‚in the North (where people are taller) as “height increasing” and all of the alleles that are slightly more common in the South as “height decreasing”‘
Respectfully, I see several perhaps fundamental problems to this kind of research, although I admit the post is really good:
1.- There is no causal model, no theoretical model using the dependent variables, with few exceptions, exceptions in which causal models are probably incomplete enough, like Height, to explain relatively small differences in it across individuals and populations. We know a lot about the Biology of things like Growth-Hormones, and some diseases like short stature or truly outstanding large ones. And the causality in those cases is easier to find because its effects are really large: they can be seen just by looking at people having those diseases.
2.- As such, doing correlational research, empirical research, without a causal model, a scientific one, looks to me like a fishing trip, even if outwardly can be justified as trying to build those models from correlations, like inspirational sources for at some point have a Scientific Theory to go along expectations and experiments, instead of having one before even designing experiments. But unfortunately, so far, I think it is not only fair, but accurate to say those correlations seem to be all over the map: from the positive, to the negative, to the non-existent, and worse: and back to a previous category, apparently. Typical observations of publishable research fishing trips...
3.- I see caveats in the heritability interpretation of some of the results:
a) Until deep into the 20th Century most progeny, most descendants, died before reaching the tender age of 1-10yrs of age. At least half, my guess is probably larger, but there are prominent disagreers with better credentials than me, and there is the issue child cadavers tend not to be well preserved in a relatively adequate time frame to study child mortality across History, let alone preHistory, and with such information studying the Evolution of things like reproductive success will be lacking, so there is a source of uncertainty which will never go away. But such kind of points to me Evolution/Natural Selection probably until around 100yrs ago had little influence on Genomes to increase some of the hypothetical outcome variables, including reproductive success.
b) Comparisons of reproductive success with something else does have a Historical Comparison Problem, maybe even an anchoring problem: Until recently most Families had way more descendants than now. So correlating reproductive success even with genetics now, will suffer perhaps the same lack of explanatory and predictive power History many times has: things changed, and as such, a Scientific Theory cannot be made because it will lack the explanatory and predictive power to include Reproductive Success in the past. It would be at best an incomplete theory.
c) Brain analogies to explain academic achievement, socio-economic class belonging, and IQ, done through a bigger brain narrative, for example, do have a strong negative selection in the disparity between a baby´s head and a woman´s birth canal, and complications arising from other stuff like spontaneous abortions, premature deliveries, placenta related diseases, etc. And such has changed too: nowadays fatalities or disabilities as outcomes of pregnancy and delivery are measured in parts per thousands. My guess in the past they were more frequent, but off my memory I don´t have a number. And it probably has the same problems as b). Even infections during pregnancy causing product fatalities or neonate disabilities has changed a lot too. It seems a moving target so to speak to me.
Things I think makes the existence of a Scientific Theory, not Humanities Models more important, more needed before doing and publishing research like this one. Sociology is not a Science, Psychology and Psychiatry are not Sciences either, even if Encyclopedias and a lot of people say and worse, claim they are.
They are not, they are Humanities and as such, they are outside of Science and its Methods. They use Hermeneutics not Logic nor Science to create narratives with explanatory power in the Hermeneutical meaning of Explanation, not in a Scientific one.
I have a Post, a comedic one, explaining those Fundamental Requirements all Sciences must have, and Humanities do not have them all, protestations to the contrary or worse claims explicit or implicit like in not saying a word about, its irrelevance notwithstanding to a Definition:
Nice work, Sasha. I predict that the behavioral/cognitive ability/psychiatric GWAS/Polygenic score era will end as the linkage/candidate gene era ended: widely recognized failure based on decades of highly publicized false positive results. Do you have any thoughts on my prediction?
Slightly off-topic, but if ADHD isn't as genetic as previous reported: to what extent is ADHD malleable through therapeutic techniques? Or, is it a pattern of behavior that, once you have it, you can't change? There are environmentally caused traits like that. For example, the effect of height on starvation. I look toward environmental pollution for the origin of many modern diseases.
I always assumed that ADHD was a static "identity," and you just had to medicate for it, or you could manage it but never "cure it." If it's more environmental, it gives me hope that we could better understand its causes (behavioral, parental, pollution) to treat it better, or even prevent cases.
