This is impressive work! I'm very pleased with the contributions you've made to the behavioral knowledge on this one.
My notes:
"It’s a myth that C^2 = 0 in adults; Cesarini found that scores of genetically unrelated siblings on the Swedish military entrance exam, taken at the age of about 18, correlate at .17 (n = 1,647 individuals). We also have not found any trait with a shared environmental effect of zero that has a heritability of below 100%, so our priors for this being the one are, well, low."
Thanks to my time researching this matter, I'm immediately suspicious of shared environment values significantly different from zero for most traits other than education or wealth. You state that you control for age, but how old are we talking? The Wilson effect appears to continue well into adulthood and I'm curious how the C effect behaves with age with many of these.
"Twin models give a much higher heritability estimation than a pedigree model that excludes twins reared together:
The twin model is probably wrong here, unless either epistasis or dominance is a large factor in determining how overweight people are. Personally, I suspect the problem is that the EEA assumption is violated"
Why can't non-additive heredity be involved with BMI here? It does appear that non-additive heredity is important on the population level (and this may be behind the low heritabilities found by molecular genetic studies). Although numerically it does look like you have a fit to a violated EEA.
With income, your method doesn't allow for teasing out the sources that could inflate C (assortative mating, the Wilson effect), so it's unclear if a C significantly different from zero is real.
"Most of the samples took the Wilson-Patterson scale of conservatism, which has a reliability of .94, or they self-reported their political views on a scale."
For political views, the fraction self-reporting their political views may be a significant factor. A "conservative" in New York means something different from a "conservative" in Texas. Any way to disaggregate this from the data?
"It’s clear that there are social transmission effects here; the political views of parents and their adoptive children are almost as highly correlated as those of biological parents and their children."
There is a cohort effect on political views. I suspect that your point about environmental confounding could very well be at play. It's suspicious that adoptive parent-child correlations are so much higher than adoptive sib correlations.
>Why can't non-additive heredity be involved with BMI here? It does appear that non-additive heredity is important on the population level (and this may be behind the low heritabilities found by molecular genetic studies). Although numerically it does look like you have a fit to a violated EEA.
I think I might have been too hasty to dismiss dominance, and I should probably have fit an ACDE model to the non-twin data, but I find the idea that BMI has a heritability of above 70% to be very hard to believe, especially given the time/location variance that exists.
>With income, your method doesn't allow for teasing out the sources that could inflate C (assortative mating, the Wilson effect), so it's unclear if a C significantly different from zero is real.
In my defense, it was surprisingly difficult to collect simple correlations between income in family members because all economists have autism, but I would find it hard to believe that there is not a shared-environmental effect for income given that income varies by location and that it is affected by education.
>There is a cohort effect on political views. I suspect that your point about environmental confounding could very well be at play. It's suspicious that adoptive parent-child correlations are so much higher than adoptive sib correlations.
I think this is just sampling. True adopted parent ~ child correlation is probably lower, true adoptee correlation is probably higher.
Correlations between adopted parents and children are also very tricky, as they are in r form, not R^2 form. So if there is an r of 0.3 between adopted parents and children, then the social transmission effect will actually be something like ~15% (it's lower than 2x(0.3)^2 because of assortative mating) instead of 30%.
I also suspect that the heritability of political views is higher in the current generation because of the internet, but I have no evidence for this, as twin studies on zoomers/late milennials on political views aren't out yet.
>For political views, the fraction self-reporting their political views may be a significant factor. A "conservative" in New York means something different from a "conservative" in Texas. Any way to disaggregate this from the data?
To be honest, no idea. But I think the use of self-reported beliefs helps to an extent. IIRC, self-id of political views and conservatism scales didn't have different heritabilities.
Also wrt C^2>0 in intelligence, the rank order of g within individuals is very stable after people reach 18, so it's unlikely that C^2 changes substantially after that point. I assume that C^2 is much closer within the elderly, however.
"I find the idea that BMI has a heritability of above 70% to be very hard to believe, especially given the time/location variance that exists."
