Stress in the womb can alter gene expression. 8) Adoptive parents are given information about the birth family which may bias their treatment. 9) Biological mothers and adopted children show reduced self-esteem and are more vulnerable to changing environments which means they basically share environment. 10) Conscious or unconscious aspects of family treatment may make adopted children different from other adopted family members. 11) Adopted children also look more like their biological parents than their adoptive parents which means they’ll be treated accordingly.
Personally, my favorite thing to discuss. Twin studies rest on the erroneous assumption that DZ and MZ environments are equal; that they get treated equally the same. This is false, MZ twins get treated more similarly than DZ twins, which twin researchers have conceded decades ago. But in order to save their field, they attempt to use circular argumentation, known as Argument A. Argument A states that MZTs (monozygotic twins reared together) are more genetically similar than DZTs (dizygotic twins reared together) and thusly this causes greater behavioral similarity. But this is based on circular reasoning: the researchers already implicitly assumed that genes played a role in their premise and, not surprisingly, in their conclusion genes are the cause for the similarities of the MZTs. So Argument A is used, twin researchers circularly assume that MZTs greater behavioral similarity is due to genetic similarity, while their argument that genetic factors explain the greater behavioral similarity of MZTs is a premise and conclusion of their argument. “X is true because Y is true; Y is true because X is true.” (Also see Joseph et al, 2015.)
We have seen that circular reasoning is “empty reasoning in which the conclusion rests on an assumption whose validity is dependent on the conclusion” (Reber, 1985, p. 123). … A circular argument consists of “using as evidence a fact which is authenticated by the very conclusion it supports,” which “gives us two unknowns so busy chasing each other’s tails that neither has time to attach itself to reality” (Pirie, 2006, p. 27) (Joseph, 2016: 164).
Even if Argument A is accepted, the causes of behavioral similarities between MZ/DZ twins could still come down to environment. Think of any type of condition that is environmentally caused but is due to people liking what causes the condition. There are no “genes for” that condition, but their liking the thing that caused the condition caused an environmental difference.
Argument B also exists. Those that use Argument B also concede that MZs experience more similar environments, but then argue that in order to show that twin studies, and the EEA, are false, critics must show that MZT and DZT environments differ in the aspects that are relevant to the behavior in question (IQ, schizophrenia, etc).
An example of an Argument B environmental factor relevant to a characteristic or disorder is the relationship between exposure to trauma and post-traumatic stress disorder (PTSD). Because trauma exposure is (by definition) an environmental factor known to contribute to the development of PTSD, a finding that MZT pairs are more similarly exposed to trauma than DZT pairs means that MZT pairs experience more similar “trait-relevant” environments than DZTs. Many twin researchers using Argument B would conclude that the EEA is violated in this case. (Joseph, 2016: 165)
So twin researchers need to rule out and identify “trait-relevant factors” which contribute to the cause of said trait, along with experiencing more similar environments, invalidates genetic interpretations made using Argument B. But Argument A renders Argument B irrelevant because even if critics can show that MZTs experience more similar “trait-relevant environments”, they could still argue that the twin method is valid by stating that (in Argument A fashion) MZTs create and elicit more similar trait-relevant environments.
One more problem with Argument A is that it shows that twins behave accordingly to “inherited environment-creating blueprint” (Joseph, 2016: 164) but at the same time shows that parents and other adults are easily able to change their behaviors to match that of the behaviors that the twins show, which in effect, allows them to “create” or “elicit” their own environments. But the adults’ “environment-creating behavior and personality” should be way more unchangeable than the twins’ since along with the presumed genetic similarity, adults have “experienced decades of behavior-molding peer, family, religious, and other socialization influences” (Joseph, 2016: 165).
Whether or not circular arguments are “useful” or not has been debated in the philosophical literature for some time (Hahn, Oaksford, and Corner, 2005). However, assuming, in your premise, that your conclusion is valid is circular and therefore While circular arguments are deductively valid, “it falls short of the ultimate goal of argumentative discourse: Whatever evaluation is attached to the premise is transmitted to the conclusion, where it remains the same; no increase in degree of belief takes place” (Hahn, 2011: 173).