- The current medications for ADHD do not target any of the (relatively few) known genetic mechanisms associated with it. This is somewhat unusual, for example the first variants associated with Schizophrenia were near genes that were already being targeted by anti-psychotic meds. This suggests existing medications may indeed be going after environmental causes.
- There is some evidence suggesting nurturing effects on ADHD (parental neuroticism, alcohol use, IQ and education; see "genetic nurture" section here: https://www.nature.com/articles/s41380-022-01863-6). These "genetic nurture" estimates are still very sketchy (they can also just capture stratification) and the effect directions are odd but it is at least some evidence that family environment can also have an impact.
- As a general rule of thumb, I look to GLP1s as an example where even an "uncurable" condition can be rapidly addressed. Obesity is moderately heritable (and the genetics points to brain function more than fat storage), it also feels like an identity to people who have it (with eating becoming a compulsion), dieting mostly doesn't work, etc. yet you target the right receptor and people suddenly feel satiated for weeks on end, stop eating almost entirely, and lose weight rapidly. I'm optimistic we will see similar drugs for other behaviors.
Lately Ive been thinking about how under appreciated (not yet fully grasped?) how GLP1s turn the nature/nurture and free will conversations on their heads. The conventional wisdom has been, if 'its' not genetic/biological then we can change it through will and behavior modification. If something is genetic or biological, it cannot be changed (historically true).
CRISPR, GLP1s, and the failure of behavior modification (dieting, ABA in autism) may be showing us that 1) the opposite its true and 2) biologic treatments may be a better bet for behavioral issues
Agreed, and I'll add another interesting genetic example for behavior: rare variants in a nicotine receptor that, crudely speaking, make cigarettes highly unpleasant and reduce heavy smoking in carriers by 35% (https://www.nature.com/articles/s41588-023-01417-8). Because these are rare variants they explain essentially zero heritability, but pharma is almost certainly going to find a way to target this receptor and enable you take a pill or a shot to quit.
That’s very helpful! Let me make sure I’ve got this concept of “genetic nurture” down:
Let’s say that a parent has a genetic predilection for alcoholism, and they pass these genes down to their children. However, the child doesn’t drink alcohol at all (due to religion or personal choice), so these genes for alcoholism are never “activated.”
However, the alcoholic parent creates an unstable life environment for the child, which results in maladaptive coping strategies or personality disorders, like ADHD. In some sense, the ADHD was a “result” of the parent’s genes, but not because of “ADHD genes,” but because of alcoholism genes.
The follow-up question I’d have is if this misidentification is solved by studies of twins raised apart.
I agree that obesity should be thought of as a neurological eating disorder. It is exciting to think that we could have a drug which could cure addictions like alcoholism, which then, due to improved “genetic nurture,” would lead to lower rates of things like ADHD!
Yeah, that's pretty much right. A genetic influence from the parent to the child that is not having a direct effect in the child. Both twin studies and twins raised apart should in principle control for "genetic nurture" (which gets counted in the "family environment" component of each method) but they suffer from their own biases. Twins raised apart studies are additionally very small and typically have unusual recruitment (as is generally the case for adoption studies).
It’s a fair reaction to be surprised by low heritability in ADHD—but it’s also a reflection of how ADHD is diagnosed: it’s behaviourally defined via DSM criteria, not biologically. Even at the clinical level, ADHD is incredibly heterogeneous—meaning we’re lumping together people with very different underlying mechanisms under a single diagnostic label.
This creates a kind of nosological gap—one that affects nearly all of psychiatry and neurology. If the categories we’re using aren’t biologically coherent, it’s not surprising that there's no consistent signal. The noise is baked into the definitions.
And of course, none of this guarantees we’ll ever find biologically coherent subtypes of ADHD. The heterogeneity may be real—not just unrecognized biology, but reflecting the fact that what we call “ADHD” is often shaped by environmental, social, and contextual factors. The genetic signal may be weak because the diagnosis is largely environmental in origin.
I suppose my next question would be, which mental diagnoses are the easiest to identify with objective rather than highly subjective standards? Schizophrenic hallucination seems pretty clear cut. I’d really like to see a chart of diagnoses with some kind of estimate of their objectivity, with ADHD being highly subjective, and schizophrenic hallucination having less ambiguity.
I think one challenge here is the assumption that biology and environment are separable—or that diagnoses exist on a spectrum from “biologically real” to “behaviourally subjective.” The reality is much messier.