I don't, given twin results and that higher quality data -> higher heritability.
"but I would find it hard to believe that there is not a shared-environmental effect for income given that income varies by location and that it is affected by education."
I don't, especially if the shared environment effect results from environmental confounding or is "empty" and impacts only the *signaling* component of educational attainment.
"self-id of political views and conservatism scales didn't have different heritabilities."
But they may have different shared environment components, if this reflects above noted contrast effects. Is it possible to tease this out?
>But they may have different shared environment components
I doubt it.
>I don't, especially if the shared environment effect results from environmental confounding or is "empty"
I thought the point of estimating the shared environment was that the component is "environmental confounding" -- environmental influences have to act at a time and a place, and siblings reared together are more likely to share one than two random people in the world. It doesn't have to be the actual house that is doing the work.
"I thought the point of estimating the shared environment was that the component is "environmental confounding" -- environmental influences have to act at a time and a place,"
What I mean is that educational attainment is *signaling*. It doesn't mean much in and of itself, rather it serves to signal the actual traits of interest (to employers, etc) about the individual, i.e., intelligence, work ethic, etc. It is those things that lead to adult income, particularly in the long run. The shared environment impact on educational attainment is in good part "real", but only impacts the signal aspect of education. The "marketing" of individual. That the shared environment of income is nonexistent–or at best, a good deal smaller than the one on education speaks to this.
The heritability of political beliefs is unintuitive. They’re clearly not “genetic” (Nick Land’s genes don’t contain a copy of Xenosystems), but they are heritable because there are heritable traits (e.g. height, intelligence, personality) cause people to adopt different political beliefs.
It makes sense. I could imagine even small aspects of personality such as disgust sensitivity influencing political views. I think I must have confused this 10% heritability with something else.
Anyways, a decently high heritability of political views means that the claims that the culture is going to become more rightwing due to the widening birth rate difference between the left and the right are not absurd.
To say that there is no underestimation of heritability for traits like weight, as you have stated, is terribly wrong. Imagine measuring a person's weight ten times consecutively every five years and calculating an average weight lifetime. You would get significant differences when comparing a single measurement with the measurement of the same person five years later. Similarly, if two identical twins had the exact same weight, calculated as the average of the ten longitudinal measurements taken every five years, you would get differences as large as the same person tested twice in studies that take single measurements at a single time point, as almost all studies do. Since the differences between identical twins are no greater than those between the same person years apart for almost everything, we can conclude that for weight, political views, income, and so on, the heritability is close to 100%, if calculated based on the average of dozens of individual longitudinal measurements taken across the adult lifespan, rather than just at a single point in time at a specific age.
No one has understood this simple truism. Human stupidity is incredible, even if only one person had understood it. Most likely, you haven't understood anything i wrote, or you won't give it the immense importance it deserves.
Your estimates of heritability for political views and income, and just about everything else, are obscenely underestimated. No twin studies have looked at lifetime income heritability from the studies I've read. You yourself have reported that heritability for twenty years of income is greater than heritability for ten years of income as a time period of income, which in turn is greater than heritability over a shorter period, and so on. If this trend of increasing heritability as a percentage of income time examined continues, and it surely does because random nonshared environmental influences cancel out over a longer time period as people experiment with more jobs, then heritability approaches 80-90% in twin studies, with zero shared environmental impact controlling for assortative mating. And studies of adoptees and twins raised apart, as well as studies of twins raised together, confirm that shared environment has near zero or zero impact on political views. Studies of adoptees report a higher impact but that is because they are based on adolescents, in fact they fully support a model in which shared environmental estimates are close to zero in adults and moderate by adolescents, as indicated by twin studies considering assortative mating. And again, differences tend to be for phenotypes like the same person tested twice years apart, so if every five years for the entire adult life you tested all the twins and calculated the heritability based on the lifetime average obtained for all the various longitudinal measures conducted the heritability would tend to be 90-100%
Correlations between adoptive siblings tend to indicate a much smaller shared environmental impact than the correlation with adoptive parents, but adoptive siblings share much more shared environment with adoptive siblings than with adoptive parents: with adoptive parents they share the exact same socioeconomic status, upbringing, cultural background, and so on, while with adoptive parents they share much less shared environment. Since the correlation between adoptive siblings fully includes shared upbringing, as well as many more shared environmental factors than the correlation with adoptive parents, it is far preferable to estimate the importance of the broad shared environment using their correlation. Sandra Scarr concluded that fascist authoritarianism is not transmitted environmentally at all, in a study of adoptees. Correlations between adoptive siblings in one study of adult adoptees I read some time ago are about 0.10, which compared to the correlation of biological siblings of 0.3 indicates a fourfold greater heritability of shared environment, in a simple and plausible additive model. All the studies of adolescent adoptees I have read demonstrate indirectly that shared environment matters zero in adults, because they provide estimates perfectly consistent with the shared environment estimates from twin studies corrected for assortative mating, which indicate zero or near zero impact in adults but moderate and significant in adolescents.