However, Hahn (2011: 180) concludes that “the existence of benign circularities makes clear that merely labeling something as circular is not enough to dismiss it; an argument for why the thing in question is bad still needs to be made.” This can be simply shown: The premise that twin researchers use (that genes cause similar environments to be constructed) is in their conclusion. They state in their premise that MZT behavioral similarity is due to greater MZT genetic similarity in comparison to DZTs (100 vs. 50 percent). Then, in the conclusion, they re-state that the behavioral similarities of MZTs is due to their genetic similarity compared to DZTs (100 vs. 50 percent). Thus, a convincing argument for conclusion C (that genetic similarity explains MZT behavioral similarity) cannot rest on the assumption that conclusion C is correct. Thus, Argument A is fallacious due to its circularity.
What causes MZT behavioral similarities is their more similar environment: they get treated the same by peers and parents, and have higher rates of identity confusion and had a closer emotional bond compared to DZTs. The twin method is based on the (erroneous) assumption that MZT and DZT pairs experience roughly equal environments, which twin researchers conceded was false decades ago.
Richardson and Norgate (2005: 347) conclude (emphasis mine):
We have shown, first, that the EEA may not hold, and that well-demonstrated treatment effects can, therefore, explain part of the classic MZ–DZ differences. Using published correlations, we have also shown how sociocognitive interactions, in which DZ twins strive for a relative ‘apartness’, could further depress DZ correlations, thereby possibly explaining another part of the differences. We conclude that further conclusions about genetic or environmental sources of variance from MZ–DZ twin data should include thorough attempts to validate the EEA with the hope that these interactions and their implications will be more thoroughly understood.
Of course, even if twin studies were valid and the EEA was true/ the auxiliary arguments used were true, this would still not mean that heritability estimates would be of any use to humans, since we cannot control environments as we do in animal breeding studies (Schonemann, 1997; Moore and Shenk, 2016). I have chronicled how 1) the EEA is false and how flawed twin studies are; 2) how flawed heritability estimates are; 3) how heritability does not (and cannot) show causation; and 4) the genetic reductionist model that behavioral geneticists rely on is flawed (Lerner and Overton, 2017).
So we can (1) accept the EEA, that the greater behavioral resemblance indicates the importance of genetic factors underlying most human behavioral differences and behavioral disorders or we can (2) reject the EEA and state that the greater behavioral resemblance is due to nongenetic (environmental) factors, which means that all genetic interpretations of MZT/DZT studies must be rejected. Thus, using (2), we can infer that all twin studies measure is similarity of the environment of DZTs, and it is, in fact, not measuring genetic factors. Accepting explanation 2 does not mean that “twin studies overestimate heritability, or that researchers should assess the EEA on a study-by-study basis, but instead indicates that the twin method is no more able than a family study to disentangle the potential influences of genes and environment” (Joseph, 2016: 181).
What it does mean, however, is that we can, logically, discard all past, future, and present MZT and DZT comparisons and these genetic interpretations must be outright rejected, due to the falsity of the EEA and the fallaciousness of the auxiliary arguments made in order to save the EEA and the twin method overall.
There are further problems with twin studies and heritability estimates. Epigenetic supersimilarity (ESS) also confounds the relationship. Due to the existence of ESS “human MZ twins clearly cannot be viewed as the epigenetic equivalent of isogenic inbred mice, which originate from separate zygotes. To the extent that epigenetic variation at ESS loci influences human phenotype, as our data indicate, the existence of ESS establishes a link between early embryonic epigenetic development and adult disease and may call into question heritability estimates based on twin studies” (Van Baak et al, 2018). In other words, ESS is an unrecognized phenomenon that contributes to the phenotypic similarity of MZs, which calls into question the usefulness of heritability studies using twins. The uterine environment has been noted to be a confound by numerous authors (Devlin, Daniels, and Roeder, 1997; Charney, 2012; Ho, 2013; Moore and Shenk, 2016).
