> Depression is just bad. I strongly recommend not having it. Don’t even have any risk genes, if you can avoid it. All of you people trying to come up with clever evolutionary benefits for depression, I still think you’re wrong, and so does the genetic correlation with cognitive and non-cognitive aspects of educational attainment.
I am not a doctor or biologist, but the fact that rates of depression are heavily correlated with ethnic background (at least that's the stereotype; I haven't validated this), I would expect that depression is genetic in nature. And if it is genetic, I would assume that there is some kind of compensatory benefit to keep it around.
I don't know what the benefit is, and it seems pretty all-around shitty to me. But why would evolution specifically give you depression genes? There has to be some evolutionary benefit
Title on the graph seems wrong, "EA FDR correction tries to impute the results from the main timeline, where Roosevelt was an effective altruist and diverted the resources of the Depression-era US into curing all diseases."
I feel like the evolutionary arguments for intellect seem to be getting weaker with each post. I think it might have mattered for a given time for humans in terms of surviving, but... then things which seem to be evolutionary vestigial at best seem to be vaguely correlated with things we denote as success, it does seem a bit... conflated with a lot of concepts. Like, for example, we have the well known IQ/Fertility inverse. We also seem to be unable to generalize g meaningfully across a variety of species [1]. Even if there have been studies regarding things like other primates and rats, to suggest g could be found in dolphins, elephants, birds, or the proud nematode seems... difficult to ask, unless one necessarily adopts a human centric model of cognition. There is also a somewhat iffy question when it comes to talking about forms of human organization over the years and the role of intellect, and how important was in various societies. The original psychometricians hypothesized humanity started being dysgenic in the developed world prior to their first studies, which... well, suggests a relatively low bound for humans today where IQ would give one an *evolutionarily*, not necessarily a societal benefit. ADHD and Depression, the two you mentioned as being effectively evolutionarily irrelevant, have some of the highest frequencies within the human population. Especially with a lot of the research today, I feel like intelligence could be possibly described not as an evolutionary benefit, but a human society beneficial "quirk".
"As of last time I checked, the leading hypothesis was that schizophrenia genes were just really bad, evolutionary detritus that we hadn’t quite managed to weed out."
Is it really? I thought the idea that schizophrenia genes/low levels of schizotypy were somewhat positively associated with creativity (when they don't result into full-blown schizophrenia, which does decrease creativity) was rather widespread. I wouldn't necessarily have expected creativity to correlate with educational attainment, but it's clearly a positive trait that is easy to think would be selected for in many environments - not quite what you'd expect from "evolutionary detritus". Eg https://onlinelibrary.wiley.com/doi/abs/10.1111/j.1601-5223.1970.tb02343.x or https://www.nature.com/articles/nn.4040 but a Google Scholar search would return much more
I'm not familiar with this type of research at all. When one says "found genes for intelligence" or "found genes for educational attainment", what does that mean? Is the claim that some portion or portions of the human genome have been identified that, when they look like 'x', 'y', or 'z', make a person unintelligent, of average intelligence, or very intelligent, respectively?
"Both cognitive and non-cognitive skills are correlated with neuroticism, which I guess makes sense - that’s probably what makes you do your homework on time."
They're both negatively correlated, though, right? So this also means that NonCog has correlation in the "good" direction for each of the Big Five, which makes sense.
Also, I'm trying to think of how to recreate this analysis, but without looking at genes. I guess it would mean that for a given educational attainment (say, college grads), people in the bottom 10% of IQ are more likely to be schizophrenic than the average college grad? Or at least have more relatives with schizophrenia?
I'd like to see some exploration of the ADHD–IQ link. To me it seems plausible that ADHD genes might affect intelligence test scores without actually affecting underlying intelligence. This could be investigated by looking at the results of the subtests: if ADHD genes have an equal effect on subtests that require attention (such as digit span) as on subtests that do not (such as vocabulary), that would be evidence against my hypothesis.
I have a kid with ADHD-Inattentive and bipolar. His ultimate level of educational attainment probably will not be high, although he is pretty bright (good writer, bad at math, horrible at history). His docs have told me that many ADHD kids are delayed in terms of maturity, which sometimes affects decision making (not him, thank god) and sometimes affects initiative, self-discipline, and attention to deadlines (definitely him). So, he may in time manage college, but it won't be soon.
By the way, there is nothing good about having bipolar disorder. What's good is the medication that keeps my son out of the hospital or worse.
Robert Sapolsky mentions in this lecture that a *little* bit of controllable oddness on the schizophrenia spectrum is useful for making usefully charismatic shamans, and that could explain its genetic advantage and persistence. The more extreme, uncontrollable end of the spectrum, on the other hand, isn't an advantage.
(Also, this is the one lecture from this series that Sapolsky has removed from his own YouTube channel. He says some highly cancelable shit in this one, which is a shame because it sure does seem to make sense to me.)
> Depression is just bad. I strongly recommend not having it. Don’t even have any risk genes, if you can avoid it.
Couldn’t depression genes be like sickle cell anemia? Eg one depression gene, say, makes you more introspective, two makes you so introspective you fixate on everything bad?
I’m struggling to understand the chart in ways that translate to common English. Is it fair to say that this chart suggests:
1) Noncognitive genes are more correlated with Academic Achievement than cognitive genes in the presence of high Big 5 Personality stats?
2) Schizophrenia and Bipolar disorder correlates to more academic success in people with more NonCog genes? So Cog genes + Schizophrenia = no correlation to academic success?
3) Noncognitive genes and academic success is correlated with significantly higher longevity as compared to Cognitive genes and academic success?
Depression is being selected FOR in Sweden. While women with clinical depression average fewer children, their almost depressed sisters have increased fertility, which outweighs the cost to the clinical cases. Thus, evolutionarily speaking, the optimal point for female depression (not male) is higher than the current rate. That is, selection is increasing the sex difference in depression.