Contrary to the constant falsehood that genetic changes happen very slowly, large genetic changes have occurred in the last century due to a large decline in homozygosity due to increased outbreeding and heterosis due to increased social mobility, and this explains the huge changes that have occurred, such as increased education, income, average IQ, height, and the reduction of extreme traits such as extreme racism, patriarchy, etc. Homozygosity increases phenotypic extremes, therefore increasing religious fanaticism, domestic violence, etc.
Your heritability estimates are grossly underestimated, as usual, by confounding factors such as assortative mating, misdiagnosis and missed diagnoses, measurement error, and transient environmental factors. In addition to the fact that many of the people sampled are children and there is a shared environmental effect in them, even if usually small. But not in adults. The fact that differences between identical twins years apart tend to be essentially the same as the same person tested twice and the prevalence of misdiagnosis and missed diagnoses and assortative mating show that true, stable heritability for phenotypes tends to be almost universally close to 100%, with very few exceptions, and shared environment is about zero or nearly zero in adults. Income and political views are no exceptions. If there were stable nonshared environmental influences, correlation differences between identical twins would be larger than test-retest differences a few years apart, but they are essentially not. Also, always keep in mind that shared environment is almost exclusively transitory, or irregular in its effect, and has an effect limited to certain periods of life, so any small impact would drop to zero in a well-designed longitudinal study that measures stable variance and correlation longitudinally.
>In addition to the fact that many of the people sampled are children and there is a shared environmental effect in them, even if usually small. But not in adults
Addressed this in the IQ section. Largest sample of UT adults has a positive effect.
Environmental factors must act at a time and a place. Siblings are more likely to share times and places, even as adults. Therefore, C^2 is probably >0 in most traits with a heritability below 100%.
>transient environmental factors.
Still an environmental factor
>misdiagnosis and missed diagnoses, measurement error
For many of these traits (e.g. BMI, height) this is a non-issue (see studies on reliability of height/BMI self-reports in surveys, it's almost always > .9). I adjusted for rxx formally in the sections on intelligence and political orientation.
>Assortative mating
Doesn't bias heritability the same way in extended family models as it does in twin models. Please learn math.
Why did you block me? Are you retarded? Why is everyone in behavioral genetics so anti-talk and loves to block for no reason? Why are you so obscenely stupid, even though you seem promising? You told me to learn math for no reason. I simply said that assortative mating leads to an underestimation of the shared environment, and therefore is a source of distortion in your data. In addition to selective placement, as you rightly acknowledged.
Extended twin studies accounting for assortative mating consistently report near zero shared environmental impact and zero vertical cultural transmission. Often they report a small impact, but it is the shared fraternal environment. But apparently no researcher is smart enough to realize that the assumption that biological siblings are related like parent and child is completely wrong, since biological siblings also share population genetic structure and mixtures of specific populations and subpopulations, so any observed phenotypic similarity between biological siblings that is greater than between parent and child is explained by genetics, not shared fraternal environment.