Adoption studies fall prey to numerous pitfalls, most importantly, that children are adopted into similar homes compared to their birth parents, which restricts the range of environments for adoptees. Adoption placement is also non-random, the children are placed into homes that are similar to their biological parents. Due to these confounds (and a whole slew of other invalidating problems), adoption studies cannot be said to show genetic causation, nor can they separate genetic from environmental factors.
Twin studies suffer from the biggest flaw of all: the falsity of the EEA. Since the EEA is false—which has been recognized by both critics and supporters of the assumption—the supporters of the assumption have attempted to redefine the EEA in two ways: (1) that MZTs experience more similar environments due to genetic similarity (Argument A) and (2) that it is not whether MZTs experience more similar environments, but whether or not they share more similar trait-relevant environments. Thus, unless these twin researchers are able to identify trait-relevant factors that contribute to the trait in question, we must conclude that (along with the admission from twin researchers that the EEA is false; that MZTs experience more similar environments than DZTs) genetic interpretations made using Argument B are thusly invalidated. Fallacious reasoning (“X causes Y; Y causes X) does not help any twin argument. Because their conclusion is already implicitly assumed in their premise.
The existence of ESS (epigenetic supersimilarity) further shows how invalid the twin method truly is, because the confounding starts in the womb. Attempts can be made (however bad) to control for shared environment by adopting different twins into different homes, but they still shared a uterine environment which means they shared an environment, which means it is a confound and it cannot be controlled for (Charney, 2012).
Adoption and twin studies are highly flawed. Like family studies, twin studies are no more able to disentangle genetic from environmental effects than a family study, and thus twin studies cannot separate genes from environment. Last, and surely not least, it is fallacious to assume that genes can be separated so neatly into “heritability estimates” as I have noted in the past. Heritability estimates cannot show genetic causation, nor can it show how malleable a trait is. They’re just (due to how we measure) flawed measures that we cannot fully control so we must make a number of (false) assumptions that then invalidate the whole paradigm. The EEA is false, all auxiliary arguments made to save the EEA are fallacious; adoption studies are hugely confounded; twin studies are confounded due to numerous reasons, most importantly the uterine environment (Van Baak et al, 2018).
Testosterone has a similar heritability to IQ (between .4 and .6; Harris, Vernon, and Boomsma, 1998; Travison et al, 2014). To most, this would imply a significant effect of genes on the production of testosterone and therefore we should find a lot of SNPs that affect the production of testosterone. However, testosterone production is much more complicated than that. In this article, I will talk about testosterone production and discuss two studies which purport to show a few SNPs associated with testosterone. Now, this doesn’t mean that the SNPs cause high/low testosterone, just that they were associated. I will then speak briefly on the ‘IQ SNPs’ and compare it to ‘testosterone SNPs’.
Complex traits are ‘controlled’ by many genes and environmental factors (Garland Jr., Zhao, and Saltzman, 2016). Testosterone is a complex trait, so along with the heritability of testosterone being .4 to .6, there must be many genes of small effect that influence testosterone, just like they supposedly do for IQ. This is obviously wrong for testosterone, which I will explain below.
Back in 2011 it was reported that genetic markers were discovered ‘for’ testosterone, estrogen, and SHGB production, while showing that genetic variants in the SHGB locus, located on the X chromosome, were associated with substantial testosterone variation and increased the risk of low testosterone (important to keep in mind) (Ohlsson et al, 2011). The study was done since low testosterone is linked to numerous maladies. Low testosterone is related to cardiovascular risk (Maggio and Basaria, 2009), insulin sensitivity (Pitteloud et al, 2005; Grossman et al, 2008), metabolic syndrome (Salam, Kshetrimayum, and Keisam, 2012; Tsuijimora et al, 2013), heart attack (Daka et al, 2015), elevated risk of dementia in older men (Carcaillon et al, 2014), muscle loss (Yuki et al, 2013), and stroke and ischemic attack (Yeap et al, 2009). So this is a very important study to understand the genetic determinants of low serum testosterone.
