The article that you link shows a researcher supporting single digit percentage heritability doesn't actually show what you claim it shows. The article says that a polygenic score for patients with neuroimaging data (with only 27k samples) explained 7.6% of the variance in g.
PGS variance explained != heritability! That's like reporting benchmark results for your machine learning model before its finished training.
The low IQ heritability estimates you do find in the literature such as https://pmc.ncbi.nlm.nih.gov/articles/PMC6411041/ all seem to have the same issue: they estimate SNP heritability based on UK Biobank's fluid intelligence test, but they fail to account for the fact that the test sucks! Gold standard IQ tests have a test-retest correlation of >0.9. UK Biobank's is short, so the test-retest correlation is 0.61. This is massively deflating estimates of SNP heritability, and thus broad sense heritability!
They calculate SNP heritability of 0.19-0.22 when actual SNP heritability (after adjusting for the crappy test) is about 0.3.
But that's just SNP heritability. A good portion of the variance in IQ comes from rare variants (population frequency <1%), and about 20% of it comes from non-linear effects that are going to be very hard to capture without much larger sample sizes.
I have yet to find a credible IQ heritability estimate that's lower than 0.5
The biggest thing I took away from my Instrument Development class in grad school was that an instument is only valid when used for it's intended purpose. The IQ test was originally created to classify French orphans with cognitive delays into categories that could be used to group their educational needs by the program supporting them. Taking IQ measurements outside of this context automatically makes them invalid; the context has changed and the instrument for deciding what an individual's IQ is is no longer applicable. Sure, we can acknowledge that defferent instruments have been designed and used over the years to try to quantify intelligence - we certainly aren't using the same test the French orphans were given- but we've completely shifted the purpose of IQ from grouping students with similar abilities to recieve education suited to them, to making IQ a predictive measure of mental ability and success! And how can a body of research be considered solid on such a shifting foundation?
False. A measurement can be valid for a variety of contexts, particularly when it's measuring something primary. Besides which mere correlation is sufficient for it to be predictively useful.
Validity does not lie in the measurement. It lies in the inferences and arguments we make about the measurement. The same test can be wonderfully reliable, while the conclusions draw are completely non valid. A conclusions drawn from good IQ test (smart student) typically have strong validity when predicting academic outcomes (including SAT scores). The predictive valid those same conclusions (smart student) increasingly decline as they move further and further from academic performance, (e.g., college completion, career success, earnings, social success). That’s because those things involve intelligence, many other personality and social variables, and the interactions between all those variables. IQ tests are a wonderful tool when used correctly, but in the wrong hands, it can do real damage… like an ax in the hands of a woodsman verses an ax in the hands of a murderer.
There are three factors; a) your raw general intellectual ability - IQ, b) what you choose to spend it on, c) what your society allows. Having the ability to be great is only the first step. Society must also not get in your way and enable you.
IQ is a good predictor of outcomes -- on aggregate, and the larger the sample, the more accurate the prediction. Unlike beauty, for an individual chosen at random, correlation is weak.
I haven't seen any good studies showing that high IQ is compensated for by low EQ, and vice versa. It's the old fallacy of implying "all" when "some" is correct. EQ seems to be fairly evenly distributed across the population, as are other personality factors like charm and selling ability.
Great piece, thank you! I just got drawn into a big rabbit hole of IQ discourse on my own Substack(partially my own fault). Arguing with the most hardcore race essentialists is like debating phrenology in the 1900s. At the same time complete IQ denialism is counterproductive, as you note. Your article is useful and I'll link it next time I write about this!
I always remember Ian Deary's studies tracking whole populations across a lifetime showing that tests taken at teenagehood predict who'll do better in school, in life and on a later test when the participants are in their 70s.
Thanks for publishing this article, it could not have done a better job at filling gaps in my understanding of the topic. Really shifted the way I think about things.