I'd like to do a twin study on the GFP derived from Big Five data (derived from either factor scores or item-level), testing heritability and dominance (Rushton found some evidence of dominance). What dataset do you recommend using?
I've seen twin studies under attack. I used to think that twin studies were uncontroversially the gold standard of heritability research. Would you point to any article in particular as a counter-argument to these criticisms?
Here's a summary of why genetic (heritable, chromsomal) factors in fluid intelligence and poverty tend to be over-estimated in MZ/DZ studies:
1. "Equal Environment Assumption" is often violated
2. Genes and environment can interact or correlate in complex ways. For example, a child’s genetic traits may shape their environment (evocative genetic influence), or children may seek out environments that fit their genetic predispositions (active genetic influence). These dynamics mean that MZ twins may experience even more similar environments than DZ twins, further biasing heritability estimates upward.
3. Neglect of Prenatal and Early-Life Environmental Factors
Twin studies often do not adequately account for critical prenatal and early-life factors—such as maternal nutrition, exposure to toxins (lead, mycotoxins, indoor smoke), maternal anaemia, and malnutrition—that can have lasting impacts on cognitive development and health. These factors can create similarities or differences between twins that are wrongly attributed to genetics if not properly controlled.
For example, MZ twins share a placenta more often than DZ twins, which can lead to more similar prenatal environments for MZ pairs, further confounding the distinction between genetic and environmental effects.
4. Sampling and Generalizability Issues
Twins, especially those who participate in studies, may not be representative of the general population, potentially limiting the generalizability of findings. Selection biases (e.g., volunteerism, socioeconomic status) can further skew results.
5. Assumptions About Genetic Sharing
While MZ twins are assumed to share 100% of their genes, post-zygotic mutations can introduce genetic differences even between identical twins, leading to overestimation of heritability.
6. Assortative Mating
If parents are more similar to each other than random (assortative mating), DZ twins may share more than 50% of relevant genes, which can lead to underestimation of shared environmental effects and overestimation of genetic effects.
7. Inability to Identify Specific Causes
Twin studies can estimate the proportion of variance due to genetic and environmental factors but cannot identify which specific genes or environmental exposures are responsible for observed differences.
Don't want to be that guy, but [citation needed]. It's probably invalid for BMI but MZT correlations are usually easy to explain with genes + twin-specific effects + shared environments.
>Genes and environment can interact
Interactions are less common than main effects, so apriori interactionism should not be reason to shun hereditarianism.
Representative samples of twins (e.g. Scottish IQ twin study) find the same results as other studies, and sampling can theoretically restrict both genetic and environmental variance; it's presence doesn't invalidate heritability estimates.
>If parents are more similar to each other than random (assortative mating), DZ twins may share more than 50% of relevant genes, which can lead to underestimation of shared environmental effects and overestimation of genetic effects.
The exact opposite is true. Do the maths.
>While MZ twins are assumed to share 100% of their genes, post-zygotic mutations can introduce genetic differences even between identical twins, leading to overestimation of heritability.
If the effect is shared between siblings/twins, then it's allocated to the shared environmental component. Otherwise, it's allocated to the unshared environmental component.
Great analysis!
This is impressive work! I'm very pleased with the contributions you've made to the behavioral knowledge on this one.
My notes:
"It’s a myth that C^2 = 0 in adults; Cesarini found that scores of genetically unrelated siblings on the Swedish military entrance exam, taken at the age of about 18, correlate at .17 (n = 1,647 individuals). We also have not found any trait with a shared environmental effect of zero that has a heritability of below 100%, so our priors for this being the one are, well, low."
Thanks to my time researching this matter, I'm immediately suspicious of shared environment values significantly different from zero for most traits other than education or wealth. You state that you control for age, but how old are we talking? The Wilson effect appears to continue well into adulthood and I'm curious how the C effect behaves with age with many of these.