Of course, this says nothing of why this might be the case. Considering that women in every place and country are more depressive than men, this pattern must have originated a long time ago. Would be interesting to study mental illness, insofar as it exists (!?), in primates. Are female primates more moody and depressive than males?
Ever since I learned that Albert Einstein’s son Eduard had schizophrenia to a level that required him to be institutionalized, I sort of assumed that schizophrenia must have something to do with genius.
Also because—take it with the usual corrections for self-reporting—I consider myself to be a genius. My family tree on my father’s side is full of schizotypal personalities. I strongly suspect that being Mennonite, chased to the coldest edges of Europe and beyond because of one’s deeply held esoteric beliefs, selects for this trait.
I remember a post from you about this, Scott (found it: reviewing Surfing Uncertainty https://slatestarcodex.com/2017/09/05/book-review-surfing-uncertainty/). You talked about schizophrenia as a hyperaffective reaction disorder to predictive modelling errors, as opposed to autism which produces hypersensitivity in the models themselves. To put it another way, schizophrenia, in the predictive-modelling paradigm, is a disease that makes it difficult to ignore surprises.
If someone lives their lives constantly having their models challenged, and having their attention pulled to every little model-error, it creates a lot of pressure to build better models, right? If your models get better, you get fewer errors and your mental life becomes easier to manage. And those are the models that, if you happen to be able to communicate them well, become PhD material.
On the other hand, if your models don’t get better, either because you were unlucky, not cognitively flexible enough, or the inundation of model-errors was just too much, the result is something that looks more like schizophrenia.
It makes sense to me that cranking up the genetic dial labeled “schizophrenia” increases the risk-reward of cognition: either you will produce surprising insights and breakthroughs that greatly simplify your (and maybe others’) predictive models, or you will struggle to cope with the basic demands of life and lurch spasmodically from breakdown to breakdown. Or both.
The hypothesis, then, is that Albert Einstein had the right mix of schizophrenia and cognition-enhancing genes to use his irritation at model-errors effectively; Eduard got the irritation but not enough of the cognitive mechanisms to handle it.
"I have been saying for years that I think some of the genes for some mental illnesses must have compensatory benefits. Everyone else said that was dumb, they’re mostly selected against and decrease IQ."
One obvious answer to the "if a gene is bad, it can't be selected for" argument is that a gene can be good in a heterozygous genotype and bad in a homozygous genotype. Sickle Cell being the classic example. One copy of the gene gives malaria resistance; two copies gives you a deadly blood disease.
I don't know much about exactly how these GWAS studies are done. But when they calculate the raw correlations between genes and phenotypes are they able to separate out the heterozygous and homozygous occurrences of the genes?
It seems pretty important. For example, in theory, a gene's hugely positive heterozygous effects might exactly cancel out its hugely negative homozygous effects. So that it's raw statistical correlation to the trait is essentially zero. (This is akin to the man with one foot in boiling water and one foot in freezing water, who is "on average" experiencing a comfortable temperature).
I feel like describing this as "intelligence" vs educational attainment may be misleading, since its actually measuring "IQ test scores" vs educational attainment as I understand it. Not to get into the whole general IQ debate, but to the extent that IQ is useful as a large scale proxy for intelligence, that becomes less meaningful when you are comparing it to another proxy for intelligence. Given the nature of IQ tests that makes some of the correlations less surprising.
Re: Math Phds and autism, I am assuming it was written as a joke, but when I read it I was initially like “sounds about right” and then I thought about all the math grad students and professors I know and realized that none of them seem actually autistic. There is definitely something atypical about most math people I know, but it seems varried and generally not autism.
Do people know stats on this? It might also be because I mainly see academics, and teaching feels like something that would select against autism, so maybe there are more autistic math PhDs out in industry?
There is a theory that the evolutionary reason for suicide is that when a person kills themselves there are more resources, (like food) for their relatives. So depression might exist to cause suicide.
I have the theory that this is right except that the gene for suicide is not in the suicidal person but in the mother of the suicidal person. That it is not a gene for being suicidal, but a gene for making your child suicidal. The child is affected when it in the mother's womb.
The benefit of my theory from the older theory is that the evolutionary drawback of the suicide is halved, as the person committing suicide only has half the genes of the mother. But the evolutionary benefit of the suicide stays the same.
Very nice post. A note on the schizophrenia result and false discovery. In figure 4 they are graphing 95% confidence intervals (plus or minus 2 standard deviations) and if you look at the confidence intervals for schizophrenia they are very far apart. If you tripled the size of the confidence intervals (plus or minus 6 standard deviations) they still would not overlap. Assuming everything is normally distributed (which they have already done, their confidence intervals rely on this) we can get a nonoptimal bound that if we run a billion trials, we expect to see a false positive in less than 2 out of a billion tries. (This number should not be taken literally because the approximation of the data by the normal distribution will not have this level of accuracy, and there are other possible errors in the study. However it accurately expresses that seeing an effect of this size simply due to a false positive from running many comparisons is exceedingly unlikely.) In the paper they also list a P_{diff_fdr}<.001 which I think is supposed to be the P value of the schizophrenia result after taking into account the risk of false discovery. I include the simple and loose analysis above simply to demonstrate that even if I am misinterpreting their P value, from the visual of the confidence intervals alone, false discovery seems very unlikely here.
There is a theory that schizophrenia is a side effect of human self domestication. Which of course means you'd need to buy into the self domestication hypothesis first but it would explain how we got schizophrenia.
"... whether you so desperately seek societal approval that you're willing to throw away your entire twenties on a PhD with no job prospects at the end of it."