Ohlsson et al (2011) conducted a meta-analysis of GWASs, using a sample of 14,429 ‘Caucasian’ men. To be brief, they discovered two SNPs associated with testosterone by performing a GWAS of serum testosterone concentrations on 2 million SNPs on over 8,000 ‘Caucasians’. The strongest associated SNP discovered was rs12150660 was associated with low testosterone in this analysis, as well as in a study of Han Chinese, but it is rare along with rs5934505 being associated with an increased risk of low testosterone(Chen et al, 2016). Chen et al (2016) also caution that their results need replication (but I will show that it is meaningless due to how testosterone is produced in the body).
Ohlsson et al (2011) also found the same associations with the same two SNPs, along with rs6258 which affect how testosterone binds to SHGB. Ohlsson et al (2011) also validated their results: “To validate the independence of these two SNPs, conditional meta-analysis of the discovery cohorts including both rs12150660 and rs6258 in an additive genetic linear model adjusted for covariates was calculated.” Both SNPs were independently associated with low serum testosterone in men (less than 300ng/dl which is in the lower range of the new testosterone guidelines that just went into effect back in July). Men who had 3 or more of these SNPs were 6.5 times more likely to have lower testosterone.
Ohlsson et al (2011) conclude that they discovered genetic variants in the SHGB locus and X chromosome that significantly affect serum testosterone production in males (noting that it’s only on ‘Caucasians’ so this cannot be extrapolated to other races). It’s worth noting that, as can be seen, these SNPs are not really associated with variation in the normal range, but near the lower end of the normal range in which people would then need to seek medical help for a possible condition they may have.
In infant males, no SNPs were significantly associated with salivary testosterone levels, and the same was seen for infant females. Individual variation in salivary testosterone levels during mini-puberty (Kurtoglu and Bastug, 2014) were explained by environmental factors, not SNPs (Xia et al, 2014). They also replicated Carmaschi et al (2010) who also showed that environmental factors influence testosterone more than genetic factors in infancy. There is a direct correlation between salivary testosterone levels and free serum testosterone (Wang et al, 1981; Johnson, Joplin, and Burin, 1987), so free serum testosterone was indirectly tested.
This is interesting because, as I’ve noted here numerous times, testosterone is indirectly controlled by DNA, and it can be raised or lowered due to numerous environmental variables (Mazur and Booth, 1998; Mazur, 2016), such as marriage (Gray et al, 2002; Burnham et al, 2003; Gray, 2011; Pollet, Cobey, and van der Meij, 2013; Farrelly et al, 2015; Holmboe et al, 2017), having children (Gray et al, 2002; Gray et al, 2006; Gettler et al, 2011); to obesity (Palmer et al, 2012; Mazur et al, 2013; Fui, Dupuis, and Grossman, 2014; Jayaraman, Lent-Schochet, and Pike, 2014; Saxbe et al, 2017) smoking is not clearly related to testosterone (Zhao et al, 2016), and high-carb diets decrease testosterone (Silva, 2014). Though, most testosterone decline can be ameliorated with environmental interventions (Shi et al, 2013), it’s not a foregone conclusion that testosterone will sharply decrease around age 25-30.
Studies on ‘testosterone genes’ only show associations, not causes, genes don’t directly cause testosterone production, it is indirectly controlled by DNA, as I will explain below. These studies on the numerous environmental variables that decrease testosterone is proof enough of the huge effects of environment on testosterone production and synthesis.