Psychiatric diagnoses like ADHD and schizophrenia are both constructed from behavioural patterns and subjective reports. While hallucinations may feel more concrete than inattention, schizophrenia is not diagnosed via blood test or imaging—like ADHD, it’s a syndrome, not a disease with a known biological origin.
The concept of “a gene-environment interaction” is misleading. There is no such thing as a gene that does something independently—it only does something in an environment. Every gene is part of a vast interaction web. The old nature/nurture dichotomy collapses under that view.
Zoom in far enough on any psychiatric condition and ambiguity abounds. This was the premise of the “endophenotype” movement in the 2000s—an attempt to biologically deconstruct psychiatric disorders into more precise components. It largely failed. Not because the goal was wrong, but because the categories themselves (e.g., schizophrenia, ADHD) may not map onto distinct biological entities. It’s akin to zooming in on a fractal: the deeper you go, the more complexity emerges. The hope was that increased biological resolution would reveal clean diagnostic lines—but instead, it revealed more overlap, more nuance, and more divergence from the categories we started with.
GLP-1s are exciting not because they reveal genetic truths, but because they work—broadly and non-specifically—on systems that shape behaviour. They’re not like Gleevec (arguably the first 'precision medicine'), which targets a specific mutation; they’re more like powerful dimmer switches for multiple brain-body circuits. Future drugs for ADHD or compulsive behaviours might work similarly—without us needing to fully decode the genetics.
And sometimes, biological clarity does arrive—but in ways that challenge our categories. I’m about to write a series (on my substack) on the discovery of anti-NMDA receptor encephalitis—a condition that can present almost identically to schizophrenia. In some cases, people with autoimmune antibodies to brain receptors may be hidden within psychiatric populations. These individuals can benefit not just from immunosuppressive therapy (e.g., steroids or IVIG), but also—perhaps surprisingly—from antipsychotic medication to manage symptoms. That doesn’t make the diagnosis “wrong”; it just shows how incomplete our current nosology is.
So rather than trying to rank psychiatric conditions by their “objectivity,” it might be more useful to ask: which diagnoses represent coherent biological syndromes, and which reflect clusters of overlapping symptoms that might arise from different causes in different people?
I'm still a layman when it comes to this science, so I appreciate your explanations. I suppose I am thinking of certain disorders like Down's Syndrome as "purely genetic," while if someone had amnesia due to a concussion, that would be "purely environmental." But even then, you could say that the risk of concussion is genetically mediated! And maybe there are environmental conditions which raise the risk of Down's Syndrome (parental age, pollution). But it does seem like there is a spectrum here, where nothing is 0 or 100, but still some variation.
So I’m not a geneticist, and I leave definitive interpretations of these data to my betters. That said, my background is in molecular biology, and I used to work on preclinical models of neurodevelopmental disorders—trying to understand how genetic and environmental risk factors shape early brain development.
Back then, we hoped that diverse risk factors—mutated genes, prenatal stress, infection, etc.—would converge on shared downstream pathways. The idea was that if these mechanisms overlapped, we might find common molecular targets. But even in genetically identical (isogenic) mouse models, that convergence was harder to demonstrate than we expected. And in humans—with all their variability—it only gets murkier.
One quick caveat about terminology: when molecular biologists talk about “environment,” we tend to mean something broader than what’s usually meant in everyday conversation. For us, environment includes everything outside the DNA sequence—like the cellular context, transcription factors, epigenetic state, hormone levels, even the microbiome—not just factors like parenting or culture. I’m not sure every geneticist uses the term this way, but it’s a helpful reminder that “environment” in biology can include both intracellular machinery and social context.
I think you’re absolutely right to imagine a spectrum, where some conditions (like Down syndrome) sit closer to the genetic end and others more clearly reflect environmental causes. What I find truly remarkable about Sasha’s post is that it throws this entire spectrum into question—our current tools and datasets simply cannot reliably parse what is “more genetic” versus “more environmental.” This is both shocking and thrilling because it means that much of our understanding needs significant revision.