"Twin models give a much higher heritability estimation than a pedigree model that excludes twins reared together:
The twin model is probably wrong here, unless either epistasis or dominance is a large factor in determining how overweight people are. Personally, I suspect the problem is that the EEA assumption is violated"
Why can't non-additive heredity be involved with BMI here? It does appear that non-additive heredity is important on the population level (and this may be behind the low heritabilities found by molecular genetic studies). Although numerically it does look like you have a fit to a violated EEA.
With income, your method doesn't allow for teasing out the sources that could inflate C (assortative mating, the Wilson effect), so it's unclear if a C significantly different from zero is real.
"Most of the samples took the Wilson-Patterson scale of conservatism, which has a reliability of .94, or they self-reported their political views on a scale."
For political views, the fraction self-reporting their political views may be a significant factor. A "conservative" in New York means something different from a "conservative" in Texas. Any way to disaggregate this from the data?
"It’s clear that there are social transmission effects here; the political views of parents and their adoptive children are almost as highly correlated as those of biological parents and their children."
There is a cohort effect on political views. I suspect that your point about environmental confounding could very well be at play. It's suspicious that adoptive parent-child correlations are so much higher than adoptive sib correlations.
Once again, great work!
>Why can't non-additive heredity be involved with BMI here? It does appear that non-additive heredity is important on the population level (and this may be behind the low heritabilities found by molecular genetic studies). Although numerically it does look like you have a fit to a violated EEA.
I think I might have been too hasty to dismiss dominance, and I should probably have fit an ACDE model to the non-twin data, but I find the idea that BMI has a heritability of above 70% to be very hard to believe, especially given the time/location variance that exists.
>With income, your method doesn't allow for teasing out the sources that could inflate C (assortative mating, the Wilson effect), so it's unclear if a C significantly different from zero is real.
In my defense, it was surprisingly difficult to collect simple correlations between income in family members because all economists have autism, but I would find it hard to believe that there is not a shared-environmental effect for income given that income varies by location and that it is affected by education.
>There is a cohort effect on political views. I suspect that your point about environmental confounding could very well be at play. It's suspicious that adoptive parent-child correlations are so much higher than adoptive sib correlations.
I think this is just sampling. True adopted parent ~ child correlation is probably lower, true adoptee correlation is probably higher.
Correlations between adopted parents and children are also very tricky, as they are in r form, not R^2 form. So if there is an r of 0.3 between adopted parents and children, then the social transmission effect will actually be something like ~15% (it's lower than 2x(0.3)^2 because of assortative mating) instead of 30%.
I also suspect that the heritability of political views is higher in the current generation because of the internet, but I have no evidence for this, as twin studies on zoomers/late milennials on political views aren't out yet.
>For political views, the fraction self-reporting their political views may be a significant factor. A "conservative" in New York means something different from a "conservative" in Texas. Any way to disaggregate this from the data?
To be honest, no idea. But I think the use of self-reported beliefs helps to an extent. IIRC, self-id of political views and conservatism scales didn't have different heritabilities.
Also wrt C^2>0 in intelligence, the rank order of g within individuals is very stable after people reach 18, so it's unlikely that C^2 changes substantially after that point. I assume that C^2 is much closer within the elderly, however.
https://www.sciencedirect.com/science/article/abs/pii/S0160289607000025
https://www.researchgate.net/publication/281377031_Interindividual_differences_in_general_cognitive_ability_from_age_18_to_age_65_years_are_extremely_stable_and_strongly_associated_with_working_memory_capacity
"I find the idea that BMI has a heritability of above 70% to be very hard to believe, especially given the time/location variance that exists."
I don't, given twin results and that higher quality data -> higher heritability.
"but I would find it hard to believe that there is not a shared-environmental effect for income given that income varies by location and that it is affected by education."
I don't, especially if the shared environment effect results from environmental confounding or is "empty" and impacts only the *signaling* component of educational attainment.
"self-id of political views and conservatism scales didn't have different heritabilities."
But they may have different shared environment components, if this reflects above noted contrast effects. Is it possible to tease this out?
>But they may have different shared environment components
I doubt it.