If functional mental disorders are primarily due to evolutionary mismatch, then there is no need to explain the benefits of "mental illness genes." While some people are genetically more prone to adult onset diabetes than are others, in the environment of evolutionary adaptation such diabetes was rare. We don't try to explain the benefits of diabetes genes. Likewise, while some people are more genetically prone to functional mental disorders than are others, in the environment of evolutionary adaptation it seems likely that such disorders were likewise rare.
Durkheim's research on the higher rates of suicide among Protestants than Catholics was the beginning of a long tradition of empirical research relating to the phenomenon of anomie, as opposed to cultural embeddedness with strong social roles and bonds, leading to adverse mental health outcomes. Here is a recent study (n = 8446) on how increased exposure to US culture increased suicide attempts among youth in the Dominican Republic,
"The increases in the propensity to attempt suicide for DR youths across these US cultural involvement indicators were both robust and large. For example, the propensity to attempt suicide ranged from 6.3% for those at the lowest end of the range of use of US electronic media and language to 13.3% for those at the highest end of the range of use of US electronic media and language. This central finding is congruent with the lower suicide or suicide attempt rates found for first-generation or less acculturated Latinos across multiple national and regional cohorts of Latinos."
Liah Greenfeld's "Mind, Modernity, and Madness" provides a neo-Durkheimian account that provides a coherent explanation for how increasing levels of anomie in modernity lead to increased rates of depression, bipolar, and schizophrenia. In traditional cultures, with humanity in the environment of evolutionary adaptation being the most "traditional" societies, there was neither need nor opportunity to construct a personal identity. A human being was one's roles. There was no "I" in the modern sense (cf. Julian Jaynes). One was unaware of the water in which one was swimming. Now we are all fish out of water, flopping around, gills desperately sucking in air in an attempt to maintain mental stability. For some of us it is easy, for others very difficult. The genetic material of fish works just fine in the water.
While Scott is not sympathetic to this explanation, I've never seen compelling evidence that shows that functional mental disorders were not due to evolutionary mismatch. Yes, whether or not functional mental orders have increased over time and across cultures or not remains contested. Depending on one's priors, the burden of proof shifts.
But studies of suicide provide less contested evidence that culture is a major influence on suicide rates. As far as I know, all such comparative studies are consistent with greater anomie, greater burden on constructing one's own identity (as opposed to a relative lack of the need to create an identity in more traditional cultures) resulting in higher rates of suicide.
Human beings evolved over many millions of years in diverse physical environments. But with respect to social structure, until the dawn of agriculture and empire, almost all adolescents:
1. Lived in a small tribal community of a few dozen to a few hundred with few interactions with other tribal groups.
2. These tribes would have shared one language, one belief system, one set of norms, one morality, and more generally a social and cultural homogeneity that is unimaginable for us today.
3. They would have been immersed in a community with a full range of ages present, from child to elder.
4. From childhood they would have been engaged in the work of the community, typically hunting and gathering, with full adult responsibilities typically being associated with puberty.
5. Their mating and status competitions would have mostly been within their tribe or occasionally with nearby groups, most of which would have been highly similar to themselves.
Could the dramatic divergence from the environment of evolutionary adaptation in any or all of these socio-cultural features result in increased "mental illness" for a genetic subset of human populations?
"Their genetic measure of non-cognitive skills... was still correlated at r = 0.31 with IQ"
Note that this is also a *genetic* correlation - the genetic influences of NonCog correlate at r = 0.31 with the genetic influences of IQ, not with actual measured IQ. The same is true for the Cog/NonCog relationship you mention with self-reported math ability and highest math class taken. (Also, assortative mating would inflate these correlations.)
Delay discounting is worse than alcoholism? I interpret this not as meaning that delay discounting is bad for success, but that high educational attainment is bad for success. Huge delay discounting = not going to college because college takes a long time (not obviously wrong); no delay discounting = being willing to go to college for 12 years to get a marginally nicer job (obviously wrong).
The way they use correlation assumes that all personality traits have linear effects on whatever they're measuring. If the function relating {number of genes "for" a behavior} to {educational attainment or IQ} has a tall U-shaped (or upside-down-U-shaped) curve, as some might, the correlation results will depend mainly on the outliers. For example, if a little conscientiousness helps you finish college, but a whole lot makes you such a perfectionist that you're likely to fail college, then the "correlation" between conscientiousness and finishing college isn't telling you how strong the effect of conscientiousness on finishing college is; it's telling you something about the skew of the U-shaped function.
For some reason, when people talk about genetics they like to forget that correlation is not causation.
Scott, you are usually vigilant for distrusting "correlation but we adjusted for confounders" studies. This is the same type of study!
What are "genes for intelligence"? In this context, they are genes that are correlated with intelligence. A gene that purely causes black skin (but does literally nothing else) would be counted as if it decreased IQ, because statistically people with black skin do worse on IQ tests.
There is no end to the number of possible confounders here; there is no exogenous source of randomness at all.
In general: I've often thought that a lot of people go to college and get phds just because they are familiar and comfortable with school and scared or unsure about entering the 'real world' (certainly true for me). I've often thought that a lot of the apparently high levels of mental weirdness in phd programs are largely related to this - who wants to effectively extend their childhood by staying in school, vs who is ready to 'grow up' and enter the real world.
I have some experience living with people diagnosed as schizophrenics. I put it that way because I'm not sure how well understood schizophrenia is and how confident we can be in a clinical diagnosis.
Other than living with diagnosed schizophrenics, I know very little about the condition. My experience is that schizophrenics live in a fantasy world. They are unable to tell the difference between the real world of experience and the world inside their head. I know one in particular who can talk at length of his day to day life in Vietnam, when in fact he has never been there. He appears to be unable to distinguish between fantasy and reality.
On the other hand, my impression of highly successful scientists, engineers, mathematicians, linguists, etc... is they can build mental maps that enable them to navigate their special practices. It seems to me they comprehend very complicated systems, such as molecular biology or microelectronics using some type of mapping onto real world experiences.