It’s not just a niche technical fix—it’s a huge cross-field correction (as far as I understand). The data highlight the limits of reductionism in human traits and underscore how much our tools can reflect cultural biases under the guise of objectivity. So many non-genetic contributors are getting lumped into what we thought were genetic effects. The fact that even family-based GWAS, which remove much of the confounding, still produce unstable and wildly divergent polygenic scores depending on modeling choices means that whatever we currently believe as the map of “this is more genetic” versus “that is more environmental” is poised for significant rearrangement as our methods improve and our datasets grow (I think;)
Some of the analysis and modeling techniques are beyond my paygrade to speak confidently on, but what would be your summarization for the educated popular science reader? Some measurement techniques might be flawed for sure and need refinement, but doesn't the differing circumstances of certain physical environments ( Ice Age NorthEast Asia, tropical Africa) at least suggest that susceptibility to different conditions ( like say ADHD as a random example out of many possible) and traits would be different between populations? Few educated people would claim environment is not an influence on the development of personality, skills and other individual characteristics.
Two things I don't understand about family GWAS's that people here in the comments can probably set me straight on:
1. Isn't the reduced inter-familial effect size expected, because families regress to the "lineage mean" of various phenotypical traits versus the "population mean?"
2. Wouldn't we also expect reduced effect sizes because of range restriction and colliders? For example, in the population overall, Math and Verbal SAT scores are extremely correlated - but within a given college, which only admits a narrow band of SAT scores, they become much less correlated, because the total SAT score was the collider. Isn't the same thing going on here?
Those two factors each seem like a reduced effect size from family GWAS's would be entirely expected, but I assume I'm misunderstanding something about the methodology.
No, the family design specifically looks at the difference between the individual genetic value and their family mean, and is properly scaled. For many (non-behavioral) traits, the within-family heritability is identical to the between family heritability. You might expect reduction in within-family estimates if there is sibling confounding, but there is so far limited evidence of this.
What really leads to decades of sensationally false genetic claims is the ideology that drives the interpretation of statistical results. Statistics are too easy to interpret in ways favorable to your point of view when you lack skepticism.
If genetic differences create environmental differences, for instance higher IQ caused by genes leads to higher education than don't we run the risk of over controlling
Human biodiversity, that human groups have distinct advantages and disadvantages shaped by their adaptation and then natural selection in different environments, especially when separated by significant distances and time spans.
‚in the North (where people are taller) as “height increasing” and all of the alleles that are slightly more common in the South as “height decreasing”‘
Not in ex-Yugoslavia. There be giants here. 😂
Respectfully, I see several perhaps fundamental problems to this kind of research, although I admit the post is really good:
1.- There is no causal model, no theoretical model using the dependent variables, with few exceptions, exceptions in which causal models are probably incomplete enough, like Height, to explain relatively small differences in it across individuals and populations. We know a lot about the Biology of things like Growth-Hormones, and some diseases like short stature or truly outstanding large ones. And the causality in those cases is easier to find because its effects are really large: they can be seen just by looking at people having those diseases.
2.- As such, doing correlational research, empirical research, without a causal model, a scientific one, looks to me like a fishing trip, even if outwardly can be justified as trying to build those models from correlations, like inspirational sources for at some point have a Scientific Theory to go along expectations and experiments, instead of having one before even designing experiments. But unfortunately, so far, I think it is not only fair, but accurate to say those correlations seem to be all over the map: from the positive, to the negative, to the non-existent, and worse: and back to a previous category, apparently. Typical observations of publishable research fishing trips...
3.- I see caveats in the heritability interpretation of some of the results:
a) Until deep into the 20th Century most progeny, most descendants, died before reaching the tender age of 1-10yrs of age. At least half, my guess is probably larger, but there are prominent disagreers with better credentials than me, and there is the issue child cadavers tend not to be well preserved in a relatively adequate time frame to study child mortality across History, let alone preHistory, and with such information studying the Evolution of things like reproductive success will be lacking, so there is a source of uncertainty which will never go away. But such kind of points to me Evolution/Natural Selection probably until around 100yrs ago had little influence on Genomes to increase some of the hypothetical outcome variables, including reproductive success.
b) Comparisons of reproductive success with something else does have a Historical Comparison Problem, maybe even an anchoring problem: Until recently most Families had way more descendants than now. So correlating reproductive success even with genetics now, will suffer perhaps the same lack of explanatory and predictive power History many times has: things changed, and as such, a Scientific Theory cannot be made because it will lack the explanatory and predictive power to include Reproductive Success in the past. It would be at best an incomplete theory.
c) Brain analogies to explain academic achievement, socio-economic class belonging, and IQ, done through a bigger brain narrative, for example, do have a strong negative selection in the disparity between a baby´s head and a woman´s birth canal, and complications arising from other stuff like spontaneous abortions, premature deliveries, placenta related diseases, etc. And such has changed too: nowadays fatalities or disabilities as outcomes of pregnancy and delivery are measured in parts per thousands. My guess in the past they were more frequent, but off my memory I don´t have a number. And it probably has the same problems as b). Even infections during pregnancy causing product fatalities or neonate disabilities has changed a lot too. It seems a moving target so to speak to me.