How testosterone is produced in the body
There are five simple steps to testosterone production: 1) DNA codes for mRNA; 2) mRNA codes for the synthesis of an enzyme in the cytoplasm; 3) luteinizing hormone stimulates the production of another messenger in the cell when testosterone is needed; 4) this second messenger activates the enzyme; 5) the enzyme then converts cholesterol to testosterone (Leydig cells produce testosterone in the presence of luteinizing hormone) (Saladin, 2010: 137). Testosterone is a steroid and so there are no ‘genes for’ testosterone.
Cells in the testes enzymatically convert cholesterol into the steroid hormone testosterone. Quoting Saladin (2010: 137):
But to make it [testosterone], a cell of the testis takes in cholesterol and enzymatically converts it to testosterone. This can occur only if the genes for the enzymes are active. Yet a further implication of this is that genes may greatly affect such complex outcomes as behavior, since testosterone strongly influences such behaviors as aggression and sex drive. [RR: Most may know that I strongly disagree with the fact that testosterone *causes* aggression, see Archer, Graham-Kevan and Davies, 2005.] In short, DNA codes only for RNA and protein synthesis, yet it indirectly controls the synthesis of a much wider range of substances concerned with all aspects of anatomy, physiology, and behavior.
Genes only code for RNA and protein synthesis, and thusly, genes do not *cause* testosterone production. This is a misconception most people have; if it’s a human trait, then it must be controlled by genes, ultimately, not proximately as can be seen, and is already known in biology. Genes, on their own, are not causes but passive templates (Noble, 2008; Noble, 2011; Krimsky, 2013; Noble, 2013; Also read Exploring Genetic Causation in Biology). This is something that people need to understand; genes on their own do nothing until they are activated by the system.
What does this have to do with ‘IQ genes’?
My logic here is very simple: 1) Testosterone has the same heritability range as IQ. 2) One would assume—like is done with IQ—that since testosterone is a complex trait that it must be controlled by ‘many genes of small effect’. 3) Therefore, since I showed that there are no ‘genes for’ testosterone and only ‘associations’ (which could most probably be mediated by environmental interventions) with low testosterone, may the same hold true for ‘IQ genes/SNPS’? These testosterone SNPs I talked about from Ohlsson et al (2011) were associated with low testosterone. These ‘IQ SNP’ studies (Davies et al, 2017; Hill et al, 2017; Savage et al, 2017) are the same—except we have an actual idea of how testosterone is produced in the body, we know that DNA is indirectly controlling its production, and, most importantly, there is/are no ‘gene[s] for’ testosterone.
Testosterone has the same heritability range as IQ, is a complex trait like IQ, but, unlike how IQ is purported to be, it [testosterone] is not controlled by genes; only indirectly. My reasoning for using this example is simple: something has a moderate to high heritability, and so most would assume that ‘numerous genes of small effect’ would have an influence on testosterone production. This, as I have shown, is false. It’s also important to note that Ohlsson et al (2011) showed associated SNPs in regards to low testosterone—not testosterone levels in the normal range. Of course, only when physiological values are outside of the normal range will we notice any difference between men, and only then will we find—however small—genetic differences between men with normal and low levels of testosterone (I wouldn’t be surprised if lifestyle factors explained the lower testosterone, but we’ll never know that in regards to this study).
Testosterone production is a real, measurable physiologic process, as is the hormone itself; which is not unlike the so-called physiologic process that ‘g’ is supposed to be, which does not mimic any known physiologic process in the body, which is covered with unscientific metaphors like ‘power’ and ‘energy’ and so on. This example, in my opinion, is important for this debate. Sure, Ohlsson et al (2011) found a few SNPs associated with low testosterone. That’s besides the point. They are only associated with low testosterone; they do not cause low testosterone. So, I assert, these so-called associated SNPs do not cause differences in IQ test scores; just because they’re ‘associated’ doesn’t mean they ’cause’ the differences in the trait in question. (See Noble, 2008; Noble, 2011; Krimsky, 2013; Noble, 2013.) The testosterone analogy that I made here buttresses my point due to the similarities (it is a complex trait with high heritability) with IQ.