What I mean to say is what schizophrenics and mathematicians, e.g., have in common is the ability to live in an alternate reality. I would say visualizing the complex folding of a protein molecule is not too far removed from imagining that you are in fact Sgt. Barry Sadler in 1969.
It's amazing how dominant UK Biobank is in genetics right now (at least behavioural genetics). I was at the IGSS conference (https://cupc.colorado.edu/conferences/IGSS_2021/) and more than half the presentations were using that data.
Overall, I'd be worried about confounds here. We have a very noisy measure of "genes for IQ" - noisy because GWAS is noisy, and because the IQ measure itself is noisy (just a quick 12 point test IIRC). Then we deduct that from "genes for educational attainment". What's left? Maybe "genes for non-cognitive skills". But maybe "genes for IQ, that we didn't measure very well". Indeed, "non-cognitive PGS" predicts IQ.... And then there are all the possible environmental confounds. I think I'd rather see a measure of non-cognitive skills and then a GWAS that targets that directly.
However, that is just a lazy first take, and I should stop shooting my mouth off about a coauthor's paper.
There's a lot of inference to worry about here, but I'm already stuck on this, from the paper: "By construction, NonCog genetic variance was independent of Cog genetic variance (r_g = 0)." What sense of "independence" follows from zero covariance? That's pretty clearly going to create some weird conditional relationships between their SNPs' imputed Cog and NonCog scores to maintain zero correlation. It seems they recognize this in the supplement but they don't really bother to interpret it.
I ask this because I follow Charles Murray on twitter in order to argue with his position that there is a race-iq causation - has anyone dared to see if there's a variance of the cognitive genes vs ethnicity?
Seems like this database would provide quite strong empirical evidence on the unfortunate topic
Am I reading the chart correctly to note that SNPs associated with higher cognition are *negatively correlated* with conscientiousness (and extraversion, and agreeableness)? That's absolutely fascinating given that educational attainment (and a whole bunch of other traits associated with success, like wealth and income) are so strongly associated with conscientiousness, and suggests at least to me that most of the difference between educational attainment-promoting and cognition-promoting genes should logically have something to do with the trait.
It also seems like a fascinating counterexample to the idea you see brought up often in population genetics that "all good things are correlated". We see a discussion of a g factor, which includes all of cognition, often, and it seems frequently mooted that g itself is part of h, a general health factor. (Also, while I would certainly associate high-cognition-but-low-educational-attainment individuals with low conscientiousness, it boggles the mind to suggest that low conscientiousness, by itself, is associated with higher cognition. Why should that be true? Where's the tradeoff?)
Assuming I am reading the graph correctly, and blue is just 'high cognition', and not something like 'SNPs promoting high cognition but not educational attainment'. I could go back and try to read the chart more carefully, or read the article the chart comes from, but I think I'm too low conscientiousness to go and do so. One read-through is enough.
I've never been tested for ADHD and I don't know if I'll ever bother as my country has a dismal record of treating it, but I strongly suspect I have it. In any case I am extremely hyperactive and have been for a long time.
My life was basically total chaos up to a point. I finished university with a very mediocre result (around 2.4 I think). I aced courses that were interesting to me, and I failed Statistics 101 TWO times until I got angry the third time and got a B+.
Then I was lucky enough to find a technique that works for me - a sort of a Zen meditation where you just sit without moving for a long time. It doesn't remove my ADHD symptoms, but it calms me down a lot and allows me to focus and work.
I am a programmer now and I think there are some advantages, but only because I am treating it in some way - otherwise it would just be a total burden.
The advantages:
- I am better than others in scanning a large amount of code in a search for some obscure problem. In more general sense, I am just better in scanning in general - if I have some list in front of me and I need to find something, I do it faster than others.
- When something gets interesting to me, I feel like my mind gets totally obsessed and get actually angry if someone distracts me - I have shouted at people for that. This might sound bad, but allows me to do fast analysis of a large amount of data.
Disadvantages:
- I have a problem when I need to slow down and focus on one thing. I just cannot motivate myself to study, I have tried for years. The only way I learn new information is when I actively write code, because it is interesting and not boring.
- If I start procrastinating even a little it is very possible that my whole day will be spent in Reddit and YouTube - basically when my brain gets interested in some bullshit many hours can pass before I can stop myself.
- I work well when my work is interesting and badly when it is not. For example, it is interesting to write some new feature, less interesting to make a strategy to test it, so that might take much more time than it has to.
Hypothesis: schizophrenics have a better connection to the supernatural, which is actually real. They're the most misunderstood mentally ill people, except maybe cluster B folks. I'm sorry to admit this hypothesis is more than a little bit inspired by stereotypes about chosen people.
There's a history of schizophrenia in my family. There's also a history of genius, creativity and outrageous financial success. The money is nice. But I miss my brother.
There is now genetic and epidemiological evidence that Autism and mental illness genes overlap (especially with ADHD, parkinsons). But this is not being communicated with the public. How long has this been covered up? I can't believe that in almost 100 years of autism research this has not been noticed or investigated until now.
> Depression is just bad. I strongly recommend not having it. Don’t even have any risk genes, if you can avoid it. All of you people trying to come up with clever evolutionary benefits for depression, I still think you’re wrong, and so does the genetic correlation with cognitive and non-cognitive aspects of educational attainment.
I am not a doctor or biologist, but the fact that rates of depression are heavily correlated with ethnic background (at least that's the stereotype; I haven't validated this), I would expect that depression is genetic in nature. And if it is genetic, I would assume that there is some kind of compensatory benefit to keep it around.
I don't know what the benefit is, and it seems pretty all-around shitty to me. But why would evolution specifically give you depression genes? There has to be some evolutionary benefit
Title on the graph seems wrong, "EA FDR correction tries to impute the results from the main timeline, where Roosevelt was an effective altruist and diverted the resources of the Depression-era US into curing all diseases."