Things I think makes the existence of a Scientific Theory, not Humanities Models more important, more needed before doing and publishing research like this one. Sociology is not a Science, Psychology and Psychiatry are not Sciences either, even if Encyclopedias and a lot of people say and worse, claim they are.
They are not, they are Humanities and as such, they are outside of Science and its Methods. They use Hermeneutics not Logic nor Science to create narratives with explanatory power in the Hermeneutical meaning of Explanation, not in a Scientific one.
I have a Post, a comedic one, explaining those Fundamental Requirements all Sciences must have, and Humanities do not have them all, protestations to the contrary or worse claims explicit or implicit like in not saying a word about, its irrelevance notwithstanding to a Definition:
https://federicosotodelalba.substack.com/p/sci-and-math-are-having-a-full-conversation?r=4up0lp
Outstanding work, as usual.
Nice work, Sasha. I predict that the behavioral/cognitive ability/psychiatric GWAS/Polygenic score era will end as the linkage/candidate gene era ended: widely recognized failure based on decades of highly publicized false positive results. Do you have any thoughts on my prediction?
Slightly off-topic, but if ADHD isn't as genetic as previous reported: to what extent is ADHD malleable through therapeutic techniques? Or, is it a pattern of behavior that, once you have it, you can't change? There are environmentally caused traits like that. For example, the effect of height on starvation. I look toward environmental pollution for the origin of many modern diseases.
I always assumed that ADHD was a static "identity," and you just had to medicate for it, or you could manage it but never "cure it." If it's more environmental, it gives me hope that we could better understand its causes (behavioral, parental, pollution) to treat it better, or even prevent cases.
I'm not a psychiatrist but a few points:
- The current medications for ADHD do not target any of the (relatively few) known genetic mechanisms associated with it. This is somewhat unusual, for example the first variants associated with Schizophrenia were near genes that were already being targeted by anti-psychotic meds. This suggests existing medications may indeed be going after environmental causes.
- There is some evidence suggesting nurturing effects on ADHD (parental neuroticism, alcohol use, IQ and education; see "genetic nurture" section here: https://www.nature.com/articles/s41380-022-01863-6). These "genetic nurture" estimates are still very sketchy (they can also just capture stratification) and the effect directions are odd but it is at least some evidence that family environment can also have an impact.
- As a general rule of thumb, I look to GLP1s as an example where even an "uncurable" condition can be rapidly addressed. Obesity is moderately heritable (and the genetics points to brain function more than fat storage), it also feels like an identity to people who have it (with eating becoming a compulsion), dieting mostly doesn't work, etc. yet you target the right receptor and people suddenly feel satiated for weeks on end, stop eating almost entirely, and lose weight rapidly. I'm optimistic we will see similar drugs for other behaviors.
Lately Ive been thinking about how under appreciated (not yet fully grasped?) how GLP1s turn the nature/nurture and free will conversations on their heads. The conventional wisdom has been, if 'its' not genetic/biological then we can change it through will and behavior modification. If something is genetic or biological, it cannot be changed (historically true).
CRISPR, GLP1s, and the failure of behavior modification (dieting, ABA in autism) may be showing us that 1) the opposite its true and 2) biologic treatments may be a better bet for behavioral issues
Agreed, and I'll add another interesting genetic example for behavior: rare variants in a nicotine receptor that, crudely speaking, make cigarettes highly unpleasant and reduce heavy smoking in carriers by 35% (https://www.nature.com/articles/s41588-023-01417-8). Because these are rare variants they explain essentially zero heritability, but pharma is almost certainly going to find a way to target this receptor and enable you take a pill or a shot to quit.
That’s very helpful! Let me make sure I’ve got this concept of “genetic nurture” down:
Let’s say that a parent has a genetic predilection for alcoholism, and they pass these genes down to their children. However, the child doesn’t drink alcohol at all (due to religion or personal choice), so these genes for alcoholism are never “activated.”