>I don't, especially if the shared environment effect results from environmental confounding or is "empty"
I thought the point of estimating the shared environment was that the component is "environmental confounding" -- environmental influences have to act at a time and a place, and siblings reared together are more likely to share one than two random people in the world. It doesn't have to be the actual house that is doing the work.
"I thought the point of estimating the shared environment was that the component is "environmental confounding" -- environmental influences have to act at a time and a place,"
What I mean is that educational attainment is *signaling*. It doesn't mean much in and of itself, rather it serves to signal the actual traits of interest (to employers, etc) about the individual, i.e., intelligence, work ethic, etc. It is those things that lead to adult income, particularly in the long run. The shared environment impact on educational attainment is in good part "real", but only impacts the signal aspect of education. The "marketing" of individual. That the shared environment of income is nonexistent–or at best, a good deal smaller than the one on education speaks to this.
The heritability of political views one was surprising. I vaguely remember studies saying it was only like 10%
The heritability of political beliefs is unintuitive. They’re clearly not “genetic” (Nick Land’s genes don’t contain a copy of Xenosystems), but they are heritable because there are heritable traits (e.g. height, intelligence, personality) cause people to adopt different political beliefs.
It makes sense. I could imagine even small aspects of personality such as disgust sensitivity influencing political views. I think I must have confused this 10% heritability with something else.
Anyways, a decently high heritability of political views means that the claims that the culture is going to become more rightwing due to the widening birth rate difference between the left and the right are not absurd.
Could you please do the same for mental disorders?
To say that there is no underestimation of heritability for traits like weight, as you have stated, is terribly wrong. Imagine measuring a person's weight ten times consecutively every five years and calculating an average weight lifetime. You would get significant differences when comparing a single measurement with the measurement of the same person five years later. Similarly, if two identical twins had the exact same weight, calculated as the average of the ten longitudinal measurements taken every five years, you would get differences as large as the same person tested twice in studies that take single measurements at a single time point, as almost all studies do. Since the differences between identical twins are no greater than those between the same person years apart for almost everything, we can conclude that for weight, political views, income, and so on, the heritability is close to 100%, if calculated based on the average of dozens of individual longitudinal measurements taken across the adult lifespan, rather than just at a single point in time at a specific age.
No one has understood this simple truism. Human stupidity is incredible, even if only one person had understood it. Most likely, you haven't understood anything i wrote, or you won't give it the immense importance it deserves.
Your estimates of heritability for political views and income, and just about everything else, are obscenely underestimated. No twin studies have looked at lifetime income heritability from the studies I've read. You yourself have reported that heritability for twenty years of income is greater than heritability for ten years of income as a time period of income, which in turn is greater than heritability over a shorter period, and so on. If this trend of increasing heritability as a percentage of income time examined continues, and it surely does because random nonshared environmental influences cancel out over a longer time period as people experiment with more jobs, then heritability approaches 80-90% in twin studies, with zero shared environmental impact controlling for assortative mating. And studies of adoptees and twins raised apart, as well as studies of twins raised together, confirm that shared environment has near zero or zero impact on political views. Studies of adoptees report a higher impact but that is because they are based on adolescents, in fact they fully support a model in which shared environmental estimates are close to zero in adults and moderate by adolescents, as indicated by twin studies considering assortative mating. And again, differences tend to be for phenotypes like the same person tested twice years apart, so if every five years for the entire adult life you tested all the twins and calculated the heritability based on the lifetime average obtained for all the various longitudinal measures conducted the heritability would tend to be 90-100%
Correlations between adoptive siblings tend to indicate a much smaller shared environmental impact than the correlation with adoptive parents, but adoptive siblings share much more shared environment with adoptive siblings than with adoptive parents: with adoptive parents they share the exact same socioeconomic status, upbringing, cultural background, and so on, while with adoptive parents they share much less shared environment. Since the correlation between adoptive siblings fully includes shared upbringing, as well as many more shared environmental factors than the correlation with adoptive parents, it is far preferable to estimate the importance of the broad shared environment using their correlation. Sandra Scarr concluded that fascist authoritarianism is not transmitted environmentally at all, in a study of adoptees. Correlations between adoptive siblings in one study of adult adoptees I read some time ago are about 0.10, which compared to the correlation of biological siblings of 0.3 indicates a fourfold greater heritability of shared environment, in a simple and plausible additive model. All the studies of adolescent adoptees I have read demonstrate indirectly that shared environment matters zero in adults, because they provide estimates perfectly consistent with the shared environment estimates from twin studies corrected for assortative mating, which indicate zero or near zero impact in adults but moderate and significant in adolescents.