I feel like the evolutionary arguments for intellect seem to be getting weaker with each post. I think it might have mattered for a given time for humans in terms of surviving, but... then things which seem to be evolutionary vestigial at best seem to be vaguely correlated with things we denote as success, it does seem a bit... conflated with a lot of concepts. Like, for example, we have the well known IQ/Fertility inverse. We also seem to be unable to generalize g meaningfully across a variety of species [1]. Even if there have been studies regarding things like other primates and rats, to suggest g could be found in dolphins, elephants, birds, or the proud nematode seems... difficult to ask, unless one necessarily adopts a human centric model of cognition. There is also a somewhat iffy question when it comes to talking about forms of human organization over the years and the role of intellect, and how important was in various societies. The original psychometricians hypothesized humanity started being dysgenic in the developed world prior to their first studies, which... well, suggests a relatively low bound for humans today where IQ would give one an *evolutionarily*, not necessarily a societal benefit. ADHD and Depression, the two you mentioned as being effectively evolutionarily irrelevant, have some of the highest frequencies within the human population. Especially with a lot of the research today, I feel like intelligence could be possibly described not as an evolutionary benefit, but a human society beneficial "quirk".
[1]: How General is Cognitive Abillity in Non-Human Animals? A meta-analytical and multi-level reanalysis approach. Porier, Marc-Antonine, et. al. https://royalsocietypublishing.org/doi/10.1098/rspb.2020.1853
"As of last time I checked, the leading hypothesis was that schizophrenia genes were just really bad, evolutionary detritus that we hadn’t quite managed to weed out."
Is it really? I thought the idea that schizophrenia genes/low levels of schizotypy were somewhat positively associated with creativity (when they don't result into full-blown schizophrenia, which does decrease creativity) was rather widespread. I wouldn't necessarily have expected creativity to correlate with educational attainment, but it's clearly a positive trait that is easy to think would be selected for in many environments - not quite what you'd expect from "evolutionary detritus". Eg https://onlinelibrary.wiley.com/doi/abs/10.1111/j.1601-5223.1970.tb02343.x or https://www.nature.com/articles/nn.4040 but a Google Scholar search would return much more
I'm not familiar with this type of research at all. When one says "found genes for intelligence" or "found genes for educational attainment", what does that mean? Is the claim that some portion or portions of the human genome have been identified that, when they look like 'x', 'y', or 'z', make a person unintelligent, of average intelligence, or very intelligent, respectively?
"Both cognitive and non-cognitive skills are correlated with neuroticism, which I guess makes sense - that’s probably what makes you do your homework on time."
They're both negatively correlated, though, right? So this also means that NonCog has correlation in the "good" direction for each of the Big Five, which makes sense.
Also, I'm trying to think of how to recreate this analysis, but without looking at genes. I guess it would mean that for a given educational attainment (say, college grads), people in the bottom 10% of IQ are more likely to be schizophrenic than the average college grad? Or at least have more relatives with schizophrenia?
I'd like to see some exploration of the ADHD–IQ link. To me it seems plausible that ADHD genes might affect intelligence test scores without actually affecting underlying intelligence. This could be investigated by looking at the results of the subtests: if ADHD genes have an equal effect on subtests that require attention (such as digit span) as on subtests that do not (such as vocabulary), that would be evidence against my hypothesis.
I have a kid with ADHD-Inattentive and bipolar. His ultimate level of educational attainment probably will not be high, although he is pretty bright (good writer, bad at math, horrible at history). His docs have told me that many ADHD kids are delayed in terms of maturity, which sometimes affects decision making (not him, thank god) and sometimes affects initiative, self-discipline, and attention to deadlines (definitely him). So, he may in time manage college, but it won't be soon.
By the way, there is nothing good about having bipolar disorder. What's good is the medication that keeps my son out of the hospital or worse.
Robert Sapolsky mentions in this lecture that a *little* bit of controllable oddness on the schizophrenia spectrum is useful for making usefully charismatic shamans, and that could explain its genetic advantage and persistence. The more extreme, uncontrollable end of the spectrum, on the other hand, isn't an advantage.
https://youtu.be/4WwAQqWUkpI
(Also, this is the one lecture from this series that Sapolsky has removed from his own YouTube channel. He says some highly cancelable shit in this one, which is a shame because it sure does seem to make sense to me.)
> Depression is just bad. I strongly recommend not having it. Don’t even have any risk genes, if you can avoid it.
Couldn’t depression genes be like sickle cell anemia? Eg one depression gene, say, makes you more introspective, two makes you so introspective you fixate on everything bad?
I’m struggling to understand the chart in ways that translate to common English. Is it fair to say that this chart suggests:
1) Noncognitive genes are more correlated with Academic Achievement than cognitive genes in the presence of high Big 5 Personality stats?
2) Schizophrenia and Bipolar disorder correlates to more academic success in people with more NonCog genes? So Cog genes + Schizophrenia = no correlation to academic success?
3) Noncognitive genes and academic success is correlated with significantly higher longevity as compared to Cognitive genes and academic success?
Depression is being selected FOR in Sweden. While women with clinical depression average fewer children, their almost depressed sisters have increased fertility, which outweighs the cost to the clinical cases. Thus, evolutionarily speaking, the optimal point for female depression (not male) is higher than the current rate. That is, selection is increasing the sex difference in depression.
Of course, this says nothing of why this might be the case. Considering that women in every place and country are more depressive than men, this pattern must have originated a long time ago. Would be interesting to study mental illness, insofar as it exists (!?), in primates. Are female primates more moody and depressive than males?
https://twitter.com/KirkegaardEmil/status/1071512978695053312
Ever since I learned that Albert Einstein’s son Eduard had schizophrenia to a level that required him to be institutionalized, I sort of assumed that schizophrenia must have something to do with genius.