That educational attainment is often taken as a factor for measuring IQ speaks volumes about the inherent lack of repeatability and objectivity of intelligence research. What is the logic behind correlating academic success to intelligence? Beats me. Furthermore, what«s never discussed is that the higher the intellectual intelligence, the lower the emotional intelligence - high IQ scores predict low EQ - lack of empathy and lower capacity for emotional processing. The narcissistic or self-interest tendencies are evident in most so-called "intelligent" people scoring high in the IQ tests. Just look at Elon Musk for an example of high IQ-low EQ individual to see my point. What does high IQ serve for if it comes at the price of flaws in emotional processing?
(For the record, I myself am one of those high-IQ-test-scoring people - but at least I am able to acknowledge my own shortcomings in understanding and processing emotions, both mine and others', and capable of recognizing my high levels self-interest.)
I enjoyed the article as someone who has a pop-sci understanding of IQ.
>If medical tests, like tests for cancer, had the same reliability as IQ tests, they’d be throwing up false signals to the point of being unusable.
Disagree. We send out tons of low sensitivity and specificity tests. Physicians, like other professionals, often operate in a bayesian manner. The more severe (and unexplained) that the patient looks, the more likely we are to send long shot tests.
Billions of dollars of policy interventions is sufficiently impactful that it should merit consideration of all useful, even if imperfect, information. It is all the more important in that setting to know its limitations, as you are writing about.
This is a good reminder to not be satisfied with a pop-sci view of IQ. Do you have any recommended books?
You have shown that the SMPY has some limitations. But you seem to be implying that we should take this as evidence against the claims that prodigies destined for greatness can be identified early, and that there is no plateau of ability. It isn't. You seem to have an unreasonably low prior on the existence of innate prodigies and super-geniuses
IQ is definitely heritable but remember how heredity looks in four siblings. Or ten siblings. We don’t all inherit the same.
Environment matters a ton. But everyone knows realistically that some people are born with a head start others born with a deficit varying by area and naturally people are drawn to areas where they more easily excel and turned off by areas that are more difficult
I think creativity matters as much as intelligence. Are they connected? To an extent. Creativity can likely be measured to a degree but I suspect people would find it offensive even more than many find iq tests offensive. I just look at it as a set of averages that may not be accurate for any individual for a variety of reasons.
Just bc most people who contribute are high iq doesn’t mean most high iq people even super high iq will make significant contributions though it likely means they’ll make a decent living
In terms of success social behavior matters too. Maybe not for figuring out a new theorem but it matters a lot in the business world
Supposedly Terence tao had advanced tutoring. But for all the aristocratic models you could name many times more people who didn’t have that advantage though many eventually had the privilege of top schools.
When I was a kid they promoted grades pretty easily. Now it’s like a hard reverse. Your kid could be years ahead they won’t even push them up a year which is unlikely to have social repercussions.
«It is not based on the hardness of some known neuroscientific mechanism we’ve pinpointed that underlies intelligence. There is no causal account we can give for the biological basis of IQ that has any serious tensile strength. Most research consists of paper tests administered for an hour or two, sometimes decades ago, followed by contortion after contortion to account for the complexities of the world.»
Not "one mechanism", as that would be absurdly reductive. But to explain intelligence via a variety of interlocking neurochemical and structural mechanisms, and as a result of various mutually reinforcing virtuous loops and cognitive strategies? Really not that hard, as we know quite enough about how the brain works. It would end up a Frankenstein-synthesis of various models and levels, but together such a thing would about explain it for the most part.
It's on my to-do list, though someone motivated enough, with more time and more knowledge than I've gotten over the years, could easily beat me to it.
The article that you link shows a researcher supporting single digit percentage heritability doesn't actually show what you claim it shows. The article says that a polygenic score for patients with neuroimaging data (with only 27k samples) explained 7.6% of the variance in g.
PGS variance explained != heritability! That's like reporting benchmark results for your machine learning model before its finished training.
The low IQ heritability estimates you do find in the literature such as https://pmc.ncbi.nlm.nih.gov/articles/PMC6411041/ all seem to have the same issue: they estimate SNP heritability based on UK Biobank's fluid intelligence test, but they fail to account for the fact that the test sucks! Gold standard IQ tests have a test-retest correlation of >0.9. UK Biobank's is short, so the test-retest correlation is 0.61. This is massively deflating estimates of SNP heritability, and thus broad sense heritability!