However, the alcoholic parent creates an unstable life environment for the child, which results in maladaptive coping strategies or personality disorders, like ADHD. In some sense, the ADHD was a “result” of the parent’s genes, but not because of “ADHD genes,” but because of alcoholism genes.
The follow-up question I’d have is if this misidentification is solved by studies of twins raised apart.
I agree that obesity should be thought of as a neurological eating disorder. It is exciting to think that we could have a drug which could cure addictions like alcoholism, which then, due to improved “genetic nurture,” would lead to lower rates of things like ADHD!
Yeah, that's pretty much right. A genetic influence from the parent to the child that is not having a direct effect in the child. Both twin studies and twins raised apart should in principle control for "genetic nurture" (which gets counted in the "family environment" component of each method) but they suffer from their own biases. Twins raised apart studies are additionally very small and typically have unusual recruitment (as is generally the case for adoption studies).
It’s a fair reaction to be surprised by low heritability in ADHD—but it’s also a reflection of how ADHD is diagnosed: it’s behaviourally defined via DSM criteria, not biologically. Even at the clinical level, ADHD is incredibly heterogeneous—meaning we’re lumping together people with very different underlying mechanisms under a single diagnostic label.
This creates a kind of nosological gap—one that affects nearly all of psychiatry and neurology. If the categories we’re using aren’t biologically coherent, it’s not surprising that there's no consistent signal. The noise is baked into the definitions.
And of course, none of this guarantees we’ll ever find biologically coherent subtypes of ADHD. The heterogeneity may be real—not just unrecognized biology, but reflecting the fact that what we call “ADHD” is often shaped by environmental, social, and contextual factors. The genetic signal may be weak because the diagnosis is largely environmental in origin.
I suppose my next question would be, which mental diagnoses are the easiest to identify with objective rather than highly subjective standards? Schizophrenic hallucination seems pretty clear cut. I’d really like to see a chart of diagnoses with some kind of estimate of their objectivity, with ADHD being highly subjective, and schizophrenic hallucination having less ambiguity.
I think one challenge here is the assumption that biology and environment are separable—or that diagnoses exist on a spectrum from “biologically real” to “behaviourally subjective.” The reality is much messier.
Psychiatric diagnoses like ADHD and schizophrenia are both constructed from behavioural patterns and subjective reports. While hallucinations may feel more concrete than inattention, schizophrenia is not diagnosed via blood test or imaging—like ADHD, it’s a syndrome, not a disease with a known biological origin.
The concept of “a gene-environment interaction” is misleading. There is no such thing as a gene that does something independently—it only does something in an environment. Every gene is part of a vast interaction web. The old nature/nurture dichotomy collapses under that view.
Zoom in far enough on any psychiatric condition and ambiguity abounds. This was the premise of the “endophenotype” movement in the 2000s—an attempt to biologically deconstruct psychiatric disorders into more precise components. It largely failed. Not because the goal was wrong, but because the categories themselves (e.g., schizophrenia, ADHD) may not map onto distinct biological entities. It’s akin to zooming in on a fractal: the deeper you go, the more complexity emerges. The hope was that increased biological resolution would reveal clean diagnostic lines—but instead, it revealed more overlap, more nuance, and more divergence from the categories we started with.
GLP-1s are exciting not because they reveal genetic truths, but because they work—broadly and non-specifically—on systems that shape behaviour. They’re not like Gleevec (arguably the first 'precision medicine'), which targets a specific mutation; they’re more like powerful dimmer switches for multiple brain-body circuits. Future drugs for ADHD or compulsive behaviours might work similarly—without us needing to fully decode the genetics.
And sometimes, biological clarity does arrive—but in ways that challenge our categories. I’m about to write a series (on my substack) on the discovery of anti-NMDA receptor encephalitis—a condition that can present almost identically to schizophrenia. In some cases, people with autoimmune antibodies to brain receptors may be hidden within psychiatric populations. These individuals can benefit not just from immunosuppressive therapy (e.g., steroids or IVIG), but also—perhaps surprisingly—from antipsychotic medication to manage symptoms. That doesn’t make the diagnosis “wrong”; it just shows how incomplete our current nosology is.
So rather than trying to rank psychiatric conditions by their “objectivity,” it might be more useful to ask: which diagnoses represent coherent biological syndromes, and which reflect clusters of overlapping symptoms that might arise from different causes in different people?