Also because—take it with the usual corrections for self-reporting—I consider myself to be a genius. My family tree on my father’s side is full of schizotypal personalities. I strongly suspect that being Mennonite, chased to the coldest edges of Europe and beyond because of one’s deeply held esoteric beliefs, selects for this trait.
I remember a post from you about this, Scott (found it: reviewing Surfing Uncertainty https://slatestarcodex.com/2017/09/05/book-review-surfing-uncertainty/). You talked about schizophrenia as a hyperaffective reaction disorder to predictive modelling errors, as opposed to autism which produces hypersensitivity in the models themselves. To put it another way, schizophrenia, in the predictive-modelling paradigm, is a disease that makes it difficult to ignore surprises.
If someone lives their lives constantly having their models challenged, and having their attention pulled to every little model-error, it creates a lot of pressure to build better models, right? If your models get better, you get fewer errors and your mental life becomes easier to manage. And those are the models that, if you happen to be able to communicate them well, become PhD material.
On the other hand, if your models don’t get better, either because you were unlucky, not cognitively flexible enough, or the inundation of model-errors was just too much, the result is something that looks more like schizophrenia.
It makes sense to me that cranking up the genetic dial labeled “schizophrenia” increases the risk-reward of cognition: either you will produce surprising insights and breakthroughs that greatly simplify your (and maybe others’) predictive models, or you will struggle to cope with the basic demands of life and lurch spasmodically from breakdown to breakdown. Or both.
The hypothesis, then, is that Albert Einstein had the right mix of schizophrenia and cognition-enhancing genes to use his irritation at model-errors effectively; Eduard got the irritation but not enough of the cognitive mechanisms to handle it.
"I have been saying for years that I think some of the genes for some mental illnesses must have compensatory benefits. Everyone else said that was dumb, they’re mostly selected against and decrease IQ."
One obvious answer to the "if a gene is bad, it can't be selected for" argument is that a gene can be good in a heterozygous genotype and bad in a homozygous genotype. Sickle Cell being the classic example. One copy of the gene gives malaria resistance; two copies gives you a deadly blood disease.
I don't know much about exactly how these GWAS studies are done. But when they calculate the raw correlations between genes and phenotypes are they able to separate out the heterozygous and homozygous occurrences of the genes?
It seems pretty important. For example, in theory, a gene's hugely positive heterozygous effects might exactly cancel out its hugely negative homozygous effects. So that it's raw statistical correlation to the trait is essentially zero. (This is akin to the man with one foot in boiling water and one foot in freezing water, who is "on average" experiencing a comfortable temperature).
I feel like describing this as "intelligence" vs educational attainment may be misleading, since its actually measuring "IQ test scores" vs educational attainment as I understand it. Not to get into the whole general IQ debate, but to the extent that IQ is useful as a large scale proxy for intelligence, that becomes less meaningful when you are comparing it to another proxy for intelligence. Given the nature of IQ tests that makes some of the correlations less surprising.
Re: Math Phds and autism, I am assuming it was written as a joke, but when I read it I was initially like “sounds about right” and then I thought about all the math grad students and professors I know and realized that none of them seem actually autistic. There is definitely something atypical about most math people I know, but it seems varried and generally not autism.
Do people know stats on this? It might also be because I mainly see academics, and teaching feels like something that would select against autism, so maybe there are more autistic math PhDs out in industry?
There is a theory that the evolutionary reason for suicide is that when a person kills themselves there are more resources, (like food) for their relatives. So depression might exist to cause suicide.
I have the theory that this is right except that the gene for suicide is not in the suicidal person but in the mother of the suicidal person. That it is not a gene for being suicidal, but a gene for making your child suicidal. The child is affected when it in the mother's womb.
The benefit of my theory from the older theory is that the evolutionary drawback of the suicide is halved, as the person committing suicide only has half the genes of the mother. But the evolutionary benefit of the suicide stays the same.
I made a comic about it: http://evolutions.thecomicseries.com/
Very nice post. A note on the schizophrenia result and false discovery. In figure 4 they are graphing 95% confidence intervals (plus or minus 2 standard deviations) and if you look at the confidence intervals for schizophrenia they are very far apart. If you tripled the size of the confidence intervals (plus or minus 6 standard deviations) they still would not overlap. Assuming everything is normally distributed (which they have already done, their confidence intervals rely on this) we can get a nonoptimal bound that if we run a billion trials, we expect to see a false positive in less than 2 out of a billion tries. (This number should not be taken literally because the approximation of the data by the normal distribution will not have this level of accuracy, and there are other possible errors in the study. However it accurately expresses that seeing an effect of this size simply due to a false positive from running many comparisons is exceedingly unlikely.) In the paper they also list a P_{diff_fdr}<.001 which I think is supposed to be the P value of the schizophrenia result after taking into account the risk of false discovery. I include the simple and loose analysis above simply to demonstrate that even if I am misinterpreting their P value, from the visual of the confidence intervals alone, false discovery seems very unlikely here.
There is a theory that schizophrenia is a side effect of human self domestication. Which of course means you'd need to buy into the self domestication hypothesis first but it would explain how we got schizophrenia.
"... whether you so desperately seek societal approval that you're willing to throw away your entire twenties on a PhD with no job prospects at the end of it."
Ouch, man. That hurts.
If functional mental disorders are primarily due to evolutionary mismatch, then there is no need to explain the benefits of "mental illness genes." While some people are genetically more prone to adult onset diabetes than are others, in the environment of evolutionary adaptation such diabetes was rare. We don't try to explain the benefits of diabetes genes. Likewise, while some people are more genetically prone to functional mental disorders than are others, in the environment of evolutionary adaptation it seems likely that such disorders were likewise rare.