They calculate SNP heritability of 0.19-0.22 when actual SNP heritability (after adjusting for the crappy test) is about 0.3.
But that's just SNP heritability. A good portion of the variance in IQ comes from rare variants (population frequency <1%), and about 20% of it comes from non-linear effects that are going to be very hard to capture without much larger sample sizes.
I have yet to find a credible IQ heritability estimate that's lower than 0.5
The biggest thing I took away from my Instrument Development class in grad school was that an instument is only valid when used for it's intended purpose. The IQ test was originally created to classify French orphans with cognitive delays into categories that could be used to group their educational needs by the program supporting them. Taking IQ measurements outside of this context automatically makes them invalid; the context has changed and the instrument for deciding what an individual's IQ is is no longer applicable. Sure, we can acknowledge that defferent instruments have been designed and used over the years to try to quantify intelligence - we certainly aren't using the same test the French orphans were given- but we've completely shifted the purpose of IQ from grouping students with similar abilities to recieve education suited to them, to making IQ a predictive measure of mental ability and success! And how can a body of research be considered solid on such a shifting foundation?
False. A measurement can be valid for a variety of contexts, particularly when it's measuring something primary. Besides which mere correlation is sufficient for it to be predictively useful.
Validity does not lie in the measurement. It lies in the inferences and arguments we make about the measurement. The same test can be wonderfully reliable, while the conclusions draw are completely non valid. A conclusions drawn from good IQ test (smart student) typically have strong validity when predicting academic outcomes (including SAT scores). The predictive valid those same conclusions (smart student) increasingly decline as they move further and further from academic performance, (e.g., college completion, career success, earnings, social success). That’s because those things involve intelligence, many other personality and social variables, and the interactions between all those variables. IQ tests are a wonderful tool when used correctly, but in the wrong hands, it can do real damage… like an ax in the hands of a woodsman verses an ax in the hands of a murderer.
There are three factors; a) your raw general intellectual ability - IQ, b) what you choose to spend it on, c) what your society allows. Having the ability to be great is only the first step. Society must also not get in your way and enable you.
IQ is a good predictor of outcomes -- on aggregate, and the larger the sample, the more accurate the prediction. Unlike beauty, for an individual chosen at random, correlation is weak.
I haven't seen any good studies showing that high IQ is compensated for by low EQ, and vice versa. It's the old fallacy of implying "all" when "some" is correct. EQ seems to be fairly evenly distributed across the population, as are other personality factors like charm and selling ability.
Great piece, thank you! I just got drawn into a big rabbit hole of IQ discourse on my own Substack(partially my own fault). Arguing with the most hardcore race essentialists is like debating phrenology in the 1900s. At the same time complete IQ denialism is counterproductive, as you note. Your article is useful and I'll link it next time I write about this!
https://darbysaxbe.substack.com/p/on-being-owned-by-the-deep-left
Chris langan is a con man
I always remember Ian Deary's studies tracking whole populations across a lifetime showing that tests taken at teenagehood predict who'll do better in school, in life and on a later test when the participants are in their 70s.
Iq research is terminally flawed because human intelligence is a mosaic not a monolith. Emotions are equally important in determining smart behavior
Thanks for publishing this article, it could not have done a better job at filling gaps in my understanding of the topic. Really shifted the way I think about things.
That educational attainment is often taken as a factor for measuring IQ speaks volumes about the inherent lack of repeatability and objectivity of intelligence research. What is the logic behind correlating academic success to intelligence? Beats me. Furthermore, what«s never discussed is that the higher the intellectual intelligence, the lower the emotional intelligence - high IQ scores predict low EQ - lack of empathy and lower capacity for emotional processing. The narcissistic or self-interest tendencies are evident in most so-called "intelligent" people scoring high in the IQ tests. Just look at Elon Musk for an example of high IQ-low EQ individual to see my point. What does high IQ serve for if it comes at the price of flaws in emotional processing?