If you can refute me, I'm listening.
Contrary to the constant falsehood that genetic changes happen very slowly, large genetic changes have occurred in the last century due to a large decline in homozygosity due to increased outbreeding and heterosis due to increased social mobility, and this explains the huge changes that have occurred, such as increased education, income, average IQ, height, and the reduction of extreme traits such as extreme racism, patriarchy, etc. Homozygosity increases phenotypic extremes, therefore increasing religious fanaticism, domestic violence, etc.
Your heritability estimates are grossly underestimated, as usual, by confounding factors such as assortative mating, misdiagnosis and missed diagnoses, measurement error, and transient environmental factors. In addition to the fact that many of the people sampled are children and there is a shared environmental effect in them, even if usually small. But not in adults. The fact that differences between identical twins years apart tend to be essentially the same as the same person tested twice and the prevalence of misdiagnosis and missed diagnoses and assortative mating show that true, stable heritability for phenotypes tends to be almost universally close to 100%, with very few exceptions, and shared environment is about zero or nearly zero in adults. Income and political views are no exceptions. If there were stable nonshared environmental influences, correlation differences between identical twins would be larger than test-retest differences a few years apart, but they are essentially not. Also, always keep in mind that shared environment is almost exclusively transitory, or irregular in its effect, and has an effect limited to certain periods of life, so any small impact would drop to zero in a well-designed longitudinal study that measures stable variance and correlation longitudinally.
>In addition to the fact that many of the people sampled are children and there is a shared environmental effect in them, even if usually small. But not in adults
Addressed this in the IQ section. Largest sample of UT adults has a positive effect.
Environmental factors must act at a time and a place. Siblings are more likely to share times and places, even as adults. Therefore, C^2 is probably >0 in most traits with a heritability below 100%.
>transient environmental factors.
Still an environmental factor
>misdiagnosis and missed diagnoses, measurement error
For many of these traits (e.g. BMI, height) this is a non-issue (see studies on reliability of height/BMI self-reports in surveys, it's almost always > .9). I adjusted for rxx formally in the sections on intelligence and political orientation.
>Assortative mating
Doesn't bias heritability the same way in extended family models as it does in twin models. Please learn math.
>Every study of adult adoptees I've read has a zero correlation for IQ
Not Cesarini, D. Essays on Genetic Variation and Economic Behavior. (Massachusetts Institute of Technology, 2010).
Why did you block me? Are you retarded? Why is everyone in behavioral genetics so anti-talk and loves to block for no reason? Why are you so obscenely stupid, even though you seem promising? You told me to learn math for no reason. I simply said that assortative mating leads to an underestimation of the shared environment, and therefore is a source of distortion in your data. In addition to selective placement, as you rightly acknowledged.
Extended twin studies accounting for assortative mating consistently report near zero shared environmental impact and zero vertical cultural transmission. Often they report a small impact, but it is the shared fraternal environment. But apparently no researcher is smart enough to realize that the assumption that biological siblings are related like parent and child is completely wrong, since biological siblings also share population genetic structure and mixtures of specific populations and subpopulations, so any observed phenotypic similarity between biological siblings that is greater than between parent and child is explained by genetics, not shared fraternal environment.
I'd like to do a twin study on the GFP derived from Big Five data (derived from either factor scores or item-level), testing heritability and dominance (Rushton found some evidence of dominance). What dataset do you recommend using?
The SATSA is public but the personality measurements are not big 5.