Durkheim's research on the higher rates of suicide among Protestants than Catholics was the beginning of a long tradition of empirical research relating to the phenomenon of anomie, as opposed to cultural embeddedness with strong social roles and bonds, leading to adverse mental health outcomes. Here is a recent study (n = 8446) on how increased exposure to US culture increased suicide attempts among youth in the Dominican Republic,
"The increases in the propensity to attempt suicide for DR youths across these US cultural involvement indicators were both robust and large. For example, the propensity to attempt suicide ranged from 6.3% for those at the lowest end of the range of use of US electronic media and language to 13.3% for those at the highest end of the range of use of US electronic media and language. This central finding is congruent with the lower suicide or suicide attempt rates found for first-generation or less acculturated Latinos across multiple national and regional cohorts of Latinos."
Liah Greenfeld's "Mind, Modernity, and Madness" provides a neo-Durkheimian account that provides a coherent explanation for how increasing levels of anomie in modernity lead to increased rates of depression, bipolar, and schizophrenia. In traditional cultures, with humanity in the environment of evolutionary adaptation being the most "traditional" societies, there was neither need nor opportunity to construct a personal identity. A human being was one's roles. There was no "I" in the modern sense (cf. Julian Jaynes). One was unaware of the water in which one was swimming. Now we are all fish out of water, flopping around, gills desperately sucking in air in an attempt to maintain mental stability. For some of us it is easy, for others very difficult. The genetic material of fish works just fine in the water.
While Scott is not sympathetic to this explanation, I've never seen compelling evidence that shows that functional mental disorders were not due to evolutionary mismatch. Yes, whether or not functional mental orders have increased over time and across cultures or not remains contested. Depending on one's priors, the burden of proof shifts.
But studies of suicide provide less contested evidence that culture is a major influence on suicide rates. As far as I know, all such comparative studies are consistent with greater anomie, greater burden on constructing one's own identity (as opposed to a relative lack of the need to create an identity in more traditional cultures) resulting in higher rates of suicide.
Human beings evolved over many millions of years in diverse physical environments. But with respect to social structure, until the dawn of agriculture and empire, almost all adolescents:
1. Lived in a small tribal community of a few dozen to a few hundred with few interactions with other tribal groups.
2. These tribes would have shared one language, one belief system, one set of norms, one morality, and more generally a social and cultural homogeneity that is unimaginable for us today.
3. They would have been immersed in a community with a full range of ages present, from child to elder.
4. From childhood they would have been engaged in the work of the community, typically hunting and gathering, with full adult responsibilities typically being associated with puberty.
5. Their mating and status competitions would have mostly been within their tribe or occasionally with nearby groups, most of which would have been highly similar to themselves.
Could the dramatic divergence from the environment of evolutionary adaptation in any or all of these socio-cultural features result in increased "mental illness" for a genetic subset of human populations?
https://flowidealism.medium.com/evolutionary-mismatch-as-a-causal-factor-in-adolescent-dysfunction-and-mental-illness-d235cc85584
"Their genetic measure of non-cognitive skills... was still correlated at r = 0.31 with IQ"
Note that this is also a *genetic* correlation - the genetic influences of NonCog correlate at r = 0.31 with the genetic influences of IQ, not with actual measured IQ. The same is true for the Cog/NonCog relationship you mention with self-reported math ability and highest math class taken. (Also, assortative mating would inflate these correlations.)
Delay discounting is worse than alcoholism? I interpret this not as meaning that delay discounting is bad for success, but that high educational attainment is bad for success. Huge delay discounting = not going to college because college takes a long time (not obviously wrong); no delay discounting = being willing to go to college for 12 years to get a marginally nicer job (obviously wrong).
The way they use correlation assumes that all personality traits have linear effects on whatever they're measuring. If the function relating {number of genes "for" a behavior} to {educational attainment or IQ} has a tall U-shaped (or upside-down-U-shaped) curve, as some might, the correlation results will depend mainly on the outliers. For example, if a little conscientiousness helps you finish college, but a whole lot makes you such a perfectionist that you're likely to fail college, then the "correlation" between conscientiousness and finishing college isn't telling you how strong the effect of conscientiousness on finishing college is; it's telling you something about the skew of the U-shaped function.
For some reason, when people talk about genetics they like to forget that correlation is not causation.
Scott, you are usually vigilant for distrusting "correlation but we adjusted for confounders" studies. This is the same type of study!
What are "genes for intelligence"? In this context, they are genes that are correlated with intelligence. A gene that purely causes black skin (but does literally nothing else) would be counted as if it decreased IQ, because statistically people with black skin do worse on IQ tests.
There is no end to the number of possible confounders here; there is no exogenous source of randomness at all.
In general: I've often thought that a lot of people go to college and get phds just because they are familiar and comfortable with school and scared or unsure about entering the 'real world' (certainly true for me). I've often thought that a lot of the apparently high levels of mental weirdness in phd programs are largely related to this - who wants to effectively extend their childhood by staying in school, vs who is ready to 'grow up' and enter the real world.
I have some experience living with people diagnosed as schizophrenics. I put it that way because I'm not sure how well understood schizophrenia is and how confident we can be in a clinical diagnosis.
Other than living with diagnosed schizophrenics, I know very little about the condition. My experience is that schizophrenics live in a fantasy world. They are unable to tell the difference between the real world of experience and the world inside their head. I know one in particular who can talk at length of his day to day life in Vietnam, when in fact he has never been there. He appears to be unable to distinguish between fantasy and reality.
On the other hand, my impression of highly successful scientists, engineers, mathematicians, linguists, etc... is they can build mental maps that enable them to navigate their special practices. It seems to me they comprehend very complicated systems, such as molecular biology or microelectronics using some type of mapping onto real world experiences.