I'm still a layman when it comes to this science, so I appreciate your explanations. I suppose I am thinking of certain disorders like Down's Syndrome as "purely genetic," while if someone had amnesia due to a concussion, that would be "purely environmental." But even then, you could say that the risk of concussion is genetically mediated! And maybe there are environmental conditions which raise the risk of Down's Syndrome (parental age, pollution). But it does seem like there is a spectrum here, where nothing is 0 or 100, but still some variation.
So I’m not a geneticist, and I leave definitive interpretations of these data to my betters. That said, my background is in molecular biology, and I used to work on preclinical models of neurodevelopmental disorders—trying to understand how genetic and environmental risk factors shape early brain development.
Back then, we hoped that diverse risk factors—mutated genes, prenatal stress, infection, etc.—would converge on shared downstream pathways. The idea was that if these mechanisms overlapped, we might find common molecular targets. But even in genetically identical (isogenic) mouse models, that convergence was harder to demonstrate than we expected. And in humans—with all their variability—it only gets murkier.
One quick caveat about terminology: when molecular biologists talk about “environment,” we tend to mean something broader than what’s usually meant in everyday conversation. For us, environment includes everything outside the DNA sequence—like the cellular context, transcription factors, epigenetic state, hormone levels, even the microbiome—not just factors like parenting or culture. I’m not sure every geneticist uses the term this way, but it’s a helpful reminder that “environment” in biology can include both intracellular machinery and social context.
I think you’re absolutely right to imagine a spectrum, where some conditions (like Down syndrome) sit closer to the genetic end and others more clearly reflect environmental causes. What I find truly remarkable about Sasha’s post is that it throws this entire spectrum into question—our current tools and datasets simply cannot reliably parse what is “more genetic” versus “more environmental.” This is both shocking and thrilling because it means that much of our understanding needs significant revision.
It’s not just a niche technical fix—it’s a huge cross-field correction (as far as I understand). The data highlight the limits of reductionism in human traits and underscore how much our tools can reflect cultural biases under the guise of objectivity. So many non-genetic contributors are getting lumped into what we thought were genetic effects. The fact that even family-based GWAS, which remove much of the confounding, still produce unstable and wildly divergent polygenic scores depending on modeling choices means that whatever we currently believe as the map of “this is more genetic” versus “that is more environmental” is poised for significant rearrangement as our methods improve and our datasets grow (I think;)
Some of the analysis and modeling techniques are beyond my paygrade to speak confidently on, but what would be your summarization for the educated popular science reader? Some measurement techniques might be flawed for sure and need refinement, but doesn't the differing circumstances of certain physical environments ( Ice Age NorthEast Asia, tropical Africa) at least suggest that susceptibility to different conditions ( like say ADHD as a random example out of many possible) and traits would be different between populations? Few educated people would claim environment is not an influence on the development of personality, skills and other individual characteristics.
Two things I don't understand about family GWAS's that people here in the comments can probably set me straight on:
1. Isn't the reduced inter-familial effect size expected, because families regress to the "lineage mean" of various phenotypical traits versus the "population mean?"
2. Wouldn't we also expect reduced effect sizes because of range restriction and colliders? For example, in the population overall, Math and Verbal SAT scores are extremely correlated - but within a given college, which only admits a narrow band of SAT scores, they become much less correlated, because the total SAT score was the collider. Isn't the same thing going on here?
Those two factors each seem like a reduced effect size from family GWAS's would be entirely expected, but I assume I'm misunderstanding something about the methodology.
No, the family design specifically looks at the difference between the individual genetic value and their family mean, and is properly scaled. For many (non-behavioral) traits, the within-family heritability is identical to the between family heritability. You might expect reduction in within-family estimates if there is sibling confounding, but there is so far limited evidence of this.
What really leads to decades of sensationally false genetic claims is the ideology that drives the interpretation of statistical results. Statistics are too easy to interpret in ways favorable to your point of view when you lack skepticism.
If genetic differences create environmental differences, for instance higher IQ caused by genes leads to higher education than don't we run the risk of over controlling
The advantage of within-family studies is they isolate a specific kind of causal effect without needing to explicitly control for any variables.
No. And I am blocked by the entire HBD crew on twitter. Interpret that how you will.
what does HBD mean?? who are “the HBD-inclined crowd”?
Human biodiversity, that human groups have distinct advantages and disadvantages shaped by their adaptation and then natural selection in different environments, especially when separated by significant distances and time spans.