(For the record, I myself am one of those high-IQ-test-scoring people - but at least I am able to acknowledge my own shortcomings in understanding and processing emotions, both mine and others', and capable of recognizing my high levels self-interest.)
Watching the followers of IQism argue small details is about as amusing as watching people argue about horoscopes, or theology.
Not as entertaining however. The astrology crowd tends to be better looking and the theology crowd has better music.
What I am saying is if you are going to have faith in mitcholodrians---sorry G you should at least be easy on the eyes or able to sing Amazing Grace.
The simple truth is that nature sets the range of possibilities and nurture/circumstance determines the actual outcome. It is not rocket science.
I enjoyed the article as someone who has a pop-sci understanding of IQ.
>If medical tests, like tests for cancer, had the same reliability as IQ tests, they’d be throwing up false signals to the point of being unusable.
Disagree. We send out tons of low sensitivity and specificity tests. Physicians, like other professionals, often operate in a bayesian manner. The more severe (and unexplained) that the patient looks, the more likely we are to send long shot tests.
Billions of dollars of policy interventions is sufficiently impactful that it should merit consideration of all useful, even if imperfect, information. It is all the more important in that setting to know its limitations, as you are writing about.
This is a good reminder to not be satisfied with a pop-sci view of IQ. Do you have any recommended books?
>Do you have any recommended books?
Hmm people who are followers of IQism might enjoy:
https://en.m.wikipedia.org/wiki/The_Turner_Diaries
If you are looking to leave your religion I would recommend this good (and very funny) primer on it
https://en.m.wikipedia.org/wiki/Extraordinary_Popular_Delusions_and_the_Madness_of_Crowds
A book from the 1800s and a fictional piece from a white nationalist in the 70s. Unserious recommendations.
IQ discourse remains unhinged.
You have shown that the SMPY has some limitations. But you seem to be implying that we should take this as evidence against the claims that prodigies destined for greatness can be identified early, and that there is no plateau of ability. It isn't. You seem to have an unreasonably low prior on the existence of innate prodigies and super-geniuses
IQ is definitely heritable but remember how heredity looks in four siblings. Or ten siblings. We don’t all inherit the same.
Environment matters a ton. But everyone knows realistically that some people are born with a head start others born with a deficit varying by area and naturally people are drawn to areas where they more easily excel and turned off by areas that are more difficult
I think creativity matters as much as intelligence. Are they connected? To an extent. Creativity can likely be measured to a degree but I suspect people would find it offensive even more than many find iq tests offensive. I just look at it as a set of averages that may not be accurate for any individual for a variety of reasons.
Just bc most people who contribute are high iq doesn’t mean most high iq people even super high iq will make significant contributions though it likely means they’ll make a decent living
In terms of success social behavior matters too. Maybe not for figuring out a new theorem but it matters a lot in the business world
Supposedly Terence tao had advanced tutoring. But for all the aristocratic models you could name many times more people who didn’t have that advantage though many eventually had the privilege of top schools.
When I was a kid they promoted grades pretty easily. Now it’s like a hard reverse. Your kid could be years ahead they won’t even push them up a year which is unlikely to have social repercussions.
And Justin Bates being a young man became Master Bates 😆😁
«It is not based on the hardness of some known neuroscientific mechanism we’ve pinpointed that underlies intelligence. There is no causal account we can give for the biological basis of IQ that has any serious tensile strength. Most research consists of paper tests administered for an hour or two, sometimes decades ago, followed by contortion after contortion to account for the complexities of the world.»
Not "one mechanism", as that would be absurdly reductive. But to explain intelligence via a variety of interlocking neurochemical and structural mechanisms, and as a result of various mutually reinforcing virtuous loops and cognitive strategies? Really not that hard, as we know quite enough about how the brain works. It would end up a Frankenstein-synthesis of various models and levels, but together such a thing would about explain it for the most part.
It's on my to-do list, though someone motivated enough, with more time and more knowledge than I've gotten over the years, could easily beat me to it.