The Minnesota twin study has personality measurements, but it’s not public. Bouchard responds to emails, last I checked.
https://sci-hub.se/10.1126/science.2218526
thanks!
I've seen twin studies under attack. I used to think that twin studies were uncontroversially the gold standard of heritability research. Would you point to any article in particular as a counter-argument to these criticisms?
I'm writing it.
Here's a summary of why genetic (heritable, chromsomal) factors in fluid intelligence and poverty tend to be over-estimated in MZ/DZ studies:
1. "Equal Environment Assumption" is often violated
2. Genes and environment can interact or correlate in complex ways. For example, a child’s genetic traits may shape their environment (evocative genetic influence), or children may seek out environments that fit their genetic predispositions (active genetic influence). These dynamics mean that MZ twins may experience even more similar environments than DZ twins, further biasing heritability estimates upward.
3. Neglect of Prenatal and Early-Life Environmental Factors
Twin studies often do not adequately account for critical prenatal and early-life factors—such as maternal nutrition, exposure to toxins (lead, mycotoxins, indoor smoke), maternal anaemia, and malnutrition—that can have lasting impacts on cognitive development and health. These factors can create similarities or differences between twins that are wrongly attributed to genetics if not properly controlled.
For example, MZ twins share a placenta more often than DZ twins, which can lead to more similar prenatal environments for MZ pairs, further confounding the distinction between genetic and environmental effects.
4. Sampling and Generalizability Issues
Twins, especially those who participate in studies, may not be representative of the general population, potentially limiting the generalizability of findings. Selection biases (e.g., volunteerism, socioeconomic status) can further skew results.
5. Assumptions About Genetic Sharing
While MZ twins are assumed to share 100% of their genes, post-zygotic mutations can introduce genetic differences even between identical twins, leading to overestimation of heritability.
6. Assortative Mating
If parents are more similar to each other than random (assortative mating), DZ twins may share more than 50% of relevant genes, which can lead to underestimation of shared environmental effects and overestimation of genetic effects.
7. Inability to Identify Specific Causes
Twin studies can estimate the proportion of variance due to genetic and environmental factors but cannot identify which specific genes or environmental exposures are responsible for observed differences.
Sources:
https://www.perplexity.ai/search/challenge-the-conclusions-of-t-dvxJbQPzTriNV38T7hjxcg
>"Equal Environment Assumption" is often violated
Don't want to be that guy, but [citation needed]. It's probably invalid for BMI but MZT correlations are usually easy to explain with genes + twin-specific effects + shared environments.
>Genes and environment can interact
Interactions are less common than main effects, so apriori interactionism should not be reason to shun hereditarianism.
https://rpubs.com/Jonatan/interactions
>Sampling and Generalizability Issues
Representative samples of twins (e.g. Scottish IQ twin study) find the same results as other studies, and sampling can theoretically restrict both genetic and environmental variance; it's presence doesn't invalidate heritability estimates.
>If parents are more similar to each other than random (assortative mating), DZ twins may share more than 50% of relevant genes, which can lead to underestimation of shared environmental effects and overestimation of genetic effects.
The exact opposite is true. Do the maths.
>While MZ twins are assumed to share 100% of their genes, post-zygotic mutations can introduce genetic differences even between identical twins, leading to overestimation of heritability.
The exact opposite is true, again. Maths please.
>Inability to Identify Specific Causes
Doesn't matter.
For most traits, EEA is not "often violated":
https://menghu.substack.com/p/sometimes-biased-but-not-systematically
Regarding assortative mating, you have it completely backwards, it's the other way around. It leads to *under*estimates of heritability.
There may be some interactions and correlations, but not very relevant to real-world implications.
Does this take into account epigenetic and congenital factors (like folate, smoke, mycotoxins, anemia, fever in early pregnancy)?
>Epigenetic
Isn't epigenetic variance technically genetic?
>Congenital factors
If the effect is shared between siblings/twins, then it's allocated to the shared environmental component. Otherwise, it's allocated to the unshared environmental component.
Amazing