What I mean to say is what schizophrenics and mathematicians, e.g., have in common is the ability to live in an alternate reality. I would say visualizing the complex folding of a protein molecule is not too far removed from imagining that you are in fact Sgt. Barry Sadler in 1969.
As a high-IQ person who withdrew from college for mental health reasons, this is really interesting personally
> you can't make a six digit number of people sit down and take IQ tests.
Are you kidding? This happens multiple times a year.
> you can't make a six digit number of people sit down and take IQ tests.
Actually: https://biobank.ndph.ox.ac.uk/ukb/field.cgi?id=20016
But in general, you're right.
It's amazing how dominant UK Biobank is in genetics right now (at least behavioural genetics). I was at the IGSS conference (https://cupc.colorado.edu/conferences/IGSS_2021/) and more than half the presentations were using that data.
Overall, I'd be worried about confounds here. We have a very noisy measure of "genes for IQ" - noisy because GWAS is noisy, and because the IQ measure itself is noisy (just a quick 12 point test IIRC). Then we deduct that from "genes for educational attainment". What's left? Maybe "genes for non-cognitive skills". But maybe "genes for IQ, that we didn't measure very well". Indeed, "non-cognitive PGS" predicts IQ.... And then there are all the possible environmental confounds. I think I'd rather see a measure of non-cognitive skills and then a GWAS that targets that directly.
However, that is just a lazy first take, and I should stop shooting my mouth off about a coauthor's paper.
Not sure if you care, but autism and ADHD are not considered "mental illnesses".
If schizophrenia genes increase education, that would also be fitness reducing.
There's a lot of inference to worry about here, but I'm already stuck on this, from the paper: "By construction, NonCog genetic variance was independent of Cog genetic variance (r_g = 0)." What sense of "independence" follows from zero covariance? That's pretty clearly going to create some weird conditional relationships between their SNPs' imputed Cog and NonCog scores to maintain zero correlation. It seems they recognize this in the supplement but they don't really bother to interpret it.
I ask this because I follow Charles Murray on twitter in order to argue with his position that there is a race-iq causation - has anyone dared to see if there's a variance of the cognitive genes vs ethnicity?
Seems like this database would provide quite strong empirical evidence on the unfortunate topic
Am I reading the chart correctly to note that SNPs associated with higher cognition are *negatively correlated* with conscientiousness (and extraversion, and agreeableness)? That's absolutely fascinating given that educational attainment (and a whole bunch of other traits associated with success, like wealth and income) are so strongly associated with conscientiousness, and suggests at least to me that most of the difference between educational attainment-promoting and cognition-promoting genes should logically have something to do with the trait.
It also seems like a fascinating counterexample to the idea you see brought up often in population genetics that "all good things are correlated". We see a discussion of a g factor, which includes all of cognition, often, and it seems frequently mooted that g itself is part of h, a general health factor. (Also, while I would certainly associate high-cognition-but-low-educational-attainment individuals with low conscientiousness, it boggles the mind to suggest that low conscientiousness, by itself, is associated with higher cognition. Why should that be true? Where's the tradeoff?)
Assuming I am reading the graph correctly, and blue is just 'high cognition', and not something like 'SNPs promoting high cognition but not educational attainment'. I could go back and try to read the chart more carefully, or read the article the chart comes from, but I think I'm too low conscientiousness to go and do so. One read-through is enough.
I've never been tested for ADHD and I don't know if I'll ever bother as my country has a dismal record of treating it, but I strongly suspect I have it. In any case I am extremely hyperactive and have been for a long time.
My life was basically total chaos up to a point. I finished university with a very mediocre result (around 2.4 I think). I aced courses that were interesting to me, and I failed Statistics 101 TWO times until I got angry the third time and got a B+.
Then I was lucky enough to find a technique that works for me - a sort of a Zen meditation where you just sit without moving for a long time. It doesn't remove my ADHD symptoms, but it calms me down a lot and allows me to focus and work.
I am a programmer now and I think there are some advantages, but only because I am treating it in some way - otherwise it would just be a total burden.
The advantages:
- I am better than others in scanning a large amount of code in a search for some obscure problem. In more general sense, I am just better in scanning in general - if I have some list in front of me and I need to find something, I do it faster than others.
- When something gets interesting to me, I feel like my mind gets totally obsessed and get actually angry if someone distracts me - I have shouted at people for that. This might sound bad, but allows me to do fast analysis of a large amount of data.
Disadvantages:
- I have a problem when I need to slow down and focus on one thing. I just cannot motivate myself to study, I have tried for years. The only way I learn new information is when I actively write code, because it is interesting and not boring.
- If I start procrastinating even a little it is very possible that my whole day will be spent in Reddit and YouTube - basically when my brain gets interested in some bullshit many hours can pass before I can stop myself.
- I work well when my work is interesting and badly when it is not. For example, it is interesting to write some new feature, less interesting to make a strategy to test it, so that might take much more time than it has to.
Hope this was interesting if not useful.
Can someone explain what "PGS analysis" means in the second graph? Maybe this is a dumb question but a Google search didn't help me at all.
Hypothesis: schizophrenics have a better connection to the supernatural, which is actually real. They're the most misunderstood mentally ill people, except maybe cluster B folks. I'm sorry to admit this hypothesis is more than a little bit inspired by stereotypes about chosen people.
There's a history of schizophrenia in my family. There's also a history of genius, creativity and outrageous financial success. The money is nice. But I miss my brother.
There is now genetic and epidemiological evidence that Autism and mental illness genes overlap (especially with ADHD, parkinsons). But this is not being communicated with the public. How long has this been covered up? I can't believe that in almost 100 years of autism research this has not been noticed or investigated until now.
Kirkegaard: "Told you about Verbal Tilt yo!" (/s)
But notice how verbal vs math is similar to non-cog vs cog or verbal tilt vs IQ, that could be elaborated upon?